Abstract

This study combines three hypotheseswhich all go against commonly held views. First, Stowell's head-initial/ head-final parameter should be replaced, but not by underlying universal head-initial order. Rather, English, French and Chinese head orderings support three conclusions: 1) Descriptive generalizations should be formulated in terms of departures from a universal default head-final order. This head-final default is equally valid in word (X-zero) and phrasal (XP) domains. 2) The language-particular head-initial domains found in English and French are derived from (somewhat different) generalizations that determine their positions of main stress. 3) Any relevant generalizations in these areas make no use of any putative boundary between morphology and syntax, and provide evidence that identical principles govern all sizes of grammatical domains. The first conclusion rests crucially on a reverse correlation between headedness and stress. French head-final stress is pervasive, leading to a wider range of head-initial structures. There are in addition 2 overrides for principles that specify head ordering: (i) Lexical specifications of bound morphemes as prefixes or suffixesmust be respected, and (ii) Specifiers (when properly defined) must precede even otherwise initial heads. The second conclusion centrally implies that independently defined phonological patterns determine syntactic word order, counter to the essentially universal tenet of generative grammar that syntactic constructs can determine phonological patterns, but not viceversa. The third conclusion is based on the superior descriptive results obtained even in domains considered "morphology proper" (some of the cited results are beyond the scope of a single article) by really fully distributing morphology, with no backtracking refuge in a special level of Morphological Structure