Illusory Causal Crescents
This page contains demonstrations of the the illusory causal crescents phenomena, which are discussed in the following paper:
Scholl, B. J., & Nakayama, K. (2004). Illusory causal crescents: Misperceived spatial relations due to perceived causality. Perception, 33(4), 455-469.
These demonstrations are provided as Quicktime movies, which can be downloaded or viewed directly in most web-browsers. These movies are a bit large and choppy, and the spatiotemporal parameters are slightly different than those used in the experiments, but they should be sufficient to illustrate the basic effect. Note that you should set each movie to 'loop' in order to best experience the effects. In each case, you should judge the amount of overlap between the two items in the upper event. (Our observers made this judgement directly in repeating displays, by manipulating the width of a separate 'crescent' in the display, to indicate the amount of the stationary disc which is never covered by the moving disc.) In all of these demos, this value is actually 100% (i.e. complete overlap), though it is not always perceived as such. (Note that in the paper reporting this illusion, we also tested several other amounts of actual overlap.)  
This study grew out of a larger exploration (reported in a separate paper) of the rules used by the visual system to infer the existence of a causal relation in simple collision events. In particular, it grew out of our studies of causal capture, a phenomenon wherein the perceived causal status of an ambiguous event can be affected by contextual information from other nearby objects and events. If you'd like you can first view demonstrations of the basic 'causal capture' effect.
The critical 100%-overlap stimulus
All of these demonstrations employ the same test event, in which the two items overlap completely before the second item begins moving. This stimulus is shown in this movie without any contextual information. Most observers perceive this as a completely non-causal 'pass' (see the causal capture demonstrations), and here they are often reasonably accurate at judging that the two items overlap completely.
Illusory causal crescents
The upper event in this movie is identical to the event in the previous movie (i.e. the two objects still overlap completely during their motion). Here, however, observers often underestimate this amount of overlap -- perceiving a thin 'crescent' of one item which is never occluded by the other moving item. As discussed in our paper, we think that this effect is driven by the fact that adding this context causes the upper test event to be perceived in causal terms, as a collision or 'launch', instead of a noncausal 'pass'. Note that this effect is weakest when you are directly fixating the area to be judged; moving the upper event into the periphery results for most observers in a stronger effect -- i.e. in wider illusory crescents.
Effects of duration
This movie demonstrates that very little of the context event is necessary to induce illusory causal crescents, so long as the actual 'impact' is visible. Here we still observe systematic underestimates of the amount of overlap in the top event, even when only a brief moment of the context event is displayed. Again, the illusory crescents will seem wider if viewed in the periphery. (Further, this movie indicates that the effect works even with 'underlaps' instead of overlaps, where the stationary item progressively occludes the moving item, instead of vice versa. In our paper we discuss illusory causal crescents obtained from both types of intersections.)
Effects of temporal asynchrony
In our previous work we observed that 'causal capture' is destroyed by introducing temporal asynchrony between the two events. As such, we expected that this would also eliminate the illusory causal crescents. This was indeed what was observed, as can be seen in this movie.
The illusion embodied in these demonstrations implies that the perception of causality does not proceed completely independently of other visual processes, but can affect the perception of other spatial properties.