Proceedings
EXPERIENCING LIGHT 2009
Proceedings
EXPERIENCING LIGHT 2009
International Conference on the Effects of Light on Wellbeing
Y. A. W. de Kort, W. A. IJsselsteijn, I. M. L. C. Vogels,
M. P. J. Aarts, A. D. Tenner, & K. C. H. J. Smolders (Eds.)
Keynotes and selected full papers
Eindhoven University of Technology,
Eindhoven, the Netherlands, 26-27 October 2009
Volume Editors
Yvonne de Kort, PhD
Wijnand IJsselsteijn, PhD
Karin Smolders, MSc
Eindhoven University of Technology
IE&IS, Human-Technology Interaction
PO Box 513, 5600 MB Eindhoven, The Netherlands
E-mail: {y.a.w.d.kort, w.a.ijsselsteijn, k.c.h.j.smolders}@tue.nl
Ingrid Vogels, PhD
Visual Experiences Group
Philips Research
High Tech Campus 34, WB 3.029
5656 AE Eindhoven, The Netherlands
E-mail: ingrid.m.vogels@philips.com
Mariëlle Aarts, MSc
Eindhoven University of Technology
Department of Architecture Building and Planning
PO Box 513, VRT 6.34
5600 MB Eindhoven, The Netherlands
E-mail: M.P.J.Aarts@tue.nl
Ariadne Tenner, PhD
Independent consultant
Veldhoven, The Netherlands
E-mail: ariadne.tenner@onsmail.nl
ISBN: 978-90-386-2053-4
Copyright:
These proceedings are licensed under Creative Commons Attribution 3.0 License (Noncommercial-No Derivative Works) This
license permits any user, for any noncommercial purpose – including unlimited classroom and distance learning use – to
download, print out, archive, and distribute an article published in the EXPERIENCING LIGHT 2009 Proceedings, as long as
appropriate credit is given to the authors and the source of the work.
You may not use this work for commercial purposes. You may not alter, transform, or build upon this work.
Any of the above conditions can be waived if you get permission from the author(s).
For any reuse or distribution, you must make clear to others the license terms of this work.
The full legal text for this License can be found at
http://creativecommons.org/licenses/by-nc-nd/3.0/us/legalcode
Reference specification:
Name Author(s), “Title of the Article”, In: Proceedings of EXPERIENCING LIGHT 2009 International Conference on the
Effects of Light on Wellbeing (Eds. Y.A.W. de Kort, W.A. IJsselsteijn, I.M.L.C. Vogels, M.P.J. Aarts, A.D. Tenner, and
K.C.H.J. Smolders), 2009, pp. X (startpage) – Y (endpage).
Preface
Perhaps Led Zeppelin said it best when they sang “Everybody needs the light”. It is hard to overestimate
the importance that light has for the human condition. From the comforting atmosphere of a quietly lit
living room, to the invigorating effects of morning light, preparing you for your day, light has dramatic
effects on mood, health and productivity, and can deeply influence the way we experience an
environment. Biologists acknowledge the powerful influence that Earth’s 24-hour light–dark cycle has on
the behaviour and physiology of animals and humans that evolved on this planet. Psychiatrists and clinical
psychologists, treating patients for sleep disorders or seasonal affective disorders, can attest to the
importance of light for psychological wellbeing. Human factors professionals too recognise light as a
significant factor in people’s health, performance, and safety in a variety of contexts, including factories,
offices, schools, and homes. Artists - from Golden Age painters to modern day cineasts - all have been
keenly aware of the aesthetic and emotional impact light has on our experience of art; its power to create
mood, suspense and mystery, to capture our gaze, and to challenge our curiosity. Similarly, in architecture
and urban planning, the importance of getting the lighting right, whether from natural or artificial sources,
is generally acknowledged. The right light enhances and improves a space; bad lighting degrades it. Light
has the power to transform the social context, creating zones of safety and comfort, making spaces more
visible, more agreeable, more habitable, and stimulating social interactions. In short, light is fundamental
to the quality of life.
Experiencing Light 2009 was the first international conference that has as its sole focus the effects of light
and light design on human wellbeing. It approaches wellbeing in its broadest sense, including mood,
emotions, subjective and objective health, comfort, atmosphere perception, productivity and performance.
Rapid developments in lighting technology are allowing for intelligent and interactive lighting designs,
dynamically illuminating public and private spaces, and embedding light in consumer electronic devices,
information displays, artistic objects, and clothing. Experiencing Light 2009 provided a timely and
necessary international forum to discuss the impact of such recent technological developments on user
experience. Experiencing Light 2009 builds on the rich multidisciplinary tradition in lighting research and
design, with inputs from perception research, environmental psychology, human factors, architecture,
lighting design and industrial design.
Experiencing Light 2009 was organized as a two-day scientific event in Eindhoven on 26-27 October
2009. In addition to our exciting keynotes, Jim Tetlow and Martine Knoop, the program of Experiencing
Light 2009 consisted of a number of selected presentations, both oral and in interactive poster format, on
new research and findings, new conceptualizations and designs, and new reflections on light and its
psychological impact. The full papers you find in these Proceedings were selected from the large
collection of submitted papers through a carefully conducted review process, using blind peer-review. We
are greatly indebted to the members of the Scientific Committee for their excellent work in reviewing the
submitted papers and selecting the best papers for presentation at the conference. Short papers that
accompany the interactive posters can be found in the Adjunct Proceedings.
Experiencing Light 2009 was hosted by the Eindhoven University of Technology (TU/e), in Eindhoven,
The Netherlands, as a joint effort between the Human-Technology Interaction (HTI) Group of the
Department of Industrial Engineering and Innovation Sciences (IE&IS), the Department of Architecture,
Building, and Planning, and Philips Research. It is no coincidence that Experiencing Light was initiated in
Eindhoven. Eindhoven has a particularly rich history as a City of Light. Starting at the beginning of the
previous century with the mass production of light bulbs at Philips, it is now a major science, technology,
and design hub, home to Philips Lighting, Philips Design, and Philips Research, Eindhoven University of
Technology, TNO (Dutch Organisation for Applied Scientific Research), SOLG (Light and Health
Research Foundation), and the Design Academy. The city hosts a range of light-oriented events, such as
the annual “Lichtjesroute” (Route of Lights) and the international light festival GLOW – Forum of Light
in Art and Architecture. A recent collaborative initiative to establish a Technological Top Institute Light
(TTIL) in Eindhoven provides a further significant impulse to the joint efforts of the university, industry,
and government in the area of lighting science, technology, and design.
We gratefully acknowledge the sponsors of Experiencing Light 2009: TU/e, HTI, TTIL, the city of
Eindhoven, KNAW (Royal Dutch Academy of Sciences), Philips Research, and Davita. Moreover, we
would like to thank all of those who supported the organization of Experiencing Light 2009 and who
worked hard to make it a successful event, including Atike Pekel, who designed the beautiful website
(http://www.experiencinglight.nl/) and conference materials, our secretarial and logistics support, and our
student volunteers. Thank you all.
October 2009
Yvonne de Kort
Wijnand IJsselsteijn
Ingrid Vogels
Mariëlle Aarts
Ariadne Tenner
Karin Smolders
Organisation
Organising Committee
General chair
Dr Yvonne de Kort, Eindhoven University of Technology
Program chairs
Dr Wijnand IJsselsteijn, Eindhoven University of Technology
Dr Ingrid Vogels, Philips Research
Ir Mariëlle Aarts, Eindhoven University of Technology
Dr Ariadne Tenner, Independent Consultant
Treasurer
Ir Karin Smolders, Eindhoven University of Technology
Sponsors
+$,*"-%-'&,.%(+-/0!"1##$
!"#$%%&'$"#()&'*#&"'
Scientific Committee
Prof Dr Emile Aarts, Philips Research, Netherlands
Dr Ir Myriam Aries, TNO Building and Geosciences, Netherlands
Prof Dr Domien Beersma, University of Groningen, Netherlands
Prof Wout van Bommel, Van Bommel Lighting Consultant, Netherlands
Dr Peter Boyce, Technical Editor, Lighting Research and Technology, UK
Prof Dr George Brainard, Thomas Jefferson University, USA
Dr Truus de Bruin-Hordijk, Delft University of Technology, Netherlands
Ing Peter van der Burgt, Philips Lighting, Netherlands
Dr Karin Dijkstra, University of Twente, Netherlands
Prof Dr Ir Berry Eggen, Eindhoven University of Technology, Netherlands
Dr Mariana Figueiro, LRC, Rensselaer Polytechnic Institute, USA
Marten Fortuin, Hogeschool Utrecht & Eindhoven University of Technology, Netherlands.
Dr Steve Fotios, University of Sheffield, UK
Dr Marijke Gordijn, University of Groningen, Netherlands
Prof Liisa Halonen, Helsinki University of Technology, Finland
Ir Hester Hellinga, Delft University of Technology, Netherlands
Prof Ingrid Heynderickx, Philips Research and Delft University of Technology, Netherlands
Dr Henri Juslén, Philips Oy, Finland
Dr Igor Knez, University of Gävle, Sweden
Dr Martine Knoop, Philips Lighting, Netherlands
Ir Marc Lambooij, Eindhoven University of Technology, Netherlands
Dr Martin Lupton, PLDA, UK
Dr Ybe Meesters, University Medical Center Groningen, Netherlands
Prof Dr Cees Midden, Eindhoven University of Technology, Netherlands
Dr Henk Herman Nap, Eindhoven University of Technology, Netherlands
Dr Guy Newsham, NRC Institute for Research in Construction, Canada
Dr Sylvia Pont, Delft University of Technology, Netherlands
Prof Dr Ad Pruyn, University of Twente, Netherlands
Prof Mark Rea, LRC, Rensselaer Polytechnic Institute, USA
Prof Dr Huib de Ridder, Delft University of Technology, Netherlands
Dr Thomas van Rompay, University of Twente, Netherlands
Prof Christoph Schierz, Ilmenau University of Technology, Germany
Ir Dragan Sekulovski, Philips Research, Netherlands
Dr Ir Pieter Seuntiens, Philips Research, Netherlands
Dr Jennifer Veitch, NRC Institute for Research in Construction, Canada
Prof Arnold Wilkins, University of Essex, UK
Contents
K1 Keynote lecture:
Creating and altering perceptions with lighting, or: How to sell with light
Jim Tetlow
1
K2 Experiencing LED: Let music lead the way?
Martine Knoop
3
1.1 Influence of ambient lighting in vehicle interior on the driver's perception
Luca Caberletti, Kai Elfmann, Martin Kümmel and Christoph Schierz
5
1.2 The effects of lighting on atmosphere perception in retail environments
Pieter Custers, Yvonne de Kort, Wijnand IJsselsteijn and Marike de Kruiff
14
1.3 Effect of lamp spectrum on perception of comfort and safety
Colette Knight
22
1.4 Light and corporate identity. Using lighting for corporate communication
Thomas Schielke
31
2.1 Tuning the spectrum of lighting to enhance spatial brightness: Investigation of
research methods
Steve Fotios and Kevin Houser
41
2.2 Ecological measurements of light exposure, activity, and circadian disruption in
real-world environments
Daniel Miller, Andrew Bierman, Mariana Figueiro, Eva Schernhammer and Mark
Rea
53
2.3 Content-based adaptation of the dynamics of estimated light source
Marc Peters, Pedro Fonseca, Lu Wang, Bas Zoetekouw and Perry Mevissen
62
2.4 Descriptions, measurements and visualizations of light distributions in 3D spaces
Sylvia Pont, Alex Mury, Huib de Ridder and Jan Koenderink
74
3.1 Flexible light sources for health and well-being
Margreet de Kok, Herman Schoo, Marc Koetse and Ton van Mol
79
3.2 Effect of glazing types on daylight quality in interiors: conclusions from three
scale model studies
Marie-Claude Dubois
86
3.3 Effect of LED-based study lamp on visual functions
Srinivasa Varadharajan, Krithica Srinivasan, Siddhart Srivatsav, Anju Cherian,
Shailaja Police and Ramani Krishna Kumar
98
3.4 Using core sunlighting to improve office illumination
Lorne Whitehead, Allen Upward, Peter Friedel, Michele Mossman, John Huizinga
and Tom Simpson
106
4.1 Effects of dynamic lighting on office workers: First-year results of a longitudinal
field study
Yvonne de Kort and Karin Smolders
114
4.2 Persuasive lighting: The influence of feedback through lighting on energy
conservation behavior
Jaap Ham, Cees Midden, Saskia Maan and Bo Merkus
122
4.3 A transformational approach to interactive lighting system design
Philip Ross, Kees Overbeeke, Stephan Wensveen and Caroline Hummels
129
4.4 Effects of colour and light on customer experience and time perception at a virtual
railway station
Mark van Hagen, Mirjam Galetzka, Ad Pruyn and Joyce Peters
137
5.1 Preliminary evidence that both red and blue lights increase nocturnal alertness
Mariana Figueiro and Mark Rea
146
5.2 Reducing light intensity and changing its spectral composition: effects on human's
sleep characteristics and melatonin rhythms under "natural conditions"
Marina Giménez, Pauline Bollen, Marijke Gordijn, Matthijs van der Linden and
Domien Beersma
151
5.3 Reflections on the eyelid: Experiencing structured light through closed eyes
Adar Pelah, Su Liu, Howard Hock, Mathew Gilbert and Philip Jepson
155
5.4 Evaluation of today's research methods for assessing light-induced alertness
Emilia Rautkylä, Petteri Teikari, Marjukka Puolakka and Liisa Halonen
162
KEYNOTE LECTURE
Creating and Altering Perceptions with Lighting
or
How to Sell with Light
Jim Tetlow
Nautilus Entertainment Design
1010 Turquoise St., Suite 215
San Diego, CA 92109
ABSTRACT
A look at how a practicing lighting designer establishes and
modifies people's perceptions of environments, products
and people through the craft of lighting.
INTRODUCTION
Perception can be defined as “the process of attaining
awareness or understanding from sensory information”.
Through the sense of sight, we can use lighting to influence
people’s perception of spaces and objects, and as a working
lighting designer, that is what I’m called upon to do in
many circumstances.
The presentation shall consist of several case studies, which
will include a combination architectural/exhibit project,
several highlights from theatrical introductions of new
automobiles, and the American Presidential Debates. The
common theme of these different projects is that lighting, in
conjunction with other disciplines, is used as a sales tool,
whether it is for products or presidential candidates.
To illustrate how lighting can be used to alter the
perception of an environment, I would like to discuss the
Hewlett Packard Exhibit at The International
Telecommunication Union, or ITU conference, which is the
major exhibit and conference for the telecommunications
industry. As is typical of these large international exhibits,
the convention floor environment was noisy, crowded, and
certainly not a conducive environment for having a serious
conversation with a potential client. Contrary to the typical
exhibit booth at the ITU, Hewlett Packard wanted to create
a different environment. One where their potential clients
perceive that they have been transported away from the
crowded and noisy convention floor to a much calmer
space where they could hold private meetings and provide
hospitality. To accomplish this, a two storey structure was
fabricated with an exhibit space on the ground floor and a
hospitality lounge and private meeting rooms on the first
floor. The lighting for this project was architectural in style,
but needed to be installed and dismantled rapidly. Lighting
was used for establishing the exterior of the exhibit
structure as a landmark that could be seen from far away in
the large convention hall. It was also used to create a sense
of privacy for the hospitality lounge and intimacy for the
meetings rooms and adjoining hallways.
To illustrate how lighting can be used to enhance our
perception of products, I would like to discuss the lighting
of several product launches in the automobile industry.
Every year, the major automobile manufacturers introduce
dozens of new car models. In order to get the public excited
about their new products, the first step is to make the
employees and salesman excited about what they will be
selling. This is especially true in dealerships where a more
than one brand of automobile is sold. Each brand needs to
make their product more attractive than the next and in an
era where many of the products are very similar in
performance and appearance; it is the perception of the
product that becomes important as a sales tool. Contrary to
the architectural style of lighting used for the Hewlett
Packard exhibit, these product launches are real shows,
many times incorporating dancers and special effects.
Specific examples will be shown from productions for
Mercedes Benz and Toyota.
Every four years America has a presidential election and
for the past 20 years the Presidential Debates have been a
critical part of the election process. Produced by the nonpartisan Commission on Presidential Debates, these forums
are the only opportunity that voters have to see and hear the
two candidates speaking with each other. There are
normally three presidential debates and one vicepresidential debates produced in several different formats.
The goal of the Commission is to provide a neutral
environment where each candidate can be comfortable and
present themselves and their platform to the American
people. For the candidates, this is a critical opportunity to
shape the public’s perception of themselves and their
beliefs, and to that end each candidate’s team is constantly
striving for the upper hand. The lighting is different from
any of the previous examples in that it is essentially portrait
lighting for television.
The solution is simplistic but
flexible enough to provide the ability to make each
candidate appear their best. As one might expect, the most
1
interesting part of this project is not the lighting, but the
politics.
Sesame Street, two other nominations, and a 1985 Monitor
Award for a music video with Jim Henson's Muppets.
BIO
He has been referred to as the guru of entertainment
systems design for his work on 29 ships for various brands
of the Carnival Corporation cruise ship fleet. He has also
worked extensively as a lighting designer on corporate
videos and live theatrical productions for such clients as
General Motors, Hewlett-Packard, Daimler-Chrysler,
Mercedes Benz, Nissan, Porsche, Michelin, Polaroid, IBM,
and an interactive live/video presentation with
Mummenschanz for AT&T.
Jim Tetlow is a theatre consultant, television and theatrical
lighting designer, and principal of Nautilus Entertainment
Design, based in San Diego, California. He was Lighting
Designer for the US Presidential Debates, and lighting
consultant for many of the Obama Inaugural Events.
Jim Tetlow is a graduate of Carnegie Mellon University
and has been working as a lighting designer and consultant
for television, theatre, and architecture since 1975. He has
been the recipient of an Emmy Award, won in 1990 for
2
KEYNOTE LECTURE
Experiencing LED:
Let music lead the way?
Martine Knoop
LiDAC International
Philips Lighting
Professional Lighting Solutions EMEA
Mathildelaan 1, Building EDW-618
5611 BD Eindhoven, the Netherlands
ABSTRACT
LEDs seem to be the promising light sources of today and
tomorrow. They are small, offer high brightness and
saturated colours. Due to the fact that it is a revolutionary
different light source, new possibilities in experiencing
light are assumed, but difficult to pin point. The
presentation will look into the evolution of another
technology, showing similarities in origin and experience,
but being in a later stage of development. Can we learn
from this? Can we expect experiences that go beyond those
that we are used to, using LED technology?
INTRODUCTION
LEDs seem to be the promising light sources of today and
tomorrow. They are small, offer high brightness and
saturated colours. Due to the fact that it is a revolutionary
different light source, new possibilities in experiencing
light are assumed, but difficult to pin point down.
The presentation will look into the evolution of the
technologies related to experiencing sound and
experiencing light. These technologies show similarities in
development as well as the way to experience it. It is
postulated that a flashlight is the equivalent of a
ghettoblaster and visiting a concert is comparable to
heliotherapy. In the same line of thought, one could
consider that the work of Olafur Eliasson can be compared
with classical music, whereas Dan Flavin and James Turell
could be seen as rock star equivalents.
Experiencing sound has changed over time due to the
development of technology. Due to recording and
reproduction possibilities it already shifted from a public
only, occasional experience to a – in general – more
private, for everyone available, experience. The most recent
and revolutionary developments though have been in
digital recording, with the development of digital audio file
formats, processors capable and fast enough to convert the
digital data to sound in real time, and inexpensive mass
storage. This again has lead to development of small music
devices containing an enormous amount of pieces of music.
The LED offers similar possibilities. It is a very small
source and, due to it’s dim and color characteristics, it gives
the opportunity to realize an enormous amount of light
settings. The comparison of both technologies shows that
‘sound’ seems to be ahead of ‘lighting’. As we have seen a
change in sound experience due to technology
development, the question is raised whether we can expect
a similar change in experiencing light.
The presentation will discuss the key learnings of the latest
technology development and it’s effect on the experience of
sound. It will present the opportunities as well as the
drawbacks and risks of adopting the ‘sound’ achievements
into lighting practice. Resulting, it will discuss whether we
can expect lighting experiences that go beyond those that
we are used. Or is the the knowledge gathered from this
related technology and experience actually not applicable?
The concluding part of the presentation will evaluate the
conclusions drawn from the above mentioned analysis in
view of social changes as well as changes in specific
applications, such as offices and elderly homes.
Interestingly enough, even in a broader perspective, there
seems to be a similarity in experiencing sound and
experiencing light. In this view, it will be discussed if LED,
possibly combined with OLED, will be the single lighting
source(s) used in the (near) future, or will we reach out to
the old fashioned ‘CD equivalent’ more often then we
think?
BIO
Martine Knoop is a senior application specialist at the
LiDAC International (Lighting Design and Application
Center) of Philips Lighting in Eindhoven, the Netherlands.
After studying architecture and building physics at Delft
University of Technology, her PhD dealt with day lighting
systems, glare from daylight and acceptance studies in daylit rooms.
Martine Knoop worked in Berlin for the Marketing
department of a manufacturer of luminaires and lighting
controls for four years. After this she started at Philips, and
3
was also part-time visiting professor at Eindhoven
University of Technology, July 2005 till December 2008.
In this position she focused on the balance of light
4
requirements for human beings and possibilities offered by
technology and architecture.
Martine Knoop now focuses on lighting solutions for
physical and mental wellbeing.
Influence of Ambient Lighting in Vehicle Interior on the
Driver!s Perception
Luca Caberletti
BMW Group
Knorrstraße 147
80788, München
+49 89 382 79005
luca.caberletti@bmw.de
Kai Elfmann
Kleefeldweg 6
06724, Kayna
kaielfmann@web.de
Dr. Martin Kümmel
BMW Group
Knorrstraße 147
80788, München
martin.kuemmel@bmw.de
Prof. Christoph Schierz
Ilmenau University of Technology.
Lighting Engineering Group.
christoph.schierz@tu-ilmenau.de
INTRODUCTION
Ambient interior lighting for vehicles is an issue of
dramatically growing relevance in the automotive industry.
In the last decade the number of light sources in the car
interior providing this illumination has drastically
increased. A steadily growing amount of cars in the high
and middle class segments are equipped with such lighting.
Ambient lighting provides an indirect illumination of the
passenger compartment in low light settings, such as during
the night. Its importance lays in the fact that it provides a
better orientation in the car, an improved sense of
spaciousness, as well as an impression of safety, value and
comfort. Furthermore it conveys an emotional and brandoriented atmosphere to the otherwise dark car interior at
night. Moreover, ambient lighting can harmonise the
luminance level between the vehicle interior and the
external environment, thus decreasing the driver’s fatigue
when driving at night [20]. Ambient lighting does not
perform a pure functional role and therefore it can be
designed in any colour, since it does not require a high
colour rendering. Indeed, car makers use different colours
also in order to give a branded image of the car interior.
It is important to notice that since ambient lighting is an
indirect illumination, the materials upon which it reflects
acquire new value and quality. Night design thus plays a
central role, since the materials and the lines of the car
interior are visible not only during daytime but at night too.
On the other hand, disability and discomfort glare caused
by ambient lighting should be avoided, in order not to
impair vision and decrease safety during drives at night.
MOTIVATION
Previous studies by Grimm [7] proved that disability and
discomfort glare originating from ambient lighting can be
eliminated by keeping maximum luminance under
0.1cd/m!. In this way, negative effects on the safety can be
neglected.
Studies by Schellinger et al. [15] and Klinger and Lemmer
[11] stated that the driver’s contrast vision won’t be
negatively affected by ambient lighting, if the driver can
control its brightness.
Other studies on vehicle interior lighting addressed the
issue of possible glare caused by reading lamps or dome
lights through veiling luminance and unwanted mirror
effects [3] [14].
However, there are no guidelines which indicate how to
correctly and consequently arrange ambient lighting in the
car interior in order to maximise its positive effects. In fact,
this procedure is based nowadays upon experts’ personal
judgement.
Many studies investigate the effects of lighting on mood
[12] [13], emotions [6] and perceptions [8] [18], within the
scope of lighting design in buildings and in officeenvironments. Of interest in this study is if such effects can
be caused even in the relatively small environment of the
vehicle and with such small luminance levels as in the case
of ambient lighting.
Thus, in order to fully understand the advantages of
ambient lighting in relationship to its characteristics and
parameters, an experimental research study has been
conducted and will be presented in this paper.
METHOD
In an immersive virtual test environment, 31 test persons
had the task of “driving” a real stationary vehicle on a
virtual highway. In the vehicle, a different ambient lighting
scenario was displayed in each run. In total twelve different
scenarios were tested, in which the following parameters
were varied: light colour, luminance and position.
5
Experimental Setup
The test took place in a static driving simulator at the
BMW Group research centre [9]. The choice of using a
simulator environment rather than leading the test on real
streets gave a complete control on the environmental
variables, guaranteed the repeatability of the experiment,
and thus increased the significance of the results.
A BMW 3 Series equipped with special interior light
features was used for the experiment. It was connected to
the simulator in a way that allowed the driver to steer the
car but not to accelerate and brake (a collision with the
preceding vehicle was impossible because of the control
mechanisms in the driving simulation software). The
driving simulation was projected on three screens placed in
front and around the car, which covered a viewing angle of
about 135°. In the simulator room, an ambient luminance
between 0.01 cd/m! and 0.1 cd/m! was present, which
caused a mesopic visual adaptation. The luminance level on
the simulated street lane was between 0.1 cd/m! and
1.5 cd/m!, a range of values which matches the measured
street luminances in reality [1] [2] [16] [19].
When the driver was unable to accomplish the secondary
task, he was allowed to refuse it.
After 3 minutes, the ambient lighting was turned off and
the vehicle was stopped by the investigator and brought on
the side-strip. The participants then completed the
questionnaire relating to the perceived lighting scenario.
This process was repeated with all twelve lighting
scenarios, which were presented in random order to each
test person.
Ambient Lighting Scenarios
In the test vehicle twelve different ambient lighting
scenarios were realised (Table 1). Three parameters were
varied: colour, position of the lighting sources and
luminance, as described in Table 2.
Table 1 Description of the tested lighting scenarios
Nr.
Lighting Scenario
1.
Everything on – bright level with accents
2.
Series (Centre console + Door trims)
3.
Doors – bright level
4.
Doors – low level
5.
Without lighting
6.
Everything on – bright level
7.
Everything on – low level
8.
Everything on – middle level
9.
Foot space – bright level
10.
Foot space – low level
Execution of the test
11.
Centre console
After the execution of the Ishihara Colour Vision Test [10]
(all the participants had a good colour vision) the room was
darkened. The test persons had 10 minutes for dark
adaptation. During this time the investigator described the
objectives and the methods of the research. Afterwards the
participants drove the vehicle a few minutes on the
simulator in order to become familiar with its steering
feeling. After this period of adaptation the test started.
12.
Everything on blue – low level
Test subjects
The investigation took place with 31 participants, 8 women
and 23 men, between 21 and 58 years-old (mean age 35
years). 18 of them had already experienced ambient
lighting while driving. 14 of them wore glasses or contact
lenses. For each participant the experiment lasted 1.5 to 2
hours.
The investigator sat in a separated room and communicated
with the test persons through a radio. After he started the
simulation, the vehicle accelerated to 100 km/h and then
remained at this speed. During the acceleration the
appropriate lighting scene was activated and then
maintained for 3 minutes. Meanwhile, the participants
drove according to their main task, which was to follow a
car on the right highway lane. Since the attention of the test
persons was focused on the driving task, the ambient
lighting was only perceived peripherally, as in reality.
Each minute the participants were asked to accomplish a
secondary task. The aim of these tasks was to give the test
persons the possibility to evaluate the functionality of the
current lighting situation in enabling normal actions that
take place while driving. For example, typical secondary
tasks were the adjustment of the climate ventilation nozzles
or the finding and operation of a specific control button.
6
Table 2 Experimental parameters
Parameter
States
Colour
Orange (605 nm)
Blue (471 nm)
Position
Centre console
Doors
Foot space
Series (Centre console + Door trims)
Complete
Mean luminance
Bright (more than 0.04 cd/m!)
Middle (0.02 – 0.01 cd/m!)
Low level (0.007 cd/m!)
The lighting colours presented in the test were orange and
blue, with dominant wavelengths of 605 nm and 471 nm
respectively. Lighting positions were selected among the
ones commonly adopted in practice in the automotive
industry. The centre console light is placed inside the roof
node and illuminates the centre console area, where usually
the gear selector lever and the controls for entertainment
and conditioning are placed. Foot space lighting was
realised with two LEDs placed in the cockpit, on both the
driver and passenger sides. The illumination of each door
consists of four LEDs and two light guides, which
combined provide a homogeneous coverage of the door
handles and of the upper part (door trims) and lower part
(map case) of the door.
Figure 2 Example of lighting scenario: series setting - centre
console and upper door trims are on.
Luminance Measurements
Figure 1 Positions of the ambient lighting. a. door trim, b. map
case, c. foot space, d. centre console. With e. and f. the accents on
the right door are highlighted (door handle and door pull
respectively)
The combination of door trims and centre console lighting
are a common setting in series vehicles and therefore was
named series lighting. The setting “everything on” included
all the above-mentioned lighting fixtures properly adjusted
so that they could provide a homogeneous appearance. The
setting “everything bright – with accents” provided a few
additional points (door handles and pulls) with higher
luminance (up to 2 cd/m!).
The luminance of the lighting fixtures in the vehicle was
measured using a luminance camera provided with fish-eye
optic (LMK Mobile Advanced, TechnoTeam, Ilmenau /
Germany). In this way, the brightness in the whole field of
view could be measured from the driver’s perspective. The
visual field has been divided into 4 zones (Figure 3). In
these 4 zones, only the measure points with a photopic
luminance between 0.003 cd/m! and 0.5 cd/m! have been
considered. These areas can be considered illuminated by
ambient lighting. Luminances below the 0.003 cd/m! have
been considered dark, while those above the 0.5 cd/m! have
been considered symbol lighting, and so not to be measured
together with ambient lighting. In Table 3, the mean
luminances LM for these areas are displayed.
Cockpit instruments, display lighting and backlit symbols
were always turned on, as in a real night drive situation.
Anyway their luminosity level was constant during the
whole research.
Figure 3 Luminance measure zones. A: left door; B: centre
console; C: right door; F: foot space.
7
Table 3 Mean Luminance LM for the different measure zones and
the different lighting scenarios [cd/m!].
whole car interior; ...causes a small impression of interior
space / causes a big impression of interior space.
•
(Perceived interior quality) ...looks cheap / looks
luxurious; ...gives a lesser quality impression / gives a
good quality impression.
•
(Interior attractiveness) ...has a really unpleasant light
colour / ...has a really pleasant light colour; ...is too dark /
is too bright; ...appears pleasant / appears unpleasant; ...is
comfortable / is uncomfortable; ...I really liked / I really
disliked.
•
(Perceived safety) ...increases the perceived safety /
decreases the perceived safety.
•
Since the lit area changes with the intensity of the
illumination, the solid angle under which the area is seen
by the driver (") has also been calculated. The product of
the solid angle and the mean luminance LM" for each
considered zone, displayed in Table 4, gives the eye
illuminance, measured in the direction of the area.
(Functionality) ...enables a better orientation in the car
interior / complicates the orientation in the car interior;
...facilitates the finding of controls / complicates the
finding of controls; ...makes me more powerful / makes
me less powerful; ...causes distracting reflections in the
windshields / does not cause reflections in the
windshields;
•
(Alertness) ...distracts me from driving / keeps my
attention on the driving; ...complicates the concentration /
enables concentration; ...makes me tired / activates me;
...makes me sleepy / animates me.
Cockpit lighting as well as backlit symbols have not been
considered in the measures, since they did not vary in
intensity for the whole experiment.
The questions were presented in random order and so
arranged that the positive sentences were equally
distributed on both sides of the questionnaire.
Table 4 Eye illuminance (measured in the area’s direction)(LM")
values for the different measures zones and the different lighting
scenarios [10-3 cd·sr/m!].
The answers were given by the test persons on a continuous
scale with a vertical line signalising the middle, as
represented in Figure 4.
Scenario
A
1
0.023
2
0.009
3
0.023
4
0.022
-
6
0.023
B
0.012
0.011
0.009
-
-
0.010
C
0.023
0.006
0.029
0.017
-
0.026
F
0.008
-
-
-
-
0.008
Scenario
7
8
A
0.021
0.015
-
-
B
0.008
0.010
-
-
0.010
0.013
C
0.017
0.017
-
-
-
0.016
F
-
-
0.008
0.004
-
-
Scenario
9
5
10
11
12
0.028
1
2
3
4
5
6
A
3.17
0.65
2.60
0.62
-
2.64
B
0.71
0.50
0.04
0.03
0.02
0.54
C
1.11
0.05
0.91
0.31
-
0.92
F
0.27
-
-
-
-
0.27
7
8
9
10
11
12
A
0.63
1.41
0.01
-
-
0.86
B
0.13
0.49
0.03
0.03
0.48
0.69
C
0.31
0.48
0.01
-
-
0.37
F
0.05
0.05
0.26
0.04
-
0.01
Scenario
Questionnaire
Subjective perception of the lighting
After each experimental run, each test person was asked to
fill out a questionnaire in the form of 18 semantic
differential pairs, which were arranged according to the
following criteria: space perception, perceived interior
quality, interior attractiveness, perceived safety, alertness
and functionality.
The questions were the following: the displayed light
situation...
•
8
(Space perception) ...allows the perception of the
whole car interior / does not allow the perception of the
Figure 4 Example of the differential pairs questionnaire
Emotional state
Influences of the three lighting parameters on the emotional
state of the test persons were also researched, using a SelfAssessment Manikin (SAM) procedure [4]. This
questionnaire method, displayed in Figure 5, is based on
the PAD Model (Pleasure-Arousal-Dominance), which has
been already adopted to describe the emotional state caused
by colours [17] and lighting situations [5] [6].
everything on low level – everything on bright level with
accents (scenarios 5 – 7 – 1).
The comparison between the scenarios “without lighting”
and that “everything on – low level” showed highly
significant (p<0.01) improvements for the second one in
five criteria: space perception, interior attractiveness,
functionality, perceived interior quality and perceived
safety. Regarding the criterion alertness, no clear trend
could be found: no degradation could be seen either.
Increasing the luminance and getting to the “everything on
- bright level” scenario brought a significant (p<0.05)
decrease in comfort, pleasantness and safety perception,
increasing the distraction and complicating the
concentration for the drive.
Figure 5 Self-Assessment Manikin (SAM) questionnaire [4].
The three independent dimensions pleasure, arousal and
dominance are assessed separately, by checking the box
under the manikin which the test person feels more to his or
her state. The pleasure dimension spans from happy,
content (corresponding to 1 on its scale) to unhappy,
displeased (9). Arousal mirrors the activity of the person,
ranging from agitated, wide awake and aroused (1) to
sleepy, calm and inactive (9). Dominance states if a person
feels controlled (1) or rather in command of the situation
(9).
The test persons were asked to fill out this form at the
beginning of the test (in order to know the emotional state
at the starting point) and after each experimental run.
Luminance variations on single lighting elements produced
no significant differences in the answer distribution, apart
from the brightness assessment, in which the test persons
recognized which scenario was actually brighter. Two
comparisons were employed for this evaluation: doors
bright – doors low level (scenarios 3 – 4) and foot space
bright – foot space low level (scenarios 9 – 10).
The comparison between the scenario without ambient
lighting and that with the centre console illumination
(scenarios 5 – 11) is also interesting, because the latter
represents the minimal ambient lighting that can be found
in today’s series cars. This kind of illumination provided
better interior attractiveness and functionality (p<0.01), and
improved perceived interior quality and space perception
(p<0.05). This means that a minimum quantity of light in
the car interior constitutes already a considerable
advantage, regarding the subjective perception, in
comparison to dark.
Effects of Colour
RESULTS
Although the influence of ambient lighting on the
emotional state of the test-persons could not be verified,
this study confirmed that the different light scenarios
significantly influenced space perception, perceived interior
quality, interior attractiveness, as well as perceived safety
and functionality. In particular the parameter colour had a
great influence on the space perception and the
attractiveness of the interiors.
Subjective perception of the car interior
In the following the results of the questionnaire on the
subjective perception will be displayed. Different scenarios
were compared in order to understand the influence of each
parameter: brightness, position and colour of the lighting.
The significance of the results was assessed using a
Wilcoxon test for two related samples of nonparametric
data. No significant differences originated from differences
in the test persons’ gender or age.
Effects of brightness
The effects of luminance variations were verified by
comparing the following settings: without lighting –
Two particular scenarios were assessed, which provided the
same luminance level and same light positions, but
different colours: orange and blue (scenarios 7 – 12).
It could be verified that the blue lighting appeared brighter
than the orange and facilitated the finding of control
elements, although being uncomfortable (p<0.01). Orange
light colour looked more luxurious and gave a better
quality perception (p<0.05). Few other effects could be told
from the comparison of the mean answers, although they
resulted not significant: blue light allowed a more complete
perception of the car interior and enhanced the orientation,
while orange light had a more pleasant light colour and was
found more appealing.
Effects of Position
Three different lighting positions were evaluated: doors,
centre console and foot space (scenarios 4 – 9 – 11). The
differences between these three scenarios were quite small.
As a trend it can be said that the more peripheral doors
lighting offered a better perception of the whole interior
and a higher perceived value, appeared more comfortable
and pleasant and offered a better orientation. On the other
hand the central illumination of the centre console
facilitated the finding of control elements. The foot space
9
The results obtained from the Self-Assessment-Manikin
test showed two aspects. On one side, there was quite a
wide variance of the answers on the Pleasure and Arousal
axis, this probably due to the different sensations and
feelings which animated the different participants,
independently from the test and the tested scenarios. On the
other side the answers on the Dominance axis concentrated
more on the middle point, this effect explained by the
apparently difficult understanding of this dimension by the
test persons.
Figure 7 Boxplot graph of the distribution of the difference in the
Arousal rating between each scenario and the answer at the
beginning of the experiment.
"Pleasure
In order to understand the change in the emotional state of
the participants, each scenario rating was compared to the
answer given at the beginning of the experiment. The
difference between these two ratings gave a dimension of
the emotional change caused by the scenario
(! " #
$; ! " #
$ ; ! " #
$%& where
$% $% $ are the values gathered at the beginning of the
test).
"Dominance
Effects on Driver!s Emotional State
"Arousal
lighting obtained slightly lower assessments than the other
two illumination places, although the differences were not
significant.
Figure 8 Boxplot graph of the distribution of the difference in the
Dominance rating between each scenario and the answer at the
beginning of the experiment.
Figure 6 Boxplot graph of the distribution of the difference in the
Pleasure rating between each scenario and the answer at the
beginning of the experiment.
The differences distributions are displayed in Figure 6,
Figure 7 and Figure 8. Small changes can be seen in the
dimensions of arousal and dominance, while in the pleasure
dimension the distribution is wider. Though, the median
value, represented in the graphs by the solid middle line,
remains in most cases 0. Moreover, this distribution should
not mislead in finding a negative trend in the influences of
ambient lighting: many test persons judged their state at the
beginning already “happy” (values 1 and 2 on the pleasure
dimension) and therefore there was no room for
improvement in the scenario ratings.
The data were analysed through a Friedman-test with
p=5%. No significant effect could be found on any of the
three dimensions. This has probably been caused by the
short time (3 minutes) in which the participants tested the
light scenario added to the lighting small luminance
(maximum 1 cd/m!) and mostly peripheral position.
10
Effects on Driver!s Performance
During the whole experiment the following data was
collected by the simulator system: elapsed time, car
position (x,y,z), absolute velocity, steering wheel angle,
road curvature, distance from the road’s edge and covered
distance. Every parameter was collected with a frequency
of 25 Hz.
The primary driver’s task was to drive in the middle of the
right lane of a three-lane highway, following another
vehicle. The aim of the task was to focus the driver’s
attention on the street, thus enabling him to perceive
ambient lighting only peripherally or through the secondary
tasks.
These secondary tasks were designed to make the driver
aware of the functionality of ambient lighting, in
recognizing controls and objects inside the car. Without a
proper lighting the test persons could not be able to push
the right button, or find the control for the air nozzle.
Since the test persons could not accelerate and brake, the
only parameter indicative of the driving performance is the
distance from road’s edge (De), measured in meters (Figure
9). Its standard deviation !(De) evaluated over the whole 3
minutes experimental run is indicative of the driver’s
performance in following the street lane in a specific
lighting scenario.
This data (shown in Table 5 and Figure 10) has been
analysed through one-way ANOVA for the lighting
scenarios. The results showed no significant dependency of
the driving performance from the lighting situation in the
car (F= 0.226 "=0.996).
However, since this measure was not the primary goal of
the research, it is difficult to assess its importance. For sure
the driver’s performance has not been influenced either
way by the lighting scenarios.
Table 5 Mean values of !(De) in meters for each lighting
scenario.
Lighting Scenario
Everything on – bright level with accents
(De)
[m]
0.45
Series
0.44
Doors – bright level
0.43
Doors – low level
0.45
Without lighting
0.44
Everything on – bright level
0.46
Everything on – low level
0.43
Everything on – middle level
0.44
Foot space – bright level
0.47
Foot space – low level
0.41
Centre console
0.41
Everything on blue – low level
0.46
CONCLUSIONS
In the following the main conclusions which can be drawn
by this experiment are listed.
Standard Deviation [m]
Distance to road’s edge –
Figure 9 Distance from the edge of the lane, as measured on the
simulator. The measure was taken from the middle of the car
bumper to the virtual white line on the right side of the street.
The presented study showed significant influences of
ambient lighting on driver’s perception. In particular the
advantages of ambient lighting concerning space
perception, functionality and perceived interior quality
were clearly stated, even with low luminance levels. These
advantages do not grow by simply using more brightness or
by employing more light sources.
1
2
3
4 5
6
7
8
9 10 11 12
Scenarios
Figure 10 Values of !(De) in relation to lighting scenarios. With
number 5 is highlighted the scenario without ambient lighting.
•
The whole perception of the car interior is improved
through the use of ambient lighting while driving. It
intensifies the space perception, enhances the perceived
quality of materials and design, facilitates the finding of
controls and the orientation in the car, and gives an
improved perceived safety.
•
A small number of light sources placed in order to
cover the whole field of view can give equal results, in
terms of perceived space and quality, as many
overlapping light sources. Thus an aimed ambient
lighting can use fewer components and reduce the
production costs and though create a welcoming pleasant
atmosphere in the car interior.
11
•
A higher luminance level (mean values of 0.04cd/m!),
while increasing the chance of creating discomfort glare
and distraction during the driving, does not bring
improvements to the driver’s perception of the car interior
or a better orientation and functionality. This means that
darker, less expensive light sources can achieve the same
comfort effects.
•
The influences of different colours affect more criteria
in different way. This has several causes: the diverse field
of view and intensity of perception for each colour in the
mesopic adaptation level (blue is perceived more
intensively and on a wider angle as orange or red), the
various emotional values and the different interaction
with interior materials through reflection. Thus the choice
of colour for ambient lighting has to meet more
requirements, nonetheless brand identity and design
compliance.
•
•
Influences on the emotional state could not be verified,
probably due to the short time available for the evaluation
and the focus that the test-persons gave to the primary
driving task. In other research studies, where the light
stimuli constituted the main focus and the test was longer,
such effects could be verified. Probably in order to
discover more on this particular aspect, a different
experimental design has to be employed.
The driver’s overall performance resulted to be
uninfluenced by the ambient lighting, although this
measure did only assess how the test persons followed the
lane line. No measurements were made on the visual
performances, since these have been already verified in
other studies.
These results can be considered and used in the future
development of such illumination systems, in order to
optimize their design, reducing costs and energy
consumption and though achieving an optimal subjective
perception by the drivers.
On a practical level, from the investigated scenarios a
guideline for developers and manufacturers, suggesting
luminance levels and their tolerance ranges for ambient
lighting systems will be derived.
Further researches should enlarge the spectrum of the
investigated colours, which in this research were limited to
only orange and blue. This comparison alone, although
juxtaposing short wave and long wave colours, cannot
describe completely the possible effects that different
lighting hues have on the driver’s perception of space and
quality. In this perspective also the influence of the interior
materials is important. Indeed, the most part of ambient
lighting comes to the eye after the reflection on completely
different kinds of material (e.g. from black plastics to beige
or white leather). Thus the perceived situation should be
considered not only in function of the lighting colour but
also of the combination lighting-material. This topic is
currently being investigated.
Moreover, dynamic interior lighting changes (in brightness,
position and colour) and their effects have to be
12
investigated. A further step in this direction will be the
connection of these changes with inputs from the
environment, the car and the passengers. This will provide
on one hand adaptation of the interior lighting to the
surrounding conditions and to the vehicle settings,
enhancing safety and possibly giving a visible feedback of
the car status. On the other side, flexibility and compliance
to the customers’ individual tastes will be ensured. The
advantages and problems arising from such systems, as
well as theirs acceptance by the drivers have still to be
tested and verified. Nevertheless, they offer a new,
interesting, emotional and much more coloured way of
understanding and developing vehicle interior lighting.
References
1 Adrian, W. and Stemprok, R. Required Visibility levels
in road Scenes at night time driving. In ISAL 2005 Proceedings of the Conference (Darmstadt 2005),
Herbert Utz Verlag.
2 Damasky, Joachim. Lichttechnische Entwicklung von
Anforderungen an Kraftfahrzeugscheinwerfer. Herbert
Utz Verlag, München, 1995.
3 Devonshire, J. and Flannagan, M. Effect of Automotive
Interior Lighting on Driver Vision. University of
Michigan, 2007.
4 Fischer, Lorenz, Brauns, Dieter, and Belschak, Frank.
Zur Messung von Emotionen in der angewandten
Forschung. Pabst Science Publishers, Lengerich, 2002.
5 Fleischer, Susanne Elisabeth. Die psychologische
Wirkung veränderlicher Kunstlichtsituationen auf den
Menschen. Dissertation, ETH, Zürich, 2001.
6 Greule, R. Emotionale Wirkung von farbiger LEDBeleuchtung im Innenraum. Hamburg, 2007.
7 Grimm, Martin. Requirements for an ambient interior
lighting system for motor vehicles. Herbert Utz Verlag,
München, 2003.
8 Houser, K.W. and Tiller, D.K. Measuring the subjective
response to interior lighting: paired comparisons and
semantic differential scaling. Lighting Research and
Technology, 35, 3 (2003).
9 Huesmann, A., Ehmanns, D., and D., Wisselmann.
Development of ADAS by Means of Driving
Simulation. (Paris 2006), DSC Europe.
10 Ishihara, Shinobu. Tests for colour-blindness. Kanehara
Shuppan Co., Tokyo, Kyoto, 1979.
11 Klinger, Karsten and Lemmer, Uli. Realisierung
ambienter Innenraumbeleuchtungen in Personenwagen
und deren Wirkungen auf den Fahrer. In LICHT
(Ilmenau 2008), LiTG, 534-538.
12 Küller, R., Ballal, S., Laike, T., Mikellides, B., and
Tonello, G. The impact of light and colour on
psychological mood: a cross-cultural study of indoor
work environments. Ergonomics, 49, 14 (2006).
Emotion. Journal of Experimental Psychology, 123, 4
(1994).
13 McCloughan, C.L.B. and Aspinall, P.A. and Webb,
R.S. The impact of lighting on mood. Lighting
Research and Technology, 31 , 3 (1999), 81-88.
18 Veitch, J.A., Newsham, G.R., and Boyce, P.R. and
Jones, C.C. Lighting appraisal, well-being and
performance in open-plan offices: A linked mechanisms
approach. Lighting Research and Technology , 40
(2008), 133–151.
14 Olson, P.L. The effect of vehicle interior lighting
systems on driver sight distance. University of
Michigan, Transportation Research Institute, 1985.
19 Völker, Stephan. Sehen in der Dämmerung - aktuelle
Forschungergebnisse zur Mesopik. In LICHT (Ilmenau
2008), LiTG, 25-31.
15 Schellinger, Sven, Dorit, Franzke, Klinger, Karsten, and
Lemmer, Uli. Advantages of ambient interior lighting
for drivers contrast vision. In Proceedings of the SPIE
6198: Photonics Europe: Photonics in the Automobile,
61980G ( 2006).
20 Wördenweber, Burkard, Wallaschek, Jörg, Boyce,
Peter, and Hoffman, Donald D. Automotive Lighting
and Human Vision. Springer Verlag, Berlin, 2007.
16 Strahlenschutzkommission, Empfehlung. Blendung
durch natürliche und neue künstliche Lichtquellen und
ihre Gefahren. Strahlenschutzkommission, Bonn, 2006.
17 Valdez, P. and Mehrabian, A. Effects of Color on
13
The Effects of Lighting on Atmosphere Perception in
Retail Environments
Pieter Custers
Philips Lighting - GOAL
Mathildelaan 1
5600 JM Eindhoven, The Netherlands
+31 40 27 55654
pieter.custers@philips.com
Yvonne de Kort
Human-Technology Interaction
Eindhoven University of Technology
5600 MB Eindhoven, the Netherlands
+31 40 247 5754
y.a.w.d.kort@tue.nl
Wijnand IJsselsteijn
Human-Technology Interaction
Eindhoven University of Technology
5600 MB Eindhoven, the Netherlands
+31 40 247 4455
w.a.ijsselsteijn@tue.nl
Marike de Kruiff
Creative Director Philips Design
Emmasingel 24, Bldg HWD, P.O. Box 218
5600 MD Eindhoven, The Netherlands
+31 40 27 96291
marike.de.kruiff@philips.com
ABSTRACT
The present study's objective was to investigate the
contribution of lighting in evoking an atmosphere in
naturalistic environments, among the extensive set of other
environmental cues. In a field study involving 57 clothing
stores, lighting attributes (e.g., brightness, contrast, glare
and sparkle) and context (i.e. the shop interior) were
assessed and quantified independently. These data were
then used to predict four dimensions of perceived
atmosphere of these stores in multiple regression analyses.
A hierarchical procedure was chosen, with context
variables entered in the first block and lighting attributes in
the second block. We were thus able to determine the
effects of lighting on perceived atmosphere, while
controlling for context effects. Both lighting attributes and
interior qualities were successfully related to perceived
atmosphere. Our most important finding was that, even
given the substantial contribution of design elements in
retail environments, lighting does play a significant role in
evoking atmospheres.
Keywords
Lighting,
environmental
assessment,
atmosphere
perception, retail environments, Multiple regression, cardsorting
INTRODUCTION
As any light designer, light researcher, and even layperson
will confirm, lighting and ambiance are intimately related.
Literature indicates that lighting characteristics can
influence emotions, mood and cognition, and atmosphere
and spatial impressions, although at times the collected
14
findings are inconclusive. With respect to emotions for
instance, some studies report more pleasant emotions with
higher light intensity levels [1], whereas others report no
significant effects [2,3]. Fleisher et al. [1] demonstrated
that a combination of high illuminance levels and a
relatively large indirect lighting component resulted in
higher feelings of dominance. Cool white light was shown
to be arousing [1], while a more complex pattern emerged
in a second study, reporting positive effects of colour
temperature on male participants’ mood, yet negative
effects on females’ moods [2].
Literature reports of several studies investigating the way
people assess lighting directly. Hawkes, Loe and Rowlands
[4] suggest that people categorize lighting using the
lighting characteristics brightness and interest (or
uniformity). Flynn and colleagues [5] added a third
dimension (overhead – peripheral). Unfortunately, both
studies [4,5] used a sample size too small for a robust factor
analysis. Veitch and Newham [6], who tackled this problem
working with 292 participants, demonstrated that people
categorize lighting in terms of the three dimensions:
brightness, visual attraction, and complexity.
Literature also describes how lighting can affect people’s
environmental impressions (for a review see [7]). As one of
the first, Flynn, Hendrick, Spencer and Martyniuk [5] used
a realistic interior (i.e. conference room) and found an
effect of lighting on subjective evaluations of the
environment, perceptual clarity and spaciousness. This
research, together with several follow-up studies,
summarized in [7], suggests that in the North American
society and culture, there are at least six broad categories of
human impression that can be influenced or modified by
lighting design: perceptual clarity, spaciousness, relaxation
and tension, public versus private space, pleasantness, and
spatial complexity (sometimes liveliness). After relating the
impression dimensions to lighting characteristics, Flynn [7]
suggested several design guidelines: For perceptual clarity,
the designer should apply bright and peripheral lighting. An
impression of spaciousness (i.e., the space is perceived as
large) is achieved when applying uniform and peripheral
lighting. Pleasant and relaxing impressions are the result of
peripheral and non-uniform lighting. And lastly, to
establish a ‘private’ impression, the designer can select
non-uniform and dimmed lighting.
Houser, Tiller, Bernecker and Mistrick [8] varied the
direct/indirect lighting ratio and concluded that walls and
ceiling contribute to the perception of overall brightness
when work plane illuminance is held constant. Also, rooms
appear more spacious with higher ratios of indirect lighting,
and rooms with relatively high levels of indirect lighting
are favoured over light settings with less than 60% indirect
lighting. Literature thus establishes that lighting is able to
influence environmental impressions.
Yet although literature reports of studies indicating that
lighting characteristics influence moods and emotions,
cognition, and environmental impressions, there are hardly
any studies that have established these effects outside the
laboratory. Although it is one thing to prove that variations
in lighting in an otherwise controlled environment have an
impact on environmental impressions, showing that
lighting actually contributes to atmosphere perception in
naturalistic environments, i.e., in the real world is quite
another, let alone ascribing this to specific lighting
attributes. This is exactly what the current study set out to
do. And it did so in a type of environment with substantial
variations in interior design, and where atmosphere has
been proven to matter significantly: retail environments.
Retail Environments
Retail environments communicate the stores’ image and
purpose to customers [9], they can evoke emotional
reactions [10], impact the customers’ ultimate satisfaction
with the service [11], and even the money and time spent in
the store [12]. Therefore, creating the right environmental
setting is of prime importance for shop owners. To create
the desired ambiance, lighting may have its contribution,
but it is only one of the numerous elements, such as
furnishing and finishing of the shop’s interior, size,
crowdedness, and music, that play a role.
Different categorizations for these environmental
characteristics are proposed. Bitner [9] suggested three
groups: ambient conditions; spatial layout and
functionality; and signs, symbols and artefacts. Berman and
Evans [13] included the exterior of the shops and came to
four groups: general interior; the layout and design; the
point-of-purchase and decoration; and the exterior of the
shop. Turley and Milliman [14], in turn, added a fifth
category: human variables. Most recently Baker,
Parasuraman, Grewal & Vos [15] proposed a model in
which the environmental cues were divided into three
categories: design, ambient, and social variables.
Since environments include such an extensive variety of
stimuli, while on the other hand consumers perceive
environments holistically [16] it is essential to seek general
variables as descriptors that grasp the main influence of the
environment [17]. Kaplan [18] suggested that four
environmental dimensions can predict preference for an
outdoor environment: complexity, mystery, coherence and
legibility. Environmental complexity refers to visual
richness, ornamentation, information rate, diversity and
variety in an environment [19], and is shown to have a
linear relationship with interest (arousal) and a curvilinear
(inverted U) relationship with preference (pleasure)
[19,20,21], meaning that moderate levels of complexity are
most preferred. Another important environmental
dimension is order [20], which is related to the extent of
coherence, legibility, organization, and clarity of an
environment [19]. In studies of urban environments
(summarized by Nasar [22]) order has been shown to have
a positive impact on pleasantness and a negative impact on
arousal. Except for the inverted U relationship between
complexity and pleasantness, all these relationships are
confirmed for retail environments [23].
We conclude that lighting has a potential contribution to
perceived ambiance, but is only one of the numerous
elements that may play a role. Our question was whether
lighting would play a role that was measurable, and if yes,
which lighting attributes would have the most substantial
contribution.
METHOD
Design
Fifty-seven clothing stores participated in a field study,
exploring the contribution of lighting to environmental
impressions, controlling for other contextual influences.
For each of these stores the three categories of variables –
perceived atmosphere, lighting attributes, and context (i.e.,
the shop’s interior design) – were assessed and quantified.
Assessments were made independently of each other, by
different groups of experts (lighting) or lay people
(atmosphere, context). We then performed multiple
regression analyses on perceived atmosphere dimensions
with lighting attributes and context as independent
variables.
Participants & Shops
For this field study 57 shops were selected. The stores were
all located in the city centre of Eindhoven, a mid-size
Dutch city, to enable participants and experts to visit all the
shops in one morning or afternoon. In order to prevent
statistical confounds caused by the type of product sold,
15
only fashion shops were selected to participate1. Low and
high-end shops were avoided for the same reason. Within
this selection of shops, which still presented a wide variety
of shop interiors and fittings we expected that structural
confounds between lighting configuration and interior
design would be limited. Nonetheless, in order to control
for this eventuality we also assessed and quantified the
style of the shops’ interiors.
To assess context, i.e., the interior design of the stores,
twenty participants were recruited from a participant
database of the university. The group consisted of ten males
and ten females, ranging in age between 19 and 44, with an
average of 28 years. The respondents were not familiar
with the shops participating in the study.
Seven lighting experts participated in the assessment of the
lighting and lighting fixtures in the stores. Their ages
ranged between 29 and 58, with an average of 46, five were
male and two female.
For quantifying perceived atmosphere, six participants
were recruited from the university’s database. The
participants did not have specific affinity to lighting or the
shops participating in this study. Three participants were
male and three were female. Their ages ranged between 22
and 29, with an average of 24.5 years.
Measurements & Procedure
Context Characterization
A card-sorting experiment was performed to characterize
the shops’ interior designs. Pictures of these interiors were
printed on A5 photo paper and served as cards. The
photographs were all taken inside the shop, from the same
position at which participants rating the atmosphere (see
below) would be standing. In taking the pictures, we
avoided photographing ceilings and lighting fixtures where
possible. Initially two pictures were taken per shop. After a
pilot study we reduced the number of cards to 87, by
removing one picture per shop if both pictures were always
categorised in the same groups. The participants performed
the experiment individually to assure independence of
grouping strategies [24].
totally applicable), based on the chosen quality. This was
repeated, until the participant could not come up with
another discriminating quality.
In total the 20 participants performed 59 categorizations.
Multiple correspondence analysis was then performed on
these data, yielding two dimensions on which the shops
varied (inter-dimensional correlation -.006). We labelled
them ‘legibility’ (order-disorder) and ‘warmth’ (warmcold), based on the labels participants had given for their
categorizations. Each shop’s scores on these dimensions
were used in the multiple regression analyses reported
below, to account for the variability of shop interiors.
Lighting Attributes
A panel of experts assessed the lighting in the shops during
a site visit. For this they used a questionnaire developed
also in cooperation with lighting experts. The questionnaire
consisted of 31 items, probing established lighting
attributes such as brightness, contrast (i.e., uniformity),
colour temperature, glare and sparkle, and modelling, as
well as the relative contribution of different types of
lighting (i.e. general, accent, architectural, decorative) and
the lighting installation (see Table 1). Each of the seven
experts filled out one questionnaire per shop (i.e., 7 times
57 in total) individually. They visited the shops between ten
o’clock in the morning and half past noon, avoiding the
busiest hours. Also, their visits were scheduled within a
period of three weeks, to minimize the chance of interiors
being redecorated. Order effects, e.g. as a result of learning,
tiredness or boredom, were controlled by varying the order
in which each expert visited the stores.
Inter-rater reliabilities were computed to determine the
level of agreement among the experts. Cronbach’s alpha’s
between experts’ scores for each individual item ranged
from .635 to .940, with an average of .804 (see Table 1).
These reliabilities were more than satisfactory, indicating a
high level of agreement among the experts in scoring the
lighting attributes of the shops. The scores of the experts
were averaged to compute each shop’s score.
Participants were instructed to think of a discriminating
quality they felt could serve as a base for sorting the shops,
e.g. ‘cluttered’. They then sorted the pictures of the shops
into five piles2 (ranging from totally not applicable to
1
Since the type of lighting often differs with the type of
product, yet product class may also influence atmosphere
perception, this could result in structural relations
between lighting and ambiance not really attributable to
the lighting per se.
2
Although a division over five piles was desired, the
participants were instructed to first create three piles – not
applicable, neutral or applicable. Then they were asked to
divide the neutral pile into three piles again – less
applicable, neutral or more applicable. This resulted in 5
piles in total. This procedure was followed because the
16
pilot study pointed out that this procedure would lead to
the most evenly spread division of the pictures over the
five piles.
Table 1. Inter-rater reliabilities of lighting questionnaire items
Atmosphere Perception
Item
In the third phase, six (new) participants also visited all the
shops (following different routes, to vary the order in which
shops were assessed) and rated the ambiance in each of
them. For measuring perceived atmosphere a short version
of Vogels’ [25] instrument was used. This questionnaire
measures perceived atmosphere in four dimensions:
cosiness, liveliness, tenseness and detachment. After
deliberation with Vogels, 18 of the original 38 items were
selected (4 or 5 per dimension), with seven-point Likert
scales ranging from totally not applicable to totally
applicable. Participants scored each shop on each of these
items. They were not aware that the study was focused on
lighting and were not instructed to pay particular attention
to lighting or lighting fixtures.
Cronbach’s
alpha
Item
Cronbach’s
alpha
General lighting
.940
Accent lighting
.942
Decorative lighting
.805
Architectural lighting
.933
Brightness
back walls
.870
Brightness
horizontal plane
.823
Brightness ceiling
.820
Brightness floor
.819
Brightness
side walls
.892
Brightness overall
.915
Colour temperature
light
.759
Colour temperature
total space
.813
Glare
.889
Sparkle
.822
Luminance ratio
back walls
.789
Luminance ratio
horizontal plane
.825
Luminance changes
back walls
.691
Luminance changes
horizontal plane
.719
Luminance ratio
ceiling
.635
Luminance ratio
floor
.765
Luminance changes
ceiling
.677
Luminance changes
floor
.638
Luminance ratio
side walls
.816
Luminance ratio
overall
.766
Luminance changes
side walls
.775
Luminance changes
overall
.773
Conspicuous
lighting installation
.628
Patterned
lighting installation
.778
Amount of fittings
.906
Different fittings
.841
Modeling
.865
Mean
.804
Internal consistencies of these atmosphere dimensions were
determined by calculating Cronbach’s alpha for each of the
six participants (see Table 3). Averaged values indicated
acceptable (>.60) to good (>.80) reliabilities. The level of
agreement between participants was determined by
calculating inter-rater reliabilities (Cronbach’s alpha) per
dimension. The values are reported in Table 3. Correlations
between the scores on the different atmosphere factors are
displayed in Table 4.
Table 3. Internal consistencies and inter-rater reliabilities of the
atmosphere scales
Factor analyses (Principal Component with Varimax
rotation) of the data resulted in six dimensions qualifying
attributes of the lighting configuration: contrast, brightness,
glare and sparkle, contrast on the ceiling, aesthetics of
lighting installation, and decorative lighting. The score for
each of the dimensions was determined by averaging the
scores of the items contributing to that particular
dimension. For instance the score for the factor glare was
calculated by averaging the scores for accent lighting, glare
and sparkle. Correlations between the six factors are
reported in Table 2.
Table 2. Lighting attributes correlation matrix
bright
ness
contrast
bright
ness
glare &
sparkle
contrast
of ceiling
lighting
install.
.402
glare &
sparkle
contrast
of ceiling
lighting
install.
decor.
Lighting
.620
-.056
-.092
.089
.399
.165
.206
-.198
-.051
.041
.047
.202
-.111
.043
Each shop’s scores on these lighting attributes were used in
the multiple regression analyses reported below, to account
for the variability of the shop lighting.
Average internal
consistency*
Inter-rater
reliability**
Cosiness
.83
.65
Liveliness
.77
.76
Tenseness
.79
.42
Detachment
.61
.84
*: averaged over 6 participants’ individual internal consistency
scores; **: between the 6 participants’ scores on that dimension.
Table 4. Correlations between scores on atmosphere dimensions
Liveliness
Tenseness
Detachment
Cosiness
.330
-.613
-.309
Liveliness
1.000
-.340
-.789
1.000
.310
Tenseness
RESULTS
Multiple regression analyses were performed predicting
perceived atmosphere dimensions with the two context
variables and the six lighting attributes as predictors. Note
that in these analyses, the 57 shops were the cases (they
made up the rows in the statistical database). Four separate
analyses were performed - one for each atmosphere
dimension.
We first performed multiple regression analyses on
atmosphere dimensions, exploring only lighting attributes
as candidate predictors in a stepwise procedure. The
17
obtained significant beta-weights are displayed in Table 5.
Brightness contributed significantly to three atmosphere
dimensions: cosiness (negatively), tenseness and
detachment. Contrast significantly decreased perceived
tenseness. Glare & sparkle contributed significantly to
liveliness and negatively to detachment.
Table 5. Significant beta coefficients of regression analyses
without context variables
Lighting
characteristics
R"
Brightness
Not controlled for context effects
Cosy
Lively
Tense
Detached
.336**
.312**
.180
.249*
-.588***
.484**
Contrast
.354*
-.362*
Glare & Sparkle
.469**
-.382*
Note: Results of 4 separate regression analyses, with the 4 atmosphere
dimensions as respective dependent variables. N=57. * p<.05, ** p<.01,
*** p<.001
Table 6A. Hierarchical regression predicting cosiness
Cosiness
! coefficients
Step 1
Warm
Table 6C. Hierarchical regression predicting tenseness
R" change
Tenseness
Legibility
.246
Warm
.384**
Block 2
(lighting)
.279 **
.058
.119
.051
-.212
-.116
.189
Block 2 (lighting)
.445 *
Glare & sparkle
.043
Contrast of the
ceiling
-.059
-.499 **
Glare & Sparkle
-.007
Contrast of
ceiling
-.206
.039
Lighting
installation
-.157
Lighting
installation
-.153
Decorative
lighting
.102
Decorative
lighting
Note: * p<.05, ** p<.01, *** p<.001
Note: * p<.05, ** p<.01, *** p<.001
Table 6D. Hierarchical regression predicting detachment
Table 6B. Hierarchical regression predicting liveliness
! coefficients
Step 1
R"
Detachment
R" change
Step 2
Legibility
-.590 ***
-.496 ***
Warm
-.247 *
-.146
Contrast
.093
Brightness
-.128
Step 1
.115
.806 ***
.765 ***
Warm
.056
.033
Brightness
R" change
.652 ***
Legibility
Contrast
R"
Step 2
.682 ***
Block 2 (lighting)
.522 ***
Block 2 (lighting)
! coefficients
Block 1 (context)
.407 ***
Block 1 (context)
.013
.170
Glare & sparkle
-.175
Glare & sparkle
.293 *
-.123
Contrast of the
ceiling
-.003
Contrast of the
ceiling
.158
Lighting
installation
-.064
Lighting
installation
-.026
Decorative
lighting
.033
Decorative
lighting
18
.130
-.298
Brightness
Brightness
Liveliness
R" change
.059
Contrast
Contrast
R"
Step 2
Block 1 (context)
-.132
.281 *
! coefficients
Step 1
.105
-.158
We then repeated the analyses, yet this time controlling for
contextual variables. A hierarchical procedure was chosen,
with context descriptors comprising the first block and
lighting attributes the second block. We could thus
determine the effects of lighting on perceived atmosphere
while controlling for context effects. In the first block,
context variables were entered (Table 6). Adding the
lighting attributes after this first block generally improved
the predicted variance. Moreover, for three atmosphere
dimensions, at least one lighting attribute had a significant
beta-weight. Brightness significantly and substantially
decreased perceived cosiness, and increased perceived
tenseness. Glare and sparkle contributed to the perceived
liveliness of fashion stores. Furthermore, the shops’
legibility was shown to significantly decrease perceived
liveliness and increase perceived detachment.
Step 2
Block 1 (context)
Legibility
R"
Controlled Regression Analyses
.030
DISCUSSION
Light and ambiance are intimately related, yet we know of
very few studies that have attempted to measure how much
lighting actually contributes to atmosphere perception in
naturalistic environments. The current study attempted to
do just that. Also, we hoped to attribute any contribution
we might find to more or less specific lighting attributes.
And indeed we did manage to verify that lighting
contributes a measurable part to atmosphere assessments.
This contribution was modest, and we did not establish
significant effects for each dimension of atmosphere, but in
view of the challenges we met, our findings were certainly
satisfactory.
Measuring light’s contribution in naturalistic settings
proved to be quite a complex exercise. For one, one is
dependent on the natural range and variance of lighting
used in ‘real’ settings, and has to find a way of categorising
or even quantifying that. In the current study, experts
scored the lighting in each of the 57 shops, using a
questionnaire specially developed to this end. Inter-rater
reliabilities between these experts indicated that this
produced a reliable and robust measure, which was more
detailed and comprehensive than what could have
realistically been possible with objective measurements.
A second obstacle in natural settings is accounting for the
substantial variance and contribution of intervening
variables. Based on the literature, we expected that
especially the shop’s interior and social variables would
play an important role in defining the atmosphere. The
social setting we tried to control by selecting time slots that
were not too long and avoided the busiest hours. The
shops’ interiors were controlled first by limiting them to a
certain type of product (clothing) and excluding the
extreme ends of the price levels. Second, since this still left
us with a huge range of different interiors – e.g. cluttered to
spacious, old-fashioned to trendy, warm wooden furniture
to cool metal racks and stands – we made an attempt to
characterise and quantify these interior styles using the
card-sorting method. These data enabled us to characterise
all 57 shops by their location in a two-dimensional space
stretching from orderly to disorderly and from warm to
cold. We were not able to control the soundscapes (e.g., the
music playing in the shop) or the shops’ exteriors.
A third obstacle in the present research was measuring
ambiance or atmosphere. We were not aware of existing
standardized instruments for measuring atmosphere in retail
environments, or other types of environments for that
matter. Instruments most often used are probably the sets of
semantic differentials, similar to the one we used in the
present study. We preferred this measure [25] to other ones,
for instance the well-known set developed by Russell,
Mehrabian, and colleagues (e.g., see [17]), since it was
specifically targeted to atmosphere perception, and its
dimensions appeared closer to what we intended to measure
than the dimensions typically coming from those sets
(generally something like evaluation, arousal and potency).
The current instrument worked well in terms of the internal
consistencies of its subscales, yet in hindsight it does not
necessarily cover all relevant aspects of atmosphere. Also,
it could have been interesting to also have probed
characteristics such as ‘spaciousness’ or ‘perceptual clarity’
directly. This would have made it easier to compare the
present study’s findings to those reviewed earlier, for
instance by Flynn [7]. However, we felt the current
measure was closer to the ‘atmosphere’ concept, and we
had to restrict the number of items, since each participant
would have to fill out the questionnaire 57 times (!), one for
every shop.
However, we feel that with these 57 shops, we have
managed to create a large enough sample to guarantee a
good variance in our core dependent and independent
variables: lighting attributes and atmospheres, and to
perform the multiple regression analyses on. We were in
fact quite happy and proud to have been able to recruit that
many shops to participate in the study. This potentially also
illustrates the interest of these shops’ owners in the role that
lighting plays in the success of their business.
The first set of regression analyses showed how several
lighting attributes were related to atmosphere dimensions.
The most important attributes were brightness, contrast,
and glare and sparkle. At least one, and sometimes two of
these attributes significantly predicted each of the four
dimensions.
In the second set of regression analyses, context variables
were entered first, before entering the lighting attributes.
This way we minimised the chance of confounds caused by
naturally occurring relationships between interior design
and lighting attributes, which might otherwise lead us to
overestimate light’s contribution to atmosphere perception.
In fact, since the lighting in the shops was also recorded on
the photographs used for the context quantifications, the
present results are probably an underestimation of the
impact of the lighting on perceived atmosphere.
Although some correlations decreased or disappeared,
others remained, showing a consistent contribution for
instance of brightness to the cosy-dimension (the brighter
the impression of the shop, the less confined/intimate/
romantic/relaxing was the atmosphere). Glare and sparkle
added most to liveliness (the more glare and/or sparkle, the
more energising/lively/stimulating was the atmosphere).
Brightness contributed positively to the tenseness
dimension (the more brightness, the more threatening,
tense, uneasy and unfriendly the atmosphere). This was in
fact quite unexpected, and not in line with earlier findings,
which generally relate brightness to more positive
evaluations. This may be specific to this type of
environment and definitely calls for more research. No
specific lighting attribute was related to detachment. This
dimension was largely predicted by the contextual variable
‘legibility’ (running from disorder to order). The more
legible the environment was, the more formal and
businesslike the atmosphere. This same legibility
19
characteristic contributed negatively to the liveliness of the
shop.
Conclusion
This study provides a better understanding of the impact of
lighting on perceived atmosphere in a retail environment.
Lighting attributes and interior qualities were successfully
related to perceived atmosphere. Granted, the amounts of
variance predicted for each of the dimensions of
atmosphere are generally modest, and typically only one of
the lighting attributes had a significant individual
contribution. However, considering the wide variety of
shop interiors, clothing collections, music played et cetera,
we nonetheless consider the findings striking and
encouraging for light designers and researchers: even in the
enormous set of visual environmental cues present in retail
environments, lighting does play a significant role in
creating an ambiance.
ACKNOWLEDGMENTS
We thank the lighting experts of Philips Lighting for
cooperating in developing and performing the lighting
questionnaire.
REFERENCES
1. Fleischer, S., Krueger, H., & Schierz C. (2001). Effect
of brightness distribution and light colours on office
staff, The 9th European Lighting Conference
Proceeding Book of Lux Europa, 76-80, Reykjavik.
2. Knez, I. (1995). Effects of indoor lighting on mood and
cognition. Journal of Environmental Psychology, 15,
39-51.
3. Hygge, S., & Knez, I. (2001). Effects of noise, heat and
indoor lighting on affect and cognitive performance.
Journal of Environmental Psychology, 21, 291-299.
4. Hawkes, R. J., Loe, D. L., & Rowlands, E. (1979). A
note towards the understanding of lighting quality.
Journal of the Illuminating Engineering Society, 8, 111120.
5. Flynn, J.E., Hendrick, C., Spencer, T.J., & Martyniuk,
O. (1973). Interim study of procedures for investigating
the effect of light on impression and behavior. Journal
of the Illuminating Engineering Society, 3, 87 – 94.
6. Veitch, J. A., & Newsham, G. R. (1998). Lighting
quality and energy-efficiency effects on task
performance, mood, health, satisfaction and comfort.
Journal of the Illuminating Engineering Society, 27,
107-129.
7. Flynn, J.E. (1992). Lighting-design decisions as
interventions in human visual space. In J.L. Nasar
(Eds.), Environmental aesthetics: theory, research and
applications (pp 156-170). New York: Cambridge
University Press.
20
8. Houser, K.W., Tiller, D.K., Bernecker, C.A., &
Mistrick, R.G. (2002). The subjective response to linear
fluorescent direct/indirect lighting systems. Lighting
Research and Technology, 34. 243–264.
9. Bitner, M.J. (1992). Servicescapes: The Impact of
Physical Surroundings on Customers and Employees.
Journal of Marketing, 56, 57–71.
10. Donovan, R.J. & Rossiter, J.R. (1982). Store
atmosphere: An experimental psychology approach.
Journal of Retailing, 58, 34-57.
11. Bitner, M.J. (1990). Evaluating Service Encounters: The
Effects of Physical Surroundings and Employee
Responses. Journal of Marketing, 54, 69–82.
12. Donovan, R. J., Rossiter, J.R., Marcoolyn, G., &
Nesdale, A. (1994). Store Atmosphere and Purchasing
Behavior. Journal of Retailing, 70, 283–294.
13. Berman, B. & Evans, J.R. (1995).Retail Management: A
Strategic Approach (6th ed), NJ: Prentice Hall
International.
14. Turley, L.W. & Milliman, R.E. (2000), Atmospheric
Effects on Shopping Behavior. A Review of the
Experimental Evidence. Journal of Business Research,
49, 193-211.
15. Baker, J., Parasuraman, A., Grewal, D. & Vos, G.B.
(2002). The Influence of Multiple Store Environment
Cues on Perceived Merchandise Value and Patronage
Intentions. Journal of Marketing, 66, 120-141.
16. Mattila, A.S. and Wirtz, J. (2001). Congruency of scent
and music as a driver of in-store evaluations and
behavior, Journal of Retailing, 77, 273–89.
17. Russell, J.A., & Mehrabian, A. (1976). Environmental
variables in consumer research, Journal of Consumer
Research, 3, 62–3.
18. Kaplan, S. (1987). Aesthetics, Affect, and Cognition.
Environment and Behavior, 19, 3–32.
19. Nasar, J.L. (2000). The evaluative image of places. In
W.B. Walsh, K.H. Craik, & R.H. Price (Eds.), PersonEnvironment Psychology (2nd ed.), Mahwah, NJ:
Lawrence Erlbaum Associates.
20. Berlyne, D.E. (1971). Aesthetics and Psychobiology,
New York: Appleton-Century-Crofts.
21. Ulrich, R.S. (1983). Aesthetics and affective response to
natural environment. In I. Altman & J.F. Wohlwill
(eds). Behavior and the Natural Environment. Human
Behavior and the Environment, Advances in Theory and
Research, Volume 6, New York: Plenum.
22. Nasar, J.L. (1997). New developments in aesthetics for
urban design. In G.T. Moore & R.W. Marans (Eds.),
Advances in Environment, Behavior and Design, 4,
New York: Plenum Press.
23. Gilboa S., & Rafaeli, A. (2003), Store environment,
emotions
and
approach
behavior:
applying
environmental aesthetics to retailing. The International
Review of Retail, Distribution and Consumer Research,
13:2, 195 – 211.
24. Kidwell, D.K. & Martin, S. (2001). A Case Study in
Cluster Analysis for Intranet Organization. Engineering
Management for Applied Technology, emat, 57-64.
atmosphere. In: J.H.D.H. Westerink, M. Ouwerkerk,
T.J.M. Overbeek, W.F. Pasveer, & B. Ruyter (Eds.),
Probing Experiences: From Assessment of User
Emotions and Behavior to Development of Products).
Dordrecht: Springer. pp. 25-41.
25. Vogels, I.M.L.C. (2008). Atmospheric Metrics:
Development of a tool to quantify experienced
21
Effect of Lamp Spectrum on
Perception of Comfort and Safety
Colette Knight
Philips Lighting B.V.
5600 JM Eindhoven, The Netherlands
+31 4027 57160
colette.knight@philips.com
ABSTRACT
In addition to improving visibility and providing
orientation, public lighting is expected to contribute to the
perception of comfort and safety of people outside after
dark. At present, high-pressure sodium (HPS) lamps are
widely used in outdoor applications due to their high
efficacy and reliable lifetime. Their use however, comes at
the expense of good color rendering and accurate color
appearance. Recently developed ceramic metal halide
(CMH) lamps provide many of the advantages of HPS in
addition to natural white light and significantly better color
rendering.
In this paper, results of quantitative research conducted in
three European countries on the effect of lamp spectrum on
visual performance and the perception of safety and
comfort outdoors are presented. The results consistently
show that at comparable light levels, the same people
perceive areas illuminated with high quality white light to
be brighter, safer and more comfortable than the same
neighborhood illuminated with yellowish high-pressure
sodium lighting.
Keywords
Perception of safety and comfort, outdoor lighting, street
lighting, white light
INTRODUCTION
Artificial outdoor lighting can play several important roles.
In addition to enabling safe movement, improving visibility
and providing orientation, public lighting is increasingly
used to contribute to the perception of safety and comfort
of people outside after dark. The perception of safety,
comfort and appreciation of an outdoor area can be strongly
influenced by the lighting used to illuminate it.
Without conditions that ensure safe movement, it would not
be possible for people to walk on the street, and without
conditions that ensure a general perception of safety,
people might choose not to walk on the streets. Factors
contributing to safe movement after dark include visual
orientation and the ability to detect obstacles on the
pavement which may otherwise be a trip hazard. Factors
contributing to the perception of safety include absence of
glare, perception of brightness in the area and the ability to
recognise the expression or faces of other road users at a
22
distance sufficient to take avoiding action if necessary.
Previous investigations have suggested that people want to
be able to recognize strangers from a distance of 4 m in
order to feel comfortable [1]. However it is extremely
likely that this “comfort zone” distance varies significantly
from one person to another and also depending on the
familiarity of the environment. Improving the distance for
and ease of facial recognition might contribute to
increasing the feeling of safety and security of pedestrians
and especially for those who feel most vulnerable. There is
certainly interaction between these factors. In general a
lighting scheme designed to meet one of these needs, such
as recognition of faces and expressions may well go some
way to meeting all of them [2].
At present, high-pressure sodium (HPS) lamps are widely
used in outdoor applications due to their high efficacy and
reliable lifetime. Their use however, comes at the expense
of good color rendering (CRI of HPS ~25) and accurate
color appearance. Recently developed ceramic metal halide
(CMH) lamps provide many of the advantages of HPS in
addition to natural white light and significantly better color
rendering (CRI > 60). Related benefits of these lamps for
the residents and pedestrians in the areas illuminated by
them might include greater ease of facial recognition and
color identification. Indeed, an earlier laboratory study
conducted by Raynham et al. [3] concluded that twice the
illuminance level of HPS is required to achieve the same
facial recognition distance as with white compact
fluorescent light sources at typical nighttime outdoor
lighting levels. The advantages of high quality white light
for facial recognition is already taken advantage of in the
British standard for road lighting, BS5489:2003, which
allows a lower lighting level to be used in residential areas
if the color rendering index (CRI) of the source used is over
60 [4,5]. Color provides important visual information.
