In this study, we explored the influence of an irrelevant translational event on the automatic capture of attention to one of two superimposed surfaces defined by transparent motion. The results showed that an irrelevant translation on one surface did not automatically capture the subjects' attention if the attentional resources have been endogenously allocated on the other surface. Moreover, the reduction in the motion-onset component of the event-related potential observed in trials where the irrelevant event affected the uncued surface supports the existence of a top-down control of early sensorial processing in this paradigm. This study provides further evidence of the interaction of stimulus-driven and goal-directed mechanisms in the control of visual attention.