The effect of usage on degrees of constituency: the reduction of don't in English

In this paper we take the position that there are many degrees of constituency and that these derive in a direct manner from the frequency with which elements are used together: elements that are frequently found next to each other show a tighter constituent structure than those that collocate less frequently. We use both phonological and functional evidence from conversation to argue that repetition conditions chunking (Haiman 1994), sometimes overriding the syntactic and semantic logic of the organization of utterances. Our study examines the reduction of don't in American English conversation. We find that don't is reduced the most in the contexts in which it occurs the most, that is, after I and before certain verbs, such as know . While a generalized constituent structure may be an emergent property arising from many analogous utterances, specific combinations that are frequently used may diverge from the general pattern because frequency conditions autonomy in storage and renders internal analysis unnecessary. This phenomenon reveals the essential role of repetition in the creation of constituent structure: while semantic and pragmatic factors determine what occurs together in discourse, the actual repetition of stretches of talk triggers the chunking mechanism that binds them into constituents.