The relatively loose idea or the general observation that dreams are social simulations needs to be turned into a theory from which testable predictions can be derived. There are several ways in which this could be done. In the rest of this paper, we will formulate some suggestions towards that end. The basic assumptions that we adopt are based on the earlier work on the definition of dreaming (and consciousness) as an internal world-simulation in general (Revonsuo 2006). Any plausible theory of social simulation should also take into consideration, and draw from, concepts and advances in the fields of social psychology and evolutionary biology, in order to create a credible theoretical context into which social simulations in dreams can be placed. We will therefore connect the idea that dreaming may function as a platform for simulating social perception and interactions to some influential evolutionary biological and social psychological theories, as well as to the earlier simulation theory of the original evolutionary function of dreaming, the TST (Revonsuo 2000).
The two generally-accepted theories in evolutionary biology that seem to be relevant for the formulation of an evolutionary SST of dreaming are the Inclusive Fitness and Kin Selection Theory (Hamilton 1964) and Reciprocal Altruism Theory (Trivers 1971). Both are general evolutionary biological theories that apply not only to humans, but to multiple other species as well. Further, both have received ample empirical support from animal and human studies, and could thus serve as solid ground in guiding our thinking about social behaviours in evolutionary biological terms.
The Inclusive Fitness Theory (Hamilton 1964) postulates that an individual’s genetic reproductive success is the sum of that individual’s direct reproduction and the reproduction of the individuals carrying identical gene alleles. An individual can improve its overall genetic success by engaging in altruistic social behaviour that is directed towards individuals carrying identical alleles. The Kin Selection Theory is a more specific form of the inclusive fitness theory, which requires that the shared alleles are identical by descent. Thus, Kin Selection Theory postulates that an individual can increase its inclusive fitness by directing acts of altruism specifically towards genetic relatives, whereas inclusive fitness as such is not limited only to cases where kin are involved. Both, however, predict that acts of altruism should more often be directed towards individuals who share identical alleles.
Reciprocal Altruism (Trivers 1971) is defined as behaviour whereby an individual acts in such a way that temporarily reduces its fitness while increasing another individual’s fitness. However, individuals engage in altruistic behaviour with the expectation that the recipient of the altruistic act will act in a similar manner at a later time. A strategy of mutual cooperation may be favoured when there are repeated encounters between the same individuals. Although cheating might be more beneficial for the individual in terms of immediate rewards, co-operation might provide net gain compared to short-term benefits.
Since selection pressures act on the typical conditions present in the history of any species, consideration of the demographics of the typical evolutionary environment of humans is crucial for understanding the evolution of social behaviours in our species. Recently, Hill et al. (2011) analyzed co-residence patterns among thirty-two present-day foraging societies, assuming that these might reflect an ancestral human group structure. They found that primary and distant kin of an adult individual accounted for approximately 25% of the co-resident adult members of a band, i.e., about 25% of adult members in the group were directly genetically related, whereas about half of the adults were related through spouse or siblings’ spouses, and the other 25% of adults were genetically unrelated.
If we accept the assumption that this observed distribution of relatedness approximates the degree of relatedness in ancestral human bands, there have been ample opportunities for ancestral humans to be subjected to selection pressures that could be explained using strategies postulated by the inclusive fitness and Kin Selection Theory, as well as Reciprocal Altruism Theory. There is ample evidence that people are more likely to help their relatives than genetically unrelated individuals (e.g., Burnstein et al. 1994), and that lethal violence is more frequently directed towards genetically-unrelated individuals than relatives (Daly & Wilson 1988). People also tend to be more altruistic towards other people in single round prisoner’s dilemma game than could be expected (Frank et al. 1993) in order to protect their reputations. This seems to be a reasonable course of action, given that the faces of individuals labelled as untrustworthy cheaters are better recalled than those labelled as cooperative (Mealey et al. 1996). There are also rather large interindividual differences in altruistic behaviour, depending on factors such as age, sex, tendency to empathize, and circumstantial conditions.
The social environment has afflicted strong selection pressures on human cognitive faculties, and there are several theories that consider our essentially social nature. Dunbar (1992, 2008) has forwarded the Social Brain Hypothesis, which states that the main factor in the increase of our neocortical volume has been the cognitive demand bestowed on us by the increase in hominid group size. Sutcliffe et al. (2012) propose the idea that the costs and benefits of social interactions have been a critical driver for cognitive evolution. While our most intimate relationships are a source of social support, they are also the most costly as the quality of these relationships is dependent on the time invested in creating and maintaining them over time. Forming weaker and less time-consuming ties with acquaintances can provide benefits such as information exchange and access to resources without exhausting an individual’s resources that are allocated for social interaction. Our individual social worlds thus consist of hierarchically-layered sets of relationships defined by relationship intimacy, and different relationship types are designed to have different kinds of functions.
