We cannot think about the world unless we assign it to categories. Categories also help us act upon the world, but are probably not essential to all kinds of activity. It is a matter for debate whether categories for thinking and for acting differ. The discussion which follows concerns classifications as objects of intellectual scrutiny (for the most part, folk classifications), rather than the classifications which anthropologists use to order their data, though these latter are ultimately subject to the same generalizations.
‘Classification’ is that activity in which objects, concepts and relations are assigned to categories; ‘classifying’ refers to the cognitive and cultural mechanisms by which this is achieved; and ‘classifications’ are the linguistic, mental, and other cultural representations which result. Problems arise when the adjectival and nominal status of the root ‘class’ are conflated. This reifies schemes as permanent cultural artefacts or mentally stored old knowledge, when they are more properly understood as the spontaneous and often transient end-product of underlying processes in an individual classifying act. We might call such a misinterpretation ‘the classifica-tory fallacy’, and there is every reason to believe that it is potentially evident whenever ethnographers try to make sense of their data, whether these be tables of symbolic oppositions, animal taxonomies or relationship terminologies.
Ways of classifying the world
Classification as an object of anthropological analysis effectively begins with the publication of Primitive Classification by Durkheim and Mauss in 1901—2. Their main argument was that social divisions provide the prior model for primitive classifications of the natural world. Durkheim and Mauss, almost as if in passing, establish a distinction between mundane (technical, descriptive) and symbolic (ritual, explanatory) schemes which well reflects, and partly determined, the subsequent history of classification studies. Before going any further it will be helpful to unpack this distinction.
Humans classify the world about them by matching perceptual images, words and concepts (Ohnuki-Tierney 1981: 453). The operations work equally in terms of unmodified sense data or their cultural representations. The cognitive and cultural tools available to do this do not distinguish between the social world and the non-social world, though in the analysis of classification this has become a conventional distinction. Similarly, classification can treat its subject in a pragmatic and mundane way or by using various symbolic allusions. Since so much of what we sense and experience is mediated by social consciousness, and since the boundary between the mundane and symbolic is often unclear, it has sometimes been difficult, in practice, to know where to divide these two axes. It may help to set out the dimensions as a matrix:
Classifications of type 1 involve the use of symbolic devices to partition and articulate social time and space: the use of material objects to represent wife-givers and receivers, or the punctuation of time through significant ritual events, are good examples of these. Classifications of type 2 use symbols to make sense of non-social space, as in representations of the cardinal directions in Javanese thought by colours. Classifications of type 3 include descriptive ways of partitioning social space, as with pronouns and relationship terminologies. Classifications of type 4 include much of what passes for biological classification, including Linnaean scientific schemes. The distinctions cannot always be neatly drawn: symbolic things are in an important sense practical, and practical classifications of the non-social world often rely on metaphors which are ultimately social, as in the use of the terms ‘genus’ and ‘family’ to organize plants and animals. Attempts to bring these aspects of classifying behaviour together have met with varying degrees of success. Those who espouse extreme formulations of the universalist (practical)— relativist (symbolic) divide sometimes claim that they are engaged in separate kinds of endeavour, and that one body of work should not invalidate the other. On the other hand, some have stressed the empirical connections between the two, and envisage an ultimate convergence of cognitive and symbolic anthropology. The distinctions are entrenched and not wholly avoidable. Here I first discuss some general principles of categorization and then examine mundane and symbolic schemes separately.
Principles of categorization
Categories may be constructed either by reference to their semantic focus, their boundaries or some combination of the two. In the first case, focus is based on some exemplary instance or cognitive prototype: thus a sparrow is a focal member of the category ‘bird’. Other instances may be non-focal (peripheral) members of their category: as with ‘emu’, an egg-laying feathered biped which does not fly, and is described in English as a ‘bird’. But not all content assigned to categories consists of physical things. Content may include attributes (colour, sound, shape, size and so forth) and abstractions such as time.
The visual sense, however, is dominant and there is a tendency to objectivize even the most intangible. The difficulties of assigning things to categories may be made easier by imposing culturally agreed boundaries. Because parts of our experience of the world are complexly continuous, it is occasionally necessary to impose boundaries to produce categories at all.
Many categories are monothetic, meaning that the defining set of features is always unique, either coded in some binary way or in terms of the clustering of criteria. Others are polythetic: single features being neither essential to group membership nor sufficient to allocate an item to a group. It may seem, therefore, that categories vary according to the complexity of their definition, rather than simply the scope of their content. Many social categories are held to be complex in this sense, though the fact that they are often vague and general (as in Polynesian tabu or taboo) might also suggest that they are semantically primitive.
The relationship between categories and words varies, although the overall developmental primacy of categories over labels is now generally accepted. Language exists within culture and society, but determines the outer parameters of neither. This highlights a major methodological impediment in the study of classification: we rely on language data as the main way of discovering classification, yet words do not always provide an accurate guide. One of the most obvious disjunctions between language and classification is represented by the existence of unlabelled, or covert categories. We know that these exist at various degrees of inclusiveness. Thus, many languages have no words for ‘plant’, yet there is plenty of evidence to suggest that the category exists. Similarly, many varieties of rice may be recognized by a people who do not consistently label them.
Any consideration of the internal structuring of categories quickly merges into a consideration of how categories relate to one another. In many cases the definition of any one category must be understood in relation to others (‘Black’, ‘White’), the part must be related to the whole. What this whole is need not always be clearly distinguished, as in the case where polythetic criteria apply. In some situations the whole may simply be two categories which mutually define each other, in others more inclusive and complex classificatory space which we describe as a semantic (or cognitive) domain. Some domains are defined by their physical boundaries, such as a house or the human body. Here the internal classificatory architecture is analytic in the sense that it decomposes a greater physical whole. The terms which label categories in such classifications are called partonyms (‘windows’, ‘doors’; ‘arms’, ‘legs’). Other domains are culturally and cognitively derived from perceived shared resemblances. These we can call synthetic, and they are exemplified by domains such as ‘animals’ and ‘plants’. Some are clearly defined emically, others are vague, and it is here that anthropologists are most tempted to impose etic distinctions, either through ignorance or convenience. Knowing whether or not to treat kinship categories apart from other social deictics represents a problem of this kind.
