We previously showed that the saccadic system could be adapted in a context-specific manner: two different adapted gains could be associated with two different context cues, with the gain state switched when the context state was switched. This was accomplished by alternating context/adaptation states several times over the course of an adaptation session, and assessing saccade gain in each context state before and after adaptation. One context cue we studied was vertical eye position; an adaptive gain increase was induced with the eyes up 10 degrees, and an adaptive gain decrease with the eyes down 10 degrees. This context cue was only partially effective: there was considerable undesired transfer of adaptation from the eyes-down condition (gain-decrease) to the eyes-up condition (gain-increase), with the result that there was little or no gain-increase adaptation. One explanation for this is that the two context/adaptation states, presented one after the other, interfered with each other. In the present study, we tested this hypothesis by interposing one-minute rest intervals between each alternation in context/adaptation state. The resulting context-specific adaptation is greatly improved (relative to the case when there are no rest intervals): both gain-increase and gain-decrease adaptations are stronger and occur more rapidly. This effect resembles that found in studies on the consolidation of motor learning, although such consolidation is believed to occur over much longer time spans (hours rather than minutes).