report resumes ed 019 655 al aw 003 universality and evolution of basic color terms. working paper number 1. by- berlin, brent kay, paul california univ., berkeley > lang.-behav. res. lab. edrs frice mf-$0.5cj hc-í3.24 79f. descriptors- ^language universals, *visual discrimination, ^cultural factors, *communication (thought transfer), slanguage research, cognitive measurement, cognitive tests, hypothesis testing, codification, anthropology, linguistic theory, basic color terms, the research reported in this working paper "strongly indicates" that semantic universals have been discovered in the domain of color vocabulary. moreover, these universals appear to be related to the historical development of all languages in a way that can properly be termed evolutionary. the research was conducted in a graduate seminar given in the department of anthropology, university of california, berkeley. students and the authors systematically collected data from several informants in each of seventeen languages from a number of unrelated language families. an additional three languages were investigated in detail after the original research seminar was comfleted. these materials from twenty languages were supplemented by comparative data from the literature, bringing the sample of languages to 80 representing a wide variety of major linguistic stocks. the seminar was designed as an experimental test of the following, loosely stated hypothesis--the prevailing doctrine of american linguists and anthropologists has, in this century, been that of extreme linguistic relativity. proponents of this view frequently offer as a paradigm example the alleged total semantic arbitrariness of the lexical coding of color. the authors feel that this allegation of "total arbitrariness" in the way languages segment the color sface is a "gross overstatement." their hypothesis was based on intuitive experience in several languages of three unrelated major stocks. their feeling was that color words translated rather too easily among various fairs of unrelated languages for the extreme linguistic relativity thesis to be valid. their results support the above hypothesis and cast doubt on the commonly held belief that each language segments the three dimensional color continuum arbitrarily and independently, they suggest that although different languages encode in their lexicons different numbers of basic color categories, there exists universally Ä total inventory of 11 basic color categories from which the 11 or fewer basic color terms of any given language are always drawn—these categories being white, black, red, green, yellow, blue, brown, fink, orange, and grey. (amm) OS q>, ü.s. department of health, education & welfare <^ office of education this document has been reproduced exactly as received from the person or organization originating it. points of view or opinions stated do not necessarily represent official office of education position or policy. Universality and Evolution of Basic Color Terms "permission to reproduce this Wmtm material has been granted by Paul kay_ to eric and organizations operating under agreements with the u.s. office of education. further reproduction outside the eric system requires permission of the mmm owner." Brent Berlin and Paul Kay University of California, Berkeley Working Paper Number 1, Laboratory for Language-Behavior Research University of California, Berkeley AL 001 003 9 Contents 0 Introduction 1. The hypothesis and general findings 1.1 Procedure 1» 2 Defining basic color terms 1.3 Mapping basic color terms 1.4 Universality of basic color terms 1.5 Inter-language vs. intra-language variability 1.6 Category foci vs. category boundaries 2. Evolution of basic color terms 2.1 Basic color lexicon and technological/cultural complexity 2.2 The seven stages in the evolution of basic color terms 2.3 Supporting data and examples 2. 3.1 Stage I systems 2.3.2 Stage II systems 2. 3. 3 Stage III systems 2.3.4 Stage IV systems 2. 3.5 Stage V systems 2.3.6 Stage VI systems 2.3. 7 Stage VII systems 2.4 Internal reconstruction of basic color terms 2.5 Problematical cases 3. Summary and indications for future research Notes References cited 0. Introduction Ethnoscience studies, and studies of color vocabulary in particular, have firmly established the point that to understand the full range of meanings of a word in any language, each new language must be approached in its own terms, without a priori theories of semantic universals. H. C. Conklin (1955) has shown, for example, that Hanunoo "color" words in fact encode a great deal of non-colorimetric information. The essentially methodological point made in such studies has been frequently misinterpreted by anthropologists and linguists as an argument against the existence of semantic universals. The research reported here strongly indicates that semantic universals have been discovered in the domain of color vocabulary. Moreover, these universals appear to be related to the historical development of all languages in a way that can properly be termed evolutionary. 1. The hypothesis and general findings Research was conducted in a graduate seminar given in the Department of Anthropology, University of California, Berkeley. Students and the authors systematically collected data from several informants in each of seventeen languages from a number of unrelated language families. An additional three languages were investigated in detail after the original research seminar was completed.1 These materials from twenty languages were supplemented by comparative data from the literature bringing our present sample of languages to 80 representing a wide variety of major linguistic stocks. The seminar was designed as an experimental test of the following, loosely stated hypothesis: The prevailing doctrine of American linguists and anthropologists has, in this century, been that of extreme linguistic relativity. Proponents of this view frequently offer as a paradigm example the alleged total semantic arbitrariness of the lexical coding of color.2 We suspect that -3- this allegation of "total arbitrariness" in the way languages segment the color space is a gross overstatement. The hypothesis was based on the intuitive experience of the authors in several languages of three unrelated major stocks. Cur feeling was that color words translated rather too easily among various pairs of unrelated languages for the extreme linguistic relativity thesis to be valid. Cur results support the above hypothesis and cast doubt on the commonly held belief that each language segments the three dimensional color continuum arbitrarily and independently of each other language.3 It appears now that, although different languages encode in their lexicons different numbers of basic color categories, there exists universally a total inventory of eleven basic color categories from which the eleven or fewer basic color terms of any given language are always drawn. The eleven basic color categories are