Journal of Modern Optics

Vol. 56, No. 13, 20 July 2009, 14331446

Kruithof s rule revisited using LED illumination

Francoise Vienota*, Marie-Lucie Duranda and Elodie Mahlerb
a

Museum National dHistoire Naturelle, Centre de Recherche sur la Conservation des Collections,
36 rue Geoffroy Saint-Hilaire, 75005 Paris, France; bSociete Nationale des Chemins de fer
Francais, Direction de lInnovation et de la Recherche, Unite Confort et Services,
45 rue de Londres, 75379 Paris Cedex 8, France

Downloaded By: [Capderou, Christine] At: 11:31 7 September 2009

(Received 21 January 2009; final version received 29 June 2009)

We have investigated the effect of colour temperature and illuminance on the visual response. Since Kruithofs
(1941) work on general illumination, it has been proposed that users would find pleasing high correlated colour
temperature (CCT) illumination at high illuminance and low CCT illumination at low illuminance, although it is
not unanimously accepted. Here, we question whether the pleasing sensation comes from facilitation of visual
performance or from subjective appraisal. We have conducted experiments at various illuminance levels and
various CCTs including performance tasks and subjective scaling of illumination quality. The change of
illumination was obtained using adjustable LED clusters with high colour rendering index (Ra490). Whereas the
performance tasks yield results which depend on luminance and only slightly on CCT, observers reported that high
CCT illuminations look brighter than low CCT illuminations. We investigate the effect of the various spectral
power distributions on the intrinsically photoreceptive retinal ganglion cells and on the pupil response.
Keywords: lighting; Kruithof; light emitting diodes; brightness; visual performance; melanopsin

1. Background
In 1941, Kruithof [1] conducted a study on the colour
of lamps for general illumination, when fluorescent
lighting had just emerged. A fluorescent lamp is a
low pressure gas-discharge emitter, the glass of which is
lined with a phosphor that absorbs ultraviolet radiations and emits light in the visible range. By modifying
the composition of the fluorescent material, one can
adjust the spectrum of the emitted light. By mixing
different luminescent substances in any desired proportion, Kruithof was able to control the colour of the
emitted light. Indeed, he used three phosphors respectively emitting in the blue, green and red domain of
the visible spectrum. Precisely, he could obtain white
lights with chromaticity close to the Planckian locus,
at correlated colour temperature above 2500 K.
Kruithof introduced a variable number of lamps in a
room, so that the observer could not guess which type
of illumination was on. All lights could be graded in
intensity while controlling the correlated colour temperature (CCT). Beneath 2850 K, he would use incandescent lamps with variable current to produce
light sources. For higher CCT, the experiment was
conducted with daylight originating from a window
or daylight luminescence lamps. There is little

*Corresponding author. Email: vienot@mnhn.fr

ISSN 09500340 print/ISSN 13623044 online
2009 Taylor & Francis
DOI: 10.1080/09500340903151278
http://www.informaworld.com

information available about the size and the geometry

of the room, on the tasks proposed to the observers, as
well as on the number of participants.
The aims of Kruithofs experiment were several
fold. As far as we are concerned, we concentrate on his
results regarding the pleasing appearance of the
illumination, results that are often quoted in lighting
engineering. Results are displayed in a two dimension
diagram where correlated colour temperature (CCT) is
reported (in MK
1) along the abscissa and illuminance
E is reported along the ordinate axis (in log 10 lx).
The scales on the axes were chosen in order to provide
significant visual steps. Two curves are drawn on
the diagram. They show the limits between which
an illumination is considered pleasing (Figure 1).
The lowest limit at any CCT and the highest limit at
low CCT were determined by experiment. The upper
curve was extrapolated above 5000 K at the highest
illuminance (104 or 105 lx), where Kruithof assumed
that daylight is never found to be unnatural. There
are no experimental points on the graph. In addition,
Kruithof reported that beneath the lower illuminance
limit, illumination is judged as dim at low CCT
and cold at high CCT. Above the upper illuminance
limit, colour reproduction is unnatural and unpleasant.

Downloaded By: [Capderou, Christine] At: 11:31 7 September 2009

1434

F. Vienot et al.

Figure 1. Reproduction of Kruithofs diagram published in 1941 [1]. Correlated colour temperature (CCT) is scaled in inverse of
megaKelvins along the abscissa. Illuminance is reported on a log scale along the ordinate axis. Two curves show the limits
between which an illumination is considered pleasing after Kruithof. The squares show the nine illumination conditions of our
experiment. Reprinted with permission from Kruithof, Philips Technical Review, 1941, 6, 6596. 1941, Koninklijke Philips
Electronics N.V.

By that time, definition of the colour rendering

index (CRI) had not been officially recommended by
the CIE [2].
Kruithofs experiment was pioneering, from a
technological point of view and as a guide to the
design of artificial lighting. In spite of the fact that
Kruithofs results have gathered a lot of recognition,
several attempts to test his statement have been
reported with inconsistent conclusions [3,4]. From the
subjective appraisals of 50 subjects, Bodmann et al. [5]
concluded that the dependence of the observers feelings
on the colour of light and the illumination level (eight
lamps covered 28007400 K and 3005000 lx) corresponds with Kruithofs findings so far as the tendency
is concerned. Later, several investigations on the
dependence of visual performance on the luminance
distribution in the visual field were analysed by
Bodmann [6]. It appeared that the performance could
be improved significantly by increasing the task
luminance. The best rating of brightness appeared
over a broad range. In general, any appreciable
spectral distortion was appraised as unnatural.
Using three types of illumination: high pressure
sodium, metal halide and cool white fluorescent, Smith
and Rea [7] invited 20 observers to perform a numerical verification task and to rate how they felt and
evaluated the task conditions. Both were unaffected by
light source differences. Performance was better and
ratings on several scales were higher as luminance
increased (from 0.96 to 1810 cd m
2). Feeling ratings

