Learning of ill-defined categories by monkeys.

Publisher Website
Google Scholar
Abstract
Three rhesus monkeys were trained on a categorization task using a successive discrimination paradigm (for a category A exemplar, go left; for a category B exemplar, go right). The category members were geometric stimuli differing in color, form, size and orientation, and the categories were ill-defined in the sense that component features of exemplars were neither singly necessary nor jointly sufficient to determine category membership. Original learning was followed by transfer tests involving both training stimuli and new exemplars. After this initial learning and transfer phase the assignment of categories to positional responses was reversed (e.g., for A exemplars, go right; for B exemplars, go left). After the reversal learning criterion was met, transfer tests were again given. This procedure of training followed by transfer tests was continued until a total of 5 reversals had been completed. Monkeys were able to learn this task involving ill-defined categories and showed both appropriate transfer to new examples and improvement across successive reversals. Both the learning and transfer data were in accord with the idea that component features are encoded or combined in an interactive rather than an independent manner and showed strong parallels with human classification data.

This publication has 3 references indexed in Scilit: