In two experiments exploring possible causes of the "weapon focus" effect, undergraduates viewed videotapes depicting interactions in business establishments. The target character was either empty-handed or held different objects that varied in both threat and unusualness. Witnesses attempted to describe the target's features and clothing, identify the object held by him (if any), and identify him in a photo line-up. The accuracy of witnesses' descriptions was affected by unusualness but not threat. Identification accuracy did not differ by condition. Witnesses had difficulty remembering the low-threat, non-unusual object; many either failed to identify it (Experiment 1) or reported seeing no object (Experiment 2). The results of both experiments imply that weapon focus, when it occurs, may do so because weapons are unexpected.