New gradients of error reinforcement in multiple-choice human learning.

Publisher Website
Google Scholar
Abstract
Detailed analysis of the multiple-choice learning records of 52 human subjects revealed not only the clear-cut operation of the Thorndike spread of effect type of error-repetition gradient, but also two new error-repetition gradients which do not seem explicable on the basis of the various criticisms levelled against the Thorndike gradient. These were (1) an error-recovery gradient according to which errors made several trials previously tended to be repeated more frequently, following the occurrence of even a completely new reinforced response, as a function of their closeness in stimulus presentation steps to the reinforcement; and (2) and intra-stimulus error-recovery gradient according to which errors tended to recur more frequently in later trials as a function of their closeness, in trial units, to a single reinforced response made to the same stimulus. It was suggested that these gradients are most effectively explained on the basis of the generalized hypothesis that responses are organized into a variety of functional systems which are susceptible to interaction effects resulting from reinforcement.

This publication has 1 reference indexed in Scilit: