1 Introduction

Humans constantly interact with other people and are considered among the most social species (Gariepy et al., Reference Gariepy, Watson and Du2014). They not only live socially but think socially and spontaneously explain and predict other people’s behaviors. They possess a Theory of Mind (ToM), which is a sophisticated cognitive ability that concerns the attribution to others of a wide range of mental states, including intentions, desires, knowledge, true and false beliefs, and emotions (Wellman, Reference Wellman2014). Sometimes called intuitive psychology, naïve or folk psychology, mentalizing, or mindreading, ToM is a foundational social cognitive ability. Understanding that others have mental states different from one’s own makes it possible to infer what others are thinking and to predict their behavior. ToM has roots in philosophy, particularly in the groundwork for a science of the mind laid down by the philosopher René Descartes. In psychology, the first scholar to study this cognitive ability was Jean Piaget, who suggested that before the age of 3 or 4 years egocentrism prevents children from understanding that other people’s thoughts and viewpoints may differ from their own (Piaget & Inhelder, Reference Piaget and Inhelder1967). But the modern impetus for ToM came from Flavell’s (Reference Flavell1968) work on “perceptual role taking” – the ability to take another’s visual perspectives – and “conceptual role taking” – the ability to take another’s mental perspective, to impute another’s knowledge or intentions. Similarly, Selman (Reference Selman1971) explored the development of conceptual role-taking and reported that conceptual role-taking is an age-related, developing social-cognitive skill.

Premack and Woodruff (Reference Premack and Woodruff1978) had coined the term ToM as it is now used in developmental science. They argued that chimpanzees and perhaps other nonhuman primates could understand intentions, suggesting that chimpanzees possessed a ToM. Although the depth of nonhuman primates’ Theory of Mind remains controversial, Premack and Woodruff’s work inspired psychologists to document Theory of Mind development in typical and atypical populations. An edited book of papers from talks presented at two conferences (Astington et al., Reference Astington, Harris and Olson1988) followed by three other books (Astington, Reference Astington1994; Perner, Reference Perner1991; Wellman, Reference Wellman1990) launched a widespread interest in ToM. To fully understand any aspect of cognitive development requires documenting its origins and change. So, over the years, research on ToM has expanded in both depth and breadth (for a history of ToM research, see Wellman, Reference Wellman2017).

Wimmer and Perner (Reference Wimmer and Perner1983) documented the development of metarepresentation (e.g., a belief about someone’s belief) with a false belief task. In the standard version of this test, a puppet named Maxi puts a chocolate in a cupboard and leaves the scene. The experimenter moves the chocolate to a new location and asks the child where Maxi will look for it. In another version called the “Sally-Anne” task, Sally and Anne put a marble in a box. Sally leaves, and Anne takes the marble out of the original box and puts it in a different box (see Figure 1). Up until about the age of 4, most children ascribe their own beliefs to Maxi and Sally and tell the experimenter that Maxi or Sally will look for the chocolate or marble in its new location. However, older children understand that Maxi will look for the chocolate where he last saw it in the cupboard, and Sally will look in the first box. The task requires that children understand that other people’s beliefs may differ from their own and then predict how those people will react based on their differing beliefs.

Figure 1 A depiction of the Sally-Anne task of false belief reasoning developed by Wimmer and Perner (Reference Wimmer and Perner1983).

Figure reproduced from Birch et al. (Reference Birch, Li, Haddock and Benson2017) with permission.

Since the landmark paper by Wimmer and Perner, ToM has become among the most studied topics in cognitive developmental science, with numerous books and thousands of research articles published on the subject. We now have a good grasp on this foundational cognitive ability and its development from research across many disciplines. The main contributions of developmental scientists have been to document when ToM develops and the factors that account for developmental changes in understanding a range of mental states. For example, the impact of ToM development on the development of language and other social cognitive skills, such as executive functions, has been documented (Wade et al., Reference Wade, Prime and Jenkins2018). Comparative psychologists have examined which mental states are uniquely understood by humans and which are also understood by other species (Krupenye & Call, Reference Krupenye and Call2019). Clinicians studying autism and other developmental disorders have also examined ToM impairments (Senju, Reference Senju2012). Finally, neuroscientists have investigated if there is a distinct and domain-specific neural system for thinking about the mind (Mahy et al., Reference Mahy, Moses and Pfeifer2014).

This Element on Theory of Mind is focused on the developmental period from infancy to the preschool years and aims to provide answers to four key questions about early ToM development: What? When? Why? and How? What are the typical steps in the development of ToM from infancy to age 5? What external and internal factors to the child influence the rate of ToM development? What are the consequences of Theory of Mind development for children’s social competence and academic achievement? When does ToM emerge? Why is it important to study ToM? And how does ToM develop? To answer these questions, I first review milestones in the development of ToM and factors that contribute to its emergence. Then, I discuss the state of science about the emergence of an automatic, nonverbal, implicit form of ToM in human infants as well as research on the neurological correlates of ToM. I conclude with suggestions for future research in this exciting area of developmental science.

The original focus on false belief reasoning, a key milestone in early ToM, was followed by research on the multiple steps leading to that sophisticated ability. The emergence of pretend play during the second year of life is one of the first pieces of evidence that children are aware of the difference between what the mind represents and the world outside the mind (Leslie, Reference Leslie1987). For example, when pretending that a stick is a phone, children show that they can distinguish between an object – the stick – and thoughts about the object – the stick as a phone (Kavanaugh, Reference Kavanaugh, Balter and Tamis-LeMonda2006). By 18 months of age, they also understand that people will display happiness when their desires are fulfilled and sadness if their desires are not fulfilled (Chiarella & Poulin-Dubois, Reference Chiarella and Poulin‐Dubois2013; Wellman & Banerjee, Reference Wellman and Banerjee1991). During the second year, children begin to grasp that there may be a difference between what they want and what another person wants (Meltzoff et al., Reference Meltzoff, Gopnik, Repacholi, Zelazo, Astington and Olson1999). Such awareness is reflected in children’s language as toddlers talk about what they and others want, like, and feel; when they are 3, they also talk about what other people think and know (Bartsch & Wellman, Reference Bartsch and Wellman1995).

As such, ToM develops gradually as a function of the complexity of mental states (e.g., from simple desires to false belief). Wellman and Liu (Reference Wellman and Liu2004) documented the gradual development of ToM by developing a step-like scale that reflects a fixed sequence in understanding mental states: diverse desires, diverse beliefs, knowledge access, false beliefs, and hidden emotion (see Figure 2). Each step was assessed with a simple scenario about a character (a human figurine), using props such as boxes and drawings of scenes, and questions asked during warm-up, test, and control phases. In several converging studies conducted worldwide with adaptations in multiple languages, a typical order of difficulty has been reported, with diverse desires being the easiest and understanding hidden emotions being the hardest to understand/produce (Wellman et al., Reference Wellman, Fang, Liu, Zhu and Liu2006, Reference Wellman, Fang and Peterson2011; Wellman et al., Reference Wellman, Lopez-Duran, LaBounty and Hamilton2008). This ToM scale has also been validated longitudinally by testing the same children between the ages of 3 and 6 years (Wellman et al., Reference Wellman, Fang and Peterson2011). Children also develop metacognitive skills about the mind during that same developmental period. Bialecka and her colleagues (Reference Białecka, Gut, Stępień-Nycz, Macheta and Janczura2024) developed a measure of children’s knowledge about the mind and thinking processes between the ages of 3 and 7 years. Children initially grasp that mental states do not possess physical attributes; that is, children understand that we do not have visual, auditory, or tactile access to thoughts. This knowledge is typically achieved by most 4-year-olds. Subsequently, children become aware of the internal localization of thoughts. The third milestone concerns the source of thoughts, with approximately half of 6-year-olds achieving such understanding. The final step in this developmental sequence involves the ability to define thoughts and the thinking process.

Figure 2 Wellman and Liu scale (Reference Wellman and Liu2004).

Figure reproduced from Wellman (Reference Wellman2014) with permission.

Nonetheless, some variations have been observed in the progression of these steps as a function of sociocultural differences (Peterson & Slaughter, Reference Peterson, Slaughter, Slaughter and de Rosnay2017; Slaughter & Perez-Zapata, Reference Slaughter and Perez‐Zapata2014). For example, both Chinese and Iranian children develop an understanding of knowledge access before any understanding of belief (Shahaeian et al., Reference Shahaeian, Peterson, Slaughter and Wellman2011; Wellman et al., Reference Wellman, Fang, Liu, Zhu and Liu2006). In the section on the impact of culture on ToM development, I discuss underlying mechanisms that might explain this deviation from the sequence observed in most linguistic and cultural groups. Finally, there are delays in addition to deviations to the scale in certain populations, such as among deaf children born to hearing parents or children with autism. For example, deaf children progress through the steps in the same sequence, but it takes them many years to progress through the scale in comparison to hearing children (Peterson et al., Reference Peterson, Park and Seligman2005; Peterson & Wellman, Reference Peterson and Wellman2009; Remmel & Peters, Reference Remmel and Peters2008). In contrast, children with autism deviate from the typical sequence by passing hidden emotions before false beliefs (Peterson et al., Reference Peterson, Wellman and Slaughter2012). Neurotypical children vary in their ability to reason about others’ mind, and these individual differences are stable over time and across tasks (Hughes & Devine, Reference Hughes and Devine2015).

Understanding false belief progresses beyond the forms typically mastered during the preschool years. A more advanced development is second-order false belief: the realization that it is possible to hold a false belief about someone else’s belief, also called recursive mentalistic reasoning (Miller, Reference Miller2009, Reference Miller2012). To illustrate how first- and second-order false beliefs differ, consider Wimmer and Perner’s (Reference Wimmer and Perner1983) Maxi scenario. The target for the belief could be the person who moved the chocolate in Maxi’s absence, and the question is where this person thinks that Maxi thinks the chocolate is. The attribution of such a belief constitutes second-order reasoning: “A thinks that B believes … .” By age 6, children pass second-order false belief tests (Miller, Reference Miller2009). What new developments are made possible by the mastery of second-order belief? A number of social and cognitive consequences have been documented (lies, jokes, irony, etc.) (see Miller, Reference Miller2009).

2 Do Infants Have a Theory of Mind?

All children’s cognitive abilities (e.g., memory, categorization) are developmental achievements that have precursors. Theory of Mind is no exception. However, the question is how rich the ToM that emerges during infancy is. As pointed out earlier, Leslie (Reference Leslie1987) proposed that the first manifestation of a capacity for metarepresentation emerges during the second year of life with pretend play. Since then, researchers have documented what specific mental states (e.g., desire, intention, belief) are understood in infancy. For some researchers, who adopt a rich, mentalistic view of early ToM, the time lag until success on standard ToM tasks can be accounted for by the complexity of the inferential process required to attribute mental states that differ from one’s own as well as by the heavy cognitive load involved in standard tasks. For others, who hold a minimalist account, attribution of mental states is beyond infants’ abilities.

2.2 Implicit False Belief Concept

A milestone in research on implicit ToM came from the violation-of-expectation (VOE) paradigm with looking time as the dependent variable to test if belief understanding develops in infancy. Poulin-Dubois et al. (Reference Poulin-Dubois, Frank, Graham and Elkin1999) (see also Sodian et al., Reference Sodian, Thoermer and Metz2007) used the preferential-looking paradigm to test infants’ concepts of knowledge versus ignorance. In one study, they reported that 18-month-old infants understand that an agent who wears a blindfold when the content of two boxes is revealed will not be cognizant of the location of an object when someone else asks for that information. This was revealed by the fact that infants looked longer when the agent pointed at the correct location. Poulin-Dubois (Reference Poulin-Dubois1999) also observed that infants looked longer at a protagonist who showed a surprised facial expression rather than a neutral expression when an object was no longer found where it was supposed to be. But Onishi and Baillargeon (Reference Onishi and Baillargeon2005) demonstrated that infants have an understanding of both true and false beliefs by using the VOE paradigm, designed initially to test infants’ naïve understanding of physical principles such as support and gravity (Kellman & Spelke, Reference Kellman and Spelke1983). The procedure required that infants first observe an agent hide an object inside one of two boxes and then reach for it repeatedly in a familiarization phase. An induction phase followed, during which the object moved from the original location to another location (the other box). The agent either witnessed the change (true belief condition) or disappeared behind a screen (false belief condition) during that critical event. During the test phase, the agent searched for the object either in the original location (congruent condition) or in the new location (incongruent condition; see Figure 3). Infants looked longer at a scene when the agent’s behavior was incongruent with their true or false belief about a toy’s location.

Figure 3 Depiction of the violation-of-expectation paradigm developed by Onishi and Baillargeon (Reference Onishi and Baillargeon2005).

Figure reproduced from Wellman (Reference Wellman2014) with permission.

According to Onishi and Baillargeon, these results suggested that infants expected the agent to behave according to where the agent believed the toy to be hidden and not where the toy was actually hidden, indicating belief understanding. Following this landmark study, other tasks based on spontaneous responses (as opposed to elicited responses) have been designed to test false belief in preverbal infants. For example, Southgate and colleagues (Reference Southgate, Senju and Csibra2007) showed that 25-month-olds correctly performed anticipatory-looking behavior in a nonverbal false belief test, a procedure similar to that developed by Clements and Perner (Reference Clements and Perner1994) except that no interview was taking place. Another type of implicit measure of Theory of Mind is the interactive paradigm in which infants must assist an agent to find an object based on their understanding of the agent’s knowledge, or lack thereof. For example, when an agent falsely believes that a desired object is in location x, 18-month-old infants spontaneously point to inform the agent that the object has been moved to another location or has been replaced with an aversive object (anticipatory-pointing tasks; Knudsen & Liszkowski, Reference Knudsen and Liszkowski2011). At the same age, infants will help an agent retrieve an object that is no longer where the agent thinks it is (Buttelmann et al., Reference Buttelmann, Carpenter and Tomasello2009). Toddlers also express more tension in their facial expressions when an agent who is approaching a container is mistaken, as opposed to ignorant, about its contents (affective-response tasks; Moll et al., Reference Moll, Khalulyan and Moffett2016). In elicited-intervention tasks, children watch a scene in which an agent forms a false belief and then are asked to perform some action for the agent. Correct performance requires children to take into account the agent’s false belief (for a review, see Scott & Baillargeon, Reference Scott and Baillargeon2017).

The road to uncovering infants’ cognitive abilities has been paved by waves of methodological advances. The study of infants’ Theory of Mind is no exception. However, measurement of early ToM has been surrounded with controversy. For example, infancy researchers have often used artificial stimuli (e.g., geometric figures) to study infants’ understanding of mental states because of the experimental control that comes with them. However, simplified, artificial stimuli have led to questions about the validity of the empirical findings. The issue is whether many tasks have sufficient construct validity, more specifically measurement (or internal) validity, which refers to whether observed results are generated by infants’ understanding of the construct that the researchers intended to study (Kominsky et al., Reference Kominsky2022). Moreover, low measurement reliability, defined as the precision or the consistency of a specific instrument when a measurement is repeated, has plagued research on ToM in infancy when individual differences are investigated (see Byers-Heinlein et al., Reference Byers-Heinlein, Bergmann and Savalei2021, for solutions for more reliable infant research).

2.4 Infants’ Theory of Mind: The Replication Crisis

The field of psychology, including developmental science, faces a replication crisis. Spearheaded by an informal compilation of published and unpublished non-replications (Kulke & Rakoczy, Reference Kulke and Rakoczy2018), a special issue of the journal Cognitive Development was devoted to ToM research. It included several studies that reported failed attempts to replicate three tasks which were designed to measure implicit false belief and other ToM constructs in infancy (Sabbagh & Paulus, Reference Sabbagh and Paulus2018). Some papers did not replicate the interactive, helping task, others did not replicate the well-known VOE task, and others failed to replicate the anticipatory-looking paradigm. Even the famous “broccoli” task assessing the understanding of subjective desires was among the failed replications (Ruffman et al., Reference Ruffman, Aitken, Wilson and Taumoepeau2018).

In their invited commentary, the authors of the original experiments (Baillargeon, Buttelmann and Southgate) identified a number of procedural differences between the original studies and the replications that could explain the failures to replicate their original findings (Baillargeon et al., Reference Baillargeon, Buttelmann and Southgate2018). Similar arguments were addressed in the Open Collaboration project at the origins of the replication crisis in psychology (Gilbert et al., Reference Gilbert, King, Pettigrew and Wilson2016). In their reply to this commentary, the authors of the papers included in the special issue acknowledged the methodological argument but also pointed out that the original findings might be false positives and that there is no justification to give them priority over later replication failures (Poulin-Dubois et al., Reference Poulin-Dubois, Hastings and Chiarella2018). They also pointed out that failed replications are informative as demonstrations of the fragility of some of the original findings. Soon after the publication of the special issue, 56 false belief conditions using the VOE, anticipatory looking, and interactive paradigms were included in a meta-analysis that included 1469 infants. Among the main findings, this meta-analysis revealed that, although infants performed these tasks above chance level, particularly VOE, there was much variability in the results (Barone et al., Reference Barone, Corradi and Gomila2019). Results showed that the VOE paradigm had a significant mean effect size, β = 1.28, 95% CI [0.86; 1.69], which indicates that correct performance on the task was about 3.58 times more likely than incorrect performance. A significant mean effect size, β = 0.31, 95% CI [0.01; 0.61], was reported for the interactive paradigm, suggesting that correct performance on the task was about 1.36 times more likely than incorrect performance. By contrast, the meta-analysis for the anticipatory looking paradigm provided no significant mean effect size, β = 0.16, 95% CL [−0.24; 0.57]. The authors of this meta-analysis also reported a possible publication bias in this body of work and less positive findings in studies that have tested larger samples and included better control conditions. In fact, large effect sizes were found in early studies on implicit false belief attribution, but this outcome could not be replicated in more recent experiments. Relatedly, those large effect sizes were found with small samples, whereas later studies used bigger samples.

As part of the massive effort to replicate and expand studies providing the foundation for a rich ToM in infancy, conceptual replications have been conducted with the original VOE paradigm. The findings from these studies have forced a revision of the rich mentalistic account. For example, Burnside and colleagues (Reference Burnside, Severdija and Poulin-Dubois2020) conducted a study that was a conceptual replication of the VOE procedure, with a human agent replaced by an inanimate agent, a toy mechanical crane (see Figure 4). Infants watched the toy crane repeatedly move toward a box containing an object. In the absence of the crane, the object changed location. When the crane returned, 16-month-old infants looked longer when it turned toward the object’s new location, consistent with the attribution of a false belief to the crane. These results suggest that infants spontaneously attribute false beliefs to inanimate agents, unlike adults who correctly expected the crane to “search” for the object in the old location. Thus, infants overattribute mental states (if this is what they do in the VOE paradigm) to any agents.

Figure 4 Depiction of the test events in the VOE task with a mechanical toy crane as the agent.

(figure reproduced from Burnside et al., Reference Burnside, Severdija and Poulin-Dubois2020, with permission)

Similarly, a conceptual replication of the classic false belief VOE task based on a switch agent paradigm revealed that 16-month-old infants attributed true and false beliefs to a naïve agent, an agent who had not been present during the familiarization phase and thus not cognizant of the original location of the object (Burnside et al., 2020). Therefore, they did not perceive beliefs as person-specific. Again, these findings indicate that the mechanisms that underlie infants’ implicit attribution of beliefs might differ from those assumed for explicit reasoning about beliefs. They challenge the view that developmental changes in false belief understanding may be described as strengthening and enriching initial adult-like representations.

3 Correlates of Theory of Mind

Many researchers assume that the development of false belief understanding during the preschool period is not an artifact of the task demands but rather a reflection of a genuine conceptual change that occurs during the preschool years (Wellman et al., Reference Wellman, Cross and Watson2001). If this is the case, then the question is how is such a progression explained? One hypothesis is that ToM achievements, from desires to second-order beliefs, reflect the gradual maturation of a ToM module, that is, an innate neural structure that has been honed and selected through evolution. Another hypothesis is that ToM development is the result of social experiences that vary across children. Decades of research have revealed that there is variability in the particular age at which, for example, individual children achieve success on false belief tasks, at 3 years of age for some children and not until 5 or 6 years for others. The fact that there is such individual variation in development has led researchers to explore possible correlates of false belief understanding in cognitive and social variables in both typically developing and clinical populations (Repacholi & Slaughter, Reference Repacholi and Slaughter2003). Thus, a substantial body of research has accumulated about the determinants of ToM development, ranging from social to cognitive variables (Jenkins & Astington, Reference Jenkins and Astington1996). The following sections attempt to cover the key findings in this vast domain of research, including the impact of language, family factors, SES, language, executive functions, and bilingualism. It is important to point out that most correlates of ToM are moderate in strength and are often based on small sample sizes. Meta-analyses on this topic have highlighted the limited number of studies using cross-lagged designs and interventions to test causal links between ToM and various correlates (Devine & Hughes, Reference Devine and Hughes2014, Reference Devine and Hughes2018).

