効果の波及に関する研究I

剌戟語の質について

抄録

It is now generally accepted that guessing sequences are sufficient to account for the ‘spread of effect’ phenomenon. This is based on experimental facts such as : (1) the gradient of number repetitions similar to the Thorndike effect was found after chance repetitions which were not rewarded (Sheffield and Jenkins), (2) the gradient disappeared when stimulus positions were changed during trials but it was found if only step positions were considered, stimulus positions being ignored (Zirkle) and (3) the effect did not appear or it was greatly reduced after non-repeated rewarded responses while it was clearly shown after repeated rewarded responses (Jenkins and Sheffield).
Unfortunately none of these findings were strong enough to refute the effect theory of the Thorndike effect on the basis of the facts that (1) this orthodox Thorndike effect and number guessings may not be principally the same, (2) the change of stimulus positions during trials does necessarily interfere with the ‘spread of effect’and would reduced the gradielt and (3) non-repeated rewarded responses may not be actually rewarded simply because they are not repeated. The purpose of this experiment was to test the theory of guessing habit when these factors were controlled.
Two groups, 50 undergraduate students each, were used as Ss. Group M was presented with two lists 24 two-syllabled Japanese nouns, for three seconds per each word. Ss were required to read them aloud and to respond to them with a number from one through 10. An orthodox instruction for the ‘spread of effect’ experiment was utilized, such as, “If you call out the right Numbers”, I shall say ‘right’and if you call out any other numbers I shall say ‘wrong’. Your task is to try to repeat the numbers called ‘right’ and to avoid the numbers called ‘wrong’ when the stimulus list is presented again .......
List 1 was shown three times in succession. Responses to the sixth and 17th stimuli were called ‘right’ and those to any other responses were called ‘wrong’ irrespective of the numbers given. This prearranged pattern was sometimes modified but was consistent for any one S. The instruction was changed after List 2 was read twice. Ss were asked to recall the numbers they gave last time and to say whether the number was called ‘right’ or ‘wrong’. Stimulus words for the two lists were interchanged from S to S.
Group N was the same as Group M except that the stimuli consisted of nonsense words rather than meaningful nouns.
Our experimental hypothesis were that (1) the mean number of correctly recalled responses would be greater for Group M than for the other group because of the very nature of the stimulus characteristics and (2) the gradient of number repetitions would be, on the other hand, greater for Group N as compared with Group M because the S-R connection would interfere with the guessing habit.
The results confirmed the predictions. The negative relation between correctly recalled responses and response repetitions strongly suggested that the Thorndike effect was very different from the ordinary S-R connections. Although the guessing sequence hypothesis would predict a high positive correlation between the repetition of rewarded responses and that of first errors, the obtained r was not large enough (-.244) to confirm the prediction for Group M, and the same r was almost zero (-.011) for Group N. This finding, however, was not very conclusive but if strengthened by further experimentation, it would require some modification of the theory of guessing sequences. It should be added that only the aftergradients were observed for both groups.