Redundant-stimulus paired associates (PAs) consisting of word-trigram stimuli and digit responses (e.g., CAT + Bwo:0) were practiced to different degrees of PA learning (either 1, 2, 3, 5, 1 Perfect, or 5 Perfect trials) by 12 subgroups of college students. Six of the subgroups were tested for intercue associations and 6 were tested for cue selection. The 5s were observed to rely almost exclusively upon the word cues at all stages of learning. Prior to overlearning (5 Perfect trials), there was no evidence for the existence of either trigram:response or trigramtword associations. Following overlearning, however, a parallel increase was observed in the availability of both trigram:response and trigram: word associations. It was concluded that the hypothesis of indirect association of lower meaningfulness (M) cues and responses (i.e., trigram:word:response associations) was sufficient to "explian" the parallel rise in availability of apparent trigram: response and trigram:word associations.