Kinterms in all systems are susceptible to several alternative formal analytic definitional systems. Of the major two, one can be seen as semantic and the other pragmatic.
Thefirst approach, a semantic one, is concerned with the distinctions among terminological categories in the set, such as the difference between a ‘table’ and a ‘chair’. In this approach to kinterm analysis, the contrasts among terms and their referents can be rigorously defined by a combination of paradigmatic contrasts (defined by the intersection of a set of distinctive features) among their focal or prototypic referents and rules for the extension of these terms to ranges of extended referents.
In kinship this distinctive features approach (usually spoken of as ‘componential analysis’) is based on the features (‘components’) such as relative generation, sex of relative, relative sex, lineal vs. collateral, mother’s vs. father’s side, etc. Some of these features such as relative generation are binary (that is, categorize ego relative to alter) while others such as sex of relative are unary (categorizing alter in absolute terms). In this analytic approach kinterm categories are defined by the intersection of distinctive (defining) features.
(1) In one variant the feature definitions are taken as applying to the full range of referents of relevant kinterms; e.g., in English, the ‘cousin’ category includes a wide range of collateral relatives. At the same time, the responses to requests to ‘describe your cousin’ or answers to questions such as‘what is a cousin’ clearly focus on a ‘first cousin’ – that is, a parent’s sibling’s child. One can speak of semantic extension here, but the extension is accomplished directly through the application of the defining features of the category, and distinguishes its prototype from the prototypes of contrasting categories. The classic articles of Goodenough (1956) and Lounsbury (1956, 1964a), as well as the important papers of Wallace and Atkins (1960) and Romney and D’Andrade (1964) employed this form of analysis. The approach was used by Romney and Epling (1958) in an early (and too much ignored) analysis of an Australian system. However, in some systems such as Fanti, the above approach to semantic extension has problems. In Fanti, ego’s father’s brother, father’s mother’s sister’s son, and father’s sister’s son,
among many other kinds of kin, fall in the egya category which includes ego’s actual father. The distinction is emically5and ethnographically clear: one’s actual father is spoken of by Fanti as the ‘real’ egya, as opposed to those other referents who are described as ‘really’ egyas (even though not one’s ‘real’ egya). The problem is that the ‘generational skewing’ represented by the presence in the category of father’s sister’s son (ego’s 0 generation relative in a basically +1 generation term) cannot be reasonably handled by such a direct application of the category’s defining features. And we further note that the preceding genealogically based usage of egya is opposed, in turn, to its use for a respected (in some sense father-like) senior friend who can be addressed, or spoken of, as egya, but who is‘not really’ one – a respect usage not unlike our use in English of ‘uncle’ for relatively senior family friends. This ‘courtesy’ usage, similarly, is not amenable to the category’s distinguishing features.
Within the preceding distinctive features frame more focused hypotheses based on work in psychology and linguistics have sometimes been explored. For instance, Nerlove and Romney’s (1967) sibling typology study was based on Bruner, Goodnow, and Austin’s (1956) work on concept formation and Greenberg’s (1966 and see 1968) work on marked vs. unmarked categories.6 It focused on true full siblings, and was followed up on with improvements by Kronenfeld (1974). The study showed that out of 4,140 logical possibilities the 12 types that occurred empirically in a sample of 245 terminologies were all conjunctively defined and 240 fit additional constraints related to marking and other specific measures of cognitive ease.
Per Hage (2001, 1999, 1998, 1997) used marking relations in the context of comparable terms from a set of genetically related languages to reconstruct the historical development of kinship terminological systems in the language families to which the languages belong.
