The Use of Modifying Terms in the Naming and Categorization of Color Appearances in Vietnamese and English

Cross-cultural studies of color naming show that basic terms are universally the most frequently used to name colors. However, such basic color terms are always used in the context of larger linguistic systems when speciŽ c properties of color experience are described. To investigate naturalistic naming behaviors, we examined the use of modiŽ ers in English and Vietnamese color naming using an unconstrained naming task (Jameson & Alvarado, in press). Monolingual and bilingual subjects named a representative set of 110 color stimuli sampled from a commonly used color-order stimulus space. Results revealed greater reliance upon polylexemic naming among monolingual Vietnamese speakers and greater use of monolexemic basic hue terms and secondary terms (object glosses) among monolingual English speakers. Systematic differences across these language groups imply that widely used monolexemic naming methods may differentially impact color-naming Ž ndings in cross-cultural investigations of color cognition. ¤Send correspondence to: Nancy Alvarado, IBM, T.J. Watson Research Center, P.O. Box 704, Yorktown, NY 10598 or [email protected]. Kimberly A. Jameson, Department of Psychology, University of California, San Diego. This research was supported by NIMH grant RO3-MH53126-01 (Jameson) and Hellman Faculty Fellowship Award (Jameson). Additional support was provided by IBM, T.J. Watson Research Center (Alvarado). The authors gratefully acknowledge the assistance of undergraduates A. Lewis, A. Nguyen, A. Garcia, I. Moy, and H. Ngo. The authors would like to thank Margaret Chan, IBM, T.J. Watson Research Center, for her assistance in the translation of Mandarin Chinese, and for general assistance on features of Asian languages. ¤¤IBM, Thomas J. Watson Research Center. ¤¤¤Department of Psychology, University of California, San Diego. c ° Koninklijke Brill NV, Leiden, 2002 Journal of Cognition and Culture 2.1 54 NANCY ALVARADO AND KIMBERLY A. JAMESON Since the 1960’s, the cross-cultural study of color-naming and color categorization has served as an in uential example of a pan-human cognitive universal (see Hardin and MafŽ , 1997 for a review). Yet, some theorists question universal models of color categorization and naming, and have warned that it makes little sense to ignore the richness of language in favor of an exclusive focus on a handful of basic color terms thought to arise from color-opponent visual processing mechanisms (Hewes, 1992; Wescott, 1992; Zimmer, 1984). Nevertheless, the search for universals in the semantic structure of the color lexicon has largely focused upon a small number of focal hues or a small number of basic color terms, or both. Even in large-scale surveys such as Berlin & Kay’s (1969) early work and the later World Color Survey (WCS), the primary aim was to determine whether color lexicons and color-naming behaviors are represented by 11 focal color categories named by basic terms arising from cognitive universals (Kay & MafŽ , 2000). This investigative emphasis on the construct of universal basic terms for color appearances continues in a number of recent cross-cultural studies of color categorization and naming. For both practical and empirical reasons, such investigations often constrain the naming behavior of informants assessed to monolexemic responses (e.g., Sturges and WhitŽ eld, 1997), or they assess naming for small subsets of color appearance stimuli (e.g., Moore, Romney and Hsia, 2000). Investigations that incorporate such constraints are essential for testing the cross-cultural validity of the theory of eleven basic color terms, but they do not always permit the assessment of systematically occurring cross-cultural differences that might be seen under more naturalistic color naming and categorization. In particular, Ž nding cross-cultural universality of color naming using Berlin and Kay’s eleven basic color terms does not automatically imply universality of color-naming behavior when assessing a different set of color naming terms, such as the color labels occurring most frequently in individual discourse. Thus, while the existence of basic color naming universals has been demonstrated many times (Hardin & MafŽ , 1997), the process of boiling a lexicon down to basic terms may obscure important cultural differences in both the use of language to name color samples and the perception of the stimuli themselves, as discussed by Jameson (2001). This “oftused constraint of forced monolexemic naming” (Guest & Van Laar, COLOR MODIFIER USE IN TWO LANGUAGES 55 2000, pg 731) using basic color terms, may empirically produce a greater appearance of universality than truly exists. Such a result is entirely possible if in daily communication individuals generally use linguistic features and vocabulary to describe their perceptual experiences that are denied them in empirical studies. This emphasis on monolexemic basic color labels, and the perceptual categories they describe, arises from the deŽ ned construct of universal basic color terms originally formulated by Berlin and Kay (1969). However, a study of unconstrained color naming by Jameson & Alvarado (in press), comparing English and Vietnamese language groups, found substantial variation in the use of polylexemic names, modifying terms, compounds, and object glosses to name color stimuli. These Ž ndings suggest an alternative perspective, explored here. We begin