Use of only a portion of the presented PA stimulus as a recall cue is now an established phenomenon (Cohen & Musgrave, 1964). Stimulus selection appears to occur whenever stimuli are poorly integrated units, e.g., GYX, which cannot easily be responded to as a whole (Leicht & Kausler, 1965). Selection circumvents learning the entire stimulus compound. Nevertheless, time spent searching the compound for a suitable recall cue should slow overall learning rate. The assumption that increasing the number of components in the stimulus slows learning by increasing selection time was examined by comparing learning of PA lists containing stimuli comprised of either two or three words. In addition, component words were either conceptually related, e.g., OAK, ELM, or were unrelated, e.g., OAK, BOY. Stimuli whose components are conceptually related tend to elicit a single category-name associate which may mediate as a recall cue, the selection process thereby bypassed. Hence, slower learning for three-than for two-word stimuli was expected only when component words were unrelated.The extent to which differential stimulus selection accounted for an effect of type of stimulus compound on rate of learning was directly assessed through a test of stimulus selection which requires that S produce the response term to components of the stimulus presented individually. In a selection test, S is presumably able to produce the response term only to component(s) which had served as recall cues in acquisition (Underwood, Ham, & Ekstrand, 1962). METHOD List stimuli were either two conceptually related words (List 2C), three conceptually related words (List 3C), two unrelated words (List 2N), or three unrelated words (List 3N). Stimulus components were said to be conceptually related if they elicited the same category-name associate in a word-association task (Cohen, Bousfield, & Whitmarsh, 1957). The same set of single-word responses was used in each list. Each of the four 14-pair lists was learned by a different group of 15 undergraduates.Lists were presented by the alternate-study-and-recall method until S achieved one perfect recall. To minimize serial learning, pair order was varied from trial to trial. A 2-sec rate applied for both study and recall portions of trials. The test of stimulus selection was given immediately upon S's achieving criterion and consisted of presenting components of the stimuli individually for 2 sec under instructions to give the response that had gone with the entire compound. For both acquisition and selection phases, materials were presented by a Carousal projector with attached interval timer.RESULTS AND DISCUSSION Mean trials to criterion (SDs in parentheses) was 5.27 (1.84), 6.73 (2.21), 4.73 (1.53), and 5.00 (2.19) for 2N, 3N, Psychon. Sci., 1968, Vol. 13 (6) 2C, and 3C conditions, respectively. The effect of relatedness of components was significant, F(1,56) = 4.68, p < .05, suggesting that selection was minimized in C conditions through use of the category name as a recall cue. The effect of number o...