ArticlePDF Available

Abstract and Figures

Several studies have recorded color emotions in subjects viewing uniform color (UC) samples. We conduct an experiment to measure and model how these color emotions change when texture is added to the color samples. Using a computer monitor, our subjects arrange samples along four scales: warm–cool, masculine–feminine, hard–soft, and heavy–light. Three sample types of increasing visual complexity are used: UC, grayscale textures, and color textures (CTs). To assess the intraobserver variability, the experiment is repeated after 1 week. Our results show that texture fully determines the responses on the Hard-Soft scale, and plays a role of decreasing weight for the masculine–feminine, heavy–light, and warm–cool scales. Using some 25,000 observer responses, we derive color emotion functions that predict the group-averaged scale responses from the samples' color and texture parameters. For UC samples, the accuracy of our functions is significantly higher (average R2 = 0.88) than that of previously reported functions applied to our data. The functions derived for CT samples have an accuracy of R2 = 0.80. We conclude that when textured samples are used in color emotion studies, the psychological responses may be strongly affected by texture. © 2010 Wiley Periodicals, Inc. Col Res Appl, 2010
Content may be subject to copyright.
Texture Affects Color Emotion
Marcel P. Lucassen,*Theo Gevers, Arjan Gijsenij
Intelligent System Laboratory Amsterdam, University of Amsterdam, Amsterdam, The Netherlands
Received 11 March 2010; revised 27 April 2010; accepted 8 June 2010
Abstract: Several studies have recorded color emotions
in subjects viewing uniform color (UC) samples. We
conduct an experiment to measure and model how these
color emotions change when texture is added to the
color samples. Using a computer monitor, our subjects
arrange samples along four scales: warm–cool, mascu-
line–feminine, hard–soft, and heavy–light. Three sample
types of increasing visual complexity are used: UC,
grayscale textures, and color textures (CTs). To assess
the intraobserver variability, the experiment is repeated
after 1 week. Our results show that texture fully deter-
mines the responses on the Hard-Soft scale, and plays
a role of decreasing weight for the masculine–feminine,
heavy–light, and warm–cool scales. Using some 25,000
observer responses, we derive color emotion functions
that predict the group-averaged scale responses from
the samples’ color and texture parameters. For UC
samples, the accuracy of our functions is significantly
higher (average R
¼0.88) than that of previously
reported functions applied to our data. The functions
derived for CT samples have an accuracy of R
0.80. We conclude that when textured samples are used
in color emotion studies, the psychological responses
may be strongly affected by texture. Ó2010 Wiley Periodi-
cals, Inc. Col Res Appl, 36, 426 436, 2011; Published online 12 No-
vember 2010 in Wiley Online Library ( DOI
Key words: color; texture; color emotion; observer vari-
ability; ranking
There is growing interest in the understanding of human
feelings in response to seeing colors and colored objects.
The so called ‘‘color emotions’’ (i.e., psychological
responses to color), involved in published studies, do
usually not refer to basic human emotions, such as hap-
piness, surprise, or fear. Rather, they capture an observ-
ers’ response on an associated affective dimension speci-
fied by the investigators, such as warm–cool and hard–
soft. Color emotion studies recently published
on the selection of emotional scales and investigate how
these scales are related by means of factor analysis.
Then, regression analysis is usually applied to reveal the
relationships of human responses on these scales with
the underlying color appearance attributes, such as light-
ness, chroma, and hue. Roughly summarizing these stud-
ies, the common finding is that the color emotions are
reasonably well described by a small number of seman-
tic factors, such as the colour weight, colour activity,
and colour heat found by Ou et al.
or valence, arousal,
and dominance by Suk and Irtel.
Of the perceptual
attributes that characterize the samples, lightness and
chroma are most frequently reported as being the rele-
vant parameters for quantitative prediction of the color
emotions, although hue cannot be ignored in scales, such
as warm–cool.
Several studies investigate whether color emotions can
be regarded as culture specific or universal.
In most
of them, it is found that the influence of cultural back-
ground is limited.
Additionally, the effect of media type (paper vs. CRT
display) upon the emotional responses to color is studied.
No effect of media type is measured.
Many color vision studies regard color as the main
experimental variable, as if it is an isolated object
property. However, real life objects are seldom
uniformly colored. Nonuniformity of object colors (tex-
ture) and their environment seems to be the rule rather
than the exception.
Therefore, a logical next step
in color emotion studies is the extension from uniform
color (UC) toward color texture (CT). So far, the role
of texture in color emotion has received only little
attention. An early study by Tinker
shows that
surface texture, as represented by coated paper or
cloth, has little or no effect upon apparent warmth or
*Correspondence to: Dr. M. P. Lucassen (e-mail:
Contract grant sponsor: Dutch Organization for Scientific Research;
contract grant number: VICI NWO grant 639.023.705.
C2010 Wiley Periodicals, Inc.
426 COLOR research and application
affective value of colors. Kim et al.
use color and
texture features to predict human emotions based on
textile images. Erhart and Irtel
indicate that surface
structure can change the emotional effect of colored
textile samples, depending upon the color. More
recently, Simmons and Russell
report that the addi-
tion of texture can significantly change the perceived
unpleasantness of colors, depending on the texture
class. This finding, however, is confined to a single
emotion, namely unpleasantness. So, although a handful
of studies exist, what is still lacking is a systematic
approach to color emotion in which the complexity of
the color stimulus is gradually increased.
This article investigates the effect upon color emotion
of adding texture to color, using a selection of four
color emotion scales. Our experiments build upon, but
differ in a number of ways from previous studies. Most
importantly, instead of only studying uniformly colored
samples, we also use samples with grayscale textures
(GTs) and samples with CTs. These textures are primi-
tive (no semantics) and synthesized to prevent strong
associations, such as reported in Simmons and Rus-
and can be fully parameterized. Second, we
introduce a method in which all samples (shown on a
computer display) remain visible during experimental
trials. The advantage is that they can be ordered con-
veniently along an emotion scale. Third, our subjects
perform the full experiment twice, with at least 1 week
in between the first and second measurement. This
allows quantification of the intraobserver variability over
time, on which we are the first to report. We believe
that repeatability information is at least as important as
the information obtained from more observers. Finally,
we systematically sample the available color gamut of
our color monitor to optimally cover the lightness,
chroma, and hue domain.
We analyze our data in terms of rank correlations
within subjects and between subjects and provide quanti-
tative descriptions. We derive color and texture emotion
formulae that predict the group-averaged responses on the
emotion scales from the samples’ color and texture
descriptors. Using these models, we present visualizations
of the arrangement of the samples used in the experi-
One of the problems we encountered in a pilot experiment
is that when samples are shown one after the other, sub-
jects tend to forget what responses they gave on the emo-
tion scales for similar samples shown earlier in the trial.
This leads to an unnecessary increase in variability in
the subjects’ responses, and therefore lower intra-
and interobserver correlations. We therefore design our
experiment in such a way that all samples remain visible
during a trial. We ask our subjects to order 105 square
samples horizontally along an emotion scale labeled with
opposite word pairs (e.g., warm-cool). They use the com-
puter mouse to drag samples from their initial location on
the top of the screen. Samples can be dragged to any
position on the screen to keep an overview of the arrange-
ment of the samples. Subjects know that only
the horizontal position of the samples on the scale will be
Four emotion scales are used: warm–cool, masculine–
feminine, hard–soft, and heavy-light. These four scales
are tested in separate experimental trials. The section
‘Selection of Emotion Scales’’ motivates the selection of
these four scales. There are three conditions differing only
in the type (complexity) of samples used. In the UC con-
dition, uniformly colored samples are used that were sys-
tematically selected from the sRGB color gamut of our
color monitor. In the GT condition, grayscale samples
have a texture created in luminance, but not in the chro-
matic domain. Textures are generated using Perlin
The samples in the CT condition are basically
blended from the UC and GT samples, thus showing the
GTs applied to a single color.