Turning our attention to the potentially relevant literature in social psychology, some further concepts and measures might be considered useful for dream theory. When it comes to the simulation of social interaction, one of the most relevant concepts is the social “Need to Belong” (Baumeister & Leary 1995). This fundamental motive towards interpersonal attachment and close, supportive social bonds pervades and influences our actions, emotions, and cognitions, and is fulfilled only by social affiliation and acceptance. To help us navigate the complex social world, and attune us to socially relevant information, two further advancements have been hypothesized in the form of the Sociometer Theory (Leary et al. 1995) and the social monitoring system (Gardner et al. 2000). Sociometer Theory proposes an internal monitoring device that feeds forward information about our level of social inclusion in the form of self-esteem or self-worth (Leary et al. 1998), whereas the social monitoring system is purported to guide the processing of social information whenever people’s needs to belong are not being met (Pickett et al. 2004). In sum, the concept of “Need to Belong” in general, and the suggested social monitoring systems in particular, might prove useful in postulating testable hypotheses for the functions of social simulation in dreams. The Sociometer, for example, might act in a similar fashion to the threat cues postulated in TST, and prompt dreams to simulate relevant social skills or interactions.
An interesting developmental suggestion about the interplay between simulation mechanisms and social deficits has recently been put forward by Oberman & Ramachandran (2007), who propose that in typically developing individuals the abilities of Theory-of-Mind (ToM), empathy, perceptual recognition, and motor mimicry might be mediated by an internal simulation mechanism or mechanisms. By taking into consideration a condition—autism—where all these abilities appear to be impaired, they make the case for a possible link between deficient simulation mechanisms and behavioural and social deficits. The exact implications of this idea for the hypothesis that dreams serve a social simulation function requires further consideration. One possibility is to test whether individuals with Autism Spectrum Disorders (ASD) dream less of social interactions, or whether their dreams of social interactions are different in content from those of other people. Thus far this line of research has not been explored in depth. Daoust et al. (2008) have looked into the dream contents of people with ASD, and found that they report significantly less dream-characters and social interactions than the control group. They note, however, possible error sources in the testing procedure, such as, for example, how the reporting of dreams itself might be affected by ASD.
There has been some research linking the effects of attachment relationships to dreaming. If, as attachment theory proposes, we use our early experiences with primary caregivers and other attachment figures as model states for future social interactions and the way we view and attune to our social world, it could be assumed that this would also affect our simulations of this world. Early attachment and bonding are, after all, quintessential for our species, and according to Fonagy & Target (1997) might also work as the basis for our abilities to mentalize or to create a ToM. McNamara (1996) has developed the idea that REM sleep is the mechanism that activates and maintains early attachment relations, as well as pair-bonding in later life. Selterman & Drigotas (2009) have found that attachment style is correlated to dream emotions when dreaming about romantic partners, so that those with anxious or avoidant attachment styles reported more stress, conflict, and negative emotions.
In an exploratory study on the dream contents of those suffering from Complicated Grief (CG) after the loss of an attachment figure, Germain et al. (2013) found the dreams containing family members to become significantly more frequent, while there was no marked increase in the occurrence of deceased characters. Males suffering from CG also reported more familiar persons in their dreams than the control group. Both male and female CG patients also exhibited fewer negative emotions and fewer instances of aggression in their dreams, and females also had decreased amounts of positive emotions and friendliness.
We can thus conclude that the inherently social nature of our species is deeply ingrained, and has likely been as important for our survival in the ancestral environment as threat perception and avoidance skills. SST can therefore be formulated in an analogous manner to TST, but in addition to the evolutionary background theory, also taking into consideration important social functions such as the need to belong, social bonding, social networking, and social support as essential ingredients.
TST (Revonsuo 2000) places the contents and the function of dreaming in an evolutionary-psychological context and proposes that dreams were selected for their ability and propensity to simulate threatening events in a safe way, thus preparing the individual to survive real-life dangers. The hypotheses and predictions of the TST, especially concerning the inclusion of threat simulations in dream content, have gained support from several independent sources, such as studies on the content of nightmares and bad dreams (e.g., Robert & Zadra 2014), recurrent dreams (Valli & Revonsuo 2006; Zadra et al. 2006), post-traumatic dreams in children and adults (Bulkeley & Kahan 2008; Valli et al. 2006), dreams anticipating a stressful experience (Arnulf et al. 2014), children’s earliest dreams (Bulkeley et al. 2005), dreams and mental contents in parasomnias (Uguccioni et al. 2013), the dreams and nightmares of new mothers (which mostly depict the infant in peril and trigger protective behaviours, Lara-Carrasco et al. 2013, 2014; Nielsen & Lara-Carrasco 2007), as well as dreams of the general population (for a review, Valli & Revonsuo 2009).
