Elamur Research⌕
☰
home / papers / 2026 / Rethinking the Idiomaticity Decomposability Hypothesis: Evidence from Distributional Learning

Rethinking the Idiomaticity Decomposability Hypothesis: Evidence from Distributional Learning

Summary

The paper tests the Idiom Decomposability Hypothesis (IDH), a classic claim that an idiom permits more syntactic flexibility the more its figurative meaning can be traced back to its parts. Usage-based theories offer an alternative, explaining idiom behaviour through exposure, familiarity and predictability. To separate these accounts, the authors treat contextualised language models as learners that receive only distributional input.

They propose a decomposability score computed from model-internal representations. This score is compared with human decomposability ratings, corpus-based syntactic flexibility measures and surprisal. They also follow how idiom representations develop across OLMo pretraining checkpoints.

The findings do not support the IDH:

  • The model-based score aligns only weakly with human judgments.
  • It shows a small negative link with flexibility, even though the IDH predicts a positive one.
  • The predicted pattern fails even for verb phrase idioms, where it should be strongest.

During pretraining, frequency alone does not explain how idiom representations stabilise. Surprisal and decomposability matter more. The work shows that probing model internals can help settle debates in linguistic theory.

Key points

  • Decomposability derived from models correlates weakly and positively with human ratings. The two measures appear to capture related but different properties.
  • Across models and layers, decomposability correlates weakly and negatively with syntactic flexibility. The effect is strongest for prepositional phrasal idioms.
  • Frequency and decomposability show a robust negative relationship, which fits the idea that frequent idioms are stored as wholes.
  • The pretraining analysis covers OLMo Stage 1 checkpoints, using decomposability scores taken from BERT-large.
    • Regressions include step × predictor interactions with z-scored variables and HC3 robust errors.
    • All interactions are negative and significant (p < .001).
    • Decomposability has the largest interaction effect.
  • The study is limited to English idioms. Code is available at github.com/mi-m1/idiom_decomp.

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

Fields: Linguistics & languages

Cite

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2026.acl-long.1579

@inproceedings{mi2026rethinking,
  title = {{Rethinking the Idiomaticity Decomposability Hypothesis: Evidence from Distributional Learning}},
  author = {Mi, Maggie and Atefi, Golzar and Yamaguchi, Atsuki and Gers, Felix and Villavicencio, Aline and Moosavi, Nafise Sadat},
  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.1579},
  url = {https://doi.org/10.18653/v1/2026.acl-long.1579}
}

