Color and Personality - 1 -

Color and Personality: Strongs Interest be driven by physiological arousal. Indeed,

Inventory and Cattells 16PF Stones (2003) findings indicate that the color
red increased individuals levels of arousal,
Rense Lange, Sigmund Testing, Dallas, TX 1 which when paired with a stimulating task,
Jason Rentfrow, University of Cambridge, UK caused deficits in cognitive performance.

Introduction More importantly, the preceding raises the pos-

sibility that the effects of color on performance
Color is an important aspect of our efforts to can have differential effects, depending on ones
create personal spaces to our own liking. More- preferences or aversions for particular colors.
over, color choices can have important social For example, Eysenck (1967, 1970) postulated
consequences as our choices are part of our that Introverted individuals are high in internal
presentation to others, and thus these choices arousal (i.e., they are preoccupied with their
may influence how others perceive us. Yet, little thoughts and feelings more than are Extraverts),
is known about why people like or dislike the and therefore prefer social environments (e.g.,
colors they do. This paper asks whether peoples where they are alone) that allow them to reduce
color preferences reflect meaningful information or maintain their optimum level of arousal.
about their personalities, interpersonal styles, Thus, when Introverts are with other people their
and behaviors. Surprisingly, relatively little re- level of arousal might rise to the point that they
search has been done to investigate the links feel uncomfortable and overwhelmed. The pre-
between such variables and individuals color ceding work on the effects of color on arousal
preferences. The research summarized here therefore suggests that color preferences and
represents our efforts to identify links between personality might be related. Specifically, indi-
peoples color preferences and their personal viduals high in internal arousal (i.e., Introverts)
characteristics as derived from two well- might prefer calm colors like blue to reduce
established psychological inventories. their level of arousal, whereas individuals low in
internal arousal (i.e., Extraverts) might prefer
Background exciting colors like red to increase their level
of internal arousal Lscher (1971).
At the most basic level, color has been shown to
affect our mood, thereby affecting the way we Perhaps the most prominent theorist arguing that
interact with our environment. A growing body color preferences and personality are linked is
of research in environmental psychology has Lscher (1971) who proposed that individuals
shown that the color of a room or work setting with similar color preferences should also pos-
can have profound effects on individual enjoy- sess similar personality characteristics. Accord-
ment and performance on a variety of tasks. For ing to Lscher, the physiological reactions that
instance, Stone (2003) showed that task per- individuals experience while viewing primary
formance varied as a function of the color of the colors (blue, red, yellow, and green) reflect basic
room in which the task was performed. In an- psychological needs of the individuals. When a
other study by Stone (2001), positive mood primary color is not liked, for example, this dis-
tended to be higher when individuals worked in like is considered to reflect a deficit or unmet
a blue carrel compared to a red carrel. Perform- physiological and psychological need. For in-
ance is also affected because individuals read stance, if an individual has a particularly strong
slower and comprehended less when performing dislike for the color red, this is believed to re-
a reading task in a red environment. This study flect unconscious anxiety within that individual.
thus provides direct evidence that color has an
effect on cognitive ability, suggesting that the Whereas Lscher (1971) regarded color prefer-
cognitive impairments produced by color could ences as a reflection of the unconscious drives
within individuals, contemporary perspectives
1
For additional information concerning this research on the color-personality relationship view color
contact the first author at: [email protected] preferences as a reflection of conscious (i.e.,

Energia White Paper # DCS - 23

 Color and Personality - 2 -

reportable) motives, drives, and values. For in- http://www.deweycolorsystem.com.

