Several recent studies have shown that attentional capture is not an automatic process. For example, abrupt peripheral onsets do not affect the processing of targets presented subsequently at that location when participants have to concurrently perform a perceptually demanding task elsewhere. This result leaves open the question of whether peripheral onsets lose their effectiveness in capturing attention or whether, instead, the performance of a perceptually demanding task entails a faster disengagement of attention from the cued location. Here, we measured exogenous spatial attentional-orienting effects either while participants performed a concurrent perceptually demanding central-monitoring task (a rapid serial visual presentation of letters for a to-be-detected digit target; Experiments 1 and 2) or in isolation (the baseline condition in Experiment 2). The results showed that peripheral onsets captured participants' attention at both the 80- and 190-ms stimulus onset asynchronies (SOAs) in the baseline condition. Crucially, however, during concurrent central monitoring, peripheral onsets were effective in capturing attention only at an 80-ms SOA, while the orienting effect disappeared as soon as a changing letter drew participants' attention back to the central stream (at an SOA of 190 ms). These findings demonstrate that task-irrelevant abrupt onsets cannot be entirely overridden by top-down attentional control, although attentional capture effects are dramatically reduced by an ongoing perceptually demanding task.