Three experiments tested the effect of verbal description on face identification accuracy. Based on verbal overshadowing research, it was predicted that enhancing verbal description of a face would reduce subsequent face identification accuracy. Experiment 1 tested and confirmed this hypothesis using the cognitive interview to enhance verbal description; face identification accuracy was reduced significantly following the cognitive interview, compared with a standard police interview. Experiments 2 and 3 tested and confirmed the hypothesis that verbal overshadowing would be reduced when a delay is inserted between verbal description and face identification, hence resulting in “release from verbal overshadowing.” These results suggest that in the verbal overshadowing task, the verbal description does not overwrite the visually based representation of the face in memory but rather makes it less accessible at the time of face identification. The cognitive interview reduces face identification accuracy only when the identification follows description immediately—a rare situation in real criminal cases.