An experiment was performed to demonstrate that natural language mediators are not merely associations to word pairs, but also playa role in learning by reducing the learning of a pairedassociate item to two simpler learning tasks. Although the study times were the same, subjects who gave one word as an association to each word pair performed more poorly on measures of retention than did subjects who gave one word which described or connected the two words of each pair. Also, evidence for some independent visual-image mediation was found.When subjects are given a paired-associate learning task in which the items have non~anguage components such as CVCs, the learning of each item is often accompanied by a natural language mediator (Bugelski, 1962; Montague, Adams, & Kiess, 1966; Underwood & Schulz, 1960). The natural language mediator is usually a word or phrase that sounds like or, when written, looks like the original item but is part of the subject's natural language. It has been found that the presence of a natural language mediator for an item indicates that the subject is more likely to have learned and is more likely to retain that item compared to an item for which no natural language mediator has been reported (for a review see Montague, 1972). Although interesting phenomena in their own right, natural language mediators may also provide an index by which the depth of processing (Craik & Lockhart, 1972) or degree of elaboration (Craik & Tulving, 1975) can be measured.Bellezza and Poplawsky (I974) found that natural language mediators facilitated learning even when the pairs being learned were randomly paired nouns which were already part of the subjects' natural language. For some of the presented pairs the subjects were asked to give one word that described, connected, or associated the two words in the pair. They found that pairs for which a one-word mediator was requested were recalled significantly better than control pairs which the subjects were told to silently study. Also, as found in other studies such as Montague et al. (I966)
460word was almost never recalled unless the mediator word was also recalled.There is agreement resulting from experimentation that if the subject can give a natural language mediator during study and also recall the natural language mediator during the test of the item, then there is a high probability that the response will also be given. However, there is disagreement as to whether or not the natural language mediator plays any functional role in learning. The mediator could be merely an epiphenomenon which accompanies learning but reflects no process involved in the actual learning of the pair (Adams & Mcintyre, 1967;Underwood, 1972). This criticism can be made of almost any study utilizing natural language mediation. BeIlezza and Pop\awsky (I974) discussed the possibility that the one-word mediator given for a word pair in their experiment may not have been a mediating link between the two presented words, but merely a strong associate to one or both of the word...