One theme I have developed in earlier chapters is that apprehending similarities and differences is basic to cognition. All natural languages provide their users with a rich haul of general terms, and the genera and species in question evidently depend on an apprehension of the similarities between their respective members. Given that this is a basic feature of language, we have no need to speculate that it has a specific origin in human social groupings (though that is not to deny the huge importance of social experience in human reasoning in general, as has been argued by Humphrey 1976, 1992, Byrne and Whiten 1988, Dupré 2002 among others). Durkheim and Mauss overdid it when they claimed that classification itself not only reflects social organisation but is indeed essentially social, as when they asserted that ‘the first logical categories were social categories; the first classes of things were classes of men’ (Durkheim and Mauss 1963 [1903]: 82). That was a thesis that was open to both logical and empirical objections as was set out in detail by their translator Rodney Needham (1963: xi–xxxix).

Thus it is somewhat of an exaggeration to suggest that classification is inherently a social act, essentially focusing on social groupings, rather than a basic and entirely general cognitive activity, and the linguistic evidence from terms for ‘genus’, ‘kind’, ‘species’ in different languages is mixed. The anthropologist Scott Atran, whose work on indigenous classifications of animals I shall be considering shortly, only managed to elicit the claimed underlying ‘natural’ classification among the Itza Maya when he bypassed their indigenous groupings and asked which creatures are ‘companions’ (et'ok) to which. In classical Greek the primary connotation of the term genos (from which our own ‘genus’ is ultimately derived) is indeed ‘family’ or ‘tribe’. But the principal ancient Greek word used for sub-groups or species, namely eidos, has ‘form’ or ‘appearance’ as its primary connotation. One of the classical Chinese terms for ‘kind’, namely zhong, refers in the first instance to what is sown, in other words to seeds. However, the basic logical point, that classification cannot but depend on the grasp of similarities, is confirmed by the principal Chinese word for categories, namely lei, which we have met several times already, for its core use is to refer to the groups or groupings of items of different sorts united by associations, resonances or resemblances. It comes in modern times to be used verbally for ‘to resemble’ and adjectivally to express the notion of ‘similar’.

It is still commonly assumed both that there are natural kinds and that they are, on the whole, adequately represented in the languages we use, though I have thrown doubt on both assumptions, and certainly the kinds actually picked out by different natural languages exhibit considerable diversity. But as is also universally acknowledged, similarities may be misleading. Some philosophies, some religions, some areas of science, develop a strong theoretical contrast between appearance and reality, but, even without that, everyone realises that things may not be as they seem. In other words, we may be taken in by a superficial or an imagined resemblance.