3.4 Culture

Culture refers to a group’s set shared values, practices, and beliefs. Culture is an overarching factor that is likely to impact most social forces reviewed so far. For example, cultural values and beliefs likely impact whether parents engage in mental talk and mind-mindedness. As Peterson and Slaughter (Reference Peterson, Slaughter, Slaughter and de Rosnay2017) pointed out, studying potential cross-cultural differences in children’s ToM can have important theoretical implications: Universality of ToM development would provide support for biological maturation as the key developmental process, whereas variability across cultural groups would provide support for the role of social factors. The bulk of cross-cultural research on ToM development has focused on two key aspects of ToM development: One concerns the universality of ToM constructs (e.g., is there a culture where false belief does not develop?), and the other is the timing of ToM development (e.g., do children in different cultural groups acquire ToM at different times in development?). The first issue was examined by Avis and Harris (Reference Avis and Harris1991), who studied the timing of false belief understanding (location) in British versus West African preschoolers. Their results provided support for the claim that belief-desire reasoning develop similarly in both groups, in terms of both content and timing. Callaghan and colleagues (Reference Callaghan, Rochat and Lillard2005) also observed synchrony in the onset of mentalistic reasoning, with children mastering a standardized test of false belief at approximately 5 years of age in five distinct cultural groups (India, Peru, Samoa, Canada, and Thailand). Regarding the issue of timing, in their meta-analysis, Wellman and colleagues (Reference Wellman, Cross and Watson2001) observed some variation in children from the Europe, North America, South America, East Asia, Australia, and Africa. Specifically, both Japanese and Peruvian children were slower to master false beliefs. In another meta-analysis, Liu and colleagues (Reference Liu, Wellman, Tardif and Sabbagh2008) compared false belief performance in four locations (the United States, Canada, Hong Kong, and Beijing) that included several thousand children. Unexpectedly, as shown in Figure 5, Chinese children from Hong Kong were approximately 2 years slower than Canadian and American children despite sharing many aspects of their culture with the Chinese children. The authors offered two plausible hypotheses about how social factors might explain such differences in ToM timing. First, the Hong Kong children were more likely than the Beijing children to have siblings. As discussed earlier, having siblings predicts faster ToM mastery. A second possibility is that the Hong Kong children were more likely to be bilingual than the other samples, and there is some evidence that bilingualism is linked to ToM. In their review, Slaughter and Perez-Zapata (Reference Slaughter and Perez‐Zapata2014) concluded that the research indicates that in all cultures studied, individuals begin to attribute false beliefs to others in childhood but that there are cultural differences in the age at which children acquire this skill. It is not yet clear what accounts for these differences. Because most studies do not systematically measure variables such as bilingualism and family size, one or more of them, rather than culture, may account for some of the cross-cultural differences.

Figure 5 Developmental trajectory of false belief reasoning in five countries.

(figure reproduced from Liu et al., Reference Liu, Wellman, Tardif and Sabbagh2008, with permission)

All the studies reviewed so far on culture and ToM focused on false belief reasoning, considered a marker of ToM. A comprehensive test of cross-cultural differences requires a similar investigation of the sequence of developmental milestones as documented by Wellman and Liu (Reference Wellman and Liu2004) in their five-step scale of ToM acquisition. Does the same developmental sequence apply to children growing up in different socio-cultural contexts? Wellman and colleagues (Reference Wellman, Fang, Liu, Zhu and Liu2006) compared Chinese children with Australian and US American children. All children passed the ToM steps in the same sequence, but a cultural difference was observed in the order of acquisition of the tasks within the scale, with the Chinese children mastering the knowledge access and the diverse belief items in reversed order from Western children. Additional evidence of cross-cultural variation in the order of the ToM scale concepts was observed when Iranian children were compared to Australian children (Shahaeian et al., Reference Shahaeian, Peterson, Slaughter and Wellman2011). Similarly to the pattern of results reported for Chinese children, Iranian children passed the diverse beliefs item before the knowledge access item despite an overall rate of ToM mastery that was identical across the two groups. A total of 10 studies have been published using the ToM scale to examine the order of emergence of ToM concepts in non-Western children (see Peterson & Slaughter, Reference Peterson, Slaughter, Slaughter and de Rosnay2017, for a review). The general conclusion of this body of work is that there is cultural variation if only false belief scores are considered. However, there is substantial cross-cultural uniformity when the overall sequence of ToM is considered, except for the ordering of knowledge access and diverse belief (Yu and Wellman, Reference Yu and Wellman2024). Researchers have also reported slight delays in the ages at which children succeed at false belief tasks in collectivist Asian societies such as Japan (Wellman et al., Reference Wellman, Cross and Watson2001), Hong-Kong (Liu et al., Reference Liu, Wellman, Tardif and Sabbagh2008), and Pacific Island countries (e.g., Samoa; Mayer & Träuble, Reference Mayer and Träuble2013). These findings have been attributed primarily to children’s low levels of exposure to different beliefs because collectivist societies value mutual agreement, group harmony, and cohesion, which require believing in the same thing or acting in the same way.

In summary, this research shows that cultural norms and children’s social experiences, which are guided by those norms, impact the age of acquisition and order of different mental state understanding. Both universal (e.g., the timing of false belief understanding) and socioculturally specific (e.g., variability in ToM sequences) findings are reported with respect to children’s developmental timing and progression of ToM (Selcuk et al., Reference Selcuk, Gonultas and Ekerim‐Akbulut2023).

4 Theoretical Accounts of Theory of Mind Development

Theoretical accounts of cognitive development typically fall into two broad categories: nativist and empiricist. Other ways to carve up theoretical accounts are domain-specific versus domain-general and one-system versus two-system accounts. The development of social cognition in general, and of ToM in particular, is also explained by such general frameworks. One of the first theoretical accounts of ToM was a nativist view based on research with children with ASD. According to that view, ToM has a specific innate basis in that the processes that determine the essential character of ToM do not apply to other cognitive domains and can be selectively impaired (Scholl & Leslie, Reference Scholl and Leslie1999). Thus, the rapid, universal development of other people’s mind understanding depends on an innate, encapsulated, and domain-specific part of the cognitive architecture, a Theory of Mind module (ToMM) (Baron-Cohen et al., Reference Baron-Cohen, Leslie and Frith1985; Carruthers, Reference Carruthers2002; Leslie, Reference Leslie1994). This view inspired researchers to conduct studies on children with autism who are high functioning to test if they display a ToM domain-specific delay or deficit (Baron-Cohen, Reference Baron-Cohen1995, Reference Baron-Cohen2000). For example, Poulin-Dubois and colleagues (Reference Poulin-Dubois, Dutemple and Burnside2021) compared children with ASD to neurotypical children on tasks measuring naïve psychology, naïve physics, and naïve biology. Only children with ASD underperformed on an implicit false belief task (anticipatory looking paradigm), with performances in naïve biology and physics equivalent across the two groups and uncorrelated to performance on the false belief task. This result confirmed that naïve physics and biological reasoning are intact in children with ASD but that tracking false beliefs is challenging for this population. This view predicts that the acquisition of ToM should be largely uniform across individuals and cultures, although the timing of the acquisition might vary.

Another form of nativist account that is not based on modularity is the rich, mentalistic account of ToM development, which argues that false belief understanding emerges in infancy. For example, Baillargeon and colleagues (Reference Baillargeon, Scott and He2010) favored a rich, mentalistic interpretation (see also Setoh et al., Reference Setoh, Scott and Baillargeon2016) and claimed that false belief understanding is already present by the second year of life but is masked in traditional explicit tasks due to a number of executive task demands. First, children may lack sufficient skills at one of the processes involved in the task; for example, they may lack sufficient inhibitory-control skills to suppress their own knowledge or resist what is called the curse-of-knowledge – when knowledge of an event’s outcome compromises the ability to reason about another person’s beliefs about that event. Second, children may be able to execute each process separately but lack sufficient information-processing resources to handle the total concurrent processing demands of the task. This interpretation is supported by studies showing that performance on traditional false belied tasks (e.g., unexpected location change or content) improves when processing demands are reduced, even in toddlers (Scott, Reference Scott2017; Scott & Baillargeon, Reference Scott and Baillargeon2017; Setoh et al., Reference Setoh, Scott and Baillargeon2016). Similar to the mentalistic account of ToM development is the developmental enrichment account, which assumes that implicit false belief understanding develops into an explicit understanding through developing language and executive functions abilities. This account is supported by the link between implicit and explicit false belief understanding that has been reported in some longitudinal studies (Sodian et al., Reference Sodian, Kristen‐Antonow and Kloo2020).