white, black, red, green, yellow, blue, brown, pink, orange, and grey. A second and totally unexpected finding is the following: if a language encodes some number n < 11 basic color categories, then there are strict limitations on which n categories it may encode. The list of 11 basic categories is partially ordered, producing seven equivalence classes^: 1. white, black < 2» red < 3. green < 4. yellow < 5. blue < 6. brown < 7. purple, pink, orange, grey 1 If a language codes a category from the mth equivalence class, (m s 1, 2, . .,,7), then it encodes all categories in each equivalence class r < m. 7hus, excluding the empirically unlikely possibility of a one-term color vocabulary, 5 there are just twenty-one possible basic color lexicons. (Cee r. able 1; we are not of course referring to the phonological or syntactic features of the color lexemes). Moreover, if a color lexicon encodes six or fewer categories, we can predict exactly which ones they will be. This fact has clear implications for the evolution of color vocabulary in all languages. (In fact, the above ordering gives considerably over-simplified picture of the detailed evolutionary sequence, which is presented below). Table I: The Twenty-one Possible Basic Color Lexicons Perceptual categories encoded in the basic color terms: No. of Type basic color terms white : black I red green yellow blue brown pink purple orange u 1 2 + 2 3 + + 3 4 + + + 4 5 + + + + 5 6 + + + + + + 6 7 + + + + + + — — 7 8 + + + + + + + — — — S 8 + + + + + — + — 9 8 + + + + + + + — - + - 10 £ + + + + + + - - — + 11 9 + + + + + + + + — — 12 9 + + + + + + + + - + — 13 9 + + + + + + + - — + 14 9 + + + + + + + — + + - 15 9 + + + + + + + _ + — + 16 9 + + + + + - - + + 17 10 + + + + + + + + — 18 10 + + + + + •í- + + + — + 19 10 + + + 4. + + + — + 20 10 + + + + + + + MM + + 21 11 + + + + + + + + + + + In sum, our two major findings are (i) the referents for the basic color terms of all languages appear to be drawn from a set of 11 universal perceptual categories, (ii) these categories become encoded hi the history of a given language in a (partially) fixed order. There does not appear to be any evidence that differences in complexity of basic color lexicon between languages reflect 6 perceptual differences between the speakers of those languages. 1.1. Procedure. Standarized color stimuli were used in conducting the research. "Jhese 7 consisted of a set of 329 color chips provided by the Munsell Color Company. The set is comprised of (i) 320 color chips of 40 equally spaced hues and eight degrees of brightness, all at maximum saturation, and (ii) nine chips of neutral hue (white, black and greys). The full array of chips was mounted on stiff cardboard and covered with acetate to form the array shown in Fig. 1. "^Ath the exception of our addition of the neutral hue series, these materials are the same as those used by Lenneberg and Roberts (1953) in their classic cross-cultural study of English-Zuni color terminology. Cur method of obtaining the individual mappings differs, however, from theirs, as may be seen by comparing the following discussion with their work. Data were gathered in two stages. First the basic color words of the P language in question were elicited according to ethnoscience techniques using as little as possible of any language other than the one under study during the interview. Secondly, each subject was instructed to map both the focal point and the outer boundary of each of his basic color terms on the array of standard color stimuli described above. 1.2. Defining basic color terms. 7 here is in every language an indefinitely large number of expressions that can denote color. Note, for example, the following English expressions: rigure 1 The Stimulus Materials Used for Obtaining Mappings of Basic Color Terms* to R * & C i—i Ph PQ HUE -ik- O J Q H K O H O O I* w P R J Ph PQ O R P PQ H H Ph P P3 J P PQ Ph j Ph P Ph H Q Ph Ph P Ph nrr : ; 1 i ) i "i i i ! 1 8 ! : 1 3 : ij f ! ■j i i • ! I ! i j —1 — I i i ! i i ! !! ! — ! : ■'! r j , —~ i : ! ! i J i ! i n s i Numerals refer to the Munsell system of color notation. F— ------- —- -5- (a) crimson, (b) scarlet, (c) blond, (d) blue-green, (e) bluish, (f) lemon-colored, (g) salmon-colored, (h) the color of the rust on mjr aunt's old Chevrolet. Cn the other hand, psychologists, linguists, and anthropologists have long operated with a concept "basic color term" or "basic color word", which excludes forms such as (a)-(h) and includes forms like black, white, red, green, and so on. "Basic color term" does not have a unique, accepted operational definition. Cur operational procedure for the determination of basic color terms is as follows. Ideally, each basic color term should exhibit the following four characteristics: (i) It is mono-lexemic, i. e., its meaning is not predictable from the meaning of its parts. This criterion eliminates examples (e)-(h) and perhaps also (d). (ii) Its signification is not included in that of any other color term. Ibis criterion eliminates examples (a) and (b) which are both kinds of red for speakers of English. (iii) Its applicability must not be restricted to a narrow class of objects. This criterion eliminates example (c) which may be predicated only of hair, furniture and perhaps a few other things,, (iv) It must be salient for informants, tending to occur at the beginning of elicited lists of color terms, stable in its reference, occurring in the ideolects of all informants, etc. This criterion eliminates ail the examples (a)-(h), most particularly (h). Criteria (i)-(iv) suffice in the vast majority of cases to determine the basic color terms in any language. The few doubtful cases that arise are handled by the following subsidiary criteria: (vi) Ihe doubtful form should have the same distributional potential as the previously established basic terms. For example, in English, allowing the suffix -ish, e.g., redish, whitish, greenish, etc., but not *scarletish, *blue-greenish, etc. (vii) Names of objects characteristically having the color in question are erIc -o- suspect, e. g., gold, silver, ash. "3 his subsidiary criterion would exclude orange, in English, if it were a doubtful case on the basic criteria (i)-(iv). (viii) Hecent foreign loan words are suspect, other things being equal. (ix) In cases where lexemic status is difficult to assess, morphological complexity is given some weight as a secondary criterion. The Engoish term blue-green might be eliminated by this criterion. 1. 3. Mapping basic color terms. No mapping of color terrrs was attempted until each investigator had elicited verbally the basic color terms in each language. Mapping was accomplished by use of acetate strips cut to the size of the stimulus board. Each informant was given a black grease pencil (china marker) and asked, for each basic color term, x: (i) Please indicate all those chips which you would under any conditions call x. (ii) Please indicate the best, r~ost typical examples of x. Cur querries were designed to get at the total area of a basic category and to determine, as well, its focus or most typical member(s). 