The organization of categories within a domain may vary. Many can be represented in terms of class inclusion and contrast, that is as taxonomies. In other cases non-hierarchical models are more appropriate. Brent Berlin has consistently argued in favour of the universality of taxonomy for ethnobiological schemes, but this only works if one also asserts the separation of general-purpose from special-purpose schemes, that is those that are logical and ‘natural’, from those that arise to meet the needs of particular cultural requirements. Despite its prevalence, the principle of taxonomy is better represented amongst some populations than others. Its effective demonstration depends on the extent of linkage between categories in (often flexible) ways; ways which undermine implicit taxonomic levels and contrasts and the general-purpose/special-purpose distinction. It also depends upon the ease with which ethnographers can elicit transitivity statements (a is a b and b is a c, therefore a is a c).
The systematic analysis of mundane schemes was first initiated during the 1950s, and in its formative phase is associated with eth-noscience methodologies. It is typified by a rigorous formal analysis of semantic domains. Early work emphasized the role of distinctive features in the allocation of things to categories and class inclusion as the means of ordering segregates. More recent work has shown a preference for core-periphery models and cognitive prototypes.
The kinds of classificatory schemes analysed in this way have been varied (colour, disease types, firewood, soil, and so on), though most work has focused on ethnobiological schemes. The model for work of this kind was established by Harold Conklin and later Brent Berlin. Their pioneering studies have been swiftly followed by much attention to the evolution of such schemes, the examination of their underlying taxonomic character, regularities in the order of appearance of different degrees of inclusiveness (ranks), and in the order of appearance of life-forms. Much of this work has sought to show the extent to which folk and scientific categories correspond and the relationship between features of a classification and kinds of society. It has also addressed the proposition that plant and animal categories at a particular classificatory level are more salient (basic) than others and have a logical primacy, though there is some dispute as to whether this operates at a consistent level between different natural kinds and across cultures.
Symbolic classification occurs when we use some things as a means of saying something about other things: for example, when totemic species stand for different social groups. It serves to express formally metaphysical and cosmologi-cal speculation, and may be translated into technical procedures which permit the efficacious manipulation of the world, as in ritual and divination. Symbols enhance the significance of important categories, such as those involved in social control (‘prohibited, non-prohibited’). In this sense, categories imply rules and rules categories.
The study of symbolic classification embarked upon by Durkheim and Mauss was given new impetus in the 1960s by the work of Levi-Strauss and a group of influential British anthropologists, notably "Edmund Leach, "Mary Douglas and "Rodney Needham. This work emphasized the centrality of "binary opposition (dualism, polarity) as a principle of social thought. With this basic human organizing idea,best exemplified by the oppositions ‘right’ and ‘left’, ‘male’ and ‘female’, the elements are usually complementary, but sometimes ambiguously asymmetric. Binary opposition is reflected in symbolic schemes where other even numbers are used as an organizing principle. Classifications partitioning semantic space by three may also be important, as in the colour triad (red-white—black) or in dual classifications where a third (mediating) element has been added.
Classification based on the principle of five is exemplified by the Javanese division of the cosmos into four cardinal directions plus a mediating centre. Here, each dimension integrates a large number of different levels of experience (time, colour, state of mind, number, and so on). The case illustrates another pervasive feature of symbolic classification, namely analogy, which explicitly occurs in some cultural genres as semantic parallelism, (Fox 1975). Apart from opposition and analogy, symbolic schemes often display principles of transition (as in marking rites of passage), exchange and transformation (such as the inversion of left and right in the interpretive logic of the Javanese shadow-play). Transformation is often a way of signifying the ambiguous, which may also be marked by anomaly. Anomalies are a means by which significant differences can be highlighted (as in the categories underlying Jewish dietary laws). However, boundaries are not always considered dangerous or polluting, and anomalies are not a necessary result of classificatory process.
While some symbolic classifications strongly reflect social groupings (binary opposition and dual organization, quaternary schemes and Kariera marriage section systems), the view that all classification (or even symbolic classification) finds its roots in social institutions is now generally considered untenable. Certain features of symbolic classification may evolve autonomously, reflecting underlying general principles of cognition. What is always striking is the consistent multivocality and economy of symbols, the same oppositions occurring again and again, amongst different peoples.
Classifications of all kinds connect culture, psychology and perceptual discontinuities of the concrete world. Confusion has arisen in the past from failure to distinguish clearly between individual instruments of cognitive process and the collective medium in which these operate, comprising belief, cultural representations and social practice. It is also crucial to distinguish information storage from representation, abstract knowledge of the world and the pragmatic schemas we use to negotiate our way through it. Our propensity to classify in the ways we do results from the possession of certain innate cognitive skills (some of which we share with non-human primates), plus an ability to organize our perceptions through culture (aided by language) based on models drawn from somatic experience (such as right and left and bodily rhythms), and from social and perceptual experience of the material world. The form a particular classification takes will sometimes be a culturally defined whole, but often as not will be the outcome of interaction in particular circumstances; the interplay of past knowledge, material context and social inputs. Classifications as things, therefore, are not the inventions of individuals, but arise through the historically contingent character of cultural transmission, linguistic constraints, metaphorical extensions, and shared social experience in relation to individual cognitive practice.