Abstract
It has long been known that frequently occurring targets are better attended than infrequent ones in visual search. But does this frequency-based attentional prioritization reflect short-term inter-trial priming or durable statistical learning? Here we show that both short-term and long-term mechanisms contribute to attentional biases for visual features. However, they are supported by different types of statistical associations between targets and features. Four experiments showed that a manipulation of the target's probability of having specific features induced only short-term prioritization of the more probable feature. In contrast, a manipulation of a feature's probability of being associated with a target rather than distractors (the feature's diagnostic value) produced a durable attentional bias towards the more diagnostic feature. Participants searched for a target, a line oriented horizontally or vertically among diagonal distractors, and reported its length. In one set of experiments we manipulated the target's color probability: targets were more often in Color1 than Color2. Distractors were in other colors. Participants found Color1 targets more quickly than Color2 targets, but this preference disappeared immediately when the target's color became random in the subsequent testing phase. In another set of experiments we manipulated the diagnostic values of two colors. Color1 more often coincided with targets than distractors; Color2 more often coincided with distractors than targets. Participants found Color1 targets more quickly than Color2 targets. Importantly and in contrast to the first set of experiments, the featural preference was sustained in the testing phase. These results show that short-term and long-term attentional biases are products of different statistical information. Finding a target momentarily activates its features, inducing short-term repetition priming. Long-term changes in attention, on the other hand, rely on learning diagnostic features of targets.
Meeting abstract presented at VSS 2017