7he mapping procedure was carried out at least three times, at one week intervals, for each informant. Cften, we had access to only one informant per language. However, in the case of Tzeltal (Mayan), it was possible to consult 40 informants. The languages we selected were relatively diverse genetically. 1 he • choice of each, however, was limited by informant availability. All informants were native speakers of their respective languages and, with the exception of the Izel-tal individuals, resident in the San Francisco Bay Area. Cur primary data include basic color categories for the following languages: 1. (Lebanese) Arabic 2. Bulgarian -7- 3. Catalan 4. Cantonese (Chinese) 5. Mandarin (Chinese) 6. English 7. Hebrew £. Hungarian 9. Ibibio 10. Indonesian 11. Japanese 12. Korean 13. Porno 14. (Mexican) Spanish 15. Swahili 16. Tagalog 17. Thai 18. Tzeltal 19. Urdu 20. Vietnamese 1.4. Universality of basic color terms. After each language was mapped;, we made a composite for all languages of the foci of all basic color terms. The composite is given in Fig. 2, where letters indicate the twenty languages for which we have systematically collected data. When several neighboring chips are marked by the same letter, it indicates that several chips were judged to be equally good representatives of the focus of a category. Fig. 2, although a brute summary of the dafta and consequently rather hard to read, nevertheless shows the considerable extent to which the foci of color categories are simtUar among totally unrelated languages. Both (i) the large blank areas (over 707. of the surface of the chart) and (ii) the close Notes to Figure 2 a) Dotted-lines on chip YELLOW RED 10, brightness 7, indicate overlap of the categories orange and yellow. It is affiliated with yellow for Tzeltal and Cantonese, with orange for Swahili. b) Arrow emanating from chip BLUE 2.5, brightness 3 indicates its affiliation with the category 'green' for Vietnamese. Likewise, chip RED PURPLE 10, bright-nese 3 is affiliated with 'purple' for Korean. D c) * indicate chips chosen as category foci for each of the 20 languages. d) Where a letter occurs more than once, several chips were judged to be equally good representatives of the focus of the category. ) Figure 2 Composite of Foci of Basic Color Terms in 20 Languages HUE W w o Q H ps I iJ Q H H PH PS a H o PS H w H PS o P w PS p3 o H t> PS p3 P.. w ps P H PS w Ph ps p.. ^ CO CD A A HI IC H Ml K V u E 8«. 6fl X J ? ft £ Y H K 6 S K 7 H A K A |S I & s S CI T V tk E 6 5 E S Tt, »81 H 5 I H H K K 3 X X K. ft r 5 H J B J r I H H V C M J u B E H6 5i E H K X X I K a K X tt H o A C il 1 «P $ V U H T I & 4 A T E V J St 9 i* $ Hi A & V S I r V E Tt V I e k~ H 3 A J H Tr n I E E M Sp S e Hfc £ E C & Ct s 3 2 I <*A H J % It V M P A •r A K n \ V r 3 sr T H E B E H \ X C f A J Key A--Arabic B--Bulgarian C--Catalan Ct--Cantonese E--English H--Hungarian Hb-Hebrew Ib--Ibibio I--Indonesian J--Japanes K--Korean M--Mandarin P--Porno S--Swahili Sp--Spanish T--Tagalog Th--Thai Tz--Tzeltal U--Urdu V- - Vietnamese -8- clustering into discrete, contiguous areas of the foci of the various languages, attest to the failure of the strict linguistic relativity hypothesis. This effect can be seen more clearly in Fig. 3. As shown in Fig. 2, informants frequently designate more than one chip as the focus of a color term. Fig. 3 is based on a calculation for each language of the center of gravity of the focus area for each basic color term in each language; each of the eleven areas, with an associated gloss (e.g., 'white', 'red') includes foci for a number of languages equal to the corresponding numeral (e. g., the area 'red' includes the foci for all 20 languages, the area 'orange' includes foci for 11 languages, the remaining 9 languages in the sample lacking a term for this category, and so on). The results shown in Fig. 3 support quite strongly our initial hypothesis: color categorization is not random and the centers of basic color terms are very si-milar in all languages. 1.5 Inter-language vs inter-*nformant variability Further evidence for the cross-language universality of color foci is the fact that variability in the location of color foci appears to be greater among the speakers of a given language than between languages. Whether or not the observed result - that inter-informant variability in a given language exceeds inter-language variability - would attain statistical significance under various sampling assumptions, from the fact that intra-language variability absolutely exceeds inter-language variability, we can confidently reject the hypothesis of greater intra-language variability in favor of the null hypothesis of no difference. The question of inter- vs intra-language variability was assessed as follows. The only language on which we have reliable data for a substantial number of informants is Tzeltal. Data were collected for forty Tzeltal informants. Cf these, thirty-one located the center of yag in the green area and nine in the blue area, a fact discussed in detail below (section 2.4.3.). Tzeltal has five basic color terms: 'white', 'black', 'red', 'green' (i.e., yai), and 'yellow'. o 'black' 20 .'gney' l£ to co BRIGHTNESS CO «--► cn w RED cn YELLOW w RED I d w YELLOW GREEN YELLOW w GREEN BLUE GREEN BLUE BLUE PURPLE PURPLE RED PURPLE H o 3 CO 2 o 3 n to a 5" ^ 3 o £ O * rt o « o öd era S3 s CO cm o co O o 1—1 o CO r—-----—- -9- Prom the thirty-one yas-normal informants a random sample of ten was selected, Ihe restriction to yas-normal informants biases the Tzeltal sample toward homogeneity and thus in the direction opposite from the point being made here. There are no other languages in the basic sample of twenty containing just five terms. However, three languages Japanese, Korean, and Cantonese can be reconstructed to five-term systems on the basis of internal evidence (see section 2.4 for detailed discussion of internal reconstruction). The foci for white, black, red, green, and yellow in these three languages and in Tzeltal were used for the inter-language comparisons. The bias, if any, introduced by using languages with varying numbers of color terms should be toward greater between-language variation, again in the direction opposite from the observed result. The foci for Cantonese, Korean, and Japanese were each obtained from a single informant. In order to introduce a finer co-ordinate system for computing inter-focus distances each unit of hue and brightness was sub-divided into four units, yielding 160 units of hue and 36 of brightness. That is, each box in Figure 1 is considered a square of 4-unit side rather than 1-unit side. For each of the forty-five pairs of Tzeltal informants, the Euclidean distance between their foci for each of the five color categories was calculated, resulting in 225 distances (45 informant-pairs x 5 color categories), The overall mean of these 225 inter-informant (intra-language) differences is 4. 