Componential analysis is a method of describing the subject matter of a language. It aims at constructing verifiable models of how specific bodies of cultural (or ideational) content are coherently organized, insofar as such content is represented by words and expressions in a people’s language. A method in both semantic and cultural description, componential analysis is perhaps best characterized as a method of ideography.
History. In Coral Gardens and Their Magic (1935), Malinowski demonstrated the immediate relevance of descriptive semantics for ethnography. But the application of rigorous method in this area began only after World War ii, inspired by the methodology of structural linguistics and developed and utilized by anthropologists trained in this discipline (Goodenough 1957). Indeed, the term “componential analysis” is taken from linguistics, where it is used to refer to the criteria by which distinctive categories of sound in a language are distinguished and, subsequently, to refer to the analysis of semantic distinctions encountered in grammatical paradigms (Harris 1948).
The first ethnographic use of the method illustrated the categorization of kinship relations in Truk (Goodenough 1951). In 1955 Conklin published the conclusions of a similar analysis of Hanunoo color categories. Then Lounsbury (1956) and Goodenough (1956), working independently, simultaneously published extended statements of the method, again utilizing kinship materials. Since then several such kinship studies have been published (see Bibliography). Wallace and Atkins (1960) were concerned with how far the results of such analysis reflected the actual cognitive organization of phenomena in the minds of the people being studied. How ideational models constructed by componential analysis actually related to “psychological reality” became a subject of debate (Frake 1962; Burling 1964). A conference to discuss this and other matters relating to componential analysis was held in June 1964 under the sponsorship of the Wenner-Gren Foundation for Anthropological Research (Hammel 1965). Conklin (1962) observed that the method is applicable to the analysis and description of folk taxonomies generally and to a wide range of problems in lexicography and ethnography. Frake began to explore its use in the analysis of such aspects of culture as the classification and diagnosis of diseases (1960) and the organization of religious ideas (1964).
As of 1964, however, the clearest expositions of the method, of the theoretical issues it raises, and of its limitations by comparison with other methods for accomplishing similar descriptive objectives have been presented in connection with the analysis of kinship terminologies (Lounsbury 1964a; 1964b). Because kinship has been studied more intensively and extensively by anthropologists than any other single aspect of culture, it is one of the few cultural domains for which anthropologists are readily able to meet the data requirements of the method. Minimally adequate notations for recording the data are available, and there is a fairly well-developed metalanguage for talking about the properties of genealogical space. To go beyond kinship is to face the arduous but scientifically important task of developing suitable notations and metalanguages.
Specific aims of the method. In the terminology of semiotics (Morris 1938), a linguistic expression designates a class of images or concepts, it denotes a specific image or subclass of images within the class on any one occasion of its use, and it signifies the criteria by which specific images or concepts are included or excluded from the class of images or concepts that the expression designates. Thus, what is signified are the definitive attributes of the class. An expression connotes other images or concepts that people associate with the expression’s designatum but which are not themselves definitive attributes of the designated class. To say that the sky is cloudy is to designate a class or type of meteorological condition, to denote a specific image of such condition, to signify the definitive attributes of the class, and to connote such things as chill and rain. Componential analysis is concerned solely with the significational aspect of meaning. Thus, it differs sharply from most word-association approaches to semantics, which deal almost entirely with connotations (e.g., the “semantic differential” technique described in Osgood et al. 1957).
In its concern with signification and definitive attributes, componential analysis starts with extensional definitions (listings of denotata) and seeks to reduce them to intensional definitions. For example, an extensional definition of the English kinship term uncle would list such denotata after it as mother’s brother, father’s brother, mother’s half-brother, father’s half-brother, mother’s sister’s husband, mother’s half-sister’s husband, father’s sister’s husband, and father’s half-sister’s husband. An intensional definition might be as follows: An uncle is any kinsman by blood or marriage who is simultaneously (a) male, (b) two degrees of genealogical distance from ego, (c) not lineal, (d) in a senior generation, and (e) not connected by a marital tie in other than the senior generation of the relationship.
As this example shows, an intensional definition is conjunctive, seeking to reduce the disjunctive extensional definition to a unitary class described as a product of the combination of several definitive attributes. That these are definitive attributes in this case is evident from the fact that to vary any one of them results in the judgment that uncle is impermissible as a term of reference. Aunt becomes the appropriate term if we vary (a) sex; great-uncle if we vary (b); grandfather if we vary (c); nephew if we vary (d); and wife’s uncle or husband’s uncle if we vary (e) . This, incidentally, illustrates a way in which the results of componential analysis are capable of verification.
