ArticlePDF Available

Language, thought and color: Recent developments


Abstract and Figures

The classic issue of color naming and color cognition has been re-examined in a recent series of articles. Here, we review these developments, and suggest that they move the field beyond a familiar rhetoric of 'nature versus nurture', or 'universals versus relativity', to new concepts and new questions.
Content may be subject to copyright.
Language, thought, and color:
Whorf was half right
Terry Regier
and Paul Kay
Dept. of Psychology, University of Chicago
Dept. of Linguistics, University of California, Berkeley
The Whorf hypothesis holds that we view the world
filtered through the semantic categories of our native
language. Over the years, consensus has oscillated be-
tween embrace and dismissal of this hypothesis. Here,
we review recent findings on the naming and perception
of color, and argue that in this semantic domain the
Whorf hypothesis is half right, in two different ways: (1)
language influences color perception primarily in half
the visual field, and (2) color naming across languages is
shaped by both universal and language-specific forces.
To the extent that these findings generalize to other
semantic domains they suggest a possible resolution
of the debate over the Whorf hypothesis.
The state of the debate
A classic debate in cognitive science concerns the relation
between language and perception. At one pole of this
debate is the relativist stance, which holds that our per-
ception of the world is shaped by the semantic categories of
our native language, and that these categories vary across
languages with little constraint - a view often associated
with Benjamin Lee Whorf. At the other pole is the uni-
versalist stance, which holds instead that there is a uni-
versal repertoire of thought and perception that leaves its
imprint on the languages of the world. Over the years,
consensus has swung back and forth between these two
The domain of color has furnished an empirical locus
classicus of the debate for the last half-century. In a 2006
review [1] we suggested that the universalist-versus-rela-
tivist debate with respect to color often conflates 2 distinct
1. Do color terms affect color perception?
2. Are color categories determined by largely arbitrary
linguistic convention?
A relativist would respond ‘yes’ to both questions,
whereas a universalist would respond ‘no’ to both. How-
ever, that review argued that empirically the answers to
the two questions do not match. Specifically, it was argued
on the basis of evidence then available that color terms do
affect color perception (yes to question 1), but that there are
also universal tendencies in color naming (no to question
2). It was argued that the universalist/relativist opposition
is unhelpful as a conceptual structuring device, since it
does not accommodate these realities.
In the years since then, new findings have arisen that
suggest a subtler view. The new evidence suggests that
Whorf was partly right with respect to each of these two
questions. With respect to question 1, color names do
influence color perception but primarily in the right
visual field, and less so in the left. With respect to question
2, color naming across languages does reflect universal
tendencies, as shown in earlier work but also some degree
of local linguistic convention. These findings suggest a way
in which the recently re-opened debate over language and
thought in the color domain might be resolved. And to the
extent that these findings generalize to other semantic
domains, they suggest a possible resolution of the Whorf
debate more broadly.
Language affects perception - in half the visual field
Does language affect perception? As noted above, several
studies suggest that the answer is ‘yes’, at least in connec-
tion with color. These studies have shown that there is
‘categorical perception’ (CP: faster or more accurate dis-
crimination of stimuli that straddle a category boundary)
for color, and that differences in color category boundaries
between languages predict where CP will occur [27].
Moreover, several of these studies, and others [e.g. 8] have
shown that color CP disappears with a concurrent verbal
interference task, confirming that color CP is language
However, this straightforward Whorfian answer - and
the yes-or-no framing of the very question ‘does language
affect perception?’ - obscure an interesting possibility: that
language might affect half of perception. Specifically,
language might be expected to shape perception primarily
in the right visual field (RVF), and much less if at all in the
left visual field (LVF). This expectation follows from the
observations that the left hemisphere (LH) of the brain is
dominant for language, and that the visual fields project
contralaterally to the brain. On this view, half of our
perceptual world might be viewed through the lens of
our native language, and half viewed without such a
linguistic filter.
This proposal was first advanced and tested in a study
by Gilbert et al. [8] that probed the perceptual discrimi-
nation of colors straddling the boundary between green
and blue, a boundary present in English but absent in
many other languages. In this study, American English
speaking participants first saw a central fixation cross, and
Corresponding authors: Regier, T. (;
Kay, P. (
The two authors contributed equally to this work..
TICS-795; No of Pages 8
1364-6613/$ see front matter ß2009 Elsevier Ltd. All rights reserved. doi:10.1016/j.tics.2009.07.001 Available online xxxxxx 1
then a ring of colored squares arrayed around it (Figure 1b).
All of the squares were of the same color except for one (the
target), and participants were required to identify whether
the target appearedon the left or right half of the display by
pressing a button with the corresponding hand. Critically,
the target color had either the same nameas the color of the
other squares (e.g. a green against a background of a differ-
ent green, as in stimuluspair A,B in Figure 1a), or a different
name (e.g. a green against a background of a blue, as in
stimulus pair B,C in Figure 1a).
One of Gilbert et al.’s experiments is depicted in
Figure 1. As shown in Figure 1c, cross-category targets
were identified faster than same-category targets in the
RVF only. When a concurrent task requiring verbal
resources (remembering an eight-digit number) was
added, the RVF color CP effect disappeared, in fact was
reversed, reinforcing the interpretation that when RVF
color CP is found, it has a verbal basis [4,5]. In a second
experiment (not shown in Figure 1) Gilbert et al. replicated
these findings, and also found that RVF CP was not dis-
rupted by a non-verbal interference task of difficulty com-
parable to the verbal interference task - again
strengthening the conclusion that the lateralization pat-
tern is linguistic in origin. Drivonikou et al.[9] reanalyzed
by visual field color search data that had been collected
earlier for a different study and found predominantly RVF
color CP; they ran a new experiment similar to that of
Gilbert et al. and found color CP in the RVF superior to
that in the LVF. However their finding of significant,
although weaker, color CP in the LVF diverged from the
Gilbert et al. result; the longer response time in the Dri-
vonikou data suggest that this apparent RH CP might be
the result of longer RTs permitting cross-callosal transfer
and/or scanning (for further discussion of this possibility in
these and other studies see [7,10,11]).
Gilbert et al. [8] also reported that a callosotomy (‘‘split-
brain’’) patient showed RVF color CP with no trace of LVF
color CP. A separate study [12] reported a similar result
with a second callosotomy patient in an experiment in the
same paradigm as [8] but using non-color stimuli, namely
silhouettes of dogs and cats. When this experiment was run
with normal adult participants, CP was found in both
visual fields but again more strongly in the RVF than
LVF, and again this pattern was disrupted by a verbal
but not a non-verbal interference task. These results
confirm that linguistic CP is localized to the LH and
suggest that the phenomenon is not limited to color.
