Building and Environment 46 (2011) 981e989

Contents lists available at ScienceDirect

Building and Environment

journal homepage: www.elsevier.com/locate/buildenv

Evening office lighting e visual comfort vs. energy efficiency vs. performance?
Friedrich Linhart*, Jean-Louis Scartezzini
Solar Energy and Building Physics Laboratory (LESO-PB), Swiss Federal Institute of Technology in Lausanne (EPFL), 1015 Lausanne, Switzerland

a r t i c l e i n f o a b s t r a c t

Article history: During the study presented in this article, we compared two highly energy-efficient lighting scenarios for
Received 23 April 2010 evening office lighting (i.e. electric lighting that is typically used for approxiamately 2 h in the evening).
Received in revised form The first of these lighting scenarios (referred to as “Reference”-scenario, Lighting Power Density or LPD of
1 October 2010
4.5 W/m2) has been successfully in use in many office rooms of the Solar Energy and Building Physics
Accepted 7 October 2010
Laboratory’s experimental building, located on the campus of the Swiss Federal Institute of Technology in
Available online 25 October 2010
Lausanne, for several years. The second lighting scenario (referred to as “Test”-scenario, Lighting Power
Density of 3.9 W/m2) is more energy-efficient, creates higher workplane illuminances but leads to an
Keywords:
Office lighting
increased risk of discomfort glare. The aim of this study was to meticulously compare the two lighting
Electric lighting scenarios in order to find a lighting solution for evening office lighting that offers an optimal trade-off
Lighting power density between energy-efficiency, visual comfort and visual performance.
Energy-efficiency For this purpose, objective visual performance tests (computer-based and paper-based) and
Perfomance subjective visual comfort assessments with 20 human subjects were carried out. The main hypothesis
Comfort of our study was that the study participants would not perform worse under the more energy-efficient
“Test” -scenario than under the “Reference”-scenario (which is extremely well accepted by the bui-
dling’s occupants).
We found that the two tested scenarios are comparable to usual lighting scenarios in other office
rooms in terms of subjective visual comfort. The study participants preferred the “Test”-scenario to the
“Reference”-scenario. Their performance in a paper-based task was significantly better under the “Test”-
scenario than under the “Reference”-scenario. No significant differences in the performance during two
computer-based tasks were found. We conclude that energy-efficient lighting with Lighting Power
Densities of less than 5 W/m2 is already achievable in today’s office rooms without jeopardizing visual
comfort and performance. Less powerfull electric lighting systems do not necessarily mean a decrease in
visual comfort and/or performance; our results even show that better visual comfort and better visual
performance can be achieved with less connected lighting power.
Ó 2010 Elsevier Ltd. All rights reserved.

1. Introduction
Large fractions of today’s buildings’ electricity consumptions
are due to electric lighting: Hinnells [1] for example states that
“lighting in domestic and commercial sectors accounts for around
16% of all UK electricity demand”. Li and Lam [2] as well as Li et al.
[3] suggest that artificial lighting can account for 20e30% of
a typical non-domestic building’s electricity consumption. Wen
et al. estimate that in the US, “roughly 40% of electricity
consumption in commercial buildings is attributable to lighting”
[4]. According to Franzetti et al., “it is generally assumed that about
30% of the energy consumption of office buildings come from
* Corresponding author. Tel.: þ41 21 6934540; fax: þ41 21 6932722.
E-mail address: friedrich.linhart@epfl.ch (F. Linhart).
artificial lighting” [5]. In 2007, Jenkins and Newborough wrote that
“the energy consumption of lighting in buildings is a major
contributor to carbon emissions, often estimated as 20e40% of
the total building energy consumption” [6]. Krarti et al. had already
put forward that 25e40% of the “energy consumption” in US
commercial buildings is due to artificial lighting [7]. Even if those
estimations are difficult to verify and and might not be completely
consistent (the difference between 20% of a building’s electricity
consumption and 40% of its “energy consumption” is likely to be
huge in most cases), these figures lead to one important finding:
around the world, the scientific community seems to agree that
discussing the articial lighting loads of buildings is extremely
important and that energy-efficient lighting solutions have to be
adopted.
However, electricity consumption and energy-efficiency are not
the only topics to consider when it comes to designing appropriate

