Learned predictiveness influences automatic evaluations in human contingency learning

Experiments used an affective priming procedure to investigate whether evaluative conditioning in humans is subject to bias as a consequence of differences in the learned predictiveness of the cues involved. Experiment 1, using brief prime presentation, demonstrated stronger affective priming for cues that had been predictive of a neutral attribute prior to evaluative conditioning than for those that had been nonpredictive. Experiment 2, using longer prime presentation, found a reversed priming effect for previously predictive cues but not for previously nonpredictive cues. The implication is that the effect observed with brief prime presentation reflects the operation of fast-acting, automatic evaluation mechanisms and hence that evaluative conditioning can be biased by our previous learning about the predictiveness of cues.