47 units on the fine scale, or roughly one and one half chip widths. The distances between foci were calculated for each of the five categories, for each pair of languages. Two slightly different sets of foci were used for Tzeltal in these comparison: (1) the five foci of a single informant selected at random from the sample of ten, and (2) the five mean foci for the ten informants - that is, the points given by the arithmetic means of the hue and brightness co-ordinates of the ten informants for each category. The different ways of treating Tzeltal did not materially affect the result. The inter-language distances for each of the five foci were then averaged for each pair of languages to give a single mean distance measure for every language pair. The results of ERJC ] -10- these computations are summarized in Table n, which shows that every inter-language distance exceeds the mean inter-informant distance for the sample of Tzeltal informants. The above results, we think, are interesting given the traditional anthropological attitude concerning the nature of human color categorization. While it can be argued that bilingualism in English affects the results to some extent (cf. Ervin, 1961), we find it hard to conceive that English could influence the placement of the foci of categories in these diverse languages in such a total fashion. Moreover, the work completed with forty Tzeltal informants varying from pure Tzeltal monolinguals to perfect Tzeltal-Spanish bilinguals indicates that our results are not skewed due to bilingualism. Finally, the fact that inter-individual differences in a given language appear to be as great or greater than inter-language differences weakens considerably the possible objection on the basis of bilingualism of the informants consulted. 1.6. Category foci vs category boundaries An immediate result of the mapping procedure was that judgements of category foci were highly reliable on repeated trials with the same informant and also across informants. It was very rare that a category focus was displaced by more than two adjacent chips on repeated trials. On the other hand category boundaries were not reliable, even for repeated trials with the same informant. This phenomenon was also reflected in the relative ease with which informants accomplished the focus-designating task in contrast with the boundary-mapping task. Many subjects engaged in long hesitations with regard to the latter, demanded clarification of the instructions, etc. In fact, in marked contrast to the foci, category boundaries proved to be so unreliable, even for a given informant, that they have been accorded a very secondary place in the analysis. Consequently, whenever we speak of color categories, either above or below, we refer ,t2 Mi?. foci of categories, rather than to their boundaries or total area (volume), exceirt when specifically stating otherwise. -11- Two alternative interpretations of this result suggest themselves. First, it may well be that the primary storage procedure in the brain for the physical reference of color categories (i. e., their meaning) is concerned with points (or very small volumes) of the color solid rather than extended volumes. Some sort of secondary processes, of considerably lower salience and inter subjective homogeneity, would then have to account for the extensions of reference to points in the color solid not equivalent to (or included in) foci. Current formal theories of lexical definition do not appear able to deal with such phenomena naturally. If empirical results of this kind accumulate, simple Boolean function theories of lexical definition may have to be revised in favor of more powerful formalisms. We do not have space here to pursue the matter further, especially since there appears some reason to suspect that color (and perhaps a few other semantic domains such as smell and noise) may have unusual lexical properties. The alternative explanation for the superior reliability of category foci to category boundaries is that this is an artifact of our experimental procedure. In retrospect, we find nothing in our procedure which might plausibly be argued to produce such a bias, although others perhaps can. Moreover, the fact that the evolutionary scheme, including the data from the 60 additional languages ordered by it, works so well in terms of foci seems to argue against interpreting the apparent reality of foci as an artifact. 2. Evolution of basic color terms Cur second major result is that there appears to be a fixed sequence of evolutionary stages through which a language must pass as its basic color vocabulary becomes enriched over time. This conclusion is based in part on the findings of universality for the eleven basic category foci, in part on the non-randomness of their distribution across contemporary languages (and certain logical consequences of the Table II Comparison of Mean distances in the Location of Five Color Foci Among Four Languages and Among Ten Speakers of One Language I Mean inter-focus difference for ten Tzeltal (yas-normal) informants 4.47 II Mean inter-focus difference for all pairs of four Languages 1. Japanese - Cantonese 4.43 2. Japanese - Korean 4.30 3. Cantonese - Korean 4.18 4. Japanese - Tzeltal (i) one Tzeltal informant 3.84 [ *■■.'•■■.;'.:'•.'.■;•;••.%•. y.\•, v.*-.".; .....•.•;: :•.*.•;.• • • "*■•■"■ ••••v" * k * • " * -.V.': <>:•: J:/j: ;'vv) J'-/.; !•«■' ::::,:vA'.:-"!'-': v-*' V-:;.'•' •//;::^'V:;^:*/V^\.. ;..:»:v::. i. «. « , • «• • • *•*•,■ *•.•• '•: "•: ■ V '•'.v.v'.vV'/■ • V //■•tf:yy::>;^ RED fi£):0? ..,«« ■... .■ i "'Vo^C RED •>:•.•:•:•■ • ■:: •: :•• .: ■;• • .• •.: •.V. V-.: ^^"';V-'.vV-%v:.-'5-• • .'.V:"'*' ..•.■'•••.'.v. -*•••.".'.••••'.•.••.•:: i: ••».... '' ..••/.•:■«•.•••••• ■ . < ....•:*" ■/.*.■ • :• J ■/•»••«■• ■.•<•!. «•• 2 »« .. •..*.: BLACK Figure 5 Typical Stage II Basic Color Lexicon ERjC WHITE V, :;; '.•/ •V- '.*."•"*"*"/•' • •. • ■*•• •* • •••• • «•*##••*( . « »«••* : • v'sV.vv "V7: »•!•":': • •„• • .•.•:■..:•.■.•.;«.•■;>.•••:••. iMy X GREEN iV^^jS^^ • • •:; .•; \ ■/yj^jjjj^-*--- RED '' •...'r-trff:: .v.*)-.' \ ; • .j.y. v \\'! RED ■.■•.•■■••J. • • BLACK Figure 6a Typical Stage III Basic Color Lexicon, Major Variant 9 EBJC r WHITE ........."'