Color differentiation and identification can contribute to
one’s ability to recognize faces or identify one’s car, for
example. Moreover, in the case of reporting a criminal act,
accurate color naming can provide key information about
the color of the suspected person’s clothing or automobile.
Research conducted by Boyce et al. in New York City and
Albany, NY suggests that there is a link in the public mind
between the perception of safety of an area after dark and
the perception of brightness of that area [6]. Of course the
perception of safety in an area depends on many factors
which are not related to lighting. Nevertheless, there is a
need for residential areas to appear appropriately brightly
illuminated at night to support the perceived safety of
people in the area at night.
Fotios and Cheal used brightness ratings, brightness
rankings and brightness matching to compare the effect of
lamp spectrum on the perceived brightness in a variety of
laboratory tests. Their results showed that at equal
illuminance, lighting from white metal halide (MH) and
compact fluorescent light sources were perceived to be
significantly brighter than from yellowish HPS. Moreover,
they found that at the typical illuminance levels
encountered on urban streets (2 – 15 Lux), the same
perception of brightness was achieved when the
illuminance ratio of metal halide to HPS (MH/HPS) was
~0.73 [7,8]. These results were consistent with early
laboratory studies conducted by Rea et al. in which subjects
were asked to adjust the illuminance on a scale model scene
illuminated with a HPS source until it matched the
brightness of the same scene illuminated by a MH source.
At illuminance of 0.1 and 1 cd/m2, the illuminance ratio
(MH/HPS) found to achieve an equal perception of
brightness was 0.71 [9]. This means that people perceived
scenes illuminated with metal halide sources to be equally
bright as scenes illuminated with HPS sources when the
measured illuminance was ~29% lower for the MH scene.
In more recent field tests conducted by Rea et al.[10],
respondents stood in the middle of a street between two
luminaires and compared the perception of brightness of
opposite ends of the street by alternatively looking at the
street scenes illuminated by each luminaire. Subjects
compared a variety of scenes where one part of the test
street was illuminated with HPS at levels between ~5 – ~15
Lux and the opposing direction of the street was
illuminated with CMH source also between ~5 – ~15 Lux.
Subjects were given written questionnaires and for each
pair of lighting conditions, they were asked to make a
forced choice for the lighting condition, under which they
would feel safer to walk at night and under which the street
and surroundings as well as objects placed on the pavement
appeared brighter. The test included pairs of lighting
conditions where the ratios of illuminance on the scene
illuminated with CMH to the illuminance on the scene
illuminated with HPS (CMH/HPS) varied between 0.33 –
3. Interpolation of the results suggested that an illuminance
ratio of CMH/HPS of 0.79 was required to create an equal
perception of brightness and a ratio of 0.66 was required to
create an equal perception of safety [10]. This opens up the
opportunity to maintain the same perception of safety with
CMH lamps while reducing the light level.
In this paper, results of field tests conducted in actual urban
streets in the Netherlands, Spain and the United Kingdom
on the effect of lamp spectrum on the perception of safety
and comfort are presented. The goal of the research was to
determine how end-users evaluate the outdoor lighting in
their neighborhoods before and after it was changed from
yellow high pressure sodium (~2000K) to warm white
CDO 2800K or neutral white CDO 4200K street lighting
and vice versa, as well as how this change affected their
perception of safety and comfort and their appreciation of
the neighborhood. At the same time, objective
measurements of the performance for facial recognition and
color identification were compared under yellow and warm
or neutral white light. Altogether, over 300 residents
participated in the experiments under both yellow and
white light.
TECHNICAL PROPERTIES OF LAMPS USED
The lamps used in the experiments were based on high
pressure sodium (HPS) and ceramic metal halide (CMH)
technologies. Some properties of the lamps are listed in
Table 1.
Table 1: Correlated color temperature (CCT) and color
rendering index (CRI) of HPS and CMH sources used in
experiments
Technology
Commercial Name
CCT (K)
CRI
HPS
SON T
2000
25
CMH
Master City White 2800
CDO-TT 2800K
2800
83
CMH
Master City White 4200
CDO-TT 4200K
4200
90
Figure 1: Spectrum of SON (based on HPS technology)
and Master City White 2800K (CMH technology)
RESEARCH SET-UP
Research was conducted in Eindhoven, NL, Navalcarnero,
Spain and St. Helens, UK by IPM International as part of a
large quantitative study commissioned by Philips Lighting
to evaluate how residents experienced the street lighting in
their neighborhoods before and after it is changed from
yellow HPS (~2000K) to warm white (~2800K) light as
well as how this change affects their perception of safety
and comfort and their appreciation of their neighborhoods.
As evident from Table 2, the tests in the UK were
conducted after those in the NL and in Spain. Additional
tests were conducted in different streets in St. Helens, UK,
where the lighting was changed from (1) HPS to neutral
white light (CDO 4200K), (2) from neutral white light
(CDO 4200K) to warm white light (CDO 2800K) and (3)
from warm white light (CDO 2800K) to HPS. The latter
23
was done to check whether or not changes seen were due to
the fact that residents expected certain changes due to a
change in lighting. In the UK, each participant performed
the test under both lighting conditions in one area. Different
participants performed the tests in the different areas.
The people responsible for public lighting in the respective
cities identified possible locations where the lighting could
be changed according to the research schedule. One of our
requirements of the test areas was that they were safe. This
was necessary to ensure that the researchers could conduct
interviews and tests at night with minimal risk. The
residents were sent or shown a letter informing them that
tests were being conducted to evaluate the perception of
safety of the area after dark. There was no mention of
lighting or the commissioner (i.e. Philips) in the letters.
The number of different respondents who participated in
the tests in each area is shown in Table 2. The respondents
were recruited from people living in the vicinity, but not in
the actual streets in the experimental area. The split over
gender and age group (below and above 40 years) is shown
in table 3. One of the recruitment criteria was that the
respondents walked or biked outside after dark at least
three times a week.
The test involved individual face-to-face interviews during
which a detailed questionnaire was filled in. In addition,
objective measurements of visual performance were
conducted. Each test lasted ~45 minutes. A mixed
research design was used in Eindhoven and Spain,
meaning that some of the respondents (55 and 60 in the
case of the Netherlands and Spain respectively) participated
in the test both under the initial lighting condition as well
as after the lighting had been changed (i.e. “before and
after”) while others only participated in the subjective
evaluations after the lighting had been changed (see Table
2). This mixed design enabled the detection of artifacts
since people might become more sensitive to lighting after
they have been interviewed about it the first time. The
lighting was changed soon after the first set of interviews
(“before” interviews) were completed and the “after”
interviews were started at least 3 weeks after installation of
the new lighting. There was no extra maintenance (e.g.
cleaning) when the lamps were changed.
During the face-to-face interviews, the respondents were
asked to
1.
2.
24
Rate their perception of safety and comfort in the
test area on a 5-pt scale
Rate the importance of street lighting to their
perception of safety and comfort
3.
List the most important aspects of lighting for
them and to evaluate the street lighting in the test
area against these and other aspects
Subsequently, the respondents were explicitly asked to
4.
Make various comparisons using a 7-point scale
with respect to the previous lighting condition.
Visual performance was evaluated on the basis of the
distance to recognize faces and colors. During the facial
recognition test, the researchers stood with their back
towards the closest pole so that the picture was only
illuminated from the distant neighboring pole. The
researchers held pictures with the faces of well-known
personalities for the particular country in front of
themselves. The pictures were printed on non-glossy A4
paper so that the size of the face was approximately lifesized. A total of 8 different pictures were used. The
pictures were divided into two groups of 4 pictures. Half of
the residents were shown one group of 4 pictures under the
initial lighting, and the second group of pictures after the
lighting had been changed.
The other half of the
participants was shown the pictures under the reverse
lighting conditions so that all pictures were observed under
both lighting conditions. The order of the 4 pictures shown
was randomized among the different participants. Only the
respondents in the “before + after” group did the facial
recognition test, and a within-subject analysis was done to
compare the performance under the different light sources.
The poles used and the position at which the researcher
stood relative to the pole was chosen under the initial
lighting. The vertical illuminance at the position of the
pictures was measured under both lighting conditions.
The protocol for the facial recognition test was as follows.
As shown schematically in figure 2, the test person started
walking slowly from a distance of ~15 m towards the
researcher holding the picture. They were instructed to
stop and say as soon as they were close enough to
1.
2.
Identify the gender of the person on the picture
See the picture well enough to guess the identity
of the person and
3. See the person well enough to be sure of their
identity
It was stressed that the focus was on seeing the picture well
enough to guess or be sure of the identity of the person on
the picture even if the respondent did not know the person
or remember their name.
All three distances were
recorded.
Figure 2: Schematic of set-up used in facial recognition and color identification tests
Table 2: Summary of the number of respondents and timing of evaluations done in 3 European cities
Location
Eindhoven,
NL
Navalcarnero,
Spain
St. Helens,
UK
Installed Lamps and Test Dates
Nr. different Initial Lighting Test Date 1
Lighting after Test Date 2
respondents Condition
Initial
Lamp Change
Condition
55
56
60
60
SON
March 2006
SON
April 2007
30
SON
33
SON
31
CDO 2800K
31
CDO 4200K
November
‘08
November
‘08
November
‘08
November
‘08
Evaluations Done
Comparison
Subjective
of Visual
Evaluation
Performance
CDO 2800K
CDO 2800K
CDO 2800K
CDO 2800K
April 2006
April 2006
May 2007
May 2007
!
CDO 2800K
January ‘09
!
!
CDO 4200K
January ‘09
!
!
SON
January ‘09
!
!
CDO 2800K
January ‘09
!
!
!
!
!
!
!
Table 3: Summary showing split over gender and age group
" 40
> 40
Eindhoven, NL
(n=111)
% male
% female
15
12
50
24
Navalcarnero, Spain
(n=120)
% male
% female
24
27
27
20
St. Helens, UK
(n=125)
% male
% female
21
20
24
35
All 3 countries
(n=356)
% male
% female
20
20
33
26
RESULTS
The results were analyzed separately for each country. In
all three countries, the most important aspect of street
lighting given by the respondents was the “brightness” of
the illuminated area. Consistent with studies referenced in
the introduction, a higher perceived brightness of the street
and pavement contributes to a higher perception of safety.
Since one of the requirements of the areas chosen was that
it was safe, it is not surprising that independent of the
lighting, people felt relatively safe in all test areas.
As illustrated by the histograms of the results from the UK,
substantially more people felt very comfortable when the
same area was illuminated with warm or neutral white light
compared to with SON (see figure 3). This trend was also
seen in Eindhoven and in Navalcarnero. The mean and
standard error of the mean is written next to the plots in
figure 3 and also by similar plots in later figures.
Table 4 summarizes how respondents in the UK answered
various questions on a 5-point scale regarding their
perception of comfort in the area, the quality of the lighting
and the effect of the street lighting on their perception of
safety and brightness of the area. The mean for the above
evaluations are given. Paired-sample T-tests (confidence
interval 95%) were used to evaluate if the mean of the
ratings were different or not under the first and second
lighting condition. There is a difference when the
corresponding value in Sig.(1-2) column in Table 4 is less
than 0,05.
25
The respondents were asked the same questions under both
lighting conditions. Moreover, at the point where they
were asked these questions under condition 1, there was no
discussion that the lighting would be changed and under
condition 2, there was no mention that the lighting had
been changed.
As seen in table 4, there were no statistically significant
differences regarding how the same people rated the area
and effect of the street lighting on their perception of safety
when the lighting was changed from CDO 4200K to CDO
2800K. However, when the lights were changed from SON
to either CDO 2800K or CDO 4200K, the perception of
safety, comfort, brightness and light quality was improved.
Question: How do you feel about the area here? After
sunset, please rate how you feel on a 5 point scale from
very comfortable/very much at ease to very uncomfortable/
very uneasy. very comfortable = 1, very uncomfortable = 5
Figure 3: Plots showing how respondents in St. Helens,
UK rated the perception of comfort before and after the
lighting had been changed.
When the lights were changed from CDO 2800K to SON,
there was a statistically significant reduction in the rated
light quality, brightness of the area and the effect of
lighting on the perception of safety. In particular, the
brightness of the area was rated to be “just right” with CDO
2800K and CDO 4200K, whereas it was rated to be “too
dark” with SON. Even when the lighting in Hereford Close
was changed from CDO 2800K to new SON lamps, the
area with the new SON lamps was evaluated to be “too
dark” (table 4). Even though a statistically significant
difference in the perception of comfort was found when the
lighting was changed from SON to CDO 2800K, there was
no statistically significant difference found regarding the
perception of comfort of the area when the reverse change
was made (i.e. from CDO 2800K to SON). This might
suggests that there is an enhancement in the subjective
ratings after changing the street lighting. Nevertheless, the
results taken together consistently indicate that for
pedestrians, streets illuminated with white light are
perceived to be brighter and safer and at least equal but
often more comfortable than the same streets with SON.
Table 4: Summary of ratings for different lighting conditions in St. Helens
Questions:
How do you feel about this area here? After
sunset, do you feel:
very comfortable / at ease (1) ! very
uncomfortable / uneasy (5)?
Cond. 1 ! Cond. 2
SON ! CDO 2800K
SON ! CDO 4200K
CDO 2800K ! SON
CDO 4200K ! CDO 2800K
Now I would like you to tell me what you think of SON ! CDO 2800K
the lighting in terms of its quality: By quality I
SON ! CDO 4200K
mean nice light, good color. Do you feel that it is: CDO 2800K ! SON
1 (very pleasant) ! 5 (very unpleasant)
CDO 4200K ! CDO 2800K
And now I would like to know whether the
lighting here makes you feel safe or not. Does it
make you feel: 1 (very safe) ! 5 (very unsafe)
SON ! CDO 2800K
SON ! CDO 4200K
CDO 2800K ! SON
CDO 4200K ! CDO 2800K
SON ! CDO 2800K
And how do you rate the brightness of the area.
For you personally, is it: 1 (much too bright) ! 5 SON ! CDO 4200K
(much too dark). 3 = just right.
CDO 2800K ! SON
CDO 4200K ! CDO 2800K
26
Mean Rating
Cond. 1 Cond. 2
1,93
1,37
1,91
1,39
1,61
1,68
1,74
1,65
2,40
1,67
2,61
1,30
1,94
2,58
1,77
1,94
2,20
1,33
2,06
1,33
1,52
2,06
1,65
1,68
3,38
2,90
3,34
3,00
3,00
3,61
2,97
3,03
Diff. (1-2)
0,567
0,515
-0,07
0,097
0,733
1,303
-0,645
-0,161
0,867
0,727
-0,548
-0,032
0,483
0,345
-0,613
-0,065
Sig. (1-2)
0,001
0,000
0,861
0,374
0,000
0,000
0,009
0,258
0,000
0,000
0,003
0,572
0,000
0,016
0,000
0,489
After answering the questions shown in table 4, the
respondents were asked during the second condition if they
had noticed any recent changes in the test area. As
mentioned in the research set-up, the respondents in general
did not live in the streets where the lighting had been
changed, but in the vicinity. About 50% of the Dutch
respondents in the “before + after” group spontaneously
mentioned the street lighting had been changed, as did
about 40% in the Dutch “after only” group.
By
comparison, in Spain, ~77% of the respondents in the
“before + after” group and ~50% of the respondents in the
“after only” group spontaneously mentioned that the street
lighting had been changed. In St. Helens, ~85% of the
respondents noticed the change from SON to CDO and
~55% still noticed the change from CDO 4200K to CDO
2800 K. When triggered to look at the street lighting, the
majority of respondents, including those in the “after only”
groups in the Netherlands and Spain who had not
spontaneously mentioned the street lighting, eventually
reported that the color of the street lights had changed or
that brighter street lights had been installed. Since the
street lighting in the Netherlands and Spain was changed on
a commonly used connecting road in the residential area,
even those respondents who did not do the test under the
first condition (i.e. “after only” group) were familiar with
the test area. In the UK, where smaller residential streets
were used, all of the respondents did the test under the
initial as well as the second lighting condition. Thus the
vast majority of respondents could make an evaluation as to
whether or not they felt equally comfortable (or safe etc.),
or less or more so than before. The comparison was done
using a 7-point scale, where “no difference” was assigned a
value of 4. A one-sample T-test was used to check the
difference between the mean of the distribution and the test
value “4” (no difference). In figures 4 – 6, results of some
of the responses are graphically shown and in Table 5, a
wide range of data is summarized.
Question: How does the present lighting compare with the
lighting before? Does it make you feel much more safe, the
same or much less safe?
( SON !CDO 2800K, CDO is the “present” lighting)
Figure 5: Plots showing how respondents in Eindhoven,
NL compare the perception of safety in the test
neighborhoods after the lighting had been changed to CDO.
much safer = 1, the same = 4, much less safe = 7
Question: How does the present lighting compare with the
lighting before in terms of brightness, quantity of light?
Figure 6: Plots showing how respondents in St. Helens,
UK compare the brightness in the test neighborhoods after
the lighting had been changed as shown on the plots.
much brighter = 1, the same = 4, much less bright = 7
Question: How comfortable is the current street lighting,
compared to the street lighting before?
( SON !CDO 2800K, CDO is the “current” lighting)
much more comfortable = 1, the same = 4, much less = 7
Figure 4: Plots showing how respondents in Navalcarnero,
Spain, compare the perception of comfort after the lighting
was changed to CDO. Mean and std. error of mean is listed.
When specifically asked to compare the lighting, the
majority of respondents in all three countries rated the
white street lighting to be equally or more comfortable than
the yellowish SON street lighting.
There were no
statistically significant differences between the “before +
after” and the “after only” group in either the Netherlands
or Spain. The most common reasons given for the
increased comfort were related to the ability to see clearer,
better and further. The reason most often given by the few
27
respondents who rated the area to be less comfortable after
the change to white light was that the area was too brightly
lit in their opinion.
Moreover, when the street lighting is changed from
yellowish HPS to warm or neutral white CDO, the
perception of safety is significantly improved. In the test
street in St. Helens, where the lighting was changed from
warm white CDO to yellowish SON, the majority of
respondents did not report any change in the perceived
safety in the area – albeit a few more respondents reported
that it had deteriorated. The fact that there was no
significant difference in this “before vs. after” test when
white CDO was offered first and yellowish HPS offered
subsequently might indicate that there is a positive
enhancement in the rating of the second lighting condition
since respondents might automatically expect an
improvement when street lighting is changed. Nevertheless
even with this “expected improvement”, yellow HPS is not
rated better than white light regarding the ambience of
perceived safety created. No difference was found between
the perceived safety under warm and neutral white light in
St. Helens. CDO 4200K was not evaluated in the
Netherlands and Spain in this test.
The main reason given in all three countries for the
increased perception of safety under white light is related to
the higher perception of brightness of the whole area. The
majority of respondents perceive the area illuminated with
white light to be brighter, even though the measured
illuminance level was not increased (Table 6). This is
consistent with previous laboratory studies referenced in
the introduction.
When the area appeared brighter, most respondents felt that
their clarity of sight was improved and this was linked to an
improved perception of safety. Though not the subject of
this paper, it should be noted that many respondents clearly
expressed that they want the area at night to appear
“bright”, but not “too bright”. The lighting levels used in
the different test locations were typical for the type of
urban streets in the particular country and were not
perceived as being “too bright” by most respondents in the
specific area. As seen subsequently in Table 6, the
installed lighting levels varied significantly in the different
countries, with the highest level being in the Spanish test
location.
In summary, in all three test locations, most respondents
appreciated the increased perception of brightness achieved
by using white CDO street lighting. This was achieved at
the same installed power and comparable illuminance
levels.
Table 5: Summary of various evaluations comparing the second to the first lighting condition
Group
Cond. 1 ! Cond. 2
Mean
b+a1
a only2
b+a
a only
b+a
b+a
b+a
b+a
SON ! CDO 2800K
SON ! CDO 2800K
SON ! CDO 2800K
SON ! CDO 2800K
SON ! CDO 2800K
SON ! CDO 4200K
CDO 2800K ! SON
CDO 4200K ! CDO 2800K
3,28
3,24
2,12
2,31
1,77
1,94
4,06
3,45
0,24
0,19
0,10
0,15
0,20
0.17
0,37
0,21
0,004
0,000
0,000
0,000
0,000
0,000
0,861
0.013
b+a
a only
b+a
a only
b+a
b+a
b+a
b+a
SON ! CDO 2800K
SON ! CDO 2800K
SON ! CDO 2800K
SON ! CDO 2800K
SON ! CDO 2800K
SON ! CDO 4200K
CDO 2800K ! SON
CDO 4200K ! CDO 2800K
3,42
3,25
2,41
2,63
2,07
1,97
4,32
3,71
0,17
0,15
0,11
0,13
0,21
0,16
0,28
0,18
0,002
0,000
0,000
0,000
0,000
0,000
0,258
0,119
b+a
a only
And what about the brightness of the
Spain
b+a
area? Does it look
a only
1 = much brighter
UK
b+a
4 = the same
b+a
7 = much less bright
b+a
b+a
1
b+a = “before + after”group, 2a only = “after only” group
SON ! CDO 2800K
SON ! CDO 2800K
SON ! CDO 2800K
SON ! CDO 2800K
SON ! CDO 2800K
SON ! CDO 4200K
CDO 2800K ! SON
CDO 4200K ! CDO 2800K
3,17
2,76
2,08
2,35
1,67
1,85
4,61
3,65
0,21
0,18
0,96
0,12
0,19
0,23
0,35
0,21
0,000
0,000
0,000
0,000
0,000
0,000
0,087
0,102
Question
Land
Now I would like to ask you how
comfortable and pleasant the present
lighting is in your personal opinion?
Compared to the lighting before, is it?
1 = much more comfortable
4 = the same
7 = much less comfortable
NL
And how about safety? How does the
present lighting compare with the
lighting before? Does it make you
feel
1 = much safer
4 = the same
7 = much less safe
Spain
UK
NL
Spain
UK
NL
28
Std. Error Test Value =4
Mean
Sig. (2-tailed)
In each area, respondents in the “before + after” group
performed facial recognition tests under 2 lighting
conditions.
In Spain and the Netherlands, the distance at which
residents were sure that they could recognize faces on the
picture was increased by more than 20% under white light.
This objective measurement was consistent with subjective
evaluation that faces were easier to recognize under white
light.
In tests conducted in the UK, independent of whether the
test was first done under white or yellow light, respondents
consistently expressed the perception that the clarity of
their visibility and ability to see expressions, faces and
details was improved under white light sources. However,
this was not consistently reflected in the results from the
facial recognition tests done in St. Helens. In Hereford
Close where the test was first done under CDO 2800K and
then under SON, the mean distance for facial recognition
was longer under SON. It should also be noted that in
Hereford Close, the mean distance for facial recognition
under CDO was lower than in other test locations in St.
Helens where the vertical illuminance on the pictures were
comparable. The reason for this is unclear. Compared to
the initial CDO condition, the difference in the mean under
the second lighting condition (SON) was just statistically
significant. This result might indicate that there was a
“learning effect” which contributed to the respondents
identifying the pictures from further away in the 2nd
lighting condition (even though they were shown different
pictures). However, this “improvement” attributed to a
learning effect is less than the improvement generally seen
when CDO 2800K or CDO 4200K is used instead of SON.
Table 6: Average Distance for Facial Recognition Measured in Tests done in Eindhoven, Navalcarnero and St. Helens
Cond.
Vert. illuminance
on picture (Lux)
Dist. to identify
person. Mean ±
std. Err. Mean (m)
Diff. Mean
Cond1-Cond 2
(m)
Sig. Cond 1_2
-1,2
0,000
-2,4
0,000
-1,1
0,015
-1,2
0,047
-1,8
0,003
-1,4
0,040
Eindhoven, The Netherlands (55 respondents)
1
SON (yellow)
3.3 ± 0.6
5.4 ± 0.5
2
CDO 2800K (warm white)
% higher with CDO 2800K rel to SON
1.4 ± 0.4
6.6 ± 0.4
+ ~22%
10 ± 0.7
8.5 ± 0,26
~10
10.9 ± 0,21
+ ~28%
Navalcarnero, Spain (60 respondents)
1
SON (yellow)
2
CDO 2800K (warm white)
% higher with CDO 2800K rel to SON
The Shires and Wedge Avenue, St. Helens, UK (30 respondents)
1
SON (yellow)
~1.6
8.7 ± 0.5
2
CDO 2800K (warm white)
% higher with CDO 2800K rel to SON
~1.6
9.8 ± 0.4
+~13%
Hereford Close, St. Helens, UK (31 respondents)
1
CDO 2800K (warm white)
~1.6
7.6 ± 0.5
2
SON (yellow)
% higher with CDO 2800K rel to SON
~1.6
8.9 ± 0.7
- ~14%
Shropshire Gardens, St. Helens, UK (33 respondents)
1
SON (yellow)
~0.6
5.7 ± 0.5
2
CDO 4200K (neutral white)
% higher with CDO 2800K rel to SON
~0.6
7.1 ± 0.6
+ ~25%
Ledger Road, St. Helens, UK (31 respondents)
1
CDO 4200K (neutral white)
~1.6
9.8 ± 0.5
2
CDO 2800K (warm white)
% higher with CDO 2800 rel to 4200K
~1.6
11.3 ± 0.5
+ ~13%
29
CONCLUSION AND DISCUSSION
REFERENCES
The results presented in this paper are based on
quantitative research exploring the effect of lamp
spectrum on people’s perception of street lighting after
dark. The results show that people experience several
benefits when high quality white light is used instead of
yellowish street lighting. In particular, the perception of
brightness, comfort and safety is significantly enhanced in
the same area as judged by respondents in three European
countries who conducted the tests in areas where the
street lighting had been changed from yellowish SON to
warm or neutral white CDO lighting. The results of these
field tests together with other published results [11, 12]
illustrate the limitations of the current practice of using
the photopic luminous efficiency function V(!) at
mesopic light levels (i.e. between 0,001 – 3 cd/m2). V(!)
is used to transform the spectral power distribution of a
light source into a single measure of the light level
(luminance and illuminance).
V(!) characterizes the
spectral sensitivity of foveal cones, which peak at 555nm
under photopic lighting conditions (i.e. > ~3 cd/m2).
However, many of the lighting levels encountered on
residential streets at night fall within the mesopic range.
At mesopic light levels, both rods and cones in the retina
may be active. This leads to changes in the spectral
sensitivity with changing light levels since the
contribution of rods and cones vary with changing light
levels in the mesopic region. The peak of the spectral
sensitivity of rods is at ~507nm. Therefore for lighting
applications at night, the effectiveness of lamps with
relatively more short wavelength emission (i.e. white light
sources compared to yellow light sources) can be
underestimated by the current system of photometry.
This is currently being addressed by various technical
committees (TC) within the CIE. In particular, CIE TC 158 is working to establish the appropriate mesopic
sensitivity functions which can serve as the foundation of
a system of mesopic photometry based on visual task
performance (e.g. detection of objects, speed of detection,
identification of the objects). This system is not expected
to correlate well with visual assessment of brightness in
the mesopic region [13]. However, another technical
committee (TC 1-37) is developing a supplementary
system of photometry for evaluation of lighting at all
lighting levels in terms of brightness.
1.
E.T. Hall, The Hidden Dimension, Doubleday,
Garden City, 1966
2.
Peter Raynham , Public Lighting in Cities,
International Conference ILUMINAT 2007 &
BALKANLICHT 2007.
3.
Peter Raynham and Toruun Saksvikr"nning, White
Light and Facial Recognition, The Lighting Journal,
pp 29-33, January/February 2003
4.
British Standards Institution (BSI) BS EN 132012:2003, Road lighting - Part 2: Performance
requirements, London: BSI, 2003 BS5489:2003
5.
Alistair Scott, White Light – The UK Balance Sheet,
The Lighting Journal, pp 18-20, January/February
2005
6.
Boyce P. R., Eklund N. H., Hamilton B. J., & Bruno
L. D., Perceptions of safety at night in different
lighting conditions, Lighting Research & Technology
32(2) 79-91 (2000)
7.
Fotios SA & Cheal C, Lighting for subsidiary streets:
investigation of lamps of different SPD. Part 1 –
Visual Performance, Lighting Research &
Technology, 2007; 39(3); 215-232
8.
Fotios SA & Cheal C, Lighting for subsidiary streets:
investigation of lamps of different SPD. Part 2 –
Brightness, Lighting Research & Technology, 2007;
39(3); 233-252
9.
Rea MS. Essay by invitation. Lighting Design and
Application 1996; 26: 15–16.
The use of a more appropriate system of mesopic
photometry for the mesopic range can encourage the use
of more visually effective and thereby energy efficient
lighting and eventually contribute to a safer, more
comfortable and pleasant feeling for people outside at
night.
30
10. Rea MS, Bullough JD & Akashi Y, Several views of
metal halide and high pressure sodium lighting for
outdoor applications, Lighting Research &
Technology, in press
11. Goodman T, Forbes A, Walkey H, Eloholma M,
Halonen L, Alferdinck J, Freiding A, Bodrogi P,
Várady G, Szalmas A (2007). Mesopic visual
efficiency IV: A model with relevance to night-time
driving and other applications. Lighting Res. Technol.
39, 365-392
12. Rea MS, Bullough JD (2007). Making the move to a
unified system of photometry. Lighting Res. Technol.
39, 393-408
13. CIE TC 1-58, Visual Performance in the Mesopic
Range, meeting June 2008, Stockholm.
Light and Corporate Identity;
Using Lighting for Corporate Communication
Thomas Schielke
Darmstadt University of Technology
Germany
www.arclighting.de
info@arclighting.de
ABSTRACT
The central focus of this study is to investigate what
potential exists for brand communication in the lighting of
retail outlets. Lighting not only facilitates the visual task,
helping to present the merchandise and contributing to the
feeling of wellbeing, but can also augment the
communication of a brand’s appearance. For this study,
computer visualisations of retail outlets with different
lighting variations are evaluated in terms of light, spatial
setting and brand impression by regional and international
groups using the semantic differential technique. A
comparison between rooms with and without luminaires yet
with the same lighting effect demonstrates the effect of
luminaires as design objects. From the results it can be
deduced that light can be used for brand communication in
order to define the image of a company more clearly.
Keywords
Retail design, Perception, Corporate identity, Lighting,
Marketing, Brand communication, Brand image
INTRODUCTION
In lighting engineering, the perception of lighting has long
been evaluated in the context of safety and efficiency at the
workplace. In recent times, however, there has been an
increase in the proportion of studies looking at the
atmosphere of the room – whether aimed at increasing the
motivation at the workplace or at generally improving the
feeling of wellbeing on the premises (Loe et al., 2000;
McCloughan et al., 1999; Knez, 2000). As a result,
quantitative lighting design has been expanded by the
addition of an important dimension, that is to say, by the
inclusion of this qualitative perspective. In the context of
brand communication, the question raises itself as to what
qualitative messages can be conveyed via architecture or
architectural light, respectively, and how is this aspect
incorporated in the marketing.
From the semiotics perspective, the architecture can be
seen as a symbol (Nöth, 1985). A window, for instance, not
only fulfils the practical function of allowing the
permeation of light, but also communicates meaning
depending on its shape and position. Accordingly, many
symbols in architecture have an intentionality and can be
deciphered if the observer knows the code – maybe using
architectural history for instance. Thus, for example,
Krampen and Kotler (1979) used the semantic differential
analysis to identify the factors of meaning that connect
people and buildings; and Eco (1972) developed his
semiotic model in which he distinguished between
denotation as a physical function and connotation as a
socio-anthropological function. Hence, the interest of brand
communication is primarily directed at the secondary
function of architecture, the connotation. Richter (2008), in
his architectural psychological work, describes how, for
instance, consumer worlds now make use of insignia from
the sphere of religion in their spatial symbolism.
Conversely, the findings of brand management form an
important framework for dealing with the concept of brand
communication. The analysis of the consumer market and
buying behaviour creates an essential pre-condition for
developing new strategies (Kotler, 2000). Cultural, social,
personal and psychological factors make a considerable
contribution to the decision to buy. Knowing the
preferences of the respective target group will simplify the
propagation of brand messages aimed at transmitting the
image of a brand from the company to the customer (Foscht
et al., 2008).
The American Marketing Association defines the term
“brand” as follows: “A brand is a name, term, sign, symbol
or design, or a combination of them, intended to identify
the goods or services of one seller or group of sellers and to
differentiate them from those of competitors.” The
dimensions of meaning embodied by the term “brand” can
go off in six directions: attributes, benefits, values, culture,
personality or user (Kapferer, 1992). In addition to the
service, which reflects the brand values, the atmosphere in
the particular retail outlet also plays a significant role and
must fit the target group (Kotler, 1973).
To investigate an effective communication strategy,
marketing uses image analysis, which in turn is often
measured using the semantic differential (Osgood et al.,
1957; Florack et al, 2007). Kotler explains that image “is
the set of beliefs, ideas and impressions a person holds
31
regarding an object.” Conversely, Stern defines the term
“image” more in terms of communication theory, when she
writes: “Image is generally conceived of as the outcome of
a transaction whereby signals emitted by a marketing unit
are received by a receptor and organized into a mental
perception of the sending unit” (Stern et al., 2001). In this
present study, the term “image” is related to the external
environment when the consumer evaluates photographs of
retail outlets – in the sense of store image.
“Psychologically-orientated definitions locate image in the
consumer’s mind and treat it as a cognitive and/or
emotional construct based on consumers’ feelings” (Stern
et al, 2001). Brand image and brand awareness together
form the two components of brand knowledge (Keller,
1993). In this context, the architecture of stores can be
categorised as a non-product-related attribute. It achieves a
symbolic benefit which is appreciated by the customer
because it corresponds to his or her self-concept. When
making the decision to buy, the emotional dimension can
even be greater than the functional aspect (Pawle, 2006).
Consumers and their emotions, social standing and value
orientation are classified using milieu studies (Florack et
al., 2007). The value orientation theory in social
psychology was developed by Kluckhohn and Strodtbeck
and assumed that understanding and communication could
be facilitated by analysing people’s orientation in a cultural
context (Kluckhohn et al., 1961). A survey consisting of
different situations with associated questions served as a
basic assessment instrument. Silberer drew the value
orientation more into the context of companies and
consumer behavior (Silberer, 1991). The allocation into
groups within this study makes use of the Sinus milieu,
which plots the value judgement on the Y-axis and social
standing on the X-axis (Florack et al., 2007).
Lighting in the form of neon advertisements has long been
used for brand communication (Schivelbusch, 1992).
Luminous texts or company logos have increased a brand’s
presence in the urban area and, as a luminous feature at a
shop’s entrance, have made it easier to identify a brandname store. Seen in terms of semantics, light is directly
used as a sign. Yet when consumers enter the store, they
are no longer confronted by the brand’s luminous signage
but are standing in the light of that brand, experiencing a
specific atmosphere that is deliberately linked with the
brand via the lighting. The consistent use of a uniform
lighting concept for all the retail outlets of a brand helps a
company to build up a uniform image for a clear brand
identity. From the marketing point of view, the lighting not
only fulfils the function of facilitating vision and of
creating a hierarchy of perception using differentiated
brightness levels for the presentation of special products,
but also reflects a brand identity. Within the corporate
architecture, the lighting then becomes an information
medium for the corporate identity (Messedat, 2007). The
value of a lighting system for salesrooms is therefore no
longer seen solely in terms of how attractive it is in the
sense of a good general sales lighting for generating more
sales turnover (Cuttle et al., 1995), but also in terms of how
32
well it conveys the brand image. The existence of uniform
design guidelines for store lighting is evidence of how light
has now become a strategic component of companies’
corporate design manuals (Scheer, 2001). The study sought
to demonstrate how the lighting can create different brand
images within the same room. The qualitative lighting
design approach helps to consider the principles of
perception-oriented lighting design as well as how
luminaries are integrated into architecture (Ganslandt et al.,
1992).
METHOD
To investigate the hypothesis that solely changing the
lighting concept is sufficient to change the brand identity of
a retail outlet, an empirical consumer investigation was
conducted. It was further assumed that the appearance of
the ceiling in a standard shop can produce a prestigious
impression all on its own. The background for this
assumption lies in the observation that heterogeneous
merchandise below eye level dominates the visual field,
whereas – speaking of architectural lighting design - the
ceiling is mainly influenced by the architecture itself and
thereby the ceiling could contribute significantly to the
appearance of a store and likewise to the corporate lighting
image. An additional assumption was that light on its own
makes classification in the sense of social milieus possible
and that luminaires are not absolutely necessary. This
aspect could clarify the role of the lighting concept in
relation to the product design of the luminaries within
corporate lighting design guidelines. A further hypothesis
was that a high-class store impression does not necessarily
equate to simply increasing the brightness.
The sample group was selected from volunteers who had
mainly little to do with architectural lighting professionally.
To analyse global differences, part of the study was
conducted with an international sample group. To obtain an
evaluation of different lighting situations, the test
participants were asked to give their judgement on the light,
spatial setting and brand. The psychophysical method of
“semantic differential” for quantifying stimulus and
subjective reaction, which is frequently used in lighting
research, was reduced to just a few dimensions in order to
reveal clearer relationships (Houser et al., 2003). Eleven
pairs of adjectives covered the different dimensions. The
light was evaluated via the following factors: “bright –
dark”, “high-contrast lighting – diffuse lighting”, “cold –
warm”. The room’s characteristics were rated using the
paired adjectives “spacious – defined”. The adjective pair
“attractive – unattractive” directly rated the subjective
emotional impression in the sense of an affective evaluation
(Schierz, 2004). Attributive components, representing a
cognitive evaluation of mental concepts, were rated using:
“natural – technical”, “dramatic – relaxed”, “uniform –
differentiated” and “unobtrusive – expressive”. The
dimension of the brand was evaluated relative to the social
milieus of the consumers and to the possible allocation of
brand fields to the attributive adjective pairs “traditional –
modern” and “low budget – high class”. The spectrum of
evaluative tasks for the participants ranged from photos
depicting real architecture, combinations of photography
and graphic art through to lighting visualisations that
enabled different lighting concepts to be created for the
same location. The test participants were surveyed online to
keep the workload and costs within appropriate limits,
especially for the international survey. The results were
evaluated using descriptive statistics and correlation
analysis.
Experiment 1:
Evaluating the photography of real projects
Experiment one aims at existing projects and reveals for an
outdoor and indoor situation that the lighting design can
influence the mood and brand appearance even if the
building structure appears similar. Surveying several
architectural situations in real environments is a highly
complex process, especially regarding the proximity of the
buildings to each other, the influence of the surroundings,
the architectural differences and the coordination of a
sufficiently high number of participants. As an initial step,
therefore, an image evaluation was conducted using
photographs in an online survey. Because petrol stations
have been using uniform lighting design principles for quite
some time now (Stichting Prometheus, 1994), night-time
photos of petrol stations were used, whereby all the specific
brand information in the form of text and logos had been
deleted using image processing (Figure C1, Situation A).
Furthermore, to test what effect the luminaires have on the
appearance of the ceiling within a store, two outline
perspectives of the room with the cut-out in the ceiling
were first given to the observer for evaluation, followed by
the complete photos. In this way, an integrated lighting
approach was set over and against an additive concept with
spotlights (Figure C2, Situation B). The personal details
collected not only included age, sex and experience in
lighting design but also the participant’s value orientation,
the size of their hometown and their current mood (Table
T1).
The online survey (n=101) used the Limesurvey software,
which worked with a seven-point scale for the semantic
differential for each question. The two ends of the scale
corresponded to “very much”; the middle was labelled
“neutral”. The first image evaluation used an image format
of 500 x 375px and the second series used 600 x 390px so
that the design and the scales could be viewed together on
one monitor.
Table T2 summarises the descriptive statistics for both
series of tests. Figure 1 provides a graphic overview of the
mean values of the eleven scales. First of all, from the
petrol-station situations A1 and A2, it can be seen that the
architecture combined with the two different lighting
concepts does indeed have an effect on the components
relevant for the social milieu since it affects both the basic
orientation of “traditional – modern” and the value rating of
“low budget – high class”. In contrast, the scales of
“attractive – unattractive” and “dramatic – relaxed” only
show marginal differences.
For situation B showing the interior of two stores, figure 2
shows that a strong analogy is evident within each of the
situations B1 and B2 when it comes to the evaluation on
the emotional and cognitive levels. Although a large spatial
area can only be recognised from its contour and only the
ceiling permits a statement about light it reveals a clear
similarity to the evaluation of the total shop image. Striking
features can be identified not only with the attributes
“traditional – modern” and “low budget – high class” but
also with “dramatic – relaxed” and the spatial perception
“spacious – defined”. These points produce a greater
contrast than the light attributes “bright – dark” and “highcontrast lighting – diffuse lighting”.
In contrast to situation A, where it could perhaps be noted
that the petrol stations differed in design and size, a
comparable differentiation of the social milieu is evident
with situation B where the perspective and proportion are
identical. The examples chosen here illustrate how the
differences with the petrol stations largely concern the “low
budget – high class” scale, whereas the two stores differ
more in the “traditional – modern” scale (figure 3). The
strong to very strong correlation (0.6-0.8 to 0.8-1
respectively) between the ceiling cut-out on its own and the
entire room (Table T3) justifies the assumption that the
appearance of the ceiling alone can be taken as an indicator
for the appearance of the store as a whole.
Table T1 Test groups for experiments 1, 2, 3
Group
1
2.1
2.2
3
N
Female %
101
48
18
38
22
50
99
67
Male %
Age average
Light experience %
48
28
18
50
28
28
45
25
41
31
31
60
No light experience %
79
61
50
38
33
Table T2 Descriptive statistics for experiment 1: Mean (M) and standard deviation (S). Situations A1 and A2 petrol-stations,
situations B1 and B2 Retail shop with a for ceiling detail and b for total shop image.
Situation
attractive | unattractive
dramatic | relaxed
spacious | defined
uniform | differentiated
natural | technical
bright | dark
cold | warm
high-contrast lighting | diffuse l.
traditional | modern
low budget | high class
unobtrusive | expressive
A1
M
-0,2
-0,2
-0,5
-1,5
1,7
-1,0
-1,5
-0,5
0,6
0,3
0,7
S
1,6
1,4
1,4
1,3
1,4
1,3
1,4
1,5
1,8
1,4
1,5
A2
M
-0,2
0,0
0,0
0,1
1,2
-0,1
-0,2
0,1
1,3
0,7
0,4
S
1,6
1,4
1,5
1,6
1,3
1,6
1,4
1,5
1,4
1,5
1,4
B1a
M
0,4
0,0
-0,4
-0,8
0,7
-0,9
0,6
0,2
0,1
0,1
0,2
S
1,7
1,4
1,5
1,8
1,5
1,3
1,4
1,4
1,6
1,7
1,5
attractive | unattractive
dramatic | relaxed
attractive | unattractive
dramatic | relaxed
spacious | defined
spacious | defined
uniform | differentiated
uniform | differentiated
natural | technical
bright | dark
natural | technical
bright | dark
cold | warm
cold | warm
high-contrast lighting | diff. l.
high-contrast lighting | diff. l.
traditional | modern
low budget | high class
traditional | modern
low budget | high class
unobtrusive | expressive
unobtrusive | expressive
Figure 1 Comparison of mean semantic differential
appearance for situations A1 and A2
B1b
M
0,4
-0,4
0,0
-0,1
0,4
-0,7
0,3
0,0
0,0
-0,3
0,4
S
1,6
1,3
1,4
1,7
1,5
1,3
1,4
1,4
1,6
1,3
1,4
B2a
M
-0,1
0,4
-0,6
-0,6
0,9
-0,4
0,6
0,2
1,5
0,8
-0,1
S
1,8
1,3
1,5
1,7
1,5
1,3
1,5
1,5
1,3
1,4
1,5
B2b
M
-0,7
0,9
-1,2
-0,7
0,6
0,0
0,8
0,5
1,4
1,4
-0,6
S
1,8
1,4
1,4
1,5
1,6
1,3
1,5
1,4
1,5
1,3
1,5
Figure 2 Comparison of mean semantic differential
appearance for situations B1 (line) and B2 (dashed line)
each with ceiling detail and total shop image
Table T3 Correlation between ceiling detail and total shop
image for situation B retail shop.
Situation
B1a
B1b
B2a
B2b
Figure 3 Relation traditional – modern (x-axis) and low
budget – high class (y-axis): ! Situation A, " Situation B1,
Situation B2
34
B1a
B1b
B2a
,758**
,634*
,398
,191
-,208
,871**
Table T4 Descriptive statistics for experiment 2: Mean (M) and standard deviation (S). Situations 1a-8a shop with luminaries
(Group 2.1) and situations 1b-8b shop with erased luminaires (Group 2.2)
Situation
Shop with luminaires
attractive | unattractive
dramatic | relaxed
spacious | defined
uniform | differentiated
natural | technical
bright | dark
cold | warm
high-contrast lighting | diffuse l.
traditional | modern
low budget | high class
unobtrusive | expressive
1a
M
0,4
-0,1
0,2
0,1
0,9
0,9
-0,8
0,5
0,6
-0,2
-0,4
Situation
Shop with erased luminaires
attractive | unattractive
dramatic | relaxed
spacious | defined
uniform | differentiated
natural | technical
bright | dark
cold | warm
high-contrast lighting | diffuse l.
traditional | modern
low budget | high class
unobtrusive | expressive
1b
M
0,9
0,2
0,0
-0,6
1,2
0,7
-0,8
1,2
0,4
0,2
-1,3
1,8
1,4
1,3
1,3
1,6
1,4
1,6
1,7
1,6
1,4
1,6
2a
M
0,1
-0,3
-1,5
-0,3
1,2
-1,4
-0,1
-1,2
0,7
0,6
1,3
1,3
1,4
0,8
1,6
1,2
1,6
0,9
1,3
1,6
1,1
1,1
2b
M
-0,2
-0,9
-0,4
-0,1
1,6
-1,9
-0,6
-1,1
1,3
0,2
1,3
S
S
1,5
1,4
0,9
1,5
1,4
1,5
1,8
1,5
1,4
1,6
1,3
3a
M
S
0,3
1,9
-0,8
1,8
1,0
1,0
1,5
1,2
1,5
1,6
1,2
1,3
-0,1
1,7
-0,8
1,7
1,7
0,9
0,0
1,6
1,2
1,6
4a
M
S
-0,2
2,0
0,2
1,6
0,3
1,5
-0,1
1,4
0,8
1,5
0,8
1,5
0,8
1,5
0,6
1,9
0,4
1,3
0,3
1,4
0,1
1,5
5a
M
S
0,5
1,9
-1,5
1,7
-0,1
1,7
-0,1
1,8
1,4
1,5
1,1
1,5
1,1
1,2
-0,4
1,6
1,4
1,2
0,4
1,8
2,1
1,3
6a
M
-0,2
-0,3
-1,6
-1,3
0,9
-2,0
-0,5
-0,1
0,6
0,7
0,9
1,8
1,2
1,2
1,4
1,3
0,9
1,3
1,2
1,1
1,5
0,8
3b
M
S
-0,1
1,6
-1,0
1,6
0,9
1,2
1,7
1,2
1,9
0,8
1,2
1,1
0,4
1,5
-1,2
1,4
2,1
0,8
0,0
1,4
1,8
1,1
4b
M
S
-0,3
1,6
0,6
1,5
0,1
1,2
0,3
1,4
-0,3
1,4
0,8
0,9
1,6
0,9
0,4
1,3
0,1
1,4
0,6
1,2
-0,5
1,3
5b
M
S
0,7
1,7
-2,1
0,8
0,0
1,6
0,0
1,7
1,5
1,3
1,1
0,8
1,4
1,4
0,4
1,2
2,1
1,0
0,1
1,1
2,3
0,8
6b
M
-0,2
-0,5
-1,5
-0,9
0,4
-2,2
-0,7
-0,4
0,2
0,4
-0,1
S
S
1,8
1,3
1,1
1,4
1,8
0,8
1,7
1,4
1,8
1,5
1,4
7a
M
-0,1
-0,9
1,3
0,8
2,7
2,6
-0,4
-0,2
1,6
0,5
1,6
1,8
1,0
1,4
1,7
1,8
1,0
1,2
1,4
1,6
1,3
1,4
7b
M
0,2
-1,9
1,1
1,3
2,4
2,7
-0,6
-1,0
2,3
0,7
2,3
S
S
2,2
2,0
1,1
1,8
0,6
0,5
1,8
2,2
1,7
2,0
2,0
8a
M
1,1
-0,8
-1,2
-1,9
1,7
-2,2
-1,7
0,3
-0,3
-0,5
0,3
2,1
1,3
1,6
2,0
1,0
0,6
1,8
1,9
0,8
1,3
1,0
8b
M
1,2
-0,7
-0,4
-1,2
1,4
-2,0
-1,9
0,3
-0,7
-1,0
0,2
S
S
S
1,6
1,5
1,8
1,5
1,4
1,0
1,4
1,6
0,9
1,5
1,7
S
1,4
0,8
1,8
1,6
1,6
1,0
1,0
1,7
1,6
1,6
1,6
Table T5 Correlation analysis for situations 1a-8a shop with luminaries (Group 2.1). *Indicates correlations coefficients that are
significant at the 5% level. **Indicates correlations coefficients that are significant at the 1% level.
P01
P01 attractive | unattractive
P02 dramatic | relaxed
P03 spacious | defined
P04 uniform | differentiated
P05 natural | technical
P06 bright | dark
P07 cold | warm
P08 high-contrast lighting | diffuse l.
P09 traditional | modern
P10 low budget | high class
P11 unobtrusive | expressive
-,531
-,134
-,310
,192
-,226
-,424
,079
-,304
-,792*
-,108
P02
-,196
-,146
-,616
-,238
-,123
,400
-,498
,053
-,738*
Experiment 2: Evaluating the lighting visualisation
To evaluate one and the same room with different lighting
situations, the study used lighting visualisations based on
Dialux Renderings. Various investigations have shown that
the comments made about computer simulations compare
favourably with observations made about the real space and
that the comparison is therefore valid and acceptable.
(Newsham et al., 2005; Mahdavi et al., 2002; Rohrmann et
al., 2002). The aim of the visualisations was to show how
the appearance of the same interior changes solely due to
the lighting. Refitting a room or constructing several
otherwise identical salesrooms would be logistically and
economically highly impractical and therefore simulations
were used here. The simulated salesroom measured
approximately ten by fifteen by three and a half metres.
P03
,846**
,477
,950**
,296
,141
,687
-,114
,127
P04
,307
,838**
,432
-,318
,853**
,150
,307
P05
,439
-,182
-,212
,455
-,005
,502
P06
,471
,081
,758*
,084
,261
P07
-,209
,527
,552
,502
P08
-,506
-,461
-,712*
P09
,424
,710*
P10
,602
Items of clothing were shown on shelves and tables. The
shop window and the background featured decorative
points with mannequins, which also gave an idea of the
room’s size. As the viewing angle, the view looking into
the room through the shop window was chosen as the
central perspective. This is a perspective that consumers
would be familiar with when walking past a store and
standing in front of the entrance. Two on-line
questionnaires were conducted in order to assess what
influence the design of the luminaires has on the
appearance. Group 2.1 (n=18) were given visualisations
with luminaires (800 x 294px); group 2.2 (n=22) received
visualisations in the same format in which the luminaires
were erased. As in experiment 1, the same semantic
differential was used with both groups. Eight different
35
lighting scenes were given to each of the two groups for
evaluation. The paired questions for each lighting situation
were randomly put in a new order each time to avoid the
effects of repetition.
Table T4 presents the results of groups 2.1 and 2.2 showing
the mean and standard deviation. Where the spatial
situation is the same but the lighting is different, great
differences between the light scenes are apparent not only
with the scales for light, but also with the attributes for the
allocation to brand fields, i.e. with “traditional – modern”,
“low budget – high class”. For group 2.1, for instance, the
relationship between the “traditional – modern” scales and
the adjective pairs “spacious – defined” and “bright – dark”
shows strong to very strong correlations. The latter has a
two-tailed significance of 0.05 (Table T5). The “low budget
– high class” attribute shows a middle correlation to the
“cold – warm” parameter.
The analysis of the mean values from the two series of
situations, 2.1 and 2.2, vividly demonstrates that a strong
correlation exists with the four factors “attractive –
unattractive”, “natural – technical”, “high-contrast lighting
– diffuse lighting”, “low-budget – high class” and all other
factors have a very strong correlation (Table T6). In seven
cases the correlation on the level of 0.01 has two-sided
significance, in three others it is 0.05.
The comparison of store situations with and without
luminaires but with the same lighting effect demonstrates
that, in the examples presented, the significant impression
can be made just with the light alone. The luminaires take
on a subordinate role. This aspect can be quite different in
real surroundings because the luminaires appear bigger in
the room due to the perspective as the observer moves
around. Nevertheless, for building a brand image, the
importance of the lighting concept compared to the choice
of luminaires should not be underestimated.
Table T6 Correlation analysis for Group 2.1 and 2.2.
Scales
attractive | unattractive
dramatic | relaxed
spacious | defined
uniform | differentiated
natural | technical
bright | dark
cold | warm
high-contrast lighting | diffuse lighting
traditional | modern
low budget | high class
unobtrusive | expressive
Light-Luminaires
,743*
,860**
,886**
,910**
,638
,995**
,952**
,718*
,907**
,772*
,891**
* Significance at 5% level. ** Significance at 1% level.
36
Experiment 3: Evaluating the lighting visualisation in
the international comparison
To analyse cultural differences in the context of global
marketing strategies, group 2.2, which originated from
Germany, was set in relation to group 3, which had an
international composition (n=99): group 3.1 = Europe
(n=24); group 3.2 = America (n=20); group 3.3 = Middle
East (n=26) and group 3.4 = Asia (n=17). Table T7 lists the
mean and standard deviation for the entire group 3.
Using the correlation coefficient, table T8 shows the
relationship of how greatly the different regions distinguish
themselves from each other or resemble each other. The
strongest analogies are present in Middle East – Europe,
followed by Europe – Asia and America – Europa. If the
values are compared with respect to the attributes, it is
shown that the strongest correlation exists for “bright –
dark”, followed by “traditional – modern” and “uniform –
differentiated”. If the mean of the correlation coefficients is
considered, overall there is a very strong correlation
between the regions.
If, for instance, only the parameters “traditional – modern”
and “low budget – high class” are considered, it then
becomes clear, as figure 4 shows, that the geographical
areas each receive a similar evaluation yet can still be
delineated from each other, and the extent to which
regional differences can arise also becomes apparent. By
dividing into groups 3.1-3.4 and 2.2, the graphic reflects
how the salient points of the evaluations arise for the
various lighting situations.
If all the data of group 3 is considered in terms of the
evaluation of “spacious – defined” and “bright – dark” in
relation to the brightness of the image (Table T9), then it
becomes evident from table T10 that the measurement of
the overall image brightness correlates very strongly with
these two factors and has a two-tailed significance level at
0.01. However, a stronger indicator for the impression of
brightness and expanse is the brightness of the vertical
surfaces in the image. These account for 70% of the image
area and produce a higher correlation coefficient than that
obtained with the overall image brightness.
If the “bright – dark” parameter is set in relation to “low
budget – high class”, it then becomes apparent that the
evaluation of the attribute for the price image remains
largely constant despite changing brightness (Figure 5).
The use of light to generate a high-price brand identity is
therefore not dependent on higher luminous flux and thus
higher energy consumption.
Table T7 Descriptive statistics for experiment 3: Mean (M) and standard deviation (S). Situations 1b-8b shop with erased
luminaires (Group 3)
Situation
Shop with erased luminaires
attractive | unattractive
dramatic | relaxed
spacious | defined
uniform | differentiated
natural | technical
bright | dark
cold | warm
high-contrast lighting | diffuse l.
traditional | modern
low budget | high class
unobtrusive | expressive
1b
M
0,6
0,0
0,3
-0,4
0,9
1,0
-0,8
0,8
0,5
-0,4
-0,6
S
1,7
1,4
1,6
1,7
1,7
1,5
1,5
1,5
1,7
1,6
1,4
2b
M
0,1
-0,8
-0,2
-0,3
1,6
-1,8
-0,4
-1,0
0,6
0,3
0,7
S
1,7
1,4
1,7
1,8
1,2
1,2
1,5
1,5
1,7
1,5
1,3
3b
M
S
-0,8
1,6
-1,5
1,3
0,7
1,6
1,5
1,3
1,9
1,1
0,4
1,4
-0,1
1,5
-1,5
1,3
2,1
0,9
0,9
1,4
1,8
1,3
4b
M
S
-0,4
1,6
0,0
1,7
0,3
1,4
0,8
1,5
0,1
1,6
0,5
1,4
1,4
1,4
-0,7
1,7
0,2
1,4
0,7
1,5
0,5
1,5
5b
M
S
0,0
2,0
-2,0
1,2
0,5
1,7
1,0
1,7
2,1
1,0
0,9
1,5
0,5
1,8
-1,0
1,7
2,1
1,1
0,6
1,5
1,7
1,3
6b
M
0,0
0,6
-1,2
-1,0
-0,2
-2,2
-0,1
-0,1
0,1
-0,1
-0,4
S
1,7
1,5
1,7
1,4
1,6
0,9
1,8
1,6
1,8
1,7
1,7
7b
M
-0,6
-2,2
1,4
1,6
2,5
2,5
-1,1
-1,6
2,5
1,2
2,0
S
2,2
1,4
1,7
1,6
0,7
0,9
1,8
1,9
0,9
1,5
1,4
8b
M
1,1
0,5
-1,0
-1,8
0,4
-1,8
-1,4
0,8
-0,5
-1,1
-0,8
S
1,8
1,2
1,8
1,5
1,7
1,1
1,4
2,0
1,9
1,8
1,5
Scales
Region
America-Asia
AmericaEurope
Europe-Asia
Middle EastAmerica
Middle EeastAsia
Middle EastEurope
GermanyAmerica
Germany-Asia
GermanyEurope
GermanyMiddle East
Table T8 Correlation analysis for different regions within Group 3 (Situations 1b-8b shop with erased luminaries).
* Significance at 5% level. ** Significance at 1% level.
Mean
attractive | unattractive
dramatic | relaxed
spacious | defined
uniform | differentiated
natural | technical
bright | dark
cold | warm
high-contrast lighting | diffuse l.
traditional | modern
low budget | high class
unobtrusive | expressive
Mean
,745*
,911**
,826*
,952**
,926**
,931**
,918**
,928**
,939**
,763*
,949**
,890
,779*
,800*
,883**
,983**
,930**
,983**
,957**
,842**
,991**
,914**
,936**
,909
,826*
,937**
,740*
,979**
,972**
,965**
,930**
,862**
,945**
,924**
,940**
,911
,714*
,956**
,912**
,941**
,938**
,981**
,879**
,910**
,937**
,678
,870**
,883
,654
,961**
,806*
,922**
,923**
,966**
,882**
,895**
,858**
,831*
,822*
,865
,927**
,924**
,942**
,938**
,960**
1,000**
,853**
,872**
,920**
,864**
,940**
,922
,529
,698
,922**
,814*
,674
,964**
,896**
,579
,955**
,501
,797*
,757
,592
,800*
,949**
,884**
,828*
,942**
,836**
,650
,950**
,677
,806*
,810
,818*
,799*
,826*
,827*
,859**
,984**
,854**
,921**
,931**
,627
,912**
,851
,763*
,734*
,865**
,879**
,721*
,985**
,969**
,643
,927**
,755*
,776*
,820
Table T9 Image brightness (Minimum 0, Maximum 255) of
situations 1b-8b: Total brightness, horizontal surfaces
(30%), vertical surfaces (70%)
1b
2b
3b
4b
5b
6b
7b
8b
Total
M
S
123,0
136,7
135,2
108,7
112,6
153,6
56,5
152,3
35,9
62,8
62,8
57,2
65,4
63,3
37,2
66,4
Horiz.
M
156,2
160,1
134,3
126,8
116,3
151,5
52,8
134,9
S
28,4
67,7
53,8
61,8
62,0
58,5
37,2
51,0
Vert.
M
S
109,2
127,1
135,6
102,4
111,0
154,7
58,3
159,5
29,1
57,9
66,7
54,1
66,8
65,7
37,0
70,6
,735
,852
,867
,912
,873
,970
,897
,810
,935
,754
,875
,862
Table T10 Correlation analysis for image evaluation factors
and image brightness within Group 3.
Scale
spacious | defined
bright | dark
brightness total
brightness horizontal
brightness vertical
spacious | defined
bright | dark
,930**
-,850**
-,678
-,861**
-,874**
-,715*
-,877**
* Significance at 5% level. ** Significance at 1% level.
37
by virtue of its lighting effect and design pattern. The
international comparison reveals that different groups
evaluate the brand image differently, although there is still
strong correlation. Uniform lighting concepts could be
implemented as global design guidelines for international
markets if global variance is included. Lighting concepts
that are able to augment the brand identity can generate
added value for the business. The financial value of a
lighting system would then no longer only consist of
investment and running costs but also of the contribution to
brand communication.
ACKNOWLEDGEMENTS
Figure 4 Relation traditional – modern (x-axis) and low
budget – high class (y-axis). Situation 1b-8b for Group 3
and Group 2.2 with separate marks for five regions:
America, Asia, Europe, Middle East, Germany
The author would like to thank those who participated in
the survey and the following institutes for their
cooperation: Darmstadt University of Technology, Dresden
University of Technology, Manipal University Dubai,
Tamasek Polytechnic Singapore, University of applied
Sciences Bochum, University of applied Sciences Cologne,
University of Nebraska-Omaha, University of Siegen and
Wismar University of Technology. This study was
supported by the IALD Education Trust Scholarship.
REFERENCES
1. Cuttle, C., and Brandston H. Evaluation of Retail
Lighting. Journal of the Illuminating Engineering
Society (Summer 1995), 21-49.
2. Eco, U. A componential analysis of the architectural
sign / column. Semiotica 5 (1972), 97-117.
Figure 5 Relation low budget – high class (x-axis) and
bright – dark (y-axis). Situation 1b-8b, Group 3
RESULTS
The different image analyses permit the conclusion that
various relationships exist between the architectural
lighting and the brand identity of a retail outlet. Groupings
for strategic marketing can undertaken based on the
attributes for social milieu in order to analyse the image of
lighting designs for target groups via aspects such as value
orientation and social standing. The evaluation of the
surveys shows that the rooms can convey a very different
image in terms of brand identity simply through having
different lighting. This phenomenon can be used for brand
communication in order to clearer define the image of a
business at the point of sale. The aspect of brightness,
although much discussed in lighting research, actually only
plays a subordinate role. A possible advantage of this is
that using light to construct a striking brand image does not
necessarily entail higher energy consumption. The
comparison of situations with and without luminaires for
the same lighting effect demonstrates that the essential
impression is already reached via light and that visible
luminaires are not strictly necessary. The appearance of the
ceiling can give an indication of the store’s image simply
38
3. Florack, A.; Scarabis, M, and Primosch, E. Psychologie
der Markenführung. Vahlen, München, 2007. 257-275,
357-391.
4. Foscht, T., and Maloles, C. The impact of culture on
brand perceptions: a six-nation study. Journal of
Product & Brand Management 17,3 (2008), 131-142.
5. Ganslandt, R., and Hofmann, H. Handbook of lighting
design. Vieweg & Sohn, Braunschweig/Wiesbaden,
1992.
6. Houser, K., and Tiller, D. Measuring the subjective
response to interior lighting: paired comparisons and
semantic differential scaling. Lighting Research
Technology 35, 3 (2003), 183-198.
7. Kapferer, J.-N. Strategic Brand Management: New
Approaches to Creating and Evaluating Brand Equity.
Kogan Page, London, 1992.
8. Keller, K. Conceptualizing, Measuring, and Managing
Customer-Based Brand Equity. Journal of Marketing 57
(January 1993), 1-22.
9. Kluckhohn, F. Variations in value
Greenwood Press, Westport, 1961.
orientations.
10. Knez, I. Effects of indoor lighting, gender, and age on
mood and cognitive performance. Environment and
Behavior 32, 6 (November 2000), 817-831.
11. Kotler, P. Atmospherics as a Marketing Tool. Journal of
Retailing (Winter 1973-1974), 48-64.
12. Kotler, P. Marketing Management. Prentice Hall, Upper
Saddle River, 2000.
13. Krampen, M. Kotler, P. Marketing Management. Pion,
London, 1979.
14. Loe, D; Mansfield, K, and Rowlands, E. A step in
quantifying the appearance of a lit scene. Lighting
Research Technology 32, 4 (2000), 213-222.
15. Mahdavi, A., and Eissa, H. Subjective Evaluation of
Architectural Lighting Via Computationally Rendered
Images. Journal of the Illuminating Engineering Society
31, 2 (2002), 11-20.
16. McCloughan, C; Aspinall, P, and Webb, R. The impact
of lighting on mood. Lighting Research Technology 31,
3 (1999), 81-88.
17. Messedat, J. Corporate Architecture. Av Edition,
Ludwigsburg, 2005.
18. Newsham, G; Richardson, C; Blanchet, C, and Veitch,
J. Lighting quality research using rendered images of
offices. Lighting Research Technology 37, 2 (2005), 93115.
19. Nöth, W. Handbuch der Semiotik. Metzler, Stuttgart,
1985.
20. Osgood, C; Suci, C, and Tannenbaum, P. The
measurement of Meaning. University of Illinois Press,
Urbana, 1957.
21. Pawle, J. Measuring Emotion – Lovemarks, The Future
Beyond Brands. Journal of Advertising Research 46, 1
(2006), 38-48.
22. Richter, P. Architekturpsychologie. Pabst Science
Publishers, Lengerich, 2008.
23. Rohrmann, B, and Bishop, I. Subjective responses to
computer simulations of urban environments. Journal of
Environmental Psychology 22 (2002), 319-331.
24. Scheer, T. Die Volkswagen Architektur: Identität und
Flexibilität als Konzept. Hatje Cantz, Ostfildern-Ruit,
2001.
25. Schierz, C. Perception and assessment of artificial
indoor lighting. Zeitschrift für Arbeitswissenschaft 58, 2
(2004), 74-83.
26. Schivelbusch, W. Licht, Schein und Wahn. Ernst &
Sohn, Berlin, 1992.
27. Silberer, G. Werteforschung und Werteorientierung im
Unternehmen. Poeschel, Stuttgart, 1991.
28. Stern, B; Zinkhan, G, and Jaju, A. Marketing images.
Marketing Theory 1 (2001), 201-224.
29. Stichting Prometheus (eds.). International Lighting
Review 1 (1994), 22-33.
39
Figure C1 Situation A.
Outdoor photos. Petrol station A1 and A2
Figure C2 Situation B.
Ceiling visualisations and photos. Shop B1 and B2
(Ralf Peters, Tankstelle,
1998. Foto: Courtesy
Bernhard Knaus Fine Art,
Frankfurt a.M., Germany.)
Figure C3 Situation C. Retail space visualisations
1a. Shop with luminaires
4a. Shop with luminaires
7a. Shop with luminaires
1b. Shop without luminaires
4b. Shop without luminaires
7b. Shop without luminaires
2a. Shop with luminaires
5a. Shop with luminaires
8a. Shop with luminaires
2b. Shop without luminaires
5b. Shop without luminaires
8b. Shop without luminaires
3a. Shop with luminaires
6a. Shop with luminaires
3b. Shop without luminaires
6b. Shop without luminaires
40
Tuning the Spectrum of Lighting to Enhance Spatial
Brightness: Investigations of Research Methods
Steve Fotios
School of Architecture
University of Sheffield
Sheffield, S10 2TN, UK
+44 114 222 0371
steve.fotios@sheffield.ac.uk
ABSTRACT
In this paper we describe research of spatial brightness at
photopic levels and how this is affected by the spectral
power distribution of the light source. Our research of
experimental methods has identified strategies for best
practice in experimental design; ignoring these leads to
results which can give a misleading estimate of the effect of
lamp spectral distribution on spatial brightness. This article
reports the on-going meta-analysis of previous work and
new experimental data pertinent to research methods. A
preliminary set of reliable data are proposed for use with
modelling to extrapolate the relationship between SPD and
spatial brightness.
Keywords
Spatial brightness, spectral power distribution, research
methods, evaluation mode, visual field.
INTRODUCTION
Spectral power distribution (SPD), along with spatial
distribution, temporal modulation, and illuminance, is one
of the fundamental variables available to the lighting
designer. This article discusses the effect of SPD on
impressions of spatial brightness at photopic levels typical
of interior lighting.
The term spatial brightness is used to imply a subjective
evaluation of the amount of light in a space. It is distinct
from object brightness - the brightness of an illuminated
surface or object - although brightness may be the term
used by naïve observers, and may be considered akin to the
visual clarity judgements investigated in some previous
work [1-4]. Spatial brightness is a dominant perceptual
attribute. Boyce and Cuttle asked test participants to
describe the lighting in a room in their own words and
found that they used mainly terms of brightness and clarity;
pleasantness and colourfulness were among those also
mentioned, but these very infrequently [5].
Interior electric lighting is a significant energy consumer.
Within the EU, lighting in the commercial sector consumes
30% of total electricity consumption [6]; lighting accounts
for up to 40% of energy costs in a typical UK office [7] and
an average of 39% of the energy use in US office buildings
[8]. Lighting recommendations are based almost entirely on
Kevin Houser
Dept. of Architectural Engineering
Pennsylvania State University
University Park, PA 16802, USA
+1 (814) 863-3555
khouser@engr.psu.edu
ensuring visibility and the data on which these
recommendations are based have not taken into account
any possible effects of the spectral content of the light
source. Visual performance models [e.g. 9] imply that
virtually all tasks done in offices and schools could be done
just as well at much lower illuminances than those
currently used. However, illuminances have not been
reduced because people like an interior to appear bright.
Dim, gloomy lighting can induce a sense of visual
discomfort which may change the observer’s mood and
motivation to carry out a task, particularly if the work is
prolonged [10]. Thus, if a perception of brightness could be
maintained at a lower illuminance, energy consumption and
carbon emissions could be reduced.
There is evidence that light source SPD affects the
perception of spatial brightness [e.g 11,12] and this
provides a means for reducing illuminances whilst
maintaining the same perception of brightness. To do this
requires a tool for predicting how lamp SPD affects spatial
brightness and hence reliable and appropriate evidence with
which to develop the tool.
There is much ongoing work to investigate effects of SPD
within the lighting community. The Illuminating
Engineering Society of North America (IESNA) has
established the Visual Effects of Lamp Spectral
Distribution committee to investigate SPD effects on
spatial brightness and visual effort at photopic levels, and
new research was presented by several groups at the 26th
Session of the CIE in Beijing, 2007.
The spatial brightness section of the IESNA committee has
two stages of work. The first stage is to identify reliable
empirical evidence that demonstrates an effect of lamp
spectrum on spatial brightness at photopic levels; the
objective of this article is to provide evidence to support
decisions necessary when identifying reliable evidence.
Reliable is here intended to mean data which are unbiased
by the experimental procedure and through this are more
consistent. Through experimentation, critical analysis of
experimental data and literature survey, the authors have
identified features of experimental design that might be
considered best practise for research of subjective
41
evaluations of lighting. The second stage is to identify a
method for predicting the magnitude of the lamp SPD
effect on spatial brightness, hence using the set of reliable
data to test and develop prediction tools.
Around sixty studies have previously investigated lamp
SPD effects on spatial brightness (or visual evaluations
considered similar to spatial brightness), some reporting a
significant effect while others report a negligible effect.
The problem encountered when comparing the outcomes of
these different studies is that each has tended to use a
unique combination of independent variables and methods lamp SPD, response task, stimulus size, illuminance and
evaluation mode [13]. A first step in interpreting these data
is an exploration of research methodologies to identify how
these differences in methodology matter, hence to identify
those methods giving reliable and appropriate estimates of
lamp SPD effects on spatial brightness.
A problem within the body of previous work is that much
of it must be considered unreliable, frequently because of
incomplete reporting. There are three reasons for
considering work to be unreliable. Firstly, the published
work either reveals an experimental or subjective bias, or
does not present sufficient data to check whether an
expected bias has been successfully countered. Secondly,
there are insufficient data to allow the results to be
analysed; one common problem is that the mean value of
the dependent variable is reported but without a measure of
dispersion, and there are no raw data or references to
further publications. Finally, descriptions of the apparatus
and methodology are insufficient.
RESPONSE TASKS
The assessment of brightness is a psychophysical task that
requires the test participant to make sensory responses to
physical stimuli. These assessment tasks are usually one of
three types:
–
adjustment, where the participant is required to adjust
the magnitude of one dimension of a stimulus (e.g.
illuminance) toward a given sensation, such as
matching the visual sensation of a reference stimulus;
–
discrimination, where the test participant is required
to make simple ordinal discrimination judgements of
stimuli, e.g. which of two stimuli is brighter; and
–
category rating, where the participant is required to
assign numbers to stimuli to represent the sensation
magnitude.
Matching
In the side-by-side matching task, two stimuli differing in
illuminance and SPD are presented simultaneously,
illuminating adjacent, identical spatial locations (Figure 1).
The illuminance of one stimulus is adjusted by the test
participant until the two appear, as near as possible, equally
bright. If the ratio of the illuminances of these stimuli is
different from unity then the brightness judgement must in
some way be affected by differences in light source SPD.
The authors recently reported on sources of experimental
42
bias in the matching task [14] and these tend to exaggerate
apparent differences between stimuli.
Figure1. Identical side-by-side rooms used in a matching
task. This is a simultaneous evaluation.
There are three elements of the matching procedure that
can affect the results. The first relates to the process of
adjustment itself. There is a tendency toward conservative
adjustment, whereby the variable stimulus is set to a lower
level than expected [15]. This can most easily be seen in
null condition data (matching using stimuli of identical
SPD), where the mean illuminance of the variable stimulus
is significantly less than that of the fixed stimulus at equal
brightness, but is also evident in matches made between
different types of lamp. When matching tests are carried
out at a range of reference illuminances, there is a tendency
for response contraction bias [16]; matches made at the
higher illuminances are set to a lower than expected
illuminance, whilst matches made at the lower illuminances
are set to a higher than expected illuminance. Bias due to
dimming can be countered by applying the dimming action
to alternate stimuli on successive trials.
The second bias in the matching task relates to the stimulus
position; whether a particular stimulus is located in the lefthand (LH) or right-hand (RH) field. Thornton & Chen [4]
used side-by-side matching to compare visual clarity under
different lamps. Their trials included four null conditions
for which an illuminance ratio (RH/LH) of unity would be
expected at equal clarity if there were no positional bias.
Subsequent analysis [17] of these null condition data
suggested a mean illuminance ratio (RH/LH) of 1.145 at
equal clarity, although there are insufficient data to
determine whether this is a statistically significant
departure from unity. A positional bias has also been
reported when using smaller fields; this was an observer
who consistently reported the top half of a horizontally split
field to be brighter than the bottom half, even when the top
and bottom stimuli were reversed [18]. Positional bias can
be countered by using lamps to illuminate alternate spatial
locations in successive trials. Some studies do analyse their
data for positional bias and in these it has been shown to
have negligible effect [e.g. 19], but the majority of studies
do not make this analysis, do not employ counterbalancing,
and a positional bias must therefore be considered possible.
In many matching studies, possible effects of conservative
adjustment bias and positional bias are compounded. Table
1 shows the results from Aston & Bellchambers’ matching
tests [1]. In these tests, Kolor-rite lamps illuminated the
left-hand booth and three test lamp types alternately
illuminated the right-hand booth. Test participants adjusted
the illuminance of the Kolor-rite booth until the visual
clarity of the two booths appeared equal, the test lamps
being set to one of three reference illuminances. Neither
dimming application nor spatial location were
counterbalanced. In every case, the median illuminance of
the Kolor-rite lamp is lower than the illuminance of the test
lamp; it is not possible to say whether these differences in
illuminance is due to lamp SPD or to experimental bias.
Illuminance
of test
lamps
Median illuminance (lux) of Kolor-rite
lamp at equal visual clarity with three
test lamps
(lux)
Daylight
Warm
White
White
200
170
130
145
400
270
230
270
800
560
460
435
Table 1. Results of Aston & Bellchambers’ side-by-side
matching test [1]. In every case the variable stimulus
(Kolor-rite) in the left-hand booth was set to the lower
illuminance.
The third bias relates to the initial illuminance of the
variable stimulus, as set by the experimenter prior to each
trial. This can be set to an illuminance either higher or
lower than that of the reference stimulus, which may
modify the observer’s internal brightness reference. An
effect of initial illuminance can be seen when an
adjustment task is used for preference judgements, an
absolute judgement carried out in the absence of a
reference stimulus. Ray asked observers to adjust the
illuminance of lighting to a level clear and comfortable to
read at [20]. This was carried out under two types of
tungsten filament (GLS) lamp, having either a clear-glass
or blue-glass envelope. Eighteen observers repeated this
twice for each type of lamp, once each starting from a high
illuminance and a low illuminance. The results are shown
in Table 2. It can be seen that the lamps were set to a higher
illuminance when the initial illuminance was high than
when the initial illuminance was low, and these differences
are statistically significant (p<0.05, t-test).
Clear-glass GLS
lamp
Blue- glass GLS
lamp
Initial illuminance
of stimulus
high
low
high
low
Mean preferred
illuminance (lux)
1123
645
806
419
Table 2. Mean illuminances of lamps set to a level clear
and comfortable to read at [20]. Note: unpublished
undergraduate thesis, raw data analysed by Fotios [17].
For the side-by-side matching task, this suggests a trend
for the variable stimulus to be set to a higher level at the
matched condition when starting from a high initial
illuminance and a lower level when starting from a low
initial illuminance. This trend can be seen in the results
from two studies [21,22] although it is not always a
significant trend, but a significant effect in the opposite
direction has also been found [23], i.e. the variable
stimulus was set to a higher level when starting from the
lower initial illuminance. It is clear that the initial
illuminance of the variable stimulus can affect the outcome
of a matching task, although the evidence is not conclusive
as to the direction of the effect, but this is sufficient to
warrant the precaution of counterbalancing the initial
illuminance of the variable stimulus.
Many studies have not employed sufficient steps of
counterbalancing, and did not include null condition trials
with which to quantify the magnitude of any bias effects.
Thus, of 18 brightness matching studies carried out at
photopic levels only five were considered to be reliable
[4,19,22,24,25]: nine were suggested to be unreliable due to
lack of counterbalancing and four studies failed to provide
sufficient data with which to make this analysis [14]. It has
been shown that in the matching task there is negligible
difference in outcome (illuminance ratio) when using
different visual objectives (e.g. equal brightness, equal
clarity or equal appearance) [26] and this conclusion
enables the findings from the five reliable studies to be
collated.
Discrimination
In the discrimination task, two stimuli of different SPD are
presented at a range of different illuminances; at each
presentation, the test participant reports which is the
brighter. Previous work has used rapid sequential
presentation of the two stimuli at the same spatial location
(Figure 2) [e.g. 12,27] or simultaneous evaluations (Figure
1) [28]. Whilst judgement of the brighter of a pair of
stimuli is a more precise and repeatable task than is
adjustment for equal brightness, the discrimination task can
be biased through the range of stimulus magnitudes
selected. Identification of relative illuminances for equal
brightness demands the discrimination task is repeated at a
range of illuminances, and two studies have shown
43
stimulus range bias is sufficient to affect the outcome of
discrimination tasks [29,30].
a higher mid-point range of luminances. Thus, the stimulus
range affected the brightness judgment; a stimulus was
made to appear brighter or dimmer than the reference by
changing the range of luminances in which it was
presented.
Investigation of the discrimination task in research of lamp
SPD and spatial brightness is on-going. It has been used in
only a few studies, and these have not tended to use nullcondition trials (stimuli of identical SPD and illuminance)
which would otherwise provide evidence to validate the
method.
Category Rating
Figure 2. A single spatial location illuminated using two
different sources of light in rapid succession. This is a
sequential evaluation.
Fotios & Cheal examined stimulus frequency bias, the
distribution of illuminances above and below that which
produces the same brightness as the reference stimulus
[29]. Biased here means there are, for example, more cases
when the test stimulus is dimmer than the reference than
when it is brighter. Consider the observation of two lamps
of different SPD at the illuminances at which they are
expected (perhaps as according to parallel studies) to
appear equally bright; a biased stimulus frequency causes
identification of brighter stimulus to be unfairly biased
toward the stimulus which has been less frequently
identified as brighter in preceding trials. This can suggest a
statistically significant difference between two stimuli
when none exists. This may arise from subjects’
preconceptions of chance, that each of a pair of stimuli
must be correct (brighter) on an equal number of trials. To
counter stimulus frequency bias, the number of stimulus
magnitudes should be equally divided about that giving
equal brightness.