General procedure in the method. We arrive at definitive attributes by a combination of two operations: inspecting the set of a term’s denotata for common attributes and contrasting the set of its denotata with the sets of denotata of other terms. The latter operation is the crucial one. It leads us, among other things, to recognize hierarchies of subject matter or of semantic domains.
To contrast the denotata of English dog and mackerel is to result in a bundle of discriminating features that also discriminate between cat and pickerel, dog and pickerel, and cat and mackerel. This provides a basis for distinguishing two semantic domains, the one to which cat and dog belong and the one to which mackerel and pickerel belong. That mammal and fish are cover terms designating these two domains shows us one kind of structural relationship between the significata of linguistic forms. The significatum of dog contains all the definitive attributes in the significatum of mammal plus some additional ones that discriminate its denotata from those of cat, horse, etc.
Hierarchical relationships among the significata of words and expressions seem obvious enough to speakers of English in an English example, but they appear to be characteristic of at least large portions of vocabulary in all languages. It is thus possible to sort vocabularies into distinct sets pertaining to different domains of experience. Componential analysis helps us to determine what “goes together” in unfamiliar languages so that we do not arbitrarily sort them on the basis of rough translations or glosses into the domains of English (or other language of description). In the language of Truk, raaw (“whale”) belongs to the domain designated by the cover term iik, along with most things that we would subsume under the cover term fish in English. There seems to be no semantic domain in Trukese that corresponds to the domain designated by English animal.
Componential analysis has been directed primarily at systematically contrasting the sets of denotata of the several expressions within single domains or subdomains. Analysis has shown that the several expressions within a domain can be sorted into sets so that all the expressions in a set have mutually exclusive denotata at a given hierarchical level and differ from one another with respect to one or several dimensions of discrimination (such as the several dimensions used to discriminate uncle from nephew, aunt, etc., in the example given above). Such a set of expressions constitutes a terminological system. The method of componential analysis has been applied almost entirely to delimiting and depicting the ideational structure of terminological systems.
Illustration of the method–Moala kinship terminology. Moala is an island in Fiji whose social organization has been described by Sahlins (1962). His published data, which appear in Table 1, provide the material for analysis here. The definitions
|Table 1 – Moala kinship data|
|a. Abbreviations used: Fa, father; Mo, mother; Br, brother; Si, sister; So, son; Da, daughter; Hu, husband; Wi, wife; Sp, spouse; Co, cousin; // Co, parallel cousin; XCo, cross cousin; and Ch, child.|
b. Correction of error in source.
Source: Adapted from Sahlins 1962.
|1. famaqu||male or female||Fa, FαBr, Fα ♂ // Co, FaFa ♂ // CoSo, FaMo ♀ // CoSo, Mo ♀ // CoHu, MoMo ♀ // CoDaHu,|
Hu of any tinaqu, Mo ♂ XCo, WiMoBr, WiMo ♂ // Co
|2. tinaqu||male or female||Mo, MoSi, Mo ♀ // Co, MoMo ♀ // CoDa, MoFa ♂ // CoDα, Fα ♀ XCo, Fa ♂ // CoWi,|
FaFa ♂ // CoSoWi, Wi of any famaqu, HuFaSi, HuFα ♀ // Co
|3. luvequ||male||Ch, BrCh, ♀ XCoCh, ♂ // CoCh, Ch of any ♂ Ch of any famaqu or tinaqu,|
Fa ♀ // CoDaCh, Mo ♂ XCoSoCh,b WiSiCh, Wi ♀ // CoCh, BrWiSiCh, BrWi ♀ // CoCh
|female||Ch, SiCh, ♀ // CoCh, $ XCoCh, Ch of ♀ Ch of famaqu or finaqu, Mo ♀ // CoDaCh,|
Mo ♂ XCoDaCh, Fa ♀ XCoDaCh, Fa ♂ XCoSoCh, Fa ♂ // CoDaChb, Fa ♀ // CoSoCh,
HuBrCh, Hu ♂ // CoCh, SiHu ♂ // CoCh
|4. taciqu||male||Br/, ♂ // Co, Fa ♂ // CoSo, Mo ♀ // CoSo, all So of all famaqu or finaqu, all Fa of luvequ|
|female||Si, ♀ // Co, Mo ♀ // CoDa, Mo ♂ XCoDa, Fa ♀ XCoDa, Da of finaqu or famaqu, Mo of luvequ|
|5. weJcaq u|