Thus, when it comes to emotionally negatively-charged dream contents that simulate some sort of dangerous situation or unfortunate event, the TST seems able to quite well predict and explain many features of the quantity and the quality of the threat simulations found in the data. Therefore, a similar theoretical approach might also prove fruitful in the case of social simulation theory. The SST, however, needs to be formulated in such a manner that its predictions can be clearly distinguished from those of the TST.
As negative and threatening events commonly occur in dreams, the TST alone already covers a fairly large proportion of dream content. But it also ignores a relatively large proportion of dream content, as it does not offer any explanation of non-threatening dreams or for the simulation of neutral and positive events in dreams. This raises the question: do types of dream events other than those that are threatening have some evolutionarily-based simulation function, independent of the threat-simulation function of dreaming? Are there events that are equally important targets for simulation as the negative, threatening situations simulated in threat simulation dreams?
TST covers threatening events in dreams, whether social in nature or not. Many threatening events of course do involve social interaction (such as verbal or physical aggression), but are explained by the TST as primarily simulations of specific types of threat, and therefore as rehearsals of threat perception and threat-avoidance behaviours, rather than as simulations of social interactions as such. A social simulation theory that explains dreams that TST does not cover should thus focus on social simulations that are largely independent of the threat-simulation function. In some dreams these two types of simulation may, however, be difficult to tease apart. For example, a social simulation theory might account for some social interactions that happen during a threatening event in a dream, such as how the Dream Self interacts with others and collaborates with them during a threatening situation. Furthermore, these two simulation theories may not be mutually exclusive but instead complement each other. Some specific types of simulations of negative social interactions are better accounted for by the TST while other, positively toned simulations can be explained by the SST. For example, from an evolutionary perspective it might make sense to simulate different kinds of interactions, friendly or aggressive, with people belonging to different layers of our social hierarchy.
We are open to the possibility that social simulation is an original evolutionary function of dreams alongside the threat-simulation function of dreaming. We believe that social simulation theories hold much promise. But before this belief can be empirically justified, a testable version of the social simulation theory needs to be formulated. Such a theory should independently cover the social simulations in dreams that fall outside the scope of the TST.
Furthermore, also the predictions of the CH must be distinguished and separated from those of the SST. Therefore, the question becomes: What aspects of human social reality might dreams be specialized in simulating in such a way that these social simulations have significant consequences for cognition and behaviour during the waking state, and in virtue of which social simulations during dreaming have fulfilled important functions in the evolutionary history of the human species? What kind of social-cognitive processes and behavioural social skills might have been both critical enough both for an individual’s survival and successful reproduction, as well as occurring frequently and universally enough in the human ancestral environment, to be selected for as a universal feature of human dreaming? Moreover, those processes and skills would have to be something that in fact can be regularly simulated by the dreaming brain, and they have to be contents that actually are being simulated frequently and universally in human dreaming, according to the evidence from content analysis studies of dreaming.
To sum up, a credible version of the SST should have predictions and explanations that are clearly different from both the TST and the CH. To be different from TST, the SST should predict and explain the social simulations that happen outside threatening events in dreams, and to be different from the CH, the SST should predict that some types of social stimuli, social cognition, or social behaviours are simulated actively and selectively, so that they are overrepresented in dreams as compared to waking life.
We will first consider some basic cognitive processes that might fulfil these roles and will then proceed to more complex social behaviours and interactions. We admit that many of these ideas are at this stage speculative. But if it is possible to formulate them in an empirically testable manner, then we can figure out later on which ideas remain mere empirically unsupported speculations, and which ones might actually predict and explain central aspects of our dream content.
Overall, there are good reasons to support the view that fast and errorless social perception abilities were universally important skills for humans during their evolutionary history, and, therefore, rehearsing them through dream simulations would have served to maintain and enhance their speed and accuracy during wakefulness. In the ancestral environment, fast and efficient social perception and recognition mechanisms were essential for telling friends and allies apart from potential enemies. Thus, detecting the presence of other human beings in the same spatiotemporal context where oneself is located, immediately classifying them in terms of familiarity, identity, and history of past interactions with them, and predicting the nature of future encounters with them must have been an important survival skill. Perhaps it was important enough that rehearsal of these social-cognitive functions through social simulations during dreaming would have increased an individual’s inclusive fitness.
The social perception system needs to quickly estimate answers to the following questions: am I alone in here or are there other humans present? Are the other humans around me familiar to me or are they strangers? Thus, the first stage of social perception is to detect other humans in the vicinity and to classify them in terms of unfamiliar people (strangers) vs. familiar people. As Diamond (2012) explains in “The World Until Yesterday”, in most traditional societies during human evolutionary history, to encounter strangers was unusual and typically considered potentially dangerous, because the social interaction that followed might not necessarily have been peaceful in nature.
