Elamur Research⌕
☰
home / papers / 2026 / Do LLMs Capture Embodied Cognition and Cultural Variation? Cross-Linguistic Evidence from Demonstratives

Do LLMs Capture Embodied Cognition and Cultural Variation? Cross-Linguistic Evidence from Demonstratives

Paper · 2026 · Yu Wang, Emmanuele Chersoni and Chu-Ren Huang

Summary

The paper asks whether large language models learn embodied, spatial knowledge and culture-specific conventions from text alone. It uses demonstratives (English "this/that", Chinese 这/那) as the probe. These words signal how easily a speaker can reach an object, but corpora rarely state such accessibility outright. The authors built a small, tightly controlled bilingual dataset of 160 four-option multiple-choice questions (80 per language). Each item describes two people facing each other and an instruction about a target object. The items vary proximity, perspective, reference cues and everyday scenarios.

A human baseline from 320 native speakers revealed opposite tendencies. English speakers separate near and far referents sharply but find perspective-taking difficult. Chinese speakers shift perspective easily but accept ambiguity for distal items. Five state-of-the-art LLMs showed no real grasp of the proximal–distal contrast, often chose the logically incoherent "All of the above", and answered the same way in both languages, which reflects English-centric reasoning. The work offers a new grounded-cognition benchmark and adds evidence to the egocentric-versus-sociocentric debate.

Key points

  • Dataset design: 160 items (80 English, 80 Chinese), arranged pair-to-pair.
    • Each scene combines proximal/distal with self/other perspective, giving four-question groups.
    • Items cover five everyday scenarios with different verbs (eat, hide, take).
  • Four reference-cue conditions:
    • demonstratives only (the main baseline);
    • pronouns only;
    • demonstratives with matching pronouns;
    • demonstratives with conflicting pronouns.
  • Human study: 320 native speakers (160 Chinese via Credamo, 160 English via Prolific) produced 6,400 responses.
    • Participants were split into four gender-balanced surveys of 40 people, each with 20 questions.
    • Answers were analysed as response distributions rather than against one correct key.
  • LLM results: pronouns improved model accuracy, yet models still failed to treat proximal and distal readings as mutually exclusive. They also showed no cross-cultural differences.
  • Limitations and future work:
    • The study uses text-only input, a small dataset and only two languages.
    • The authors propose multimodal and 3D-simulation data, more languages (e.g., Spanish with its three-way system), and models that account for individual variation among users.

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

🎓Hong Kong Polytechnic UniversityHong Kong · China84 works · 4,956 citations

 Fields, topics and concepts

📚 Shelves 6The field of science this work belongs to (OpenAlex fields).Arts and Humanities · 16,039 worksLanguage and Linguistics · 15,891 worksPsychology · 1,552 worksExperimental and Cognitive Psychology · 870 worksNeuroscience · 165 worksCognitive Neuroscience · 164 works

🧭 Topics 1Research subjects and languages the work is about.Language models & pretraining · 3,320 works

Cite

Cite as Yu Wang, Emmanuele Chersoni and Chu-Ren Huang. 2026. Do LLMs Capture Embodied Cognition and Cultural Variation? Cross-Linguistic Evidence from Demonstratives. 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.461

@inproceedings{wang2026do,
  title = {{Do LLMs Capture Embodied Cognition and Cultural Variation? Cross-Linguistic Evidence from Demonstratives}},
  author = {Yu Wang, Emmanuele Chersoni and Chu-Ren Huang},
  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.461},
  url = {https://doi.org/10.18653/v1/2026.acl-long.461}
}