|male||Si, ♀ // Co, Mo ♀ // CoDa, Fa ♂ // CoDa, Fa ♀ XCoDa, all Da of famaqu or finaqu|
|female||Br, ♂ // Co, Fa ♂ // CoSo, Fa ♀ XCoSo, Mo ♀ // CoSo, Mo ♂ XCoSo,b|
all So of famaqu or finaqu
|6. vugoqu||male or female||MoBr, FaSi, Mo ♂ // Co, Fa ♀ // Co, Br of any finaqu, Si of any famaqu, ♂ XCo of any|
famaqu, ♀ XCo of any finaqu, Ch of any wefcaqu, SoWi, DaHu, So of any luvequ
|male||SiCh, ♀ // CoCh|
|female||BrCh, ♂ //CoCh|
|7. tavalequ||male||MoBrSo, FaSiSo, Mo ♂ // CoSo, ♂ // CoSo of any finaqu, Fa ♀ // CoSo, ♀ // CoSo of any|
famaqu, WiBr, Wi ♂ // Co, any wefcaqu of Wi, SiHu, SiHuBr, SiHu ♂ // Co, any taciqu of SiHu
|8 . dauvaqu||female||MoBrDa, FaSiDa, Mo ♂ // CoDa, ♂ // CoDa of any tinaqu, Fa ♀ // CoDa, 9 // CoDa of any|
famaqu, HuSi, Hu ♀ // Co, any wekaqu of Hu, BrWi, BrWiSi, any taciqu of BrWi
|male||MoBrDa, FaSiDa, Wi, Mo ♂ / / CoDa, $ // CoDa of any tinaqu, Fa ♀ // CoDa,|
♀ 9 // CoDa of any famaqu, WiSi, Wi ♀ // Co, any taciqu of Wi, BrWi, BrWiSi,
BrWi ♀ // Co, any taciqu of BrWi wafiqu
|female||MoBrSo, FaSiSo, Hu, Mo ♂ // CoSo, ♂ // CoSo of any tinaqu, Fa ♀ // CoSo,|
♀ // CoSo of any famaqu, SiHu, HuBr, SiHuBr, SiHu ♂ // Co, Hu ♂ // Co,
any taciqu of Hu or SiHu
|10. tukaqu||male or female||FaFa, MoFa, Fa of any famaqu or tinaqu, Fa of any relative of parents’ generation|
|11. tubuqu||male or female||MoMo, FaMo, Mo of any famaqu or tinaqu, Mo of any relative of parents’ generation|
|12. maJcafauqu||male or female||SoCh, DaCh, BrChCh, SiChCh, Ch of any luvequ, Ch of any relative of ego’s child’s generation|
in Table 1 are extensional, listing after each term a number, but not all, of the conceivable kin relationships or kintypes that can be denoted by it. Any kin relationship, no matter how remote, can be denoted by one of the terms given. Because kinship terms are infinitely extendable to the kinsmen of kinsmen, it is not possible to exhaust the universe of kintypes that may be denoted by any one of them. Analysis must work with a sample of kintypes such as that provided in the data.
It seems evident from direct inspection of the data that 6 of the 12 kinship terms are used reciprocally, in the sense that if A is taciqu to B then B is taciqu to A. The remaining 6 terms fall into 2 reciprocating sets of 3 terms each: 1 and 2 reciprocate 3; and 10 and 11 reciprocate 12. This observation enables us to expand the data of analysis; e.g., we can assume that SiDaHu (man speaking) is a denotatum of 3 (luvequ) because the reciprocal WiMoBr is given as a denotatum of 1 (tamaqu).
Direct inspection of the data also allows us to conclude that the sets of denotata of the several terms are differentiated, among other things, according to
4. Seniority of the alter’s generation. Thus:
4.1 alter’s generation senior (1, 2, 10, 11),
4.2 alter’s generation junior (3, 12).
Leaving aside distinctions involving variables 2, 3, and 4 above, we show the distribution of terms according to variable 1 (degree of generation difference) in Table 2. What remains to be accounted for is the distinction between the reciprocal relationships involving terms 1, 2, and 3 and that involving term 6 and also the distinction between the reciprocal relationships involving terms 4 and 5 and those involving terms 7, 8, and 9.
|Table 2 – Moo/a kinship terms by distance in generations|
|GENERATIONS DISTANT||KINSHIP TERMS|
|Two||10, 11, 12|
|One||1, 2, 3||6|
|Zero||4, 5||7, 8, 9|
If Moala society were divided into two intermarrying patrilineal or matrilineal moieties, there would be no problem. All of these relationships would be readily distinguishable according to whether ego and alter were in the same or different moieties. But there are no moieties in Moala, and depending on how ego and alter choose to trace their relationship they may find themselves in a category covered by terms 1–5 or in a category covered by terms 6–9. Casual inspection of the data provided will not give us a clue as to what the discriminating factors are. But if we diagram every one of the denotata given—or that can be inferred by virtue of reciprocation—for terms 1, 2, 3, and 6, as briefly illustrated in tables 3, 4, and 5, patterns inherent in the data are much easier to discern.