• I -ii GREEN BLACK Figure 6b Typical Stage III Basic Color Lexicon Minor Variant eric UBBS1QS9 (minor variant). RED continues to encompass the areas of English red, some yellow reds, purple and purple reds. Presumably, BLACK and WHITE continue to be deprived of hue reference at this time, becoming more and more restricted to the neutral values. Stage IV is seen in Pig. 7. At Stage V the focus of blue emerges, normally from the GREEN area but perhaps occasionally also from BLACK . In either case, GREEN now becomes green. At this stage, BLACK and WHITE are fully reduced to black and white; i.e., to neutral values. The RED area is probably also reduced with respect to purples and violets. Stage V is depicted in Fig. 8. Stage VI is the last at which a single focus appears, brown. At Stage VI both RED and YELLOW become even more restricted in area although it is not until Stage VII that they become red and yellow. Stage VI is seen in Fig. 9. When the color lexicon expands beyond seven terms, i. e., beyond Stage VI, there is apparently a rapid expansion in filling out the full roster of eleven basic color categories. This conclusion is suggested by the fact that, for the 80 languages investigated, only 4 color lexicons belong to types other than 1, 2, 3, 4, 5, 6 and 21. (See Tables III and IV). Apparently, at Stage VII, the remaining basic categories pink, orange, grey and purple are added to the lexicon very rapidly and, as far as we have been able to ascertain at the mo-in no particular/ ment, /order. Our data now suggest that purple and pink probably arise from RED while orange becomes isolated from YELLOW . There is, however, some evidence to indicate that orange may have, in some cases, arisen from RED . Grey represents simply the encoding of mid-brightness neutral hues between black and white. Stage VII systems include all eight-, nine-, ten-, and eleven-term systems and thus include types 7-21. As shown in Table IH, of the 20 Stage VII systems so far encountered, 13 are of type 21, i.e., contain all eleven basic categories while 11 of the fifteen possible types of Stage VII are not represented, (cf.note 5, also section 2.3.). An eleven term Stage VH system is seen in Fig, 10. WHITE ;.'v{Y ELLOW^: • «• .v;. •. ■ ■..: •.:y; y. BLACK Figure 7 Typical Stage IV Basic Color Lexicon black Figure 8 Typical Stage V Basic Color Lexicon erJc Figure 9 Typical Stage VI Basic Color Lexicon d ERIC J white Figure 10 Typical Stage VII Basic Color Lexicon, Eleven Term System* * The eleventh category, grey, cannot be depicted on the above diagram given the conventions discussed in note 15. Table III Distribution of 80 Basic Color Lexicons Among the Twenty-one Theoretically Possible Types, with Indication of Evolutionary Stage. Type No. of Basic Stage No. of Examples Color Terms 1 2 I 5 2 3 II 21 3 4 III 10 4 5 IV 13 5 6 V 7 6 7 VI 4 7 8 VII - 8 8 VII 2 9 8 VII - 10 8 VII 1 11 9 VII 12 9 VII m 13 9 VII - 14 9 VII - 15 9 VII 1 16 9 vn m 17 10 VII - 18 10 VII - 19 10 vn - 20 10 VII 21 11 VII 13 Total Stage VII 20* Grand Total 80 ♦Catalan, Cantonese, and Vietnamese are Stage VII systems but are not typed. See Söc*,ic:i 2.5. 16 2.3. Supporting data and examples Our search of the literature for reports on color terminologies is not complete. To date we have gathered reasonably reliable information on 60 languages in addition to the twenty languages for which we have experimental data. The results from all, reliably reported languages are considered here and conform almost totally to our proposed evolutionary sequence. First, we give a few examples of each stage, with emphasis on the earlier, more interesting stages. 2.3.1. Stage I systems Originally, we had no hope of discovering an extant example of Stage I. We were thus pleasantly surprized to receive from K. F. Koch (1966) the following report on a New Guinea Highland group called the Jale, whose language has tentatively been affiliated with the Dani (Non-Austronesian) group. In a report made in our seminar, Koch, who was totally unaware of the theory, stoutly resisted our suggestions that Jale might have more than two true color terms. Jale is Stage I, having basic color terms only for 'BLACK1 and 'WHITE' (see Fig. 11). There are other terms which, in highly restricted contexts, refer to certain hues. These terms, however, are restricted almost exclusively to particular substances or objects, e.g., mut 'red soil', piano 'name of plant whose leaves are used to rub yarn, dying yarn a green color', etc. Koch reports that when he requested a Jale native to do something with a 'green' object, by using the term piano, he was consistently misunderstandood. He subsequently learned to use the term sin, 'BLACK' or nolo 'WHITE' depending on the degree of brightness that the particular 'green' represented. This was made even more obvious when he reported that the appearance of blood is sin_ 'BLACKi exactly what 'blood (red)1 should be at Stage I due to its low brightness. That we should find a Stage I system in Highland New Guinea is consonant with the association of simple color lexicon with simple technological and cultural development. holo Figure 11 Inferred Color Categories for Jale, Representing Stage I -17- A very interesting variant of Stage I is represented by the Paliyans, a technologically marginal group of Southern India. Our data comes from a recent, unpublished manuscript by P. M.Gardner (1966a) who recently returned from field-work among this group. The Paliyans speak a dialect of Tamil, a major Dravidian language with about 30,000,000 speakers. Plains Tamil has a clear Stage V color terminology, given by Gardner as veUai 'white1, karuppu •black', sivappu 'red1, paccai 'green', manjal 'yellow' and nilam 'blue'. However, in Paliyan Tamil, cognates of five of these six terms are retained, but with radically altered meanings. Discrimination is encoded uniquely 16 on the dimension of brightness. Paliyan color terms with their glosses and standard Plains Tamil sources are given in Table IV. Gardner indicates that there is considerable overlap in usage between each term in the series and its neighbor(s). The effect on the usage of the terms of the extent to which brightness comes from light source as against surface properties is not entirely clear from Gardner's preliminary manuscript. The most extreme terms, velle and karuppu are not reserved exclusively for very extreme degrees of brightness. Gardner says "The usual leaf on a tree is velle on its upper surface and karuppu on its lower surface." (1966a). He also notes, however, that the same leaf may well be nilam or sihappu (presumably on both sides) if seen in different light. In any case, whether or not Paliyan can be construed as a perfect exam-plar of Stage I color lexicon it is certainly a variant of this basic type. For example, "sihappu was elicited for dark shades of red, yellow, green, purple and black" (1966a). Of particular interest for the general evolutionary hypothesis is Gardner's conviction that these people have a minimum of shared culture. He speaks of "imprecision and lack of elaboration in the most basic aspects of Paliyan subsistence related classifications... highyly idiosyncratic taxonomy... de-emphasis on both verbal communication and formality of expression [sic}" (1966a, see also Gardner 1966b:397-399).17 Another example of a Stage I system is taken from W.H.R. Rivers' treatment of the Western tribes of the Torres Straits of New Guinea (including the islands of Mabuiag, Muralag, Badu, Moa and Saibai) (1901). Of the six terms which Rivers elicited as "names in general use", four are clearly descriptive D Table IV Paliyan and Plains Tamil Basic Color Lexicons Paliyan Plains Tamil velle •illuminated (sometimes 'bright')' < vellai ■white' manj a •bright' < manjal 'yellow nilam •of medium brightness1 < nilam 'blue' sihappu •dark1 < sivappu 'red' karuppu 'dark or in shadow1 < karuppu 'black' I L 'JZ ERIC -18- expressions formed by the name of some natural object plus a productive suffix -dgamulnga most appropriately glossed 'it looks like1. Thus, we see kul-kadgamulnga 'red and purple' < kulka 'blood', murdgamulnga 'yellow and orange' < mur 'yellow ochre', ildegamulnga 'green and blue'< il^ 'gall-bladder, bile', maludgamulnga 'blue and green' < malu 'sea'. The forms for 'black' and 'white' are less ameable to analysis, the derivation of 'white' being questionable, i.e., miakalunga merkalunga < merkai? 