0360-1323/$ e see front matter Ó 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.buildenv.2010.10.002
982 F. Linhart, J.-L. Scartezzini / Building and Environment 46 (2011) 981e989

lighting scenarios for buildings: good visual comfort is of course are presented in Section 3 and discussed in Section 4. A final
equally important. In order to guarantee an appropriate visual conclusion is given in Section 5 where scope for future work is
comfort in office rooms, the horizontal illuminances (especially on also discussed.
the workplanes) must be sufficiently high, the light on the work-
plane has to be properly distributed (appropriate illuminance
2. Detailed experiment description
uniformities) and discomfort glare (e.g. from luminaires or through
windows) must be avoided. It is obvious that it is not an easy task
The performance and visual comfort tests took part in evenings
to reach the two objectives “energy-efficiency” and “good visual
(external darkness) from April to May 2008. For each subject, the
comfort” at the same time. A significant reduction of the lighting
experiment was scheduled in such a way that he or she arrived at
load will in most cases result in lower illuminances and might
the laboratory around sunset. The subjects were young and healthy
therefore have a negative influence on the occupants’ visual
males and females (10 males, 10 females) with an academic back-
comfort. However, it is not an impossible task: Akashi and Boyce [8]
ground. Their average age was 23.7  3.5 years.
for example have shown ways to reduce the illuminances in an office
environment without significantly jeopardizing visual comfort.
In one of our previous articles, we have discussed possibilities 2.1. Experimental setup
of minimizing the lighting power density (LPD) in office rooms
equipped with Anidolic Daylighting Systems (ADS). We concluded The workplace used during our study is shown in Fig. 1. It’s
that LPDs of 5 W/m2 and less are possible in the particular case of dimensions were approximately 1.7 m  1.9 m and it was located
ADS-equipped office rooms because the artificial lighting installa- near the window of an office room with 4.7 m of depth and 3.4 m
tion in such office rooms is mainly complementary. It is typically of width [10,11].
only used under extremely overcast sky conditions or for a limited In the “Reference”-scenario, it was illuminated with two ceiling-
amount of time in the evening hours. During large parts of normal mounted “Lip” luminaires by Regent [12]. In the “Test”-scenario, it
working days, the ADS are able to supply workplane illuminances was illuminated with two “Zen3” luminaires by Tulux [13], which
that are highly sufficient [10]. As a matter of fact, a detailed were positioned above the workplane by means of a ceiling-
assessment of the occupant satisfaction in ADS-equipped office mounted rail system. The “Lip” luminaires were each equipped
rooms has shown that insufficient illuminances are not causing with one T8 36 W/830 fluorescent tube, leading to an LPD of 4.5 W/
major annoyance to these offices’ occupants [9]. m2 (electronic control gear included). The horizontal workplane
In our opinion, the electric lighting systems in office rooms that illuminance under this lighting condition was 232 Lux (measured
typically benefit from large daylight flux (and where artificial with a calibrated luxmeter) and the corresponding uniformity g1
lighting is therefore mainly complementary) should be designed (i.e. the lowest illuminance value measured on the workplane
differently than those in office rooms that are more dependent on divided by the average illuminance) was 0.79. The “Zen3” lumi-
artificial lighting. In order to “push” the office occupants to make naires were each equipped with one T5 28 W/830 fluorescent tube,
good use of the available daylight flux, the electric lighting systems leading to an LPD of 3.9 W/m2 (electronic control gear included).
should be dimensioned as thriftily as possible. Economically The horizontal workplane illuminance under this lighting condition
dimensioned electric lighting systems keep a building’s artificial equalled 352 Lux and the corresponding uniformity g1 was 0.9 [10].
lighting load low in a simple but effective way: Sloppily spoken, One important difference between the two different luminaire
power that is not connected (i.e. not installed) cannot be switched types is their optical efficiency: the “Zen3”-luminaire achieves
on and can therefore not consume any electricity. a value of 96% whereas the “Lip” luminaire only achieves 69%.
During the study presented in this article, we compared two low-
LPD lighting scenarios for evening office lighting (i.e. electric lighting
that is typically used for approxiamately 2 h in the evening). The first
of these lighting scenarios (referred to as “Reference”-scenario) has
been successfully in use in many office rooms of the Solar Energy and
Building Physics Laboratory’s experimental building (LESO solar
experimental building), located on the campus of the Swiss Federal
Institute of Technology in Lausanne, for several years. The second
lighting scenario (referred to as “Test”-scenario) is more energy-
efficient, creates higher workplane illuminances but leads to an
increased risk of discomfort glare. The aim of this study was to
meticulously compare the two lighting scenarios in order to find
a lighting solution for evening office lighting that offers an optimal
trade-off between energy-efficiency, visual comfort and visual
performance. For this purpose, objective visual performance tests
(computer-based and paper-based) and subjective visual comfort
assessments with 20 human subjects were carried out. The main
hypothesis of our study was that the 20 human subjects would not
perform worse under the more energy-efficient “Test” -scenario
than under the “Reference”-scenario (which is extremely well
accepted by the buidling’s occupants [9]). Furthermore, we expected
to find a positive impact of the “Test”-scenario on the study
participants’ visual performance during the paper-based task,
compared to the “Reference”-scenario (based on our prior experi-
ence in the LESO solar experimental building).
A detailed description of our study’s experimental setup and
the applied methods is given in Section 2 of this article. The results Fig. 1. Workplace used during our study in the test office room.
 F. Linhart, J.-L. Scartezzini / Building and Environment 46 (2011) 981e989 983

Table 1 Table 2
Overview of the two tested low-LPD lighting scenarios. Overview of the Karolinska Sleepiness Scale (KSS).