Alert, Rested, Satisfied, Good, Comfortable,

Relaxed and rating task evaluations Good,
Relaxing, Pleasant, Satisfying gradually stabilised
(levelled off) at the higher level. Other task evaluation
ratings Brightness, Clarity, Stimulation rose sharply along the full luminance range. Varying illuminance (350, 600 lx) and fluorescent lamp colour
characteristics (CCT in the 35006500 K range and
CRI in the 5695 range), Boyce [8] verified that the
higher luminance produced ratings of greater colourfulness and greater clarity, and that the better the
colour properties of the lamp, the higher the rating of
colourfulness. Fotios [4] summarising his previous
studies (1995, 1997) where high discharge lamps,
fluorescent lamps, and incandescent lamps were used,
emphasised the positive aspects of high colour temperature and high colour rendering index in terms of
subjective brightness and observers preference.
We admit that studies using different lamp types
have had importance for the practice of lighting [9],
nevertheless studies that deal with similar lamp types
with high CRI offer a better methodology to investigate the effect of illuminance and CCT. Thus, a gain in
apparent brightness caused by higher colour temperature of filtered incandescent lamps was found by most
of the 73 observers tested by Harrington [10]. The gain
was confirmed by Berman in the 1990s who noted the
close correlation between brightness perception and
scotopic response for the first time, and later with the
pupillary response [11]. Comparing several fluorescent

Downloaded By: [Capderou, Christine] At: 11:31 7 September 2009

Journal of Modern Optics

lamp types, with different colour rendering, as defined
by the nominal percentage output in six spectral bands,
Bellchambers and co-authors [12,13] showed that the
balance in spectral emission is necessary to offer visual
clarity and provide high satisfaction. Comparing
eleven different illuminants (10 fluorescent lamps plus
one incandescent lamp) for their brightness impression
on five observers, Kanaya et al. [14] showed that the
effect of correlated colour temperature was negligible,
and that the major effect could be explained by the
colour rendering properties.
A thorough study was conducted by Cuttle and
Boyce [15,16] who collected colour discrimination
performance results and subjective assessments of the
lighting under 16 combinations of illuminance (30, 90,
225 and 600 lx) and CCT (2700, 3500, 4200 and
6300 K). These authors carefully designed the experiment to eliminate or control any potential influential
factors or peripheral cues on the experimental environment. To ensure that differences of the subject
response would be due to variation of CCT and not to
other differences of spectral output, they chose four
fluorescent lamp types of high and consistent Colour
Rendering Index (Ra [82, 85]), a condition that
probably could not be achieved by Kruithof in the
early 1940s. The goal of their experiment was the
subjective assessments of lighting through free choice
descriptive words and ratings on a semantic scale.
Sixteen of the 19 semantic scales showed significant
effects of the illuminance. Increasing illuminance was
assessed altogether favourably. Low illuminances were
found not to be pleasing. This was not dependent upon
CCT. They concluded that once the subject is adapted
to the viewing conditions, the CCT of good CRI lamps
has little effect on peoples impressions of the lighting
of the room. For them, Kruithofs boundaries are not
determinant. Davis and Ginthner [17] seriously
opposed Kruithof. In an experiment with a colourbalanced environment and sources with colour rendering index equal to 8990, they demonstrated that the
subjective ratings of preference were influenced only
by light level from 270 to 1345 lx but not by colour
temperature between 2750 and 5000 K. The subjective
ratings that were related to brightness were not
influenced by colour temperature. Only the warm
cool scale was affected by CCT. An additional index,
in terms of scotopic/photopic (S/P) ratio, has been
considered by Boyce et al. [18] who compared the effect
of two fluorescent lamps with different correlated
colour temperature, moderate colour rendering index
and contrasting S/P ratio (3000 K, Ra 78, S/P 1.3
and 6500 K, Ra 72, S/P 2.1) on the performance of a
black and white Landolt ring task. They concluded
that the speed and accuracy of performance of the task
was determined by the Landolt C gap size, and at a

1435

much lesser extent, by the illuminance and not by the

lamp spectrum. The results of all 79 participants
somewhat indicated that the 6500 K lighting was
perceived as brighter than the 3500 K lighting at the
same light level.
Today, novel approaches can be used to study the
effect of light on human well-being. Progress in
physiology has shown us that visual effects are twofold [19]. Most of the information originating from the
retina is processed through the geniculostriate pathway
and devoted to image analysis. In parallel, a part of the
information that originates from the retina is directed
to the pretectum and controls other visual functions
such as the circadian rhythm and the pupil response.
Nowadays, fluorescent lighting has matured, new
lighting technologies have been developed, and individual experience has considerably changed in terms
of quality of available sources and management of
illumination. Progress in solid state lighting technology
has made possible the generation of valuable white
light sources in terms of efficiency and colour
rendering. In particular, light emitting diode (LED)
lighting which is adjustable in flux and correlated
colour temperature allows us to revisit Kruithofs rule.
We question the interpretation of pleasing. There
could be several reasons for which an observer can
judge the illumination pleasing. Task performance
could be facilitated, and/or colour appearance could
contribute to beautification, and/or global subjective
impression could improve, and/or cognitive appreciation could interact with direct sensation. For the
present study, we set several LED illuminations at
controlled CCT and illuminance levels. We asked
observers to perform tasks and to enunciate subjective
appraisals of the illumination.

2. Materials
We built a light booth which could house several LED
clusters and provide controlled white illumination at
various correlated colour temperatures and various
illuminance levels. Then we selected several visual tests
in order to investigate performance, colour appearance, cognitive responses and subjective feelings of
observers.

2.1. Illumination system

Several LED clusters were mounted on heat radiators
and housed inside the ceiling of a light booth
(41 cm  35 cm  38 cm). We could modulate blue,
cyan, green, amber, orange, red, one-phosphor
cool light and/or two-phosphor warm white LEDs
(Figure 2). Each family of LEDs was controlled by its

Downloaded By: [Capderou, Christine] At: 11:31 7 September 2009

1436

F. Vienot et al.

Figure 2. Relative spectral power distribution of every LED type used in the experiments. Ordinate in arbitrary units. (The
colour version of this figure is included in the online version of the journal.)

own driver through a pulse width modulation (PWM)

power supply. Heat radiators were dimensioned so as
to avoid heating that could modify emission during the
experiment. The full system was interfaced with a
computer and each family of LEDs was controlled
using proprietary software (LedToLite).
In order to investigate Kruithofs rule, we planned
nine illumination scenarios which could be encountered in interior lighting. In order to cover Kruithofs
diagram, all combinations of illuminance E 150, 300,
600 lx and correlated colour temperature Tcp 2700,
4000, 6500 K were prepared. Figure 3 shows the
spectral power distributions of the light for the nine
illumination conditions. Figure 1 shows where the
illuminations fall into Kruithofs diagram.
Each illumination was obtained using as many
LED varieties as necessary to obtain the target
illumination with the highest available Colour
Rendering Index [20]. For this, we wrote a MatLab
program to predict E, Tcp and CRI values, using all the

possible combinations of LED families. Some manual

correction was necessary to achieve precise chromaticity and CRI adjustment. Finally, all spectral power
distributions were measured using a Minolta CS-1000
spectroradiometer aimed at a white spectralon
tile positioned on the booth floor (Figure 3).
Measurements are given in terms of luminance
expressed in cd m
1. They correspond to about 150,
300 and 600 lx illuminance (Table 1).

