Compound trial-and-error learning as a function of response availability (nr).

Abstract

The study involved paced four-link heterogeneous compound trial-and-error learning with immediate serial reinforcement under seven conditions of response availability (Nr) and task length held constant. 336 human subjects were used in a simple randomized design and the ratios of correct to available responses were 4/4, 4/5, 4/6, 4/7, 4/8, 4/9, and 4/10. To a series of four stimuli presented for 2 sec. each at a 3.44-sec. rate with a 6.88-sec. intertrial interval, S was required to learn an in-variant sequence of four pushbutton responses. The criterion of mastery was two successive error-free trials. Response availability was manipulated by means of a set of detachable covers for the apparatus response keys. It was found that Nr exerted a highly significant retarding effect upon the rate of trial-and-error learning, in accordance with expectation. Probability of correct first choice (p+) for three representative values of Nr yielded ogival acquisition curves as functions of trials. The curves originated at hypothetical p+ values consistent with the initial chance solubility of each problem. The apparent inflection points were a direct function of Nr while the acquisition rates varied inversely.