Single-cue probability learning as a function of the sign and magnitude of the correlation between cue and criterion

Abstract The effects of the sign and magnitude of the cue-criterion correlation, rCE, the slope of the cue-criterion regression line, bce, and the magnitude of the unaccounted for variance in the task system Sce2, upon single-cue probability learning were studied in three experiments. Positive cue-criterion relations were learned faster than negative relations, and tasks with high values of Sce2 were learned faster than tasks with low values of Sce2. The final level of performance was, however, determined only by rce. At this level, the cue-response correlations exceeded rce. The amount of overshooting was an inverse function of rce, due to the fact that the subjects' prediction strategies were more extreme for low levels of rce than for high levels. The relative consistency of their response systems was, however, not affected by rce.