Читать книгу A Companion to Chomsky - Группа авторов - Страница 16

There are, of course, many ways in which grammar can be studied. Chomsky's methodology has been remarkably influential and successful, but it is important to compare it with other perspectives. For reasons of space, we obviously could not do justice to the many perspectives in the larger field of linguistics. We have chosen comparisons with three prominent approaches to syntax: “nonderivational,” but broadly generative approaches, general statistical learning, and usage‐based approaches.

A hallmark of Chomskyan approaches has always been that they are what is called “derivational,” that is, that they assume that non‐local dependencies are created by what is typically called movement (e.g., in What did you eat for breakfast?, what has moved from its original position as the object of the verb eat). However, in the late 1970s, nonderivational approaches started to emerge, where rules of grammar are stated as constraints that apply to representations, i.e. ruling that they are allowed (well‐formed) or barred (ill‐formed). Such approaches are often called “declarative” (sometimes “representational”) as they do not have a procedural component – the grammar is not characterized in terms of derivational steps.⁹ However, they share with Chomsky the goal of providing explicit and structured accounts of a well‐formed expression in a given language and for any language. Chapter 16 by Peter Sells discusses three different declarative frameworks and shows in detail how they are compatible with Chomsky's early work, and then outlines their design features, and provides examples of grammatical analyses within such frameworks.

As we noted at the beginning of this introduction,another approach that is often assumed to be incompatible with Chomsky's approach is statistical learning in its many different guises. In Chapter 17, Lisa Pearl argues that this is too simplistic, and that statistical learning can both complement UG and help refine our models of its contents. Her focus is in particular on how statistical modeling provides a better understanding of how learners navigate the hypothesis space they are faced with, and she argues that within an approach that remains generative, statistical learning might even replace UG as an explanation for at least certain properties of the adult grammar and how they are acquired.

The last chapter in this section, Chapter 18, written by Frederick Newmeyer, scrutinizes differences between Chomsky's linguistics and what can broadly be labeled “usage‐based linguistics.” The latter is an umbrella term for a variety of approaches that share the commitment to study the use of language as opposed to focusing on the underlying computational systems of syntax that enable such use. This includes “constructionist” work that sees the basic units of language as grammatical constructions (e.g. active and passive) and the “functionalist” perspective that regards the use of language (primarily in communication) as a major influence on grammar. Comparing Chomsky's work to the shared commitments of usage‐based linguistics is complicated by the fact that Chomsky has rarely engaged with published work in this tradition. However, Newmeyer solves this problem by selecting what he takes to be the most important issues separating the two traditions and carefully outlining the positions that they have taken on those issues.