Sample Selection
All samples were square patches of 100 3100 pixels.
Below we discuss the selection of the three types of sam-
ples used, in order of increasing visual complexity: UC,
GT, and CT.
Uniform Color. Our goal is a systematic sampling of
the available color gamut. The color monitor that we
use to display the samples is calibrated to the sRGB
color space.
Details hereof are presented in the section
‘Monitor.’’ Within the sRGB color gamut, we select
100 chromatic samples and five achromatic samples. The
chromatic samples are selected at five lightness levels
(L*¼10, 30, 50, 70, 90). For each level in L*, 10 hue
angles are selected at 36 degree interval (h¼0, 36,
72, ... , 288, 324). Finally, for each of these hue angles
two levels in C* are selected, being the maximum value
max) within the sRGB gamut and half the maximum
value (C
max/2). Figure 1 shows the positions of the sam-
ples in the a* and b* plane of CIELAB color space.
Note that different C
max values are obtained for the dif-
ferent hue angles, typical for any color gamut. Five
additional achromatic samples are selected at L*¼20,
40, 60, 80, 100. A specification of these samples in
terms of CIE L*, C*, and h
is presented in Table AI
in Appendix A. With the above sample selection, we
cover the lightness, chroma, and hue domain of our
monitor’s color gamut.
Grayscale Texture. Being aware of the fact that texture
is one of the characteristics that may alter surface percep-
at this point in our research we do not want
to use natural textures to avoid the possibility of strong
inherent emotional associations. Therefore, we use tex-
tures that are synthesized on the basis of Perlin noise
using the open-source libnoise library (http://libnoise. Perlin noise is a primitive structure used
Volume 36, Number 6, December 2011 427
in procedural texture generation, and is pseudo-random in
appearance. All visual details in Perlin noise are the same
size, which means that theoretically such an image can be
said to truly represent a single texture. Perlin noise can be
fully parameterized implying that we can reliably generate a
random sample of textures by randomly sampling from the
Perlin parameter space. Through controlling the number of
octaves, the frequency of each octave and the amplitude of
each octave we can respectively control the level of detail,
the granularity and the contrast of the resulting texture. For a
more detailed description of Perlin noise-based textures, we
refer to the aforementioned libnoise library. The GTs are
achromatic, showing only spatial variations in lightness.
Figure 2 shows an example of how changes in the individual
Perlin parameters (number of octaves, frequency, persist-
ence, lacunarity) affect the visual appearance. The extent to
which changes in these parameters result in changes in vis-
ual appearance depends on the actual position in this param-
eter space. Summarizing, we have created textures that have
no semantics, which can be systematically controlled by the
Perlin noise parameters, form a subset of all possible tex-
tures and have a natural appearance.
Color Texture. Our CT samples are colored versions of
the GTs. They are not multicolored, but consist of light-
ness variations in a singe color. Although more complex
(multicolor) textures exist in reality, we used the simpler
FIG. 1. Positions of the uniform color samples in CIELab
color space, covering the sRGB gamut of our color moni-
tor. One hundred chromatic samples are selected at 5
lightness levels (L*), 10 hue angles (h
), and 2 chroma lev-
els (C*). Another five achromatic samples are selected at
intermediate lightness levels. A specification of these sam-
ples is presented in Table AI in Appendix A.
FIG. 2. The effect of varying the Perlin noise parameters is demonstrated here. From left to right: number of octaves, fre-
quency, persistence, and lacunarity. The middle row shows identical samples at parameter values (6, 0.5, 0.6, and 2.5).
The top row shows lower values for the parameter in question, while keeping the other parameters fixed. For example, the
top left sample has values (1, 0.5, 0.6, and 2.5). The bottom row shows higher values for the parameter in question, while
keeping the others fixed. For example, the bottom left sample has values (12, 0.5, 0.6, and 2.5). Note that there is some
overlap in visual effect for the higher parameter values (bottom row), but this depends on the position in parameter space.
For this particular example, the upper row shows clear differences in visual effect.
428 COLOR research and application
monochrome textures as a first step. The advantage is that
this way, the effect of adding texture to color can be stud-
ied in stages involving an increasing visual complexity.
Figure 3 illustrates how the colored textures are created.
Selection of Emotion Scales
Contrary to a number of preceding studies,
our pri-
mary aim is not to find out which scales are most appro-
priate to capture color emotions, but rather to explore the
effects of adding texture to color samples. We therefore
select four scales with opposite word-pairs that have been
frequently used in previous studies
and for
which we also gained experimental confidence in our pilot
studies. These four scales are warm–cool, masculine–fem-
inine, hard–soft, and heavy–light. The warm–cool scale is
not used for the GT samples, because our subjects found
this combination very hard, if not impossible. With the
exception of the masculine–feminine scale, quantitative
descriptions of the scales on the basis of CIELAB param-
eters are available from previous studies, which enable us
to compare our results with that of other investigators.
Ten subjects participated in the experiments, six men and
four women. Their ages range from 26 to 53, with an aver-
age of 31.9. Subjects are from seven different nationalities:
Dutch (4), Chinese (1), Russian (1), Italian (1), Spanish (1),
Polish (1), and German (1). All subjects have normal color
vision and normal or corrected to normal visual acuity. Sub-
jects are screened for color vision deficiencies with the
HRR pseudo-isochromatic plates (4th edition), allowing
color vision testing along both the red-green and yellow-blue
axes of color space.
The HRR test is viewed under pre-
scribed lighting (CIE illuminant C) using the True Daylight
Illuminator (Richmond Products), whereas illumination by
other light sources is reduced to a minimum. The first author
also participates as a subject in the experiment; the other
subjects are unaware of the purposes of the experiment.
Subjects participate on voluntary basis and do not receive a
financial reward; they are all employed or studying at the
institute where the experiment is carried out.
Monitor and Calibration
Stimuli are presented on a high-resolution (1600 3
1200 pixels, 0.27 mm dot pitch) calibrated LCD monitor,
an Eizo ColorEdge CG211. The monitor is driven by a
computer system having a 24-bit (RGB) color graphics
card operating at a 60 Hz refresh rate. Before each experi-
mental session, a colorimetric calibration of the LCD is
performed using a spectrophotometer (Eye-one, Gretag-
Macbeth [now X-Rite]). The monitor is calibrated to a
D65 white point of 80 cd/m
, with gamma 2.2 for each of
the three color primaries. The CIE 1931 x,ychromaticities
coordinates of the primaries were (x,y)¼(0.638, 0.322)
for red, (0.299, 0.611) for green and (0.145, 0.058) for
blue, respectively. With these settings of our monitor, we
closely approximate the sRGB standard monitor profile.
Spatial uniformity of the display, measured relative to the
center of the monitor, was DE
ab \1.5, according to the
manufacturer’s calibration certificates.
Subjects were seated in front of the monitor at a viewing
distance of about 60 cm. The screen size extended 39.683
30.28of visual angle, and a sample (square patch of 100 3
100 pixels) 2.6832.68. Samples were initially displayed in
random order at the top of the screen. Subjects dragged the
samples away from their initial position to give them a rela-
tive ordering along the horizontal emotion scale. Subjects
knew that only the horizontal position would be analyzed,
the vertical space could be used to keep an overview of the
samples. After ordering the first group of 50 samples, sub-
jects pressed a button after which the second group of 55
samples was shown (the first 50 samples remained visible).