Theoretical accounts of ToM development that focus on conceptual changes fall under the “constructivist” perspective. A theoretical account of ToM development that stands in contrast to the nativist modular view is so-called theory theory (Wellman, Reference Wellman1990, Reference Wellman2014). According to this constructivist, domain-specific view, scientific theories carve the world into different domains of knowledge, as is the case for naïve or folk theories, such as naïve psychology and naïve physics. The theory “predicts the three empirical signatures of constructivist learning: a) learning progresses in orderly conceptual steps, b) both the sequence and timetable of these steps depend on experience, c) prior conceptual knowledge constraints the development of each new step” (Wellman, Reference Wellman2014, p. 159). Another early account of ToM development that is also constructivist and domain-specific is the simulation theory (Goldman, Reference Goldman1992; Harris, Reference Harris1992, Reference Harris2000). According to the simulation approach, we have privileged access to our own mental states. This approach suggests that we adopt the perspective of others through simulation, hypothesizing their mental states based on our own experiences. For example, we may predict how someone else will respond to an emotion-arousing stimulus by imagining how we ourselves would respond. Or we may make sense of someone else’s behavior by imagining what would lead us to behave in a similar way. Clearly, such ability depends on a capacity for imaginative pretense. Fortunately, in simulation theory such a capacity develops early in development – even 2- and 3-year-olds are remarkably skilled at pretense (Harris, Reference Harris1992). However, pretense abilities improve with age, and such improvements contribute to developmental changes in ToM. To explain the sequence in the development of mental states, such as shown by the ToM scale, it is proposed that there are levels of difficulty in carrying out simulations for different mental states. Thus, some developments (e.g., reasoning about beliefs) are slower to emerge in childhood than others (e.g., reasoning about desires). The link between social pretense and performance on ToM tasks has been confirmed (Astington & Jenkins, Reference Astington and Jenkins1995; Lohman & Tomasello, Reference Lohmann and Tomasello2003). Additional evidence for this view comes from analyses of mental state language which show that children almost always make assertions about want, know, and think that are first focused on themselves (Harris et al., Reference Harris, Yang and Cui2017).

Other theoretical views that can be classified as constructivist and psychosocial (à la Bronfenbrenner) also posit that a fundamental change occurs in ToM development around 4 years of age. Implicit false belief tasks may be passed by reliance on behavior rules (Perner & Ruffman, Reference Perner and Ruffman2005) or other low-level processes (perceptual novelty; Heyes, Reference Heyes2014; statistical learning; Ruffman, Reference Ruffman2014). In line with the bioecological theoretical framework, a full-fledged Theory of Mind emerges out of interactions with the proximal and distal social environment, with exposure to mental state talk and regularities in people’s behavior as precursors (Ruffman, Reference Ruffman2023). Only a few studies have directly tested these minimalist accounts of the evidence for early forms of ToM skills. As discussed earlier, in a study attempting to directly test the perceptual novelty account, Zmyj and colleagues (Reference Zmyj, Prinz and Daum2015) found no link between a working memory test (A-not-B task) and false belief understanding (anticipatory looking task). Yott and Poulin-Dubois (Reference Yott and Poulin‐Dubois2012) tested whether teaching infants a new rule might affect their reasoning on a false belief task. They taught 18-month-old infants over 8 trials that an object placed in one container always reappeared in a different container. They then examined whether learning this new rule affected infants’ reasoning on a false belief task by making infants expect that the agent will search at the new location in the violation of expectation paradigm (i.e., show surprise when the agent searches at the old location). The researchers found no effect of training, which is inconsistent with the idea that infants use simple rules to succeed on implicit false belief tasks. If infants were processing false belief scenarios on learned behavioral rules (e.g., people always search for an object where they last saw it), they would have shown the opposite behavior at the test after learning that people never search at the last location. However, training might have been insufficient to overwrite the well-learned rule that people search for objects in the last place they saw them. (As discussed in the section on the stability of ToM, a link has been established between exposure to repeated behaviors and later mental state language; Ruffman et al., Reference Ruffman, Chen and Lorimer2023.)

A two-system account of belief-ascription was put forward by Apperly and Butterfill (Apperly & Butterfill, Reference Apperly and Butterfill2009; Butterfill & Apperly, Reference Butterfill and Apperly2013; Low et al., Reference Low, Apperly, Butterfill and Rakoczy2016). There are two systems involved in understanding others’ beliefs: a fast but inflexible one (Type 1) and a later-emerging more flexible one (Type 2). According to this view, the spontaneous responses in false belief tasks come from the Type 1 system, which helps us think about basic belief-like states called “registrations,” but only in certain situations. The Type 2 system develops later, is flexible, and supports the attribution of genuine beliefs. This system is effortful and explicit but also inefficient and supports children’s success in explicit false belief tasks. Such a flexible system allows children to overcome limitations posed by the early system and track more complex beliefs. Such a two-system account has been proposed to explain the apparent discrepancy between infants’ early success in nonverbal mindreading tasks, on the one hand, and failures of children younger than 4 to pass verbally mediated false belief tasks, on the other. It assumes that two types of ToM develop independently. Type I ToM system operates implicitly and unconsciously and remains functional throughout the lifespan in the context of spontaneous-response tasks such as violation of expectation and anticipatory looking. Type II ToM – representational ToM – develops later, requires the development of language and executive function capacities, and operates in explicit and conscious ways across the lifespan. As discussed earlier, findings with infants and adults that show some dissociations between implicit and explicit tasks provide evidence for this dual process framework (Edwards & Low, Reference Edwards and Low2017; Grosse Wiesmann et al., Reference Grosse Wiesmann, Schreiber, Singer, Steinbeis and Friederici2017; Low & Watts, Reference Low and Watts2013).

Another theory for the dissociation between the implicit and explicit ToM is the altercentric viewpoint developed by Southgate and colleagues (Kampis & Southgate, Reference Southgate2020; Southgate, Reference Southgate2020). An altercentric bias results from focusing on what others see or think, especially in the absence of a concept of self. Once one has developed one’s own perspective, it can create a conflict when reasoning about false beliefs that requires executive functions, such as inhibitory control, to overcome. Thus, infants’ apparent success in ToM tasks that require mature executive functions (which are sometimes challenging to adults) could be explained by the fact that they have an altercentric bias. A self-perspective emerges with the development of cognitive self-awareness sometime in the second year of life, at which point it leads to competition between one’s own perspective and that of others. For example, in the seminal violation of expectation paradigm, the infant encodes that the agent witnesses a toy moving into the yellow box and maintains the agent’s perspective despite the fact that the infant later sees that the toy moves to the green box, an event that would not be encoded with the same representational strength because it is not co-witnessed by another agent (Southgate, Reference Southgate2020). As a consequence, infants would correctly predict the agent to search in the yellow box in line with the agent’s beliefs, although this would not entail that the infant represents this as the other’s belief. The altercentric account thus predicts infants’ surprise when an agent acts in a belief-incongruent way, explaining data from violation of expectation false belief tasks. In a similar way, encoding the agent’s, rather than their own perspective, predicts infants’ correct anticipation in previous anticipatory looking studies (e.g., Southgate et al., Reference Southgate, Senju and Csibra2007) and their correct interpretation of others’ communication and action prediction in interaction-based studies (e.g., Knudsen & Liszkowski, Reference Knudsen and Liszkowski2011; Southgate et al., Reference Southgate, Chevallier and Csibra2010).