References

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

  1. Inbal Arnon and Neal Snider. 2010. More than words: Frequency effects for multi-word phrases. Journal of memory and language, 62(1):67–82. find
  2. Nyssa Z. Bulkes and Darren Tanner. 2017. “going to town”: Large-scale norming and statistical analysis of 870 american english idioms. Behavior Research Methods, 49(2):772–783. find
  3. Joan Bybee. 2010. Language, Usage and Cognition. Cambridge University Press. find
  4. Cristina Cacciari and Sam Glucksberg. 1991. Chapter 9 understanding idiomatic expressions: The contribution of word meanings. In Greg B. Simpson, editor, Understanding Word and Sentence, volume 77 of Advances in Psychology, pages 217–240. North-Holland. find
  5. Cristina Cacciari and Maria Chiara Levorato. 1998. The effect of semantic analyzability of idioms in metalinguistic tasks. Metaphor and Symbol, 13(3):159–177. find
  6. Cristina Cacciari and Patrizia Tabossi. 1988. The comprehension of idioms. Journal of Memory and Language, 27(6):668–683. find
  7. Tyler A. Chang and Benjamin K. Bergen. 2022. Word acquisition in neural language models. Transactions of the Association for Computational Linguistics, 10:1–16. find
  8. Noam Chomsky. 1980. Rules and representations. Behavioral and Brain Sciences, 3(1):1–15. find
  9. Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. 2019. BERT: Pre-training of deep bidirectional transformers for language understanding. In Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies,… find
  10. Filippo Ficarra, Ryan Cotterell, and Alex Warstadt. 2025. A distributional perspective on word learning in neural language models. In Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Lon… find
  11. Bruce Fraser. 1970. Idioms within a transformational grammar. Foundations of language, pages 22–42. find
  12. Raymond W Gibbs and Nandini P Nayak. 1989. Psycholinguistic studies on the syntactic behavior of idioms. Cognitive Psychology, 21(1):100–138. find
  13. Raymond W Gibbs, Nandini P Nayak, John L Bolton, and Melissa E Keppel. 1989a. Speakers’ assumptions about the lexical flexibility of idioms. Memory & cognition, 17(1):58–68. find
  14. Raymond W. Gibbs, Nandini P. Nayak, and Cooper Cutting. 1989b. How to kick the bucket and not decompose: Analyzability and idiom processing. Journal of Memory and Language, 28(5):576–593. find
  15. Raymond W Gibbs Jr and Herbert L Colston. 2012. Interpreting figurative meaning. Cambridge University Press. find
  16. Adele Goldberg. 2005. Constructions at Work: The Nature of Generalization in Language. Oxford University Press. find
  17. John Hale. 2001. A probabilistic Earley parser as a psycholinguistic model. In Second Meeting of the North American Chapter of the Association for Computational Linguistics. find our card
  18. James T. Heringer. 1976. Idioms and Lexicalization in English, pages 205 – 216. Brill, Leiden, The Netherlands. find
  19. Ryan Hubbard, Nyssa Bulkes, and Vicky Tzuyin Lai. 2023. Predictability and decomposability separately contribute to compositional processing of idiomatic language. Psychophysiology, 60(8):e14269. find
  20. Milo Jakubíek, Adam Kilgarriff, Vojtěch Kovář, P. Rychlý, and Vít Suchomel. 2013. The tenten corpus family. find
  21. Ganesh Jawahar, Benoît Sagot, and Djamé Seddah. 2019. What does BERT learn about the structure of language? In Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, pages 3651–3657, Florence, Italy. Association for Computational Linguistics. find our card
  22. L. Kantorovitch. 1958. On the translocation of masses. Management Science, 5:1–4. find
  23. Jerrold J. Katz. 1973. Compositionality, idiomaticity, and lexical substitution. In Stephen R. Anderson and Paul Kiparsky, editors, A Festschrift for Morris Halle, pages 357–376. Holt, Rinehart and Winston, New York. find
  24. Simon Kornblith, Mohammad Norouzi, Honglak Lee, and Geoffrey Hinton. 2019. Similarity of neural network representations revisited. In International conference on machine learning, pages 3519–3529. PMlR. find our card
  25. Roger Levy. 2008. Expectation-based syntactic comprehension. Cognition, 106(3):1126–1177. find
  26. Maggie Mi, Aline Villavicencio, and Nafise Sadat Moosavi. 2025a. From input perception to predictive insight: Modeling model blind spots before they become errors. In Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing, pages 34328–34341, Suzhou, China. Association… find
  27. Maggie Mi, Aline Villavicencio, and Nafise Sadat Moosavi. 2025b. Rolling the DICE on idiomaticity: How LLMs fail to grasp context. In Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 7314–7332, Vienna, Austria. Association for Com… find
  28. Emily Nordmann, Alexandra A. Cleland, and Rebecca Bull. 2014. Familiarity breeds dissent: Reliability analyses for british-english idioms on measures of familiarity, meaning, literality, and decomposability. Acta Psychologica, 149:87–95. Including Special section articles of Temporal Processing With… find
  29. Geoffrey Nunberg. 1978. The Pragmatics of Reference. Indiana University Linguistics Club, Bloomington, IN. find
  30. Geoffrey Nunberg, Ivan A Sag, and Thomas Wasow. 1994. Idioms. Language, 70(3):491–538. find
