Abstract
In dynamic environments, encoding information in working memory (WM) depends on allocating attention to the relevant object at the right time. In rapid serial visual presentation (RSVP) tasks, failures in attentional selectivity are frequently observed when a target is followed by a potentially reportable distractor. However, in tasks with two targets, accuracy for both targets is typically high when they are presented in immediate succession (lag-1 sparing). To account for this disparity, we tested whether expectations about the number of targets in RSVP streams gates their access to WM. Colored target digits were embedded among grey letters and digits in two lateralized RSVP streams. The first target was followed either by a grey digit, or a second target (another colored digit). To manipulate expectations, the ratio of one-target and two-target trials (75%-25% or vice versa) was varied between blocks. Participants were much more likely to report seeing two targets when two targets were expected, even on trials where only a single target was present. To rule out response bias, we measured ERP markers of attentional selection (N2pc) and WM storage (CDA) in a second experiment. Both components were larger when two targets were expected, regardless of the actual number of targets, demonstrating that expectations modulated attentional selection as well as the number of items encoded in WM. In a third experiment, participants always provided two guesses about the targets’ identity, and blocks with 100% one-target trials were followed by blocks with 100% two-target trials, or vice versa. The number of reported items varied with expectations only for observers who started with one-target blocks, plausibly reflecting the strategic advantage of maintaining an expectation for two targets in this task. Together, these findings reveal that attentional selectivity and WM encoding is modulated by expectations about the amount of task-relevant information.