In summary, several theoretical accounts have been proposed to explain Theory of Mind development (see Ruffman, Reference Ruffman2023, for a review of the main accounts). They all share the view that what allow the changes in ToM from infancy to preschool years are cognitive (e.g., executive functions, language) and social (e.g., social interactions) achievements. They diverge on the initial ToM competences that children initially possess, ranging from an implicit understanding of mental states to the detection of regularities in behaviors. After decades of research, is there an account that is most robustly supported by the data? Each theory is consistent with some research findings, although there is inconsistency in the findings linked to each account, as illustrated by the conflicting results of studies on the stability of Theory of Mind. Two environmental factors (repeated behaviors and maternal talk) and two developing insights (a sense of self and language) likely help children transition from an understanding of behavior to a ToM. To test the minimalist account’s hypothesis that the foundations of Theory of Mind are parsimonious domain-general mechanisms, future research should adopt designs that expose infants to behavioral regularities and examine their impact on later Theory of Mind. Similarly, the impact of increased exposure to maternal talk on ToM development should be investigated. With regard to the rich, mentalist account is the wonder if it is falsifiable. Additional conceptual replications of implicit Theory of Mind tasks must be conducted to assess the depth of infants’ reasoning in implicit ToM tasks. For example, infants’ failure to infer false belief in the VOE paradigm when the agent turns her back or wears a veil when the object changes location provides key information about the fragility of ToM understanding in infancy (Poulin-Dubois et al., Reference Poulin-Dubois, Azar, Elkaim and Burnside2020).

5 Neural Foundations of Theory of Mind Development

Regardless of the theoretical perspective that one adopts, a consensus exists that neural changes support ToM development in addition to the cognitive and social correlates discussed earlier. Frith and Frith (Reference Frith and Frith1999) described the first major efforts to document neural bases of mentalizing. Their studies were conducted with positron emission tomography (PET), where the brains of small samples of adults were scanned while reading stories, which required either mental state reasoning (e.g., false beliefs) or physical state reasoning (Fletcher et al., Reference Fletcher, Happé and Frith1995). They revealed a set of brain regions that were activated when people had to reason about mental states: in particular the medial prefrontal cortex (mPFC), superior posterior temporal sulcus (pSTS), and posterior cingulate cortex (PCC). Remarkably, this network of regions has been repeatedly identified in studies of mentalizing and is sometimes loosely referred to as the “social brain.” In a review, Frith and Frith (Reference Frith, Frith, Gilead and Ochsner2021) reported the progress in mapping the brain areas involved in ToM.

The neuroscience of ToM has contributed to clarifying the dissociation between implicit and explicit mentalizing. If the two types of tasks measure similar cognitive processes, similar brain regions should be implicated in their development. In contrast, if different brain regions are involved in successful performance on these tasks, this provides critical evidence for distinct underlying cognitive processes. Furthermore, a comparison of the relevant brain regions observed in children to those involved in ToM in adults would help identify which of these processes reflect mature ToM reasoning. Verbal, explicit ToM reasoning recruits a consistent network of brain regions, including the temporoparietal junction (TPJ), middle temporal gyrus (MTG), precuneus (PC), and medial prefrontal cortex in both adults (Schurz et al., Reference Schurz, Radua, Aichhorn, Richlan and Perner2014) and children (Grosse Wiesmann et al., Reference Grosse Wiesmann, Schreiber, Singer, Steinbeis and Friederici2017; Liu et al., Reference Liu, Wellman, Tardif and Sabbagh2008; Richardson et al., Reference Richardson, Lisandrelli, Riobueno-Naylor and Saxe2018; Saxe et al., Reference Saxe, Whitfield‐Gabrieli, Scholz and Pelphrey2009). This brain network shows increased connectivity, both functional and structural, at the age that children pass explicit ToM tasks. In contrast, research on the neural structures involved in implicit ToM tasks is scarce. Which brain regions underlie young children’s success in implicit ToM tasks, and are they the same or different from those involved in the explicit ToM tasks? To address this issue, Grosse Wiesmann and colleagues (Reference Grosse Wiesmann, Friederici, Singer and Steinbeis2020) tested children with MRI and a battery of implicit and explicit ToM tasks at the critical age of 3–4 years, corresponding to a developmental period before and after explicit ToM reasoning emerges. Markers of cortical brain structure (i.e., cortical thickness and surface area) were related to children’s performance in novel nonverbal and traditional verbal ToM tasks. Explicit ToM reasoning was supported by cortical surface area and thickness of the precuneus and temporoparietal junction, which are brain areas typically involved in ToM in adults. In contrast, implicit ToM reasoning was supported by the cortical structure of an independent neural network, including the supramarginal gyrus. This double neural dissociation was also independent of other cognitive domains (language, executive function, general intelligence), providing strong support for two systems for reasoning about mental states – an explicit, verbal ToM that emerges around 4 years of age and an implicit nonverbal earlier-developing ToM (see Figure 6).

Figure 6 Distinct and independent brain regions associated with success on explicit (blue) and implicit ToM tasks (red, orange) for surface area (A) and cortical thickness (B).

(figure reproduced from Grosse Wiesmann et al., Reference Grosse Wiesmann, Friederici, Singer and Steinbeis2020 with permission)

Regarding brain imaging research with even younger children, one study has used fNIRS to test the specific hypothesis that RTPJ is preferentially activated when young infants are reasoning about other people’s mental states (Hyde et al., Reference Hyde, Simon, Ting and Nikolaeva2018). A replication of the false belief test based on the violation of expectation procedure with 7-month-olds was conducted (Onishi & Baillargeon, Reference Onishi and Baillargeon2005). Participants watched a protagonist place an object in a container, leave, and then return to retrieve the object. As in the original VOE procedure, the object’s location was changed during the induction trial when the protagonist was absent. Hyde and colleagues (Reference Hyde, Simon, Ting and Nikolaeva2018) reported that the RTPJ showed increased blood oxygenation in adult participants, specifically when the target with the false belief returned to retrieve the object, compared to control conditions that did not involve a false belief. They observed a similar response in 7-month-old infants to that observed in adults: preferential increases in oxygenated blood over the RTPJ, specifically when the protagonist returned to search for the object. An fMRI study tested the link between neural development and behavioral changes in reasoning about others’ minds by 4 to 5 years (Richardson et al., Reference Richardson, Lisandrelli, Riobueno-Naylor and Saxe2018). Children aged 3–12 years and adults watched a short movie while undergoing fMRI. The movie highlighted the characters’ bodily sensations (often pain) and mental states (beliefs, desires, emotions). There were three main findings: (1) ToM and pain networks are functionally distinct by age 3 years, (2) functional specialization increases throughout childhood, and (3) functional maturity of each network is related to increasingly anti-correlated responses between the networks. Furthermore, the most studied milestone in ToM development, passing explicit false belief tasks, did not correspond to discontinuities in the development of the social brain. When it comes to the neural underpinnings of ToM achievements linked to emotion, a review by Bachmann et al. (Reference Bachmann, Munzert and Krüger2018) reports that emotional stimuli evoke greater neural responses than neutral stimuli and that the brain regions observed to have a significant activation during emotional processing encompass the action observation network (formed by the inferior frontal gyrus, the premotor cortex, and the inferior parietal lobe) and the mentalizing network (the temporoparietal junction, the temporal pole, the lateral orbitofrontal cortex, and the dorsomedial prefrontal cortex), making it possible for the individual to infer the other’s emotions as mental states (Frith & Frith., Reference Frith and Frith2006).

