Psycholinguistic Studies of Entrenchment CATHERINE L. HARRIS Boston University 1. Introduction Intuitively, an entrenched expression is one familiar to speakers from frequent use. Langacker (1987) and other cognitive linguists have observed that linguistic structures fall along a continuous scale of entrench- ment in cognitive organization. Every use of a structure is thought to increase its degree of entrenchment. An event type has unit status when it is sufficiently well entrenched that it is easily evoked as an inte- grated whole, i.e., when it constitutes an established routine (Langacker 1987, p.100). If language structures are cognitive structures and are stored in memory, then psycholinguistic techniques can be used to investigate them. However, there is no established tradition in psycholinguistics of studying how humans process units larger than single words but smaller than sentences (see review in Tanenhaus 1988). (An exception is the recently burgeoning work on idioms, as in Cacciari & Tabossi 1993, Everaert et al. 1995.) My own view is that there is considerable storage of common word combinations and other entrenched items (Harris 1994b). An initial step towards validating this view is to show that mental struc- tures representing units larger than words are activated when reading entrenched expressions. Two types of entrenched constructions are investigated in this paper: multi-word idioms (Great minds think alike, Til the cows come home) and common word combinations (last chance, faced with, good job). My question about idioms is whether an “idiom-level” of representation exists. For common word com- binations (which I will call collocations) my question is whether collocation-level representations exist. An intuitive explanation of “idiom-level representation” is a cognitive structure which bundles together the information activated when an idiom is processed. The idea of a unified structure which bundles together relevant information comes from laboratory experiments on processing of letters, words and non-word letter strings. Under conditions of brief exposure duration and masking, subjects are faster and more accurate when processing words than when processing non-word letter strings. This was called the word superiority effect. Carr (1986) has summarized psycholinguistic explanations of the WSE as follows: “Words benefit from higher-order, unitized codes that bundle all the available stimulus information together in a form that is safe from visual masking and memorable for long enough to support all of the decision and response selection processes required by the task (Carr 1986). The laboratory findings thus support the introspection that words (compared to word fragments and random letter strings) have a cohesive, unitized feel. Given that speakers recognize the cohesiveness of idioms and common word combinations, psycholinguistic tasks may reveal evidence for the existence of unitized struc- tures supporting these constructions. 2. Does Exposure to One Part of an Idiom Activate the Whole Idiom? A common experience is anticipating the completion of an idiom after hearing an initial word or two. This suggests that the first part of an idiom may be able to activate a representation of the whole idiom. A helpful