References

54 references, 7 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Anthropic. 2025. Claude sonnet 4.5 model card. https://www.anthropic.com/. Accessed 3 Jan 2026. ↗
  2. Merlijn R. E. Breunesse. 2019. Iconicity in spatial deixis: A cross-linguistic study of 180 demonstrative systems. In Caroline Gentens, Lobke Ghesquière, William B. McGregor, and An Van Linden, editors, Reconnecting Form and Meaning: In Honour of Kristin Davidse, volume 230 of Studies in Language Co… find
  3. Tom Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared Kaplan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, Sandhini Agarwal, Ariel Herbert-Voss, Gretchen Krueger, Tom Henighan, Rewon Child, Aditya Ramesh, Daniel Ziegler, Jeffrey Wu, Clemens Winter, and 12… find
  4. Karl Bühler. 1934. Sprachtheorie: Die Darstellungsfunktion der Sprache. Fischer. find
  5. Eve V. Clark. 1978. From Gesture to Word: On the Natural History of Deixis in Language Acquisition. In Jerome Seymour Bruner and Alison Garton, editors, Human Growth and Development, pages 85–120. Oxford University Press, Oxford, United Kingdom. find
  6. Potential Risks This study presents minimal risk. No sensitive personal information is collected, and all tasks involve non– sensitive linguistic judgements. Participants could choose to withdraw at any time without consequence. Participant Recruitment and Compensation Chinese participants are recru… find
  7. Kenny R. Coventry, Angelo Valdés, Jonathan Castillo, and Michael Guijarro-Fuentes. 2008. Language within your reach: Near-far perceptual space and spatial demonstratives. Cognition, 108(3):889–895. find
  8. DeepSeek. 2025. Deepseek v3.1: Open-source model release. https://www.deepseek.com/. Accessed 3 Jan 2026. ↗
  9. Holger Diessel. 1999. Demonstratives: Form, function, and grammaticalization. Typological Studies in Language. find
  10. Holger Diessel. 2013. Where does language come from? some reflections on the role of deictic gesture and demonstratives in the evolution of language. Language and Cognition, 5:239–249. find
  11. Holger Diessel. 2014. The Grammar of Demonstratives. Oxford University Press. find
  12. Holger Diessel and Sergei Monakhov. 2023. Acquisition of demonstratives in cross-linguistic perspective. Journal of Child Language, 50(4):922–953. Publisher: Cambridge University Press. find
  13. R. M. W. Dixon. 2010. Basic Linguistic Theory: Volume 1, Methodology. Oxford University Press, Oxford. find
  14. Guy O Dove. 2023. Rethinking the role of language in embodied cognition. Philosophical Transactions of the Royal Society B: Biological Sciences, 378(1870). find
  15. Jiale Duan, Shuai Yu, Hao L Tan, Hongbo Zhu, and Cheston Tan. 2022. A Survey of Embodied AI: From Simulators to Research Tasks. arXiv preprint arXiv:2103.04918. — [arXiv:2103.04918](https://arxiv.org/abs/2103.04918) find
  16. Patricia González-Peña, Martin J. Doherty, and Pedro Guijarro-Fuentes. 2020. Acquisition of Demonstratives in English and Spanish. Frontiers in Psychology, 11. find
  17. Google DeepMind. 2025. Gemini 2.5 Pro. https://deepmind.google/models/gemini/pro/. Accessed: 2025-05-25. ↗
  18. William B. Gudykunst and Stella Ting-Toomey. 1988. Culture and Interpersonal Communication. Sage Publications. find
  19. Edward T. Hall. 1976. Beyond Culture. Anchor Books. find
  20. John A Hawkins. 1984. A note on referent identifiability and co-presence. Journal of Pragmatics, 8(5–6):649– 659. find
  21. Nikolaus P. Himmelmann. 1997. Deictic Expressions, Articles, and Nominal Phrases: On the Emergence of Syntactic Structure, volume 362 of Linguistische Arbeiten. Max Niemeyer Verlag, Tubingen. find
