対連合学習の研究-1-

概要

論文の詳細を見る
This paper is the first report on the Paired-associated learning process which are investigated by the method of forward-backward recall gradient (FB-RclG). This term is defined operationally as difference between forward recalls (S→R) and backward recalls (R→S). after paired associated learning. The present study intends to examine some variables of FB-RclG. Exper. I. FB-RclG as a function of paired materials. Materials are three in kind. Difficult materials, "N" (two letters, non-sense syllables with about 20% association values). Medium materials, "M" (two letters, meaningful words). Easy materials, "L" (one letter of Alphabet). Conditions of presentation are (1) N-N (S-diff., R-med.), (2) M-M (both med.), (3) L-M (S-easy, R-med.), (4) M-N (S-med., R-diff.), and (5) M-L (S-med., R-easy). Methods : The anticipation method is used. Forward and backward recall tests are performed at about 30 sec. after one perfect trial. Results show that the difference between cond. 1 and 3 is smaller than that between 4 and 5, as regard to number of trials. This fact supports Umemoto's finding-the difficulties of paired associated learning are determined largely by R factor rather than S factor (18). FB-RclG of cond. 1 is greater than that of 4, as well as 5greater than 3. When S is more difficult than R, FB-RclG is greater than when S easier than R. It is confirmed that FB-RclG is a function of materials in pairs. Exper. II FB-RclG as both function of degree of learning and interpolated recall effects. Materials are N-M, 9 pairs in a list. Methods are the same as Exper. I, except that recall tests are interpolated immediately after 1 presentation and after criteria of 3/9, 6/9, 1 perfect, and 3 and 5 successive perfect learning. Results show that FB-RclG increases gradually in early stage, and reaches its maximum in middle stage and then decreases (Fig. 1). The interpolated recall test during learning makes the amount of backward recalls increasing (but not statistically significant), while its effects to forward recalls are negligible. Exper. III FB-RclG as a function of difficulty of learning. Conditions are 3, 5, 7, and 9 pairs of N-N in one list. Except recall criteria, other conditions are the same as Exper. I, Recall tests are performed 3 times after about 1/2, 1 perfect, and 3 successive perfect learning for each list. Results show that EB-RclG increases as the length of list increases, when lists are overlearned. Exper. IVa, IVb FB-RcLG as a function of pair presentation methods. Methods : In stead of anticipation method, simultaneous pair presentation method is used. And after each trial, forward and backward recalls are tested. (Fig. 3). Results show that even under the situation which differs from anticipation learning (S-recognition and R-recall), the existence of FB-RclG is found (Fig. 4, 5). The Vincent curves show that FB-RclG is not unique phenomenon characteristic of anticipation method. Exper. Va, Vb Purpose is the same as Exper. IVa, IVb. Learning method is simultaneous pair presentation, Recall tests are interpolated not after each trial, but after 6, 12, and 18 presentations. Each list contains 3 kinds of presentation set, i.e. N-M, M-N, and N_??_M (alternative presentation). Results are consistent with the writer's previous experiments (Fig. 6). These facts suggest that the first member of pair may be called S, and the second R. Conclusion The differences in FB-RclG may be explained by the following three hypotheses : a) R is emphasized by the mode of anticipation method, in which S is only discriminative and R acquisitive. b) Based on the facts of Exper. IVa-Vb, one may assume a temporal (porality) factor in learning of paired association.
公益社団法人日本心理学会の論文

著者

森川弥寿雄
京都大学文学部

対連合学習の研究-1-

スポンサーリンク

概要

著者

関連論文

スポンサーリンク