Two experiments using a temporal occlusion paradigm (the first with expert and novice participants and the second with participants of intermediate skill) were conducted to examine the capability of tennis players to predict the direction of an opponent's service in situ. In both experiments two different response conditions, reflecting differing degrees of perception-action coupling, were employed. In a coupled condition players were required to make a movement-based response identical to that which they would use to hit a return of service in a game situation, whereas in an uncoupled condition a verbal prediction of service direction was required. Experiment 1 provided clear evidence of superior prediction accuracy under the coupled response condition when ball flight was available, plus some limited evidence to suggest that superior prediction accuracy under uncoupled response conditions might hold true if only advance (pre-contact) information was available. Experiment 2 showed the former finding to be a robust one, but was unable to reveal any support for the latter. Experiment 1 also revealed that expert superiority is more apparent for predictions made under natural (coupled) than uncoupled response-mode conditions. Collectively, these findings suggest that different perceptual processes may be in operation in anticipatory tasks which depend on skill level, the type of information presented, and degree of perception-action coupling inherent in the task requirements.