Previous work has found that search principles derived from simple visual search tasks do not necessarily apply to more complex search tasks. Using a Multi-element Asynchronous Dynamic (MAD) visual search task, where high numbers of stimuli could either be moving, stationary and/or changing in luminance, Kunar and Watson (2011) found that, unlike previous work, participants missed a high number of targets with search for moving items worse than for static and no benefit for finding targets that showed a luminance onset. Here we investigate why luminance onsets do not capture attention and whether luminance onsets can ever capture attention in MAD search. Experiment 1 investigated whether blinking stimuli, which abruptly offset for 100 ms before re-onsetting again -conditions known to produce attentional capture in simpler visual search tasks -captured attention in MAD search, while Experiments 2 -5 investigated whether giving participants advance knowledge and pre-exposure to the blinking cues produced efficient search for blinking targets. Experiments 6 -9 investigated whether unique luminance onsets, unique motion or unique stationary items captured attention. The results found that luminance onsets only captured attention in MAD search when they were unique, consistent with a top-down unique feature hypothesis.