Elamur Research⌕
☰
home / papers / 2026 / Language Statistics and False Belief Reasoning: Evidence from 41 Open-Weight LMs

Language Statistics and False Belief Reasoning: Evidence from 41 Open-Weight LMs

Paper · 2026 · Sean Trott, Samuel M. Taylor, Cameron Robert Jones et al.

Summary

This paper addresses a methodological weakness in studies that examine how language models reason about others' minds. Most such studies rely on a few closed-source systems, which hides training details, makes data contamination hard to check, and breaks reproducibility when models are retired. The authors re-run the false belief experiment of Trott et al. (2023) on 41 open-weight LMs from five families: Gemma, Llama, OLMo, Pythia and Qwen. The original study tested only text-davinci-002, which has since been deprecated. The authors compare model behaviour to trial-level human responses on 192 novel passages. Roughly a third of the models respond to whether a character witnessed an object being moved, but none accounts for the full human effect. Bigger models are both more sensitive and better at predicting human responses. The models also suggested a new hypothesis, then confirmed in humans: belief probes using "thinks" push both groups toward false-belief answers. The human size of this effect falls within the range of LM effects. Language statistics may therefore explain this cue effect but not the core knowledge-state sensitivity.

Key points

  • The stimuli come from Trott et al. (2023): 12 scenarios × 16 variants, giving 192 passages. The variants cross Knowledge State, Knowledge Cue (explicit vs. implicit probe), First Mention and Recent Mention.
  • Of the 41 open-weight models, 34% showed reliable sensitivity to the true- vs. false-belief manipulation. No model statistically "explained away" the human knowledge-state effect.
  • Parameter count was the LM property linked to both knowledge-state sensitivity and psychometric predictive power.
  • A non-factive verb cue ("John thinks...") produced more false-belief attributions than an indirect action cue ("John looks in the..."), in both LMs and humans. The human effect size lay inside the LM distribution.
  • A supplementary modified false belief task (Appendix E) gave qualitatively similar results. The authors present large open-weight model samples as a template for testing theories of cognition.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

 Institutions

🎓Massachusetts Institute of TechnologyCambridge · United States211 works · 69,000 citations🎓University of California San DiegoSan Diego · United States73 works · 51,948 citations🎓Rutgers, The State University of New JerseyNew Brunswick · United States36 works · 12,026 citations🎓Stony Brook UniversityStony Brook · United States43 works · 2,571 citations

 Fields, topics and concepts

📚 Shelves 6The field of science this work belongs to (OpenAlex fields).Computer Science · 16,667 worksArtificial Intelligence · 16,194 worksArts and Humanities · 16,039 worksLanguage and Linguistics · 15,891 worksPsychology · 1,552 worksDevelopmental and Educational Psychology · 499 works

🧭 Topics 1Research subjects and languages the work is about.Reasoning · 1,426 works

Cite

Cite as Sean Trott, Samuel M. Taylor, Cameron Robert Jones, James A. Michaelov and Pamela D. Rivière. 2026. Language Statistics and False Belief Reasoning: Evidence from 41 Open-Weight LMs. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers).

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2026.acl-long.1849

@inproceedings{trott2026language,
  title = {{Language Statistics and False Belief Reasoning: Evidence from 41 Open-Weight LMs}},
  author = {Sean Trott, Samuel M. Taylor, Cameron Robert Jones, James A. Michaelov and Pamela D. Rivière},
  booktitle = {Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)},
  year = {2026},
  doi = {10.18653/v1/2026.acl-long.1849},
  url = {https://doi.org/10.18653/v1/2026.acl-long.1849}
}

