Journal of Vision (2014) 14(2):17, 1–25 http://www.journalofvision.

org/content/14/2/17 1

The color lexicon of American English

Department of Psychology, Ohio State University,
Mansfield, OH, USA
College of Optometry, Ohio State University,
Delwin T. Lindsey Columbus, OH, USA $
College of Optometry, Ohio State University,
Angela M. Brown Columbus, OH, USA $

This article describes color naming by 51 American English– languages. The third issue concerns the evolution of a
speaking informants. A free-naming task produced 122 language’s color lexicon as color categories change. The
monolexemic color terms, with which informants named present study seeks to address some of these issues by
the 330 Munsell samples from the World Color Survey. analyzing the color-naming behavior of a group of native
Cluster analysis consolidated those terms into a glossary of English-speaking informants drawn from the relatively
20 named color categories: the 11 Basic Color Term (BCT) culturally homogeneous population of Ohio State Uni-
categories of Berlin and Kay (1969, p. 2) plus nine nonbasic
versity faculty, staff, and students.
chromatic categories. The glossed data revealed two color-
naming motifs: the green–blue motif of the World Color
Survey and a novel green–teal–blue motif, which featured
peach, teal, lavender, and maroon as high-consensus terms. Berlin and Kay
Women used more terms than men, and more women
expressed the novel motif. Under a constrained-naming The context for this work is the classic theoretical
protocol, informants supplied BCTs for the color samples analysis of cross-cultural differences in color categories
previously given nonbasic terms. Most of the glossed by Berlin and Kay (1969). On the basis of their study of
nonbasic terms from the free-naming task named low- 98 world languages, these authors advanced two
consensus colors located at the BCT boundaries revealed by conjectures about the differences they observed. Their
the constrained-naming task. This study provides evidence first conjecture was that there is a limited set of basic
for continuing evolution of the color lexicon of American color terms (BCTs) in most languages, which are distinct
English, and provides insight into the processes governing from other color terms that an individual might use to
this evolution. name colors. According to this first conjecture, the
colors in the lexicon of each language are a subset drawn
from a universal set of 11 color categories, which are
Introduction closely related to the BCTs of English and other
languages spoken in technologically advanced societies.
Berlin and Kay’s second conjecture was that color
Humans can discriminate on the order of 106 different
colors, many more colors than any individual can name lexicons evolve from simple to complex, along highly
reliably. These colors fall into a much smaller number of constrained paths, starting from two BCTs correspond-
categories that speakers in a language community can ing to warm-or-light and dark-or-cool categories in the
name and can use among themselves to communicate simplest lexicons and ending with the 11 BCTs of
about color. People around the world differ greatly in the languages like English.
number of these named color categories. However, despite
more than 150 years of research, several unresolved issues
persist regarding cross-cultural differences in color The first conjecture: The basic color terms and
lexicon. One of these issues concerns how best to their universality
characterize the relative importance of terms in a
language’s color lexicon. Another is how to compare and Berlin and Kay proposed that most world languages
contrast color lexicons across the world’s 7,000 living include a set of BCTs in their lexicons. According to

Citation: Lindsey, D. T., & Brown, A. M. (2014). The color lexicon of American English. Journal of Vision, 14(2):17, 1–25, http://
www.journalofvision.org/content/14/2/17, doi:10.1167/14.2.17.
doi: 10 .116 7 /1 4. 2. 1 7 Received August 12, 2013; published February 25, 2014 ISSN 1534-7362 Ó 2014 ARVO

Downloaded from jov.arvojournals.org on 06/23/2024

 Journal of Vision (2014) 14(2):17, 1–25 Lindsey & Brown 2

their definition, the BCTs are monolexemic (single, Boynton and Olson to Japanese color terms, the results
noncompound words that lack modifying prefixes or were generally similar. Particularly, Uchikawa and
suffixes) and are used principally in reference to the Boynton also found that similar terms—hada (skin),
colors of things, without constraint as to what thing is meaning tan, and mizu (water), meaning light blue—
being described. Moreover, BCTs are present in the may be making their way into the Japanese basic color
idiolects of all informants speaking a given language, lexicon. Similarly, several investigators have discussed
are used in a consistent way across all informants, and terms for light blue, which might be a 12th BCT in
can be used to partition color space exhaustively. By several other languages (Al-Rasheed, Al-Sharif, Thabit,
these and other criteria, Berlin and Kay proposed 11 Al-Mohimeed, & Davies, 2011; Borg, 2007; Friedl,
English BCTs: black, white, red, yellow, green, blue, 1979; Ozgen & Davies, 1998; Thierry, Athanasopoulos,
brown, orange, pink, purple, and gray. In contrast to Wiggett, Dering, & Kuipers, 2009; Winawer et al.,
the BCTs, most languages, including English, have 2007).
additional color terms that fail to meet one or more of The issue of consensus, which was central to the
these criteria: Either not everybody uses them (e.g., definition of the BCTs, has turned out to be unex-
chartreuse in English), they are not monolexemic (e.g., pectedly complex. Boynton and Olson (1987) discov-
light blue in English) or they are restricted as to what ered clear individual differences among their observers,
they can name (e.g., blond in English). and Uchikawa and Boynton (1987) found that there
BCTs name basic color categories. While the terms were no 100% consensus colors among the 430 OSA
themselves are specific to a language (the same samples samples corresponding to the Japanese colors akai
might be called red in English, rouge in French, akai in (red), kuroi (black), kiroi (yellow), and aoi (blue).
Japanese, and so forth), Berlin and Kay’s first Similarly, Sturges and Whitfield (1995) reported no
conjecture was that the color categories these terms 100% consensus samples for yellow, pink, orange, and
refer to are universal across languages. While not every white. Furthermore, the issue of consensus is compli-
language has every color category named within its cated by the existence of synonyms for many colors:
lexicon, Berlin and Kay proposed that ‘‘a total Different words might be used by different informants
universal inventory of exactly eleven basic color speaking the same language to name the same or highly
categories exists from which the eleven or fewer basic similar color categories (violet and purple might be
color terms of any language are always drawn’’ (1969, synonyms in English, hairoi and guree (gray) might be
p. 2). synonyms in Japanese).
In the years since Berlin and Kay’s work, an Moreover, Lindsey and Brown (2006, 2009) have
enormous amount of research has been done to shown that, strictly speaking, high consensus may be
determine whether their first conjecture is correct. Some the exception rather than the rule among the color
investigators have addressed the issue of whether the lexicons in world languages. They examined the World
inventory of terms listed, and only those color terms, Color Survey (WCS) data set (Kay, Berlin, Maffi,
fulfill Berlin and Kay’s definition of BCTs in every Merrifield, & Cook, 2010), a large database of color
known language. This literature as a whole suggests naming by 2,616 informants, each speaking one of 110
that at least some BCTs exist in every language that has unwritten languages and living a traditional lifestyle far
been examined, but that there might be more than 11 of from daily influences of modern technology. Each WCS
them in some cases. Boynton and Olson (1987, 1990; informant was tested with a standard set of 330
for a review, see Boynton, 1997, pp. 144–145) used Munsell color samples of varying hue, value, and
performance-based measures of color naming to chroma (shown in Figure 1a), one at a time, in a fixed
evaluate the special status of Berlin and Kay’s English pseudorandom order, and provided a color name for
BCTs. When Boynton and Olson’s American English– each. Lindsey and Brown (2006) used cluster analysis to
speaking subjects were allowed to use any monolexemic extract a glossary of universal terms used by WCS
terms to name colors in the Optical Society of America informants. A second cluster analysis (Lindsey &
(OSA) Uniform Color Space, they used BCTs with Brown, 2009) on the glossed color-naming systems of
significantly greater speed, consensus, and consistency WCS informants revealed that the color vocabularies of
than nonbasic terms, much as Berlin and Kay WCS informants clustered into four distinct vocabulary
predicted. Boynton and Olson also noted that a 12th types (‘‘motifs’’), where each motif had its own
term—peach—might, over time, assume BCT status in characteristic set of color terms. Crucially, multiple
terms of naming speed, consensus, and consistency. motifs occurred side by side within most WCS
Sturges and Whitfield (1995) found similar results using languages. This meant, for example, that some speakers
Munsell color samples and British English–speaking of a language might use only color terms glossed as
subjects, except that they suggest that the 12th term black, white, and red, while others might use five color
might be cream (Sturges & Whitfield, 1997). When terms, and still others might use 10 color terms. The
Uchikawa and Boynton (1987) applied the methods of lack of consensus revealed by Lindsey and Brown’s

Downloaded from jov.arvojournals.org on 06/23/2024

 Journal of Vision (2014) 14(2):17, 1–25 Lindsey & Brown 3

Figure 1. (a) Color chart approximating the samples used in this study, arranged in the order of their Munsell hues (horizontal
direction of the chart) and values (vertical direction). (b) Sample holders, shown on Color-aid N4.5 background.

2009 analysis was not merely quibbling about where the to languages that have fewer color terms by ‘‘succes-
boundaries are located in the stimulus set. Rather, it sive differentiation of previously existing color cate-
indicates a profound failure of consensus among the gories’’ into smaller, more accurately named
speakers of most WCS languages. subcategories (Kay & McDaniel, 1978, p. 640). This
The work of Lindsey and Brown (2009) revealed a process follows a series of stages, in a fairly
new kind of universality in addition to the one constrained evolutionary trajectory. According to the
proposed by Berlin and Kay (1969). Whereas the WCS second conjecture, this continues until the lexicon
languages differed from one another in how many reaches a stage equivalent to the 11 BCTs of English
individuals used each of the four motifs, the motifs and other languages spoken in industrialized societies.
themselves occurred worldwide, even though the Thus, Berlin and Kay’s second conjecture is that color
informants who used each of the motifs spoke lexicons evolve, that they follow a prescribed trajec-
languages with no known historical linguistic ties. This tory, and that color terms are added by ‘‘partition’’ of
suggested that analysis of color naming must be existing named color categories.
conducted at the level of each informant’s idiolect, In a sense, the strict ordering proposed by Berlin and
rather than at the level of the language shared among a Kay resembles a theory of biological development, in
community of informants. Furthermore, the results which maturation of the organism occurs in stages
revealed the usefulness of cluster analysis as an following a single prescribed trajectory, with minor
objective means of comparing color naming across differences from individual to individual. More recent
languages, thus avoiding many of the pitfalls associated work by Kay et al. (2010) has relaxed and generalized
with glossing color terms by traditional lexicographic the evolutionary hypothesis considerably, to allow for
techniques. some languages that do not fit neatly into one of Berlin
and Kay’s original stages and to suggest a much less
constrained, more diverse range of evolutionary path-
The second conjecture: The evolution of BCTs ways. The more diverse range of trajectories proposed
by Kay et al., 2010 resembles ontological evolution
Berlin and Kay’s second conjecture was that more closely, where a population of organisms can
languages evolve over time by adding new color evolve in any of a number of directions, subject to the
categories (see also similar concepts proposed by Darwinian principles of natural selection.
Gladstone, 1858; Rivers, 1901; Hugo Magnus, trans- Berlin and Kay’s second conjecture poses two
lated in Saunders & Marth, 2007; and Schontag & important questions pertaining to languages like
Schafer-Priess, 2007). According to the second con- American English, which are spoken in technologically
jecture, new color terms are continually being added advanced cultures. First, is the current state of modern

