To appear in Catching Language : Issues in GrammarWriting

Linguists often distinguish work they characterize as descriptive from work they characterize as theoretical. Similarly, linguists often characterize certain work as atheoretical. This label is sometimes applied, not only to descriptive work on particular languages, but also occasionally to crosslinguistic typological work. I argue in this chapter that this way of talking represents a fundamental confusion about the relationship between theory and description. First, there is no such thing as atheoretical description. Second, although it is confused to talk about theory and description as contrastive notions, it does make sense to talk about a contrast between description and explanation. I further argue that there is a need for both descriptive theories and explanatory theories. Descriptive theories (or theoretical frameworks) are theories about what languages are like. They are theories about what tools we need in order to provide adequate descriptions of individual languages. Explanatory theories (or theoretical frameworks), in contrast, are theories about why languages are the way they are. The distinction between descriptive theories and explanatory theories is not widely recognized in linguistics, although it is not hard to identify the historical explanation for this. First, pregenerative theories, such as American structuralism, explicitly disavowed the goal of constructing an explanatory theory. As such they were examples of descriptive theories, but the underlying assumption was that that was the only type of theory needed. Generative grammar, in contrast, has aimed at being an explanatory theory. Furthermore, a central tenet of generative grammar, especially clear in the work of Chomsky since the mid-1970s (e.g. Chomsky 1973), has been the idea that a single theory can serve simultaneously as a descriptive theory and as an explanatory theory. Such a view follows from Chomsky's ideas about innateness: if one believes that languages are the way they are because of our innate linguistic knowledge, then a theory about that innate linguistic knowledge will simultaneously serve as a theory about what languages are like and as a theory about why they are that way. Curiously, however, many linguists who reject Chomsky's views about innateness seem to implicitly accept the Chomskyan view that a single theory will serve both theoretical goals. Many functionalists, in particular, propose kinds of explanations for why languages are the way they are that are radically different from those of Chomsky, yet they often see questions of how to describe languages as the domain of formal linguists, confusing issues of descriptive theory with issues of explanatory theory. In this chapter, I examine the implications of rejecting the Chomskyan view of a single theory serving both theoretical goals, and examine the question of what sort of theory will serve as an adequate descriptive theory. I argue that what Dixon (1997) calls “basic linguistic theory” will serve as such a descriptive theory. This paper is primarily directed at linguists who can be construed as functionalist, using the term in a broad sense that includes most work in typology and work by descriptive linguists. The central issue discussed in this paper is what sort of theory we need for linguistic description, if one adopts a functionalist view of language, which for the purposes of this paper can be characterized as the view that functional or grammar-external principles play a central role in explaining why languages are the way they are.