Expectation is jointly shaped by global and transitional probabilities

Humans implicitly extract statistical regularities from sensory input across multiple perceptual dimensions. But it is unclear how statistical learning builds expectations that in turn may affect perceptual decisions, especially in the context of co-occurring regularities. In audition both transitional probabilities (across successive sounds) and global probabilities (sounds’ overall frequency of occurrence) shape perception. Here, we investigated how these two forms of statistical structure jointly influence perceptual judgments. Participants heard a 70-ms cue tone followed by a 50-ms or 90-ms target tone and judged whether the target tone was short or long. Statistical regularities were embedded along a task-irrelevant dimension, tone frequency, allowing for assessment of how implicit statistical learning unfolds when regularities are orthogonal to task goals. A sensory match in cue and target tone frequencies provided no advantage to duration decision response times. However, duration decisions were faster when cue tone frequency predicted target tone frequency across a high transitional probability. Target tones with high global probability drove faster duration decisions compared to those with low global probability. The combined evidence indicates that global target probability and cue–target transitional probability exerted similar magnitude effects, but their combined influence was not additive. Thus, even in active tasks demanding attention to a specific dimension (duration), listeners implicitly learn statistical structure along an irrelevant dimension (frequency) and track multiple regularities. This work advances our understanding of how layered statistical regularities guide prediction in goal-directed listening contexts.