効果の波及に関する研究 (III)

前勾配と数の推測習慣

抄録

This research was intended to explain in terms of the guessing-sequence hypothesis (a) the fore-gradient of the ‘spread of effect’ and its relative weakness compared with the after-gradient and (b) the higher level of repetition in the Thorndike situation than that in the non-1earning situation.
Experiment I was designed to find gradients around chance repeats in a number-guessing situation. Ss wrote a number one through 10 in response to a 3 sec. signal without given any rationale other than an experiment in psychology. No systematic responses such as 1, 1, 1, ..... or 1, 2, 3, ..... were allowed. Two series of 240 such responses (24 rows and 10 columns) were obtained. Numbers of repetitions in the three positions around chance repeats on 6th, 13th and 20th rows were computed and the percentages of repetition are shown in Table1 (see Text). A clear fore-gradient as well as an ordinary after-gradient obtained, can be explained by number-guessing habits. If a chance repeat was, say, 3-3 and if an S had a tendency to say 2 before 3 or 3 after 2, a repetition of 2-2 would be expected before 3-3. The fore-gradient can be explained by the same principle as the after-gradient. No supra-chance stimulus-number connections may be hypothesized as suggested by Jenkins and Cunningham.
In Experiment II, two numers in every column were presented by E after an extra 3 sec. pause in order to avoid any voluntary number responses after an S was given numbers. Only an after-gradient was obtained around pairs of the same numbers specified by E (see Table 3). The lack of a fore-gradient was assumed to be due to interruption of S's guessing sequences by being given specified numbers by E. This was somewhat comparable to the Thorndike situation in which S's guessing sequence would be discontinued when a rewarded word was presented and he responded to it with the number he knew correct. A higher after-gradient than that in Exp. I was interpreted to be caused by the fact that guessing sequences were strengthened after the numbers given by E, because these numbers were isolated.
The same effect of isolation was obtained in Experiment III with only extra 3 sec. pauses and no numbers given by E (see Table 5). Experiment IV was fundamentally a repetition of Exp. II without extra pauses before specified numbers. S's were carefully instructed to write down a number given by E immediately after it was heard. The results were similar to those of Exp. II, indicating an extra pause was not essential to the increase of repetition level (Table 8). In the same experiment, gradients of repetition around pairs of the same numbers, one member of which was given by E and the other by S himself were computed. A lower after-gradient appeared (Table 10) suggesting that even though guessing habits after specified numbers were strong, those after numbers chosen by S were weak.
Experiment V was different from Exp. II only in that two responses in every column were called ‘immportant’ and no numbers were specified by E. A weak gradient preceding chance repeats which were called ‘important’ was shown. This was because S's guessing sequence was not discontinued. On the other hand, a strong after-gradient appeared (Table 11) and was explained that E's calling ‘important’ isolated the numbers and thus guessing habits the numbers and thus guessing habits thereafter would be strengthened.
In conclusion, the fore-gradient can be explained by the guessing-sequence hypothesis and the lack of it in most of the previous studies was caused by interruption of guessing sequences by S's responding to rewarded numbers with specified numbers he learned on previous trials. The higher level of repetition after rewarded repeats than after chance repeats was attributed to the fact guessing sequence were strengthened after isolated numbers and rewarded responses were isolated as suggested by Zirkle.