'spirit', kubikubinga 'black' < kubi 'charcoal, night and darkness'. If we interpret Rivers correctly, only the terms for black and white can be considered as basic color terms while the remaining expressions are best treated as descriptive phrases. In fact, Rivers notes this when he states that "names for unfamiliar colours were apparently invented for the occasion by adding the usual suffix to the name of some natural object and once or twice a native omitted the termination and simply gave the name of the object" (1901:59). Murray Island, for which Rivers examined 107 individuals, may also be interpreted as representing Stage I. Unlike the Torres Straits materials, where a "color term" is formed by suffixation, in this group we find color adjectives formed by "reduplication from the names of various natural objects" (1901:56). Thus, there occur mammam 'red' < mam 'blood', bambam 'orange, yellow' < bam 'turmeric', siusiu 'yellow* < siu 'yellow ochre', s^sk^pus^ske^p 'green' < s8sk^p 'bile, gall-bladder', bulubulu 'blue' < English 'blue', kake-kakek 'white' < ?, golegole 'black' < gole 'cuttlefish' [sicj, pipi 'grey' < p_i_ 'ashes'. The term for white was the only common expression for which Rivers found it impossible to obtain a derivation, and the derivation he offers for black is suspicious. By eliminating bulubulu, clearly a recent English loan, Murray Island falls into Stage I with the Torres Straits data. One point should be made at this juncture concerning what might be considered as "incipient color categories" and Rivers' contribution in this respect. While Rivers noted several times that "many of these... names were devised on the spur of the moment" (1901:56).he was nevertheless impressed by the ap- -19- parent high reliability of many of the more common descriptives. He suggests that this material may be taken as illustrative of vocabulary accretion in the domain of color, and that it should not be surprizing that when a new color category is linguistically recognized that it should be labelled by the names of natural objects. Rivers concludes as follows: "It is probable what when primitive man began to use names for colours, he used the names of natural objects either simply or modified in some way, and that definite generic terms have evolved out of these. The Mabuiag vocabulary (Western Torres Straits tribe given above] is a good example of the coexistence of a large number of special names with a few which have become definitely abstract terms for colour [i. e., black and white]". (1901:63-64). This fact is born out in many languages where, for example, the word for red may be seen historically to derive from a form of the word for blood. (Greenberg 1963:154). 2. 3.2. Stage II systems [introduction of RED] Stage II is exemplified by the African languages Tiv, Tonga and a dialect 18 of Shona. P. Bohannan notes that "In Tiv...all green, some blues, and some grays are ii. But very light blues and light grays are pupu. Nyian, which covers brown, also covers all warm colors through red to yellow. The distinction between ii and pupu actually is not in terms of color, but in term of what we would call shade—darkness and lightness. Very light blue, gray, or white are all pupu. H_ means dark and covers all dark colors and black—unless there is a warm color present; brown, red and yellow are all nyian. Tiv can distinguish colors and do color-blind tests, but their culture does not require--or allow— that they make some of the color distinctions that Westerners make. Westerners are the most color-conscious of peoples" (1963:35-36). A schematic representation of Tiv is seen in Fig. 11. Shona, a Bantu language of Rhodesia, may represent Stage II, if we interpret the report in H. Gleason correctly. Apparently, there exi3ts a unitary pupu Figure 12 Inferred Categories for Tiv, Representing Stage II -20- term including reds and purples. The two remaining terms not only encompass the mid-spectrum hues, but black and white as well. Thus, Gleason notes "interestingly enough, citema also includes black and cicena white" (1961:4) The basic color terms of the Tonga group studied by Colson consist of three apparently cognate forms with the Shona materials reported above: čisia •black and all other dark colors', čituba 'white and all other light colors1, and čisubila 'red, including oranges and dark yellows'. The apparent cognate pairs citema/Čituba and cicena/čisia suggest that the glosses for these forms may be interchanged in Gleason's account. On very speculative evidence, we may interpret Kirchoff's report (1883: 546) on certain Australian languages as State II systems, having definite color names for only white, black and red. This fact, however, is tentative in that it is impossible to determine if any of the terms are descriptives. Some of Rivers' materials indicate that the term for red, oti, elicited by Kirchoff from several Queensland natives, is really the word for blood which would make these groups Stage I. Other State H groups for which we have early historical data are the Tshi of West Africa: fufu 'white', tuntum 'black' and koko 'red', the Todas of India's Nilgiri Hills, the Bantu reported in Buchner (1883) and a New Caledonian group.* The Nasioi of Bougainville are a clear contemporary example of Stage II if the term for red is indeed a basic term and not simply a descriptive. Other-wize, they are Stage I. E. Ogan reports: "I worked a total of twenty-six months among Nasioi speakers. The only words I heard in regular use which might be described as "color terms" were kakara 'white', mutana 'black, dark' and ereren 'red'. The etymology of the last word is clear: erer) 'blood'. I know of no such etymology for the other two words" (1967). Nasioi is a non-Malayo-Polynesian language. -21- 2.3.3. Stage III systems (Introduction of GREEN] 20 Apparent Stage III terminologies are Hanunoo, Eassa and Ibibio. We use Ibibio as the examplary case of Stage III, since E. Kaufman gathered the data using our experimental method. Ibibio basic color terms glossed 'WHITE', 'BLACK', 'RED'and 'GREEN! and are seen displayed in Fig. 13. Hanunoo, also Stage III, is a minor variant of this stage. The term for •BLACK' in Hanunoo, (ma)biru ranges over black, violet, indigo, blue, dark green, dark gray and deep shades of other colors and mixtures; 'WHITE' (ma)lagti? ranges over white and light tints of other colors and mixtures; 'RED1 (ma)rara? includes maroon, red, orange, yellow, and mixtures in which these qualities are seen to predominate; 'GREEN' (ma)latuy includes green and mixtures of green, yellow and light brown. Bassa, a member of the Kwa branch of the Niger-Congo family found in Liberia, is also probably Stage III. According to Gleason (1961:4) Bassa has one term including purples, greens, and blues while another term encompasses reds, yellows and oranges. As Gleason's chart refers only to non-neutral hues, we assume that Bassa also has terms for neutral 'BLACK' and 'WHITE'. The appearance of purple as well as blue in 'GREEN' show Bassa to be an extreme form of the major variant of Stage III. 2. 