References

76 references, 9 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Ian A Apperly. 2012. What is “theory of mind”? concepts, cognitive processes and individual differences. Quarterly Journal of Experimental Psychology, 65(5):825–839. find
  2. Douglas Bates, Martin Mächler, Ben Bolker, and Steve Walker. 2015. Fitting linear mixed-effects models using lme4. Journal of Statistical Software, 67(1):1– 48.
  3. Andrew M Bean, Ryan Othniel Kearns, Angelika Romanou, Franziska Sofia Hafner, Harry Mayne, Jan Batzner, Negar Foroutan, Chris Schmitz, Karolina Korgul, Hunar Batra, and 1 others. 2025. Measuring what matters: Construct validity in large language model benchmarks. arXiv preprint arXiv:2511.04703. — [arXiv:2511.04703](https://arxiv.org/abs/2511.04703) find
  4. Stella Biderman, Hailey Schoelkopf, Quentin Gregory Anthony, Herbie Bradley, Kyle O’Brien, Eric Hallahan, Mohammad Aflah Khan, Shivanshu Purohit, Usvsn Sai Prashanth, Edward Raff, Aviya Skowron, Lintang Sutawika, and Oskar Van Der Wal. 2023. Pythia: A Suite for Analyzing Large Language Models Across… find
  5. Paul Bloom and Tim P German. 2000. Two reasons to abandon the false belief task as a test of theory of mind. Cognition, 77(1):B25–B31. find
  6. Elisabeth EF Bradford, Victoria EA Brunsdon, and Heather J Ferguson. 2020. The neural basis of belief-attribution across the lifespan: False-belief reasoning and the n400 effect. cortex, 126:265–280. find
  7. Jane R. Brown, Nancy Donelan-McCall, and Judy Dunn. 1996. Why Talk about Mental States? The Significance of Children’s Conversations with Friends, Siblings, and Mothers. Child Development, 67(3):836– 849. find
  8. Josep Call and Michael Tomasello. 2008. Does the chimpanzee have a theory of mind? 30 years later. Trends in cognitive sciences, 12(5):187–192. find
  9. Emmanuel Chemla. 2008. An epistemic step for antipresuppositions. Journal of Semantics, 25(2):141– 173. find
  10. Javier Conde, María Grandury, Tairan Fu, Carlos Arriaga, Gonzalo Martínez, Thomas Clark, Sean Trott, Clarence Gerald Green, Pedro Reviriego, and Marc Brysbaert. 2025. Adding llms to the psycholinguistic norming toolbox: A practical guide to getting the most out of human ratings. arXiv preprint arXiv… — [arXiv:2509.14405](https://arxiv.org/abs/2509.14405) find
  11. Ionut Constantinescu, Tiago Pimentel, Ryan Cotterell, and Alex Warstadt. 2025. Investigating critical period effects in language acquisition through neural language models. Transactions of the Association for Computational Linguistics, 13:96–120. of Models. arXiv preprint. ArXiv:2407.21783 [cs] vers…
  12. Jill De Villiers. 2007. The interface of language and theory of mind. Lingua, 117(11):1858–1878. find
  13. Yuling Gu, Oyvind Tafjord, Hyunwoo Kim, Jared Moore, Ronan Le Bras, Peter Clark, and Yejin Choi. 2024. Simpletom: Exposing the gap between explicit tom inference and implicit tom application in llms. arXiv preprint arXiv:2410.13648. — [arXiv:2410.13648](https://arxiv.org/abs/2410.13648) find
  14. Jill G. de Villiers and Peter A. de Villiers. 2014. The Role of Language in Theory of Mind Development. Topics in Language Disorders, 34(4):313–328. find
  15. Courtney Melinda Hale and Helen Tager-Flusberg. 2003. The influence of language on theory of mind: A training study. Developmental Science, 6(3):346–359. find
  16. Vittoria Dentella, Fritz Günther, and Evelina Leivada. 2023. Systematic testing of three language models reveals low language accuracy, absence of response stability, and a yes-response bias. Proceedings of the National Academy of Sciences, 120(51):e2309583120. find
  17. Marta Halina. 2015. There is no special problem of mindreading in nonhuman animals. Philosophy of Science, 82(3):473–490. find
  18. David Dodell-Feder, Sarah Hope Lincoln, Joseph P Coulson, and Christine I Hooker. 2013. Using fiction to assess mental state understanding: a new task for assessing theory of mind in adults. PloS one, 8(11):e81279. find
  19. Guy Dove. 2020. More than a scaffold: Language is a neuroenhancement. Cognitive neuropsychology, 37(5-6):288–311. find
