144 Ss learned a list of 24 noun-noun pairs in which the image-arousing value, or concreteness, of stimulus and response members was varied in a factorial design. Different groups were given instructional sets to use imagery, verbal mediators, or rote repetition to link the members of pairs. A 4th group was a no-set control. Within each condition, different subgroups were stopped after 1, 2, or 3 trials and queried concerning their use of the different strategies to learn individual pairs. The findings were: (1) recall was facilitated by stimulus and response concreteness on all trials, and by imaginal and verbal mediation sets relative to both no-set and repetition conditions on the 1st 2 trials; (2) imaginal and verbal mediators were most frequently reported by Ss given the corresponding sets, and imagery reports increased generally over trials regardless of set; (3) reported use of imagery was also most affected by noun concreteness, the pattern over trials closely resembling that obtained for learning scores; and (4) repetition-set Ss reported frequent use of repetition on Trial 1, with a sharp decrease on subsequent trials. These findings are discussed in terms of the hypothesis that Ss abandon the use of ineffective associative strategies over trials and increasingly adopt mediation strategies appropriate to the meaning of the to-be-learned items. Imagery appeared to be a preferred type of mediator when at least 1 member of the pair was concrete and high in its capacity to evoke images. (PsycINFO Database Record (c) 2006 APA, all rights reserved)