Outcome Predictability Biases Cued Search.

Within the domain of associative learning, there is substantial evidence that people (and other animals) select amongst environmental cues on the basis of their reinforcement history. Specifically, people preferentially attend to, and learn about, cueing stimuli that have previously predicted events of consequence (a predictiveness bias). By contrast, relatively little is known about whether people prioritize some (to-be-predicted) outcome events over others on the basis of their past experience with those outcomes (a predictability bias). The present experiments assessed whether the prior predictability of a stimulus results in a learning bias in a contingency learning task, as such effects are not anticipated by formal models of associative learning. Previously unpredictable stimuli were less readily learned about than previously predictable stimuli. This pattern is unlikely to reflect the use of strategic search processes or blocking of learning by the context. Instead we argue that our findings are most consistent with the operation of a biased learning mechanism.