The second stage of social perception deals in more detail with the familiar people that are detected. If the people in my presence are familiar to me, who exactly are they? What is my relationship with them? What have my past interactions with them been like? What should I expect the interaction between us to be like this time around? To answer these questions, familiar people need to be quickly identified. Based on semantic and autobiographical memory information that we have about people familiar to us, we quickly activate expectations and strategies as to how we should interact with the people around us in the most constructive way.
But so far this idea is mere speculation. What kind of testable hypotheses and predictions could be derived from this theory? How could we derive predictions that clearly distinguish the SST from the CH? The CH does not attribute any evolutionary simulation functions to dream content; according to CH, dreaming simply and passively mirrors whatever experiences have recently been encountered in the dreamer’s waking life (and thus impressed on long-term memory). Obviously, therefore, it would not lend sufficient (or specific) support to the SST to predict that social perception should be found in dreams in the same proportions as in waking life, because the CH predicts and explains exactly the same observation and, moreover, does it more parsimoniously, without postulating any just-so-story of evolutionary functions to social dream content.
The SST must thus go beyond the CH and make the risky prediction that, if social perception is the original evolutionary function of dreaming and it is therefore still expressed in our dream contents, then dreams are specialized in simulating social perception. If dreams are specialized in simulating social perception, then perceptual contents, cognitive processes, and behaviours relating to social perception skills should occur (as simulations) in a selective or exaggerated form in our dreams. The testable prediction derived from this is that during dreaming, social perception occurs more frequently than in waking life (shows quantitatively an increased frequency) and/or qualitatively in a more difficult or challenging form than in waking life.
Quantitatively, dream simulations could exaggerate the proportion of the types of stimuli that were most important to recognize quickly and accurately during evolutionary history (e.g., strangers vs. familiar people; enemies vs. friends). It is important to process this information quickly because the information had high survival value in ancestral environments. Furthermore, dream simulations could present qualitatively challenging stimuli for the social perception system; for example, more variety of different kinds of stimuli (different kinds of familiar and unfamiliar simulated people), or ambiguous stimuli that are more difficult to perceive or interpret than real life stimuli (vague or unstable simulations of people).
Conversely, if the social stimuli in dreams simply mirror the social stimuli during wakefulness (and memory representations of them), quantitatively and qualitatively, then the CH gains support: dream experiences merely copy the patterns and rates of social stimulation encountered during wakefulness, but do not selectively and actively simulate them in ways and proportions that would reflect some original evolutionary functions and would therefore have supported important survival skills in ancestral environments.
To test these two opposing theories, SST and CH, against each other empirically, we need detailed information not only about the quantity and quality of social perception in dreams, but also about the quantity and quality of social perception during wakefulness in the same subjects’ lives during the same period of their lives. Some studies already exist that provide us with this kind of data, but most of the hypotheses remain to be tested in future studies that should be explicitly designed to test the opposing hypotheses and predictions of the two theories.
McNamara et al. (2005) conducted an interesting study that can be interpreted as testing the SST prediction that social perception is quantitatively exaggerated in dreams as compared to waking life. They conducted experience sampling from fifteen individuals over two weeks across waking, REM sleep, and Non-Rapid Eye-Movement (NREM) sleep states. The participants recorded verbal reports of their perceptual and other experiences when paged at random intervals during sleep or wakefulness.
The results showed that more characters appeared in dreams than in wake reports. Unfortunately McNamara et al. (2005) do not report the exact descriptive statistics of this finding, so we do not know how large this difference exactly was. In any case, this finding is better in accordance with the predictions of the SST than CH: Stimuli requiring social perception (human characters) are present at higher frequencies during dreaming than during wakefulness, when experiences from both states are sampled and reported in a similar manner.
This important finding suggests that the basic processes and skills required in social perception are more engaged during dreaming than during an equal stretch of time in wakefulness. This lends support to the hypothesis that dreaming is specialized in the simulation and rehearsal of social perception, which may thus be one of the original evolutionary functions of dreaming. It has to be added, however, that McNamara et al. (2005) is the only study so far that provides us with this kind of data, where the frequencies of the social contents of dreaming and waking experiences have been directly compared with each other. Replications are obviously required in different populations and in larger samples of dreams and waking experiences. But so far, so good for SST.
The same study can be taken to test the additional prediction of SST, namely that dream simulations of human characters should exaggerate the proportion of the particular types of stimuli that were, during evolutionary history, most important to recognize quickly. Meeting strangers posed a threat in the original evolutionary context; thus, the SST predicts that strangers or unfamiliar people should be overrepresented in dreams as compared to waking life, to simulate and rehearse the type of perceptual categorization (familiar vs. unfamiliar) that was most important in the evolutionary context. McNamara et al. (2005) report that the proportion of strangers (or unfamiliar people) encountered in dreams is indeed significantly higher than in waking life. Only 25% of people present in the waking episodes were unfamiliar, whereas about 50% of the (simulated) people in dreams were unfamiliar. Again, this discrepant pattern is well predicted by and accounted for by the SST, but goes against the predictions of the CH.
