The short-term retention of minimal numbers of paired associates was studied in four experiments. Pairs of words were presented in a memory drum and tested for recall by presenting a stimulus after some interval no longer than 16 sec. during which Ss counted backward. Recall decreased as number of pairs increased, but a primacy effect complicated an interpretation in terms of competing responses. This primacy effect was present with either one or two presentations of a pair. When the stimulus situation at recall was extended to include the alternate pair of words, the primacy effect was reduced. Errors at recall were largely words from the set of pairs just presented.