stance, French and Alexander (1972) found that
individuals preferring the color blue were calmer Secondly, Lschers own research, as well as
and preferring yellow was related to "positive" later research inspired by his work, relied exclu-
feelings (e.g., happiness). However, the hy- sively on standard statistical tests and correla-
pothesis that red reflects "negative" feelings tions. Thus, hypotheses were tested mainly
(e.g., tension) was not supported. Additionally, based on piece-meal analyses of counts and
Seefeldt (1979) investigated sex differences in average preferences, which were then analyzed
color preference, and found that yellow was according to the analytical options provided by
preferred more highly by women than men. Yet, standard statistical methodology which treats
Stimpson and Stimpson (1979) found no sex each choice as the basic unit of analysis. By
differences in color preferences, nor did they contrast, we hypothesize that the identification
observe a relationship between color preferences of global patterns of color preferences is far
and personality. Finally, Picco and Dzindolet more informative than a series of piece-meal
(1994) failed to show that color preferences are analyses. In particular, individual choices should
related to self-descriptions, even when control- be combined into patterns that simultaneously
ling for social desirability (e.g., participants encompass respondents likes and dislikes, while
favoring green and blue were not more intro- not giving undue weight to the evaluation of a
verted than those favoring red or yellow). Taken particular color sample. Thus a contextual ap-
together then, support for Lschers theories proach is needed that identifies and captures
about the correlates of color preference is mixed specific color combinations, while ignoring ir-
at best. relevant variation in isolated choices. It is pro-
posed that such an approach is provided by the
The present research use of neural nets (cf., Galant, 1994), and details
concerning the present use of such nets are pro-
The following addresses topics similar to the vided in later sections.
above, however, with two major differences.
From a practical perspective, the present re-
Firstly, a different color test is used that derives search promises to assess personality and occu-
from the patented Dewey Color System (Sadka, pational qualifications on the basis of tests that
2000), which is currently in commercial use (cf., have no perceived face validity to the respon-
Sadka, 2004). This Dewey Color System Test dents. Thus, it will not be readily apparent to
uses distinctive spectrum divided color hues. In respondents how their color preferences relate to
particular, starting with yellow, blue and red, their personality characteristics or occupational
they are mixed to visually create shades of proclivities. Accordingly, issues of socially de-
green, purple and orange with no visible charac- sirable responding should be less of a concern
teristics of the primary shades. Likewise, the than in standard paper-and-pencil tests, thereby
intermediates are fabricated from a primary and decreasing response bias.
a secondary color. Black, white, and brown are
also added. Although the Dewey Color System Study I: Strongs Interest Inventory
Test covers preferences among primary, secon-
dary, achromatic colors, and intermediate colors, The latest version of the Strong Vocational In-
the major task to be performed in this test is the terest Blanks called the Strong Interest Inven-
ordering of 15 colors according to respondents tory (Harmon, Hansen, Borgen, and Hammer,
preferences. In fact, it is the preference order of 1994) is a questionnaire consisting of 317
these fifteen choices that constitutes the basic questions which inquire about respondents in-
predictor variable in this study. These colors can terest in a wide rang of items associated with
be described as teal, purple, brown, red-orange, occupations, occupational activities, hobbies,
yellow, magenta, orange, white, lime green, leisure activities, school subjects, and preferred
blue, gold, green, black, indigo, and red. The types of people. While the Strong Interest Inven-
entire test can be seen in color at the website tory provides a variety of other information as

Energia White Paper # DCS - 23

 Color and Personality - 3 -

well, this research focused on respondents gen- (.NET) component.

