Abstract : Two opposing views prevail on the visual control of actions. One supports predictive models of future states of the environment and the other the use of online visual information. The latter contemplates prediction only if online input is missing, otherwise it disregards past information. I will present different experiments based on both published and more recent unpublished data, in which people had to catch real and virtual balls. These experiments provide evidence that the distinction between these two control modes does not correspond to foundamentally different processes. In the these experiments the use of both predictive cues and late online visual information is revealed. In addition, in one of the experiments it is shown that early predictions and late visual information can be combined, when both are available, to improve the temporal precision in the interception task. This is consistent with past information being integrated with online vision rather than being discarded.