Developmental Science 4:4 (2001), pp 389± 398

REPORT
Preferences for colours and colour ± emotion combinations in
early childhood
Marcel R. Zentner
University of Geneva, Switzerland

Abstract
Previous research has shown that, by the age of 3 to 4 years, children rely not only on perceptual similarity but also on
shared category or other underlying structures to draw analogies. The present study extends this work by showing that
children as young as 3 years old detect consistent relationships between colours and facial expressions of emotions ± two
phenomena that share no physical characteristics, violate conventional categories and have no obvious environmental
contiguity. Two explanatory hypotheses are put forward: (a) learning by convention, which is explored against the
standard of adults' and older children's matching patterns, and (b) reliance upon a common underlying but perceptually
unavailable dimension ± operationalized in terms of emotion and colour preferences in the context of the present study.
Both explanatory approaches are discussed and avenues for future work are suggested.

While the basic purpose of human colour vision is to may already exist at an early age. First, research on
discriminate objects, at a more elaborate level this sensory conceptual development in young children has shown
capacity is used to attribute salience and meaning to that by the age of 3±4 years children rely not only on
chromatic stimuli. A prime example is that individuals perceptual similarity but also on shared category or other
not only show specific colour preferences but also underlying structures to draw analogies (e.g. Brown,
attribute emotional characteristics to colours in consistent 1989; Markman, 1989; Goswami, 1992). This capacity to
ways from school age on (e.g. Guilford & Smith, go beyond the mere appearance of the objects, relating
1959; Karp & Karp, 1988; Whitfield & Wiltshire, 1990; them on some deeper grounds, might be a conceptual
Boyatzis & Varghese, 1994; Valdez & Mehrabian, 1994; prerequisite to appreciating more abstract relationships,
Meerum Terwogt & Hoeksma, 1995). However, with the such as relating `bright' to `happy'.
exception of studies on colour discrimination in infancy Further evidence comes from research on metaphor
(Bornstein, 1975; Werner & Wooten, 1979; Adams, 1987), comprehension in young children. One could consider
surprisingly little is known about the early development analogies such as `anger is like red' and `sadness is like
of perceptual attractiveness and emotional connotations blue' to be metaphors since they concur with the
of colours. The present research extends previous work defining feature of a metaphor as a `similarity between
back into early childhood. Such an extension raises objects and events that violate children's conventional
important issues of both theoretical and practical interest. categories' (Vosniadou, 1987, p. 873). Although it is
The most important question to be addressed here is generally acknowledged that preschool children have
whether young children can detect a relationship between some ability to comprehend metaphors (Gardner &
a colour and an emotional expression ± two phenomena Winner, 1986; Vosniadou, 1987), it has also been found
that share no physical characteristics, violate conven- that this ability is fragile, especially in relation to
tional categories, and have no obvious environmental metaphors describing psychological phenomena such
contiguity. Different research traditions in cognitive as emotions (Winner, Rosenstiel & Gardner, 1976).
development have provided evidence suggesting that the More recent work, on the other hand, has suggested that
prerequisites for the construction of these relationships even preschool children do have some basic intuitive

Address for correspondence: Department of Psychology, University of Geneva, 40 Bd. du Pont d'Arve, CH-1205 Geneva, Switzerland; e-mail:
[email protected]

# Blackwell Publishers Ltd. 2001, 108 Cowley Road, Oxford OX4 1JF, UK and 350 Main Street, Malden, MA 02148, USA.
390 Marcel R. Zentner