eral orientation to work as reflected by six sepa-
rate factors, dubbed Basic Interest Scales (BIS). Results
In particular, a distinction is made (Harmon, et
al., 1994, pp. 70-78) among Realistic BIS (in- The six panels of Figure 1 are scatter plots of the
cluding agriculture, nature, military, athletics, relation between the actual but squashed
and mechanical activities), Investigative BIS Strong BIS (X-axis) and the squashed BIS val-
(science, mathematics, and medical science), ues as predicted by the neural net (Y-axis) from
Artistic (music/ dramatics, art, applied arts, writ- the color test inputs. Visual inspection reveals
ing, and culinary arts), Social BIS (teaching, that prediction is quite successful for each of the
social and medical service, and religious activi- six BIS. As is shown in Table 1, the finding of
ties), Enterprising (public speaking, law/politics, sizeable correlations between the actual and
merchandising, sales, and organizational man- predicted squashed BIS values supports this
agement), and Conventional BIS (data manage- interpretation. For instance, the median correla-
ment, computer activities, and office services). tion across all BIS is 0.68, and the lowest corre-
lation, i.e., for Social BIS, is 0.54. Further, given
Procedure. A commercial testing center headed the very nature of the color test, it is at least
by a former Human Resource Director at a lead- suggestive to note that the highest correlation (r
ing aerospace company was charged to adminis- = 0.73) obtains for the Artistic BIS.
ter the Strong Interest Inventory to volunteers
who participated in a series of Career Transition We conclude therefore that respondents color
Clinics provided by an Atlanta, GA church. As preferences as assessed by the Dewey Color
part of a battery of occupational tests (including System Test are indeed powerful predictors of
Cattells 16PF as used in Study II), four profes- all Basic Interest Scales of the Strong Interest
sional psychologists administered Strongs In- Inventory. These findings thus open the exciting
terest Inventory together with a paper-and-pencil possibility that peoples vocational interests can
version of the Dewey Color System Test. These be inferred quite accurately from their color
psychologists processed and evaluated the an- preferences.
swers, and respondents returned after two weeks
to learn about the results. The resulting sample Study II: The 16PF
of 885 potential career changers consisted of 524
women and 359 men, and 2 persons with un- The Sixteen Personality Factor ques-
known gender. The respondents average age tionnaire (16PF) is a 185-item instrument that
was 32.1 years (Median = 29.0, SD = 12.9 years) comprises 16 personality primary factor scales
with missing age information for 28 individuals. originally identified by Raymond Cattell (cf.,
Catell, Eber, & Tatsuoka, 1970). Historically,
Analyses. Although other methods were consid- letters only identified these factors, but it has
ered as well, the results reported here derive become customary to add descriptive labels also
exclusively from analyses based on standard (Russel & Karol, 2002). The complete list of
neural nets using backpropagation based on factor designation is: Factor A (Warmth), Factor
standard logistic squashing, momentum, and B (Reasoning), Factor C (Emotional Stability),
varying learning rates (cf. Gallant, 1994). The Factor E (Dominance), Factor F (Liveliness),
software used was a Delphi implementation of Factor G (Rule-Consciousness), Factor H (So-
that described in Lange (1996). Throughout, one cial Boldness), Factor I (Sensitivity), Factor L
intermediate layer was used. To facilitate de- (Vigilance), Factor M (Abstractedness), Factor
ployment of the resulting neural net, this soft- N (Privateness), Factor O (Apprehension), Fac-
ware also produced a standalone Delphi program tor Q1 (Openness to Change), Factor Q2 (Self-
that provides an efficient implementation of the Reliance), Factor Q3 (Perfectionism), and Factor
forward pass of the neural net for use as a web Q4 (Tension).

Energia White Paper # DCS - 23

 Color and Personality - 4 -

Figure 1: Actual vs. Predicted Values for All Six Strong BIS (Study I)

-0.9 -0.5 -0.1 0.3 0.7

0.7

0.3

-0.1

-0.5
Enterprising BIS Conventional BIS
Predicted BIS (squashed)

-0.9
0.7

0.3

-0.1

-0.5
Artistic BIS Social BIS
-0.9
0.7

0.3

-0.1

-0.5
Investigative BIS
Realistic BIS
-0.9
-0.9 -0.5 -0.1 0.3 0.7
Actual BIS (squashed)

Energia White Paper # DCS - 23

 Color and Personality - 5 -

Table 1: Pearson Correlations Between Actual and Predicted BIS Values (Study I)

Pearson Cor-
Strong Basic Interest Scale (BIS) relation
Realistic 0.70
Investigative 0.64
Artistic 0.73
Social 0.54
Enterprising 0.66
Conventional 0.70

Note: All correlations shown are statistically significant at p < .001.

From these sixteen factors, as well as from par-

ticular combinations of low or high factor scores, It will be clear that respondents color preferences
it is possible to derive other personal characteris- provide powerful predictors of their personalities.
tics as well. However, as such combinations can The highest correlation is obtained for Factor H
be always derived once the sixteen primary fac- (Social Boldness), and the magnitude of this cor-
tors are known, this topic is not further pursued relation (r = 0.68) is comparable to the highest
here. value obtained in Study I for the Strong Interest
Inventory. Also, the median correlation in this
Procedure. The 16PF was administered in the study (0.51) is similar to that found earlier. How-
same context as the Strong and details are there- ever, the correlation of the worst performing vari-
fore not repeated. A subset of 1010 volunteers, able Factor O (Apprehension) with r = 0.33
which partially overlapped with those taking the falls well below the worst performing BIS (i.e.,
Strong Interest Inventory in Study I, completed Social) of the Strong Interest Inventory.
both the 16PF and the Dewey Color System Test.
This sample consisted of 602 women and 403 Nevertheless, given the results for the Strong In-
men, and 5 individuals with unknown gender. The terest Inventory, the findings for the 16PF clearly
average age was 32.7 years (Median = 30.0, SD = reinforce the notion that color preferences as as-
13.6 years) with missing age information con- sessed by the Dewey Color System Test are in-
cerning 27 individuals. deed valid indicators of personality.