Downloaded from jov.arvojournals.org on 06/23/2024

 Journal of Vision (2014) 14(2):17, 1–25 Lindsey & Brown 4

color naming in these languages a proper end state of Berlin and Kay were allowed. This constrained-naming
color-term evolution, as Berlin and Kay proposed? phase of the protocol was used to establish each
There is some evidence that it is not. First, several informant’s BCT category boundaries. Then, the
modern languages with more than 11 basic color terms deployment of nonbasic color names—within versus
have been identified, which otherwise satisfy Berlin and between BCT categories—in the free-naming phase
Kay’s criteria. These include Russian (Davies & could be gauged in relation to these boundaries.
Corbett, 1994; Winawer, 2007), Greek (Thierry et al., The data analysis in the present project followed the
2009), and Turkish (Ozgen & Davies, 1998). Second, two-stage cluster-analysis methodology of Lindsey
there is some evidence that other English-like color and Brown (2006, 2009). In the first stage, each color
lexicons may be continuing to evolve. For example, term used by each informant was encoded by a
Zollinger (1984) proposed that turquoise may be a separate binary feature vector, which represented the
nascent color category in German, and Boynton and subset of color samples associated with each color
Olson (1987) proposed that even English itself might be word. These feature vectors were then partitioned into
currently evolving, adding peach as a possible new distinct clusters, which represented a glossary of
color category. distinct color categories identified in the data set
The second question arises if we grant the likelihood irrespective of the actual words associated with the
of the continuing evolution of the English color clustered feature vectors. This step avoided the
vocabularies: How are the new categories formed? Is potential pitfalls of synonymy in the data analysis. In
the process constrained by the partition principle of the second stage, the clustered feature vectors were
Kay et al., or by some other process? An alternative used to reconstruct a representation of the glossed
process has been advanced by Levinson (2000) and color-naming system for each American English–
Lyons (1995), who have challenged both of Berlin and speaking informant, and a second cluster analysis of
Kay’s conjectures. Here, we focus on the second the data set was performed at the level of the
conjecture and Levinson’s idea that in the earliest
informants. This step allowed us to determine whether
stages of color-term evolution, color vocabularies do
distinct subpopulations of American English–speak-
not exhaustively name all colors. Rather, according to
ing informants express different color-naming motifs.
Levinson, each ancient color term referred to a
This two-stage cluster analysis permitted an analysis
restricted range of colors that were identified with a
of the American English color lexicon that was more
particular item in the environment, for example a
certain animal or plant, or a substance such as blood or nuanced and powerful than one based on simple
bile. Over time, the original terms generalized to the tabulation of subject color-naming responses.
colors of the substances to which they originally This analysis was designed to address the issues
referred. However, great gaps remained where the outlined previously. Under the first conjecture, are
colors were either unnamed or else were named with there color terms in American English that fulfill
great difficulty. According to Levinson, additional Berlin and Kay’s definition of the BCTs? And are the
color terms came into use as the need arose to name most common American English color terms equiva-
colors in the gaps, colors that previously had no names. lent to the universal color terms that have previously
Thus, according to Levinson, ‘‘color terms [emerge] out been identified for the WCS, or do they differ from
of noncoloric expressions’’ (2000, p. 8); this view of those universal terms in important ways? Under the
color-term evolution has come to be known as the second conjecture, is American English in the process
‘‘emergence hypothesis.’’ of evolving to higher numbers of BCTs, as suggested
by Boynton and Olson? And if new color terms are
evolving in American English, do they appear by
This project partitioning existing categories into smaller ones, as
Kay et al. proposed? Or is Levinson closer to the
In the present study, we examined Berlin and Kay’s mark, with these new terms appearing de novo,
two conjectures in light of a new color-naming data set popping up in places where no high-consensus color
that we obtained from American English–speaking term exists and informants find the colors hard to
informants. To facilitate comparisons between Amer- name? Finally, we used the data set to examine the
ican English and the 110 languages represented in the relationship between informant gender and color
World Color Survey, we used the set of Munsell color naming. Several studies have proposed that females
samples used in the WCS. We used a ‘‘free-naming’’ have larger color vocabularies than males, possibly for
protocol, in which informants used whatever single genetic reasons, and women might have a finer
color term they wished, subject to a few simple rules, appreciation of the differences between colors and
and we added a ‘‘constrained-naming’’ phase to the their identities because of their roles in modern and
data-collection protocol, in which only the 11 BCTs of traditional cultures.

Downloaded from jov.arvojournals.org on 06/23/2024

 Journal of Vision (2014) 14(2):17, 1–25 Lindsey & Brown 5

This data set also allowed us to examine prospectively (Kay et al., 2010). Ten samples were achromatic, with
the generality and replicability of the motifs of Lindsey values from 1.5/ to 9.5/ in the notation of the Munsell
and Brown (2009). Their conclusion that multiple motifs color-order system. The remaining samples were the 40
exist side by side within the lexicons of most world equally spaced Munsell hues (2.5 R to 10 RP, in hue
languages was based on an analysis of the WCS, a data steps of 2.5) sampled at each of eight values from 2/ to
set that had already been collected. This unexpected 9/ (hence, 320 hue/value combinations). The chromatic
result suggested that the color lexicons of many world samples generally had high chroma, except for some
languages are currently undergoing linguistic change, hues at the lowest and highest values, where the
albeit over more trajectories than the simple path Munsell Book of Color does not contain high-chroma
originally proposed by Berlin and Kay. Therefore, it was samples. Each sample was placed in a 51-mm · 28-mm
important to determine whether multiple motifs occur in holder, which was covered in Color-aid N4.5 paper
a data set that was collected to reveal them if they exist. exposing a 20-mm · 20-mm (2.38 · 2.48 at an
On one hand, American English is a written language approximate viewing distance of 500 mm) portion of its
spoken by a large number of people, which suggests that Munsell sample.
it might not still be evolving in a basic, common lexicon
such as the names of colors. On the other hand,
American culture is highly industrialized, which suggests Apparatus
that there might be a need for more color terms as more
and more artifacts in American culture differ from one The color samples were presented in a custom-made
another only in their color. light box, with white walls, a floor covered with Color-
aid N 4.5 paper, and a 42-cm · 142-cm opening
through which subjects viewed the illuminated color
Methods samples. Illumination was provided by a bank of four
full-spectrum fluorescent lights (F40T12 Spectralite;
CRI 90) suspended from the top of the light box. Color
Informants calibrations were performed using a PhotoResearch
PR-670 spectrophotometer at regular intervals
Fifty-one native American English–speaking infor- throughout the duration of the study. Illumination
mants (24 men and 27 women; ages 19–58 years) varied between 1970 and 2216 lux during the course of
participated in the study. All were born and raised in the study and had a correlated color temperature
the United States, none had spoken any language other (CCT) of between 5200 and 5400 K during the same
than English at home before the age of 12 years, and all time period. This CCT is near that of direct sunlight.
resided in the Columbus, Ohio, metropolitan area at
the time of testing. None of the informants were aware
of the hypotheses being tested in this study. All subjects Procedure
reported that they were free of visual pathology except
for refractive error, and all were color normal, as At the beginning of each experimental session, the
assessed using the D-15 screening test. The informants informant was briefed on the nature of the task. After
were tested following a protocol previously approved the informant provided informed consent for partici-
by the Ohio State University’s Institutional Review pation in accordance with the Declaration of Helsinki
Board, and all gave informed consent prior to and under the approval of the Ohio State University
participating in this study. Institutional Review Board, we obtained the infor-
Data on three additional male informants, one who mant’s age and sex, the languages he or she spoke and
spoke British English and two who were color-vision at what age he or she learned them, where he or she was
specialists and well aware of the hypotheses being born and lived as a child, and where he or she lived
tested in this study, have been eliminated from the data presently. We also administered the D-15 panel to
set reported here. None of the results or conclusions screen for color-vision deficiency.
from this project are materially affected by the The color-naming part of the experiment consisted
exclusion of these informants. of two phases: free naming and constrained naming.
The free-naming phase of the experiment was always
completed first, and informants were not apprised of
Color samples the second, constrained-naming, phase until after the
first phase was completed. All but one informant
The 330 color samples were taken from the Munsell completed both color-naming phases in one 1.5-hr
Book of Color Glossy Edition and corresponded closely experimental session. The remaining informant re-
to those used in the World Color Survey color chart quired two 1.5-hr sessions to complete the two color-

Downloaded from jov.arvojournals.org on 06/23/2024

 Journal of Vision (2014) 14(2):17, 1–25 Lindsey & Brown 6

naming phases of this study. Subjects were paid with a required the informant to provide a BCT from the list.
$10 gift card for their participation. The colors provided in the constrained-naming phase
In the free-naming phase of the experiment, the were added to the data set obtained in the free-naming
experimenter presented each of the 330 Munsell phase to create a second complete constrained-naming
samples in the light box in a fixed, pseudorandom data set.
order. The subject named each color sample in turn.
The instructions were to name the colors of the
samples, based on the following three criteria:
Results
1. The color name must be a single word. (Phrases like
light blue and dark green, and phrases with intrinsic
modifiers like yellowish are not acceptable.) Color terms elicited under free-naming
2. The word must be a general color name, applicable instructions
to anything of that color. (Blond, for example, is not
such a word, as it is used to name the color of hair, Every informant succeeded in naming every sample
furniture, or beer, but not, for example, a car or a with a monolexemic color term in the free-naming
potato.) phase of the study. The 51 native American English–
3. The word must be the one that you would normally speaking informants used a total of 122 color terms to
use to name the color of something in your everyday name the 330 Munsell samples (Tables 1 and 2). Figure
life. (We are not looking for a unique name for each 2a and Table 3a present the most commonly used basic
color. We are not testing for how many different and nonbasic color terms in this study. The mean
color names you know or can dream up, or how number of free-naming color terms per informant was
many subtle distinctions in color you can name. We 21.9 (SD ¼ 7.6) and the minimum number of color
want just to know how you naturally name the terms was 12 (Figure 3a).
colors, when you can use only a one-word name.) Among the nonbasic terms, peach was used by the
largest number of informants (40 out of 51 informants).
Once we had instructed the informant and answered Teal, which was used by 32 of 51 informants, was the
any questions, the informant was allowed to use only other nonbasic term used by over half of
whatever color terms he or she chose: Informants informants. Fifty-nine of the 122 color terms (48.4%)
complied with criterion #1 without exception, and we were used by only one informant each, and of those, 25
did not interrupt to object to terms that might have were used only a single male informant. The inventory
violated criteria #2 or #3. Color terms that were not of frequently used nonbasic color terms in Figure 2a
among the 11 BCTs of Berlin and Kay were flagged on and Table 3a is qualitatively similar to that reported by
the data sheet. The free-naming data set was the full set Boynton and Olson (1987, 1990) and by Sturges and
of monolexemic color terms provided by this sample of Whitfield (1995). However, none of those previous
informants. studies mentions teal in its nonbasic color inventory;
The instructions in the free-naming phase of data instead, they all report the use of turquoise, which we
collection differed somewhat from those used in the found to be synonymous with teal (we deal with color-
WCS protocol. In that study, field-workers were term synonymy later). None of Sturges and Whitfield’s
encouraged to elicit short, single-word BCTs from their informants used lavender, though they did report the
informants, in their native language. However, there term lilac, which we found to be synonymous with
was probably considerable variation in how these lavender. Finally, Boynton and Olson (1990) reported
instructions were actually followed by field-workers that both peach and tan were used by all nine of their
(see Cook, Kay, & Regier, 2005). Thus, we believe our informants, compared to 62.7% and 45.1%, respec-
color-naming protocol adhered reasonably well to that tively, of informants in this study. Boynton and Olson
of the WCS as actually implemented in the field. (1987, 1990) speculated that peach might be an
Moreover, by using a relatively unconstrained color- emerging 12th English BCT. We will return to this
naming protocol in the first phase of data collection, we point in the Discussion section of this article.
could compare our results to those of prior studies of While every informant used all the BCTs of Berlin
English monolexemic color naming by Boynton and and Kay, the BCTs differed considerably in their
Olson (1987, 1990) and by Sturges and Whitfield (1995, frequency of usage (Table 3a). On average, 24.0% of
1997). the sample presentations (4,035 of 16,830 total re-
In the constrained-naming phase of the experiment, sponses) elicited green as a color term, followed by blue
we gave the informant a printed list of the 11 BCTs. We (16.9%). Red, with a usage rate of 3.3%, was the least
then presented, for a second time and in reverse order, elicited of the chromatic BCTs. Of the neutrals, black
the samples corresponding to the color terms that were was elicited by 1.9% of the 330 color samples across the
flagged in the free-naming phase of the experiment, and 51 informants, white was elicited by 4.7%, and gray was

