ABSTRACT
Two models of context effects on lexical processing are described; (a) the Construction-integration (CI) model (Kintsch, 1988) and (b) the Lexical Distance (LD) model (Sharkey, N.E. in press; 1989). Both can account for some well known effects in the priming and ambiguity literature. However the mechanisms by which they operate differ. The CI model presumes connections between related items in the lexicon. In this model, it is assumed that, during the initial stages of processing, the associates of a word in the lexicon are always activated in a context-independent manner. It is postulated that textual priming effects can only occur after this phase of sense activation. Lexical priming effects and textual priming effects are the result of the operation of different processes. In the LD model, on the other hand, there are no associations between items in the lexicon. Words are represented in the lexicon as vectors of microfeatures. Context effects are conceptualised as a measure of network distance from an initial state to a target state. Both lexical priming effects and textual effects exert their influence in the same way; reducing the time taken to move to the target state. An experiment is reported in which textual priming effects are examined in an attempt to test the predictions made by the two models. In this experiment, a facilitatory influence of knowledge-based texts on a lexical decision task was demonstrated, despite short SOAs and the absence of an immediately preceding associatively related word. According to the CI model the SOAs involved were too short to allow anything but associative priming; therefore the results of this experiment favour the LD model. This research clears up a conflict in the literature between two sets of experimental findings (Kintsch & Mross, 1985; Sharkey & Mitchell, 1985). It also illustrates the use of psychological experimentation in allowing a principled choice to be made between two models.