Roberson, Pak & Hanley [7] examined a color boundary
in Korean that does not correspond to any English color
boundary in a setup very similar to that of [8]. Their eight
fastest subjects showed color CP in the RVF only, whereas
their eight slowest subjects showed color CP in both visual
fields. Roberson et al. infer that the slower subjects’ LVF
CP probably reflects cross-callosal transfer. The literature
so far shows a general tendency for color CP to be restricted
to the RVF for short RTs, to appear in the LVF as well but
to a lesser degree in medium length RTs, and to show up
more or less equally in both VFs with very long RTs,
Figure 1. Lexical categories influence perception in the RVF. (a) Print-rendered versions of the four colors used. (b) Sample display for the visual search task. Participants
were required to press one of two response keys, indicating the side containing the target color. (c) In the no-interference condition, category effect in the RVF only. (d)
Effect reversed with verbal interference. *,P<0.05, two-tailed ttest; df =10; ns, nonsignificant. Values are mean SEM. Source: [8].
Review Trends in Cognitive Sciences No.x
TICS-795; No of Pages 8
reinforcing the conclusion that LVF CP in adults reflects
cross-callosal transfer, and perhaps scanning in some
cases. Roberson, Pak & Hanley [7] is the only published
study at this time to have tested lateralized color CP at a
color category boundary other than those found in English.
So far, everything we have considered points to
language-driven CP originating in the LH manifesting
itself in the RVF. However, this picture is complicated
by evidence of prelinguistic - and thus non-linguistic -
CP in infants [e.g. 13,14,15] and the finding that color
term knowledge does not affect color CP in toddlers [16].
The existence of prelinguistic CP has recently been both
contested [17,18] and defended [19]. If confirmed, it raises
an important question: What relation does prelinguistic
CP bear to the language-driven RVF-lateralized color CP
universally observed in speaking adults? Prelinguistic
color CP has been observed along boundaries correspond-
ing to some of those in familiar European (and many other)
languages. Thus, a natural possibility is that prelinguistic
categories serve as a starting-point for the later elabor-
ation of linguistic categories.
Franklin et al. [20] compared infant and adult perform-
ance on a visual search task much like that of [8]. Since
infants cannot reliably respond through a button press, a
different response measure was used: the time it took to
initiate an eye movement from central fixation to a target
dot, displayed in one color against a uniform background of
a different color. As in previous experiments the target and
background were either of the same category (e.g. two
blues) or of different categories (one green, the other blue).
This method was used with adults (with less easily dis-
criminable colors) and showed the expected RVF-dominant
CP of color. In contrast, prelinguistic infants showed no CP
in the RVF, and clear CP in the LVF.
Is the migration of the category effect - from the RH/LVF
in infants to the LH/RVF in adults - caused by the acqui-
sition of color terms? To pursue this question more precisely,
Franklin et al. [10] tested toddlers aged two to five with the
same procedure. The children’s grasp of the relevant color
terms blue and green was assessed and the toddler subjects
sorted readily into two groups: ‘‘learners’’, who had not yet
acquired these color terms, and ‘‘namers’’, who had. In the
visual search task the learners patterned like infants: they
displayed color CPin the LVF but not theRVF; whereas the
namers patterned like adults: CP in the RVF but not the
LVF. These results remain when age is added as a covariate
to the analysis, as shown in Figure 2.
This study suggests strongly that it is acquisition of
color terms per se that causes the shift of color CP from RH
to LH, a conclusion that raises several questions. What is
the nature of this shift? Are the RH prelinguistic categories
transported to the LH, where they serve as a starting-point
for the elaboration of LH linguistic categories? Or are LH
linguistic categories constructed de novo by language,
without reference to the RH prelinguistic categories? In
either case, why is there no trace left of the RH prelinguis-
tic categories once color terms have been learned - are they
permanently erased [21], or merely suppressed on-line by
We can presently answer only one of these questions,
and only tentatively. It seems likely that RH prelinguistic
categories are indeed permanently erased by language,
rather than merely temporarily suppressed by it. Tellingly,
split-brain patients, in whom the hypothetical channel of
suppression has been severed, show no evidence of LVF/
RH CP [8,12]. Further research is needed to settle this
question, along with the other questions opened by these
hemisphere-switching findings from infants and toddlers.
So far, our discussion of the brain bases of color CP, and
its lateralization, has been based on inference from beha-
vioral experiments. Recently however a number of ERP
and fMRI studies have probed this issue more directly.
Three studies deal with lateralization per se. Liu et al. [22]
performed a lateralized visual search experiment using the
procedure of [8], while recording brain activity by EEG.
They focused on a specific ERP (event-related potential)
component that is typically evoked in visual search for a
perceptually unique target among distractors. They found
that this component was present in the contralateral hemi-
sphere for both within- and cross-category targets - but
only in the left hemisphere did the amplitude of this
component for cross-category targets exceed that for
within-category targets, providing electrophysiological
confirmation of the (behaviorally established) lateraliza-
tion of the categorical perception of color. In a lateralized
event related fMRI study, Siok et al. [23] found that
discriminating colors of different lexical categories (versus
the same category) elicited faster and stronger response in
the left hemisphere language regions, especially when the
colors were presented in the right visual field. They found
further that only for these same stimuli was activation
significantly enhanced in the visual areas responsible for
color perception, suggesting that the language areas might
act as a top-down control source on the visual areas in color
perception. Another fMRI study [24] found that speeded
perceptual discrimination of colors which are easy to name,
relative to colors that are harder to name, activates left
hemisphere regions that mediate language processes:
these same regions were independently shown to be acti-
vated when subjects named colors aloud.
Figure 2. Category effect in the LVF for toddlers who have not yet learned the
words blue and green, and in the RVF for those who have. The dependent measure
is the estimated mean of the log transformation of initiation time (ms), estimated
for when variance due to age is accounted for. Error bars are within-subjects 95%
confidence intervals calculated by using the error term from the three-way
interaction. Source: [10].
Review Trends in Cognitive Sciences No.x
TICS-795; No of Pages 8
Three other recent ERP studies examine neural corre-
lates of color CP more generally, in the context of beha-
vioral tasks requiring the visual discrimination of colors.
Fonteneau & Davidoff [25] found that the peak latency in
ERP for a small color difference that bridged a color term
boundary (195 ms) was shorter than that for an equally-
sized color difference that did not bridge a color term
boundary (214 ms) - providing evidence for a neural corre-
late of categorical perception of color. Thierry et al. [26]
found that speakers of Greek, which has separate basic
color terms for light and dark blue, showed visual mis-
match negativity, an index of automatic and preattentive
change detection, for luminance modulations in the blue
area, whereas English speakers did not, a difference that
appeared at around 200 ms. These results imply a role of
the language-specific category boundary early in color
discrimination. Holmes et al.[27] compared English-
speaking subjects’ brain responses to same- versus cross-
category color differences; ERPs showed shorter latencies
for early components to cross-category differences than for
within-category differences, mirroring RT results on a
behavioral version of the task and providing evidence for
an early role for categorical differences in color perception.
Later components also differentiated between- and within-
category differences, suggesting influence of color
categories on post-perceptual processing as well.
