Journal of Experimental Child Psychology 94 (2006) 334–338

www.elsevier.com/locate/jecp

Color terms and color concepts

Centre for Cognition, Computation, and Culture, Goldsmiths, University of London, London SE14 6NW, UK

Received 23 February 2006

Abstract

In their lead articles, both Kowalski and Zimiles (2006) and O’Hanlon and Roberson (2006)
declare a general relation between color term knowledge and the ability to conceptually represent
color. Kowalski and Zimiles, in particular, argue for a priority for the conceptual representation
in color term acquisition. The complexities of the interaction are taken up in the current commen-
tary, especially with regard to the neuropsychological evidence. Data from aphasic patients also
argue for a priority for abstract thought, but nevertheless it may still be that the use of color terms
is the only way in which to form color categories even if both linguistic and attentional factors play
an important role.
Ó 2006 Elsevier Inc. All rights reserved.

Keywords: Color concepts; Color terms; Neuropsychology

The issue addressed in the lead articles by both Kowalski and Zimiles (2006) and
O’Hanlon and Roberson (2006), but in particular by Kowalski and Zimiles, is the devel-
opmental relation between color terms and color concepts. The issue has been of consid-
erable interest to Davidoff, Roberson, and their colleagues. In their work (e.g., Davidoff,
Davies, & Roberson, 1999; Roberson & Davidoff, 2000; Roberson, Davidoff, Davies, &
Shapiro, 2005), possession of a color concept goes hand-in-hand with the demonstration
of within-category similarity and between-category dissimilarity, that is, categorical
perception. Davidoff and colleagues (1999) found that memory and perceived
(category) similarity were simply predicted by the color terms in a speaker’s language.

*
Fax: +44 2070785145.
E-mail address: j.davidoff@gold.ac.uk.

0022-0965/$ - see front matter Ó 2006 Elsevier Inc. All rights reserved.
doi:10.1016/j.jecp.2006.03.001
 J. Davidoff / Journal of Experimental Child Psychology 94 (2006) 334–338 335

Their neo-Whorfian stance (Davidoff, 2001) derived from the cases where categorical per-
ception is or is not found and has provoked some controversy (Franklin, Clifford, Wil-
liamson, & Davies, 2005; Pilling, Wiggett, Özgen, & Davies, 2003).
Pilling and colleagues (2003) disputed the apparent contention in Roberson and
Davidoff (2000) that categorical perception for color derives from matching the stimulus
to a color label. However, simple verbal labeling as an explanation of categorical percep-
tion is not really what those authors claimed. The current commentary gives a needed
opportunity for clarification of their position. On the one hand, there is reason to believe
from their data (Roberson et al., 2005) that color labels are much involved. To borrow a
phrase from visual neurophysiology, Roberson and colleagues (2005) showed an ‘‘exqui-
site tuning’’ between color labels and categorical perception within a narrow range of
categories. On the other hand, their other data would imply that color labels per se were
not important for production of categorical perception. For example, one of Roberson
and Davidoff’s (2000) experiments showed that interference from noncolor words was
equally effective as that from color words in removing categorical perception. Further-
more, O’Hanlon and Roberson (2006) give several examples, particularly from develop-
mental studies, where there is no clear relation between performance on color tasks and
the speaker’s basic color knowledge. A reflection on these conflicting directions is given
here from a consideration of what might at first seem to be a quite different field of
research.
A similar conflict about the role of color labels was shown in a patient, LEW, who
could neither produce nor comprehend color terms, and although he was unable to under-
stand color category similarity, he nevertheless showed normal categorical perception
(Roberson, Davidoff, & Braisby, 1999). Thus, evidence from neuropsychology may be crit-
ical to understanding the relation between color terms and abstract color concepts. The
important evidence comes from studies of patients with language impairment (aphasia)
in categorization tasks. These tasks have been used by many neuropsychologists because
an inability to categorize was considered crucial to the debate concerning the relations
between impaired language and thought. During the early days of modern neuropsycho-
logical research, Wernicke considered aphasia to be a lexical impairment (see Davidoff &
Roberson, 2004). In contrast, Hughlings Jackson (cited in Zangwill, 1964, p. 261) declared
that aphasics, by having lost language, were ‘‘lame in thinking.’’ The argument was con-
sidered one of the most crucial in neuropsychology and surfaced many times during the
subsequent 100 years. Indeed, Goldstein (1948) considered that a loss of abstract process-
ing was the core deficit in aphasia.
As Goldstein (1948) noted, there are two main types of categorization task. One type
can be ‘‘solved’’ simply by episodic knowledge and what he called a concrete attitude,
and the other type required an abstract attitude. According to Goldstein, after certain
types of aphasic damage, the patient maintained only a concrete attitude. For color tasks,
the consequence is that judgments are based on perceptual similarity. For other tasks, the
patient’s judgments are based on what are now termed thematic associations (Markman &
Huchison, 1984) or on personal associations. For example, Goldstein (1948) described a
patient who kept both a hammer and a saucepan in the kitchen and therefore erroneously
placed them together in a categorization task. Another example, where the use of episodic
(visual) associations arrives at correct categorization, is from LEW (Roberson et al.,
1999), who was able to sort animals into British versus foreign by virtue of whether he
had seen them in a zoo.
336 J. Davidoff / Journal of Experimental Child Psychology 94 (2006) 334–338