Downloaded from jov.arvojournals.org on 06/23/2024

 Journal of Vision (2014) 14(2):17, 1–25 Lindsey & Brown 7

Terms/informant
Number of Number of Type of Number of
Study informants color samples samples terms Mean SD Median I.q.r.* Mode

Present study 51 330 Munsell 122 21.88 7.58 20 7.5 18

(World Color Survey)
Boynton and Olson (1990) 9 424 OSA 82 32.89 14.84 31 29.5 N/a**
Sturges and Whitfield (1995) 20 446 Munsell N/a*** N/a*** N/a*** N/a*** N/a*** N/a***

Table 1. Basic data from three studies. Notes: *Interquartile range. **No two subjects used the same number of terms. ***Not
reported.

elicited by 2.2%. The nonbasic term used to name the Color terms elicited under constrained-naming
most samples was teal (2.0% of samples), followed by instructions
peach (1.5%) and lavender (1.3%). The only term for
light blue was sky, which was used by only four Under the constrained color-naming instructions, all
subjects, to name 0.21% of samples. the informants succeeded in naming all of the samples
Many investigators have found that the frequency of for which they had provided nonbasic color terms in
word use conforms to a power law—that is, the the free-naming part of the protocol. This was
logarithms of the frequency with which words occur fall
on a line when graphed as a function of the logarithm of
their rank order. This power-law relation is sometimes
called Zipf’s law (see Mitzenmacher, 2003, for a review).
Contrary to that general result, when we graphed the
number of informants using each term (the term’s
‘‘popularity’’) as a function of the sorted rank order of
that term’s popularity (Figure 4, lower data set), the data
were broken quite sharply into three regimes. First, there
was a ceiling effect, as the BCTs were all used by all 51
informants and are therefore fitted perfectly by a constant
function. For the next 17 most popular terms, the power
law had an exponent of 1.2, whereas the power law for
the less popular terms was 3.32 (gray circles in Figure 4).
The slopes of these functions depended somewhat on how
ties were treated in the rank ordering (here, tied
frequencies have consecutive ranks) and how the BCTs
were shown on the graph (here, included as 11 tied
frequencies). However, no matter how we treated the ties,
there was always a break in the function after the 28th
term (11 BCTs plus 17 additional popular terms; numbers
listed in Table 2; colors listed in Table 3a). Double-power-
law behavior is common in language corpora, where two
exponents often ‘‘divide words in two different sets: a
kernel lexicon formed by about N versatile words and an
unlimited lexicon for specific communication’’ (Ferrer i
Cancho & Solé, 2001, p. 170).
Criterion All Chromatic Nonbasic
Used by . 0 informants 122 122 111
Used by . 1 informant 63 63 52
Used by . 2 informants 51 51 40
Used by . 3 informants 43 43 32
Basic color terms (Berlin & Kay, 1969) 11 8 0
Most common color terms* 28 25 17 Figure 2. Histograms of color-term usage in the free-naming
Glossed color categories** 20 17 9 phase of the experiment. (a) The number of informants using
each of the 43 color terms used by four or more informants. (b)
Table 2. Free-naming color terms. Notes: *White disks in Figure The free-naming data consolidated by cluster analysis into
4. **See Figures 2b and 3b and Table 3. glossed categories.

Downloaded from jov.arvojournals.org on 06/23/2024

 Journal of Vision (2014) 14(2):17, 1–25 Lindsey & Brown 8

Number of Number of
Rank by informants samples
popularity (popularity) (usage)

(a) Unglossed color terms

GREEN 1 51 4,035
BLUE 2 51 2,843
PURPLE 3 51 1,668
PINK 4 51 1,302
BROWN 5 51 930
WHITE 6 51 788
ORANGE 7 51 716
YELLOW 8 51 628
RED 9 51 558
GRAY 10 51 375
BLACK 11 51 315
PEACH 12 40 247
TEAL 13 32 338
LAVENDER 14 25 215
MAROON 15 24 109
TAN 16 23 58
GOLD 17 21 100
AQUA 18 20 136 Figure 3. (a) The distribution of the number of informants using
TURQUOISE 19 20 131 each total number of color terms in the free-naming phase of
BURGUNDY 20 20 72 the experiment. (b) The distribution of the free-naming data
VIOLET 21 19 163 after synonymous color terms were consolidated into color-
OLIVE 22 17 112 category clusters. The red arrows indicate 11, the number of
MAGENTA 23 17 84 BCTs according to Berlin and Kay (1969). Even after consolida-
SALMON 24 17 40 tion, no informant used only 11 color terms, and the modal
FUCHSIA 25 16 66 number of glossed color terms was 18.
LIME 26 15 40
LILAC 27 14 81 accomplished with small amounts of grumbling from
PERIWINKLE 28 14 57 some informants, who found some of the peach-colored
SKY 42 4 35 samples particularly challenging.
(b) Glossed color categories
All informants used all of the BCTs to name at least
GREEN 1 51 4,154
some of the color samples, so the median/average
BLUE 2 51 2,997
PURPLE 3 51 1,731
number of basic color terms was 11, and the
PINK 4 51 1,366 interquartile range/standard deviation was nil. On
BROWN 5 51 962 average, 27.9% of samples (4,693 of 16,830 total
WHITE 6 51 788 responses by 51 subjects) elicited green as a color term,
ORANGE 7 51 752 followed by blue (19.3%), with red, once again, being
YELLOW 8 51 655 the least used chromatic term (4.2%). Black was elicited
RED 9 51 591 by 1.9% of total responses, white was elicited by 4.9%,
GRAY 10 51 375 and gray was elicited by 2.3% of trials.
BLACK 11 51 315
PEACH 12 44 312
TEAL 13 43 589 Consensus in color-term usage
LAVENDER 14 41 447
MAROON 15 39 232 In their 1969 monograph, Berlin and Kay wrote that
GOLD 16 32 174 the BCTs in any language, including English, are
MAGENTA 17 28 119 ‘‘psychologically salient.’’ By this, they meant that the
BEIGE 18 23 86
BCTs are used with high consensus and consistency by
OLIVE 19 20 127
all competent speakers of the language. What were the
LIME 20 20 58
salient colors in this data set? Figure 5 shows consensus
Table 3. Usage and popularity of color terms and categories. (a) maps for the free-naming (left) and constrained-naming
Unglossed color terms ranked 1–28 in the order of their (right) phases of this study. Each of the 330 small
popularity (white circles in Figure 4), plus sky (see Discussion for rectangles within each map in Figure 5 corresponds to a
details). (b) Glossed color categories (triangles in Figure 4). color sample (its true color is illustrated in the

Downloaded from jov.arvojournals.org on 06/23/2024

 Journal of Vision (2014) 14(2):17, 1–25 Lindsey & Brown 9

corresponding location in Figure 1a), and its false-color

hue encodes its most frequent color term. In the case of
black, the false-color code was orange (the actual
orange category is in a lighter shade of orange in the
figure), and for gray it was chartreuse. The BCTs
constitute the most frequent names for 99% (326/330)
of the color samples named by the informants in the
free-naming phase of this study. Only four color
samples, which were in the peach area of the color
chart, received nonbasic color terms most of the time.
The maximum consensus for peach was 0.61 of
informants (see Boynton & Olson, 1987; Jameson &
Alvarado, 2003, for a similar result). Of course, the
BCTs were the modal color terms for all the samples in
the constrained color-naming data set.
The false-color light-to-dark values of the samples
shown in Figure 5a and d encode the consensus—that
is, the fraction of informants who used the modal
name, normalized to the overall maximum value for
visibility in the figure. The consensus in the con-
Figure 4. The number of informants using each of 43 unglossed
strained color-naming data set (Figure 5d) was higher
color terms used by four or more informants (circles), and 20 within color categories and lower near the color-
glossed color terms (triangles), plotted against their sorted rank category boundaries, indicating variation across
order. The first 11 terms, used by all 51 informants, are the BCTs subjects in the locations of the boundaries between
(horizontal red lines). Both graphs show reliable changes in the BCTs. The consensus was generally lower overall
slope (arrows), with rare color terms being used by very few in the free-naming task (average consensus ¼ 0.74;
informants (gray symbols). Circles are displaced downward for Figure 5a) than in the constrained-naming task
clarity. (average consensus ¼ 0.85), because informants used
nonbasic terms in addition to the BCTs in the free-
naming task.

Figure 5. Diagrams of color-naming consensus in the free-naming (a–c) and constrained-naming (d–f) tasks. See text for details of
color coding. (a, d) Consensus maps of the usage of the BCTs of Berlin and Kay (1969). Second and third rows indicate samples for
which the consensus reached or exceeded two threshold criteria: 1.0 (b, e) and 0.8 (c, f). Red dots: focal colors from Berlin and Kay
(1969); orange dots: focal colors from Sturges and Whitfield (1995).

