Familiarity, context, and the distinction between literal and nonliteral language

Six experiments investigating the relationship of familiarity and context to the processing of metaphor are presented. Experiment 1 compares time to understand sets containing idiom or metaphor targets following 1-2 sentence contexts in four conditions: NONLITERAL, where the context was completed by an idiom or metaphor, PARAPHRASE, where the same context was completed by a literal target with the same meaning, SURPRISE, where the context was completed by a less-anticipated literal target, and LITERAL, where the target from the NONLITERAL condition was used in its literal sense in a different context. In Experiment 2, sets of dead metaphors replaced the idiom sets, and the metaphor sets had novel metaphors. Experiment 3 matched the targets in each condition for overall printed word frequency, to investigate whether word familiarity was interacting with type of usage. It also included an UNFAMILIAR condition, where the same context was completed by a much less familiar word used literally. Experiment 4 took 20 contexts from Experiment 3 and asked subjects to generate their own endings. Experiment 5 replicated Experiment 3 but with a two-target semantic choice instead of a single response. Experiment 6 shortened contexts and reduced their information content. Its purpose was to see how much context was contributing to understanding, and whether some conditions would be more affected than others. The results may be interpreted as indicating that familiarity with the use of a word is important in determining speed of understanding. Dead metaphors could be understood just as quickly as words used literally, but novel metaphors took longer. Contextual expectations are also a powerful adjunct to the understanding process. When expectations are thwarted, errors and understanding time increases. Metaphor understanding is interpreted as a class-inclusion process in the manner described by Glucksberg and Keysar (1990), where a word used metaphorically is viewed as a prototypical exemplar of a hierarchically superordinate class that becomes extended to incorporate the context topic. This process takes time, but metaphors have a response latency advantage over surprising or unfamiliar literal words encountered in context. When context is reduced, metaphors are still advantageous in terms of time, but are less useful to depth of understanding.Arts, Faculty ofPsychology, Department ofGraduat