Purchase this article with an account.
Fabian Soto, Edward Wasserman; The role of error-driven learning in object categorization by primates and birds. Journal of Vision 2011;11(11):839. doi: 10.1167/11.11.839.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Research in object categorization has focused on how the primate visual system extracts shape information from images and how that system subsequently constructs object representations, with little emphasis on the way in which task demands affect the categorization process via learning. Here, we show that a simple model of associations between representations of stimulus properties and responses based on error-driven learning can account for many findings in the literature on object categorization by nonhuman animals. Furthermore, the model generates predictions that have been confirmed in both people and pigeons, suggesting that these learning processes are widespread across distantly related species. In Experiment 1, people and pigeons showed impaired learning to sort objects according to their basic-level category if subjects were pretrained to memorize the correct responses associated with each individual object in the task; in other words, discrimination learning ‘blocked’ categorization learning. Such ‘blocking’ is usually taken as a hallmark of error-driven associative learning. Experiment 2 used a Go/No Go task in which pigeons' responses to objects from a single category were rewarded and nonrewarded on different trials. As predicted by an error-driven learning account, generalization of responding to new objects from the category was lower when the individual training objects were informative about reward than when those individual objects were uninformative. In Experiment 3, object categorization learning was impaired when a second category of objects provided redundant information about correct responses. The same impairment was not observed when single objects provided redundant information, but the category to which they belonged was uninformative. This ‘overshadowing’ effect was also predicted by our model. These results indicate that, across distantly related species, prediction error is necessary for categorization learning to occur. These results thus add to a growing body of evidence implicating common mechanisms in object categorization by birds and primates.
This PDF is available to Subscribers Only