understanding of metaphors involving emotions (Wag- emotions. This, in turn, would presuppose contiguities
goner & Palermo, 1989; Broderick, 1991). of emotions and colours in the environment of the
Finally, it has also been shown that preschool children young child. Because adults largely provide this
are quite proficient in mapping emotions onto inanimate environment, comparing patterns in early childhood
stimuli such as museum art (Callaghan, 1997) or music with the patterns manifested by adults could shed light
(see Zentner, 1999, for a review) in strikingly consistent on the role of learning. Consequently, an adult sample
ways. For years, the analogies children draw between was also included for the purpose of comparison.
psychological characteristics and inanimate stimuli were Beyond these theoretical issues, there is also an applied
recognized (e.g. Werner & Kaplan, 1967) but often interest to young children's affective responses to colours.
interpreted as accidental errors of categorization or as a For example, prevailing gender stereotypes such as `pink
result of imagination or animistic thinking character- is for girls, blue is for boys' are frequently used by
istics of the preoperational child (Piaget, 1962; Chu- advertisers and parents (e.g. Picariello, Greenberg &
kovsky, 1968). However, although animism and magical Pillemer, 1990). In addition, clinicians claim to possess
thinking might facilitate the extraction of emotional knowledge concerning the emotional significance of
information from inanimate stimuli, these concepts do colours for young children but little is known about
not explain the consistency of the results. how young children themselves feel about colour. Finally,
While a systematic explanation of this phenomenon is experimental work with toddlers and preschoolers
beyond the aims of this report, the study of how children regularly involves coloured stimuli (e.g. toys, animals)
map colours onto emotions may provide a valuable without asking what impact different colours might have
research strategy to identify what children attend to on young children's attention and behaviour.
when relating emotions to perceptual stimuli. One In summary the aims of this study were threefold: first,
possibility, raised by a number of authors, is that from and most importantly, to investigate whether 3- to 4-year-
early childhood individuals organize their perceptual old children show reliable colour preferences and map
input around some basic dimensions, such as suggested colours onto emotions in non-random ways; second, to
by Osgood, Saci and Tannen Gaum (1957). More compare how these early patterns are related to the
specifically, the perception of an event would include results found with adults and older children; finally, to
processing both on a modality-specific dimension (such explore the role of colour and emotion valence (pre-
as having a particular colour or a particular emotion) ference) as a possible mediator between the two
and on a more basic amodal dimension (such as being phenomena. In designing this research, two methodolo-
more or less pleasant or intense). In this view, some gical issues were of particular concern. First, colours can
translation of modality-specific input into an amodal be defined in terms of hue, brightness and saturation.
code is taking place and this, in turn, provides the key However, many studies in the psychology of colour have
for understanding why analogies between perceptually failed to provide adequate specification of colour samples
and=or categorically dissimilar phenomena can some- yielding results that are difficult to interpret (see Valdez &
times be drawn with a striking degree of consistency ± Mehrabian, 1994, for a critique). The present colour
even by very young children (e.g. Gardner, 1974; samples were therefore specified in terms of a standar-
Gardner & Winner, 1986; Marks, Hammeal & Born- dized system of notation (Munsell). Second, a procedure
stein, 1987; Smith & Sera, 1992; Collier, 1996). had to be devised that would not overtax the verbal and
Colours are an interesting variable for the investigation attentional capacities of 3-year-old children. Accordingly,
of this hypothesis because they can be described using a subjects' colour preference ratings were obtained by
series of abstract dimensions that are similar to the basic asking them to choose their preferred one from an array
dimensions used to describe emotions (Osgood et al., of nine coloured cardboard rectangles. In the colour±
1957; Russell, 1980). Concurrent with this explanatory emotion mapping task, children were instructed to match
framework, Meerum Terwogt and Hoeksma (1995) six coloured cardboard rectangles to three cartoon-like
recently suggested and found partial support for the drawings of faces designed to express emotions of
hypothesis that colours and emotions may be related to happiness, sadness and anger.
each another due to the preference (valence) given to each
of them within their own domain. Thus, in order to
further examine the role of valence as a possible mediator Method
between colours and emotions in young children, colour
Participants
preference ratings were added to our design.
Another possibility is that even young children draw 127 children, predominantly of Caucasian middle-class
on cultural conventions when mapping colours onto parents, were recruited from daycare centers and pre-

# Blackwell Publishers Ltd. 2001

 Preferences for colours and colour ± emotion combinations in early childhood 391

schools in a major Swiss urban area. Of these 127

children, 106 (50 males, 56 females) successfully com-
pleted the colour preference task, and 103 (53 males, 50
females) showed successful performance on the colour±
emotion matching task (see below for details). The mean
age was 45 months (age range 30±58 months).
In addition, we included a sample of 65 under-
graduates (31 males, 34 females). The mean age was 23.4
years (age range 18 ± 39 years). The participants were
primarily Caucasian and middle class.

Figure 1 Stimuli used in the practice matching task (targets in

Materials row above; standards in row below).
Materials used in the colour preference task consisted of
nine 35 cm  50 cm cardboard rectangles coloured in
red, yellow, dark blue, bright blue, dark green, bright
green, pink, brown and black. We chose fully saturated
colours that varied on hue and value (see Table 1 for the
precise colour notation).