Downloaded from jov.arvojournals.org on 06/23/2024

 Journal of Vision (2014) 14(2):17, 1–25 Lindsey & Brown 10

In Figure 5b, c, e, and f, the data from Figure 5a was defined as the average coordinates of the samples
and d were examined by applying two threshold levels that were named with the corresponding term by one
of consensus: 1.00 (Figure 5b, e) and 0.80 (Figure 5c, f; subject, specified within the 2-D Cartesian coordinate
0.80 was one of the criterion levels used by Kay et al., frame of the color chart shown in Figure 1. The faint
2010, in their description of the World Color Survey colors in the backgrounds of Figure 6a through d are
data set]. In those diagrams, the chromatic color the false colors from Figure 3f, corresponding to the
categories named at or above the critical value of modal BCTs used at consensus  0.80. Here, we
consensus appear as islands of at-or-above-threshold provide this map as a guide for examining the color-
agreement, separated by black boundary regions of term centroids. The black dots in Figure 6a are the
below-threshold agreement. Agreement among sub- averages of the centroids across all 51 subjects. For
jects was perfect (consensus ¼ 1.0) for 6.3% of samples comparison, the white dots are the average centroids
in the free-naming data set and for 31% of samples in obtained from 20 University of Teesside (U.K.)
the constrained-naming data set. Fifty-two percent of undergraduates by Sturges and Whitfield (1995),
the samples in the free-naming data set and 69% of the whose 446 Munsell color samples spanned a greater
samples in the constrained-naming data set reached or range of chromas than were used in the present
exceeded a consensus criterion of 0.80. In the free- study. The centroids from the two studies agree fairly
naming data set, none of the samples were called well.
white, red, yellow, pink, or orange by all informants, A striking feature of the individual data is the
and even in the constrained-naming data set, no informant-to-informant variation in the usage of the
samples were called white or red by all informants. nonbasic color terms. Informants often used different
However, at the 0.80 consensus threshold, all the color names to label similar regions of color space.
BCTs were represented in both the free-naming and For example, in the central region of the color chart
constrained-naming data sets. The dots in Figure 5 are
(Figure 6b), there were seven different nonbasic color
the ‘‘focal colors’’ of Berlin and Kay (1969) and
terms: teal, turquoise, aquamarine, aqua, jade, ocean,
Sturges and Whitfield (1995), that is, the colors that
and seafoam. These terms reliably denoted colors that
were chosen by their informants to be the best
fall near the boundary between the blue and green
examples of each of the color categories. The focal
BCT categories, and the distributions of the centroids
colors correspond closely to the regions of high color-
naming agreement, which indicates that color samples for these seven terms were broad and showed
were named with high consensus if they were, on considerable overlap. However, the centroids were
average, particularly good examples of named color not quite identical, suggesting that informants might
categories. Conversely, as colors deviated more and differ slightly in the meanings they associate with
more in appearance from the focal colors, subjects these nearly synonymous color terms. The centroids
were more variable in their responses and consensus for maroon and burgundy showed almost perfect
agreement declined. overlap, and the centroids for lavender and lilac also
The features of the 0.80 threshold consensus map for overlapped greatly within the upper lightness range
the constrained-naming task generally agree with the of samples called purple on the constrained-naming
results of prior studies of English color naming by task. The centroids for violet, like those for teal,
other investigators (Boynton & Olson, 1987, 1990; covered a large range of lightnesses, suggesting that
Sturges & Whitfield, 1995). Green and blue were the violet was generally not synonymous with lavender
only named categories that extended vertically and lilac.
throughout the lightness range. All the other basic In contrast, some color terms were used to name
color categories were confined to restricted ranges of quite different colors by different individual infor-
lightness: Pinks and yellows were light compared to the mants. For example, the centroids for tan (Figure 6b)
neutral background against which the colors were appeared in two disjoint areas of the color diagram:
viewed, and reds and browns appeared among the One area overlapped with the centroids for peach and
lower lightness warm colors; orange was at an beige, and the other was close to the centroids for olive.
intermediate lightness value. Similarly, informants used chartreuse to mean either
greenish-yellow or desaturated green (synonymous with
lime). Puce is also interesting, as all three of the
Color-term centroids informants who used this term applied it to yellowish-
green-colored samples. Apparently informants using
To examine the individual data, we calculated the puce did not know that it refers to a dark, highly
color-term centroids for each term for each infor- saturated purplish red or purplish-brown, and they
mant from the free-naming phase of the study assigned the color term instead to the color of vomit
(Figure 6). Each centroid (shown by the colored dots) (purple dots in Figure 6b).

Downloaded from jov.arvojournals.org on 06/23/2024

 Journal of Vision (2014) 14(2):17, 1–25 Lindsey & Brown 11

The nonbasic color terms: Partition or boundary disks in Figure 4; see Table 2) fall into the overlap
colors? region between 1 and 1.5 chips; the average distances
for the other colors were all cleanly divided between the
These data sets allowed us to examine the second BCTs, which were centered in their BCT categories,
conjecture and the two hypotheses about the origins of and the nonbasic color terms, which appeared at the
new color categories: the partition hypothesis of Kay et boundaries of the color terms. The three nonbasic
al. and the emergence hypothesis of Levinson. We terms that fell in the overlap region were lavender, lilac,
examined the locations of the centroids of the color- and lime. The average distance between the BCTs and
naming patterns in the free-naming data set relative to their nearest color boundary was 1.78 chips (SD ¼
the boundaries of the BCTs obtained from the 0.89); for the nonbasic color terms it was 0.65 chip (SD
constrained-naming data. If the partition hypothesis is ¼ 0.34), a statistically significant difference: t(7) ¼ 3.50,
correct, then each nonbasic color term will tend to be df ¼ 8, p ¼ 0.008.
located within one of the BCT categories, with its In summary, the nonbasic color terms generally
centroid at some distance from the nearest BCT appeared at the boundaries between the BCT catego-
boundary. In contrast, if the emergence hypothesis is ries. This result was broadly in agreement with
correct, and new terms intrude into the areas between Levinson’s emergence hypothesis. However, the color
named categories, then the centroids for the nonbasic terms for light purple (lavender and lilac) and light and
color terms will be located near the nearest BCT dark yellowish-green (lime and olive) appear to be
boundary, and the average distance to the nearest partition colors in the sense of Kay et al. These results
boundary will be near zero. show that both processes can occur, although the
For each color-naming pattern for each informant, intrusion of new colors in between established catego-
we calculated the unsigned distance between its ries may be more frequent in modern American
centroid and the nearest BCT boundary, expressed as English. Previous work by Sturges and Whitfield (1997)
the number of samples between the centroid and the has suggested qualitatively that British English might
closest boundary (above, below, to the right, or to the also have more intrusion colors than partition colors.
left of the centroid). Inasmuch as the centroids were not
integers, the separation between the centroids and the
boundaries were not integers either. Figure 7 shows the American English glosses
results of this analysis in two ways. In the line graphs,
the distance data from the informants who used a given Inspection of the list of color terms from the free-
color term were binned into half-chip bins. Each line naming phase of the experiment (Table 3a), and
shows, for a given color term, the number of examination of the individual data outlined previously,
informants who placed it within 0.5 chips of the nearest suggested that many of the terms that subjects used
boundary, between 0.5 and 1 chip from the nearest might be synonyms. Perhaps the very large number of
boundary, and so forth. Not surprisingly, the centroids color terms shown in Figures 2a and 3a would be much
of the BCTs were well centered within their respective smaller if those synonym groups were to be consoli-
categories (Figure 7a; see also Figure 6a), so the dated into larger color categories, much as Lindsey and
distances to their nearest boundaries were generally Brown (2006) did in their cluster analysis of the WCS.
greater than 1 chip. The closest bin (under 0.5-chip Therefore, we applied a similar k-means analysis to the
separation) was never the most frequent separation present data set. Briefly, we expressed each chromatic
between a BCT centroid and its nearest boundary. In color term (i.e., each term not used to name any of the
contrast, the distance data for 13 of the 17 most 10 achromatic color samples in the WCS chart)
frequently used nonbasic chromatic colors were at their deployed by each of the 51 informants as a 320-element
maximum within 1 chip of zero, as predicted by the binary feature vector, representing the 320 chromatic
emergence hypothesis. Figure 7b shows the results for color samples in the WCS chart. For a given color term
four representative nonbasic chromatic color terms. (say, yellow) used by a particular informant, each
However, the distance data of four of the 17 most element of the term’s feature vector was assigned the
common nonbasic terms peaked at a distance of more value 1 if that informant called the corresponding
than 1 chip. These were lavender, lilac, olive, and lime chromatic sample yellow and 0 otherwise. Applying this
(Figure 7c; compare to Figure 6c, arrows). encoding to all of the chromatic words used by our
The bar graph (Figure 7d) shows the average of the informants yielded a total of 963 binary feature vectors.
unsigned nearest boundary distances for each color We then performed a k-means cluster analysis,
term. The average distances for the BCTs and the computing a partition of the feature vectors into k ¼ 2
nonbasic chromatic color terms overlap only slightly: clusters, then k ¼ 3, then k ¼ 4, and so forth. The k-
Seven of the 25 most frequently used chromatic color means process works by assigning each feature vector
terms (the chromatic color terms shown with the white to the ‘‘nearest’’ cluster in feature space. We used a

Downloaded from jov.arvojournals.org on 06/23/2024

 Journal of Vision (2014) 14(2):17, 1–25 Lindsey & Brown 12

Figure 6. Centroids of named color categories provided by individual informants in the free-naming task. (a) Centroids of the 11 BCTs
of Berlin and Kay (1969), with the group average centroids (black disks) and the centroids of Sturges and Whitfield (1995; white disks).
Color key for the chromatic color terms is above and below the diagram. (b) Individual differences in usage of nonbasic color terms.
The centroids for tan are disjointly distributed in the warm-color region of the chart. Informants used several terms—some of which


Downloaded from jov.arvojournals.org on 06/23/2024

 Journal of Vision (2014) 14(2):17, 1–25 Lindsey & Brown 13

were uncommon—in the area between the green and blue BCTs: teal, turquoise, aquamarine, aqua, jade, ocean, and seafoam. The dark
purplish centroids within the green region (arrow) are for the color puce. (c) Centroids of the 17 most common color terms (Figure 4,
Tables 2 and 3a). Most nonbasic color terms in the free-naming task fall near the boundaries between the BCTs, where consensus for
the BCTs is low. However, lime and olive and lavender and lilac (arrows) are generally proper subsets of green and purple, respectively.
Color key for (b) and (c) is below (c), and the asterisks refer to the centroids indicated with arrows. (d) Informants used color terms
for the light colors peach, yellow, lime, lavender, and pink, but not for light blue. Color key above the diagram.

distance metric of (1  r) to determine nearness, where r

is the Pearson correlation coefficient between a feature
vector and a cluster centroid.
The k-means process is an iterative clustering
algorithm, and the initial positions of the centroids in
feature space must be assigned. A total of k feature
vectors, chosen at random from the data set, served as
the initial centroids at the start of each run. The resulting
k-means partition can be somewhat sensitive to these
initial conditions. Therefore, each k-means partition
reported here is based on the best of 100 replications per
value of k, where ‘‘best’’ was defined as the clustering
that produced the smallest within-cluster distances
among cluster members, summed across all k clusters.
All computations were performed in Matlab (Math-
Works, Natick, MA), using its ‘‘kmeans( )’’ function.
An important goal of our cluster analysis was to
determine an optimal value for k. At what value of k
are the feature vectors partitioned into their ‘‘natural’’
clusters, that is, the clusters that correspond to
distinctly different color terms? In particular, we
wondered whether there were more than the eight
chromatic BCTs specified by Berlin and Kay. This is a
difficult question to answer with cluster analysis
because there is no ground truth against which the
solution corresponding to a particular value of k can be
assessed (if there were such a ground truth, it would not
be necessary to perform the cluster analysis, and the
problem would be merely to classify the terms into their
known clusters).
To deal with this issue, we followed an approach
based on the gap statistic described by Tibshirani,
Walther, & Hastie (2001). This approach is based on a
comparison between the clustering of the data and the
corresponding clustering of a synthesized reference set
of feature vectors. The reference set is created by
sampling from a uniform distribution of vectors in
feature space that have all the statistical properties of
the data, except that by design they have no natural
clustering. For each k-means clustering of j data points
into k clusters, let Di(k) be the sum of the distances
between points within the ith cluster (i ¼ 1, . . ., k) and
the corresponding cluster centroid. Then TDD(k) is the Figure 7. The positions of the centroids of the chromatic color
total dissimilarities across all k clusters. Now, consider terms, expressed as the unsigned distance (in number of chips)
the corresponding clustering of a reference set of j to the nearest BCT boundaries. (a) The BCT centroids. (b–c) The
feature vectors into k clusters. The gap g(k) is the centroids of eight nonbasic color terms. (d) The average
difference between log TDD(k) and the mean of log locations of the centroids for the 25 chromatic color terms from
TDRn(k), obtained from clusterings of n sample sets of the free-naming data set on the upper limb in Figure 4.