  20. Sarah Fairchild and Anna Papafragou. 2021. The role of executive function and theory of mind in pragmatic computations. Cognitive Science, 45(2):e12938. find
  21. Kanishk Gandhi, Jan-Philipp Fränken, Tobias Gerstenberg, and Noah Goodman. 2023. Understanding social reasoning in language models with language models. Advances in Neural Information Processing Systems, 36:13518–13529. find
  22. Elizabeth O Hayward and Bruce D Homer. 2017. Reliability and validity of advanced theory-of-mind measures in middle childhood and adolescence. British Journal of Developmental Psychology, 35(3):454– 462. find
  23. Joan B Hooper. 1975. On assertive predicates. In Syntax and Semantics volume 4, pages 91–124. Brill. find
  24. Jennifer Hu and Michael Frank. 2024. Auxiliary task demands mask the capabilities of smaller language models. In First Conference on Language Modeling.
  25. Jennifer Hu, Kyle Mahowald, Gary Lupyan, Anna Ivanova, and Roger Levy. 2024. Language models align with human judgments on key grammatical constructions. Proceedings of the National Academy of Sciences, 121(36):e2400917121. find
  26. Jennifer Hu, Felix Sosa, and Tomer Ullman. 2025. Re-evaluating theory of mind evaluation in large language models. Philosophical Transactions B, 380(1932):20230499. find
  27. Gemma Team, Morgane Riviere, Shreya Pathak, Pier Giuseppe Sessa, Cassidy Hardin, Surya Bhupatiraju, Léonard Hussenot, Thomas Mesnard, Bobak Shahriari, Alexandre Ramé, Johan Ferret, Peter Liu, Pouya Tafti, Abe Friesen, Michelle Casbon, Sabela Ramos, Ravin Kumar, Charline Le Lan, Sammy Jerome, and 178… find
  28. Ann Huang, Satpreet H Singh, Flavio Martinelli, and Kanaka Rajan. 2025. Measuring and controlling solution degeneracy across task-trained recurrent neural networks. ArXiv, pages arXiv–2410. find
  29. Morton Ann Gernsbacher and Melanie Yergeau. 2019. Empirical failures of the claim that autistic people lack a theory of mind. Archives of scientific psychology, 7(1):102. find
  30. Anna A Ivanova. 2025. How to evaluate the cognitive abilities of llms. Nature Human Behaviour, 9(2):230– 233. find
  31. Alice Ann Howard Gola. 2012. Mental verb input for promoting children’s theory of mind: A training study. Cognitive Development, 27(1):64–76. find
  32. Aaron Grattafiori, Abhimanyu Dubey, Abhinav Jauhri, Abhinav Pandey, Abhishek Kadian, Ahmad Al-Dahle, Aiesha Letman, Akhil Mathur, Alan Schelten, Alex Vaughan, Amy Yang, Angela Fan, Anirudh Goyal, Anthony Hartshorn, Aobo Yang, Archi Mitra, Archie Sravankumar, Artem Korenev, Arthur Hinsvark, and 542 o… find
  33. Cameron R Jones, Sean Trott, and Benjamin Bergen. 2024. Comparing humans and large language models on an experimental protocol inventory for theory of mind evaluation (EPITOME). Transactions of the Association for Computational Linguistics, 12:803– 819.
  34. Jared Kaplan, Sam McCandlish, Tom Henighan, Tom B Brown, Benjamin Chess, Rewon Child, Scott Gray, Alec Radford, Jeffrey Wu, and Dario Amodei. 2020. Scaling laws for neural language models. arXiv preprint arXiv:2001.08361.
  35. Paul Kiparsky and Carol Kiparsky. 1970. Fact. In Manfred Bierwisch and Karl Erich Heidolph, editors, Progress in Linguistics, pages 143–173. Mouton, The Hague. find
  36. Michal Kosinski. 2023. Theory of mind may have spontaneously emerged in large language models. arXiv preprint arXiv:2302.02083.
  37. Michal Kosinski. 2024. Evaluating large language models in theory of mind tasks. Proceedings of the National Academy of Sciences, 121(45):e2405460121.
  38. Christopher Krupenye, Fumihiro Kano, Satoshi Hirata, Josep Call, and Michael Tomasello. 2016. Great apes anticipate that other individuals will act according to false beliefs. Science, 354(6308):110–114. find
  39. Yair Lakretz, Dieuwke Hupkes, Alessandra Vergallito, Marco Marelli, Marco Baroni, and Stanislas Dehaene. 2021. Mechanisms for handling nested dependencies in neural-network language models and humans. Cognition, 213:104699. find
  40. Martha Lewis and Melanie Mitchell. 2025. Evaluating the robustness of analogical reasoning in large language models. Transactions on Machine Learning Research. find