  22. Anna A. Ivanova. 2025. How to evaluate the cognitive abilities of llms. Nature Human Behaviour, 9:230– 233. find
  23. Julian Jara-Ettinger and Paula Rubio-Fernandez. 2024. Demonstratives as attention tools: Evidence of mentalistic representations within language. Proceedings of the National Academy of Sciences, 121(32):e2402068121. find
  24. Samer O. Jarbou. 2010. Accessibility vs. physical proximity: An analysis of exophoric demonstrative practice in spoken jordanian arabic. Journal of Pragmatics, 42(11):3078–3097. find
  25. Hongru Ji, Yuyin Fan, Meng Zhao, Xianghua Li, Lianwei Wu, and Chao Gao. 2026. Stride-ed: A strategy-grounded stepwise reasoning framework for empathetic dialogue systems. Preprint, arXiv:2604.07100.
  26. Carina Kauf, Emmanuele Chersoni, Alessandro Lenci, Evelina Fedorenko, and Anna A Ivanova. 2024. Log probabilities are a reliable estimate of semantic plausibility in base and instruction-tuned language models. In Proceedings of the 7th BlackboxNLP Workshop: Analyzing and Interpreting Neural Networks…
  27. Carina Kauf, Anna A Ivanova, Giulia Rambelli, Emmanuele Chersoni, Jingyuan Selena She, Zawad Chowdhury, Evelina Fedorenko, and Alessandro Lenci. 2023. Event knowledge in large language models: the gap between the impossible and the unlikely. Cognitive Science, 47(11):e13386. find
  28. Krishnaram Kenthapadi, Mahima Sameki, and Amit Taly. 2024. Grounding and evaluation for large language models: Practical challenges and lessons learned. arXiv preprint. find
  29. Takeshi Kojima, Shixiang Shane Gu, Machel Reid, Yutaka Matsuo, and Yusuke Iwasawa. 2022. Large Language Models Are Zero-Shot Reasoners. In Advances in Neural Information Processing Systems.
  30. Aoxue Kong, Shuofei Zhao, Hong Chen, Qian Li, Yujia Qin, Ruoxi Sun, Yuxian Zhang, and Xin Dong. 2023. Better zero-shot reasoning with role-play prompting. arXiv preprint arXiv:2308.07702. find
  31. Stephen C. Levinson, Sarah Cutfield, Michael J. Dunn, N. J. Enfield, and Sérgio Meira, editors. 2018. Demonstratives in Cross-Linguistic Perspective, volume 14 of Language, Culture and Cognition. Cambridge University Press, Cambridge. find
  32. Xiaoyuan Li, Keqin Bao, Yubo Ma, Moxin Li, Wenjie Wang, Rui Men, Yichang Zhang, Fuli Feng, Dayiheng Liu, and Junyang Lin. 2025. Mtr-bench: A comprehensive benchmark for multi-turn reasoning evaluation. arXiv preprint arXiv:2505.17123.
  33. Hongye Liu and Ricardo Henao. 2025. Learning to substitute words with model-based score ranking. In Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers), pages 11551–11565. find
  34. Shuo Lu, Yinuo Xu, Jianjie Cheng, Lingxiao He, Meng Wang, and Jian Liang. 2025. Deepresearch-slice: Bridging the retrieval-utilization gap via explicit text slicing. arXiv preprint arXiv:2601.03261. — [arXiv:2601.03261](https://arxiv.org/abs/2601.03261) find
  35. Christopher Lyons. 1999. Definiteness. In Concise Encyclopedia of Grammatical Categories. Elsevier. find
  36. Chunlei Meng, Jiabin Luo, Zhenglin Yan, Zhenyu Yu, Rong Fu, Zhongxue Gan, and Chun Ouyang. 2026. Tri-subspaces disentanglement for multimodal sentiment analysis. arXiv preprint arXiv:2602.19585. — [arXiv:2602.19585](https://arxiv.org/abs/2602.19585) find
  37. OpenAI. 2025. Gpt-5.1 model card. https://openai.com. Accessed 3 Jan 2026. ↗
  38. David Peeters, Peter Hagoort, and Asli Özyürek. 2015. Electrophysiological evidence for the role of shared space in online comprehension of spatial demonstratives. Cognition, 136:64–84. find
