Purchase this article with an account.
Brandon Barakat, Aaron Seitz, Ladan Shams; There is more to statistical learning than associative learning: Predictable items are enhanced even when not predicted. Journal of Vision 2012;12(9):694. doi: 10.1167/12.9.694.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Statistical learning refers to automatic and unconscious learning of associations between stimuli based on their repeated temporal and/or spatial relationships. Previous studies that have used an RSVP detection paradigm have demonstrated that subjects respond more quickly to target stimuli when they follow stimuli to which they have been temporally associated. These results were interpreted as reflecting priming of 2nd items by 1st items within a learned pair. However, an alternative interpretation is that 2nd items are more easily detectable in general after learning. To test between these two competing hypotheses, we conducted a visual statistical learning experiment (Experiment 1) in which subjects’ reaction times were measured in response to 2nd items when they were preceded by associated items (match condition) vs when they were preceded by non-associated items (mismatch condition). Our results showed that subjects responded with faster reaction times to 2nd items of pairs compared to 1st items in both the match and mismatch conditions. These data suggest that the reaction time results observed in statistical learning studies of this kind are not exclusively due to associative learning. We conducted two follow-up experiments to examine whether the temporal predictability of the 2nd items was leading to an increase in their general saliency. Specifically, after inducing statistical learning by exposing subjects to temporally associated visual pairs, we tested their ability to detect the stimuli in a visual search task (Experiment 2) and a masking task (Experiment 3). Our results indicated that subjects had a significant change in response bias and a trend in increased detection sensitivity for 2nd items compared to 1st items. Altogether, these results suggest that statistical learning leads to representational changes for the individual items and that previous findings of associative effects are at least partially confounded by increased sensitivity and perceptual biases for statistically reliable stimuli.
Meeting abstract presented at VSS 2012
This PDF is available to Subscribers Only