By Roger Martin, Juan Uriagereka
In this paper we treat the phenomenon of syntactic preference as a grammaticality judgment that emerges in situations of structural ambiguity. There seem to be instances where, despite the fact that several relevant structures for some terminal string are grammatical, speakers somehow feel that they do not all have the same grammatical status: one or more are preferred over the others. To account for this phenomenon, we presumably need some sort of comparison set as well as a procedure to establish the appropriate choice (s). In this paper we will concentrate only on one well-studied example of syntactic preference, which emerges in the interpretation of scopally ambiguous sentences.
Before presenting our account of preferences, we need to consider the way a chain is constructed. For Chomsky (1995), a chain is an n-tuple of phrase-markers linked up by way of a moved item. If L, a sister to K, moves to merge to K’, the object thus formed is the chain {{L, K},{L, K’}}. If something happens to L in one position (link) in the chain, it happens to L in all positions. For example, when the ‘head’of a chain moves, the entire chain expands. Implicit in this is the idea that chains are objects encoding several derivational stages, all of which co-exist in an array of phrasemarkers, and that the computational system somehow has access to all of the stages simultaneously, at least within given domains.