Abstract
Perceptual learning (PL), improvements in perception due to practice (Gibson, 1969; Kellman, 2002), appears to be influenced by effects found in non-perceptual learning (NPL) domains. For example, spacing and interleaving effects in factual learning also appear in perceptual category learning (Mettler & Kellman, 2014; Carvalho & Goldstone, 2015). As in NPL domains, learning in PL can be either passive (requiring only observation) or active (requiring active responses followed by feedback). Here we study the effects on PL of active vs. passive presentations and their combination, previously investigated primarily in NPL domains. Method: Stimuli were organized into 12 categories of butterfly genera with 9 exemplars per genus. Participants learned to match genus names to exemplar images. Training conditions consisted of either: a) passive presentations only, b) 2 initial blocks of passive presentations followed by active, adaptive learning, c) 1 initial passive presentation per category, followed by active, adaptive learning, or d) active adaptive learning only. Adaptive schedules adjusted spacing based on learner performance (Mettler, Massey & Kellman, 2016) and trained learners to a criterion; Passive Only schedules had a fixed number of trials. For each training category, one exemplar was withheld from training and used as a test of generalization at posttest. Participants received a pretest, immediate posttest, and one week delayed posttest. Efficiency, accuracy adjusted for trials invested in training, was compared across conditions. Results: Initial blocks of passive trials followed by active adaptive learning resulted in the most efficient learning. Active Only, Passive Only, and Passive Initial Exemplar conditions fared worse. Results were similar across familiar and novel exemplars. Conclusion: PL with initial blocks of passive trials followed by active adaptive learning enhanced PL. These results are similar to effects found in factual learning, suggesting that common learning principles and/or mechanisms underlie passive-active synergies in PL and NPL domains.
Acknowledgement: Supported by NSF Grant DRL-1644916