Probability learning with four event classes of schematic stimuli

Describes 2 experiments with a total of 80 college students in which classification performance was explored using distorted patterns within a sequentially dependent temporal context. Visual stimuli representing 4 prototype classes were used. All Ss classified patterns by identifying each with 1 of 4 prototypes, and rated the similarity of each pattern to the prototypes. In addition, Ss in 2 groups predicted which pattern would occur next and indicated their confidence in that prediction. Stimulus classes were arranged so that there were sequential dependencies to be learned. Results indicate that both classification and prediction performance improved with sequence redundancy. Prediction performance improved during training, demonstrating that probability learning can occur even when the S is dealing with distorted patterns. Moreover, sequential dependencies in the stimulus sequence facilitated Ss' accurate classification of the patterns. (French summary) (17 ref)