A study of disjunctive reasoning

Abstract

Whilst it is currently fashionable to construct sequential models of reasoning, the results of recent experiments on conditional reasoning tasks are apparently incompatible with this notion. There is evidence of two orthogonal statistical components, one logical and the other non-logical. It has been suggested that non-logical tendencies are ‘response biases’ independent of the linguistic structure of the rule. However, certain results suggest that the ‘matching bias’ observed in conditional tasks may not occur with disjunctives. Two experiments are reported in which subjects generated ‘subjective truth tables’ for disjunctive rules, in which the presence and absence of negative components was varied. Experiment I provided a powerful test of the matching bias hypothesis and found no evidence of its presence. The presence of negative components did, however, systematically affect performance. Experiment II, with the aid of latency measurements, provided support for the hypothesis that negatives were affecting the ease of interpretation of the disjunctive sentences, with one negative component causing at least as much difficulty as two. In discussion, it is pointed out that the present results can be reconciled with those on conditional tasks, on the assumption that ‘matching bias’ is a special case of associational biases which are dependent on the linguistic form of the rule. It still appears necessary to retain the distinction between logical and non-logical components of the performance. However, this does not necessarily imply parallel processes underlying the data. A single muddled thought process, in which a rational sequential strategy fails to be consistently applied, could give rise to the type of data obtained.