Abstract
There is growing consensus that selection history plays an important role in guiding spatial attention and should be distinguished from current goals and physical salience. Here, we focused on a selection-history phenomenon known as statistical learning of target location: when the target is more likely to appear in one (high-probability) region than in other (low-probability) regions, search performance gradually improves for targets appearing in that region. This effect is thought to reflect an implicit, long-lasting and inflexible attentional bias towards the high-probability region. However, strong evidence for these claims is lacking. The objective of the present research was to test them. In three experiments (N=32, each), the target was more likely to appear in a high-probability region than in the other low-probability regions during the learning phase, and all regions were equiprobable during the extinction phase that followed. The number of distractors was manipulated and participants’ awareness of the probability manipulation was measured. We found that statistical learning reduced search slopes, and that its influence was independent of participants’ awareness of the probability manipulation. Although inter-trial priming of the target location strongly affected both overall response times and search slopes several trials back, this effect did not account for all the statistical learning effect. In addition, the bias during the extinction phase was equally large irrespective of whether or not participants had been informed of the probability imbalance after the learning phase and of its discontinuation in the upcoming extinction phase. We conclude that statistical learning of target location is indeed an implicit, long-lasting and inflexible attentional bias towards the high-probability region.