Some situational determinants of hypothesis-testing strategies

Abstract

An elementary hypothesis-testing model is advanced to account for behavior across a wide spectrum of social inference problems. According to the model's test conditions assumption, positive test strategy will be employed whenever the accessible hypothesis set consists of a single working hypothesis; however, when the accessible set consists of more than one, nonexclusive hypothesis and test choices are unconstrained, diagnostic test strategy will be employed. Two experiments were conducted to test these predictions. Experiment 1 utilized a set of category induction problems for which pretesting had established the existence of a single salient hypothesis. Such problems are predicted to elicit positive test strategy and, whenever the salient hypothesis is a subset of the true hypothesis, positive test strategy will yield only confirming feedback and will lead to the erroneous conclusion that the salient hypothesis is correct. As predicted, solution rates were low when the salient hypothesis was a superset of the true hypothesis. Instructions to construct multiple hypotheses increased solution rates only in the latter condition, suggesting that multiple-hypothesis subjects employed diagnostic test strategy. In Experiment 2, the universe of allowable hypotheses was constrained to two alternatives, one being a subset of the other. The majority of subjects in this situation demonstrated effective use of diagnostic test strategy, and 84% of all problems were solved in only one test trial. The results of both experiments help refute the notion that people have an absolute bias towards employing positive test strategy and provide support for the proposed situation-centered model.