Practice matching task

Figure 2 Stimuli used in the colour ± emotion matching test.
The materials for the practice matching task consisted of Emotional expressions of sadness, happiness and anger.
four stimuli (standards) which were designed to corre-
spond to a second set of four stimuli (targets). The four
target pictures were a bird, a shoe, a baby and a Santa (1975) and because they are easily understood even by
Claus, each painted on a 33 cm  50 cm cardboard very young children (e.g. Stein & Levine, 1987; Harris,
rectangle. The standard pictures consisted of a tree, 1989). The six standards to be matched with these target
a sock, a milk bottle and a gift, presented on pictures were 16 cm  16 cm cardboard rectangles, co-
16 cm  16 cm cardboard rectangles (see Figure 1). loured in red, yellow, bright green, dark blue, brown and
black (see Table 1). Considering our young age group,
the number of colours was reduced from nine to six for
Colour± emotion matching task
this task, yielding a 3 (emotion)  6 (colours) array.
In analogy to the practice task, the materials for the
colour±emotion matching task consisted of six colour
Mother questionnaire
stimuli (standards) and three targets displaying three
schematic faces, expressive of happiness, sadness and To obtain demographic information and the mother's
anger, which were painted on 33 cm  50 cm cardboard perception of their child's preferred colour, a ques-
rectangles (see Figure 2). These emotions were chosen tionnaire was given to each mother. The relevant
because they are basic as defined by Ekman and Friesen question asked: `Does your child have a preferred
colour? If yes, which one?'
Table 1 Colour samples specified according to the Munsell
system of colour notation (Munsell, 1988) Procedure
Colour name Hue Value Chroma
Colour seriation task
Red 7.5 R 4 14
Yellow 10 Y 8.5 12 Children were tested individually in a quiet room in their
Green (light) 10 GY 6 12 daycare center under daylight conditions. Before the
Green (dark) 2.5 G 3 8 child entered the room, the nine colours were laid out in
Blue (light) 10 B 7 8
Blue (dark) 8 PB 2 10 a half-circle, about 1 m away from the child, with the
Pink 7.5 PR 6 10 position of colours in the half-circle randomized across
Brown 10 R 3 10 subjects. The experimental session began by drawing the
Black N2 ± ±
child's attention to the nine colours and by asking the

# Blackwell Publishers Ltd. 2001

392 Marcel R. Zentner

child to pick the colour he=she liked most and to bring it Adults
back to the experimenter. The same procedure was then
Adults were tested in a group session. The same nine
repeated for the eight remaining colours until the last,
coloured cardboards used for the children were mounted
and least preferred one, was left. Thus, each time
on a wall. On a response sheet, subjects were asked to
the child returned the colour to the experimenter, a
indicate their order of preference and to then match
preference rank for that colour was recorded by an
the colours red, yellow, green, blue, brown and black
assistant of the experimenter who was unobtrusively
with the emotions of happiness, sadness and anger. To
seated in a corner of the testing room. Children who
control for response bias, order of presentation of
picked the colours in the preset order were not included
colours and of emotions was inverted for half of the
in the analyses (N = 21).
subjects.

Practice matching task

Results
After completion of the colour seriation task, the child
was introduced to a practice matching task. The child
was shown the four target pictures and the correspond- Colour preferences
ing standards and told that it was his=her task to find
out which one would go with which (see Figure 1). Agreement with mothers
Boxes were placed under each of the four targets, such
Maternal reports of their child's preferred colour agreed
that the items to be matched with the targets could be
with the child's first or second choice 71% of the time
placed by the child into the corresponding box. A
( = 0.67), providing evidence for the validity of our
response was recorded whenever the child put a
measurement.
standard into the box of one of the target alternatives.
Successful performance on this practice task was
expected to be a prerequisite for understanding the Colour ranking
colour ± emotion matching task. Therefore, children who
matched less than three out of four items correctly were Mean ranks of each colour are given in Table 2. The
not tested on the subsequent task.1 existence of overall colour preferences was revealed by
a highly significant Friedman analysis of variance,
 2 (8, N = 106) = 65.48, p < 0.001. The result was signifi-
Colour ±emotion matching task cant for both males ( 2 (8, N = 57) = 18.53, p = 0.018)
and females ( 2 (8, N = 49) = 56.60, p < 0.001). As deter-
After a short break, the children were shown the three mined by a multivariate repeated measures analysis of
faces displaying a happy, a sad and an angry expression variance (MANOVA) with sex as the between-subjects
and were asked to identify the expression of the `fellows' factor and colour as the repeated measure, there was
(`bonhommes') in the pictures. Out of the 127 children, no sex  colour preference interaction, F(8, 97) = 1.54,
only six (4.7%) needed help with identifying the
expressions. These children were prompted with ques-
tions such as `Does one of them seem happy? Show me Table 2 Mean ranks of colour preferences in children and
which one'. The children were then shown the six colour adults (1 corresponds to most preferred, 9 to least preferred)
stimuli which were positioned approximately 50 cm in
Mean Mean
front of the child with order of presentation randomized Children rank (SD) Adults rank (SD)
across subjects. Then, the child was asked to choose a
colour for each of the three `fellows', a colour that 1. Red 3.73 (2.33) Dark blue 2.89 * * (2.08)
2. Pink 4.23 * * (2.68) Bright blue 3.71 * * (2.36)
`would go well with' the character in question, and to 3. Dark blue 4.53 * * (2.57) Red 3.88 (1.87)
put the colour into the box of the specific `fellow'.2 4. Yellow 4.82 (2.45) Dark green 4.74 (2.48)
5. Bright green 5.00{ (2.52) Yellow 5.28 (2.38)
1
6. Bright blue 5.24 * * (2.61) Black 5.54 (2.62)
The following routes could lead to three correct answers: (1) the 7. Dark green 5.36 (2.40) Bright green 5.72{ (2.53)
match of the sock with Santa Claus was counted as correct; (2) some 8. Brown 5.95{ (2.46) Pink 6.46 * * (1.34)
children put more than one item in the same box; (3) some children 9. Black 6.02 (2.40) Brown 6.65{ (1.83)
stopped after three correct mappings.
2 Notes: Colours in bold held significantly different orders of preference in the two
After the colour was put in a box it was not available for future
samples.
selections. A small number of children put a second colour into the * * Reliable between-group differences (df = 169), p < 0.001.
same box. In these cases the first choice was counted. { Marginally significant between-group differences, p < 0.10.

