Explaining the Time-Course of Literal and Nonliteral Comprehension

Over the last few decades, a considerable number of experimental studies have been made on the time-course of literal and nonliteral comprehension. These studies have been devoted to test two incompatible models of nonliteral processing: serial model and parallel (direct) model. The serial model states that literal interpretation is an obligatory process and nonliteral interpretation is triggered only when the literal interpretation is rejected as incongruous. Thus, it predicts longer processing times for nonliteral sentences as compared to literal sentences. On the other hand, the parallel/direct model claims that nonliteral interpretation is computed directly, without the rejection of the literal interpretation. As a result, it predicts no processing time difference between literal and nonliteral sentences. However, both models explain only a limited number of the time-course findings: some studies have shown equal (or shorter) processing times, while other studies have shown longer processing times for nonliteral interpretation. What these results indicate is that literal interpretation does not always have priority over nonliteral one and nonliteral interpretation is not always derived directly. Recently, Giora (1997) proposes that the serial/parallel debate can be reconciled by the graded salience hypothesis: nonliteral and literal comprehension can be viewed as governed by a simple principle of salience that more salient meaning is processed first. The view underlying the model, which we agree with, is that no priority is assumed with respect to literality. However, Giora’s (1997) explanation of the time-course findings by the graded salience hypothesis is inconsistent mainly because she takes a static view of salience which decides priorities for multiple interpretations. Her analysis of the findings is based on the salience of lexical meanings of a sentence, and assumes that the priorities decided by the lexical salience remain static during language comprehension. The purpose of this paper is to consistently explain a variety of apparently incompatible psycholinguistic findings on the time-course of nonliteral comprehension. For this purpose, we argue that (a) three levels of language comprehension — i.e., lexical access, sentence meaning construction, discourse interpretation — should be separately considered, (b) a priority of one interpretation to an alternative interpretation at one level does not entail the same priority at higher levels because of a shift in salience caused by contextual support, and (c) this dynamic view of salience based on the multilevel analysis best accounts for the time-course findings of nonliteral versus literal comprehension.