Efficiency and Complexity in Grammars: Three General Principles

An interesting general correlation appears to be emerging between variation data in performance, in languages that dispose of several structures of a given type, and the fixed conventions of grammars, in languages with fewer structures of the same type. The performance data come from corpus frequencies and processing experiments, the grammatical data from our expanding samples of languages and from the growing number of languages that have now been subjected to in-depth analysis. The Keenan-Comrie Accessibility Hierarchy was an early example of such a correlation, whereby an implicational hierarchy in cross-linguistic relativization possibilities (SU>DO> IO/OBL>GEN) was argued to correlate with ease of relativization and with relativization preferences in a language (English) with many relativizable positions (Keenan and Comrie 1977, Keenan 1975, Keenan and Hawkins 1987). The preferred word orders in languages with choices have also been argued to be those that are productively grammaticalized in languages with fixed orders, and in almost exact proportion to their degree of preference, in Hawkins (1994, 1998a, 2001). The goal of this paper is to define a hypothesis concerning the relationship between performance and grammars, and to propose three general prin-