with the idea that color naming typically incorporates linguistic features beyond monolexemic labels, and that such features can also be examined for cross-cultural similarities that may exist due to speciŽ c linguistic constructions shared by some languages. In other words, we suggest that the lexical color labels that best describe the central exemplars of a culture’s color category partitions may not always be monolexemic terms. Whether they are is a matter for empirical investigation. To evaluate this, we studied similarities across language groups in the use of modiŽ ers and other linguistic constructions to name color appearances. It is important to note that the need for a strong test of the monolexemic assumption is not widely discussed in the existing literature, although it is crucial to demonstrating universal basic color term generality. At times the monolexemic assumption is found implicit even in investigations that use unconstrained naming methods. For example, Guest and Van Laar (2000) studied unconstrained naming in English participants using computer-diplayed color appearances. In accord with the Kay-BerlinMafŽ theory, Guest and Van Laar found greater use of basic names than non-basic names and greater use of monolexemic terms than polylexemic terms. Upon close examination, however, these results cannot be generalized cross-culturally. This is because the index of naming consistency they deŽ ne incorporates as parameters characteristics they observed in their English color-naming results. (SpeciŽ cally, a “modal term” is deŽ ned as consisting of a hue term with or without some modifying term or com56 NANCY ALVARADO AND KIMBERLY A. JAMESON pound. The additional words in a polylexemic name are assigned a lower value and weight the modal name in calculations of naming consistency.) As deŽ ned, their index privileges both hue stem terms and monolexemic naming in its calculations. We believe that this can only be assumed in cross-cultural research if the relative frequencies of use of monolexemic stem terms and polylexemic, basic and non-basic terms are known to be the same across languages. Thus, by noting their overall high frequency of use of unmodiŽ ed basic terms (63.7% of English responses) and generally concluding: “This Ž nding suggests that the oft-used constraint of forced monolexemic naming may not be as constrictive as one might fear” (pg 731), they fail to acknowledge the ethnocentric assumption inherent in their analysis which implies that all languages will make use of hue denotata and monolexemes with similar frequencies. Further, although polylexemic naming may indicate difŽ culty in naming samples at the boundaries of categories for English speakers, we do not know whether polylexemic naming indicates difŽ culty in naming in other languages, or only in those languages emphasizing hue. Universal Patterns of ModiŽ er Use and Object Glosses in Color Naming Several previous investigators have studied the use of modifying terms in color naming. Burgess, Kempton & Maclaury (1983) studied the use of modiŽ ers as support for a theory of evolution toward a hue-based naming system. They use Maclaury’s three-stage mapping process to assign modiŽ ers to regions of the Munsell color space. However, their results are not easily compared to the results of modiŽ er use presented here for English or Vietnamese due to structural differences between the languages tested and differences in the types of distinctions captured by modiŽ ers. Thus their research further illustrates the problems involved in assuming that English color naming patterns are found in other ethnolinguistic cultures. As described by Burgess et al. (1983), the Tarahumara language employs a postposed bound modiŽ er for each color stem term that speciŽ es the relation of the currently named color to the center of the category. For each color name, the modiŽ er KAME is used for colors near the center of the category, the modiŽ er NAME refers to colors further from the center, and the modiŽ er NANTI refers to colors on the fringes of the category. COLOR MODIFIER USE IN TWO LANGUAGES 57 Thus, for red (SITA), these stem term and modiŽ er combinations would translate as “very red” (SITAKAME), “somewhat red” (SITANAME), and “slightly red” (SITANANTI). No color names appear without these postposed modiŽ ers and the modiŽ ers are also used in naming things other than colors. This system of naming is very different than that used in Chinese, Vietnamese, and English where modiŽ ers consist of one or more separate words preceding a color stem term (in English) or following it (in Vietnamese), and relational distinctions are subordinated to other distinctions (although the modiŽ er VERY and its translation were used in both languages). Together with our own previous unconstrained naming results (Jameson & Alvarado, in press), these Ž ndings of Burgess et al. (1993) seem to justify a closer examination of modiŽ er use in color naming across different ethnolinguistic cultures. Among the identiŽ cation criteria originally proposed by Berlin and Kay (1969) to deŽ ne basic color terms is the requirement that candidate basic terms are words which have acquired a meaning independent of both: (a) the object whose color is being named and (b) the context in which naming occurs. Even in cultures with few abstract color names (Berlin & Kay’s early stages), an abundance of color names bound to objects or contexts are used (Schirillo, 2001). In the present study, we refer to color