R1.1. The entertainment hypothesis

Using the standard framework of social evolution theory (Hamilton, Reference Hamilton1964), we first consider the point of view of the producers of fictions (the agent), before turning to the consumers of fictions (the recipient).

R1.2. Comparing the entertainment hypothesis and the information hypothesis

If fictions mostly exist because they entertain people (which brings benefits to both the producers and the consumers), then it means that the main causal factors of the contents of fiction are people's preferences. Here, the situation is very similar to the evolution of signaling in non-human animals (e.g., courtship parades, feathers, and nests). In non-human animals, the main causal factors of the content of signals are the sensory preferences of the receiver (Barrett, Reference Barrett2010; Enquist & Arak, Reference Enquist and Arak1994; Krebs & Dawkins, Reference Krebs and Dawkins1978; Lorenz, Reference Lorenz1966; Verpooten & Nelissen, Reference Verpooten and Nelissen2010). For instance, the female frog Physalaemus pustulosus had pre-existing preferences for lower-frequency chuck sounds, and then males evolved the ability to produce such sounds to exploit this sensory preference (Ryan, Fox, Wilczynski, & Rand, Reference Ryan, Fox, Wilczynski and Rand1990). In non-human animals, this recycling usually emerges by natural selection. In humans, it can emerge through cultural evolution: Producers target and refine stimuli that are already appealing to consumers. Thus, we completely agree with Lightner et al. that, when consumers have strong exploratory preferences, producers should consider investing time in the creation of imaginary worlds.

This hypothesis differs markedly with the informational hypothesis according to which fictions exist because consuming fictions leads to improving cognitive capacities, transmitting fitness-related information, or simulating real-life events (Gabora & Gomez; Goldy & Piff; Moore & Hills; Nissel & Woolley; Pianzola, Riva, Kukkonen, & Mantovani [Pianzola et al.]; Scrivner & Clasen; Scalise Sugyiama; Sitek & Konieczna; Wilbanks et al.; Wylie et al.). To take a concrete example retrieved from a commentary, Beck and Harris write that “when children read Harry Potter, they are learning about personal relationships and morality, as well as the rules of Quidditch.”

First, it seems to us that if the goal had been to teach things about personal relationships and morality, there would be more straightforward ways than inventing a whole imaginary world (and rules for an imaginary sport). More precisely, the information hypothesis fails to explain why producers of fictions invent and exaggerate stimuli to the point that they resemble less and less reality (more on the exaggeration of stimuli in sect. R2). For instance, while pedagogy may sometimes require simulations with exaggerated situations (e.g., a flight simulator should train the individuals to difficult but plausible flight situations), flying on broomsticks is a fictional feature that cannot be explained as a simulation device, because it does not train consumers to any possible situation. Yet, the existence and cultural success of Quidditch deserves a causal explanation. We argue more generally that the invention and exaggeration of stimuli in fictions can be evidence that fictions might not be suited to simulate the real world (Morin, Acerbi, & Sobchuk, Reference Morin, Acerbi and Sobchuk2019).

Second, the empirical results in favor of the information hypothesis are ambiguous. Sure, people who read fictions tend to have higher mindreading abilities, but the direction of causality is unclear: It could be that consuming fiction leads to developing mindreading abilities, which corresponds to the information hypothesis (Black & Barnes, Reference Black and Barnes2015; Castano, Reference Castano2021; Kidd & Castano, Reference Kidd and Castano2013; Zunshine, Reference Zunshine2006), or more parsimoniously, that people who are good in mindreading and like understanding other's lives are more entertained by fictional stories about the lives of fictional people, and thus are more likely to read fictions (Panero et al., Reference Panero, Weisberg, Black, Goldstein, Barnes, Brownell and Winner2016). The entertainment hypothesis makes the latter interpretation of such correlational results.

To take an example closer to our article, Scrivner, Johnson, Kjeldgaard-Christiansen, and Clasen's study (Reference Scrivner, Johnson, Kjeldgaard-Christiansen and Clasen2021) provided interesting evidence of correlations between consuming horror fictions and psychological resilience toward the COVID pandemic (Scrivner & Clasen; Wilbanks et al.). They conclude that consuming horror films benefits the consumers “through preparation and practice of both specific skills relevant to particular situations and more general skills associated with emotion regulation.” But there is another explanation, that is fully compatible with their results. It could be the case that people already more psychologically resilient to stress (e.g., because of genetic and ecological factors) would be fonder of horror movies.

