Poster E64, Monday, March 27, 2:30 – 4:30 pm, Pacific Concourse
The Effect of Instruction on People’s Ability to Learn Simultaneous Statistical Inputs
Tess Allegra Forest1, Taraz Lee2, Ashkan Kiyomarsi1, Amy Finn1; 1The University of Toronto, 2The University of Michigan
While Statistical Learning is a powerful and largely automatic mechanism, it is important to know its limitations. We are specifically interested in 1) whether it’s possible to learn two statistical patterns at the same time, and 2) if instructions to attend to one pattern boost learning for either. We exposed 54 adults to a 6-minute stream of black dots inside coloured circles; circles were one of 9 colors and dots occurred in one of 9 locations within them. The stream was comprised of 3 color triplets and 3 dot-location triplets. Color and dot triplets were offset so that the onset of each streams’ triplets did not coincide. During exposure, participants completed a detection task, pressing X when a circle appeared and Y for a randomly-appearing blank square (10% of trials, no relationship to either pattern). Half the participants were given explicit instructions to learn the pattern of the dot locations, while the others received no stream-related instructions. Results showed that without explicit instructions, adults learned the triplets of each stream with minor success, while explicit instruction to attend to a particular stream (dots), improved learning for that stream. The poor learning in the no-instruction condition could stem from attempts to learn the relationship between, or bind, the streams, rather than learn each structure separately. In support of this, individuals with a preference for bound sequences over sequences that maintained triplet structures, displayed less knowledge of dot triplets.
Topic Area: LANGUAGE: Other