In tables 3 and 4, every consanguineally related pair in the same generation in every chain of genealogical connection between ego and alter are (a) of the same sex, (b) of different sex an even num-
ber of times, or (c) of different sex an odd number of times when there is a marital tie in the most junior generation in the relationship. In Table 5 we find the complementary pattern exemplified. Every consanguineally related pair in the same generation in the chain of genealogical connection between ego and alter are (a) of different sex an odd number of times or (b) of different sex an even number of times (or not at all) when there is a marital tie in the most junior generation in the relationship.
This pattern also appears in the denotata of terms 4, 5, 7, 8, and 9, if we diagram them in a similar way. It is even more clearly evident that sex differences of consanguineal pairs in the same generation in the genealogical chain do not affect the most junior generation in the relationship, whereas the number of intervening marital ties count only if they occur in the most junior generation. Therefore, the discriminant variable may be described as
5. Number of sex differences among the consanguineally related pairs in the same generation in the genealogical chain between ego and alter, excluding the most junior generation in the relationship, and/or the number of marital ties in the genealogical chain that are in the most junior generation. This is represented as
5.1, when the number of sex differences and/ or marital ties is an even one (1, 2, 3, 4, 5),
5.2, when the number of sex differences and/ or marital ties is an odd one (6, 7, 8, 9).
The matrix of variables and kinship terms resulting from this analysis is shown in Table 6. The variables by which the sets of denotata were discriminated have a definite structural relationship to one another. Variable 1 partitions the entire universe of kinship, being applicable to all of the terms, and is the only variable that is so applicable. The other variables subdivide the universe further. Thus the variables relate to one another as in a taxonomic hierarchy, but the terminology lacks cover terms for such implicit entities as “kinsman of my own generation.”
Taxonomies and paradigms. The arrangement of variables in Table 6 reflects the hierarchical or implicit taxonomic structure. Since variable 1 (generation distance) ranks highest in the structure, it is in the extreme left-hand column. Variable 5 (odd or even number of sex differences and/or marital ties in the genealogical chain) is the next highest. Variables 2 and 4 are in complementary distribution in the matrix and occupy equivalent positions in the taxonomic structure. Variable 3 ranks lowest of all.
If all the variables applied equally to the discrimination of all terms, the arrangement of variables in columns in a matrix would be entirely optional. The matrix would have the structure of a paradigm instead of a taxonomy.
The kinship systems that have been analyzed so far produce matrices that are partially taxonomic and partially paradigmatic in structure. When several alternative models (involving a choice among several discriminant variables) can be constructed for a terminological system, they are usually alike
in this respect. Most of them differ at the middle and lower levels of the taxonomic structure rather than at the higher levels. This has obvious significance for students of cognition, quite apart from the question of whether any particular model is the one from which the participants derive their “feel” of how things work.
|DISCRIMINANT VARIABLES||KINSHIP TERMS|
|1.1||5.1||2.2||˙||˙||5. wekaqu or ganequ|
|1.1||5.2||2.2||˙||˙||9. davoloqu or watiqu|
Some terminological systems in the domain of kinship contain fifty or more terms and a dozen or more discriminant variables. Such systems produce large, complicated componential matrices. The procedural rules for grouping terms in rows and variables in columns in such matrices are an important feature of the method of analysis.
Componential analysis and anthropological theory. The Moala kinship terminology illustrates how componential analysis clarifies traditional problems in social anthropology. Anthropologists have regularly sought to explain kinship terminologies like the Moalan one—and there are a number of generally similar ones—in terms of a dual division of society into exogamous moieties or in terms of preferred or mandatory cross-cousin marriage. Moieties are entirely absent from Moalan society. Nor can cross-cousin marriage explain the terminology, for although one must marry a davolaqu, it is explicitly reported that by tracing different routes people can convert a wekaqu into a davolaqu. And they manage things so that they have many more davolaqu than wekaqu. Furthermore, marriage with a close davolaqu, a real cross-cousin, is prohibited. Neither moieties nor cross-cousin marriage can be said, therefore, to provide the criteria by which people sort one another into different kinds of kin. What maintains the terminological system conceptually is the simple odd-even pattern in variable 5. For it is not necessary to count the actual number of marriages or sex differences in the genealogical chain in distant relationships. If A and B have a relative in common, they can immediately relate to one another by the rule that an even of my even or an odd of my odd is my even, while an even of my odd or an odd of my even is my odd. Aside from this, ego’s and alter’s sex and generation are the only things to take note of. This is but one example of how componential analysis calls into question some of the explanations of kinship terminology that have had wide currency among anthropologists (for another example, see Lounsbury 1964b).