Downloaded from jov.arvojournals.org on 06/23/2024

 Journal of Vision (2014) 14(2):17, 1–25 Lindsey & Brown 14

j feature vectors drawn from the uniform reference of kopt. The minimum value of kopt was 11, so there
distribution: were always at least three more clusters in the data set
    than the eight chromatic BCTs listed by Berlin and
gðkÞ ¼ log TDRn ðkÞ  log TDDðkÞ ð1Þ Kay. The mode of the kopt distribution (which was
close to its median and its mean) at kopt ¼ 17 was the
The gap statistic is based on the intuition that if a best estimate of the statistically significant chromatic
data set contains exactly kopt clusters, then g(k) will clusters in the free-naming data set.
increase for k  kopt, since k-means is doing an Thus, the first cluster analysis of the free-naming data
increasingly better job of reducing within-cluster yielded a glossary of 20 color categories (17 chromatic
distances in the data set as k approaches kopt when color categories plus black, white, and gray; Table 2).
compared to the distances obtained by clusterings of This glossary included nine more chromatic clusters in
the reference sets, which by design have no clusters. For the free-naming data set than there are BCTs, according
k . kopt, the partitions must split one or more of the to the first conjecture of Berlin and Kay. All informants
kopt clusters, and g(k) will not continue to improve and used more than 11 glossed color categories; the most
may even decline relative to g(kopt). frequent number was 18 (Figure 3b).
Let G(k) represent the change in gap between the kth Consensus diagrams of the 17 chromatic color
and the kth þ1 clustering, terms appear in Figure 9. In this article, we call them
GðkÞ ¼ gðk þ 1Þ  gðkÞ  sTDR ðk þ 1Þ; ð2Þ by the names most commonly used by informants
(above each diagram in Figure 9; see also Figure 2b).
where sTDR(k þ 1) is an error term related to the The second result was that even when all the colors
standard deviation of the log TDRn(k). Then kopt is were consolidated into their kopt clusters, none of the
defined as the largest k before the first zero crossing of nine statistically significant nonbasic chromatic cate-
G(k) (see Tibshirani et al., 2001, for details). Formally, gories was used by 100% of informants (Table 3b,
this rule is stated as follows: categories of rank 12–20). The clusters illustrated in
kopt ¼ argmaxfkjGðkÞ . 0g: ð3Þ Figure 9 were used in the motifs analysis described
k later in this article.
Among the virtues of the gap statistic is that it can The glossary derived from the k ¼ 17 solution varied
test for the absence of any clusters in the data set (i.e., slightly from run to run. Repeated runs of k-means
kopt ¼ 1). revealed essentially the same glossary, but occasionally
Our gap-statistic analysis was based on n ¼ 20 the lime cluster shown in Figure 9 was replaced by a
reference sets. To create a reference set, we took the 963 rust (dark red) category. Most k-means runs produced
color-naming patterns from our data set. The centroid cluster centroids that were all confined to contiguous
of each pattern was then randomly relocated to a new regions of feature space, like those illustrated in Figure
location in the coordinate frame of our WCS color- 9. Occasionally, however, one of the 17 consensus maps
sample space, and a feature vector for this new pattern derived from cluster analysis covered two disjoint
was created based on the new location. Thus, our regions of the color chart. One of the regions was
reference sets preserved exactly the distributions of the always more prominently represented than the other,
sizes and shapes of regions of color space associated and the corresponding cluster centroids were easily
with the color terms observed in the original data set. associated with one of the observed nonbasic color
However, in each reference set, the locations of the terms. In any event, the minor run-to-run perturbations
centroids of the feature vectors were drawn from a in the k-means derivation of the American English
uniform distribution of centroids falling within the glossary did not affect the main conclusions drawn
coordinate frame of the WCS color chart, and therefore from the motifs analysis discussed later.
had no natural clusters.
Preliminary studies indicated that despite adopting a
best-of-100-replications criterion for the clustering of American English motifs
our data, several independent clusterings of the data for
a given value of k still tended to produce small In their analysis of the WCS data set, Lindsey and
differences in TDD(k). In order to assess the impact of Brown (2009) found that the color-naming systems of
this variation on our gap-statistic analysis, we ran 1,000 individual informants around the world fell into about
separate analyses. For each analysis, we created k- four motifs, and that multiple motifs were present in the
means clusterings of the data for k ¼ 1, . . ., 25 and data sets of the great majority of the WCS languages. To
compared those to the corresponding clusterings on a determine whether these results also apply to American
new ensemble of 20 reference sets. English, we performed a second k-means analysis on the
Figure 8 shows the values of the gap statistic G(k) present data set, based on the glossary of 20 terms
over the 1,000 runs, and the inset shows the distribution revealed by the first k-means analysis.

Downloaded from jov.arvojournals.org on 06/23/2024

 Journal of Vision (2014) 14(2):17, 1–25 Lindsey & Brown 15

For each of the 51 informants, we created a feature

vector consisting of 20 elements, one element for each
of the glossed terms (the 17 chromatic color terms in
Figure 9 plus black, white, and gray; see Tables 2 and
3b). Each of the 20 elements had a value corresponding
to the fraction of samples (out of 330) in the WCS chart
assigned to that particular gloss by the first cluster
analysis. For example, if a given informant named 20
samples using one or more words synonymous with
teal, then the teal element of that informant’s feature
vector was assigned the value of 0.061 (20/330).
We performed a series of k-means analyses on the set
of 51 informants’ feature vectors to create partitions of
k ¼ 1, . . ., 5 clusters. There was almost no between-run
variation in the k-means solutions. To determine how
many clusters were statistically significant, we created
reference feature vectors in ensembles of 20 sets by
Figure 8. The gap statistic (see Equation 3) as a function of the creating scrambled versions of the informants’ feature
number k of chromatic clusters. Each individual line in the main vectors. We then used the results of the k-means
graph represents the results for one of 1,000 comparisons analysis of the informants’ feature vectors and the sets
between the k-means clusterings of the data for k ¼ 1, . . ., 25 of reference feature vectors to perform a gap-statistical
and corresponding clusterings of ensembles of 20 reference null analysis as outlined previously (Tibshirani et al., 2001).
sets. See text for details. Inset: histogram of kopt derived from This analysis revealed two statistically significant
the 1,000 functions shown in the main graph. This analysis motifs in American English. Figure 10 shows the
indicates that there are about 17 chromatic color terms in consensus maps for the two motifs, as well as the 0.80
American English (red arrow in main figure). threshold consensus maps. The informants whose data
fell into the first motif used primarily the eight chromatic

Figure 9. Consensus diagrams of the 17 chromatic color terms identified by the first k-means cluster analysis.

Downloaded from jov.arvojournals.org on 06/23/2024

 Journal of Vision (2014) 14(2):17, 1–25 Lindsey & Brown 16

Figure 10. The results of the second cluster analysis, based on the glossed color-naming patterns from the free-naming task. (a–b)
Seventy-three percent of 51 informants fell into a cluster that corresponded closely to the green–blue motif of Lindsey and Brown
(2009). This motif is similar to the Stage-VII pattern of Berlin and Kay (1969), being composed primarily of the 11 BCTs. (a) Consensus
map. (b) 0.8 threshold consensus map. (c–d) The remaining 27% of 51 informants fell into a second, green–teal–blue motif, which is
new here. The new motif included the high-consensus nonbasic terms maroon, peach, teal, and lavender. These new colors (see
arrows) appear in both the consensus map (c) and the 0.8 threshold consensus map (d).

BCTs of Berlin and Kay (plus black, white, and gray). performed a separate analysis employing 330 element
Thus, the first motif is similar to the green–blue (GB) vectors, where each element was assigned a nominal
motif in the WCS, which was so-named after the color value representing one of the 20 glossed terms. In this
terms corresponding to the cool colors (Lindsey & approach, which Lindsey and Brown (2009) used in their
Brown, 2009). Informants expressing this motif tended analysis of the WCS motifs, the dissimilarity metric was
to use the nonbasic color terms in the glossary a modified Jaccard coefficient (Leisch, 2006). In yet
idiosyncratically and with low frequencies. The infor- another version of the motif analysis, the 20-element
mants whose data fell into the second motif also used the feature vectors were populated with z scores representing
11 BCTs of Berlin and Kay, but they also used some deviations of each informant’s usage of each glossed
additional terms extensively and consistently, particu- term from the mean. The most striking result was that
larly teal, peach, lavender, and maroon. Figure 10d shows all these approaches agreed very well on the identity of
that consensus for each of these terms equaled or the first two motifs: a green–blue motif and a second
exceeded 0.8 for some of the color samples. We will call motif—green–teal–blue—with high informant usage of
this the green–teal–blue (GTB) motif because of the teal, peach, maroon, and lavender. Beyond two motifs,
names given to the cool colors. The GTB motif is new, the various cluster analyses mostly generated minor
and did not appear in the WCS analysis. Fourteen variations on the green–blue motif and variations of the
informants used the GTB motif, whereas 37 informants green–teal–blue motif that emphasized various subsets
used the GB motif. Partition of informants’ data into the of the four additional categories in the green–teal–blue
GB and GTB motifs increased color-naming consensus motif. The main differences between the various
from the overall value 0.74 for the free-naming data set approaches were in their statistical power. The 330-
as a whole to an average within-motif consensus of 0.79 dimension approach revealed only one statistically
for the glossed terms (GB consensus ¼ 0.81, GTB significant motif, whereas the 20-dimension approach
consensus ¼ 0.77). We created 10,000 partitions of the involving z scores revealed four statistically significant
informants’ data into two random ‘‘motifs’’ with 37 motifs. The 330-dimension approach from Lindsey and
individuals in one and 14 individuals in the other. The Brown (2009) worked on the WCS data set because of
highest average consensus from this simulation was 0.77. the enormous number of observations, where statistical
Therefore, the statistical significance of our k-means- power was not an issue. However, that approach was
based partition is p , 104. apparently underpowered for the present application.
The results shown in Figure 10 were remarkably Figure 10c and d also reveals that three of the four
robust, as they were essentially independent of the high-consensus nonbasic color terms appear in the low-
precise representation of informant feature vectors that consensus regions of the chart between the BCTs: Peach
was chosen, the dissimilarity metric, or even the size of appears in the dark, low-consensus area between pink,
the glossary (12  k  22) extracted from the first orange, yellow, and white; teal appears between green
cluster analysis. For example, in addition to the 20- and blue; and maroon appears between red and black.
element informant feature vectors described above, we These color terms evidently name new categories that