FIG. 3. Color Textures are created by combining uniform
color patches with the grayscale textures.
FIG. 4. Experimental result (data from a single observer)
for the uniform color samples, ordered horizontally along
the masculine–feminine scale. Only the horizontal position
matters. At a viewing distance of 60 cm, the screen
size extends 39.68330.28visual angle, and one sample
2.6832.68. The 100 chromatic patches systematically
sample the sRGB color gamut at 5 lightness levels, 10 hue
levels, and 2 chroma levels. Additionally, five achromatic
samples are used.
Volume 36, Number 6, December 2011 429
During a trial, all samples could be reordered if desired. One
trial of 105 samples took about 5–10 min. All subjects
repeated the experiment with at least 1 week in between the
first and the second measurement.
Examples of the results for a single observer on the three
sample types UC, GT, and CT are shown in Figs. 4, 5,
and 6, respectively. The emotion scale arbitrarily extends
from 24 (outer left) to þ4 (outer right), with value zero
being neutral (center). Actual scale values for the samples
are calculated from their horizontal midpoints. Through-
out this article we use ranks (i.e., a relative order from
the left side to the right side of the scale) and rank corre-
lations rather than the absolute scale values, because the
scales are not expected to be linear. An additional advant-
age of using ranks is that it corrects for individual differ-
FIG. 5. Experimental result (data from a single observer) for the grayscale texture samples, ordered horizontally along the
heavy–light scale. Heavy extends to the left from the center, Light to the right side from the center. Neutral (neither Heavy
nor Light) is at the center of the horizontal scale. Textures are made using Perlin noise.
FIG. 6. Experimental result (data from a single observer) for the color texture samples, ordered horizontally along the
warm–cool scale. The neutral point is in the center, warm extends to the left side of the scale, cool to the right side of the
scale. The color textures are made from blending the uniform color samples with the grayscale textures.
430 COLOR research and application
ences in the used scale range. For instance, one subject
may use the full scale range to position the samples,
whereas another subject may use only 75% of that range.
Statistical analyses are performed with the Statgraphics
Centurion XV software package.
Quantitative Analysis: Observer Variability
Intraobserver Agreement. How well do observers agree
with themselves? For each observer, sample type and
emotion scale, we determine the rank correlation between
the first and second measurement (Table I). This correla-
tion is a measure for the intraobserver agreement, or in
other words, the repeatability. For 105 samples, the criti-
cal value of the correlation coefficient is about 0.195 at
the 95% confidence level. Table I shows that the correla-
tion between the first and second measurement is highly
significant, for all subjects and all conditions, except for
subject 6 on the Heavy-Light scale for the GT samples.
For the UC samples, the correlation averaged over sub-
jects and emotion scales is 0.73, which is higher than the
corresponding values for the GT samples (r¼0.66) and
the CT samples (r¼0.65). A paired t-test on the UC and
the CT data shows that the difference is significant at the
95% confidence level (P¼0.015). The same test on the
UC and GT data reveals that the difference is not signifi-
cant (P¼0.23), but this is based on less data because the
warm–cool scale was not measured for the GT samples.
Apparently, subjects reproduce their color emotional
responses on UC samples better than on the CT samples.
Averaged over the three sample types, the highest intraob-
server agreement is found for the warm–cool scale (r¼
0.74), followed by heavy–light (r¼0.70), masculine–
feminine (r¼0.69), and hard-soft (r¼0.60). Consider-
ing that the second measurement is made about 1 week
after the first measurement, these intraobserver values
TABLE II. Interobserver agreement.
Sample type Emotion scale
1 2 345678910Average
Uniform color WC 0.41 0.83 0.90 0.20 0.82 0.27 0.59 0.51 0.47 0.73 0.57
MF 0.91 0.81 0.79 0.76 0.86 0.85 0.74 0.83 0.73 0.54 0.78
HS 0.77 20.04 0.56 0.25 0.81 0.86 0.77 0.49 0.02 0.06 0.46
HL 0.91 0.95 0.97 0.95 0.84 0.68 0.90 0.88 0.96 0.87 0.89
Average 0.75 0.64 0.81 0.54 0.83 0.67 0.75 0.68 0.55 0.55 0.68
Grayscale texture WC –––––––– –
MF 0.94 0.89 0.89 0.73 0.81 0.75 0.74 0.87 0.93 0.75 0.83
HS 0.88 0.94 0.84 0.81 0.88 0.85 0.83 0.44 0.92 0.79 0.82
HL 0.85 0.77 0.79 0.55 0.87 0.63 0.79 0.30 0.44 0.70 0.67
Average 0.89 0.87 0.84 0.70 0.85 0.74 0.79 0.54 0.77 0.75 0.77
Color texture WC 0.63 0.87 0.79 0.11 0.81 0.63 0.82 0.78 0.63 0.60 0.67
MF 0.79 0.80 0.78 0.45 0.71 0.73 0.59 0.60 0.33 0.22 0.60
HS 0.53 0.88 0.76 0.79 0.54 0.83 0.67 0.64 0.78 0.69 0.71
HL 0.77 0.83 0.28 0.80 0.79 0.56 0.84 0.06 0.78 0.70 0.64
Average 0.68 0.85 0.65 0.54 0.71 0.69 0.73 0.52 0.63 0.55 0.65
Shown are the correlation coefficients between rank orders of a single observer with the average rank orders of the nine other observ-
ers. WC, Warm–Cool; MF, Masculine–Feminine; HS, Hard–Soft; HL, Heavy–Light.
TABLE I. Intraobserver agreement.
Sample type Emotion scale
1 2 3 4 5 6 7 8 9 10 Average
Uniform color WC 0.73 0.83 0.78 0.74 0.81 0.82 0.76 0.78 0.79 0.79 0.78
MF 0.82 0.72 0.77 0.79 0.74 0.92 0.76 0.79 0.40 0.65 0.74
HS 0.86 0.67 0.42 0.32 0.66 0.66 0.33 0.77 0.86 0.56 0.61
HL 0.83 0.87 0.87 0.90 0.85 0.89 0.54 0.75 0.93 0.60 0.80
Average 0.81 0.77 0.71 0.69 0.77 0.82 0.60 0.78 0.74 0.65 0.73
Grayscale texture WC – – – – – – – – – –
MF 0.87 0.79 0.76 0.57 0.61 0.67 0.67 0.68 0.85 0.38 0.68
HS 0.73 0.82 0.71 0.78 0.72 0.76 0.80 0.68 0.87 0.58 0.74
HL 0.66 0.54 0.69 0.62 0.76 0.10 0.68 0.42 0.50 0.47 0.54
Average 0.75 0.71 0.72 0.66 0.70 0.51 0.72 0.59 0.74 0.47 0.66
Color texture WC 0.82 0.67 0.85 0.74 0.88 0.66 0.87 0.50 0.76 0.66 0.74
MF 0.75 0.58 0.73 0.76 0.38 0.63 0.29 0.43 0.78 0.61 0.60
HS 0.54 0.66 0.78 0.75 0.59 0.67 0.29 0.59 0.87 0.68 0.64
HL 0.76 0.74 0.67 0.84 0.76 0.70 0.52 0.37 0.68 0.35 0.64
Average 0.72 0.66 0.76 0.77 0.65 0.67 0.50 0.48 0.77 0.57 0.65
Shown are the correlation coefficients between rank orders of the first and second (after 1 week) measurement. WC, Warm–Cool; MF,
Masculine–Feminine; HS, Hard–Soft; HL, Heavy–Light.
Volume 36, Number 6, December 2011 431
seem satisfactory. It is impossible to compare this result
with other studies because previous color emotion studies
did not repeat experiments to assess the level of intra-ob-
server variability.
