The hypothesis that free recall curves reflecting the effects of serial position, presentation time, and delay of recall are attributable to subjects' pattern of rehearsal was explored in four experiments. In Experiment I, unrestricted overt rehearsal during list learning reveals that a list item's rehearsal time and item retention interval (the interval between an item's last rehearsal and the signal to begin recall) produced a multiple correlation of .86 with probability of recall. In Experiment II, subjects' spontaneous pattern of rehearsal was disrupted with instructions to either rehearse items equally, or else to selectively rehearse beginning, middle, or terminal items. Although free recall curves were grossly altered, the multiple correlation remained high (.89). Experiments III and IV show that better-than-predicted recall of beginning items in certain conditions in the first two experiments is largely attributable to subjects' overt rehearsal underestimating concurrent covert rehearsal of beginning items.