The 'scanning hypothesis' of rapid eye movements during REM sleep: a review of the evidence.

Rapid eye movements (REMs) and visual dreams are salient features of REM sleep. However, it is unclear whether the eyes scan dream images. Several lines of evidence oppose the scanning hypothesis: REMs persist in animals and humans without sight (pontine cats, foetus, neonates, born-blinds), some binocular REMs are not conjugated (no focus point), REMs occur in parallel (not in series) with the stimulation of the visual cortex by ponto-geniculo-occipital spikes, and visual dreams can be obtained in non REM sleep. Studies that retrospectively compared the direction of REMs to dream recall recorded after having awakened the sleeper yielded inconsistent results, with a concordance varying from 9 to 80%. However, this method was subject to methodological flaws, including the bias of retrospection and neck atonia that does not allow the determination of the exact direction of gaze. Using the model of RBD (in which patients are able to enact their dreams due to the absence of muscle atonia) in 56 patients, we directly determined if the eyes moved in the same directions as the head and limbs. When REMs accompanied goal-oriented motor behaviour during RBD (e.g., framing something, greeting with the hand, climbing a ladder), 90% were directed towards the action of the patient (same plane and direction). REMs were however absent in 38% of goal-oriented behaviours. This directional coherence between limbs, head and eye movements during RBD suggests that, when present, REMs imitate the scanning of the dream scene. Because REMs index and complexity were similar in patients with RBD and controls, this concordance can be extended to normal REM sleep. These results are consistent with the model of a brainstem generator activating simultaneously images, sounds, limbs movements and REMs in a coordinated parallel manner, as in a virtual reality.