Altogether, the recent electrophysiological and imaging
studies (1) support earlier behavioral findings of linguistic
category influence in colordiscrimination via involvement of
left hemisphere language regions, (2) demonstrate that CP
influence for color can be exerted both early and late in
processing, and is thus likely to be partially perceptual as
well as post-perceptual, and (3) show that lexical differences
between languages (Greek vs. English) are reliably repro-
duced in a standard ERP indication of pre-attentive novelty
of a stimulus. Also, (4) one study suggests that regions of the
left hemisphere that process language act as a top-down
influence on the function of visual areas in color perception.
More broadly, it now appears uncontroversial that once
language is learned, its categories shape perceptual dis-
crimination primarily in the left hemisphere/right visual
field and less so if at all in the right hemisphere/left visual
field - a specific and unexpected sense in which Whorf was
half right.
Where do color categories come from?
If linguistic color categories do affect perception, at least in
half the visual field, where do those categories come from?
Why do languages have the color categories they do? An
exploration of this question suggests another way in which
Whorf was half right.
According to an influential universalist view, color
categories across languages are organized around the uni-
versal focal colors black, white, red, green, yellow and blue
[2830]. This view gained support from the finding that
color memory appeared to be privileged for these colors,
even in speakers of a language with a color naming system
quite different from that of English [29,30]. For many years
the prevailing consensus was that these focal colors, and
prelinguistic color categories organized around them,
reflected in prelinguistic CP as discussed above, consti-
tuted a cognitive foundation for universals of color naming,
and the debate seemed settled.
Starting in 1999, however, the debate was re-opened by a
study of the Berinmo language (Sepik-Ramu family, Papua
New Guinea) by Debi Roberson and colleagues [31,32].In
this study, Roberson and colleagues failed to replicate the
finding of privileged memory for the proposed focal colors; in
addition, they found categorical perception of color at
language-defined boundaries. They concluded that there
are no universal foci, that categories therefore cannot be
organized around them, and that ‘‘color categories are
formed from boundary demarcation based predominantly
on language’’ [32], subject to the constraint of ‘grouping by
similarity’: namely, that categories must form contiguous
regions of color space. The implication is that apart from
that rather loose constraint, category boundaries are deter-
mined exclusively by local linguistic convention. These
authors held up Berinmo as a counterexample to universals
of color naming: ‘no evidence was found for [Berinmo color
categories] to correspond to a limited set of universal basic
color categories’ [33] and they interpreted their results as
providing ‘evidence in favor of linguistic relativity’ [32].
Thus, this study constituted a Whorfian challenge to the
then-reigning universalist consensus concerning why
languages have the categories they do.
Since then, considerable evidence has been advanced for
universal tendencies in color naming [e.g. 34,35], as well as
evidence supporting relativity [e.g. 6,17,36,37]. This evi-
dence has often been presented as unambiguously support-
ing one side or the other in the debate, in articles with titles
such as ‘Focal colors are universal after all’ [35] and ‘Color
categories vary with language after all’ [38]. Below we
briefly review recent contributions to this exchange, and
then discuss subsequent proposals that might help to
reconcile these two positions, and perhaps resolve the
Kay and Regier [39] compared the boundaries of Ber-
inmo color categories with those of the 110 languages of the
World Color Survey (WCS) [40]. They reasoned that if the
only constraint on color categories is that they must occupy
contiguous regions of color space, as had been suggested,
there should be nothing privileged about the location of
category boundaries in Berinmo the boundaries could
just as easily have been drawn elsewhere as long as each
category remained a connected region. To probe this, they
created artificial variants of the Berinmo color naming
system by rotating that system in the hue dimension by
2,4,... etc. hue columns, all the way around the hue circle,
as illustrated in Figure 3(a) - preserving the shape and
thus contiguousness of the categories, but altering their
actual location in color space. For each version of Berinmo,
it was determined to what extent category boundaries in
that system matched those in each language of the WCS. It
was found that the unrotated (attested) version of Berinmo
matched boundaries in the WCS better than did any
rotated (hypothetical) version of Berinmo, as shown in
Figure 3(b). This finding demonstrates that color naming
in Berinmo reflects universal forces that constrain the
location, not just the connectedness, of categories.
Lindsey and Brown [41] tested for universal tendencies
in color naming in a different way, using the same WCS
Review Trends in Cognitive Sciences No.x
TICS-795; No of Pages 8
data. They considered the color naming systems of individ-
uals, rather than summary measures capturing the systems
of entire languages, as previous analyses had used, and
performed cluster analyses based on these individual-level
data. They found clusters that ‘generally correspond to
readily identifiable English categories ... or their compo-
sites’. For instance, a cluster analysis with two clusters
revealed categories ‘readily described as WARM (including
red, yellow, orange, and pink) and COOL(including blue and
green)’ [41,p. 16609]. Cluster analyses with more clusters
revealed more fine-grained categories that again closely
resemble those isolated on a less rigorous basis in the earlier
literature [e.g. 42,43]. Further recent studies also establish
universal constraints on color naming [44,45].
However, there is also evidence consistent with the
relativist view that boundaries are determined by
language. Roberson, Davidoff, Davies, and Shapiro [6]
compared color naming and cognition in English, Berinmo,
and in Himba, a language of Namibia. They found that
Berinmo and Himba had similar but distinct systems of
color naming. On a universal-foci account, Himba and
Berinmo would receive the same analysis: each language
has terms that appear to be focused in black, white, green,
red, and yellow. Yet some of the boundaries among these
categories differ noticeably between the two languages,
and speakers of each language exhibit categorical percep-
tion of color at their native language’s boundaries. This
finding suggests strongly that there is more to the deter-
mination of boundaries than groupings of universal foci,
and that linguistic convention might indeed play some role
in determining where boundaries are drawn. This view is
confirmed by recent evidence from GreekEnglish bilin-
guals, in which the foci themselves shift their position
somewhat, under the influence of the categories of the
newly-learned language [37]. Finally, and most generally,
qualitative inspection of the 110 languages of the WCS
reveals some language-specific idiosyncrasies against the
backdrop of broad universal tendencies [46].
On the whole, then, the universalist and relativist views
regarding color naming are each partly supported and
partly challenged by existing empirical data. This empiri-
cally complex picture can be accounted for in theoretically
simple terms. Jameson & D’Andrade [47] suggested that
patterns of color naming could be attributed to irregula-
rities in the shape of perceptual color space combined with
general human cognitive tendencies toward constructing
informationally efficient systems of concepts:
One possible explanation [for patterns of color nam-
ing in languages] is...the irregular shape of the color
space. ...Hue interacts with saturation and lightness
to produce several large ‘‘bumps’’; one large bump is
at focal yellow, and another at focal red. ... We
assume that the names that get assigned to the color
space...are likely to be those names which are most
informative about color [47, p. 312].
This proposal can be seen as a natural generalization of
the universal-foci account, in which every color is focal to
some extent - the extent to which it protrudes from the
irregular outer skin of color space; the originally proposed
universal foci are then simply ‘more focal’ than other colors.