Teller, Pereverzeva & Civan [30] sought brightness
judgments of small red and blue targets presented on a
white monitor screen. For each colour, a range of targets
varying in luminance were presented in random order, and
observers reported whether the target was brighter or
dimmer than the surround.
Three ranges of target
luminance were used in successive trials – for the red target
these ranges had mid-point values of -0.6, -0.3 and 0.1 log
luminance relative to the white surround. Typically 11
target stimuli were used in each range, increasing in steps
of 0.05 log units. It was found that a stimulus judged
brighter than the surround on 100% of trials with a target
range of lower mid-point luminance, was also judged
dimmer than the (identical) surround on 100% of trials with
44
In the category rating task different lighting conditions are
evaluated separately (Figure 3) and attributes of the visual
environment are rated using a scale that gives only a
limited range of fixed numbers. Poulton [31] discusses
many potential causes of bias within this task. A recent
review applied Poulton’s ideas to research using the
category rating method, and found that this method can
understate the effect of lamp spectrum [32].
Figure 3. A single space is illuminated by one type of lamp.
Judgements are made of this in isolation before proceeding
to the next stimulus. This is a separate evaluation.
Previous lighting studies have tended to use seven-point
rating scales, for example a scale ranging from 1 (dim) to 7
(bright). There is some evidence that test participants are
able to reliably distinguish between approximately seven
categories of a uni-dimensional stimulus, and this is
apparent for a broad range of sensory judgements, but with
more than seven categories confusions become more
frequent [33]. Green and Rao [34] demonstrated that a
response range of around seven categories is able to
adequately represent intended responses; fewer categories
(2 or 3) lead to poor recovery and there are diminishing
returns beyond six categories. The seven-point response
range has commonly been used to define the semantic
differential rating task, e.g.;
–
The semantic differential consists of a set of bipolar,
seven-category rating scales [35].
–
Semantic differential rating scales – a seven category
range between the extremes [36].
There is a tendency for respondents to avoid using the ends
of a scale, to underestimate large sizes and overestimate
small sizes, and this response contraction is enhanced if the
response range has an obvious middle value such as with
the seven-point scale [31]. Such an outcome can be
observed in the findings of previous lighting research:
Wake et al [37] used 7-point scales, and for their brightness
rating they concluded “the differences among lamps are
extremely small”; Akashi & Boyce [11] used 5-point scales
(-2 to +2) with a middle neutral point marked ‘0’ and found
“The mean ratings … do not indicate any strong opinions,
i.e. all mean responses are around neutral”. Because of
potential response contraction bias it is not clear whether
there really is no difference of brightness between the
lamps used in these studies, under the particular conditions
used, or if the mid-point value in the response range
contributed to the test failing to reveal a difference. This
bias can be countered by using a response range with an
even number of response points.
The rating task is affected by the relative numbers of
response categories and stimulus magnitudes [31]. If the
response scale has fewer categories than there are stimuli,
several stimuli will need to be grouped within each
category, and this may hide the difference between two
stimuli when this difference may be small but is
nonetheless real. Consider the study by Boyce & Cuttle
(their Experiment 1) which used 22 stimulus conditions,
including four types of lamp and four illuminances, and a
five-point response range [5]. Their participants would thus
need to group several of the 22 stimuli within each
response category. Their results reveal that only one of the
19 rating items (dim) was found to be significantly affected
by lamp type, and this at p<0.05 may be a Type I error (i.e.
erroneous rejection of the null hypothesis). Differences in
brightness due to illuminance were significant; these may
be more prominent than differences due to SPD and would
thus dominate the response category decision. The use of
too few response categories does not give observers the
opportunity to report whether two SPDs are differently
bright. This response grouping bias can be countered by
using similar numbers of stimulus magnitudes as there are
response categories.
Response contraction in the category rating task can also be
induced by failure to randomise or balance the order in
which stimuli are presented and by failure to anchor the
response range to the stimulus range by visual
demonstration [32].
In a recent review of 17 studies using category rating at
photopic levels to compare brightness effects of lamp
spectrum, only three were considered to provide reliable
data, the remainder having suspected experimental bias or
provided insufficient data to check for such bias [32].
STIMULUS SIZE
Three further aspects of experimental design pertinent to all
psychophysical methods in lighting research are discussed
below: stimulus size, evaluation mode and design of the
illuminated field.
Previous studies have used visual stimuli of a wide range of
sizes, from remote viewing of a bipartite field subtending
2° at the eye [38] or booths subtending around 40° at the
eye [4], to tests placing subjects within lit rooms [5] and
thus giving stimulation of the whole retina - full-field
stimulation. Stimulus size is expected to matter because the
relative distribution of the long, medium and shortwavelength sensitive photoreceptors varies with retinal
location [10]. Whilst full fields are representative of most
real world conditions, it is often easier to set up and
characterise smaller fields in laboratory trials.
Experimental evidence demonstrates that a 10° field
produces different colour matching judgement to a field of
size 102° wide and 50° high [39]; that the difference in
sensitivity between fields of size 9° and 64° is small
relative to the difference between 3° and 9° fields [40]; and
that the average luminance of the horizontal band 40° wide
centred at normal eye height relates well to subjective
ratings of spatial brightness [41]. These data suggest that
subjective evaluations of lighting for full field vision can be
made using scale models, and that the minimum size is
somewhere in the region of 10° to 40°. This proposal will
be examined in further work.
EVALUATION MODE
There are two primary modes of evaluation, joint and
separate [42]. In the separate mode (Figure 3) stimuli are
presented individually, whilst in the joint mode two or
more stimuli are presented in juxtaposition. The joint mode
can be subdivided into simultaneous and successive modes.
In the simultaneous mode (Figure 1), two stimuli are
presented at the same time in adjacent spatial locations; in
the successive mode (Figure 2) the two stimuli are
presented in temporal juxtaposition at the same spatial
location.
Chromatic adaptation
Joint and separate modes of evaluation lead to different
degrees of chromatic adaptation. Chromatic adaptation is
the neutralisation of activity in the opponent colour
channels as the eyes acclimatise to the stimulus. Activity in
the opponent colour channels contributes to brightness [43]
and thus the degree of chromatic adaptation will affect the
size of this contribution.
The time course of chromatic adaptation has been measured
using colour appearance judgements following a change in
adaptation. The data suggest two stages of adaptation. The
initial rapid stage gives approximately 60% chromatic
adaptation in the first five seconds, and is followed by the
slower stage where approximately 90% chromatic
adaptation is reached after 60 seconds; it takes almost two
minutes to reach 100% chromatic adaptation [44,45].
In separate evaluations which allow adaptation to a single
stimulus for two minutes or more, an observer’s white point
becomes the chromaticity of the stimulus. This complete
45
chromatic adaptation reduces the chromatic contribution to
brightness although experimental results suggest it does not
completely eliminate any effect of SPD [32,46].
In simultaneous evaluations the chromatic adaptation state
of the observer is difficult to define. The observer does not
adapt to the individual stimuli but to the mixed spectrum,
giving a white point somewhere between the chromaticities
of the two adapting conditions being considered [47]. In
sequential evaluations the same spatial location is
illuminated by different stimuli in rapid sequential
presentations. Berman et al [12] illuminated a wall
alternately by two different sources, presented for 5
seconds each, and with three alternations between the two
sources. Vrabel et al [27] illuminated a room for three
seconds per source, with a two second dark interval
between them, this cycle being repeated as many times as
required by the observer. With each stimulus presented for
approximately five seconds before alternating to the second
stimulus, the observer’s white point would move toward
the chromaticity of the first stimulus, without actually
reaching it, then towards the chromaticity of the second
stimulus when that is presented, again without reaching it,
and so on. The white point would therefore eventually lie
somewhere between the chromaticities of the two
individual stimuli and the state of chromatic adaptation
would be similar to that for the side-by-side presentation.
Hence the simultaneous and sequential modes of evaluation
will yield similar results when other parameters are also
similar. Studies using joint modes of evaluation tends to
exaggerate differences between stimuli compared to
findings using separate evaluation [46].
Interval bias
While simultaneous evaluations may suffer from positional
bias, the preference for one spatial location over another,
successive evaluations may be affected by interval bias
[48], the preference for one temporal interval over the
other. In brightness discrimination judgements this would
be a tendency to report a particular interval as being
brighter when it is not. Yeshurun et al present experimental
data exhibiting large interval bias in visual judgements,
some favouring the first interval and some the second
interval [48]. Needham defines interval bias as the
overestimation (negative time-error) or underestimation
(positive time-error) of the second of two stimuli presented
in succession [49] and suggests that it changes with
variation of the time interval, or pause, between
presentations of the stimuli: intervals of up to
approximately three seconds tend to result in an
underestimate of the second stimulus, whilst intervals
above approximately three seconds tend to result in an
overestimate of second stimulus.
In their detection task, Jäkel and Wichmann [50] found a
strong bias to the second interval from three of their five
observers, including the expert observer, when using
successive evaluation whilst the simultaneous evaluation
task was virtually unbiased. In their discrimination task,
46
Jäkel and Wichmann found similar sensitivity with
simultaneous and successive tasks but each of their four
naïve observers was still better at the simultaneous
discrimination task than the successive discrimination task
after 20,000 detection trials [50]. Uchikawa and Ikeda
found that matching and discrimination tasks using side-byside brightness comparisons gave more precise results than
did successive presentations [51]. Doubts about the
successive discrimination task lead a recent study to report
that it should be used with caution, if at all [48].
There are two issues regarding use of sequential and
simultaneous evaluation modes that need further
investigation before discussions of previous lighting
research can be resolved. The first relates to the dominant
visual mechanism though which lamp SPD affects spatial
brightness. If this is through the opponent colour channels
[43] then chromatic adaptation is of interest and the
sequential and simultaneous evaluation modes lead to
similar states of chromatic adaptation. Alternatively, it has
been suggested that the spatial brightness response is
mediated by control of pupil size [12] in which case the
sequential evaluation is preferable to the simultaneous
mode because it would allow the pupil to respond to the
SPD of the individual stimuli rather than to the mixed SPD
of both. The second issue is that of interval bias in
sequential evaluation tasks. It is unfortunate that previous
studies of lamp SPD and brightness using discrimination
between successive stimuli have tended not to include a
null condition trial so there are no data with which to
quantify the magnitude of any such bias.
Further research has been carried out at Sheffield
University and Pennsylvania State University to compare
the simultaneous and sequential modes of evaluation and
preliminary results are reported below.
Experimental data: Sheffield
Fotios & Cheal previously reported the results of brightness
matching and brightness discrimination tests, both using
simultaneous evaluations [21]. This work is currently being
repeated using sequential evaluation.
The simultaneous evaluations [21] used a pair of side-byside booths, with separate light sources simultaneously
illuminating each booth. Light was transported to the top of
each booth through a light pipe, using an iris in the pipe to
adjust illuminance and avoid any effect on the SPD or
spatial distribution of light in the visible chamber. The
sequential evaluations used only one of these booths,
presenting a visual field of approximately 37° high and 36°
wide. Light from two different lamps was transported to the
top through separate light pipes, again using irises to adjust
the illuminance. Luminance measurements show negligible
differences in spatial distribution between lamps, between
light from the two light pipes and between levels of
dimming.
The two stimuli were presented in rapid succession:
stimulus A (5s); dark interval (300ms); stimulus B (5s);
dark interval (300ms); stimulus A (5s) etc. These durations
were chosen to repeat the conditions used by Berman et al
[12]. For the matching test this procedure was followed
until the test participant was satisfied with their brightness
match. For the discrimination test the number of repeats
was limited to three.
illuminance stimulus, thus avoiding a stimulus frequency
bias [29].
Results from the ten test participants used to date are shown
in Table 3, in comparison with results from the previous
trials using simultaneous evaluation [21]. The fourparameter logistic equation was used to derive the
illuminance ratio for equal brightness from the results of
the discrimination tests.
Four lamps were used; a standard high pressure sodium
(HPS 70W), a compact fluorescent (CFL) and two types of
metal halide (MH1, MH2), as defined in Table 2, these
being the lamps used in previous work [21]. Using the HPS
as the reference source gave four lamp combinations
including a null condition. The order in which lamp pairs
were presented was balanced between subjects.
Evaluation
mode
Lamp
Illuminance ratio at equal brightness
HPS/
HPS
CFL/
HPS
MH1/
HPS
MH2/
HPS
CCT (K)
CRI
Brightness matching
Sequential
0.99
0.68
0.74
0.70
Simultaneous
0.99
0.72
0.73
0.71
HPS
70W/150W SON-T
Pro
2000
25
CFL
55W PL-L
3000
82
Brightness discrimination
MH1
70W CDO-TT
2800
83
Sequential
1.01
0.67
0.69
0.66
MH2
70W CDM-T
4200
92
Simultaneous
1.00
0.59
0.68
0.64
MH3
150W CDM-TT
4200
92
Table 2. Lamps used by Fotios & Cheal in brightness
matching and discrimination tests.
Table 3. Comparison of illuminance ratios for equal
brightness determined using matching and discrimination
tasks with simultaneous (n=21) and sequential (n=10)
modes of evaluation. Simultaneous data as previous
reported [21]; sequential data not previously reported.
In sequential brightness matching trials one of the two
lamps in a pair was set by the experimenter to the reference
illuminance. The test participant used the dimming control,
a three-turn rotary dial, to match the lighting as-near-aspossible for equal brightness. This procedure was repeated
by each test participant to counterbalance dimming
application and dimming direction. When the HPS lamp in
a pair was used as the stimulus of fixed illuminance the
reference illuminance was 7.5 lux, measured at the centre
of the floor of the booth, this being a pilot study for further
research of street lighting. When the MH and CFL lamps
were used as the stimulus of fixed illuminance the
reference illuminance was 5.0 lux, this expected to be
approximately equally bright as the HPS at 7.5 lux and thus
maintain a similar state of adaptation in both cases.
Two observations are drawn from Table 3. Firstly, there
appears to be little difference in illuminance ratio for a
particular lamp pair between sequential and simultaneous
evaluation modes, for both the matching and discrimination
tasks, and thus that the evaluation mode does not
significantly affect operation of the visual mechanism(s)
responsible for SPD effects on spatial brightness. Secondly,
brightness discrimination appears to suggest illuminance
ratios that depart slightly further from unity than those from
the matching task. Data from the null condition trials, the
HPS/HPS lamp pair, suggest negligible experimental bias.
In sequential brightness discrimination trials, lighting from
one lamp in each pair was set to the reference illuminance
and lighting from the other lamp was set to a range of
illuminances. At each presentation the test participant
reported which interval appeared brighter, a forced choice
task. This procedure was repeated by each test participant
to counterbalance lamp nomination as reference and
variable stimulus. When the HPS lamp in a pair was used
as the stimulus of fixed illuminance, this being 7.5 lux, the
CFL and MH lamps were presented at 2.0, 3.0, 5.0, 7.5,
10.0 lux. When the MH or CFL lamps in a pair was used as
the stimulus of fixed illuminance, this being 5.0 lux, the
HPS lamp was presented at 3.0, 5.0, 7.5, 10.0 and 15.0 lux.
These ranges were chosen with expectation that the middle
value would tend to appear equally bright as the fixed
Results of the sequential brightness judgements and
comparison of these with results of the simultaneous tests
will be submitted for peer reviewed publication upon
completion of the trials. In addition to using illuminance
ratios to compare the size of any SPD effect upon spatial
brightness, this analysis will also analyse precision and
interval bias.
Experimental data: Penn State
Brightness judgements at photopic levels were made using
side-by-side and rapid sequential discrimination tasks. The
visual field in each case was one, or both, of a pair of
identical empty rooms with approximate dimensions of
3.0m (wide) x 3.6m (deep) x 2.7m (height). All surfaces
within the subject’s field-of-view were neutral gray. The
rooms were fitted with indirect luminaires, suspended about
400mm from the ceiling, and these had continuous rows of
RGB LEDs. Four stimulus conditions were used, these
47
being the four possible combinations of two correlated
colour temperatures (3000K, 7500K, both on the blackbody
locus) and two luminances (24 and 30 cd/m2) as measured
at eye height on the surface of the wall directly in front of
the subject. The ten paired combinations of these four
stimuli included four null conditions and the left/right
stimulus position (simultaneous evaluations) and the
first/second stimulus interval (sequential evaluations) were
counterbalanced for the between-stimulus pairs, giving
sixteen paired comparisons.
In the simultaneous evaluations the rooms were observed
from a seated position just outside of the rooms, with the
partition between the rooms aligned with the subject’s
sagittal plane. In the sequential evaluations the subject was
seated within the left-hand room, In all cases a
chin/forehead rest was used to maintain consistency in the
viewing field across trials and subjects. For the
simultaneous evaluations, presentation durations were not
limited. For the sequential presentations each stimulus was
presented for 5s with a 25ms dark interval and subjects
were instructed to view at least three sets of alterations (i.e.
ABABAB) before making their choice about which light
setting was brighter. This is comparable to the method
employed by Berman et al [12].
The tests were carried out by 47 participants using a
repeated measures procedure. Full results will be submitted
for publication in a peer reviewed journal and here we
focus on the comparison of results from the sequential and
simultaneous evaluations.
Stimulus pair
Distribution of judgements of brighter
stimulus
Simultaneous
evaluation
VISUAL FIELD
In previous work, visual fields have ranged from uniform,
neutral surfaces, to interior spaces with coloured surfaces
and containing objects. Whilst the neutral field enables
analysis of brightness effects purely due to differences in
SPD, the coloured environment better represents most real
world interiors. Two questions are raised. Firstly, are
results obtained in studies using coloured environments
transferable to other settings? Secondly, are test
participants attracted to objects in the observed field such
that their response is dominated by foveal vision rather than
full field vision? Brightness matching trials were carried
out using four different field designs to explore the
transferability of results from one setting to another [52].
These tests were carried out at mesopic levels, this again
being a pilot study for work investigating lighting for
residential streets.
Method
Sequential
evaluation
The four illuminated fields are shown in Figure 4. These
are:
Achromatic: These are two side-by-side booths. The
interior surfaces of the booth were painted matt grey
(Munsell N5, r = 0.2).
1
2
1
2
1
2
A
D
6
88
4
90
B
C
72
22
90
4
A
C
28
66
39
55
B
D
31
63
49
45
C
D
2
92
0
94
A
B
2
92
0
94
Table 4. Comparisons of brightness discrimination
judgements obtained using sequential and simultaneous
modes of evaluation. (n=94). These stimuli are A (3000K,
24 cd/m2), B(3000K, 30 cd/m2), C(7500K, 24 cd/m2), D
(7500K, 30 cd/m2).
Table 4 summarises the results. The frequency with which
the stimulus in a pair was reported to be brighter is similar
for both the sequential and simultaneous evaluations.
McNemar’s test suggests the difference is significant
(p<0.01) only for the BC and BD lamp pairs. Conclusions
48
drawn about statistical significance related to effects of
SPD and luminance were identical with both methods. The
room with the higher luminance was selected as brighter
irrespective of CCT, and at equal luminance CCT was
unrelated to brightness perception. There are subtle
differences in some of the contrasts that were studied and
these are presently under further investigation, but the
general conclusion is that both experimental methods will
lead to comparable results. This is not unexpected since
both methods place the subject in a state of mixed
adaptation. Preliminary analysis suggests that any bias
between the right-hand and left-hand rooms in the
simultaneous evaluation, or between the first and second
intervals in the sequential evaluation, was negligible.
Coloured Objects: Pyramids made from coloured card
(red, green, blue and yellow) were placed on the floor of
the achromatic environment. This is the field design used in
previous work [21].
Coloured Surfaces: Approximately one third of the
visible interior surfaces of the achromatic booths were lined
with unglazed quarry tiles in three colours (red, beige and
black) simulating brick, stone and asphalt surfaces. The
proportion of colour was determined from a brief survey of
residential streets in Sheffield, a city in the UK.
Uniform Field: The front openings of the achromatic
booths were covered with two sheets of acrylic diffuser of
neutral transmittance. This provided a neutral and uniform
stimulus field.
Within these treatments, data for each subject are the mean
of the eight trials carried out per lamp pair and field design.
Field
design
Achromatic
Coloured Objects
Coloured
surfaces
Coloured
objects
Achromatic
Uniform
field
Coloured Surfaces
Uniform Field
Figure 4. Visual fields used to compare effect of colour and
objects on the results of brightness matching tests. Note:
only the left-hand field is shown; the right-hand field was a
mirror image.
The test participant’s task was to adjust the illuminance in
one booth to match the brightness produced by the
reference illuminance (7.5 lux) in the other booth. For the
uniform field design the reference illuminance was set to
achieve an average luminance of the front surface of the
reference field equal to the average luminance (0.38 cd/m2)
of the walls in the other field designs. In trials, both sides
were of identical design, the left-hand booth being a mirror
image of the right-hand booth.
Four lamps were used, these being similar to the lamps
used in previous work [21]. These were high pressure
sodium (HPS 150W), compact fluorescent (CFL), and two
types of metal halide (MH1, MH3) as defined in Table 2.
Lamp MH1 was used as the reference stimulus, and thus
there were four lamp pairs including a null condition.
Each of the four lamp pairs were matched four times,
counterbalancing the initial illuminance of the variable
stimulus (set by the experimenter to be obviously higher or
lower than the fixed stimulus) and counterbalancing the
designations of fixed and variable booth. Each trial was
repeated twice. The order in which the four lamp pairs were
used and the booth in which the reference lamp (MH1) was
located were balanced across the ten test participants (age
range 25-54 years; 7 female, 3 male). Each participant saw
all experimental conditions, a repeated-measures
procedure, and hence made 128 brightness matches.
Results
The mean illuminance ratios for the four lamp
combinations and the four field types are shown in Table 5.
Mean illuminance ratio for each lamp
combination
HPS/
MH1
CFL/
MH1
MH1/
MH1
MH3/
MH1
1.35
0.93
0.99
0.92
(0.13)
(0.09)
(0.08)
(0.08)
1.24
0.90
0.96
0.92
(0.11)
(0.09)
(0.07)
(0.07)
1.24
0.88
0.98
0.93
(0.14)
(0.12)
(0.06)
(0.14)
1.22
0.90
0.96
0.90
(0.27)
(0.08)
(0.08)
(0.12)
Table 5. Mean illuminance ratios
(and standard
deviations) from side-by-side brightness matching trials (n
= 10) using four different visual fields.
The effect of field design can be seen by comparing
illuminance ratios for the four field designs under each
lamp combination. The mean illuminance ratios in Table 5
suggest all four field designs yield similar illuminance
ratios under the MH1/MH1, MH3/MH1 and CFL/MH1
lamp pairs; under the HPS/MH1 lamp pair there appears to
be a difference between the coloured surfaces field and the
other three field designs. Two-way repeated measures
ANOVA (lamp pairs x field design) suggests that the effect
of field design is not statistically significant, although it is
close (p=0.082). Differences between field-designs were
examined using paired t-tests on all combinations of field
design within each lamp pair. Of these 24 analyses, only
two differences are significant, and both of these are for the
HPS/MH1 lamp pair; coloured surfaces vs. coloured
objects (p=0.003), and coloured surfaces vs. achromatic
(p=0.008).
Effects of field design on brightness judgements were
considered using the current results and also the results
from two previous studies at photopic levels [5,19] in
which surface colour and the presence of an object were
varied. Three conclusions were drawn:
1. Brightness matching using illuminated achromatic
interior environments produces the same outcome
(illuminance ratio at equal brightness) as brightness
matching using illuminated flat surfaces of neutral spectral
reflectance.
2. The insertion of coloured objects into an achromatic
environment does not affect the outcome.
3. An environment with coloured surfaces produces the
same outcome as an achromatic environment, and there is
no significant effect with the level of colourfulness.
49
These findings will be submitted to a peer reviewed journal
for publication.
SUMMARY
All experimental methods contain bias. This is not
necessarily a problem if there are data, such as nullcondition data, that enable bias effects to be estimated.
Robust conclusions demand the same stimuli are compared
using a variety of psychophysical methods and if these tend
to agree then greater confidence can be placed in the
results. Whilst a few studies have done this [11,19,24,27],
most do not, hence the meta analysis being carried out by
the authors.
The consideration of research methods discussed in this
article suggests that much of the previous work provides an
unreliable estimate of lamp SPD effects on brightness.
Frequently, this is because the reported method reveals
experimental error, or because there are insufficient data
reported to determine whether a potential experimental
error has been countered. At present, the analysis suggests
that data from only 14 of 60 previous studies are reliable;
these are shown in Table 6.
The next stage of this research is to develop a tool to enable
prediction of lamp SPD effects on spatial brightness, hence
to guide the selection of lamp type and illuminance. A
common limitation of the experimental work is that lamps
are selected from those commercially available using
coarse indicators of lamp spectral characteristics, such as
Colour Rendering Index, Correlated Colour Temperature or
the ratio of scotopic to photopic lumens (S/P). It is less
common for researchers to create custom illuminants that
have spectra intentionally designed to manipulate an
underlying mechanism of vision; only two studies appear to
have done so [12,28].
Three categories of prediction tool are colour appearance
models; the S/P ratio, e.g. consideration of the intrinsically
photoreceptive retinal ganglion cell (ipRGC); and lamp
colour characteristics, e.g. relative values of CCT, CRI,
and gamut area. Allied discussions include consideration
of how the effect of lamp SPD might be applied in practice
and comparison with effects on visual effort and circadian
response.
Study
Response task
Field size
Evaluation mode
Akashi & Boyce, 2006 [11]
Yes/No response to statements.
Full field
Separate
Berman et al, 1990 [12]
Discrimination
Full field
Sequential
Boyce, 1977 [19]
Matching
Full field
Simultaneous
Boyce, Akashi, Hunter & Bullough, 2003
[53]
Yes/No response to statements.
Full field
Separate
Boyce & Cuttle, 1990 (Experiment 2) [5]
Category rating
Full field
Separate
Flynn & Spencer, 1977 [54]
Category rating
Full field
o
Separate
o
Fotios & Gado, 2005 [26]
Matching
40 high, 72 wide
Simultaneous
Fotios & Levermore, 1997 [22]
Side-by-side Matching
22o high, 38o wide
Simultaneous
Houser, Tiller & Hu, 2004 [28]
Discrimination
Full field
Simultaneous
Hu, Houser & Tiller, 2006 [24]
Matching
Full field
Simultaneous
Ray, 1989 [20]
Adjust to preferred illuminance
Full field
o
Separate
o
Thornton & Chen, 1978 [4]
Matching
30 high x 50 wide
Simultaneous
Vrabel, Bernecker & Mistrick, 1998 [27]
Discrimination
Full field
Sequential
Vrabel, Bernecker & Mistrick, 1998 [27]
Category rating
Full field
Separate
Table 6. Tests suggested to give reliable demonstration of SPD effect on brightness at photopic levels
ACKNOWLEDGMENTS
We would like to thank Chris Cheal (University of
Sheffield) and Mike Royer (Pennsylvania State University)
for their contributions to work reported in this article.
Experimental work at Sheffield University was carried out
with support from the Engineering and Physical Sciences
Research Council (EPSRC) grant reference EP/F035624/1.
The experimental work at The Pennsylvania State
University was carried out with support from Project
50
CANDLE, which is an industry-university partnership with
primary support from the IALD Education Trust.
REFERENCES
1. Aston, S.M. and Bellchambers, H.E. Illumination,
colour rendering and visual clarity. Lighting Research
& Technology, 1,4 (1969), 259-261.
2. Bellchambers, H.E. and Godby, A.C. Illumination,
colour rendering and visual clarity. Lighting Research
and Technology 4,2 (1972), 104-106.
3. Worthey, J.A. An analytical visual clarity experiment.
Journal of the Illuminating Engineering Society 15,1
(1985), 239-251.
Journal of the Optical Society of America 45,7 (1955),
507-514.
19. Boyce, P.R. Investigations of the subjective balance
between illuminance and lamp colour properties.
Lighting Research & Technology, 9 (1977), 11-24.
4. Thornton, W.A. and Chen, E. What is visual clarity?
Journal of the Illuminating Engineering Society, 7
(1978), 85-94.
20. Ray, S. The evaluation of a daylight tungsten lamp for
task lighting. Under-graduate research project,
Department of Human Sciences, Loughborough
University. (1989).
5. Boyce, P.R. and Cuttle, C. Effect of correlated colour
temperature on the perception of interiors and colour
discrimination. Lighting Research & Technology 22,1
(1990), 19-36.
21. Fotios, S.A. and Cheal, C. Lighting for subsidiary
streets: investigation of lamps of different SPD. Part 2 –
Brightness. Lighting Research & Technology, 39,3
(2007), 233-252.
6. International Energy Agency, Lights Labours Lost:
Policies for energy efficient lighting, (2006).
22. Fotios, S.A. and Levermore, G.J. The perception of
electric light sources of different colour properties.
Lighting Research & Technology, 29,3 (1997), 161-171.
7. Department of the Environment, Transport and the
Regions. Energy Consumption Guide 19; Energy use in
offices, (2000).
8. US Energy Information Administration. Table E5,
Electricity Consumption (kWh) by End Use for All
Buildings, 2003. (Accessed April 30, 2009)
http://www.eia.doe.gov/emeu/cbecs/cbecs2003/detailed
_tables_2003/detailed_tables_2003.html#enduse0.
23. Houser, K.W., Tiller, D.K. and Hu, X. Prototype
demonstration of vision-tuned fluorescent lamps,
Energy Innovations Small Grant (EISG) Program, Final
Report for the California Energy Commission. (2003).
24. Hu, X., Houser, K.W. and Tiller DK. Higher colour
temperature lamps may not appear brighter. Leukos 3,1
(2006), 69-81.
9. Rea, M.S. and Ouellette, M.J. Relative visual
performance: A basis for application, Lighting Research
& Technology, 23,3 (1991), 135-144.
25. Fotios, S.A. and Levermore, G.J. Visual perception
under tungsten lamps with enhanced blue spectrum.
Lighting Research & Technology, 27,4 (1995), 173-179.
10. Boyce, P.R. Human Factors in Lighting. London:
Taylor & Francis, (2003).
26. Fotios, S. and Gado, T. A comparison of visual
objectives used in side-by-side matching tests. Lighting
Research & Technology, 37,2 (2005), 117-131.
11. Akashi, Y. and Boyce, P.R. A field study of illuminance
reduction. Energy & Buildings 38,6 (2006), 588-599.
12. Berman, S.M., Jewett, D.L., Fein, G., Saika, G. and
Ashford, F. Photopic luminance does not always predict
perceived room brightness. Lighting Research &
Technology 22,1 (1990), 37-41.
13. Fotios, S.A. Experimental conditions to examine the
relationship between lamp colour properties and
apparent brightness. Lighting Research & Technology,
34,1 (2002), 29-38.
14. Fotios, S.A., Houser, K.W. and Cheal, C.
Counterbalancing needed to avoid bias in side-by-side
brightness matching tasks. Leukos, 4,4 (2008), 207-223.
27. Vrabel, P.L., Bernecker, C.A. and Mistrick, R.G. Visual
performance and visual clarity under electric light
sources: Part II - Visual Clarity. Journal of the
Illuminating Engineering Society 27,1 (1998), 29-41.
28. Houser, K.W., Tiller, D.K. and Hu, X. Tuning the
fluorescent spectrum for the trichromatic visual
response: A pilot study. Leukos 1,1 (2004), 7-22.
29. Fotios, S.A. and Cheal, C. The effect of a stimulus
frequency bias in side-by-side brightness ranking tests,
Lighting Research & Technology, 40,1 (2008), 43-54.
30. Teller, D.Y., Pereverzeva, M. and Civan, A.L. Adult
brightness vs. luminance as models of infant
photometry. Journal of Vision, 3 (2003), 333–346.
15. Fotios, S.A. An error in brightness matching associated
with the application of dimming. Lighting Research &
Technology, 33,4 (2001), 223-231.
31. Poulton, E.C. Bias in quantifying judgements. Lawrence
Erlbaum Associates Ltd; London, (1989).
16. Fotios, S.A. and Cheal, C. Evidence for response
contraction bias in side-by-side matching tasks. Lighting
Research & Technology, 39,2 (2007), 159-169.
32. Fotios, S.A. and Houser, K.W. Research methods to
avoid bias in categorical ratings of brightness. Leukos,
5,3 (2009), 167-181.
17. Fotios, S.A. Lamp colour properties and apparent
brightness: a review. Lighting Research and Technology
33,3 (2001), 163-181.
33. Miller, G.A. The magical number seven, plus or minus
two: Some limits on our capacity for processing
information. The Psychological Review 63,2 (1956), 8197.
18. Kinney, J.A.S. Sensitivity of the eye to spectral
radiation at scotopic and mesopic intensity levels.
51
34. Green, P.E. and Rao, V.R. Rating scales and
information recovery – how many scales and response
categories to use? Journal of Marketing 34,3 (1970),
33-39.
44. Fairchild, M.D. and Reniff, L. Time course of chromatic
adaptation for color-appearance judgements. Journal of
the Optical Society of America, series A, 12,5 (1995),
824-833.
35. Tiller, D.K. and Rea, M.S. Semantic differential scaling:
Prospects in lighting research. Lighting Research &
Technology 24,1 (1992), 43-52.
45. Shevell, S.K. The time course of chromatic adaptation.
Color Research & Application, Supplement 26 (2001),
S170-S173.
36. Houser, K.W. and Tiller, D.K. Measuring the subjective
response to interior lighting: paired comparisons and
semantic differential scaling. Lighting Research &
Technology 35,3 (2003), 183-198.
46. Fotios, S.A. Chromatic adaptation and the relationship
between lamp spectrum and brightness. Lighting
Research & Technology, 38,1 (2006), 3-17.
37. Wake, T., Kikuchi, T., Takeichi, K., Kasama, M. and
Kamisasa, H. The effects of illuminance, color
temperature and colour rendering index of light sources
upon comfortable visual environments in the case of the
office. Journal of Light & Visual Environment 1,2
(1977), 31-39.
38. Alman, D.H. Errors of the standard photometric system
when measuring the brightness of general illumination
light sources, Journal of the Illuminating Engineering
Society, (October 1977), 55-62.
47. Braun, K.M., Fairchild, M.D. and Alessi, P.J. Viewing
techniques for cross-media image comparisons. Color
Research & Application, 21,1 (1996), 6-17.
48. Yeshurun, Y., Carrasco, M. and Maloney, L.T. Bias and
sensitivity in two-interval forced choice procedures:
Tests of the difference model. Vision Research, 48
(2008), 1837-1851.
49. Needham, J.G. The time-error as a function of
continued experimentation. The American Journal of
Psychology, 46,4 (1934), 558-567.
39. Hu, X. and Houser, K.W. Large-field colour matching
functions. Colour Research & Application, 31,1 (2006),
18-29.
50. Jäkel, F. and Wichmann, F.A. Spatial four-alternative
forced-choice method is the preferred psychophysical
method for naïve observers, Journal of Vision, 6 (2006),
1307-1322.
40. Kokoschka, S. and Adrian, W.K. Influence of field size
on the spectral sensitivity of the eye in the photopic and
mesopic range. American Journal of Ophthalmic &
Physiological Optics, 62,2 (1985), 119-126.
51. Uchikawa, K. and Ikeda, M. Accuracy of memory for
brightness of colored lights measured with successive
comparison method. Journal of the Optical Society of
America, series A, 3 (1986), 34-39.
41. Loe, D.L., Mansfield, K.P. and Rowlands, E.
Appearance of lit environment and its relevance in
lighting design: Experimental study. Lighting Research
& Technology, 26,3 (1994), 119-133.
52. Fotios, S.A. and Cheal, C. A Pilot Study Investigating
How the Design of the Illuminated Field Affects
Brightness Matching. (article in progress).
42. Hsee, C.K., Loewenstein, G.F., Blount, S. and
Bazerman, M.H. Preference reversals between joint and
separate evaluations of options: A review and
theoretical analysis. Psychological Bulletin 125,5
(1999), 576-590.
43. Yaguchi, H. and Ikeda, M. Contribution of opponentcolour channels to brightness. Eds. Mollon, J.D. and
Sharpe, L.T. Colour Vision: Physiology &
Psychophysics. Academic Press Inc. (London) Ltd;
London. 353-360, (1983).
52
53. Boyce, P.R., Akashi, Y., Hunter, C.M. and Bullough,
J.D. The impact of spectral power distribution on the
performance of an achromatic visual task. Lighting
Research & Technology, 35,2 (2003), 141-161.
54. Flynn, J.E. and Spencer, T.J. The effects of light source
colour on user impression and satisfaction. Journal of
the Illuminating Engineering Society, 6 (1977), 167179.
Ecological Measurements of Light Exposure,
Activity, and Circadian Disruption
in Real-world Environments
Daniel Miller, MS,1 Andrew Bierman, MS,1 Mariana G. Figueiro, PhD,1
Eva S. Schernhammer, MD, DrPH,2 Mark S. Rea, PhD1
1
Lighting Research Center
Rensselaer Polytechnic Institute
21 Union Street
Troy, NY 12180 USA
+1 518 687-7100
ream@rpi.edu
ABSTRACT
Presented is an overview of the recently developed
Daysimeter, a personal measurement device for recording
activity and circadian light-exposure. When the Daysimeter
is worn on the head, two light sensors near the eye are used
to estimate circadian stimulus (CSA) exposures over
extended periods of time. Phasor analysis combines the
measured periodic activity-rest patterns with the measured
periodic light-dark patterns to assess behavioral circadian
entrainment/disruption. As shown, day-shift and rotatingshift nurses exhibit remarkably different levels of
behavioral circadian entrainment/disruption. These new
ecological measurement and analysis techniques may
provide important insights into the relationship between
circadian disruption and well-being.
Keywords
Circadian rhythms, phasor analysis, circadian disruption,
Daysimeter, circadian light, phototransduction
INTRODUCTION
The synchronization of the endogenous master clock to
local time on earth is governed by a robust and regular 24hour pattern of light and dark on the retina. If the light-dark
pattern is not above threshold, or if its period is something
other than 24-hours, human biology is disrupted, from
single cells to overt behavior [6, 14]. Circadian disruption
can manifest itself in poor sleep, digestion, reduced
attention and performance [8, 9, 15]. Over time this
disruption can lead to serious maladies such as
cardiovascular disease [24], diabetes and obesity [1], and
cancer [17, 19, 20]. Despite the plethora of animal and
epidemiological research demonstrating the negative
impact of circadian disruption on health, surprisingly little
is known about the levels of circadian disruption actually
experienced by people in different types of living
conditions and environments.
2
Brigham and Women’s Hospital and
Harvard Medical School
Channing Laboratory, 3rd Floor
181 Longwood Avenue
Boston, MA 02115 USA
+1 617 525-4648
This paper describes a measurement tool for collecting
ecological data on circadian light-dark stimuli and restactivity patterns along with methods of analyzing such data
so that circadian entrainment and disruption can be
assessed. Results of using such techniques to study a cohort
of nurses engaged in both day- and rotating-shift work is
also presented.
INSTRUMENTATION
The Daysimeter™ was developed as a personal circadian
light exposure and activity meter to measure circadian
light-dark and activity-rest patterns in the field [3]. It was
necessary to develop a sophisticated, small photometer
calibrated in terms of the spectral-spatial-intensity response
of the human circadian system. Moreover, since light must
be incident on the retina to be effective, the Daysimeter had
to be designed to place the light sensors near the plane of
one cornea. Extending from a number of basic studies on
circadian phototransduction [2, 4, 21], and physiological
optics [22], a model was developed for human circadian
phototransduction [12], and this model forms the basis of
the Daysimeter’s photometric response characteristics. To
measure
activity,
the
Daysimeter
incorporates
accelerometers that respond to head movements and
orientation with respect to Earth’s gravitational field.
Activity and light are measured together at regular time
intervals and electronically stored. In addition, the
Daysimeter logs its operating temperature. Using practical
power management techniques, the Daysimeter can gather
light, activity and temperature data for up to 30 days for
subsequent analysis.
Two calibrated photosensors are employed to measure
optical radiation in close proximity to the cornea. Based
upon the spatial sensitivity model of the retina developed
by Van Derlofske et al. [22], both sensors have a nearly
cosine spatial response meaning that the sensors are most
sensitive to light at normal incidence with sensitivity at
53
other angles decreasing proportional to the cosine of the
incident angle. One sensor is a conventional glass-filtered
silicon photodiode (Hamamatsu model S1286, custom glass
filter) having a spectral sensitivity closely matching the
standard photopic luminous efficiency function, V(!) [5].
The other sensor is a short-wavelength (blue) sensor
fabricated from a gallium arsenide phosphide (GaAsP)
photodiode (Hamamatsu model GA5645), having an
intrinsic long wavelength response cutoff at approximately
580 nm, together with a UV blocking glass filter (Schott
GG395). This “blue” sensor has a spectral response
peaking at approximately 460 nm with an 80 nm fullwidth-half-maximum (FWHM) bandpass.
Once the optical radiation data obtained by the Daysimeter
from both sensors are downloaded to a computer,
“circadian light” levels can be approximated using postprocessing algorithms based upon the model of circadian
phototransduction by Rea et al. [12]. The values of
circadian light levels are scaled so that 1000 lux of CIE
Illuminant A (an incandescent blackbody radiator at 2856
K) is equivalent to 1000 circadian light units (CLA). Four
spectral sensitivity functions are used in the model: the
scotopic luminous efficiency function, V’(!) [5], based on
rod sensitivity, V10(!) based upon the S, M and L cone
fundamentals [16], the S-cone fundamental [18], and a
standard photo-opsin emulating melanopsin contained
within the intrinsically photosensitive retinal ganglion cell
(ipRGC; [2]) and having a peak spectral response at 480
nm and a half-bandwidth of 95 nm. Briefly, in the model
the cone fundamentals form a spectrally opponent (blue vs.
yellow [7]) input to the ipRGC which sends circadian light
signals to the SCN. The modeled rod response suppresses
output from the ipRGC when the blue-yellow opponent
signal is positive, with diminishing suppression at higher
irradiance levels as rods become more fully bleached. A
negative blue-yellow opponent signal, however, produces a
response determined solely by the ipRGC (no cone input
and no rod suppression). The Daysimeter system uses its
two spectral channels to approximate the four spectral
channel input in the model. Discrepancies were minimized
between the two-channel estimates and the four-channel
model calculation of CLA for several practical light sources
using a conventional least-squares technique. Based upon
those results, algorithms were developed to minimize
measurement errors of CLA from the Daysimeter; the
results of that analysis are shown in Figure 1. Spectral
mismatch errors of the photopic channel for the light
sources in Figure 1 are less than 2% except for the 470 nm
LED which has a photopic spectral mismatch error of 8%.
The human circadian system response to light (CLA), as
measured by acute nocturnal melatonin suppression [10],
follows a logistic function [25]. This response function is
used to transform the CLA values into circadian stimulus
values (CSA). CSA is considered to be a better measure of
the effectiveness of the light stimulus for the human
circadian system because it is defined in terms of the
54
Figure 0. Errors in estimating a fixed level of
circadian stimulus (1000 CLA) from the model of
circadian phototransduction by Rea et al. [12]
generated by several common light sources using the
algorithms applied to the two-sensor data acquired
by the Daysimeter.
circadian system’s input-output relationship, including both
threshold and saturation.
The Daysimeter model used in the present study had two
orthogonally oriented accelerometers contained within a
single electronic sensor package (Analog Devices, model
number ADXL330) and mounted on the Daysimeter’s
circuit board. The outputs of the sensor package are
voltages that are proportional to the instantaneous
acceleration of each accelerometer. These voltages are
converted to digital values using the 12-bit analog-todigital converter of the microprocessor (Texas Instrument
MSP430F169) that controls operation of the Daysimeter.
The digital data are acquired once per second, and then
used to calculate an activity index every 30 s using the
following equation:
,
where SSx and SSy are the sums of the squared deviations
from the mean digital value for each accelerometer (x and
y) over the 30-second logging interval, n is the number of
samples (30), and k is a calibration factor converting the
measured output voltage of the accelerometers in arbitrary
analog-to-digital converter counts to units of g-force (1 gforce = 9.8 m/s"). In other words, the activity index is the
root-mean-square (rms) deviation in acceleration in two
dimensions measured for every 30-second logging interval.
ANALYSIS METHODS
Phasor analysis, a technique based on signal processing
techniques [11], makes it possible to interpret the light and
activity data, sampled together over consecutive multiple
days (usually seven days for the data presented here), in
terms of the phase and magnitude of the joint 24-hour
patterns. The correlations between the periodic changes in
light and in activity are first determined by calculating the
circular correlation function of the light and activity time
series. The circular correlation function reveals how the
correlation (r, not r!) between light and activity change as a
function of the timing difference, or phase between them
(Figure 2). The circular correlation function is then
decomposed into its temporal frequencies and phase angles
using Fourier analysis techniques, from which the 24-hour
frequency component is selected as a measure of circadian
rhythmicity. The 24-hour phasor magnitude is used as the
metric for behavioral circadian entrainment/disruption; the
greater the magnitude, the greater the level of behavioral
circadian entrainment of activity to light. The phasor angle
reflects the phase relationship between the periodic lightdark exposure pattern and the periodic activity-rest pattern
in the correlations (Figure 2). Figure 2 shows examples of
Figure 2: Circular correlation functions and associated phasors for two nurses, one day-shift and one
rotating-shift, and for two nocturnal rodents, a rat exposed to a regular pattern of 12L:12D and a rat
exposed to a “jet lag” pattern of light and dark (the 12L:12D pattern was reversed every 48 hours).
55
the circular correlation functions and associated phasors for
two nurses, a day-shift nurse and a rotating-shift nurse, as
well as for two nocturnal rodents, one exposed to a regular
pattern of 12 hours light and 12 hours dark (12L:12D) and
one exposed to a simulated “jet lag” pattern (discussed later
in the text).
Two techniques of parsing the Daysimeter data can be used
for phasor analysis. The “all-at-once” technique computes
the circular correlation function of the entire data set (e.g.,
seven days) in one operation. The resulting circular
correlation function provides the correlation coefficients
between light and activity for time shifts in activity ranging
from zero to the total length of time over which data were
collected (usually seven days for the data collected here).
Because the circular correlation extends over multiple days,
this technique is sensitive to day-to-day variations as well
as variations within each day. Fourier analysis applied to
the all-at-once circular correlation extracts the magnitude
and the angle of the 24-hour component, separating it from
the longer, infradian, and shorter, ultradian, variations. If
the power in these longer or shorter periods in the circular
correlation is large, the magnitude of the 24-hour phasor is
reduced. To facilitate Fourier harmonic analysis the time
duration, or total length of the data set, is truncated to an
integer number of days.
The “sliding-window” technique parses the data into a
progression of overlapping 24-hour segments and
calculates a circular correlation for every segment. Fourier
analysis is applied to each of these circular correlation
functions to extract the 24-hour, fundamental component.
The resulting 24-hour phasors for each data segment are
then averaged (vectorally) to arrive at a phasor representing
the entire data collection period. The advantage of the
sliding-window technique is that it utilizes all the available
data without truncation; therefore, it is better than the all-atonce technique for short collection protocols because it
utilizes all of the limited data available. A limitation of this
technique, however, is that there is no ability with the
sliding-window technique to identify rhythms slightly
longer or shorter than 24 hours. Analyzing seven days of
data using the all-at-once technique has a precision of
approximately 3.4 hours for resolving the expressed
circadian periods in the circular correlations, whereas the
sliding-window technique cannot reliably resolve periods
up to approximately 12 hours. This higher degree of period
resolution with the all-at-once technique will necessarily
result in shorter 24-hour phasor magnitudes than those
obtained for the same data using the sliding-window
phasors unless there is no significant power in the infradian
or ultradian rhythms. This difference is most evident with
disrupted subjects who do not necessarily show a strong
24-hour rhythm but still exhibit relatively large slidingwindow phasor magnitudes. Because the period resolution
is worse with the sliding-window technique than with the
all-at-once technique, the power of the other non-24-hour
periods also contribute to the phasor magnitude.
56
The activity index values can be used to compute two
statistics developed by Van Someren et al. [23] to estimate
the day-to-day consistency of activity over the recording
session (interdaily stability, IS) and the hour to hour
consistency of activity over the recording session
(intradaily variability, IV). High values of IS indicate that
the subject’s 24-hour activity and rest pattern was
consistent over the entire recording session; high values of
IV indicate that the subject’s pattern of activity and rest
was highly fragmented with intermittent and inconsistent
intervals of movement and no-movement. Neither statistic
accounts for levels of light exposure or changes in light
exposure levels.
DATA COLLECTION
The Nurses’ Health Study II is a prospective cohort study
that began in 1989, when 116,671 registered female nurses
in the United States between the ages 25 to 42 were
enrolled. The study was designed to prospectively examine
the effects of oral contraceptive use and other life style
factors on chronic diseases, particularly cancers and cardiovascular diseases. Surviving nurses within this cohort
were contacted to serve as potential subjects between November 2006 and April 2008; 138 nurses volunteered to
participate. The study was approved by the Committees on
the Use of Human Subjects in Research at the Brigham and
Women’s Hospital and the Harvard School of Public
Health, as well as Rensselaer’s Institute Review Board;
written informed consent was obtained from each
participant.
Reported here are results from Daysimeter measurements
for 38 day-shift and 61 rotating-shift nurses; the rotatingshift nurses worked from one to five nights over the recording period. For 16 other nurses in the study who were
categorized as rotating-shift, the Daysimeter data indicated
that they did not work any nights during the study period,
so their data are not included. The larger number of rotating-shift nurses reflects the initial assumption that day-shift
nurses would be more homogenous with respect to circadian entrainment than the rotating-shift nurses. Data from
23 of the 138 nurses were unusable due to recording irregularities or non-compliance with the seven-day protocol. It
was possible to identify data from those nurses who did not
comply with the protocol by using the data from the Daysimeter’s on-board temperature sensor; room-temperature
readings, as opposed to elevated temperatures when the
Daysimeter is in close contact with the body, and extended
periods of inactivity were certain signs that the nurse did
not wear the device when required. Visual inspection of the
temperature and activity data plotted against time was used
to identify noncompliance. Only those protocol departures
lasting about a day or more could be unambiguously identified and were removed. The data from 13 of the remaining
99 nurses showed non-compliance during only one to three
days of the seven-day recording period. For these nurses,
only the data for the noncompliant days were removed; 10
nurses provided six days of useful data and three provided
five.
RESULTS
activity-rest behavior was much greater for the day-shift
nurses than for the rotating-shift nurses based upon the
highly significant difference between the two groups in
terms of their respective phasor magnitudes determined
using either the sliding-window or the all-at-once techniques. There was no difference between the two groups in
terms of their phasor angles using the sliding-window technique, but there was a significant difference between the
phasor angles of the two groups using the all-at-once technique. Following the earlier discussion, the all-at-once
technique is much more sensitive to consistency in the light
and activity patterns over the entire recording period than
the sliding-window technique. Thus, there is a much larger
difference between the two groups in terms of their phasor
angles using the all-at-once technique.
Table 1 summarizes the results of the measurements. Twotailed student t-tests revealed that, on average, day-shift
nurses were exposed to statistically higher levels of light
(both photopic and CLA) than rotating-shift nurses during
the recording sessions. It is interesting to note, however,
that the ratio of the average photopic light level to the average circadian level is similar for both day-shift and rotating-shift nurses. This suggests that the types of light
sources, natural and electric, seen by both groups were not
remarkably different for both groups. Interesting too, the
average activity levels of the two groups were nearly
identical. The phasor analysis of the continuous light and
activity data provided by the Daysimeter indicated that the
synchronization between the light-dark exposures and the
Table 1.
Photopic Circadian
Activity
light
light
Index
(lux)
(CLA)
(!g-forcerms)
Interdaily
Stability
(IS)
Phasors
Intradaily
Sliding-window
All-at-once
Variability
Angle
Angle
(IV)
Magnitude
Magnitude
(hours)
(hours)
Day-shift
(N=38)*
302
[188]
369
[227]
0.0094
[0.0014]
0.692
[0.144]
0.447
[0.170]
0.50
[0.11]
0.65
[0.74]
0.46
[0.12]
0.68
[0.71]
Rotating-shift
(N=61)*
188
[152]
209
[166]
0.0097
[0.0018]
0.252
[0.129]
0.458
[0.135]
0.33
[0.10]
0.64
[0.73]
0.12
[0.10]
2.3
[3.2]
t-test p-value
(two-tail)
0.0028
0.94
2.7 x10
3.9 x 10
-4
0.26
2.1 x 10
-24
0.72
6.4 x10
-11
-24
2.4x10
-4
* – Mean values shown; standard deviation values in [brackets].
Photopic light exposures: The Daysimeter system utilizes a fully calibrated (spectral, spatial, intensity) photopic light
sensor measuring in lux (lm/m!). Group means [standard deviation] are computed from individual subject means of each
subject"s entire recording session.
Circadian light exposures: Using the model from Rea et al. [12], circadian light exposure values are determined from the
photopic and blue light sensor data measured in CLA. Group means [standard deviation] are computed from individual
subject means of each subject"s entire recording session.
Activity index: Values from two orthogonally oriented accelerometers (measuring up/down and forward/back head motions)
are used to compute an activity index that is logged at regular time intervals along with the light readings. Each logged
measure of activity is the rms combination of the standard deviation of acceleration taken once per second over a 30-second
interval of both accelerometers. Group means [standard deviation] are computed from individual subject means of each
subject"s entire recording session.
Interdaily stability: The IS statistic developed by Van Someren et al. [23] measures the consistency of activity among days
and ranges from 0 to 1. A value of 1 results when every day’s activity is identically to the other days, while conversely a
value of 0 results from no similarity among days.
Intradaily variability: The IV statistic developed by Van Someren et al. [23] measures the fragmentation of rhythm between
rest and activity based on a scale from 0 (no variability from hour to hour) to upwards of 2. A larger value indicates more
fragmentation of rest and activity, or conversely, less consolidation of rest/activity patterns.
Phasor magnitude: A correlation between circadian light exposure and activity (calculated using either the sliding-window
or the all-at-once technique) for the observation period, in this case 5-7 days. A higher magnitude indicates the subject has a
consistent, 24-hour schedule with respect to activity and light. Lower magnitudes indicate low correlation between daily
cycles of light and activity irrespective of phase differences.
Phasor angle: A phase relationship between circadian light exposure and activity (calculated using either the sliding-window
or the all-at-once technique) for the observation period, in this case 5-7 days. A positive angle (first quadrant) indicates a
delay in activity with respect to light and a negative angle (fourth quadrant) indicates an advance in activity with respect to
57
intermittently, but, of course, there are too few samples to
reach any reliable conclusions concerning the health of
these nurses.
Figure 3. Average phasors for day-shift and for
rotating-shift nurses using the sliding-window (left)
and all-at-once (right) techniques. See text for
discussion of the two methods.
Figure 3 illustrates on polar coordinates the average
phasors for both groups using both techniques, slidingwindow and all-at-once. For diurnal species the phasors
typically plot in the first and fourth quadrants, indicating
activity and light exposure are positively correlated. Phasor
angles for humans are almost always in the first quadrant,
indicating that activity is delayed with respect to light
exposure, typically because people become active with the
bright, morning light but continue to be active after sunset
under dim, electric light; see later discussion on this point.
Figure 4 shows the distribution of the phasor magnitudes
for the two groups of nurses on a linear scale, but broken
down in terms of the number of night-shifts they worked
during the recording period. Obviously the day-shift nurses
did not work any nights; two rotating-shift nurses worked
five nights. Figure 4 shows that, in general, as the number
of working nights increase, phasor magnitudes decrease.
This suggests that behavioral circadian disruption increases
with the number of nights worked. It should be noted,
however, that one of the two nurses who worked five nights
had a higher phasor magnitude than all those working three
or four nights and most working two nights, and that the
highest average circadian disruption was for nurses
working three night shifts during the week. This suggests
that disruption might be less for people who continually
work the night shift compared to those who do so
Figure 4: All-at-once phasor magnitudes for day-shift
and rotating-shift nurses plotted as a function of the
number of nights worked
58
Figure 5 compares the sliding-window and the all-at-once
techniques for computing phasor magnitudes for the dayshift and rotating-shift nurses. For the day-shift nurses,
either method returns similar results (r!=0.91). For the
rotating-shift nurses, the similarity is reduced (r!=0.18).
The most obvious explanation being that the day-shift
nurses are on a regular schedule with every day similar to
every other day. Since their behavior conforms to a stable
24-hour pattern both techniques give the similar phasor
magnitudes. Rotating-shift nurses, on the other hand, have
variable schedules so they do not have stable 24-hour
patterns of light and/or of activity. They exhibit a complex
variety of rhythm periods, some slightly shorter and some
longer than 24 hours. As previously discussed, the slidingwindow technique lumps together a broader range of
periods in the 24-hour phasor resulting in relatively larger
phasor magnitudes than those obtained by the all-at-once
technique.
Figure 5: Comparison of phasor magnitudes for the
sliding-window and the all-at-once techniques
Figure 6: Comparison between the IS statistic based
upon activity and the two techniques for determining
phasor magnitude based on activity and light exposures
Figure 6 is a comparison of the IS statistic and the phasor
magnitudes obtained using the sliding-window and the allat-once techniques. The all-at-once technique yields results
that correlate highly with the IS statistic (r!=0.94). The IS
statistic is inherently based upon an assumed 24-hour
activity rhythm, but since it does not depend on light
exposures, the high correlation indicates that light exposure
maintained a similar pattern with respect to activity
throughout the week. The relatively weaker correlation
between the IS statistic and the sliding-window phasor
magnitude implies that, although light and activity might be
highly correlated, the period of oscillation did not conform
precisely to a 24-hour period. That is, the 24-hour based IS
statistic is low but the sliding-window phasor magnitude is
high, relative to the all-at-once phasor magnitude. It is not
known at this time which phasor technique is more
predictive of health and well-being; must people maintain
high regularity across all days (measured with the all-at-
once phasor) to be healthy or can people vary behavior
(activity) across days (measured with the sliding-window
phasor) without consequences to health and well-being?
Figure 7 illustrates sliding-window phasors for four different day-shift nurses paired with their average daily circadian stimulus (CSA) and activity profiles. Because average
activity and average light are plotted independent of one
another, they are not strictly comparable to the corresponding phasors. Nevertheless, the paired diagrams do help
illustrate how activity and light tend to affect phasor magnitudes and angles. Based upon examinations of a wide
range of activity and light exposure profiles from day-shift
nurses, the data from nurse A can be considered as “typical.” Generally speaking, a “typical” day-shift person exhibits a delayed phasor angle of between one and two hours
with a phasor magnitude between 0.5 and 0.6. “Typical”
individuals tend to be consistently active throughout their
waking period, but are exposed to relatively higher light
Figure 7: Average activity and average circadian stimulus (CSA) values across a day with corresponding slidingwindow phasors for four different day-shift nurses
59
levels in the early part of their activity period than toward
the end. This asymmetry in relative light exposure from the
early to the late part of their activity period, in fact, produces delayed phasor angles in Quadrant I. The phasor
angle for the “typical” nurse is nearly the same as that for
day-shift nurse B, but the phasor magnitude for nurse B is
lower. Comparing the average activity and light exposure
data for the two nurses, it is clear that for both nurses light
exposures are lower in the late part of the activity period
than in the early part of the activity period, thus leading to
their nearly identical phasor angles. It is even clearer that
activity is more variable and less well correlated with light
exposure for nurse B than for nurse A. This yields a shorter
phasor magnitude for nurse B than for nurse A. It is important to point out that the absolute time of activity and
light do not affect the phasor angle or the phasor magnitude, at least not directly. Again, the phasor is based upon
the synchronization between activity and actual light exposure, not between activity and clock time. The phasor magnitudes for nurses A and C are nearly identical, the major
difference is in their phasor angles. It can be readily appreciated from this figure that nurse C does not exhibit the
relative asymmetry in light exposure from the early to the
late activity period as is seen with nurse A. Consequently
the phasor angle for nurse C is advanced with respect to
nurse A to near 0 hours. The average activity and light exposure data for nurse D are very much like those for nurse
C except they are less well correlated, yielding the same
phasor angle, but a lower phasor magnitude. Compared to
the “typical” nurse A then, nurse D is more variable in her
activity and light exposure patterns and does not exhibit the
asymmetry of activity from the early to the late parts of her
activity period. Of potential interest, nurse D appears to be
exposed to bright light toward the end of her activity period
relative to any of the other nurses, including the “typical”
nurse. Also it should be noted that the period of time with
no light and activity (presumably sleep) is less for nurses B
and D than for nurses A and C who both have relatively
large phasor magnitudes. Again, whether these differences
are predictive of health and well-being remains to be studied systematically.
DISCUSSION
It is becoming clearer that human health and well-being are
dependent upon the synchronization of biological systems
[13]. The master clock in the SCN sets the pace for
peripheral systems, but if these complex systems become
asynchronous with one another (e.g., following jet lag) we
experience poor sleep, indigestion, and performance errors.
These, in turn, may lead to chronic problems such as
insomnia, diabetes and obesity, and injury.
Since the pattern of light and dark drives the timing of the
master clock, it is obviously important for our evolving
understanding of human health that we begin to measure
the light-dark cycles actually experienced by individuals in
their own living environments as affected by electric
lighting and modern day work schedules. The Daysimeter
was developed for this purpose. Measuring the light-dark
60
patterns experienced by people in their living environments
is important, but those data are not particularly valuable
unless it is possible to interpret them in the context of
human well-being and health. Phasor analysis was
developed to quantify the synchronization of oscillating
light-dark patterns and activity-rest patterns, thereby giving
an individual-specific measure of behavioral circadian
entrainment or disruption. In addition to activity, but not
examined here, other measures of circadian regulated
processes, such as core body temperature, heart rate, and
hormone levels, can be made together with light exposure
measurements and then analyzed using the phasor analysis
technique described here.
It is also readily possible to relate individual-specific
measures of circadian disruption to individual-specific
measures of performance, affect and fatigue. It is also
theoretically possible to develop individual-specific
treatments to correct circadian disruption and measure their
efficacy through changes in phasor magnitude and angle as
they relate to changes in medically meaningful outcomes.
Obviously genetic differences as well as differences in
environmental stressors must also be considered. In the
near future, however, it should be possible, using tools like
the Daysimeter and phasor analysis, to begin to bridge
ecological measurements of circadian disruption to
controlled studies of circadian disruption using animal
models for such diseases as cancer, cardiovascular disease,
and diabetes.
Recently we were able to show the feasibility of this
bridge. Figure 2 shows the circular correlations and
associated phasors for a day-shift nurse and a rotating-shift
nurse together with the circular correlations and associated
phasors for a rat placed on a regular 12L:12D schedule and
a rat placed on a regularly reversing (i.e., continuous jet
lag) pattern of light and dark. As can readily be appreciated
from Figure 2, and despite the differences in two species
with regard to their photic niche (i.e., diurnal nurses versus
nocturnal rats), it should now be possible to parametrically
study the impact of circadian disruption actually
experienced by individuals in different living environments
with any one of several animal models for human diseases
and disorders. These envisioned studies can then serve as
the next logical step in understanding the impact of
circadian disruption on human health, complementing the
pioneering epidemiological studies that raised our
collective concern for how circadian disruption might
impact human health [17, 19].
ACKNOWLEDGEMENTS
This work was supported in part by CDC Grant 1R01
OH008171 to Dr. Eva Schernhammer at Harvard Public
Health and by the Trans-NIH Genes, Environment and
Health Initiative Grant 1U01 DA023822-01 to Dr. Mark
Rea at the Lighting Research Center. The authors would
like to thank Terry Klein and Dennis Guyon of the Lighting
Research Center for their assistance with the study and
manuscript.
REFERENCES
1.
Bechtold, D.A. Energy-responsive timekeeping. J
Genet. 2008 Dec;87(5):447-58.
2.
Berson, D.M., Dunn, F,A,, Takao, M. Phototransduction by retinal ganglion cells that set the circadian
clock. Science 2002, 295:1070-1073.
3.
Bierman, A., Klein, T.R., Rea, M.S. The Daysimeter:
A device for measuring optical radiation as a stimulus
for the human circadian system. Measurement Science
and Technology 2005, 16:2292-2299.
4.
Brainard, G.C., Hanifin, J.P., Greeson, J.M., Byrne, B.,
Glickman, G., Gerner, E., Rollag, M.D. Action
spectrum for melatonin regulation in humans:
Evidence for a novel circadian photoreceptor. J.
Neurosci. 2001, 21(16):6405-6412.
5.
Commission International de l’Eclairage (CIE). Light
as a True Visual Quantity: Principles of Measurement.
Commission Internationale de l’Eclairage 1978, Paris.
6. Erren, T. C. and Reiter, R.J. Defining chronodisruption. J Pineal Res. 2009 Apr: 46(3):245-7. Epub
2009 Feb 9.
7.
Ingling, C.R., Martinez-Uriegas, E. The relationship
between spectral sensitivity and spatial sensitivity for
the primate r-g X-channel. Vision Research 1983,
23:1495-1500.
8.
Kohsaka, A., Bass, J. A sense of time: How molecular
clocks organize metabolism. Trends Endocrinol Metab
2007, 18(1):4-11.
9.
Mallis, M.M., DeRoshia, C.W. Circadian rhythms,
sleep, and performance in space. Aviat Space Environ
Med 2005, 76(6 Suppl):B94-107
10. McIntyre, I., Norman, T.R., Burrows, G.D.,
Armstrong, S.M. Human melatonin suppression by
light is intensity dependent. J Pineal Research 1989.
6(2):149-156.
11. Rea, M., Bierman, A., Figueiro, M., Bullough, J.D. A
new approach to understanding the impact of circadian
disruption on human health. J Circ Rhy 2008. 6:7.
12. Rea, M.S., Figueiro, M.G., Bullough, J.D., Bierman,
A. A model of phototransduction by the human
circadian system. Brain Res Rev 2005, 50(2):213-228.
13. Reddy, A.B., Wong, G.K., O'Neill, J., Maywood, E.S.,
Hastings, M.H. Circadian clocks: neural and peripheral
pacemakers that impact upon the cell division cycle.
Mutat Res 2005 Jul 1; 574(1-2):76-91. Epub 2005 Apr
15.
14. Reppert, S. M., and Weaver, D.R. Coordination of
circadian timing in mammals. Nature 418, 935-941 (29
August 2002).
15. Sack, R.L., Auckley, D., Auger, R.R., Carskadon,
M.A., Wright, K.P., Jr., Vitiello, M.V., Zhdanova, I.V.
Circadian rhythm sleep disorders: part II, advanced
sleep phase disorder, delayed sleep phase disorder,
free-running disorder, and irregular sleep-wake
rhythm. (An American Academy of Sleep Medicine
review.) Sleep 2007, 30(11):1484-1501.
16. Schanda, J., Morren, L., Rea, M., Rositani-Ronchi, L.,
Walraven, P. Does lighting need more photopic
luminous efficiency functions? Lighting Res. Technol.
2002. 34,1: 69-78.
17. Schernhammer, E.S., Laden, F., Speizer, F.E., Willett,
W.C., Hunter, D.J., Kawachi, I. Rotating night shifts
and risk of breast cancer in women participating in the
Nurses' Health Study. J Natl Cancer Inst 2001,
93:1563-1568.
18. Smith, V.C., Pokorny, J. Spectral sensitivity of the
foveal cone photopigments between 400 and 500 nm.
Vision Research 1975, 15:161-171.
19. Stevens, R.G., Blask, D.E., Brainard, G.C., Hansen, J.,
Lockley, S.W., Provencio, I., Rea, M.S., Reinlib, L.
Meeting report: the role of environmental lighting and
circadian disruption in cancer and other diseases.
Environ Health Perspect 2007, 115(9):1357-1362.
20. Straif, K., Baan, R., Grosse, Y., Secretan, B.,
Ghissassi, F.E., Bouvard, V., Altieri, A., BenbrahimTallaa, L., Cogliano, V. Carcinogenicity of shift-work,
painting, and fire-fighting. The Lancet Oncology 2007,
8(12):1065.
21. Thapan, K., Arendt, J., Skene, D.J. An action spectrum
for melatonin suppression: Evidence for a novel nonrod, non-cone photoreceptor system in humans. J
Physiol 2001, 525(Pt 1):261-267.
22. Van Derlofske, J., Bierman, A., Rea, M.S., Ramanath,
J., Bullough, J.D. Design and optimization of retinal
flux density meter. Meas. Sci. Technol. 2002, 13:821.
23. Van Someren, E.J.W., Hagebeuk, E.E., Lijzenga, C.,
Scheltens, P., de Rooij, S.E., Jonker, C., Pot, A.M.,
Mirmiran, M., Swaab, D.F. Circadian rest-activity
rhythm disturbances in Alzheimer's disease. Biol
Psychiatry 1996, 40:259-270.
24. Young, M.E., and Bray, M.S. Potential role for
peripheral circadian clock dyssynchrony in the
pathogenesis of cardiovascular dysfunction. Sleep Med
2007. 8(6):656-67.
25. Zeitzer, J.M., Dijk, D.J., Kronauer, R., Brown, E.,
Czeisler, C. Sensitivity of the human circadian
pacemaker to nocturnal light: melatonin phase
resetting and suppression. J Physiol 2000, 526(Pt
3):695-702.
61
Content-based Adaptation of the Dynamics of Estimated
Light Sources
Marc Peters*, Pedro Fonseca*, Lu Wang*, Bas Zoetekouw*, Perry Mevissen#
*Philips Research
#Philips Consumer Lifestyle
Experience Processing Group
Advanced Technology
High Tech Campus 34, 5656AE Eindhoven
High Tech Campus 37, 5656AE Eindhoven
{marc.a.peters, pedro.fonseca, l.wang, bas.zoetekouw, perry.mevissen}@philips.com
ABSTRACT
Lighting is a very important aspect in film-making. Using a
technique known as light source estimation, it is possible to
estimate the color properties of the light sources used while
filming scenes of films or television series. One very
important—but often unaddressed—aspect of light source
estimation is related to temporal control. In this paper, we
propose a novel method for temporal control of the
estimated light source of a video scene. After describing
the method, we will explain the results of a user study
which shows that it is superior when compared with
traditional temporal control techniques.
Keywords
Movie lighting, content analysis, light source estimation,
temporal control.
INTRODUCTION
Lighting is a very important aspect in film-making. In
modern movies and television series, film makers and
cinematographers carefully use light to accentuate certain
aspects of the story, to change the atmosphere that is
conveyed, or to establish a certain mood. For example,
candlelight suggests romance and harmony, high contrast
lighting achieves accentuated dramatization, and moving
light can invoke fear, chaos and madness [17]. Colored
light is also used to accentuate certain aspects of the story
or to help convey certain emotions. Although there are no
specific rules on how to associate colors with emotions, red
is often found associated with love or hatred, yellow with
happiness and joy, and blue with peace and tranquility [18].
For regular film watchers, these lighting aspects usually
have an implicit influence: although they are of crucial
importance to help convey the story, most people don’t
actually realize that they are “manipulated” by lighting
changes during a movie. On the other hand, the presence
and characteristics of these elements have very often been
used in the area of video content analysis; interpretation of
cinematographic rules — such as information about the
lighting of a scene and the color of the light source that
illuminates it — can give important semantic information
about that scene or even about the entire movie. For
example, Rasheed et al. use scene lighting characteristics,
along with other visual features, to automatically classify
62
the genre of a movie [11]. Light source estimation can thus
be a very important technique to extract high-level,
semantic information about a scene.
One particular application of light source estimation is the
creation of a lighting atmosphere which is rendered while
users watch a movie on a television screen. If the light
source is estimated correctly, the rendered atmosphere will
resemble the light settings of the scene in the movie and
increase the user’s immersion in the content.
The topic of light source estimation has been well studied
in the past and several well described techniques are in
common use. The techniques range from simple MPEG-7style dominant color extraction [15] to more advanced
systems based on white point extraction. In [1,2], some of
these methods are reviewed. As we are mostly interested in
the dynamic control of the estimated light source rather
than the actual light source estimation itself, a full
comparison between the different known methods is
beyond the scope of this paper. For simplicity reasons, we
will only use a single light source estimation algorithm in
this paper to test our dynamic filtering technique. For each
content frame, we construct an RGB space in which each
the color of each pixel is represented by a point. We then
use least squares estimation to determine the best fit of the
data from the linear RGB color space into a vector in that
space. Because the light source is reflected on objects on
the screen, this vector has the property that it reflects
precisely the color that is reflected on the different surfaces.
For the application mentioned above, the temporal
dynamics of the estimated light source are very important.
The actual lighting set-up of a mise-en-scène — i.e. the
arrangement of actors and scenery on a setting filmed for a
motion picture — filmed for a calm movie scene is usually
constant for an entire scene. However, light source
estimation techniques are not perfect, and camera action
and object movement often cause the estimated light source
to vary throughout such a calm scene, even though the
actual lighting during recording of the scene didn’t change
at al. If the raw results of light source estimation are used to
render a lighting atmosphere, the atmosphere will be much
more dynamic than the actual content as the user sees it on
the screen. This effect not only reduces immersion, but it
might even distract and annoy the viewer. On the other
hand, a very dynamic lighting atmosphere might be
desirable if scene under consideration is equally dynamic,
for example when many special visual effects are used. In
that case, a dynamic atmosphere will match the content on
the screen and will therefore contribute to the viewer’s
immersion.
This temporal aspect of lighting has not been so much
addressed. This is surprising, as dynamics of lighting and
light effects are a fundamental part of the cinematographic
experience Most existing work on temporal control of
estimated lighting focuses on the use of (advanced) lowpass digital filters which “smoothen” the estimated signal
over time. In [16] an advanced implementation of such a
system is described, which uses substantial time
subsampling and employs the spatial detection of special
effects on small regions to reset a low-pass filter. Whereas
this system is able to react to very localized visual effects, a
drawback is that it will often fail to reflect the global
dynamic properties of an entire scene. This drawback
actually applies to most filters described in literature:
however advanced these filters are, unless they take into
account the fact that global and local dynamic properties of
the content, on a frame by frame basis, are inherently a
very important part of the content itself, they will
unavoidably decrease the intensity of these elements for the
viewer.
Let’s look at two extreme cases as an example. During
reasonably static, dialog-based scenes — such as those
which occur quite often in soap operas and popular comedy
television series like “Friends” and “Will and Grace” —
the resulting estimated light source should be rather static
throughout the entire scene. However, at the exact frame
this scene ends and a completely different scene starts, the
light source should reflect the properties of the new scene
immediately and should not be slowly smoothed
throughout. At the other extreme, consider a scene in a war
movie where the battle takes place at night. Due to the dark
lighting conditions, any special effect (small such as a
gunshot or intense such as an explosion) should be
reflected in the estimated light source. In this case,
smoothing with a low-pass filter is simply not acceptable.
In this paper, we propose a novel method for temporal
control of the estimated light source of a video scene. The
goal is to smooth the estimated light source color when the
content is static but to allow special, abrupt effects to be
instantly reflected without latency in the resulting
estimation. This will allow an atmosphere to be rendered
while viewers watch a movie on a television. This
atmosphere not only reflects the color of the lighting of a
scene but also its temporal dynamics. All of this will help
increase the immersive experience.
In the next section we will introduce the algorithm.
Afterwards we will describe the test content and objective
criteria for characterizing the dynamics of that content. We
will then describe the user test that we have carried out to
evaluate the proposed algorithm. Finally, we will discuss
the results and conclude the paper.
TEMPORAL CONTROL ALGORITHM
In the context of this paper, by “temporal control”, we
mean the process through which the results of light source
estimation are modified or filtered to change its temporal
characteristics. An example of a very simple temporal
control algorithm is a low-pass filter. When applied to the
output of a light source estimator, it simply eliminates (or
“smoothes”) all abrupt color variations given by that
estimator. As can be easily imagined, the resulting colors
vary slowly in time, regardless of how dynamic the content
might be.
For the reasons explained in the introductory section, it is
not always desired that the estimated light source follows
the changes in the content on a frame-by-frame basis.
Particularly for the application described earlier where an
atmosphere is rendered in real-time along with a movie
being watched by the viewer, the rendered light source
should only change significantly when the same happens
with the content. With the approach described in this paper
we attempt to let the dynamics of each scene dictate how
“smooth” the variations of the resulting light source color
should be. Very dynamic scenes will lead to fast variations
in the light, whereas static, slowly varying scenes will lead
to results that are calm and smooth in time.
It should be noted that the temporal control algorithm
proposed in this paper does not depend on the light source
estimation algorithm used. In fact, it is suitable for any kind
of raw color signal input, for example expressing an
approximation of the color properties of the dominant light
source in a scene. The method described in this section can
be used without loss of generality for any such input.
To achieve automatic smoothing as described above, we
need to characterize the dynamics of the content in a
feature which is simple to use. As physical light sources
that illuminate the scene during recording naturally
influence both the colors and the luminance of that portion
of the motion picture, we will estimate the dynamics based
on color- and illumination-based features. From these
features, we then compute the dynamics for each frame of
the content.
To calculate these features, we make use of a so-called
“combined HSV histogram”. HSV is a relatively simple
three-component color space, characterized by the hue (H),
saturation (S) and brightness (value, V) [6]. We use an
HSV color space here, rather than the simpler RGB color
space, because HSV more accurately describes perceptual
color relationships than RGB. On the other hand, it is still
computationally simpler to implement than real
perceptually uniform color spaces like CIE 1976 La*b*.
However, because the HSV color space is not completely
perceptually uniform, in particular in the low brightness
colors, we need to use a non-trivial distance measure as we
will define below.
We define a 256-bin HSV histogram as follows:
• The 256 bins are ordered in a cube of dimensions
16!4!4. The first dimension corresponds to the hue
63
values (discretized in 16 bins), the second dimension
corresponds to the saturation (discretized in 4 bins), and
the third dimension corresponds to the value
(discretized in 4 bins). We use more bins for the hue
component than for the saturation and value
components because we want to give more importance
to the differences in hue than to differences in saturation
or brightness.
• For each video frame, we calculate the HSV values for
each of the pixels and fill each bin of the histogram with
the number of pixels that have an HSV value in the
corresponding range.
• If the hue value is expressed as a number in [0,360) and
the saturation and value are numbers in [0,1], the first
bin will thus contain the number of pixels that have an
hue between 0 and 22.5 (=360/16), a saturation between
0 and 0.25, and a value between 0 and 0.25.
Mathematically, we can express this as follows:
1.
Dark regions of an image do not convey much
information about the light settings of a scene, other
than the fact that the light source did not strongly
illuminate that area;
2. A small difference between the histograms of two dark
video frames might hide the fact that the light settings
captured on those two frames are completely different.
For example, consider the situation in which a very
dark scene is illuminated by a small green light in the
first frame, and that that light suddenly changes to red
in the next frame. In that case the light condition as
determined by the histograms should clearly be very
different, even though the majority of the pixels are
black or very dark in both frames.
In order to make the distance measure of Equation (2) more
robust to such conditions we introduce an alternative
distance measure , that takes into account this problem
by not counting dark pixels in the frames. We define this
alternative distance measure as follows:
(3)
(1)
where i and j correspond to the rows and columns of pixels
in each video frame, and Hij, Sij, and Vij are the hue,
saturation and value components, respectively, of the pixel
at position
.
The histogram is then normalized by dividing each value
by the total number of pixels in the video frame, in order to
make the feature independent of the dimensions of the
video frame.
The HSV histogram is created for each frame in a video
between
sequence. Next we define a distance measure
two of such histograms
The numerator counts the bin-to-bin difference between the
histograms, but leaves out bins for which the brightness
component is low (i.e., only pixels with brightness
components in the three highest bins are taken into
account). The minimum in the denominator should be taken
between the sums of the bins with the lowest brightness of
the two consecutive frames. The denominator leaves out
the dark parts that are common to both frames. It is bound
to a minimum value of 1/4, to avoid extremely large
distance values between the histograms of two video
frames when both have very large dark areas. The distance
measure
of Equation (3) thus emphasizes the difference
between the non-dark parts of the images.
Next, for each frame !we compute two values:
1.
(2)
as the sum of the absolute differences between
corresponding bins in two consecutive frames. As the
histograms are normalized, this definition of the distance
measure
will always yield a value between 0 and 2.
However, large parts of video frame often have very low
brightness. Whenever such dark regions are present in two
consecutive frames, the distance between the corresponding
HSV histograms will also be very small. This is caused by
the fact that HSV is not perceptually uniform as explained
before. For our application, this is undesirable for two
reasons:
64
The first is the estimated light source
(expressed as RGB values) for that frame. The exact
nature of the algorithm used to find the light source of
the current frame does not matter as long as for each
frame we find a color vector (in the linear RGB color
space) that reflects some properties of the dominant
light source illuminating the scene:
(4)
2.
The second value we calculate is the resulting light
source,
after temporal filtering:
(5)
!
It is computed as a linear combination of the estimated light
source for the current frame and the color that was
calculated from the previous frame. The smoothing factor
is defined as
!
(6)
For dynamic scenes the smoothing factor is small, giving
a high weight to the light source estimated for the current
frame. For calm, static scenes the smoothing factor is large,
leading to a calm and gradual transition between colors
because the previous value has a large weight. The
minimum smoothing factor is larger than 0 to make sure
that the filtered light color will always converge to the
if the
estimated light source of the current frame
video freezes or becomes completely static (i.e., if the
distance computed between several consecutive frames
is 0).
