In this paper I will be reviewing literature, mostly from cognitive science, related to the human brain’s processes of categorization

B.Natural kinds are classical categories

The argument that there is a distinct set of categories, natural kinds, with a very peculiar structure, can be traced back to Locke (1964[1690]), whose distinction between nominal and real essences has led to the one between nominal and natural kinds (Keil 1989:36-37). Briefly summarized, nominal kinds (e.g. pencil) reflect more or less arbitrary conventions of usage, whereas natural kinds (e.g. fish) at least carry the intuition that they carve out causally relevant domains of the natural world, and when they are revised––or rather, when we revise our ideas about which objects out there in the world should be excluded from, or included in, a natural kind––we do not feel this as a definitional shift so much as a gain in understanding of the category itself, whose definition is something we discover, not arbitrate.³

Such intuitions require that there be a principled distinction between criteria of inclusion in the category, and criteria of identification (or typical appearance) of its members. Consider that humans seem readily to accept, when the information is conveyed by an expert source, that a whale and a dolphin are not fish even though a cursory inspection suggests that they are. Similarly, we gleefully leave ‘fool’s gold’ to the fools who have not figured out it is not the real thingeven though we ourselves need an expert to decide which is which. Now try imagining someone easily accepting the statement by a librarian that what looks like a book is really a pencil. Malt (1989) has systematically investigated these intuitions and found that for borderline cases on the basis of appearance, people feel that they need experts to sort natural kinds, but with artifacts deciding where such cases belong is felt to be a matter of opinion (e.g. the statement “According to experts, this is a shirt” is silly, but “According to experts, this is a fish” is not).

The criteria of identification—i.e. appearances––supply, for most instances (and in particular those typical of the category) a rough-and-ready way of guessing that an object is ‘inside’ the category. The guesses are probabilistic, graded by the number of category-relevant characteristics that are readily apparent in any particular object. However, these characteristics are not determining of inclusion, which is not probabilistic––natural kinds do not have graded membership. Whales and dolphins, for example, are not ‘less fishy’ than fish, they are not fish at all; and fool’s gold is not ‘partly’ gold, it is not gold at allthis, despite appearances, and despite the probabilistic results of naïve, perceptually-based guesses concerning inclusion in the categories. Given some relevant information, then, one could include in a given category a member that was initially identified as lying outside, and the probabilistic identifying guess would then become a judgment of typicality, not of graded inclusion. This disjunction between appearance and determining criteria of inclusion in natural kinds has led to the intuition of ‘essences’.

Almost everyone has had the intuition that things are not always what they seem and that there is something deeper and more basic to a kind than what is immediately apparent. One way to capture this intuition is to argue that things have essences that are often difficult to discern immediately.—Keil (1989:36)

Thus, a necessary and sufficient condition governs inclusion in a natural kindpossession of the ‘essence’. No stipulation of surface characteristics will be a statement of the ‘essence’, since these are just the typical manifestations or consequences of the essence, and not the essence itself. It is these manifestations, and not the inclusion criteria, that are responsible for the typicality effects. Thus, natural kinds are classical categories in one sense, although they certainly have a richer structure than earlier philosophers may have recognized.

This brings us full circle to a reinstatement of classical categories, at least for some domains. But this insight has not percolated completely, and what may be termed the ‘Wittgensteinian bias’ that sees categorization always as a consequence of degree of similarity, where similarity is on a fuzzy continuum (exemplar, prototype, and dual models among others; cf. Rips 1989) is still dominant.

What is common to them all?—Don’t say: “There must be something common, or they would not be called ‘games’”—but look and see whether there is anything common to all.—For if you look at them you will not see something that is common to all, but similarities, relationships, and a whole series of them at that. To repeat: don’t think, but look!—L. Wittgenstein, Philosophical investigations, quoted in Medin & Ortony (1989).

Medin & Ortony (1989) argue that, pace Wittgenstein, there may after all be necessary and sufficient conditions for some categoriesin fact, for many, and in the case of natural kinds we cannot see this analytically unless we do some thinking. ‘Looking’ will not suffice because the appearances may not be shared by all members of a category; the essence, which we cannot see, on the other hand, may be taken by individuals as necessary and sufficient for inclusion in the category.

III.Identification vs. categorization in Mongolia

My study population consists of Torguud semi-nomadic pastoralists, in the main (Torguuds are a small Mongol ethnic group in Western Mongolia). They move around in the district of Bulgan Sum, in Hovd province, Republic of Mongolia. In the winter they are not for away from the district ‘center’, which is a town lying by the Bulgan river. The town itself has about 500 families and around 2500 inhabitants. The land is quite fertile on the river’s banks and the sedentary residents of the ‘city’ grow all manner of fruits and vegetables in their small horticultural gardens. Another 2500 people make a living as farmers beyond the town, and are considered as part of the district ‘center’. Beyond these is the steppe, where a total of about 5000 nomads eke out a living. The nomads winter in the valley, which is also the large floodplain of the Bulgan river, not far away from the district center. The male herders make regular two-week trips out into the Gobi where the snow is less thick and the sparse grass more accessible. They also assist their livestock with fodder collected in late August/September. In the summer months they move to high ground, in the Altai Mountains, changing their location constantly as pastures get depleted (they may make as many as 10 migrations in a four-month period). The high ground to which the pastoralists move in the summer is very green high-altitude forest-steppe, crisscrossed by innumerable glacial rivers and streams. Apart from Torguuds, there are other Mongol ethnic groups in the area, as well as a large population of Kazakhs (perhaps 30% of the local population), with the biggest local ethnic contrast that between Mongols and Kazakhs

I constructed a short questionnaire and administered it verbally to 59 Torguud subjects.