Interobserver Agreement. How well do observers agree
with each other? We calculate the rank correlation between
each observer (averaged rank from the first and second
measurement) and the average of all other observers. This
data is shown in Table II. From the data in Table II, we
note that the average interobserver correlation is r¼0.68
for the UC samples, r¼0.77 for the GT samples and r¼
0.65 for the CT samples, respectively. Apparently observ-
ers agree best on the GT samples. One salient result on the
UC samples is that subjects 2, 4, 9, and 10 have low corre-
lations with the group average on the hard–soft scale. This
is partly attributable to the positioning of the dark samples
along the scale. Further analysis shows that the standard
deviation in the subject responses shows a minimum at
L*þC*¼100 and a more than two-fold increase at
lower and higher values. Obviously, dark colors and satu-
rated colors lead to lower agreement among subjects. This
is found to apply to both the warm–cool and hard–soft
scale. We do not consider the four observers as outliers.
Their correlation coefficients calculated between the first
and second measurement (r¼0.67, 0.32, 0.86, and 0.56,
respectively) indicate that three of the four observers are
able to replicate their results fairly well.
Before discussing the results of adding texture to the
color samples, we first present the results of regression
analysis. This provides color emotion formulae with
which we can more easily explain the effects of texture.
Quantitative Analysis: Modeling
The goal of this section is to derive quantitative formu-
lae that describe the color and texture emotions as a func-
tion of the samples’ color and texture parameters. As a first
step, one-way ANOVA’s are performed to find out which
of the parameters are significantly connected to the emo-
tion scales. Using both the results of the one-way
ANOVAs and formulas derived in previous studies
as a
starting point, we search for the analytical functions giving
the highest amounts of variance explained on the color
emotion scales. This is done using our statistical software
that indicates the significance of each parameter in the non-
linear regression. The resulting functions are shown in
Table III. These functions predict the activity on the emo-
tion scales, based on the color parameters L*, C*, h,and/
or the texture parameters number of octaves, frequency,
persistence, and lacunarity. Before the functions for
the CTs are derived, we first recalculate the L*, C*, and h
values as obtained from averaging over each samples’ 100
3100 pixels. This is done because the blending procedure
used to create the CTs as sketched in Fig. 3 results in
somewhat darker samples compared with the UC samples.
The models are derived on group averaged scale values,
that is, averaged over 10 observers. A negative scale value
indicates a response toward the left word of the opposite
word-pair (e.g., warm on the warm–cool scale), a positive
value indicates a response toward the right word (e.g., cool
on the warm–cool scale). A value of zero, corresponding
to the scale center, indicates neutral response, that is, nei-
ther warm nor cool on the warm–cool example.
Table III reports the adjusted R
as a goodness-of-fit
measure for the regression functions. This measure cor-
rects R
(variance explained) for the number of free pa-
rameters in the regression models. The table shows that
for the UC samples, the functions based on the CIELAB
parameters L*, C*, and hgive rise to high values of
adjusted R
, with an average of 0.88. For the Grayscale
and CT samples, the average adjusted R
is 0.82 and
0.80, respectively.
All in all, the color and texture emotion functions pro-
vide a reasonably accurate description of the average ob-
server response on the emotion scales. UC samples are
best described, followed by GT and CT. In Figs. B1, B2,
and B3, we show visualizations of the samples used in
our experiments, ranked along the emotion scales as pre-
dicted from the functions in Table III.
In the following section, we return to our main research
question: what is the effect of texture on the color emotion
TABLE III. Color and texture emotion formulae and percentages of explained variance.
Function predicting
absolute scale values Adjusted R
adjusted R
Uniform color WC 20.59 þ0.017 L20.21 C
cos(h- 45) 0.90 0.88
MF 22.47 þ0.035 Lþ0.80 C
20.018 h20.000021 h
þ0.00000023 h
HS 210.26 þ7.35 L
þ0.053 C20.0019 C
þ0.000011 C
þ0.42 cos(h- 30) 0.82
HL 24.41 þ0.30 L
20.26 cos(h- 130) 0.98
Grayscale texture WC 0.82
MF 101.36 þ9.27 L
230.06 oct
26.06 freq
253.38 pers
225.15 lac
HS 116.12 þ6.10L
232.30 oct
213.13 freq
248.81 pers
229.33 lac
HL 42.67 þ0.064 L212.46 oct
211.35 freq
25.84 pers
217.23 lac
Color texture WC 20.80 þ0.015 L20.2 C
cos(h- 40) þ0.056 oct 0.84 0.80
MF 0.84 L
þ0.022 C20.017 hþ0.00000014 h
20.57 oct
20.70 freq
HS 586.33 2178.78 oct
284.20 freq
2106.83 pers
2213.89 lac
HL 0.33 L
þ0.020 C
22.57 oct
21.41 freq
þ0.015 lac
The adjusted R
measure accounts for the number of free parameters in the formulae. The functions predict the activity on the emotion
scales based on the CIELAB color parameters L*, C*, h, and/or the Perlin noise texture parameters (oct, number of octaves; freq, fre-
quency; pers, persistence; lac ¼lacunarity). WC, warm–cool; MF, masculine–feminine; HS, hard–soft; HL, heavy–light.
432 COLOR research and application
The Effects of Texture on Color Emotion. We already
noted that the intraobserver agreement for the UC samples
is higher than for the textured samples. At the same time,
the interobserver agreement is better for the grayscale sam-
ples (average R
¼0.77) than for the uniform samples
¼0.68) and the CT samples (R
¼0.65). This may be
due to the fact that the GT samples have no variations in hue
or chroma, and so the observers have to deal with less color
dimensions as they arrange the samples along the scales.
The analytical functions presented in Table III reflect
the dependencies on the samples’ color and texture
parameters. The color parameters L*, C*, and hplay an
important role in all functions for UC samples, with the
exception that C* does not appear in the function for the
heavy–light scale. With respect to the functions for
the CT samples, three things are noted. First, all color
parameters L*, C*, and happear in the warm–cool and
masculine–feminine scales. Second, only L* and C*
appear in the heavy–light scale, and third, no color pa-
rameters appear in the function for the hard–soft scale.
So, when texture is added to the UC samples, only the
hard–soft scale loses its dependency on color parameters.
In other words, hard–soft is fully dominated by texture.
Warm–cool, masculine–feminine, and heavy–light are
dominated by color parameters (in order of descending
dominance), adding the texture parameters explains for
another 2.9, 36.2, and 27.5%-point of the variance in the
data, respectively, as shown in Table IV. This Table
presents a comparison of model performances on the CT
samples. For example, when the function for warm–cool
derived on the UC samples is applied to the CT data,
already 82% of the data variance is explained. Adding
texture parameters to this function increases the model
performance by 2.9%-point. Likewise, for the hard–soft
scale, the model derived on the UC samples has no
explanatory power at all (R
¼0) on the CT samples and
adding texture parameters results in R
¼0.73. The last
column shows the ‘‘added value’’ of texture parameters,
calculated as the difference between the adjusted R
obtained with- and without texture parameters.
In conclusion, when texture is added to UC samples, color
emotions change. Responses along the hard–soft scale are
fully determined by texture, and in decreasing extent for the
masculine–feminine, heavy–light, and warm–cool scales.
The impact of this is that when textured samples are
involved in color emotion studies, texture cannot be ignored.
