It is a well-known fact that reproduction is easier when it proceeds along the lines of original learning than when it reversely regresses the line. So we reproduce “d, être” easier from “raison” h than “raison” from “d'être” (1). But that is not exclusively due to the direction which accords with the line of habit, as ordinarily presumed. An example will show how reverse is also true:
In our first experiment, 2 series of 10 pairs of alphabet nonsense word type materials were offered to 10 student subjects, under the condition of an “anticipating right associates method.” The repetitions were 4, at a speed of 2sec. per item. Half of the subjects reproduced the first series progrssively, and then the second series in the reverse order. The other half worked vice versa. The average number of pairs reproduced were 5.3 for progression, and 6.5 for regression, with a reliability of difference expressed in Student's t=4.47, . 01>p>.001.
Now let us analyse this experimental situation. We found there were 2 conditions which determine the facility of reproduction. One is the alleged “direction” or status of the items in the S→R sequence, the other is the facility grade inherent in the material. Concerning this material factor, Kuraisi (2) has shown that reproduction was easier when it started from a member of poorly organized knowledge region, than from a member of a well organized one. So we may infer that this factor was stronger than the direction factor in our example mentioned above, for alphabet letters can be considered easier to memorize than nonsense words.
But how about the role which these 2 factors share with each other? In several experiments excecuted by Kuraisi only 2 types of series, such as A→B and B→A type, were compared. In this constellation, however, both conditions direction and material change at the same time. Hence the problem is insoluble. To be able to determine the role of these conditions, there must be a third medium. For this reason we resorted to the follwing experimental device:
Four series of 10 pairs in different mateial construction were prepared.
I. nonsense word→word (difficult→middle)
II. alphabet→word (easy→middle)
III. word→nonsense word (middle→difficult)
IV. word→alphabet (middle→easy)
Ten subjects learned these 4 series in different order. The conditions were equal to thoseof the above experiment, except that they learned up to complete anticipation, of which the number of repetitions was computed. The table 6 shows this value.
From the table we see clearly that the difference of easy and difficult materials is more markedly accentuated in the response status (III/IV) than in the stimulus status (I/II) of items. So it is to be concluded that the status of response is far more important to the reproduction of the materails learned by method of righr associates. But it might be suspected that the oral responses made by Ss in the course of the original anticipating learning was the cause of the better reproduction rate of response items. To answer this question, we presented to Ss the paired two items simultaneously. Nonetheless, we found here also that the effect of response items was greater than that of stimulus items.