In summary, more research on the overlap in brain areas involved in ToM reasoning across ages is needed. Furthermore, future research will need to directly contrast brain activation during ToM tasks with activation involved during other social-cognitive processes in infants and young children. According to Grosse-Wiesmann et al. (Reference Grosse Wiesmann, Friederici, Singer and Steinbeis2020, p. 6932), this remains a challenge as “it requires task-based functional MRI with very young children because the relatively low spatial resolution of infant neuroimaging methods (such as fNIRS) does not allow reliably measuring and dissociating the relevant brain regions (i.e., TPJ versus SMG and different portions of the PC).” Despite these challenges, understanding the neural processes underlying performance on implicit ToM success and their exact function must remain an approach in acquiring a better understanding of ToM development. Neuroscience could be a particularly useful tool to document Theory of Mind in infancy, and infant cognition in general, given that the behavioral measures used with infants are difficult to interpret and have weak psychometric properties (Turk-Browne & Aslin, Reference Turk-Browne and Aslin2024). This is well illustrated in a recent longitudinal study that examined the relationship between functional sensitivity of the right temporal-parietal junction (TPJ) to mental state processing in 7-month-old infants (as measured with the anticipatory looking paradigm) and explicit Theory of Mind reasoning in the same group several years later (Liu et al., Reference Liu, Moss, Ting and Hyde2025). Recall that TPJ is the brain region known to underlie Theory of Mind in older children and adults. Unexpectedly, the link was not between infant’s brain activity during false belief processing but during a direct perception condition (false belief condition but with transparent boxes so the agent could see the new location of the object). Such evidence of a longitudinal brain-behavioral link from infancy to childhood provides support for stability of Theory of Mind across development, specifically between the understanding of seeing = knowing and later Theory of Mind. Thus, infant neuroscience provides a valuable contribution to the science of Theory of Mind development.

6 Summary and Future Directions

Now thousands of articles and books have been published on ToM, spanning the development of ToM across the lifespan. The universality of ToM and its evolutionary roots are also well documented, as is individual variability in reaching ToM milestones. We have seen that such variability has triggered a large body of work on the antecedents and correlates of Theory of Mind competence, including having siblings, engaging in conversations, and growing up bilingual. In addition, research has revealed the neural underpinnings of ToM, showing that different brain regions in the so-called ToM network are activated at different developmental stages, reflecting changes in the way ToM is used across the lifespan (Bowman & Wellman, Reference Wellman2014). To fully document ToM across the lifespan, a large number of measures have been developed over the years that assess the whole range of Theory of Mind constructs (for a review of adult measures, see Quesque & Rossetti, Reference Quesque and Rossetti2020, and Yeung et al., Reference Yeung, Apperly and Devine2024; for measures for children, see Beaudoin et al., Reference Beaudoin, Leblanc, Gagner and Beauchamp2020, and Osterhaus & Bosacki, Reference Osterhaus and Bosacki2022). A large literature has also documented how ToM is relevant in other aspects of child development, such as pretend play, lying, academic achievement, and peer relationships.

This Element provides a review of the research on children’s ToM that is necessarily selective, given the scope of the literature. This Element strives to balance breadth and depth, and the review reports interpretations that researchers offer for their findings and only occasionally critiques those findings. All studies have limitations, leaving their findings open to alternative explanations. Thus, the conclusions that are offered here are tentative until more research confirms these findings through new designs as well as strict and conceptual replications. As is typical with areas in which a large number of researchers explore a topic from different vantage points, variability across studies makes the comparison of findings challenging.

Given the scope of extant research on ToM, what future directions will this well-researched topic take? Despite progress made in better understanding this foundational ability, many aspects of ToM development and its determinants are still not well understood. First, the nature of ToM in infancy remains uncertain. The internal and external validity of all implicit ToM tasks remain to be demonstrated, casting doubt on whether there is a precocious form of advanced mindreading (e.g., false belief) in infancy that is implicit and unconscious (Poulin-Dubois et al., Reference Poulin-Dubois, Hastings and Chiarella2018). Indeed, the collective effort of the ManyBabies consortium in which the authors of original and replication studies test whether implicit ToM tasks are replicable across laboratories is an important step to address that issue (Schuwerk et al., Reference Schuwerk, Kampis and Baillargeon2025). More longitudinal studies will also need to be conducted with a wide range of ToM tasks both in infancy and in childhood. Importantly, control tasks that assess cognitive skills unrelated to ToM will be required to establish conceptual continuity of ToM. Nonetheless, it is important to note that evidence for stability does not provide conclusive proof for a full-fledged ToM in infancy but at least shows that implicit tasks measure abilities that are the building blocks for Theory of Mind development (Tomasello, Reference Tomasello2018).

Among the promising avenues to clarify the status of infants’ behaviors in implicit ToM tasks is the altercentric theoretical proposal (Grosse Wiesmann & Southgate, Reference Grosse Wiesmann, Southgate, Ochsner and Gilead2021; Kampis & Southgate, Reference Southgate2020; Southgate, Reference Southgate2020). Altercentrism in human cognition reflects that people’s behaviors are spontaneously influenced by the presence and perspective of others, even when the task requires them to focus on their own point of view (Elekes et al., Reference Elekes, Varga and Király2017; Kovács et al., Reference Kovács, Téglás and Endress2010; Samson et al., Reference Samson, Apperly, Braithwaite, Andrews and Bodley Scott2010). Although this altercentric influence has been well studied in adults, there is little research on its developmental origins. According to this view, when an agent with a diverging perspective is present, infants’ representation is impacted by the agent’s perception of the situation, which means that infants do not focus on their own representation of the situation but simply encode events that are witnessed in common with another agent. This stronger encoding when another agent is present entails that infants will update their representation only when a change occurs that is witnessed together with the agent. In contrast, if a change occurs in the absence of the agent, infants will not update the representation that they encoded together with the agent. This would allow infants to make correct action predictions in line with the agent’s belief without having to hold two conflicting perspectives. Before the development of a concept of self and with low executive functions, the altercentric bias might be an effective mechanism, which in most cases generates correct predictions while avoiding the complex cognitive demands of meta-representation. Support for this view is emerging. It remains to be seen how infants’ predictions change as a function of self-concept development, as measured by tasks such as the mirror recognition test (see Calmette & Meunier, Reference Calmette and Meunier2024, for a review of self-awareness and Theory of Mind).