# Blackwell Publishers Ltd. 2001

 Preferences for colours and colour ± emotion combinations in early childhood 393

p = 0.154 (Wilk's  = 0.89). Although this indicates that Table 3 Patterns of colour ± emotion mappings in children
boys and girls did not differ significantly in their colour Red Yellow Blue Brown Green Black
preferences, there was a marginal difference in the
strength of the patterns, with girls being slightly more Happiness
Males 21 11 8 3 6 4
sensitive (z = 1.35, p = 0.088). Females 9 13 8 6 6 8
This was particularly clear in examining girls' and Total 30 24 16 9 12 12
boys' reactions to brightness. This was done by creating
Sadness
an overall dark versus bright preference score which was Males 9 3 16 12 8 5
obtained by adding, for each child, the ranks for dark Females 10 6 14 5 4 11
(dark blue, brown, dark green, black) and bright (pink, Total 19 9 30 17 12 16
red, bright green, bright blue, yellow) colours. If the Anger
overall dark score was lower than the overall bright Males 8 12 10 9 8 6
score, a dark preference was noted, and vice versa (since Females 11 7 13 5 5 9
Total 19 19 23 14 13 15
lower ranks denote greater preference). The majority of
children (68.1%) manifested a bright preference, but this
was relatively more true for girls (77.8%) than for boys
(59.2%),  2 (1, N = 106) = 3.73, p = 0.053. Happy=sad
First, one could expect that children tend to match
Comparison with adults bright colours with a happy emotional expression and
dark colours with a sad emotional expression. This
Overall differences in colour preference between children prediction was confirmed by comparing the number of
and adults were analysed using a MANOVA with age as children choosing a bright colour (yellow, red, green) for
the between-subjects factor and colour as the repeated the happy expression and a dark colour (blue, brown or
measure. This analysis showed a highly significant black) for the sad expression with the number of
age  colour preference interaction, F(8, 162) = 8.62, children choosing the opposite pattern, namely a dark
p < 0.001 (Wilk's  = 0.70), indicating that the profile of colour for the happy and a bright colour for the sad
colour preferences differs reliably between 3-year-old expression: 48=21, p < 0.001, by the binomial theorem.
children and adults. The descriptive statistics listing the Similarly, the number of children choosing either red or
mean ranks for each colour in children and adults is yellow for the happy expression and either blue or
shown in Table 2. brown for the sad expression was three times the number
of children choosing the opposite pattern: 30/10,
Colour ±emotion association p < 0.001.

Practice matching task Sex differences

65.4% of the children in our sample matched all four Whereas blue and yellow had similar emotional con-
items correctly, and 80.4% met our criterion of notations for boys and girls, the associations for red and
matching three out of four items correctly. Accordingly, brown were gender-specific. Boys matched red more
the remaining 19.6% (N = 25) of the children were not than twice as often with the happy than with the sad face
tested on the colour ± emotion matching task. (21=9, p = 0.021), whereas girls did not do so (9=10, ns).
Similarly, boys matched brown four times as often with
the sad than with the happy face (12=3, p = 0.018),
Colour± emotion matching task whereas girls' matching patterns for the colour brown
The descriptive statistics for the colour ±emotion asso- were random (5=6, ns). The colours green and black had
ciation are given in Table 3.  2 computed across these no particular emotional connotation for either girls or
values was significant at (10, N = 103) = 17.66, boys.
p = 0.061. This value, however, has to be treated with
caution given that the assumption of independence of
Age differences
observations is not met. Moreover, as an `omnibus test'
it is not very informative. Thus, more focused analyses Due to the relatively wide age range in our sample, we
in which the assumption of independence was met were were able to split the sample into a younger age group
carried out. (mean age 3.3 years, N = 53) and an older age group