Downloaded from jov.arvojournals.org on 06/23/2024

 Journal of Vision (2014) 14(2):17, 1–25 Lindsey & Brown 17

arose between categories that previously existed, along color terms, whereas women used 12.3 nonbasic color
the lines proposed by Levinson, and were not the result terms. A multiple regression of the log-transformed
of partitioning existing color categories into smaller sets. frequency data (to normalize their distribution) against
In contrast, the samples called lavender in the free- age and gender revealed that this difference between men
naming task were a proper subset of the purple category, and women was statistically significant (r ¼ 0.304, p ¼
consistent with the partition hypothesis of Kay et al. 0.0301) but that age was not associated with the number
of nonbasic terms (r ¼ 0.079, p . 0.5). Apparently
women’s color vocabularies contained more terms than
The popularity of glossed color terms those of men. Women also distinguished more color
categories than men did and were more likely to use the
The number of informants using each of the terms motif that contained more color terms. In contrast to the
corresponding to the 20 glossed color categories from clear effects of gender, this data set showed no
the first k-means cluster analysis (Table 3b) appears as a statistically significant age effect.
function of the rank order of their popularity as the There is a sizable literature on the subject of
upper graph (triangles) in Figure 4. As before, the BCTs gender and color naming (e.g., DuBois, 1939; Now-
were at the ceiling and were fitted with a constant line. aczyk, 1982; Simpson & Tarrant, 1991; see Biggam,
There was a clear break in the function fitted to the 2012, for a recent review), which generally shows
nonbasic terms no matter how we dealt with ties in rank larger color vocabularies among women than among
ordering the data. The four most popular nonbasic men (but see also Machen, 2002; Sturges & Whitfield,
terms were fitted with a power law of exponent 0.79 1995). However, data like these do not indicate
(white triangles), whereas the remaining five terms were whether this difference is biological or social in
best fitted with exponent 2.3 (gray triangles). The four origin. On the biological side, Jameson, Highnote,
nonbasic categories on the second limb of the function and Wasserman (2001) have argued that women
were teal, peach, lavender, and maroon, the same terms identify more color categories than men do because
that appeared in the GTB motif. This clear distinction of well-understood sex-linked genetic differences in
between the popularity of the four new terms in the GTB their long- or middle-wavelength-sensitive (L or M)
motif and the remaining five statistically significant cone pigments (Nathans, Merbs, Sung, Weitz, &
terms provides additional evidence, independent of the Wang, 1992). Many heterozygous females carry the
second k-means motifs analysis, that those four addi- genes for four types of cone: In addition to the three
tional glossed terms are well integrated into the color normal cone pigments, some have the gene for an
lexicons of many informants. It also invites the additional (normal) L cone pigment, and about 10%
speculation that the color lexicon of American English is of females carry the gene for an additional (anoma-
currently undergoing change, and that those four terms lous) L or M cone pigment. According to Jameson et
are in the process of taking their place along with the al. (2001), women who are heterozygous for the two
original 11 BCTs of Berlin and Kay. versions of the L cone pigment may divide the
spectrum into more color bands than men or women
with only three cone-pigment genes. However, very
Gender, age, and color naming few women who are heterozygous for anomalous
trichromacy are actually tetrachromats, in the sense
We also examined the free-naming data set to of being able to use the normal and anomalous
determine whether there was an effect of age and pigments together to discriminate between colors
whether the American men and women in this sample (Jordan, Deeb, Bosten, & Mollon, 2010), although
differed in the number of terms in their color some apparently experience a subtle influence of their
vocabularies. Figure 11a shows the number of men and anomalous cones on color appearance under condi-
women using each of the nonbasic terms used by three tions where the influence of the normal cones is
or more informants. The terms in Figure 11a were minimized. Thus, while a well-documented L-cone
generally used more frequently by women than by men gene polymorphism might, in principle, provide a
[average difference ¼ 0.070, t(39) ¼ 2.50, p ¼ 0.017, two- basis for explaining some or all differences between
tailed]. When the nonbasic color terms were consoli- males and females in color naming, the behavioral
dated into categories, this gender difference persisted data obtained from heterozygous women do not
[Figure 11b; average difference ¼ 0.117, t(8) ¼ 2.72, p ¼ provide straightforward, unambiguous support for
0.027, two-tailed]. Furthermore, men and women were this explanation. Also on the biological side, there are
unevenly distributed across the two motifs (Figure 11c), other biological differences between males and
with women significantly less likely to use the GB motif, females, for example due to testosterone receptors in
and more likely to use the GTB motif, than men: t(49) ¼ the cerebral cortex, that may explain subtle, quanti-
2.30, p ¼ 0.026. On average, men used 9.71 nonbasic tative differences between men and women in the

Downloaded from jov.arvojournals.org on 06/23/2024

 Journal of Vision (2014) 14(2):17, 1–25 Lindsey & Brown 18

Figure 11. (a) The fraction of men and women who used each nonbasic color term shown in Figure 2a. Vertical dashed line: the break
point between the two power-law functions that were fitted to the disks in Figure 4 (Table 2). (b) The pattern of women using more
color terms persisted when the nonbasic color terms were consolidated into their corresponding glossed color categories. Vertical
dashed line divides the four nonbasic color terms (to the left of the line) in the second motif from the other nonbasic color terms; it is
also the break point between the two power-law functions fitted to data in Figure 4 (triangles). (c) The fraction of men and women
whose data fell into each of the two motifs. GB: the green–blue motif; GTB: the green–teal–blue motif. Error bars: 6 one standard
error of the dividing line between the two motifs.

appearance of colors (Abramov, Gordon, Feldman, a measurable influence on the appearance and
& Chavarga, 2012). The present difference between naming of colors.
men and women is probably larger than the subtle Some investigators have espoused the alternative
sex-related differences of color appearance that were view that women’s role in society as consumers of the
reported by Abramov et al. (2012). These two types decorative arts has honed their color discrimination
of biological difference between males and females into a finer sense of color appearance, resulting in
might have small combined effects that together exert superior color-category naming ability among women

Downloaded from jov.arvojournals.org on 06/23/2024

 Journal of Vision (2014) 14(2):17, 1–25 Lindsey & Brown 19

than among men (e.g., Rich, 1977; Swaringen, Layman, categories revealed that their corresponding color
& Wilson, 1978). The difficulty with the social terms were generally deployed to name colors that fell
hypothesis is that it does not make specific quantitative in the low-consensus regions between the chromatic
predictions that could be falsified. BCTs of the constrained-naming data set, suggesting
There is evidence in the field of sociolinguistics that an ‘‘emergence’’-like mechanism of color-term evolu-
language change begins with women and young people, tion. However, two terms, lime and lavender, were
especially in the lower-middle socioeconomic class, with proper subsets that partitioned their corresponding
the language of men, older people, and people of other BCTs, green and purple, suggesting a ‘‘partition’’-like
socioeconomic classes changing later (Labov, 1990; mechanism.
Tagliamonte & D’Arcy, 2009). The American men and A second cluster analysis, based on the glossary
women in this sample differed in the number of terms in derived in the first analysis, revealed that this sample
their color vocabularies, which provides further evidence of American English–speaking informants expressed
suggesting that the color lexicon of American English is two motifs. The first motif, expressed by 73% of
still changing. However, there was no reliable effect of informants, was similar to the green–blue (GB) motif
age. Furthermore, data on socioeconomic status were observed in the World Color Survey (Lindsey &
not collected, and our informants probably represented Brown, 2009). It was also similar to the color lexicon
a relatively narrow range along this dimension, so of BCTs listed by Berlin and Kay. The second, green–
language-change effects related to social class could not teal–blue (GTB) motif was expressed by 27% of
be examined. informants, and included four nonbasic terms that
It is not immediately clear that the gender difference were used with high consensus: peach, teal, lavender,
we report here is directly related to color. In addition to and maroon. Women were statistically more likely
the pervasive gender effect outlined by the sociolin- than men to use the GTB motif.
guists, Laws (2004) reported other domain-specific This prospectively designed study shows that the
differences in vocabulary size between the genders. key features of color naming in the WCS are general
These approaches suggest that the difference in the size to a written language spoken in the United States.
of the color lexicons of men and women might be a These key features are the diversity among the
specific instance of more general, and less-color-related, speakers of a single language in how colors are to be
phenomena. Of course, it is also possible that subtle named, and the way in which the language’s color
biological differences between males and females, terms are deployed across the range of colors that
combined with the social differences between men and any individual might encounter. Based on the
women, are jointly responsible for the reliable gender- published literature, one might suspect that diversity
related effect. Such a combined explanation would across individuals in their color vocabularies might
require a model with many free parameters, and it be restricted to unwritten languages such as those in
would be even harder to test quantitatively than either the WCS, which are spoken in nonindustrialized
explanation alone. societies far from Western influence. Contrary to
that supposition, prominent diversity in color
vocabulary among individuals who speak American
English persisted, even after cluster analysis consol-
Discussion idated the 122 color terms elicited in a free-naming
protocol into a glossary of 20 distinct named color
The data for this project were the color names categories. This indicates that diversity among
provided for 330 Munsell color samples by 51 native informants is common even in American English.
speakers of American English under two instructions: One might also suspect that this apparent within-
free naming, where any monolexemic color term could language diversity might be an artifact of each
be used, and constrained naming, where only the 11 informant’s haphazard color-term choices (from a
basic color terms (BCTs) of Berlin and Kay (1969) much larger color idiolect) on the spur of the
were allowed. This sample of informants used a total moment, and might not reflect true individual
of 122 color terms under free-naming instructions, differences in color cognition. On the contrary, the
with an average consensus of 0.74, and 11 color terms diversity reported here was not haphazard: Instead,
under constrained-naming instructions, with an aver- there were two distinct motifs, which repeated
age consensus of 0.85. When the free-naming data set themselves with minor variation across the idiolects
was subjected to a cluster analysis similar to that of of these 51 informants. Thus, the phenomenon of
Lindsey and Brown (2006), 20 statistically significant multiple within-language motifs observed in the
color categories were discovered: the BCTs of Berlin unwritten languages of the WCS also generalizes to
and Kay plus nine nonbasic chromatic categories. at least one written language spoken in an industri-
Examination of the centroids of these nine nonbasic alized society, namely American English.