2.2. Visual tests

Visual tests were chosen to evaluate the pleasing-ness
of illumination, according to aspects such as performance on a visual task, colour appearance, subjective
appraisal. Visual performance was tested in terms
of acuity, contrast threshold, reading speed. In order
to prevent the observers from using memory cues
whenever they were screened repeatedly under several
illuminations, it was necessary to use a variety of

1437

Downloaded By: [Capderou, Christine] At: 11:31 7 September 2009

Journal of Modern Optics

Figure 3. Spectral power distribution of the light for the nine illumination conditions. Measurements were taken with a Minolta
CS-1000 spectroradiometer aimed at a white spectralon tile placed on the floor of the light booth. Precise correlated colour
temperature and luminance values are given in Table 1. (The colour version of this figure is included in the online version of the
journal.)

Table 1. Photometric parameters of the nine illuminations used in the experiment. Measurements are given in terms of
luminance expressed in cd m
2. They correspond to about 150, 300 and 600 lx.

2700 K
2700 K
2700 K
4000 K
4000 K
4000 K
6500 K
6500 K
6500 K

150 lx
300 lx
600 lx
150 lx
300 lx
600 lx
150 lx
300 lx
600 lx

Correlated
colour
temperature (K)

Luminance
(cd m
2)

Colour
rendering index
Ra

Average total
colour rendering
index Ra14

2711.2
2703.5
2699.3
4019.5
4014.0
3997.4
6489.7
6503.7
6491.7

50.4
99.6
198.0
50.7
98.9
196.5
48.7
105.2
192.2

91.7
92.0
93.2
94.3
93.6
96.1
90.7
92.0
91.0

88.2
90.9
91.5
92.5
92.2
93.3
86.6
89.1
90.6

1438

F. Vienot et al.

charts or texts based on a unique scheme, which

yielded equivalent performance scores.

Downloaded By: [Capderou, Christine] At: 11:31 7 September 2009

Accuracy number of correctly cancelled rings

and

2.2.1. Acuity test

We measured morphoscopic acuity using a chart with
logarithmic scaling of the size of the letters. The model
has been designed by Galinier [21]. It consists of 20
lines of five letters, each line being bounded by a
diamond symbol at both ends in order to offer similar
neighbourhood to every letter. Letter size increases
from one line to the next by 1.22%. The chart allows
one to measure near acuity at 40 cm distance. Only the
original Galinier frame is commercially available.
We had to manufacture a series of copies by permuting
letters within each line. The Galinier chart is available
in black and white and at low contrast. Thus, we
printed black and white copies, as well as grey and
white (Weber contrast 0.1243) copies on non-fluorescent white paper using an office printer (HP Business
Inkjet 2280). We checked the details of the print using
an enlarging lens.
Acuity charts were presented at 45 from the
horizontal, facing the observer. We noted the smallest
letters that the observer could read. Precise instructions, in French, were given to the observer.
2.2.2. Visual performance test
We used the Weston Landolt Ring performance chart
described by Hopkinson and Collins [22]. The chart
consists of 256 Landolt rings printed in a 16  16
pattern, with positions of the gap evenly distributed.
The size of the gap provides 3/10 acuity at 40 cm
distance, so as not to impair detection. The task was to
cancel with a pen every ring with the gap at a specified
oblique orientation as illustrated by a model.
The subject should perform the test as rapidly as he
could, with as few errors as possible. The chart was
presented to the observer at one of four orientations of
the total pattern.
After Hopkinson and Collins [22], Performance
depends on both accuracy and speed. . . . Westons
method of scoring took into account the time taken in
the purely non visual part of the task. So we set-up a
separate test which involved the cancellation of a
similar number of rings regardless of the orientation of
their gaps, and we subtracted this mechanical time
from the total time taken by the observer during the
visual task. The unitary mechanical time necessary to
tick one ring was computed from the time used to
cancel all rings of a 4  4 ring matrix, divided by 16.
Finally, the performance score was computed as [23]
Visual performance Accuracy  Speed

with

Speed 1=Total time
Mechanical time:

2.2.3. Reading performance

To assess reading performance, we chose standardised
texts carefully prepared by a European consortium [24]
to offer equivalent difficulties in terms of length and
syntactic complexity. Ten texts are available in French.
After Legge et al. [25] maximum reading rates are
achieved as soon as acuity is higher than 3/10. We
followed the layout recommendations given by the
authors and printed the 10 texts in high contrast with
a laser jet printer in Times New Roman 12 which
corresponds to a 3/10 visual acuity. After the authors,
for 1835-year-old subjects, the maximal oral reading
speed difference between texts is 6.5%.
2.2.4. Colour appearance tests
Three additionnal tests for colour appearance were
also proposed: colour preference, binary hue selection,
landscape reminiscence, the results of which are not
given in this paper. We just mention the colour
appearance tests to give an exhaustive representation
of the sequence of experiments.
2.2.5. Questionnaires
Following the visual tests, a questionnaire was
proposed in order to collect subjective judgements
related to illumination. The questionnaire started with
an open question (in French): As far as you are
concerned, how would you qualitatively assess this
illumination? Please, use three words or expressions.
Responses to the open question are not used here,
although the responses from a pilot study were
analysed to select subjective scales. Further, several
subjective scales1 were proposed to characterise the
illumination.
Non glaring Glaring
Dark Bright
Cold Warm
Artificial colour rendering Natural colour rendering
Crepuscular Clear
Dull Cheerful
Tiring Relaxing
Uncomfortable Comfortable
Unpleasant Pleasant

The observer answered using a 7-point scale from 1 to

7, rating 4 being the neutral answer.