The recognition and identification of familiar people as who exactly they are could also potentially be a target of useful simulation in dreams. It might be argued from SST that quick and correct recognition of familiar people enhances the quick selection of the appropriate social strategies and behaviours when we interact with them. As about 50% of simulated people in dreams are familiar, there are still plenty of opportunities to rehearse these recognition skills. There are, however, no studies that would have directly and quantitatively compared the frequency of face recognition during dreaming and wakefulness. But still, there are some studies that question whether face recognition is engaged during dreaming and to what extent.
Kahn et al. (2002) report, in a character recognition study, that about 45% of familiar dream characters were recognized through their appearance (including facial features), and an additional 12% by their observable behaviour. Thus, nearly 60% of dream characters are recognized perceptually. However, about another 12% of dream characters are recognized intuitively, by “just knowing” who they are, which suggests that in those cases, the “recognition” happens in a top-down manner and is therefore independent of the perceptual and facial features of the dream character.
If familiar persons are not overrepresented in dreams to begin with (as the McNamara et al. 2005 study suggests), and only well under 50% of the familiar people simulated in dreams are recognized through their facial features, this pattern of data does not particularly support the idea that dreams are specialized in rehearsing familiar face recognition. However, we still lack knowledge about the frequency of face recognition in waking vs. dreaming, and only a study directly making that comparison could properly test this idea. So, the case remains open, but the expectations are not particularly high that this prediction of the SST will gain strong support in the future.
In addition to the processing of familiarity and identity, another aspect of social perception is called Theory-of-Mind (ToM) or “mindreading”. This refers to the interpretations we automatically make about the internal mental states of the people around us. We not only categorize the people around us as familiar and unfamiliar, and assign an identity to familiar persons, we also attribute thoughts, beliefs, motives, and emotions to them. As mindreading is crucial for our ability to predict and explain other people’s behaviours, our mindreading abilities could potentially have been a target of simulation during simulated social perception in dreams (Kahn & Hobson 2005; McNamara et al. 2007).
The study by Kahn & Hobson (2005) quantifies the frequency of mindreading activities in dreams. In one sample of thirty-five participants and about nine dream reports per participant, about four dream characters per report were observed on average. In over 80% of these dreams, the participants reported having had engaged in mindreading (at least one of) the other dream characters’ internal mental states. In another sample, 24 subjects reported on average six dreams per participant. Each dream was divided into separate dream events (on average four events per report were found), and the participants were asked to report, concerning each event, whether or not they were engaged in mindreading the other dream characters. In 50% of the episodes, mindreading was reported to have occurred. Thus, on the basis of these results, we may say that mindreading frequently occurs during dreaming. Kahn & Hobson (2005) in fact suggest that this may be evidence for a specific simulation function being at work:
The two studies undertaken here support the idea that dreaming may provide a simulation of waking life as suggested by Revonsuo (2000), though not restricted to only threatening events. Instead, the data of these studies suggest that if dreaming is a simulation process, it is a simulation that provides a way of knowing and dealing with the intentions of others, both positive and negative. (p. 56)
The above studies show that mindreading is well represented in dreams, but they cannot tell us whether mindreading is overrepresented in dreams, as its frequency of occurrence cannot be directly compared to waking life. However, McNamara et al. (2007) have conducted a direct comparison of the frequency of mindreading between waking experiences, REM dreams, and NREM dreams of the same subjects. This is what they found:
REM reports were three times as likely to contain instances of mind-reading as were wake reports and 1.3 times as likely as NREM reports. Of 100 reports per state, there were 39 instances of mind-reading in REM reports, 29 in NREM reports, and 12 in wake reports. (McNamara et al. 2007, p. 211)
In conclusion, from looking at these studies, we may say that mindreading activities frequently occur in dreams, and that their frequency of occurrence is significantly greater during dreaming than during wakefulness: Mindreading is overrepresented or exaggerated during dreaming. Thus, this data supports the SST prediction that dreaming specifically simulates mindreading in order to maintain and rehearse our mindreading abilities, rather than the CH prediction that dreaming simply reflects the amount of mindreading we engage in during wake experiences.