The appeal of the proposal lies in the possibility that
simple principles of categorization, operating over this
irregular surface, might account for both the universal
tendencies and the deviations from them that we see in
the world’s languages. Regier, Kay & Khetarpal [48]
formalized this idea and tested it against the WCS data.
They defined a well-formedness measure that captures the
extent to which a given categorical partition of the outer
skin of color space maximizes perceptual similarity within
color categories and minimizes it across categories [49].
Given this, they created theoretically optimal color naming
systems for n=3,4,5,6 categories, through steepest ascent
in well-formedness. The results are shown in Figure 4,
compared with selected color naming systems in the WCS.
This demonstrates that some languages have color naming
systems that are near optimal in this sense [cf. 50].
While other languages deviated considerably from these
templates, across the languages of the WCS there was a
strong tendency for color naming to be shaped in part by
well-formedness. Specifically, for each language in the
WCS, rotated variants were created as in Figure 3(a),
and for 82 of the 110 WCS languages the unrotated system
was more well-formed than any rotated variant thus
Figure 3.(a) The Berinmo color naming system (top) unrotated, and rotated four (middle) and eight (bottom) hue columns. Each colored region denotes the extensionofa
named color category, against a standard color naming grid. (b) Boundary match between each version of Berinmo, including the unrotated version, and WCS languages
overall. For each version of Berinmo, the dot shows the mean boundary match with all WCS languages, and the bar shows the 95% confidence interval of the mean. Source:
Review Trends in Cognitive Sciences No.x
TICS-795; No of Pages 8
well-formedness provides an explanation for universal
tendencies. At the same time, this account also suggests
where linguistic convention might get some wiggle room.
There are often several similar but different partitions that
are roughly equally well-formed. Hence, local factors, in-
cluding arbitrary linguistic convention, might select from
among a group of highly-ranked systems that differ some-
what from each other - such as Berinmo and Himba. This
recent account suggests a middle ground between nature
and nurture in color naming across languages, and high-
lights another sense in which Whorf might have been half
A competing explanation holds that color naming pat-
terns reflect the statistical distribution of colors in the
world, rather than the shape of perceptual color space in
our minds. Yendrikhovskij [51] showed that k-means clus-
tering of colors obtained from images of natural scenes
produced color clusters similar to those observed in the
world’s languages. Is color naming shaped by exogenous
color diet, or by endogenous perceptual structure? The two
explanations might be partially reconcilable. Shepard
[52,53] argued that universal aspects of perceptual-cogni-
tive structure might correspond to invariants in the
environment, internalized over the course of evolution
[cf. 54]. Thus, it is possible that color naming reflects
the structure of perceptual color space in our minds - which
in turn reflects the distribution of colors in the world.
There is an obvious fact that our discussion so far has
not engaged: color naming systems are used for the social
process of communication about color, and they are passed
on socially from one generation to the next within a speech
community. What is the role of such social forces in
accounting for patterns of color naming? This question
has been increasingly addressed recently, often in agent-
based simulations [5561]. For example, Steels and Bel-
paeme [55] argued on the basis of such simulations that
color categories might in principle reflect both perceptual
structure and social coordination through language; Dow-
man [57] accounted for patterns of color naming in the
world’s languages in broadly similar terms, through cul-
tural evolution in an idealized color space; and Komarova
et al.[58] show that, given certain simple assumptions, a
population of agents communicating about color will con-
verge to a system of near-optimal color categories.
Of particular interest in such simulations is the ques-
tion of how much explanatory force is carried by social
coordination and transmission per se, and how much by the
individual agents’ biases and expectations concerning the
color domain. In other words, how much of the explanation
lies between individuals, and how much within individ-
uals? A possible answer is suggested by some recent work
by Griffiths and Kalish [62]. They presented a Bayesian
analysis of the social transmission of language, and showed
that under certain assumptions, this process converges to a
distribution over languages that reflects the shared prior
bias of each learner in the social chain. Thus, the eventual
effect of socially transmitting language across generations
is that language itself takes on the form of the learning bias
in each learner’s mind; in the case of color naming, this
would be our prior expectations concerning the shape of
color categories. And as argued above, this mental struc-
ture might in turn reflect invariants in the distribution of
colors in the world. Taken as a whole, such an account
might help to explain the success of some models that are
based on color diet, or on the mental representation of
color, with no social component at all: environmental
Figure 4. Theoretical optima (left) for n=3,4,5,6categories, compared withcolor naming systems (right,top to bottom) of Ejagam (Nigeria/Cameroon),Culina (Peru/Brazil), Iduna
(Papua New Guinea), Buglere (Panama). All results mapped against a standard color grid, in which columns denote hues and rows denote lightness levels. Source:[48].
Review Trends in Cognitive Sciences No.x
TICS-795; No of Pages 8
structure might be internalized in the mind over evolution-
ary time, and then externalized in the form of language
through iterated social transmission, with the result that
models capturing this structure either in the environment,
or in the mind, can fit color naming data well. At the same
time, there are insights that seem less likely to be easily
capturable in these terms. For example, Jameson and
Komarova [61] examined agent populations in which not
all agents had the same biases, capturing various mixtures
of normal and anomalous color vision. They concluded that
a rather small percentage of agents with anomalous color
vision could greatly affect the color category boundaries of
the population - suggesting a possible explanation for some
of the cross-language variation observed.
We have reviewed two broad recent findings here: that
language affects color perception primarily in the right
visual field probably via activation of language regions of
the left hemisphere, and that color naming reflects both
universal and local determinants. Neither of these findings
is anticipated by the traditional universalist-versus-rela-
tivist framing of the debate over language and perception,
and neither sits particularly comfortably with it. Instead,
these findings suggest novel perspectives on the relation of
language and perception. There is some evidence already
available that these findings generalize to domains beyond
color. The replication of the initial results on the latera-
lized Whorf effect in color [8,9] to dog and cat silhouettes
[12] indicates that RVF-lateralized categorical perception
might not be restricted to the color domain (and ongoing
research suggests that optimization of well-formedness
might generalize to the spatial domain). Just which of
the findings reported here for color will generalize to other
domains of language or to cognition more broadly, and
which are restricted to color, should furnish significant
research questions in the future.
This work was supported by NSF grants 0418283 (to TR) and 0418404 (to
1 Kay, P. and Regier, T. (2006) Language, thought, and color: Recent
developments. Trends in Cognitive Sciences 10 (2), 5154
2 Kay, P. and Kempton, W. (1984) What is the Sapir-Whorf hypothesis?
American Anthropologist 86, 6579
¨zgen, E. and Davies, I.R.L. (1998) Turkish color terms: Tests of Berlin
and Kay’s theory of color universals and linguistic relativity.
Linguistics 36, 919956
4 Roberson, D. and Davidoff, J. (2000) The categorical perception of
colours and facial expressions: The effect of verbal interference.