LEW was completely unable to sort colors (Roberson et al., 1999). A further examina-
tion of the patient (Davidoff & Roberson, 2004) showed clearly that the difficulty in sort-
ing colors was part of a general difficulty in taxonomic classification. His difficulty was not
explained by a general intellectual deficit. LEW was also asked to do analogical reasoning
tasks of the type used by Gentner (see Kowalski & Zimiles, 2006). Although his level of
performance on analogical reasoning was only that of a 4- or 5-year-old, it far surpassed
his ability to follow rules of perceptual classification. It was therefore argued that taxo-
nomic classifications cannot be driven by the development of analogical reasoning. How-
ever, LEW behaved like a 3-year-old in a task that directly contrasted thematic and
taxonomic classification (Markman & Huchison, 1984). LEW showed a preference for the-
matic classification. In fact, there was no evidence of any substantial ability to make tax-
onomic color classifications despite evidence for good preservation of the associated
object–color knowledge (e.g., red went with strawberry). The reason is that tasks involving
colors, as opposed to object–colors, allow only a limited number of procedures to achieve
correct categorization. For example, artifacts and animals, because of their multiple asso-
ciations and properties, allow multiple routes for categorization. Indeed, allocation of
objects to categories is always context dependent; there are an infinite number of ways
in which they may be divided into categories. Colors are not like objects because there
is not the variety of episodic knowledge to use in color categorization tasks. Thus, the fail-
ure of the patients is more extreme for perceptual categories because there are no associ-
ations, such as are seen in a zoo, that can be used in the task of color categorization.
These important differences between objects and colors with respect to categorization
are laid down in the model of color processing proposed by Davidoff (1991). That model
proposed an internal color space that would organize together perceptual hues defined as
similar. Thus, access to the color space was required to produce a common lexical entry
to many different hues. However, the model may be incorrect in showing that color
terms can be accessed only by reference to the internal color space. It must be possible
to ‘‘parrot’’ a color term without its understanding. So, it might be better put to say that
the internal color space is used only when, to use Goldstein’s (1948) terms, we adopt an
abstract attitude for color naming. Goldstein contended that when a person ‘‘truly
‘names’ an object, he has the experience of a word which ‘means’ this object. . . . Other-
wise, he experiences the word as a sound complex belonging to an object’’ (p. 61). Gold-
stein referred to the type of naming that is preserved without access to the semantic
features of an object as ‘‘pseudonaming,’’ in contrast to true naming that requires an
abstract attitude.
Goldstein (1948) claimed that when pseudonaming, the names do not affect thought.
The consequence is that the label given to the color patch does not influence categoriza-
tion. For example, Goldstein reported a patient who placed colors in the same group
despite an unwillingness to call them by the same name. When asked whether the colors
in the group could not all be called red, the patient replied ‘‘No, these are red, this one
is pink, this one is maroon.’’ LEW’s poor color sorting was carried out in much the same
way and likewise gave away his inability to employ an abstract attitude. Goldstein warned
that the patient’s placing of color stimuli in a line according to some similarity dimension
should not lead the examiner to believe that the colors were categorized; rather, the reverse
is the case. He argued that the patient’s sorting behavior differed from that used in normal
categorization. When asked to sort many colors, we normally make piles of colors; this is
possible because the task is driven by some internal categorization mechanism, possibly
 J. Davidoff / Journal of Experimental Child Psychology 94 (2006) 334–338 337