3.4. Stage IV systems [Introduction of YELLOW] Stage IV is represented by many languages of the world. For Africa we find this stage attested in a dialect of Shona distinct from that reported above with terms for 'BLACK' nema, 'WHITE' cena, 'RED' cuku, 'GREEN' pfumbu and 'YELLOW' sara. Ibo and Urhobo, Nigerian languages, probably represented pure examples of Stage IV until recent times. In Ibo we find basic terms for 'BLACK' pji, 'WH*TE' nzu, 'RED' uhui, and 'YELLOW' odo. The aborigi- Figure 13 Ibibio, Representing Stage III o IC -22- nal term for 'GREEN' has apparently been replaced and we find merely a descriptive phrase agwokwondu meaning roughly 'it has the color of leaves'. In Urhobo we find basic terms for 'BLACK' pbyibi, 'WHITE' afuafu, 'RED' ababare, and 'YELLOW odo. The term for 'GREEN' has apparently been lost under pressure from English and replaced by the English loans grini 'green' and blu 'blue'. In North America we find Stage IV represented by Eskimo, with terms for 'BLACK' girnitak, 'WHITE' gakurktak, 'RED' aupaluktak, 'GREEN' tunayuk-tuk, and 'YELLOW quksutak. (Graburn, 1967). In Central America we find Stage IV systems in many of the aboriginal languages of the area. To our knowledge, all of the Mayan languages of Mexico and Guatemala exhibit Stage IV color terminology. We have chosen Tzeltal, a Mayan language of Chiapas, Mexi co, as the exemplary case of Stage IV as we have collected rather extensive data from this language utilizing the experimen- 22 tal methods discussed earlier. Tzeltal has five basic color terms which are ^ihk' 'BLACK', sak 'WHITE', cah 'RED', yas 'GREEN' and kjan 'YELLCW. The distribution of these terms may be seen in Pig. 13. The treatment of the category yas 'GREEN' in Tzeltal is of particular interest in that some of the data suggest that this language may now be transitional from Stage IV to Stage V. Of the 40 Tzeltal informants for which we have gathered experimental data, 31 indicate that the focal point of yas falls precisely in the area of the spectrum which corresponds to English green. In general usage.- the maximum extension of yas includes greens, blue-greens, blues and some blue purples. However, when greater specification of yas is requested, many informants restrict the term almost exclusively to greens and some blue greens. 'Blues' and 'purple blues' are recognized as a distinct area on the spectrum and are designated by a descriptive phrase, 9ihk' '■'ihktik syaaal 'blackish green' or sim-ply ?ihk'?ihk'tik 'blackish'. In at least one instance, an informant referred to sak Figure 14 Tzeltal, Representing Stage IV -23- fhis area by the Spanish term asul 'blue1. The remaining 9 informants in our sample of 40 have essentially the same maximal extension of yas as the previous 31 individuals (i.e./ over greens and blues) but the foci of the category is squarely in the blue area. When greater specificity was requested for the greens and blue-greens, descriptive phrases were often utilized, e.g., saksaktik syasal 'whitish green1. How might these data concerning the foci of 'GREEN' in Tzeltal best be interpreted? Our most plausible explanation is one which suggests that Tzeltal may be moving from Stage IV to V and the ambiguity of the foci for yas reflects this transitional period. It is apparent to all Tzeltal speakers that yas includes two major perceptual centers, green and blue. In contact with speakers of Spanish over the last 400 years* this fact has most likely been accentuated many times. Speakers of Tzeltal respond by reducing the extension of yas in instances where specificity is required either to greens (for most informants) or to blues (for the minority) and treat the remaining area with descriptive phrases. Tzeltal may continue for many generation to rely on such descriptives to designate what is clearly an incipient color category best glossed as 'blue'. It is our prediction, however, that as Tzeltal speakers become more exposed to Spanish in the schools yas will eventually be restricted entirely to greens and that asul or some other Spanish term will be encoded for the perceptual category 'blue', making Tzeltal a legitimate V system. 2.3.5, Stage V systems [Introduction of blue] Thus far. Stage V color lexicons have been found only in Africa and Southern India, although Mandarin may also, on further research, prove to be Stage V as well. We have already discussed Stage V Plains Tamil (see above) which exhibits terms for 'white' vellai, 'black' karuppu* 'RED' sivappu* 'green' ...... 23 paccai, 'YELLOW' manjal and 'blue' nilam. The inferred distribution of these terms is seen in Fig. 15, Figure 15 Inferred Categories for Plains Tamil, Representing Stage V" o ERIC -24- In Africa, Stage V vocabularies are found for the Nupe, the Massai, 24 Bedauye and at least some dialects of Hausa. The basic color terms of these African languages are presented in Table V. 2.3.6. Stage VI systems [Introduction of brown] Stage VI systems are rather sparsely represented in our sample but thus far we have found examples in Southern India, Africa, and North America. The color terminology of Nez Perce, an Indian language of the state of Wash- 25 ington, is depicted in Fig. 15. Stage VI is also seen for some dialects of Malayalam of Southern India, e.g., vella 'white, kadup:- 'black1, cuwj?pp:> 'RED1, pacca 'green', manna 'YELLOW_ntta 'blue', tavfta 'brown' (Goodman 1963:9-10). We have at least two examples of this stage in Africa, the Bari and the Siwi. Bari terms are -kwe 'white', -ruo 'black', -tor 'RED', -ngen 'green1, -forong 'YELLOW', -murye 'blue' and -jere 'brown'. Siwi terms are aztuf 'white', amilal 'black1, ozgahh 'RED', owrarr 'green', lasfarr 'YELLOW', asmawee 'blue' and lasmarr 'brown'. 