Growing evidence suggests that angry faces do not “pop-out” of crowds, and that the evidence for such effects has tended to arise from methodological issues and stimulus confounds. In contrast, evidence that angry faces exert special influence at later stages of information processing is accumulating. Here we use two common paradigms to show that participants have difficulty disengaging attention from angry faces relative to happy faces. Experiment 1 used a visual search task to show that angry crowds took longer to search. Experiment 2 used an exogenous cueing paradigm to show that brief onset angry faces held attention and delayed responses on a primary task. This suggests that when seen, they engage attention for longer time, but they do not have the preattentive features that would allow them to pop-out. Together, these two different experimental paradigms and realistic stimulus sets suggest that angry faces resist attentional disengagement.