  41. Andreas Liesenfeld and Mark Dingemanse. 2024. Rethinking open source generative ai: open-washing and the eu ai act. In Proceedings of the 2024 ACM Conference on Fairness, Accountability, and Transparency, pages 1774–1787. find
  42. James A Michaelov, Roger P Levy, and Benjamin K Bergen. 2025. Language model behavioral phases are consistent across architecture, training data, and scale. arXiv preprint arXiv:2510.24963. — [arXiv:2510.24963](https://arxiv.org/abs/2510.24963) find
  43. Kanishka Misra and Najoung Kim. 2024. Generating novel experimental hypotheses from language models: A case study on cross-dative generalization. arXiv preprint arXiv:2408.05086. — [arXiv:2408.05086](https://arxiv.org/abs/2408.05086) find
  44. Aida Nematzadeh, Kaylee Burns, Erin Grant, Alison Gopnik, and Tom Griffiths. 2018. Evaluating Theory of Mind in Question Answering. In Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing, pages 2392–2400, Brussels, Belgium. Association for Computational Linguistics… find
  45. Chris Olah. 2023. Interpretability dreams. Transformer Circuits. Https://transformer-circuits.pub/2023/interpretability-dreams/index.html. find
  46. Chris Olah, Nick Cammarata, Ludwig Schubert, Gabriel Goh, Michael Petrov, and Shan Carter. 2020. Zoom in: An introduction to circuits. Distill. Https://distill.pub/2020/circuits/zoom-in. find
  47. OLMo Team, Pete Walsh, Luca Soldaini, Dirk Groeneveld, Kyle Lo, Shane Arora, Akshita Bhagia, Yuling Gu, Shengyi Huang, Matt Jordan, Nathan Lambert, Dustin Schwenk, Oyvind Tafjord, Taira Anderson, David Atkinson, Faeze Brahman, Christopher Clark, Pradeep Dasigi, Nouha Dziri, and 21 others. 2025. 2 OL… find
  48. Long Ouyang, Jeff Wu, Xu Jiang, Diogo Almeida, Carroll L Wainwright, Pamela Mishkin, Chong Zhang, Sandhini Agarwal, Katarina Slama, Alex Ray, and 1 others. 2022. Training language models to follow instructions with human feedback. arXiv preprint arXiv:2203.02155.
  49. Damian KF Pang, Severin KY Pang, Marianne D Broeker, and Alix Hibble. 2025. Do large language models have a theory of mind? Proceedings of the National Academy of Sciences, 122(28):e2507080122. find
  50. David Premack and Guy Woodruff. 1978. Does the chimpanzee have a theory of mind? Behavioral and Brain Sciences, 1(4):515–526. find
  51. Qwen Team, An Yang, Baosong Yang, Beichen Zhang, Binyuan Hui, Bo Zheng, Bowen Yu, Chengyuan Li, Dayiheng Liu, Fei Huang, Haoran Wei, Huan Lin, Jian Yang, Jianhong Tu, Jianwei Zhang, Jianxin Yang, Jiaxi Yang, Jingren Zhou, Junyang Lin, and 24 others. 2025. Qwen2.5 Technical Report. arXiv preprint. Ar… find
  52. R Core Team. 2025. R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria. find
  53. Maarten Sap, Ronan Le Bras, Daniel Fried, and Yejin Choi. 2022. Neural theory-of-mind? on the limits of social intelligence in large LMs. In Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing, pages 3762–3780, Abu Dhabi, United Arab Emirates. Association for Compu… find
  54. Michael Saxon, Ari Holtzman, Peter West, William Yang Wang, and Naomi Saphra. 2024. Benchmarks as microscopes: A call for model metrology. In First Conference on Language Modeling. find
  55. Dana Schneider, Virginia P Slaughter, Stefanie I Becker, and Paul E Dux. 2014. Implicit false-belief processing in the human brain. NeuroImage, 101:268–275. find
  56. Natalie Shapira, Mosh Levy, Seyed Hossein Alavi, Xuhui Zhou, Yejin Choi, Yoav Goldberg, Maarten Sap, and Vered Shwartz. 2024. Clever hans or neural theory of mind? stress testing social reasoning in large language models. In Proceedings of the 18th Conference of the European Chapter of the Associati… find
  57. Curt Tigges, Michael Hanna, Qinan Yu, and Stella Biderman. 2024. Llm circuit analyses are consistent across training and scale. Advances in Neural Information Processing Systems, 37:40699–40731. find
  58. Sean Trott. 2025. Toward a theory of generalizability in LLM mechanistic interpretability research. In Mechanistic Interpretability Workshop at NeurIPS 2025. find
