Numerous suggestions for solving the mystery of sleep and dream function can be found in the literature. In the previous sections I have reviewed four clusters of proposed functions of sleep and dreaming: 1) consolidation of recently acquired memories, including procedural motor skill rehearsal, replay of recently acquired memories, and integration of memory episodes into autobiographical memory schemas; 2) emotion regulation, including both an enhancement of emotionally-tagged information and a decoupling of this information from its associated emotional tone; 3) creativity and problem solving; and (4) preparation and simulation of waking life, including iterative genetic programming, virtual world model optimization, the simulation of threats (TST), and the simulation of social interactions (SST). The question thus remains what the real or primary function of sleep and dreaming is—and what the relationship between the different candidates might be. SST aims to independently cover the social simulations that fall outside the scope of TST, thereby describing an “original evolutionary function of dreams alongside with the threat simulation function of dreaming” (Revonsuo et al. this collection).
The concept of evolutionary function has been one of the main topics in the philosophy of biology (Mahner & Bunge 2000) and philosophy of mind (Millikan 1984; Neander 1991). Several notions of biological functions exist (Wouters 2003); however a general idea is that the biological function of a trait is determined by its contribution to evolutionary fitness (Walsh & Ariew 1996). Darwin (1871) differentiated between selection occurring as a consequence of ecological factors that directly threaten the organism’s survival, such as predators or other potentially life-threatening dangers of nature, and interactions with members of the same species in order to compete for mating partners. Both principles, dubbed natural and sexual selection respectively, eventually determine reproductive success as the ultimate decision points for selection. In contemporary accounts, sexual selection was generalized to the concept of social selection, of which the former is considered a subtype (Lyon & Montgomerie 2012; West-Eberhard forthcoming). The concept of runaway selection, famously illustrated by the evolution of the peacock’s tail, was thought to also be applicable to the evolution of social skills in higher animals, eventually leading to the development of theory of mind, language, dance, or artistic creativity in humans (Flinn & Alexander 2007). This process of an arms race of social skills would require increasing cognitive capacity—and in fact, at least in primates, relative brain size has been related to social group size (Dunbar 1992; Dunbar & Shultz 2007).
It is tempting to associate natural and social selection as the main principles of evolution with TST and SST, respectively. This interpretation would strongly support TST and SST, as it would equate the function of dreaming with two main principles of evolution in general. In this broad sense, however, certain attributes like learning capacity or motor skills increase fitness in terms of natural selection, but do not necessarily serve to help us avoid direct threats. Likewise, certain attributes such as emotion regulation or artistic creativity increase fitness in terms of social selection, but are not necessarily themselves social in a strict sense. Ultimately, of course, all these functions serve reproductive success—however, if any skill ultimately helping us to acquire sexual partners is interpreted as social and any possible obstacle to reproduction is interpreted as a threat, then TST and SST would be trivial, as a biological function is by definition one that supports reproductive success. In contrast, if TST and SST are interpreted in a more narrow, non-trivial way, there is ample space in dreams for further functions: consolidation of navigational information acquired during exploration; rehearsal of a recently learned motor sequence; facilitation of a behavior recently rewarded with food; incubated creative insight into the solution of a recent unsuccessful attempt to build a helpful tool; refinement of the discriminative skills regarding recently perceived pattern, etc.—all these potential benefits of sleep and dreaming increase inclusive fitness of the individual, but do not directly refer to the simulation of threats or social interactions.
This problem can further be illustrated by Revonsuo’s (1995, 2006) approach, where he considers any phenomenal experience as a virtual world model: what is the function of waking consciousness, threat avoidance, or social interaction? Both threat avoidance and social interaction, of course—and many others. That this rather uninformative answer can also be transferred back to the function of dreaming might be illustrated with another ubiquitous example of simulation: in child’s play, simulation of real life and the practice of skills needed therein is considered one of the main functions—play allows children to simulate coping with threats in a safe environment, and to develop the social skills needed later in life (Mellou 1994; Pellegrini & Bjorklund 2004). However, these aspects, while important, are not the only functions of play—it also offers the rehearsal of motor and sensory skills, training in predatory behavior, and general intellectual development. Hence, child’s play can be considered multifunctional, as can waking or dreaming consciousness.