names derived from these contextualized color terms as “object glosses.” In English, which has a full complement of 11 basic color terms, unmodiŽ ed object glosses (e.g., BRICK, OLIVE) are widely used to capture Ž ne distinctions in naming (Jameson & Alvarado, in press). Critics of the concept of “basic terms” point out that many English basic color names began as object glosses. For example, Hewes (1992) states that PURPLE began as the name of an “exotic commodity derived from a mollusk” (p. 163). Compounding of object glosses with basic stem terms is frequent in English and Vietnamese, constituting a form of modiŽ er (e.g., TURQUOISE BLUE, SKY BLUE, XANH LA CAY translated as “leaf green”). Lin, Luo, MacDonald, and Tarrant (2001a, 2001b) compare color modiŽ er use for Mandarin Chinese speakers with British speakers of English. In addition to constructing a database of Mandarin color terms (Lin et al., 2001a), they use subject judgments to map speciŽ c modiŽ ers to regions of the Swedish Natural Color System space for both languages 58 NANCY ALVARADO AND KIMBERLY A. JAMESON (Lin et al., 2001b). Moreover, as noted by Lin et al. (2001a), features of the Chinese language may encourage this compounding of object glosses as modiŽ ers or as modiŽ ed contextualized color names. First, Chinese characters are frequently constructed by compounding several characters into a single character with a new meaning. All three of the “secondary terms” listed by Lin et al. (2001a, Table II) contain the character for “sky” as a constituent element. The terms DAI, BE and ZAN (translated as JADE or DYE by Lin et al., 2001a) could also be more poetically translated as references to a dark sky before a rain, a clear sky, and a blue sky near sunset, respectively. Second, in Chinese, all color terms consist of two-character pairs. The Ž rst character denotes the color category and the second character speciŽ es that it is a noun referring to the color appearance itself. These pairs might be more strictly translated as “the red color” rather than simply “red.” Thus each of these names qualiŽ es as an abstract color name not bound to any object or situation and each meets Berlin & Kay’s (1969) abstractness criterion for basicness (if not the other criteria). In Chinese, when the color name is used as an adjective, the second character is the name of the object taking that color. This practice also occurs in Japanese, where a different second character is used to differentiate chromatic and achromatic color names, as discussed by Lin et al. (2001a). Vietnamese does not afŽ x a second word meaning “color” to color names in its phonetic writing system, but like Japanese, Vietnamese employs both a phonetic and a Chinese character-based writing system. Thus, when using characters, both languages may habitually denote even basic colors using multiple-word phrases. Lin et al. (2001a) also list several alternative basic names that are more metaphorical, Ž gurative, or poetic, expanding the 11 basic color terms. They argue on the basis of frequency of use that these deserve consideration as basic terms. For example, HONG is red in standard Chinese but JU refers to the color red of royal clothing. It brings with it associations of majesty and privilege, and is thus a special red. However, these alternative terms are arguably more culture speciŽ c and more likely to be used differently in different dialects. For example, Moore et al. (2000) include as their name for basic brown the character for “coffee” accompanied by a second character denoting that this is a color name. It is an object gloss characteristic of the Taiwanese Mandarin dialect, not used COLOR MODIFIER USE IN TWO LANGUAGES 59 to name brown in standard Chinese. Lin et al. (2001a), who also studied Taiwanese Mandarin, do not list this as a high frequency alternative term for brown, though it was used by some of their subjects (along with “tea”). By Berlin and Kay’s criteria, this complicates evaluation of the basicness of the Taiwanese Mandarin term for brown used by Moore et al. (2000), and underscores the difŽ culty in untangling whether Ž ndings of cross-cultural universality originate from the few basic terms they assessed, coupled with large differences among color appearance stimuli, or from consistencies in basic color naming and perception. The impact of culture is more obvious in use of these alternative, object-derived color names, richer in connotation (Schirillo, 2001; Hewes, 1992), but that impact may become evident only when comparisons are made that reveal Ž ner distinctions. To systematically describe the frequency and use of non-monolexemic linguistic constructions in color naming, different researchers have applied different taxonomies for classifying the color naming behaviors from different linguistic cultures. These differing classiŽ cation strategies, typically dictated by different assumptions, make comparisons difŽ cult across studies, much less languages. Lin et al. (2001a) applied one version of Simpson and Tarrant’s (1991) seven-category taxonomy: (1) basic color terms; (2) modiŽ ed basic terms (e.g., DARK RED); (3) compound terms (e.g., BLUISH-RED); (4) qualiŽ ed basic terms (e.g., DARK BLUISH RED; (5) secondary terms (e.g., CARDINAL RED); (6) idiosyncratic (no obvious pattern); (7) unnamed (no name given). This scheme classiŽ es object glosses as “secondary terms.” They then boiled these seven categories down to four, for further comparisons: (1) basic; (2) modiŽ er; (3) compound; (4) secondary. Guest and Van Laar (2000) used a different version of Simpson and Tarrant’s seven-category system, one that differentiates hue modiŽ ed terms from lightness modiŽ ed terms, but classiŽ es secondary terms as “other monolexemic” and similarly divides them by modiŽ er type. This latter scheme is more similar to the one used in the present study, though lightness and “other” modiŽ ers (e.g., BRIGHT, VERY) are combined in our study and only hue-based modiŽ ers are counted separately (as compounds or object glosses). Because we were interested in directly examining use of object glosses (secondary terms), we classiŽ ed them both with and without modiŽ ers, separately from compound names and modiŽ ed basic terms. Former object glosses now used to name basic 60 NANCY ALVARADO AND KIMBERLY A. JAMESON color categories in English (e.g., ORANGE, PURPLE) were not considered object glosses in either language. Universal Mapping of ModiŽ ers and Object Glosses to Color Appearances Another method used to compare color naming across linguistic cultures involves examining the spatial mappings of color terms from different languages as they correspond to color appearances in a common color ordered space (e.g., Moore et al., 2000). While such mappings have been used liberally in studies of universal basic color categories and basic terms, the application of such analyses to study cross-cultural similarities of modiŽ er use has been infrequent. In their discussion, Lin et al. (2001a, 2001b) relate their results to Kelly’s Universal Color Language (UCL) for describing surface color appearances (Kelly & Judd, 1976). Kelly’s classiŽ cation system spatially organizes color descriptions by their correspondence to color space regions and by levels of precision. The lexical classiŽ cation speciŽ es that the Ž rst and least precise level consists of the generic hue names or basic color terms studied in most cross-cultural research on color naming. Kelly’s second level consists of the entire range of names in the lexicon, including compound names such as “yellowishbrown” and object glosses such as “coffee” or “chocolate.” Kelly’s level three consists of all of the level two hue names together with modiŽ ers. In Kelly’s taxonomy, modiŽ ers include any word used as an adjective to modify a hue term. Such modiŽ ers typically refer to saturation or lightness (e.g., “light,” “bright,” “pastel”) but might also describe other dimensions of the stimulus (e.g., “ uorescent,” “peaceful,” “dirty”). These latter forms of modiŽ ers were not studied by Lin et al. (2001b) though they note their importance (2001a). Guest and Van Laar (2000) classiŽ ed them as “Other, complex”. Kelly’s proposed naming system assumes that the use of modiŽ ers permits increased Ž neness of description using the lexicon and thus greater precision in naming. Levels four through six consist of color notation systems (e.g., Munsell) measured at increasing levels of precision. (For a review of models of color space, see Kuehni, 2001.) Using this naming system, Kelly (Kelly & Judd, 1976) assigned modiŽ er names to regions of Munsell space (the same space sampled by Berlin and Kay), as shown in Figure 1. While Kelly did not intend them to COLOR MODIFIER USE IN TWO LANGUAGES 61 Figure 1. Universal Color Language map of modiŽ er use by Munsell value and chroma within hue. Note: From “Color: Universal language and dictionary of names,” by K. Kelly and D. Judd, 1976, U.S. Department of Commerce, NBS Special Publication 440. apply to languages other than English, these maps were used as predictions for naming behavior observed in our study. The questions explored in our research include the following. First, are there differences in the naming strategies of the English and Vietnamese languages and if so, how do they relate to the use of monolexemic color 62 NANCY ALVARADO AND KIMBERLY A. JAMESON terms? Second, are modiŽ ers applied to samples according to the maps proposed by Kelly and Judd (1976), Guest and Van Laar (2000) and Lin et al. (2001b)? Third, what is the effect of bilingualism on modiŽ er use? After considering these questions we return to discussing what our Ž ndings imply for studies of color naming universality. The data reported here was collected under precisely controlled viewing conditions in a study of unconstrained naming behaviors for a set of standardized color papers sampled to represent a wide range of color appearances. In Jameson & Alvarado (in press) we hoped to discover whether the previously noted perceptual salience of certain category focal colors would in general contribute to greater agreement in naming under the less constrained naming conditions used. To examine this issue across language groups, we compared color-naming behaviors in Vietnamese and English, two languages with different linguistic categories for green/blue and yellow/orange. We also investigated the impact of the level of access to terms on naming behavior by including bilingual Vietnamese living in the U.S.A. as a comparison group. We found that when subjects were unconstrained by empirical demands, their naming behavior (1) did not differentiate the salience of focal colors from that of non-focal colors, and (2) showed group differences in the use of polylexemic names, modifying terms, compounds, and object glosses. Here we present new analyses of the color-naming behaviors listed in (2), and present an explanation of the important implications of such differences for theories of color cognition.