Question (1) If the father is Kazakh and the mother Mongol, what is the ethnicity of the child?

Question (2) The father is Kazakh, the mother Mongol, but everybody around the family is Mongol and the child has never even seen a Kazakh, outside of the father. The child will learn Mongol customs and language. What is the ethnicity of this child?;

Question (3) A Kazakh couple has a child that they do not want. They give it in adoption to a Mongol couple when the child is only a year old. Around the Mongol family there only are Mongols and the child grows up never meeting a single Kazakh. Since he was a baby when adopted, he knows nothing and thinks that his biological father and mother are the Mongol adopters. He grows up learning Mongol customs and language. What is the ethnicity of this child?

The results of this study have been reported in full elsewhere (Gil-White 1999), so I limit myself here to a summary of the relevant points.

Question 1 tested whether the patrilineal rule apparent in exploratory pretests was a strong one. Questions 2 and 3 pitted this rule against enculturation in an ethnic outgroup, presenting a conflict of classification between a descent criterion and an environmental/cultural one. Given that the environment of development/enculturation was quite extreme, in that the child grew up not even seeing any other Kazakhs (the ethnicity of the biological father), the results (see Table 1 below) indicate a very strong preference for classification according to descent.

Table 1

Answer	Question 1	Question 2	Question 3
Child is Kazakh	56 94%	49 83%	35 59%
Child is Mongol	0	10 17%	24 41%
Child is erlets “half-breed”	3 6%	0	0
Total	59 100%	59 100%	59 100%

N=59
The column for question 1 shows that Mongols (like other pastoralists) are patrilineal (Khazanov 1994:143). However, this usually refers to clan and sub-clan ascription, and material inheritance; here we see that fathers also transmit ethnic ascription. The question was ‘open’ in that they were not forced to choose among predetermined options (“half-breed” was their idea), but in another sense it was forced by presuming that children are born with an ethnic status. However, if actors believe membership in these categories is a matter of one’s absorbed culture perhaps they should object that “it depends”, and explain what on. No such answer was ever given.

The wording in question 2 presumes the opposite: ethnic ascription will depend on circumstances of enculturation—which are quite extreme here. The “but” in this question was highly emphasized sociolinguistically by raising my voice along with my index finger, while making big eyes that looked straight into the interviewee’s in what I hope was an ominous expression. I hoped thereby to draw close attention to a set of circumstances absent in the first question that could result in a different answer from question 1, and to imply that this was the expected answer (I was trying to bias the experiment against my favored hypothesis). However, the great majority of respondents were unfazed by this implication and insisted that the child in the second question was Kazakh.

Question 3 is perhaps the most extreme rearing-bias scenario possible. If respondents insist this child is also Kazakh, they will imply that one can be Kazakh and not know it (à la ugly duckling). More than half responded in this way. To them, the child takes the biological father’s ethnicity no matter what. Notice that here the child has two fathers, one of whom he never knew. If the latter is the one that matters for ethnic ascription, the underlying model of recruitment is extremely descent-based.

Of particular importance here is that biological descent is for most a necessary and sufficient criterion of inclusion in a category (Kazakh) even though members are typically identified by recourse to cultural traits that they display. In other words, despite the fact that my informants probably will agree that this Kazakh child will not look or behave anything like a Kazakh⁴and thus will violate the phenomenological expectations associated with a Kazakhthey nevertheless include him in the Kazakh category. This suggests that ethnies may be ‘natural kinds’ to human cognition.

Natural kinds are categories of objects and substances that are found in nature (e.g., tiger, water, cactus). Not all categories are natural kinds, for example, human artifacts (e.g. chairs, mittens) or categories that are defined by arbitrary properties (.e.g. the set of green things) (. . .) natural kind terms capture regularities in nature that go beyond intuitive similarity (. . .) Natural kinds have a deep, nonobvious basis; perceptual features, though useful for identifying members of a category, do not always serve to define the category. For example, ‘fool’s gold’ looks just like gold to most people, yet we accept the statement of an expert that it is not gold (. . .) Because natural kinds capture theory-based properties rather than superficial features, some of the properties that were originally used to pick out category members can be violated, but we will still agree the object is a member of the kind if there is reason to believe that ‘deeper’, more explanatory properties still hold.—Gelman & Markman (1987:1532)

The child in my questions is human ‘fool’s gold’. If he existed, and was encountered by a naïve local traveler, the traveler would no doubt classify the child—on perceptual input alone—as a Mongol. But told that the father was Kazakh, he would probably change this classification, realizing that surface features had misled him as to the “‘deeper’, more explanatory properties” (much like the baby swan’s foster siblings at first mistook him for a duck). I do not believe these results from Mongolia are idiosyncratic; the ethnographic literature suggests that all over the world, however culturally marked ethnic actors may be, the ‘rule’ for making ethnic ascriptions is based on blood much more than enculturation (Gil-White 1999).