There has been much debate over whether salient but irrelevant items ('singletons') ubiquitously capture attention in a bottom-up fashion. There seems to be general agreement, however, that abrupt onsets capture attention. Most of the evidence has come from cueing or visual search paradigms, where presentation of the target display is effectively an abrupt onset (potentially making an abrupt onset singleton relevant). However, in the natural environment, objects of interest may move and vary in a continuous fashion. Here, we used a multiple object tracking paradigm to study the effect of irrelevant singletons on a continuous attentionally-demanding task. Our stimuli comprised a set of identical, independently moving disks. Observers tracked a subset of target disks for several seconds, and reported at the end of the trial whether a randomly selected disk was a target or not. In Experiment 1 (N = 16), an irrelevant stationary singleton appeared at some point during the trial, and remained visible until the end of the trial. Control singletons were present from the start of the trial, before the tracking phase, while onset singletons appeared abruptly during the tracking phase. Singletons could be unique in colour, shape, both, or only in being stationary. We observed no difference between onset and control singletons, regardless of salience. In Experiment 2 (N = 16), singletons moved according to the same algorithm as the tracking disks, and could be identical to the tracking disks (additional distractor) or different in colour. We also added a no-singleton baseline condition. Here, the colour singleton impaired performance relative to the baseline condition, while there was no effect of an additional distractor. Again, however, there was no difference between onset and control singletons. These results demonstrate that performance on a continuous tracking task is surprisingly robust to the abrupt onset of irrelevant singletons.