The purpose of the present experiments is to confirm the hypothesis that the stimulus (S) in paired-associate learning is learned mainly discriminative as the cue for the response (R), and R is acquired as the actual goal to be responded. As meaningfulness (defined in terms of association values) and familiarity (defined in terms of the number of previous presentations either experimentally or experientially) are expected to have a greater effect upon the difficulty of item-acquisition than upon the difficulty of item-discrimination, it is predicted that the number of trials to a learning criterion should be a function of meaningfulness or familiarity of R, while the amount of backward recalls (R→S) and of forward-backward recall gradients (FB-RclG) (defined as the difference between forward recalls (S→R) and backward recalls), should depend on the degree of meaningfulness or familiarity of S. More precisely, it is predicted that the lower the degree of meaningfulness or familiarity of R members are, the more the trials to criterion are required, and on the other hand, the lower the degree of meaningfulness or familiarity of S members are, the fewer the backward recall become, and the more the FB-RclGs become.In Experiment I, each list which consisted of nine pairs of N-N, N-M, M-N, or M-M (N: nonsense syllables, M: meaningful words) was learned by the anticipation method to a criterion of three successive errorless trials. Immediately after either 6 out of 9 correct responses, or one errorless, or 3 succcessive errorless trials on each list, forward and backward recall tests were performed. In Experiment II, the procedures were the same as those of Expreiment I, except that familiarized N (F) which was presented in the pre-learning lists was used in place of M. Experiment III, in which subjects were junior high school students (13-15 years old) and slightly different procedures were used, was performed in order to obtain further evidence for the effects of familiarity found in Experiment II. Our procedures of familiarization using paired-associate method (Exp. II) or unaided reproduction method (Exp. III), might be somewhat similar to Hovland & Kurtz's which seems to emphasize item-acquisition as well as item-predifferentiation.The results were summarized as follows: (1) The difficulty of learning expressed either in terms of trials (Exp. I), or of errors (Exp. II) until learning criteria, or of the number of items correctly recalled after 2 presentations (Exp. III) was generally determined by meaningfulness or familiarity of R. The change in S also yielded significant differences in some measures, but yet their differences were very small as compared with those of R change under the same conditions, except in Experiment II (see Figs. 1, 2, 4, and 6, and Tables 1, 2, 3, and 4). (2) The forward recalls which were measured using three learning criteria were N-N<M-N≅N-M<M-M in Experiment I, but these differences were not obtained in Experiment II. (3) The backward recalls were a function of meaningfulness and familiarity of S, and the lists in which S consisted of either N or U (unfamiliarized N) were consistently inferior to the lists consisting of either M or F withregard to backward recalls. (4) The FB-RclGs were also a function of meaningfulness and familiarity of S (see Figs. 3, 5, and 6, and Tables 1, 2, 3, and 4).These results were interpreted as confirming the hypothesis presented above. All the results on backward recalls suggest that S items are incidentally acquired in paired-associate learning using the anticipation method. Some discrepancies shown between the results of Experiments I and II, in terms of the number of trials, and of forward recalls under corresponding conditions of M and F, might suggest the different functions between meaningfulness and familiarity in learning. It was discussed that these discrepancies might depend upon an effect of mediating