The strategic basis of performance in binary classification tasks: Strategy choices and strategy transitions

Three experiments were conducted using a probe procedure in binary classifi cation tasks to track strategies and strategy transitions over blocks of tr ials and to determine the effects of task difficulty (easy or hard classifi cation rule) and task rule familiarity (artificial rule or natural language rule) on classification performance and strategy use. Two primary response strategies were identified, one based on the use of a classification rule and one based on the retrieval of instances from memory. There was clear ev idence of rule use during the early trials, but this rule strategy eventual ly was replaced by the instance strategy except when the easy natural langu age rule was in effect. Response time speed-up followed a power function th at was specific to the strategy in use. (C) 1999 Academic Press.