Memory and Cognition 28, 977986
5 Winawer, J. et al. (2007) Russian blues reveal effects of language on
color discrimination. Proceedings of the National Academy of Sciences
104 (19), 77807785
6 Roberson, D. et al. (2005) Colour categories in Himba: Evidence for the
cultural relativity hypothesis. Cognitive Psychology 50, 378411
7 Roberson, D. et al. (2008) Categorical percept ion of colour in the left and
right visual field is verbally mediated: Evidence from Korean.
Cognition 107, 752762
8 Gilbert, A.L. et al. (2006) Whorf hypothesis is supported in the right
visual field but not the left. Proceedings of the National Academy of
Sciences 103, 489494
9 Drivonikou, G.V. et al. (2007) Further evidence that Whorfian effects
are stronger in the right visual field than the left. Proceedings of the
National Academy of Sciences 104, 10971102
10 Franklin, A. et al. (2008) Lateralization of categorical perception of
color changes with color term acquisition. Proceedings of the National
Academy of Sciences 105 (47), 1822118225
11 Kay, P. et al. (2009) Lateralized Whorf: Language influences perceptual
decision in the right visual field. In Language, Evolution, and the Brain
(Minett, James W. and Wang, William S.-Y., eds), The City University
of Hong Kong Press, (Hong Kong
12 Gilbert, A.L. et al. (2008) Support for lateralization of the Whorf effect
beyond the realm of color discrimination. Brain and Language 105, 91
13 Bornstein, M.H. et al. (1976) Color vision and hue categorization in
young human infants. Journal of Experimental Psychology: Human
Perception and Performance 2 (1), 115129
14 Franklin, A. and Davies, I.R.L. (2004) New evidence for infant color
categories. British Journal of Developmental Psychology 22, 349377
15 Franklin, A. et al. (2005) The nature of infant color categorization:
Evidence from eye movements on a target detection task. Journal of
Experimental Child Psychology 91, 227248
16 Franklin, A. et al. (2005) Color term knowledge does not affect
categorical perception of color in toddlers. Journal of Experimental
Child Psychology 90, 114141
17 Goldstein, J. et al. (2009) Knowing color terms enhances recognition:
Further evidence from English and Himba. Journal of Experimental
Child Psychology 102, 219238
18 Davidoff, J. et al. (2009) Nature vs. nurture: The simple contrast.
Journal of Experimental Child Psychology 102, 246250
19 Franklin, A. et al. (2009) What can we learn from toddlers about
categorical perception of color? Comments on Goldstein, Davidoff,
and Roberson. Journal of Experimental Child Psychology 102, 239245
20 Franklin, A. et al. (2008) Categorical perception of color is lateralized to
the right hemisphere in infants, but to the left hemisphere in adults.
Proceedings of the National Academy of Sciences 105, 32213225
21 Hanley, R. and Roberson, D. (2008) Do infants see colors differently?
Scientific American May 14
22 Liu, Q. et al. (2009) The N2pc component in ERP and the lateralization
effect of language on color perception. Neuroscience Letters 454, 5861
23 Siok, W-T. et al. (2009) Language regions of brain are operative in color
perception. Proceedings of the National Academy of Sciences 106, 8140
24 Tan, L-H. et al. (2008) Language affects patterns of brain activation
associated with perceptual decision. Proceedings of the National
Academy of Sciences 105, 40044009
25 Fontenau, E. and Davidoff, J. (2007) Neural correlates of colour
categories. NeuroReport 18, 13231327
26 Thierry, G. et al. (2009) Unconscious effects of language-specific
terminology on preattentive color perception. Proceedings of the
National Academy of Sciences 106, 45674570
27 Holmes, A. et al. (2009) Neurophysiological evidence for categorical
perception of color. Brain and Cognition 69, 426434
28 Berlin, B. and Kay, P. (1969) Basic Color Terms: Their Universality and
Evolution. University of California Press, (Berkeley)
29 Heider, E.R. (1972) Universals in color naming and memory. Journal of
Experimental Psychology 93, 1020
30 Heider, E.R. and Olivier, D.C. (1972) The structure of the color space in
naming and memory for two languages. Cognitive Psychology 3,
Outstanding questions:
In what other semantic domains, if any, are categories near-
In what other semantic domains, if any, does language affect
perception in half the visual field?
What becomes of prelinguistic color categories in the right
hemisphere, and by what mechanism?
Are linguistic color categories elaborations of prelinguistic
Are non-universal aspects of color naming attributable solely to
arbitrary linguistic convention, or to environmental or other non-
universal forces?
Review Trends in Cognitive Sciences No.x
TICS-795; No of Pages 8
31 Davidoff, J. et al. (1999) Colour categories in a stone-age tribe. Nature
398, 203204
32 Roberson, D. et al. (2000) Colour categories are not universal:
Replications and new evidence from a Stone-age culture. Journal of
Experimental Psychology: General 129, 369398
33 Roberson, D. et al. (2002) Colour categories are not universal:
Replications and new evidence. In Theories, technologies,
instrumentalities of color: Anthropological and historical perspectives
(Saunders, B. and van Brakel, J., eds), pp. 2535, Lanham, Maryland,
University Press of America Inc
34 Kay, P. and Regier, T. (2003) Resolving the question of color naming
universals. Proceedings of the National Academy of Sciences 100, 9085
35 Regier, T. et al. (2005) Focal colors are universal after all. Proceedings
of the National Academy of Sciences 102, 83868391
36 Agrillo, C. and Roberson, D. (2009) Colour language and colour
cognition: Brown and Lenneberg revisited. Visual Cognition 17,
37 Athanasopoulos, P. (2009) Cognitive representation of colour in
bilinguals: The case of Greek blues. Bilingualism: Language and
Cognition 12, 8395
38 Roberson, D. and Hanley, R. (2007) Color vision: Color categories vary
with language after all. Current Biology 17, R605R607
39 Kay, P. and Regier, T. (2007) Color naming universals: The case of
Berinmo. Cognition 102, 289298
40 Cook, R.S. et al. (2005) The World Color Survey database: History and
use. In Handbook of Categorization in Cognitive Science (Cohen, H.
