Hostname: page-component-76d6cb85b7-f97m6 Total loading time: 0 Render date: 2026-07-15T15:41:55.851Z Has data issue: false hasContentIssue false
  • English
  • Français

Studies with impossible languages falsify LMs as models of human language

Published online by Cambridge University Press:  01 July 2026

Jeffrey S. Bowers Open the ORCID record for Jeffrey S. Bowers [Opens in a new window]  and
Jeff Mitchell Open the ORCID record for Jeff Mitchell [Opens in a new window]
Jeffrey S. Bowers*
Affiliation:
School of Psychology and Neuroscience, University of Bristol, Bristol, UK j.bowers@bristol.ac.uk
Jeff Mitchell
Affiliation:
Department of Informatics, University of Sussex, UK j.j.mitchell@sussex.ac.uk
*
*Corresponding author.
Get access Rights & Permissions [Opens in a new window]

Abstract

According to F&M, both infants and language models (LMs) find attested languages easier to learn than “impossible languages” that have unnatural structures. We review the literature and show that LMs often learn attested and many impossible languages equally well. Difficult to learn impossible languages are simply more complex (or random). LMs are missing human inductive biases that support language acquisition.

Information

Type
Open Peer Commentary
Information
Behavioral and Brain Sciences , Volume 49 , 2026 , e202
Check for updates
Copyright
© The Author(s), 2026. Published by Cambridge University Press

Access options

Get access to the full version of this content by using one of the access options below. (Log in options will check for institutional or personal access. Content may require purchase if you do not have access.)

Article purchase

Temporarily unavailable

References

Bowers, J. S. (2025a). The successes and failures of artificial neural networks (ANNs) highlight the importance of innate linguistic priors for human language acquisition. Psychological Review. Advance online publication. https://doi.org/10.1037/rev0000595 CrossRefGoogle Scholar
Bowers, J. S. (2025b). Does distilling Bayesian priors into language models support rapid language learning? Comment on McCoy and Griffiths (2025). PsyArxiv. Preprint. https://doi.org/10.31234/osf.io/zeprf_v1 CrossRefGoogle Scholar
Chomsky, N. Roberts, I, & Watumull, J. (2023, March 8th). Noam Chomsky: The False Promise of ChatGPT. New York Times. https://www.nytimes.com/2023/03/08/opinion/noam-chomsky-chatgpt-ai.html 10.3917/phimag.hs057.0009CrossRefGoogle Scholar
Kallini, J., Papadimitriou, I., Futrell, R., Mahowald, K., & Potts, C. (2024). Mission: Impossible language models. In Proceedings of the 62nd annual meeting of the association for computational linguistics (Volume 1: Long Papers, pp. 14691–14714). Association for Computational Linguistics Google Scholar
Luo, X., Ramscar, M., & Love, B. C. (2024). Beyond human-like processing: Large language models perform equivalently on forward and backward scientific text. arXiv preprint arXiv:2411.11061.Google Scholar
McCoy, R. T., & Griffiths, T. L. (2025). Modeling rapid language learning by distilling Bayesian priors into artificial neural networks. Nature Communications, 16(1), 4676.10.1038/s41467-025-59957-yCrossRefGoogle ScholarPubMed
Mitchell, J., & Bowers, J. (2020, December). Priorless recurrent networks learn curiously. In Proceedings of the 28th International Conference on Computational Linguistics (pp. 5147–5158).10.18653/v1/2020.coling-main.451CrossRefGoogle Scholar
Moro, A., Greco, M., & Cappa, S. F. (2023). Large languages, impossible languages and human brains. Cortex, 167, 82–85.10.1016/j.cortex.2023.07.003CrossRefGoogle ScholarPubMed
Wolpert, D. H., & Macready, W. G. (2002). No free lunch theorems for optimization. IEEE Transactions on Evolutionary Computation, 1(1), 67–82.10.1109/4235.585893CrossRefGoogle Scholar
Xu, T., Kuribayashi, T., Oseki, Y., Cotterell, R., & Warstadt, A. (2025). Can language models learn typologically implausible languages? arXiv preprint arXiv:2502.12317.Google Scholar
Yang, X., Aoyama, T., Yao, Y., & Wilcox, E. (2025). Anything goes? A crosslinguistic study of (im) possible language learning in LMs. arXiv preprint arXiv:2502.1879510.18653/v1/2025.acl-long.1264CrossRefGoogle Scholar
Ziv, I., Lan, N., Chemla, E., & Katzir, R. (2025). Large language models as proxies for theories of human linguistic cognition. arXiv preprint arXiv:2502.07687.Google Scholar