1439

Journal of Modern Optics

2.3. Observers and experimental layout

Downloaded By: [Capderou, Christine] At: 11:31 7 September 2009

Thirty volunteer French native observers took part in

the experiment, aged from 18 to 36 years. Their mother
tongue was French. Three extra observers served for
planning the sequence of tests. Normal colour vision
was checked using Lanthonys desaturated D-15
screening test [26].
The observer was seated facing the light booth
which covered a viewing angle of about 80 . He wore
white gloves to manipulate the tests. The room was
dark. All tests were performed under a given illumination. The sequence started with 3 minutes adaptation
to the illumination. Then the observer was screened
sequentially for:
Low contrast acuity
High contrast acuity
Reading performance
Colour preference
Landolt ring visual performance
Binary hue selection
Landscape reminiscence
Questionnaire.
A sequence of tests would last about 15 min.
The questionnaire was presented at the end of the
sequence after the observer had practiced all activities
under the given illumination. Illumination was changed from one sequence to the next. The new sequence
started with three minutes adaptation to the new
illumination.
Only six illuminations could be evaluated during a
2-hour session. A break was allowed after three
illumination scenarios. The layout of the total experiment was balanced along a Latin square scheme,
so that each observer would appraise six illuminations
among which two CCTs at every illuminance and two
illuminance levels at every CCT. In consequence,

every CCT and illuminance

appraised once by 20 observers.

combination

was

3. Results
For every illumination we could average the results of
20 observers.
The aim was to assess whether the visual response
varies with illuminance and/or with correlated colour
temperature. Nevertheless, whichever family of tests
performance, or subjective judgements there might be
systematic variations between observers. Therefore, we
performed a multivariate ANOVA on the results with
three factors: illuminance, CCT and observers.
We used the anovan command from MatLab which
can accommodate missing elements. A 30  3  3
matrix was prepared with individual results, in which
NaN elements were entered when the ith observer had
not been screened under the jth CCT and kth
illuminance lighting combination.
Detailed information about means and standard
deviations of the results and about the statistical
analysis are given in Tables 2 to 5.
3.1. Examples of results
As an example, we present in Figure 4 the results for
low contrast acuity and high contrast acuity. Low
contrast acuity significantly depends upon the illuminance ( p50.0001) but not upon the CCT ( p40.05)
(Table 3). High contrast acuity significantly depends
upon the illuminance ( p50.001) and upon the
CCT ( p50.05). Further, we could investigate which
CCT had a significant effect on high contrast acuity.
Indeed, we checked that high contrast acuity improves
between 2700 and 6500 K ( p50.05), but not between
2700 and 4000 K ( p40.05) or between 4000 and
6500 K ( p40.05) (Table 6).

Table 2. Results of the performance tests. Mean and standard deviation of 20 observers.

Low contrast
acuity
High contrast
acuity
Visual performance
with Landolt rings
Reading speed

Test
condition

2700 K
150 lx

4000 K
150 lx

6500 K
150 lx

2700 K
300 lx

4000 K
300 lx

6500 K
300 lx

2700 K
600 lx

4000 K
600 lx

6500 K
600 lx

Mean
Standard
deviation
Mean
Standard
deviation
Mean
Standard
deviation
Mean
Standard
deviation

0.1660
0.0724

0.1805
0.0764

0.1475
0.0767

0.1547
0.0717

0.1305
0.0727

0.1325
0.0647

0.1105
0.0620

0.0930
0.0440

0.1095
0.0707

0.0465
0.0455
0.0620
0.0542
0.0700
0.0650
0.0662
0.0745
0.0825

0.0483
0.0525
0.0562
0.0587
0.0513
0.0482
0.0473
0.0466
0.0497
0.3308
0.0882
42.70
5.77

0.3189
0.0926
41.20
5.51

0.3380
0.0699
41.70
3.69

0.3184
0.0754
42.11
3.65

0.3098
0.0664
42.65
5.41

0.3163
0.0695
42.05
6.82

0.3302
0.0881
41.00
6.76

0.3124
0.0897
41.85
4.22

0.3104
0.0968
41.10
5.57

1440

F. Vienot et al.
Table 3. Results of the analysis of variance for performance tests. F-values and significance. Degree of
freedom for correlated colour temperature, illuminance and participant is, respectively, 2, 2 and 29.

Low contrast acuity

High contrast acuity
Visual performance with Landolt rings
Reading speed

F
p
F
p
F
p
F
p

CCT

Illuminance

Participant

1.76
0.1761
3.28
0.0404
0.81
0.4466
0.38
0.6857

33.4
0.0000
8.82
0.0002
1.05
0.3522
1.84
0.1629

11.78
0.0000
11.71
0.0000
7.78
0.0000
31.61
0.0000

Downloaded By: [Capderou, Christine] At: 11:31 7 September 2009

Table 4. Responses to the questionnaire. Mean and standard deviation of 20 observers.

Bright
Clear
Glaring
Pleasant
Comfortable
Relaxing
Warm
Cheerful
Colour
rendering

Test
condition

2700 K
150 lx

4000 K
150 lx

6500 K
150 lx

2700 K
300 lx

4000 K
300 lx

6500 K
300 lx

2700 K
600 lx

4000 K
600 lx

6500 K
600 lx

Mean
Standard
deviation
Mean
Standard
deviation
Mean
Standard
deviation
Mean
Standard
deviation
Mean
Standard
deviation
Mean
Standard
deviation
Mean
Standard
deviation
Mean
Standard
deviation
Mean
Standard
deviation

