We examined the time course of preview search, using stimuli that were defined by color, but not by luminance changes. We demonstrate that, under these conditions, search performance in a preview condition improved selectively over time, relative to a baseline condition in which all the items appeared together. The data confirm earlier reports from Humphreys, Kyllinsbæk, et al. (2004) and Watson and Humphreys (1997), who used luminance-defined stimuli and showed a long time course to preview search. The data contradict Donk and Verburg (2004), who argued that the preview benefit was instantaneous but did not include baseline conditions with which to test for any influence of distractors equivalent to the old items in preview search, even under nonpreview conditions. The data support the proposal that the prioritization of new items in preview search is a time-consuming business.