The formal features of dream characters were studied in a sample of 320 dream reports submitted by 33 adult subjects (13 male, 20 female) of varying ages in a university extension course. Subjects were queried by questionnaire about dream characters immediately after recording their dreams upon awakening in their normal home setting. It was found that 48% of characters represented a named personage known to the dreamer, 35% were generically identified by their social role (e.g., policeman) or abstract relation to the dreamer (e.g., a friend) while only 16% were wholly novel. Seventy-seven percent of characters were pseudosensorily present in the dream whereas 23% were present only by mention or thought. Subjects were allowed to endorse one or more of four bases of recognition and, among named characters, 32% were identified by 'appearance', 21% by 'behavior', 45% by 'face', and 44% by 'just knowing' (with the respective percentages for generic characters being 39%, 38%, 9% and 40%). Fourteen percent of named and generic characters had associated some element of bizarreness most frequently consisting of an incongruous feature. Comparing the 25 longest and 25 shortest reports, named subjects were significantly more common in the shortest reports whereas generic and unknown characters were more common in the longest reports. Results are interpreted in neurocognitive terms as possibly reflecting a decrease during dreaming relative to waking in the exchange of information between inferotemporal face identification areas and prefrontal areas subserving logic and working memory.