Comparison with Other Studies. We can evaluate the
performance of color emotion functions derived by others on
our experimental data, but only for the UC samples. Func-
tions for grayscale or CT samples have not been published
previously. For the scales warm–cool, hard–soft, and heavy–
light, we determine the adjusted R
for models derived in
studies by Xin and Cheng
and Ou et al.,
see Table V. The
results show that our experimental data for the heavy–light
scale (which strongly depend on lightness L*) is very well
described by all three models. For the warm–cool scale, the
model by Ou et al.
is reasonably good (R
¼0.70), but the
model by Xin and Cheng
completely fails. For the hard–
soft scale, both models by Ou et al.
and Xin and Cheng
fail. An explanation for this may be the different methods
used for obtaining the observer scores. In our experiments,
the subjects put the samples in relative order along the scale,
whereas the other investigators only record the preference
for one of the scale directions (for instance warm or cool). In
the latter case, a final scale value is obtained by performing
some sort of averaging over the scores of the observers, and
therefore many observers are necessary.
We demonstrate a systematic approach to the study of color
emotions and the effect thereupon of adding texture to the
color samples. A limited number of scales (four) are used,
because we are mainly interested in the specific effect of
adding texture, and not so much in factor analysis that
reveals how different scales may combine into new
descriptors. Nevertheless, we have gathered a valuable set
of experimental data using an improved method in which
subjects order the samples along the scale while maintain-
ing a view on all samples. Another methodological
improvement in comparison with other studies is that our
subjects repeat the experimental trials after 1 week, which
provides us with an estimate of the intraobserver agree-
ment. We derive analytical functions that predict the
group-averaged scale responses, with a precision exceeding
TABLE IV. Comparison of the performance (adjusted
) of color emotion models on our data for the
Color Texture samples, when using the Uniform
Color functions (derived on the Uniform Color
samples) or the Color Texture functions including
texture parameters.
Adjusted R
Added value
of texture
Warm–Cool 0.82 0.84 0.029
Masculine–Feminine 0.40 0.76 0.36
Hard–Soft 0 0.73 0.73
Heavy–Light 0.59 0.86 0.28
TABLE V. Performance (adjusted R
) of color
emotion models by different investigators on our
experimental data for the Uniform Color samples.
Adjusted R
Ou et al.
Xin and
Cheng (2000)
Warm–Cool 0.90 0.70 0.14
Hard–Soft 0.82 0.16
Heavy–Light 0.98 0.96 0.96
Excluding the five achromatic samples. When including these
samples (having C*¼0), there is no correlation between our data
and the model prediction by Ou et al. (2004).
Volume 36, Number 6, December 2011 433
that reported in other studies. Also, our functions outper-
form the functions derived by Xin and Cheng
and Ou
et al.
when applied to our data, which is probably
explained by the differences in methods. We note that the
adjusted R
measure is the preferred measure to report,
because that one corrects for the number of free parameters
in the functions.
Our subjects are from seven different nationalities.
Testing on cross-cultural effects, as done in other stud-
is not performed as that would require more
subjects. Neither do we test on possible gender differen-
ces. Again, our focus is on the effect of adding texture,
not on other issues. In the experimental design we adopt
the minimum number of observers (10) as discussed in
As long as the desired scale precision is
unknown it is impossible to make precise estimates on
the required number of observers. All that can be said is
that the use of more subjects leads to lower standard devi-
ations in the estimates. Scale accuracy increases with
about the square root of the number of observers. Other
studies have used more subjects (e.g., Ou et al.
used 31
observers, Gao and Xin
used 70 observers, Gao et al.,
used 50–70 observers per cultural group) but we prefer to
perform a repetition of the full experiment, which we
regard equally important. In this respect, an interesting
question is what the subjects’ long term repeatability on
the color and texture emotion scales is. That kind of in-
formation would greatly help to assess the validity and
applicability of the color and texture formulae derived
here. From our study it is clear, though, that whenever
textured samples are used, texture may play an important
role in color emotion.
As for future experiments, there are several routes to go. In
addition to the four scales we study, other scales used in color
emotion studies may be selected. Moreover, scales from other
studies (not necessarily color studies) may be adopted to
better capture the responses for texture classes. Although we
already enhance the complexity of our stimuli by adding
lightness textures to UCs, the textures that we use are still
rather primitive. Using chromatic textures, having chromatic
variations around the average, would be a logical next step.
When the chromatic distribution is mainly in one direction in
color space, discrimination thresholds for natural and synthe-
sized textures were found to be identical,
which allows us to
continue working with synthetic textures. It may be expected
that when using natural textures, certain color–texture combi-
nations will fit prototypical templates like green grass, and ini-
tiate strong associations. Recent findings from neuroimaging
studies suggest that the cerebral processing of form, texture,
and color may be independent.
Yet, these studies provide
no answer to the question how those object features interact
when subjects have to respond on emotion scales.
When texture is added to UC samples, color emotions
change. Texture fully determines the responses on the
hard–soft scale, and plays a role of decreasing weight for
the masculine–feminine, heavy–light, and warm-cool
scales. We conclude that when textured samples are used
in color emotion studies, the psychological responses may
be strongly affected by texture.
TABLE AI. CIELAB L*, C*, and h
specification of
the uniform color samples.
Sample L*C*h
Sample L*C*h
1 10 31 0 51 10 13 180
2 10 15.5 0 52 10 6.5 180
3 30 55 0 53 30 24 180
4 30 27.5 0 54 30 12 180
5 50 78 0 55 50 34 180
6 50 39 0 56 50 17 180
7 70 50 0 57 70 45 180
8 70 25 0 58 70 22.5 180
9 90 14 0 59 90 56 180
10 90 7 0 60 90 28 180
11 10 26 36 61 11 216
12 10 13 36 62 10 5.5 216
13 30 65 36 63 30 20 216
14 30 32.5 36 64 30 10 216
15 50 94 36 65 50 29 216
16 50 47 36 66 50 14.5 216
17 70 52 36 67 70 38 216
18 70 26 36 68 70 19 216
19 90 14 36 69 90 26 216
20 90 7 36 70 90 13 216
21 10 15 72 71 10 14 252
22 10 7.5 72 72 10 7 252
23 30 41 72 73 30 25 252
24 30 20.5 72 74 30 12.5 252
25 50 60 72 75 50 35 252
26 50 30 72 76 50 17.5 252
27 70 78 72 77 70 46 252
28 70 39 72 78 70 23 252
29 90 22 72 79 90 16 252
30 90 11 72 80 90 8 252
31 10 15 108 81 10 29 288
32 10 7.5 108 82 10 14.5 288
33 30 40 108 83 30 52 288
34 30 20 108 84 30 26 288
35 50 57 108 85 50 75 288
36 50 28.5 108 86 50 37.5 288
37 70 75 108 87 70 48 288
38 70 37.5 108 88 70 24 288
39 90 92 108 89 90 15 288
40 90 46 108 90 90 7.5 288
41 10 24 144 91 10 42 324
42 10 12 144 92 10 21 324
43 30 43 144 93 30 74 324
44 30 21.5 144 94 30 37 324
45 50 62 144 95 50 107 324
46 50 31 144 96 50 53.5 324
47 70 82 144 97 70 77 324
48 70 41 144 98 70 38.5 324
49 90 79 144 99 90 25 324
50 90 39.5 144 100 90 12.5 324
101 20 0
102 40 0
103 60 0
104 80 0
105 100 0
The first 100 samples are chromatic, and the last five samples
are achromatic.
434 COLOR research and application
Here, we show arrangements of the samples used in our
experiments. For each emotion scale (except for the
warm–cool scale for the GTs), the samples are ranked
on scale values as calculated using the functions given
in Table 3. The UC samples are displayed as vertical
bars to save some space. For the same reason, we left
out half the samples in the arrangements of the GTs and
the CTs. The arrangements are illustrative; accuracy of
color reproduction is limited.