and Lefebvre, C., eds), pp. 223242, Amsterdam, Elsevier
41 Lindsey, D.T. and Brown, A.G. (2006) Universality of color names.
Proceedings of the National Academy of Sciences 103, 1660916613
42 Kay, P. et al. (1997) Color naming across languages. In Color Categories
in Thought and Language (Hardin, C.L. and Maffi, L., eds), pp. 2156,
Cambridge, Cambridge University Press
43 Kay, P. and Maffi, L. (1999) Color appearance and the emergence and
evolution of basic color lexicons. American Anthropologist 101, 743760
44 Kuehni, R.G. (2007) Nature and culture: An analysis of individual focal
color choices in World Color Survey languages. Journal of Cognition
and Culture 7, 151172
45 Webster, M.A. and Kay, P. (2007) Individual and population
differences in focal colors. In Anthropology of Color:
Interdisciplinary multilevel modeling (MacLaury, R.E., Paramei,
G.V. and Dedrick, D., eds), pp. 2953, Amsterdam, John Benjamins
46 Kay, P. et al. (in press). The Worl d Color Survey. Center for the Study of
Language and Information: Stanford University. Stanford, CA
47 Jameson, K. and D’Andrade, R. (1997) It’s not really red, green, yellow,
blue: An inquiry into cognitive color space. In Color categories in
thought and language (Hardin, C.L. and Maffi, L., eds), pp. 295319,
Cambridge, UK, Cambridge University Press
48 Regier, T. et al. (2007) Color naming reflects optimal partitions of
color space. Proceedings of the National Academy of Sciences 104, 1436
49 Garner, W.R. (1974) The Processing of Information and Structure.
Erlbaum, (Potomac, MD)
50 Griffin, L.D. (2006) The basic colour categories are optimal for
classification. Journal of the Royal Society: Interface 3, 7185
51 Yendrikhovskij, S. (2001) Computing color categories from statistics of
natural images. Journal of Imaging Science and Technology 45, 409
52 Shepard, R. (1992) The perceptual organization of colors: An
adaptation to regularities of the terrestrial world? In The Adapted
Mind: Evolutionary Psychology and the Generation of Culture (Barkow,
J.H., Tooby, J. and Cosmides, L., eds), pp. 495532, New York, Oxford
University Press
53 Shepard, R. (2001) Perceptualcognitive universals as reflections of the
world. Behavioral and Brain Sciences 24, 581601
54 Regan, B.C. et al. (2001) Fruits, foliage, and the evolution of primate
colour vision. Philosophical Transactions of the Royal Society B:
Biological Sciences 356, 229283
55 Steels, L. and Belpae me, T. (2005) Coordinating perceptually grounded
categories through language: A case study for colour. Behavioral and
Brain Sciences 28, 469489
56 Belpaeme, T. and Bleys, J. (2005) Explaining universal color
categories through a constrained acquisition process. Adaptive
Behavior 13, 293310
57 Dowman, M. (2007) Explaining color term typology with an
evolutionary model. Cognitive Science 31, 99132
58 Komarova, N., Jameson, K. and Narens, L. (2007) Evolutionary models
of color categorization based on discrimination. Journal of
Mathematical Psychology 51, 359382
59 Komarova, N. and Jameson, K. (2008) Population heterogeneity and
color stimulus heterogeneity in agent-based color categorization.
Journal of Theoretical Biology 253, 680700
60 Jameson, K. and Komarova, N. (2009a). Evolutionary models of color
categorization. I. Population categorization systems based on normal
and dichromat observers. Journal of the Optical Society of America, 26,
61 Jameson, K. and Komarova, N. (2009b). Evolutionary models of color
categorization. II. Realistic observer models and population
heterogeneity. JournaloftheOpticalSocietyofAmerica,26,
62 Griffiths, T.L. and Kalish, M.L. (2007) Language evolution by iterated
learning with Bayesian agents. Cognitive Science 31, 441480
Review Trends in Cognitive Sciences No.x
TICS-795; No of Pages 8
... Second language colours also effects the colour perception of the speakers. And this shows that for colours perception stud Kay, & Regier, (2006), studied colour categorization in 2006 they finalized that Sapir Whorf hypothesis is half proved right and half is not proved. They said that Sapir Whorf Hypothesis is that we filtered the world thorough our language that is used by the speaker as a native language. ...
This study explores the impact of language on temporal progression, color perception, and categorization in L1 Punjabi speakers, focusing on gender and education level differences. The goal of the study is to investigate whether language shapes our thoughts. 60 L1 speakers were selected by using simple random sampling method. Data from 60 L1 speakers was taken by giving them some pictures. The researcher found that educated male speakers were less proficient at naming colors compared to uneducated males, while female speakers were better at color perception and naming than educated females. The study concludes that the L1 language greatly influences how we perceive and categorize colors and temporal progression, and that gender differences also play a role in this phenomenon. The study suggests that language shapes our worldview and gender influences perception of the same concept, indicating the importance of considering language and gender in communication.
... This is an abstraction or words contained in the meaning. The process of speaking and writing is a cerebral process which means the process of verbal expression, comprehension, and competence formed by the brain's nerve cells in neurons (Kay & Regier, 2006;Regier & Kay, 2009). ...
Full-text available
In essence, in communication activities, there is a process of producing and understanding speech. It can be said that psycholinguistics is the study of mental mechanisms that occur in people who use language, both when producing or understanding speech. In other words, in language use, there is a process of changing thoughts into codes and changing codes into thoughts. Speech is a synthesis of the process of converting concepts into code while understanding the message is the result of code analysis. Language as a form or result of a process and as something that is processed in the form of spoken or written language psycholinguistics is the study of humans as language users, namely the study of language systems that exist in humans who can explain how humans can capture other people's ideas and how they can express their ideas through language, either in writing or orally. Language skills that must be mastered by someone, this is related to language skills, namely listening, speaking, reading, and writing.
... In subsequent decades, the Linguistic Relativity Hypothesis (LRH) was developed, which advances the Sapir-Whorf hypothesis by stating that both human cognition and behavior can be shaped by languages. While LRH was regarded as misguided by some linguists and cognitive scientists , a substantial and ever-growing body of literature emerged starting in the 1990s testifying to the validity of the theory (Levinson and Wilkins 2006;Kay and Regier 2006;Boroditsky et al. 2003;Slobin 2003;Levinson 1996). ...
Full-text available
Using cross country data, we explore the role of linguistic structures for the financial development of countries. Specially, we investigate if future time reference (FTR), the requirement of an obligatory future tense marking in languages, matters for financial development or not. Our results show that countries speaking weak FTR language or a language not needing a dedicated future tense marking have enhanced financial development relative to countries speaking strong FTR language. Discounting the future less or having a connection between the present and the future—characteristics of weak FTR languages—has implications for caring about saving and investment, having efficient property rights, protection of shareholders and cost of acquiring information. Our results are robust to multiple measures of financial development and inclusion of determinants of the same. Finally, results show that weak FTR language speaking countries benefit more when their financial development is low.
... Studying crosslinguistic differences can provide insight into the feasibility of a universal cognitive theory of coherence relations. 9. Consider findings from colour research: some languages distinguish between certain colour concepts that other languages do not distinguish (see, e.g., Kay & Regier, 2006;Roberson et al., 2005). Of course, colour research is not located in the same domain as discourse coherence research, but it illustrates that languages and cultures do not necessarily conceptualize the same distinctions. ...