3.85
1.27

3.50
1.10

5.25
1.52

4.35
1.04

5.20
1.28

5.60
1.10

5.05
1.05

5.95
0.83

6.16
0.69

3.55
1.32

3.35
1.31

5.10
1.55

4.00
1.34

4.50
1.40

5.70
1.42

4.25
1.37

6.05
1.00

6.21
0.63

5.70
1.30

5.05
1.43

4.00
1.81

5.00
1.52

4.60
1.39

4.00
1.38

5.15
1.66

4.35
1.69

3.16
1.74

4.55
1.36

3.60
1.60

3.30
1.42

4.75
1.52

4.85
1.42

4.35
1.73

4.80
1.28

4.45
1.39

3.79
1.51

4.95
1.23

3.45
1.73

3.45
1.43

4.45
1.39

4.85
1.46

4.35
1.63

5.15
1.09

4.15
1.50

3.68
1.57

4.75
1.41

3.60
1.60

3.20
1.44

4.40
1.31

4.60
1.50

3.85
1.39

4.75
1.48

3.60
1.57

3.11
1.29

5.15
1.42

3.85
1.53

2.75
1.29

5.55
1.15

4.25
1.48

3.25
1.45

5.05
1.43

3.75
1.62

3.42
1.26

3.70
1.42

2.85
1.18

3.55
1.54

4.20
1.32

4.00
1.17

4.05
1.47

4.05
1.10

4.65
1.14

4.84
1.01

4.40
1.14

3.70
1.13

4.25
1.65

4.10
1.41

4.40
1.31

4.25
1.48

3.95
110

500
1.56

4.26
1.28

Another example is presented in Figure 5 which

shows the mean score for darknessbrightness judgments. The statistical analysis shows that observers feel
that brightness increases with illuminance as well as
with CCT (Table 5). A detailed analysis shows that it
increases from 2700 to 4000 K and from 4000 to
6500 K (Table 7).
Table 8 gives an overall survey of the performance
results. Table 9 gives an overall survey of the subjective
scale results. For most of the screening tests, results
vary between observers. This gives reason for entering
the observer as a factor in the multivariate analysis of
variance.

With respect to the effect of illuminance and/or

CCT, tests can be classified as follows.
3.2. No effect of CCT
3.2.1. No effect of illuminance nor CCT
. Landolt ring test
. Reading
3.2.2. Significant effect of illuminance but no effect of
CCT
. Low contrast acuity
. Cheerfulness versus sadness

1441

Journal of Modern Optics

Table 5. Results of the analysis of variance for the responses
to the questionnaire. F-values and significance. Degree of
freedom for correlated colour temperature, illuminance and
participant is, respectively, 2, 2 and 29.

Bright
Clear
Glaring
Pleasant
Comfortable
Relaxing

Downloaded By: [Capderou, Christine] At: 11:31 7 September 2009

Warm
Cheerful
Colour rendering

F
p
F
p
F
p
F
p
F
p
F
p
F
p
F
p
F
p

CCT

Illuminance

Participant

20.41
0.0000
28.5300
0.0000
18.2200
0.0000
5.6900
0.0042
7.6700
0.0007
10.7000
0.0000
37.7600
0.0000
1.3700
0.2506
0.380
0.686

29.39
0.0000
21.2000
0.0000
3.5800
0.0302
5.2800
0.0061
2.6500
0.0743
1.8500
0.1610
1.6200
0.2019
15.9100
0.0000
0.790
0.455

1.65
0.0294
1.4400
0.0839
2.1200
0.0020
1.4100
0.0974
1.3300
0.1397
0.8900
0.6324
1.5900
0.0401
2.5100
0.0002
0.960
0.525

Figure 4. Low contrast and high contrast acuity.

Illumination parameters are reported on the abscissa axis.
Average and standard deviation of 20 observers.

Clearly, there is little effect of correlated colour

temperature on visual performance, except a slight
effect on high contrast acuity. We will come back to
this slight effect in the discussion.

3.3. Significant effect of CCT

3.3.1. Significant effect of CCT but not of illuminance
. Comfortable versus uncomfortable
. Relaxing versus exciting
. Warm versus cool

Table 6. Detailed analysis of variance for the high acuity

test. F-values and significance. Degree of freedom for
temperature, illuminance and participant is, respectively,
1, 2 and 29.
CCT

F
p
High contrast
F
p
F
p

Between 2700 Illuminance Participant

and 4000 K
2.02
6.96
8.73
0.1586
0.0018
0.0000
Between 4000 Luminance Participant
and 6500 K
1.42
4.93
8.45
0.2371
0.0094
0.0000
Between 2700 Luminance Participant
and 6500 K
6.14
4.24
7.54
0.0151
0.0175
0.0000

Figure 5. Perceived brightness. Average score on a 7-point

scale from 1 to 7 (rating 4 being the neutral answer) and
standard deviation of 20 observers. Illumination parameters
are reported on the abscissa axis.

3.3.2. Significant effect of CCT and of illuminance

.
.
.
.
.

High contrast acuity

Bright versus dark
Clear versus crepuscular
Non-glaring versus glaring
Pleasant versus unpleasant

Almost all subjective scales are modified by the

CCT. One class of scales (pleasant, comfortable,
relaxing) is related to the mood, the well-being. It
was constructed to address the pleasing-ness quality.
Results indicate an increase of comfort, pleasantness

1442

F. Vienot et al.

Table 7. Detailed analysis of variance for the darkness

brightness judgments. F-values and significance. Degree of
freedom for temperature, illuminance and participant is,
respectively, 1, 2 and 29.
Between 2700 Illuminance Participant
and 4000K
F
5.6
29.07
1.41
High contrast p
0.0202
0.0000
0.1147

Downloaded By: [Capderou, Christine] At: 11:31 7 September 2009

Between 4000
and 6500K
F
15.77
p
0.0001

Luminance

Participant

19.24
0.0000

1.53
0.0692

Table 8. Effect of correlated colour temperature and effect

of illuminance on subjective sensations. Up arrow means that
sensation significantly increases with an increase of CCT or
illuminance. Bottom arrow means decreasing. The p-value is
given next to the arrow.
Effect of
CCT
Bright
Clear
Glaring
Pleasant
Comfortable
Relaxing
Warm
Cheerful
Colour
rendering

"
"
"
#
#
#
#

0.0000
0.0000
0.0000
0.0042
0.0007
0.0000
0.0000

Effect of
illuminance
"
"
"
50"100

0.0000
0.0000
0.03
0.002

"

0.0000

Effect of
participants
Yes
Yes

Yes
Yes

Table 9. Effect of correlated colour temperature and effect

of illuminance on visual performance. Up arrow means that
performance significantly improves with an increase of CCT
or illuminance. Bottom arrow means impairing. The p-value
is given next to the arrow.
Effect of
CCT
Low contrast
" 0.0000
acuity
High contrast
" 0.04
acuity
Visual performance
with Landolt rings
Reading speed

Effect of
Effect of
illuminance participants
Yes
"

0.0002

Yes
Yes
Yes

and impression of relaxation, when the CCT decreases.

The other scales (clear, bright, non-glaring) are related
to the amount of perceived light. As expected, brightness increases with luminance. Results also show that
at a given luminance level, brightness increases with
CCT. We will come back to this point in the discussion
section.

4. Discussion and conclusion

The aim of our study was to re-examine Kruithofs
rule. We chose to investigate the effect of CCT on
performance and subjective percepts.