(2) The skewing problem led Lounsbury (1964b– and for the general case, beyond kinship, see Lounsbury 1969) to the other major variant of the distinctive features approach, in which the feature definitions of kinterm categories are taken only as applying to the prototypic referents (often spoken of in the literature as‘kernel’ or ‘core’ referents); other (more distant) referents are linked to the category by some form of equivalence (or extension) rule (such as ‘a mother’s brother’s son is terminologically equivalent to a mother’s own son’). These equivalences were created by the analyst in the analyst’s analytical language, and were not necessarily directly equivalent to native speaker statements. Their justification was that they worked – and in parti- cular that a very small set of very simple ones worked powerfully across a great many different systems. Lounsbury used three basic rules for consanguines: a‘merging rule’ (making same-sex- siblings terminologically equivalent when appearing in a string as linking relatives) of which the immediately preceding example is a partial version, a ‘half-sibling rule’ (in which half-siblings are made terminologically equivalent to full siblings) and a‘skewing rule’ which moves specific relatives up or down a generation. The half-sibling rule seems general in its application; the merging rule applies to the large set of systems defined as ‘classificatory’; the skewing rule applies to a subset of those in which some central terminological categories include relatives from several generations. This approach was developed by Lounsbury (1964b, 1965), and further elaborated in various ways by others including Scheffler, Kronenfeld, and Trautmann. In this approach the courtesy usage can be included via extension from the prototype based on features of apparent generation, sex, and a kinship-like attitude towards the courtesy alter.
Within the kin domain, in the context of the above approach, I have looked at the functional and communicative bases and uses of the different kinds of extension (Kronenfeld 1996:172–6, and see 2009:137–40 and in press). For kinship terminologies denotative extension is based on formal extension rules, while connotative extension is based directly on the functional relations implied by the term;figurative extension applies the kin contrasts to another domain. ‘Essential properties’ seem to apply necessarily only to prototypic referents.
The second approach, a pragmatic one has to do with how the referents of the terms interrelate or interconnect– for example, how are chairs used with tables in which contexts. In the kinterm domain, one basic system of pragmatic relations among terms (related particularly to how native speakers calculate kinterm assignments of relatives) can be formally defined as an algebraic system based on relative products (e.g., in English, ‘brother’s son’ equals ‘nephew’, but ‘son’s brother’ equals ‘son’). Such systems can be ethnographically defined directly on the basis of lexemes in the language or ethnologically (i.e. comparatively) defined in terms of genealogical abstractions from the lexemes’ focal definitions. The link to genealogical abstraction stems from the dependence of all kinship terminological systems on the core axiomatic categories (such as father, mother, spouse, and their reciprocals) on which the relative product calculus gets based. Terms in the parent–children–spouse set are taken as axiomatic in the sense of being defined by events outside the terminology itself– such as birth and marriage – and relative products of these terms are used to define other terms. Sometimes the set of non-axiomatic terms is split into a set of ones (such as‘brother’ and ‘sister’ in English) which can be used in the definition of other terms (as illustrated above) and a set (such as ‘grandfather’ in English) which are not so used. There exist two major variants of this approach.
(1) One approach, an emic one (for example, see Read 2011, 2001 and Kronenfeld 1980b– reprinted as Kronenfeld 2009:ch. 3), aims at precise ethnographic accuracy. It is focused on the lexical categories of the language system being analysed, and constructs the relative product formalism totally out of those categories– whether literally in the words of that language or in a direct representation of them in some symbolic or graphic meta-language.
(2) The other approach (for example, Gould 2000; Kronenfeld 2001b, 2009:ch. 11, and 2013) is more etic. It aims at precisely representing the structural relations among kinterm categories that define the major ethnological types. This approach represents the categories of the analysed systems (including the distinction between axiomatic and non-axiomatic terminological categories) and tightly models the relations among them with a set of externally defined analytic symbols, but it sometimes leaves out structurally irrelevant labelling anomalies.