Downloaded from jov.arvojournals.org on 06/23/2024

 Journal of Vision (2014) 14(2):17, 1–25 Lindsey & Brown 20

The first conjecture of Berlin and Kay: The BCTs understood to reveal an evolutionary trade-off between
and their universality the need of the speaker to spend the least effort
necessary in communication, at the risk of reduced
Berlin and Kay’s (1969) first conjecture was that clarity (producing a small vocabulary), and the need of
every language includes a vocabulary of no more than the hearer for conciseness and clarity in the received
11 BCTs, which are distinct from other ordinary color message (requiring a larger vocabulary; see Ferrer i
terms that an informant might use. In English, Berlin Cancho & Solé, 2003, for further discussion). Double-
and Kay’s BCTs are black, white, red, yellow, green, power-law behavior is common and is thought to reflect
blue, brown, orange, pink, purple, and gray. According two processes: one process that governs the creation of a
to Berlin and Kay, the BCTs in any language are kernel lexicon of limited size consisting of versatile
monolexemic, abstract terms that are used by all or words designed for general but imprecise communica-
nearly all competent speakers of that language, with tion, and the other generating an unlimited lexicon for
high consensus and consistency, to name the color of specific communication (Ferrer i Cancho & Solé, 2001).
any type of object, including color samples such as The unglossed data set shows a break in the function
those used in the present study. The 11 BCTs of Berlin between the regimen for the more popular color terms
and Kay were indeed used by all 51 informants here, (ranked 12–28) and the less-popular color terms (ranked
and they were the only terms that showed such full after 28). This discontinuity may be an instance of this
usage. This general result is certainly consistent with distinction between common color terms and those
Berlin and Kay’s first conjecture. chosen on the spur of the moment from a larger color
However, the free-naming data set shows several lexicon that each informant has in his or her mind but
other features that are less obviously consistent with does not use routinely in everyday communication. Even
Berlin and Kay’s first conjecture. First, only seven of the glossed data have a steep-slope section, which occurs
the BCTs were applied to any samples with 100% after four glossed categories. Taken together, the
consensus, omitting white, red, yellow, pink, and segmented structure of the frequency data suggested that
orange (Figure 5b), so the requirement that the 11 the BCTs are not the whole story when it comes to color
BCTs show high consensus is not perfectly observed. naming in American English: Even after consolidating
Second, every informant used at least 12 terms, and the data into glossed categories, at least four more
the modal number of terms was 18, so the minimum glossed color categories are commonly used and
number of terms is greater than 11. A third finding understood by many informants.
that challenges Berlin and Kay’s first conjecture is the
frequency-versus-rank power-law functions derived
from the popularity data. If the 11 BCTs of Berlin and The second conjecture: Color-term evolution
Kay were the only commonly used terms, and if the
nonbasic terms were entirely idiosyncratic in their use, Berlin and Kay’s (1969) second conjecture was that
the power-law functions should fall off very steeply for color lexicons evolve over time by adding new color
ranks greater than 11. Instead, both unglossed data terms. This evolution occurs as societies become
and glossed data (white disks and triangles, respec- technologically more complex, and as distinctions
tively, in Figure 4) fell off with steepness near 1.0 for among similar colors become more crucial in the
the first 17 unglossed and the first four glossed everyday lives of their individual members. We
nonbasic terms. The expected precipitous decline examined this data set to determine whether it provided
followed (gray disks and triangles in Figure 4). These evidence in favor of the ‘‘successive differentiation of
results suggest that there are about 28 common terms existing categories’’ specifically suggested by Berlin and
that are used and understood as part of the core color Kay’s partition hypothesis, against the alternative view,
vocabulary of American English; after glossing, 20 of which is Levinson’s emergence hypothesis (2000),
these terms are statistically significant, and about four whereby new color terms are added to cover hard-to-
of them seemed on their way to frequent use. Those name colors that for one reason or another have
four glossed terms are the key components of the GTB become particularly salient.
motif, which was expressed by 14 of the 51 informants. The data set reported here is certainly compatible
The flow of information from informant to listener is with the view that the color lexicon of American
important to understanding the use of color terms in English is currently evolving. For example, the high
communicating about color. Zipf’s law expresses the popularity of the four most common nonbasic color
reciprocal relationship between the frequency of inde- categories and their inclusion in the second motif
pendent observations (e.g., the usage of words in a suggest that those terms are on their way to joining the
language, the population of cities; in this case, the BCTs of Berlin and Kay to form a new lexicon of basic
popularity of color terms) and the rank order of those terms. However, it is logically impossible for the
frequencies. In the case of language, Zipf’s law is number of color categories to continue to increase ad

Downloaded from jov.arvojournals.org on 06/23/2024

 Journal of Vision (2014) 14(2):17, 1–25 Lindsey & Brown 21

infinitum. Yendrikhovskij (2001) proposed that 16 2007), modern Greek ghalazio (Thierry et al., 2009),
named color categories is on the high end of basic (some forms of) Spanish celeste (Bolton, 1978, p. 294),
color-lexicon size, on the basis of his information- and Farsi asamuni (Friedl, 1979)—as well as some non-
theoretic analysis of color terminology. Our Zipf’s-law Indo-European languages—(some forms of) Arabic
analysis also reveals a steep decrease in color-term celesti (Al-Rasheed et al., 2011; Borg, 2007), and
popularity beyond 15 glossed terms (triangles in Figure Turkish may distinguish dark blue from light blue,
4). Thus, 15 or 16 terms may be an upper limit on the lacivert versus mavi (Ozgen & Davies, 1998). Of the
size of basic color idiolects, at least in the milieu of common light color terms in English, pink and yellow
early-21st-century American English. are BCTs, peach is an ‘‘emergent’’ color, and lime and
The difference between the men and the women in lavender are ‘‘partition’’ terms. Thus, a light blue color
the present sample also suggests that the language term could appear either as a partition of blue or as a
related to color is changing, and that women are in the boundary term between blue and white. The fact that
vanguard. However, the lack of a reliable age effect in no common color term, basic or nonbasic, exists for
the present data set suggests that the American English light blue suggests that the ‘‘emergence’’ and ‘‘parti-
color lexicon is changing slowly compared to the age tion’’ mechanisms do not necessarily predict, univer-
range of our sample, and it does not suggest that sally, which color terms will occur. This case study
individual informants recapitulate the history of color- illustrates how little is really understood about how
lexicon change over their lifetimes. Without historical color terms are added to the lexicons of world
data collected using consistent methodology, it is not languages.
possible to examine these issues definitively. There has been considerable speculation over the
The results of these analyses provide some evidence years about what universal processes guide color-term
for both Levinson’s and Kay et al.’s (2010) views of evolution. Kay and his colleagues have emphasized the
color-term evolution. Much as Levinson suggested, 13 importance of universal aspects of the perceptual
of the 17 frequently used nonbasic color terms (from representation of color appearance; particularly, Kay et
Figure 4, listed in Table 2) appeared in the low- al. (2010) emphasized the salience of the Hering
consensus regions between the BCTs, and their fundamental hue sensations of red, green, blue, and
centroids were very near the nearest BCT boundaries yellow, plus black and white, which they supposed to
(Figures 6 and 7). For example, peach appears in the have a well-understood physiological basis. While this
hard-to-name region between orange, pink, white, and account is at least qualitatively in line with modern
yellow. In contrast, four of the 17 nonbasic color terms accounts of human color vision, the physiology
were clearly ‘‘partition’’ colors, as predicted by Kay et underlying the Hering sensations remains obscure
al. (2010). For example, lavender appeared as a proper (Lindsey & Brown, 2014). Furthermore, this account
subset of purple, suggesting that it partitions the large, does not provide any insight into the order in which
previously undifferentiated purple category into two color terms should appear as color lexicons change
smaller, more articulately named units. These results (however, see Ratliff, 1976). Recent theories that are
generally suggested that both ‘‘emergence’’ and ‘‘suc- based on human perception of color differences predict
cessive differentiation’’ probably occur as American that color terms should be added in a manner that
English adds new color terms. optimally partitions color space by minimizing color
Considering Berlin and Kay’s evolution conjecture, differences within the new contiguous color categories
it is instructive to examine the terms for the light while maximizing color differences between adjacent
colors. BCTs and nonbasic color terms exist in categories (Jameson & D’Andrade, 1987; Regier, Kay,
American English to name the light colors in much of & Khetarpal, 2007). The simulations by Regier et al.
the diagram: pink, peach, yellow, lime, and lavender (2007) based on this principle resemble Berlin and
(Figures 6d and 12). In contrast, there is no commonly Kay’s evolution trajectory of color names up to six
used, high-consensus word that means light blue (Table terms, but the authors do not show simulations for
3a), a finding consistent with Sturges and Whitfield’s lexicons greater than this number. Therefore, it is not
(1995) report of nonbasic-term usage by informants of clear that their simulations will generalize to more than
British English, and Boynton and Olson’s (1987, 1990) six terms. Interestingly, the explanation that Boynton
studies of American English. Sky, the closest term to and Olson (1987, 1990) give to explain the possible
light blue, is included in the unglossed data set in emergence of peach as a new BCT is similar to the
Figures 2b and 12, but it ranks 43rd in popularity in optimal partition principle of Regier et al.
this data set (Table 3a), and does not reach significance Other accounts of color-term evolution have focused
in the cluster analysis that created the glosses. In on the importance of the statistics of color in the
contrast to English, light blue has been reported to be a natural environment. Philipona and O’Regan (2006)
standard (perhaps basic) color term in some Indo- argued that the red, green, blue, and yellow universal
European languages—Russian goluboj (Winawer et al., color categories extracted from the World Color Survey

Downloaded from jov.arvojournals.org on 06/23/2024

 Journal of Vision (2014) 14(2):17, 1–25 Lindsey & Brown 22

color-term evolution: Clearly, not every conceivable

color term could come into being, and not every
possible motif could evolve at every possible juncture
in color-term change. These deterministic accounts
resemble theories of biological development in some
ways, where change occurs from one state to the next
along a single pathway, subject to limited variability
across individuals. Contrary to that view, we believe
that the trajectory of color-lexicon change is rather
more haphazard and evolution-like than is suggested
by the more development-like deterministic accounts
reviewed previously. It seems more likely that color-
lexicon change in general, and particularly the
addition of color terms beyond Berlin and Kay’s 11
BCTs, is better understood in light of the principles
that govern cultural evolution (Henrich, Boyd, &
Richerson, 2008). In this view, cultural change is
analogous to Darwinian evolution, where change from
a given state can occur along any one of many
trajectories, subject to a few well-understood princi-
ples (see, for example, Xu, Griffiths, & Dowman,
2013). Over time, these principles lead to changes in
Figure 12. The number of informants using terms for light red the relative prevalences of the different ‘‘species’’
(pink), light orange (peach), yellow, light green (lime), light blue (motifs) of color naming within a language commu-
(sky), and light purple (lavender), from Figure 2. All but light blue nity. Consistent with this view, the languages in the
are statistically significant color categories in the cluster WCS differ from one another in the relative frequency
analysis, and appear in the set of glossed color categories. of the expressed motifs (Lindsey & Brown, 2009), just
as isolated biological populations differ from one
another in the fraction of individuals with certain
are special because the early visual responses to the genetic alleles. Several recent studies—Steels and
light from the corresponding surfaces are least per- Belpaeme (2005), Komarova, Jameson, & Narens
turbed by changes in the environmental illuminant. On (2007), and Xu et al. (2013), among others—have
the basis of this account, one might expect all the attempted to model color-term evolution along these
Hering primary-color categories to appear consistently lines. Kay and his colleagues are increasingly pro-
early in color-term evolution. Contrary to that posing a less linear and deterministic (and less
expectation, the WCS data set shows many informants development-like), and more reticulate and stochastic
who use brown, gray, pink, or especially purple (and more evolution-like), account of color-term
categories, while lacking a blue category (Lindsey & evolution based on the WCS data set. Consistent with
Brown, 2009). Yendrikhovskij (2001) has shown that k- this view of color-term change as evolution, the results
means clustering (k ¼ 2, 3, . . ., 11) of color samples of the present study have provided clear evidence for
obtained from pictures of natural scenes and repre- both color-naming diversity among speakers of
sented in the CIELUV uniform chromaticity space American English and the organization of this
recapitulates Berlin and Kay’s evolutionary sequence of diversity into two distinct color-naming motifs.
color-term acquisition. However, Steels and Belpaeme Keywords: color naming, color categories, color
(2005) have shown that Yendrikhovskij’s analysis is appearance
very sensitive to the perceptual space in which color is
represented. Therefore, they suggest that Yendrikhov-
skij’s results are due more to his choice of color model
than to the statistics of the natural-scene samples Acknowledgments
themselves.
Thus there are many deterministic accounts of how Mr. Kevin Guckes assisted with data collection. We
human color perception, possibly coupled with color acknowledge A. C. Hurlbert and M. W. Ridley for
statistics in the natural environment, could guide helpful discussions, and the constructive suggestions of
color-term evolution. The factors that underlie these an anonymous reviewer. This project was supported by
accounts undoubtedly place important constraints on a Fred Brown Award from the Ohio State University