VALIDATION FRAMEWORK
In this section, we present a framework to quantize the
dynamic properties of video sequences in order to
characterize the properties and the behavior of the proposed
temporal control algorithm. We start by describing our test
set; this test set is used both for the quantization in this
section as well as for the user study of the temporal
filtering algorithms that is described in the next section.
We then describe which features we extract from the video
sequences to characterize their dynamic properties. We
continue by analyzing the effects of temporal filtering on
the dynamics of estimated light source.
Test content
The test set consists of six video clips of 30 seconds each.
They were selected based on the presence of very different
lighting conditions which characterize different genres in
film and TV series. The clips from the test set can be
described as follows:
• Walk the Line — the first sequence is a clip from the
movie “Walk the Line”; this particular scene depicts a
concert with different illumination sources: the
backstage lighting, with a relatively dark and saturated
color and the non-saturated, bright illumination of the
singer which dominates the scene. The scene has a very
high contrast, is very calm and shots are relatively long.
• Hellboy — the second sequence is a clip from the
movie “Hellboy II: The Golden Army”; this particular
sequence has a filmed part and a computer generated
part. Both are highly saturated and the filmed part has
the additional particularity of having a very distinctly
colored light source in the left and the right side of the
screen.
• Friends — the third sequence is a clip from the episode
“The One with the Kips” (season 5, episode 5) of the
popular soap opera “Friends”; like most soap operas,
Friends has a flat appearance: there is little contrast, and
it is filmed in high-key, i.e., it has an abundance of
unsaturated light, and the scenes are free from shadows.
The colors are mostly pastel and although there is not a
lot of movement on the scene, the shots are relatively
short as it mainly consists of dialogues.
• Hulk — the fourth sequence is a clip from the movie
“The Incredible Hulk”; this particular clip takes place in
a cave, at night, during a thunderstorm. Although the
shots are long and dark and the contrast is high, the
lightning strikes and the rain add a very dynamic
element to the scene lighting.
• Wall-E — the fifth sequence is a clip from the
animation movie “Wall-E”; this particular computer
generated clip depicts an indoors scene, illuminated by
one of the characters (a robot). During the scene the
illumination varies drastically, from very well lit, highkey, to very dark and saturated.
• Platoon — the sixth and last sequence is a clip from the
movie “Platoon”; this particular scene depicts a combat
situation at night; apart from being a very dark scene,
contrast is relatively low and gunshots and explosions
dominate the scene, making it very dynamic and
intense.
Visual feature extraction
The variety of clips and genres will help us explore the
behavior of the temporal control algorithm proposed in this
paper. As the clips we use are not part of any public
domain test set and therefore are not available for free, it is
important to characterize them as well as possible. Not only
will this offer the reader a better description of the test set
used in our study, but it will also help us explain the results
of that study later on in this paper.
In this section, we describe this characterization of the
dynamics of the video sequences in terms of temporal
changes of their visual properties. To quantify these visual
properties, we extract a number of video descriptors from
the content. These descriptors offer a numerical
representation of visual properties which can be analyzed in
terms of their dynamic behavior in time.
Video descriptors
In order to characterize the dynamics of each sequence in
our test set, we will use three different descriptors: the shot
duration, the HSV histogram and the light source color.
The first descriptor we will use is the shot duration. A shot
is defined as a sequence of video frames, captured
uninterruptedly by a movie camera. It is delimited by a shot
transition (commonly called a “shot cut”) at its beginning
and at its end. The transition between shots can be abrupt,
in which case the new shot will start on a frame
immediately after the last frame of the previous shot, or
gradual, in which case the actual transition lasts for a
number of frames (e.g. cross-fade, fade in, fade out).
65
Figure 1 – (left) Number of shots in each of the test sequences and (right) average intra-shot feature differences.
Note that the HSV histogram difference values were scaled up so both features can be visually compared.
Different shots often correspond to different points of view
within the same scenario; other times, they belong to
distinct scenes and thus have completely different visual
properties. These shot transitions represent visual
discontinuities and should be taken into account when
characterizing the video sequences. An important way to
characterize the dynamics of film content is by measuring
the average shot length. Since most of the shots end in
abrupt transitions, shot boundaries are moments when the
visual properties (including the lighting of a scene) change
drastically from one moment to another. This does not
necessarily mean, however, that the sequence is very
dynamic. In a soap opera, mainly comprised of dialogues,
shots are typically very short and alternate between two or
more view points within the same scene. However, the
properties of each shot are very similar because the
dialogues take place in the same physical location.
Furthermore, within each shot, there is little or no change
since usually, during dialogues, only the face of the actors
is shown in the image. In an action movie, on the other
hand, directors usually keep the shots short to indicate
action and induce a high tempo. In this case, the visual
properties of each shot are very different from each other,
offering the viewer very much different visual information
within a short period of time.
The second video descriptor we will use is the HSV
histogram. This descriptor was introduced in the previous
section. It describes the color properties of each frame of
the video in terms of the hue, saturation and brightness
(value) of all its pixels.
The third and final descriptor we will use is the light source
color. It gives an approximation of the color of the light
entering a scene. As remarked before, the temporal control
algorithm that is the main focus of this paper is independent
of the algorithm used to estimate the light source, and a full
description of the different methods available is out of
scope here. We use a method based on a least-squares fit in
the RGB-space representing all pixels in a single video
frame.
66
Characterization of visual properties
The number of shots in each sequence of our test set is
illustrated in Figure 1 (left).
To express the temporal behavior of the two remaining
visual descriptors, we compute the corresponding average
feature differences. The average HSV histogram difference
is given by:
(7)
where the histogram HSVt[h,s,v] was described in the
previous section. The outer summation is over all N
, and the inner
frames, the first frame being
summation is over all the 16!4!4 bins of the histogram.
The average light source difference is given by:
(8)
with
(9)
i.e., the Euclidian distance in RGB space between the light
source colors in two consecutive frames, and where ,
are the RGB values of the light source computed for
and
frame .
As was mentioned above, shot cuts represent visual
discontinuities. To better assess the dynamic properties of
the sequences, while at the same time excluding the
influence of the shots boundaries from this process, the
computation of the average feature difference will be
restricted to all frames that are neither the starting frame of
a shot nor the frames that make up a gradual transition.
Figure 1 (right) illustrates the intra-shot average feature
differences for the sequences in the test set, ordered by
increasing average light source differences.
As can be clearly seen, the sequence Friends has one of the
lowest variations in terms of both visual descriptors, even
though the clip contains the highest number of shots in the
Figure 2 – Cumulative histogram of light source differences, filtered with the three different settings for Hellboy
(left) and Platoon (right); note that the horizontal scales are different for the two graphs. The fact that the avg
setting is in both cases above the unfiltered result, shows that simple averaging always decreases the variations in
the light source. The auto setting, on the other hand, is first above and later below the avg setting: this indicates
that small changes are smoothed even more than in the avg case, while large changes are kept (see text). Note that
the maximum possible light source difference is 442; however, all of the graphs saturate much earlier.
clips in our test set. This is typical of soap operas: short
shots are necessary for dialogues but the scene is kept
rather static to avoid inducing a notion of action and
activity to the viewer. In contrast, Platoon is the sequence
with the most relative activity, as expressed by the high
variation of visual features. This is also expected from an
action scene in a war movie: a combination of short shots
and high intra-shot visual variation help convey a notion of
high activity and intense action.
Characterization of temporal control techniques
In this section we will characterize the properties of three
different temporal filtering strategies for the estimated light
source estimation. In the next section, these are compared
in a user study.
The three strategies (or “settings”) that we examine here
are the following:
• No smoothing, or none: no filtering is done on the
results of light source estimation. Light source is
estimated frame by frame without any sort of temporal
filtering;
• Low-pass filtering with a windowed average, or avg: a
low-pass filter smoothes out abrupt transitions and
sudden light source changes, as well as small variations
that can occur on certain frames. In particular, the
estimated light source of the past 20 frames (i.e., 0.8
seconds in a video with a frame rate of 25 frames per
second) is averaged.
• Content-based temporal filtering, or auto: as described
in the temporal control algorithm section: small
variations that might occur from frame to frame are
smoothed out, but abrupt transitions with high
amplitude are retained.
In order to visualize the different behavior of each of these
temporal filtering mechanisms, we first apply each method
to the light source computed for each frame of a sequence.
Then, we compute a histogram of light source differences
for each setting.
This histogram gives an indication of the type of variations
that occur for the filtered light source for each video and
helps us compare the different temporal filtering strategies.
The light source difference is computed as defined by
Equation (8). If ,
and
can have any value in the
then the maximum light source difference
range
will be
. The light source difference
will therefore be a value in the range
.
in an entire
The distribution of light source differences
movie sequence gives important insight into the dynamics
of the light source: if all light source differences are small,
this means that the light source changes very gradually,
whereas a more homogeneous distribution would point
towards a case where both small and larger changes are
present. To characterize this distribution, we look at the
histogram of light source differences, which is computed as
follows:
(10)
for each frame in the sequence except for
where
the first,
and with
. In order to make it
easier to visualize, we compute a cumulative histogram
as:
based on
(11)
is normalized, i.e., the sum of
Note that the histogram
all bins in the histogram will add up to 1. Conversely, the
last bin on the cumulative histogram will also be 1, i.e.,
.
67
Sequence
Walk the line
Hellboy
Friends
Hulk
Wall-E
Platoon
Setting
75%
90%
99%
none
6
9
54
avg
2
5
9
auto
2
3
6
none
5
9
52
8
avg
3
6
auto
2
3
7
none
5
9
69
avg
3
5
8
auto
2
3
62
none
5
10
264
avg
2
3
11
auto
2
3
272
none
12
25
102
avg
5
8
13
auto
4
7
44
none
15
35
284
avg
5
8
11
auto
5
13
270
Table 1 – Light source difference below which
75%, 90%, or 99% of all light source difference are
accounted for, i.e., when compared to the graphs in
Figure 2, we look for the light source difference
values for which the graphs cross the 75%, 90%, or
99% point, respectively. Note that the maximum
possible light source difference is 442, and that
almost all of the sequences and settings saturate
much earlier than that.
Figure 2 illustrates the cumulative histograms computed for
each temporal control setting for two very distinct video
clips: Hellboy and Platoon.
As can be clearly seen for Hellboy, – Figure 2 (left) –
without any type of filtering (setting none), the cumulative
histogram saturates much later than for the other two
settings. This means that with avg and auto, most light
source variations are simply filtered out.
When comparing this with Figure 2 (right), notice that the
horizontal scale is different – light source differences in
Hellboy are much smaller (for all settings) than for Platoon.
It can be easily seen that the cumulative histogram for
setting avg saturates very quickly; this is expected, as this
algorithm is nothing more than a low-pass filter which cuts
out any large sudden variation.
More interesting is the difference between the settings none
and auto. The latter has a steeper curve for low difference
values – this means that small consecutive differences in
light source are simply smoothed out. However, after this
initial point, the curves of auto and none are similar for
higher difference values. This means that large light source
differences are kept. This behavior is characteristic for the
the temporal filtering technique proposed in this paper. In
68
scenes with rather static content most variations of
estimated light source are smoothed out. In scenes with
very dynamic content, on the other hand, variations are
sharp and pronounced, reflecting the amount of dynamics
of the content on the screen.
In order to characterize the behavior of the temporal
filtering techniques for the remaining sequences, we
compute the lowest bin in the histograms for which a
certain percentage of light source differences are found:
(12)
This measure is computed for the percentages of 75%, 90%
and 99%, i.e.,
. Table 1 lists these values
for all sequences, for the three different temporal filtering
settings. Compare this to cumulative histograms like those
of Figure 2: we look for the light source differences
(horizontal scale) for which the curves cross a horizontal
line at p=75%, p=90% and p=100%, respectively.
As can be easily seen in Table 1, for sequences with little
visual variation (e.g, Walk the Line, Hellboy), the saturation
point for both avg and auto settings occurs quite early, with
99% of the light source differences occurring already below
bin 9 for both settings. This means that 99% of the
differences as defined in Equation (8), after each of these
temporal control settings were applied, are smaller than a
value of 9. On the other hand, for sequences with high
visual variation (in particular Hulk and Platoon), this early
saturation stays low for the avg setting but is much higher
for setting auto. This again reflects the characteristic of the
setting auto, which smoothes out small variations but not
large variations in light source.
Based on the analysis done in this section, we expect that
the setting none will be appropriate for sequences which
are very dynamic, because all the variations in detected
light source are maintained, but won’t be very useful for
less dynamics scenes, for which small but potentially
disturbing changes in the estimated light source will also be
present in the filtered result.
We expect the setting avg to be appropriate for sequences
which are calm, since most small and large variations are
smoothed out. It will probably not work very well for
dynamic scenes, particularly those with special effects such
as lightning and explosions, as these will be averaged out
of the filtered result.
Finally, we expect the setting auto to be appropriate for
most sequences, both calm and dynamic, because it is able
to match the dynamics of the filtered light source to the
actual dynamics of the content.
In the next section we will present the results of our user
study, performed to test these hypotheses.
USER STUDY
In this section we describe the user study that we have
performed in order to evaluate the perceived quality of the
temporal filtering method described earlier.
To test the settings, we created a system that plays a movie
clip and analyses the light source of the content in realtime. The estimated light source is then filtered in one of
the three ways (“settings”) described in the previous
section. The resulting, filtered light source color is
projected into the user’s living room using four Philips
LivingColors lamps, creating an effect from here on
designated as “surround light”. In this way, the atmosphere
of the movie clip is brought into the user’s living room,
potentially increasing the user’s immersion in the content.
Two basic questions arise:
1. Does this new atmospheric context improve the users’
viewing experience?
2. How do the three different settings influence the
viewing experience?
Answers to these questions will help us explore the
temporal filtering algorithms to further tune and develop
them. For this purpose, we need to answer the following
research questions:
!
Do the surround light settings match the video
content?
!
Do the surround light settings help increase the level
of immersion?
!
Do the three different temporal filtering settings help
increase the feeling of presence and engagement in
different ways?
!
Which of the three temporal filtering setting for the
surround light system do users prefer?
We expect that the presence of surround light settings will
improve the level of presence and engagement for the
users, and as explained in the previous section, we expect
the auto temporal filtering setting to best reflect the
dynamics of the video content. Hypotheses are therefore:
1 The level of immersion while watching video with the
surround light turned on is higher than without
surround light.
2 The level of immersion while watching video using the
auto temporal filtering setting for the surround light
system is higher than when the other two settings
(none and avg) are used.
Based on the research questions and the hypotheses, the
surround light settings were evaluated with 25 participants,
using a within-subjects design in which participants
watched six video clips in four variations: without surround
light and with the three different light settings.
Additionally, before the six regular clips, an additional
movie sequence from the movie “Shrek” was used as a
training for the participant. A Presence and Engagement
questionnaire [10] was used to measure the level of
presence and engagement. Additionally, the participants
were asked to rank their preference for the three settings.
Participants
For the user test, 25 voluntary participants from Philips
Research with ages ranging from 22 to 33 (mean=26.4,
sd=3.3), were recruited (11 males and 14 females). The
participants were selected not to suffer from color
deficiency in red and green hue. Each participant received a
5-euro voucher as a surprise at the end of the experiment.
Material
We developed a surround light system as described before,
which can operate using the three different temporal filter
settings described in the previous section. The three light
settings constitute experimental conditions. In the control
condition, no surround light is used.
The six test video clips used were described in the previous
section. A seventh 30-second video clip, extracted from the
movie “Shrek”, was used to explain the procedure and to let
the participants become familiar with the questionnaires.
Each video clip is presented to each participant four times.
The first viewing uses the control condition (i.e., without
surround light), and is followed by three experimental
conditions (i.e., the three light settings) in a randomized
order. The first of the seven video clips shown to the
participants (the clip from ‘Shrek’) is used as a test video.
The test video provides a training opportunity of the
experimental setting, and allows the participants to
calibrate their rating scales for the following test video
clips. The results from the test clip are not used in further
analysis of the measurements. The order of remaining six
video clips is randomized. The order in which the three
light settings and the six video clips are shown to the
participants was pre-edited in order to ensure a balanced
distribution over the 25 participants. A script program was
used to play the video clips and light settings according to
the pre-edited order with a single key press by the
experiment leader.
In human computer interaction (HCI), immersion [4,13],
presence [8,9,13,14,19], engagement [3,9], and flow [5] are
often related to the experience of interacting with virtual
environments. Various questionnaires [9,10,12,19] were
developed to measure immersion or engagement. Most of
these questionnaires were developed for interactive virtual
environments, whereas our experimental context is passive
television watching. This makes some common factors in
the aforementioned questionnaires not applicable for our
experiment, for example, the ‘Control’ factor.
Although there is an ongoing scientific debate on the notion
of immersion and presence (technology space or subjective
experience) [12,19], our goal is to test the participants’
subjective experience. The difference between the two
notions lies in whether it is measured by objective
parameters (e.g., the amount of the virtual space the user
can interact with) or by subjective ratings of his/her own
experience.
69
Figure 3 – (left) Map of the lab room used, (right) photo of the setup from angle (top) 2 and (bottom) 1.
For the test, we used a 13-item Presence and Engagement
questionnaire1 [10] which was originally developed for the
context of 3D-TV watching, and which measures the
subjective experience of the participant. The questionnaire
consists of a number of questions which should be
answered on a five-point Likert scale, ranging from
‘strongly disagree’ to ‘strongly agree’. The question items
contribute to two factors: Feeling of Presence (5 items) and
Engagement (8 items). However, since the questionnaire
was originally developed for 3D-TV, two items
(contributing to Feeling of Presence) from the
questionnaire are not applicable for our experiment. These
two items are ‘I had a strong sense that the characters and
objects were solid’, and ‘I felt I could have reached out and
touch things’. We excluded these from our questionnaires.
Based on the research questions as stated above, we would
like to test whether the dynamics of the lights in the living
room match the video content. We developed one question
item to measure this aspect: ‘The surrounding setting
matches the video clip’. This question item was not asked
after the control condition, i.e., when no surround light is
used.
Procedure
The test started by handing the questionnaire booklet to the
participant. The participant was asked to fill in his/her
personal information and TV watching experiences. The
purpose of this experiment was not explained beforehand in
order to avoid biasing participants, thus preventing that
they would pay too much attention to the surround light.
The experiment leader explained the procedure of the
following steps.
The following steps encompassed seven sessions, in which
the first session was a training session. Each session
consisted of four sub-sessions, where one video was played
with one light setting from the four variations. The starting
time of each sub-session was manually controlled by the
1
This questionnaire has not yet been validated.
70
experiment leader with a remote keypad. After each subsession, the participant had to fill in a questionnaire about
the movie–lighting setting combination that he/she had just
watched. In addition, he/she was encouraged to write down
his/her comments on the same page of the questionnaire.
When the participant finished filling in the questionnaire,
the experiment leader pressed the keypad to proceed to the
next sub-session. At the end of each session, the participant
had to rank the light settings based on his/her preference.
To make sure the participant could follow the experiment,
the experiment leader (only during the first session) asked
whether the questions and the procedure were clear. All
participants in the experiment understood the questions and
the procedure after the first session. The same process was
repeated for the remaining sub-sessions.
After the seven sessions, the experiment leader had a short
interview with the participant. The conversation was noted
down by the experiment leader.
Data analysis
In order to avoid inconsistencies across different raters [7],
a within-subject design with two factors was chosen. The
two factors are: type of video and type of setting (6!4),
where Walk the Line is used as the baseline video (as it
constitutes the calmest sequence) and the auto setting is
used as the baseline setting. The questionnaire data on the
test video was not used.
On each questionnaire item, we collected 600 (6 video clips
! 4 settings) data points from 25 participants. A two-way
repeated measures ANOVA was used to analyze data. Main
effects were analyzed by multivariate tests.
The interview was conducted in a semi-structured way.
Participants were given printed posters of the movies and
TV series they had watched on an A4 paper. This helped
them remember the video clips they had watched and easily
start the conversation. The experiment leader started the
conversation by talking about the movies. This helped
observe their emotional reaction to the movies and later on
the light settings which were not captured in the previous
sessions. During the conversation, the experiment leader
asked their general impression about the settings, whether
any settings made them uncomfortable or distracting, and if
they had any wishes to improve the settings. The
experiment leader tried to ask these questions in a
spontaneous way, thus not following a particular order. For
example, if a participant started talking about annoying
settings, the experiment leader followed by asking ‘so did
any other settings made you annoyed or made you feel
uncomfortable?’
Matching question data
An extra question ‘The surrounding setting matches the
video clip.’ was used with the none, avg and auto setting.
Adjustment for multiple comparisons: Bonferroni was
used.
Sig
type of video
F(6,19) = 6.35
0.001
type of setting
F(3,22) = 78.636
<0.001
type of video !
type of setting
F(18,7) = 5.174
0.006
Factor
F
Sig
type of video
F(6,19)= 7.627
<0.001
type of setting
F(3,22)= 16.045
<0.001
type of video !
type of setting
F(18,7)=1.934
0.147
Table 3 - Result of main effect analyses on
Engagement score
Factor
F
Sig
type of video
F(6,19)= 8.475
<0.001
type of setting
F(3,22)= 23.873
<0.001
type of video !
type of setting
F(18,7)=5.027
0.003
Table 4 - Result of main effect analyses on the
matching score.
On the matching score (range from 1 to 5), all main effects:
type of video (F(6,19)=8.475, p<0.001), type of setting
(F(3,22)=23.873, p<0.001) and interaction (F(18,7)=5.027,
p=0.003), were significant at p<0.05.
Paired comparisons revealed that comparing to the none
setting and the avg setting, the auto setting had
significantly (p<0.001) higher score. Figure 6 shows the
mean scores of matching, from which one can observe that
the avg setting had very similar matching score comparing
to the auto setting on the first three videos with lower
amount of dynamics (i.e., Walk the Line, Hellboy and
Friends), whilst the effect of the avg settings became much
less comparing to the auto setting on the next three videos
with higher amount of dynamics (i.e., Hulk, Wall-E and
Platoon).
Preference data
As introduced in the previous section, if difference among
the experimental light settings was found, participants were
asked to rank the three settings3 none setting, avg setting
and auto setting, to their preference on a ranked order scale
where 1 is most preferred and 3 is least preferred. After all
light settings corresponding to the video were presented,
150 groups (6 video clips ! 25 participants) of orders were
collected from the 25 participants, of which 141 groups
3
2
F
Table 2 - Result of main effect analyses on Feeling of
Presence score.
Presence and Engagement questionnaire data
The Presence and Engagement questionnaire we used in
this experiment was comprised of eleven items,
contributing to two factors: Feeling of Presence and
Engagement.
On the Feeling of Presence score (range from 3 to 15),
which was aggregated from three question items, all main
effects: type of video (F(6,19)=6.35 p=0.001), type of
setting (F(3,22)=78.636 p<0.001), and interaction type of
video ! type of setting (F(18,7)=5.174, p<0.01) were
significant at p<0.05. The significant interaction effect
indicates that type of setting had different effects on the
Feeling of Presence score depending on which type of
video was used. Figure 4 shows the mean scores for all the
light settings on each video.
Paired comparisons2 were performed comparing the three
settings to their baseline setting (i.e., no light setting). It
revealed that the three settings: none setting, avg setting
and auto setting resulted significantly (p<0.001) higher
ratings on the Feeling of Presence score than the baseline
setting. Further, paired comparisons comparing the none
and avg setting to the auto setting revealed that the auto
setting resulted significantly (p<0.005) higher rating than
the none and avg setting.
On the Engagement score (range from 8 to 40), which was
aggregated from eight question items, significant main
effects were found on: type of video (F(6,19)=7.627,
p<0.001) and type of setting (F(3,22)=16.045, p<0.001).
The interaction main effect (F(18,7)=1.934, p=0.147) was
found not significant.
Paired comparisons comparing the three settings to their
baseline settings showed that the avg setting and the auto
setting resulted significantly (p<0.01) higher rating than the
baseline setting, whilst the none setting did not result
significantly higher rating (p=0.601). This can be observed
from Figure 5 as well, where the mean score of none setting
on two video clips (i.e., Walk the Line and Friends) were
lower than the baseline setting.
Factor
The actual setting labels were replaced with the labels
‘setting 1’, ‘setting 2’ and ‘setting 3’, which were
randomized over the video clips.
71
Figure 4 - Mean scores of Feeling of Presence for all the
settings on each video clip.
Figure 6 - Mean scores of matching for the three
experimental settings on each video clip.
Figure 5 - Mean scores of Engagement for all the
settings on each video clip
To break down this aggregated result, we categorized it for
each video (Figure 8). Figure 8 shows a trend where the
auto setting is more preferred (when compared with the
other two settings) when higher dynamics are present in
video clips, whilst preference on the avg setting is close to
the auto setting when lower dynamics are present. This
trend is consistent with the results from the matching
question item (see Figure 6).
Interview
Figure 8 - Number of occurrence in the highest
preference for the three experimental settings on each
video.
were found different and were indicated with preference
order. In general, the auto setting got the highest number of
highest preference votes from the participants (see
Figure 7).
72
Most participants liked the idea of surround light settings.
They mentioned particularly positive impressions about the
settings with cartoon movies, such as Wall-E. One
participant expressed his feeling as:
Wall-E impressed me the best, because the settings were
bright and gave me the feeling of presence in space.
Participants found that the none setting is distracting or
annoying in most cases, but they accepted it more when it
is used with a fighting or action scene, such as Platoon.
This is confirmed with the ordering of preference shown in
Figure 8. A representative comment made by one
participant:
Whenever the lights flicker too much, it's very distracting,
especially for static moments! However, on Platoon, the
flicker at the same time as the bullet was a nice touch!
On calm scenes such as Walk the Line, the participants did
not express much difference in preference between the avg
and the auto settings. Recalling Table 1 from an earlier
section, this is not surprising since the behavior of these
two settings is very similar for this particular video clip.
Discussion
The results of the test suggest that the surround light
settings helped increasing the feeling of presence. The avg
and auto settings increased the level of engagement, but the
effect from the none setting was not significant. This may
be explained by the comments made by the participants in
the final interview, that is, the none setting is distracting in
most cases. Moreover, the auto setting resulted in higher
feeling of presence and engagement comparing to the none
and the avg setting. Similarly, tests showed that the auto
setting also resulted in a better matching effect than the
none and the avg settings.
From qualitative analysis on the preference ordering, the
auto setting is in general the most preferred one comparing
to the none and the avg setting. However, the avg setting
seemed to be an equally preferred setting when the
dynamics present in the video clips are low.
CONCLUSIONS
We have described a novel method for temporal filtering of
light source colors that are extracted from video content.
The resulting dynamics of the detected light source fits
much better with the content on the screen than previous
methods: dynamic, action scenes or scenes with special
effects result in dynamic lighting, whereas slow and static
scenes result in calm lighting.
We have tested the new algorithm in a user test in which
the lighting conditions from video content were projected
into the living room. The results of the test show that the
users liked the effect of the proposed new algorithm better
than the control condition (no lights) and also better than
two other tested algorithms for temporal dynamics. In
addition, the user test suggests that the novel method
provides increased immersion in the video content when
compared to the two other algorithms or to the situation
without surround light.
References
1. Barnard, K. et al., A Comparison of Computational
Color Constancy Algorithms, Part One; Theory and
Experiments with Synthetic Data, IEEE Transactions on
Image Processing, Vol. 11, No. 9, pp. 972-984, 2002
2. Barnard, K. et al., A Comparison of Computational
Color Constancy Algorithms; Part Two: Experiments
with Image Data, IEEE Transactions in Image
Processing, Vol. 11. No. 9. pp. 985-996, 2002
3. Bentham, J. The Theory of Legislation. Oceana
Publications, 1975.
4. Brown, E. and Cairns, P. A grounded investigation of
game immersion. CHI 2004, ACM Press, 1279-1300.
5. Czikszentmihalyi, M. Flow: The Psychology of Optimal
Experience, Harper Perennial, 1991.
6. Foley J., van Dam A., Feiner S., Hughes J., Computer
Graphics, Principles and Practice, second edition,
Addison-Wesley, Reading, MA, 1990.
7. Freeman, J., Avons, S.E., Pearson, D.E. and IJsselsteijn,
W.A. Effects of sensory information and prior
experience on direct subjective ratings of presence,
Presence: Teleoperators and Virtual Environments, 8,
1, (1999), 1-13.
8. Heeter, C. Being There: The Subjective Experience of
Presence. Teleoperators and Virtual Environments, 1, 2,
(1992), MIT Press, 262-271.
9. Lessiter, J., et al. A Cross-media Presence
Questionnaire: The ITC-Sense of Presence Inventory.
Presence, Teleoperators and Virtual Environments, 10,
3 (2001), 282-297.
10. Rajae-Joordens, R. J. E., Measuring Experiences in
Gaming and TV Applications – Investigating the Added
Value of a Multi-View Autostereoscopic 3D Display.
Probing Experience – From assessment of User
Emotions and Behaviour to Development of Products.
Springer Netherlands, 2008, 77-90.
11. Rasheed, Z. et al., On the Use of Computable Features
for Film Classification, IEEE Trans. On Circuits and
Systems for Video Technology, vol. 15, no. 1, Jan 2005.
12. Slater, M. Measuring Presence: A Response to the
Witmer and Singer Questionnaire. Presence, 8, 5
(1999), 560-566.
13. Slater, M., Usoh, M. and Steed, A. Depth of Presence in
Virtual Environments. Presence, Teleoperators and
Virtual Environments, 3, 2 (1994), MIT Press, 130-144.
14. Slater, M., et al. Taking Steps: The Influence of a
Walking Metaphor on Presence in Virtual Reality. ACM
Transactions on Computer Human Interaction (TOCHI)
Special Issue on Virtual Reality, September 1995.
15. Text of ISO/IEC 15938-3/FDIS Information technology
– Multimedia content description interface – Part 3
Visual
16. Van den Broek, H.A. Automatic living light effect
generation, Master’s thesis, Technische Universiteit
Eindhoven, Dept. of Mathematics and Computer
Science, 2004.
17. Van Sijll, J, Cinematic storytelling, Michael Wiese
Productions, 2005.
18. Wei, C. Dimitrova, N., and Change S., Color-mood
analysis of films based on syntactic and psychological
models, IEEE International Conference on Multimedia
and Expo (ICME), Taipei, Taiwan, June 2004.
19. Witmer, B. G., and Singer, M. J. Measuring presence in
virtual environments: A presence questionnaire.
Presence, Teleoperators and Virtual Environments, 7,3
(1998), 225-240.
73
Descriptions, Measurements and Visualizations of Light
Distributions in 3D Spaces
Sylvia C. Pont, Alex Mury, Huib de Ridder
Delft University of Technology
Industrial Design
Landbergstraat 15, 2628 CE Delft
The Netherlands
s.c.pont@tudelft.nl
ABSTRACT
The aim of our studies is to understand the physical
structure and human perception of natural light fields. The
light field depends on the primary illumination, the
scattering properties of the environment and the scene
geometry. We present our newly developed methods to
describe, measure and visualize visually complete
descriptions of the light field, the 5-dimensional “plenoptic
function”. The structures of natural light fields were found
to be rather smooth and built up of just a few possible
topologies. We show that our visualizations by means of
light tubes represent the well-known “flow of light”
lighting design concept in a surprisingly intuitive way.
Keywords
Light field, appearance, plenoptic function, visualization,
flow of light, scale of light, light-zones.
INTRODUCTION
Lighting influences our perceptions of our surroundings,
including the “visual light field” [6]. We investigate which
aspects of the appearance of scenes underlie these
perceptions and how the appearance changes with lighting
variations. In this paper we focus on the physical
description, measurement and visualization of the light
field in three-dimensional (3D) spaces in order to
scientifically assess the spatial and form-giving
characteristics of light, and we analyze the structure of
natural light fields.
On a most basic level one might draw relations between
lighting and how well people can see details around them.
Such maximization of luminance contrast of visual detail
forms the basis of many lighting recommendations and
standards. However, lighting may vary in many more ways
and influences the appearance of scenes and our
perceptions in a very complicated manner [2]. The
directional properties of the illumination strongly affect the
appearance of an object. For instance, in fully diffuse
illumination even a specular object looks rather matte.
Diffuse illumination can have directional properties –
74
Jan J. Koenderink
Delft University of Technology
Electrical Engineering, Mathematics and
Computer Science
Mekelweg 4, 2628 CD Delft
The Netherlands
illumination from an overcast sky is directed vertically
downwards. However, the properties of diffuse and highly
directional (collimated) illumination are very different. In
collimated illumination the shading is dominated by cast
and body shadows, whereas in diffuse illumination the
shading gradients are much more gradual.
Examples of light properties which artists (light designers,
architects, photographers, painters, etcetera) know to be
important aspects of scene appearance are the diffuseness,
density variations, flow of light, and 3D modeling
properties of the light. Such light properties depend in a
complicated way on the primary illumination, the scattering
properties of the environment and the scene geometry and
typically cannot be represented by just a simple number.
Moreover, they vary from point to point in 3D scenes.
Therefore we first need a visually complete description of
the luminous environment, or “the plenoptic function” [1].
If we constrain ourselves (for simplicity) to static scenes
and if we ignore color, we can describe the plenoptic
function, or the (white) light distribution in 3D space, by
the 5D “light field” [4] (the radiance as a function of the 3D
position and 2D direction). The light fields of natural
scenes are often highly complicated functions. The angular
variations can be almost arbitrary, ranging from smooth
(such as under an overcast sky) to very spiky (such as on a
sunny day on the beach or under forestry). In
contradistinction, the surface irradiance is typically fairly
smooth, because surface elements of a convex object are
illuminated from half spaces.
In this paper we will address our newly developed methods
to describe, measure and visualize light fields in 3D spaces.
In the discussion we will draw a relation between our
visualizations and well-known concepts from lighting
design and architecture, namely “the flow of light” [2], “the
scale of light”[3], and “light-zones” [8].
Figure 1 A panoramic image (a local light field measurement)
and its first three spherical harmonic components (see text for
explanation).
FORMAL DESCRIPTIONS OF THE LIGHT FIELD
The light field is a complicated 5D function of position and
direction. At any point in space the light field is a function
of direction, a spherical function (a 360° ! 180° panoramic
view), which may well be described by a superposition of
components of different angular frequencies. At first sight
it might seem important to include all those frequencies in
light field studies. However, most objects around us are
quite matte and low-pass or “diffuse” the illumination.
Thus, only the low frequencies of the light field influence
their appearance. This suggests that a decomposition of the
light field in components of different frequencies might be
useful. For a spherical function such as the light field this
comes down to spherical harmonics, usually known as a
multipole development in physical context.
(mathematically: the second order spherical harmonic
approximation), and of matte spheres rendered in those
high-resolution and approximated low-pass local light
fields. The spheres indeed look the same. Thus, only 9
numbers are sufficient to describe the appearance of a
matte convex object.
Mathematically, it was shown that a second order spherical
harmonics description, which is a superposition of just
three qualitatively different, low angular frequency
components, is sufficient to describe the appearance of
perfectly matte convex objects [15]. These three
components are physically equivalent to:
•
the flux density (a number representing the
monopole contribution or average radiance from
all directions in the point under consideration),
•
the light vector (a number and direction
representing the dipole contribution or magnitude
and direction of the net maximum transport of
light in the point), and
•
the squash tensor [12] (two numbers and a
direction representing the quadrupole contribution
or magnitudes and orientation of a set of
orthogonal light and dark two-fold lobes).
The flux density describes a constant illumination from all
directions, which is usually known as “ambient
illumination” in computer graphics, or Ganzfeld
illumination in psychology. Light fields in which this
monopole component dominates are rare in nature. An
example is an overcast sky over a snow cover (“polar
white-out”). The combination of a monopole and a dipole
term yields what is known as the “point source at infinity
with ambient term” in computer graphics. A natural light
field that approximates a dipole dominated field is the
overcast sky or hemipsherically diffuse source. Quadrupole
dominated light fields occur in the case of ring sources or
two point sources at opposite sides of the region of interest
– therefore we called it the “squash tensor”.
In figure 1 we show a local light field measurement (a
panoramic image) and its first three spherical harmonic
components; the first component represents the flux
density, (it is clearly a constant); the second component
represents the light vector (here it is clearly oriented
vertically); the third component represents the squash
tensor (the two light and two dark lobes are clearly visible).
In figure 2 we show spherical maps of the panoramic
image, of the sum of the first three components
Figure 2 A panoramic image (a local light field measurement;
upper left image) mapped on a sphere, which may be thought of
as photographed on a specular sphere. The upper right image
shows the superposition of the first three spherical harmonic
components, which were shown separately in figure 1. The second
row shows renderings of matte spheres in the high-resolution local
light field and its low-pass approximation. It is clear that the
superposition of flux density, light vector and squash tensor is
sufficient to describe the appearance of such a matte convex
object.
MEASUREMENTS OF THE LIGHT FIELD
The enormous luminance range that is common in natural
scenes forms a second challenge in light field studies. This
so-called very high dynamic range (HDR) cannot be
covered by photographical methods, even if the dynamic
range is extended by techniques such as photographical
composition from multiple exposures. This range can
however be covered by HDR sensors consisting of a photodiode and a logarithmic amplifier. The combination of the
low-pass approach and HDR sensor finally resulted in the
design for our light field measurement system, the
“Plenopter”, see figure 3. This custom-made apparatus
allows for local light field measurements in the order of a
second. The Plenopter contains 12 sensors in a regular
configuration. A single, local measurement results in 12
numbers, which allow for the estimation of the local low
order approximation (9 numbers) which was described in
the previous section.
Then, from a set of such local measurements on a suitable
matrix of points in a 3D space we can reconstruct the
global structure of the light field in that 3D space [10] (by
75
interpolation). This reconstruction thus gives the flux
density, light vector and squash tensor at each point within
the 3D space. These data can be used to make computer
graphics renderings of matte convex objects, e.g. spheres,
at arbitrary points in this space (see figure 4 bottom row,
for a maybe somewhat more interesting shape). It is known
from artistic practice, e.g. in lighting design and
architecture [2, 9] that such renderings give a good
impression of the visual quality of light in a scene and
therefore this method may be very useful in applications.
Figure 3 The plenopter: 12 high dynamic range sensors in a
regular dodecahedron configuration. Local measurements result in
12 numbers from which we can estimate the second order
spherical harmonic approximation. Sets of local measurements
over an array of positions in 3D space allow for reconstructions of
the light field in that space by interpolation.
We used our methods to measure 24 different light fields in
an empty office room (daylight was screened off) in the
Light Lab at Philips Research. We made reconstructions of
each of these light fields, which consist of 9 numbers, or
the three components depicted in figure 1, at each point of
the finite 3D space that we covered with our measurements.
Since these basic data of which each of the 24
reconstructed light fields exist are rather abstract and bulky
we need intuitive visualizations in order to get some insight
into the global structure of the light fields.
VISUALIZATIONS OF LIGHT FIELDS
We visualized the light fields through it's “light tubes” [4],
see figure 4, which represent the net flow of light (not the
rays of light - light tubes can be curved and light rays
76
cannot). The tubes directions are locally tangential to the
light vector (the direction of maximum net energy transfer)
and their widths are locally inversely related to the
magnitude of those vectors (the larger the light transport,
the smaller the tube). The tubes usually start at light
sources, where they are quite narrow, and end on light
absorbing surfaces, where they tend to be quite wide.
In figure 4 we show three light fields with quite typical
structures. The upper image shows a case for primary
illumination existing of three quite diffuse lamps on the
ceiling in a row close to one of the long walls of the empty
office room. The tubes diverge out from the sources
towards the floor and opposite wall. In the second case the
tubes diverge from a diffuse lamp in the middle of the
ceiling towards the walls and floor. In the third case the
tubes diverge from four quite narrow beams towards the
walls and the floor, where they curve upwards due to
interreflections from the floor. We rendered white, matte
bunnies at three points along one of these curved tubes. The
right bunny was rendered closest to the primary
illumination in the right front corner of the room. It is
clearly visible from its appearance that the light comes
from above and slightly towards the right. The shading and
shadowing contrasts over the bunny are quite strong,
though a small effect of secondary illumination is visible at
the bunnies’ breast. In the middle case the bunny is clearly
illuminated from both primary and secondary illumination.
The shading and shadowing contrasts are quite weak. The
left bunny is primarily illuminated from below due to
interreflections.
In natural scenes the light fields are due to complicated
combined effects of primary illumination, scattering and
screening by the objects in the scene, and scene geometry.
However, the global structures of natural light fields show
perhaps surprisingly smooth behavior and can be modeled
in a very simple way [10]. Moreover, for 2D light field
descriptions we showed which generic topological
configurations are possible. These configurations could be
described by a small range of singular points [12].
CONCLUSION AND DISCUSSION
Our methods allow low order, HDR measurements of a
light field in a finite 3D space. The low order
representations exist of the flux density, light vector and
squash tensor. We visualized our measurements by means
of reconstructions of light tubes, which represent the net
transport of flux in the space. These visualizations give
intuitive pictures of the light fields, allowing insightful
inferences about the light quality in that space. For
instance, it could be a scientific tool in the quantitative
assessment of “light-zones” [8], or the areas where light
from roughly different directions “meet” [9].
In lighting design the concept of the “flow of light” [2]
describes the potential of lighting to produce distinct
shading patterns. The associated metrics of the flow of light
are the vector/scalar ratio of the illuminance, the
illuminance vector direction and the flow of light ratio.
Note that our methods deal with the radiance, not the
illuminance. Nonetheless, the illuminance vector/scalar
ratio and illuminance vector direction correspond to the
light vector magnitude and direction up to some
normalization factor. Thus, our tubes visualizations directly
represent the flow of light.
Figure 4 Light fields visualizations by means of light tubes. The
local tubes’ directions are tangential to the light vectors and the
local widths are inversely related to the vectors’ magnitudes. The
box represents the room and the squares and circles on the ceiling
represent the primary light sources, which were quite diffuse in
the upper two cases and quite narrow beams in the bottom case.
Cuttle [2] proposed to examine the flow of light using a
small matte white sphere to reveal the shading pattern.
Madsen and Donn [9] did experiments with a “light-flowmeter” consisting of a grid of matte white spheres that was
placed vertically in scenes. They used this method for the
visual assessment of the spatial and form-giving character
of (day-)light, that is, simultaneous judgments of the flow
of light and Frandsen’s “scale of light” [3]. The “scale of
light” is a measure of the diffuseness of the illumination,
ranging from fully collimated to hemispherically diffuse. In
psychophysical studies on light field and material
perception we found that for smooth matte spheres human
observers
confuse
illumination
diffuseness
with
illumination direction [14] and that they confuse
illumination with material properties [13], due to basic
image ambiguities. In figure 5 we illustrate the diffusenessdirection ambiguity. The figure shows matte white spheres
rendered under illumination which was more or less diffuse
(from left to right) coming from the right to almost frontal
directions (from top to bottom). Note the similarities of
images along the above left to below right diagonals. This
diffuseness-direction ambiguity causes interactions of
visual judgments of the “scale of light” and the “flow of
light”. This problem was implicitly noted by Madsen and
Donn [9]. The “scale of light” estimates might improve if
they would be done for views of matte spheres
perpendicular to their average illumination directions or
light vectors, since Frandsen’s illustrations [3] were made
under this condition.
77
Figure 5 Rendered Lambertian spheres for which the illumination
diffuseness and direction were varied systemati- cally: from left to
right the diffuseness varies from halfway between fully diffuse
and hemispherically diffuse to fully collimated, and from the top
to the bottom the direction varies from 90 to 22.5 degrees in steps
of 22.5 degrees. Notice that variations along the diagonals from
above left to below right result in illuminance patterns which are
more similar than along other directions (confusing judgments of
the diffuseness and direction).
In figure 6 we demonstrate effects of type of illumination
on the appearance of a rough white sphere. The golf-ball
shows harsh body shadows and strong texture gradients for
collimated illumination (left photograph), medium strong
texture gradient for hemispherically diffuse illumination
(middle image) and hardly any contrast for totally diffuse,
Ganzfeld illumination (right image). The roughness texture
provides cues about the illumination, which are additional
to the shading and which human observers use to resolve
confounds in material and illumination judgments [13].
Therefore, we propose to examine the flow of light with a
rough sphere instead of a smooth one. The development of
simple intuitive probes for the examination of lighting
qualities stays an interesting challenge for future research.
Models of Visual Processing, (M. Landy, and J.
Movshon, eds., 1991), MIT Press, pp. 3-20.
2. Cuttle C. Lighting by design. Architectural Press,
Oxford, UK, 2003.
3. Frandsen S. The scale of light. International Lighting
Review, 3 (1987), 108-112.
4. Gershun, A. The Light Field. Transl. by P. Moon and G.
Timoshenko, J.Math.Phys. 18, 51 (1939).
5. Kahrs, J., and Calahan, S., and Carson, D., and Poster,
S. Pixel Cinematography; a lighting approach for
computer graphics. Siggraph '96 course 30 (1996).
6. Koenderink, J.J., and Pont S.C., and Doorn A.J. van,
and Kappers A.M.L., and Todd J.T. The visual light
field. Perception, 36 (2007), 1595-1610.
7. Lynes J.A. Principles of natural lighting. Elsevier
Publishing Company LTD, Barking, UK, 1968.
8. Madsen M. Light-zones: as concept and tool; An
architectural approach to the assessment of spatial and
form-giving characteristics of daylight.
9. Madsen, M., and Donn, M. Experiments with a digital
“light-flow meter” in daylit art museum e-buildings. 5th
International Radiance Workshop, Leicester, UK.
10. Mury, A.A., and Pont S.C., and Koenderink J.J. Light
field constancy within natural scenes. Applied Optics,
46, 29 (2007), 7308-7316.
Figure 6 Photographs of a golf-ball in different light fields. The
effects on the appearance are huge. The images show, from left to
right, harsh effects in collimated illumination, medium in
hemispherically diffuse illumination, and very weak in totally
diffuse (Ganzfeld) illumination.
Currently we are investigating relations between light field
descriptions and visually relevant, known and new
measures of lighting.
ACKNOWLEDGMENTS
This work was supported by the Netherlands Organisation
for Scientific Research (NWO). We thank Markus
Reisinger and Ingrid Vogels of the Visual Experiences
Group at Philips Research, Eindhoven, the Netherlands, for
collaboration and providing their outstanding facilities.
REFERENCES
1. Adelson, E.H., and Bergen J.R. The plenoptic function
and the elements of early vision, in Computational
78
11. Mury, A.A., and Pont S.C., and Koenderink J.J.
Representing the light field in finite 3D spaces from
sparse discrete samples. Applied Optics, 48, 3 (2009),
450-457.
12. Mury, A.A., The light field in natural scenes. Thesis,
Delft University of Technology, 2009.
13. te Pas, S.F., and Pont, S.C. Comparison of material and
illumination discrimination performance for real rough,
real smooth and computer generated smooth spheres.
Proceedings APGV 2005, ACM SIGGRAPH, (2005),
75-81
14. Pont S.C., and Koenderink, J.J. Matching illumination
of solid objects. Perception & Psychophysics, 69, 3
(2007), 459-468.
15. Ramamoorthi, R., and Hanrahan, P. On the relationship
between radiance and irradiance: determining the
illumination from images of a convex Lambertian
object. J. Opt. Soc. Am. A. 18, 10 (2001), 2448-2459.
Flexible Light Sources for Health and Well-being
Margreet de Kok, Herman Schoo,
Marc Koetse, Ton van Mol
Holst Centre / TNO
Hightech Campus 31
5605 KN Eindhoven
The Netherlands
ABSTRACT
The availability of flexible and conformable light sources
will enable applications that demand minimal distance
between light source and body, such as therapeutic use of
light and monitoring of personal health by wearable optical
sensors. Organic light emitting diodes are large area, low
voltage, thin light sources which can be processed on
flexible substrates like foil in Roll-to-Roll process
technology and are therefore potentially low cost. All of
these characteristics render OLED highly suitable for
applications on the body. The Holst Centre is developing
Systems in Foil, both OLEDs and sensors. Lighting
applications in health and well-being will benefit from both
functionalities.
Keywords
OLED, foil, sensors, light source
INTRODUCTION
In recent years it has become clear that there are many
positive effects of light on health and well-being of people.
Blue light for example can be used for phototherapy of
dermatological diseases like psoriasis and neonatal jaundice
[1] or even skin rejuvenation [2, 3]. Acne [4] and seasonal
affective disorder (SAD) [5] can be successfully treated
with light. Pain can be relieved with light treatment [6].
For extra-clinical treatment and home-use wearable or at
least conformable light sources are optimal. Not only the
comfort of patients will be increased, also a serious cost
reduction will be possible with these conformable light
sources as the burden on clinical personnel and space will
become less for cases which do not need continuous
supervision.
With these wearable light sources, phototreatment can not
only be used in healthcare, but mainly in well-being
applications such as pain relief, anti-wrinkle measures and
perfusion enhancement.
Integration of inorganic LEDs in textile is developed at the
moment. Organic LEDs are an interesting alternative light
source, because of their intrinsic large area and the
possibility to be produced in a cost effective Roll-to-Roll
processing. Furthermore they operate at a low voltage and
will be energy efficient. When in contact to the skin,
permeability to ensure regulated humidity at the skindevice interface and sterile contact areas are prerequisites
and to be taken into account when developing these
systems for health and well-being.
Light can also be of use for well-being and health in optical
sensors. The condition of the body in blood perfusion,
saturation, skin tone, etc. can be determined when light
sources are combined with photodetectors. Both
functionalities can be achieved in foil based structures with
organic active materials.
These conformable light sources with integrated sensors are
yet to be developed for commercialisation. The research at
Holst Centre is focusing on enabling technology for these
applications. In an open innovation environment, together
with industrial and academic partners We develop
technology platforms for autonomous wireless transducer
solutions and systems-in-foil. The Systems-in-Foil Program
Line targets to develop new device architectures,
technologies and production processes for foil based
electronic devices that will revolutionize the electronics
industry. It will enable new ultra-light, ultra-thin, flexible,
easy-to-wear electronic products such as lighting and
signage devices, reusable and disposable sensor devices,
foldable solar and battery panels and displays. The research
addresses batch-wise and web-based processing (Roll-toRoll), encompassing processes like printing, vacuum
deposition, lithography, lamination and interconnection,
which will enable the manufacturing of these devices in
large sizes and quantities at low costs.
We will hereafter highlight the components that can be
realised in foil and which are of most interest in light for
health and well-being: organic light emitting diodes,
organic photo diode, and the assembled sensors.
Organic Light-emitting diodes
An organic light-emitting diode (OLED) consists of an
active material, sandwiched between two electrodes. At
least one of the electrodes must be transparent. By applying
a voltage, charge carriers are injected into the light-emitting
material. Upon recombination excitons are formed and by
subsequent radiative decay photons are emitted from the
device. With a transparent cathode a top emissive OLED is
formed, whereas a reflective cathode with transparent
anode results in a bottom emissive OLED is the result (see
figure 1). With two transparent electrodes a transparent
OLED can be constructed. By this concept light sources
79
can be realised which are transparent in the off state, i.e.
they can be integrated imperceptibly. Transparent windows
generating light in the evening is one example of an
envisioned application.
a)
top emissive OLED
b)
to ensure a homogeneous current distribution over the
complete area. At Holst Centre a combination of printed Ag
shunt lines and PEDOT:PSS, (poly (3,4-ethylene dioxy
thiophene) : poly (styrene sulphonic acid), was recently
presented to be a viable alternative[7].
Because of the reactive materials used in OLED devices
encapsulation is key to ensure long lifetimes and reliable
systems. If water is allowed to come into contact with the
cathode, it is oxidised and the electron injection blocked
which becomes visible as dark areas in the OLED. This can
be prevented by thin film barrier technology. Holst Centre
has developed a barrier with a water vapour transmission
rate through the barrier well below 10-5 g/m2day under
ambient conditions without visible defects in a Ca-mirror
test for 67 days. These barriers are capable of bending to
radii of 20 mm which allows roll to roll processing [8].
b) bottom emissive OLED
Figure 1 OLED cross-section
In the foil based OLED program of Holst Centre,
challenges like design-layout are studied. Flexibility,
encapsulation and related lifetime are being optimised.
Figure 3 Top emissive OLED (100 cm2) on metal foil
thin film encapsulated
Figure 2 Bottom emissive ITO-less OLED (144 cm!)
with inkjet printed metal grid
!
Recent advances in the Holst Centre include improved light
homogeneity and increased reliability. In case a transparent
anode is applied, a transparent conductive oxide (TCO) like
indium tin oxide (ITO) can be used. ITO is however
relatively expensive and shows only limited compatibility
with roll to roll processes. Such high speed, low cost
processing is considered to be required to meet the low cost
per area for these light sources. Moreover in order to fully
exploit the large area, the TCO must be conductive enough
80
Nowadays energy consumption is a very important topic
and therefore the increase of efficiency of OLEDs is given
a lot of attention. Reported record value for white light by
Konica Minolta is 64 lm/W at 1000 cd/m2 and a lifetime
(50 % luminance decrease) of 10.000 hrs. Kido et al. also
reported similar high power efficiency for white OLEDs:
63 lm/W and 64 cd/A at a luminance of 100 cd/m2 at the
MRS meeting in 2006. For blue light emitting devices,
record value is 50 lm/W [10], for green OLEDs 130 lm/W
[9].
Small molecule OLEDs show currently higher efficiencies
and lifetimes than polymer OLEDs. However there is also
progress in the latter field: a luminous efficacy of 25 lm/W,
39 cd/A was reported for a white OLED based on a host
polymer with phosphorescent dye [11]. A fluorescent,
blended polymer system showed 16 lm/W and an external
quantum efficiency of 6 % [12]. For roll to roll process
technology solution processing has substantial advantages
and solution processing of small molecules is therefore
gaining interest.
Current challenges for OLEDs on foil lie in increasing
efficacy, flexibility, lifetime and reliability.
For on the body applications a new generation of OLEDs
are needed to adjust to the shape of the underlying person.
The next generation of conformable OLEDs demand to be
truly stretchable and not only bendable. This will cost
considerable efforts. Up till now several stretchable
electronic circuits have been published including
conductive wiring into a stretchable matrix of PDMS or
thermoplastic polyurethane [13-15]. The method to render
the light generating area stretchable is still to be explored.
Sensors in Foil
The system in foil approach as taken by the Holst Centre
will allow to add more functionalities like sensors to the
light generating foil. With sensors, feed-back of the
effectiveness of the phototherapy can be derived and active
control of the phototherapy will become possible. Sensor
based functionalities might be focusing on the perfusion of
tissue, saturation of the tissue or biological processes
associated with healing of f.e. burn wounds. Abnormalities
in the healing process, like infection, should at best be
detected at a very early stage allowing for the proper action
to be taken to minimize negative effects and promote the
healing process. These biological processes can be detected
by several physical parameters: colour of the skin (e.g. red
for infection), a temperature or chemical substances in the
wound fluids as indicators. These parameters can be
monitored by dedicated sensors of which several examples
will be discussed.
The in plane optical sensor (IPOS) platform, developed at
the Holst Centre as a platform for many application areas
(Figure 4). It can serve as a chemical sensor but can also be
used for direct optical measurements of physiological
parameters on the skin.
Figure 5 Artist impression of a smart bandage,
integrating organic electronic devices with a wound
dressing.
A foil based sensor capable of measuring the perfusion of
the microvascular tissue in the wound area by means of
photoplethysmography has already been shown [16] (PPG).
Traditionally, a PPG is recorded with a pulse-oximeter
giving additional information on the blood oxygen
saturation (SpO2). This is achieved by measuring the
change in absorption due to the pulse (blood volume) at
two wavelengths, typically in the red and near infrared
from which the ratio between haemoglobin and
oxyhaemoglobin is derived (Figure 6). For perfusion only
one wavelength is sufficient, thereby simplifying design
and manufacturing of the sensor device.
Apart from being a platform for use in many application
areas, the IPOS may also be seen as a platform for the
development and testing of manufacturing technologies.
These include printing technologies for the active materials,
barrier development for encapsulation of the devices,
lamination and interconnection technologies for the final
device assembly, printing of conductive structures, and
lithography on foil.
Figure 4 Principle of In plane Optical Sensor (IPOS)
For our sensor application we aim at an array of organic
photo detectors (OPD) and a compatible array of organic
light emitting diodes (OLED). The OPDs are based on a
blend of poly(3-hexylthiophene) (P3HT, Merck Chemicals
Ltd) and [6,6]-phenyl C61butyric acid methyl ester
(PCBM, Solenne BV). This blend is used for photovoltaic
research but has also been well studied for use in
photodetectors [17, 18]. This blend has an optical band gap
of 650 nm which is able to detect part of the red light.
An artist impression of such an application, a smart
bandage, is given in Figure 5. The fact that printing can be
used as a processing technology allows for the construction
of arrays of optical elements, which may be advantageous
for large wound areas or a combination of functions.
The light emitting polymers (LEP) used for this study are a
red emitting [19] and yellow emitting [20] material,
obtained from Merck OLED Materials GmbH. The
emission of the yellow material (!max 575 nm) has a strong
emission shoulder in the red spectral region and has a
strong overlap with the diode spectrum.
81
Figure 6 Absorption of haemoglobin and oxyhaemoglobin in the part of the spectrum that is useful
for pulse-oximetry
In order to increase the flexibility with regard to the
fabrication and design of these devices, we decided to
fabricate the detectors and LEDs on separate foils. For
example, this allows comparison of various printing and
coating
technologies
and
substrates,
poly(ethylenenaphthalene) (PEN) and glass. A further
implication of this approach is that one is essentially free to
choose the order of functional foils in the final device.
B
C
A
Figure 7 Mock up version of the sensor node showing
the optical array (A), interconnection (B) and wireless
node (C).
For the device discussed in this study we opted for a three
foil assembly containing an OLED foil with an OPD foil
laminated on it. The device is finished with a flexible
circuit board containing noise filters (band pass 0.5 – 17
Hz), logarithmic amplifier and DC/DC converter for the
power supply of the OLEDs. Data collection and OLED
driving is controlled with a microprocessor embedded on a
multifunctional wireless node [21]. The interconnection
between the foils is achieved using a propriety lamination
and interconnection technology. Figure 7 shows a mock up
version used for testing the interconnection technology and
of the attached node as well as a picture of the OLED and
OPD foils.
82
Figure 8 OLED (foreground) and OPD (background)
foils.
The functional foils were processed in a batch wise manner.
A typical work flow for both the OLED and OPD foils
involves: lamination of a substrate to a carrier; deposition
of the barrier; deposition and patterning of the anode and
shunt lines; deposition of PEDOT:PSS and the active layer
by means of spin coating or inkjet printing; evaporation of
the cathode; and, finally thin film encapsulation [85]. Both
foils can be made either bottom or top emissive (receptive)
giving a large design freedom. Although processing on
glass is a well know procedure, our devices were designed
to be double sided. In such a device light leakage through
the substrate can be avoided by using a top emissive OLED
on one side of the substrate en a bottom receptive OPD on
the other. This light will be emitted close to the skin and
only reflected and scattered light will pass through a single
substrate. Classical encapsulation with a metal lid is
therefore not useful and a transparent thin film
encapsulation was used. A further advantage of using a
double sided device is that spin-coating can be used
without the risk of contamination of the devices. A
schematic overview of the architecture of both device
designs is given in figure 4. The photodiodes with an area
of 1 mm2 are placed behind the LEDs (8 mm2) and receive
the reflected light via an opening in the middle.
Lambertian mode but has a maximum between 40° and
50°. Control of this angle dependence, would give an
interesting tool for light incoupling.
A
B
C
Figure 9 Schematic overview of the buildup (A) and
cross section (B,C) of the devices. B shows the
architecture in case a top emissive OLED (light gray) is
used in combination with a bottom receptive OPD
(black). The dotted line in the middle may be seen as the
adhesive in case foil is the substrate. C depicts the
combination of both bottom emitting and receptive
devices. For clarity, the size of the actual devices are
strongly exaggerated.
Top-emissive OLEDs
Bottom-emissive OLEDs are considered Lambertian
emitters. As indicated before this might reduce the amount
of light that is able to penetrate the skin because of
“leakage” into the substrate. Clearly, the loss of light is
strongly dependent on the thickness of the substrate. For
the devices on glass and the first generation on foil (Figure
9, B) we therefore chose to use top emissive OLEDs. Here
the light has to pass only through the thin film
encapsulation, minimising this loss.
The red emitting material shows a peak emission at 670nm
and has a large overlap with the absorption spectrum of the
photodiode blend (Figure 10). Also the emission overlaps
with the prime wavelength for the PPG measurement, 650
nm. This makes this material particularly useful for the
saturation sensor (Figure 6).
The bottom-emissive devices had a maximum efficiency of
1.5 Cd/A. In top emissive devices this was reduced to a
maximum efficiency of 0.45 cd/A at 8V and 320 cd/m2 as
measured with a luminance meter. The area of the LED
was 8 mm2, implying a current of 5.6 mA. These results
show that the luminescence of top emissive devices is
much lower than their bottom emissive counterpart. This is
partly due to the reduced transparency of the cathode (6070%) and partly due to the angle dependent emission by a
cavity effect. Figure 10 shows a conoscopic measurement
(Eldim, EZ-Contrast L160D) of a top-emission device
driven at 9 V. Clearly the light is not emitted in a
Figure 10 Emission spectrum of the red emitting OLED
(top) and Cross section of a conoscopic image in
Luminance (Cd/m2) vs. the viewing angle (bottom).
Organic photodiodes
The photodiodes used in this study were bottom-receptive
and had an area of 1 mm2. The active layer was spin coated
from a chlorine free solvent. No separate thermal annealing
was required. A TEM picture (Figure 11) of an inkjet
printed layer, using the same solvent, showed similar
features as reported in literature [23]. This indicates that the
morphology is very similar to that of films obtained by
chlorinated solvents such as ortho dichloro benzene. This
was corroborated with IV-measurements of both spin
coated and inkjet printed devices using LiF/Al as cathode.
Both showed high short circuit currents (Jsc) of 10 and 8.7
mA/cm2, respectively (Figure 11) under approximately 100
mW/cm2 white light illumination. The open circuit voltage
(Voc) was 0.58 V with a fill factor (FF) of 0.56 for the spin
coated device. The printed device showed a slightly lower
Voc of 0.54 V but a dramatically lower FF. We attribute
this to the inhomogeneity of the printed layer. Optimisation
of the printing process including ink formulation is
currently being carried out.
83
Photoplethysmography with OLEDs and OPDs
For the proof of principle we made use of a double side
glass based device using two red emitting OLEDs and one
photodiode. The OLEDs were driven at 9V. The measured
photocurrents were filtered (band pass 0.5-17 Hz) and the
AC component was amplified using a logarithmic amplifier
(AD8304). The measurement was performed on the right
index finger of a test person. A simultaneous measurement
with a commercial pulse-oximeter (Nelcor N200) on the
middle finger served as control. The resulting
photoplethysmograms are shown in Figure 12.
Figure 12 Photoplethysmogram showing the pulse of a
test person measured with the organic device (top) and
a commercial pulse-oximeter (bottom). The signals have
been shifted on the y axis for clarity.
The signal of the organic device and the control match
perfectly showing that our sensor can be used for
measuring the pulse. The PPG is typical for a measurement
on a finger [16]. The commercial device delivers a strongly
smoothed signal, whereas the signal of the organic device
was only slightly smoothed. The total signal measured by
the photodiode was 10 !A, the relevant modulation (AC
signal) was approximately 50 nA, estimated from the
amplifier characteristics [24].
Figure 11 TEM image of an inkjet printed film of
P3HT/PCBM blend (top) and IV curves of a spin coated
device (bottom) and inkjet printed device (middle).
The noise levels of our devices can be relatively easily
estimated from the shunt resistance (Rsh) measured in the
dark. The measured thermal noise currents are in the order
of 20 fA/Hz1/2. This is slightly higher than Si diodes.
However, these levels are well below the measured
currents, during operation (nA).
84
CONCLUSIONS
Homogeneous flexible light sources can be produced at
large areas by OLEDs comprising thin film encapsulation
and shunting lines for homogeneous current density. Recent
advances at the Holst Centre have shown that areas of more
than 100 cm2 become feasible and compatible with Roll-toRoll processing, thereby allowing low cost production
methods. Holst Centre’s strategy furthermore focuses on
sensors incorporated in foils. Both functional components
are ultimately suited to serve in devices for health and wellbeing in phototherapeutic applications like dermatological
disorders and pain relief. Future developments are expected
to increase conformability, area and efficacy of light
sources.
Sensors based on organic electronic devices allow for such
applications but can also be used for direct measurements
of physiological processes on the body. We have shown a
functional prototype comprising an in plane optical sensor
node containing OLEDs and OPDs on foil that can be
integrated with existing electronics using lamination
technologies. More specifically the sensor was designed for
photoplethysmography, measuring the perfusion in the
skin. The sensor produced a signal from measurement on
finger extremity which was comparable to a commercial
pulse-oximeter, This shows that organic optoelectronic
devices can well be used for the direct measurement of
physiological parameters such as the perfusion in skin.
We foresee a bright future for light sources and associated
sensor systems based on organic electronics in the
application fields of health and well-being.
ACKNOWLEDGMENTS
We thank our colleagues at the Holst Centre in the
development of lighting and sensor systems for the health
applications as well as the industrial residents involved in
this project especially our Philips colleagues; in particular
Liesbeth van Pieterson.
REFERENCES
1. H. Vreman, Seminars in perinatology (2004). 326-333.
2. M.A. Trelles, S. Mordon, R. Glen Calderhead, Facial
rejuvenation and light: our personal experience Lasers
Med. Sci. 22, 93-99 (2007)
3. S.Y. Lee, A prospective, randomized, placebocontrolled, double-blinded, and split face clinical study
of LED phototherapy for skin rejuvenation J.
Photochem and Photobiol. B: Biology 88 51-67 (2007)
4. G. Munavalli, R.A. Weiss, Evidence for Laser and
Light-Based Treatment of Acne Vulgaris,Semin Cutan
Med Surg 27,: 207-211, (2008)
5. G. Tonello, Seasonal Affective disorder: Lighting
research and Environmental Psychology Lighting Res.
Technol. 40, 103-110. (2008)
6. R.T. Chow, L. Barnsley Systematic Review of the
Literature of Low-Level laser Therapy (LLLT) in the
management of Neck Pain Lasers in Surgery and
Medicine 37:46-52 (2005)
7. E. Rubingh, P. Kruijt, R. Andriessen, Proceeding
LOPE-C
2009,
press
release:
http://www.holstcentre.com/NewsPress
8. F. van Assche, H. Rooms, E. Young, J. Michels, T. van
Mol. G. Rietjens, P. van de Weijer, P. Bouten,
Proceedings AIMCAL (2008)
9. J. Kido, IMR Workshop on organic light emitting
devices (23/24 jan 2009) Sendai Japan conference
proceedings
10. S. Watanabe et al. J. Photopolymer Science and
Technology 18, 83(2005)
11. B.C. Krummacher, V.E. Choong, M.K. Mathai, S.A.
Choulis, F. So Appl. Phys. Lett. 88, 113506(2006)
12. J. Huang, G. Li, E. Wu, X. Xu, Y. Yang, Adv. Mater.
18, 114(2006)
13. T. Loeher, D. Manessis, R. Heinrich, B. Schmied, Jan
Vanfleteren, J. DeBaets, A. Ostmann, Herbert Reichl,
Electronics Packaging Technology Conference 271-276
(2006)
14. D.H. Kim, J.A. Rogers, Adv. Mater., 20, 4887-4892
(2008)
15. T. Skitani, H. Nakajima, H. Maeda, Takanori
Fukushima, T. Aida, K. Hata, T. Someya, Nature
Materials 8, 494-499 (2009)
16.
J. Allen, “Photoplethysmography and its application in
clinical physiological measurement,” Physiol. Meas., 28
R1–R39 (2007)
17. P. Schilinsky, C. Waldrauf, J. Hauch, C. Brabec,
“Polymer photovoltaic detectors: Progress and recent
developments,” Thin Solid Films, 451-452, 105-108
(2004)
18. M. Ramuz, L. Bürgi, C. Winnewisser , P. Seitz, “High
sensitivity organic photodiodes with low dark currents
and increase lifetimes,” Org. Elect. 9, 369-376 (2008)
19. A. Falcou H. Becker, E. Breuning, A. , Buesing, S. ,
Heun, A. Parham H. , Spreitzer,J. Steiger, P, Stoessel, ,
“Light Emitting Polymer Materials: The working base
for flexible full color displays,” Proceedings of the third
International Meeting on Information Display, 10531056 (2003)
20. H. Becker H., Spreitzer H., Kreuder W., Kluge E.,
Schenk H., Parker I., Cao Y., “Soluble PPVs with
enhanced performance, a mechanistic approach,” Adv.
Mater. 12, 42-48 (2000)
21. B. Gyselinckx, J. Penders,, R. Vullers “Potential and
challenges of body area networks for cardiac
monitoring,” J. Electrocardiol., 40 S165-S168 (2007)
22. M. Hemerik, R. Van Erven, R. Vangheluwe, J. Yang, T.
Van Rijswijk, R. Winters, B. Van Rens, “Lifetime of
Thin-Film Encapsulation and its Impact on OLED
Device Performance,” SID Symposium Digest 37, 1571
(2006)
23. X. Yang, J. Loos S. Veenstra, W. Verhees, M. Wienk,
J. Kroon, M. Michels, R. Janssen, “Nanoscale
morphology of high-Performance polymer solar cells,”
Nano Lett., 5, 579 -583 (2005)
24. Analogue
Devices
AD8304,
http://www.analog.com/static/importedfiles/data_sheets/AD8304.pd
85
Effect of Glazing Types on Daylight Quality in Interiors:
Conclusions from Three Scale Model Studies
Marie-Claude Dubois
École d’architecture
Université Laval
1 Côte de la Fabrique
Québec, QC, Canada
G1R 3V6
+1 418 656 2131 ext 5010
marie-claude.dubois@arc.ulaval.ca
ABSTRACT
This paper reports the results of three scale model studies
about the effect of glazing types on daylight quality in
interiors. This paper emphasizes on the constancy in the
results of the three studies, which indicate that glazing
types have a statistically significant effect on the perception
of brightness, naturalness, beauty-pleasantness and
precision. Glazing types with higher visual transmittance
yield brighter, more natural, beautiful-pleasant and sharp
views of the interior. The three studies also indicate that the
glazing visual transmittance is negatively correlated with
glare comfort: higher transmittance glazing types result in
more glaring views of the interior. Finally, the three studies
show that the glazing type has no effect on the perception
of shadows.
Keywords
Daylight quality, windows, glazing, visual transmittance,
glare, tinted glazing, reflective coating, low-e coating.
INTRODUCTION
Fifteen generations ago, most of our ancestors spent the
majority of their waking hours outdoors and buildings
primarily provided only shelter and security during the
hours of darkness [1]. Today, people spend nearly 85-90%
of their time indoors [2] and the interior of buildings is the
main scenery supporting daily lives.
In interior environments, a contact with the exterior, natural
world is of prime importance, and this contact is made
possible by the window glazing material. Window glazing
is the primary filter of daylight in a building and the main
interface between the interior and exterior worlds. A large
field study carried out in Denmark [3], indicated that
“being able to see outside” was the most important benefit
of windows for office workers.
Recently, the need for energy conservation in buildings has
spurred innovations in window technologies. The use of
coated and tinted glazing is one of the strategies that can
improve energy efficiency of buildings [4]. Window
86
coatings and tints alter the quantity and spectral quality of
daylight, which may have an effect on user satisfaction,
daylight utilization and even photobiological responses in
humans. According to Chain et al. [5], glazing types which
are thermally efficient are rarely evaluated according to
their visual impact: grey or green tints can lead to the
impression of being sick. Glazing types which are
thermally efficient may also produce a colour distortion of
the natural light spectrum which may affect pupillary
reflex, alertness, mood and performance in fully daylit
buildings. A recent discovery [6, 7] about circadian retinal
photoreceptors (photosensitive retinal ganglion cells:
ipRGCs) suggests that short-wavelength (blue) light is
associated with the good functioning of neuro-endocrinal
systems and circadian cycles, with evidence for the
involvement of these ipRGCs cells in pupillary reflex,
alertness, mood and performance [8].
This paper presents the results of three studies of the effect
of glazing types on daylight quality in interiors. The three
studies were achieved in scale models under artificial as
well as natural skies in Denmark and Canada, using a
within-subject experimental design. The objective of the
artificial and natural sky studies was to examine the
relationship between the optical properties and colour
coordinates of different glazing types and various
qualitative factors related to daylight quality: brightness,
glare, naturalness, beauty and pleasantness, precision, light
distribution and shadows. This paper emphasizes on the
constancy in the results obtained in the three studies.
LITERATURE REVIEW
There are a large number of researches about artificial light
sources in terms of spectral characterization [9] and effects
on occupant satisfaction, performance, mood [e.g. 10, 11].
Previous research on electric lighting strongly suggests that
both brightness and spectral distribution are contributing to
the visual experience, to perception and performance in a
space. One study [12] indicated a relationship between
desktop daylight illuminance and the preferred colour
temperature: low daylight levels (500 lux) cause preferred
CCT around 3300 °K, while higher daylight levels (1500
lux) result in increased CCT to 4300 °K. This is in
agreement with the Curve of Amenity for artificial lighting
(Kruithof Diagram [13]), which shows that the higher the
overall lighting level, the higher its colour temperature
should be. Conversely, high colour temperatures under low
luminance tend to make the space look cold and dark, while
low colour temperatures under high lighting level tend to
make the space look artificial [14]. In relation with this
research on electric lighting, older research on glazing
types [15, 16] indicated that solar bronze glass (warm shift)
had been found to give an enhanced perception of the same
transmittance while solextra glass had been found to give a
reduced perception of brightness relative to a spectrally
neutral glass of the same transmittance.
Overall, research specifically focused on the effect of
window glazing type on daylight quality is scarce, dated or
confounding:
In a recent doctoral thesis [17], computer simulations with
Lightscape were used to assess the effect of two tinted
glazing (bronze and green) on indoor correlated colour
temperature (CCT). The study showed that tinted glazing
greatly affected the interior average CCT but the author
concluded that this effect would not be important since the
occupant would be chromatically adapted to the scene. One
research [18], which examined the attitudes towards the use
of heat rejecting or low-light-transmission glasses in highrise office buildings, supports this statement. This research
concluded that tinted glass had little or no effect on the
visual environment.
On the contrary, another study [19] indicated that people
were clearly able to distinguish between a standard threepane clear glass window and a super insulated four-pane
window (green shift) in a full-scale laboratory experiment
where two identical rooms furnished alternately as office
and bedroom were evaluated by 95 subjects using a
between-subject, random order experimental design. The
room with the four-pane window felt more enclosed, and
the daylight felt less strong and clear. The four-pane
window also affected colour perception, making the colours
of the room and of the view look drabber.
In a series of experimental model studies [20], where room
and window size, room décor, illuminance (total incident
light) level and light colour were manipulated, the
responses of office staff to the appearances of windows
with variable glazing transmission characteristics were
analysed. The study showed that the acceptability of an
office can be increased by the use of reduced transmittance
glazing, and that generally, there is a preference for a
colour effect that gives a warm shift, but the author
concluded that these preferences can be influenced by room
and window sizes and by room décor.
In another experimental study [21] about the minimum
acceptable transmittance of glazing where three types of
glass (spectrally neutral, brightness enhancing solar bronze
and brightness reducing solextra) were tested under a range
of conditions by subjects viewing a real sky and scenery
through the window of a model office, the authors
concluded that the minimum acceptable glazing visual
transmittance lied in the range 25-38%. The study also
pointed no statistically significant difference between the
spectrally neutral glass and the brightness reducing solextra
glass regarding the minimum acceptable transmittance.
In summary, this literature review yields the following
conclusions:
! Two studies [17, 18] suggest that tinted glazing has no
effect on the visual environment due to adaptation of the
visual system.
! In contradiction, another study [19] indicates that the
glazing type has a significant effect on the perception of
the visual environment: a four-pane window with two
coatings (green shift) makes the room feel more enclosed
and gives an appearance of weaker daylight and drabber
colours.
! On the other hand, Cuttle [20] found that the acceptability
of an office can be increased by the use of reduced
transmittance glazing. Studies by Boyce et al. [21] even
concluded that quite low glazing visual transmittance is
acceptable.
! Cuttle [20] found a preference for a warm shift, a result
which seems to agree with the results of Boyce et al. [15,
16] who found that a brightness enhancing solar bronze
glazing is perceived to result in a brighter room
environment than the spectrally neutral glazing. This
result does not necessarily disagree with Bülow-Hübe’s
experiment [19] where a window with a green shift,
(which is opposite to a warm shift) was studied.
METHOD
This paper presents the results of three separate scale model
studies where the effect of window glazing types on
daylight quality in interiors was investigated. The first
study [22] was carried out at the Danish Building Research
Institute in Hørsholm, Denmark (lat. 55,4° N) and the
second [23] and third [24] studies were achieved at the
École d’Architecture of Université Laval, Quebec City,
Canada (lat. 46,5° N).
Scale models
The three studies were achieved using scale models of a
regular rectangular room. The scales used were 1: 7,5 in the
first study and 1:6 in the second and third studies.
Previous research [25, 26] has shown that scale model
studies are a quick and reliable method provided that the
scale is not too small and that great care is taken to
represent the details and décor of the real environment.
Room geometries and furniture
In all three studies, a regular rectangular room with a single
87
window was simulated. In the first study, no specific décor
was represented; the room only contained a scaled table
and some objects that research participants could observe
(Fig. 1). In the second and third studies, the room was fully
furnished as a typical residential living room with sofa,
table, book shelf, curtains, etc. (Fig. 2).
In the first (Danish) study, paired comparisons were used.
Athough all efforts were made to make the two rooms look
exactly the same, some unintentional differences between
the Test and Reference rooms did appear, which may have
introduced some bias in this first study. The two identical
scale models with interior dimensions of 3,5 * 6,0 * 3,0 m3
(width * depth * height, full-scale) were built and placed
next to one another. The scale models thus measured 0,47 *
0,8 * 0,4 m3 (w * d * h). Each scale model had a unique
opening for the window measuring 0,17 * 0,24 m2 (height *
width) placed 0,18 m above the floor (window full-scale
dimensions were 1,2 * 1,8 m2, located 1,35 m above the
floor). Opposite the window, a small horizontal hole
allowed the research participants to make their
observations. The research participants thus looked straight
ahead towards the window when making their assessments.
The interior of both scale models was painted a diffuse
white colour (refl. > 85%) and furnished with a brown,
scaled table, a silver key (on the table), a piece of broccoli,
a baby tomato, a pine cone, a staple remover and a yellow
tennis ball (Fig. 1). There was no electric lighting in the
scaled rooms.
In the second and third studies, paired comparisons were
abandoned and thus a single scale model (1:6) was built of
a typical living room measuring 0,92 m by 0,66 m (width x
depth, full scale: 5,5 m x 3,9 m) with a single, centrally
placed window measuring 0,41 m by 0,22 m (width x
height, full scale: 2,4 m x 1,3 m), with window sill height at
0,14 m (full scale: 0,8 m) from the floor (Fig. 2). The scale
selected for the study allowed a detailed and faithful
representation of furniture, interior finishes of various
colours and textures. The walls were painted a light beige
(70% reflectance), the ceiling was white (74% reflectance)
and the floor was covered with a veneer similar to a
wooden floor (52% reflectance). There was no electric
lighting in the scaled room. The research participants
observed the room through an opened door on the lateral
wall of the model.
Fig. 1 Photograph showing the interior of the scaled
room in the first (Danish) study.
Fig. 2 Photograph showing the interior of the scaled
room in the second and third (Canadian) studies.
Orientation, sky conditions and view out
In the first (Danish) study, both scale models (Reference
and Test rooms) were placed behind the window of an
empty office room at the Danish Building Research
Institute (SBI), Hørsholm, Denmark. This window, which
had a north orientation, was replaced by a single, iron free
window pane (daylight transmittance = 91%). This study
was entirely achieved during January and February 2002,
between 09.30 and 15.00 hours, under overcast sky
conditions, in order to make sure that exterior daylight
conditions were as constant as possible. The window
allowed a view of a white sculpture placed on a grass lawn
and surrounded by trees and shrubs (Fig. 1).
In the second study, the experiments took place during May
2007, between 10.00 and 19.00 hours. The scale model was
placed next to the artificial sky of the École d’Architecture
of Université Laval, Québec, Canada. The light coming
from the artificial sky penetrated through the window
opening of the scale model. This artificial sky consists of a
mirror box measuring 1,22 m x 1,22 m x 1,22 m, which has
the distribution of a typical CIE overcast sky. The
illumination is achieved with “daylight” fluorescent tubes
placed above a diffuser (acrylic white sheet). In this second
experiment, the visible part of the artificial sky was
88
simulated as a typical landscape of a suburb using small
plants and shrubs (Fig. 3).
because it was the most neutral in colour and had a high
transmittance.