Of perhaps greater theoretical interest is the attention componential analysis directs to the relationship between the ethnographer and what it is he seeks to describe. As advances are made in the rigor with which ethnographic data are analyzed and the coherency with which they are presented— a goal toward which componential analysis will have been, in retrospect, but one early step—cultural theory will of necessity be considerably transformed.
Ward H. Goodenough
Burling, Robbins 1964 Cognition and Componential Analysis. American Anthropologist New Series 66: 20–28.
Conant, Francis P. 1961 Jarawa Kin Systems of Reference and Address: A Componential Comparison. Anthropological Linguistics 3, no. 2:19–33.
Conklin, Harold C. 1955 Hanunóo Color Categories. Southwestern Journal of Anthropology 11:339–344.
Conklin, Harold C. 1962 Lexicographical Treatment of Folk Taxonomies. Pages 119–141 in Conference on Lexicography, Indiana University, 1960, Problems in Lexicography: Report. Bloomington: Indiana Univ., Research Center in Anthropology, Folklore, and Linguistics.
Frake, Charles O. 1961 The Diagnosis of Disease Among the Subanun of Mindanao. American Anthropologist New Series 63:111–132.
Frake, Charles O. 1962 The Ethnographic Study of Cognitive Systems. Pages 72–85 in Anthropological Society of Washington, Washington, D.C., Anthropology and Human Behavior. Washington: The Society.
Goodenough, Ward H. 1956 Componential Analysis and the Study of Meaning. Language 32:195–216.
Goodenough, Ward H. 1957 Cultural Anthropology and Linguistics. Pages 167–173 in Georgetown University, Washington, D.C., Institute of Languages and Linguistics, Report of the Annual Round Table Meeting on Linguistics and Language Studies, Seventh. Washington: Georgetown Univ. Press.
Goodenough, Ward H. 1964 Componential Analysis of Könkämä Lapp Kinship Terminology. Pages 221–238 in Ward H. Goodenough (editor), Explorations in Cultural Anthropology. New York: McGraw-Hill.
Hammel, Eugene A. 1965 Formal Semantic Analysis. American Anthropologist New Series 67, no. 5, part 2.
Harris, Selig S. 1948 Componential Analysis of a Hebrew Paradigm. Language 24:87–91.
Lounsbury, Floyd G. 1956 A Semantic Analysis of the Pawnee Kinship Usage. Language 32:158–194.
Lounsbury, Floyd G. 1964a A Formal Account of the Crow- and Omaha-type Kinship Terminologies. Pages 351–393 in Ward H. Goodenough (editor), Explorations in Cultural Anthropology. New York: McGraw-Hill.
Lounsbury, Floyd G. 1964b The Structural Analysis of Kinship Semantics. Pages 1073–1093 in International Congress of Linguistics, Ninth, Cambridge, Mass., 1962, Proceedings. The Hague: Mouton.
Malinowski, Bronislaw 1935 Coral Gardens and Their Magic. 2 vols. London: Allen & Unwin.
Morris, Charles W. (1938) 1955 Foundations of the Theory of Signs. Volume 1, pages 77–137 in International Encyclopedia of Unified Science. Univ. of Chicago Press.
Osgood, Charles E.; Suci, G. J.; and Tannenbaum, P. H. 1957 The Measurement of Meaning. Urbana: Univ. of Illinois Press.
Pospisil, Leopold 1964 Law and Societal Structure Among the Nunamiut Eskimo. Pages 395–431 in Ward H. Goodenough (editor), Explorations in Cultural Anthropology. New York: McGraw-Hill.
Sahlins, Marshall D. 1962 Moala: Culture and Nature on a Fijian Island. Ann Arbor: Univ. of Michigan Press.
Wallace, Anthony F. C.; and Atkins, John 1960 The Meaning of Kinship Terms. American Anthropologist New Series 62:58–80.