# Blackwell Publishers Ltd. 2001

394 Marcel R. Zentner

(mean age 4.1 years, N = 50). The ratio of `predicted' p = 0.97, ns, indicating no relation between colour and
(bright ± happy=dark ±sad) to `unpredicted' (bright ± emotion preference.
sad=dark± happy) matches was above chance in both
age groups (younger 24=12, p = 0.03; older 24=9,
p = 0.007), suggesting no substantial difference in the Discussion
ability to detect similarity between colour and emotion
between 3 and 4 years of age. In general, our results suggest that chromatic stimuli
carry different degrees of perceptual attractiveness
and different emotional meanings for 3- to 4-year-old
Anger
children. We can therefore anticipate that even very
Overall, the results for anger were weak and non- young children have the capacity to detect relationships
significant. Only yellow was matched significantly more between the perceptually unrelated phenomena of
often with the angry than with the sad facial expression, colours and emotions. This result adds to the literature
12=3, p = 0.018. in several ways. First, and compared to related studies,
this study included an unusually young age group, an
unusually large sample size, and colour sample specifi-
Comparison with adults
cations in terms of a standardized system. Second, the
Adults' colour ±emotion matches are in several ways data extended research that has shown that, by the age
different from 3-year-olds' associations (see Table 4). of 4 years, children can make sense of metaphors
For example, red is rarely matched with happiness and involving emotions and can map emotions onto paint-
becomes the colour of anger. In addition, sadness is ings and music (see introduction). Finally, the method
primarily associated with black (rather than blue). devised underscores the feasibility of empirical work on
Similar in both samples was the choice of yellow for affective responses to perceptual stimuli even with very
happiness and the non-emotionality of the colour green. young children. The following discussion will highlight
the most important findings from both tasks, compare
them to the results from previous studies, and discuss
Relation of colour preference to emotion preference
some avenues for future research.
Although emotion preferences were not measured
directly in this study, the literature on emotion
Colour preferences
preference in preschoolers suggests that happiness is
preferred over both sadness and anger (Russell & Red was the preferred colour for both girls and boys.
Bullock, 1985). Thus, the respective emotions were This result is consistent with infant colour preferences
dummy coded as 0, 1, 1 (lower values denoting higher (Bornstein, 1975; Adams, 1987) but contradicts the well-
preference). For each subject, the three emotions coded established preference for blue in adults (see Eysenck,
as such were then paired with the rank of the specific 1941; Valdez & Mehrabian, 1994), a preference that was
colour the three emotions had been associated with. The replicated with our own adult sample. Two recent
resulting point biserial correlation was rpb ˆ 0:001, studies suggested that a preference for blue over red
might already be established by 7 years of age (Boyatzis
& Varghese, 1994; Meerum Terwogt & Hoeksma, 1995).
Table 4 Patterns of colour ± emotion mappings in adults Thus, an important developmental transition in the
Red Yellow Blue Brown Green Black perceptual attractiveness of colours seems to occur
during the preschool and early elementary school years.
Happiness The developmental decrease in preference for red might
Males 1 16 10 0 1 0
Females 4 20 7 0 1 0 be due to the fact that red is a more ambivalent colour
Total 5 36 17 0 2 0 than blue across cultures (Adams & Osgood, 1973),
representing such unpleasant things as blood, danger or
Sadness
Males 0 0 1 6 1 20 corrections (see Safuanova & Safuanov, 1992). Appar-
Females 0 0 5 5 2 20 ently, these connotations are not present in early
Total 0 0 6 11 3 40 childhood but are gradually acquired through increased
Anger social contact and schooling.
Males 22 1 0 2 0 3 No evidence was found for widespread gender
Females 22 0 0 2 0 8 stereotypes such as `pink is for girls and blue is for
Total 44 1 0 4 0 11
boys' (see Picariello et al., 1990; Pomerleau, Bolduc,