Results

Analogous to the procedure followed for the

Strong, respondents scores on each of the sixteen
primary factors were transformed to z-scores, and
then squashed using a logistic function. These
sixteen quantities were then predicted from re-
spondents color choices on the Dewey Color
System Test. The sixteen panels of Figure 2 sum-
marize the observed relations between respon-
dents actual (squashed) factor scores (X-axis) and
the squashed predicted values produced by the
neural net (Y-axis). (Note: In this figure the points
are offset along the X-axis by random amounts to
better display their local frequencies). The
correlations between the actual and predicted
values of these factors are listed in Table 2.

Energia White Paper # DCS - 23

 Color and Personality - 6 -

Figure 2: Actual vs. Predicted Values for All Sixteen Primary Factor Scales (Study II)

Energia White Paper # DCS - 23

 Color and Personality - 7 -

Figure 2: Continued

Note: The letters in the sub-panels of this figure refer to the 16 factors listed in Table 2.

Energia White Paper # DCS - 23

 Color and Personality - 8 -

Table 2: Pearson Correlations Between Actual employed in previous research, or whether the
and Predicted Primary Scale Values (Study II) particular test and analysis combination used here
is responsible. Also, while it is true that neural
Pearson nets form a holistic oriented approach in which
16 PF Primary Factor Scale Correlation choices are combined into patterns that simultane-
Factor A (Warmth) 0.57 ously encompass respondents likes and dislikes,
Factor B (Reasoning) 0.50 neural nets have the disadvantage of being a black
Factor C (Emotional Stability) 0.59 box only, i.e., they must be used as is. In par-
ticular, neural nets weights are difficult to inter-
Factor E (Dominance) 0.51
pret directly (Gallant, 1994), and even knowing
Factor F (Liveliness) 0.65
all its weights provides little guidance in identify-
Factor G (Rule-Consciousness) 0.52 ing which patterns of color preferences are associ-
Factor H (Social Boldness) 0.68 ated with which particular personality traits.
Factor I (Sensitivity) 0.50
Factor L (Vigilance) 0.57 It should be kept in mind that we have essentially
Factor M (Abstractedness) 0.63 bridged the relation between two types of tests,
Factor N (Privateness) 0.50 i.e., color preferences and the personality related
Factor O (Apprehension) 0.33 variables assessed by the Strong and the 16PF.
Factor Q1 (Openness to Change) 0.49 However, what we have not yet established is that
Factor Q2 (Self-Reliance) 0.63 the Dewey Color System Test also predicts the
Factor Q3 (Perfectionism) 0.49 behaviors for which these personality tests are
Factor Q4 (Tension). 0.47 typically used. Thus, more extensive validation
should consider using color preferences directly to
Note: All correlations shown are statistically predict variables such as job satisfaction, leader-
significant at p < .001 ship potential, etc. Given the promising findings,
and given the advantages of having non-verbal
tests, we believe that the present research justifies
Summary and Discussion the cost and effort involved in conducting such
studies.
As was anticipated, the present findings clearly
indicate that peoples color preferences, as as- Finally, the present findings may have implica-
sessed via the Dewey Color System Test, indeed tions beyond standard personality assessment. For
provide meaningful information about their per- instance, observers can (and do) infer characteris-
sonalities, interpersonal styles, and behaviors. In tics of people based on such cues as physical ap-
particular, the data indicated that this test pre- pearance, clothing, nonverbal behavior, facial
dicted with considerable precision all six of the features, appearance of bedrooms and offices
Basic Interest Scales of Strongs Interest Inven- (Gosling, Ko, Mannarelli, & Morris, 2002), web-
tory (Harmon, et al., 1994), and nearly all of the sites (Vazire & Gosling, 2004), and music prefer-
sixteen Primary Factors of Cattells 16PF (Russel ences (Rentfrow & Gosling, 2003). Thus, to the
& Karol, 2002). As such, the present findings extent that peoples theories about others use of
show far greater consistency than do those ob- color play similar roles, the study of color prefer-
tained in earlier research inspired by Lschers ences could be expanded to areas such as advertis-
(1971) work (cf., Stone, 2001, 2003; French & ing, web-design, on-line dating, and interpersonal
Alexander, 1972; Seefeldt, 1979; Stimpson & perception in general.
Stimpson, 1979; Picco & Dzindolet, 1994).
A number of issues remain, however. For in-
stance, it is not clear at this point whether our
findings must be attributed to the differences be-
tween the colors contained in the Dewey Color
System Test and Lschers color test, whether
neural nets simply provide a superior method of
analysis relative to the simpler statistical methods

