Repetition When Learned Paired-Associate Items are Replaced

Abstract

Rock (1957) reported no difference in number of trials to criterion between a control group which learned a constant list of paired associates in the usual fashion and an experimental group for which new pairs replaced those missed on any trial. Several reports of replications have since appeared (Clark, et al., 1960; Rock & Heimer, 1959; Williams, 1961). Rock and others have concluded that associations are formed on an all-or-none basis, and interpret their results as weakening the traditional incremental conception of the formation of associations. The current study was designed to show that the all-or-none position is roo extreme. Subthreshold associative bonds present before the first correct associacion need special conditions to manifest themselves precisely because they are subthreshold, which is to say, weak. In the experimenc reported here, the Rock procedure was reversed: new pairs replaced those which had been covect on any trial. This experiment used easy pairs, and a work rather than a performance criterion. Similar methods have been reported by Battig ( 1962 ) and Williams ( 1962 ). Battig used 12 shape-number pairs and compared intra-subject differences between six pairs which were always replaced after a single trial and six pairs replaced only when they had been responded to correctly on the preceding trial. Williams analyzed data for 10 Ss using 12 letter-number pairs in a n experimenc where the items responded to correctly were replaced by new items on the subsequent trial. The results of both studies support an incremental theory of associative learning. The current study employs an easier list of materials to be learned, more Ss, and attempts to eliminate response learning as a confounding factor by familiarizing Ss with the stimulus and response items. The hypothesis examined was that the probability of acquiring an item increases with the number of presentations of chat item before acquisition. We used dropping of correct pairs as our special condition because we hoped this would eliminate two possible hindrances to the detection of weak subthreshold associative bonds. The first is the competition with the learning of unlearned items which results from the attempt to retain already learned items. And the second, which is related to the first, is the usual negatively accelerated learning curve.