Consolidation, integration, regulation, and re-evaluation of acquired information during sleep prepare the organism for its waking life. However, such processes do not necessarily need to be purely reactive, depending solely on the experiences of the preceding day: several authors propose that a major function of sleep and dreaming might include primarily preparational mechanisms. Since REM sleep dominates sleep more during early developmental periods in comparison to later in life, some researchers have argued that REM sleep plays a role in early brain maturation (Roffwarg et al. 1966; Marks et al. 1995; Mirmiran 1995); however, another life-long preparational function of REM sleep has been proposed. One of the first approaches in this direction was offered by Jouvet (1979), who combined the brain maturation hypothesis with a metaphor offered by Dewan (1970), in which he claims that the brain is a computer that is programmed during REM sleep—suggesting that innate behaviors are rehearsed during REM-sleep dreaming in order to prepare the organism for their application in waking life. Jouvet later revised his approach, assuming that REM sleep constitutes an iterative genetic programming that helps to maintain the process of psychological individuation (Jouvet 1998). In a similar vein, Hobson (2009) proposed that REM sleep may constitute a “protoconscious” state, preparing the organism for waking conscious experiences. The development of consciousness during ontogenetic development in this view is a gradual and lifelong process, building on the more primitive innate virtual reality generator, which is phenomenally experienced as dreaming. With the recent integration of Friston’s (2010) predictive coding approach into this theory, the brain is thought to run a virtual world model (see also Revonsuo 1995, 2006; Metzinger 2003) that is continuously updated by processing prediction errors during wakefulness. Freed from external sensory constraints, processing of prediction errors in the dreaming brain actively refines intermediate hierarchy levels of the virtual world model. Dreaming thereby minimizes internal model complexity in order to generate more efficient predictions during subsequent wakefulness (Hobson & Friston 2012; Hobson et al. 2014).
One of the first and today the most widely discussed preparational approach is based on the observation that during dreaming particularly threatening experiences are overrepresented: the Threat Simulation Theory (TST) proposes that one function of sleep is to simulate threatening events, and to rehearse threat perception and threat avoidance (Revonsuo 1995, 2000). Such a mechanism of simulating the threats of waking life over and over again in various combinations would be valuable for the development and maintenance of threat-avoidance skills. Several empirical studies support TST (Revonsuo 2006; Valli & Revonsuo 2009), however some inconstant findings have been reported (Zadra et al. 2006; Malcolm-Smith et al. 2008, 2012). In a variation of TST, Revonsuo et al. (this collection) propose the Social Simulation Theory (SST), according to which the function of dreaming consists in the simulating of “the social skills, bonds, interactions and networks that we engage in during our waking lives”. The SST aims to predict and explain the simulations of social interaction of dream avatars that happen outside threatening events in dreams. Like the TST, predictions of the SST are supported by a number of studies, but face inconsistent data (Revonsuo et al. this collection).
On a neurobiological level, empirical support for simulation theories of dreaming comes from a recent study demonstrating that the ventromedial prefrontal cortex subserves the simulation and evaluation of possible future experiences, integrating arbitrary combinations of knowledge structures to simulate the emergent affective quality that a possible future episode may hold (Benoit et al. 2014). As the ventromedial prefrontal cortex is known to be activated in REM sleep (Nir & Tononi 2010), this mechanism might also underlie episodes of reality simulation during dreaming. Further neurobiological support for the preparational role of sleep comes from recent research demonstrating a neural “preplay” of future learning-related place-cell sequences in the hippocampus (Dragoi & Tonegawa 2011, 2013). In contrast to the intuitive view that such activation patterns are established for the first time during a novel experience, according to these findings the specific temporal firing sequence during learning seems rather to be selected from a larger repertoire of preexisting activation patterns, thus suggesting that sleep plays a role not only in the subsequent consolidation, but also in the preceding preparation for new experiences. It has been demonstrated that sleep preceding the learning experience indeed influences memory acquisition during the following day (van der Werf et al. 2009). Interestingly, support for the hypothesis that sleep mentation constitutes a virtual reality model preparing for waking life comes also from research outside of sleep neuroscience: approaches probing artificial intelligence demonstrate that robots perform better in navigational tasks if they create and update models of their own structure and actions during a state of motoric inactivity (Bongard et al. 2006). Not surprisingly, this process of evaluation and simulation of prior and future actions was interpreted as dream-like (Adami 2006).
In summary, a fourth important function of sleep and dreaming is preparation for waking life. This includes proposals of REM sleep as an iterative genetic programming system, dreaming as a state of protoconsciousness and virtual world model optimization, and dreaming as a simulation of threats (TST) and social interactions (SST).