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

Abstract

It is well established that associative learning, such as learning new cue–outcome pairings, produces changes in attention: cues that are good predictors of relevant outcomes become prioritised compared with those that are non-predictive or redundant. However, there is controversy about whether such a learnt attentional bias results from a controlled orientation of attention, or whether it can be involuntary in nature. In three experiments, participants learned that cues of certain colours were predictive or non-predictive, and we assessed attention to cues using a dot-probe task. On dot-probe trials, participants were instructed to control attention by orienting towards a cue of a certain shape (target), while trying to ignore another cue (distractor). Although the colours of the cues were critical for the associative learning task, they were irrelevant for the dot-probe task. The results show that, even though participants’ controlled attention was focused on the target shape (as evident in response times and accuracy data), response times to the probe were slower (Experiments 1 and 2) and error rates were higher (Experiments 2 and 3) when the distractor was of a (previously) predictive colour. These data suggest that attention was captured involuntarily by the predictive value of the distractor, despite this being counterproductive to the task goal.