Generalization in auditory perceptual learning: Experimental and theoretical perspectives
Wisniewski, Matthew G.
MetadataShow full item record
Theories of perceptual learning make different predictions about the direction of generalization (positive or negative) and how that generalization changes over time. Positive generalization, where learning benefits an untrained condition, has been well studied. In contrast, negative effects of perceptual learning on an untrained task are rarely reported. The current behavioral experiments tested auditory learning's effect on generalization along a continuum (Experiment 1), to an irrelevant auditory dimension (Experiment 2), and to an untrained task (Experiment 3). Results indicate that the relationship between trained and untrained sounds, and the tasks used in training and testing, affect the direction and extent of generalization. Simulations using a connectionist model incorporating standard computational specifications of representational modification and reweighting learning mechanisms replicated the behavioral data. Furthermore, simulations argue for hybrid models of perceptual learning because neither mechanism on its own accounted for all results.