Another finding that might indirectly lend support to the SST-mindreading idea is that the behaviours and communications of dream characters are often bizarre (Kahn et al. 2002; Revonsuo & Salmivalli 1995; Revonsuo & Tarkko 2002); that is, they are unusual, unexpected, and thus unpredictable on the basis of our waking expectations. Studies on intentional social interactions between the Dream Self and other avatars in lucid dreaming suggest that dream characters are largely independent of the dreamer and behave autonomously (Stumbrys et al. 2011; Tholey 1989). Unusual and unpredictable behaviours could be interpreted simply as failures of the dream simulation to produce credible sequences of real-life behaviour. But they could also be interpreted as particularly engaging and activating social stimuli that serve to challenge our mindreading skills. That is, bizarreness in this case could be functional in the sense that it makes the simulation more challenging. Perception of unexpected behaviours may trigger a reconsideration of what is going on in the character’s mind in order to produce such unexpected behaviour, and thus present a frequent need to engage in mindreading as we interact with unpredictable characters in our dreams. This idea could be empirically tested by studying whether bizarre behaviours on the part of dream characters tend to trigger mindreading in the Dream Self, and whether this feature of dreams might partially explain the apparently frequent engagement in mindreading in dreams.
Humans are an essentially social species and an individual’s survival in the ancestral environment was most likely entirely dependent on the individual’s ability to form long-lasting positive social bonds with close kin and other group members who offered protection, access to nutrition and other crucial resources for survival, collaboration, friendship, social support, mating opportunities, and opportunities to gain a better social status within the group.
Social interaction in dreams is a more complex affair than simple social perception. There need to be some behaviours that link dream characters and the Dream Self, where the intentional behaviour of one character (or the Dream Self) is directed at another character (or at the Dream Self), and the recipient somehow registers it or reacts to it. Traditionally, in the Hall & Van de Castle (1966) content analysis system, social interactions have been classified into three different categories: aggression, friendliness, and sexual interactions. It may be, however, that these three categories are too broad, and do not cover or identify all theoretically-interesting types of social interaction.
When it comes to the simulation of social interactions, the predictions of the SST should, again, be contrasted with the predictions derived from competing theories. In this case the SST needs to be distinguished from two other theories: CH and the TST. The TST is a simulation theory that describes and explains the simulation of aggressive behaviours in dreams, by including them under the category of “threatening events”. The function of dreaming, according to TST, is not to specialize in the simulation of social interactions per se, but in threatening events; thus, any social interactions are simulated in dreams not because they are social events but because they are threatening events. No independent social simulation theory is required to explain the simulation of social interactions involving a threat; and aggressive behaviours between dream characters are, obviously, social interactions where the wellbeing of the Dream Self or some other dream character is potentially threatened.
Compared to CH or SST, the TST can account for the overrepresentation of threatening events and aggressive interactions in dreams (as compared to waking life, McNamara et al. 2005; Valli et al. 2008). The TST, however, gives no description or functional explanation for neutral and positive types of social interactions (unless they occur as parts of a threatening event). The TST assumes that neutral and positive events in dreams are either parts of a threat simulation (e.g., responding to a threat by helping others who are targets of a threat) or that they represent some kind of superfluous, non-functional dreaming that simply goes on automatically even if the threat simulation mechanisms are not activated. Thus, when it comes to social interactions, the SST should in particular predict and explain the neutral and friendly types of social interactions, and show that some of them are actively selected as targets of dream simulation. In contrast, the CH predicts that neutral and positive types of social interactions should only occur in the same proportions as they occur in real life, passively reflecting their waking-life frequencies.
If, according to SST, the simulation of neutral and positive social interactions in dreams serve to represent and strengthen important social connections and to rehearse prosocial behaviours in relation to those connections, then these types of interactions should frequently occur in dreams. This would serve the function of maintaining, rehearsing, or strengthening our waking life social bonds and networks, and would satisfy our social need to belong to groups that enhance our survival. After dreaming about prosocial behaviours, our social bonds during wakefulness would automatically be experienced as stronger and we would be more likely to engage in behaviours that further strengthen those bonds. Some tentative steps towards examining how the affects and contents of social dreams predict subsequent waking behaviour have been taken by Selterman et al. (2014). They discovered that an increased frequency of dreams involving significant others was associated with higher levels of intimacy and interaction the following day, whereas dream infidelity predicted less intimacy. Reported arguments in dreams were also found to be correlated with subsequent conflict in waking life. They leave open the question whether this is due to the conscious reflection of the reporting procedure, a more implicit association, or a mixture of the two.
Again, there are no detailed content analysis studies that have investigated the exact nature of social interaction in dreams by taking into account the social context of the interaction; that is, by studying who is engaged in what type of interaction and with whom. From previous studies based on home dream diaries we know that dreamer-involved aggression, adjusted to take into account all social interactions except sexual interactions, is present in 60% of male dreams and half (51%) of female dreams (Domhoff 1996). When male strangers appear in a dream, the likelihood that physical aggression will occur in that dream far exceeds what would be expected on the basis of chance. Basically this means that male strangers signal physical aggression. The dreamer, however, is an aggressor in 40% of male dreams and a third of all female dreams (Domhoff 1996).
