The question has been raised whether an associative mechanism is responsible for the persistence of the McCollough effect. Since current estimates of its rate of decay are derived from procedures in which subjects are repeatedly tested, it was hypothesized that the measured effects might be attenuated by the testing process. Accordingly, a comparison was made between repeated testing and time-elapse testing. A conventional group of 16 subjects had repeated testing at 0, 8, 24, 56, and 120 hr. after induction. Five other groups of 16 were run, each at one of the time delays, with no intervening tests; an additional measure was taken where appropriate at 120 hr. A magenta-green nulling procedure was used to assess the aftereffect. The repeated-test group showed a linear decrease of effect against the stated delays, reaching zero at 120 hr. In contrast, the time-elapse groups showed little decline up to 120 hr. Those groups retested at 120 hr. showed declines due to prior testing. When four more groups were subsequently tested at intervals up to 2,040 hr., the effect remained at better than half strength.