There is preliminary evidence for interference between subsequently encoded velocities. We explored this effect using a motion extrapolation paradigm. In Experiments 1 and 2, moving targets disappeared behind an occluder. Participants responded at the time when they thought the target should reappear. Participants responded as if the current velocity was faster after a previous trial with a fast velocity and vice versa. In Experiment 3 the targets reappeared either on-time, early or late. A greater proportion of early trials were reported 'correct' following a fast previous trial. These experiments indicate that blending occurs between successively encoded velocity representations.