The Need for Explicit Inferential Methods in Linguistics

Like every discipline, theorizing in linguistics crucially involves drawing (often tentative) conclusions about the relation between the available evidence and the theories under consideration. At many points in their research, linguists must decide whether the available evidence (theoretical, empirical, etc.) supports a given theory well enough to accept it, possibly over various competing theories. Surprisingly, however, virtually no attention has been paid to this aspect of linguistic theorizing, which contrasts with the standard practices of the social, behavioral and psychological sciences. In the latter fields, whole subareas – including dedicated faculty positions, journals, societies, etc. – are devoted to the quantitative, frequently statistical, details of drawing inferences from data. But in linguistics, theoretical inferences typically have the form of informal, verbal, holistic judgments made by professional linguists. I begin by characterizing the issue of linguistic inferences, showing why the current methods are cause for concern. The psychological literature on professional judgment strongly suggests that professional judgments aren’t as accurate as the experts believe, and that they are inferior to more explicit methods. In fact, linguistic inferences are particularly (although not uniquely) suspect in this respect. I then consider several responses linguists have made to the criticism of their reliance on expert judgment. These replies, I argue, do little to mitigate the concerns about how evidence is compiled and assessed in linguistics.