  31. OpenAI. 2022. Chatgpt: Optimizing language models for dialogue. https://openai.com/blog/ chatgpt. Accessed: 2026-01-05. ↗
  32. Barbara H. Partee. 1995. Lexical semantics and compositionality., pages 311–360. An invitation to cognitive science. The MIT Press, Cambridge, MA, US. find
  33. Susanne Riehemann. 2001. A constructional approach to idioms and word formation. find
  34. Ivan A. Sag, Timothy Baldwin, Francis Bond, Ann A. Copestake, and Dan Flickinger. 2002. Multiword expressions: A pain in the neck for nlp. In Proceedings of the Third International Conference on Computational Linguistics and Intelligent Text Processing, CICLing ’02, page 1–15, Berlin, Heidelberg. Sp… find our card
  35. C. E. Shannon. 1948. A mathematical theory of communication. The Bell System Technical Journal, 27(3):379–423. find
  36. Maya R Libben and Debra A Titone. 2008. The multidetermined nature of idiom processing. Memory & cognition, 36(6):1103–1121. find
  37. Livnat Herzig Sheinfux, Tali Greshler, Nurit Melnik, and Shuly Winter. 2019. Verbal multiword expressions: Idiomaticity and flexibility. Representation and parsing of multiword expressions: Current trends, 3:35. find
  38. Jiacheng Liu, Sewon Min, Luke Zettlemoyer, Yejin Choi, and Hannaneh Hajishirzi. 2024. Infini-gram: Scaling unbounded n-gram language models to a trillion tokens. arXiv preprint arXiv:2401.17377. find
  39. Anna Vogel Sosa and James MacFarlane. 2002. Evidence for frequency-based constituents in the mental lexicon: Collocations involving the word of. Brain and language, 83(2):227–236. find
  40. Simone A Sprenger, Willem JM Levelt, and Gerard Kempen. 2006. Lexical access during the production of idiomatic phrases. Journal of memory and language, 54(2):161–184. find
  41. Kevin Stowe, Prasetya Utama, and Iryna Gurevych. 2022. Impli: Investigating nli models’ performance on figurative language. In Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 5375– 5388. find our card
  42. Jurg Strassler. 1982. Idioms in english. A Pragmatic Analysis. Tubingen: Narr. find
  43. Patrizia Tabossi, Rachele Fanari, and Karoline Wolf. 2009a. Why are idioms recognized fast? Memory & Cognition, 37(4):529–540. find
  44. Patrizia Tabossi, Rachele Fanari, and Kristine Wolf. 2009b. Processing idiomatic expressions: Effects of semantic compositionality. Journal of Experimental Psychology: Learning, Memory, and Cognition, 35(2):313–327. find
  45. Team Olmo, Allyson Ettinger, Amanda Bertsch, Bailey Kuehl, David Graham, David Heineman, Dirk Groeneveld, Faeze Brahman, Finbarr Timbers, Hamish Ivison, Jacob Morrison, Jake Poznanski, Kyle Lo, Luca Soldaini, Matt Jordan, Mayee Chen, Michael Noukhovitch, Nathan Lambert, Pete Walsh, and 49 others. 20… find
  46. Ian Tenney, Dipanjan Das, and Ellie Pavlick. 2019. BERT rediscovers the classical NLP pipeline. In Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, pages 4593– 4601, Florence, Italy. Association for Computational Linguistics. find our card
  47. Debra A. Titone and Cynthia M. Connine. 1994. Descriptive norms for 171 idiomatic expressions: Familiarity, compositionality, predictability, and literality. Metaphor and Symbolic Activity, 9(4):247–270. find
  48. Debra A. Titone and Cynthia M. Connine. 1999. On the compositional and noncompositional nature of idiomatic expressions. Journal of Pragmatics, 31(12):1655–1674. Literal and Figurative Language. find
  49. Hugo Touvron, Thibaut Lavril, Gautier Izacard, Xavier Martinet, Marie-Anne Lachaux, Timothée Lacroix, Baptiste Rozière, Naman Goyal, Eric Hambro, Faisal Azhar, Aurelien Rodriguez, Armand Joulin, Edouard Grave, and Guillaume Lample. 2023. Llama: Open and efficient foundation language models. Preprint… find
  50. L. N. Vaserstein. 1969. Markov processes over denumerable products of spaces, describing large systems of automata. Problemy Peredachi Informatsii, 5(3):64– 72. find
  51. Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N Gomez, Ł ukasz Kaiser, and Illia Polosukhin. 2017. Attention is all you need. In Advances in Neural Information Processing Systems, volume 30. Curran Associates, Inc. find
  52. Mila Vulchanova, Evelyn Milburn, Valentin Vulchanov, and Giosuè Baggio. 2019. Boon or burden? the role of compositional meaning in figurative language processing and acquisition. Journal of Logic, Language and Information, 28(2):359–387. find
  53. Evan 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, Allyson Ettinger, and 23 others… find
  54. Benjamin Warner, Antoine Chaffin, Benjamin Clavié, Orion Weller, Oskar Hallström, Said Taghadouini, Alexis Gallagher, Raja Biswas, Faisal Ladhak, Tom Aarsen, Griffin Thomas Adams, Jeremy Howard, and Iacopo Poli. 2025. Smarter, better, faster, longer: A modern bidirectional encoder for fast, memory e… find our card
  55. Thomas Wasow, Ivan Sag, and Geoffrey Nunberg. 1983. Idioms: An interim report. In Proceedings of the XII-Ith international congress of linguistics, volume 29. Tokyo. find
  56. Marta Wierzba, Jessica MM Brown, and Gisbert Fanselow. 2023. Sources of variability in the syntactic flexibility of idioms. Glossa: a journal of general linguistics, 8(1). find

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/mi-2026-rethinking-idiomaticity-decomposability-hypothesis-evidence/