Bending the Power Law: The Transition from Algorithm-Based to Memory-BasedPerformance

摘要

Two theories of tasks that exhibit a transition from algorithm-based to retrieval-based performance were compared. The instance theory of automaticity assumes parallel strategy execution and instance-based memory representation and predicts power function reduction in mean reaction time and standard deviation of mean reaction time (SD) with practice. An alternative theory is proposed that assumes nonparallel strategy execution and strength-based memory representation and that predicts, among other things, power function speed-up and reduction in SD within each strategy, and systematic deviations from power functions in both of these variables when strategy transition occurs. Two of these experiments employing pseudoarithmetic are reported. The results of these experiments are consistent with the new theory that assumes nonparallel strategy execution and strength-based memory representation. These results also constitute the first convincing demonstration of a class of adult skill acquisition tasks for which the power law of practice does not apply overall, a finding that should have notable implications for a variety of human skill acquisition theories.