Philosophers have always thought of nominalism as a doctrine, not as a practice. They may therefore be excused for having trouble seeing the relation of nominalistic linguistics to the doctrine of nominalism, which is that the former is a way of doing linguistics to which doctrinal nominalists could not object, but that would seem deficient to those who are doctrinal realists. For if there are no classes in reality, but they exist in name only, as doctrinal nominalists claim, then any way of dividing up phenomena, including linguistic phenomena, is as good––or at least as true––as any other. And by ‘nominalistic linguistics’ I mean the practice of imposing an arbitrary taxonomy on linguistic phenomena.

This use of terms and concepts from the history of philosophy to make headway in linguistic theorizing may be interesting but also possibly confusing, the latter for the following reason. The linguistic phenomena classified might include linguistic universals (the Peircean ‘types’) as well as linguistic individual events (the Peircean ‘tokens’). And one who is familiar with the nominalist/realist distinction as a matter of  doctrine only might naturally suppose that by ‘nominalist linguist’ is meant one who denies the reality of linguistic universals. That, of course, would be an application of the nominalist doctrine to linguistic phenomena; but that, one can see now, is distinct from nominalist linguistics as a practice or method. Nominalism as a practice would not necessarily deny that universals are real; rather, it consists in deciding their classification arbitrarily––both their classification into subtypes, if they are segregated from individuals, and whether to so segregate them. Even their classification as real or unreal would be quite arbitrary.

The Chomskyan (= mainstream linguistics) search for deep structure and generative principles looks relatively realist from a doctrinal point of view. For whether or not surface phenomena are conceptualized in terms of types as well as tokens, the deep structure and principles look like universals, and especially so the way Chomsky and his followers speak of them. Chomsky and his school are nominalist linguists, not realist linguists, because their taxonomy of surface phenomena––the phenomena they wish to explain as following from deeper principles––is arbitrary. (It would follow that the hypothetical structure must be arbitrary too, for it is justified only by its capacity to explain those phenomena.)

‘Realism’, of course, is used to designate the opposite of phenomenalism as well as the opposite of nominalism. With respect to doctrine exclusively, not method, Jakobson and  his structuralist continuators (like me) look like phenomenalists in contrast to Chomsky and his followers, since the former seem much more concerned with the description of what is here being called surface phenomena, whereas the latter plunge quickly to the (putative) underlying realities that explain them. One could say that Chomsky et al. are in error for proceeding too quickly: after all, how can they abduce explanatory realities when they are wrong about the explanandum? But this is not so simple an issue as that. For if the classification of phenomena is to be real, not nominal, then it is often impossible to know what that classification is until the underlying realities have been identified.

As an example from a domain other than language, consider whether it was possible to know that rusting, fire, and metabolism should be classed together as members of the same natural kind before they were all explained as different forms of oxidation. The circle here is like the hermeneutic circle: the explanans and the explanandum are found together, not first one and then the other.

But there is another way of looking at this which can be identified, mutatis mutandis, with that of semeiotic neostructuralism in linguistics. Realism in contradistinction to nominalism (doctrinally) is connected with teleology—or so, at least, Peirce appears to have thought. A natural class is one the members of which exist because each satisfies the same idea. That idea has a certain potency, and hence the class exists independently of anyone’s having named it. This idea is consistent with the argument of the preceding  paragraph, according  to which some natural classes may be those classes entailed by a true explanatory theory. But it is not limited to cases where the explanatory structures lie beneath the surface phenomena.

Suppose language qua phenomenon has a history, and suppose that history can be understood by postulating goals not involving any underlying mechanisms. For example, linguistic change might be seen as tending toward a more adequate diagrammatization (as it is, in fact, by semeiotic neostructuralists). Then we have a teleological basis for identifying natural linguistic classes, namely those that we have to attend to in order to understand language as diagrammatization. (This too involves a hermeneutic circle: neither the right description of the process nor the goal that explains it can be discovered without also discovering the other.)

If the preceding is a roughly correct account of the linguistic practice of semeiotic neostructuralism, then it would seem that one who espouses the latter is in method, if not in doctrine, a realist as opposed to a nominalist, but a phenomenalist as opposed to a realist, and a teleologist to boot. One may doubt whether a semeiotic neostructuralist is a phenomenalist in doctrine. For such a linguist does not deny, in fact, he presupposes that there are realities beyond or beneath language but for which his teleological  account of linguistic change would make no sense. That is, there must be flesh-and-blood bodies that speak and listen, and it is their  desires and needs that explain why ever more adequate diagrammatization is an inevitable if unintended goal. If the research program subtended  by semeiotic neostructuralism can be made to work, then it will indeed conflict with Chomskyan (= mainstream) linguistics––and prove superior to it.