Recognition without identification is the finding that, among recognition test items that go unidentified (as when a word is unidentified from a fragment), participants can discriminate those that were studied from those that were unstudied. In the present study, we extended this phenomenon to the more life-like situation of discriminating known from novel stimuli. Pictures of famous and nonfamous faces (Exp. 1), famous and nonfamous scenes (Exp. 2), and threatening and nonthreatening images (Exp. 3) were filtered in order to impede identification. As in list-learning recognition-without-identification paradigms, participants attempted to identify each image (e.g., whose face it was, what scene it was, or what was in the picture) and rated how familiar the image seemed on a scale of 0 (very unfamiliar) to 10 (very familiar). Among the unidentified stimuli, higher familiarity ratings were given to famous than to nonfamous faces (Exp. 1) and scenes (Exp. 2), and to threatening than to nonthreatening living/animate (but not to nonliving/nonanimate) images (Exp. 3). These findings suggest that even when a stimulus is too occluded to allow for conscious identification, enough information can be processed to allow a sense of familiarity or novelty with it, which appears also to be related to the sense of whether or not a living creature is a threat. That the sense of familiarity for unidentified stimuli may be related to threat detection for living or animate things suggests that it may be an adaptive aspect of human memory.