Full-text available
A central issue in linguistics concerns the relationship between theories and evidence in data. We investigate this issue in the field of discourse coherence, and particularly the study of coherence relations such as causal and contrastive. Proposed inventories of coherence relations differ greatly in the type and number of proposed relations. Such proposals are often validated by focusing on either the descriptive adequacy (researcher’s intuitions on textual interpretations) or the cognitive plausibility of distinctions (empirical research on cognition). We argue that both are important, and note that the concept of cognitive plausibility is in need of a concrete definition and quantifiable operationalization. This contribution focuses on how the criterion of cognitive plausibility can be operationalized and presents a systematic validation approach to evaluate discourse frameworks. This is done by detailing how various sources of evidence can be used to support or falsify distinctions between coherence relational labels. Finally, we present methodological issues regarding verification and falsification that are of importance to all discourse researchers studying the relationship between theory and data.
... See e.g. Lucy (1996); Slobin (2003); Boroditsky et al. (2003); Oh (2003); Levinson and Wilkins (2006); Kay and Regier (2006); Winawer et al. (2007) gender norms. As a consequence, existing gender norms, which often legitimise IPV, are being reinforced, and may acquire a greater influence on the behaviour of speakers of a gendered language. ...
Full-text available
This study establishes the influence of sex-based grammatical gender on gendered violence. We demonstrate a statistically significant relationship between gendered language and the incidence of intimate partner violence in a cross-section of countries. Motivated by this evidence, we conduct an individual-level analysis exploiting the differences in the language structures spoken by individuals with a shared religious and ethnic background residing in the same country. We show that speaking a gendered language is associated with the belief that intimate partner violence is justifiable. Our results are consistent with the theoretical possibility that gendered language activates gender schemata in the minds of speakers, increasing the salience of gender distinctions and existing gender norms which legitimize gendered violence.
Robert Motherwell is regarded as one of the great American abstract expressionists. He was highly intelligent and articulate about his art. In this essay, I explore the thesis that the ability to make fine category discriminations, which can be indexed by language, is necessary to produce great art. I argue that Motherwell might not have been as great an artist if he were not so articulate. Relying on a constructivist view, I argue that fine-grained categories of human emotions can be represented in language; language carves out affective space in a way that makes these states explicit and easier to communicate. Ineffability in art implies exhausting the effable. Being articulate about emotions allows one to reach for higher states of ineffability and aspire to great art.
Purpose Through examination of the Library Reference Model (LRM) specifications for nomen and the potential challenges visual nomen might present for their description and use in information systems, the purpose of this study was to investigate two questions: (1) how do nonlinguistic or nonalphanumeric signs or symbols act as nomen to identify entities? and (2) what details or attributes are relevant to describe and classify such nomen to integrate them into information systems? Design/methodology/approach This research was built on an exploratory, qualitative instrumental case study design using multiple (or comparative) cases. Using the International Federation of Library Associations and Institutions LRM conceptualization of nomen as the basis, this research explored the similarities and differences between the LRM definition, its attributes and the use of nonlinguistic and nonalphanumeric “strings” for visual nomen to represent a res , moving iteratively between the LRM documentation, visual nomen identified in previous research and additional examples. This study used a constant comparative method to conduct a structured, focused comparison across different cases found in the source survey. Findings A close review of the history of the development of the nomen entity was made to understand the semiotic relationship between entities and their symbolic representation, how those symbols are then reified to be further classified and described and how such definitions in the LRM offer a path forward for better understanding the role and function of visual nomen. Based on the foundation of the nomen entity and its attributes established in the LRM, this research then looked at visual representations of concepts and entities to suggest a nascent framework for describing aspects of visual nomen which may be relevant to their use and application Originality/value This exploratory study of the use of supralinguistic ways of referencing entities delineates novel insights into a potential framework for describing and using visual nomen as a way of labeling or naming entities represented in information systems. By examining the specifications of the nomen entity and its attributes as delineated by the LRM, this study reinforces the applicability of LRM-defined attributes in the use of visual nomen in addition to offering other attributes or dimensions.
Synopsis The research problem This paper assesses whether and how people’s perceptions of time — strong future time reference (FTR) versus weak FTR — affect corporate default risk. Motivation or theoretical reasoning Studies have shown that default risk varies across firms, regions, and countries, highlighting the need for a comprehensive understanding of the contributing factors. Traditional studies focus on how firm-level, industry-level, national and international economic and financial variables shape corporate default risk, but they fail to explain cross-country and cross-regional differences in corporate default risk from the perspective of informal institutions, particularly, language. This study takes the first step to examine whether and how future-oriented language shapes corporate default risk. The test hypotheses We first tested whether strong-FTR language decreases corporate default risk. We further tested whether the effect of strong-FTR language on default risk depends on firms’ level of information transparency. In addition, we tested whether the effect of strong-FTR language on default risk depends on a country’s disclosure requirements. Lastly, we tested whether the effect of strong-FTR language on default risk depends on a country’s control of corruption. Target population We find that corporate default risk is significantly higher in regions dominated by speakers of weak-FTR languages, using a comprehensive sample of firms in 36 countries with 180,013 observations spanning from 1988 to 2017. Adopted methodology Ordinary least square regressions were used in this study. Analyses Corporate default risk is measured by two proxies of firm probability of default, following Merton [(1974) Journal of Finance, 29(2), 449–470] and Lee and Lin [(2012) Journal of International Financial Markets, Institutions, and Money, 22(4), 973–989]. Our independent variable is Strong FTR, which equals 1 if a language belongs to the strong-FTR language family, as defined by the European Science Foundation’s Typology of Languages in Europe (EUROTYP) project. If a language does not require “obligatory [FTR] use in (main clause) prediction-based contexts” [Dahl (2000)Tense and Aspect in the Languages of Europe, O. Dahl (Ed.), pp. 309–328], then we put this language into the weak-FTR group. On the other hand, if a language does have the above-mentioned requirement, then it belongs to the strong-FTR group. Findings We found that corporate default risk is significantly higher in regions dominated by speakers of weak-FTR languages. Furthermore, the FTR effect on default risk is weakened in countries with stronger formal institutions (e.g., high disclosure quality, greater transparency, and less corruption). Our results introduce a new explanation for heterogeneity in corporate default risk, provide insights about whether language is an economic institution, and adds to research on the effects of languages on economic and financial outcomes.
This is a book about numbers – what they are as concepts and how and why they originate – as viewed through the material devices used to represent and manipulate them. Fingers, tallies, tokens, and written notations, invented in both ancestral and contemporary societies, explain what numbers are, why they are the way they are, and how we get them. Overmann is the first to explore how material devices contribute to numerical thinking, initially by helping us to visualize and manipulate the perceptual experience of quantity that we share with other species. She explores how and why numbers are conceptualized and then elaborated, as well as the central role that material objects play in both processes. Overmann's volume thus offers a view of numerical cognition that is based on an alternative set of assumptions about numbers, their material component, and the nature of the human mind and thinking.