Yet, as the Hall and Van de Castle norms indicate, there are friendly interactions in dreams—slightly more often in female (42%) than male (38%) dreams (Domhoff 1996). Females also dream more often of familiar people (58%) than of strangers (42%) while the opposite is true for males (45% vs 55%, respectively); which might suggest that when there are more familiar people in dreams, there is also more friendliness. The dreamer participates in the majority of interactions that involve friendliness (84% for females, 90% for males), and the befriender proportion is 50% for males and 47% for females. Thus, both sexes initiate friendly interactions in their dreams approximately as often as they are befriended. Helping and protecting is the most frequent type of friendly behaviour in both sexes, followed by friendly remarks and compliments, and giving gifts or granting loans. Surprisingly, however, there is very little mutual or reciprocal friendliness, so although friendly interactions are initiated in dreams by the Dream Self or other characters, in less than 10% of friendly interactions the act is reciprocated immediately. This observation goes against any social simulation theory that predicts reciprocal friendliness should be highly represented in dreams: this does not seem to be the case.
McNamara et al. (2005) investigated whether types of social interaction are different in REM than in NREM dreams compared to wakefulness, and noticed that aggressive interactions were more often simulated in REM dreams, whereas friendly interactions were more often simulated in NREM dreams. Furthermore, dreamer initiated friendliness was more typical for NREM than REM dreams. What is most interesting in this study, however, is that they also found that social interactions in general are more often depicted in both REM and NREM dreams than in wake reports. While aggression was more often simulated in dreams than encountered in waking life, the number of reports with at least one occurrence of friendliness did not differ significantly across sleep–wake states. Thus, these observations imply that dreams do not seem to overrepresent friendly interactions as compared to waking experiences.
In sum, aggressive interactions seem to be more prominent in dreams than neutral or friendly interactions, which would lend more support to the TST than to SST, and friendly interactions are not more prominent in dreams than in waking life, which would lend support to CH and the TST. Nevertheless, if simulations are biologically functional, and if these two types of simulation functions are not mutually exclusive, might there be enough room in the dream content for simulation of neutral and positive interactions, in such a way that it could have contributed to the inclusive fitness of our dreaming ancestors?
Let us see how this general approach to social simulation in dreams could be translated into some directly testable hypotheses. Now, a general thesis derived from the SST could be formulated as follows:
Dreams are specialized in simulating the most important social connections and networks of the dreamer to give an additional selective advantage and to enhance the survival of the dreamer in waking life. The simulations of particular people (the frequency of their presence in a person’s dream life), and the simulations of positive interactions with particular people, should focus on the people closest to us in waking life and on the social bonds most important for our inclusive fitness in the real world.
This thesis could be directly tested by deriving some empirical predictions from it, telling us what kind of simulations of social interactions and to what extent they should appear in dreams. If dreams are specialized in the way predicted by SST, then the most important social networks and the people in them should appear more frequently in dream life than in a corresponding stretch of waking life. That is, their frequency of occurrence should be targets of active selection and inclusion into dreams, and hence over-represented and exaggerated in dreams.
This empirical prediction could be tested by identifying a person’s most important social networks in waking life, and by quantifying the frequency of interactions of the dreamer with those people during dreaming vs. during wakefulness. In the already existing literature, there are some data relevant to the hypothesis, but data that directly compares waking social life and dream life in the manner required to test the hypothesis seems to be lacking.
The data scattered in the literature describes the relative frequency of dreams in which a certain type of close person appears on average in the dreams of the general (or the student) population. For example, romantic partners occur in 20% of dreams and this frequency correlates with the time spent together in wakefulness (Schredl 2011; Schredl & Hofmann 2003). Core family members occur in 10%–30% of dreams; parents in about 8%–20% of dreams, and siblings from 2%–7.5% of dreams (see Schredl 2013). Friends occur in about 20% of dreams (Roll & Millen 1979), but during long-term isolation from social contacts with friends in one case (Merei 1994) this declined to 10%. In studies of long dream series from a single person, a close family member or spouse has been found to be the person most often dreamed about. In a sample of over two hundred dream reports, reported by a married woman (Arlie) with four grown-up children, the most frequently occurring character is her husband; whereas in a sample of over three hundred dreams from an unmarried woman in her thirties (Merri), the most frequently occurring character is her sister, who was no longer alive at the time when the dream reports were collected (Schweickert 2007).
In Schredl’s studies, interesting analyses of a long dream series from a single dreamer were conducted, revealing the proportions of schoolmates (2012) and family members (2013) simulated in dreams across a period stretching over twenty years. Old school mates continued to appear in about 5% of dreams over the years when the dreamer had nothing to do with them any more in real life. Similarly, family members, even when the participant was not living with them anymore, still retained a strong if somewhat reduced presence in the same dream series, being present in approximately 15–20% of the dreams over a twenty-year period.