A second line of research that requires attention concerns how children’s Theory of Mind skills are expressed as a function of the characteristics of the agent. Infants appear to attribute mental states to any agents, with agents being defined as objects with a stable disposition to display mechanical (e.g., autonomous motion) and actional properties (i.e., goal-oriented, contingent, reactive-at-a-distance behavior) (Carey, Reference Carey2009; Csibra, Reference Csibra2008; Leslie, Reference Leslie1994; Scholl & Tremoulet, Reference Scholl and Tremoulet2000; Surian & Caldi, Reference Surian and Caldi2010). Thus, infants presented with chasing triangles, blobs, or mechanical cranes anticipate the actions of those agents consistently with their mental states (Burnside et al., 2020; Surian & Franchin, Reference Surian and Franchin2020; Surian & Geraci, Reference Surian and Geraci2012). Although the assumption that behaviors in implicit tasks reflect a rich Theory of Mind is contentious, the developmental progression of ToM reasoning across a wide range of agents remains to be more fully explored. A first wave of studies has used interviews as the main method and animals, plants, and artifacts as agents (for a review, see Goldman & Poulin-Dubois, Reference Goldman and Poulin-Dubois2024). ToM reasoning with social robots has gained attention, as robots allow for the manipulation of morphological features and dynamic cues (for reviews, see Stower et al., Reference Stower, Calvo-Barajas, Castellano and Kappas2021; Marchetti et al., Reference Marchetti, Manzi, Itakura and Massaro2018). There is some evidence that younger children tend to anthropomorphize more than older children but that they attribute fewer mental states to robots than humans (Goldman et al., Reference Goldman, Baumann and Poulin-Dubois2023; Manzi et al., Reference Manzi, Peretti and Di Dio2020). An important step would be to test ToM with robots using scenarios such as those in the Wellman and Liu scale.

Yet another promising avenue of research is computer modeling. In order to confirm the causal role of many of the determinants of ToM, experimental studies are not ethically possible. For example, social interactions have been identified as impacting Theory of Mind development, as suggested by slower ToM in deaf children deprived of sign language, children with fewer siblings, and so on. Also, a comprehensive review of how child-level social tendencies can shape ToM showed that children’s tendencies to attend to social information and engage in social interaction both played critical roles in Theory of Mind development (Lane & Bowman, Reference Lane and Bowman2021). To shed light on social interactions and ToM, Yu and Wellman (Reference Yu and Wellman2023) conducted a computational experiment using agent-based modeling. They simulated agents (hypothetically, children) and assigned them to 10 groups, with a different amount of social interaction assigned to each group, from 1 to 10. In each group, agents randomly encountered one another and then interacted. The amount of social interaction strongly influenced how fast agents developed ToM, where more social interactions led to more rapid development. They also explored how other factors – such as having a prior, established social network or agents’ network centrality – could influence the social interaction–ToM link. Then, they tested the model against real-world data from 84 deaf children and showed that the modeling results could explain the social interaction–ToM link observed in deaf children.

7 Conclusions

In the introduction to this Element, I wrote that I aimed to answer four central questions about the development of a Theory of Mind: What, When, Why, and How? The large body of work reviewed provides some answers to these questions. Concerning the What question, the key constructs of a Theory of Mind are now well known, and there is a consensus that there is a progression in the complexity of mental constructs that children develop over the first few years of life, from simple mental states such as goals to complex constructs such as second-order false belief. The dominant approach to measure preschoolers’ understanding of these concepts has been through everyday scenarios enacted with props (Wellman & Liu, Reference Wellman and Liu2004). There is also effort devoted to measuring these constructs through parental reports, such as the Children’s Social Understanding Scale, which appears to have good validity (Tahiroglu et al., Reference Tahiroglu, Moses and Carlson2014). Adaptations in a wide range of languages are desirable so that populations around the globe can be assessed, but whether it can substitute for laboratory-based tasks when relating ToM to other cognitive abilities remains to be clarified.

The When question remains one of the most contentious issues in the Theory of Mind literature. How rich is the reasoning that preverbal infants need to solve some implicit ToM tasks? Clearly, more cross-sectional and longitudinal research is needed once the validity and reliability of implicit tasks are established. New behavioral and neurological methodological approaches should also be developed. Nonetheless, building blocks are in place in infancy, including some core knowledge about agents, that provide the foundations for Theory of Mind development. These building blocks include joint attention, statistical learning, and inferential abilities that are the starting points of a cascade of more advanced socio-cognitive skills, including Theory of Mind.

Why is it important to study ToM development? We hope to have convinced the reader that deficits in ToM have negative consequences in other developmental domains: ToM abilities are associated with social skills (e.g., Banerjee & Yuill, Reference Banerjee and Yuill1999; Cassidy et al., Reference Cassidy, Parke, Butkovsky and Braungart1992; Slaughter et al., Reference Slaughter, Dennis and Pritchard2002; Watson et al., Reference Watson, Nixon, Wilson and Capage1999), and = academic performance (e.g., Doudin et al., Reference Doudin, Martin and Albanese2001; Jones et al., Reference Jones, Brown and Aber2011; Lecce et al., Reference Lecce, Caputi and Hughes2011). Given that ToM procedures have been found to be effective in improving children’s ToM skills, past and future studies on ToM will benefit children by offering them the possibility to reach their full potential in domains known to impact their well-being across the lifespan (Hofmann et al., Reference Hofmann, Doan and Sprung2016).

Finally, the How question has received much attention from researchers over the past decades. Exciting new techniques, such as computer simulations, are starting to validate the main conclusions from behavioral studies. What remains to be clarified are the specific mechanisms involved in ToM. For example, there is solid empirical evidence that executive function and language both play crucial roles in acquiring metarepresentation. However, it is still unclear what underlying proximal mechanisms permit executive function and language to make metarepresentation possible. Continued research on the development of a ToM will enable a better understanding of how children explain and predict people’s behaviors by inferring mental states and how such foundational skills could be supported to reach their optimal level.

Theory of Mind is the capacity for people to understand others and themselves as sentient agents with subjective perspectives on the world who act based on their mental states. This foundational form of social cognition rests on the ability to form metarepresentations, that is, having beliefs about others’ beliefs. Although some primitive forms of ToM are evolutionarily ancient, are ontogenetically precocious, and appear to be shared with non-human primates, a mature, full-fledged metarepresentational ToM develops gradually over childhood. There appear to be milestones in ToM capacity during the second year of life and around 4 years of age. General neurocognitive capacities such as executive function and linguistic experience seem to facilitate that development, as well as a wide range of other social determinants, with many social and educational consequences. Still many key aspects of ToM development and its underpinnings are still poorly understood. For example, what mechanisms allow metarepresentation to emerge through executive function and language remain to be identified. The current debate concerning the role of language in ToM research should, therefore, be part of a much broader research program that aims to understand how language shapes thought during development. Likewise, the neural underpinnings of adult ToM are well documented, and studies are underway to reveal those relevant to developmental change. Future research will need to aim at developmental implementation theories, which detail how neural changes implement or realize cognitive transitions. New computational theories of the underpinnings of ToM (Baker et al., Reference Baker, Jara-Ettinger, Saxe and Tenenbaum2017; Jara-Ettinger, Reference Jara-Ettinger2019) might have a crucial role in this endeavor. Finally, a central question for future study concerns the status of a nascent implicit ToM in infancy. Establishing the reliability and validity of implicit ToM tasks is underway as one of the projects of the ManyBabies consortium (Frank et al., Reference Frank, Bergelson and Bergmann2017), a platform for replicability research in infant cognition. This consortium is currently testing whether the implicit ToM task based on anticipatory looking is replicable across laboratories (Schuwerk et al., Reference Schuwerk, Kampis and Baillargeon2025). Needless to say, even if this effort of replicability is successful, it will be insufficient in reaching a consensus about the richness of infants’ reasoning. More conceptual replications of original studies as well as new experimental procedures are needed to clarify the richness and limitations of Theory of Mind during the first 2 years of life. The basic, theoretical question of the developmental origins of this foundational form of human social cognition could have practical applications, including the design of intervention programs that foster the development of social skills.