Value “Reference”-scenario “Test”-scenario Value English rating French rating

Luminaire type 2  “Lip” (Regent) 2  “Zen3” (Tulux) 1 Extremely alert Extrêmement alerte
Luminaire optical efficiency 0.69 0.96 2 Very alert Très alerte
Lamp type One T8 fluorescent tube One T5 fluorescent 3 Alert Alerte
per luminaire (T8 36 W/830) tube per luminaire 4 Rather alert Assez alerte
(T5 28 W/830) 5 Neither alert nor sleepy Ni alerte, ni somnolent
LPD 4.5 W/m2 3.9 W/m2 6 Some signs of sleepiness Signes de somnolence
NPD 1.9 W/(m2*100 Lux) 1.1 W/(m2*100 Lux) 7 Sleepy, no effort to stay awake Somnolent, mais reste
Horizontal illuminance 232 Lux 352 Lux éveillé sans effort
Horizontal uniformity g1 0.79 0.9 8 Sleepy, some effort to stay Somnolent, efforts pour
UGR 13.84 14.75 awake rester éveillé
9 Very sleepy, great effort to Très somnolent, gros efforts
keep awake, fighting sleep pour rester éveiller, lutte contre
The “Reference”-scenario leads to a normalized power density le sommeil
(NPD) of 1.9 W/(m2*100 Lux) whereas the “Test”-scenario leads to
an NPD of 1.1 W/(m2*100 Lux). These NPD values visualize that
the “Test”-scenario is more energy-efficient than the “Reference”- 2.2. Methods
scenario. Since using the LPD instead of the NPD is much more
common in Switzerland when it comes to talking about lighting 2.2.1. Alertness monitoring
installations, mainly the LPD values of the two scenarios are being During the entire duration of each experiment, the test persons’
used in the following discussions. alertness was monitored using the Karolinska Sleepiness Scale
At first sight, it may seem surprising that two lighting scenarios (KSS). It is a subjective rating during which the persons have to
creating workplane illuminance levels of much less than 500 Lux state their actual alertness level on a 9-stage scale going from
were compared in this study, since most lighting standards suggest “extremely alert” (¼ 1) to “very sleepy, great effort to keep awake,
horizontal workplane illuminances of 500 Lux. However, as fighting sleep” (¼ 9). The KSS was validated against electroen-
previously mentioned, prior research on visual comfort in office cephalography (EEG) data by Akerstedt and Gillberg in 1990 [16].
buildings equipped with ADS has shown that under certain Table 2 gives an overview of the corresponding scale, the original
circumstances much lower illuminance levels seem to be sufficient ratings, as well as the translated French ratings used during our
[10]. study.
Table 1 summarizes the characteristics of the two different
lighting scenarios. 2.2.2. Performance at computer-based task
Fig. 2 shows photographs of the two different lighting scenarios, One main objective of the study was to test the influence of
taken with an HDR camera equipped with a fish eye objective. the two different lighting scenarios “Reference” and “Test” on the
In terms of glare risk, the two scenarios are comparable: the subjects’ performance during a computer-based visual perfor-
“Lip” luminaires lead to a Unified Glare Rating (UGR) [14] of 13.84 mance task. For this purpose, the “Freiburg Visual Acuity & Contrast
versus a UGR of 14.75 in the case of the “Zen3” luminaires. The UGR Test” (FrACT) [17] has been used. It allows to determine a person’s
values have been computed using the software EVALGLARE [15]. visual acuity and contrast threshold through correct recognition
The two lighting scenarios are thus comparable in terms of glare of Landolt rings on a PC screen. The computerized test can be
risk, the latter being however slightly higher in the case where downloaded free of charge on www.michaelbach.de/fract. We have
the “Zen3” luminaires are used. used two different methods to assess the occupants’ visual

