ABRUPT LEARNING AND RETINAL SIZE SPECIFICITY IN ILLUSORY-CONTOUR PERCEPTION

Citation
N. Rubin et al., ABRUPT LEARNING AND RETINAL SIZE SPECIFICITY IN ILLUSORY-CONTOUR PERCEPTION, Current biology, 7(7), 1997, pp. 461-467
Citations number
24
Categorie Soggetti
Biology,Biology
Journal title
ISSN journal
09609822
Volume
7
Issue
7
Year of publication
1997
Pages
461 - 467
Database
ISI
SICI code
0960-9822(1997)7:7<461:ALARSS>2.0.ZU;2-A
Abstract
Background: In behavioral studies of learning, a distinction is common ly made between gradual and abrupt improvements in performance. The le arning of perceptual and motor skills is often characterized by gradua l, incremental improvement, and is found not to generalize over stimul us manipulations such as changes in the size or location of the retina l image. In contrast, marked improvement in performance can occur sudd enly - a phenomenon which has been termed 'insight'. Consequently, the brain mechanisms subserving the two types of learning are commonly th ought of as distinct. Here, we examine learning of a perceptual task i n which improvement appears to exhibit characteristics of both gradual and abrupt learning. Results: We describe experiments on illusory-con tour perception in which the observers underwent an abrupt, dramatic i mprovement in performance, resembling an incident of insight. Al the s ame time, however, the phenomenon showed a degree of stimulus-specific ity that was previously thought to characterize incremental, gradual l earning. The improvement was triggered only by specific visual stimuli , whereas other, quite similar, stimuli were found to be ineffective f or training; the learning did not generalize to a new retinal image si ze, and re-training was necessary for different-sized images. Conclusi ons: The juxtaposition of abrupt and stimulus-specific learning that w e observed suggests that the distinction between the two forms of lear ning needs to be revised. Rather than postulating two distinct mechani sms, incremental and insightful learning need to be addressed within a single framework. In particular, the findings suggest that learning m ay involve interactions between multiple levels of representations of the stimulus.