the name. The person has no need to see the colors already classified. In contrast, the
patient with only concrete abilities needs to make visual, and apparently only pairwise,
comparisons. The patient’s concrete sorting procedure requires that stimuli that have
already been classified be in sight.
LEW’s sorting of colors (and shapes) was restricted to that which can be achieved by
perceptual similarity. The normal person also has the ability to sort categorically but
can adopt either mode. Kay and Kempton (1984), by varying task instructions and pro-
cedures, were able to induce normal adults to make ‘‘odd one out’’ color judgments based
either on perceptual or linguistic (categorical) differences. Recently, these two types of
judgment have been again related to a language versus a perceptual attitude to processing
with consequent implication for hemispheric differences (Gilbert, Regier, Kay, & Ivry,
2006). The categorical effects were found only for left hemisphere presentations.
The neuropsychological evidence would seem to give priority for a general semiotic
ability responsible for color concepts that could precede the acquisition of color terms.
Therefore, as both lead articles in this issue show, children may understand about color
before knowing color terms. However, things are not that simple. One needs to explain
the origin of the particular color concepts. Roberson and colleagues (1999) argued that
there was no other source for the origin of color categories besides labels unless the con-
cepts had some innate basis as, for example, proposed by Franklin and colleagues (2005).
One argument against an innate basis is that it would mean a considerable amount of
unlearning for speakers of most of the world’s languages where categories (e.g., blue,
green) do not exist in the adult’s language.
Recently, Fagot, Goldstein, Davidoff, and Pickering (in press) examined the question of
whether color categories are innate from a different approach by observing the similarity
matching of baboons and humans. They used as standards a typical green and a typical
blue. The interest was in the match of the intermediary colors. None of the baboons,
despite having color discrimination as good as that of humans, showed any inclination
to match any but very close colors to the standard. Humans showed quite different perfor-
mance, with a sharp division of the intermediary colors at the appropriate boundary
between green and blue. At the very least, Fagot and colleagues showed that the posses-
sion of a trichromatic visual system like that of humans does not by itself produce color
categories. So, one is drawn back to the argument that, at the very least, labels generate
the range of each color category.
In light of this work, what are the conclusions that we ought to draw from the Kowalski
and Zimiles article and the O’Hanlon and Roberson article? Kowalski and Zimiles show
that 3-year-olds have an idea about color and can extract color as a feature and use it in an
identity match. Specific color labels do not seem to map neatly onto those specific color
concepts, but overall there is a good relation between the two abilities. Importantly, there
is a priority for the abstract representation. The current commentary does not disagree
with these conclusions, but it cautions that the relation between concepts and terms
may be more profound than is shown in their study. A similar caution may be given to
the studies of O’Hanlon and Roberson in which they assign the relative importance of lin-
guistic or attentional factors to the acquisition of new color labels. Their tasks, like those
of Kowalski and Zimiles, do not allow us to know whether children really understand
about redness, blueness, beigeness, and the like. To be sure about that level of abstraction,
we need to examine the range of the match. So, if the interest is to examine the question of
whether labels are necessarily required for color category instantiation, a more demanding
338 J. Davidoff / Journal of Experimental Child Psychology 94 (2006) 334–338

task—one that uses similarity rather than identity matching—could show a precise
connection.

Acknowledgment

The author is supported by an EU Grant (12984), Stages in the Evolution and Devel-
opment of Sign Use (SEDSU), and by a Leverhulme major research Grant (F/07 605/D).

References
Davidoff, J. (1991). Cognition through color. Cambridge, MA: MIT Press.
Davidoff, J. (2001). Language and perceptual categories. Trends in Cognitive Science, 5, 382–387.
Davidoff, J., Davies, I., & Roberson, D. (1999). Colour categories of a stone-age tribe. Nature, 398, 203–204.
Davidoff, J., & Roberson, D. (2004). Preserved thematic and impaired taxonomic categorisation: A case study.
Language and Cognitive Processes, 19, 137–174.
Fagot, J., Goldstein, J., Davidoff, J., & Pickering, A. (in press). Cross-species differences in color categorization.
Psychonomic Bulletin and Review.
Franklin, A., Clifford, A., Williamson, E., & Davies, I. R. L. (2005). Color term knowledge does not affect
categorical perception in toddlers. Journal of Experimental Child Psychology, 90, 114–141.
Gilbert, A. L., Regier, T., Kay, P., & Ivry, R. B. (2006). Whorf hypothesis is supported in the right visual field but
not the left. Proceedings of the National Academy of Sciences, 103, 489–494.
Goldstein, K. (1948). Language and language disturbances. New York: Grune & Stratton.
Kay, P., & Kempton, W. (1984). What is the Sapir–Whorf hypothesis? American Anthropologist, 86, 65–78.
Kowalski, K., & Zimiles, H. (2006). The relation between children’s conceptual functioning with color and color
term acquisition. Journal of Experimental Child Psychology, 94, 301–321.
Markman, E., & Huchison, J. (1984). Children’s sensitivity to constraints on word meaning: Taxonomic vs.
thematic relations. Cognitive Psychology, 16, 1–27.
O’Hanlon, C., & Roberson, D. (2006). Learning in context: Linguistic and attentional constraints on children’s
color term learning. Journal of Experimental Child Psychology, 94, 275–300.
Pilling, M., Wiggett, A., Özgen, E., & Davies, I. R. L. (2003). Is color ‘‘categorical perception’’ really perceptual?
Memory & Cognition, 31, 538–551.
Roberson, D., & Davidoff, J. (2000). The categorical perception of colors and facial expressions: The effect of
verbal interference. Memory & Cognition, 28, 977–986.
Roberson, D., Davidoff, J., & Braisby, N. (1999). Similarity and categorisation: Neuropsychological evidence for
a dissociation in explicit categorisation tasks. Cognition, 71, 1–42.
Roberson, D., Davidoff, J., Davies, I. R., & Shapiro, L. R. (2005). Color categories: Evidence for the cultural
relativity hypothesis. Cognitive Psychology, 50, 378–411.
Zangwill, O. L. (1964). Intelligence in aphasia. In A. V. S. de Reuck & M. O’Connor (Eds.), Disorders of language
(pp. 261–284). London: J. & A. Churchill.