Full-text available
Color is a prime example of categorical perception, yet it is unclear why and how color categories emerge. On the one hand, prelinguistic infants and several animals treat color categorically. On the other hand, recent modeling endeavors have successfully utilized communicative concepts as the driving force for color categories. Rather than modeling categories directly, we investigate the potential emergence of color categories as a result of acquiring visual skills. Specifically, we asked whether color is represented categorically in a convolutional neural network (CNN) trained to recognize objects in natural images. We systematically trained new output layers to the CNN for a color classification task and, probing novel colors, found borders that are largely invariant to the training colors. The border locations were confirmed using an evolutionary algorithm that relies on the principle of categorical perception. A psychophysical experiment on human observers, analogous to our primary CNN experiment, shows that the borders agree to a large degree with human category boundaries. These results provide evidence that the development of basic visual skills can contribute to the emergence of a categorical representation of color.
Full-text available
We report a study of Turkish color terms with four main aims: to establish the inventory of BASIC color terms; to compare this inventory with Berlin and Kay's 11 color universals; to see if Turkish is an exception to the theory by having two basic terms for blue; and if it is, to explore whether there are cognitive effects of the two blue terms. Eighty children aged from eight to 14 years and 153 adults performed a color-term list task (write down as many color terms as you can) and a subset of these two samples went on to perform a color-naming task. In the naming task, they were asked to name 65 representative color "tiles." Measures of salience and consensus derived from the two tasks converge to suggest that Turkish has 12 basic color terms. The denotations of these terms and the glosses provided by dictionaries and Turkish-speaking consultants are consistent with 11 of the terms being Turkish tokens of Berlin and Kay's 11 universal categories. The twelfth term - lacivert 'dark blue' - lies between the foci of the universal blue and purple and its range overlaps with the dark-blue term of Russian, sinij. However, in a third phase of the investigation, the majority of informants said that lacivert 'dark blue' was a kind of mavi 'blue', thus violating one of Berlin and Kay's criteria (noninclusion) for basicness. Thus we have the unusual, but logically possible, case of a term being used with prevalence, consensus, and specificity, while at the same time being acknowledged as a subset of another term. Whatever the status of the two blue terms, however, we found evidence that the cognitive representation of the blue region of color space may reflect the salience of the two blue terms using color grouping, similarity judgments, and same-different tasks.
Full-text available
The World Color Survey (WCS) is a research project that was undertaken to validate, invalidate, or, most likely, modify the main findings of Berlin and Kay (B&K), which said that there exist universal crosslinguistic constraints on color naming, and that basic color terminology systems tend to develop in a partially fixed order. This chapter reviews the history of the World Color Survey, including the creation of the online data archive, and describes the recent use of the archive to test the universality of color naming across languages. The WCS data archives are a publicly accessible resource, which is available to all who wish to pursue questions related to color categorization across languages. The chapter provides this background to orient the potential users of the archive, to give them a sense for where the data came from, how the data were compiled into an archive, and what sorts of questions the data can be used to investigate.
This article presents a framework for understanding, modeling, and computing color categories on the basis of knowledge from the color imaging science. One of the main assumptions advocated in this article is that the structure of color categories originates from the statistical structure of the perceived color environment that was observed throughout an individual's life. The perceived color environment can be modeled as color statistics of natural images in some perceptual and approximately uniform color space (e.g., the CIELUV color space). The process of color categorization can be modeled as the grouping of the color statistics by clustering algorithms (e.g., K-means). The proposed computational model enables one to predict the location, rank, and number of color categories. The model is examined on the basis of K-means clustering analysis of statistics of 630 natural images in the CIELUV color space. In general, the model predictions are consistent with data from psycholinguistic studies. The model might be applied in different areas of imaging science such as color quantization, image quality, and gamut mapping.
The universality, invariance, and elegance of principles governing the universe may be reflected in principles of the minds that have evolved in that universe - provided that the mental principles are formulated with respect to the abstract spaces appropriate for the representation of biologically significant objects and their properties. (1) Positions and motions of objects conserve their shapes in the geometrically fullest and simplest way when represented as points and connecting geodesic paths in the six-dimensional manifold jointly determined by the Euclidean group of three-dimensional space and the symmetry group of each object. (2) Colors of objects attain constancy when represented as points in a three-dimensional vector space in which each variation in natural illumination is canceled by application of its inverse from the three-dimensional linear group of terrestrial transformations of the invariant solar source. (3) Kinds of objects support optimal generalization and categorization when represented, in an evolutionarily-shaped space of possible objects, as connected regions with associated weights determined by Bayesian revision of maximum-entropy priors.
Languages are transmitted from person to person and generation to generation via a process of iterated learning: people learn a language from other people who once learned that language themselves. We analyze the consequences of iterated learning for learning algorithms based on the principles of Bayesian inference, assuming that learners compute a posterior distribution over languages by combining a prior (representing their inductive biases) with the evidence provided by linguistic data. We show that when learners sample languages from this posterior distribution, iterated learning converges to a distribution over languages that is determined entirely by the prior. Under these conditions, iterated learning is a form of Gibbs sampling, a widely-used Markov chain Monte Carlo algorithm. The consequences of iterated learning are more complicated when learners choose the language with maximum posterior probability, being affected by both the prior of the learners and the amount of information transmitted between generations. We show that in this case, iterated learning corresponds to another statistical inference algorithm, a variant of the expectation-maximization (EM) algorithm. These results clarify the role of iterated learning in explanations of linguistic universals and provide a formal connection between constraints on language acquisition and the languages that come to be spoken, suggesting that information transmitted via iterated learning will ultimately come to mirror the minds of the learners.
More than 100 indigenous languages are spoken in Mexico and Central America. Each language partitions the color spectrum according to a pattern that is unique in some way. But every local system of color categories also shares characteristics with the systems of other Mesoamerican languages and of languages elsewhere in the world. This book presents the results of the Mesoamerican Color Survey, which Robert E. MacLaury conducted in 1978-1981. Drawn from interviews with 900 speakers of some 116 Mesoamerican languages, the book provides a sweeping overview of the organization and semantics of color categorization in modern Mesoamerica. Extensive analysis and MacLaury's use of vantage theory reveal complex and often surprising interrelationships among the ways languages categorize colors. His findings offer valuable cross-cultural data for all students of Mesoamerica. They will also be of interest to all linguists and cognitive scientists working on theories of categorization more generally.
This article presents a framework for understanding, modeling, and computing color categories on the basis of knowledge from the color imaging science. One of the main assumptions advocated in this article is that the structure of color categories originates from the statistical structure of the perceived color environment that was observed throughout an individual's life. The perceived color environment can be modeled as color statistics of natural images in some perceptual and approximately uniform color space (e.g., the CIELUV color space). The process of color categorization can be modeled as the grouping of the color statistics by clustering algorithms (e.g., K-means). The proposed computational model enables one to predict the location, rank, and number of color categories. The model is examined on the basis of K-means clustering analysis of statistics of 630 natural images in the CIELUV color space. In general, the model predictions are consistent with data from psycholinguistic studies. The model might be applied in different areas of imaging science such as color quantization, image quality, and gamut mapping.