Conservative cutoff placement in a numerical decision task

Abstract

Conducted 2 experiments with 40 university students who were required to make binary decisions on 5-digit numbers drawn from 2 overlapping normal distributions. In Exp I, there were 2 conditions of discriminability determined by the degree of the overlap of distributions. Although observed cutoff locations for uneven prior probabilities of stimulus distributions were conservative in both conditions, the Ss in the low discriminative condition changed their cutoff placement less than those in the high discriminative condition. In both conditions, Ss' response probabilities were close to the prior probabilities. In Exp II, the verbal protocol during the task showed that ambiguous impressions were used for judgments and that responses were made as accurately as possible to both stimulus distributions. It is concluded that the heuristics used by Ss in the task could account for the conservative cutoff placement and that the heuristics confirmed their judgments. (PsycINFO Database Record (c) 2012 APA, all rights reserved)