In the second and third studies, paired comparisons were
abandoned and thus a single scale model was used
throughout. The window opening of the scale model was
alternately covered by different glazing samples presented
in random order during the experiments. A set of glazing
samples commonly used in residential buildings was
chosen from a stock of samples provided by local glazing
manufacturers. A total of seven glazing types were selected
based on their optical properties (Table 1). Glazing type
A83 was selected because it was an iron-free combination.
The other glazing types were selected because of the
availability of measured optical data.
Fig. 3 Photograph showing the view of the interior of the
room towards the window in the second study.
Fig. 4 Photograph showing the view of the interior of the
room towards the window in the third study.
In the third study, the same scale model (as in the second
study) was moved so as to expose the window to the
natural climate. The window of the scale model was
oriented facing the south-east direction. The experiments
took place during October 2007 between 12.00 and 16.00
hours to avoid the presence of any direct sunlight
penetration due to the south-east orientation of the window.
The position of the observation hole was the same as in the
second study (i.e. via a lateral door). The exterior scene
viewed through the window of the scale model consisted of
one of Quebec City’s most beautiful views of a park
overlooking St-Lawrence River (Fig. 4).
In the third study, only five glazing samples were selected
from the second study (Table 1, see *). Glazing B82 was
selected because it is one of the most common glazing
assemblies in older buildings. Apart from glazing G38, all
glazing samples looked almost the same in all three studies;
the differences between the samples were subtle and the
research assistant needed to look at the code written on the
side of each sample to be able to identify it.
In the second and third studies, the effect of glazing type on
the transmitted light colour was determined using a digital
photographic image technique [see 24], which consisted of
subtracting colours from two digital images: a referenceand a test-photo. Since colour temperature of daylight
varies according to sky type and time of day, four series of
photos were taken, each one corresponding to a sky type:
(1) clear sky, (2) partly sunny, (3) partly cloudy and (4)
overcast. The colour subtractions were then performed
using Photoshop CS2 according to the three channels (red,
green and blue) of the RGB model. The RGB data were
then converted to CIE-L*a*b coordinates using a colour
calculator (see Fig. 5). The more salient points of this
analysis are summarized below:
! Glazing A83 yields a negligible green shift and small
yellow shift;
! Glazing B82 is significantly greener and slightly yellower
than A83;
! Glazing types C74 and F65 exhibit colour shifts in two
directions (similar values for both axes);
! Glazing G38 yields the strongest green shift but the
weakest yellow shift amongst all glazing types studied. A
blue shift appears under clear sky conditions.
Glazing samples tested
In the first (Danish) study, paired comparisons were used
with one Reference and one Test room. The glazing
samples tested (Table 1) were selected because they are
widely used in Denmark according to the glazing
manufacturer who provided the samples and thermaloptical data. The Reference glazing (Ref77) was selected
89
Table 1 Glazing samples tested in the three studies with their thermal and optical properties.
Name
Description
U-value
cog
Study 1 (Denmark)
Solar
Tr
CIE*Lab
Rext
Rint
Tr
(W/m °C)
(%)
a
b
(%)
(%)
(%)
2
Study 2 and 3* (Canada)
Daylight
A79
1 cl. + 1cl. low-e (s)
1.12
79
-2.8
2.3
11
12
63
B76
1 cl. + 1 cl. low-e (h)
1.45
76
-2.7
3.2
17
16
72
C70
1cl. low-e (s) + 1 cl. + 1cl. low-e (s)
0.46
70
-4.3
4.1
14
14
46
D66
1 solar low-e + 1 cl.
1.12
66
-7.0
8.8
20
18
42
E50
1 solar low-e + 1 cl.
1.04
50
-8.9
3.2
18
15
26
Ref77
1 ironfree + 1 cl. low-e (h)
1.45
77
-1.8
2.2
17
16
79
A83*
1 iron-free + 1 iron-free
n/a
83
n/a
n/a
7
7
79
B82*
1 cl. + 1 cl.
n/a
82
n/a
n/a
n/a
n/a
n/a
C74*
1 cl. + 1 low-e (Ti-PS)
n/a
74
n/a
n/a
11
12
46
D73
1 cl. + 1 low-e (h)
n/a
73
n/a
n/a
16
15
54
E68
1 Ti-AC 40 + 1 cl.
n/a
68
n/a
n/a
9
11
34
F65*
1 Ti-AC 36 + 1 cl.
n/a
65
n/a
n/a
10
10
31
G38*
1 Ti-AC 23 + 1 cl.
n/a
38
n/a
n/a
13
11
18
n/a= data not available
Research participants
In all three studies, the research participants were recruited
by email. In the first study, the participants were mainly
from the administrative staff of the Danish Building
Research Institute while in the second and third studies, the
participants were mainly students from the school of
architecture. The participants were not paid to take part in
the experiments and they were not aware of the real goal of
the research. The participants over 45 years old or with
important visual problems, as well as those with knowledge
about windows or glazing were excluded from the study. A
total of 18 participants (9 males) took part in the first study,
15 (7 males) in the second study and 30 (16 males) in the
third study.
Questionnaire
In the first study, a two-page questionnaire (see [22]) which
used semantic, seven-grade, bipolar scales was developed
to cover most of the following dimensions of light:
brightness, distribution, shadows, reflexes, glare, light
colour and colours. These factors were identified in the
literature as the most important for the description of light
quality in interior spaces [27]. The questionnaire was
adapted from Bülow-Hübe [19] and focused more
specifically on daylight intensity and colour, colours in the
interior and in the view out, glare, shadows and textures.
After the first study, a principal component analysis (PCA)
was carried on the 27 questions of the questionnaire. Using
90
Jolliffe’s criteria [28] which consists of retaining only the
factors with associated eigenvalue larger than 0,70, the
PCA retained seven (F1-F7) factors explaining 79,5% of
the variance in the scores. The number of factors was also
confirmed by the scree plot test [29]. The results of the
PCA (with a Varimax rotation) and the interpretation of
each factor retained is presented in detail in the relevant
paper [22].
The questionnaire for the second and third studies was
developed from the first questionnaire, taking into
consideration the seven factors identified previously: (F1)
brightness, (F2) glare comfort, (F3) naturalness, (F4)
beauty pleasantness, (F5) precision, (F6) distribution and
(F7) shadows. The factor “colour temperature” (from the
first study) was abandoned since the research participants
did not know what it referred to and we introduced a factor
called “distribution”, since this parameter is often discussed
in lighting design and research. The other qualitative
factors were used when designing the questionnaire and
throughout the statistical analysis of the results. The
questionnaire, which was identical for the second and third
study, contained a total of seven questions (14 sub
questions) and also used seven-grade bipolar scales (see
Table 2).
of presentation and no particular order of presentation
prevailed. Latin Squares [30] were used to conserve the
random order of presentation while narrowing the required
number of participants.
In the first study, one of the scale models was used as a
Reference Room and fitted with a double-pane window
with an iron free and a low-emissivity coated glass (glazing
“Ref77”, see Table 1). The other scale model, called the
“Test Room”, was alternately fitted with one of the other
glazing types (A79, B76, C70, D66 or E50, see Table 1).
The first study relied on paired comparisons such that
during each visit to the lab, the research participant was
first asked to look into the Reference Room and fill in the
questionnaire concerning the visual conditions in this room.
The participant was then asked to look into the Test Room
and fill in an identical questionnaire. S/he was allowed to
look back into the Reference Room and at the first
questionnaire to make sure that the evaluation of the Test
room was consistent with the previous one. Once this
second questionnaire was completed, the subject was asked
to leave the room. The researcher then went into the
laboratory, changed the glazing of the Test Room, and told
the subject to come back into the laboratory and evaluate
the conditions in the Test Room again, filling in a third
questionnaire, identical to the two previous ones. The
participant was never told that the glazing of the Test Room
had been changed and s/he could not see the researcher
change the glazing. The exact same procedure was repeated
each time the subject had to evaluate a new glazing type.
The evaluation of all glazing types was completed using
two sessions per subject, each session lasting about 40-45
minutes and covering only three glazing types (apart from
the Ref77 glazing). The subjects used three minutes to adapt
and the remaining time to fill the questionnaire. Prior to
these two sessions (sessions 2 and 3), each subject was also
invited to perform one whole session (session 1) to get
some training and make sure that there was no
misunderstanding in the questionnaire and procedure.
The glazing types were divided into two groups
corresponding to each session:
Fig. 5 Resulting CIE-L*a*b coordinates for (1) clear, (2)
partly sunny, (3) partly cloudy and (4) overcast sky.
Experimental design and research procedure
The three studies used a within-subject experimental design
so every research participant evaluated every glazing
situation. Each participant had a unique and random order
Session 2:
glazing types A79, C70, E50
Session 3:
glazing types B76, D66, E50
Glazing E50 was evaluated during each session, to check
that the ratings were consistent from one session to the
next. Moreover, each session included glazing types with a
high light transmittance (A79 and B76), an intermediate light
transmittance (C70 and D66) and a low light transmittance
(E50). A multivariate Wilk’s lambda statistic analysis was
carried out on the evaluation of glazing Ref77 and E50,
which showed no difference between sessions 2 and 3 for
both glazing types (Ref77: Wilk’s lambda = 0.426, F(9.9) =
1.345, p = 0.333 / Glazing E50 : Wilk’s lambda = 0.441,
F(9.9) = 1.270, p = 0.364). For two glazing types, the
average value of the factor scores for sessions 2 and 3 was
thus used for the rest of the analysis.
91
Table 2: Questionnaire filled by the research subjects
(translated from French) in the second and third studies
and associated factor.
simply closed the door giving access to the observation
hole. The time laps for the glazing switch was fairly rapid
since the participant could remain seated without being
aware of the change. The evaluation of the five glazing
types took around 15 to 20 minutes to complete in the third
study. The subjects used three minutes to adapt and one to
two minutes to fill the questionnaire.
Measurements
In the first study, the following quantities were recorded:
! the interior horizontal illuminance,
! the exterior global illuminance,
! the exterior vertical illuminance (on the north facade),
! the vertical spectral irradiance.
Specific details regarding these measurements are carefully
reported in the relevant paper [22].
In the second and third studies, the interior horizontal
illuminance was the only physical quantity recorded during
the experiments. Details regarding the illuminance
measurements are reported in the relevant papers [23, 24].
RESULTS
Danish study (first study)
In the second study, the large amount of glazing types
studied also required that the experimentation be conducted
in two sessions to avoid visual fatigue of the participants.
The validation of the results between session 1 and 2
required that glazing type B82 was evaluated each time.
Subsequently, the scores of the first session were used for
the analysis since a univariate repeated measure ANOVA
showed that the scores were statistically equivalent
between sessions 1 and 2 (F(1,6) = 0,415 and p = 0,867).
In the second study, the artificial sky was lit at least 30
minutes before the beginning of each experiment to
stabilize the light output from the fluorescent tubes. At the
beginning of each session, the participant entered the
laboratory and was instructed to sit in front of the
observation point from which s/he looked inside the scale
model and filled a questionnaire. Once the first
questionnaire was completed, the participant gave it to the
researcher and left the laboratory. In the meantime, the
researcher changed the glazing type. Once the second
glazing was in place, the participant was asked to come
back in the laboratory to repeat the exact same procedure.
These steps were repeated for each glazing type and each
laboratory session.
In the third study, a similar procedure was used as in the
second study except that once a first questionnaire was
completed, the participant gave it to the researcher and
92
The statistical analysis was performed using the SPSS 12.0
software. In the first (Danish) study, a statistical analysis
was carried out by identifying seven important factors to
which the research questions related (also drawn from the
results of a principal component analysis (PCA) and
confirmed by scree plot test):
! shadows (F1);
! brightness (F2);
! naturalness and colouring (F3);
! colour temperature (F4);
! beauty and pleasantness (F5);
! comfort (glare) (F6);
! sharpness (F7);
The average of scale scores for each factor was calculated
and the factors were classified with respect to their capacity
to explain the variance (from F1 to F7).
Subsequently, a randomized complete block design
ANOVA [31] was carried out on the average scores for
each factor with the different glazing types as the within
subject effect. This analysis showed that for all the factors
except F1 (shadows), the glazing type had a statistically
significant effect on the subjective scores (see [22] for
detailed results).
Table 3 Glazing type (Daylight Tr), mean, standard error,
and results for the specific orthogonal comparisons
(Dunnett) and multiple comparisons (LSD method adjusted
with Bonferroni), first (Danish) study.
The means obtained for each factor and each glazing were
also ordered with bipolar scales all presented as negativepositive (1–7) in the analysis (see Table 3). A rating of 7
corresponds to the highest (most positive) score; a rating of
1, to the lowest (most negative) score, and a rating of 4, to a
neutral score. This analysis indicated that glazing types of
higher transmittance generally obtained higher (more
positive) scores than the glazing types with lower
transmittance. However, there are two exceptions to this
trend. For F1 (shadows), the means were almost constant
thus indicating a weak effect of glazing type on the
perception of shadows (also showed by the statistical
analysis). For F6 (comfort glare), the means (Table 3) are
higher for lower transmittance glazing types, indicating that
lower transmittance glazing types resulted in less glare.
To improve clarity in the results, a first approach in the
statistical analysis consisted of examining five planned
specific orthogonal comparisons (single degree of freedom
tests) so the average ratings for glazing types A79– E50 were
contrasted with the results obtained for the reference
glazing, which corresponded to the way subjects arrived at
their ratings. The results of this analysis are presented in
Table 3 (“Dunnett”) and discussed in detail in [22]. The
main conclusions from this analysis are summarised below:
! The glazing type did not have any statistically significant
effect on the ratings for questions related to shadows
(F1).
! In terms of brightness (F2), all glazing types resulted in a
statistically significant difference compared to the
reference glazing.
! In terms of naturalness and colouring (F3) and beauty and
pleasantness (F5), glazing types A79– B76 did not result in
a statistically significant difference compared to the
reference glazing, but glazing types C70–D66–E50
produced a statistically significant difference with respect
to the reference glazing.
! In terms of colour temperature (F4), glazing B76 was the
only one rated as statistically different compared to the
reference glazing.
! In terms of comfort (glare) (F6) and sharpness (F7), all
glazing types except A79 resulted in statistically
significant differences compared to the reference glazing.
Following finding a significant effect of glazing types, a
second approach in the statistical analysis was explored
which consisted of performing multiple comparisons with
the protected LSD method adjusted with Bonferroni
correction to control the type I error rate. Detailed
discussion of these results are presented in [22]. The LSD
tests indicated, among other things, that for all factors
studied, except brightness (F2), there was no statistically
significant difference between glazing types Ref77, A79 and
B76 (see Table 3, last column to the right).
Second study (artificial sky, Canada)
In the second study, the factors were slightly modified from
the first questionnaire in the Danish study after we realised
93
that questions regarding colour temperature were difficult
to understand for research subjects who did not really
understand the term “colour temperature” (in French
“temperature de couleur” is familiar only to lighting
researchers). Moreover, it appeared that this factor was
already covered by the factor “naturalness”. In addition, a
factor called “distribution” was added since this parameter
is often discussed in the literature about lighting. The
factors were ordered as follows according to their capacity
to explain the variance:
shadows. We should however emphasize that the difference
between both analyses (ANOVA and correlation between
scores and visual transmittance) may be related to the low
participation rate. The negative correlation for F2 (glare
comfort) (rF2= – 0,185) suggests that a higher visual
transmittance yields more glare. Although this is consistent
with the previous (Danish) study, the low correlation and
absence of statistically significant effect (ANOVA) reduces
the overall importance of this finding in this case.
!
!
!
!
!
!
!
Third study (natural sky, Canada)
brightness (F1);
glare comfort (F2);
naturalness (F3);
beauty pleasantness (F4);
precision (F5);
distribution (F6);
shadows (F7).
A randomized complete block design ANOVA [31] was
performed using the SPSS 12.0 software. This analysis
allowed identifying the factors for which there were
statistically significant differences between glazing types (p
< 0,05). According to this analysis (outlined in Table 4),
the glazing type influenced the perception of brightness
(F1), beauty pleasantness (F4) and precision (F5) but had
no statistically significant effect on glare comfort (F2:
p=0,280), naturalness (F3: p=0,920), distribution (F6:
p=0,460) and shadows (F7: p=0,056). The results are thus
consistent with the previous (Danish) study concerning the
perception of brightness (F1), beauty pleasantness (F4),
precision (F5) and shadows (F7). Note that the absence of
statistically significant effect of glazing type on F2, F3, F6
and F7 may be explained by the low participation rate.
A multiple comparison ANOVA (Tukey’s test) was also
performed in order to compare each pair of glazing type
according to each factor for which a statistical difference
was revealed. A detailed discussion of this analysis is
presented in the relevant paper [23].
Of interest for the present paper is the study of the relation
between the average scores for each factor and the glazing
visual transmittance. Positive correlations between the
glazing visual transmittance and scores were obtained for
all factors except F2 (glare comfort). These results are
generally consistent with results of the previous (Danish)
study. The correlation with factor F1 (brightness) is very
strong (rF1=0,955). This result was expected because
transmittance actually corresponds to the quantity of light
transmitted. However, it is surprising to obtain that this
particular correlation is weaker than the correlations for
beauty pleasantness (rF4=0,972) and precision (rF5=0,986).
Although the ANOVA did not reveal any statistically
significant differences among glazing types for F7
(shadows, p=0,056), the correlation between F7 and the
visual transmittance was fairly high (rF7=0,829) indicating
that higher transmittances result in a superior perception of
94
In the third study, a randomized complete block design
ANOVA [31] was also performed using SPSS 12.0. This
analysis allowed identifying the factors for which there
were statistically significant differences between glazing
types (p < 0,05, Table 5). According to this analysis, the
glazing type influenced the perception of brightness (F1),
naturalness (F3), beauty pleasantness (F4) and precision
(F5) but had no statistically significant effect on glare
comfort (F2: p=0,580), distribution (F6: p=0,316) and
shadows (F7: p=0,050). Except for the factor naturalness,
these results are consistent with the results of the second
study [23]. The only difference between these new results
and the results obtained in the artificial sky study [23]
concerns the perception of naturalness (F3). In the artificial
sky study, the ANOVA did not indicate that the glazing
type had a statistically significant effect on the perception
of naturalness. It is possible that this is attributable to the
use of an artificial sky, which made the room look rather
artificial. Furthermore, the present study also identified the
perception of shadows as a qualitative factor not influenced
by glazing type, repeating results of the two previous
studies [22, 23].
To elaborate on these results, a multiple comparison
ANOVA (Tukey’s test) was performed in order to compare
each pair of glazing type according to each factor for which
a statistical difference was revealed. This analysis is
presented and discussed in detail in the relevant paper [24].
Subsequently, an analysis of correlation between average
scores for each factor and the glazing visual transmittance
was also performed. This analysis revealed that the only
negative correlation was obtained for glare comfort (F2),
which means that a higher visual transmittance yields more
glare. This result is consistent with the two previous
studies. On the other hand, the positive correlations indicate
that a higher transmittance glazing yields more positive
ratings for brightness, naturalness, beauty pleasantness,
precision, light distribution and shadows. Five correlations
(F1, F2, F3, F4 and F5) were found to be very high
(r>|0,95|). The only low correlation concerns the perception
of distribution (rF6= 0,347).
We should however emphasize that the difference between
the ANOVA and the correlation study may be related to the
low participation rate, especially in the second study. The
negative correlation for F2 (rF2=-0,996) suggests that higher
visual transmittance glazing type yields more glare.
Although this is consistent with earlier research [22, 23],
the absence of statistically significant effect (ANOVA)
reduces the overall solidity of this finding.
Table 4 Results of the second study. The ANOVA indicates the statistical significance of a perceptible difference between
glazing types for values of p < 0,05. Correlations between the glazing visual transmittance and the seven qualitative factors
indicate the direction and strength of the relationship between the variables.
F1
Brightness
F2
Glare
comfort
F3
Naturalness
F4
Beauty
pleasantness
F5
Precision
F6
Distribution
F7
Shadows
ANOVA
p < 0,001
p = 0,280
p = 0,920
p < 0,001
p < 0,001
p = 0,460
p = 0,056
Correlation
r = 0,955
r = – 0,185
r = 0,373
r = 0,972
r = 0,986
r = 0,586
r = 0,829
Table 5 Results of the third study. The ANOVA indicates the statistical significance of a perceptible difference between
glazing types for values of p<0,05. Correlations between the glazing visual transmittance and the seven qualitative factors
indicate the direction and strength of the relationship between the variables.
F1
Brightness
F2
Glare
comfort
F3
Naturalness
F4
Beauty
pleasantness
F5
Precision
F6
Distribution
F7
Shadows
ANOVA
p < 0,001
p = 0,580
p < 0,001
p < 0,001
p < 0,001
p = 0,316
p = 0,050
Correlation
r = 0,985
r = – 0,996
r = 0,984
r = 0,976
r = 0,985
r = 0,347
r = 0,596
DISCUSSION AND CONCLUSION
Three studies of the effect of window glazing types on
daylight quality were presented in this paper. The first
study was carried out at the Danish Building Research
Institute in Hørsholm, Denmark. The second and third
studies were achieved at the École d’architecture of
Université Laval, Québec, Canada. All three studies used
scale models and within-subject experimental design. The
first study used a paired comparison with two identical
1:7,5 scale models of an unfurnished rectangular room with
a single, north oriented window exposed to overcast skies.
A total of 18 research participants evaluated daylight
conditions in the two rooms, by looking straight ahead
towards the window through an observation hole placed at
the back of the room. The second study used a single 1:6
scale model of a rectangular room, which was fully
furnished as a typical residential living room and provided
with different glazing types presented in random order.
This room had a single window facing an artificial,
overcast sky. The third study used the same scale model as
the second study but the model was exposed to the natural
sky with a sunlight free, south-east oriented window. A
total of 15 and 30 research participants took part in the
second and third study respectively. In both studies, the
participants evaluated the light conditions in the rooms
(after three minutes of adaptation) from an observation hole
located in one of the lateral walls of the model. Their view
of the interior thus had a diagonal direction with respect to
the window.
This paper focuses on the constancy in the results obtained
in the three studies, which were achieved using different
experimental designs. In all three studies, the results
indicated the following:
The glazing visual transmittance was positively correlated
with brightness, naturalness, beauty pleasantness, and
precision or sharpness. Previously, Cuttle [20] found that
the acceptability of an office can be increased by the use of
reduced transmittance glazing, which is in contradiction
with the findings of our three studies. In general we found
that increasing the glazing visual transmittance produced
higher scores for beauty pleasantness, naturalness,
precision, brightness. However, in our three studies, all
glazing types had a dominantly green or green-yellow shift
in colour. We did not have any bronze glazing (with a
warm shift). Cuttle [20] found that there is a preference for
a colour effect that gives a warm shift and Boyce et al [15,
16] indicated that solar bronze glass (warm shift) had been
found to give an enhanced perception of the same
transmittance. One explanation for these confounding
results may be the fact that a warm shift is preferred
because it gives an impression of sunshine (sunny day)
even on overcast days while dominantly greenish shifts
tend to flatten out the colours in the scene, as shown by
Bülow-Hübe’s experiment [19]. Since the human spectral
sensitivity curve (v(!)) peaks in the green-yellow region
(550 nm) in the photopic state, it may be that more stimulus
is needed in the other colour bands (e.g. red) in order to
produce a colourful, lively image. Also, Hårleman et al.
95
[32] pointed out that reddish colours are associated with
human skin, facial colour, strong emotional expressions
such as affection and defiance and other mental
characteristics.
The glazing visual transmittance was negatively correlated
with glare comfort. In other words, glazing types with
higher visual transmittance created more glaring views of
the interior. The difference between the glazing types was
statistically significant only in the first study but the
participants looked at the window directly, which may have
emphasized the difference between the participants’ ratings
in this case. These results show that glazing with lower
transmittance may contribute to reduce glare to some
extent. However, the same glazing types also create
interiors that are darker, less beautiful, pleasant, natural and
precise. This is very interesting because it means that
someone may experience an interior as more beautiful,
pleasant, natural and precise, although more glaring. In the
recent years, lighting research has focused on glare as the
single most important parameter of lighting quality in
interiors. The results of these three studies emphasize the
importance of considering glare in parallel with other
parameters.
Finally, the three studies indicated that the glazing type did
not have a statistically significant effect on the perception
of shadows. In the last studies, the perception of shadows
was positively correlated with the glazing visual
transmittance but the difference between the glazing types
was not statistically significant according to the ANOVA.
We did not obtain statistically significant differences
between glazing types for a factor called distribution either.
This can be explained by the fact that the glazing type do
not affect contrasts between different surfaces, it reduces
the luminance of both bright and dark surfaces
proportionally, with no resulting effect on shadows or
distribution.
Our future research in this field will investigate the effect
of colour shifts, including warm shifts, for glazing types of
constant visual transmittance. We also plan to include
energy and photobiology related effects, and to take into
consideration the use of electric lighting together with
daylighting for various compass orientations and types of
skies. The use of longer adaptation times and full-scale
experiments are also two important aspects that should be
improved in this research on window glazing types.
ACKNOWLEDGEMENTS
This paper is dedicated to Rikard Küller, who got me
interested in this type of research in the first place and who
was a passionate and inspiring professor. I also express my
gratitude to Jens Christoffersen, from the Danish Building
Research Institute and Gaétan Daigle from the Département
de mathématiques et de statistique of Université Laval, who
took time to read the final version of this paper. I thank the
Danish Energy Agency for funding the first study and the
Social Sciences and Humanities Research Council of
96
Canada (SSHRC) for providing extra support for the
statistical analyses of all three studies. I also sincerely
thank architects Richard Lafontaine and Richard Langford
for providing funding for the experimental setting of the
second and third studies, local glazing manufacturers
Multiver, Robover and Thermos de la Capital for providing
free glazing samples. I thank Gaétan Daigle, who wisely
supported the statistical analysis of all three studies. I thank
every participant in the three studies for their time and
involvement. Finally, last but not least, I thank my Master’s
student Nathalie Pineault for her great involvement in the
second and third studies and Professor Claude MH Demers
from the École d’architecture of Université Laval, who
provided support for the digital photo analysis.
REFERENCES
[1] Baker, N. We are all outdoor animals. Proc. of PLEA
2000: Architecture City Environment. (London, James &
James, 2000), 553-555.
[2] Jenkins, P. L., Phillips, T. J., Mulberg, E. J. & Hui, S.
P. Activity patterns of Californians: Use of and proximity
to indoor pollutant sources. Atmos. Environ. 26A (1992),
2141-2148.
[3] Christoffersen, J., Petersen, E., Johnsen, K., Valbjørn,
O., & Hygge, S. Vinduer og dagslys - en feltundersøgelse i
kontorbygninger (Windows and Daylight - a Postoccupancy Evaluation of Offices). SBI-rapport 318 (1999).
Statens Byggeforskningsinstitut. Hørsholm (Denmark), 71
pages (in Danish).
[4] Bülow-Hübe, H. Energy-efficient window systems:
effects on energy use and daylight in buildings. PhD thesis.
Lund University, 2001.
[5] Chain, C., Dumortier, D. & Fontoynont, M.
Consideration of daylight’s colour. Energy and Buildings.
33 (2001), 193-198.
[6] Brainard, G.C., Hanifin, J.P., Rollag, M.D., Greeson,
J.M., Byrne, B., Glickman, G., Gerner, E. & Sanford, B.
Human melatonin regulation is not mediated by the three
cone photopic visual system. J Clin Endocrinol Metab; 86,
1 (2001), 433-436.
[7] Thapan, .K, Arendt, J. & Skene, D. J. An action
spectrum for melatonin suppression: evidence for a novel
non-rod, non-cone photoreceptor system in humans. J
Physiol. 535 (2001), 261-267.
[8] Rautkylä, E., Tetri, E., Elholma, M., Halonen, L. Effects
of the correlated colour temperature of light on alertness –
field studies in lecture environment. Report 46 (2008).
Helsinki University of Technology, 40 p.
[9] Veitch, J.A., Tosco, A.-M., Arsenault, C.D.
Photometric issues in healthy lighting research and
application. Proc. of the CIE Symposium '04 – Light and
Health: Non-Visual Effects. Vienna, Austria, Sept. 30-Oct.
2 (2004), 146-149.
[10] Knez, I. Effects of indoor lighting on mood and
cognition. J of Env Psychol, 15 (1995), 39-51.
[11] Knez, I. & Enmarker, I. Effects of office lighting on
mood and cognitive performance and a gender effect in
work-related judgment. Environment and Behavior, 30
(1998), 553-567.
[12] Begemann, S.H.A., van den Beld, G., Tenner, J.A.D.
Daylight, artificial light and people in an office
environment, overview of visual and biological responses.
Industrial Ergonomics, 20 (1997), 231–239.
[13] Kruithof, A. A. Tubular Luminescence Lamps for
General Illumination, Philips Technical Review, vol.6
(1941), 65-96.
[14] Fonseca, I.C.L. Quality of light and its impact on
man’s health, mood and behaviour. Proc of the PLEA 2002
(Toulouse, July 2002).
[15] Boyce, P. R., Beckstead, J. W., Gutkowski, J. M., Fan,
J. & Strobel, R. W. Brightness- enhancing glazing:
perception, performance and energy. Report from MSR
Scientific to Public Works Canada, Ottawa, Public Works
Canada, 1992.
[16] Boyce, P. R., Gutkowski, J. M., Beckstead, J. W. &
Rea, M. Brightness-enhancing glazing. Report from MSR
Scientific to Public Works Canada, Ottawa, Public Works
Canada, 1991.
[17] Chain, C. Caractérisation spectrale et directionnelle
de la lumière naturelle : application à l’éclairage des
bâtiments. Doctoral thesis. Institut National des sciences
appliquées de Lyon, France, 2004.
[18] Cooper, J. R., Wilshire, T. J. & Hardy, A. C. Attitudes
towards the use of heat rejecting/low light transmission
glasses in office buildings. Proc. of the CIE TC-42
Daylight Symposium, Architectural Design Proc, Istanbul,
1973.
[19] Bülow-Hübe, H. Subjective reactions to daylight in
rooms: effect of using low-emittance coatings on windows.
Lighting Res Technol, 27, 1 (1995), 37-44.
[20] Cuttle, K. Subjective assessments of the appearance
of special performance glazing in offices. Lighting Res and
Technol, 11, 3 (1979), 140-149.
[21] Boyce, P. R., Eklund, N., Magnum, S., Saalfield, C. &
Tang, L. Minimum acceptable transmittance of glazing.
Lighting Res Technol, 27, 3 (1995), 145-152.
[22] Dubois, M.-C., Cantin, F. & Johnsen, K. The effect of
coated glazing on visual perception: A pilot study using
scale models. Lighting Res and Technol, 39, 3 (2007), 283–
304.
[23] Pineault, N., Dubois, M.-C., Demers, C.M.H., Briche,
M. Effect of coated and tinted glazing on daylight quality
(...). Proc. of Challenging Glass Conf., TU Delft, Holland,
22 May 2008.
[24] Pineault, N. & Dubois, M.-C. Effect of Window
Glazing Type on Daylight Quality: Scale Model Study of a
Living Room under Natural Sky. LEUKOS: J of the Illumin
Eng Soc of North America, 5, 2 (2008).
[25] Cowdroy, R. A Study of the scalar effect in the use of
models for the assessment of glare, M. Blg. Sc. Thesis of
the Macquaire University, Sydney, Australia, 1972.
[26] Lau, J. Use of scale models for appraising lighting
quality. Lighting Res and Technol, 4, 4 (1972), 254-262.
[27] Liljefors, A., Ejhed, J. Bättre belysning [in English;
Improved lighting]. Report T17. Stockholm: Statens råd för
byggnadsforskning, 1990.
[28] Jolliffe, I. T. Discarding variables in a principal
component analysis I: Artificial data. Applied Statistics 21
(1972), 160-173.
[29] Cattell, R. B. The scree test for the number of factors.
Multivariate Behavioral Research, 1(1966), 245-276.
[30] Keppel, G. Design and analysis: a researcher’s
handbook (2nd edition). Englewood Cliffs (NJ), PrenticeHall, 1982.
[31] Crowder, M. J., Hand, D. J. Analysis of Repeated
Measures. Monographs on Statistics and Applied
Probability. New York, Chapman and Hall, 1990(41).
[32] Hårleman, M., Werner, I.-B. & Billger, M.
Significance of Colour on Room Character: Study on
Dominantly Reddish and Greenish Colours in North- and
South-Facing Rooms. J Colour Design and Creativity, 1, 1,
9 (2007), 1-15.
97
Effect of LED-based Study-Lamp on Visual Functions
Srinivasa Varadharajan, Krithica Srinivasan, Siddhart Srivatsav, Anju Cherian, Shailaja Police,
Ramani Krishna Kumar
Elite School of Optometry, Unit of Medical Research Foundation
8, G. S. T. Road, St. Thomas Mount
Chennai – 600 016, INDIA
+91 44 2232 1835
drlsv@snmail.org
ABSTRACT
Changes in visual functions following near vision tasks
under lighting provided by an LED-based study lamp were
analysed. Visual performance and basal tear production
before and after reading and painting tasks were assessed in
the light provided by an LED and a CFL based study lamps
on thirty volunteers with normal vision. Measurements
were made for each light with room lights on and off.
Visual comfort was assessed using a questionnaire.
Statistically significant but clinically insignificant changes
were seen only in basal tear production in three conditions.
Unexplainable changes were seen in the near visual acuity
for two contrast levels in certain conditions. No other
parameters showed any significant change in any condition.
Keywords
LED lamp, visual functions, Munsell chips, Near vision
tasks
INTRODUCTION
Reading is a complex visual process involving visual and
environmental variables [19, 9]. The predominant factors
that influence reading performance are luminance [13],
uniformity of illumination, contrast of the task [8]. Color of
the source and/or the target does not affect performance
[10, 11, 5]. Berman et al studied the effect of lighting color
temperature and luminance on near visual acuity in children
and found that higher the color temperature the better the
acuity and that lower the luminance the lower the acuity at
higher color temperatures [2].
Reading speed and critical print size at which the subject
has the maximum reading speed are usually measured with
MNRead acuity charts [17]. Reading performance can be
improved in illumination levels of 100-300 lux. AgeRelated Macular Degeneration (ARMD – an ocular
condition which affects the central part of the retina called
macula that aids in fine vision) patients are known to prefer
yellow filters to improve their reading speed [5]. The
reading rates for normally sighted subjects are greatest for a
range of intermediate character sizes ranging from 0.3
degree to two degree. Reading speed declines for characters
smaller than 0.13 degrees and characters larger than 4
degrees [1].
Traditional incandescent lamps use high amount of energy
to produce standard amounts of indoor lighting and also
sodium light is known to cause visual fatigue [3] after
prolonged reading [12]. Fluorescent (FL), compact
sources use progressively less amounts of energy to
produce the same amount of light [15]. Since LEDs are low
energy but directional sources, the visual performance
under these light sources could be different. Our aim was to
estimate the efficacy of LED lamp for continuous and/or
demanding near vision tasks. Therefore we compared the
effect of LED based reading lamp and CFL on various
visual tasks and also estimated the visual comfort.
METHODS
The study adhered to the tenets of Declaration of Helsinki
and was approved by the institutional review board (IRB).
Signed informed consent was obtained from all subjects.
All subjects underwent complete optometric and orthoptic
evaluations [4]. These included determination of monocular
visual acuity (resolving ability) for distant and near targets,
refractive error, action of the eye muscles, alignment of the
two eyes (phoria status), ability and speed of shifting gaze
from distant to near targets (accommodation amplitude and
facility), ability of the two eye to work together for near
objects (convergence). In addition, their color vision,
stereopsis (ability to perceive depth using the two eyes) and
basal tear production were tested. Screening for color
vision was done using Ishihara pseudo-isochromatic plates,
stereopsis using Wirt circles and basal tear production
using Schirmer’s test II. Only subjects who met our
inclusion criteria were included. The inclusion criteria
were:
• Age: 13 – 25 years
• Read and write English at 8 grade level
• Best corrected distance visual acuity – equal to or better
than 6/6
• Best corrected near visual acuity – equal to or better
than N6
• Near point of accommodation as per Hoffstetter’s
average formula [6, p70]
• Accommodative facility better than or equal to 10
cycles per minute using ±1.75D flippers
• Near point of convergence ! 10 cm
• Distance and near Phoria as per Morgan’s values [14]
• Basal tear production using Schirmer’s Test II " 10mm
• Stereopsis using Wirt circles - 40 arc sec
• Normal findings in the anterior and posterior segment
evaluations
Those who had the following were excluded:
• Severe dry eyes (< 10mm wetting length in Schirmer’s
test II)
• Overaction/underaction of any extraocular muscle
• Any ocular pathologies/diseases.
Study Lamps:
The LED based lamp consisted of an array of 24 white
LED-s spaced equally on the circumference of a circle of
diameter 15 cm (Fig 1a). Figure 2 displays the
manufacturer supplied power spectrum of the LEDs used in
the lamp. The CFL lamp consisted of a single circular CFL
source of the same radius (Fig 1b). We were not able to get
the power spectrum of the CFL from the manufacturers nor
did we have the facility to measure the same. However,
spectral power distribution of common fluorescent light
sources could be easily found on the internet [20]. Figure 3
shows the primary and secondary task areas as defined in
the study. Uniformity index was calculated as the ratio of
the illuminance of the light falling at the boundary between
the primary and secondary task area and the illuminance at
the center of the primary task area.
m. Since all subjects who participated in the study were
right handed, the study lamp was placed on the left side of
the table so that light from the lamp illuminated the center
of the table. The subjects were allowed to adjust the
position of the lamp. The subjects were instructed to keep
the task materials where the maximum light was falling on
the table, i.e., on the primary task area. A video camera
focused on the face of the subject was placed without
obstructing the light falling on the task area. Two
fluorescent lamps fitted on the ceiling directly above the
reading table provided illumination of approximately 200
lux on the table.
Figure 3: Primary and secondary task areas defined in the
study. The shaded portion is the primary task area and the
non-shaded portion is the secondary task area. The primary
task area measured 1.25 ft x 1.25 ft and the secondary task
area measured 3ft (length) x 2 feet (depth). Area outside the
secondary task area is known as the tertiary task area and it
is not depicted in the figure.
Experiment:
(a)
(b)
Figure 1: (a): The LED based light study lamp; (b): CFL
study lamp. For description refer text.
In an attempt to study the interaction of the study light with
the environmental lighting, the experiments were done
under four different lighting conditions as shown in table 1.
Table 1: Definition of the four conditions used in this study
Condition name*
Room lights
Lamp used
I
On
CFL lamp
II
Off
CFL lamp
III
On
LED lamp
IV
Off
LED lamp
*Conditions II and IV were called “Dark Conditions” since the
room lights were off. Similarly, conditions I and III were called
“Light Conditions”.
Figure 2: The dark continuous line in the upper figure
denotes the relative spectral power distribution of the LED
used in the LED based study lamp. The dashed curve
denote the human photopic sensitivity function, commonly
known as the V(!) curve. The graph was supplied by the
manufacturer of the LEDs.
Study area:
A standard study table and chair was placed in the middle
of a windowless room that measured 4.2 m x 4.2 m x 3.1
During each condition, the same set of experimental
procedures was performed. The procedures were done in
the following order: (i) ten minutes of adaptation to the
lighting condition – the standard and LED lamps were kept
on and only the room lights were switched either on or off,
(ii) evaluation of basal tear function using Schirmer’s strip,
(iii) achromatic point estimation using Munsell chips, (iv)
Near visual acuity at various contrast levels using a
Landolt-C based near vision chart, (v) stereopsis estimation
based on Wirt circles, (vi) reading speed measurement
using variations of MNREAD chart (which we named
SNREAD, to avoid confusion with MNREAD), (vii)
reading task for ten minutes, (viii) coloring task for ten
minutes, (ix) procedures (ii), (iv) and (v) mentioned above
(post-task measurements), and (x) administration of a five-
point Likert scale questionnaire. Procedures (vi), (vii) and
(viii) (i.e., reading speed measurement with SNREAD
charts, reading and paining tasks) were video recorded to
extract the reading speed, critical print size and blink rate.
Basal Tear Production:
Basal tear function is a measure of normal production of
tears and hence is also a measure of dry eyes. It is usually
quantified using Schirmer’s test II. This test uses a thin
strip of Whatman filter paper #40 called the Schirmer’s
strip. The Schirmer’s strip is 5mm x 35mm in dimension
and has graduations along its length at every millimeter.
The subject’s eye is anesthetized using a single drop of
proparacaine 0.5%. The Schirmer’s tear strip is inserted
into the temporal part of the lower cul-de-sac (the area
under the lower eye lid) in both the eyes. The strip remains
in the eye for 5 minutes. Due to capillary action, the tear
from the eye wets the Schirmer’s strip. The wetting length
at the end of 5 minutes is noted. If the wetting length is 15
mm or more, the tear production is considered as normal.
Wetting lengths less than 10 mm are considered indicative
of severe dry eyes. Basal tear production was measured
using Schirmer’s test II before and after the reading and
painting tasks in each condition. The Schirmer’s test II is
conventionally done only with room lights turned off. But
in our experiment it was done under the lighting provided
for each condition to study the effect of the light on tear
production.
Achromatic Point Estimation:
Achromatic setting was measured using 40 plates of
Munsell chips. Achromatic point as defined by Werner et al
(1993) is “Typically called the white point, … more
accurately called the achromatic point, as it may appear
dark gray, light gray or white, depending upon its
luminance and surrounding conditions of illumination”
[18]. Each plate consisted of 7 chips that varied from one
hue to its opponent hue and arranged randomly on the plate.
Of the 7 chips, one would be achromatic. The task would
be to identify the chip that looks “hueless” or “colourless”
or “the chip that is devoid of the hues in the opponent axes
of that particular plate”. A practice session was given using
few randomly chosen plates. The response was recorded in
the scoring sheet that accompanies the Munsell chips. Each
chip has a score attached to it ranging from -3 to +3 with 0
denoting the achromatic point and values closer to zero
denoting chromaticities closer to the achromatic point on
that axis. For our experiment, we only noted the number of
errors made in the 40 plates irrespective of the direction on
error. We did this because we were interested how the
different lighting conditions affected this task.
Near Visual Acuity at Various Contrast Levels:
Near vision acuity was measured using a variation of the
VALiD kit [16]. To avoid confusion with VALiD kit we
called our chart the SVIS chart (Fig 4). The SVIS chart was
designed for use at 40 cm. The chart was constructed using
the Landolt-C optotypes facing up, down, right and left.
The chart contained ten sets of three rows of C-s.
Orientations of C-s were randomized using the
pseudorandom number generator in Microsoft Excel. Each
row contained C-s of various sizes that decreased from 1.0
logMAR to -0.3 logMAR in steps of 0.1 logMAR. Each set
of C-s had a fixed contrast value. The contrast decreased
from 100% to 4% in steps of 0.15 log units down the chart.
The chart was placed in the primary task area such that the
light from lamp under consideration fell on the chart. The
subject was instructed to speak aloud the orientation of the
C from the top-most line. At any contrast level, the acuity
will be the smallest size of C that was correctly identified.
Each subject was asked to read only one of the three lines
at each contrast level. For each contrast level, the visual
acuity was thus noted. We use the term visual acuity to
mean visual acuity at 100% contrast. For all other contrasts,
we mention the contrast value.
Stereopsis:
Stereopsis is the ability to perceive depth using the two
eyes together. We measured stereopsis using Wirt circles
illuminated by the lighting of the given condition. In this
procedure the subject will be asked to wear a polarizing
spectacle and asked to view a polarizing sheet. The
polarizing sheet contains groups of four circles. In each
group one circle will appear to float above the rest at some
distance. The subject’s task is to point out the floating
circle. This distance is given in terms of what is called the
retinal disparity measured in arc seconds. Because of the
laterality of the two eyes, the image on the retinae of two
eyes will be slightly laterally displayed. This is known as
retinal disparity [7]. Wirt circles are useful for measuring
stereopsis from 800 arc seconds and 40 arc seconds.
Reading Speed Estimation:
Reading speed was calculated using the SNREAD chart.
SNREAD is a variation of the MNREAD near vision
reading chart that contains eleven lines of continuous text.
Each line has 60 characters and the size of the lines
decreased down the chart. There are two versions of the
chart that are available. We constructed 12 versions of the
chart. These charts were called SNREAD chart. The
SNREAD charts had the same construction design as the
MNREAD chart, but the sentences in these charts were
different. The sentences used in these charts were selected
from books recommended for 8th grade students.
Essentially designed for use at 40 cm, the chart was placed
in the primary task area illuminated using the lamp. The
same version of the chart was not given to a subject more
than once. The subjects were asked to read the chart aloud
clearly with minimum mistakes. Video recording of the
procedures was started at this point. Reading errors and
reading time was calculated from the recording. The lines
in the chart vary in size from 4.0M to 0.4M. M notation is a
metric measure of the Visual Acuity. Each mm of letter
height is set equal to 0.7M. The measurement is done with
lower case letters without any ascending or descending
limb, such as e, o and c. If the visual acuity is 1M it means
that the letter subtends 5 arc minutes at a distance of 1 m.
Reading Task:
The subject was given a reading task for ten minutes. The
reading material was kept at their habitual working
distance. The subjects were instructed to read at their usual
reading speed. The text in the reading material was printed
in 8 point Times New Roman font with 1.5 line spacing.
The contents of the reading task varied across experimental
conditions. All reading materials had a side box that
highlighted the salient point of the material. This highlight
was printed in 10 pt Times New Roman.
Colouring Task:
A set of drawings were chosen from a collection of
colouring book. One of the investigators coloured a
randomly chosen drawing with crayons and the subject was
asked to colour another copy of the same drawing with this
as the template. The crayon set that was used had totally 36
shades. Hue variations were quite small and not easy to
make out. For example, shades in green varied as “Olive
green”, “Emerald green”, “Deep green”, “Virdian hue”,
“Light green” and “green”. Subjective responses for the
shades that were difficult to match were noted. The same
crayon set that was used by the investigator to colour the
template was given to the subjects for the colouring task
too. This was done for ten minutes. If the subject finished
the task within ten minutes, a new drawing was given to be
completed in the remaining time.
Blink Rate:
Blink rate was calculated from the video recording when
the reading and coloring tasks were in progress. The total
number of blinks over the period of 20 minutes was
determined and from that the number of blinks per minute
was calculated.
Questionnaire:
The questionnaire aimed to assess visual discomfort during
various tasks. At the end of each experimental condition, a
5-point Likert scale questionnaire was given to the subjects
to fill out. This questionnaire had 14 questions six of which
dealt with visual comfort (such as glare, eye strain, dry
eyes, eye fatigue, eye pain and headache) and the remaining
eight were fillers (such as hunger, back ache, anxiety,
questions from the text given for reading and painting, etc).
From the responses, the visual discomfort score was
calculated [3].
Other Procedures:
The order of the conditions was randomized for each
subject. Different versions of the same chart were used for
different experimental conditions for both reading speed
and visual acuity measurements. The contents of the
reading task and objects for the coloring task were varied
across experimental conditions. Not more than three
sessions per day were done for each subject. Minimum of
half-hour breaks were given between conditions.
Analysis:
Changes in visual functions in a single condition before and
after the reading and painting tasks were analysed. These
are variously called “within condition changes” or “prepost changes” or just “changes”. Differences in these
changes across lighting conditions were also analyzed.
Unless stated otherwise, all comparisons were done using
Wilcoxon signed-rank test. Results were considered
significant when p < 0.05. All analysis was done using
SPSS 15, MATLAB 7.2 and Microsoft Excel.
Figure 4: SVIS chart. The contrast varies after every three rows. All the letters in a given triplet have the same contrast. The
contrast values were: A-100%; B-71%; C-50%; D-35%; E-25%; F-18%; G-13%; H-9%; I-6%; J-4%. The acuity level varies
along the row of every line in every triplet in steps of 0.1 logMAR starting from 1.0 logMAR to -0.3 logMAR. The chart is
designed for use at 40 cm reading distance
RESULTS
The illuminance values due to the LED lamp alone in the
primary task area for various subjects were around 200
lux and with the CFL lamp the value was around 500 lux.
The uniformity index was found to be 0.73 for the LED
lamp and 0.50 for the CFL lamp.
Thirty subjects participated in the experiments. The
number of subjects, however, for following variables was
reduced as given in parenthesis: Visual Discomfort Score
(29); Blink Rate (20); Reading Speed (26); Critical Print
Size (26). The reduction in numbers was due to either
incomplete response or failure of video recording. All
subjects were college students, doing their undergraduate
or postgraduate studies. The age of the students ranged
from 18 to 23.5 years. There were 25 female subjects and
5 male subjects who participated in the study.
Changes Within a Condition:
All subjects had zero change in stereopsis in all the four
conditions and hence we did not do any statistical analysis
on this parameter.
Differences Across Conditions:
Comparison of changes in conditions I and III was done
to study the behavior of LED lamp as compared to the
CFL lamp in a bright environment and between II and IV
to study the same in a dark environment. Analysis
between conditions I and II was done to understand the
effect of the room lighting on the CFL lamp; similarly,
comparison of changes in conditions III and IV was done
to find the effect of external illumination on the LED
lamp.
Table 2: Changes in near visual acuity across four
conditions
Contrast
(%)
Basal Tear Production:
The mean changes in tear production in various
conditions are shown in fig 5. Clinically, changes in
Schirmer’s test are said to be significant when the
difference between two readings is 5mm or more. In
condition I, the change was found to be statistically
insignificant (mean change; 0.65 mm; p=0.45). In
condition II (mean change = 1.75 mm; p=0.01), III (mean
change = 2.13 mm; p=0.01) and IV (mean change = 2.12
mm; p=0.01) though statistically significant changes were
found, these changes were clinically insignificant. The
maximal mean change was 2.13 mm in the third
condition. The median changes in all these four conditions
were 0 mm.
100
71
50
35
25
18
Figure 5: Mean change in basal tear production in the
four conditions. The boxes denote the mean values and
the lines denote ± 1 std error of means.
Near Visual Acuity at Various Contrast Levels:
The changes in the near visual acuity at various contrast
levels for the four conditions are shown in table 2. As can
be seen, most changes were statistically insignificant.
Only the changes for condition IV (i.e., LED lamp on
with the room lights turned off) at contrast values of 13%
and 9% were statistically significant. Since there is no a
priori reason why only these should be statistically
significant changes, we propose that these changes are
spurious in nature.
Stereopsis:
102
13
9
6
4
Condition
Mean
Median
I
II
III
IV
I
II
III
IV
I
II
III
IV
I
II
III
IV
I
II
III
IV
I
II
III
IV
I
II
III
IV
I
II
III
IV
I
II
III
IV
I
-0.04
0.02
0.02
0.03
-0.02
0.00
0.00
0.00
0.01
-0.02
0.03
-0.01
0.01
0.01
-0.01
0.03
0.00
0.02
-0.01
0.02
0.02
0.00
0.03
-0.01
0.04
0.01
0.01
0.04
-0.03
0.03
0.02
0.05
0.02
0.04
0.02
0.02
-.06
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.05
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
p–
value
0.10
0.20
0.48
0.10
0.18
0.68
0.84
1.00
0.43
0.40
0.19
0.55
0.62
0.49
0.82
0.13
0.98
0.20
0.85
0.51
0.24
0.78
0.18
0.78
0.09
0.98
0.56
0.03
0.18
0.16
0.32
0.01
0.51
0.07
0.30
0.38
0.47
II
III
IV
0.02
-0.03
0.00
0.00
0.00
0.00
0.42
0.22
0.86
Near Visual Acuity at Various Contrast Levels:
The differences in near visual acuity at various contrast
levels across conditions are shown in the table 4. As can
be seen most differences are statistically insignificant.
Significant differences were seen only between conditions
I and III at 100 % contrast and between conditions I and II
at 9 % contrast. The first difference could be indicative of
a real difference in the light provided by the two lamps
when the room lights were kept on. However, the
difference between conditions I and II at 9% contrasts
level has no rationale to be believed. Moreover, the
differences were only 0.05 logMAR which is clinically
insignificant.
Figure 6: Mean differences in error scores in Munsell
colour chips across various conditions.
Table 4: Difference in changes in visual acuity at various
contrast levels compared across various conditions
Contrast
(%)
Basal Tear Production:
100
Differences in changes in tear production across the
various conditions were found to be statistically
insignificant (table 3). Since the maximal mean change
was 2.13 mm in the third condition, these differences
across conditions were neither clinically significant. The
median difference value was found to be 0 mm for all the
four comparisons.
71
Table 3: Change in basal tear production across
conditions.
Conditions compared
I and III
II and IV
I and II
III and IV
Mean Difference (mm)
-1.48
-0.37
-1.10
0.02
p-value
0.13
0.79
0.09
0.57
50
35
25
Stereopsis:
The amount of change in depth perception (stereopsis) in
each of the lighting condition is 0 arc seconds. Therefore
the amount of change across lighting conditions was of no
difference.
18
Achromatic Point Estimation:
Mean error scores were 3.66 (± 3.85), 3.5 (± 4.14), 5.33
(±5.83), and 5.2 (± 4.77) for conditions I, II, III, and IV
respectively. Under the LED lamp, the average error
scores were around 5 irrespective of whether the room
lights were kept on or off, while it was around 4 for the
CFL lamp. Comparison of error values in achromatic
point estimation using the Munsell chips across the four
conditions are shown in figure 6. None of the differences
were found to be statistically significant (p > 0.05 for all
the four comparisons). Since there is no standard for
clinical usage of achromatic setting we cannot comment
about the clinical significance of the differences.
However, we surmise that the differences are clinically
insignificant since the magnitude of difference is only
about 1.5 out of 40 plate which translates to an error rate
difference of 3.75%.
13
9
6
4
Conditions
compared
I and III
II and IV
I and II
III and IV
I and III
II and IV
I and II
III and IV
I and III
II and IV
I and II
III and IV
I and III
II and IV
I and II
III and IV
I and III
II and IV
I and II
III and IV
I and III
II and IV
I and II
III and IV
I and III
II and IV
I and II
III and IV
I and III
II and IV
I and II
III and IV
I and III
II and IV
I and II
III and IV
I and III
II and IV
I and II
III and IV
Mean
Median
-0.05
-0.01
-0.05
-0.01
-0.02
0.01
-0.03
0.00
-0.01
-0.01
0.04
0.04
0.01
-0.02
-0.01
-0.04
0.01
0.01
-0.02
-0.02
-0.01
0.00
0.02
0.04
0.03
-0.03
0.03
-0.03
-0.04
-0.02
-0.05
-0.03
-0.01
0.02
-0.02
0.00
-0.04
0.02
-0.08
-0.02
-0.10
0.00
-0.05
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
-0.05
0.00
0.00
0.00
-0.10
0.00
0.00
0.00
0.00
0.00
pvalue
0.05
0.39
0.08
0.52
0.39
0.70
0.22
0.71
0.82
0.83
0.11
0.35
0.61
0.62
0.75
0.18
0.88
0.92
0.48
0.55
0.59
0.93
0.32
0.23
0.39
0.21
0.27
0.21
0.09
0.37
0.03
0.26
0.78
0.43
0.51
0.93
0.74
0.64
0.35
0.41
103
Maximum Reading Speed and Critical Print Size:
Maximum reading speed (MRS) measured as number of
words correctly read per minute and critical print sizes
(CPS – critical print size is one acuity level above the size
at which the maximum reading speed was obtained) were
estimated using recommended methods. Comparison of
these two quantities across the four conditions revealed no
statistically significant differences (table 5) except for
critical print size when compared between conditions II
and IV; even this was only of marginal significance. Both
of these conditions are “Dark conditions”. We
hypothesize that the light provided by the LED lamp was
such that better reading performance was obtained with
larger print sizes when using CFL lamp. This is justified
by the illuminances provided by the two lamps. A plot of
reading speed against font size did not come up as an
inverted U for all subjects.
Table 5: Maximum Reading speed and critical print size
on comparing between various conditions
Conditions
compared
MRS difference (wpm)
CPS (logMAR)
I and III
Mean
3
p-value
0.92
Mean
-0.1
p-value
0.74
II and IV
-10
0.10
-0.2
0.05
I and II
6
0.33
0.0
0.24
III and IV
-7
0.27
0.1
0.94
Blink rate:
Blink rate was reduced from normal across all condition
and had a value of around 5 per minute. None of the
comparisons across conditions showed any significant
difference.
Visual Discomfort Score:
Visual discomfort score was obtained using Rasch
analysis. Different weights were given for each of the
visual comfort variable. The response to a given question
had values ranging from 0 to 5. For each question, the
answer was multiplied by the weight for that question and
these were summed to get the total score. Maximum score
(14 out of 85) was obtained for condition 3. Most people
responded ‘no discomfort’ for all the tested parameters,
namely, fatigue, pain, glare, headache, eyestrain and
dryness. Among those who had discomfort, glare was the
most common visual discomfort across all conditions.
Comparison of visual discomfort score across conditions
revealed no significant difference.
LED – CFL Comparison: Pooled Analysis:
Since we did not find substantial differences in the visual
performance under the two lamps under the two lighting
conditions, we decided to pool date from the two lighting
conditions for each of the lamps to see any difference in
these two lamps. Statistically significant differences in
104
changes were seen in the visual acuity at 100% contrast
using the SVIS chart. Under the CFL lamp, the visual
acuity improved by 0.02 logMAR unit and deteriorated by
0.01 logMAR unit under the LED lamp. However, both
these values are way too small compared to be of any
clinical significance. The only other parameter that
showed any statistically significant difference between the
two lamps was the achromatic point setting. The mean
setting for the CFL lamp was 3.58 and 5.27 for the LED
lamp. These translate to an error rate of 8.95% for the
CFL lamp and 13.18% for the LED lamp.
DISCUSSION AND CONCLUSION:
Statistically significant change was not seen in most of the
visual/ocular parameters tested. Where statistically
significant change was seen, the magnitude of change was
not clinically significant. Basal tear secretion was
statistically significantly reduced in all but the first
condition. However, none of these reductions were
clinically significant. Blink rate was observed to be
subnormal across all conditions. Therefore, the changes
that were seen could be not large enough to show
statistical significance.
Reading speed could not be taken as a reliable measure
since the variation of reading speed with font size did not
come up as an inverted U. The critical print size was
statistically significantly larger for the LED lamp than for
the CFL lamp when the room lights were kept off. The
difference was two logMAR sizes which could also be
clinically significant. Under the “Light condition”,
however, the difference was only one logMAR size which
was not found to be statistically significant. Our LED
lamp provided on average 200 lux at the primary task area
while the CFL lamp provided 2.5 times that amount.
Therefore, this difference in critical print size could be
due to the glare produced by the CFL lamp due to its
larger light level. On the other hand, at 100% contrast, in
the “Light Condition”, (i.e., when the room lights were
kept on), the visual acuity change was ! a line smaller
under the CFL lamp than under the LED lamp. This
difference in change however is not clinically significant
but its statistical significance could be due to the less light
level provided by the LED lamp.
Glare was the most commonly complained visual
discomfort using both lamps and in both lighting
conditions. However, complaint of glare was reported by
more number of people when using the CFL Lamp under
“Dark Condition” and minimum number of people
complained of glare when using LED lamp in the “Light
Condition”. In both the dark and light conditions, the LED
lamp had the least number of complaints with respect to
glare. This could be attributed to the low light level
provided by the LED lamp.
Pooled data from the dark and light conditions for both
the lamps showed expected difference in the change in
visual acuity under the two lamps. However, to our
surprise, difference in the achromatic setting was also
seen. For these visual parameters, the CFL lamp seemed
to have fared well. While it is possible that the effect on
visual acuity could be explained by the higher illuminance
provided by the CFL lamp, we are not in a position to
speculate on the reason behind the difference observed in
the achromatic setting. Measurement (or the availability)
of the colour rendering index of the light sources used in
the two lamps could have thrown some light on this issue.
From the results, we find that there is not much of a
difference in the effects produced by the LED based study
lamp and the CFL lamp on most visual functions,
irrespective of whether the room lights were kept on or
off. The performance of the subjects in discrimination of
various hues, resolution at various levels of contrast,
perception of depth across all four lighting conditions was
not much affected in any condition.
In conclusion, the two lamps that we used in our
experiment did not produce statistically or clinically
significant different effects for the most of the visual
parameters we studied. The small number of statistically
significantly different affect that we observed could
possibly be explained by the vast differences in the
illuminances provided by the two lamps. Therefore, we
speculate that equalising the illuminances could probably
have shown some significant differences. In addition, the
near vision tasks were done only for 20 minutes. The task
and its duration might not have stressed the visual system
to bring out the differences in the effect the two lamps
had.
7.
8.
9.
10.
11.
12.
13.
14.
15.
REFERENCES
1.
2.
3.
4.
5.
6.
Akutsu, H., Legge, G.E., Ross, J.A., Schuebel, K.J.
Psychophysics of reading: X - Effects of age related
changes in vision. J Gerontol: Psychol Sci 46,
(1991), 325-331.
Berman, S.M., Navvab, M., Martin, M.J., Sheedy, J.,
Tithof, W. A comparison of traditional and high
colour temperature lighting on the near acuity of
elementary school children. Lighting Res Tech 38,
(2006), 41-49.
Borsting, E., Chase, C. H., Ridder, III W.H.
Measuring visual discomfort in college students.
Optom Vis Sci, 84, (2007), 745–751.
Carlson, N.B., Kurtz, D. Clinical Procedures for
Ocular Examination, 3rd Edn. McGraw-Hill
Companies, Inc., New York, USA, (2004).
Eperjesi, F., Fowler, C.W., Evans, B.J.W. Effect of
light filters on reading speed in normal and low
vision due to age-related macular degeneration.
Ophthal Physiol Opt 24, (2004), 17–25.
Eskridge, J. B., Amos, J. F., Bartlett, j, D. Clinical
Procedures in Optometry. J. B. Lippincott Company,
Philadelphia, USA, (1991).
16.
17.
18.
19.
20.
Grosvenor, T. Primary Care Optometry, 4th Edn.
Butterworth-Heinemann, Massachussetts, USA,
(2002), 93-94.
Legge, G.E., Parish, D.H., Luebker, A., Wurm, L.H.
Psychophysics of reading: XI. Comparing color
contrast and luminance contrast. J Opt Soc Am A, 7,
(1990), 2002-2010.
Legge, G.E., Pelli, D.G., Rubin, G.S., Schleske,
M.M. Psychophysics of Reading I: Normal Vision.
Vision Research, 25, (1985), 239-252.
Legge, G.E., Rubin, G.S. Psychophysics of Reading
IV. Wavelength effects in normal and low vision. J.
Opt. Soc. Am. A, 3, (1986), 40-51.
Lightstone, A., Lightstone, T., Wilkins, A. Both
coloured overlays and coloured lenses can improve
reading fluency, but their optimal chromaticities
differ. Ophthal. Physiol. Opt, 19, (1999), 279-285.
Le!nik, H,, Poborc-Godlewska, J. The relationship
between ciliary muscle fatigue and the type of
artificial light used to illuminate the area of visual
work. Pol J Occup Med Environ Health. 6, (1993),
287-292.
Rabin, J. Luminance effect on visual acuity and
small letter contrast sensitivity. Optom Vis Sci, 71,
(1994), 685- 688.
Scheiman, M., Wick, B. Clinical Management of
Binocular Vision. 2nd Edn. Lipppincott Williams and
Wilkins, Philadelphia, USA (2002); 58.
Steigerwald, D.A., Bhat, J.C., Collins, D., Fletcher,
R.M., Holcomb, M.O., Ludowise, M.J., Martin, P.S.,
Rudaz, S.L. Illumination with solid state lighting
technology. IEEE J. Select. Topics Quantum
Electron, 8, (2002), 310 – 320.
Veitch, J.A., McColl, S.L. Modulation of fluorescent
light: Flicker rate and light source effects on visual
performance and visual comfort. Lighting Res Tech
27, 1995, 243-256.
Virgili , G., Cordaro, C., Vigoni, A., Crovata, S.,
Cecchini, P., Menchini, U. Reading acuity in
children: evaluation and reliability using MNREAD
charts. Invest Ophthalmol Vis Sci, 45, (2004), 3349 –
3354.
Werner, J. S., Schefrin, B. E. Loci of achromatic
points throughout the life span. J Opt Soc Am A, 19,
(1993), 1509 – 1516.
Whittaker, S.G., Lovie- Kitchin, J.E. Visual
Requirements for reading. Optom Vis Sci, 70,
(1993), 54-65.
http://www.gelighting.com/na/business_lighting/edu
cation_resources/learn_about_light/distribution_curv
es.htm (Last Accessed: July 15, 2009).
105
Using Core Sunlighting to Improve Office Illumination
Lorne Whitehead
Allen Upward
Peter Friedel
Michele Mossman
University of British Columbia
6224 Agricultural Rd
Vancouver, BC, V6T 1Z1, Canada
+1 604 822 3075
lorne.whitehead@ubc.ca
John Huizinga
Tom Simpson
3M Company
3M Center
St. Paul, MN, 55144
+1 651 733 4175
jtsimpson@mmm.com
ABSTRACT
Today’s lighting industry is challenged to deliver high
quality illumination while using less electrical energy and,
to achieve this, it is generally agreed that it is useful to
incorporate daylighting techniques. We present a new
system that is capable of delivering sunlight deep into the
core of multi-floor buildings. This is the first direct
sunlighting system that has the potential for sufficiently
low net lifecycle costs to enable widespread
implementation. While the energy savings are important,
the impact of sunlight on the visual environment is perhaps
even more important, given possible effects on health and
well-being. A demonstration of this core sunlighting
system has been installed in an existing building located at
the British Columbia Institute of Technology in Burnaby,
Canada. The demonstration shows a substantial reduction
in the electrical lighting load and enables an evaluation of
the spectral quality of the sunlight illumination from the
perspective of the occupants.
lumen) of these systems has not been established. As a
result, this is the first sunlighting system with potential for
widespread adoption.
SYSTEM DESIGN
The sunlighting system combines two structures, one for
collecting sunlight and another for distributing it within the
building.
Collecting the sunlight
The sunlight collection system shown in Fig. 1 is housed
inside an enclosure that extends along the south-facing
facade, above the windows and adjacent to the plenum
space. The enclosure protects the components from wind,
precipitation and dirt, enabling the use of relatively
inexpensive and lightweight materials and eliminating the
need for regular maintenance.
Keywords
Illumination, daylighting, energy efficiency, light guides
INTRODUCTION
We have developed a sunlighting collection and
distribution system to deliver sunlight to the core, or
interior regions, of multi-floor office buildings, in order to
substantially reduce the need for electric lighting and
improve the quality of the illumination [11, 12]. This
system has the potential to reduce energy for standard
commercial building lighting by at least 25%, replace
electric lighting 75% of the time each day that the sun
shines within six core daylight hours, reduce peak electrical
power demand when it is needed (i.e. midday on clear,
sunny days) and provide high quality illumination with
excellent colour rendering properties. The recognized value
of core daylighting is not new [2, 4, 14, 17], and some
recent systems using optical fibers have been under
development [8, 16]. However, the cost-effectiveness and
performance (in terms of expenditure rate per delivered
106
Fig. 1. Design of the sunlight collection module
The key feature of the collection system is an array of thin,
approximately square mirrors that move to track the motion
of the sun across the sky. The mirrors reflect the sunlight
directly toward the building so that it can efficiently be
concentrated throughout most the day, in order to enter the
interior light guide system and thus illuminate the building
interior. They are interconnected so that they can be
reoriented in unison using only two simple, inexpensive
motors. To achieve this motion, low cost universal joints
positioned in the centre of the top and bottom mirrors in
each column allow the mirrors to rotate easily about two
axes to adjust to the changing altitude and azimuth of the
sun, but prevent neighbouring columns from interfering
with one another. The mirrors in a single column are
attached by wires in each of the four corners, as shown in
Fig. 2, and two mirror arrays, each consisting of 35 mirrors,
are enclosed within each 3 meter wide collection module.
Fig. 2. The collection system includes an array of small
computer-controlled tracking mirrors.
As with any moving system, it is important to consider the
lifetime of the device. Since moving parts are
commonplace in many devices, such as automobiles,
elevators and computers, moving components are generally
designed to run, maintenance-free and often unprotected
from environmental elements, at a rate of many cycles per
second for at least hundreds of thousand of cycles. In
contrast, the mirror array in this system is protected from
the weather and dirt by the exterior canopy, and it need
only move at a rate of one cycle per day. As a result, the
lifetime reliability requirements are greatly reduced. An
accelerated-aging test of a prototype unit has demonstrated
performance that far exceeds the typical 30-year
requirement for building installations.
The demonstration devices use machined metal
components, but they have been designed in such a way
that they can be fabricated in plastic using standard
injection molding processes. These processes will require a
capital investment in tooling, but will result in the
inexpensive components that are required for a
commercially-viable device.
Distributing the sunlight
The concentrated and still partially collimated sunbeam is
directed through the small window and into light
distribution structures housed within the ceiling cavity.
These specially-designed interior light guides are integrated
with dimmable electric lamps. Sensors monitor the
illumination to supplement the sunlight as required [13]. In
the demonstration project, the prototype fixtures were
integrated into the existing t-bar drop-ceiling configuration.
For simplicity, a rectangular cross-section, as shown in Fig.
3, was chosen in order to be compatible with the existing
ceiling configuration, but this approach could be readily
adapted to other luminaire designs, including discrete
fixtures, direct/indirect luminaires or other suspended
fixtures.
Fig. 3. (a) Single light guide section and (b) modules
installed in the ceiling
DEMONSTRATION SYSTEM
The first demonstration of this core sunlighting system has
been installed in an existing building located at the British
Columbia Institute of Technology in Burnaby, Canada.
This building, constructed in the 1980’s, is representative
of a substantial fraction of the existing commercial building
stock, so it showcases the potential of the technology to
substantially reduce the electrical lighting in existing
buildings as well as new construction, and consequently
reduce greenhouse gas emissions. The demonstration
shows a substantial reduction in the electrical lighting load
and enables an evaluation of the spectral quality of the
sunlight illumination from the perspective of the occupants.
For this demonstration, half of the existing building canopy
on the third floor was removed in order to install five
collection modules, as shown in Fig. 4.
Following its reflection from the mirror array, the sunlight
is concentrated and recollimated by off-axis paraboloidal
mirrors and then directed into the building through a small
window.
107
portion is visible to the occupants in the room; it has been
designed to appear like a standard fluorescent fixture,
including a diffuser panel that covers the prismatic film.
Fig. 4. Collection modules installed at BCIT’s
For ease of installation, four of the collector modules, three
of which are shown in Fig. 5, redirect sunlight through the
existing topmost windows, which previously were not used
for view but rather to passively introduce daylight into the
room. Only one new window needed to be installed in the
building as the aperture for the fifth module, since the
existing windows did not extend to the far end of the
building. In all installations, the window will meet typical
structural and fire safety standards and in practice it can be
readily designed into a new building or added as part of a
retrofit project. In the case of this project, a double-glazed
low iron window with an anti-reflective coating was
selected.
Fig. 6. The original prismatic troffers were replaced with
dual-function light guides.
For simplicity, a rectangular cross-section was chosen for
this design since it is compatible with the standard 0.61 m
by 1.22 m (2 ft by 4 ft) t-bar drop-ceiling configuration, as
shown in Fig. 6. The guide has a depth of 0.25 m, which is
about twice that of a standard fixture, but as will be
explained in the integration section, there is sufficient space
within the ceiling cavity in this building to easily
accommodate the thicker profile.
In addition to providing core sunlight, this system, when
mounted above the windows on each, provides shade for
the windows at high sun elevation, as illustrated by the
building profile in Fig. 7. This shading feature reduces the
direct solar heating within the building, which
commensurately reduces the required air conditioning load,
resulting in additional significant energy savings.
Fig. 5. Collectors concentrate the sunlight into light guides
for distribution within the building.
The initial design is the same as a recessed luminaire, with
a rectangular cross-section. Although this is not the only
possible design, these fixtures can be directly mounted in
the place of the standard recessed troffer fixture that is
common in many office buildings. The fluorescent lamps
are installed within the light guide at specific locations to
ensure that an even distribution of electric lighting is
provided. In this configuration, only the light emitting
108
Fig. 7. The canopies also provide shade for the windows
below, reducing glare and air conditioning cost.
The shading feature also further improves the quality of the
interior illumination by substantially reducing the
discomfort glare caused by the direct light penetrating
sideways into the room through the windows. The result is
a workspace with bright, comfortable and pleasant
illumination at all times, and requires no electrical lighting
whenever direct sunshine is available.
In this demonstration, the canopies were installed in this
manner both for ease of installation and to take advantage
of the shading feature provided by the exterior canopy. In
other installations, depending on the architecture of the
building, it may be desirable to integrate the canopies into
the building so that they are not visible from the exterior.
This architectural flexibility enables the system to be used
in many existing buildings as well as new construction, as
will be demonstrated in subsequent installations.
watt” to 74 lumens per watt. (Although fluorescent lamps
have been used in this demonstration project, it is worth
nothing that higher efficacy sources, such as light emitting
diodes, can be used in the same system as soon as they
become a cost-effective alternative.)
Fig. 8 shows the measured illuminance distribution
provided by the guided sunlight, as measured on a
horizontal plane 0.8 m above the floor in one of the rooms
in the demonstration space, with the electric lights
completed turned off.
The results show that the
illuminance is greater than the required 500 lx everywhere
in the room, except for the extreme rear corners. The
illuminance is highest near the windows, since sunlight
enters through the window in this region, as well as via the
overhead light guide.
The installation of an additional five modules is currently
underway to complete the demonstration on the third floor.
Although this demonstration installation will only occupy a
single floor, the technology has been developed to provide
sunlight to all floors of a multistory building, by mounting
the modules adjacent to the plenum space on each floor.
EVALUATING THE SYSTEM PERFORMANCE
A preliminary assessment of the resulting electrical energy
savings as well as the spectral characteristics and colour
rendering properties of the illumination has been
completed.
Performance of the electric lighting system
The interior space is largely an open-plan office space with
individual workspaces and one interior meeting room that
does not have access to the windows. The layout of the
light guides ensured that the light is uniformly distributed
and the aesthetics suit the size and shape of the rooms.
Originally, 31 1x4 prismatic troffers with 2/34 Watt T8
lamps delivered 44 lumens per watt (lpw) at desk height. In
the new installation, 43 hybrid luminaires with 2/T5 28
Watt lamps and dimming ballasts, operating at 60%, yield a
net efficacy of 56 lpw when no sun is available. The colour
temperature of the lamps was selected to be 4100 K, which
is consistent with common illumination trends in North
America. These lamps [9] have specified to have a colour
rendering index (CRI) value of 85.
Performance of the sunlighting system
The light guides distribute up to 65,000 lumens of sunlight
over a distance of 12 m, and illuminance levels provided by
the guided sunlight are well above typical standards of 500
lux.
Based on the average annual local sunshine
probability, it is anticipated that this demonstration system
will enable the electric lights to be turned off at least 25%
of the time. This 25% displacement by sunlight increases
the average net effective efficacy of the lighting system, in
terms of the ratio of work plane lumens to average “wall
Fig. 8. Illuminance distribution measured in the room
These illuminance measurements were taken under clear
sky conditions, with direct sunlight incident on the
collection modules. The overall illuminance levels vary
somewhat depending on the time of day and year; these
measurements are representative of the average illuminance
levels achieved in sunny conditions. On overcast days, the
diffuse sunlight incident on the collection module does
provide a small amount of interior illumination via the light
guides, achieving average illuminance levels of 30 lux.
This low level of illumination is adequate for occupants to
see their surroundings and one another, for example for
emergency egress, but it is not adequate for activities such
as reading, so supplemental electric light is required on
overcast days.
The efficiency of the core sunlighting system is a result of
the use of the highly reflective material lining the interior
surfaces of the light guide. Without this high reflectance,
over the entire visible spectrum, it would not be possible to
guide the light deep within the building since it would
instead by absorbed as it was transported only a short
distance.
109
building. As the graph in Fig. 10 shows, the irradiance for
wavelengths less than 375 nm shows only statistically
insignificant noise, meaning that essentially all of the
ultraviolet light is absorbed by the front UV-absorbing
front window of the canopy enclosure and does not enter
the guide.
Fig. 9. The transmission efficiency of the light guide
depends on the reflectance of the interior surfaces.
In the guides used in the demonstration project, the sunlight
interacts with the inner surface of the guide roughly once
during each 2m section of the guide. (This distance
between reflections is estimated from the roughly 8°
collimation half angle of the guided sunlight, and the 0.25
m minimum guide dimension.) Based on this estimate, Fig.
9 illustrates the efficiency with which light is transported
down the guide, for different reflectance values.
The spectral distribution of the light emitted by the guide
was then evaluated in the visible band using a
spectrophotometer. Fig. 11 shows that the relative intensity
of the light as a function of wavelength for the light emitted
at the start and the end of the guide (1m and 11m,
respectively, from the collector aperture). A comparison of
the two sunlight distributions shows that there is a very
small amount of absorption of light toward the blue end of
the spectrum, but the distribution is largely preserved along
the guide. Importantly, the longer reddish wavelengths are
efficiently transported along the guide, and these
wavelengths are especially important for yielding high
quality colour rendering for familiar surfaces, including
human complexion.
This is a consequence of the
continuity and uniformity of the natural black-body
emission provided by the sun.
It is apparent that the reflectance of the surface is the
dominant factor in determining the guide efficiency; 90%
reflective aluminum sheeting is unacceptable since it
causes 60% of the light to be lost after 20 m, compared to
only a 10% loss along the same distance for a 99%
reflective material. Until recently, there was no costeffective way to achieve this high reflectance, but the
availability of a suitable highly reflective material now
makes this a practical solution.
Fig. 11. Spectral distribution of the light emitted from the
light guide
Using these spectral distributions and a software program
provided by the CIE [3], the value of the colour rendering
index (CRI) was determined for each. The CRI for sunlight
measured outside the building, at the start of the light guide
and at the end of the light guide was calculated to be
between 95 and 100, as would be expected for an
approximately black-body source. The measurements after
light had traveled 12m to the end of the guides showed that
the high colour rendering quality was retained..
Fig. 10. Irradiance measurements inside the light guide
In addition to the efficiency, the spectral characteristics of
the light in the workspace were evaluated in detail. First, a
UV spectrometer was used to measure the irradiance within
the light guide, next to the light guide aperture from the
canopy and through a short pass filter, to establish the
degree to which ultraviolet light is transported into the
110
The correlated colour temperature (CCT) of the unfiltered
sunlight was measured to be 5002 K, whereas the CCT was
measured to be 4700 K at the start of the guide and 4437 K
at the end. While this was a measurable difference, the
change in the CCT was not perceptible to occupants in the
room. If it were desirable to do so, however, a small
amount of filter material within the guide could be used to
make the colour temperature more uniform.
These spectral measurements confirmed that the sunlight
emitted from the light guide is full-spectrum, natural
sunlight along the entire length of the guide and therefore
provides a high colour rendering quality throughout the
workspace.
Comparison to optical fibers
As mentioned briefly in the introduction, there has been
substantial interest in using solid core optical fibers to
guide sunlight from a collector module into the building.
Optical fibers have the tremendous advantage that they are
flexible and therefore can be relatively easily routed
through the ceiling cavity to avoid HVAC ducts and other
services, in order to provide light where it is needed.
However, there are a number of fundamental disadvantages
that make optical fibers an impractical method for
transporting sunlight deep into buildings.
concentration factor also poses a fire hazard within the
building, if the appropriate (and costly) precautions are not
taken to safely transport and distribute the radiation.
Finally, the spectral quality of the transported light in a
plastic optical fiber is undesirable, since the absorption is
highly wavelength dependent. Fig. 13 shows the spectral
transmission of one particular type of fiber [10]. This
means that although the light entering the fiber has the
natural spectrum of sunlight, by the time it has traveled into
the core regions within a building, the light emitted from
the end of the fiber no longer has a natural distribution.
Rather, it has a noticeable greenish hue and as a result does
not provide the high colour rendering that is the appeal of
illumination by natural sunlight.
First, in order to minimize the cost of the fiber, it is
desirable to use as small a diameter fiber as possible, to
minimize the material cost. However, the smaller the fiber,
the more concentrated the sunlight must be, which
substantially increases the required precision, complexity,
and, ultimately, the cost, of the collector module.
Second, to reduce cost and increase flexibility, a single
solid core plastic fiber is generally proposed for this
application rather than a bundle of glass fibres, such as
those used in the telecommunications industry. Solid core
plastic fibers generally include additives that increase their
flexible and durability, but at the expense of optical clarity.
It is illustrative to compare this transmission efficiency for
a hollow light guide to that of a comparable length of
optical fiber. The attenuation of light per meter of optical
fiber typically ranges from 3% to 10%. These attenuation
values result in unacceptably low transport efficiencies for
long fiber lengths, as illustrated by the graph in Fig. 12.
Fig. 13. The transport efficiency of 10 m of optical fiber
shows that it preferentially absorbs red light.
While advances in plastics engineering may help to
overcome these limitations, at the moment optical fibers do
not appear to be a practical method for transporting
sunlight deep within a building.
