「効果の波及」に関する研究: VI

Abstract

The “spread” hypothesis proposed by Thorndike to explain the “spread of effect”, states that the influence of a rewarding state of affairs acts not only on the connection to which it belongs but on adjacent connections around the rewarded connection as a function of distance from reward. In an orthodox Thorndikian design of spread of effect, the effect is found to act to strengthen punished connections or more precisely to repeat error responses in the neighborhood of the rewarded one. However, just the opposite result would be predicted strictly following the Thorndike hypothesis, in that if his hypothesis is true, S-wrong R connections around reward will be strengthened and thus the “wrong” R's will not be repeated on the following trial in comparison with those away from reward.A clear cut test of the “spread” hypothesis will be made if S-R connections near reward are not punished and number of repetitions of correct responses rather than errors is calculated. Several such studies were attempted with rats (5, 11, 12, 13, 15) and one with human (6) although their own objectives were different from ours. However, these works have been criticized because of a number of uncontrolled and/ or confounding factors.In the present experiment, 12 pairs (S-R's) of two syllables Japanese nonsense letters were presented individually to S every three seconds. Only one of the 12 such pairs which is called the key item, was written in red. The experimental Ss were required to the key item with special attention among the rest items. A recall test was given immediately after presentation of the 12 paired-associates but the ordering of the stimulus letters in recall test was randomly changed from trial to trial in order to rule out response-response learning. Five recall tests were administered per S. The control Ss with 12 homogenious paired-associates, i. e., without the red item, were used to control serial position effects in the experimental results. As the key item was expected to recall better and thus to be “rewarded” more often that the rest, a Thorndikian gradient in percent recalls should appear if the “spread” hypothesis is correct. The results shown in terms of the group difference (Table 1, Fig. 1 in Text) failed to prove the prediction but the correct recall score appeared to be greater for the response immediately preceding the key item than immediately following it. The trend was enhanced not only with trials (Table 3, Fig. 2) but also when the key item was correctly recalled (Table 4, Fig. 3). The results were explained in terms of increased motivation resulting from the anticipation of the key item with respect to the better recall of the first-before item and the decreased motivation as well as the rehearsal-distraction effect suggested by Jenkins & Sheffield, with respect to the first-after response.Similar results (Table 5) were obtained by the group experiments (Groups AE & AC) where the almost identical procedures as in the previous individual test were used but no observable effects were obtained by Groups BE & BC in which the ordering of the stimulus items in recall test was not varied.