We have previously shown that Os in multiple object tracking (MOT) experiments can successfully track objects that disappear for up to 500 ms (Alvarez et al. in press; see also Keane & Pylyshyn VSS 04). Last year, we demonstrated that Os recover those tracked objects based on some memory representation, rather than tracking amodally during the gap. (Horowitz et al. VSS 04). How quickly can targets be recovered after they reappear? Are targets recovered one by one, or simultaneously? Here we employed a version of the MOT paradigm in which Os indicated whether or not a probed object was a target with a speeded response. Eight Os tracked 4 of 8 moving gray disks. After 1000 to 3000 ms of tracking, the disks disappeared for a fixed gap duration, then reappeared at updated positions as if they had continued to move during the gap. After a stimulus onset asynchrony (SOA) of 0, 80, 160, 320, or 1280 ms, one of the disks (the probe disk) turned red until response. The gap duration was either 0 ("no-gap" condition) or 133 ("gap") ms. At an SOA of 0 ms, response times (RTs) to target disks were elevated in the gap condition relative to the no-gap condition; however, at SOAs of 80 ms and greater, there was little or no difference between RTs on the gap and no-gap conditions. In a second experiment (8 Os) that included shorter SOAs, the difference in RT between gap and no-gap conditions disappeared by 40 ms. Os can apparently recover targets within 40 ms after reappearance. This rapid timecourse argues against a target-by-target, serial recovery process. Instead, Os may be matching a single template of expected target locations to the display.