EXPERIENCING SUNLIGHT INDOORS
In addition to its cost-effectiveness advantage, the core
sunlighting system demonstrated at BCIT is unique in its
ability to deliver concentrated and collimated sunlight,
essentially a bright sunbeam, deep within each floor of a
building. With the new core sunlighting system operating,
the workspace, including the interior meeting room shown
in Fig. 14, is entirely illuminated by sunlight whenever
direct sunshine is available between roughly three hours
before and three hours after noon (true local time).
Fig. 12. The transmission efficiency of optical fibers with
different attenuation constants.
Even for the most optically clear fibers, the transmission
efficiency is considerably less than that of a hollow light
guide lined with a highly reflective material. The high
There are a number of factors associated with the system
that can be adjusted based on occupant preference. For
example, in the demonstration installation, there is a
perceptible shift in colour temperature, depending on
whether the dual-function light guide emits sunlight
(CCT~4500 K) or fluorescent light (CCT~4100 K). If it is
111
desirable to match the colour temperatures for both sources,
this could be achieved by either using so-called “daylight”
fluorescent lamps or alternately by filtering the transmitted
sunlight at the collector aperture to slightly lower the CCT,
albeit with a slight decrease in the overall system
efficiency.
Similarly, the current control algorithm allows for a slight
and temporary decrease in the illumination level when a
cloud passes in front of the sun, before the fluorescent
lights are fully powered to compensate for the decreased
sunlight. This results in a gentle adjustment of the
illumination level at certain times. While these fluctuations
in brightness and colour are perceptible, thus far occupants
have not found them to be unpleasant or disturbing. On the
contrary, many have found that it gives them a desirable
feeling of contact with the outdoors and the changing
exterior lighting conditions, especially when they do not
have access to a window and would otherwise be
completely cut off from the outdoors.
illuminated by sunlight for about six hours during the
workday, whenever direct sunshine is available. The
demonstration also shows that the system can integrate
readily into standard architectural designs for new
commercial buildings and retrofit opportunities. This initial
demonstration shows the potential for the technology to be
cost-effective through energy savings and this economic
viability is critical to widespread adoption of any
sustainable technology. A detailed study of feedback from
the building occupants regarding the quality of the visual
environment is currently underway. The evaluation and
analysis of both the energy savings and the lighting quality
will be ongoing for several years to enable an accurate
assessment that is independent of unusual weather patterns.
ACKNOWLEDGMENTS
The authors are grateful for the support for this project that
has been provided by the Natural Sciences and Engineering
Research Council of Canada, BC Hydro Power Smart, the
British Columbia Institute of Technology, the Ontario
Power Authority, CEATI International, Inc., the Province
of British Columbia, Natural Resources Canada, Public
Works and Government Services Canada, Vancity, the Real
Estate Foundation of British Columbia and Busby Perkins
and Will Architects.
REFERENCES
1. Boyce, P., Akashi, Y., Hunder, C., and Bullough, J.,
“The impact of spectral power distribution on the
performance of an achromatic visual task”, Lighting
Research and Technology, 35, 141-161, 2003.
Fig. 14. The guide delivers sunlight into a windowless
conference room.
The resulting psychological impact of this increased
contact with the outdoors combined with the high colour
rendering illumination may prove to be significant in terms
of the perceived value of core sunlighting in buildings.
Some experts feel that the productivity and human factors
value of daylighting systems increase their value far
beyond the raw energy savings alone [1, 15, 7, 6, 5]. As a
result, the unique appeal of sunlight in the visual
environment is likely to encourage early adoption of core
sunlighting systems before they are able to be cost-effective
solely in terms of their energy savings. While such human
factors studies are well outside the scope of this paper, the
demonstration at BCIT provides an opportunity for more
extensive analyses in this area. The results of these human
factors studies will be presented in subsequent publications
on this work.
CONCLUSION
As a result of the demonstration system, the workspace,
including an interior windowless meeting room, is entirely
112
2. Callow, J., Shao, L., “Air-clad optical rod daylighting
system”, Lighting Res. Technol. 35, 31-38, 2003.
3. CIE 013.3-1995, “Method of measuring and specifying
colour rendering properties of light sources”,
Commission Internationale de L'Eclairage, Jan 1995.
4. Greenup, P. and Edmonds, I., “Test room
measurements and computer simulations of the microlight guiding shade daylight redirecting device”, Solar
Energy 76(1), 99–109, 2004.
5. Hedge, A., “Reactions of computer users to three
different lighting systems in windowed and
windowless offices”, Work and Display Units, B54B56, 1994.
6. Heerwagen, J. and Heerwagen, D., “Lighting and
psychological comfort”, Lighting Design and
Application 6, 46-51, 1986.
7. Heschong Mahone Group, “Windows and offices: a
study of office worker performance and the indoor
environment”, Report prepared for Public Interest
Energy Research, California Energy Commission,
2003.
8. Muhs, J. D., “Design and Analysis of Hybrid Solar,
Lighting and Full-Spectrum Solar Energy Systems”,
American Solar Energy Society, 2000.
9. Philips "Silhouette Series" 4100K T5 lamps (Product
Number 23086-2 F28T5/841).
10. Poly-optical Lumileen acrylic optical fibers.
11. Rosemann, A., Mossman, M. and Whitehead, L.,
“Development of a cost-effective solar illumination
system to bring natural light into the building core”,
Solar Energy 82, 302–310, 2008.
12. Rosemann, A., Mossman, M. and Whitehead, L.,
“Development of a cost-effective solar illumination
system to bring natural light into the building core”,
Solar Energy 82, 302–310, 2008.
13. Rosemann, A., Cox, G., Friedel, P., Mossman, M. and
Whitehead, L., “Efficient dual-function solar/electric
light guide to enable cost-effective core daylighting”,
Leukos Vol. 3(4), 259-276, 2007.
14. Schlegel, G., Burkholder, F., Klein, S., Beckman, W.,
Wood, B. and Muhs, J, “Analysis of a full spectrum
hybrid lighting system”, Solar Energy 76(4), 359–368,
2004.
15. Veitch, J. and McColl, S., “Full-spectrum fluorescent
lighting effects on people: A critical review”, National
Research Council Canada, Ottawa, ON, 1994.
16. Shah, V., “Flexible Sunlight—The History and
Progress of Hybrid Solar Lighting”, Emerging
Environmental Technologies, Springer, Netherlands,
83-104, 2008.
17. Whitehead, L., Brown, D., Nodwell, R., “A new device
for distributing concentrated sunlight in building
interiors”, Energy and Buildings 6, 119-125, 1984.
113
Effects of Dynamic Lighting on Office Workers:
First-year Results of a Longitudinal Field Study
Yvonne de Kort
Human-Technology Interaction
Eindhoven University of Technology
5600 MB Eindhoven, the Netherlands
+31 40 247 5754
y.a.w.d.kort@tue.nl
ABSTRACT
Dynamic lighting is designed to have positive effects on
wellbeing and performance. In a field experiment we tested
whether these effects are detectable and stable over time
when employed in actual work settings. This 2-year study
consists of two tranches, one following a monthly
alternating experimental design, the other a yearly
alternating one. This paper reports on the first tranche. In a
fully counterbalanced design, office workers experienced
dynamic or static lighting according to an a-b-a scheme
over 3 consecutive periods (N=142, 90, and 83).
Questionnaire data suggest no significant differences for
need for recovery, vitality, alertness, headache and
eyestrain, mental health, sleep quality, or subjective
performance, although employees were more satisfied with
dynamic lighting. Yet it is too early to discard the
hypotheses and claims made about dynamic lighting
altogether. Its effects may still emerge in environments
with limited daylight, over a longer time period, or when
more pronounced or differently shaped lighting patterns are
applied.
Karin Smolders
Human-Technology Interaction
Eindhoven University of Technology
5600 MB Eindhoven, the Netherlands
+31 40 247 3163
k.c.h.j.smolders@tue.nl
Tenner [1] showed that peoples’ preferences for artificial
lighting vary with weather conditions and time of the day.
Recent research has indicated that new lighting solutions
may actually have an impact on biological and
psychological processes.
Dynamic lighting is one of these innovative solutions, in
which lighting characteristics such as colour temperature
and intensity vary during the day according to a preset
protocol. This should have a positive effect on users'
wellbeing, health and performance. The rationale behind
dynamic office lighting is that it supports the natural
rhythm of employees’ alertness [2]. An exemplary protocol
– also applied in the present study – is presented in Figure
1. It is based on the idea that it stimulates workers during
the (work) day by exposing them to a high lighting level
and colour temperature in the morning and after lunchtime,
and creating a relaxing environment with lower and warmer
white light during the late morning and afternoon.
700 lux
4700 K
3000 K
Keywords
Dynamic Lighting, Wellbeing, Health, Performance
INTRODUCTION
500 lux
Figure 1. Protocol Philips dynamic lighting, source [2].
Office work isn’t all it’s cracked up to be. Although not
always physically challenging, having to deal with heaps of
paperwork, incessantly incoming emails, and the constant
buzz of phones, office humour, and printers rattling does
take its toll on one’s mental resources. On a more serious
note, stress and attention fatigue are all too common in the
office, so any environmental or ambient feature that holds
the potential to revive office workers or help them
recuperate from stress or fatigue throughout the day
deserves our attention. In the current study we explore
lighting as a potential environmental feature impacting
office workers’ wellbeing.
Light can influence the regulation of the biological clock,
and the secretion of hormones such as melatonin and
cortisol. During daytime the secretion of melatonin is low
and therefore the influence of light on its suppression
minimal [3]. Research has shown that the level of cortisol
increases when exposed to high lighting levels in the
morning, but not in the afternoon [3,4] or evening [5].
These biological effects are dependent on the colour
temperature, lighting level, duration and timing of
exposure, as well as on the size and position of the light
source [2,6,7,8] and probably have an influence on
individuals’ wellbeing, health and performance [9].
Artificial office lighting typically is constant in both
intensity and colour temperature, whereas natural light
varies throughout the day as a result of weather conditions
and the position of the sun. Begemann, Van den Beld and
Light also has a direct effect on people’s alertness and
sleepiness [2], apart from its indirect effect via the
biological clock. Research into the psychological effects of
lighting suggests that both a high intensity and a high
114
colour temperature can have positive effects on people’s
wellbeing, health and performance. For instance Fleischer,
Krueger and Schierz [10] showed that exposure to higher
colour temperature lighting (5600K) is more stimulating
than warm white lighting (3000K). Participants did indicate
that they experienced the warm white lighting as more
pleasant. Some smaller studies also showed an activating
effect of a higher colour temperature (6500-7500K)
compared to 3000 K lighting [11,12]. Other studies,
however, failed to demonstrate comparable effects (see e.g.
[13,14]), so overall the literature is still inconclusive.
Employing more extreme lighting conditions, Viola and
colleagues [15] found an effect of high colour temperature
(17000K) on workers’ ability to concentrate, level of
fatigue, alertness, daytime sleepiness and subjective
performance compared to a lower colour temperature
(2900K). Lastly, Mills et al. [16] found comparable effects
of colour temperature on wellbeing and performance of
employees in a call centre. Yet the range of colour
temperatures used in the current study is substantially
lower.
Aries [17] reported an inverse correlation between lighting
level and employees’ level of fatigue and sleep quality. In
an earlier experiment by Grünberger and colleagues [18],
participants were exposed to either a high lighting level
(2500 lux) or a lower lighting level (500 lux) for four hours
between 9.00am and 5.00pm. The results showed that the
higher lighting level had a positive effect on participants’
alertness, their ability to concentrate, the number of errors
they made on a performance test, and their mood compared
to lower intensity lighting. Other studies also showed
positive effects of a high lighting level on people’s
wellbeing and performance [e.g., 3, 19, 20]. It should be
noted that in most of these studies the difference in lighting
level between the high and low intensity lighting condition
was large (>2000 lux).
Practically all of the rigorous scientific research into the
biological and psychological effects of high intensity or
high colour temperature office lighting was performed in
laboratories, where participants are exposed to – sometimes
extreme – lighting conditions for only short periods of time
– typically several hours. Studies into the effects of
dynamic lighting are scarce both in the field and in the lab
and often involve only limited numbers of participants.
User evaluation in realised projects shows anecdotal proof
for increased wellbeing and performance amongst office
employees (e.g., Interpolis and Trigion in the Netherlands,
VUB bank in Slovakia). Whether these effects are
detectable and whether they are stable over time when
actually employed in the work setting has not been
thoroughly investigated to date.
The present paper will report on intermediate results of the
first large-scale field test into the effects of dynamic
lighting for office workers. The longitudinal study follows
an experimental design in two tranches, in which four
groups of about 100 to 200 employees each are alternately
exposed to dynamic and static lighting. In one tranche,
which we are reporting on here, lighting conditions change
on a monthly basis during winter months, counterbalanced
over two groups. In the second tranche the lighting
conditions remain stable during winter, dynamic for one
group, static for the other. Then during summer both
groups switch to the alternate condition. The advantage of
this design is that we can both explore the relatively short
and long-term effects of dynamic lighting compared to
constant lighting. In addition, we can compare the two
lighting conditions both between and within groups. In this
paper, we describe the results of data gathered during the
first winter for the two short-term groups (see Smolders &
de Kort [21] for preliminary results of the second tranche).
METHOD
Design
The current study is a field experiment, with Lighting
condition (dynamic vs. static) within, and Group (A vs. B)
between groups, and three consecutive measurement
periods (d-s-d and s-d-s scheme respectively, in January,
February and March1). In other words, two groups of
participants were exposed to dynamic or static lighting,
alternating on a monthly basis and counterbalanced
between groups. In the dynamic lighting condition,
employees experienced a gradually changing lighting
scenario with a higher lighting level (700 lux) and colour
temperature (4700 K) in the morning and after lunchtime
(see Fig1). The static condition had a 500 lux level of
3000K lighting.
The study was performed in a recently renovated high-rise
office building, with a large daylight contribution (see
Figure 2), in which a flex-working concept is applied.
Daylight-dependent control was applied in both conditions.
Investigation of the weather in first, second and third
measurement period – weekdays of the two weeks before
and one week during the survey – showed that during
January there were more sun hours than in February and
March (approximately 60, 35 and 40 respectively) [22].
Participants
In the first month of the field study, a questionnaire was
distributed among 414 office employees from 7
departments, of which 147 were completed and returned
(response rate: 35.5%). The data of five participants were
removed because they indicated that they were only rarely
at their workplace in the high-rise office building, that they
were ill during the measurement period, or that they filled
out the questionnaire at home. Of the remaining 142
participants (83 in the static and 59 in the dynamic
condition), 111 were male and 31 female (mean age 45, SD
= 10.23, range: 23 to 65).
1
In the original design there were four measurement periods and
the lighting condition would change three times. Due to
technical problems, the study was delayed and it was not
possible to have four measurement periods.
115
or from (1) ‘never’ to (5) ‘at least once a day’. Other
restorative activities had dichotomous response scales with
either (1) ‘It happens never or rarely’ and (2) ‘It happens
sometimes or often’ as response options, or with (1) ‘yes’
and (2) ‘no’ options. The evaluative statements are
dichotomous items with (1) ‘yes’ and (2) ‘no’ as response
options. Separation reliability of the scale was .83 in each
consecutive month. The separation reliability matches with
a classical definition of reliability; it represents the ratio
between the true and estimated variance of people’s
recovery needs [25]. The reliability score of this scale thus
indicates that scale’s internal consistency is satisfactory.
Mental health and vitality
Figure 2. Picture of indoor environment
In the second measurement period, the questionnaire was
again distributed and 96 employees (43 in the static and 47
in the dynamic condition) filled out the questionnaire
completely (response rate: 23.2%). The data of six
participants were removed because they indicated that they
were only rarely at their workplace in the office or that they
had filled out the questionnaire at home. Of the remaining
90 participants, 67 were male and 23 female with a mean
age of 48 (SD = 9.73, range: 25 to 63).
In the third measurement period, 84 employees (42 in the
static and 41 in the dynamic condition) completed the
questionnaire (response rate: 20.3%). One participant filled
out the questionnaire at home and his data was removed
from the dataset. Of the remaining 83 participants, 68 were
male and 15 female (mean age 48, SD = 9.45, range: 25 to
65).
Measures
The questionnaire consisted of measures for need for
recovery (i.e., the need to recuperate from attention fatigue
and stress), vitality, alertness, headache and eyestrain,
mental health, sleep quality, and subjective performance.
Subjective evaluations of lighting conditions were also
assessed. In addition, attitudes towards the job and work
environment and personal characteristics were included as
control variables. Objective measures such as days of sick
leave and coffee consumption were collected on
department level to corroborate subjective findings.
Need for Recovery
Need for recovery was measured with a behaviour-based
scale consisting of 34 items2 describing behaviours at office
employees’ discretion to recover from mental strain,
psychological distress, motivational deficits, and/or mental
fatigue [23], combined with 11 evaluative statements by
Van Veldhoven and Broersen [24]. Some items had 5-point
response scales ranging from (1) ‘never’ to (5) ‘very often’
2
The original scale consists of 35 items. The item "I take care of
plants in the office" was dropped due to a lack of variance as it
was not allowed to have plants in this office.
Mental health and vitality were assessed with two subscales
from the Dutch version of the SF-36 Health Survey
(RAND-36) [26]. The mental health subscale consists of 5
items, such as ‘Have you been a very nervous person?’ and
had an internal consistency between ! = .75 and ! = .81.
The vitality subscale consists of 4 items (e.g. ‘Did you have
a lot of energy?’) with Cronbach’s alpha between ! = .76
and ! = .87. The response options of both subscales ranged
from (1) never to (5) very often.
Headache and Eyestrain
Headache and eyestrain were measured with 8 items
adopted from Viola et al. [15], which describe symptoms,
such as ‘headache’ and ‘eye fatigue’, with response options
ranging from (1) ‘absent’ to (4) ‘severe’. The scale had an
internal reliability ranging from ! = .84 to ! = .89.
Alertness and sleep quality
Alertness was assessed with the Karolinska Sleepiness
Scale [27] with ‘today’ instead of ‘at this moment’ as time
frame. The response options ranged from (1) ‘extremely
alert’ to (9) ‘extremely sleepy - fighting sleep’. Sleep
quality was measured with the Pittsburgh Sleep Quality
Index [28] consisting of 18 items concerning subjective
sleep quality, sleep latency, sleep duration, sleep efficiency,
sleep disturbances, sleeping medication and daytime
dysfunction. The scale has an internal consistency between
! = .61 and ! = .70.
Subjective performance
Subjective performance was measured with the question
‘On a scale from 0 to 10, how would you rate your
performance on the days you worked during the last 2
weeks?’ derived from the World Health Organization
Health and Work Performance Questionnaire (WHO-HPQ).
Subjective evaluations
Subjective evaluations of lighting conditions concern
pleasantness of the lighting, experienced lighting level,
experienced disturbances of the artificial lighting and of
daylight, and satisfaction with the lighting. Pleasantness of
the lighting was measured with two semantic differential
adjective items (pleasant – unpleasant, comfortable uncomfortable). These items were internally consistent
116
with Cronbach's alpha ranging from ! = .79 to ! = .90.
Experienced lighting level was measured with three items
about lighting level (artificial light and daylight) on the
workplace, on the screen and in the office space from
Hellinga and de Bruin-Hordijk [29]. The response scale
ranged from (1) ‘too little light’ to (5) ‘too much light’ and
the scale was internally consistent with alphas ranging from
! = .72 to ! = .84. Experienced disturbance of the artificial
lighting was assessed with two items adopted from
Hellinga and de Bruin-Hordijk [29]. The 5-point response
scale ranged from (1) ‘never’ to (5) ‘very often’ and these
items had an internal consistency of alpha ranging from ! =
.75 to ! = .91. Experienced disturbance of daylight was
measured with similar items. This scale was internally
consistent with alpha ranging from ! = .69 to ! = .77.
Satisfaction with the lighting was assessed with the
question: ‘How satisfied are you with the lighting at your
workplace?’ with response options ranging from (1) ‘very
dissatisfied’ to (5) ‘very satisfied’.
Job and work-related evaluations
Job-related questions concern evaluation of the work
atmosphere, job satisfaction, commitment to the company,
work diversity, decision authority and job demands. To
assess work atmosphere, four evaluative statements were
employed, such as ‘The work atmosphere is good.’ The
response scale was a 5-point scale from (1) ‘never’ to (5)
‘very often’. The internal consistency of the four statements
ranged from ! = .81 to ! = .83. Three dichotomous (yes/no)
statements were employed to assess job satisfaction (‘I am
satisfied with my job’), commitment to the company (‘I
feel committed to the company’) and work diversity (‘my
work is diverse’), respectively. Decision authority and job
demands were measured with two subscales of the Job
Content Questionnaire [30]. Decision authority was
assessed with three statements, such as ‘I have freedom to
make decisions about my job'. The subscale is internally
consistent with alpha ranging from ! = .64 to ! = .69. Job
demands were measured with four statements, such as ‘My
job requires I work fast’. This subscale had an internal
consistency of alpha between ! = .68 and ! = .76. Both
subscales had a 4-point response scale ranging from (1)
‘totally disagree’ to (4) ‘totally agree’.
Work-condition related questions concerned the impression
of the office environment, pleasantness of the indoor
climate and satisfaction with the indoor climate. Impression
of the office environment was assessed with nine
adjectives, such as ‘pleasant‘, ‘orderly’ and ‘quiet’ from
Aries et al. [17]. The unipolar response options ranged
from (1) ‘not at all to’ (5) ‘extremely’. The internal
consistency of the 9 adjectives ranged from ! = .78 to ! =
.91. Pleasantness of the indoor climate was measured with
two semantic differential adjective items (pleasant –
unpleasant, comfortable - uncomfortable). This scale was
internally consistent with alpha ranging from ! = .84 to ! =
.92. To assess satisfaction with the indoor climate two
items concerning satisfaction with the temperature and
ventilation at the workplace were employed with response
options ranging from (1) ‘very dissatisfied’ to (5) ‘very
satisfied’. This scale was internally consistent with alpha
between ! = .73 and ! = .77.
Personal characteristics
Questions regarding personal characteristics concerned
gender, age, light sensitivity, and mean number of working
hours per week. Light sensitivity was measured with the
items 'How much trouble do your eyes give when you are
exposed to bright light?' and 'How much do you suffer from
headaches when you are exposed to bright light?' on a 5point scale from (1) ‘not at all’ to (5) ‘extremely’. The
reliability of this scale ranged from ! = .73 to ! = .78.
Procedure
In January, the lighting condition was dynamic for half of
the participants (group A) and static for the others (group
B). In the third week of this first month, all potential
participants received an e-mail with a hyperlink to the
questionnaire. A reminder was sent one week later. It took
about 15 minutes to fill in the questionnaire. A Living
Colors lamp from Philips was raffled every measurement
period as an incentive for participants to complete the
questionnaire. In February, the lighting condition was
switched from dynamic to static and vice versa. In March,
the lighting condition was again switched to the same
lighting condition as in January. During the second and
third measurement periods, the same procedure as in
January was used.
RESULTS
To investigate the effect of lighting condition (dynamic vs.
static lighting) on employees’ well-being, health and
performance, Linear Mixed Model analyses were
performed on need for recovery, vitality, mental health,
alertness, headache and eyestrain, sleep quality and
subjective performance (separately), with Lighting
condition and Month as fixed factors and participant
number as random factor. Light sensitivity, impression of
the office and work atmosphere were included as
covariates3.
The results showed that there was no significant effect of
Lighting condition on need for recovery, vitality, mental
health, alertness, headache and eyestrain, global sleep
quality and subjective performance (all F<1, except
alertness, F=1.31, NS). In Table 1, the F-statistics for
Condition and Month are shown. Table 2 shows the
estimated means for all dependent variables in both the
static and the dynamic condition.
3
We first assessed the Pearson’s correlations between potentially
confounding variables and dependent variables and added only
those covariates that had significant correlations with the
dependent measures for wellbeing, health and performance.
117
Table 1. Results linear Mixed Model analyses: F-statistics for Wellbeing, health and performance measures.
Need for recovery
F
Lighting condition
Month
Vitality
df
F
Mental Health
df
F
df
Alertness
headache &
eyestrain
F
df
F
df
sleep quality
Subjective
performance
F
F
df
df
.06
(1,167)
.08
(1,190)
.01
(1,179)
.01
(1,193)
1.31
(1,202)
.63
(1,151)
.35
(1,210)
13.27**
(2,153)
.34
(2,169)
2.56†
(2,154)
.45
(2,172)
1.01
(2,180)
2.81†
(2,135)
1.19
(2,190)
* p < .05, ** p < .01 and † p < .10
Table 2. Estimated marginal means of wellbeing, health and
performance measures.
Dynamic
Need for recovery
Static
M
SD
M
SD
-0.76
0.05
-0.77
0.05
Vitality
3.59
0.04
3.58
0.04
Mental Health
4.10
0.03
4.10
0.03
Headache and eyestrain
1.53
0.03
1.53
0.03
Alertness
3.74
0.12
3.59
0.11
Sleep quality
4.98
0.18
4.84
0.17
Subjective performance
7.42
0.06
7.46
0.06
The factor Month did show an effect on need for recovery
[F(2,153.0) = 13.27; p < .01]. Pair-wise comparisons
indicated that workers’ recovery needs were lower in
January (M = -.95; SD = .77) than in February (M = -.64;
SD = .71) and March (M = -.78; SD = .76) with p < .01 for
both contrasts. There was no difference in recovery needs
between February and March (p = .16). The effects of
Month on remaining dependent variables did not reach
significance.
We also performed Linear Mixed Model analyses with
scales probing the subjective evaluation of the lighting as
dependent variable, Lighting condition and Month as fixed
factors, participant number as random factor, and light
sensitivity, impression of the office environment and work
atmosphere as covariates. The results of these analyses
showed that Lighting condition had a significant effect on
satisfaction with the lighting [F(1,211.5) = 5.16; p < .05].
Office workers were more satisfied with the lighting in the
dynamic lighting condition (M = 3.69 and SD = .87) than in
the static condition (M = 3.53 and SD = .91). In addition,
Lighting condition had a significant effect on the
experienced disturbances of artificial lighting [F(1,196.3) =
4.44; p < .05]. Unexpectedly, workers reported fewer
disturbances of artificial lighting in the static condition (M
= 1.71 and SD = .72) than in the dynamic lighting condition
(M = 1.80 and SD = .78). Note that disturbances were
measured on a 5-point scale, thus office employees in both
conditions, on average never (1) or rarely (2) experienced
disturbances of the artificial lighting. There was no
significant effect of Lighting condition on experienced
disturbances of daylight [F<1, NS]. In addition, the
Lighting condition had no significant effect on the
evaluation of pleasantness of the lighting [F>1, NS]. The
effect of Lighting condition on experienced lighting level
approached significance [F(1,247.2) = 3.01; p = .08]:
indicating a trend for employees to evaluate the lighting as
brighter in the dynamic lighting condition (M = 3.06 and
SD = .48) than in the static condition (M = 2.98 and SD =
.52). Table 3 reports the F-statistics for Lighting condition
and Month concerning the subjective evaluation of the
lighting; Table 4 reports the mean scores on all subscales
for both experimental conditions.
Month had a significant effect on disturbances of daylight
[F(2, 192.0) = 4.98; p < .01]. Pair-wise comparisons
indicated that workers experienced more disturbances of
daylight in January (M = 2.69; SD = .91) than in February
(M = 2.54; SD = .91) and March (M = 2.52; SD = .82) with
p < .05 for both contrasts. There was no significant
difference between February and March concerning
disturbances of daylight (p = .55).
Table 3. Results of Linear Mixed Model analyses: F-statistics of subjective evaluation of the lighting condition.
Pleasantness
lighting
F
df
Satisfaction lighting
F
df
Lighting level
Disturbances
daylight
Disturbances
lighting
F
F
F
df
df
df
Lighting condition
1.87
(1,242)
5.16*
(1,211)
3.01†
(1,247)
.93
(1,215)
4.44*
(1,196)
Month
1.09
(2,220)
.21
(2,192)
2.28
(2,223)
4.98**
(2,192)
1.31
(2,178)
* p < .05, ** p < .01 and † p < .10
118
Table 4. Estimated marginal means of subjective evaluations of
lighting conditions.
Dynamic
Static
M
SD
M
SD
Pleasantness lighting
3.66
0.06
3.55
0.06
Satisfaction lighting
3.73
0.07
3.57
0.06
Lighting level
3.06
0.04
2.97
0.04
Disturbance daylight
2.64
0.07
2.56
0.07
Disturbance lighting
1.82
0.06
1.69
0.06
DISCUSSION
We are investigating the effect of dynamic lighting
compared to static lighting on workers' wellbeing, health
and subjective performance in a longitudinal field study. In
this paper, the results of the linear mixed model analyses on
the data of the short-term groups are reported (first
tranche). The results showed no significant difference in
workers' need for recovery, vitality, sleep quality, mental
health, headache and eyestrain, or subjective performance
between the dynamic and static lighting condition,
controlled for relevant personal, job and work-related
characteristics.
Interestingly, in spite of us not finding the beneficial effects
that were hypothesized, workers in the dynamic lighting
condition did report being more satisfied with the lighting
condition, although at the same time they reported more
disturbances from the lighting than did workers in the static
lighting condition.
Need for recovery showed a significant effect of month of
measurement, with employees reporting a lower need in
January than in February and March. A lower need for
recovery indicates a lesser degree of attention fatigue and
stress. This is in line with weather reports, indicating more
hours of sun on the workdays during the measurement
period in January than in February and March, but may also
be related to the fact that most employees had taken time
off in December on account of the holidays. The higher
number of disturbances of daylight in January may also be
explained by the fact that there were more hours of sun in
the first measurement period than in the other two.
The question we now need to address is what conclusions
could or should be drawn from these data. For this we must
consider not only the data, but also the methodology. We
had hoped to conduct the study in four consecutive months,
running four full-month measuring periods. Yet instead we
saw ourselves compelled to cut one period and shorten the
remaining periods from four to three weeks. This
unfortunately is the reality of doing field studies. However
we did manage to uphold a sound experimental design.
Also, considering the fact that in the questionnaires
participants were always asked to reflect on the last two
weeks, the procedure still worked well in the three-timesthree-week period compromise that resulted.
Furthermore, we employed a range of measurements, none
of which showed significant beneficial effects of dynamic
lighting. All scales repeatedly showed good reliability and
had been successfully used in earlier studies and although
response rates were only modest, participant samples were
still large enough to enable testing of these effects. Yet in
spite of the robust design, methodology and procedure, we
were not able to establish beneficial effects of dynamic
lighting when compared to static lighting.
On the other hand, a few considerations caution us to not
discard the potential of dynamic lighting just yet. First, a
possible reason for the lack of expected findings is the
substantial daylight contribution in the renovated building
of our study, especially in combination with the daylight
responsive lighting control. Dynamic lighting is said to be
most effective in situations with low daylight contribution
[31], so the building in this study – even if the study was
performed during the darker months of the year - may not
have been the best candidate for studying the effects of
artificial lighting. Moreover, the dynamic pattern of the
lighting itself my have attenuated the findings. As was
already reflected in the introduction, there as yet exists only
little research on dynamic lighting. The pattern employed
in the current study employs fairly subtle changes, both in
intensity and colour temperature, especially in comparison
to changes outdoors, or manipulations applied in
laboratory-based studies (e.g. [3,15,16,18,19,20]). These
design choices have been based on state-of-the-art insights
into human alertness curves, yet we are still far from fully
understanding light’s effects on humans’ psychological and
physiological states. The exact height of colour temperature
and intensity of the lighting, the exact timing and shape of
the curve and the range of wavelengths employed are all
still under investigation.
We conclude that in the first tranche of this longitudinal
research we have not been able to establish beneficial
effects of dynamic lighting on individuals’ need for
recovery, vitality, sleep quality, mental health, headache
and eyestrain, or subjective performance, although office
workers did report higher satisfaction with dynamic than
static lighting. Yet it is too early to discard the hypotheses
and claims made about dynamic lighting altogether. Its
effects may well emerge in more long-term applications,
environments with limited daylight contribution, or when
more pronounced, or differently shaped curves are applied
in terms of intensity and/or colour temperature.
ACKNOWLEDGEMENT
We
are
grateful
to
Rijkswaterstaat,
the
Rijksgebouwendienst, Philips and Ariadne Tenner for their
assistance and support during this project. We would also
like to thank Martine Knoop for her comments on an earlier
version of this paper.
119
REFERENCES
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
Begemann, S. H. A., van den Beld, G. J., & Tenner, A.
D. (1997). Daylight, artificial light and people in an
office environment, overview of visual and biological
responses. International Journal of Industrial
Ergonomics, 20, 231-239.
Knoop, M. (2006). Dynamic lighting for well-being in
work places: Addressing the visual, emotional and
biological aspects of lighting design. Proceedings of
the
15th
international
Symposium
Lighting
Engineering 2006 (63-74), Beld.
Rüger, M., Gordijn, M. C. M., Beersma, D. G. M., de
Vries, B., Daan, S. (2006). Time-of-day-dependent
effects of bright light exposure on human
psychophysiology: comparison of daytime and
nighttime exposure. American Journal of Physiology Regulatory, Integrative and Comparative Physiology,
290(5), 1413-1420.
Rüger, M. (2005). Lighting up the human circadian
clock. Ph.D thesis, Groningen: University of
Groningen.
Scheer, F. A., Buijs, R. M. (1999). Light affects
morning salivary cortisol in humans. Journal of
Clinical Endocrinical Metabolism, 84(9), 3395-3398.
van Bommel, W. J. M. (2006). Non-visual biological
effect of lighting and the practical meaning for lighting
for work. Applied Ergonomics, 37(4), 461–466.
Rea, M. S., Figueiro, M. G., Bullough, J. D. (2002).
Circadian photobiology: an emerging framework for
lighting practice and research. Lighting Research and
Technology, 34(3), 177-190.
Morita, T., & Tokura, H. (1998). The influence of
different wavelengths of light on human biological
rhythms. Applied Human Science, 17, 91-96.
van Bommel, W. J. M., van den Beld, G. J. (2004).
Lighting for work: A review of visual and biological
effects. Lighting Research and Technology, 36(4), 255269.
Fleischer, S., Krueger, H., Schierz, C. (2001). Effect of
brightness distribution and light colours on office staff,
results of the ‘Lighting harmony’ project. The 9th
European Lighting Conference ‘Lux Europa 2001’
(77-80), Reykjavik.
Mukea, H., & Sato, M. (1992). The effect of color
temperature of lighting sources on the autonomic
nervous functions. The Annals of Physiological
Antropology, 11, 533-538.
Deguchi, T., & Sato, M. (1992). The effect of color
temperature of lighting sources on mental activity
level. The Annals of Physiological Antropology, 11,
37-43.
McColl, S. L., & Veitch, J. A. (2001). Full-spectrum
lighting: a review of its effects on physiology and
health. Psychological Medicine, 31, 949-964.
Veitch, J. A., & McColl, S. L. (2001). A critical
examination of perceptual and cognitive effects
15.
16.
17.
18.
19.
20.
21.
22.
23.
24.
25.
26.
27.
28.
attributed to full-spectrum fluorescent lighting.
Ergonomics, 44, 255-279.
Viola, A. U., James, L. M., Schlangen L. J. M., Dijk,
D-J. (2008). Blue-enriched white light in the
workplace
improves
self-reported
alertness,
performance and sleep quality, Scandinavian Journal
of Work, Environment and Health, 34(4), 297-306.
Mills, P. M., Tomkins, S. C., Schlangen, L. J. M.
(2007). The effect of high correlated colour
temperature office lighting on employee wellbeing and
work performance. Journal of Circadian Rhythms,
5(2), 2-10.
Aries, M. B. C. (2005). Human lighting demands,
healthy lighting in an office environment. Ph.D thesis,
Eindhoven: Eindhoven University of Technology.
Grünberger, J., Linzmayer, L., Dietzel, M., Saletu, B.
(1993). The effect of biologically-active light on the
noo- and thymopsyche and on psychophysiological
variables in healthy volunteers. International Journal
of Psychophysiology, 15(1), 27-37.
Badia, P., Myers, B., Boecker, M., & Culpepper, J.
(1991). Bright light effects on body temperature,
alertness, EEG and behavior. Physiology and
Behavior, 50(3), 582-588.
Kaida, K., Takahashi, M., & Otsuka, Y. (2007). A
short nap and natural bright light exposure improve
positive mood status. Industrial Health, 45(2), 301308.
Smolders, K. C. H. J. and de Kort, Y. A. W. (in press).
Light up my day: A between-group test of dynamic
lighting effects on office workers’ wellbeing in the
field. Proceedings of Lux Europa 2009, 9-11
September 2009, Istanbul.
http://www.knmi.nl/klimatologie/maand_en_seizoenso
verzichten/index.html, retrieved 27 April 2009.
Smolders, K. C. H. J., de Kort, Y. A. W., Kaiser, F. G.,
Tenner, A. D. (2009). Need for Recovery in Offices:
Behavior-based Assessment. Unpublished manuscript,
in preparation.
Van Veldhoven, M., Broersen, S. (2003).
Measurement quality and validity of the “Need for
Recovery scale”. Occupational Environmental
Medicine, 60(supplement 1), 3-9.
Bond, T. G., & Fox, C. M. (2001). Applying the Rasch
model: Fundamental measurement in the human
science. Mahwah, NJ: Erlbaum.
Van der Zee, K. I., Sanderman, R. (1993). Het meten
van de algemene gezondheidstoestand met de RAND36, een handleiding. Groningen: Rijksuniversiteit
Groningen.
Åkerstedt, T., Gillberg, M. (1990). Subjective and
Objective Sleepiness in the Active Individual,
International Journal of Neuroscience, 52(1), 29-37.
Buysse D. J., Reynolds C. F., Monk T. H. Berman S.
R., Kupfer D. J. (1989). The Pittsburgh Sleep Quality
Index: A new instrument for psychiatric practice and
research. Psychiatry Research, 28(2), 193–213.
120
29. Hellinga, H.IJ., de Bruin-Hordijk, G.J. (2008). The
influence of daylight and view out on the visual
comfort of workplaces. Balkanlight 2008 Proceedings
(233-242), Ljubljana.
30. Karasek R., Brisson C., Kawakami N., Houtman I.,
Bongers P., Amick B. (1998). The job content
questionnaire (JCQ): an instrument for internationally
comparative assessments of psychosocial job
characteristics. Journal of Occupational Health
Psychology, 3(4), 322-355.
31. Schriek M.H.J. (2007). Dynamic Lighting – Energy
Savings by Use of Atypical Daylight Responsive
Lighting Control. Master Thesis, Eindhoven:
Eindhoven University of Technology, Faculty
Architecture.
121
Persuasive Lighting:
The Influence of Feedback through Lighting on Energy
Conservation Behavior
Jaap Ham, Cees Midden, Saskia Maan and Bo Merkus
Human-Technology Interaction
Eindhoven University of Technology
PO Box 513, 5600 MB Eindhoven
The Netherlands
+31 40 247 4210
j.r.c.ham@tue.nl
ABSTRACT
Earlier research has investigated persuasive technology:
Technology designed to influence human behavior or
attitudes. The current research investigates lighting as
persuasive technology. In an experimental study,
participants could conserve energy while carrying out tasks
and received feedback about their energy consumption in
each task. We tested the effect of feedback through a lamp
that gradually changed color dependent on energy
consumption and compared these effects to more widely
used factual feedback. Results indicated that feedback
through lighting has stronger persuasive effects than factual
feedback. Furthermore, factual feedback seemed more
difficult to process than lighting feedback, because
cognitive load interfered with processing factual feedback,
but not with processing lighting feedback. Implications for
theory and design of persuasive lighting, and (ambient)
persuasive technology are discussed.
Keywords
Lighting feedback, factual feedback, interactive feedback,
energy consumption behavior, ambient persuasive
technology
INTRODUCTION
The threats of growing CO2-emissions and climate change
effects and the exhaustion of natural resources have urged
nations worldwide to seek for substantial reductions in
energy consumption. Next to important technological
solutions like more efficient systems and devices and the
development of renewable energy sources, consumer
behavior plays a crucial role in bringing down the level of
energy consumption.
Influencing consumer behavior to promote energy
conservation has become an important target of national
and international policy efforts. Thereby, the question
which instruments should be applied to promote energy
conservation behavior has become highly relevant.
Recent reviews [e.g., 2, 15] have evaluated the effects of
interventions to promote energy efficient behavior. In
122
general, mass media public campaigns seem to lack
precision in targeting and message concreteness to achieve
behavioral change. By contrast, raising people’s awareness
of energy consumption by providing tailored feedback
about their energy consumption (for example in kWh) can
promote the achievement of behavioral change [see, e.g., 2,
15]. The results are mixed though. Weak linkages between
specific actions and energy outcomes caused by low
feedback frequencies (e.g. once per month) and insufficient
specificity of the feedback (e.g. household in general vs.
specific person or specific devices) are underlying these
mixed findings.
Recently, technological solutions have created new
opportunities to improve feedback efficacy by embedding
feedback in user-system interactions. That is, energy use is
in essence always the outcome of an interaction between a
user and some energy-consuming device. Intervening in
these specific interactions might improve the quality of
feedback substantially. Some evidence supports this claim.
McCalley and Midden [14] demonstrated in several studies
that interactive forms of feedback could be effective in
enhancing energy-efficient use of devices like washing
machines. By adding an energy meter to the user interface
of a washing machine they achieved 18% of energy
conservation both in lab and field studies. Basically, their
approach entailed giving factual feedback in terms of kWh
consumed as a function of programming choices made by
the user, like water temperature, spinning speed or the
duration of the washing cycle.
However, in many day-to-day situations people might not
be motivated or lack the cognitive capacity to consciously
process relatively complex information [see e.g. 5]. Factual
feedback (e.g., the numbers representing kWh
consumption) might be that kind of relatively complex
information. In the current research, we will investigate the
persuasive effects of a form of feedback that is easier to
process. We argue that (interactive) feedback using lighting
is simpler to process than (interactive) factual feedback
because it can directly express evaluative meaning whereas
factual feedback still needs to be processed and evaluated
by the user. For example, red lighting might be defined as
meaning “high energy consumption”, which does not need
to be evaluated further, whereas factual feedback that 120
kWh was used does. Also, feedback through (diffused)
lighting can be perceived easily without focusing, in
contrast to factual feedback. For example, (part of) the
environment of the user can be used for lighting feedback,
whereas the user needs to focus on factual feedback (e.g.,
in the form of numbers).
In addition, we argue that lighting has specific qualities that
make it particularly suitable for providing user feedback.
For example, lighting can be very cheap, is easy to install,
lighting can be very energy friendly, lighting can be seen
by other people present in a room as well (inducing social
pressure as a persuasive mechanism), and lighting might
have an emotional appeal or even direct emotional effects.
Also, the low conspicuity of light and color changes sets
lighting apart from other feedback mechanisms.
Furthermore, lighting can be calm (in the sense of ‘calm
computing’). Other feedback mechanisms often lack these
characteristics. For example, feedback mechanisms like
factual feedback or feedback that uses sound, smell, or
tactile feedback cannot easily be calm in that sense.
Therefore, we argue that lighting can be particularly suited
as a persuasive agent.
Earlier research indicates that energy consumption
feedback that does not consists of specific facts, but rather
of lighting changes can influence consumer behavior [see
7, 23, 3, 9, 20, 18, see also 17]. For example, in the eighties
of the previous century Becker and Seligman [6]
investigated the effectiveness of a light that went on “in a
highly visible part of the home” whenever the air
conditioner was on, but the outside temperature was 20°C
or lower. In homes that contained the signaling device, an
average of 15% savings in energy consumption was found.
More recently, a device called an energy orb was used that
changed color dependent on the time-of-use tariff in
operation. This type of information helped users save some
energy [12] and the usefulness of the device was positively
evaluated by users [20, 12].
The current research will investigate the effects of feedback
through lighting on energy consumption and compare them
to the effects of factual feedback. The feedback (lighting
feedback and factual feedback) that we will investigate in
this research will be of a highly interactive nature. Earlier
research of lighting feedback has already employed
feedback that contained elements of interactivity (e.g., in
Becker & Seligman’s research [6]). For example, Becker
and Seligman’s participants received feedback about their
action, although not in direct response to those actions. In
the current research, participants will receive feedback
about consequences of an action in direct response to that
action. More specifically, the current research will give
users lighting feedback about their current energy
consumption in a specific task, and this lighting feedback
will change directly when they use more or less energy.
Furthermore, the current research will investigate the
assumption that lighting feedback is easier to process than
factual feedback.
The Current Research
In the present study, we examine whether interactive
feedback through lighting can stimulate energy
conservation behavior. That is, we will use lighting color as
feedback to indicate the absolute level of energy
consumption (more green = lower energy consumption, vs.
more red = higher energy consumption). We set up an
experiment in which participants had the opportunity to
conserve energy in a series of tasks and received feedback
about their energy consumption during these tasks. We
tested the effect of lighting feedback and compared these
effects to more widely used factual feedback. More
specifically, we compared the effects of lighting feedback
using lighting color to indicate energy consumption, to the
effects of factual feedback using a number to indicate
energy consumption in Watts. When giving lighting
feedback, low consumption was indicated by completely
green lighting and high consumption by completely red
lighting. So, people can quite easily understand whether a
specific lighting (e.g. light-green) indicates high or low
consumption. However, when factual feedback would
consist of only one number (representing energy
consumption in Watts), it would be a lot more difficult to
know whether that number indicates high or low
consumption. Therefore, when giving factual feedback,
next to the number indicating the current energy
consumption level, two additional numbers were presented
indicating low and high consumption. Thereby the amount
of information present in lighting feedback and factual
feedback is comparable.
As argued above, we expect that feedback through lighting
has stronger persuasive effects (leading to lower energy
consumption) than factual feedback. In addition, we
expected that lighting feedback would be easier to process.
To test this, we manipulated cognitive load: Half of the
participants performed an additional task. We expected that
for participants processing factual feedback, performing
this additional task would interfere with the persuasive
effects of that feedback, leading to more energy
consumption than without the additional task. At the same
time, we expect that for participants processing lighting
feedback, performing this additional task would not
interfere with the persuasive effects of that feedback,
leading to the same energy consumption as without the
additional task. Also, we expected that for participants
processing factual feedback, performing this additional task
would lead to slower processing of that feedback, while for
participants processing lighting feedback, performing this
additional task would not lead to slower processing of that
feedback.
METHOD
Participants and Design
Fifty-seven participants (39 men and 18 women) were
randomly assigned to one of the four cells of a 2 (feedback
123
type: lighting feedback versus factual feedback) x 2
(cognitive load: load vs. no load) experimental design. All
participants were student at Eindhoven University of
Technology, were recruited on campus to participate in a
study on ‘How to program a heating thermostat’, and
received ! 5 for a participation of 30 minutes.
(indicating consumption of a medium level, halfway
between low and high) or a color between white and red
(indicating high consumption).
Procedure and Materials
Upon arrival, participants were seated in front of a
computer. For all participants, a simulated programmable
thermostat panel was presented on the computer screen (see
Figure 1). This heating thermostat was modeled to look like
a commercially available heating thermostat. It contained a
virtual LCD display (with a background that was always
green) on which all relevant information and clickable
buttons were presented. For participants in the lighting
feedback condition, a computer-controlled power-led lamp
was positioned behind the participants' desk that reflected
its lighting on the wall behind the desk (see Figure 2). For
participants in the factual feedback condition, next to this
thermostat panel we presented a number indicating the
participant’s energy consumption in Watts, and also two
numbers indicating low and high consumption levels in
Watts.
Figure 1 -- The simulated programmable
thermostat panel
More specifically, for each of the ten scenarios (described
below) we calculated a low consumption score in Watts
(based on a setting of 17°C in relevant rooms) and a high
consumption score in Watts (based on a setting of 26°C in
all rooms). In the lighting feedback condition, these
numbers were used to determine the lighting color. That is,
when a participant’s energy consumption caused by his or
her setting of the thermostat were at the low consumption
level or lower, the lamp was given a completely saturated
green color, and when energy consumption was at the high
level or higher, the lamp was given a completely saturated
red color. When a participant’s thermostat settings lead to
an energy consumption in between the low level and the
high level, the light the lamp emitted was set to a color
between green (indicating low consumption) and white
124
Figure 2 – Feedback through lighting on the
wall behind the monitor
After general introductions, participants were asked to
program the programmable thermostat in ten different
tasks. Also, all participants were given two specific goals to
strive for while programming the thermostat. First, they
were instructed to strive for optimal comfort levels within
each specific task. More specifically, they were asked to
“program the programmable thermostat such that your
house would be comfortable to live in.”1 Second,
participants were instructed to use as little energy as
possible. That is, they were told that heating your house
costs energy (fuel) and diminishing the level of the
temperature settings for specific rooms would lead to lower
energy consumptions. We included the first goal to
motivate participants to use energy (to heat the house to
comfortable levels). Had we only included the second goal,
all participants might have chosen to use as little energy as
possible by simply not turning the heating on at all, and any
feedback about energy consumption would have been
irrelevant.
Next, the thermostat and the energy consumption feedback
(factual or ambient) it provided were explained. In each
task, participants were instructed to program the thermostat
for a different scenario. For this, we used 10 different, short
scenario descriptions (e.g., “It is evening and you are
having a party at home tonight”, “It is night and you are
going to bed. It is -10°C outside”, “On a Sunday afternoon
you are at home and outside temperature is 18°C”). In each
1
As in real-life programming of programmable heating
thermostats, participants did not experience physical
effects of changes (e.g., changes in heat) during the
programming tasks. So, participants had to judge the
comfort level corresponding to their settings of the
thermostat based on earlier experiences and their current
settings.
task, one of the ten scenarios was displayed above the
programmable thermostat panel. Scenarios were drawn
randomly from the set of ten and each scenario was used
only once. Participants received feedback after each change
of settings, until they pressed the “ready” button. For each
task, we registered the energy consumption corresponding
to the final setting, and the total amount of time a
participant used for that task.
and 14 received factual feedback) to an identical 2
(feedback type: lighting feedback versus factual feedback)
x 2 (cognitive load: load vs. no load) ANOVA. This
analysis showed results completely comparable to the
previous one: a main effect of feedback type, F(1, 39) =
4.63, p < .05, but no interaction of feedback type and
cognitive load nor a main effect of cognitive load, both F’s
< 1.
Participants in the cognitive load conditions performed an
additional task while setting the thermostat. This task was
comparable to cognitive load tasks used in earlier research
(e.g., [22]). Participants heard numbers (one to thirty) read
out aloud on headphones. As a manipulation check, we
registered the number of correct responses (pressing the
space bar after an odd number). Finally, participants were
debriefed and thanked for their participation.
Finally, to assess whether lighting feedback would be
easier to process, we analyzed the time it took these
remaining participants to program the thermostat. This
dependent variable was calculated by averaging the times
they needed on each of the 10 tasks. This analysis showed
the expected interaction of Feedback Type X Cognitive
Load, F(1,39) = 7.20, p = .011 (see Figure 4). Further
analyses indicated that participants who received factual
feedback needed more time to program the thermostat
under cognitive load (M = 55.0 seconds, SD = 15.1) than
without cognitive load (M = 38.7 seconds, SD = 7.0), F(1,
40) = 6.02, p = .019, whereas this difference was not found
for participants who received lighting feedback, F<1. In
general, programming the thermostat using lighting
feedback was faster (M = 39.3 seconds, SD = 8.0) than
when using factual feedback (M = 44.1 seconds, SD =
12.7), F(1,41) = 9.24, p < .01.
RESULTS
Averaged energy consumption scores (over the 10 tasks)
were submitted to a 2 (feedback type: lighting feedback
versus factual feedback) x 2 (cognitive load: load vs. no
load) ANOVA. As expected, participants who had received
feedback through lighting used a lower amount of energy
on average on the tasks (M = 544 Watt, SD = 208) than
participants who received factual feedback (M = 692 Watt,
SD = 202), F(1,53) = 7.16, p = .01 (see Figure 3). This
analysis did not indicate the expected interaction of
Feedback Type X Cognitive Load, F < 1. Also, this
analysis did not show a main effect of cognitive load, F <
1.
Finally, we also explored the effect of cognitive load on
energy consumption scores, but found no significant results
of cognitive load, all F’s<1.
Figure 3 – Energy consumption by type of feedback
However, the manipulation check of the cognitive load task
indicated that approximately half of the participants in the
cognitive load conditions had not performed the load task
in line with instructions (had pressed the space bar for less
than 10% of odd numbers). Therefore, to assess whether
the effect of feedback type on energy consumption was
qualified by cognitive load (indicated by an interaction of
feedback type x cognitive load), we submitted the average
energy consumption scores of the remaining participants
(14 in the load conditions, of whom 7 received lighting
feedback and 7 received factual feedback, and 29 in the no
load conditions, of whom 15 received lighting feedback
Figure 4 – Time to program thermostat by type of
feedback and cognitive load
DISCUSSION
Results indicated that participants who received feedback
through lighting used less energy in thermostat
programming tasks than participants who received factual
feedback. Thereby, the current research suggests that
lighting feedback can have stronger persuasive effects than
factual feedback (approximately 27%). Also, the current
results suggest that for participants processing factual
125
feedback, doing an additional task led to slower processing
of that feedback. For participants processing lighting
feedback, results suggest that adding cognitive load did not
lead to slower processing. This finding fits our suggestion
that lighting feedback is more easy to process and use in
goal-striving processes than factual feedback.
In contrast to expectations, the current results did not show
evidence for effects of cognitive load on energy
consumption, or of different effects of cognitive load on
energy consumption for participants who received lighting
feedback compared to those who received factual feedback.
An important reason for this might be that the setup of the
current study may not have been ideal for finding such an
effect because of the lack of time constraints when setting
the thermostat. That is, because there were no time
constraints, participants who received factual feedback and
who performed an additional task, may have been able to
use more time to set the thermostat (and did so, as indicated
by the analysis of response times). It seems quite
straightforward that these participants used this additional
time to process the factual feedback. Thereby these
participants may have processed the factual feedback well,
even though they also had to spend processing capacity on
the additional task. Future research might continue the
investigation of whether cognitive load can increase energy
consumption for factual feedback. Importantly, the current
results indicate that setting time constraints might be
important to find those effects.
Another possibility is that cognitive load could exert an
effect on energy consumption even without time
constraints, especially in a goal-setting paradigm, since it
leaves less cognitive capacity for considering the various
goals (i.e., ‘comfort’ and ‘energy saving’). Cognitive load
might make people forget secondary goals (‘energy saving’
would often be considered secondary), or process
additional cues (e.g., light feedback) in a more peripheral
rather than central way. Interestingly, both paths would
have implications for the most optimal design of feedback
cues, and future research could investigate both pathways.
Future research might also investigate using other forms of
cognitive load. That is, because the current cognitive load
task contained numerical elements (as participants had to
identify odd numbers in a spoken list of numbers), it could
have interfered especially with processing the factual
feedback because that also consisted of numbers (indicating
energy consumption). Therefore, cognitive load may not
have been equal in both cognitive load conditions. That
said, we argue that the numbers in the current load task
were only of secondary importance, as the main task
participants had to do was to identify specific element in an
array of elements (and these could just as easily have been
arrays of letters, in which participants would have to
identify consonants). In line with this argument, theories
that account for effects of information processing demands
generally do not identify different effects of processing
demands caused by different types of information (for an
126
overview, see [16]). So, these theories would not predict
fundamentally different mental load effects of a cognitive
load task that consisted of a more numerical load task
versus another type of load task. Likewise, cognitive load
theory [21] indicates that limitations of human cognitive
processing become especially pronounced when dealing
with complex tasks [4]. Based on cognitive load theory, we
argue that adding an additional task (our load task, which
indeed contained numbers) could have revealed limitations
of cognitive processing also in the lighting feedback
condition, independent of the specific nature of that
additional task. In other words, because our load task
added to the complexity of the overall task participants in
the lighting feedback conditions had to perform, it therefore
could have revealed limitations of cognitive processing.
And indeed results did not indicated these limitations (in
terms of slower processing) in lighting feedback conditions,
but only revealed these limitations (slower processing) in
factual feedback conditions. Still, future research
replicating the current findings with different cognitive
load tasks would certainly strengthen the evidence for our
argument that lighting feedback is easier to process and use
in goal-striving processes than factual feedback.
Furthermore, future research could also investigate which
other differences between lighting feedback and factual
feedback may underlie the stronger persuasive effects of
lighting feedback in addition to the higher ease of
processing of lighting feedback that the current research
suggests. For instance, lighting feedback might be more
conspicuous, have specific physiological consequences, or
may have stronger emotional or moral effects.
Overall, the current research indicates that diffuse lighting
can be used successfully as persuasive technology. These
technologies can be incorporated into everyday life in many
forms to change different types of behavior or attitudes. For
example, the data about energy consumption provided by
smart meters might be used to deliver interactive lighting
feedback in the living room. The current research suggests
that such an application could successfully influence
energy consumption behavior, even when users do not
spend cognitive attention to this lighting feedback. The
current research indicates that lighting can have a particular
aptitude as a medium for persuasive communications. Next
to being very cheap, or easy to install (and other fitting
characteristics, as discussed in the Introduction), the current
research suggests that persuasive lighting can have stronger
persuasive effects than other forms of persuasion (i.e.,
factual persuasion), especially under (day-to-day)
circumstances of high cognitive load.
In addition, we argue that lighting has specific qualities that
make it particularly suitable for providing user feedback.
For example, lighting can be very cheap, is easy to install,
lighting can be very energy friendly, lighting can be seen
by other people present in a room as well (inducing social
pressure as a persuasive mechanism), and lighting might
have an emotional appeal or even direct emotional effects.
Also, the low conspicuity of light and color changes sets
lighting apart from other feedback mechanisms.
Furthermore, lighting can be calm (in the sense of ‘calm
computing’). Other feedback mechanisms often lack these
characteristics. For example, feedback mechanisms like
factual feedback or feedback that uses sound, smell, or
tactile feedback cannot easily be calm in that sense.
Therefore, we argue that lighting can be particularly suited
as a persuasive agent.
In general, persuasive technologies are generic
technologies which are “intentionally designed to change a
person’s attitude or behavior or both” [7, see also, 12].
Based on current results, we argue that lighting in various
modalities can serve as Ambient Persuasive Technology
[see also 6, 8, 10, 11, 19]. We propose that Ambient
Persuasive Technologies are generic technologies that are
intentionally designed to change a person’s attitude or
behavior or both, that can be integrated unobtrusively into
the environment and exert an influence on people without
the need for their focal attention. The current research
suggests that ambient persuasive technology can have
important advantages over more focal persuasive
technologies without losing its persuasive potential.
ACKNOWLEDGMENTS
We thank the Persuasive Technology Lab Group for
comments and ideas, and Martin Boschman for technical
assistance.
Pervasive
Technology
and
Environmental
Sustainability, Workshop at the Sixth International
Conference on Pervasive Computing (Pervasive 2008),
Sydney, May 19–22.
8. Fogg, B. J. Persuasive Technology: Using Computers to
Change What We Think and Do. Morgan Kaufmann,
San Francisco, 2003.
9. Froehlich, J. (2009). Promoting energy efficient
behaviors in the home through feedback: The role of
human-computer interaction. Paper presented at
HCIC2009.
10. IJsselstein, W., De Kort, Y., Midden, C., Eggen, B., &
Van den Hoven, B. (2006). Persuasive technology for
human well-being: Setting the scene. In: IJsselstein, de
Kort, Midden, Eggen, van den Hoven (Eds.): Persuasive
Technology. First International Conference on
Persuasive Technology for Human Well-Being
(Proceedings), Berlin, Heidelberg and New York:
Springer, p. 1-5.
11. Martinez, M. S., & Geltz, C., R. (2005). Utilizing a preattentive technology for modifying customer energy
usage. Proceedings of the European Council for an
Energy-Efficient Economy 2005.
REFERENCES
12. Martinez, M.S. 2006. “Residential demand response
technologies: A consumer’s guide.” Presentation at
National Town Meeting and Symposium on Demand
Response.
1. Aarts, E. H. L., Markopoulos, P., Ruyter, B. de (2007).
The persuasiveness of ambient intelligence. In M.
Petkovic, & W. Jonker (Eds.), Security, privacy, and
trust in modern data management. (pp. 367-381) Berlin,
Germany: Springer.
13. Mathew, A. P. (2005). Using the environment as an
interactive interface to motivate positive behavior
change in a subway station. CHI ’05 Extended Abstracts
on Human Factors in Computing Systems, pp. 1637–
1640.
2. Abrahamse, W., Steg, L., Vlek, C., & Rothengatter, T.
(2005). A review of intervention studies aimed at
household
energy
conservation.
Journal
of
Environmental Psychology, 25, 273-291.
14. McCalley, L. T., & Midden, C. J. H. (2002). Energy
conservation through product-integrated feedback: The
roles of goal-setting and social orientation. Journal of
Economic Psychology, 23, 589-603.
3. Arroyo, E., Bonanni, L. & Selker, T. Waterbot:
Exploring feedback and persuasive techniques at the
sink. In Proceedings of the SIGCHI Conference on
Human Factors in Computing Systems (CHI 2005),
pages 631–639, 2005.
15. Midden, C.J.H., Kaiser, F.G., & Mccalley, L.T. (2007).
Technology's four roles in understanding individuals'
conservation of natural resources. Journal of Social
Issues, 63, 1, 155-174.
4. Ayres P., & van Gog, T. (2009). State of the art
research into Cognitive Load Theory. Computers in
Human Behavior, 25, 253-257.
5. Bargh, J. A. (1997). The automaticity of everyday life.
In R. S. Wyer, Jr. (Ed.), The automaticity of everyday
life: Advances in social cognition (Vol. 10, pp. 1-61).
Mahwah, NJ: Erlbaum.
6. Becker, L. J., & Seligman, C. (1978). Reducing airconditioning waste by signaling it is cool outside.
Personality and Social Psychology Bulletin, 4, 412-415.
7. Davis, J. (2008). Towards participatory design of
ambient persuasive technology. Presented at Persuasive
16. Nietfeld, J. L., Finney, S. J., Schraw, G., & McCrudden,
M. T. (2007). A test of theoretical models that account
for information processing demands. Contemporary
Educational Psychology, 32, 499-515.
17. Riva, G., Vatalaro, F., Davide, F., Alcaniz, M. (eds.)
Ambient Intelligence. IOS Press, Amsterdam (2005).
18. Roth, K., & Brodrick, J. (2008). Home energy displays.
ASHRAE Journal, July 2008.
19. Schmidt, A. (2005). Interactive Context-Aware Systems
Interacting with Ambient Intelligence. In G. Riva, F.
Vatalaro, F. Davide, M. Alcaniz, (eds.). Ambient
Intelligence. IOS Press, Amsterdam.
127
20. Stein, L.F. 2004. “California Information Display Pilot:
Technology Assessment.” Final Report by Primen to
Southern California Edison.
21. Sweller, J. (1988). Cognitive load during problem
solving: Effects on learning. Cognitive Science: A
Multidisciplinary Journal, 12, 257-285.
22. Winter, L., Uleman, J. S., & Cunnif, C. (1985). How
automatic are social judgments? Journal of Personality
and Social Psychology, 49, 904-917.
128
23. Wisneski, C., Ishii, H., Dahley, A., Gorbet, M.G.,
Brave, S., Ullmer, B., & Yarin, P. (1998). Ambient
Displays. In, N. A. Streitz, S. Konomi, H. J. Burkhardt
(eds.). CoBuild 1998, LNCS, vol. 1370. Springer,
Heidelberg.
A Transformational Approach to
Interactive Lighting System Design
Philip Ross, Kees (C.J.) Overbeeke, Stephan (S.A.G.) Wensveen & Caroline Hummels
Department of Industrial Design
Eindhoven University of Technology
Den Dolech 2, 5612 AZ Eindhoven, the Netherlands
+31 40 2475966
p.r.ross@tue.nl
ABSTRACT
Light affects our behaviors and experiences. Research into
this field mainly focuses on the effects of lighting
conditions on people. The current paper focuses on human
interaction with lighting systems, and the way this
interaction transforms people’s behaviors and experiences.
Technological developments, such as Solid State Lighting
and increasingly powerful and economic sensing and
control electronics, open up a myriad of possibilities for
incorporating interactivity and intelligence in lighting
systems design. How can we design an interactive lighting
system that influences people’s behaviors and experiences
in a positive way? This paper explores this area from an
industrial design research point of view. It introduces a
transformational approach to interactive lighting design,
combining frameworks of Technological Mediation,
Human Values and Kansei design. In a research-throughdesign process, a set of interactive lighting systems are
designed based on this transformational approach and
empirically evaluated. Results indicate that it is indeed
possible to invite specific behaviors and experiences
through interactive lighting system design.
Keywords
Interactive lighting systems, transformational design,
human values.
INTRODUCTION
A growing body of research studies how light influences
human behaviors and experiences. Such research mainly
focuses on the effect of specific artificial lighting
conditions on people, e.g., [8], [12] and [14]. But artificial
lighting becomes ever more dynamic. Technological
developments, such as Solid State Lighting, and
increasingly small, cheap and powerful sensing and control
electronics, open up new possibilities for incorporating
interactivity and intelligence in lighting systems design [4].
Increasingly intelligent lighting systems are envisioned to
integrate into the everyday environment, playing a role in
everyday life that goes well beyond task lighting [1][9]. In
view of these developments, the current paper focuses on
human interaction with lighting systems and the way this
interaction affects behaviors and experiences, rather than
on the influence of given lighting conditions on people. Our
focus on interaction entails that we treat situations in which
lighting systems and humans respond to each other’s
actions in a meaningful way. These lighting systems are
typically equipped with electronics that enable them to
sense human actions, process the data, and respond
accordingly with lighting actuators. How can we design
interactive lighting systems that influence people’s
behaviors and experiences in a positive way? The current
paper explores this question from an industrial design
research point of view.
Technological mediation, ethics and light
The theory of Technological Mediation [13] is used in the
current research to conceptualize the influence of
interactive light on our behaviors and experiences. The
theory states that every technology in use transforms our
experiences and behaviors. This transformation has a dual
structure. Each technology on the one hand amplifies
specific experiences, and on the other hand reduces others.
Compare for example how an mp3 player amplifies the
experience of music and reduces the experience of the
environment, by immersing the listener in music and
blocking other sounds. The theory also states that
technology in use always invites specific behaviors while
inhibiting others. The mp3 player, when used in a busy
train, invites a person to concentrate on his work, while at
the same time it inhibits social interaction with people in
the vicinity. These mechanisms can also be applied to
interaction with lighting systems. When we do this, the
question arises for designers of interactive lighting systems
what experiences their system should amplify or reduce,
and what behaviors they should invite or inhibit. This
question has an ethical dimension: People with different
ethical beliefs might prefer to engage in different behaviors
and might prefer to have different experiences in a given
context.
A research-through-design process
This paper presents design research that explores how to
design interactive lighting systems that aim to invite
specific behaviors in interaction. We call this approach to
lighting system design transformational. In a researchthrough-design process [3][5], actual lighting systems are
129
designed using a combination of design techniques and
auxiliary theoretical frameworks. The aim of these lighting
systems is to invite specific behaviors in human-system
interaction. These designs are evaluated in an empirical
study. Central in the current process is design work from a
40-hour bachelor course called Personality in Interaction
[10], conducted at the department of Industrial Design at
Eindhoven University of Technology [6]. In this course,
students designed interactive lighting systems with the aim
to invite behaviors that fitted the personality of a specific
fellow student.
A framework for ethical beliefs
Before elaborating on the course, we treat an auxiliary
theory that was used to operationalize people’s ethical
beliefs, namely the theory of Human Values [11]. This
theory offers a way to understand what kind of behaviors
and experiences a specific person would desire to engage
in. Human values are defined as follows: ’Values (1) are
concepts or beliefs, (2) pertain to desirable end states or
behaviors, (3) transcend specific situations, (4) guide
selection or evaluation of behavior and events, and (5) are
ordered by relative importance’ [11]. Examples of values
are Creativity, Helpfulness and Social Power. Empirical
research in 20 countries identified a set of 57 values
considered near-universal. This research allowed Schwartz
to meaningfully locate the 57 values on a plane with four
quadrants, labeled Self-Enhancement, Conservation, SelfTranscendence and Openness-to-Change. Figure 1 shows a
selection of 13 of the 57 values plotted on this plane. In this
value scheme, the distance between values represents their
mutual compatibility. Figure 1 shows, for example, that the
closely located values Helpful and Loyal are more
compatible than Helpful and Social Power. The behaviors
these values motivate are compatible (or not) in a similar
manner. Schwartz developed a survey to measure
individual people’s value priorities. The instrument is
called the Schwartz Value Survey [11] and consists of the
57 value items that can be scored on a 9-point scale.
A large body of research exists that relates people’s value
priorities to certain behaviors, attitudes and personalities.
Several research projects demonstrate the relevance of
Human Value theory to design research. For example,
Allen and Ng [2] show how values could be related to
choice for products as varied as different sunglasses and
different cars. The fact that values guide selection and
evaluation of behaviors connects ethical beliefs of people
and specific kinds of behaviors. The definitions of values
can serve as a characterization of desired behaviors a
lighting system should invite. For example, for people that
value creativity, we could aim to design an interactive
lighting system that invites creative behaviors.
Figure 1: 13 out of 57 value items arranged according to
the research of Schwartz and placed in the four quadrants
(adapted from [11]). The distance between values indicates
motivational compatibility.
DESIGNING INTERACTIVE LAMPS: THE PERSONALITY
IN INTERACTION COURSE
Research into the influence of interactive lighting systems
on human behavior and experience requires evaluation of
actual lighting systems. These lighting systems were
designed and built in the Personality in Interaction course.
The students’ design assignment was to create an
interactive lamp or lighting system that invited behaviors
and experiences that corresponded to the most important
values of a fellow student. So if a fellow student prioritized
Creativity highly, the assignment was to create an
interactive lighting system that invited creative behaviors
from the person interacting with it. Note that the
assignment was not to create a lamp that acted creatively
itself: It was about inviting creative behaviors from the
person interacting with the lamp. The lamps did not need to
be functional in the sense of providing task lighting.
Course set-up
The course’s design process followed a Kansei design
approach [7] that was adapted for this specific course. It
included the following steps:
1.
2.
3.
4.
130
Students (voluntarily) completed the Schwartz
Value Survey [11] to learn about their own
personality. Pairs of students with contrasting
personalities were created with the test results.
Relevant theories (Human Value theory, Kansei)
were introduced in a lecture and students read
accompanying papers.
The students created a one-minute ‘dynamic
personality collage’ on video of their assigned
fellow student. This collage had to display
behaviors of the fellow student that expressed his
or her values.
The personality collages were analyzed to find
interaction qualities for design.
5.
6.
The next step was to design and prototype an
interactive living room lamp or lighting system
that invited behaviors that related to the fellow
student’s top priority values.
The course ended with a final presentation, in
which the students interacted with the prototypes
designed for them, and the design and design
process were evaluated.
Resulting lighting designs
This section treats three designs resulting from the course,
to illustrate the nature of the design work. See Figure 2 to 4
for images of the lighting system interactions and
accompanying explanations. Film clips of these lamps and
the other nine lamps used in the current research are
available at http://www.philipross.nl/thesis.
Figure 2: This staircase lighting system targets Creativity
related behaviors. It consists of several light balls hanging
from the ceiling above the staircase. When the balls are
moved, they light up and create a dynamic light and
shadow play in the staircase. The balls stick to each other
with magnets when they touch, allowing a person to
rearrange the layout of light balls as desired. The system’s
easy interaction, combined with the beautiful, dynamic
light and shadow effects that each action creates, invites a
person to be creative while walking the stairs.
Figure 3: This decorative lamp is designed to invite curious
behavior. The lamp’s main interaction elements are three
semi-transparent light cubes, placed in a cubic space
delimited by three mirrors. The cubes are equipped with
colored LED’s but do not give away their lighting effects
until they are combined with each other. Different ways of
stacking or aligning the cubes result in different dynamic
colored lighting effects. The lamp triggers curiosity in
interaction through its intentional absence of feedforward
for actions, combined with the reward of beautiful effects
after each interaction.
Figure 4: The Throw Ball light object targets the value
Pleasure. This design is conceived for a person that likes to
have fun in social setting. The final design is a ball the size
of a soccer ball with holes in it that transmit light. The ball
tries to stimulate people to throw it by blinking when it is
held longer than 0.5 seconds. When it is thrown, it lights up
fully. When held longer than 2 seconds, the light dies out
which could mean the game is over.
THE EVALUATION EXPERIMENT
An evaluation experiment was conducted to see how people
naïve to the design intentions would experience the
interactive lighting systems. In this experiment, participants
viewed film clips of interactions with twelve different
lamps (including one trial) and rated them in terms of
values. Twenty people participated, thirteen male and seven
female. All participants were architecture students, coming
from both the bachelor and the master program.
131
Architecture students were chosen since they have no
education in interaction design, but are still sensitive to
design in general.
Procedure
The experiment procedure was as follows:
1.
The participant received an introduction in which the
experiment was explained.
2.
A participant watched a film clip showing interaction
with a given lamp.
3.
The participant filled out a value rating form. Details
about this form are treated further on in this paper.
4.
Step two and three were repeated for all eleven film
clips, preceded by a trial clip.
There were 8 separate sessions with 1 to 5 participants
simultaneously. The clips were show in three different
orders. Order 1 and 3 were randomized, order 2 was
counterbalanced with order 1. The participants received
!5,-.
Stimuli
The designs from the Personality in Interaction course were
only partly functional prototypes. It was impossible to test
them live with participants in an experiment, so film clips
of these interactions were shown to the participants. In
these film clips, the prototypes seemed to be truly
interactive.
A set of eleven lamps (plus one for the trial clip) served as
the stimuli. Two of these lamps were not explicitly
designed for a value. The students that designed these
lamps deviated from the course assignment, and used other
personality traits as input. These lamps were still included
in the study to explore how they would be rated in terms of
values. Ideally, each of the four quadrants of the Schwartz
Value Structure was targeted by at least one lamp. This
could however not be realized. There were only a few
course students with highest priority values in the
‘Conservation’ quadrant or the ‘Self-Transcendence’
quadrant. So these values were rarely targeted in the
course. The result was that there were no usable designs
targeting the Conservation and Self-Transcendence
quadrants. Explanations and pictures of all eleven lamp
interactions and the trial lamp interaction are available in
[9].
One of the clips was selected as the trial clip. The clip
duration ranged from 15 seconds to 39 seconds.
Screenshots of these clips are shown in Figure 2 to 4. The
clips were numbered and shown on a 37’’ Flat Screen TV.
Rating form
To measure the way people characterized the interactions in
terms of values, a rating form was devised including a list
of Human Value rating scales. The form was originally
created in Dutch, but treated here in English translation.
The participant was asked to imagine they would interact
with the lamp themselves. Then they placed a tick mark on
the value scale to indicate to what extent a particular value
132
description matched the interaction in the film clip. The
value scales looked like this:
Imagine you are interacting with the lamp yourself. Use a
tick mark to indicate to what degree the interaction evokes
the following terms in you:
Creativity (uniqueness, imagination)
Does not
describe it
at all
o
o
o
o
o
o
o Describes
it perfectly
The value descriptions used in the scales were copied from
the value descriptions in the Schwartz Value Survey [11].
A selection of 13 of the 57 values was made to include on
the form, to keep the rating task feasible for the
participants. These selected values were spread out over all
four quadrants of the value plane. Furthermore, the list
contained all the values that were targeted by the selection
of lamps. The value rating list contained the following
items:
•
Inner harmony (at peace with myself)
•
Curious (interested in everything, exploring)
•
Humble (modest, self effacing)
•
Freedom (freedom of action and thought)
•
Social power (control over others, dominance)
•
Capable (competent, effective, efficient)
•
Pleasure (gratification of desires)
•
Loyal (faithful to my friends, group)
•
Politeness (courtesy, good manners)
•
An exciting life (stimulating experiences)
•
Sense of belonging (feeling that others care about me)
•
Creativity (uniqueness, imagination)
•
Helpful (working for the welfare of others)
The distribution of the corresponding values over the 2D
structure is depicted in Figure 1.The forms were filled in on
a laptop running SPSS Data Entry Station.
Hypotheses
If the design of the lamps has any effect measurable with
the value scales, the ratings on the value scales should
differ between lamps targeting different values. Formally
put:
Hypothesis 1
H0: The mean ratings on the value scales are equal
between lamps
H1: The mean ratings on the value scales are not equal
between lamps
This effect should have a certain pattern for the lamps that
targeted a specific value. One would expect that a target
value would always have a significantly higher score on the
scales than all other values. This leads to the second
hypothesis.
Hypothesis 2
H0: The mean rating of the target values are not higher
than those of all other values
H1: The mean rating of the target values are higher than
those of all other values
Human value theory predicts a structure in the relation of
the score of the target value scale to the scores of the other
value scales. As treated earlier in this paper, the mutual
distance of value items on Schwartz’ value structure is a
measure of ‘motivational compatibility’. If two values are
located close to each other on the value structure, they are
compatible. The larger the distance between them, the less
compatible they are. For example, the values Helpful and
Loyal (closely co-located) are more compatible than
Helpful and Social Power (large distance in between). See
the locations of these values in Figure 1. This degree of
compatibility between values is expected to have a
systematic effect on the scores on the value scales. For
example, if a lamp in the current experiment succeeds in
eliciting the value Helpful, the value scale Helpful would
receive the highest mean scores. The value scale Loyal (the
most compatible value in this experiment) would receive
the second highest score, and the value scale Social Power
(the least compatible value) would receive the lowest score.
So it is possible to determine a theoretical rank order of the
means of all value scale scores, based on the targeted value
score. The occurrence of this rank order in the data would
be an indication that the ratings are in line with value
theory and that the interaction is really relevant in terms of
values. The ‘fit’ of the measured rank order of value scale
scores with the theoretical rank order of scores is
determined here by a correlation analysis of both rank
orders. Put in terms of a hypothesis:
Hypothesis 3
H0: The correlation between the measured and theoretical
rank orders of the value scores is not significant
H1: The correlation between the measured and theoretical
rank orders of the value scores is significant
Figure 5: The mean ratings of the three lamp designs
explained in this paper. The values are placed in order
according to the value structure quadrants along the x-axis.
The vertical lines indicate the borders of the quadrants.
Each lamp’s target values are highlighted with a large,
filled dot.
Results for Hypothesis 1:
H1: The mean ratings on the value scales are not equal
between lamps
Figure 5 show differences between the scores on the value
scales. An 11 (Lamp) x 13 (Scale) repeated measures
Analysis of Variance (ANOVA) was performed on scores
for the value scales for all 11 lamps. The results are
reported in Table 1. Significant main effects were obtained
for Lamp, F(10, 2717) = 7.7, p < .001, and for Scale, F(12,
2717) = 47.7, p < .001. In addition, the interaction effect
was significant, F(120, 2717) = 2.2, p < .001. Simple main
effects analyses (Dunnett T3) were performed to examine
the nature of the significant interaction. It was found that
the means of 9 of 11 lamps were significantly different
from one or more of the other lamps’ means. The
conclusion is that H(0) is rejected. (Note: Homogeneity of
variance could not be assumed. Non-parametric test, the
Friedman Two-way Analysis of Variance by Ranks and
Kruskal-Wallis tests were performed on the value scale
scores. The same significant effects were obtained from
these tests.)
Results
Figure 5 shows the ratings of the three lamps treated in the
current paper. Most of the evaluated lamps targeted values
in the Openness to Change quadrant. This shows in the
ratings. The highest scores are generally located in the
Openness to Change quadrant. This section continues with
a treatment of the three hypotheses in light of the
experiment results.
Table 1: Results of the ANOVA. Independent Variables are
Lamp and Scale, the Dependent Variable is Score.
Source
Type III df
Mean F
Sum
of
Square
Squares
Lamp
202.8
10
20.3
7.7
Scale
1515.9
12
126.3 47.7
Lamp * Scale 704.4
120
5.9
2.2
Error
7199.5
2717
2.7
Total
52975.0
2860
R Squared = 0.252 (Adjusted R Squared = 0.213)
Sig.
0.001
0.001
0.001
133
Table 2: Ranks of each lamp’s target value scores compared to the other values.
Lamp name
Target
value rank
Staircase
lighting
system
Mirror
Blocks
Flower
Lamp
Throw
Ball
High
Five
Segmen
ted Ball
Stacker
Lamp
Spring
Lamp
Puzzle
Lamp
Color
Box
Tree of
Light
2
2
2
1
n.a.
n.a.
1
2
1
3
5
Results for Hypothesis 2:
H0: The mean rating of the target values are not higher
than those of all other values
Nine of eleven lamps tested in this experiment actually
targeted a value. The other two designs targeted other
aspects of personality, since the designers deviated from
the course design brief. Three of the nine lamps targeting
values actually received the highest ratings on their target
value, i.e., Light Ball for Pleasure, Stacker lamp for
Freedom and Puzzle Lamp for Curiosity (See [9] for a
description all the experiment’s lamps). In four lamps, the
target value was rated second highest, one was rated third
and one was rated fifth. See Table 2 for an overview. In
almost all cases, H(0) cannot be rejected.
