REM sleep behavior disorders (RBD) are violent dream-enactments caused by loss of normal muscle atonia during REM sleep. RBD are idiopathic or associated with neurological diseases (parkinsonism, narcolepsy) and represents an exceptional window on dream content and motor control during REM sleep. The characteristics of the movements during RBD have been poorly studied. In patients with Parkinson’s disease, we recently found that movements were faster, and speech was more intelligible during RBD than during wakefulness. The disappearance of parkinsonism during REM sleep suggests a transient ‘levodopa-like’ reestablishment of the basal ganglia loop. Alternatively, the jerky movements could be the expression of the primary motor cortex, relieved from the filtering, smoothing control of the basal ganglia. To support this hypothesis, using questionnaire and video-polysomnography, we will measure (i) if the head, mouth, arms and hands are more used than the legs during the RBD, as they have a larger cortical representation ii) examine the synchronization between the direction of eye movement (the major characteristics of REM sleep) with other movements during RBD in relation to dream imagery in REM sleep. To further understand the role of REM sleep-dependent procedural learning, we will also analyze in RBD patients if a specific procedural task is replayed during REM and the context of the replay. The first, precise description of movements during RBD and their synchronization with rapid eye movement real-time correlated with visual dream imagery would bring important information to recognize them in complex clinical cases, to infer their source, and to understand why parkinsonism disappears during RBD. Finally, the exploration « in live » of the potential replay of a procedural learned task in RBD patients represents a new approach that could carry some evidence supporting the role of REM sleep in procedural learning.