  59. Sean Trott and Benjamin Bergen. 2019. Individual differences in mentalizing capacity predict indirect request comprehension. Discourse Processes, 56(8):675–707. find
  60. Sean Trott and Benjamin Bergen. 2020. When Do Comprehenders Mentalize for Pragmatic Inference? Discourse Processes, 57(10):900–920. find
  61. Sean Trott, Cameron Jones, Tyler Chang, James Michaelov, and Benjamin Bergen. 2023. Do large language models know what humans know? Cognitive Science, 47(7):e13309. find
  62. Tomer Ullman. 2023. Large language models fail on trivial alterations to theory-of-mind tasks. arXiv preprint arXiv:2302.08399. find
  63. Oskar van der Wal, Pietro Lesci, Max Müller-Eberstein, Naomi Saphra, Hailey Schoelkopf, Willem Zuidema, and Stella Biderman. 2025. Polypythias: Stability and outliers across fifty language model pre-training runs. In Proceedings of the Thirteenth International Conference on Learning Representations… find
  64. Alex Warstadt, Aaron Mueller, Leshem Choshen, Ethan Wilcox, Chengxu Zhuang, Juan Ciro, Rafael Mosquera, Bhargavi Paranjabe, Adina Williams, Tal Linzen, and Ryan Cotterell. 2023. Findings of the BabyLM challenge: Sample-efficient pretraining on developmentally plausible corpora. In Proceedings of the… find
  65. Taylor Webb, Keith J Holyoak, and Hongjing Lu. 2023. Emergent analogical reasoning in large language models. Nature Human Behaviour, 7(9):1526–1541. find
  66. Hadley Wickham. 2016. ggplot2: Elegant Graphics for Data Analysis. Springer-Verlag New York. find
  67. Ethan Wilcox, Clara Meister, Ryan Cotterell, and Tiago Pimentel. 2023. Language model quality correlates with psychometric predictive power in multiple languages. In The 2023 Conference on Empirical Methods in Natural Language Processing. find
  68. Heinz Wimmer and Josef Perner. 1983. Beliefs about beliefs: Representation and constraining function of wrong beliefs in young children’s understanding of deception. Cognition, 13(1):103–128. find
  69. Thomas Wolf, Lysandre Debut, Victor Sanh, Julien Chaumond, Clement Delangue, Anthony Moi, Pier-ric Cistac, Tim Rault, Rémi Louf, Morgan Funtowicz, Joe Davison, Sam Shleifer, Patrick von Platen, Clara Ma, Yacine Jernite, Julien Plu, Canwen Xu, Teven Le Scao, Sylvain Gugger, and 3 others. 2020. Transf…
  70. Yufan Wu, Yinghui He, Yilin Jia, Rada Mihalcea, Yulong Chen, and Naihao Deng. 2023. Hi-ToM: A benchmark for evaluating higher-order theory of mind reasoning in large language models. In Findings of the Association for Computational Linguistics: EMNLP 2023, pages 10691–10706, Singapore. Association f… find
  71. Yuheng Wu, Wentao Guo, Zirui Liu, Heng Ji, Zhaozhuo Xu, and Denghui Zhang. 2025. How large language models encode theory-of-mind: a study on sparse parameter patterns. npj Artificial Intelligence, 1(1):20. find
  72. Hainiu Xu, Runcong Zhao, Lixing Zhu, Jinhua Du, and Yulan He. 2024. OpenToM: A comprehensive benchmark for evaluating theory-of-mind reasoning capabilities of large language models. In Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pa… find
  73. Nicolas Yax, Pierre-Yves Oudeyer, and Stefano Palminteri. 2024. Phylolm: Inferring the phylogeny of large language models and predicting their performances in benchmarks. arXiv preprint arXiv:2404.04671. — [arXiv:2404.04671](https://arxiv.org/abs/2404.04671) find
  74. Rosie Zhao, Tian Qin, David Alvarez-Melis, Sham Kakade, and Naomi Saphra. 2025. Distributional scaling laws for emergent capabilities. arXiv preprint arXiv:2502.17356. — [arXiv:2502.17356](https://arxiv.org/abs/2502.17356) find
  75. Kaitlyn Zhou, Dan Jurafsky, and Tatsunori Hashimoto. 2023. Navigating the grey area: How expressions of uncertainty and overconfidence affect language models. In Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing, pages 5506–5524, Singapore. Association for Comput… find
  76. Rolf A Zwaan, Mark C Langston, and Arthur C Graesser. 1995. The construction of situation models in narrative comprehension: An event-indexing model. Psychological science, 6(5):292–297. find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/trott-2026-language-statistics-false-belief-reasoning/