Downloaded from jov.arvojournals.org on 06/23/2024

 Journal of Vision (2014) 14(2):17, 1–25 Lindsey & Brown 23

Department of Psychology, the Ohio Lions Eye Categorisation in the Cognitive Sciences (pp. 224–
Research Foundation, and NSF BCS-1152841. 242). Amsterdam: Elsevier.
Davies, I., & Corgett, G. (1994). The basic color terms
Commercial relationships: none. in Russian. Linguistics, 32, 65–89.
Corresponding author: Delwin T. Lindsey.
DuBois, P. H. (1939). The sex difference on the color-
Email: lindsey.43@osu.edu.
naming test. American Journal of Psychology, 52(3),
Address: Delwin Lindsey, PhD, Professor of Psychol-
380–382.
ogy, Ohio State University-Mansfield, Mansfield,
Ohio, USA. Ferrer i Cancho, R., & Solé, R. V. (2001). Two regimes
in the frequency of words and the origins of
complex lexicons: Zipf’s law revisited. Journal of
Computative Linguistics, 8(3), 165–173.
References Ferrer i Cancho, R., & Solé, R. V. (2003). Least effort
and the origins of scaling in human language.
Abramov, I., Gordon, J., Feldman, O., & Chavarga, A. Proceedings of the National Academy of Sciences,
(2012). Sex and vision II: Color appearance of USA, 100(3), 788–791.
monochromatic lights. Biology of Sex Differences, Friedl, E. (1979). Colors and culture change in
2(1), article 21. southwest Iran. Language in Society, 8(1), 51–68.
Al-Rasheed, A., Al-Sharif, H. H., Thabit, M. J., Al- Gladstone, W. E. (1858). Studies on Homer and the
Mohimeed, N. H., & Davies, I. R. L. (2011). Basic Homeric age (Vol. III). London: Oxford University
colour terms in Arabic. In C. P. Biggam, C. A. Press.
Hough, C. J. Kay, & D. R. Simmons (Eds.), New
directions in colour studies (pp. 53–58). Amsterdam: Henrich, J., Boyd, R., & Richerson, P. U. (2008). Five
Benjamins. misunderstandings about cultural evolution. Hu-
man Nature, 19, 119–137.
Berlin, B., & Kay, P. (1969). Basic color terms: Their
universality and evolution. Berkeley, CA: University Jameson, K. A., & Alvarado, N. (2003). Differences in
of California Press. color naming and color salience in Vietnamese and
English. Color Research and Application, 28(2),
Biggam, C. P. (2012). Locating basic colors in the OSA 113–138.
space. The semantics of colour: a historical ap-
Jameson, K. A., & D’Andrade, R. G. (1987). It’s not
proach. Cambridge, UK: Cambridge University
really red, green, yellow, blue: An inquiry into
Press.
perceptual color space. In C. L. Hardin & L. Maffi
Bolton, R. (1978). Black, white and red all over: the (Eds.), Color categories in thought and language (pp.
riddle of color term salience. Ethnology, 17, 287– 295–319). Cambridge, UK: Cambridge University
311. Press.
Borg, A. (2007). Towards a history and typology of Jameson, K. A., Highnote, S., & Wasserman, L. (2001).
color categorization in colloquial Arabic. In R. E. Richer color experience in observers with multiple
MacLaury, G. V. Paramei, & D. Dedrick (Eds.), photopigment opsin genes. Psychonomic Bulletin
Anthropology of color (pp. 263–294). Amsterdam: and Review, 8(2), 244–261.
Benjamins. Jordan, G., Deeb, S. S., Bosten, J. M., & Mollon, J. D.
Boynton, R. M. (1997). Insights gained from naming (2010). The dimensionality of color vision in
the OSA colors. In C. L. Hardin & L. Maffi (Eds.), carriers of anomalous trichromacy. Journal of
Color categories in thought and language (pp. 135– Vision, 10(8):12, 1–19, http://www.journalofvision.
148). Cambridge, UK: University of Cambridge. org/content/10/8/12, doi:10.1167/10.8.12.
Boynton, R. M., & Olson, C. X. (1987). Locating basic [PubMed] [Article]
colors in the OSA space. Color Research and Kay, P., Berlin, B., Maffi, L., Merrifield, W. R., &
Application, 12, 94–105. Cook, R. (2010). The World Color Survey. Berke-
Boynton, R. M., & Olson, C. X. (1990). Salience of ley, CA: CSLI Publications.
chromatic basic color terms confirmed by three Kay, P., & McDaniel, K. (1978). The linguistic
measures. Vision Research, 30, 1311–1317. significance of the meanings of basic color terms.
Cook, R. S., Kay, P., & Regier, T. (2005). The world Language, 54, 610–646.
color survey database: History and use. In Cohen, Komarova, N. L., Jameson, K. A., & Narens, L.
Henri, & Claire Lefebvre (Eds.), Handbook of (2007). Evolutionary models of color categorization

Downloaded from jov.arvojournals.org on 06/23/2024

 Journal of Vision (2014) 14(2):17, 1–25 Lindsey & Brown 24

based on discrimination. Journal of Mathematical Regier, T., Kay, P., & Khetarpal, N. (2007). Color
Psychology, 51(6), 359–382. naming reflects optimal partitions of color space.
Labov, W. (1990). The intersection of sex and social Proceedings of the National Academy of Sciences,
class in the course of linguistic change. Language USA, 104(4), 1436–1441.
Variation and Change, 2, 205–254. Rich, E. (1977). Sex-related differences in colour
Laws, K. R. (2004). Sex difference in lexical size across vocabulary. Language and Speech, 20, 404–409.
semantic categories. Personality and Individual Rivers, W. H. R. (1901). Vision. In A. C. Haddon
Differences, 36, 23–32. (Ed.), Reports on the Cambridge anthropological
Leisch, F. (2006). A toolbox for K-centroids cluster expedition of the Torres Straits. Vol. II: Physiology
analysis. Computational Statistics and Data Analy- and psychology (pp. 1–132). Cambridge, UK:
sis, 51(2), 526–544. Cambridge University Press.
Levinson, S. C. (2000). Yélı̂ Dnye and the theory of Saunders, B. M., & Marth, I.-T. (2007). In B. M.
basic color terms. Journal of Linguistic Anthropol- Saunders (Ed.), I.-T. Marth (Trans.), The debate
ogy, 10, 3–55. about colour-naming in 19th century German philol-
Lindsey, D. T., & Brown, A. M. (2006). Universality of ogy. Leuven, Belgium: Leuven University Press.
color names. Proceedings of the National Academy Schontag, R., & Schafer-Priess, B. (2007). Color term
of Sciences, USA, 103, 16608–16613. research of Hugo Magnus. In R. E. MacLaury, G.
Lindsey, D. T., & Brown, A. M. (2009). World Color V. Paramei, & D. Dedrick (Eds.), Anthropology of
Survey color naming reveals universal motifs and color: Interdisciplinary multilevel modeling (pp. 107–
their within-language diversity. Proceedings of the 124). Amsterdam: John Benjamins.
National Academy of Sciences, USA, 206, 19785– Simpson, J., & Tarrant, A. W. S. (1991). Sex- and age-
19790. related differences in colour vocabulary. Language
Lindsey, D. T., & Brown, A. M. (2014). Color and Speech, 34(1), 57–62.
appearance, language, and neural coding. In J. S. Steels, L., & Belpaeme, T. (2005). Coordinating
Werner & L. M. Chalupa (Eds.), The new visual
perceptually grounded categories through lan-
neurosciences. Vol. 1 (pp. 511–521). Cambridge,
guage: A case study for colour. Behavioral and
MA: MIT Press.
Brain Sciences, 28(4), 469–489.
Lyons, J. (1995). Colour in language. In T. Lamb & J.
Bourrieau (Eds.), Colour: Art & science (pp. 194– Sturges, J., & Whitfield, T. W. A. (1995). Locating
224). Cambridge, UK: University of Cambridge. basic colours in the Munsell space. Color Research
and Application, 20, 364–376.
Machen, V. (2002). Color naming by boys and girls.
Perceptual and Motor Skills, 94, 348–350. Sturges, J., & Whitfield, T. W. A. (1997). Salient
features of Munsell colour space as a function of
Mitzenmacher, M. (2003). A brief history of generative monolexemic naming and response latencies. Vision
models for power law and lognormal distributions.
Research, 37(3), 307–313.
Internet Mathematics, 1(2), 226–251.
Swaringen, S., Layman, S., & Wilson, A. (1978). Sex-
Nathans, J., Merbs, S. L., Sung, C.-H., Weitz, C. J., &
differences in color naming. Perceptual and Motor
Wang, Y. (1992). Molecular genetics of human
visual pigments. Annual Review of Genetics, 26, Skills, 47(1), 440–442.
403–424. Tagliamonte, S. A., & D’Arcy, A. (2009). Peaks beyond
Nowaczyk, R. H. (1982). Sex-related differences in the phonology: Adolescence, incrementation, and lan-
color lexicon. Language and Speech, 25, 257–265. guage change. Language, 85(1), 58–108.
Ozgen, E., & Davies, I. R. L. (1998). Turkish color Thierry, G., Athanasopoulos, P., Wiggett, A., Dering,
terms: Tests of Berlin and Kay’s theory of color B., & Kuipers, J.-R. (2009). Unconscious effects of
universals and linguistic relativity. Linguistics, language-specific terminology on preattentive color
36(5), 919–956. perception. Proceedings of the National Academy of
Sciences, USA, 106(11), 4567–4570.
Philipona, D. L., & O’Regan, J. K. (2006). Color
naming, unique hues, and hue cancellation pre- Tibshirani, R., Walther, G., & Hastie, T. (2001).
dicted from singularities in reflection properties. Estimating the number of clusters in a data set via
Visual Neuroscience, 23, 331–339. the gap statistic. Journal of the Royal Statistical
Ratliff, F. (1976). On the psychophysical basis of Society: Series B, 63(2), 411–423.
universal color terms. Proceedings of the American Uchikawa, K., & Boynton, R. M. (1987). Categorical
Philosophical Society, 120(5), 311–330. color perception of Japanese observers: Compari-

Downloaded from jov.arvojournals.org on 06/23/2024

 Journal of Vision (2014) 14(2):17, 1–25 Lindsey & Brown 25

son with that of Americans. Vision Research, term universals. Proceedings of the Royal Society,
27(10), 1825–1833. Series B, 280(1758), 20123073.
Winawer, J., Witthoft, N., Frank, M. C., Wu, L., Yendrikhovskij, S. (2001). Computing color categories
Wade, A. R., & Boroditsky, L. (2007). Russian from statistics of natural images. Journal of
blues reveal effects of language on color discrimi-
nation. Proceedings of the National Academy of Imaging Science and Technology, 45, 409–417.
Sciences, USA, 104, 7780–7785. Zollinger, H. (1984). Why just turquoise? Remarks on
Xu, J., Griffiths, T. L., & Dowman, M. (2013). Cultural the evolution of color terms. Psychological Re-
transmission results in convergence towards colour search, 46(4), 403–409.

Downloaded from jov.arvojournals.org on 06/23/2024