Purchase this article with an account.
Everett Mettler, Philip Kellman; Adaptive Sequencing in Perceptual Learning. Journal of Vision 2010;10(7):1098. doi: 10.1167/10.7.1098.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Question: In real-world perceptual learning (PL) tasks learners come to extract distinguishing features of categories, enabling transfer to novel instances. This kind of learning can be accelerated by structured interventions involving a series of classification trials (e.g., Kellman, Massey & Son, 2009, TopiCS in Cognitive Science). Little is known about practice schedules that optimize PL, nor their relation to laws of learning for factual items. Method: We tested an adaptive sequencing algorithm for PL that arranged spacing for categories as a function of the individual learner's trial-by-trial accuracy and reaction time. Participants learned to classify images from 12 butterfly genera. Each genus contained 9 exemplars from 3 species (Experiment 1) or 9 exemplars from 1 species (Experiment 2 - low variability categories). 1 of the 9 exemplars was not presented in training and was used as a test of novel transfer. Training trials were 2AFC where participants matched one of two images to a genus label. During training participants received either: 1) random presentation, 2) adaptive sequencing, or 3) adaptive sequencing with sets of 3 sequential category exemplars (mini-blocks). Participants completed pre and post-tests immediately before and after training, and an additional post-test after a 1-week delay. Results: Learning efficiency (accuracy per learning trials invested) was reliably greater for adaptive sequencing. Effects persisted over a 1-week delay and were larger for novel items. In experiment 2 where the variability of category exemplars was lower, adaptive sequencing resulted in even greater learning efficiency gains. Mini-blocks hurt efficiency in both experiments, especially for novel items. Conclusion: Results suggest that, across a range of category distributions, adaptive sequencing (but not blocking) increases the rate of learning and benefits novel transfer – key components of PL and fundamental aspects of learning in many domains.
This PDF is available to Subscribers Only