To take yet another example, let's think of love stories: It seems much more parsimonious to state that more romantic people enjoy more reading romantic fictions than to hypothesize that consuming romantic fictions makes people more romantic. Therefore, we argue that, within this debate about causal effects, the burden of proof falls on those who advocate that consuming cultural artifacts has causal effects on other behavioral or cognitive traits (and not the other way around). And we argue that such causal evidence is as of now far from being convincing.

R1.3 Mixed products: Edutainment, religious myths, and folktales

So far, we have opposed the “entertainment hypothesis” and the “information hypothesis.” But they are not mutually incompatible. Several commentators observed that imaginary worlds are also present in works that are less fictional: oral traditions, folktales, myths, and religious narratives (Arnett; Dunk & Mar; Moore & Hills; Scalise Sugiyama; Sugiyama; Wiessner). We agree with them that oral traditions, folktales, myths, and religious narratives often include imaginary worlds. Yet we think that they are not produced and consumed for the same reasons, that is, to fulfill the same fitness-related goals.

For instance, religious beliefs about supernatural agents (e.g., gods, spirits, and ancestors) are probably produced by specialists (e.g., shamans) who gain in selling their services to people who believe in their supernatural capacity to communicate with supernatural agents (Boyer, Reference Boyer2020; Singh, Reference Singh2018). In the same way, religious myths about supernatural punishment are probably produced, transmitted, and supported by the members of the community because they evaluate that the threat of supernatural punishment can deter cheating and increase group solidarity (just as the same people would support a police force or a judicial system; Baumard & Chevallier, Reference Baumard and Chevallier2015; Fitouchi & Singh, Reference Fitouchi and Singh2021). Producers of fictions can also invent narratives to transmit some fitness-related information, leading to educational narratives (Scalise Sugiyama, Reference Scalise Sugiyama2011) or what is called today edutainment (Anikina & Yakimenko, Reference Anikina and Yakimenko2015; Singhal, Reference Singhal2004). Crucially, imaginary worlds invented to control others' cooperation, transmit some fitness-related information, and entertain people won't be successful for the same reasons and won't be composed of the same content features.

Of course, narratives which do not aim at entertaining can contain entertaining features, such as supernatural entities, because they make the narratives overall more attention-grabbing (Fig. R1). But the strategic incentives of the producers (e.g., to control other's cooperation) lead them to preferably include features that more specifically fulfill their goal (e.g., punishing deities; Fitouchi and Singh, Reference Fitouchi and Singh2021). By contrast, when the strategic goal of producers is to entertain other people, they preferably include features that tap into people's preferences.

Figure R1.

Examples of functions of fictional narratives related to recurrent adaptive challenges, and their possible interactions. The red area represents purely entertaining fictions (even if, locally, there can be moralistic or educative features in Harry Potter). At the intersections, we observe mixed products. For instance, Santa Claus bringing gifts to well-behaved children is undoubtedly a moralistic fictional narrative, which ultimately aims at disciplining. As a matter of fact, in most oral narratives from early modern and modern Europe (in the Holy Roman Empire), Saint Nicholas was accompanied by a foil threatening to trash disobedient children (e.g., “Knecht Ruprecht” in Germany; “Krampus” in Austria; “Parkelj” in Slovenia, Croatia, and Hungary). Dora the Explorer is at the intersection of education and entertainment because Dora directly teaches children how to speak and count. Finally, some narratives can be moralistic, educative, and entertaining, as exemplified by La Fontaine's Fables.

To sum up, we believe that the “information hypothesis” alone does not explain why humans invent fictional narratives that depart from real facts, real social events, real persons, and real settings. If information transmission was the sole (or most important) goal of producers, they wouldn't incorporate fictional entertaining contents to their narratives. Scalise Sugiyama (Reference Sugiyama, Gottschall and Wilson2005, Reference Sugiyama2021) acknowledged this is a puzzle, and proposed that facts are mixed up with invented features because such features are (1) fully recognized as such (through the use of pragmatic cues observable across cultures and in small-scale societies) and (2) attention-grabbing and memorable (providing better learning opportunities). Our general framework, in fact, agrees with this view and asks the question why and how such specific features are attention-grabbing to the human mind.