Energia White Paper # DCS - 23

 Color and Personality - 9 -

References ity and underlying assumptions. Perceptual &

Motor Skills, 79, 1555-1558.
Catell, R.B., Eber, H.W., & Tatsuoka, M.M. Rentfrow, P. J. & Gosling, S. D. (2003). The do-
(1970). Handbook for the 16PF. Champaign, re-mis of everyday life: The structure and
IL: Institute for Personality and Ability Test- personality correlates of music preferences.
ing, Inc. Journal of Personality and Social Psychology,
Eysenck, H. J. (1967). The Biological Basis of 84, 1236-1256.
Personality. Springfield: C. C. Thomas. Russel, M., and Karol, D. (2002). 16PF, Fifth
Eysenck, H. J. (1970). The Structure of Human Edition with Updated Norms. Institute for
Personality (3rd Edition). London: Methuen. Personality and Ability Testing, Inc.: Cham-
French, C. A. & Alexander, A. B. (1972). The paign, IL.
Lscher Color Test: An investigation of valid- Sadka, D.G. (2000). Patent Number 6,863,534
ity and underlying assumptions. Journal of Color Preference Self-Help System
Personality Assessment, 36, 361-365. Sadka, D. G. (2004). The Dewey Color System
Gallant, S.I. (1994). Neural network learning. Choose Your Colors, Change Your Life. Three
Cambridge, Mass.: MIT Press. Rivers Press/Crown Publishing, a Division of
Gosling, S. D., Ko, S. J., Mannarelli, T., & Random House; 1st Rev edition.
Morris, M. E. (2002). A room with a cue: Seefeldt, F. M. (1979). Cross-cultural validity,
Judgements of personality based on offices reliability, and stimulus characteristics of the
and bedrooms. Journal of Personality and So- Lscher Color Test. Perceptual & Motor
cial Psychology, 82, 379-398. Skills, 48, 896-898.
Harmon, L.W., Hansen, J-I. C., Borgen, F.H., and Stimpson, D. V. & Stimpson, M. F. (1979). The
Hammer, A.L. (1994). Strong Interest Inven- Lscher Color Test as a measure of anxiety in
tory: Application and Technical Guide. Stan- juvenile delinquents. Perceptual & Motor
ford, CA: Stanford University Press. Skills, 49, 60-62.
Lange, R. (1996). An empirical test of the Stone, N. J. (2001). Designing effective study
weighted effect approach to generalized pre- environments. Journal of Environmental Psy-
diction using recursive neural nets. In: E. Si- chology, 21, 179-190.
moudis, J. Han, and U. Fayyad Proceedings of Stone, N. J. (2003). Environmental view and color
the Second International Conference on for a simulated telemarketing task. Journal of
Knowledge Discovery and Data Mining. Environmental Psychology, 23, 63-78.
Menlo Park: AAAI Press. Pp. 183-188. Vazire, S. & Gosling, S. D. (2004). e-Perceptions:
Lscher, M. & Scott, I. (1971). The Lscher Color Personality Impressions Based on Personal
Test. New York: Pocket Books. Websites. Journal of Personality and Social
Picco, R. D. & Dzindolet, M. T. (1994). The Psychology, 87, 123-132.
Lscher Color Test: An investigation of valid-

Energia White Paper # DCS - 23