To achieve perceptual alignment between a flashed target and a moving one, subjects typically require the flashed target to be aligned with a position that the moving target will only reach some time after the flash (the flash-lag effect). We examined how the magnitude of this misalignment changes near an abrupt change in velocity. The magnitude of the misalignment turns out to depend on the target's velocity after, rather than before, the flash. Thus, the misalignment cannot be caused by motion extrapolation. Neither can it be the inevitable consequence of a difference between the time it takes to process flashed and moving stimuli, because the magnitude of the misalignment is influenced by the extent to which subjects can anticipate the flash. We propose that it is the consequence of having to 'sample' the moving target's position in response to the flash.