對連合學習の機構

Abstract

In 1956, the author presented the following hypothesis concerning a process of paired-associates; “In paired-associate learning by anticipation method, subjects may learn, concurrently, to discriminate stimulus items (S), to acquire response items (R), and to pair S and assigned R in a list.” Paired-associate learning can not be completed till learning of these three modes is done perfectly (102, 112). This paper is attempted to present further evidences for the hypothesis from the more extensive point of view (Chap. I-V), and to clarify some functions implicitly referred by the author's previous reports (Chap. VI-IX). I. Function of S and R There had been two contradictory views about the nature of paired-associate learning ; One is Gibson's view, that “a major necessity of verbal learning is the establishment of discrimination among the (S) items...(32).” The another is Umemoto's view, that “the status of response is far more important to the reproduction of the materials learned ...(185).” On the other hand, it was found from a review of the literatures that the rate of paired-associate learning depended mainly upon the degree of intralist similarity of S items and upon the degree of meaningfulness (familiarity) of R items. Then, this contradiction was resolved by the present view, that the different dimensions of the materials from which those two views are derived, made such apparent conflict. II. Four dimensions of verbal material (meaningfulness, familiarity, emotionality, and intralist similarity) Meaningfulness was operationally defined in terms of the mean frequency of association made by subjects within a definite time, according to Noble's definition (124). Familiarity was defined as a psychological correlate for frequency of exposures of items to subjects (experienced or experimentally introduced). Highly correlated relationships among meaningfulness, familiarity, and emotionality were found in the literatures. The influence of these four dimensions of material on verbal learning was discussed. III. Discriminative learning and item-acquisitive learning Discriminative learning means to differentiate each item in connection with its assigned response, and item-acquisitive learning means to remember an item itself independent of its proper position or association in a list. The degree of the former is indicated by the measure of reconstruction method and the latter by the measure of free recall testing. It has been shown by a number of experimental results that as intralist similarity increases, discriminative learning is much inhibited and item-acquisition is somewhat facilitated, while with an increase in familiarity, item-acquisition increases significantly but discrimination insignificantly. IV. The role of S and R (a proof of the hypothesis) The hypothesis stated at beginning of this paper is regarded as logically substantiated from the findings in previous sections ; That is, from that the more similar intralist S items are, the more difficult is paired-associate learing, and the more similar items are, the more difficult is discriminative learning, it is induced that S items in paired-associates will be learned discriminatively. While, as the less familiar R item is, the more difficult is paired-associate learning, and the less familiar an item is, the more difficult is item-acquisitive learning, hence, we come to the conclusion that R item in paired-associates will be learned item-acquisitively. Further tests of the validity of the hypothesis was examined by means of a descriptive situational analysis and from the data of some transfer experiments. V. The another confirmation of the hypothesis was done on the basis of the measure of forward-backward recall (or recognition) gradients, with resect to the variables on meaningfulness of S, familiarity of S, intralist similarity of S, isolation of S.

VI. A hypothesis that S item is acquired incidentally was verified in terms of the comparison between the data of R-S recall or recognition and that of incidental learning, concerning the variables on meaningfulness, intralist similarty, isolation, method of testing, and degree of learning. After that, it was suggested that backward recall (or recognition) should not be regared as the resultant of incidental R-S associative learning from S-R learning, but as the resultant of incidental acquisitive learning of S item. VII. Temporal factor in association and directional inhibition hypothesis In the case of learning of nonsense syllable pairs, paired-associate might be formed as an unsymmetrical association which had 'polarity' from S to R, and then the strength of polarity might be rapidly diminished after the termination of learning. Reminiscence that appeared on backward recall within 5-30 minutes after reaching a criterion of learning was interpreted as a recovery from directional inhibition (Id), which was a hypothetical construct with the same explanatory mechanism as Hull's Ir. Polarity factor in association was supported with respect to method of learning (successive vs. simultaneous presentations), method of testing (recall vs. recognition), and the procedure of modified free recall. Directional inhibition hypothesis was examined by a function of retention time, of distribution of practice, of modified free recall, and of delayed memory span. VIII. (1) Though a S-R connection should be formed on one presentation of S-R together, an overt correct anticipation might not always appear on the next trial, because the S-R connection suffer interference or inhibition from the following pairs to be learned within a list. Consequently, learning of paired-associates would progress gradually as a function of the number of presentations. (2) A brief review have been made about some intra-pair variables, such as associative prepotency, meaningfulness, mediated-association, similarity between S and R, S-R homogeneity, S-R compatibility, S-R organization, clustering, and proximity. IX. After an introduction of some interpretations concerning a process of paired-associate learning by anticipation method suggested in several experimental reports, the present author's theory was summarized as a conclusion of the above discussions (see Fig. 1 and 3). X. As a summary, the relationships among the dimensions of verbal materials, the hypothetical modes of learning, and the measures of learning were shown in Fig. 4.