EXPLORATION OF RELEVANCE EFFECTS IN REASONING

Abstract

The study examines possible underlying mechanisms that may be responsible for generally observed biased response patterns in two conditional reasoning tasks: the Wason selection task and the conditional inference evaluation task. It is proposed that memory processes that may account for priming phenomenon, may also account for the phenomena of matching bias and double-negation effects in reasoning. A new mental activation model is proposed, based on distributed theories of memory, which models relevance effects of the problem materials by way of a simple algorithm. The model is seen to parsimoniously predict previous general response patterns found using the two reasoning tasks and makes unusual predictions concerning the size of the concepts used in the reasoning problems. The findings show that matching bias can occur between materials that do not lexically match but correlate on a semantic basis, which clarifies a previously uncertain area in the literature. It is also shown that previously deemed 'irrelevant' or mismatching cards on the selection task can interfere with the perceived relevance of matching cards if they are semantically related. The findings also show a weak but significant effect of concept size on matching bias in the inference task, supporting the proposed mental activation model. Issues concerning the notion of relevance perceptions being measured by particular response choices are raised with respect to both the selection and inference tasks.