  39. David Peeters and Asli Özyürek. 2016. Social grounding of demonstratives: An experimental study of communicative and cognitive factors. Cognitive Science, 40(1):255–276. find
  40. R. O. Robinson, W. Herzog, and B. M. Nigg. 1987. Use of force platform variables to quantify the effects of chiropractic manipulation on gait symmetry. European Journal of Applied Physiology and Occupational Physiology, 54(5):524–530. find
  41. Renato Rocca, Kenny R. Coventry, and David Peeters. 2019. The interplay of egocentric and sociocentric factors in demonstrative choice. Journal of Pragmatics, 140:1–16. find
  42. Larry A. Samovar and Richard E. Porter. 2004. Communication Between Cultures. Wadsworth/Thomson Learning. find
  43. Ben Shneiderman. 2020. Human-Centered AI. Oxford University Press, Oxford. First edition. find
  44. Yuxi Sun, Aoqi Zuo, Wei Gao, and Jing Ma. 2025. Causalabstain: Enhancing multilingual llms with causal reasoning for trustworthy abstention. In Findings of the Association for Computational Linguistics: ACL 2025, pages 14060–14076. find
  45. Emanuela Todisco, Mila Vulchanova, and Valentin Vulchanov. 2020. Demonstratives are frequent in early child language: Evidence from longitudinal corpora. Journal of Child Language, 47(6):1205–1233. find
  46. Mingyu Wan, Qi Su, Kathleen Ahrens, and Chu-Ren Huang. 2024. Perceptional and actional enrichment for metaphor detection with sensorimotor norms. Natural Language Engineering, 30(6):1181–1209. find
  47. Chengbing Wang, Yang Zhang, Wenjie Wang, Xiaoyan Zhao, Fuli Feng, Xiangnan He, and Tat-Seng Chua. 2025. Think-while-generating: On-the-fly reasoning for personalized long-form generation. arXiv preprint arXiv:2512.06690. — [arXiv:2512.06690](https://arxiv.org/abs/2512.06690) find
  48. Xiaoyang Wang, Liang Wu, Tianyi Ma, and Bing Liu. 2024. Fac 2 e: Better understanding large language model capabilities by dissociating language and cognition. In Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing (EMNLP), pages 13245–13259, Miami, Florida, USA. A… find
  49. Qihui Xu, Yingying Peng, Samuel A Nastase, Martin Chodorow, Minghua Wu, and Ping Li. 2025. Large language models without grounding recover non-sensorimotor but not sensorimotor features of human concepts. Nature human behaviour, 9(9):1871–1886. find
  50. Chao Xue, Di Liang, Pengfei Wang, and Jing Zhang. 2024. Question calibration and multi-hop modeling for temporal question answering. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 38, pages 19332–19340. find
  51. Chao Xue, Yao Wang, Mengqiao Liu, Di Liang, Xingsheng Han, Peiyang Liu, Xianjie Wu, Chenyao Lu, Lei Jiang, Yu Lu, Haibo Shi, Shuang Liang, Minlong Peng, and Flora D. Salim. 2026. Reason only when needed: Efficient generative reward modeling via model-internal uncertainty. Preprint, arXiv:2604.10072.
  52. An Yang, Anfeng Li, Baosong Yang, Beichen Zhang, Binyuan Hui, Bo Zheng, Bowen Yu, Chang Gao, Chengen Huang, Chenxu Lv, and 1 others. 2025. Qwen3 technical report. arXiv preprint arXiv:2505.09388.
  53. Shuang Zeng, Xinyuan Chang, Mengwei Xie, Xinran Liu, Yifan Bai, Zheng Pan, Mu Xu, Xing Wei, and Ning Guo. 2025. Futuresightdrive: Thinking visually with spatio-temporal cot for autonomous driving. arXiv preprint arXiv:2505.17685. — [arXiv:2505.17685](https://arxiv.org/abs/2505.17685) find
  54. Qianchi Zhang, Hainan Zhang, Liang Pang, Hongwei Zheng, and Zhiming Zheng. 2026. Stable-rag: Mitigating retrieval-permutation-induced hallucinations in retrieval-augmented generation. arXiv preprint arXiv:2601.02993.

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/wang-2026-do-llms-capture-embodied-cognition/