# Blackwell Publishers Ltd. 2001

 Preferences for colours and colour ± emotion combinations in early childhood 395

Malcuit & Cossette, 1990). However, although both that yellow was mapped onto the happy and angry facial
genders preferred bright to dark colours, the overall expressions is consistent with the finding that, for adults,
brightness preference was relatively stronger for girls the hue spectrum of green-yellow is associated with
than boys. This finding is consistent with previous emotional responses of arousal and dominance (Valdez
research which suggested general similarities in male and & Mehrabian, 1994). Furthermore, the link of yellow
female preferences for various colours, while noting with happiness has been repeatedly found in work with
slight sex differences in the strengths and consistencies adults (e.g. Wexner, 1954; Collier, 1996) and with
of those preferences (Norman & Scott, 1952; Whitfield children (e.g. Lawler & Lawler, 1965; Karp & Karp,
& Wiltshire, 1990; Boyatzis & Varghese, 1994; Valdez & 1988). The link of blue with sadness is so pervasive that
Mehrabian, 1994; Hemphill, 1996). Our findings suggest it has entered the English language and numerous
that this gender bias towards greater chromatic sensi- studies with both adults and children report blue to be
tivity in females is already present in early childhood. matched with sadness (e.g. Karp & Karp, 1988; Collier,
The data on colour preferences in young children is 1996). The sadness connotation of blue might be related
potentially relevant to an old issue ± the relation to the fact that the hue spectrum of purple-blue (similar
between the linguistic evolution of colour terms and to the dark blue used in the present study, see Table 1) is
the organization of colour perception in early childhood. perceived as low in arousal (Valdez & Mehrabian,
In their well-known study concerning the universal 1994).4
acquisition of colour terms, Berlin and Kay (1969) In contrast to the developmental consistencies in the
argued that colour terms evolved in the following fixed emotional connotations of blue and yellow, differences
order (achromatic colours omitted): (1) red, (2) green ± emerge for black and red. There is evidence that by 10
yellow, (3) blue, (4) brown, (5) pink ± orange ± purple. If years of age children categorize black as sad and red as
the perceptual attractiveness of colours played a role in angry (Karp & Karp, 1988) ± connotations that appear
learning, e.g. by making certain colours more salient or to be absent in early childhood. The developmental shift
memorable, then the same perceptual attractiveness for black and red lends further support to the argument
might have influenced the linguistic evolution of colour that certain meanings of colours, perhaps particularly
terms. If this was true we should expect a correspon- negative ones, are gradually acquired throughout the
dence between the evolutionary order of colour terms preschool and elementary school years. Some degree of
and early emerging colour preferences. In line with past caution, however, is needed in interpreting these
research (see Bornstein, 1985, for a review), however, the comparative results. First, past research often provided
present results were inconclusive in this regard.3 insufficient stimulus descriptions (e.g. such as `blue' and
`green') and so only approximative comparisons are
possible. Second, the present stimulus set was limited.
Colour ±emotion mappings Future research would need to clarify to what extent the
The association of happy with bright colours and sad present findings are resistant to slight variations in hue
with dark colours is established by 3 years of age. More and brightness.
specifically, whereas yellow was most often matched Having demonstrated that young children can detect
with the happy and, to a minor extent, angry facial consistent relationships between colours and facial
expressions, blue was most consistently associated with expressions of emotions, how children relate these two
the sad facial expression. Furthermore, red and brown phenomena still needs to be explained. Clearly, children's
had emotional connotations for boys, but not for girls. performance can be attributed neither to perceptual nor
The results for green and black were inconclusive. to categorical similarity. As anticipated in the introduc-
In comparing these findings with the results of the tion, there are at least two non-mutually exclusive
adult sample and with data from previous research, the explanations that can now be elaborated on the basis of
similarities are stronger for yellow, dark blue and brown the present findings. One possibility is that children draw
than for red and black. For example, the current finding on cultural conventions when projecting colours onto
emotions. A certain difficulty with this explanation is that
3
As can be seen in Table 2, there is only partial correspondence it presupposes a contiguity between colours and expres-
between the two. However, as indicated by Rosch (1972) long ago, a sions of emotions in the environment of the very young
poor or inconsistent match between the linguistic evolution of colour
terms as described by Berlin and Kay and the early organization of
4
colour perception can indicate different things ± an independence of Interestingly, the present adult sample chose black rather than blue
the two phenomena, limitations in the developmental work ± but it for sadness. In contrast to past studies, however, both blue and black
could also reflect shortcomings of the classic anthropological work. At were presented simultaneously. Given a choice of black and blue,
present, not enough is known to resolve this question. adults from this particular cultural background might favour black.