These results show that the probability of occurrence of a character in dreams is to some extent related to the amount of real life contact with that person and to the closeness of the relationship in real life, thus supporting the CH. However, people who have at some point in life been close and important do not seem to disappear totally from the dream simulations even though they have long ago totally disappeared from the real life of the dreamer. This feature of the already-existing data suggests that simulations of social contact might serve the function of maintaining or strengthening close relationships over time. When the frequency of a previously close and important social contact falls to zero in waking life, and the person is no longer encountered in waking life (like old school mates after leaving school, or after the death of a family member), the simulation of such a person seems never to totally disappear from dream life, even if the frequency of dream simulations of that person to some extent diminishes. Social simulations in dreams thus seem to maintain an active storage and rehearsal of the most important and closest social relationships of our entire lives, even when those relationships are broken or discontinued for good, or are temporarily on hold in our waking lives.
What happens if a relationship that has disappeared from waking life is reactivated after years of disconnection? In Schredl’s (2012) study, old schoolmates met for a reunion twenty years after going their separate ways. Interestingly, when the same relationships are re-activated in real life for just one day, the dream simulation of those social relationships is increased significantly and for a long period of time (compared to the time of actually meeting). The mechanism that reactivates old targets of simulation might be analogous to that proposed in TST for the re-activation of old threats. The frequency with which the most important real threats are simulated (e.g., in post-traumatic nightmares) increases when, during wakefulness, new cues are encountered that are associated with the old threat possibly reoccurring in real life.
These considerations suggest a more precise function of social dream simulations that could be formulated along the following lines. We may call it the Strengthening Hypothesis: the function of social simulations in dreams is to maintain and strengthen the dreamer’s most important social bonds from waking life. Consequently, a prediction derived from the Strengthening Hypothesis can be formulated as follows: if strengthening important social bonds is a function of social dream simulations, then dreaming should include with high frequency social interactions in which the (current or past) most important social bonds are strengthened through various types of simulated positive social interactions and prosocial behaviours. Thus, the frequency of prosocial, positive interactions (bond-strengthening) with the most important persons should clearly surpass the frequency of negative (bond-weakening) interactions within dreams, and also be more frequent in dreams than in a corresponding stretch of waking life.
Schredl’s (2012, 2013) findings are to some extent consistent with both the CH and the SST, but do not allow any firm conclusions about which theory better predicts the occurrence of the most important social connections in dreams. Studies that collect data from both waking life and dream life during the same period of life from the same people, as well as from the life history of these individuals, are necessarily required to test whether the representation of the most important connections is exaggerated in dreams, or if they just reflect the waking frequency. In practice, this prediction could be tested by identifying all the interactions between the dreamer and the people in his or her most important social networks, in both dream and waking reports. Then the interactions could be classified according to whether they tend to strengthen or weaken the relationship with that particular person. If the frequency with which dreaming simulates positive interactions surpasses the frequency of those interactions in real life, then the SST would gain credence over the CH.
Another potential simulation function to consider can be called the Practise and Preparation Hypothesis. According to this hypothesis, the function of social simulations in dreams is to force the dreamer to practise important social bonding skills, such as how to give social support to others. The prediction derived from this hypothesis states that if practising social bonding skills is a function of dreaming, then the dreamer should frequently offer various types of social support to other dream characters, for example emotional, instrumental, or informational support. Furthermore, the types of social support offered should be dependent on the degree of relationship intimacy, i.e., the distance between the self and the recipient in the hierarchy of the social world of the individual. If the Practise and Preparation Hypothesis is correct, then the frequency of simulating social support should be higher than comparable behaviours in real life.
These ideas are testable, but dream content studies are to be carefully designed with the specific aim of testing them. In the literature already published, friendliness percentages in different dream samples and descriptive statistics concerning who initiates friendliness in dreams might shed some light on these questions. However, without any data about the frequency of occurrence of these same behaviours in the waking state of the same person, the purely descriptive findings from dreaming alone will not be able to separate CH predictions from SST predictions. The comparable waking data is crucial as a baseline against which the dream data can be evaluated and in relation to which the CH predictions can be contrasted with the SST predictions.
In an ideal setting the hypotheses for the SST and its proposed functions would also be tested cross-culturally and in particular, as the theory makes bold evolutionary claims, in traditional small-scale human societies. As Henrich et al. (2010) have pointed out, the concentration of behavioural research into the so-called Western, Educated, Industrialized, Rich, and Democratic (WEIRD) societies are highly unrepresentative of the species, and might pose problems for the generalizability of the results. Furthermore, by contrasting, for example, the differences between the social simulations of small-scale and Western societies, we might uncover useful information about the plasticity and ontogenetic mechanisms of the social simulation function.