Fig. 2. Photographs of the two different lighting scenarios, taken with an HDR camera and a fish eye objective. These photos were used to calculate the UGR-values with Evalglare
[15].
984 F. Linhart, J.-L. Scartezzini / Building and Environment 46 (2011) 981e989
performance for each of the two lighting conditions. The first
method consisted of showing the subject a sequence of 36 Landolt
rings of different size and orientation (see Fig. 3). The participant
had to determine the orientation of the ring and give the appro-
priate answer via the PC’s keyboard as quickly as possible. Each
participant repeated this “acuity” -task three times under each
lighting condition. After each sequence (36 Landolt rings), the
participant took a short break which was used to store the results
of the test (e.g. correct answers, response time per ring) in an
Excel spreadsheet. For each sequence of 36 rings, a performance
indicator hperf was determined as follows:
ncorrect
hperf ¼ (1)
ssequence
where ncorrect stands for the total number of correctly identified
Landolt rings per sequence and ssequence stands for the total dura-
tion of the sequence.
Using the FrACT to determine a person’s contrast threshold
is another possible way to assess visual performance. Instead of
looking at Landolt rings of different size, the contrast between the
ring and the screen background is modified during the test. Yet
again, each subject performed three sequences of this “contrast”-
task of 36 Landolt rings for each of the two lighting scenarios.
The entire FrACT test is illustrated in Fig. 3.
2.2.3. Performance at paper-based task
Another main objective of this study was to test the influence of
the two different lighting scenarios on the subjects’ performance
during a paper-based task. For this purpose, we have used a test
Fig. 3. Overview of the computer-based FrACT test used during our study. (a) visual-
izes the use of the four direction keys during the FrACT, (b) the “acuity”-task and (c) the
“contrast”-task.
suggested by Courret [18] in 1999. Yet again, Landolt rings are used
during this test: the study participants receive a piece of white
paper on which 96 Landolt rings are printed in clear grey. They
are asked to determine, as quickly as possible and without
writing on the paper, the correct orientations of the 96 rings by
writing down the number of counted rings for all four possible
orientations (open on top, open on bottom, open left, open right).
The contrast between the paper and the grey color of the rings
is very weak which makes this quite a difficult task. The test is
shown in Fig. 4.
We have used two different versions of this visual performance
test (i.e. with different ring orientations) in order to avoid any
bias from people who remember the number of rings previously
counted during the first lighting scenario while they are doing the
test under the second lighting scenario.
2.2.4. Subjective visual comfort assessment
Additionally, the study participants had to rate their visual
comfort during the two different lighting scenarios. A simple and
reliable questionnaire-based assessment method for occupant
satisfaction regarding office lighting (Office Lighting Survey - OLS)
was presented by Eklund and Boyce in 1996 [19]. Many questions
within the OLS however only allow an answer on a symmetrical,
two-stage “Yes/No” scale. Akashi and Boyce, as well as Ramasoot
and Fotios, have more recently used slightly modified versions of
the OLS [8,20]. The original OLS has been adapted to the specific
situation of our study: a questionnaire with a mix of general and
artificial lighting-specific statements was set up for that purpose.
Occupants were asked to rate their agreement with each statement
on a symmetric answering scale (i.e. without neutral choice) in
order to avoid possible interpretation problems associated with
neutral choices. The different statements and questions used in the
questionnaire are shown in Fig. 9 (“Results”- Section of this article)
and are therefore not outlined in detail at this point.
2.3. Study Procedure
After arriving at the laboratory at the scheduled appointment
time (i.e. sunset of the particular day), each participant was offered
a quick tour around the building: this typically took between 5
and 10 min. Then, a quick introduction concerning the study was
given, questions were answered and study participants were offered
Compter les anneaux en fonction de leur orientation
et inscrire ci-dessous le nombre trouvé pour chaque catégorie
(Effectuez cet exercice le plus rapidement possible
sans faire de marque sur les anneaux)
: : : :
Fig. 4. Paper-based Landolt-ring task as suggested by Courret [18]. The study partici-
pants are asked to count the number of rings for every orientation as quickly as
possible and without marking the rings.
 F. Linhart, J.-L. Scartezzini / Building and Environment 46 (2011) 981e989 985

to use the bathroom. After that, the visual performance tests began.
Each subject was tested once under the “Reference”-scenario and
once under the “Test”-scenario. Half of the participants started
under one condition, the second half under the other condition
(cross-over design). Fig. 5 shows the detailed study schedule.
In the beginning of each session, the participants filled out a KSS-
test. Then, they performed a computer-based acuity-test and
a computer-based contrast-test (determining the correct direction of
36 Landolt rings in both cases). After that, the paper-based task (see
Fig. 4) was carried out. The latter was followed by another sequence of
acuity- and contrast-tests, a KSS-test and a short break during which
participants were shown a humorous video clip (for relaxing
reasons). Participants then took a third sequence of acuity- and
contrast-tests and filled out the questionnaire for subjective visual
comfort assessment. Each session ended with a third KSS-test. One
such session took approximately 40 min. Between the two sessions,
the lighting scenario inside the office room had to be modified.
Meanwhile, the participants took a break of approximately 10 min
during which they were offered light snacks, soft drinks and water.
They were also allowed to use the bathrooms. During the break, the
study participants were exposed to significantly less light than
during the two test sessions. The lighting conditions during the break
were more or less the same for all participants since all of them spent
the break in the same hallway with non-dimmable luminaires. Then,
the second session started. The only difference between the two
sessions was the fact that the questionnaires for subjective visual
comfort assessment filled out by the participants at the end of each
session were slightly different: while the first one only contained 16
the participants’ average alertness over the duration of the entire
experiment (i.e. 90 min), based on six KSS-tests. The observed
differences could not be shown to be statistically significant
(2-sample dependent t-Test between the first and the last rating
yields p-value of 0.77). The average alertness evolution during the
“Reference”-session was tested is plotted in Fig. 6b, the alertness
evolution during the “Test”-session session in Fig. 6c. The observed
differences were not statistically significant either (2-sample
dependent t-Tests between the first and the last rating yield
p-values of 0.24 for the “Reference”-scenario 0.79 for the “Test”-
scenario.)
3.2. Computer-based task
Fig. 7 gives an overview of the results obtained during the
computer-based tasks. The average performance efficiency hperf
of all “acuity”-tests is shown in Fig. 7 (left) for the two different
lighting conditions. For both of them, the average performance
efficiency is close to 0.9 correct decisions per second. There was
no significant difference between the two lighting scenarios (t-Test
yields p ¼ 0.161  0.05). Fig. 7 (right) shows a comparison of the
average contrast thresholds obtained during the “contrast”-tests
for the two lighting scenarios. We measured values of 0.37% for
9
8 Experiment
7
KSS rating

