Pseudocontingencies

Abstract

In 3 experiments, the authors explored a contingency illusion termed pseudocontingency (PC) that produces logically unwarranted but potentially useful inferences. PCs arise when bivariate contingencies are inferred from univariate distributions via heuristic alignment processes. For example, in the absence of information about the co-occurrence of TV habits and aggressive behavior within a school class, when the prevalence of both attributes is high, a teacher may infer a positive PC as if students who often watch TV were highly aggressive. Throughout 3 experiments, predictions of the level of 1 variable from the level of another served as a measure of PCs. The illusion could be evoked reliably whether information about target attributes was presented successively or simultaneously, whether common-cause or common-effect models were activated, and whether attributes involved 2 or more levels. The discussion centers on the cognitive processes underlying PCs and their origin and adaptive value.