Testing the automaticity of an attentional bias towards predictive cues in human associative learning

The process of learning an associative relationship between two events, such as a cue-response pairing, can be shaped by changes in attention. It is well established that predictive cues are prioritized for attentional processing compared to those that are non-predictive. However, there is controversy about whether this attentional bias is controlled or automatic in nature. In two experiments, participants learnt about cue-response relationships in an associative learning (AL) task. Attention to the cues was assessed using a dot-probe (DP) task. In AL trials, participants learned to make appropriate responses to compounds of two cues that differed in color and shape, but only color predicted the correct response. The same cues were used on DP trials, but shape was the relevant feature, since probes appeared more frequently over one specific shape. Even though participants were informed that the only relevant feature for the DP task was shape, and their controlled attention was focused on the relevant shapes (as evident in response times and accuracy data), response times to the probe were faster (Experiments 1 and 2) and error rates were lower (Experiment 2) when it appeared over a predictive color (for the AL task); the opposite pattern was found when the predictive cue was displayed as a distractor. These data suggest that attention is captured automatically by predictive cues, despite this being counterproductive to the task goal.