Abstract
Attention fluctuates between optimal and suboptimal states. How do these changes affect what we learn from our environments? In particular, do they affect the degree to which we learn visual regularities? This question seems nearly impossible to answer: Although we learn regularities across repeated pattern exposure, we may be attentive at one exposure but inattentive the next. To overcome this challenge, we designed a task in which visual regularities are presented contingent on attentional state. In an online study (N=150), participants performed a continuous performance task with shape stimuli (1200 trials, 800 ms/trial). They were instructed to press a button in response to shapes from a frequent (90%) but not infrequent category (10%). We measured correct-trial response times (RTs) in real-time, and inserted distinct shape triplets in the trial stream when RTs indicated that a participant was attentive (>1 SD above the participant’s mean RT) or inattentive (>1 SD below the participant’s mean) (deBettencourt et al., 2018; 2019). In other words, participants saw one sequence of three shapes when they were attentive (M=17 triplet repetitions) and another when they were inattentive (M=17.8). Participants next performed a task in which they responded to target shapes selected from the regular triplets. Demonstrating that participants learned regularities, we observed a main effect of intra-triplet position in the target detection task, such that shapes drawn from the third position of the regular triplets were detected faster than shapes from the earlier positions. Furthermore, we observed an interaction between attentional state and intra-triplet position, such that this RT facilitation was greater for the triplet encountered in the attentive vs. the inattentive state. Together these results demonstrate statistical learning for regularities that are not explicitly task-relevant and show, for the first time, consequences of sustained attention fluctuations for visual statistical learning.