Himberger, K. D. 1 , Finn, A. S. . 2 & Honey, C. J. . 1
1 Johns Hopkins University
2 University of Toronto
Introduction: People are believed to continually extract regularities from the environment, a process known as statistical learning (Turk-Browne, 2012). For example, when one image (?A?) reliably precedes another image (?B?) then we may learn to predict B after seeing A (e.g. Turk-Browne et al., 2005). How much do pre-existing perceptual properties of stimuli affect learning which images follow one another?
Methods: Two stimulus sets were created using the same 12 unique stimuli (Schapiro et al., 2012). Each set contained 4 groups (?triplets?) of 3 stimuli. Crucially, one set (?Similar Set?) included a single triplet containing two consecutive ?similar? stimuli. The other (?Non Similar Set?) lacked any triplets containing similar stimuli. Stimuli were positionally matched across the two sets: an image that was second in its triplet in the Similar set would also be second in its triplet in the Non Similar set.
While participants performed a cover task, they were exposed to a sequence of triplets drawn from either the Similar or Non Similar sets. Stimuli within each of the 4 triplets appeared sequentially with absolute fidelity and transitions between triplets were random. After exposure, learning was assessed using implicit (response time facilitation) and explicit measures (two-alternative forced choice and triplet creation).
Results: Participants whose stimulus set contained the ?similar triplet? exhibited better evidence of learning the set structure in both the implicit and explicit learning measures. Importantly, this learning was not restricted to the similar triplet, but extended to other items within the Similar set.
Conclusion: Our results suggest that the ability to learn associations in a stream of stimuli is enhanced with only a pair of similar consecutive stimuli. Thus, the pre-existing associations between items may provide an entryway to superior learning, even while participants report no awareness of set structure.