lighting-related questions, the second one had an additional section 6

with general questions (e.g. age). 5
4
3
3. Results 2
1
3.1. Alertness 0
0 20 40 60 80 100
Fig. 6 shows the participants’ alertness monitored by means of a Time elapsed [%]
the KSS throughout the experiment. Fig. 6a shows the evolution of
9
1st lighting 2nd lighting 8 Reference
scenario scenario 7
KSS rating

KSS KSS
6
5
Acuity Test 1A Acuity Test 2A
4
Contrast Test 1A Contrast Test 2A
3
2
Paper Test Paper Test
1
0
Acuity Test 1B Acuity Test 2B
Break 0 20 40 60 80 100
Contrast Test 1B Contrast Test 2B
(light snacks,
softdrinks, water,
b Time elapsed [%]
KSS possibility to use KSS
bathrooms)
9
Test
Short break Short break 8
7
KSS rating

Acuity Test 1C Acuity Test 2C 6

Contrast Test 1C Contrast Test 2C 5
4
Questionnaire 1 Questionnaire 2 3
2
KSS KSS 1
0
0 20 40 60 80 100

00:00 00:40 00:50 01:30 c Time elapsed [%]

Time elapsed [hh:mm] Fig. 6. Evolution of the subjects’ average KSS ratings over the duration of the entire
experiment (panel a)), over the duration of only the “Reference”-session (panel b)) and
Fig. 5. Detailed schedule of our study. over only the duration of the “Test”-session (panel c)).
986
1.0
Performance [cor. decisions/s]
0.90
0.8
0.6
0.4
0.2
0.0
Reference
Fig. 7. Avereage results of the “acuity”-test (left) and the “contrast”-test (right)
obtained during the computer-based tasks. SEMs ¼ 0.024, 0.024, 0.012, 0.012.
the “Reference”-scenario and 0.36% for the “Test”-scenario. Yet
again, the small difference was not statistically significant (t-Test
yields p ¼ 0.158  0.05).
3.3. Paper-based task
Fig. 8 shows the results of the paper-based task for the two
different lighting conditions.
The average number of mistakes per ring orientation is shown in
Fig. 8 (left). One can observe that the subjects made in average more
mistakes under the “Reference”-scenario than under the “Test”-
scenario. The differences between the average number of mistakes
for each orientation that are displayed in Fig. 8 (left) could, however,
not be shown to be statistically significant (t-Tests yield p  0.05). In
Fig. 8 (right), the difference between the “Reference”- and the “Test”-
scenario becomes more evident. This Figure shows the average (over
20 participants) number of mistakes for the sum of all Landolt-ring
orientations. Under the “Reference”-scenario, the participants got in
average 24.95 out of 96 Landolt-ring orientations wrong, while they
made only an average of 18.68 mistakes under the “Test”-scenario,
the difference being statistically significant (p ¼ 0.032  0.05). The
Figure also shows that the average study participant made an
average of 6.24 mistakes per ring orientation under the “Reference”-
scenario and only an average of 4.67 mistakes per ring orientation
under the “Test”-scenario. This difference is also statistically
significant (p ¼ 0.009  0.05).
3.4. Visual comfort assessment
Fig. 9 now shows the results of the subjective visual comfort
assessment. Nineteen out of 20 study participants have properly
filled out the corresponding questionnaires. Only the differences in
statements S4 (“This office seems too dim.”) and statement S10
Average number of mistakes
9 8.53
8
7
6 4.95
5 4.37
4
3 10
2 5
1
0 0
F. Linhart, J.-L. Scartezzini / Building and Environment 46 (2011) 981e989
0.4 (“The ceiling-mounted luminaires are too bright.”) were statisti-
0.37 0.36
0.87
cally significant (t-Tests performed on test data yielded p-values of
Contrast threshold [%]
0.036 and 0.011, respectively).
0.3
Fig. 10 shows the differences between the “Reference”- and the
“Test”-scenario for the statements S1 to S14. A negative value means
0.2
that the “Test”-scenario performs worse than the “Reference”-
scenario in this specific regard, a positive value corresponds to an
0.1
enhancement.
Fig. 11 shows an additional result of the subjective visual
0.0 comfort assessment. In addition to the 14 statements listed above,
Test Reference Test the study participants were asked to state the maximum amount
of time per day during which they could imagine themselves
working under the given lighting condition.
At the end of the experiment, each participant was asked
whether he or she preferred the “Reference”-scenario or the “Test”-
scenario. Fig. 12 shows the participants’ answers to this question.
4. Discussion
These results show that the two tested low-LPD lighting
scenarios are comparable in terms of subjective visual comfort.
Overall, the study participants preferred the “Test”-scenario over the
“Reference”-scenario. Their performance in our paper-based task
was significantly better under the “Test”-scenario than under the
“Reference”-scenario. No significant differences in the computer-
based tasks under the two different lighting scenarios were found.
There was no significant decrease in the participants’ alertness,
neither over the entire duration of the experiment, nor in one of the
two different sessions (i.e. “Reference”- or “Test”-session).
Fig. 6 shows that the average alertness was indeed always
situated between “rather alert” and “alert”. Nevertheless, one could
get the impression that a slight decrease in alertness occurred (i.e.
an increase of the average KSS rating) during the “Reference”-
sessions and a very slight increase of alertness (i.e. a decrease of
the average KSS rating) during the “Test”-sessions. These small
differences are, however, not statistically significant. The small KSS
rating variations over the entire experiment are also not statisti-
cally significant. This means that there is a considerable probability
that the variations observed in Fig. 6 have occurred by chance.
It therefore makes sense to assume that the participants’ alertness
did not change throughout the study and that it is not possible
to distinguish the two lighting scenarios regarding the participants’
alertness.
Since the (very small) differences between the “Reference”- and
the “Test”-scenario displayed in Fig. 7 were not statistically signif-
icant, it can be concluded that there was no measurable influence of
the lighting scenario on the study participants’ performance during
the computer-based tasks. In other words: the participants did not
perform better or worse in the computer task under one or the
other lighting scenario. This could of course indicate that the FrACT
[ *
Average number of mistakes
30 Reference Test
Reference Test 24.95
7.11 25
5.84 20 18.68
5.26
15
[
3.68 3.89
*
6.24
4.67
Bottom Top Left Right Sum of all orientations Average over all
orientations
Ring opening orientation
Fig. 8. Results of the paper-based task for the two different lighting conditions. * ¼ p  0.05.
 F. Linhart, J.-L. Scartezzini / Building and Environment 46 (2011) 981e989 987

