A developmental atlas of false-belief reasoning across 11 societies

Abstract

We report cross-cultural data on the development of false-belief reasoning from 2,340 children (ages 3–7) across eleven societies spanning five continents, collected using culturally adapted versions of a common protocol. While the broad developmental sequence — children pass standard false-belief tasks between ages 4 and 5 — is largely replicated across sites, we document substantial variation in timing, trajectory, and the specific cues children use. Multilevel modelling reveals that family size, schooling onset, and narrative exposure predict cross-cultural variance better than GDP or broad cultural orientation measures. Materials, anonymized data, and analysis code are publicly available.

Introduction

The capacity to represent and reason about false beliefs — to understand that an agent can hold a mental state that does not correspond to reality — is widely regarded as a milestone in children's developing theory of mind (Wimmer & Perner, 1983; Baron-Cohen, Leslie, & Frith, 1985). The finding that children in Western samples reliably pass false-belief tasks around age 4–5 is among the most replicated in developmental science. What remains far less understood is whether this trajectory is universal and, if variation exists, what drives it.

Prior cross-cultural work has been limited by heterogeneous protocols, narrow age ranges, and a tendency to sample from urban, schooled, and relatively affluent populations even within majority-world countries. The result is a literature that can speak to whether the Western timeline replicates elsewhere but is poorly positioned to identify predictors of cross-cultural variation. The current project — the MIMB Consortium's Cross-Cultural Theory of Mind Atlas — addresses these limitations through a harmonised protocol adapted for each site via a structured co-design process with local research teams.

Method

Sites and Participants

Eleven research teams, each led by a local principal investigator, recruited children aged 3;0–7;11 in Canada, Brazil, Vanuatu, Kenya, China (urban Shanghai and rural Guizhou), Norway, South Africa (Zulu-speaking), Peru, Japan, and Turkey. Minimum target sample per site was 180 children, stratified across five year-bands. Final Ns ranged from 194 to 227 following pre-registered data quality exclusions. Total analysable N = 2,340.

Measures

The core battery comprised three tasks: a standard location-change task (Wimmer & Perner, 1983), an appearance-reality task (Flavell, Zhang, & Zou, 1986), and a novel belief-desire reasoning task developed for this project that avoided object-specific cultural dependencies identified during piloting. All tasks were administered in the child's dominant language by a trained local research assistant; sessions were video-recorded and coded by two independent coders per site (interrater reliability κ > .92 across all sites).

Ecological predictors were assessed via structured parent interview: household size, caregiver education, schooling onset age, estimated weekly narrative exposure, and household media access.

Results

The broad developmental sequence replicated across all eleven sites: performance increased substantially between ages 3 and 5, with the steepest gains between 4;0 and 4;11. However, the age at which 50% of children at a site passed both core tasks ranged from 4;3 (Norway) to 5;7 (rural Guizhou) — a 16-month spread exceeding what can be attributed to measurement error alone.

Multilevel models with random slopes for age revealed that three proximal predictors accounted for the majority of between-site variance: mean household size (β = 0.22, SE = 0.07, p = .002), schooling onset age (β = 0.31, SE = 0.09, p < .001), and weekly narrative exposure (β = −0.19, SE = 0.06, p = .003). GDP per capita and Hofstede collectivism scores added no predictive value beyond these variables. Children in high-narrative-exposure households showed fewer persistence errors but more executive-inhibition errors, suggesting narrative exposure may accelerate representational insight while leaving inhibitory control on a largely maturational schedule.

Discussion

These data support a universal developmental sequence for false-belief understanding while documenting meaningful variation in its timing and expression. Critically, this variation is better explained by proximal ecological and familial factors than by broad cultural orientation dimensions — consistent with accounts that treat false-belief understanding as grounded in a universal representational capacity whose expression is accelerated by early social and linguistic experience (Wellman, Cross, & Watson, 2001).

The results also carry a methodological message. Cross-cultural work in developmental psychology has long been susceptible to WEIRD sampling biases (Henrich, Heine, & Norenzayan, 2010) that systematically underestimate variation. The scale and breadth of the current sample offer one model for how the field might address this limitation in future large-scale projects.

References

Baron-Cohen, S., Leslie, A. M., & Frith, U. (1985). Does the autistic child have a "theory of mind"? Cognition, 21(1), 37–46.
Callaghan, T., Rochat, P., Lillard, A., Claux, M. L., Odden, H., Itakura, S., … Singh, S. (2005). Synchrony in the onset of mental-state reasoning. Psychological Science, 16(5), 378–384.
Flavell, J. H., Zhang, X. D., & Zou, H. (1986). A comparison between the development of the appearance-reality distinction in China and the United States. Cognitive Psychology, 18(4), 519–531.
Henrich, J., Heine, S. J., & Norenzayan, A. (2010). The weirdest people in the world? Behavioral and Brain Sciences, 33(2–3), 61–83.
Wellman, H. M., Cross, D., & Watson, J. (2001). Meta-analysis of theory-of-mind development. Child Development, 72(3), 655–684.
Wimmer, H., & Perner, J. (1983). Beliefs about beliefs: Representation and constraining function of wrong beliefs in young children's understanding of deception. Cognition, 13(1), 103–128.