R2.1. Exploratory preferences explain the existence and the content of imaginary worlds

R2.1.1 Exploratory preferences explain the existence of imaginary worlds

Some commentators argued that our definition of an imaginary world was somehow vague (Arnett; Llewellyn; Norman & Goldstein; Salmon & Burch; Shtulman; Sobchuk). This is not because we do not want to be analytical, it is because imaginary worlds are a product of the human mind: Our definition corresponds to the “actual domain” (Sperber & Hirschfeld, Reference Sperber and Hirschfeld2004) of exploratory preferences, that is, to all the settings that trigger them (Fig. R2). This preference for exploration has evolved in all humans because exploring the local environment leads to discover new resources, new mates, or new habitats, for example, and therefore brings about fitness benefits to the individuals. We hypothesized that this preference is activated by cues indicating that the environment is unknown and that these cues are exaggerated in fictions with imaginary worlds. Such cues include landscapes that are visually different from any landscape one knows (e.g., Hyrule's landscapes in Zelda), representations of a delimitation between the known and the unknown (e.g., the walls from Attack on Titan), new location names (e.g., Hogwarts in Harry Potter), and novel world-related information (e.g., the nine-headed phoenix from The Classic of Mountains and Seas). We contend that these cues (indicating to the consumers that the fictions they are consuming are taking place in an imaginary world) trigger exploratory preferences.

Figure R2.

Exploratory preferences, with their proper domain (i.e., cues that an environment is unknown, e.g., unknown landscapes) and their actual domain (i.e., all stimuli mimicking cues that a setting is unknown, e.g., imaginary landscapes). Based on Sperber and Hirschfeld (Reference Sperber and Hirschfeld2004).

This makes the definition specific enough to distinguish imaginary worlds from other (sometimes related) stimuli. For instance, we believe that the world of Balzac is not an imaginary world. Balzac invented 3,000 characters, but his novels do not inspire the same degree of curiosity, because the world in which they take place (e.g., its geography, its functioning, and its rules) is already well known to the readers.

Note that this apparent problem of definition is not specific to imaginary worlds. We believe that the appeal of fictions all around the world is explained by the presence, in fictions, of a myriad of already appealing stimuli that producers use and exaggerate to enhance the attraction of the overall product. Such specific stimuli are appealing because they tap into specific preferences. The level of granularity of a fictional content feature (Sobchuk), therefore, depends on the specificity of the cognitive preference it is associated with. For instance, romance can be defined as a fiction based on a long-term relationship (Baumard, Huillery, & Zabro, Reference Baumard, Huillery and Zabroin press). What is a “long-term relationship”? This seems loose. It is not because it is based on the notion of pair-bonding, a specific adaptive behavior that evolved among monogamous species for which parental care is important (Fletcher, Simpson, Campbell, & Overall, Reference Fletcher, Simpson, Campbell and Overall2015).

R2.2 Exploratory preferences also predict the content features of imaginary worlds that are unrelated to exploratory preferences

Imaginary worlds are a difficult stimulus to isolate, in that respect: Producers can mix imaginary worlds with virtually any other content feature, making it even more difficult to disentangle them. Besides, some stimuli are often found associated together in fictions (more than chance would predict). For instance, imaginary worlds are often associated with dichotomic representation of good and evil. In our view, this does not mean that they constitute the same stimulus, or that they tap into the same cognitive preference. Rather, we hypothesize that some cognitive preferences are evoked by the same factors (e.g., ecological cues and the life stage), so that stimuli that activate such preferences are likely to be found in the same fictions. For instance, because adolescents are (1) reaching puberty (Del Giudice, Angeleri, & Manera, Reference Del Giudice, Angeleri and Manera2009), (2) more risk-taking (Bakerjr & Maner, Reference Bakerjr and Maner2008; Steinberg et al., Reference Steinberg, Icenogle, Shulman, Breiner, Chein, Bacchini and Takash2018), and (3) still highly explorative (Ciranka & van den Bos, Reference Ciranka and van den Bos2021), we predict that love-related stimuli, risk-related stimuli, and imaginary-world-related stimuli will tend to be associated in fictions (way more than chance would predict it), because a love story with a dangerous imaginary world would be a popular combination among a big potential audience (i.e., the adolescents). The evolutionary study of clustered features in fictions is a different research program which could be very promising in the near future. On that note, this is how we explain the bigger success of The Lord of the Rings over The Silmarillion (a puzzle rightfully highlighted by Pianzola et al.). While the “imaginary-world stimulus” may be stronger in The Silmarillion, The Lord of the Rings succeeds in combining a greater variety of appealing stimuli (e.g., Frodo's quest and Frodo's friendship with other protagonists).