70
63
61
60 58 Reference Test 57
54
49
50

Agreement [%]
41
*
40 36 36
35 33 33
32 32 32 32
30 29 29
30 28 27
25 24 24

20
* 15
21

13
11
10

Statement

Fig. 9. Results of the subjective visual comfort assessment. * ¼ p  0.05.

test used during our study was inappropriate for our means; it absence of an effect. In summary, the analysis of the paper-based
might, for example, not be sensitive enough. However, we believe task shows that the subjects performed on average better under
that our results rather illustrate the fact that the differences the new “Test”-scenario than under the old “Reference”-scenario.
between the two lighting scenarios were not strong enough to The better performance is most likely due to higher workplane
have a measurable influence on the participants’ performance illuminance and a brighter room achieved under the “Test”-
during a computer-based task. This means that it is possible to scenario. This leads us to the conclusion that when working under
reduce the LPD in such office rooms from 4.5 W/m2 to 3.9 W/m2 artificial lighting conditions during a limited amount of time in the
without having a negative influence on the performance in evening hours, the positive effects of elevated workplane illumi-
computer-based tasks. nances are stronger than the negative effects of a slightly higher
The fact that the differences between the “Reference”- and the UGR from luminaires. Reducing the LPD in such office rooms
“Test”-scenario in terms of performance in the paper-based task from 4.5 W/m2 to 3.9 W/m2 will therefore not only minimize the
displayed in Fig. 8 (left) could not be shown to be statistically electricity consumption but can also be expected to have a positive
significant but that the results become significant in Fig. 8 (right) influence on the office occupants’ performance on paper-based
(where the calculations are based on the sum of all Landolt rings) tasks.
might be a hint that the apparent insignificance in Fig. 8 (left) is During the subjective visual comfort assessment, the statement
rather due to the small amount of samples than to the actual “In general, the lighting in this office is comfortable.” scored
agreement values of 63% for the “Reference”-scenario and 61%
-100% -80% -60% -40% -20% 0% 20% 40% 60% 80% for the “Test”-scenario. For both conditions, these values are
S1
slightly lower than the typical values described by Akashi and
7%
Boyce (69%) [8]. This is not surprising at all because our experi-
S2 -3%
ments took place during external darkness whereas the values
S3 -13%
S4 57%
7
S5 6 6 6
Number of participants

