How Do We Construct Convergent Numerations?

概要

論文の詳細を見る
Current versions of Chomskyan syntax take for granted that syntactic derivations depend on prior specification of complete ``numerations'' of lexical items, which then combine according to the principles of syntax. However, competence models have provided no ways to choose such numerations; they are either chosen randomly or based on the intuition of (ultimately native) speakers. In neither case is there any scientific characterization of these objects, and so syntactic derivations lose their status as a scientific model. This essay claims that numerations in a plausible formal model of language can be conceived as random (or, formally equivalently, pragmatically determined) only if syntactic derivations can supplement them in a highly constrained way: by adding to them items from a special lexical subcomponent of purely grammatical or ``closed class'' items. Items from this subcomponent, dubbed the Syntacticon in some recent work, seem to have precisely the grammatical properties (insertion into already processed structures, possibly null phonology) needed to make the otherwise randomly selected numerations ``converge'' to well formed Logical Forms.
神戸松蔭女子学院大学の論文
2004-03-21