R2.3 Fictions with imaginary worlds trigger several preferences, but only exploratory preferences are specific to imaginary worlds

We do not deny that fictions with imaginary worlds also tap into other preferences (see Fig. R3), but we contend that fictions with imaginary worlds require the existence of exploratory preferences. Threat detection, romantic love, and social skills, for example, can lead to the invention of all kinds of fictions, but not to the full development of imaginary worlds.

Figure R3.

Three examples of three entertaining features in fictions (cultural level) with the proximate mechanisms they co-opt (psychological level) and the ultimate functions of these mechanisms (evolutionary level), and their possible interactions.

R3.1. Changes in consumers’ preferences, not producers’ skills

The line of argument in the previous sections clears up some misunderstandings about our general hypothesis. For instance, Wolf rightfully mentions that imaginary worlds have actually existed for a long time (see also Dunk & Mar; Moore & Hills; Scalise Sugiyama; Wiessner). Yet this observation does not contradict our general hypothesis that the appeal for imaginary worlds relies on exploratory preferences and that this appeal increased as people's exploratory preferences increased. Importantly, we never argued that producers “lacked the cognitive abilities to create elaborate and inventive imaginary worlds of substantial size and with considerable amounts of world data” (we agree with Wolf that they did not lack such abilities), but rather that consumers lacked strong exploratory preferences to find such imaginary worlds entertaining. We, therefore, argue that past literary authors could (and did) produce inventive imaginary worlds, but that such fictions were relatively fewer in number, and relatively poorer in details, than in modern societies, precisely because they were less popular given the preferences of the people at the time. Let's note also that what matters for our theoretical framework is the cultural success at the time the fictional work is released, not their later success or influence.

As a matter of fact, the question is: Who consumed such fiction with imaginary worlds from earlier times? In our article, we explained why and how, in most species (including humans), the strength of exploratory preferences is linked to the level and steadiness of resources in local ecologies. At the individual level, such ecologies differ within a given society, because economic and material resources greatly vary from one family to another, with the best proxy being their economic status. Our hypothesis makes the prediction that imaginary worlds could be culturally successful in past societies, but only with richer individuals (i.e., only with a small fraction of the potential audience). Of course, this prediction is hard to test because, until recently, richer people were the only ones who could both buy books and read, and therefore the only ones who consumed literary works (Kaestle, Reference Kaestle1985; Schofield, Reference Schofield1973; Stone, Reference Stone1969), so that we can't compare their consumption of fictions with the consumption of fictions of poorer people (as Winner rightfully suggests we should do). More precisely, there might not be enough variance in the economic status of readers from the past to test our hypothesis with historical data at the individual level. However, we can test that, at the population level, when a given society gets richer, people in this society express on average stronger exploratory preferences (because the overall environment is both more secure and more affluent), and eventually become fonder of imaginary worlds. This prediction is, therefore, fully consistent with the observations that imaginary worlds culturally emerged (1) before contemporary times, with the examples of More's Utopia and Defoe's Robinson Crusoe (mentioned by Wolf), and (2) outside Western countries, with the examples of the tales of Coyote's travels (mentioned by Wiessner) and the Chinese Classic of Mountains and Seas (fourth century BC).

Finally, while Buttrick and Oishi agree with our hypothesis that consumers' exploratory preferences explain the evolution and success of imaginary worlds, they challenge the hypothesis that the variability of such preferences can be ultimately explained with adaptive phenotypic and developmental plasticity. However, the disagreement does not seem that deep: We agree with the commentators that our drive to explore tracks proximate socio-ecological cues and adapt to them. And we agree that the motivation to go sightseeing and visiting foreign places should be correlated with the motivation to consume fictions with imaginary worlds (both behaviors being driven by exploratory preferences). However, we disagree with the hypothesis that the preference (or possibility) to move causes the preference for imaginary worlds. According to our theoretical framework, both are effects of a cognitive preference for spatial exploration, which adaptively varies according to ecological conditions. Both hypotheses make more or less similar diachronic predictions about correlations between different variables (e.g., the evolution of the rate of tourism and the evolution of the prevalence of imaginary worlds in fictions). However, our evolutionary hypothesis makes at least two predictions that their socioecological hypothesis doesn't: (1) that, synchronically, in the same society, people with higher socio-economic status are more exploratory and hence consume more fictions with imaginary worlds than people with lower socio-economic status, even if they don't travel more (a variable that can be controlled for); and (2) that children are more explorative and hence consume more fictions with imaginary worlds (even if they don't travel; see sect. R3.2). Only further empirical research about the variability of fiction consumption across modern population can settle this debate.