# Blackwell Publishers Ltd. 2001

396 Marcel R. Zentner

child for which there is no direct evidence at present. Of Colours offer two additional possibilities: saturation
course, some indirect evidence comes from adults who (chroma) and brightness (value). The link between
attribute emotions to colours in highly consistent ways. emotion and brightness, which this study demon-
However, if children simply adapted the preferential strated, clearly encourages extension of research on
and associative patterns from adults, then their patterns brightness. In addition, to the extent that variations in
should conform to those of adults. The partial lack of saturation are perceived as variations in intensity or
resemblances found in this study undermines this strength, they would provide an obvious conceptual
interpretation and suggests a role for alternative learning bridge between a colour and an emotion. Thus,
mechanisms. Keeping the robust association of happy systematically varying the colour components of
with bright colours and sad with dark colours in mind, it brightness and saturation while examining how these
may be that young children develop this association due variations map onto different emotional expressions is
to the way happy and sad characters are coloured in an important future avenue for discovering what
children's books, to colour±emotion metaphors nested in young children attend to when overriding the category
everyday language or to personal experiences with light boundaries of colours and emotions.
and darkness (Williams, Boswell & Best, 1975). More
studies are needed to examine the potential impact of
these learning mechanisms.5 Acknowledgements
However strong the impact of learning, what remains
remarkable from a cognitive point of view is the ease I am grateful to Kathrin Lygren for her help in
with which even 3-year-old children are able to over- developing the stimuli used in this research and in
come the perceptual and conceptual gap between a testing during the initial stages of the research. The
coloured cardboard rectangle and a facial expression of contributions to testing from Stacy Woolston, Rachel
an emotion. Although the relation between this ability Laff and Samantha Rukert are also gratefully ac-
and other cognitive skills emerging at this age remains knowledged. Thanks is extended to Tara Callaghan,
to be explored, research has shown that an ability to Philipp Rochat, Ellen Winner and two anonymous
categorize objects relying on underlying structure reviewers for their insightful comments on an earlier
rather than on perceptual similarity develops rapidly draft of this manuscript. Tanja Wranik-Odenahl
between 3 and 4 years of age (e.g. Goswami, 1992). helped with editing. Above all, I would like to thank
Interestingly, however, the underlying structure in the all of the daycare centers and their children for their
present study is of a more abstract nature than in most cooperation and patience.
standard classification tasks. Thus, the possibility was
raised at the beginning that when children classify
certain colours with certain emotions they might refer
to a perceptually unavailable dimension or to an References
`amodal code'. Building upon Meerum Terwogt's
study, and drawing upon the dimensional approach Adams, F., & Osgood, C.E. (1973). A cross-cultural study of
the affective meaning of colour. Journal of Cross-Cultural
within emotion research, we explored the role of
Psychology, 4, 135± 156.
valence as one possible amodal dimension underlying Adams, R.J. (1987). An evaluation of colour preference in
young children's projections of colours onto emotions. early infancy. Infant Behaviour and Development, 10, 143±
The fact that little support was found for this particular 150.
assumption may not come as a surprise to emotion Berlin, B., & Kay, P. (1969). Basic color terms: Their
theorists who have argued in favour of treating universality and evolution. Berkeley, CA: University of
preferences and emotions as separate entities (Scherer, California Press.
2000). In this view the former might simply reflect an Bornstein, M.H. (1975). Qualities of colour vision in infancy.
aesthetic judgement that has little to do with emotional Journal of Experimental Child Psychology, 19, 401± 419.
content. Bornstein, M.H. (1985). On the development of color naming
However, valence is not the only dimension to in young children: data and theory. Brain and Language, 26,
72 ± 93.
be investigated within this theoretical framework.
Boyatzis, C.J., & Varghese, R. (1994). Children's emotional
associations with colours. Journal of Genetic Psychology,
5
Work on `metaphorical mapping' in infants (see Wagner, Winner, 155, 77 ± 85.
Cicchetti & Gardner, 1981; Philipps, Wagner, Fells & Lynch, 1990) Broderick, V. (1991). Young children's comprehension of
provides useful suggestions on how the reported phenomena might be similarities underlying metaphor. Journal of Psycholinguistic
investigated with still younger children. Research, 20, 65 ±81.

