Five groups composed of 24 college students each were run in a three-choice probability learning experiment in which information was given only concerning correct responses. Different groups received different assignments of reinforcement probabilities to the three choices. Major findings of this investigation were (i) asymptotic response probability approximated probability matching if reinforcement probabilities summed to unity, and (ii) for all reinforcement schedules, statistics of the data were well described by the linear or n-element models of statistical learning theory using the assumption that following incorrect responses, response probability was redistributed in a constant ratio of the probabilities of the reinforcements associated with the remaining alternatives.

A continuous technique for studying short-term memory of paired-associates was used. For 4, 6, or 8 stimuli, recall was found to be a decreasing function of the number of pairs of items interpolated between study and test on a given item. Reliable differences between the functions for 4, 6, and 8 stimuli were found.