| Conditions under which costs and benefits of the prior processing of an ambiguous word on subsequent occurrences of that word or semantically related words are explored. Ambiguity resolution can be influenced by immediate context in which the ambiguity appears, and context no longer in conscious awareness. Several theories have been proposed to account for difficulties in processing when meaning is changed from the first occurrence of an ambiguous word to a second occurrence. An inhibition model (Simpson & Kang, 1994) and a suppression model (Gemsbacher, 1990) suggest selection of a homograph meaning inhibits an alternative meaning causing a decrement in performance. The activation-selection model (Gorfein, 1987, 2001) unlike these views does not require inhibition, and allows for processing effects in situations when the homograph is not present. Three experiments using the relatedness decision task demonstrate that deciding seal is related to dolphin facilitates the decision that walrus is related to otter. A homograph paired on different occurrences with associates related to one of two meanings (seal-DOLPHIN and the contrasting meaning seal-ENVELOPE), shows a performance decrement on the latter pair. However, associate pairs related to either meaning of the homograph presented after this observed decrement show facilitation. This implies that neither meaning is less available in the absence of the homograph. Two additional experiments explore cross-task transfer effects using the levels of processing framework. Homograph-associate word pairs are initially processed under two different instructions, lexical decision and judging sentence sensibility. Transfer effects are assessed using a second task (relatedness decision). Homograph-associate words were presented on two occurrences, once with each meaning. Homograph-associate word pairs identical to either occurrence were facilitated across tasks, regardless of the initial processing task. The same results were not observed across tasks when the homograph-associate word pair was altered by presenting a new associated word. Across tasks a third occurrence resulted in facilitated responses when identical pairs were presented. This was inconsistent with performance for three occurrences of the homograph within the relatedness decision task. The results are interpreted as requiring a two-process model, a slower process as described by the activation-selection model, and a faster process based on retrieval. |