In the 1960s there was a big debate about the psychological reality of various componential analyses (modelled on the debate in linguistics concerning the psychological reality of the phoneme). Roger Brown’s (1964) statement to the effect that ‘you keep asking more and more questions of your analysis until only one is left standing’ pointed the way. The additional questions sometimes involved psychological measures of inter-term closeness and the structural models implied by those (Romney and D’Andrade 1964); sometimes they involved native statements about definitions (Keen 1985; Kronenfeld 1980b; Read 2011, 2001, based on earlier work); sometimes they involved psychological/logical measures of simplicity (Wallace and Atkins 1960; Lounsbury 1964a). It eventually became clear, as more analyses and kinds of analysis emerged, that the kinship domain was an overdetermined one – i.e. that the combination of shaping constraints and the regularities that native speakers pulled out of their experience as they learned their terminology allowed of alternative, quite different analytic approaches, as long as each produced more or less the same allocation of referents to terms. And it became clear that sometimes one was asking a comparative analytic question for which any narrow construction of psychological reality was impossible and meaningless (e.g. Gould 2000; Kronenfeld 1992).
The psychological reality debate in kinship studies related closely to another presupposition from linguistics– to the effect that there should exist some one single best analysis that won at the expense of all competitors and that served all relevant analytic purposes. The splintering of the psychological reality argument also ended the kinship version of that presupposition. The game was still– maybe more than ever– an empirical and systematic one, but different kinds of approaches, when suc- cessful, were found to serve different analytic ends, and thus potentially to be incommensurable.
Another, third, kind of approach, offered by Wierzbicka (1992: chs 9, 10), uses culture-specific supposedly folk-based definitions of kinterms constructed out of what she sees as universal semantic primes (i.e. universally basic concepts). Her presentation is too brief and minimal for the reader to see how it might apply to the kinds of terminological problems and issues addressed in this chapter; on the face of her presentation of it, there would seem to exist serious logical problems with any such application. See Kronenfeld 1996: 20 and 2000: 211–214 (notes 3 and 4) for discussions of these problems along with some tentative suggestions about how they might be dealt with.
IV Behavior
Pragmatic relations of terms (i.e. categories) to their cultural and communicative uses (including figurative usage) can also be described and analysed. Systematic analytic attention to behavioural implications in the analysis of kinterm systems was proposed by Romney and Epling (1958).
In one study within the Fanti kinship project (Kronenfeld 2009: ch. 4; Kronenfeld 1975 [Kronenfeld 2009:ch. 8]) a set of behaviours was elicited that informants considered relevant to various kinds of kinfolk, or that the ethnographer observed as being relevant. For a sample of informants, each was then asked for each behavior on the list whether or not that behavior could be directed (1) towards each kintype off of a list of kintypes that were relevant to terminological contrasts and (2) towards each of an individually constructed (for each informant) list of indi- viduals who were known to be in relevant terminological categories. The kintype list included ‘stranger’ and ‘friend’ as a way of assessing any possible kin-related boundary. The ascribed and remembered data sets were found to be similar to each other in terms of incidence of behavior across kintype categories and across kin-relevant attributes. This behavioural data was compared with kinterm categories and with kin group (i.e., for Fanti, matrilineage) membership, and the ascribed kintype data was directly analysed using a variety of scaling and regression techniques to see what accounted for the incidence of ascribed (to the kintypes) or remembered (for actual specific kin) behavior to kin. In general no necessary or tight relationship was found between ranges of referents of a given kinterm and those behaved towards in the manner associated with the kinterm. Similarly, no necessary or tight relationship was found between behavior patterns and lineage membership– with one exception. Fanti kin groups control inheritance (including succession), and so inheritance itself is limited to lineage members; responses to the‘can inherit from’ question did isolate lineage membership. The inheritancefinding showed that the methodology was capable of picking up behavioural isolation where it existed, and thus in turn implied that the general lack offit of behavior with kin categories was not any artefact of that methodology. Instead, the great bulk of behaviours were found to depend directly on a kintype’s position on relevant variables– mostly generalized ‘seniority’ and ‘closeness’, but sometimes more specific aspects of the relationship.
Informant linking of the behaviours to kinterm categories was not random or casual, however. Where the behavior and the terminology came together was for the category’s prototypic (i.e. kernel) referents. In the absence of other, inconsistent, information the term was taken as referring to its prototype and not to its full range.