Effects of Hypothesis Saliency on the Use of Positive and Diagnostic Test Strategies

Abstract

The situation-centered model of inference under conditions of uncertainty (McDonald, 1990) is expanded to account for changes in test strategy employment over problems. Basic concepts regarding the transfer of learning between problems are incorporated into the model and lead to the prediction that diagnostic test strategy, rather than positive test strategy, will be the norm on problems presented after subjects have learned to access sets of hypotheses for which there is no identifiable salient (obvious) hypothesis. Two experiments were conducted to test this prediction. Both utilized a set of category discovery problems in which subjects were given an exemplar and were asked to discover the correct category to which the exemplar belonged by testing the membership status of additional exemplars. In Experiment 1, 83% of subjects solved a target problem in which the set of learned hypotheses was comprised of subsets of the correct category. This is a condition in which positive test strategy is completely ineffective, suggesting that most subjects employed diagnostic test strategy. In Experiment 2, 89% of subjects who listed an analyzable best guess tested exemplars that were not members of their best-guess category, behavior which is wholly inconsistent with the use of positive test strategy. The results of both experiments refute the idea that people have a bias toward using positive test strategy and specify some of the conditions under which positive test strategy and diagnostic test strategy will be employed.