44%
6
S6 -19% 5
5
S7 -15% 4 4 4
4
S8 34% 3
3
S9 18%
2
S 10 -81%
1
S 11 13%
0
S 12 -43%
Less than 2 2 to 4 hours 4 to 6 hours more than 6
S 13 55% hours hours
S 14 -13% Maximum working time under this lighting condition

Fig. 10. Differences between the “Reference”- and the “Test”-scenario for the state- Reference Test
ments S1 to S14. A negative value means that the “Test”-scenario performs worse than
the “Reference”-scenario in this specific regard, a positive value corresponds to an Fig. 11. Maximum working time under each lighting scenario as indicated by the
enhancement. subjects.
988
10
Number of participants
8 “Reference”-scenario. It is moreover possible to say that occupant
8 satisfaction under the “Test”-scenario is not significantly worse
6 5
4 creating a negative impact on the office occupants’ visual comfort.
2
0
Preference for Preference for
"Reference" "Test" lighting lighting designs lighting design
lighting design design
Fig. 12. Study participants’ subjective preferences for one or the other low-LPD
lighting scenario.
described by Akashi and Boyce are most likely to describe agree-
ments tested during normal office hours (with at least some
daylight and its associated positive effects). In addition to that,
Akashi and Boyce have only used a 2-stage scale (yes/no), whereas
we used a 4-stage scale (yes/rather yes/rather no/no). If we trans-
form our 4-stage scale into a 2-stage scale (i.e. “rather yes” becomes
“yes” and “rather no” becomes “no”), we find agreement values of
65% and 70% for the “Reference”- and the “Test”-scenario, respec-
tively. We therefore assume that the two tested lighting scenarios
are comparable to lighting conditions in other office rooms. It
might seem surprising that the “Test”-scenario scores better
agreement values in statement S1 (“I like the lighting in this
office.”) than the “Reference”-scenario and that this trend is
inversed in statement S2 (“In general, the lighting in this office is
comfortable.”). However, the small differences in the average
agreements to these two questions are not statistically relevant. As
a matter of fact, only the differences in statements S4 (“This office
seems too dim.”) and statement S10 (“The ceiling-mounted lumi-
naires are too bright.”) are statistically significant (see “Results” -
section of this article). Nevertheless, it seems somewhat surprising
that a significant difference between the two lighting designs for
statement S4 (“This office seems too dim.”) and none for statement
S5 (“There is not enough light to carry out the different tasks.”)
was found. It is possible that the absence of significance is in many
cases the result of the small number of study participants rather
than the result of an absence of a difference. It therefore makes
sense to take a closer look at various statements where a large
difference between the two lighting designs has occurred, even if
some are not statistically significant.
If we take into account only those statements where a difference
of þ/ 20% is observed, we find an enhancement for the statements
S4, S5, S8 and S13 and deterioration for the statements S10 and
S12 (cf. Fig. 10). The facts that the new “Test”-scenario creates higher
illuminances on the workplane (S5) and makes the room in general
appear brighter and less gloomy (S4, S8 and S13) are perceived
positively, whereas the aggressiveness (S12) and especially the
increased luminance of the luminaires (S10) are perceived negatively.
The results displayed in Fig. 11 suggest a maximum working
time that is higher for the “Test”-scenario than for the “Reference”-
scenario.
In Fig. 12, one can observe that five (out of 19) participants had
a preference for the “Reference”-scenario, whereas eight partici-
pants preferred the “Test”-scenario; four participants liked both
solutions equally well. Only two participants felt that neither of the
two environments was comfortable enough.
The results of the subjective visual comfort assessment outlined
in Figs. 9e12 can be interpreted as follows: despite the higher
luminaire luminance and the more “aggressive” lighting style, the
F. Linhart, J.-L. Scartezzini / Building and Environment 46 (2011) 981e989
participants slightly preferred the new “Test”-scenario over the old
than under the “Reference”-scenario. It is thus possible to reduce
the LPD in such office rooms from 4.5 W/m2 to 3.9 W/m2 without
4
2
5. Conclusion and outlook
Like both Like neither The results of our tests on visual performance and subjective
visual comfort under two different low-LPD lighting scenarios lead to
several interesting conclusions. First of all, the results show that the
two tested low-LPD lighting scenarios are comparable to usual
lighting scenarios in other office rooms in terms of subjective visual
comfort. Overall, the study participants preferred the “Test”-scenario
(LPD of 3.9 W/m2) over the “Reference”-scenario (LPD of 4.5 W/m2).
In addition to that, their performance in our paper-based task was
significantly better under the “Test”-scenario than under the
“Reference”-scenario. This leads us to the conclusion that, when
working under artificial lighting conditions during a limited amount
of time in the evening hours, the positive effects of elevated work-