However, the target value is in most cases ranked second or
third. Value theory says that the values are part of a
motivational continuum. When values are located close to
each other in the structure, they are similar in motivation.
This means that behaviors motivated by a value very near a
target value are still highly compatible with the behaviors
motivated by the target value. An analysis considering the
order of the ranks of all values gives a more nuanced view
on how successful the lamps are, as explained for
hypothesis 3.
target value. And the structure of gradually increasing and
decreasing compatibility is present as well. The
approximate sinusoid lines in Figure 5 visually depict this.
The results of this analysis indicate that these lamps elicit
interactions that are actually relevant in terms of values.
Figure 6: Determining the first six rank orders for
Creativity. The circles indicate the different distances from
the values to Creativity. The circles have the Creativity
value as their centre, and have a radius that corresponds to
the distance to another value.
Results for Hypothesis 3:
H1: The correlation between the measured and theoretical
rank orders of the value scores is significant
To test whether the rank orders of the values as they are
rated are equal to the theoretical rank orders, based on their
mutual compatibility, a correlation analysis is conducted. In
this analysis, the scored rank orders are compared with the
theoretical rank orders. The theoretical rank orders are
calculated by determining the distance between the target
value and all other measured values on the structure. See
Figure 6 for a graphical representation of this process.
Table 3 shows the table of correlation coefficients.
The table shows that the value scores of 6 of 9 lamps that
target a value correlate significantly with the theoretical
rank orders. This indicates that the interactions they elicit
show the same ‘motivational structure’ as the values they
try to elicit. So although the target values are not in all
cases rated highest, the values that motivate similar
behaviors score higher than the values that conflict with the
134
Table 3: Correlations of scored value rank orders with
theoretical rank orders (all N=13). Continued on the next
page.
Correlations – Kendall’s tau
Staircase lighting
system
(Creativity)
Correlation
Coefficient
0.538
Sig. (2-tailed)
0.01
Mirror Blocks
(Curious)
Correlation
Coefficient
0.564
Sig. (2-tailed)
0.007
Correlation
Coefficient
0.641
Sig. (2-tailed)
0.002
Correlation
Coefficient
0.538
Sig. (2-tailed)
0.01
Flower lamp
(Creativity)
Throw Ball
(Pleasure)
Table 3: continued.
Stacker Lamp
(Freedom)
Spring Lamp
(Pleasure)
Puzzle Lamp
(Curious)
Colour Box
(Hedonism)
Tree of Light
(Self-Direction)
Correlation
Coefficient
0.641
Sig. (2-tailed)
0.002
Pearson Correlation
0.445
Sig. (2-tailed)
0.128
Correlation
Coefficient
0.513
Sig. (2-tailed)
0.015
Correlation
Coefficient
0.308
Sig. (2-tailed)
0.143
Correlation
Coefficient
0.359
Sig. (2-tailed)
0.088
character of the resulting lamps indicate that taking a
targeted value related behavior as an input for the design
process is a fruitful approach to come to innovation in
interactive lighting design.
On a general level, the results show the relevance and
potential of design research specifically directed at
interaction with lighting systems, taking the way they
transform our behaviors and experiences into account. The
current value-based transformational design approach can
help designers create lighting systems that influence our
behaviors and experiences in a positive way.
ACKNOWLEDGEMENTS
We would like to thank SeungHee Lee for her help setting
up the first run of the Personality in Interaction course and
Paul Locher for his methodological support. Many thanks
also to the students participating in the Personality in
Interaction course.
REFERENCES
Discussion of the experiment
The experiment results are encouraging. However, there are
reservations that need to be made. The lamps were tested
using video-clips of interaction. Experiencing an
interaction captured on video may be different than experiencing interaction live. It is unknown how this difference
manifests itself in the measurements. Because of the low
number of participants and their specific background,
caution is required in generalizing the results to a larger
population. All lamps in this test focused on values in the
Openness-to-Change quadrant and the Self-Enhancement
quadrant. It is therefore still unknown if values in the other
quadrants could be targeted. Although the rating form
makes use of the exact formulations of the Schwartz Value
Survey, it is not a validated measuring instrument.
GENERAL CONCLUSION AND DISCUSSION
The outcomes of this study indicate that it is possible to
design interactive lighting systems that invite behaviors
that relate to a specific range of values. ‘Range of values’ is
mentioned since the lamps in the experiment invite a range
of compatible values, rather than only one isolated value.
Quantitative analysis of the value scale scores indicated
that the behaviors and experiences invited by the lamps in 6
of 9 cases corresponded significantly to the values these
lamps targeted. The authors interpret the outcomes of the
study as a stimulus to continue this line of research. A
follow up research question is to see if people evaluate
lamps that invite behaviors that correspond to their own
high priority values more positively than lamps that invite
conflicting behaviors.
The theoretical frameworks of Technological Mediation
and Human Values serve as useful input for design, helping
designers define what they would like to achieve with their
interactive lighting system. The creative and novel
1. Aarts, E., and Marzano, S. The new everyday. Views on
ambient intelligence. 010 Publishers, Rotterdam, 2003.
2. Allen, M.W. and Ng, S.H. The direct and indirect
influences of human values on product ownership.
Journal of Economic Psychology 20, 1 (1999), 5-39.
3. Archer, B. The nature of research. Co-Design Journal, 2
(1995), 6-13.
4. Fred Schubert, E. and Kim, J.K. Solid-State Light
Sources Getting Smart. Science 308 (2005), 1274–1278.
5. Frens, J.W. Designing for rich interaction: Integrating
form, interaction and function. Unpublished doctoral
dissertation, Eindhoven University of Technology,
Eindhoven,
2006.
Available
at
http://www.richinteraction.nl.
6. Hummels, C.C.M. and Frens, J.W. Designing for the
unknown: a design process for the future generation of
highly interactive systems and products. Proceedings of
the 10th International Conference on Engineering and
Product Design Education (2008), 204-209.
7. Lee, S.H., Harada, A. and Stappers, P.J. Pleasure with
products: Design based on Kansei. In Green, W.S. and
Jordan, P.J. eds. Pleasure with products: Beyond
usability, Taylor and Francis, London, 2002, 219-230.
8. Partonen, T. and Lönnqvistc, J. Bright light improves
vitality and alleviates distress in healthy people.
Journal of Affective Disorders 57 (2000), 55-61.
9. Ross, P.R. Ethics and aesthetics in intelligent product
and system design. Unpublished doctoral dissertation.
Eindhoven University of Technology, Eindhoven, 2008.
Available at http://www.philipross.nl/thesis.
10. Ross, P.R. and Lee, S.H. Integrating Interactive Product
Design Research and Education: the Personality in
Interaction Assignment. In Rodgers, Brodhurst and
Hepburn eds. Crossing Design Boundaries (2005),
Taylor & Francis, 151-156.
135
11. Schwartz, S.H. Universals in the content and structure
of values: Theory and empirical tests in 20 countries. In
Zanna, M. Ed. Advances in experimental social
psychology, 25, Academic Press, New York, 1992, 165.
12. Veitch, J.A. and Gifford, R. Assessing beliefs about
lighting effects on health, performance, mood and social
behavior. Environment and Behavior 28, 4 (1996), 446470.
136
13. Verbeek, P.P. What things do - Philosophical
reflections on technology, agency, and design. Penn
State University Press, Penn State, 2005.
14. Wirz-Justice, A., Bucheli, C., Graw, P., Kielholz, P.,
Fisch, H.-U. and Woggon, B. Light treatment of
seasonal affective disorder in Switzerland. Acta
Psychiatrica Scandinavica, 74, 2 (1986), 193-204.
Effects of Colour and Light on Customer Experience and
Time Perception at a virtual Railway Station
Mark van Hagen
NS Commercie Dutch Railways
Utrecht, the Netherlands
+31 30 2357781
Mark.vanHagen@NS.NL
ABSTRACT
Various studies have shown that colour and light influence
our emotions and behaviour. In this paper the results will
be presented of research into the combined effects of 5
different colours and 2 different intensities of light for
Leiden station.
Two experiments in a virtual Leiden station show that
although colour and light are perceived subconsciously, the
combination of the two does in fact have significant effects.
Most of the passengers appeared to overestimate waiting
time on the platform, which concurs with results from
earlier fieldwork. Moreover, time would appear to pass
more quickly with low intensity lighting as opposed to high
intensity lighting.
The second experiment showed that passengers prefer
warm colours in combination with dimmed lighting and
estimate the waiting time as being shorter than when cooler
colours and a more intense lighting are used. Practical
implications will be discussed.
Keywords
Colours and lighting, railway station, virtual simulation,
customers’ evaluation
INTRODUCTION
In many public spaces, such as railway stations, shopping
malls and healthcare institutions, colours strongly
determine how we feel and act as service customers. As is
also the case with temperature, smell, sound and décor,
changing these environmental factors can influence both
perceptual and emotional reactions as well as the actual
behaviour [28].
For the Dutch Railways Corporation colour and lighting are
important instruments to manage the overall impression of
the service environment. They are expected to affect
customer experiences (via perceived pleasantness, feelings
of safety and even (waiting) time perception) and thus
customer satisfaction. The key question in this paper is:
How can Dutch Railways specifically deploy colour and
light on platforms in stations so as to positively influence
emotions? The objective is to win more happy customers
by improving the ‘servicecape’ [13] and its potential to
Mirjam Galetzka, Ad Pruyn, Joyce Peters
Marketing Communication and Consumer
Psychology
University of Twente
Enschede, The Netherlands
+31 53 4893329
m.galetzka@gw.utwente.nl
‘signal’ service quality. and customer care. Of course, in
considering the environmental design of public services
such as railway stations, many factors may play an
important role, and a variety of quality dimensions may be
affected by clever design. In this study we specifically
focused on the intangible factor ‘time’, because the train’s
departure is scheduled and passengers have to get on the
train in time. So time is one of the predominant processes
that passengers are dealing with during their stay at the
platform. Time also has a crucial impact in quality surveys
in public transport.
This paper presents the results of 2 studies into the
combined effects of colours and intensities of light on
emotions and time perception in a virtual simulation of
Leiden station.
SERVICE ENVIRONMENT
According to Parasuraman, Zeithaml and Berry [37], three
aspects play a role in the service environment: intangibility,
the simultaneous course of the production and
consumption, and the heterogeneity of the service. Through
the intangibility of the service, people cannot feel, taste, see
or smell it. They can only experience the service. Owing to
a lack of tangible proof, customers perceive other aspects
of the environment to evaluate the service and determine its
quality [2], [13], [14]. A service environment comprises all
the objective factors that can be controlled for by the
organization with the aim of prompting employees and
consumers to a specific behaviour. Baker [5] divides the
physical environment into three components: design
elements that are visually and tangibly present, ambient
elements that are intangible and often processed
subconsciously, and social elements, other people present
in the service environment, such as customers and
personnel. Colour and lighting belong to the category of
ambient and intangible elements of the environment.
STIMULUS, ORGANISM, RESPONSE
This research employs the model of Mehrabian and Russell
[32] to investigate whether colour and light influence the
degree of pleasure, arousal and dominance that determine
137
behaviour. The relationship between environmental
variables and approach or avoidance behaviour in a service
setting can be modeled (after Mehrabian and Russell, [32])
as a stimulus-organism-response (SOR) chain:
•
Stimulus (environment): all ambient aspects such
as colour, light, smell, sound etc.
• Organism (emotions): emotional reactions on the
basis of pleasure, arousal and dominance (PAD
model).
• Response (behaviour): the degree to which
consumers show approach or avoidance
behaviour.
Many studies have focussed on the influence of pleasure on
behaviour (e.g. [9]). Also the relationship between arousal
and behaviour has been empirically demonstrated.
However, little attention has been paid in literature to the
degree of dominance [9]. Imperative for a station
environment is a sense of control, and thus dominance;
likewise for emotional aspects such as feelings of
uncertainty and pressure, how easy it is to orient oneself
and how one experiences the wait. These aspects will thus
be included in this research.
LITERATURE OVERVIEW OF COLOUR AND LIGHT
Colours with a short wavelength are specified as cool
colours (blue and green), whereas those with a long
wavelength are warm (red and yellow). Light comprises the
light intensity and the spreading of the colour tone. Bright
or dimmed light is determined by the light intensity. Little
research has been conducted on the combination of colour
and light ([16]; [43]). The majority of (published) studies of
the effects of colour in retail environments was conducted
in laboratory settings. To our knowledge, no research has
yet been published on the usage of light and colour in a
railway station.
will exude a warm, festive aura whereas the same colour in
a hospital can have a negative influence on the state of
mind of the already anxious visitor.
Research on the use of colour in retail environments has
shown that it influences buying behaviour [9], purchasing
speed [9], time spent in the shop [9], pleasure [9], [18],
arousal ([18], image of shop and merchandise [8] [18] and
the potential to draw customers into the shop [8]. Blue and
green are perceived to be the most pleasant in a retail
environment [20], [26] and are also evaluated higher than
shops with a warm (orange) interior [8], [18]. The results
for pleasure strongly resemble the scores for arousal. From
research by Kwallek et al. (1988; in [42]), it appeared that
people who performed a business task in red surroundings
later scored higher for stress and anxiety. Colours with a
short wavelength cause a person to be more externally
oriented and to show forceful and extrovert behaviour.
From the study by Belizzi et al. [8], it appeared that
respondents, irrespective of their colour preference, felt
more drawn physically to warm colours yet perceived
surroundings in warm colours as less pleasant [8]. Warm
colours are apparently successful when it comes to drawing
people in (entrances, shop windows), but less so when it
comes to making them feel at ease. In situations where
people experience mental pressure, it is better to keep the
colours cool; with their calming effect people are prepared
to remain longer in such surroundings. Brengman [15; 16]
showed respondents photos of a shop in which the colours
were manipulated. She concludes that people will spend
more time and money in a shop if they find the colours
agreeable [15]. Blue and yellowish red are perceived as
pleasant, as are light colours. Such atmospheres invoke
approach behaviour and the desire to explore. According to
Brengman [15], red and yellowish green, just like bright
and dark colours are perceived as less pleasant; these
colours lead to tension and stress and cause a distasteful
feeling. Such negative stress leads to avoidance behaviour
[15].
Colour
In public environments there is often a need for the right
colour that incorporates the element of ‘pleasantness’.
Although the optimal design may strongly differ across
service contexts and situations (and even across individual
customers), it appears that specific colours, generally
perceived as pleasant, may result in very specific emotions.
Cool colours, such as blue and green, have a relaxing
effect, whereas colours with a long wavelength, such as
orange and red, are stimulating [1], [26], [43], [47]. Warm
colours are perceived as being protective [46]. Clear and
saturated colours are generally experienced as more
pleasant [24], but are also more strongly associated with
fear than cool colours [26]. Dark colours are perceived to
be more dominant and more strongly provoke hostility and
aggression. So, with the environment and state of mind
determining the effects of colour, red in the cinema foyer
138
Light
Psychologists state that light has a tremendous influence on
human behaviour. Baker and Cameron [5] and Küller et al.
[29] indicate that there is a basic level of how people
experience light as the most pleasant. A preference for light
intensity depends on the situation, the task and one’s
surroundings [7], [10], [43].
Light has a strong effect on the degree of arousal [7], [19],
[22], [27], [34]. Light also influences a shop’s image and
the stimulus to look at and scrutinize the merchandise [6],
[16].
Colour and light
Valdez and Mehrabian [43] have shown that it is not only
colour hue that determines the evoked emotions but also
the saturation and brightness (i.e. intensity) thereof. It
appears, for example, that although there is hardly any
difference in the way men and women react to colour,
women are more sensitive to the colours’ brightness. In a
study of non-chromatic colours (black-white-grey), it
appeared that the brightness strongly determines their
degree of stimulation and dominance [43]. Mehrabian
suggests that “brightly lit rooms are more arousing than
dimly lit ones” and that light, besides colour, has a strong
influence on arousal [32]. From a scenario study [4], in
which a blue and an orange shop were compared, it
appeared that the blue shop was preferred the most and that
it generated a greater willingness to shop or buy there. A
brightly-lit orange shop was perceived as having the
greatest adverse effect. However, when soft lighting was
introduced to this orange shop, it became almost as
positively rated as the blue one. With a blue shop the
effects are even more positive in a brightly-lit variation.
The combination of light and colour seem to qualify the
perceived effects quite convincingly. A restriction,
however, is that this was a scenario study and its results
should preferably also be tested in a realistic setting [4].
Generally speaking, the studies of the effects of colour have
predominantly focussed on the wavelength of the colour
and hardly at all on the brightness and the saturation of the
colour ([15], [43]). Light and colour combined were
seldomly investigated.
COLOUR BRIGHTNESS, LIGHT AND TIME PERCEPTION
Smets [41] demonstrated how people estimate the length of
an interval as being shorter after having seen a red as
opposed to a blue colour. Under red light, time would
appear to pass more slowly and objects seem bigger and
heavier, whereas under blue light time seems to pass more
quickly and objects look smaller and lighter. Casinos use
this information and opt for red as a basic colour which
excites the customers without their realizing that they are
spending a lot of time there [40]. Research into the waiting
time of downloading internet pages in various colours, with
different levels of saturation and brightness, revealed that
respondents felt more relaxed by particularly the bright
colours and that time seemed to pass more quickly.
Conversely, tension and stress when downloading seems to
slow down the subjective experience of time. Analogous to
other studies of the usage of colour, it would appear that
blue screens have a more calming effect than red or yellow
ones [23].
In the context of traditional –off-line- shopping, Markin et
al. [31] suggest that dimmed light calms customers, causing
them to move more slowly through the shop, which means
that they then take their time to pay attention to and
scrutinize the merchandise. This suggests that the
shopkeeper can use the intensity of light to keep customers
in the shop for a longer or shorter period of time. As
pleasant and stimulating colours combined with bright
lighting appears to lengthen the perceived waiting time [5],
it would be better to opt for softer lighting to prevent
overestimation of the actual wait.
STUDY 1: VIRTUAL LABORATORY
Method
A 2 (colour hue: red vs. blue) x 2 (light: high vs. low
intensity) between-subjects design was used. At the VR
Laboratory at the University of Twente, 130 participants
were asked to navigate through a virtual simulation of
Leiden station.
Procedure
The experiment ran for four executive days, during which
the different conditions were arbitrarily distributed among
the respondents. Participants who indicated they wished to
take part in the experiment were first subjected to a test for
colour blindness, after which they were invited -in a
separate room- to practise with the navigation system that
was used in the experiment. Subsequently the respondents
entered the VR lab where the final instructions were given.
After the simulation the respondent was requested to fill
out a questionnaire. Following completion, (s)he was
thanked for his/her time.
Participants
In total, 142 respondents, all Master/PhD students at the
University of Twente, took part in the experiment. Of these,
130 (65 men and 65 women; average age 22; range 18-29
years) questionnaires were included in the final analyses.
Twelve respondents dropped out because of colour
blindness, or because they experienced a mild nausea in the
virtual environment.
Stimulus material
The virtual simulation was projected on a 10 meter screen.
Figure 1 depicts one of the participants at the Virtual
Reality Laboratory at the University of Twente and Figure
2 depicts two stills of the simulation. After reading an
instruction on the start page, participants could navigate
through an animation of Leiden station with a mouse and
scroll arrows on a keyboard. They were instructed to:
“...get the first train to Amsterdam. Find out at which
platform and at what time your train leaves. Wait on the
platform until your train arrives. You already got your
ticket. Please, try to imagine the situation, and try to behave
as you would in a real life situation”. Then, the ‘avatar’
could enter the station and freely navigate through the
station from a first person perspective. From this
perspective they were able to ‘walk’ through the station,
climb the stairs and enter the platform. Real life
139
background noises were played during the session to
enhance imaginative power.
simulate low intensity of light. A high intensity of light was
simulated by using a omni 0.5. spot 1.0 for the roof and
omni 0.6 spot 0.2 for the platform.
Measures
A questionnaire was used to measure the overall evaluation
of the railway station.
Emotions were measured on the basis of the Pleasure
Arousal Dominance (PAD) scale [31] with 19 semantic
differential items. Pleasure was measured with 6 items
(unhappy-happy, annoyed-pleased, unsatisfied-satisfied,
melancholic-contended, despairing-hopeful, unpleasantpleasant; coefficient alpha = .88). Arousal was measured
with 7 items (stimulated-relaxed, exited-calm, frenziedsluggish, jittery-dull, wide awake-sleepy, arousedunaroused; fit-tired; coefficient alpha = .71). Dominance
was measured with 6 items (controlled-controlling,
influenced-influential, cared for-in control, awedimportant, submissive-dominant, guided-autonomous;
coefficient alpha = .78).
Figure 1. One of the participants at the Virtual Reality
Laboratory
Evaluation of the platform was measured on the basis of a
combination of 3 scales [12], [39] which resulted in a 12item scale. Participants could indicate to what extent they
felt the platform was attractive, comfortable or messy (1 =
totally disagree, 7 – totally agree; coefficient alpha = .86).
Attitude to the waiting time was measured with 4 items
based on a study by Pruyn and Smidts [38] on waiting time.
Examples of items are ‘I was annoyed because of the time I
had to wait’ and ‘I felt bored during the waiting time’ (1 =
totally disagree, 7 – totally agree; coefficient alpha = .76)
Time perception - Measures included subjective estimations
of time spent in the station and on the platform (‘‘How long
do you estimate the time (in minutes)” and the experience
of time (‘How long did you think this time took’: 1 = very
long, 7 = very short)).
Cognitive preference was measured by asking the
participant which colour they thought was best appropriate
for a station (grey, green, yellow, red or blue).
Perceived colour was measured by asking participants what
the main colour was they saw on the platform. Also
included were a number of demographic variables such as
age, gender and gaming experience.
RESULTS
Figure 2. Two stills of the simulated platform
The colours were manipulated on the platform: blue (colour
code 000.128.255) and red (colour code 255.075.075).
Level of satuaration was held constant for both conditions.
Light was simulated by using a omni 1.0 spot 1.5 for the
platform roof and omni 1.0 spot 0.4 for the platform to
140
On the basis of a multivariate analysis of variance, we then
inspected the main and interaction effects of colour and
type of light on the different dimensions of the overall
evaluation of the station.
Table 1 shows the results per colour and type of light for all
aspects of the judgement.
Table 1
3B.
Analysis of variance (interaction) effects colour and light1
Main effect
Main effect
Colour
Light
F
P
F
Pleasure
<1
<1
Arousal
<1
1.08
Dominance
1.96
Attitude to the
platform
Attitude to the
waiting time
ns
Interaction
effect
Colour x Light
P
ns
F
P
3.55
.06
<1
<1
<1
<1
<1
2.72
.10
<1
<1
4.95
.03
Note: ns = not significant
3C.
Although no main effects came to the fore with regard to
either colour (F < 1) or light (F < 1), the analysis does
show a marginally significant interaction effect
(F(5,99)=1,93, p=.09) for colour x light. Because this
interaction effect is leaning toward significance, we
decided to conduct univariate analyses to further explore a
possible tendency.
3A.
Figure 3. Interaction effects between colour and light for
pleasure (A), attitude to the platform (B), attitude to the
waiting time (C)
The presence of the two aspects together appear to
determine the scores for pleasure, the attitude with regard
to the appearance of the platform and attitude to the waiting
time. Figure 3 shows the interaction effects that were
found. The results reveal a tendency. Figure 3 (panels a, b
and c) shows similar patterns of results for pleasure and the
two attitude measurements. With pleasure, the attitude to
the environment and the attitude to the time spent on the
platform, the blue platform tends to score highest with the
dimmed lighting whereas the opposite effect is the case for
the red platform. In other words, with a low intensity of
light one prefers blue as opposed to red surroundings. As
the intensity of light increases, a shift occurs and the red
environment is found to be more pleasant. The score for the
three aspects of station perception is highest with the red
platform with the high intensity of light; even higher than
the blue platform with the lower intensity.
1
Because it is possible that game experience influenced
participants’ responses, we conducted ANCOVA’s with
game experience as covariate. The reported results
remained (marginally) significant when controlled for
game experience.
Additional analyses were performed to gain insight into
participants’ preference. When asked which colour they
thought was best appropriate for a station (grey, green,
yellow, red or blue), participants indicated a cognitive
preference for the colour blue (36.2%) followed by a the
colour grey (21.5%). Various one-way ANOVA tests show
that this preference has no influence on the overall
evaluation of the station. That is to say that when one
141
prefers (e.g.) blue on a platform, one does not necessarily
appear to appreciate that platform better than someone who
has a preference for another colour. It also appeared that
only one third of the participants could actually indicate
which colour was dominant on the platform, suggesting
that the effects of colour occur subconsciously. This result
is in line with studies on automatic consumer behaviour
which suggests that consumers are often unaware of
environmental factors influencing their behaviour (e.g.,
[21]).
Time perception
Time perception was included as a specific focus of interest
in this study. Generally speaking, respondents estimate
their time spent on the platform as significantly longer
(M=4:54, SD=1:57) than the actual time of stay (M=3:18,
SD=0:49; t(129)=-11.03, p<.00). Four univariate analyses
of variance with the objective and subjective time, the
overestimated length of time and how the time was
experienced were carried out as dependent variables. No
main effect for light comes to the fore from these analyses,
nor does an interaction effect occur for colour x light. A
main effect for colour did appear, however, for how the
time was experienced (F(1,104)=4,63, p=.03). Results show
that the time on the blue platform (M=4,66, SD=1.46) was
perceived as being significantly longer than the time on the
red platform (M=3.98, SD=1.79)). These results show that
on a blue platform time is experienced to pass relatively
slower than on a red one.
platform (F(4,2310)=2.74, p=.03)2. The significant
interaction effects ‘dominance’ and ‘attitude to platform’
are specified in Figure 4 (panels A and B).
For the degree of ‘dominance’ the results show that colours
with an extreme wavelength (blue and red) achieve the
highest score with a lower intensity of light. With a
platform with a medium wavelength (green), however, the
highest score is reached with a higher intensity of light.
With the baseline, or rather the grey platform, the intensity
of light makes no difference. The results for the attitude to
the platform show that the intensity of light with the blue
and green platforms makes little difference. However, the
red platform is deemed better with a higher intensity of
light. This effect is the opposite of the yellow platform, i.e.
on a platform with a yellow colour the platform is
appreciated more when the light is less bright. Also
noticeable here with the baseline measurement is that the
intensity of light has little influence and causes no major
differences. The marginal effects for the attitude to the
waiting time show another picture: a platform with a short
to medium wavelength (blue and green) is valued more
positively than a platform with a higher intensity of light.
On a platform with colours that have a longer wavelength
(yellow and red) the attitude is better with less brightness.
4A.
In the study 1 we were able to explore the effects of two
colours in combination of intensity of light in a virtual
laboratory. To further explore the effects of various colours
in combination with intensities of light we need to expand
the design of study 1. In a virtual laboratory, the actual
performance of a study is time consuming. Therefore the
number of participants is limited. Study 2 investigated the
interactive effects of 5 different colours and two intensities
of light in an online environment which allowed us to
include a larger number of participants.
STUDY 2: ONLINE STUDY
Method
4B.
A 5 (colour: grey vs blue vs red vs green vs yellow) x 2
(light: high vs low intensity) between-subjects design was
used. The virtual reality environment and the questionnaire
from study 1 were converted to an online version. In total
2,360 respondents (56,9% men, 43.1% women) were asked
to navigate through the online simulation.
Results
In order to ascertain interaction effects between colour and
light, a multivariate analysis was carried out on the various
aspects of station perception with colour and intensity of
light as independent variables. The analysis yields an
interaction effect for colour x light (F(20,9236) = 2.35, p =
.001). The interaction effect was found in the degree of
dominance (F(4,2310)=2,62, p=.03), and the attitude to the
142
Figure 4 Interaction effects colour and light for dominance
(A) and attitude platform (B)
2
Given the large d.o.f., the statistical significance of these
results should be considered with caution.
Aspects of time were also included in study 2. On average
respondents spent 7:09 minutes (SD=3:50) at the station, of
which an average 3:54 minutes (SD=2:58) were on the
platform. A t-test revealed a significant difference between
the objective and subjective time on the platform
(t(2244)=44,88, p<.001). The perceived time at the station
appears to be significantly longer than the actual or
objective time.
Again, univariate analyses of variance were conducted with
objective and subjective time and time experience as
dependent variables. These analyses produce a main effect
for the intensity of light and applies to the time perception
of the station (F(4,2329)=4,37, p=.04). This main effect
demonstrates a difference between the station with the
higher intensity of light (M=4,11, SD=1,52) and the station
with less light (M=4,26, SD=1,49). The time spent at the
station with the lower intensity of light was perceived as
being significantly shorter than at the more brightly lit
station. The interaction effect shows how time perception
(F(4,2329)=2,41, p=.05), as well acceptability of time spent
at the platform (F(4,2329)=3,18, p=.01) is influenced by
the combined effects of colour x light. Panel A in Figure 5
shows that time was perceived as being shortest on the
blue, yellow and red platform when the intensity of light
was higher.
5A.
5B.
Figure 5 Interaction effects colour and light for perception
platform (A) and time acceptable (B)
DISCUSSION
When we look at the results we can conclude that almost no
main effects were found for colour and light. But we did
find some interesting interaction effects for light and colour
conditions. The results indicate that passenger respond
more positively to warm colours in combination with
dimmed (low intensity) lighting but at the same time
estimate the waiting time as shorter when cooler colours
and a more intense lighting are used.
As for time perceptions, most passengers appeared to
overestimate the waiting time on the platform, which
concurs with results from earlier work [45].
Although passengers have a definite cognitive preference
for the colour blue in a well-lit environment, it appeared
that only one third of the respondents could indicate which
colour was dominant on the platform. In all situations the
colour one thought to have seen most often was grey.
Despite people indicating they also preferred well-lit
surroundings, the results particularly show effects with
dimmed situations. Apparently, passengers cling to the
image they have of a platform. This confirms that colours
and intensity of light are perceived subconsciously. For
station evaluation, affective effects are thus more important
than cognitive ones.
Both experiments also show the strong influence of waiting
time perception in a station environment. Most people tend
to overestimate the waiting time on the platform, as was
also found in earlier research [25], [30], [35], [45] and can
be explained by the attentional model of Zakay [49]. Zakay
stated that people divide their attention in a prospective
time judgment between the time and other activities. When
time gets more attention, time seems to go slower [49],
known as “a watched pot never boils” [36]. In a station
environment especially daily commuters are focussed on
the time. How the wait is evaluated and how useful
passengers find it seems to be related to both the attitude to
the platform and the impression thereof. In most situations,
time in dimly lit surroundings appears to pass more quickly
than when the lights are brighter. This confirms the results
of Baker and Cameron [5]. In contrast to results found in
the literature, time in a blue environment appears to pass
more slowly than in a red one. One explanation might be
that passengers who feel stressed not only desire cooler
colours which are less arousing and distractive, but also pay
more attention to the time itself, which makes it seem to
pass more slowly [49].
The results show that manipulations in a virtual public
environment successfully allow effects with colour, light,
crowding and time pressure to be demonstrated. These
findings offer an initial insight into the way colour and light
work in a station. However, both experiments were
conducted in a virtual station which might influence the
outcome. Moreover, the significant results obtained in the
online study should be considered with caution. Due to the
large number of participants any difference in a dependent
143
variable can be considered a statistical artefact evoked by
large degrees of freedom. The question arises whether these
findings would also be found in a real station. Recent
developments in new techniques of lighting (e.g.
‘ambilight’) make it easier to study these effects of colour
and light in real-life public spheres.
LITERATURE
1.
2.
Adams, F. M., & Osgood, C. E. (1973). A CrossCultural Study of the Affective Meanings of
Colour. Journal of Cross-Cultural Psychology,
4(2), 135.
Aubert-Gamet, V., & Cova, B. (1999).
Servicescapes From Modern Non-Places to
Postmodern Common Places. Journal of Business
Research, 44(1), 37-45.
12. Bitner, M. J. (1990). Evaluating Service
Encounters: The Effects of Physical Surroundings
and Employee Responses. Journal of Marketing,
54(2), 69-82.
13. Bitner, M. J. (1992). Servicescapes: The Impact of
Physical Surroundings on Customers and
Employees. Journal of Marketing, 56(2), 57-71.
14. Brady, M. K., & Cronin, J. J. (2001). Some new
thoughts on conceptualizing perceived service
quality: a hierarchical approach. Journal of
Marketing, 65(3), 34-49.
15. Brengman, M. (2002). The impact of colour in the
store environment. An environmental psychology
approach [Dissertation]. Gent: University of Gent.
3.
Babin, B. J., Chebat, J. C., & Michon, R. (2004).
Perceived appropriateness and its effect on
quality, affect and behavior. Journal of Retailing
and Consumer Services, 11(5), 287-298.
16. Brengman, M., & Geuens, M. (2004). The four
dimensional impact of color on shoppers’
emotions. Advances in Consumer Research, 31,
122-128.
4.
Babin, B. J., Hardesty, D. M., & Suter, T. A.
(2003). Color and shopping intentions The
intervening effect of price fairness and perceived
affect. Journal of Business Research, 56(7), 541551.
17. Butler, D. L., & Biner, P. M. (1987). Preferred
Lighting Levels: Variability among Settings,
Behaviors, and Individuals. Environment and
Behavior, 19(6), 695.
5.
Baker, J., & Cameron, M. (1996). The effects of
the service environment on affect and consumer
perception of waiting time: an integrative review
and research propositions. Journal of the Academy
of Marketing Science 24, 338-349.
6.
7.
Baker, J., Grewal, D., & Levy, M. (1992). An
Experimental Approach to Making Retail Store
Environmental Decisions. Journal of Retailing,
77(2), 273-289.
Baron, R. A., Rea, M. S., & Daniels, S. G. (1992).
Effects of indoor lighting (illuminance and
spectral distribution) on the performance of
cognitive tasks and interpersonal behaviors: The
potential mediating role of positive affect.
Motivation and Emotion, 16(1), 1-33.
8.
Bellizzi, J. A., Crowley, A. E., & Hasty, R. W.
(1983). The effects of color in store design.
Journal of Retailing, 59(1), 21-45.
9.
Bellizzi, J. A., & Hite, R. E. (1992).
Environmental color, consumer feelings, and
purchase likelihood. Psychology & Marketing,
9(5), 347-363.
10. Biner, P. M., Butler, D. L., Fischer, A. R., &
Westergren, A. J. (1989). An Arousal
Optimization
Model
of
Lighting
Level
Preferences: An Interaction of Social Situation and
Task Demands. Environment and Behavior, 21(1),
3.
144
11. Birren, F. (1969). Principles of Color. New York:
Van Nostrand Reinhold Company.
18. Crowley, A. E. (1993). The two-dimensional
impact of color on shopping. Marketing Letters,
4(1), 59-69.
19. Daurat, A., Aguirre, A., Foret, J., Gonnet, P.,
Keromes, A., & Benoit, O. (1993). Bright light
affects alertness and performance rhythms during
a 24-h constant routine. Physiol Behav, 53(5),
929-936.
20. Dijksterhuis, A., Smith, P. K., Van Baaren, R., &
Wigboldus, D. (2005). The Unconscious
Consumer: Effects of Environment on Consumer
Behavior. Journal of Consumer Psychology 15(3):
193-202.
21. Eysenck, H. J. (1941). A Critical and
Experimental Study of Colour Preferences. The
American Journal of Psychology, 54(3), 385-394.
22. Gifford, R. (1988). Light, decor, arousal, comfort
and communication. Journal of Environmental
Psychology, 8, 177-189.
23. Gorn, G. J., Chattopadhyay, A., Sengupta, J., &
Tripathi, S. (2004). Waiting for the Web: How
Screen Color Affects Time Perception. Journal of
Marketing Research, 41(2), 215–225.
24. Guilford, J. P., & Smith, P. C. (1959). A system of
color-preferences. Am J Psychol, 72, 487-502.
25. Hornik, J. (1984). Subjective vs objective time
measures: a note on the perception of time in
consumer behavior. Journal of Consumer
Research, Vol. 11, pp. 615-18.
26. Jacobs, K. W., & Suess, J. F. (1975). Effects of
four psychological primary colors on anxiety state.
Percept Mot Skills, 41(1), 207-210.
27. Kallman, W. M., & Isaac, W. (1977). Altering
arousal in humans by varying ambient sensory
conditions. Percept Mot Skills, 44(1), 19-22.
28. Kotler, P. (1973). Atmospherics as a marketing
tool. Journal of Retailing, 49(4), 48-64.
29. Küller, R., Ballal, S., Laike, T., Mikellides, B., &
Tonello, G. (2006). The impact of light and colour
on psychological mood: a cross-cultural study of
indoor work environments. Ergonomics, 49(14),
1496-1507.
30. Maltha, J (2006). Stationsbeleving. Hoe ervaren
reizigers hun verblijf op een klein station.
Afstudeerscriptie Strategisch Product Ontwerpen.
Technische Universiteit Delft, Faculteit Industrieel
Ontwerpen.
31. Markin, R. J., Lillis, C. M., & Narayana, C. L.
(1976). Social psychological significance of store
space. Journal of Retailing, 52(1), 43-54.
32. Mehrabian, A., (1976). Public Spaces and Private
Spaces: The Psychology of Work, Play, and
Living Environments. New York: Basic Books.
33. Mehrabian, A., & Russell, J. A. (1974). An
approach to environmental psychology: MIT Press
Cambridge, MA
34. Miwa, Y., & Hanyu, K. (2006). The Effects of
Interior Design on Communication and
Impressions of a Counselor in a Counseling
Room. Environment and Behavior, 38(4), 484.
35. Moreau, A. (1992). Public transport waiting times
as experienced by customers, marketing research
involving the Grenoble system. In: Public
Transport International, Volume 41, no 3 (AugNov, pp. 52-68).
36. Ornstein, R.E. (1969). On the experience of tiem.
Penguin Harmondsworth.
37. Parasuraman, A., Zeithaml, V. A., & Berry, L. L.
(1985). A Conceptual Model of Service Quality
and Its Implications for Future Research. Journal
of Marketing, 49(4), 41-50.
38. Pruyn, A. & A. Smidts (1998). Effects of waiting
on the satisfaction with the service: Beyond
objective time measures. International Journal of
Research in Marketing, 15(4), 321-334.
39. Senta. (2005). Beleidsvertaling Monitor Integrale
Kwaliteit Stations: Samenhang en prioriteiten.
Onderzoek in opdracht van NS/Prorail.
40. Singh, S. (2006). Impact of color on marketing.
Management Decision, 44(6), 783-789.
41. Smets, G. (1969). Time Expression of Red and
Blues. Perceptual and Motor Skills, 29 (AugustDecember), 511-514.
42. Stone, N. J., & English, A. J. (1998). Task type,
posters, and workspace color on mood,
satisfaction, and performance. Journal of
Environmental Psychology, 18(2), 175-185.
43. Valdez, P., & Mehrabian, A. (1994). Effects of
color on emotions. J Exp Psychol Gen, 123(4),
394-409.
44. Van Bommel, W. J. M. B., G.J. van den. (2004).
Werkverlichting: visuele en biologische effecten.
Nederlandse Vereniging voor Veiligheidskunde,
Info, februari.
45. Van Hagen, M. (2007). Wachttijdperceptie- en
waardering op NS stations. Colloquium
Vervoersplanologisch Speurwerk, November.
46. Walters, J., Apter, M. J., & Svebak, S. (1982).
Color preference, arousal, and the theory of
psychological reversals. Motivation and Emotion,
6(3), 193-215.
47. Wexner, L. B. (1954). The degree to which colors
(hues) are associated with mood-tones. Journal of
Applied Psychology, 38(3), 432–435.
48. Wilson, G. D. (1966). Arousal properties of red
versus green. Perceptual and Motor Skills, 23(3),
947-949.
49. Zakay, D. (1989). Subjective time and attentional
resource allocation: an integrated model of time
estimation. In: Zakay, D. (Eds) and Levin, I.
(1989), Time and Human Cognition: A Life Span
Perspective, Advances in Psychology, No. 59,
Elsevier, Amsterdam. Pp. 365-397.
145
Preliminary Evidence That Both Red and Blue Lights
Increase Nocturnal Alertness
Mariana G. Figueiro, Ph.D., & Mark S. Rea, Ph.D.
Lighting Research Center,
Rensselaer Polytechnic Institute
21 Union Street, Troy, NY 12180, USA
(518) 687-7100
figuem@rpi.edu
ABSTRACT
Retinal blue light exposures impact nocturnal alertness,
implicating participation by the circadian system, which is
maximally sensitive to short-wavelength light. We
investigated the impact of two levels of both blue and red
light on nocturnal alertness, as measured by
electroencephalogram (EEG). Exposures to both levels of
the blue and red lights increased beta and reduced alpha
power relative to preceding dim-light conditions. These
results suggest that the circadian system is not the only
light-sensitive pathway that can affect nocturnal alertness
because the levels of red light used here are not effective
for stimulating the circadian system.
Keywords
Circadian, alertness, light, electroencephalogram
BACKGROUND
The impact of light on alertness has gained recent attention
in the scientific community because of the now wellestablished role that retinal light exposure plays in
regulating extra-visual functions like the circadian timing
system. Bright white light (greater than 2500 lx at the eye)
has been shown to increase alertness at night [2, 3, 5, 7-10,
12, 17], but the mechanisms associated with the alerting
effects of light have not been unambiguously established.
The human circadian system is known to be maximally
sensitive to short-wavelength radiation (blue light) [4, 18,
25]; thus, the efficacy of “moderate” blue light should be
comparable to that of “bright” white light for evoking
measures of alertness at night [18]. Results from recent
studies using blue light [6, 13] are entirely consistent with
the neurophysiological evidence that neural pathways from
the suprachiasmatic nuclei (SCN) affect sleep and alertness,
as recently elucidated by Saper and colleagues [20-22],
adding weight to the inference that the SCN, through retinal
stimulation by short-wavelength light, play a role in human
nocturnal alertness.
One way to test the hypothesis that the alerting effects of
nocturnal light exposures are mediated only by the
circadian system is to compare the impact of red and blue
light on alertness at night. Long-wavelength (! > 600 nm)
light does not stimulate the human circadian system [4, 25]
except perhaps at very high levels [14]. Therefore, if the
light-induced stimulation of the circadian system at night is
146
solely responsible for light-induced nocturnal alertness,
then red light should be an ineffective stimulus.
The present experiment was designed to look at the impact
that two “moderate” levels (10 lx and 40 lx) of both
narrow-band blue (!max = 470 nm) and narrow-band red
(!max = 630 nm) light might have on electroencephalogram
(EEG) recordings during the night. Electrodes affixed to
the human scalp are able to sense changes in brain activity
when subjects engage in different types of mental tasks.
The relative electrical power recorded at these different
frequencies (from 1 to 50 Hz) is used to infer mental states
in these subjects. Changes in the power at specific different
frequency bands have been used as markers of mental
alertness. Alertness is associated with lower levels of
power in the alpha frequency band (8-12 Hz) and is
associated with higher levels of power in the beta
frequency band (12-30 Hz). If the circadian pathway is
solely responsible for light-induced alertness at night, then
only the blue light should reliably evoke an alerting response. Further, there should be a graded response in the
EEG recordings with increasing levels of the blue light, as
long as their irradiances are above threshold and below
saturation for the circadian system response.
METHODS
Procedures and apparatus
Sixteen subjects (21 to 33 years of age) were recruited to
participate in the study from an electronic posting at
Rensselaer Polytechnic Institute in Troy, N.Y. All subjects
were screened for major health problems and except for
women taking birth control pills, subjects reported not
taking any pharmaceuticals or medications. Every subject
completed a Munich Chronotype Questionnaire (MCTQ)
prior to the study [19]. In order to have a more homogenous
sample of subjects, those who were late or extremely late
chronotypes were excluded from the experiment. All
subjects provided an informed consent approved by
Rensselaer’s Institute Review Board. Subjects were asked
to refrain from alcohol and caffeine on the days of the
experiment and were asked not to sleep after awakening for
the day. Of the sixteen subjects, nine males and five
females completed the entire experiment, and the results of
their EEG data are reported here.
Four experimental lighting conditions, two spectra (blue
and red) each at two light levels (10 and 40 lx) were deli-
vered to individual subjects from 0.6 ! 0.6 ! 0.6 m light
boxes, each fitted with arrays of light-emitting diodes
(LEDs). The arrays (ICove, Color Kinetics) were located
behind the front box apertures to be outside the subject’s
direct view, thereby creating a uniform, non-glaring distribution of light within the box. During light exposures,
subjects placed their chin on a chinrest mounted near the
front of a box, ensuring delivery of the prescribed light
exposure. When sitting at the light box, the subject’s head
was aligned with the aperture of the box, so that subjects
were always exposed to full-field, diffuse light. The
spectral emissions of the blue LEDs peaked at 470 nm with
a full width at half maximum (FWHM) of 25 nm. The red
LEDs peaked at 630 nm with a FWHM of 25 nm. Before
the experiment, each of the light boxes was calibrated using
a Gigahertz illuminance photometer to provide the
prescribed corneal illuminance levels when the subjects
were positioned on the chinrest. The spectral irradiance of
the red and blue conditions were measured prior to the
experiment
with
a
calibrated
spectroradiometer
(Photoresearch model PR705a) and diffuse white reflectance standard (Labsphere model SR 099) and used to calibrate the Gigahertz illuminance readings. Two boxes provided blue light (40 µW/cm2 at 40 lx and 10 µW/cm2 at 10
lx) and two emitted red light (19 µW/cm2 at 40 lx and 4.7
µW/cm2 at 10 lx); light levels could be adjusted with an
electronic dimmer to reach the prescribed light levels without significantly affecting the relative spectral distributions
of the LED emissions. Measurements of pupil area completed after the experiment with a different group of
subjects (N = 5) were: red at 10 lx, 34 mm2; red at 40 lx, 22
mm2; blue at 10 lx, 10 mm2; blue at 40 lx, 6.5 mm2.
Groups of four subjects participated in two sessions separated by at least one week. Subjects were asked to arrive at
the laboratory at 22:00 to receive instructions and be fitted
with scalp electrodes for EEG recordings. Because only
one EEG machine was available, data collection was staggered. The first subject in a session started at 23:00, the
second at 23:10, the third at 23:20, and the last at 23:30; the
last subject completed the experiment at 03:45. During
every session, each subject was presented a high (40 lx)
and a low (10 lx) light exposure condition of the same
spectrum (blue or red). The presentation order of the light
levels was counterbalanced across sessions for a given
subject; light spectra were counterbalanced across subjects
within sessions. Every 45-minute experimental lighting
condition was preceded by a 45-minute period of inactivity
in a dim-light anteroom (< 1 lx of red light at the cornea).
During the inactive, dim-light periods, subjects remained
quiet and were not allowed to perform any task (e.g., talk,
read, or computer work) except for the prescribed data
sampling specified in the experimental protocol. Each
nighttime session consisted of four, 45-minute light-anddim conditions (a dim-light condition always preceded one
of the four experimental lighting conditions), plus a 15minute period for data collection prior to each lighting condition (in addition to EEG recording, performance
measures and saliva melatonin were also collected, but are
not reported here).
Data Collection
The Biosemi ActiveTwo system with active electrodes was
used for EEG recordings. This system is battery powered,
minimizing electrical interference from alternating current
(ac) during recording sessions. Electrodes were placed on
subjects’ scalps according to the International 10-20 system
at Oz, Pz, Cz, and Fz [1]. Two additional electrodes serving
as virtual reference electrodes for those attached to the
scalp were attached to the right and to the left earlobes.
Near the end of each 45-min dim light and each 45-min
light exposure period, the scalp electrodes on each subject
were attached to the EEG recording system. Six minutes of
data were collected: three one-minute periods with the
subject’s eyes closed alternating with three one-minute
periods with the eyes open. When the eyes were open and
subjects were not sitting at the light box (dim-light
condition), the subjects were asked to fixate on a specific
marked point approximately one meter away. Similarly,
when sitting at the light box, subjects fixated on specific
point on the far wall of the box approximately 0.6 m away.
Subjects were visually monitored by an experimenter to
ensure compliance with the protocol.
The EEG signals were sampled at 16384 Hz and then lowpass filtered and downsampled to 2048 Hz for electronic
storage by the Biosemi system. All subsequent EEG data
processing and analyses were performed with Matlab version R2008a by The MathworksTM. The signals recorded
from the two reference channels were averaged and these
values were subtracted from those obtained from all of the
other channels. The direct current (dc) offset of each channel was eliminated by subtracting the mean value of each
channel from itself. A low-pass finite impulse response
(FIR) filter (f-3dB = 50 Hz) was applied and the data were
downsampled to 512 Hz. Then a high-pass, third-order
Butterworth filter (f-3dB = 4 Hz) was applied to the downsampled signals from each channel to eliminate slow
trending in the data.
Another program divided the filtered data into 5-second
epochs, segregated by periods when the eyes were open and
when they were closed during the six-minute recording
period. Eye blink artifacts were eliminated by removing
epochs from all channels where voltage fluctuations of any
epoch exceeded ±100 µV. A Blackman window followed
by a fast Fourier transform (FFT) was then applied to the
data segments. This process yielded spectral power
distributions from 1-50 Hz. The power spectra for each
one-minute segment were then combined to give an
average spectral power distribution for each trial. The
relative power levels for eyes open in the alpha (8-12 Hz),
beta (12-30 Hz), gamma (30-50 Hz), theta (5-8 Hz), and
alpha-theta (5-9 Hz) ranges were calculated as a percentage
of overall power from 1-50 Hz. These calculations were not
performed for those intervals when the eyes were closed.
Reported here are the results from the percentages of power
147
in the alpha and the beta range of frequencies because they
have been used as measures of alertness in previous studies
(e.g., [6]).
RESULTS
Two-way (eight light-and-dim conditions and four recording channels), repeated measures ANOVAs were employed
using the percent power in the alpha frequency range (8-12
Hz) and using the percent power in the beta frequency
range (12-30 Hz) recorded from four scalp electrode channels (Oz, Pz, Cz and Fz) in the EEG recordings. Both the
main effects of light-and-dim conditions and of recording
channels were significant for alpha (respectively, F7,91 =
2.15, p = 0.046 and F3,39 = 44.7, p <0.0001) and for beta
(respectively, F7,91 = 3.91, p = 0.0009 and F3,39 = 5.36, p =
0.0035); the interaction between the light-and-dim conditions and the channels was not statistically significant for
either the alpha or the beta frequencies, indicating that the
alpha and the beta frequencies from every channel exhibited similar patterns among the eight light-and-dim lighting conditions. Post-hoc, paired two-tail t-tests were performed for the alpha and for the beta frequencies using the
combined data from all four channels.
As illustrated in Figure 1, relative alpha power recorded
from all channels after light exposure decreased compared
to relative alpha power recorded in the dim light just prior
to the light exposure. Alpha power after exposures to both
levels of blue light (i.e., 10 lx and 40 lx) and to red light at
10 lx was statistically significantly lower than alpha power
recorded in the previous dim-light condition. Consistent
with Figure 1, Figure 2 illustrates the increase in relative
beta power following light exposures compared to relative
beta power recorded in the dim light just prior to the light
exposures. Mirroring the statistical inferences for the alpha
frequencies, there was a significant difference between the
previous dim-light condition and the two blue-light
conditions and for the red-light condition at 10 lx.
It was hypothesized that the blue light conditions would
follow a dose response such that relative alpha power
would be lower for the 40 lx condition than for the 10 lx
condition. It was also expected that the relative beta power
would be significantly higher for the blue-40 lx condition
than for the blue-10 lx condition. This expectation was met
for the alpha frequencies, (p = 0.01) but, although in the
right direction, not for the beta frequencies (p = 0.1).
Although the red-light condition resulted in dose
intransitivity for both the alpha and the beta frequencies
(i.e., the red-10 lx condition produced lower relative alpha
power and higher relative beta power than for the red-40 lx
condition), this difference was not statistically different for
either frequency band (p = 0.21 for alpha power and p =
0.13 for beta power).
148
Figure 1. Relative alpha power after the four dim and the four
experimental lighting conditions. Statistically significant (*)
lower levels of relative alpha power were associated with blue-10
lx (B10; p = 0.007), blue-40 lx (B40; p<0.0001), and red-10 lx
(R10; p< 0.0001), than with the previous dim-light exposures.
There was no significant difference in alpha power between red40 lx (R40) and the previous dim-light exposure which must, in
part at least, reflect the significantly lower (p < 0.05) alpha power
level associated with the dim condition preceding the R40
condition than with any of the other the dim condition.
Figure 2. Relative beta power after the four dim and the four
experimental lighting conditions. Statistically significant (*)
higher levels of relative beta power were associated with blue-10
lx (B10; p = 0.0006), blue-40 lx (B40; p<0.0001), and red-10 lx
(R10; p< 0.0001), than with the previous dim-light exposures.
There was no significant difference in beta power between red-40
lx (R40) and the previous dim-light exposure.
DISCUSSION
Nocturnal alertness as measured by EEG is affected by
light, but it does not seem to be affected only by light
stimulation of the circadian system. Exposures to both red
and blue light reduced alpha power and increased beta
power levels relative to their preceding dim-light condition.
However, there was only an apparent dose response for the
blue light. That is, as levels of blue light increased from 10
lx to 40 lx, alpha power decreased and beta power
increased (although not statistically significant), as would
be expected if blue light served as alerting stimuli.
Quizzically, the reverse was true for the red light. Although
not significant, alpha levels were higher and beta levels
were lower for 40 lx than for 10 lx of red light. If in fact
reliable, the dose intransitivity for the red-light conditions
remains unexplained and, indeed, somewhat implausible. It
is conceivable that there is an optimum irradiance of red
light for alertness (i.e., red-10 lx), but this inference seems
rather unlikely and these results definitely demand further
study. Nevertheless, these results indicate that colored light
of ”moderate” corneal irradiance levels can induce
alertness at night, but that light-induced alertness at night is
not mediated only by the circadian system.
It is not completely clear, however, how light-induced
alertness can arise from other neural pathways. Some
evidence suggests that red light, which is ineffective for
stimulating the circadian system at “low” and “moderate”
light levels, can be more stimulating than blue light [15,
24]. Studies have reported that perception of red color prior
to executing an important task impairs performance relative
to the perception of green or achromatic color [11, 15].
Elliot et al. [11] performed a series of studies to investigate
the impact of color red on performance in achievement
contexts, that is, in situations in which competence is
evaluated and positive and negative outcomes are possible.
They hypothesized that red color is associated with danger
of failure, and therefore, an automatic, unconscious
decision to avoid the object, situation or event occurs.
According to their hypothesis, red color impairs
performance because it evokes motivational tendency to
avoid failure, which, according to the authors, undermines
performance. The results of their experiments supported
their hypothesis that perception of red color prior to an
achievement task impairs performance compared to a green
and an achromatic color. Similar findings have been
reported by Stone [23]. These results are not consistent
with findings by Hill and Barton [15], however, who
reported that red enhances performance of athletes who
wore red color. In general, the studies of color on emotions
and performance are conflicting and not well-grounded in
neurophysiology. The explanation for this lack of
consistency may be due to random, non-systematic effects
of color on human perception or psychology or to
individual differences in preference and cultural
associations [16]. Notwithstanding this last point, these
results are then, to a limited extent, consistent with some
previous studies suggesting that red light acts as a stimulant
through some unspecified neural pathway. It is important to
note, however, that these earlier studies (e. g., [11, 15, 23,
24]) were probably conducted in the daytime, not at night,
although the times for these studies were not documented.
Further, these studies do not always provide quantitative
descriptions of the colored stimulus.
Clearly, more
research is needed to elucidate the light-sensitive
mechanisms affecting alertness during the day and during
the night.
CONCLUSIONS
The present results are consistent with previous findings
showing that light of sufficient corneal irradiance increases
alertness at night. There is previous compelling evidence
that light-induced stimulation of the circadian system
increases alertness at night, but the present results implicate
other mechanisms through which light can also increase
alertness. It is important then to determine if these inferred
mechanisms are independent of the circadian system or
interact with it by conducting more systematic studies of
light spectra and light levels during the night as well as
during the day.
ACKNOWLEDGEMENTS
The study presented here was supported by the Office of
Naval Research through the Young Investigator Program
awarded to MGF. The authors would like to acknowledge
Dr. Vodyanoy of the Office of Naval Research for his
support. Dr. Christopher Steele of the Naval Research
Medical Laboratory, Andrew Bierman, John Bullough,
Dennis Guyon, Bonnie Morgan, Chris Munson, Barbara
Plitnick, Jennifer Taylor, and Dan Wang of the Lighting
Research Center, and Lauren Schramek of Russell Sage
College, Troy, N.Y., are acknowledged for their support
and contributions to the study.
REFERENCES
1.
American Encephalographic Society. Guidelines for
Standard Electrode Position Nomenclature. J Clinical
Neurophysiology 1991, 8:200-202.
2.
Badia, P., Myers, B., Moecker, M., Culpepper, J.
Bright light effects on body temperature, alertness,
EEG and behavior. Physiol Behav 1991, 50(3):583588.
3.
Boyce, P., Beckstead, J., Eklund, N., Strobel, R., Rea,
M. Lighting the graveyard shift: The influence of a
daylight-simulating skylight on the task performance
and mood of nightshift workers. Light Res Technol
1997, 29(3):105-134.
4.
Brainard, G., Hanifin, J., Greeson, J., Byrne, B.,
Glickman, G., Gerner, E., Rollag, M. Action spectrum
for melatonin regulation in humans: Evidence for a
novel circadian photoreceptor. J Neurosci 2001,
21:6405-6412.
5.
Cajochen, C., Krauchi, K., Danilenko, K.V., WirzJustice, A. Evening administration of melatonin and
bright light: interactions on the EEG during sleep and
wakefulness. J Sleep Res 1998, 7(3):145-157.
149
6.
7.
8.
9.
Cajochen, C., Munch, M., Kobialka, S., Krauchi, K.,
Steiner, R., Oelhafen, P., Orgul, S., Wirz-Justice, A.
High sensitivity of human melatonin, alertness,
thermoregulation and heart rate to short wavelength
light. J Clin Endo Met 2005, 90:1311 – 1316.
Cajochen, C., Zeitzer, J., Czeisler, C., Dijk, D. Doseresponse relationship for light intensity and ocular and
electroencephalographic correlates of human alertness.
Behav Brain Res 2000, 115:75-83.
Campbell, S.S., Dawson, D. Enhancement of nighttime
alertness and performance with bright ambient light.
Physiol Behav 1990, 48(2):317-320.
Daurat, A., Foret, J., Benoit, O., Mauco, G. Bright
light during nighttime: effects on the circadian
regulation of alertness and performance. Biol Signals
Recept 2000, 9(6):309-318.
10. Eastman, C.I., Liu, L., Fogg, L.F. Circadian rhythm
adaptation to simulated night shift work: effect of
nocturnal bright-light duration. Sleep 1995, 18(6):399407.
11. Elliot, A.J., Maier, M.A., Moller, A.C., Friedman, R.,
Meinhardt, J. Color and psychological functioning: the
effect of red on performance attainment. J Exp Psychol
Gen 2007, 136(1):154-168.
12. Figueiro, M., Rea, M., Boyce, P., White, R., Kolberg,
K. The effects of bright light on day and night shift
nurses' performance and well-being in the NICU. Neonatal Intens Care 2001, 14(1):29-32.
13. Figueiro, M.G., Bullough, J.D., Bierman, A., Fay,
C.R., Rea, M.S. On light as an alerting stimulus at
night. Acta Neurobiol Exp (Wars) 2007, 67(2):171178.
14. Hanifin, J.P., Stewart, K.T., Smith, P., Tanner, R.,
Rollag, M., Brainard, G.C. High-intensity red light
suppresses melatonin. Chronobiol Int. 2006; 23(1-2):
251-68.
150
15. Hill, R.A., Barton, R.A. Psychology: Red enhances
human performance in contests. Nature 2005,
435(7040):293.
16. Lee, T-R, Tang, D-L and Tsai, C-M. Exploring color
preference through eye tracking, Proceedings AIC
Color 05 – 10th Congress of the International Colour
Association, 333-336, 2005.
17. Lowden, A., Akerstedt, T., Wibom, R. Suppression of
sleepiness and melatonin by bright light exposure during breaks in night work. J Sleep Res 2004, 13(1):3743.
18. Rea, M., Figueiro, M., Bullough, J., Bierman, A. A
model of phototransduction by the human circadian
system. Brain Res Rev 2005, 50(2):213-228.
19. Roenneberg, T., Wirz-Justice, A., Merrow, M. Life
between clocks: daily temporal patterns of human
chronotypes. J Biol Rhythms 2003, 18(1):80-90.
20. Saper, C.B., Cano, G., Scammell, T.E. Homeostatic,
circadian, and emotional regulation of sleep. J Comp
Neurol 2005, 493(1):92-98.
21. Saper, C.B., Lu, J., Chou, T.C., Gooley, J. The
hypothalamic integrator for circadian rhythms. Trends
Neurosci 2005, 28(3):152-157.
22. Saper, C.B., Scammell, T.E., Lu, J. Hypothalamic
regulation of sleep and circadian rhythms. Nature
2005, 437(7063):1257-1263.
23. Stone, N. Designing effective study environments.
Journal of Environmental Psychology 2001, 21:179190.
24. Stone, N.J. Environmental view and color for a
simulated
telemarketing
task.
Journal
of
Environmental Psychology 2003, 23:63-78.
25. Thapan, K., Arendt, J., Skene, D.J. An action spectrum
for melatonin suppression: evidence for a novel nonrod, non-cone photoreceptor system in humans. J
Physiol 2001, 535(Pt 1):261-267.
Reducing Light Intensity and Changing its Spectral
Composition: Effects on Human!s Sleep Characteristics
and Melatonin Suppression Under “Natural Conditions”
Marina. C Giménez1! , Pauline Bollen1, Marijke C.M1. Gordijn, Matthijs L. van der Linden2,
Domien G.M. Beersma1
1
2
Department of Chronobiology
Center for Life Sciences
University of Groningen, The Netherlands
m.c.gimenez@rug.nl
ABSTRACT
Light has a great impact in our everyday life for vision, but
also for non-visual processes. The recent discovery of
photosensitive retinal ganglion cells triggers new studies on
the non-visual effects of light’s spectral composition. In the
present study via the use of soft orange contact lenses we
investigated how, under natural conditions, a reduction in
exposure to the short wavelengths (blue) light affects sleep
and the suppression of melatonin concentrations to a
standard light stimulus in healthy young subjects. The
orange lenses were effective in reducing light input. If worn
only during the light pulse melatonin suppression in
response to a 2h 600 lux white light pulse was reduced
from 29% in the control condition to 17.3% (p<0.05). No
significant differences in melatonin suppression were
observed between the control condition (29%) and after
wearing the orange lenses for 16 days (34.1%). These
results indicate that the non-visual response of melatonin
suppression to light adapted. While wearing the orange
lenses the amount of sleep was reduced, somewhat similar
to the sleep changes that occur with ageing.
Keywords
Light intensity, Light spectral composition, Melatonin,
Sleep, Humans.
Oculenti
University Medical Center
Groningen, The Netherlands
and activating effects and acutely suppresses melatonin at
night [3,4,5]. Synchronization of the biological clock
depends on several aspects related to the light signal; its
intensity and the time of exposure. The recent discovery of
intrinsically photosensitive retinal ganglion cells (ipRGCs,
maximal sensitive to short wavelengths) [6] has triggered
new studies on the role of not only light intensity but also
its spectral composition. Suppression of melatonin has been
shown to be higher to light pulses of short wavelengths
than to light pulses of other wavelengths [7,8,9]. Alertness
was also shown to be more sensitive to short wavelengths
[9]. These studies however, compared the effects of
monochromatic light sources, which although informative
to understand mechanisms is far from being a natural
situation.
Considering the natural phenomenon of cataract (yellowish
of the lens with ageing) [10], in the present study we
investigate what the effects are of a reduction in (blue) light
via the use of orange soft contact lenses. This situation
resembles at least qualitatively what happens with ageing.
We hypothesized therefore a disruption of sleep patterns
and a reduction in the suppression of the nocturnal
melatonin to nocturnal light exposure.
MATERIALS AND METHODS
INTRODUCTION
Due to the earth’s daily rotation around its axis a temporal
pressure is imposed on almost all organisms; a 24h day
with light and dark cycles (day and night). In order to
anticipate the temporal changes along the 24h day,
organisms have evolved circadian clocks. The circadian
clock generates cycles with an approximate period of 24h
that needs to synchronize with the external environment.
Light is the signal that sets the phase of our biological
clock, which in turn synchronizes our physiological and
psychological rhythms to the 24h rhythm of the
environment [1, 2]. Furthermore, light has acute alerting
Subjects
In total 50 subjects enrolled for the study. Only those
subjects who were healthy, non-smoker, non-color blind,
and with an intermediate chronotype [11, modified for
Dutch population] were selected. Subjects who worked
night shifts or travelled through more than 2 time zones
during the 2 weeks prior the study were also excluded.
Because subjects have to wear soft contact lenses during 2
consecutive weeks, 24h per day a check-up by a contactlenses specialist was conducted at the University Medical
Center of Groningen (UMCG) in order to assess subject’s
151
eyes condition. After screening, 22 subjects were selected
from whom 15 completed the study (7m:8f, mean age ±
sem: 24.5 ± 4.6 years old). The study was conducted
between December 2007 and September 2008. The Medical
Ethical Committee of the UMCG, The Netherlands,
approved the study protocol. All subjects signed a written
informed consent form prior to their participation.
Soft orange contact lenses (OL)
The OL (CE: 0120, with UV protection) were supplied by
Oculenti at the UMCG, The Netherlands from Ultravision
International Ltd., UK. In the visible range of the spectrum
(from 400-700 nm), the OL reduced transmitted light for
37% (calculated as the area under the curve) while in the
short wavelengths (400-530 nm) the reduction was 56%.
The reduction in light transmission through the OL can be
seen in Figure 1.
110
100
90
80
70
60
50
40
30
20
10
0
400
hour until 2:00 h, and 2-hourly from 3:00 until 9:00 h. On
the second night the same protocol was followed until
00:00. From 00:00 until 2:00 h subjects were asked to sit in
front of 2 light boxes with full spectrum white light (600
lux, Pharos Max, Osram Dulux-L tubes, ©Lumie) to
investigate the suppression of melatonin. During these two
hours subjects watched a movie on a TV situated in
between both light boxes so that they could keep their level
of gaze constant. Light intensity at eye level was regularly
checked during the 2 hours light pulse and adjusted if
necessary. On a separate night from each 16-days session,
subjects came for an extra night at which the acute effect of
the OL on suppressing melatonin was measured (S-OL).
For this purpose, the protocol of the second night was
repeated but in this session subjects put the OL in only 30
minutes before the light pulse (in contrast with 15 days of
continuously wearing the OL; L-OL). Subjects were
carefully instructed about the collection of saliva samples
for melatonin assessment. Eating was restricted to the 15
minutes after each sample, chocolate, bananas, coffee or
black tea were not allowed during the whole time. Samples
were centrifuged immediately after its collection and stored
at -20ºC until its analysis.
Analysis
450
500
550
600
650
700
wavelength
Figure1: Relative light transmittance through the OL from a
halide light bulb (MDS 200.2, Philips).
Experimental design
A control condition (subjects wore their own contact lenses
n=13, or no lenses n=2) and an experimental condition
(subjects wore the OL) were assigned to each subject in a
randomized order. The OL were worn 24h per day. Each
condition lasted for 16 days, they were planned at least 2
weeks apart to avoid any potential carry-over effects, and
they started on the same day of the week in order to have a
similar pattern of behavior within each subject.
Melatonin concentration measured in saliva was
determined by means of radio-immunoassay (RK-DSM,
Bühlmann laboratories AG, Siemens Medical Solutions
Diagnostics, Breda, The Netherlands). The area under the
curve was calculated from time point 00:00 until time point
2:00 for the control profile, and the control (CL), the L-OL
and the S-OL suppression conditions to estimate the
nocturnal melatonin suppression by light. A repeated
measurements ANOVA was used to test the effects of these
conditions.
Sleep parameters were assessed by means of actiwatches
and sleep diaries. For this purpose only the first 14 days of
each condition were used since sleep during the last two
nights in our lab was disturbed by the sampling protocol.
Actual sleep (the percentage of assumed sleep minus the
time being awake), sleep efficiency (percentage of time
spent asleep while in bed) and sleep fragmentation (the
percentage of immobility phases of 1 minute as a
proportion of the total number of immobility phases) were
calculated. The effects of the OL were tested with Paired-T
test.
Measurements
During the 16 days that each condition lasted, subjects
wore an actiwatch® (Cambridge Neurotechnologies, UK)
to measure sleep-wake cycles and filled in sleep diaries.
During the last two nights of each condition, subjects came
to the lab in order to assess a dim light melatonin profile on
the first night and melatonin suppression on the second one.
During the first night, in order to assess melatonin profiles,
light levels were dimmed (<10 lux) and saliva samples
were taken using cotton swabs (Sarstedt BV, Etten-Leur,
The Netherlands) hourly from 19:00 to 00:00 h, every half
152
RESULTS
The suppression of melatonin during light exposure
measured as the area under the curve relative to the control
profile condition (= 0 level in figure 2) can be seen in
figure 2. The repeated measurement ANOVA revealed a
significant effect of condition (F (4,9) = 5.694, p < 0.05).
Post hoc comparisons showed no significant differences
between suppression after wearing the OL for 16 days (LOL) with the control suppression (CL) (F (1,13) = 0.26 p =
0.62). However, when compared to the control suppression
the acute suppression achieved in the S-OL condition was
significantly different (F (1,12) = 10.427, p < 0.01).
Wearing the lenses for 16 days lead to a small but
significant reduction in the actual sleep percentage (t =
3.41, p < 0.01), no significant differences were found
however in the sleep efficiency nor in the fragmentation
index (table 1).
0
-5
-10
-15
-20
-25
-30
-35
-40
-45
-50
b
a
CL
a
L-OL
S-OL
Figure 2: Melatonin suppression (calculated as area under
the curve) ± s.e.m. relative to the control profile (0 level).
No significant differences in suppression were found
between the control (light grey bar) and the OL (dark grey
bar) condition (denoted by a), while the acute suppression
of the OL (black bar) was significantly different from the
control condition (denoted by b).
Actual sleep %
Sleep
efficiency
Fragmentation
Index
Control
L-OL
85.05 ±
1.49
80.77 ±
2.05
27.54 ±
3.13
83.64 ±
1.44
80.10
±1.68
28.71 ±
3.00
pvalue
< 0.01
ns
ns
Table 1: Mean ± s.e.m. and p values of the sleep parameters
measured in this study for both conditions.
DISCUSSION
The aim of this study was to investigate the effects of
diminishing the light input, in particular in the short
wavelengths range of the visible spectra. In order to do this
and trying to simulate a natural situation [10], subjects
wore soft orange contact lenses (OL) for 16 consecutive
days, 24h per day.
Melatonin suppression by light is a way to estimate the
effects of light input to the biological clock [12]; the
smaller the suppression to the same stimulus the less
sensitive the system is. In the present study we clearly
showed that wearing the OL only during the light pulse (SOL) reduced the light input to the system; the suppression
of melatonin to a white light stimulus was less then without
the lenses (CL). When the OL were worn for 16 days (L-
OL) the suppression of the nocturnal melatonin was not
different from the suppression without the OL (CL). It can
be concluded that the system has become more sensitive
after 16 days of reduced (short wavelengths) light input by
wearing the OL. It has already been shown that exposure to
dim light during one week by staying inside and using dark
goggles (2% light transmission) increased the suppression
of the nocturnal melatonin due to a higher sensitivity of the
system [13]. Our study represents a more realistic situation
in terms of both the reduction in light intensity as well as
the “bright light exposure” condition. In Hébert et al.[13],
subjects exposed themselves to bright light boxes in the
bright light week condition while in the present study they
exposed themselves to natural and artificial light in
accordance to their personal behaviour. The mechanisms by
which adaptation occurs and at which level in the circadian
system it happens is not know. At the retinal level several
possibilities could be hypothesized. Photostasis, gradual
changes in the retina to keep a constant number of photons
absorbed per day, has been already shown in rats [14],
although there is still no proof of such processes happening
in the human retina. A shift to the responsive form of the
bistable melanopsin molecule due to a reduction in
exposure to short wavelengths and a relative increase in
exposure to long wavelengths while wearing the OL is
another possibility [15].
Regarding sleep parameters a reduction in the actual
percentage of time that subjects spent sleeping was found
as a result of wearing the OL. Although also sleep
efficiency and fragmentation index showed minor changes
in the direction of a more disturbed sleep pattern while
wearing the OL these differences were not significant.
Obviously the reduction of light exposure due to the OL
was not big enough to induce sleep disturbances in these
young people, or the adaptation process was fast enough to
normalize the overall sleep pattern over the 16-days period.
The present study does not support the idea that the
changes seen in the circadian system with ageing can be
explained by the development of cataract in the elderly.
However, the present study has been conducted in young
healthy subjects and probably extrapolating these data to
the elderly might not be possible. With ageing the circadian
input system might become less flexible and looses its
capability of adapting to the different situation.
In conclusion short wavelengths do play an important role
in the suppression of melatonin [7,8,9, and the present
study], but exposure to changes in the spectral composition
of “natural light” on the long-term lead to adaptation of the
non-visual responses to light in young subjects. These
results are important both for understanding individual
differences in non-visual responses to light, and for
artificial indoors lighting developments.
153
ACKNOWLEDGMENTS
Our work is supported by the 6th Framework Project
EUCLOCK (No. 018741) and the first SLTBR grant
sponsored by Outside In, 2007.
REFERENCES
1. Wever, R.A. Phase shifting effects of human circadian
rhythms due to shifts of artificial Zeitgebers. Chrnobiol
Int (1980), 7:303-327.
8. Thapan, K., Arendt, J., and Skene, D. An action
spectrum for melatonin suppression: evidence for a
novel non-rod, non-cone photoreceptor system in
humans. Journal of Physiol (2001), 535(1): 261-267.
9. Cajochen, C., Münch, M., Kobialka, S., Kräuchi, K.,
Steiner, R., Oelhafen, S.O., and Wirz-Justice, A. High
sensitivity
of
human
melatonin,
alertness,
thermoregulation, and heart rate to short wavelength
light. The Journal of Endocrinology & Metabolism
(2005), 90(3):1311-1316.
2. Czeisler, C.A., Kronauer, R.E., Allan, J.S., Duffy, J.F.,
Jewett, M.E., Brown, E.N., and Ronda, J.M. Bright light
induction of strong (type 0) resetting of human
circadian pacemaker. Science (1989), 244:1328-1333.
10. Charman, W.N. Age, lens transmittance, and the
possible effects of light on melatonin suppression.
Ophthal Physiol Opt (2003), 23:181-187.
3. Campbell, S.S., Dijk, D.J., Boulos, Z., Eastman, C.I.,
Lewy, A.J., and Terman, M. Light treatment for sleep
disorders: consensus report. III. Alerting and activating
effects. J Biol Rhythms (1995), 10:129-132.
11. Roenneberg, T., Wirz-Justice, A., and Merrow, M. Life
between the clocks: daily temporal patterns of human
chronotypes. Journal of Biol Rhythms (2003), 18(1):8090.
4. Cajochen, C., Zietzer, J.M., Czeisler, C.A., and Dijk,
D.J. Dose-response relationship for light intensity and
ocular electroencephalographic correlates of human
alertness. Behav Brain Res (2000), 115:75-83.
12. Brainard, G.C., Rollag, M.D., and Hanifin, J.P. Photic
regulation of melatonin in humans: ocular and seural
signal transduction. J Biol Rhythms (1997), 12(6): 537546.
5. Lewy, A.J., Wehr, T.A., Goodwin, F.K., Newsome,
D.A., and Markey, S.P. Light suppresses melatonin
secretion in humans. Science (1980), 210:1267-1269.
13. Hébert, M., Martin, S.K., Lee, C., and Eastman, C.I.
The effects of prior light history on the suppression of
melatonin by light in humans. J Pineal Res (2002),
33(4): 198-203.
6. Provencio, I., Rodriguez, I.R., Jiang, G., Hayes, W.P.,
Moreira, E.F., and Rollag, M.D. A novel human opsin
in the inner retina. J Neurosci (2000), 20:600-605.
7. Brainard, G.C., Hanifin, J.P., Greeson, J.M., Byrne, B.,
Glickman, G., Gerner, E., Rollag, M.D. Action
spectrum for melatonin regulation in humans: evidence
for a novel circadian photoreceptor. The Journal of
Neurosc (2001), 21(16):6405-6412.
154
14. Penn, J.S., Williams, T.P. Photostasis: regulation of
daily photon-catch by rat retinas in response to various
cyclic illuminances. Exp Eye Res (1986), 43:915-928.
15.Mure, L.S., Rieux, C., Hattar, S., Cooper, H.M.
Melanopsin-dependet nonvisual responses: evidence for
photopigment bistability in vivo. Journal of Biol
Rhythms (2007), 22(5): 411-424.
Reflections on the Eyelid:
Experiencing Structured Light through Closed Eyes
Adar Pelah
Department of Electronics
University of York
Heslington, York YO10 5DD
ap23@york.ac.uk
Siyuan Liu
St John’s College
University of Cambridge
Cambridge CB2 1PT
Howard Hock
Department of Psychology
Florida Atlantic University
777 Glades Road
Boca Raton, FL 33431-0991
Mathew Gilbert
Department of Electronics
University of York
Heslington, York YO10 5DD
Philip Jepson
Department of Electronics
University of York
Heslington, York YO10 5DD
Also at:
Department of Engineering,
University of Cambridge,
Trumpington Street,
Cambridge CB2 1PZ.
ABSTRACT
It is generally taken that closure of the eyes for periods
longer than a blink blocks visual perception due to the
presumed diffusion of image structure by the eyelids.
Although as much as 14.5% of the light incident at the
eyelid may reach the retina, a capacity to visually perceive
meaningful structure has not previously been proposed. We
report on visual experiments through the closed eyelid,
demonstrating the presence of both spatial and temporal
sensitivity. By Rayleigh’s criterion, we found a mean
spatial resolution of 21° for the closed eye (N=17), in
comparison with optimal open eye resolution of
approximately 0.008°. In addition, we found that motion
direction discrimination was qualitatively comparable to
performance with an open eye that was perceptually
matched with the closed eye for blur and brightness (N=8).
Confidence in making closed eye observations was
significantly lower than with open eyes, and subjects’ mean
blur and brightness matching using the open eye
overestimated (based on previous measurements of
transmission through the eyelid) the attenuation of light in
the closed eye by more than 50 times. A further observation
indicates that colour naming can also be made accurately
through closed eyes. Applications of the findings are
considered in the context of how light and pattern may be
experienced on the dark side of human vision.
Keywords
Closed eyes, eyelid, motion sensitivity, direction
discrimination, Rayleigh’s criterion, colour naming
INTRODUCTION
The present investigation challenges the widely-held
assumption that the world cannot be perceived visually
when the eyes are closed. Closing the eyes attenuates
significantly the characteristics of the retinal image
affecting the consequent reception by conscious (i.e.
perceptual, cognitive, cortical) and non-conscious (i.e.
circadian, reflexive, subcortical) neural mechanisms. The
factors leading to the attenuation of the retinal image when
the eyes are shut may be broken down into the spectral
filtering and diffusion applied by the skin of the eyelid
prior to the light reaching the retina. We examine
experimentally the extent of the spatial and temporal
attenuation. We consider whether, under controlled lighting
and display conditions, perceptual processing may still take
place through closed eyes, and if thus, what are the
implications, and what advantage can be made for lighting
and well being applications.
Spectral Filtering
Sensitivity when the eyes are closed is heavily reduced,
especially at short and medium wavelengths, with most of
the uniform light reaching the retina radiating in the ‘red’
region of the spectrum [2, 9, 13, 14]. Subjective
observation confirms readily the band-pass nature of the
eyelid filter at long wavelengths: for example, looking at an
intense, broad spectral light source (such as the sun)
normally produces an appearance through closed eyes of a
broadly homogenous field of light with a reddish hue.
The observation favouring a red coloured filtering effect by
the eyelid has also been confirmed by formal investigation.
Ando & Kripke’s [2] threshold measurements for the
detection of light passing through the eyelid indicated that
there was 94% attenuation for monochromatic red light,
compared with 99% for blue and green light. Physical
measurements find spectral characteristics for the eyelid
that are similar to other blood-bearing biological tissue.
Robinson et al [13] delivered monochromatic light through
a fibre-optic that was mounted onto a contact lens, with the
output detected on the outside skin of the eyelid using a
photodiode. Their data from 5 adult subjects indicate that
the eyelid acts as a predominantly red-pass filter with mean
155
transmissions at 700nm and with as much as 14.5% of the
light transmitted across the skin of the eyelid. Similar
measurements with 9 preterm neonates indicated the
transmission of 21.4% if the light.
