The fear response hypothesis and the associated claim that humans have an evolutionary propensity to detect threats automatically in their immediate visual environment are critically appraised. This review focuses on reports of visual search experiments in which participants were tested with speeded oddball tasks in which the search displays contained photographic images of naturally occurring entities. In such tasks, participants have to judge whether all the images are from one category or whether the display contains a distinctive image. The evidence, which has been used to support the fear response hypothesis, is assessed against a series of concerns that relate to stimulus factors and stimulus selection. It is shown that when careful consideration is given to such methodological details, it becomes very difficult to defend the fear response hypothesis. It is concluded that, at present, the fear response hypothesis has no convincing empirical support, and it is urged that, in the future, researchers who wish to study visual threat detection take stimulus selection much more seriously.