効果の波及についての研究 (III)

3つの対立した仮説の実験的吟味

抄録

For the purpose of examination of the three conflicting hypotheses with regard to the spread of reward-effect-the Thorndike's, Zirkle's, and Jenkins-and-Sheffield's “guessing-sequence” hypothesis-the following experiment was carried out.
Ss : 32 College students, divided equally into 4 groups A, B, C and D.
Stimulus items : 18 Japanese nonsense, syllables.
Response-materials : 9 simple figures (See Fig. 2).
Number of trials : 9 times.
Time-interval of responses : 3 seconds.
The responses of the Ss of A and B groups were given reward-and-punishments respectively according to the ordinary experimental procedures. The Ss of C and D groups were requested, instead of rewarding, to keep using those response-materials which they first placed on some definite key stimulus items to the final trial (the“fixed” response), and, at the same time, in lieu of punishment, not of use the same response-materials twice to other many stimulus items (the “liquid” response). The Ss of Group C, in all trials, were shown a card, on which all of the 9 response-materials were arranged in an identical manner, and to each of the rest of the groups were shown a card of the response-materials on each trial on which materials were arranged in different ways (See Table 1). The key stimulus items of groups A, C and D were isolated from other stimulus items in terms of style of printed letters and of color.
The principal results were as follows :
1. Percentage of repetition of Group A's correct responses was markedly more dominant than that of Group B's, but neither of the groups indicated any gradient of repetition in the incorrect responses neighboring the correct ones (See Table 3).
2. The “fixed” responses of Groups C and D were fully repeated, but Group C alone showed a gradient of repetition in the “liquid” responses behind the “fixed”ones (See Table 3).
It is rightfully argued through these findings that, even though guessing of the Ss are likely affected by the spatial sequencehabits during the course of each trial, no repetition-gradient appeared in the incorrect or “liquid” responses under conditions in which the spatial sequence habits do not affect the guessing of the Ss during the course of any two continuous trials as a whole, though correct or “fixed” responses were fully repeated.
On the basis of the results of the present experiment, findings of the many investigators have been carefully re-examined. The repetition-gradient of incorrect responses following the correct ones known as the spread of reward-effect appears neither to depend on the spread of reward-effect nor on the spread of isolation-effect of the correct responses. The gradient is the function of the degree of repetition of the correct responses and the degree to which the guessings are affected by sequence-habits (both in time and in space).
Thus far, it is reasonably concluded that the “guessing-sequence” hypothesis is the most dependable.