Dream reports are often not accurate representations of experiences during sleep, according to the narrative fabrication thesis. The received view that dreams are experiences during sleep that are accurately reported upon waking is challenged. Dream reports are frequently confabulated due to poor memory, bizarre content, and cognitive deficits. Narratives can be altered between initial REM sleep awakenings and subsequent reports. Experiments show subjects rationalize strange elements, omitting supernatural or bizarre components in waking memories. Inaccuracies are worsened by rapid memory loss and inability to reality-test dream memories. Lucid dreams with higher cognitive functions may yield more accurate reports, but generally dream reports are less accurate than waking reports. The theory cannot be tested with current methods but may be in the future.
Dreams usually feature a protagonist who represents the dreamer, but sometimes self-representation is displaced, disrupted, or absent. Examining dream reports from established databanks reveals two forms of this polymorphism: dreams where the dreamer takes an external, observer perspective on themselves, and dreams where the dreamer takes someone else's perspective on events. These cases parallel the phenomenon of observer perspective in waking memory and imagination. In other dreams, the dreamer does not appear at all, and the first-person perspective is occupied by another character—termed 'vicarious dreams.' These puzzling cases pose empirical and conceptual questions about the nature of self-representation with implications beyond dreaming.
The sense of agency over one's own thoughts can be altered in dreams in ways that resemble symptoms of schizophrenia, particularly thought insertion and auditory hallucination. Analysis of dream reports shows that while both thought insertion and auditory hallucination occur in dreams, thought insertion is very rare, unlike in schizophrenia where it is common. The rarity may be explained by cognitive differences between schizophrenia and dreaming, or by problems with how dreams are reported—vague reports often fail to distinguish whether 'hearing voices' refers to sound or thought phenomenology. Better experimental conditions could disambiguate these reports and deepen understanding of schizophrenia through the study of similar experiences in dreams.
A pluralist approach to dream content argues that dreams vary along a spectrum from bizarre and incoherent to wake-like and mundane, rather than being exclusively one or the other. Theorists disagree on defining dreams as primarily bizarre or convincing simulations partly because rating scales can over- or underestimate bizarreness depending on their variables and interpretation of contextual factors. Double-blind analysis of dream reports is considered the most accurate method, but only the dreamer can judge whether an event would be bizarre in their own life; dreamers may find elements bizarre after waking that seemed normal during the dream. Some dreams are too bizarre or incoherent to report accurately, while others are indistinguishable from waking life.