Event Tracking in Probability Learning
Arthur S. Reber, Richard B. Millward · The American Journal of Psychology · 1971
After 1,000 observation trials in which the event schedule followed a sawtooth wave with a period of 50 trials, subjects tracked the events very closely on periodic test trials but showed no periodic tracking on no-event or constant-probability test trials. During the early periodic-event cycles the subjects' response proportions tracked the event proportions approximately as predicted by statistical learning theory, but during later cycles, their response proportions peaked and bottomed at the same points as the event proportions, a relationship nicely described by the averaging model of Friedman, Carterette, and Anderson.