# Blackwell Publishers Ltd. 2001

 Preferences for colours and colour ± emotion combinations in early childhood 397

Brown, A.L. (1989). Analogical learning and transfer: what Do infants recognize emotion in facial expression?: cat-
develops? In S. Vosniadou & A. Ortony (Eds), Similarity and egorical and `metaphorical' evidence. Infant Behavior and
analogical reasoning (pp. 369± 412). Cambridge: Cambridge Development, 13, 71 ± 84.
University Press. Piaget, J. (1962). Play, dreams and imitation in childhood. New
Callaghan, T.C. (1997). Children's judgments of emotions York: Norton.
portrayed in museum art. British Journal of Developmental Picariello, M.L., Greenberg, D.N., & Pillemer, D.B. (1990).
Psychology, 15, 515± 529. Children's sex-related stereotyping of colors. Child Develop-
Chukovsky, K. (1968). From two to five. Berkeley, CA: ment, 61, 1453± 1460.
University of California Press. Pomerleau, A., Bolduc, D., Malcuit, G., & Cossette, L. (1990).
Collier, G.L. (1996). Affective synesthesia: extracting emotion Pink or blue: environmental gender stereotypes in the first
space from simple perceptual stimuli. Motivation and two years of life. Sex Roles, 22, 359± 367.
Emotion, 20, 1 ± 32. Rosch, E. (1972). Universals in color naming and memory.
Ekman, P., & Friesen, W.V. (1975). Unmasking the face: A Journal of Experimental Child Psychology, 93, 1 ± 10.
guide to recognizing emotions from facial clues. Englewood Russell, J.A. (1980). A circumplex model of affect. Journal of
Cliffs, NJ: Prentice-Hall. Personality and Social Psychology, 39, 1161± 1178.
Eysenck, H.J. (1941). A critical and experimental study of Russell, J.A., & Bullock, M. (1985). Multidimensional scaling
colour preference. American Journal of Psychology, 45, 385± of emotional facial expressions: similarity from preschoolers
394. to adults. Journal of Personality and Social Psychology, 48,
Gardner, H. (1974). Metaphors and modalities. How children 1290± 1298.
project polar adjectives onto diverse domains. Child Devel- Safuanova, O., & Safuanov, F. (1992). Connotative meanings
opment, 45, 84 ± 91. of colors and color names. In L. Dorfman (Ed.), Emotions
Gardner, H., & Winner, E. (1986). Attitudes and attributes: and art. Perm: Perm Institute for Arts and Culture.
children's understanding of metaphor and sarcasm. In M. Scherer, K.R. (2000). Psychological models of emotion. In J.
Perlmutter (Ed.), Perspectives on intellectual development: Borod (Ed.), The neuropsychology of emotion (pp. 137± 162).
The Minnesota Symposia on Child Psychology (pp. 131± Oxford: Oxford University Press.
152). Hillsdale, NJ: Erlbaum. Smith, L.B., & Sera, M.D. (1992). A developmental analysis of
Goswami, U. (1992). Analogical reasoning in children. Hills- the polar structure of dimensions. Cognitive Psychology, 24,
dale, NJ: Erlbaum. 99 ± 142.
Guilford, J.P., & Smith, P.C. (1959). A system of colour Stein, N.S., & Levine, L.J. (1987). Thinking about feelings.
preferences. American Journal of Psychology, 72, 487± 502. The development and organization of emotional knowledge.
Harris, P. (1989). Children and emotion. Oxford: Blackwell. In R. Snow & M. Farr (Eds), Aptitude, learning, and
Hemphill, M. (1996). A note on adults' colour± emotion instruction, Vol. 3: Conative and affective process analyses
associations. Journal of Genetic Psychology, 157, 275± 280. (pp. 165± 197). Hillsdale, NJ: Erlbaum.
Karp, E.M., & Karp, H.B. (1988). Colour associations of male Valdez, P., & Mehrabian, A. (1994). Effects of colour on
and female fourth-grade school children. Journal of Psy- emotions. Journal of Experimental Psychology: General, 123,
chology, 122, 383± 388. 394± 409.
Lawler, C.O., & Lawler, E.E. (1965). Color ± mood associa- Vosniadou, S. (1987). Children and metaphors. Child Devel-
tions in young children. Journal of Genetic Psychology, 107, opment, 58, 870± 885.
29 ± 32. Waggoner, J.E., & Palermo, D.S. (1989). Children's compre-
Markman, E. (1989). Categorization and naming in children. hension of emotion-descriptive metaphors. Developmental
Problems of induction. Cambridge, MA: MIT Press. Psychology, 25, 152± 163.
Marks, L.E., Hammeal, R.J., & Bornstein, M.H. (1987). Wagner, S., Winner, E., Cicchetti, D., & Gardner, H. (1981).
Perceiving similarity and comprehending metaphor. Mono- `Metaphorical' mapping in human infants. Child Develop-
graphs of the Society for Research in Child Development, 52 ment, 52, 728± 731.
(1, Serial number 215). Werner, H., & Kaplan, B. (1967). Symbol formation: An
Meerum Terwogt, M., & Hoeksma, J.B. (1995). Colours and organismic-developmental approach to language and the
preferences: preferences and combinations. Journal of expression of thought. New York: Wiley.
General Psychology, 122, 5 ±17. Werner, J.S., & Wooten, B.R. (1979). Human infant colour
Munsell (1988). The Munsell book of color: Glossy finish vision and colour perception. Infant Behaviour and Devel-
collection. Baltimore, MD: Munsell Color. opment, 2, 241± 274.
Norman, R.D., & Scott, W.A. (1952). Color and affect: a Wexner, L. (1954). The degree to which colors are associated
review and semantic evaluation. Journal of General Psychol- with mood-tones. Journal of Applied Psychology, 38, 432±
ogy, 46, 185± 223. 435.
Osgood, C.E., Saci, G.J., & Tannen Gaum, P.H. (1957). The Whitfield, T.W., & Wiltshire, T.J. (1990). Color psychology: a
measurement of meaning. Urbana, IL: University of Illinois critical review. Genetic, Social and General Psychology
Press. Monographs, 116, 387± 405.
Philipps, R.D., Wagner, S.H., Fells, C.A., & Lynch, M. (1990). Winner, E., Rosenstiel, A.K., & Gardner, H. (1976). The

# Blackwell Publishers Ltd. 2001

398 Marcel R. Zentner

development of metaphoric understanding. Developmental Paper given at the symposium `The developmental psychol-
Psychology, 12, 289± 297. ogy of music'. Meeting of the Society for Research in Child
Williams, J.E., Boswell, D.A., & Best, D. (1975). Evaluative Development, Albuquerque, NM.
responses of preschool children to the colours white and
black. Child Development, 46, 501± 508. Received: 22 August 1999
Zentner, M.R. (1999). Children's affective responses to music. Accepted: 11 July 2000

# Blackwell Publishers Ltd. 2001