R3.2. Children's preferences, not their abilities

Now that our framework is better defined, we can address a point of divergence between our general hypothesis and commentaries from developmental psychologists (Beck & Harris; Norman & Goldstein; Nyhout & Lee; Weisberg & Sobel). According to us, this point of divergence stems from the same misunderstanding that we reviewed in the latter section: They consider (more or less explicitly) that children's skills should somehow drive the cultural evolution of children fictions. It is all the more important to clarify this point as we believe that developmental predictions derived from our framework could be both highly specific and straightforwardly testable. As we argue elsewhere (Dubourg & Baumard, Reference Dubourg and Baumardin press), this framework (bringing together cultural attraction theory and adaptive developmental plasticity) could explain the presence and absence of a myriad of content features in fictions, by considering the age of their targeted audience. Why? Each human life stage from infancy to old age, and including childhood, juvenility, adolescence, and adulthood, is endowed with age-specific preferences (Bjorklund & Pellegrini, Reference Bjorklund and Pellegrini2000; Del Giudice et al., Reference Del Giudice, Angeleri and Manera2009), because natural selection has favored individuals who are able to adopt an optimal scheduling of preferences (Gangestad & Kaplan, Reference Gangestad, Kaplan, Buss and Giudice2015; Hill, Reference Hill1993; Kaplan & Gangestad, Reference Kaplan, Gangestad and Buss2005). Because our general framework suggests that the cultural evolution of fictions is driven by the consumers' preferences, we can derive a series of predictions about the distribution of content features in fictional narratives by considering the age of the people who find them entertaining (Dubourg & Baumard, Reference Dubourg and Baumardin press).

Coming back to the specific case of the appeal for imaginary worlds, we maintain that our hypothesis doesn't make any prediction about “how children imagine” (Weisberg & Sobel) or about whether children are “imaginative” or not (Beck & Harris; Norman & Goldstein; Nyhout & Lee; Weisberg & Sobel). We don't claim that “children appear to be both highly exploratory and highly imaginative” (Beck & Harris) but rather that children appear to be highly exploratory and hence highly receptive to content features triggering exploratory preferences in fictions. This is why we make the prediction that, if imaginary worlds do co-opt exploratory preferences, they should be popular among children. Studies reported in the commentaries never test the specific predictions that children enjoy imaginary worlds more than adults do. For instance, Barnes, Berstein, and Bloom's (Reference Barnes, Bernstein and Bloom2015) seminal study and the following related studies (e.g., Taggart, Heise, & Lillard, Reference Taggart, Heise and Lillard2018) show that the attraction toward fiction varies with age, with younger people being less drawn to fictional over realistic narratives (or activities). This doesn't go against our theory, which would rather state that when children actually decide to consume fictions, their cognitive preferences (such as their exploratory preferences) drive what they like and want to consume. Nyhout and O'Neill’ (Reference Nyhout and O'Neill2017) study shows that following the characters' movements in a story can be difficult for children. Likewise, this is no evidence against our hypothesis because the study focuses on children's abilities, not children's preferences. This is, we argue, one major issue that prevents us from using results from such otherwise important studies to investigate children's specific preferences.

We could not agree more with the limitations highlighted in the commentaries with regard to the possible ways to empirically test developmental predictions derived from our hypothesis. Studying children's actual preferences with fictional content is difficult. Parents influence children's fiction consumption (Nyhout & Lee; Weisberg & Sobel), and further research should investigate to what extent before drawing any conclusion from the analysis of consumption data. Besides, market data (1) often don't include data on children's consumption (Nyhout & Lee), (2) never include indicators of success such as ratings by children, and (3) are hard to find in developing countries and in small-scale societies (Norman & Goldstein). Only laboratory research with children could allow studying actual children preferences, and to test more specifically the prediction that children are actually fond of imaginary worlds when they are consuming fictional stories.