Performance in 2 versions of a computer-animated task was compared. Participants either indicated the time of arrival of a target that rolled off a horizontal surface and fell--hidden from view--onto a landing point (production task) or judged flight time on a rating scale (judgment task). As predicted, performance was significantly better in the production task (Experiment 1), in which imagery of object motion probably replaced reasoning processes. Participants who exhibited eye movements suggesting mental tracking performed particularly well in the production task (Experimenl 2). There was, however, no decrement in performance when participants were asked to fixate the point where the target disappeared. For motion duration estimations, eye movements seem to be only a by-product of mental tracking.