Cue Predictiveness and Uncertainty Determine Cue Perception During Statistical Learning

This study investigated the key mechanisms by which cue predictiveness impacts cue perception under different levels of uncertainty during statistical learning. Although previous studies have demonstrated the impacts of cue predictiveness on cue processing during explicit associative learning tasks, it is still unclear whether the same phenomenon can be observed during implicit learning tasks. Using a novel probabilistic cueing-validation paradigm administered to 35 adults (Experiment 1) and 31 adults (Experiment 2), we systematically manipulated cue predictiveness through the conditional probability of a target following a cue, i.e., high (75%), medium (50%, only in Experiment 1), low (25%), and zero (only in Experiment 2). Different levels of uncertainty were raised by high-predictable (75%), medium-predictable (50%, only in Experiment 1), and low-predictable (25%) targets appearing after cues. Following half of these targets, we probed the perception of cues with various predictiveness. Our study demonstrated 1) impeded perception of high-predictive cues after high-predictable targets, indicating exploration-like cue processing after low uncertainty inputs; and 2) enhanced perception of high-predictive cues after low-predictable targets, indicating exploitation-like cue processing after high uncertainty inputs. These results suggest that uncertainty may regulate the dominance of attention-dependent and attention-independent learning systems during statistical learning.