plane illuminances are stronger than the negative effects of
discomfort glare from luminaires. Another interesting finding of this
study is the fact that no significant differences in the computer-based
tasks under the two different lighting scenarios where found. This
suggests that the lighting environment might have a much smaller
influence on the performance during computer work than on the
performance during paper work. In general, we conclude that
energy-efficient lighting with LPDs of less than 5 W/m2 is already
achievable in today’s office rooms without jeopardizing visual
comfort and performance. Less powerfull electric lighting systems do
not necessarily mean a decrease in visual comfort and performance;
our results even show that better visual comfort and better visual
performance can be achieved with less connected lighting power.
Further work could look at the long-term effects of different
low-LPD lighting scenarios on performance and visual comfort of
office workers. Another interesting question is whether the light
exposure during the normal working day influences performance
and visual comfort in the evening. Recent research in the field of
chronobiology indicates that, amongst other factors, light history
should be taken into consideration during office lighting design
[21]. In this regard, it is also very important to get a sound insight
into typical ocular light exposures in office buildings. We have
recently started activities pointing in this direction at our labora-
tory [22,23].
References
[1] Hinnells M. Technologies to achieve demand reduction and microgeneration
in buildings. Energy Policy 2008;36:588e99.
[2] Li D, Lam J. An analysis of lighting energy savings and switching frequency for
a daylit corridor under various indoor design illuminance levels. Applied
Energy 2003;76(4):363e78.
[3] Li D, Lam T, Wong S. Lighting and energy performance for an office using
high frequency dimming controls. Energy Conversion and Management
2006;47(9e10):1133e45.
[4] Wen Y, Granderson J, Agogino A. Towards embedded wireless-networked
intelligent daylighting systems for commercial buildings. In: IEEE Interna-
tional Conference on Sensor Networks, Ubiquitous, and Trustworthy
Computing, 2006, vol. 1; 2006.
[5] Franzetti C, Fraisse G, Achard G. Influence of the coupling between daylight
and artificial lighting on thermal loads in office buildings. Energy and Build-
ings 2004;36(2):117e26.
[6] Jenkins D, Newborough M. An approach for estimating the carbon emissions
associated with office lighting with a daylight contribution. Applied Energy
2007;84(6):608e22.
[7] Krarti M, Erickson P, Hillman T. A simplified method to estimate energy
savings of artificial lighting use from daylighting. Building and Environment
2005;40(6):747e54.
 F. Linhart, J.-L. Scartezzini / Building and Environment 46 (2011) 981e989
[8] Akashi Y, Boyce P. A field study of illuminance reduction. Energy and Buildings
2006;38(6):588e99.
[9] Linhart F, Scartezzini J. Occupant satisfaction in office rooms equipped with
anidolic daylighting systems, In: Int’l Conf. Sol. Heating Cooling Build.
(EUROSUN); 2008, p. 355e362.
[10] Linhart F, Energetic, visual and non-visual aspects of office lighting, Ph.D.
thesis. Lausanne; 2010. doi:10.5075/epfl-thesis-4634, http://library.epfl.ch/
theses/?nr¼4634
[11] Linhart F, Scartezzini J. Minimizing lighting power density in office rooms
equipped with anidolic daylighting systems. Solar Energy 2010;84(4):
587e95.
[12] Regent. Website of Regent Beleuchtungskö, http://www.regent.ch; 2009.
[13] Tulux. Website of Tulux AG, http://www.tulux.ch; 2009.
[14] Cie. Tech. Rep. CIE. Cie collection on glare: commission international de
l’clairage, 147. CIE; 2002.
[15] Wienold J, Christoffersen J. Evaluation methods and development of a new
glare prediction model for daylight environments with the use of CCD
cameras. Energy and Buildings 2006;38(7):743e57.
989
[16] Åkerstedt T, Gillberg M. Subjective and objective sleepiness in the active
individual. International Journal of Neuroscience 1990;52(1e2):29e37.
[17] Bach M. The Freiburg Visual Acuity test-automatic measurement of visual
acuity. Optometry and Vision Science 1996;73(1):49.
[18] Courret G. Systmes anidoliques d’clairage naturel, Ph.D. thesis, Swiss Federal
Institute of Technology in Lausanne; 1999.
[19] Eklund N, Boyce P. The development of a reliable, valid, and simple office
lighting survey. Journal of the Illuminating Engineering Society.
[20] Ramasoot T, Fotios S. Lighting for the classrooms of the future, electronic
classrooms: a new challenge for school lighting guidance. Balkan Light.
[21] Webb A. Considerations for lighting in the built environment: non-visual
effects of light. Energy and Buildings 2006;38(7):721e7.
[22] Linhart F, Wittkopf SK, Münch M, Scartezzini J. Recent research on anidolic
daylighting systems: highly reflective coating materials and chronobiological
properties (Proceedings Paper), In: Proceedings of SPIE, vol. 7423; 2009, p. 74230K.
[23] Linhart F, Wittkopf S, Scartezzini J. Performance of Anidolic Daylighting
Systems in tropical climates-parametric studies for identification of main
influencing factors. Solar Energy 2010;84(7):1085e94.