Lampros Perogamvrosa and Sophie Schwartz, of the University of Geneva, Switzerland have given us a significant advance in the theory of REM sleep and dreams in their recent paper “The roles of the reward system in sleep and dreaming.” published in Neuroscience and Biobehavioral Reviews (volume 36, 2012, pages 1934–1951).
In that paper Perogamvrosa and Schwartz present what they call their ‘Reward Activation Model’ (RAM) of sleep and dreams. The authors integrate recent neurophysiological, neuroimaging, and clinical findings that point to significant activation of the mesolimbic dopaminergic (ML-DA) reward system during both NREM (N2 in humans, SWS in rats) and REM sleep. Phasic bursting of ventral tegmental neurons (VTA) in the ML-DA system is known to signal reward prediction error such that rate increases over a tonic baseline signals reward and rate decreases signal lack of reward or punishment/threat. Given that this system is activated in N2 and REM what is its contribution to sleep and dreams?
The authors propose that that this system functions to select emotionally significant memories during the N2 phase and then facilitates consolidation during the REM phase thus enhancing overall learning and synaptic plasticity. The system can also contribute to triggering REM via the projection from the VTA to the sublateral dorsal nucleus (SLD) of the pons. While generation of REM enhances the likelihood of dreams it does not guarantee their occurrence (you can have a REM episode without recalling a dream). Nevertheless once REM occurs and a dream occurs, or conversely if REM comes online due to RAM and contributes to ongoing dream content then RAM suggests that some elements of the dream must be related to phasic bursting of VTA neurons in REM. Threat scenes inducing avoidant states may be related to reduced bursting while pleasurable scenes inducing approach states may be related to phasic increases.
One of the welcome benefits of RAM is that it simultaneously allows for a role of sleep in memory processing but avoids reducing dreams to mere flux of memory processing systems. Activation of the ML-DA reward system during sleep likely facilitates simulations of worlds or environments that invite motivationally relevant novelty-seeking behaviors (i.e. activation of Panksepp’s SEEKING system) since that is one of the known functions of the ML-DA system. This accords well with our intuitions concerning the imaginative, creative, exploratory aspects of dreams.
Another benefit of the RAM theory is that it helps to explain a host of clinical findings with regard to sleep disorders. Clinicians have known for some time that hyper-dopaminergic states tend to enhance dreams while hypodopaminergic states reduce dream recall; and that sleep deprivation and some sleep parasomnias enhance appetitive drives. Because the RAM model places the dopaminergic-driven motivational and emotional drives of the individual at the center of the dreaming state, changes in these drives and in dreams would be expected if dopaminergic activity and/or REM sleep were perturbed.
The only criticism I have of the RAM is that the authors do not really address WHY we find intense phasic activation of the ML-DA system during REM. The answer that it promotes selection of salient memories for consolidation is surely plausible but it cannot be the whole story. The authors propose a test of RAM by assessing via neuroimaging whether a salient, emotionally intense memory is preferentially consolidated versus a neutral memory element but surely no-one will be surprised if salient memory elements are preferred over non-salient elements. To really ask why REM is associated with intense activation of the salience monitoring system would be to ask what kinds of salient memories are preferred during REM. A better test would be to compare salient memories of equal intensity but addressing differing values and then see what particular elements REM prefers. This would speak directly to the real functions of REM and dreams.