Linguistic Determinants of Bias in Conditional Reasoning

Abstract

An experiment is reported to test the hypothesis that a non-logical tendency known as “matching bias”, which occurs in conditional reasoning tasks, is linguistically determined. It is argued that the bias occurs because items which are not mentioned in the conditional rules appear to be “irrelevant” since they do not conform to the linguistic “topic” of the sentence. Logical cases can, however, be produced which always “match” the items named in the rules by introducing explicit negatives into the descriptions of the instances. As predicted, when this was done, the usual “matching bias” effect was significantly reduced as compared with the normal use of affirmative instances which implicitly negate components of the rules. The results show that linguistic factors have an important influence on subjects’ processing of problem information, and not simply on their initial representations of the logical rules.