Abstract
Regularities in the environment can help optimise attentional guidance in dynamic and perceptually complex environments. For example, when cycling through town, we may rely on prior experience with traffic lights, cars and pedestrians to carefully choose when - and where - to attend. Recent studies have shown that observers can instantly pick up task-embedded spatiotemporal regularities to predict the appearance of targets in a dynamic visual search design. Such predictions improve the accuracy and speed of performance. In the current study, we make a step forward to understanding whether anticipation-led attention also changes how we learn new information. Participants searched for eight unique targets that belong to a single category and appeared among visually similar distractors in a dynamic search display. Stimuli were letters and numbers (Exp 1; N=120) or pseudo-letters in different colours (Exp2; N=120), which continuously faded in and out on a textured background in extended trails lasting ~15 seconds. Critically, four out of eight targets were predictable in their onset and approximate location, allowing participants to form predictions. The first ten experimental trials were designed to facilitate learning of regularities. Subsequently, a memory task was introduced: participants recalled the targets they detected at the end of each trial (25 trials and 200 targets altogether). When analysing the data, we first identified that, as seen before, participants were more likely to find predictable targets. Then, we compared the probability of remembering selected targets based on whether they appeared predictably or randomly. Participants were significantly more likely to remember the recent predictable targets compared with the unpredictable targets. Items that occurred at the beginning of the trial were recalled equally irrespective of predictions. Therefore, predictions lead to enhanced memory representation of items only for the short-term, potentially most relevant for immediate behavioural guidance.