2.3.7. Stage VII systems Stage VII is represented by 20 of the 80 languages in our sample, and varying tvoes of this stage are found widely in the world's languages. It may eventually prove possible to establish some partial ordering of terms as described for the preceeding stages. We cannot, at this time, however, offer a sequence for the appearance of basic color terms after stage VI. Many more languages must be examined in detail, both synchronically and historically, before we can present a more definite report in this regard. Noting only languages in our experimental sample, we may suggest tentatively that Urdu is "early" Stage VII having terms for black, white, red, cimu- xcimux ;;;; xayxayx Figure 16 Inferred Categories for Nez Perce, Representing Stage VI :rjc Table V Basic Color Terms in Four Stage V African Languages Gloss Massai Bedauye Hausa Nupe white eborr era baki bokun black erok hadal fari zukb RED enyki adaro dzuru green mbusth sotay algashi YELLOW ngirro asfa rawaya wonjin blue ainyori delif shudi dofa -25- green, yellow, blue, brown, and purple, but lacking orange, pink and grey. Likewise, Cantonese has yet to add brown, purple or orange to its basic inventory, (cf. section 2.5.). Tagalog lacks a term for orange as does Vietnamese. Finally, Catalan lacks pink and orange terms. Hungarian presents a special case. It has bi sic terms for the ten basic categories exclusive of red and two basic terms for red. Should this finding be born out on further research, it may be possible to suggest additional developmental stages other than those already mentioned. Similarly, Russian, as well as several other Slavic languages is reported to have two basic terms for blue. A total summary of the available data relevant to the evolutionary hypothesis is given in Table VI. All interpretable reports found are summarized in this Table which gives for each language listed, its stage* type, and the source of the data. (The actual categories for each color lexicon type are given in Table 1). All languages examined confirm the evolutionary hypothesis in each detail except as noted in section 2.5. English, a typical Stage VII eleven term system has been depicted in Fig. 10, 20 4. Internal reconstruction of basic color terms The principles of internal linguistic reconstruction outlined by Edward Sapir (1916), and employed by Romney (1967) in his treatment of Yuman kinship suggest that several of the languages in our sample have only recently acquired basic color terms characteristic of Stage VII. Korean is an interesting example which illustrates the effect of foreign influence on the formation of new color terminology. Korean has basic color terms (bound forms accompanied by a suffix meaning roughly 'color') for 'BLACK', 'WHITE', 'RED', 'GREEN', and 'YELLCW. These expressions are clearly indigenous Korean forms. Terms for pink, orange, (chestnut) brown, brown, green, blue, purple and grey, however, are of obvious Chinese derivation as can be observed in the following forms: Old Korean terms — kamata 'BLACK', hayata 'WHITE', palgata'RED' 9 ERLC Table VI Classification of 80 Languages in tarms of Evolutionary Stage of Basic Color Lexicon Stage I (BLACK, WHITE) Language Jale Murray Island Ngombe Paliyan Torres Straits Area New Guinea New Guinea Africa South India New Guinea Source K. - F. Koch, (1966) Rivers (1901) Stapleton (1903) Gardner (1966a) Rivers (1901) Stage II (BLACK, WHITE, RED) Arawak South America van Wijk (1959) Baganda Africa van Wijk (1959) Bantu Africa Rivers (1901) Bulu Africa von Hägen (1914) Ila Africa Smith (1907) Kongo Africa Stapleton (1903) Lingala Africa Anderson (1966) Nasioi Bouganville Ogan (1967) Ndembu Africa Turner (1966) [New Caledonia group] South Pacific Rivers (1901) Ngbandi Africa Lekers (1908) Pomo California Corson (1966) Poto Africa Stapleton (1903) [Queensland group] Australia Rivers (1901) o ERJ.C Table VI (Page 2) Language Area Source Sango Africa Anderson, (1966) Shona Africa Gleason (1961) Tlv Africa Bohannan (1963) Todas India Rivers (1901) Tonga Africa Colson (1966) Tshi Africa Rivers (1901) Yibir Africa Kirk (1905) Stage HI (BLACK, WHITE, RED, GREEN) Bagirmi Africa Gaden (1909) Bangui Africa Staplcton (1903) Eassa Africa Gleason (1961) Batak Malaya van Wijk (1959) Bullom Africa Nylander (1814) Hanunoo Philippines Conklin (1955) Ibibio Africa E. Kaufman (1966) Poul Africa Faidherbe (1882) Somali Africa Kirk (1905) /Xam Africa Bleek (1956) Stage IV (BLACK, WHITE, RED, GREEN, YELLOW) Chinook Jargon Canada T. Kaufman (1967) Eskimo Canada Graburn (1967) Ewe Africa Migeod (1908) Daza Africa Le Coeur (1956) Duhomeen Africa Delafosse (1894) Ibo Africa Goldberg (1966) Table VI (Page 3) erjc Language Papago Shona Songhai Tzeltal Tzotzil Urhobo Wolof Area Southwest U. S. Africa Africa Mexico Mexico Africa Africa Source C'Neale and Dolores (1943) Goldberg (1966) Prost (1956) Berlin (1967) Collier (1963) Goldberg (1966) Rambaud (1903) Stage V (black,, white, RED, green, YELLOW, blue) Bedauye Hausa Masai Mandarin Nupe Samal Plains Tamil Africa Africa Africa China Africa Philippines South India Reinesch (1895) Robinson (1925) Hinde (1901) McClure (x966) Banfield (1915) Geoghegan (1967) Gardner (1936a) Stage VI (black, white, RED, green, YELLOW, blue, brown) Bari Nez Perce Malayalam Siwi Africa Northwest V, S. South India Africa Owen (1908) Aoki (1967) Goodman (1963) Walker (1921) Stage VII (all eight-, nine-, ten- and eleven- term lexicons) Arabic Bulgarian Lebanon Europe Kay (1967) Forman (1967) Table VI (Page 4) Language Catalan Cantonese Dink a English Hebrew Hungarian Indonesian Japanese Korean Nandi Russian Spanish Swahili Tagalog Thai Urdu Vietnamese Zuni Area Europe China Africa U. S. Israel Europe S.E.Asia Japan Korea Africa Soviet Union Mexico Africa Philippines Thailand India Vietnam Southwest TJ. S. Source Corson (1967) Stross (1967) Nebel (1948) Berlin and Kay (1967) McClaren (1967) Madarasz(1967) Madaraszd 96 7) McClure (1967) Steger (1967) Madaraszd 96 7) Hollis (1909) Slobin (1967) Stross (1967) Madarasz(1967) Frake (1967) Forman (1967) McClaren (1967) Madarasz(1966) Lenneberg & Roberts (1953) -26- norata •YELLOW, and parata 'GREEN'. Chinese loans — punhon-sek 'pink' tyn-sek 'orange', kal-sek 'brown', pam-sek '(chestnut) brown', nok-sek 'green', tson-sek 'blue', tsa-sek 'purple1, and ke-sek 'grey'. Cantonese Chinese is also only recently Stage VI, judging by internal reconstruction of its color vocabulary. In dictionaries of 100 years ago, the term for 'pink' sui does not occur. The present meaning can be shown to be best translated as 'water colored'. The terms lok 'jade colored' and la;m 'artificial blue' are also recent category labels which now segment GREEN. V/e may, therefore, reconstruct a Stage IV Cantonese with the following terms: bak 'WHITE', hek 'BLACK', hon 'RED', cen_ 'GREEN' and won. 'YELLOW'. That we find fui_ 'grey' is somewhat anomalous. However, there is some evidence that it refers to 'ashes', and, if so, can be eliminated. The internal reconstruction for the African Creole Swahili is relatively interesting in that it may be Stage n, having ancient terms only for 'BLACK' neusi* 'WHITE' neupe and 'RED' nekundu. The term for green kidjani may be new in that it might be glossed as 'leaf green'. The remaining terms are des-criptives or loan words, i.e., kidjivu 'grey' < 'ashes'* kicunwa 'orange' < 'orange fruit', hudhuruni 'brown' < Arabic 'brown', kimandjano 'yellow' < 'turmeric', bulu 'blue' < English 'blue'* arudjwani 'purple' < Persian 'purple'. The appearance of new terms in some of the European languages is also indicative that on internal evidence these languages reconstruct to earlier stages. Bulgarian, for example, has borrowed terms oranzh