4.1. Visual performance

The general finding that performance is not affected
by correlated colour temperature but only by luminance is largely acknowledged by previous studies on
lighting [7,16,18]. The lack of effect of CCT on visual
performance might be explained by the fact that
visual performance depends mainly on luminance
contrast. The improvement of low contrast acuity
with illuminance goes along with the well-known
improvement of contrast sensitivity function with
illuminance [27]. It is also known that at threshold,
contrast sensitivity is an achromatic response and that
the chromatic content does not influence the luminous
contrast sensitivity [28].
It is usually reported that colour temperature
conditions only slightly affect performance [29]. Also,
it has been shown that, under photopic conditions, the
reading rates of normal subjects are independent of
wavelengths [30].
In our study, suprathreshold tasks (Landolt ring
test, reading) are affected neither by CCT nor by
illuminance which may indicate that observers compensate for any difficulty arising from poor illumination conditions.

4.2. Brightness versus luminance

The most fruitful experiment is the series of perceptual
scaling. It clearly appears that brightness, as well as
other quantities dealing with the amount of perceived
light such as glare or clarity increases with CCT at
constant luminance. This effect has already been
diversely appraised by other authors [10,11].
Figure 5 allows a comparison of the gain of
brightness B due to correlated colour temperature
Tcp and the gain of brightness B due to luminance L.
In the absence of many experimental points, we could
simply assume a linear relationship. Nevertheless, the
model would need a constant term to predict at best
the experimental data. Therefore, we have chosen to
propose a non-linear model2 with one term that
converges toward zero at zero luminance. We write
B
Lm ncp :

Regression applied to experimental results yields

0.859994,  0.000321, m 0.31 and n 0.1.

Downloaded By: [Capderou, Christine] At: 11:31 7 September 2009

Journal of Modern Optics

Equation (2) means that, within the context of the
mode of illumination of our experiment, increasing
CCT from 4000 to 5000 K is equivalent to increasing
luminance (we restate that the photometric measurements were done in terms of luminance) from 100 to
134 cd m
2.
This gain of brightness is slightly larger than the
gain mentioned by Harrington [10] who used filtered
incandescent light. It is much more than the figure
modelled by Ware and Cowan [31] or Nakano [32] and
opposite to the figure yielded by Guth and Lodges [33]
model. We stress the fact that establishing a model
linking brightness B and correlated colour temperature
T makes sense only when the variation of CCT is
accompanied by a regular change of spectral power
distribution and not a change in lamp technology
which, for instance, was the case in Harringtons [10]
work or is the case in our work.

4.3. Hypothesis concerning psycho-physiological

effects
We propose to establish a parallel between our results
and the recently discovered melanopsin photopigment
that is included within the cell body and dendrites
of a subset of ganglion cells of rodents [34] and
primates [35]. The so-called intrinsically photosensitive
retinal ganglion cells (ipRGCs) are directly hit by light
entering the eye. They project onto the lateral
geniculate nucleus as well as to the subcortical
brain [34]. It has been proposed that they control the
circadian system, that they mediate the sustained pupil
response [36], and also that their response increases
with the amount of light entering the eye, with or
without an adaptation mechanism. The peak wavelength of their action spectrum, measured in vitro in the
primate retina, is about 480482 nm [35].
Using LED illumination, our scenarios have
allowed us to control the spectral power distribution
of the light. So, knowing the spectral power distribution of the light, it is possible to calculate the intensity
of the stimulus of these ipRGCs. We have hypothesised
that the spectral template of the melanopsin photopigment is the same as the template of rod and cone
rhodopsin. Thus, starting with the low density
absorbance spectra of the visual pigments proposed
by Stockman and Sharpe [37], we applied a shift of the
template along the frequency axis [38], so as to position
the peak at 482 nm. Then, we reconstructed the action
spectrum of the melanopsin at the corneal plane.
Because dendrites and the cell body of ganglion cells
are thin compared with the cone external segment, we
could assume a small density (equal to 0.1) for the
photopigment. No correction was introduced for

1443

Figure 6. Stimulus intensity for S-cones, melanopsin and

rods. Calculation from the spectral power distribution of
light. Normalisation relative to the average value for each
photoreceptor type. Illumination parameters are reported on
the abscissa axis. (The colour version of this figure is
included in the online version of the journal.)

macular pigmentation because the ipRGCs are supposed to be evenly distributed over the total retina. The
lens spectral absorbance proposed by Stockman and
Sharpe could apply to our young observers. Finally,
the stimulus that elicits the response of melanopsin is
proportional to the convolution of the spectral power
distribution of the LED illumination and the action
spectrum of ipRGCs. Our calculation implies additivity of the menalopsin phototransduction, a property
that has been assessed in the circadian response in
rodents [39].
To explain our results about brightness, a fair
scrutiny requires us to discuss possible mechanisms to
explain the brightness response other than the melanopsin stimulation such as the S-cones participation
or the rods participation. We calculated the S-cone
stimulus using the fundamental response curve proposed by the CIE [40] and we calculated the rod
response using the V0 () scotopic luminous efficiency
function. Results of the calculation of the S-cones, the
melanopsin and the rods stimuli are displayed in
Figure 6 where every stimulus is normalised to its
average value. The figure clearly shows the increase of
the stimulus with CCT when luminance is constant
whichever photoreceptor is involved. We note the same
tendency as for the subjective scaling of brightness
(Figure 5). Therefore, we propose to consider as a
possible explanation of the relative increase of brightness with CCT an increase of intensity of the stimulus
for the melanopsin, which occurs when luminance
increases as well as when CCT increases at constant
luminance.

