Associative transfer in verbal paired-associate learning as a function of stimulus similarity 1 Associative transfer was explored in the A-B, A-C and A-B, A-Br paradigms under three levels of stimulus similarity using an A-B, C-D control comparison. Three groups, differing in degree of structural similarity, learned two mixed lists of CVC trigrams by the anticipation method. A total of 90 sixth grade Ss were used. The effect of paradigms was highly siAnificant, with both the A-B, A-C and A-B, A-Br paradigms demonstrating negative transfer. Some evidence was found for increasing negative transfer as stimulus similarity increased. The purpose of the present investigation was to assess the effects of stimulus similarity on transfer in the A-B, A-C and A-B, A-Br paradigms, both of which generally result in associative inteference when evaluated against an A-B, C-D control fOr nonspecific transfer (Besch & Reynolds,1958). A secondary purpose was to develop a technique for using nonsense syllables in a paired-aSSOCiate learning task with children. Stimulus similarity, defined in terms of common elements in CVC trigrams, has been shown to retard learning within a single list (Goss & Nodine, 1965, Ch. 5; Goss, Nodine, Gregory, Taub, & Kennedy, 1962). Kazura (1961), using an A-B, A-C paradigm, found a condition of high similarity to result in lowered performance when compared with that of low similarity, however, since a control for nonspecific transfer was not included in the design, the amount of interference due to increased similarity cannot be assessed. The present study, then, represents an attempt to determine the relative effects of stimulus similarity on transfer in both the A-C and A-Br paradigms with nonspecific transfer controlled. Method Ninety Ss, ranging in approximate age from 10 to 13 years, were selected from the sixth grade of a school in Chicopee, Massachusetts. A Hunter Card Master was used to present the stimuli which were displayed on white plastic cards for this purpose. The stimuli and responses were nonword CVCs with Archer (1960) association values from 84 to 94. stimulus members of high similarity shared two letters in common (e.g., NAL, JAL), those of medi!1Ill similarity had one letter in common (e.g., DEK, MES) and those of low similarity shared no elements (e.g., WIP, ROG). The response members were of low similarity with one or no letters in common. Similarity between the stimulus and response members of a pair was minimal. Within each level of similarity, manipulated between