Salience-Based Selection: Attentional Capture by Distractors Less Salient Than the Target
Figure 5
Stochastic model of salience-based selection.
(a) For each location in the visual field, salience is accumulated over time t = {t1, t2,…, tk} by leaky accumulators. Gray jagged lines represent sample paths of sensory evidence accumulation over time, influenced by noise. Mean accumulation behavior is indicated by solid black lines. Salience asymptotes s (st = target salience, sd = distractor salience, snt = non-target salience) indicate maximum salience when time is infinite and noise absent; asymptotes correspond to the salience values of map locations computed by deterministic models. (b) Selection time distributions (t = target, d = distractor) indicate selection time variation due to noise. Overlap of these distributions (red area) marks the range within which a distractor may be selected first even if it is less salient than the target. (c) The final salience pattern evolves over time, as illustrated by heat maps at different points in time.