Episodic Specialization of Classification Rules

Abstract

<p>This thesis reports research investigating the effect of practice on the operation of an explicitly stated rule. Subjects were furnished with an easy, sufficient additive rule and given practice classifying pictures of imaginary animals into two categories. Following practice, subjects showed a strong similarity effect: They were faster and more accurate at classifying positive matches (new items in the same category as the most similar old item) than negative matches (new items in the opposite category to the most similar old item), suggesting that memory for specific prior episodes played an important role in applying the rule. Subjects were not always wrong on the negative matches, and were usually immediately aware of any mistakes they did make, suggesting that the rule had not been abandoned but may have served a monitoring function. Mnemonically distinctive stimuli are probably important for episode-based responding since these effects did not occur when the items were presented as written lists of features. When subjects were stressing accuracy, and paying a large premium in response times, the negative match effect remained, while fewer than half the subjects considered similarity to old items when questioned about the cause of their errors, suggesting that the episodic information can have effects without subjects' awareness. However, a similarity strategy can be under strategic control as shown when the rule was manipulated so that information from prior experience was in opposition to the present rule information for nearly all the items. Additionally, negative matches were more disrupting if seen in the context of predominantly old items, these episodic effects were helpful when the rule was not perfectly predictive, and the classification decisions of third grade children were strikingly similar to those of the adults suggesting that children of this age group were employing a similar balance between the use of the abstract and episodic information. While some previous accounts of categorization suggest that episode-based responding is important only before subjects develop a rule or some other basis of categorization the current results suggest that, for some common types of stimuli, similarity to prior episodes may become more important with practice.</p>