Downloaded By: [Capderou, Christine] At: 11:31 7 September 2009

1444

F. Vienot et al.

Were the melanopsin responsible for the relative

increase of brightness with CCT, would the mechanism
operate directly or indirectly? Histological studies of
ipRGCs have shown that they project onto the
pretectum which is known to control pupillary
constriction. Whenever ipRGCs absorb light, it
induces a sustained pupil constriction. Indeed we
checked on three control observers who had not
participated in the experiment that the pupil constricts
when the illuminance increases as well as when the
CCT increases. We took photographs of their eyes, in
which we marked the border of the pupil and the
border of the iris. We plotted the relative diameter of
the pupil in Figure 7. On these control observers, we
have verified that the pupil diameter decreases when
the CCT increases at a given illuminance. Thus,
because the brightness increases even though the
amount of light impinging on the retina decreases, a
possible indirect mechanism might be operating. The
CCT increases, so the melanopsin stimulus increases,
so the pupil constricts, then, although less light reaches
the retina, the observer feels that the illumination is
brighter. The way that the indirect mechanism operates
is probably complex. On the one hand, the photoresponse of ipRGCs encodes quantal irradiance over a
34 log-unit range that encompasses most of daylight
vision [35]. On the other hand, some experiments [41]
indicate that ipRGCs display both light and

dark adaptation. Alternately, whereas rods saturate

at photopic level, it has been shown that rods dominate
the pupil behaviour during dark adaptation [42].
Indeed, the interaction between inputs from rods,
cones and ipRGC signals deserves further study.
An aside consequence of the pupillary constriction
is the modification of the point spread function that
accompanies the change of pupil area [43]. It could be
possible to explain the significant improvement of high
contrast acuity at high CCT. Such a possibility has
already been mentioned by other authors while it has
been seriously opposed by others. On the one hand, in
a number of papers, Berman and co-authors [11,44]
have found that light with higher correlated colour
temperature allowed better visual acuity and greater
brightness perception compared to light of lower
correlated colour temperature, with both light conditions controlled to be at the same photopic illuminance. They checked that a variation of the pupil size
was accompanied by an improvement of visual acuity.
On the other hand, Boyce and co-authors [18] have
demonstrated that the lamp spectrum affected neither
the speed nor the accuracy of performance on an acuity
task. We note that at that time, only fluorescent lamps
with moderate CRI were used which provides little
information about the spectral power distribution of
light. The use of new light sources such as LED clusters
with high CRI, the spectrum of which has been
measured, brings some useful information on this
aspect. Also, part of the discussion by these authors
bears the assumption that the rod/cone ratio controls
the pupil size, an argument which could be reconsidered in parallel with the melanopsin/cone ratio.
Establishing a link between pupil size and achromatic
visual task performance deserves further investigation.

4.4. About Kruithofs rule

Figure 7. Relative diameter of the pupil of three control

subjects measured on photographs taken after one minute of
adaptation to each illumination. Results of observer JA are
derived from one photograph. Results of observers SB and
JLR are derived from three photographs. (The colour version
of this figure is included in the online version of the journal.)

In the end, have we validated Kruithofs rule?

The responses should be examined with caution. In
one sense, we have validated Kruithofs statement that
high CCT at low illuminance is unpleasant.
Nevertheless, we cannot conclude that low CCT
should be confined to low illuminance to arouse
pleasant sensations. Our observers, after having been
invited to perform various visual tasks, have responded
that illumination is significantly less pleasant, less
comfortable, less relaxing at high CCT than at
low CCT. In Figure 8, we have reported the average
rating chosen on the UnpleasantPleasant scale.
It shows that low CCT are judged the most pleasant
at any illuminance 150, 300 or 600 lx.
Several studies have yielded a similar conclusion [1517], although a few studies are partially

1445

Downloaded By: [Capderou, Christine] At: 11:31 7 September 2009

Journal of Modern Optics

contradictory. For instance, low illuminances were
described by Cuttle and Boyce [15,16] as dull and
hazy regardless of CCT. By contrast, only the dim
scale out of their 19 scales showed a significant effect
due to CCT, in a somewhat inconsistent pattern. The
authors concluded that there was no apparent trend
due to CCT. We emphasise that a questionnaire is
subject to interpretation and that the observers
response depends upon the context. In Manavs study
[45], the room was furnished and emphasis was put on
the objects of the office environment and the wall
opposite from the viewpoint. Regrettably, our experiment could not be conducted in a real room which
would need much more flux than available with LEDs.
Thus the illuminated field of view was restricted to
about 80 and the large surround was in the dark. This
could explain the highest score for pleasantness and
comfort at the lowest correlated colour temperature
whether at the lowest illuminance or at the highest
illuminance (Figure 8). We stress the fact that unlike
previous studies conducted with fluorescent lamps, our
experiments were conducted with LED clusters that
deliver continuous light spectra with a very high CRI.
The quality of our illumination may have favoured the
consistency of the subject responses.
Finally, we conclude that visual performance
probably cannot predict subjective impressions
aroused from lighting. Under the conditions used in
the present study, we can say that Kruithofs rule has
been only partially validated. When the colour rendering index is very high and the light spectrum is under
control, there is no indication that high colour
temperature is judged more pleasant than low colour
temperature at higher illuminance levels.

Figure 8. Perceived pleasantness. Average score on a 7-point

scale from 1 to 7 (rating 4 being the neutral answer) and
standard deviation of 20 observers. Illumination parameters
are reported on the abscissa axis.

Acknowledgements
We would like to thank Corinne Talotte from the Direction
de la recherche de la Societe Nationale des Chemins de fer
Francais for initiating this study, Alain Bricoune and Frantz
Dennery from LedToLite for providing the LED illumination system, Eric Castet and Francois Vital-Durand for
counselling us on how to choose reading texts, Danie`le
Dubois and Gaelle Delepeau from Laboratoire dAcoustique
Musicale for guidance in preparing the questionnaire, Jean
Le Rohellec for photographs and comments and Kenneth
Knoblauch for comments on the manuscript.

Notes
1. The English version is a translation from French of the
subjective scales proposed to evaluate light quality: Non
eblouissante Eblouissante; Sombre Lumineuse;
Froide Chaude; Rendu des couleurs artificiel
Rendu des couleurs naturel; Crepusculaire Claire;
Triste Gaie; Fatigante Relaxante; Inconfortable
Confortable; Desagreable Agreable.
2. Introducing a reference CCT Tcp,o would yield the model

B
Lm  Tcp
Tcp,o

n

with
0.821918,  0.000467, m 0.31, n 0.96 and

Tcp,o 4. It only slightly modifies the prediction.

References
[1] Kruithof, A.A. Philips Techn. Rev. 1941, 6, 6596.
[2] Commission Internationale de lEclairage. Method of
Measuring and Specifying Colour Rendering Properties
of Light Sources, (CIE Pub. 13.2-1974), CIE Pub.13.
3-1995.
[3] Tiller, D.K. Qualite de leclairage. http://irc.nrccnrc.gc.ca/pubs/bsi/92-5-print_f.html (accessed 2009).
[4] Fotios, S.A. Lighting Res. Technol. 2001, 3, 163178.
[5] Bodmann, H.W.; Sollner, G.; Voit, E. Bewertung von
Beleuchtungniveaus bei verschiedenen lichtarten.
Proceedings of the CIE 15th Session, Vienna, 1963.
[6] Bodmann, H.W. Trans. Illum. Eng. Soc. 1967, 32, 2240.
[7] Smith, S.W.; Rea, M.S. Relationships between Office
Task Performance and Ratings of Feelings and Task
Evaluations under Different Light Sources and Levels.
Proceedings of the CIE 19th Session, Kyoto, 1979.
[8] Boyce, P.R. Lighting Res. Technol. 1977, 9, 1124.
[9] Flynn, J.E.; Spender, T.J. J. Illum. Eng. Soc. 1979, 6,
167179.
[10] Harrington, R.E. J. Opt. Soc. Am. 1954, 44, 113116.
[11] Berman, S.M. Lighting Res. Technol. 2008, 40, 373376.
[12] Aston, S.M.; Bellchambers, H.E. Lighting Res. Technol.
1969, 1, 259261.
[13] Bellchambers, H.E.; Godby, A.C. Lighting Res. Technol.
1972, 4, 104106.
[14] Kanaya, S.; Hashimoto, K.; Kichize, E. Subjective
Balance Between General Color Rendering Index Color

1446

[15]
[16]
[17]
[18]
[19]
[20]
[21]

Downloaded By: [Capderou, Christine] At: 11:31 7 September 2009

[22]

[23]
[24]

[25]
[26]
[27]
[28]
[29]

F. Vienot et al.

Temperature and Illuminance of Interior Lighting.

Proceedings of the CIE 19th Session, Kyoto, 1979.
Boyce, P.R.; Cuttle, C. Lighting Res. Technol. 1990, 22,
1936.
Cuttle, C.; Boyce, P.R. Lighting in Australia 1988, 8,
1728.
Davis, R.G.; Ginthner, D.N. J. Illum. Eng. Soc. 1990,
19, 2738.
Boyce, P.R.; Akashi, Y.; Hunter, C.M.; Bullough, J.D.
Lighting Res. Technol. 2003, 35, 141156.
Boyce, P.R. Presented at the 2nd CIE Expert
Symposium on Lighting and Health, 2006.
Mahler, E.; Ezrati, J.-J.; Vienot, F. Color Res. Appl.
2009, 34, 817.
Galinier, L.; Clouzet, A. Une Nouvelle echelle
dacuite, Inform optique, 1998, 160. Test dacuite,

Laurent Galinier, 2005, Fax International, www.fax.fr

(accessed 2007).
Hopkinson, R.G.; Collins, J.B. The Ergonomics of
Lighting; McDonald Technical and Scientific: London,
1970; pp 50, 7079.
CIE. The Correlation of Models for Vision and Visual
Performance; CIE Publication 145-2002; Vienna, 2002.
Hahn, G.A.; Penka, D.; Gehrlich, C.; Messias, A.;
Weismann, M.; Hyvarinen, L.; Leinonen, M.; Feely,
M.; Rubin, G.; Dauxerre, C.; Vital-Durand, F.;
Featherston, S.; Dietz, K.; Trauzettel-Klosinski, S.
Br. J. Ophthalmol. 2006, 90, 480484.
Legge, G.E.; Pelli, D.G.; Rubin, G.S.; Schleske, M.M.
Vision Res. 1985, 25, 239252.
Lanthony, P. Doc. Ophthalmol. 1978, 46, 185189.
Van Nes, F.L.; Bouman, M.A. J. Opt. Soc. Am. 1967,
57, 401406.
Lennie, P.; Pokorny, J.; Smith, V. J. Opt. Soc. Am. A
1993, 10, 12831293.
Yamagishi, M.; Yamaba, K.; Nokura, K.; Nagata, M.
The Effect of the LED Lighting Characteristics at the
Visual Performance. Proceedings of the 1st International

[30]
[31]
[32]

[33]
[34]
[35]

[36]

[37]
[38]
[39]

[40]

[41]
[42]
[43]
[44]
[45]

Conference on White LEDs and Solid State Lighting,

Tokyo, November 2630, 2007.
Legge, G.E.; Rubin, G.S. J. Opt. Soc. Am. A 1986, 3,
4051.
Ware, C.; Cowan, W.B. CIE J. 1986, 5 (2), 5759.
Nakano, Y.; Yamada; Suehara, K.; Yano, T. A Simple
Formula to Calculate Brightness Equivalent Luminance.
Proceedings of the CIE 24th Session, Warsaw, 1999.
Guth, S.L.; Lodge, H.R. J. Opt. Soc. Am. 1973, 63,
450462.
Hattar, S.; Liao, H.-W.; Takao, M.; Berson, D.M.;
Yau, K.-W. Science 2002, 295, 10651070.
Dacey, D.M.; Liao, H.; Peterson, B.; Robinson, F.;
Smith, V.; Pokorny, J.; Yau, K.-W.; Gamlin, P.D.
Nature 2005, 433, 749754.
Gamlin, P.D.; McDougal, D.H.; Pokorny, J.; Smith,
V.C.; Yau, K.W.; Dacey, D.M. Vision Res. 2007, 47,
94654.
Stockman, A.; Sharpe, L.T. Vision Res. 2000, 40,
17111737.
Vienot, F.; Serreault, L.; Pardo Fernandez, P. Visual
Neurosci. 2006, 23, 419427.
Bullough, John D.; Figueiro, Mariana G.; Possidente,
Bernard P.; Parsons, Robert H.; Rea, M.S. Zoolog. Sci.
2005, 22, 223227.
Commission Internationale de lEclairage. Fundamental
Chromaticity Diagram with Physiological Axes Part 1;
Technical Report for CIE: Austria, January 2006; CIE
170-1: 2006.
Wong, K.; Dunn, F.; Berson, D. Neuron 2005, 48,
10011010.
Alpern, M.; Campbell, F.W. J. Physiol. Lond. 1962, 165,
5P7P.
Campbell, F.W.; Green, D.G. J. Physiol. Lond. 1965,
181, 576593.
Berman, S.M.; Navvab, M.; Martin, M.J.; Sheedy, J.;
Tithof, W. Lighting Res. Technol. 2006, 38, 4152.
Manav, B. Build. Environ. 2007, 42, 979983.