Diffusion
In addition to heavy spectral filtering, a second attenuation
factor present when closing the eyes is the diffusion of the
spatial structure of the image incident at the eyelid, blurring
the retinal image prior to photoreception. The extent of the
blur is assumed implicitly in the literature, and commonly
by lay people, to be total and therefore the main cause for
blocking visual perception with closed eyes. It has never
been measured, to our knowledge, probably for this reason
Our first objective, therefore, is to estimate the blur due to
the eyelid by measuring spatial resolution to determine if it
is instead finite in extent and quantifiable under controlled
conditions.
Spatio-temporal Structure
Whether or not meaningful visual perception can take place
with closed eyes can be determined by measuring the
psychophysical performance of human observers in spatial
and temporal visual tasks. Comparing the results to
appropriately matched open eye equivalents would indicate
whether the underlying mechanisms are the same for openand shut-eye vision. Spatial resolution is often measured in
optical studies by the Rayleigh criterion [16], which is
defined as the minimal retinal angle subtended for which
the separation between two point sources can be resolved.
The minimal resolvable angle thus measured defines the
line-spread function (the inverse of the modulation transfer
function), which is readily converted to spatial resolution
(see, for example, Fig. 1).
Our second objective is to assess whether vision with
closed eyes can include functional perceptual properties in
the temporal domain. Motivated by the evolutionary need
for predator avoidance, perhaps the simplest task to
examine for temporal sensitivity across the eyelids would
be direction discrimination during motion. With closed
eyes under bright illumination - conceptually, for instance,
for predator avoidance, a shadow cast on the eyelid would
firstly need to be detected, and secondly, its direction of
motion would need to be discriminated (e.g. in order to take
the correct evasive action). This was addressed
experimentally by testing our subjects’ ability to detect
motion direction for the brief presentation of a vertical bar
drifting to the left or right, measured separately for the shut
and open eye, with brightness and blur perceptually
matched beforehand.
Our final objective, if visual perception can be shown to
take place through closed eyes, is to consider what use can
be made of the reported visual property, especially for the
benefit of human health and wellness. We consider in our
discussion categories of eye closure and discuss possible
directions for applying the finding within each of them.
METHODS
Experimental methods are described below for the overall
setup, and separately for the spatial resolution (Experiment
1), perceptual matching (Experiment 2) and direction
discrimination (Experiment 3) experiments.
Figure 2. The experimental setup. Left: A subject is seated in the
experimental booth; during experiments the curtain would be
drawn for complete darkness. Right: The subject’s chin rest and
stimulus display array. An illuminated vertical column of the red
LEDs is shown. For Experiment 1, the viewing distance is
changed by moving the display array along calibrated track (partly
visible on the lower left).
Figure 1. Illustration comparing the Snellen optometric chart
visual acuity for an open eye (barely visible ‘A’, left) to the
theoretical Snellen acuity of a closed eye through the eyelid (large
‘A’, right). The relative letter sizes are derived from the ‘worst’
Snellen acuity (open eye), and ‘best’ Snellen acuity (closed eye)
calculated from the mean Rayleigh resolution (see Experiment 1).
156
Experimental Setup
Physiological optics generally neglects the optical
properties of the eyelid (or treats it as opaque), though it is
rightly the first stage of the visual pathway when the eyes
are closed. The eyelid can be modeled as a spectral filter,
attenuating light differentially across the visible spectrum,
placed in series with a diffuser that blurs the image.
Spectral attenuation by the eyelid is highest at the blue and
green regions of the spectrum, with most of the energy at
the photoreceptors remaining in the red (due to the optical
properties of skin and blood) [2, 9, 13, 14]. As our
objective for obtaining resolution thresholds was to
measure the blur that includes the degrading effects of the
eyelid we maximised effective transmission in the
apparatus by restricting stimuli to light in the red region of
the spectrum. We used a custom, computer controlled
stimulus grid designed and built in-house consisting of 9
rows by 9 columns of bright, red LEDs spaced 1.0 cm apart
along each of the vertical and horizontal axes of the grid
(see Fig. 2, Right Panel).
The spectral peak of each LED in the display array was 635
nm with a bandwidth of 45nm at half height. Experimental
observation was done monocularly, with one eye closed
and the other patched to full darkness at a given trial, using
a chin-rest for stability and a Maxwellian view that allowed
subjects to sit at comfortably close distances to the stimulus
grid centred at the eye (see Left and Right Panels of Fig. 2).
LEDs that were lit were displayed at equal, high brightness
levels. The total luminance flux per area was determined by
measuring, with a photometer, the luminance of a single
LED (269 cd/m2), multiplying this by the area of an LED
(circular, 3mm diameter) to get the amount of luminous
flux that the LED would contribute to the display. The
single LED’s value is then multiplied by the number of
LEDs illuminated in a given stimulus configuration to
obtain the total flux, which is then divided by the area
encompassed by the illuminated cluster. In Experiment 1,
the 4 x 7 LED display (Fig. 3, Left Panel) has a luminance
of 30.0 cd/m2, while the respective luminance values for the
display in Experiments 2 and 3 (see Fig. 4) is 33.3 cd/m2.
Informed consent was obtained from subjects prior to their
participation; all subjects were between the ages of 20-23
years and had normal or corrected to normal vision through
contact lenses. Subjects first dark adapted for at least 20
min prior to the commencement of trials and were allowed
at least 10 min of practice to become familiar with the
nature of the apparatus and the experiments. Measurements
for spatial resolution were made using the method of limits
(the viewing distance was manually manipulated by the
experimenter in order to vary the size of the retinal angle
subtended by the test stimulus), while those for directional
discrimination were done using a 2-alternative, forced
choice (2AFC) paradigm by randomly varying the motion
direction of the stimulus. Subject responses were made
verbally in Experiment 1 and 2, and by button presses
corresponding to ‘left’ or ‘right’ direction with a hand-held
keyboard in Experiment 3.
Experiment 1 – Rayleigh resolution
Spatial resolution thresholds were determined on the basis
of the Rayleigh criterion [14], defined as the minimum
resolvable detail, as limited by factors such as diffraction,
blur and noise. Functionally, the Rayleigh criterion is
measured as the smallest retinal angle at which a gap is
resolved between two adjacent point sources. To enable
better performance through brighter stimuli, we used
columns of LEDs rather than point sources, thus obtaining
resolution based on the alternative 1-D optical line-spread
function rather than the 2-D point-spread function [see 5}.
The Rayleigh criterion was thus taken as the maximal
distance between the eyelid and the array at which a 2.0 cm
gap between two vertical LED columns was resolved and
discriminated from 1.0 cm gap (see Fig. 3). The two frames
were presented alternately for a 1.0 s duration each
interleaved by a dark frame of the same duration. Viewing
distance was used to estimate thresholds for resolving the
fixed stimulus gap, owing to the limitations in the range of
resolutions possible with our custom built bright LED
display apparatus. A confound with stimulus luminance
therefore occurs owing to light attenuation with increasing
distance; superior resolution values are likely to be
achieved using more advanced displays with higher
luminance values that are held constant, while varying
stimulus resolution directly. 17 subjects participated in this
experiment and each subject’s testing session required
approximately 30 min.
Figure 3. The display frames used for measuring Rayleigh
criterion for Experiment 1. The two frames were alternated
repeatedly, interleaved at 1 Hz by a blank frame (not shown). The
physical gap between adjacent columns was 1cm (which was not
detectable at the range of viewing distances used), thus the gap to
be resolved in the second frame was 2cm wide (see arrow, right
frame). The subjects were required to correctly identify and
verbally confirm the visibility of the gap, with viewing distance
manually adjusted to the largest distance, and therefore the
smallest retinal angle, for which subjects correctly named the
frame containing the gap.
Experiment 2 – Perceptual matching
Quantitatively comparing open and closed eye vision
required that the stimulus properties of the former be
perceptually matched to that of the latter. Subjects were
asked to wear a custom-made facemask to which layers of
neutral density (ND) filters and sheets of tracing paper
could be attached in front of each eye’s view, for light
attenuation and, in addition, blur. Each ND filter attenuated
light intensity by 75.0%, while tracing paper attenuated
light intensity by 97.0% per sheet. We then presented
subjects with a bright, steady LED array stimulus
157
consisting of three vertical columns, asking subjects to
match the percept seen in the closed left eye with that of the
open right eye, by covering each eye alternately while
observing the stimulus with the other. We added and
removed filters and sheets of tracing paper to the open
eye’s view until the subject reported no difference between
the percepts for both eyes.
A subject’s settings therefore consisted of a finite number
of ND filters and tracing paper sheets, from which the
viewed luminance may be computed for the open eye as the
perceptual match to the closed eye. If settings were based
on veridical luminance perception, as measured for light
transmission through the eyelid by other studies [2, 13],
then we would expect filter settings for the open eye’s view
that achieve between 5% and 14.5% light transmission.
13 subjects participated in this experiment and each
subject’s testing session required approximately 15 min.
One additional subject reported difficulty doing the
matches and also produced extraordinary results. These
data were therefore excluded from the sample as an outlier.
Figure 5. The distribution of viewing distance for spatial
resolution measured according to Rayleigh’s criterion in
Experiment 1. The mean, shown at the vertical red line, is
5.409cm, which is equivalent to 20.95°±8.44° of visual angle
subtended at the nodal point of the eye. The standard deviation,
shown by the blue lines, is ± 1.96,. N=17.
Experiment 3 – Direction discrimination
Experiment 2 – Perceptual matching
Directional discrimination thresholds were determined as
the percentage of correct responses (‘left’ or ‘right’) for
three vertical columns of illuminating LEDs drifting in the
horizontal direction over a sequence of three 400 ms frames
(2.5 Hz), as illustrated in Fig. 4. The viewing distance used
to present the stimuli was 2.0 cm, resulting in a temporal
drift velocity of ±70 °/s. Stimulus brightness was varied by
placing the appropriate number of layers of ND filters in
front of the LED array.
Figure 4. The display frames used for discriminating direction of
drift in Experiment 3. The frames were presented in sequence at a
rate of 2.5 Hz from left to right, as above; the illustrated sequence
corresponds to rightward motion. Blank frames appear as shown
at the start and end of each presented sequence.
RESULTS
Experiment 1 – Rayleigh resolution
Spatial resolution thresholds were obtained from onedimensional Rayleigh criterion judgments. Expressed in
degrees of visual angle at the nodal point of the eye, the
mean (± SD) Rayleigh resolution was 20.95° (±8.44°). The
viewing distance values across all subjects are normally
distributed, as shown in Fig. 5. The mean Rayleigh
resolution may be compared to the equivalent resolution for
open eyed vision. Under optimal conditions, the
diffraction-limited, optical resolution [e.g. 5] has been
shown to match that of the neural pathways [6] to a value
of approximately 0.0083°.
158
Perceptual matching was obtained, as explained in
Methods, between the percept from a bright light as viewed
by an open eye through filters to that seen in the closed eye.
The filter settings (also used for Experiment 3) show a
considerable mean ((± SD) over-estimation of the amount of
brightness attenuation taking place across the closed eyelid
of 0.09% (± 0.21) transmission (i.e. 99.01% attenuation);
this may be compared with published estimates of
approximately 5% transmission (i.e. 95% attenuation) for
similar perceptual measurements [2]. This is more than a
50-fold over-estimation. In addition, there are large
differences in subjects’ individual perceptual settings. (see
Table 1, column 3).
Sub
ID
Matched
filters
(ND & TP)
Transmissi
on
(%)
1
4&3
1.05E-05
2
2&2
5.63E-03
3
2&2
5.63E-03
4
2&1
1.88E-01
5
4&2
3.52E-04
6
1&8
1.64E-11
7
1&4
2.03E-05
8
2&6
4.56E-09
9
2&2
5.63E-03
10
1&1
7.50E-01
11
2&2
5.63E-03
12
2&2
5.63E-03
13
2&1
1.88E-01
Table 1. Subjects’ individual filter settings in Experiment 2
selected for the open eye to match the perceived appearance of a
bright light source seen through the eyelids in the closed eye. The
first number in the middle column indicates the number of neutral
density filters, each with 75% light transmission. The second
number in the middle column indicates the number of sheets of
tracing paper, each with a 97% light transmission. The mean
transmission percentage (± 1 SD) through the matched open eye
filter for all subjects was 0.09% (± 0.21), a large overestimation of
the attenuation (compared to earlier reports), or equivalently, an
underestimation of brightness through the eyelid.
Experiment 3 – Direction discrimination
Directional discrimination thresholds for 8 subjects for
open and closed eye conditions are shown in Fig. 6. Results
show clearly that motion direction can be discriminated
through a closed eyelid. This appears to take place in a
qualitatively similar manner as a function of stimulus
luminance as it does through a perceptually matched open
eye. However, it is also clear that performance in the open
eye remains superior to that of the closed eye by
approximately 15-20% under the conditions in the present
experimental setup. The quantitative difference is seen
despite the filtering applied individually for each subject
(see Table 1, and Experiment 2) to match the percepts in
the two conditions. As indicated earlier, the matching made
the open eye’s view darker than would have been expected
from previous measurements [2, 13]. So the difference in
direction discrimination between closed and open eyes may
have been under-estimated in this experiment.
Figure 6. Mean subject performance (% correct) in discriminating
the direction of motion monocularly through an open or closed
eyelid. In the open eye condition subjects viewed the stimulus
through tracing paper and a series of neutral density filters,
selected individually as a perceptual brightness and blur match to
their closed eye percept. Discrimination is easier through an open
eye despite the matched percept and probable overestimation of
attenuation in the perceptual matching (see Table 1), though
performance appears qualitatively similar in the two conditions.
Errors bars are shown as ± standard errors of the mean. N=8.
A Colour Naming Observation
The spectral filtering of the eyelid is well known and
although colour perception behind the closed eyelid has not
explored, the filtration is thought largely to impoverish the
retinal image of chromatic information except in the red.
The experiments we report make use of a band limited red
LED display panel in order to maximize the brightness at
the most effective spectral range, namely in the red.
However, we also make the following observation. We
presented single, bright (uncalibrated) LED key-chain type
flashlights in red, green, white and blue, applied in random
order at the surface of the eyelid to one closed eye (the
other eye was patched). The five subjects tested, with 10 or
more presentations of each colour, were able to name the
colour to near complete accuracy. Thus, human perception
of colour, and thus the presence of cone vision, is not
precluded when the eyes are closed.
DISCUSSION
The investigation began with questioning the commonplace
presumption that active visual perception is extinguished
when the eyes are shut. Indeed, in basic psychophysical,
neurophysiological, and applied vision research, a closed
eye condition is often used as the controlled ‘no-vision’
condition. There undoubtedly is extensive attenuation of
visual information through closed eyes. However, casual
observation (through introspection) of a small uniform light
source through closed eyes has motivated us to consider
whether structural spatiotemporal image information may
be detected and perceived under certain circumstances; and
if this is so, whether such shut-eye visual processing may
serve to benefit novel applications in medicine, lighting
and imaging research, as well as improve our
understanding of visual processing.
Our preliminary survey of visual capacity through closed
eyes of two visual spatiotemporal parameters, spatial
resolution and directional discrimination, demonstrate that
spatiotemporal structure can be perceived in visual images
seen through the closed eyelids. The novelty of making
conscious observations with closed eyes, and probably the
reduced control of eye position sense [1], leads subjects to
report a reduced confidence in their judgments. When
questioned at the end of experimental runs, subjects
reported mean (± SD) confidence1 in their direction
discrimination of 7.63 (±1.97), with the eye open,
compared with only 4.76 (±2.48) with it closed, a reduction
in confidence of 38%. The reduced confidence cannot
1
Confidence scale 0 to 10; 0 = ‘no confidence’, 10 =
‘complete confidence.’
159
explain the underestimation of perceived brightness
through the eyelid (see Table 1), as uncertainty would more
likely lead to a large variance but not necessarily an order
of magnitude or two of underestimation in perceived
brightness of the closed-eye.
Subject accuracy, particularly for direction discrimination,
is nevertheless good. Percentage correct for the most
visibly (i.e. brightest) stimulus is only 14% lower for the
closed eye (see Fig. 6). A likely reason for the inferior
direction discrimination of the closed eye is the absence of
position information due to the severe loss in spatial
resolution through the eyelid. Indeed, the discrimination of
motion direction despite such severely impaired spatial
resolution, provides a clear instance of objectless motion
perception [15], as also observed for blindsight patients [3].
There is no indication, however, that performance would
not continue to improve with further increases in image
brightness.
Spatial resolution, as measured by the Rayleigh criterion, is
extremely poor compared with open-eye vision (see Fig. 1),
yet surprisingly good compared with what might be
expected, namely, an absence of any resolvable detail when
the eyes are shut. The poor resolution is due to the blur and
brightness reduction of the eyelid caused by diffusion. The
expected dominance of lower-resolution, rod-mediated
vision operating at the eyelid-attenuated light levels
following dark adaptation would not be expected to be a
significant factor, as the rods perform poorly for red light.
In addition, we found that colour perception, the hallmark
of cone-mediated vision, can take place with closed eyes.
Our estimates are therefore conservative, given the greater
resolving power of the cones, and we believe that
significantly higher resolutions than reported here are likely
to be detected with improved conditions.
Extended periods of eye closure are normally associated
with sleep, when only relatively large and abrupt changes
in ambient light level seem to affect behaviour (e.g. by
waking or stirring the sleeper). On the other hand,
sufficiently slow, gradual changes in ambient level are not
likely to be perceived consciously through closed eyes (or
even open eyes), although some light transmission through
the closed eyelid would presumably be advantageous to
assist circadian training as dawn approaches. For structured
light, it may be important, for ecological reasons of
predator avoidance for instance, to collect at least the edges
and direction of movement of an approaching shadow cast
on the eyelids. It is therefore not unreasonable on first
principles to expect that eyelids should be designed through
evolution to allow for transmission of light and some
images.
It is promising and attractive to learn that a richer and more
complex appreciation of the environment can be obtained
from light incident on the closed lids of the eyes. While the
underlying mechanisms for visual perception during eye
closure are not fully understood, our finding that structured
spatiotemporal light is perceptible behind the eyelid could
160
have applications in medicine, architecture, education or
entertainment.
In relation to our final objective for this study, we therefore
consider that people close their eyes within three categories
of experience, and suggest possible thinking on
applications within each. Naturally, each will present its
own challenges and unanswered questions on
implementation are outside the scope of this paper.
1) Reflex and maintenance, as with natural blinks for
moistening the eye surface, and protection from bright light
or physical objects nearing the eye. The prospect that useful
visual information can be delivered to the eyes during brief
periods of closure could be applied to high-speed physical
and informational activities, for instance in sports, battle, or
other time-critical safety monitoring scenarios, such as car
racing or air-traffic control.
2) Communications and emotions, as in facial expressions,
or responses to enjoyment, fear etc. Within this category,
one could envisage additional modalities of communicative
information conveyed visually during eye closure, for
instance, to enhance or modify the emotional state or
convey more subjective information during these intervals
during video telephony (e.g. Skype) type communications.
The third category to consider is 3) Sleep, relaxation and
related states of longer duration eye closure. During eye
closure brain activity is distributed differently, and eye
movement velocities are greater, as reported by. Marx et al
[8]. Interestingly, they also argue from brain imaging
(fMRI) results concerning the presence of two distinct
mental states: an ‘exteroceptive’ state when the eyes are
open, characterized by attention and oculomotor activity,
and an ‘interoceptive’ state during extended periods of eye
closure, dominated by imagination and multisensory
activity. Perhaps applying custom-structured light displays
during eye closure to awake individuals could cross such
normal boundaries, by instigating a mixture of perceptual
processes and unusual experiential effects?
Differences are found for certain medical conditions that
may support this view. For example, the eye movement
patterns of schizophrenics are different when their eyes are
open compared to when their eyes are closed [10].
Applications in this category may thus include visual
stimulation during eye closure for schizophrenic patients,
for epilepsy patients [7], who demonstrate abnormal brain
patterns during eye closure [4], fro coma patients, who
have abnormal sleep-awake patterns, and for other patients
with ‘disorders of consciousness’ (i.e. coma, vegetative
state and minimally conscious state) [11, 12]. It is
conceivable that stimulation during eye closure periods for
such patients could stimulate different brain areas and
trigger alternative pathways to facilitate diagnosis or
treatment.
ACKNOWLEDGMENTS
The work was completed in Cambridge as part of the final
year Neuroscience part 2 NST project by SL. We thank
Prof John Mollon and Prof Roger Carpenter for helpful
discussions; Dr Joan Lasenby and Dr Jonathan Cameron for
assistance and facilities; and our experimental subjects for
their enduring patience during lengthy trials.
REFERENCES
1. J. Allik, M. Rauk, A. Luuk, Control and sense of eye
movements behind closed eyelids, Perception, 10, 1,
39-51. 1981.
2. K. Ando & D.F. Kripke, light attenuation by the human
eyelid, Biological Psychiatry, 39, 1, pp 22-25, 1996.
8. E. Marx, T. Stephan, A. Nolte, A. Deutschlander, K.C.
Seelos, M. Dieterich & T. Brandt, Eye closure in
darkness animates sensory systems. NeuroImage, 19,
924-934. 2003.
9. M.J. Mosely, S.C. Bayliss & A.R. Fielder, Light
transmission through the human eyelid: In vivo
measurement. Ophthalmic and Physiological Optics, 8,
229-230, 1988.
10. K. Nakajima, Eye movements in schizophrenia-relationships among eye movements under three
experimental conditions; closed-eye, pursuit, and
exploratory Seishin Shinkeigaku Zasshi. (Article in
Japanese) 92(8):509-31. 1992.
11. A.M. Owen & M.R. Coleman, Functional neuroimaging
of the vegetative state. Nature Reviews Neuroscience, 9:
235-243. 2008.
3. P. Azzopardi, & H.S. Hock, Dissociation of featurebased motion and 'objectless' motion energy for
direction discrimination within the sighted and
blindsighted visual fields of a hemianope [Abstract].
Journal
of
Vision,
9(8):761,
761a,
http://journalofvision.org/9/8/761/, doi:10.1167/9.8.761.
2009.
12. A.M. Owen, M.R. Coleman, M. Boly, M.H. Davis, S.
Laureys, J.D. Pickard, Detecting awareness in the
vegetative state.” Science, 313(5792):1402. 2006.
4. C. Di Bonaventura, A.E. Vaudano, M. Carn`ı, P.
Pantano, V. Nucciarelli, G. Garreffa, B. Maraviglia, M.
Prencipe, L. Bozzao, M. Manfredi, & A.T. Giallonardo,
Long-term Reproducibility of fMRI Activation in
Epilepsy Patients with Fixation Off Sensitivity.
Epilepsia, 46(7):1149–1151, 2005.
14. E. Simonson, S.S. Blankstein & E.J. Carey, The
efficiency of the glare reduction by the eyelids,
American Journal of Physiology, 143, pp 541-547,
1945.
13. J. Robinson, S.C. Bayliss & A.R. Fielder, Transmission
of light across the adult and neonatal eyelid in vivo,
Vision Research, 31, 10, pp1837-1840, 1991
5. F.W. Campbell & D.G. Green, Optical and retinal
factors affecting resolution, Journal of Physiology, 181,
576-593. 1965.
15. G. Sperling & Z.-L. Lu, A systems analysis of visual
motion perception. In T. Watanabe (Ed.), High-level
motion processing. Cambridge, MA: MIT Press. Pp.
153-183. 1998.
6. F.W. Campbell & R.W. Gubisch, Optical quality of the
human eye, Journal of Physiology, 186, 558-578. 1966.
16. R. Yuste & A. Konnerth, Imaging in neuroscience and
development, CSHL Press, p 45. 2005.
7. F.S.S. Leijtena,*, E. Dekkerb, H. Spekreijsec, D.G.A.
Kasteleijn-Nolst Trenite´ b, W. Van Emde Boasb, Light
diffusion
in
photosensitive
epilepsy.
Electroencephalography and clinical Neurophysiology
106, 387–391. 1998.
161
Evaluation of Today!s Research Methods for Assessing
Light-Induced Alertness
Emilia Rautkylä, Petteri Teikari, Marjukka Puolakka, and Liisa Halonen
Lighting Unit
Helsinki University of Technology
P.O. Box 3340
02015 TKK Finland
+358 50 5469255
emmi.rautkyla@tkk.fi
ABSTRACT
Daytime alertness has not established the same type of
research routines as nighttime alertness. This paper
evaluates today’s research methods to provide help in
choosing suitable methods for assessing light-induced
daytime alertness in future research. The evaluation is done
according to two main criteria; the method’s ability to
reflect alertness physiologically and its suitability for use in
a lighting study. The methods under evaluation are
subjective ratings, reaction tests, brain imaging,
pupillometry, measurements of heart rate and skin
conductance, and the use of an electro-oculogram and
electroencephalogram. On the basis of a literature review
and practical testing of some of the methods, the writers
suggest that in low-cost studies, where detecting the effects
is enough, autonomic nervous system activation should be
measured. To gain broader knowledge of the mechanisms,
central nervous system responses need to be studied.
Keywords
Alertness, sleepiness, central nervous system, autonomic
nervous system activation, lighting research
INTRODUCTION
With growing interest in the impact of light on health [60,
70] and wellbeing [84], accelerated by the introduction of
the novel photoreceptive melanopsin-containing retinal
ganglion cell [10], the relation of light and alertness during
nighttime hours has been the subject of considerable
examination [for a review, see 17]. Daytime alertness, on
the other hand, has not yet attracted the same amount of
research interest. In general there are two approaches to this
field of interest. One aims at detecting the effects of light
[6] so that the current lighting standards used in the
building business could be updated so as also to consider
wellbeing instead of only the visual performance [81]. This
type of research is clearly application-oriented and of high
commercial value [79], which can be seen from the high
level of interest in e.g. bright light treatment. The other
type of research tries to reveal the mechanisms behind the
detected effects [4]. This can be thought to be a more
rigorous approach, trying to find out the scientific basis for
the phenomenon.
162
The current weakness in both types of research is that the
methods used for evaluating light-induced daytime
alertness are not always suitable for such research. First of
all, it is often forgotten what alertness is, physiologically,
and hence which methods measure human alertness
quantitatively. Second, it is not always considered that
using light as a stimulus can create special demands for the
method. Finding proper methods is challenging because in
lighting research papers the problems related to the
usability of the study methods chosen are reported too
seldom.
The object of this paper is to provide help in choosing
eligible methods for assessing light-induced daytime
alertness by evaluating the research methods used today.
This is done according to two main criteria: the ability of
the method to reflect alertness physiologically and its
suitability for use in a lighting study. Besides evaluating
today’s research methods theoretically, the paper also
reports on the practical testing of some of the methods,
namely subjective evaluation, and measuring pupil size,
heart rate, and the skin’s ability to conduct electricity. The
methods are chosen for the practical testing on the basis of
their potential value for alertness research and on the
available resources.
EVALUATION CRITERIA
Physiology of Alertness
In 1949 Moruzzi and Magoun presented a theory of the
activation system of the brainstem which suggests that
stimuli travelling through the brainstem give rise to the
level of alertness [58]. This was thought to be the missing
explanation of why external stimuli increase alertness but a
lack of them reduces it. Soon the concept of the “ascending
reticular activating system” achieved wide currency.
Anatomically speaking, the activating system is located in
the reticular formation. The reticular formation is a broad
and netlike formation in the core of the brainstem running
through the mid-brain, pons, and medulla oblongata. The
ascending reticular activating system is connected to areas
in the thalamus, hypothalamus, and cerebral cortex, while
the descending reticular activating system is connected to
the sensory nerves of the cerebellum. [31]
The regulation of alertness is based on the messages sent
between the nuclei in the reticular formation in the
brainstem and the cerebral cortex [68]. Of the nuclei in the
brainstem, the locus coeruleus (LC) is the most essential
one when alertness is considered, because it reacts easily to
stimuli and improves alertness by increasing its
noradrenalin secretion to the cortex. It has been said that a
single noradrenergic neuron can innervate the entire
cerebral cortex via its branches, mediating arousal and
priming the brain’s neurons to be activated by stimuli [6].
That supports the belief that the LC is one of the key
components in light-induced alertness [4].
How the function of the LC is seen in practice is through
the secretion of noradrenalin, which enables the body to
perform well in stressful situations [8]. Noradrenalin
normally produces effects such as increased heart rate,
blood pressure, and sweat gland activity, and the dilation of
the pupils and of air passages in the lungs. Hence, the LC is
in direct connection to the autonomic nervous system
(ANS). In fact, the activation of the ANS is often used as a
conceptual definition of alertness. Because the activation of
the ANS increases as arousal increases, it is reasonable to
claim that by observing the changes in the ANS, it is
possible to see how the LC reacts to stimuli and activates
the body.
This section concludes that one potential way to assess
daytime alertness is by observing the brain and
characterising the neural correlates of the alerting effects of
light. Another option is to observe the activation of the
autonomic system. Although the pathways have not yet
been fully identified, there is evidence that light stimulates
the ascending arousal system and eventually the cortex in
order to enhance alertness [65]. It should be noted that it
still remains unclear whether light induces alerting effects
in daytime, when the homeostatic sleep pressure is low and
there is no circadian drive for sleep. Therefore the methods
need to be even more sensitive than the methods used in
nighttime studies.
Demands Set by Lighting Research
In general a good method for use in a scientific study is
something that can be applied both in laboratory and field
studies. This permits the best further use of the results in
real-life applications. To gain objective and reliable data
the experiment should be repeatable and as independent of
the subject and the researcher as possible. Setting a baseline
or altering the testing conditions often helps in analysing
the data and verifying the results.
In addition to these general guidelines, lighting research
sets some special demands for the study method. The most
essential demand is that the method allows light to be used
as a stimulus. Often it is also important to be able to alter
the lighting conditions, hence changing the exposure time,
light source, spectrum of the light, irradiance etc. [62]. One
crucial criterion for the method is that it presents the data in
such a way that the effect of light can be distinguished from
effects caused by other stimuli such as caffeinated
beverages [30], indoor climate [14], and auditory stimuli
[27], among others. There are two ways to do this: either
eliminating other stimuli from the set-up or separating them
out in the analysis.
One practical factor to consider in choosing a good method
is that in research done on humans there is considerable
variation between individuals [65]. Therefore the number
of subjects has to be big enough to gain reliability. This
results in the fact that the test cannot be too complicated to
conduct. Practicalities such as the costs and availability of
the method may also limit the use of some methods. In
studies of long-term light exposure and its effects, it is
important to make sure that the lighting conditions and the
method are not too burdensome and uncomfortable for the
subject.
THEORETICAL EVALUATION OF METHODS
Subjective Methods
Subjective evaluation is a commonly used research method
because it is easy to conduct both in laboratory and field
conditions. Sleepiness is typically assessed on a Likert-type
discrete scale [i.e. 32] or a continuous visual analogue scale
(VAS) anchored by word descriptors at each end [76].
Perhaps the most popular subjective measure is the
Karolinska Sleepiness Scale (KSS) [88], which uses a
discrete scale from 1 to 9, where 1 = very alert and 9 = very
sleepy, great effort to stay awake or fighting sleep. KSS has
been validated to significantly correlate with EEG and
behavioural variables [39] and is therefore considered a
reliable measure of alertness or sleepiness. However, as
Cajochen points out [17], the precise meaning of the terms
alertness and sleepiness may differ between languages and
situations. Furthermore, approaching alertness through
sleepiness can be inadequate because alertness is not
always the inverse of sleepiness [57].
The main criticism of any type of subjective assessment
arises from the fact that it relies on self-reporting, leaving it
open to misinterpretation, unintended bias, and falsification
for any number of reasons (e.g. the act of rating itself can
affect sleepiness [42]). It is possible that the subjects may
evaluate their alertness differently in light than in darkness,
even though there was no real difference in their level of
alertness. In fact, there is no real placebo control for light,
but it can only be hoped that the subject assesses his
alertness time after time following the same logic. Another
weakness of subjective assessment in a study of lightinduced alertness is that it can only be used to point out the
changes in the subject’s way of responding to light, but it
does not show anything about the reasons or mechanism
behind the changes. Therefore it can never give as much
input to the study as objective measures that are linked to
physiology.
It is also hard to be sure that the effect is caused by light
and not something else. Using subjective assessments thus
requires a very strictly controlled test environment where
there are no other factors that could affect the person’s way
of answering. Finally, one major disadvantage of using
163
subjective assessment in lighting studies is that the data
recording is not continuous. Because self-reports are
produced after certain time periods, the information about
the state of alertness between the measuring points is
automatically lost. One could say that the data expire as
soon as they are recorded. This is a big problem because it
hinders one in detecting whether it is a question of a fast or
slow response to light.
Objective Methods
Reaction Tests
Performance is often used to evaluate how alert a person is
[85]. This is based on the assumption that alertness is
involved with increased reactivity to external stimuli; thus
an alert person reacts fast to stimuli. Using reaction tests to
evaluate alertness is, however, not as problem-free and easy
as it would at first seem. For example, the test itself can act
as an activating stimulus and thus affect alertness.
The reaction tests need to be well designed in terms of the
complexity of the test because the subject should be able to
perform the test without too great an effort. Another
important factor is how the subject manages to retain their
motivation throughout the whole test. That depends on
whether the subject is being rewarded after a successful
test, but also on the duration of the test. For a long time it
was thought that a reaction test should last no less than 10
minutes because studies indicated that shortening a
performance task resulted in reduced sensitivity to changes
in performance [12]. However, recently the study of Roach
et al. [64] showed that a 5-minute test correlates well with a
10-minute test. They tested the psychomotor vigilance test
(PVT) [83], which has been shown to be a reliable indicator
of decreased alertness [19] and is commonly used for
assessing neurobehavioural performance.
The advantage of the PVT is that it reflects the tirednessrelated reduction in performance without being confounded
by the learning effect, a factor that often causes bias in the
experimental data. In the traditional study protocol a visual
stimulus appears on the display and the subject is instructed
to press the response button as fast as possible after
detecting the stimulus.
Another and more modern option is to use auditory stimuli
in a psychomotor vigilance test. In the auditory PVT the
subject presses a button after hearing the stimuli in the
same way as in the visual PVT. By using two buttons and
two different stimuli instead of one it is also possible to add
complexity to the test. Today there are portable, palm-held
devices that make it possible to conduct experiments in real
environments such as workplaces, instead of only
laboratories [83]. The Walter Reed Army Institute of
Research, Maryland, offers test and analysis software for
this kind of field-portable reaction time tester for free [75].
However, their PalmPVT does not allow auditory stimuli to
be used.
In theory, these kinds of reaction tests can be used in
lighting research in two ways. First of all, if light with the
specific characteristics under study acts as the stimulus, the
164
reaction time will show how easy it is to detect that
stimulus. In practice, however, this only shows that the
person reacts to light but not whether the light induces any
alerting effects. Another option is to use an exogenous
stimulus in the PVT to assess vigilance after being exposed
to light for a certain amount of time. Following e.g.
Lockley’s example [44], in this kind of protocol it is better
to use an auditory stimulus instead of a visual one to
prevent the PVT stimulus from masking the light-induced
effect under study. This has potential for revealing how the
exposure to light affects the reaction times.
The biggest disadvantage of using a PVT in studies of
light-induced alertness is that it measures sustained
attention rather than alertness and therefore it is not a
proper method to evaluate the activation system in detail.
Furthermore, it does not measure the functioning of the LC
or other body parts that take part in light-induced alertness
but instead it exhibits the circadian and homeostatic
processes that take care of the natural asleep/awake rhythm.
Therefore it can be concluded that a PVT can be broadly
used in chronobiology research [11] but it does not
necessarily make a good assessment method for alertness in
lighting research.
Pupillometry
The pupil provides control over the retinal illumination and
depth of focus [50]. In addition to constricting as a
response to increased light flux and vice versa, the pupil
also responds e.g. to accommodative changes [41] and to
anticipating effects for an instructed task [87], illustrating
the wide range of confounding factors involved in pupil
recordings. Given that pupil size modulates the retinal
illuminance, precautions are needed to control the exact
retinal illuminance. These precautions include monitoring
the pupil size via a video-based infrared pupillometer [73]
with or without dilating the pupil to a constant size during
the recording. Additionally, one can use a Maxwellian view
[82], as opposed to a free view in which the stimulus sizes
are smaller than the smallest physiological pupil diameter;
hence pupil size has no modulating effect on retinal
illuminance.
The pupil size can be measured using a direct approach
with binocular light stimulation or by a consensual
approach, where only one eye is stimulated and the
response of the unstimulated eye is recorded. There is no
vast literature on pupillometric hardware but it should be
noted that many of the approaches are similar to that used
in the eye tracking literature [20]. Typical temporal
resolutions range from 30 Hz in low-cost setups [43] to 612 kHz in more customised setups [74], with spatial
resolutions going down to 0.008 mm [29] depending on the
sensor resolution, quality of the optics, and the signal-tonoise ratio of the video signal. The increase in temporal
resolution can be achieved by using complementary metal
oxide semiconductor (CMOS) sensors instead of chargecoupled device (CCD) sensors, in which this is not possible
because of technical limitations [47].
Typically, the human pupillary light reflex (PLR) exhibits
roughly three phases, rapid phasic constriction in response
to light onset, which is followed by a steady-state pupil,
and finally, depending on the light stimulus, there can be
the post-stimulus persistence of a constricted pupil even
after light offset [26]. Additionally, pupil size exhibits
spontaneous fluctuations called hippus or pupillary noise,
which is characterised by a random noise in the frequency
range of 0.05 to 0.3 Hz [67]. To avoid contamination of
pupillary measurements by spontaneous fluctuations of the
pupil, a continuous monitoring of the pupil is preferred.
The exact origin of hippus is not fully understood, but it has
been suggested that it is an indicator of the state of
vigilance of a person. There is evidence that if a drowsy
subject spends daytime in darkness, the ‘fatigue waves’
start to occur with an increasing amplitude at the
frequencies 0.025–0.25 Hz [45], whereas in an alert subject
pupil size remains stable for a long time, oscillating mainly
with a frequency of 1 Hz.
For this reason, pupillary fluctuations have been widely
exploited as an easy and non-invasive measure to track
changes in autonomic nervous system activity. One
example of such an approach is the Pupillary Unrest Index
(PUI) which measures the cumulative changes in pupil size,
typically during periods ranging from 25 seconds to 15
minutes [53] in darkness or under light. PUI was used, for
example, by Szabó et al. [72] to measure the changes in the
vigilance levels of subjects during bright light exposure.
Among others, Nikolau et al. [59] found pupillary
assessment to be a promising objective tool to detect
pharmacologically induced changes in alertness. However,
it should be noted that the majority of studies on pupillary
fluctuations have been carried out in darkness and the
relationship between fatigue and oscillations in daylight
requires further validation, adding some restraints to reallife lighting studies with pupillometric alertness
assessment.
Considering that the pupillometer is comfortable for the
subject and the protocol does not include any tasks to be
performed, it might work as a good objective indicator for
light-induced alertness. The method operates with a fairly
delicate apparatus and requires the subject to sit still
without extra blinking and head movements. There are,
however, some indications that it could be used in field
studies too [78]. Recent studies suggest that pupil size
measurements could offer a simpler way to estimate
autonomic nervous system activity than the commonly used
heart rate [54]. Therefore it is reasonable to suggest that the
reactivity of the pupil could well be used in lighting-related
psychophysical experiments.
Heart Rate
The heart responds to psychological stress via the
autonomic nervous system [52]. Over the years a
correlation between heart rate and arousal caused by light
exposure has been found both with rats and with humans
[56,77]. Heart rate variability (HRV) has become the
conventionally accepted term to describe the variations of
interbeat (RR) intervals that represent autonomic nervous
activity [66].
Heart rate variability is normally recorded by placing 10
electrodes on the skin on the subject’s arms, legs, and
chest. They measure the activity of different parts of the
heart muscle and transmit it to an electrocardiogram (ECG)
machine. The machine produces an ECG tracing of these
cardiac electrical impulses. In clinical studies the heart
rates or cycle intervals are recorded over long time periods,
traditionally 24 hours, allowing more reliable calculations
of the measures. Because the analysis of HRV data is more
complex than generally appreciated, there is a potential for
incorrect conclusions and unfounded generalisations [25].
The experimental procedures and analysis of the results
should be carried out in accordance with the
recommendations of the European Society of Cardiology
and the North American Society of Pacing and
Electrophysiology [49]. In fact, Peña et al. [61] remind that
caution should be exercised concerning the use of short
recording segments, a circumstance not fully considered in
several studies. Therefore, although heart rate is easily
measured in the presence of a light stimulus, the method
does not meet the needs of detecting the effects of shortterm light exposure.
From the subject’s point of view, studying light-induced
alertness by observing the heart rate is an easy research
method because there is no task to be performed. However,
real clinical equipment contains detectors and wires that
can hinder the subject from conducting other things, as is
often required in field studies. Fortunately, there are
commercial and cost-effective heart rate monitors that can
be used in studies where it is possible to reduce accuracy in
order to gain mobility. There is evidence that motion does
not contaminate the signal too much [36]. It should,
however, be considered that in field studies the heart rate
data are even more sensitive to distractions than in
laboratory studies. Hence, the effects of the light stimulus
are easily masked by other unintended stimuli.
Skin conductance
Because of the connection between the autonomic nervous
system and locus coeruleus, arousal has long been assessed
through the skin’s ability to conduct electricity [28]. In fact,
activation theorists long considered skin conductance to be
the most appropriate measure of a generalised arousal
response [21]. Skin conductance, galvanic skin response,
and electrodermal response are different terms for the same
physiological measure. It is known that as a person
becomes more or less stressed, the skin’s conductance
increases or decreases proportionally [18].
The easiest way to measure electrodermal activity (EDA) is
by strapping two electrodes to two fingers, namely the little
finger and pointing finger of the non-dominant hand. The
skin acts as a resistor whose conductance (inverse of
resistance) changes with time according to the changes in
hydration in the sweat glands [23]. Changes in EDA occur
165
with even a slight rise or decrease in the amount of sweat
within the glands [69]. Therefore a typical signal recorded
from a skin conductor sensor shows relatively rapid
increases and slower decreases.
From the lighting research point of view measuring the
skin’s ability to conduct electricity can be considered a
good research method for light-induced alertness because it
can be used within and between light exposures with both
continuous and short-term light stimuli. It has, however,
not been used often in lighting studies. The recording
apparatus is small and the experimental protocol does not
involve any kind of task performance by the subject. One
major disadvantage is that the wiring hinders its use in reallife settings.
The analysis is rather easy as long as the recordings are
time-locked to specific events to allow the analyser to
select the right blocks of data from the general data [40].
The analysis has the potential to show the intensity of the
effect of the light on a human. However, it is important to
note that changes in the signal may be elicited by external
stimuli or internal events [46]. Hence, it might become hard
to distinguish the effects of different stimuli from one
another. Therefore, to make electrodermal activity a proper
indicator of the intensity of light-induced alertness, all
other emotional cues that might mask the effect of light
have to be eliminated.
Electro-oculogram
Eye movements react to a decrease in alertness. The
attenuation of blinking is often a marker of the fact that the
person is losing interest. At the same time the duration of a
blink becomes longer and the eyelids become lazier. When
the eyelid closes, the eyeball makes slow roll-like
horizontal movements that are called slow eye movements
(SEM) [5]. From these visible neurophysiological factors it
can be seen when the person is transiting from being awake
to asleep. Therefore eye blink rate and SEMs are
considered reliable correlates of human alertness [16].
Clinical alertness evaluation takes advantage of the
knowledge that a person who is not alert finds it hard to
follow targets. In the electrophysiological test called an
electro-oculogram (EOG) two skin electrodes are placed as
close as possible to both eyes. Moving the eyes induces a
voltage between them. The voltage varies from one to
several millivolts, depending on the ambient retinal
illumination. The subject is instructed to look back and
forth at a steady fixation rate between two fixation targets
to generate consistent saccades. These saccades are
amplified and registered to be considered for analysis [51].
Normally, EOG amplitude increases significantly if the eye
is first kept in darkness and then in light [3]. However, it
has been shown that in electro-oculogram analysis it is
better to use the light-peak to dark-trough amplitude ratio
instead of the actual amplitude values because the
amplitude varies widely among individuals [13].
The method is well suited to use in lighting research, both
during and between light exposures. However, when
166
designing the light stimulus, it is important to make sure
that the entire visual field is evenly illuminated and that
there is no direct glare on the subject that could hinder the
subject from focusing on the targets [51]. As the eyes
alternate direction every 1 to 2.5 seconds, the test soon
becomes uncomfortable and tiresome for the subject.
Therefore it is advisable to record the movements in sets
and let the eyes rest between the sets. According to the
international standard approved by the International Society
for Clinical Electrophysiology of Vision (ISCEV), one set
of 10 saccades per minute is enough to recognise the
relevant peaks and troughs in the EOG data. The standard
for EOG technology and protocol also offers other valuable
recommendations for the recording technique. facilitating
the comparability of the EOG data throughout the world.
A drawback in using an EOG to study light-induced
alertness is that it does not allow the subject to concentrate
on other tasks at the same time. That, and the presence of
skin detectors and recording apparatus, makes the method
unsuitable for real-life settings. Despite its few
impracticalities, the EOG technique is quite commonly
used to asses alertness objectively [34], either alone or
together with brain activity measures.
Electroencephalogram
A number of observations suggest that there is a possible
causal link between the activity of the locus coeruleus and
electroencephalographic (EEG) activation [24]. Because
the activation of the LC has been shown to induce EEG
signs of cortical and hippocampal activation [9], it is
reasonable to claim that by observing the forebrain EEG
activity it might be possible to monitor the alerting process.
Electroencephalographic activation is a direct measure of
the general cortical activation level. A set of electrodes is
placed on the subject’s skull to detect and amplify the small
electrical voltages that are generated by brain neurons when
they fire. Similarly to muscle fibres, neurons in different
locations can fire at different rates. The EEG is typically
described in terms of rhythmic activity and transients. The
rhythmic activity is divided into bands by frequency.
Jung and Makeig state that it is possible to use the EEG
power spectrum to estimate alertness [38]. The spectrum
Beta band (15-20 Hz) is generally regarded as a normal
rhythm, which explains why changes in Beta activity are
often used to reflect different levels of arousal [7]. An
decrease in Alpha activity (8-13 Hz) has also been reported
to be associated with a drop in alertness and cognitive
performance across the waking day [86]. This means that
high levels of EEG Alpha activity could indicate a high
level of alertness during an eyes-open condition [15],
similarly to Beta. Theta (4-8 Hz) and Delta (2-4 Hz)
activity are linked to increased drowsiness and reductions
in performance [48]. However, Theta and Delta activity are
rarer in awake adults.
EEG has both advantages and limitations in alertness
research. One of the advantages as a correlate to human
alertness is that it measures the brain's electrical activity
directly, while other methods record the responses of the
autonomic system. Another advantage is that EEG is
capable of detecting changes in electrical activity in the
brain on a millisecond time scale. Compared to techniques
such as functional magnetic resonance imaging (fMRI) that
have a time resolution between seconds and minutes, EEG
has a much higher temporal resolution. However, the
spatial resolution of EEG is poor and therefore it is not able
to indicate the location of the activity of the brain. One
possibility is to use EEG simultaneously with fMRI, so that
data with a high temporal resolution can be recorded at the
same time as data with a high spatial resolution. However,
there are technical difficulties associated with analysing the
activity of the brain in exactly the same time frame.
Furthermore, currents can be induced in moving EEG
electrode wires as a result of the magnetic field of the MRI.
As a research method EEG is fairly comfortable for the
subject, because it records spontaneous brain activity in the
absence of tasks. Therefore light can easily act as a shortterm or continuous stimulus. Despite the easiness of the
study protocol, using it in real-life settings is complex
because of the wiring and its interference-prone nature.
Brain Imaging
Brain imaging provides an opportunity to study what is
really happening in a human as a result of light exposure.
There are two techniques, namely functional magnetic
resonance imaging (fMRI) and positron emission
tomography (PET), which provide an anatomical and a
functional view of the brain and are commonly used for
brain imaging [71].
fMRI measures changes in the blood flow to particular
areas of the brain. Through a process called the
hemodynamic response, blood releases oxygen to neurons,
creating magnetic signal variation. This variation can be
detected using an MRI scanner. PET, for one, detects
radioactive material that is injected or inhaled. The material
collects in the area of the brain being examined, where it
gives off energy in the form of gamma rays. [63]
The procedure and analysis of both techniques is rather
hard and requires knowledge of fields such as physics,
psychology,
neuroanatomy,
statistics,
and
electrophysiology. That is why it is not within the scope of
this paper to go deeper into the measurements. Instead, the
most important characteristics of the two techniques from
the lighting research point of view will be discussed.
With brain imaging it is rather easy to identify the precise
areas that are activated in the brainstem as a result of the
light. A conventional 1.5-Tesla fMRI scanner has a spatial
resolution of 3 mm and higher-strength magners may
decrease it down to 1 mm. PET is not as accurate; the
effective spatial resolution of PET remains 8-15 mm when
standard image processing procedures, such as a
smoothening filter, are used. Temporal resolution is also
superior with fMRI. However, compared to EEG, which
has a time resolution of only a single millisecond, PET and
fMRI are slow, because they can detect a new stimulus
only some seconds after the first stimulus. As a matter of
fact, with PET it is not at all possible to pick out neural
activation patterns associated with individual stimuli
measures, so event-related phenomena, such as the effect of
a short exposure to light, can only be detected with fMRI.
From the lighting research point of view this hinders the
use of subsequent light pulses as stimuli.
The strong magnetic field around the functional magnetic
resonance imaging scanner also causes other limitations on
using light as a stimulus. The light source cannot be
installed in the study room because the electricity will
interfere with the magnetic field. Instead, the light stimulus
has to be transmitted by an optic fibre, as was done recently
by Vandewalle and his colleagues [e.g. 80]. The same
problem arises when trying to measure other physiological
measures during the scans to help in the interpretation of
the brain imaging data. Generally speaking, it is possible to
measure EEG, EOG, EMG, ECG, or skin conductance only
during the scans to prevent the magnetic field from
inducing a current in the electrode wires. However, several
techniques are under development to deal with these issues
and there are already good experiences of recording fMRI
and EEG simultaneously [e.g. 55]. Positron emission
tomography, for one, is free of this kind of physical
limitations.
If the problems mentioned above are to be overcome there
are still many practical issues that impede the use of brain
imaging in a typical lighting study. First of all, there are
only 100-200 MRI centres and 20-30 PET centres
worldwide where the studies can be conducted. Needless to
say, not only can they not be used in field studies but in
laboratory studies too their use is very limited because of
their huge expense, which is around $500 per session with
fMRI and $1500-2000 with PET [63]. In a lighting study it
is often necessary to have many subjects and run various
sessions with each one, which makes the costs enormous.
These two requirements can also be hard to realise for
safety reasons. With fMRI the suitability of the subject for
the test is very restricted (e.g. no pregnancy, tattoos,
pacemaker, or claustrophobia) and with PET the repeated
studies are limited by the annual permissible radiation
exposure. The number or duration of fMRI tests is not
limited but since the scanner is very sensitive to motion, the
subject can only be expected to hold still for some hours.
PRACTICAL EVALUATION OF METHODS
Subjects
Twelve healthy young volunteers (5 women and 7 men; age
range 20-28; mean age 24.4 ± 2.4 SD years) and nine
healthy older volunteers (5 women and 4 men; age range
50-62; mean age 56.6 ± 3.7 SD years) participated in the
study. Before the study the subjects’ chronotypes were
assessed using the Morningness-Eveningness questionnaire
(MEQ) [33]. Extreme chronotypes that scored below 31
(Definitely Evening type) or above 69 (Definitely Morning
type) were excluded from the study. The chronotypes were
lower in the younger than in the older group (range, mean ±
167
SD: 32-56, 47.3 ± 7.6 vs. 53-63, 57.1 ± 3.3; t test: p =
0.001). The duration of sleep before the study did not differ
significantly between the groups (young: 7:56 ± 0:59 hours
vs. older: 7:36 ± 1:05 hours; mean ± SD; t test: p = 0.240).
The subjects were instructed to avoid alcohol, coffee,
energy drinks, and teas (and other drinks containing
caffeine) for 3 hours prior to the study.
resolution of 320 x 240 and a sample rate of 30 Hz. The
camera was mounted at the back of the Goldman perimeter
at a distance of 30 cm from the subject’s eye. The camera
was equipped with a telephoto lens and an IR bandpass.
The minimum focusing distance was reduced with homemade extension tubes which, at the same time, made the
depth of the field narrower.
Protocol and Study Design
The pupil was illuminated with infrared LEDs (Everlight
HIR204/H0, !max = 850 nm, hbw = 45 nm, beam angle =
60°) positioned off-axis close to the eye. The pupil size was
to be determined from a recorded uncompressed video file
using an edge-based segmentation program developed by
the authors under Matlab (Mathworks, USA). The corneal
irradiance of the infrared LEDs was below the safety levels
of 10 mW/cm2 for chronic infrared exposure at ! = 7201400 nm as defined by ICNIRP [35].
The experiment took place at the Lighting Unit of Helsinki
University of Technology. The subjects were exposed to
conditions of dark and lightness, the light exposure being
provided by a Goldman perimeter (diameter 60 cm). One
experimental session took 2 hours (15:00-17:00). As
presented in Figure 1, the lighting conditions were as
follows:
15:00-15:25 darkness
15:25-15:30 quasimonochromatic blue light
15:30-15:45 darkness
15:45-15:55 broadband orange-red light
15:55-16:05 darkness
16:05-16:10 quasimonochromatic blue light
16:10-16:40 minutes of darkness
16:45-16:50 quasimonochromatic blue light
16:50-17:00 minutes of darkness.
Heart Rate
Heart rate was monitored continuously during the whole
experiment using a Polar Rs800sd heart rate monitor (Polar
Electro, Vantaa, Finland). Heart rate was analysed with
Kubios HRV Analysis Software [37] by dividing raw heart
rate data into 5-minute bins. The mean of each bin was
calculated for heart rate, low-frequency power (LF), highfrequency power (HF), and LF/HF ratio, which is
considered to be a good index of cardiac activity [2].
Skin Conductance
Figure 1: One 2-hour experimental session between 15:00 and
17:00. Grey = light, white = darkness, black = recording
period of the pupil.
Between the recording periods the subjects were free to
stretch the legs by moving around in the experimental
room, which was light-proofed with dark curtains. The
quasimonochromatic “standard 5 mm” blue LEDs used in
the study had a peak wavelength of !max = 468 nm and a
half-bandwidth of hbw = 26 nm and they provided corneal
illuminance of 40 lx (corresponding to a photon density of
~1.5 x 1014 photons/cm2/s). The broadband orange-red light
that was used between the blue light pulses was a mixture
of two types of Luxeon Star III LEDs: Red-Orange (!max =
617 nm, hbw = 20 nm) and Amber (!max = 590 nm, hbw =
14 nm). The corneal illuminance provided by this
broadband red-orange light was 83 lx (corresponding to a
photon density of 1014 photons/cm2/s). All the LEDs were
mounted on a Goldman perimeter providing uniform light
distribution. The luminance distribution was measured
using a Nikon Coolpix 8400 digital camera equipped with a
Nikon FC-E9 fisheye lens. The acquired images were
analysed using the PHOTOLUX 2.1 software [22] which
had a calibration profile for the camera that was used.
Measurements
Pupil Size
The pupil size of the subject was recorded during the
periods illustrated in Figure 1. It was recorded using a
Unibrain Fire-I OEM (Unibrain Inc., San Ramon,
California, USA) digital monochrome board camera with a
168
Skin conductance was measured continuously using a
ProComp Infiniti Encoder (ThoughtTechnology, Montreal,
Canada). 256 samples were recorded per second with
BioGraph Infiniti software [1]. Mean skin conductance was
determined for the same 5-minute bins as with heart rate.
Karolinska Sleepiness Scale
Subjective sleepiness was assessed using the Karolinska
Sleepiness Scale (KSS) [88] every 20 minutes during the
experiment. The mean subjective sleepiness every five
minutes was calculated by extrapolating the data.
Data Analysis and Statistics
For all the analysis, the Statistical Package for the Social
Science (SPSS) was used. The significance level was set to
0.05 in all comparisons. To analyse the values in different
lighting or recording conditions, the Student t-test for
independent variables was used. This t test is a special case
of ANOVA that assesses whether the means of two groups
are different (if p < 0.05 the means are different). The
correlations of different methods within the age group and
of the same methods between the age groups were tested
with Pearson's correlation coefficient (r = 0.00 = no
correlation and |r| = 1.00 = perfect correlation). Pearson's
correlation was also used to investigate the time correlation
of the measures.
Results
The mean values of normalised skin conductance,
normalised heart rate, LF/HF ratio, and subjective
sleepiness every 5 minutes for the young and older test
groups during the 2-hour test period are illustrated in
Figure 2. Unfortunately, the pupil size values could not be
calculated because of the low image quality caused by an
unfocused lens, camera vibration, and the absence of a
fixation point of the eye.
The subjective sleepiness increased significantly by time
with both the young (r = 0.74, p = 0.000) and older (r =
0.80, p = 0.000) test groups. The LF/HF ratio decreased
somewhat (young: r = -0.26, p = 0.113; older: r = -0.26, p =
0.112; not significant) and normalised heart rate values
decreased significantly (young: r = -0.56, p = 0.002; older:
r = -0.35, p = 0.046), corresponding to reduced arousal. In
contrast, the skin conductance values supported the
increase in arousal with time, but only with the young
subjects (young: r = 0.47, p = 0.010; older: r = 0.07, p =
0.379).
Exposure to light (either quasimonochromatic blue or
broadband orange-red) did not cause any significant effect
on the values in the young group (p > 0.05 with all
methods; t test). In the older group there were differences
in the heart rate and skin conductance values during the
light period compared to darkness (heart rate: p = 0.045;
skin conductance: p = 0.021; t test). However, the effect of
the recording period appeared much stronger in the values.
In both age groups the heart rate, LF/HF ratio, and skin
conductance were significantly higher during the periods
when the subject could move freely in the dark
experimental room compared to the periods when he or she
was attached to the Goldman perimeter (young: p = 0.000,
0.007, 0.050; older: p = 0.000, 0.022, 0.000; p = heart rate,
LH/HF ratio, skin conductance; t test). Within the recording
periods there was no difference in the responses to light
exposure compared to darkness in any of the methods in
either of the age groups (p > 0.05 with all methods in both
age groups; t test). Sleepiness acted independently and did
not follow the recording or the lighting conditions. The
behaviour of the skin conductance correlated negatively
with the behaviour of the LF/HF ratio in both age groups
(young: r = -0.50, p = 0.006; older: r = -0.27, p = 0.043).
With the young subjects KSS correlated with skin
conductance (r = 0.54, p = 0.003), giving conflicting
information about the changes in alertness. However, the
negative correlation of KSS and HR (r = -0.50, p = 0.006),
implied that their alertness did indeed decrease with time.
These inter-method correlations were not found significant
in the older test group. However, all the measures showed
corresponding behaviour in both age groups (heart rate: r =
0.73, p = 0.000; LF/HF ratio: r = 0.35, p = 0.049; skin
conductance: r = 0.70, p = 0.000; KSS: r = 0.99, p =
0.000).
Figure 2: Time course of subjective sleepiness (top), LF/HF
ratio (panel 2), normalised skin conductance (panel 3) and
normalised heart rate (bottom). Mean values per 5-min bin ±
SEM.
Discussion
Subjective sleepiness ratings and heart rate measures
showed that the subjects became sleepier during the test.
That is in conflict with the skin conductance responses,
which suggest that the subjects became more aroused. It is
possible that the KSS ratings reflected reduced motivation
rather than alertness. However, a more likely explanation
for the inconsistency is that while becoming sleepier the
subjects were trying harder to fight against the desire to
sleep. That appeared in the skin conductance data as
arousal.
The effect of light exposure was shown in the older test
group as a change in skin conductance and heart rate.
However, the variations of the autonomic nervous system
functions were mainly detected when the subject was able
to move freely in the experimental room without having his
or her head attached to the Goldman perimeter. This
indicates that the presence of the pupil camera played a
strong role in the ANS responses masking the effect of the
light stimulus. This is a practical illustration of the
sensitivity of the ANS methods to external stimuli. It shows
169
that more effort has to be put into the study protocol to
either exclude everything that could appear as external,
unwanted stimuli or choose methods that can be used in the
presence of such stimuli. Apparently, sitting still in front of
the Goldman perimeter eye facing towards the camera was
a task that did not go together with skin conductance and
heart rate measurements. Hence, in this type of study
setting the measurement of pupil size cannot be used at the
same time as other methods recording the activation of the
autonomic nervous system. Furthermore, the recording
protocol has to be designed so comfortable for the subject
that no difference between the recording period and the rest
of the experiment can be detected.
The biggest drawback of the study was the unsuccessful
recording of the pupil size, which appeared despite the fact
that the protocol had been tested in a pilot study. It was not
possible to adjust the focus of the camera, so the focusing
had to be done manually. In future studies one could try to
adjust the distance to the eye by attaching the camera to a
microrail on which the camera can move back and forth. To
reduce the noise in the image, more infrared light should be
applied. That is challenging because the light should be
kept invisible to the subject. In this study difficulties were
encountered in keeping the camera still. During some of the
experimental sessions the heavy camera could not be held
in a constant position, causing the eye to change its position
in the image. Therefore not all the data could be read and
processed with the Matlab program. This could be
corrected by using a separate stand for the camera instead
of attaching the camera directly to the Goldman perimeter.
CONCLUSIONS
The theoretical and practical examination of the methods
showed that using subjective evaluation to assess alertness
is an easy method to conduct. However, it should be kept in
mind that ratings on scales such as the Karolinska
Sleepiness Scale (KSS) do not necessarily indicate changes
in alertness. Therefore subjective evaluation should always
be used together with an objective test. Objective
evaluation can be done by using either central (CNS) or
autonomic nervous system (ANS) variables or, in the best
case, a combination of those two. Reaction tests are also
often used in lighting studies, but they measure sustained
attention rather than alertness. In addition, the tasks can
mask the light stimulus. For measuring ANS activity
through skin conductance and heart rate, there is relatively
cheap, low-tech equipment available. As the practical
testing showed, it is, however, very sensitive to external
stimuli, which can limit its use in lighting studies. The
current study encountered some difficulties in the
measurement of pupils. However, it is foreseen that with
more careful study design pupillometry could be a suitable
method for use in light-induced alertness research.
In the current study it was not possible to test CSN
variables. Previous studies show, however, that electrooculogram (EOG) measurements could suit lighting
research if the tiresome tasks did not limit their use to short
170
recordings. The electroencephalogram (EEG) is popular
because of its high temporal resolution. However, as a
result of the low spatial resolution a lot of interference can
occur in the data. The authors are of the opinion that of the
research methods presented in this paper, the greatest
potential lies in brain imaging, because it can reveal the
mechanisms behind the (hypothetical) light-induced
daytime alertness by spotting the neural correlates. The
protocol is expensive and hard to design because of
numerous restrictions. What is common to all CNS
methods is that they are not suitable for field studies.
By using validated methods and designing the experiments
in accordance with the standards, the data analysis and the
comparison of the results with other studies can be made
easier. There is already good software available for
numerous methods.
REFERENCES
1. Anon. Thought Technology Ltd. Available
http://www.thoughttechnology.com/software.htm/.
at
2. Arai. Y.C., Ushida, T., Matsubara, T., Shimo, K., Ito,
H., Sato, Y., Wakao, Y. and Komatsu, T. The Influence
of Acupressure at Extra 1 Acupuncture Point on the
Spectral Entropy of the EEG and the LF/HF Ratio of
Heart Rate Variability. Evid Based Complement
Alternat Med (2008).
3. Arden, G.B. and Kelsey, J. Changes produced by light
in the standing potential of the human eye. J Physiol
161 (1962), 189-204.
4. Aston-Jones, G. and Chen, S. A neural circuit for
circadian regulation of arousal. Nature Neurosci 4
(2001), 732-738.
5. Atienza, M., Cantero, J.L., Stickgold, R. and Hobson,
A. Eyelid movements measured by Nightcap predict
slow eye movements during quiet wakefulness in
humans. J Sleep Res 13, 1 (2004), 25-29.
6. Badia, P., Culpepper, J., Myers, B., Boecker, M. and
Harsh, J. Psychophysiological and behavioral effects of
bright and dim light. Sleep Res 19 (1990), 387.
7. Badia, P., Myers, B., Boecker, M., Culpepper, J.,
Harsch, J.R. Bright light effects on body temperature,
alertness, EEG and behavior. Physiol Behav 50, 3
(1991), 583-588.
8. Baron, M. G., Groden, J. Groden, G. and Lipsitt, L.P.
Stress and coping in autism. Oxford University Press,
NY, 2006.
9. Berridge, C.W. and Foote, S.L. Effects of Locus
Coeruleus Activation on Electro-encephalographic
Activity in Neocortex and Hippocampus. J Neurosci 10,
10 (1991), 3135-3145.
10. Berson, D.M., Dunn, F.A. and Takao, M.
Phototransduction by Retinal Ganglion Cells That Set
the Circadian Clock. Science 295, 5557 (2002), 10701073.
11. Blatter, K. and Cajochen, C. Circadian rhythms in
cognitive performance: Methodological constraints,
protocols, theoretical underpinnings. Physiol Behav 90,
2-3 (2007), 196-208.
26. Gamlin, P.D.R., McDougal, D.H., Pokorny, J., Smith,
V.C., Yau, K-W. and Dacey, D.M. Human and macaque
pupil responses driven by melanopsin-containing retinal
ganglion cells. Vis Res 47, 7 (2007), 946-954.
12. Bonnet, M.H. Performance during sleep. In Webb,
W.B. (eds.) Biological rhythms, sleep, and performance,
Chichester UK, 1982 (pp. 205-237).
27. Goel, N. and Etwaroo, G.R. Bright light, negative air
ions and auditory stimuli produce rapid mood changes
in a student population: a placebo-controlled study.
Psychol Med 36, 9 (2006), 1253-1263.
13. Byron, L. Electrophysiology of vision-clinical testing
and applications. Informa Healthcare, London UK,
2005.
14. Cable, N.T., Drust, B. and Gregson, W.A. The impact
of altered climatic conditions and altitude on circadian
physiology. Physiol Behav 90, 2-3 ( 2007), 267-273.
15. Cajochen, C. and Dijk, D-J. Electroencephalographic
activity during wakefulness, rapid eye movement and
non-rapid eye movement sleep in humans: Comparison
of their circadian and homeostatic modulation. Sleep
Biol Rhythms 1 (2003), 85-95.
16. Cajochen, C., Khalsa, S.B.S., Wyatt, J.K., Czeisler,
C.A. and Dijk, D-J. EEG and ocular correlates of
circadian melatonin phase and human performance
decrements during sleep loss. Am J Physiol Regul Integr
Comp Physiol 277 (1999), R640-R649.
17. Cajochen, C. Alerting effects of light. Sleep Med Rev
11, 6 (2007), 453-464.
18. Dawson, M.E., Schell, A.M., Filion, D.L. The
electrodermal system. In Cacioppo, Tassinary and
Berntson (eds.), Handbook of Psychophysiology,
Cambridge University Press, Cambridge UK, 2000 (pp.
200-223).
19. Dinges, D.F., Grace, R. PERCLOS: A Valid
Psychophysiological Measure of Alertness As Assessed
by Psychomotor Vigilance. Federal Highway
Administration, Washington WA, 1998.
20. Duchowski, A.T. Eye Tracking Methodology: Theory
and Practice. Springer-Verlag, New York NY, 2007.
21. Duffy, E. The psychological significance of the concept
of 'arousal' or 'activation'. Psychol Rev 64 (1957), 265275.
22. Dumortier, D, Coutelier, B. and Faulcon, F and Roy,
F.V. PHOTOLUX: a new luminance mapping system
based on Nikon Coolpix digital cameras, in Lux Europa
(Berlin GE, September 2005), 308-311.
23. Edelberg, R. Electrical properties of the skin. In Brown
(eds.) Methods in psychophysiology. Williams and
Wilkins, Baltimore MA, 1967 (pp. 1-53).
24. Foote, S.L., Aston-Jones, G., and Bloom, F.E. Impulse
activity of locus coeruleus neurons in awake rats and
monkeys is a function of sensory stimulation and
arousal. Proc Natl Acad Sci USA 77 (1980), 3033-3037.
25. Galbrecht, C.R. and Dykman, R.A. On the scoring of
heart rate: reply to criticism. Psychophysiology 2, 1
(2007), 42 – 46.
28. Grings, W.W. and Dawson, M.E. Emotions and bodily
responses: a psychophysiological approach. Academic
Press, New York NY, 1978.
29. Hachol, A., Szczepanowska-Nowak, W., Kasprzak, H.,
Zawojska, I., Dudzinski, A., Kinasz, R. and
Wygledowska-Promienska, D. Measurement of pupil
reactivity using fast pupillometry. Physiological
Measurement 28, 1 (2007), 61-72.
30. Hindmarch, I., U. Rigney, N. Stanley, P. Quinlan, J.
Rycroft, and Lane, J. A naturalistic investigation of the
effects of day-long consumption of tea, coffee and water
on alertness, sleep onset and sleep quality.
Psychopharmacology 149, 3 (2000), 203-216.
31. Hobson, J.A. and Brazier, M.A.B (eds.) The Reticular
Formation Revisited. Raven Press, New York NY,
1980.
32. Hoddes, E., Zarcone, V. and Smythe, H. Quantification
of Sleepiness: a new approach. Psychophysiol 10
(1973), 431-436.
33. Horne, J A, and O Östberg. A self-assessment
questionnaire to determine morningness-eveningness in
human circadian rhythms. Int J Chronobiol 4, 2 (1976),
97-110.
34. Hyoki, K., Shigeta, M., Tsuno, N., Kawamuro, Y. and
Kinoshita, T. Quantitative electro-oculography and
electroencephalography as indices of alertness.
Electroencephalog Clin Neurophysiol 106, 3 (1998),
213-219.
35. ICNIRP Guidelines. Guidelines on Limits of Exposure
to BroadBand Incoherent Optical Radiation (0.38 to
3!m). Health Physics 73, 3 (1997). Available at
http://www.icnirp.de/documents/broadband.pdf.
36. Jaywoo, K., Mi-hee, L., Hyoung-Ki. L., Kiwan, C.,
Seokwon, B. and Sangryong, K. Heart Rate Monitor for
Portable MP3 Player. Engineering in Medicine and
Biology Society, in 27th Int Conf of the IEEE EMBS,
(Shanghai China, September 2005), 5207 - 5210.
37. Niskanen, J.P., Tarvainen, M.P. Ranta-aho, P.O. and
Karjalainen, P.A. Software for advanced HRV analysis.
Computer Methods and Programs in Biomedicine 76, 1
(2004), 73-81.
38. Jung, T-P., Makeig, S., Stensmo, M. and Sejnowski, J.
Estimating alertness from the EEG power spectrum.
IEEE transactions on biomedical engineering 44, 1
(1997), 60-69.
171
39. Kaida, K., Takahashi, M., Åkerstedt, T., Nakata, A.,
Otsuka, Y., Haratani, T. and Fukasawa, K. Validation of
the Karolinska sleepiness scale against performance and
EEG variables. Clin Neurophysiol 117, 7 (2006), 15741581.
40. Karenbach, C. Ledalab - a software package for the
analysis of phasic electrodermal activity. Technical
report. Institut für Psychologie, 2005.
41. Kasthurirangan, S. and Glasser, A. Characteristics of
pupil responses during far-to-near and near-to-far
accommodation. Ophthal Physiol Opt 25, 4 (2005),
328-339.
42. Kosuke, K., Åkerstedt, T., Kecklund, G., Nilsson, J.P.
and Axelsson, J. The effects of asking for verbal ratings
of sleepiness on sleepiness and its masking effects on
performance. Clin Neurophysiol 118, 6 (2007), 13241331.
43. Dongheng, L. and Parkhurst, D.J. OpenEyes: an openhardware open-source system for low-cost eye tracking.
J Mod Opt 53 (2006), 1295-1311.
44. Lockley, S.W., Evans, E.E., Scheer, F.A., Brainard,
G.C., Czeisler, C.A. and Aeschbach, D. Shortwavelength sensitivity for the direct effects of light on
alertness,
vigilance,
and
the
waking
electroencephalogram in humans. Sleep 29, 2 (2006),
161-168.
45. Loewenfeld, I.E. The pupil: anatomy, physiology and
clinical applications. Iowa State University Press, Ames
IA, 1993.
46. Lykken, D.T. and Venables, P.H. Direct measurement
of skin conductance: A proposal for standardization.
Psychophysiol 8, 5 (1971), 656-672.
47. Magnan, P. Detection of visible photons in CCD and
CMOS: A comparative view. Nuclear Instruments and
Methods in Physics Research Section A 504, 1-3 (2003),
199-212.
48. Makeig, S. and Jung, T-P. Tonic, phasic, and transient
EEG correlates of auditory awareness in drowsiness.
Cogn. Brain Res 3 (1996), 15-25.
49. Malik, M., Bigger, J.T., Camm, A.J., Kleiger, R.E.
Malliani, A., Moss, A.J., and Schwartz, P.J. Heart Rate
Variability - Standards of Measurement, Physiological
Interpretation, and Clinical Use. European Heart
Journal 17 (1996), 354-381.
50. Marcos, S., Moreno, E. and Navarro, R. The depth-offield of the human eye from objective and subjective
measurements. Vis Res 39, 12 (1999), 2039-2049.
51. Marmor, M.F. and Zrenner, E. Standard for Clinical
Electro-oculography. Documenta Ophtalmologica 85
(1993), 115-124.
52. Masaoka, Y., Hirasawa, K., Yamane. J., Hori. T. and
Homma, I. Effects of Left Amygdala Lesions on
Respiration, Skin Conductance, Heart Rate, Anxiety,
172
and Activity of the Right Amygdala During
Anticipation of Negative Stimulus. Behav Modif 27
(2003), 607-619.
53. McLaren, J.W., Hauri, P.J. Lin, S-C. and Harris, C.D.
Pupillometry in clinically sleepy patients. Sleep Med 3,
4 (2002), 347-352.
54. McNames, J., Thong, T. and Goldstein, B. Reliability
and accuracy of heart rate variability metrics versus
ECG segment duration. Engineering in Medicine and
Biology Society, in 25th Int Conf of the IEEE EMBS,
(Cancun Mexico, September 2003), 212-215.
55. Menon, V. and Crottaz-Herbette, S. Combined EEG and
fMRI studies of human brain function. Int Rev
Neurobiol 66 (2005), 291-321.
56. Michimori. A., Araki, K. and Hagiwara, H. Effects of
illuminance level on alertness levels sympathetic
activities. Sleep Res 26 (1997), 677.
57. Moller, H.J., Devins, G.M., Shen, J. and Shapiro, C.M.
Sleepiness is not the inverse of alertness: evidence from
four sleep disorder patient groups. Exp Brain Res 173,
2 (2006), 258-66.
58. Moruzzi, G. and Magoun, H.W. Brain stem reticular
formation
and
activation
of
the
EEG.
Electroencephalogr Clin Neuro 1 (1949), 455-73.
59. Nikolaou, A., Schiza, S.E., Giakoumaki, S.G., Roussos,
P., Siafakas, N. and Bitsios, P. The 5-min pupillary
alertness test is sensitive to modafinil: a placebo
controlled study in patients with sleep apnea.
Psychopharmacology 2196, 2 (2008), 167-175.
60. Pauley, S.M. Lighting for the human circadian clock:
recent research indicates that lighting has become a
public health issue. Medical Hypotheses 63, 4 (2004),
588-596.
61. Peña, M., J. Echeverría, M. García, and R. GonzálezCamarena. Applying fractal analysis to short sets of
heart rate variability data. Medical and Biological
Engineering and Computing 47, 7 (2009).
62. Rea, M.S., Figueiro, M.G. and Bullough, J.D. Circadian
photobiology: an emerging framework for lighting
practice and research. Light Res Tech 34, 3 (2002), 177187.
63. Reiman, E., Lane, R.D., Van Petten, C., Bandettini, P.A.
Positron emission tomography and functional magnetic
resonance imaging. In Cacioppo, Tassinary and
Berntson (eds.), Handbook of Psychophysiology,
Cambridge University Press, Cambridge UK, 2000 (pp
85-118).
64. Roach, G.D., Dawson, D. and Lamond, N. Can shorter
psychomotor vigilance task be used as a reasonable
substitute for the ten-minute psychomotor vigilance
test? Chronobiol Int 23, 6 (2006), 1379-1387.
65. Saper, C.B., Scammell, T.E. and Lu, J. Hypothalamic
regulation of sleep and circadian rhythms. Nature 437
(2005), 1257-1263.
66. Stanciu, M., Faur, D., Boev, A. and Stanciu, A. ECG
measurements and acquisition system for heart rate
variability
analysis,
in
6th
Int
Conf
on
Electromechanical and Power Systems (Chisinau,
Moldova, October 2007)
67. Stark, L. Stability, Oscillations, and Noise in the Human
Pupil Servomechanism, in Proceedings of the IRE 47,
11 (1959), 1925-1939.
68. Steriade, M. and McCarley, R.W. Brain control of
wakefulness and sleep. Kluwer Academic/Plenum, New
York NY, 2005.
69. Stern, R.M., Ray, W.J. and Davis, C.M.
Psychophysiological recording. Oxford University
Press, New York NY, 1980.
70. Stevens, Richard G. Light-at-night, circadian disruption
and breast cancer: assessment of existing evidence. Int J
Epidemiol (2009), dyp178.
71. Sturm, W., Willmes, K. On the Functional
Neuroanatomy of Intrinsic and Phasic Alertness.
Neuroimage 14, 1 (2001), 76-84.
72. Szabó, Z., Tokaji, Z., Kalman, J. Oroszi, L., Pestenacz,
A. and Zoltan Janka. The effect of bright light exposure
on pupillary fluctuations in healthy subjects. J Affect
Disord 78, 2 (2004), 153-156.
73. Szczepanowska-Nowak, W., Hachol, A. and Kasprzak,
H. System for measurement of the consensual pupil
light reflex. Optica Applicata XXXIV 4 (2004).
74. Talukder, A. Morookian, J-M. Monacos, S., Lam, R.,
LeBaw, C. and Lambert, J.L. Eye-tracking architecture
for biometrics and remote monitoring. Applied Optics
44, 5 (2005), 693-700.
75. Thorne, D.R., Johnson, D.E., Redmond, D.P., Sing,
H.C., Belenky, G. and Shapiro, J.M. The Walter Reed
palm-held psychomotor. Behav Res Meth 37, 1 (2005),
111-118.
76. Torrance, G.W., Feeny, D, and Furlong, W. Visual
Analog Scales: Do They Have a Role in the
Measurement of Preferences for Health States? Medical
Decision Making 21, 4 (2001), 329-334.
77. Troncoso, E., Rodriguez, M., and Feria, M. Lightinduced arousal affects simultaneously EEG and heart
rate variability in the rat. Neurosci Lett 188, 3 (1995),
167-170.
78. Ukai, K., Tsuchiya, K. and Ishikawa, S. Induced
pupillary hippus following near vision: increased
occurrence in visual display unit workers. Ergonomics
40, 11 (1997), 1201-1211.
79. van Bommel, W.J.M. Non-visual biological effect of
lighting and the practical meaning for lighting for work.
Applied Ergonomics 37, 4 (2006), 461-466.
80. Vandewalle, G., Schmidt, C., Albouy, G., Sterpenich,
V., Darsaud, A., Rauchs, G., Berken, P.Y., Balteau, E.,
Degueldre, C., Luxen, A., Maquet, P. and Dijk, D.J.
Brain responses to violet, blue, and green
monochromatic light exposures in humans: prominent
role of blue light and the brainstem. PLoS ONE 2, 11
(2007), e1247.
81. Veitch, J.A. Investigating and influencing how
buildings affect health: Interdisciplinary endeavours.
Canadian Psychology 49, 4 (2008), 281-288.
82. Westheimer, G. The Maxwellian view. Vis Res 6, 11-12
(1966), 669-682.
83. Wilkinson, R.T., Houghtong, D. (1982). Field test of
arousal: a portable reaction timer with data storage.
Human factors 24: 487-493.
84. Viola, A.U, James, L.M. Schlangen, L.J.M. and Dijk,
D-J. Blue-enriched white light in the workplace
improves self-reported alertness, performance and sleep
quality. Scandinavian Journal of Work, Environment &
Health 34, 4 (2008), 297-306.
85. Wright, K.P., Hull, J.T. and Czeisler, C.A. Relationship
between alertness, performance, and body temperature
in humans. Am J Physiol Regul Integr Comp Physiol
283 (2002), R1370-R1377.
86. Wyatt, J.K., Ritz-De Cecco, A., Czeisler, C.A. and Dijk,
D.J. Circadian temperature and melatonin rhythms,
sleep, and neurobehavioral function in humans living
20-h day. Am J Physiol Regul Integr Comp Physiol46, 4
(1999), 1152-1163.
87. Young, R.S.L., and Kimura, E. Pupillary correlates of
light-evoked melanopsin activity in humans. Vis Res 48,
7 (2008), 862-871.
88. Åkerstedt, T. and Gillberg, M. Subjective and objective
sleepiness in the active individual. Int J Neurosci 52, 12 (1990), 29-37.
173