FIG. B1. Arrangement of the uniform color samples used in the experiments, based on the scale values predicted from the func-
tions for uniform colors in Table III.
FIG. B2. Arrangement of the grayscale texture samples used in the experiments, based on the scale values predicted
from the functions for grayscale texture in Table III.
Volume 36, Number 6, December 2011 435
1. Ou L-C, Luo MR, Woodcock A, Wright A. A study of colour emo-
tion and colour preference. Part I. Colour emotions for single col-
ours. Col Res Appl 2004;29:232–240.
2. Gao XP, Xin JH. Investigation of human’s emotional responses on
colors. Col Res Appl 2006;31:411–417.
3. Gao XP, Xin JH, Sato T, Hansuebsai A, Scalzo M, Kajiwara K,
Guan S-S, Valldeperas J, Lis MJ, Billger M. Analysis of cross-cul-
tural color emotion. Col Res Appl 2007;32:223–229.
4. Suk H-J, Irtel H. Emotional response to color across media. Col Res
Appl 2010;35:64–77.
5. Adams FM, Osgood CE. A cross-cultural study of the affective
meanings of color. J Cross Cult Psychol 1973;4:135–156.
6. D’Andrade R, Egan M. The colors of emotion. Am Ethnol 1974;1:49–63.
7. Xin JH, Cheng KM, Taylor G, Sato T, Hansuebsai A. Cross-regional
comparison of colour emotions Part I: Quantitative analysis. Col Res
Appl 2004;29:451–457.
8. Clarke T, Costall A. The emotional connotations of color: A qualita-
tive investigation. Col Res Appl 2008;33:406–410.
9. Ekroll V, Faul F, Niederee R. The peculiar nature of simultaneous
colour contrast in uniform surrounds. Vision Res 2004;44:1765–1786.
10. Nascimento SM, Foster DH, Amano K. Psychophysical estimates of
the number of spectral-reflectance basis functions needed to repro-
duce natural scenes. J Opt Soc Am A 2005;22:1017–1022.
11. Hansen T, Giesel M, Gegenfurtner KR. Chromatic discrimination of
natural objects. J Vision 2008;8:1–19.
12. Tinker MA. Effect of stimulus-texture upon apparent warmth and
affective value of colors. Am J Psychol 1938;51:532–535.
13. Kim S-J, Kim EY, Jeong KJ, Kim J-I. Emotion-based textile indexing
using colors, texture and patterns. Proc ISVC 2006;LNCS4292:9–18.
14. Erhart L, Irtel H. Der Einfluss von Oberfla
¨cheneigenschaften auf die
emotionale Wirkung von Farbe. FarbInfo, Dresden; 2006.
15. Simmons DR, Russell CL. Visual texture affects the perceived
unpleasantness of colours. Perception37 ECVP Abstr Suppl 2008:146.
16. Perlin K. An image synthesizer. Comp Graph 1985;19:287–296.
17. Stokes M, Anderson M, Chandrasekar S, Motta R. A standard
default color space for the Internet–sRGB. Available at: http://; November, 1996.
18. Anderson BL, Kim J. Image statistics do not explain the perception
of gloss and lightness. J Vision 2009;9:10.1–17.
19. Kingdom FAA. Perceiving light versus material. Vision Res
20. Ou L-C, Luo MR, Woodcock A, Wright A. A study of colour emo-
tion and colour preference. Part III: Colour preference modeling. Col
Res Appl 2004:29:381–389.
21. Xin JH, Cheng K. Quantitative evaluation of colour emotion. Japan
Soc Promotion of Sci Fund Meet 2000:71–86.
22. Bailey JE, Neitz M, Tait D, Neitz J. Evaluation of an updated HRR
color vision test. Visual Neurosci 2004;22:431–436.
23. Engeldrum P. Psychometric Scaling: A Toolkit for Imaging Systems
Development, Winchester, Massachusetts, USA: Imcotek Press; 2001.
24. Giesel M, Hansen T, Gegenfurtner KR. The discrimination of chro-
matic textures, J Vision 2009;9:1–28.
25. Cant JS, Goodale MA. Attention to form or surface properties modulates
different regions of human occipitotemporal cortex. Cereb Cortex
26. Cavina-Pratesi C, Kentridge RW, Heywood CA, Milner AD. Separate
channels for processing form, texture, and color: Evidence from FMRI
adaptation and visual object agnosia. Cereb Cortex 2010;20:433–446.
FIG. B3. Arrangement of the color texture samples used in the experiments, based on the scale values predicted from the func-
tions for color texture in Table III.
436 COLOR research and application
... Empirical studies, focused on colour preferences and colour connotations, have demonstrated that pink is considered to be a feminine colour and blue a masculine colour [3,8,10,11]. Pink further represents groups of low social power and low social status [12][13][14]. ...
... The femininity bias of pink mirrors previous empirical findings [3,10,11,49,50]. In contrast, findings on blue did not confirm a masculinity bias in our corpus, unlike previous empirical studies would have suggested [1,3,10,11]. ...
... The femininity bias of pink mirrors previous empirical findings [3,10,11,49,50]. In contrast, findings on blue did not confirm a masculinity bias in our corpus, unlike previous empirical studies would have suggested [1,3,10,11]. Indeed, blue seems gender-neutral, equally liked by males and females of all ages [9,[51][52][53][54][55][56]. ...
Full-text available
In Western societies, the stereotype prevails that pink is for girls and blue is for boys . A third possible gendered colour is red . While liked by women, it represents power, stereotypically a masculine characteristic. Empirical studies confirmed such gendered connotations when testing colour-emotion associations or colour preferences in males and females. Furthermore, empirical studies demonstrated that pink is a positive colour, blue is mainly a positive colour, and red is both a positive and a negative colour. Here, we assessed if the same valence and gender connotations appear in widely available written texts (Wikipedia and newswire articles). Using a word embedding method (GloVe), we extracted gender and valence biases for blue , pink , and red , as well as for the remaining basic colour terms from a large English-language corpus containing six billion words. We found and confirmed that pink was biased towards femininity and positivity, and blue was biased towards positivity. We found no strong gender bias for blue , and no strong gender or valence biases for red . For the remaining colour terms, we only found that green , white , and brown were positively biased. Our finding on pink shows that writers of widely available English texts use this colour term to convey femininity. This gendered communication reinforces the notion that results from research studies find their analogue in real word phenomena. Other findings were either consistent or inconsistent with results from research studies. We argue that widely available written texts have biases on their own, because they have been filtered according to context, time, and what is appropriate to be reported.
... While motifs emphasize the games' themes, "colors are associated with certain qualities which enables them to evoke specific feelings of individuals" (Roohi and Forouzandeh, 2019). Adding textures, perceived color emotions vary (Lucassen et al., 2011). Lastly, light not only fulfills functional aspects, such as making the VE perceivable and drawing the players' attention to game-relevant objects, but also performs emotion-inducing tasks (El-Nasr, 2006;Knez and Niedenthal, 2008). ...
Full-text available
Virtual environments (VEs) can evoke and support emotions, as experienced when playing emotionally arousing games. We theoretically approach the design of fear and joy evoking VEs based on a literature review of empirical studies on virtual and real environments as well as video games’ reviews and content analyses. We define the design space and identify central design elements that evoke specific positive and negative emotions. Based on that, we derive and present guidelines for emotion-inducing VE design with respect to design themes, colors and textures, and lighting configurations. To validate our guidelines in two user studies, we 1) expose participants to 360° videos of VEs designed following the individual guidelines and 2) immerse them in a neutral, positive and negative emotion-inducing VEs combining all respective guidelines in Virtual Reality. The results support our theoretically derived guidelines by revealing significant differences in terms of fear and joy induction.
