On Psycholinguistic Grammars

Abstract

It has long been known that language acquisition is only possible if information is available above and beyond the mere presence of a set of strings in the language. One commonly postulated source of such information is a (possibly innate) constraint on the syntactic forms that a grammar can take. This paper develops and presents a set of formalisms based on the Marker Hypothesis that natural languages are “marked” for complex syntactic structure at surface form. It further compares the expressivity and restrictedness of these formalisms and shows that, first, not all constraints are actually restrictive, and second, that the Marker Hypothesis, and its implicit function/content word distinction, provide strong restrictions on the form of allowable grammars. These restrictions may in turn provide evidence about its actual psychological reality and salience. In particular, the class of strongly marked languages can be demonstrated not to admit all finite languages and thus not be subject to the hangman's noose of Gold's learnability proofs, and it is conjectured that these languages may provide a computable method of inferring human-like languages.