Under a recall model in which presentations and rehearsals are treated as equivalent encoding events, we investigated whether rehearsal efficiency differences explain the effects of word frequency and bilingual proficiency on the temporal dynamics of rehearsal and free recall. Experiments 1 and 3 were conducted with monolingual English speakers, and Experiments 2 and 4 were conducted with Spanish-English bilinguals with matched age, education, and socioeconomic status. In Experiments 1 and 2, lower word frequency, lower proficiency, and bilingualism were associated with less accurate free recall of items from early serial positions, beginning recall with items from later serial positions, and making fewer transitions to items from later or adjacent serial positions. These effects were replicated and rehearsal-based explanations were validated in Experiments 3 and 4 using a rehearse-aloud protocol. With lower frequency words or lower language proficiency, rehearsal was less efficient with fewer rehearsals between item presentations. As a result, items from early serial positions had fewer rehearsals that stopped earlier in the study sequence, less spacing between repeated rehearsals, and fewer transitions to items from later or adjacent serial positions. Rehearsal-contingent analyses revealed that these rehearsal patterns were associated with less accurate recall, beginning recall with items from later serial positions, and consistent transition patterns from rehearsal to recall. These patterns support a model in which presentations and rehearsals are treated as equivalent encoding events and the effects of word frequency and language proficiency on recall accuracy are mediated by less efficient rehearsal.