Testing in the Minimalist Program

In many respects, hypothesis testing in the minimalist program is not very different from what it is for anything that aspires to be a scientific theory of an observable natural phenomenon. The normal practices of science should inevitably apply. Nevertheless, it is true that different sorts of natural phenomena present somewhat different opportunities and challenges, and this may cause good scientific practice to take on somewhat different flavors and emphases in different disciplines. For example, language behaviors are ubiquitous in the lives of most of us, so plenty of potentially relevant data is available. On the other hand, these behaviors are only produced by human beings which have intrinsic value and rights, so there are certain kinds of controlled experiments that might otherwise be desirable that we dare not do (intrusive surgical procedures, raising children in unusual environments, etc.). These factors give linguistics a particular character. More subtly, how hypotheses are tested in the minimalist program depends to some degree on what the minimalist program takes language to be, and hence what an adequate theory of language will be. And in this there can be some differences between the minimalist program and other approaches to the study of language that are worth discussing. In this article, then, I say a bit about both facets of the question: both what I take to be good scientific practice applied to natural language, and what may be specific to the minimalist program in light of its characteristic take on the subject. At its roots, the Minimalist Program—like earlier versions of generative or “Chomskian” linguistics—takes a language to be a system of knowledge that humans use to construct and interpret sentences (and other units, both smaller and larger). As such, it is inherently about the relationship between the finite and the infinite (Chomsky, 1965:4; 1980:220-224; 1995:3). It is concerned with the infinite, because most people can easily create and interpret an unbounded number of distinct sentences and sentence types. For example, it is likely that most of the sentences contained in this article are new to both its author and to you its reader; perhaps many of them have never been used before in the history of the world. That does not necessarily make the article particularly hard to read. At the same time, the Minimalist Program is concerned with the finite, because every person who knows a language learned it from a finite exposure to that language, and each one has finite mental capacity for storing their knowledge of the language. Therefore, we have the challenge of explaining how a finite amount of experience and knowledge can be used to construct and interpret an infinite (unbounded) range of new sentence types. The answer must be some sort of recursive rule system—a generative grammar in the broadest sense. The quest of finding that system underlies the generative program in all its various incarnations, even as its specific terminologies and technologies have changed and developed over time. Chomsky has often summed up this fundamental point by quoting Wilhelm von Humboldt, that language is “the infinite use of finite means” (Chomsky, 1965:8; 1995:4, 15) Not all approaches to language see it this way, or at least they do not emphasize this part of the project. For example, usage-based approaches to linguistics emphasize that we list much of what we hear, and are rather conservative in how we go beyond the list. Construction grammars emphasize that