The "spacing effect" is the commonly observed phenomenon that memory for spaced repetitions is better than memory for massed repetitions. To further investigate the role of rehearsal in spacing effects, three experiments were conducted. With pure lists we found spacing effects in free recall when spacing intervals were relatively long (Experiments 1, 2 and 3), but not when spacing intervals were relatively short (Experiments 2 and 3). In contrast, with mixed lists spacing effects emerged at both short spacing intervals and long spacing intervals (Experiment 3). Additional analyses on the combined pure-list data revealed that the correlation between the primacy advantage and the spacing effect in Quadrants 2 through 4 was positive for all-massed lists and negative for all-spaced lists. This provides some first evidence for the zero-sum nature of the spacing effect in pure lists. The need to incorporate assumptions about rehearsal in theories of spacing is discussed. [Research for this article was supported in part by Erasmus University.]