Association chains and paired-associate learning

Abstract

We have proposed the hypothesis that degree of remoteness of a pair of words is related to the difficulty with which that pair can be established as a paired associate. Word chains were derived by combining stimulus words with their frequent responses, and as third elements frequent responses to the second elements. Free-association norms were used for this purpose. The results indicate that the hypothesis is tenable and may be summarized in three major conclusions. (1) Pairs in which the response member has the property of being frequently emitted as a free associate to the stimulus word are learned more rapidly than pairs of words not so related. (2) Second-order associates in which the response is related to the stimulus by being a common free associate to a common free associate to the stimulus are learned more rapidly than pairs of words in which the response member has few associations with the stimulus member. (3) First-order mediated pairs are learned more rapidly than second-order mediated pairs. The results were discussed in terms of a possible general model for paired-associate learning.