... Using fMRI studies, Echavarria et al. found that much rougher textures will evoke stronger activation of the human lateral occipital cortex, and smooth objects are more preferred over those with textured surfaces (Echavarria et al. 2016). Further research indicated that texture affects color emotions and that both texture and color strongly affect psychological responses to visual art (Lucassen et al. 2011). ...
Full-text available
The exploration of the potential relationship between computable low-level texture, such as features extracted from color and texture, and the perceived high-level aesthetic properties, such as warm or cold, soft or hard, has been a hot research topic of neuroaesthetics. First, the selection and clustering of aesthetic antonyms used to represent the aesthetic properties of visual texture are completed through two semantic differential experiments. Subsequently, 151 visual textures are rated according to the selected aesthetic antonyms by participants in a third semantic differential experiment. Third, 106 textural features are extracted using four different image analysis algorithms to describe the low-level characteristics of visual textures. Finally, the construction and evaluation of the visual aesthetic perception model based on multiple linear and nonlinear regression algorithms are discussed. We analyzed the frequency of each aesthetic antonym selected from 20 pairs of semantic antonyms, and the most frequently mentioned 8 pairs of semantic antonyms were selected as the core set for model building. The extracted low-level features are highly correlative. Of the correlation coefficients based on absolute values, 3383 are less than 0.75, accounting for 14.84% of the total. The correlation coefficients were larger than 0.5 accounts for 27.29% of the total. Through neighborhood component analysis, the top 10 low-level features are selected with lower correlation. The gap between low-level calculated features and high-level perceived aesthetic emotions can be bridged by a brain-inspired model of visual aesthetic perception.
... Textures may influence the affective meaning of color (Lucassen et al., 2011;Ebe and Umemuro, 2015). This has been demonstrated by pairing colors with computer-generated textures and asking participants to rate the color-texture pairs using four scales: Warm-Cool, Masculine-Feminine, Hard-Soft, and Heavy-Light. ...
Full-text available
A cluster of research in Affective Computing suggests that it is possible to infer some characteristics of users’ affective states by analyzing their electrophysiological activity in real-time. However, it is not clear how to use the information extracted from electrophysiological signals to create visual representations of the affective states of Virtual Reality (VR) users. Visualization of users’ affective states in VR can lead to biofeedback therapies for mental health care. Understanding how to visualize affective states in VR requires an interdisciplinary approach that integrates psychology, electrophysiology, and audio-visual design. Therefore, this review aims to integrate previous studies from these fields to understand how to develop virtual environments that can automatically create visual representations of users’ affective states. The manuscript addresses this challenge in four sections: First, theories related to emotion and affect are summarized. Second, evidence suggesting that visual and sound cues tend to be associated with affective states are discussed. Third, some of the available methods for assessing affect are described. The fourth and final section contains five practical considerations for the development of virtual reality environments for affect visualization.
... The emotional meaning of a color can change according to the texture (Lucassen et al., 2011). This has been demonstrated by pairing colors with computer-generated textures and asking participants to rate the color-texture pairs using four scales: Warm-Cool, Masculine-Feminine, Hard-Soft, and Heavy-Light. ...
Full-text available
A cluster of research in Human-Computer Interaction (HCI) suggests that it is possible to infer some characteristics of users' mental states by analyzing electrophysiological responses in real-time. However, it is not clear how to use the information extracted from electrophysiological signals to create visual representations of the emotional states of Virtual Reality (VR) users. Visualization of users' emotions in VR can lead to biofeedback therapies for training emotion self-regulation. Understanding how to visualize emotions in VR requires an interdisciplinary approach that integrates disciplines such as psychology, electrophysiology, and audiovisual design. Therefore, this review aims to integrate previous studies from these fields to understand how to develop virtual environments that can automatically create visual representations of users' emotional states. This manuscript addresses this challenge in three sections: First, theories related to emotion and affect are compared. Second, evidence suggesting that specific visual and sound cues tend to be associated with particular emotions are discussed. And third, some of the available methods for assessing emotions are described.
... Over at least the last 30 years a number of studies have reported evidence that there are universal wordcolour associations. Most studies show that there is more similarity between colour meanings than differences between cultural groups [20,34]. Nevertheless, when cultural differences do exist they can be significant and it is recommended that care needs to be taken before assuming that a word-colour association found in one cultural group applies to another group. ...
Full-text available
It is known that colours evoke memories and emotions and that these associations are valuable in design. However, it is difficult to ascertain universal associations because often the word-colour associations are culture-and context-specific. Nevertheless, some important universal colour associations have been shown to exist, at least if the variable of culture is removed. Laboratory studies, however, although effective struggle to collect sufficient information to allow the data to be useful in a practical setting. In this study, a crowd-sourcing internet experiment was carried out to find colours that are associated with different words and the results were compared with those from a previous laboratory study. Approximately 150 participants made a total of 1350 word-colour associations to generate palettes for 30 target words. Comparison with corresponding data from the laboratory study suggests that crowd-sourcing is a promising route and should be explored further.
... Their later study (Hewawalpita and Perera 2019) assessed whether or not and how multimodal interactions can be used to improve the consumer experience in e-commerce, including speech inputs, gesture inputs and interactive 3D product presentations. Lucassen et al. (2011) identified that the psychological responses of a product user may be strongly affected by the product's texture and color. They explored how to consolidate interactive design concepts so that the designer can establish a user style experience model in line with their demands. ...
Full-text available
This research investigates the possibility of incorporating affective factors in 3D model creation using handbags as an exemplary product. We demonstrate a crowd-based experimental approach that establishes the correlations between human emotional responses described by bi-polar adjectives and design factors of a bag. A factorial experiment is conducted to understand which design factors independently and/or interactively influence the responses by systematically varying 3D bag models. In-depth statistical analysis of the experimental results indicates that increasing the bag height makes it look more casual; a black bag tends to look luxurious and formal; a bag made of linen feels more classic and sophisticated. Those insights provide a basis for implementing a prototyping design tool that enables users to create 3D bag models with certain affective features using free-form deformation. The proposed approach demonstrates the preliminary feasibility of improving Kansei Engineering applications with big data obtained from crowdsourcing.
Both studies on color emotion and color harmony have been developed for many years. For designers, creating harmonious color combinations that satisfy specific color emotions has always been designers' pursuit. However, most of the studies used flat color combinations and lacked the discussion of the influence of three‐dimensional color configuration on color emotion and harmony. This study used the two‐color combination applied to the 3D color configuration to explore the relationship between color emotion and color harmony, intending to enhance designers' insights on the color schemes. The Multi‐dimensional Scaling (MDS) was used to obtain the perceptual map, giving an overview of the relationships among color emotion, color harmony, and color scheme techniques. The results showed that the emotion scales of “warm‐cool,” “soft‐hard,” and “heavy‐light” are similar. The scales of “complex‐simple,” “active‐passive,” and “warm‐cool” can be used to predict color harmony. The harmonious color combinations come along with “simple,” “passive,” and “warm” emotions.
Full-text available
Basic design is the origin of many design‐related fields and covers different concepts such as pattern, colour, texture and relief. Because colour is rarely seen as a uniformly plain element, current research embraces colour and texture as equally important variables. Investigations consider the association of colour and texture in terms of preference. One hundred individuals trained in design, as well as 96 individuals not trained in design, participated in an experiment in a virtual and abstract environment. We chose four primary colours from Natural Colour System colour space (yellow, red, green and blue) with the same saturation and brightness values. They were associated with three different texture types (plain, basket and broken twill), which we obtained by scanning the physical textile samples. In the second part of the experiment, we attempted to find a relationship between colour and one texture variable, in this case texture strength. The results of the study do not verify the existence of a dependency between colour and texture in either sample group. Also, there is no verification that a relationship between colour and texture strength exists. Furthermore, the results show that overall blue is the most preferred colour. Moreover, all participants preferred fine textures to coarse ones. Design‐trained participants often associate fine strength level with basket texture whereas the non‐design–trained participants more often associate it with plain textures. Lastly, the majority of participants state that blue is their most preferred colour in fine rather than coarse textures.
Full-text available
This study investigates the relationship between color perceptual attributes and color emotions, as well as the influence of different cultural backgrounds. Totally 214 color samples were evaluated on 12 emotion variables by subjects from seven different region groups in the psychophysical experiment. By factor analysis, it was found that three factors were sufficient to represent 80 “region-emotion” variables. For each variable, there is no distinct difference among different region groups. The 12 emotion variables could be divided into four categories, namely, activity index, potency index, definition index, and temperature index. Factor scores were further calculated to study the determinants on each factor. The analysis showed that the three factors were mainly related to chroma, lightness, and hue, respectively. It was concluded that chroma and lightness were the most important factors on color emotion, whereas the influences of hue and cultural background were very limited. © 2007 Wiley Periodicals, Inc. Col Res Appl, 32, 223–229, 2007
Full-text available
Typically, the research on the connotations of colors has presented isolated color samples to the participants, and/or restricted their responses to a limited set of alternatives (e.g., basic emotions). In the present study, the participants were asked to imagine their own examples of the target colors, thereby allowing the participants to select their own salient examples, including the context. In addition, a qualitative method, the semi-structured interview, was employed, giving them freedom to choose their own terms of description, and also identify issues which they themselves regarded as relevant. To a large extent, the results were consistent with previous research about the emotional significance and arousing effects of different colors. But this alternative approach also threw light upon more subtle connotations of colors, the multiple meanings of a single color, and also the participants' reasons for these connotations, ranging from individual experiences to cultural conventions and stereotypes. © 2008 Wiley Periodicals, Inc. Col Res Appl, 33, 406–410, 2008
Full-text available
In this study three colour preference models for single colours were developed. The first model was developed on the basis of the colour emotions, clean–dirty, tense–relaxed, and heavy–light. In this model colour preference was found affected most by the emotional feeling “clean.” The second model was developed on the basis of the three colour-emotion factors identified in Part I, colour activity, colour weight, and colour heat. By combining this model with the colour-science-based formulae of these three factors, which have been developed in Part I, one can predict colour preference of a test colour from its colour-appearance attributes. The third colour preference model was directly developed from colour-appearance attributes. In this model colour preference is determined by the colour difference between a test colour and the reference colour (L*, a*, b*) = (50, −8, 30). The above approaches to modeling single-colour preference were also adopted in modeling colour preference for colour combinations. The results show that it was difficult to predict colour-combination preference by colour emotions only. This study also clarifies the relationship between colour preference and colour harmony. The results show that although colour preference is strongly correlated with colour harmony, there are still colours of which the two scales disagree with each other. © 2004 Wiley Periodicals, Inc. Col Res Appl, 29, 381–389, 2004; Published online in Wiley InterScience ( DOI 10.1002/col.20047
Full-text available
Colour emotion is a feeling or emotion induced in our brains when we look at a colour. In this article, the colour emotional responses obtained by conducting visual experiments in different regions, namely Hong Kong, Japan and Thailand, using a set of 218 colour samples are compared using a quantitative approach in an attempt to study the influence of different cultural and geographical locations. Twelve pairs of colour emotions described in opponent words were used. These word pairs are warm–cool, light–dark, deep–pale, heavy–light, vivid–sombre, gaudy–plain, striking–subdued, dynamic–passive, distinct–vague, transparent–turbid, soft–hard, and strong–weak. These word pairs represent the fundamental emotional response of human beings toward colour. The influences of lightness and chroma were found to be much more important than that of the hue on the colour emotions studied. Good correlations of colour emotions among these three regions in East Asia were found, with the best ones for colour emotion pairs being light–dark and heavy–light. © 2004 Wiley Periodicals, Inc. Col Res Appl, 29, 451–457, 2004; Published online in Wiley InterScience ( DOI 10.1002/col.20062
Conference Paper
At the Gottingen meeting of the International Colour Vision Society, I reported oil a comparison of the second edition of the American Optical Hardy-Rand-Rittler Pseudoisochromatic plates (AO HRR) with the Richmond Products third edition of the same test and concluded that the chromaticities were exceptionally poorly matched and that the new edition was a "pale imitation of the real thing" (unpublished). This conclusion led to our abandoning a clinical trial. In 2002. Richmond Products has published a fourth edition and, in 2003, Waggoner has published a modified HRR with additional (Ishihara style) plates and the tetartan confusion figures removed. As a precursor to any clinical trial, the colors used in the plates have been measured and comparisons drawn between the four editions. While the two most recent editions much more closely resemble the original AO HRR and the chromaticities are much better aligned on the dichromatic confusion lines, the excitation purities (and therefore the degree of difficulty) of the plates are less well matched in the Richmond Products editions. In addition, there is a significant degree of metamerism in the third edition and Waggoner edition that makes variations in illuminant more critical to performance.
Color data from the Osgood et al. 23-culture semantic differential study of affective meanings reveal cross-cultural similarities in feelings about colors. The concept RED is affectively quite salient. BLACK and GREY are bad, and WHITE, BLUE, and GREEN are good. YELLOW, WHITE, and GREY are weak; RED and BLACK are strong. BLACK and GREY are passive; RED is active. The color component Brightness, as determined by comparing data on WHITE, GREY, and BLACK, is strongly associated with positive Evaluation, but also with negative Potency. Eighty-nine previous studies of color and affect were analyzed. They generally support these findings, and, together with the fact that there are very few exceptions in our data or the literature, lead one to believe that there are strong universal trends in the attribution of affect in the color domain.
This article investigates human's emotional responses on colors based on a psychophysical experiment. Totally 218 color samples were evaluated by 70 subjects based on 12 basic descriptive variables including “warm–cool,” “weak–strong,” and “dynamic–passive.” By using factor analysis, these 12 variables were split into two orthogonal factors (activity index and potency index) and one correlative factor (definition index), which may be used for description of color emotion. Based on these three indexes, a color emotion map in CIELCH color space was obtained by cluster analysis. A well-regulated distribution of human's emotion was observed in CIE L*C* plane with a neutral feeling region around the point of chroma C*=30.5 and lightness L*=53.3. Colors scattered at opposite direction of this neutral region possess the opposite feelings. The detailed relationship between color emotion indexes and color perception attributes, i.e., hue, lightness, and chroma, were further studied by correlation analysis and graphical representation. The results indicated that the activity index was dependent on chroma, the potency index was dependent on lightness, and the definition index was dependent on both chroma and lightness. It was also observed that the influence of hue on emotional response was not as significant as those in previous studies even for the variable “warm–cool.” © 2006 Wiley Periodicals, Inc. Col Res Appl, 31, 411–417, 2006; Published online in Wiley InterScience ( DOI 10.1002/col.20246
Research on synesthesia has consistently found an association between colors and emotions. In order to try to determine whether the basis for this phenomenon is culture specific or universal, a test to determine color and emotion associations was administered to a sample of Tzeltal-speaking adults from highland Chiapas and U.S. college undergraduates. The results of the color-emotion test indicate that within the limits of translation equivalence color chips and emotion terms show very similar patterns of association in both cultures. These results seem to indicate that the phenomenon of synesthesia is mediated by the universal factors uncovered by Osgood's semantic differential test. The results also indicate that saturation is a major factor in color-emotion synesthesia.