Segmentation of reality (including dream reality) is possible along numerous lines. In a sense, TST and SST could be interpreted as expressing two orthogonal dimensions of dream space: a security dimension with the directions threat vs. safety, and a sociality dimension with the directions social vs. individual. Dreamed accidents or natural disasters would be characterized by low security and sociality, dreamed experience of bullying by high sociality and low security, and dreamed bonding by high sociality and security, etc. Threat and social interactions in a narrow sense are important aspects both of waking and dreaming life, however they are not the only aspects. Other segmentations are also possible, e.g., by a dimension of motor activity vs. inactivity, or emotional vs. neutral dream content, or a novelty dimension. In the broad sense of natural and social selection, threat and social interaction would be the two main drivers of evolution, however to the cost that the answer to the question of the function of dreaming becomes a trivial “to support reproductive success”. Of note is that the other discussed functions might be interpreted within a simulation framework: e.g., simulation visuomotor activity after learning a respective task in the memory function, simulating affective experiences in the emotion regulation function, and simulating problem solving attempts in the creativity function. These different functions are neither mutually exclusive nor strictly independent from each other. In particular the emotion-processing function largely overlaps with both TST an SST—all threats and at least the most important social interactions induce strong emotions, and successful coping with these emotions would be of considerable help when facing threats or social situations. Also other functions of dreaming overlap with TST and SST: consolidation of threat-related information or social gossip improves threat avoidance or social skills, as does creative incubation on threat-related or social problems. On a more abstract level, all these simulations serve the integration of recently experienced information into the behavioral repertoire in order to adapt it to the current waking environment (Hobson et al. 2014).
Identifying the original function of a given trait has proven to be a notoriously difficult issue in the philosophy of biology (Wouters 2013). Dreaming might have originally developed as an epiphenomenon of rather basal neurophysiological sleep functions, and this phenomenological level might eventually have acquired additional functions. Such exaptations (Gould & Vrba 1982) might have been further adapted and in turn developed further neurophysiological exaptations without phenomenological correlates, etc. The original function of dreaming might be unimportant today compared to subsequently evolved functions. Instead of singling out one or two functions of dreaming as original, dreaming might be best seen as a multifunctional general reality simulator, including the simulation of motor skills, emotional processing, problem solving attempts, threats, and social interactions. To follow specific research questions, of course certain functions still could be highlighted and followed as research heuristics with a given purpose. All functions of sleep and dreaming serve reproductive success ultimately, even though some might be more important than others from a selection point of view. For all dream functions discussed in this chapter, there are convincing supporting but also inconsistent data. The fact that dreaming is not an unselective simulation of the waking world as, e.g., the continuity hypothesis suggests (Schredl & Hofmann 2003), is a sign that some simulation functions might be more important than others. We should note, however, that quantitative overrepresentation of a specific function does not necessarily prove the primacy of this function: different functions might rely on different processes with different timescales, with a highly important function potentially requiring only seconds to be processed, while an unimportant function might take hours. In times of sufficient sleep, dream content related to the relatively unimportant function might thus be overrepresented. The relative importance of one function over another might be tested in cases of scarcity of sleep, e.g., under sleep deprivation, when different functions would have to compete for restricted simulation time. Also of interest in this regard is a comparative approach: it has been demonstrated that sleep propensity, and particularly REM sleep, negatively correlates with predatory risk across species (Lima et al. 2005), which would rather speak against TST. Concerning SST, the tendency to sleep in groups has been reported to negatively correlate with sleep time, which, however, has been interpreted either in terms of social sleep being more efficient due to reduced predatory risk, or as more social species sacrificing sleep to service social relationships during wakefulness (Capellini et al. 2008). Against this background, sleep and dreaming pose an optimization problem: how much time is best spent asleep, spent in specific sleep stages, and spent engaging in specific dream mentation in order to optimize the interplay between the different functions of sleep and dreaming? Dreaming as a general reality simulator might dynamically change its functional priorities, favoring one over the other of its several functions, depending on the current requirements and constraints of the environment.