Automatic category search and its transfer.

Experiments examined practice and transfer effects in consistently mapped (CM) and variably mapped (VM) semantic search. Experiment 1a examined improvements in reaction time in detecting words from a category as a function of the number of exemplars (4-12) in the category. All CM conditions showed improvement, but there was no significant effect of the number of exemplars. Experiment 1b examined the extent to which training on a subset of exemplars transferred to untrained members of the category. Results showed substantial positive transfer (60%-92%) to untrained exemplars from the trained category. The transfer was better if there were more exemplars in the training set. Experiment 2a showed practice reduced resource sensitivity in CM category search but did not benefit VM category search. Experiment 2b showed that under high workload, untrained exemplars of the trained CM category were detected when first presented to exhibit substantial positive transfer (70%). We conclude that many of the practice effects observed for CM category search take place at either the category level or the category feature level. We suggest that practice results in context activation of the category node or category features. This context activation hypothesis is evaluated with respect to major phenomena relating to automatic and controlled processing.