---
title: "Syntactic Variation Across the Grammar — paper by Dunn (2023)"
description: "Most studies of syntactic variation pick out one or two constructions and treat them as independent… 19 references, cited by 1 paper here."
authors: ["Jonathan Dunn"]
year: 2023
venue: "arXiv:2309.11869"
source: "https://arxiv.org/abs/2309.11869"
doi: "10.48550/arxiv.2309.11869"
arxiv: "2309.11869"
licence: "CC BY 4.0 (arXiv)"
licence_url: "http://creativecommons.org/licenses/by/4.0/"
summary_licence: "Elamur Cosmic License"
canonical: "https://research.elamur.ai/free/en/papers/dunn-2023-syntactic-variation-across-grammar-modelling/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/dunn-2023-syntactic-variation-across-grammar-modelling/)

[home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [2023](https://md.research.elamur.ai/free/en/papers/2023/page/14/) / Syntactic Variation Across the Grammar: Modelling a Complex Adaptive System

# Syntactic Variation Across the Grammar: Modelling a Complex Adaptive System

**Authors:** [Jonathan Dunn](https://md.research.elamur.ai/free/en/authors/jonathan-dunn/) · **Year:** 2023 · **Venue:** arXiv:2309.11869 · **Publisher:** [Frontiers Media](https://md.research.elamur.ai/free/en/publishers/frontiers-media/) · **Licence:** [CC BY 4.0 (arXiv)](http://creativecommons.org/licenses/by/4.0/)

**Links:** [Publisher's page](https://arxiv.org/abs/2309.11869) · [PDF](https://arxiv.org/pdf/2309.11869) · [DOI 10.48550/arxiv.2309.11869](https://doi.org/10.48550/arxiv.2309.11869) · [arXiv 2309.11869](https://arxiv.org/abs/2309.11869) · [OpenAlex](https://openalex.org/W4387187950)

**Citations:** 6 citations (OpenAlex, 2026-10-14) · 1 in our library, of the 149,439 papers whose bibliographies we read

## Summary
Most studies of syntactic variation pick out one or two constructions and treat them as independent variables. That ignores the fact that, in usage-based theory, a grammar is a connected network of constructions. Dunn tests how much this simplification distorts the picture. He models English dialects at three spatial levels: 7 regions, 16 countries and 49 local areas. The data is a geo-referenced tweet corpus of about 200M words. The features come from an unsupervised Computational Construction Grammar, learned on separate tweets from the same register.

Dialect classifiers are trained in two ways: once on the full grammar, and once on single macro-clusters (hundreds of constructions) or micro-clusters (dozens of constructions). Many clusters do carry dialect signal, but none matches the whole grammar. Dialect similarities are read off the classifiers' confusion errors, and these similarities shift sharply depending on which part of the grammar is used. The author concludes that much variation lives in interactions spread across the grammar. Single-construction studies therefore risk misleading conclusions about how dialects relate to one another.

## Key points
- **Corpus:** 200,727,956 words of tweets from 16 countries, split into 49,107 samples. Each sample is roughly 3,910 words and shares a fixed keyword distribution (Table 10), which controls for topic. Language ID was done with two existing models.
- **Spatial levels:** 7 regions, 16 countries and 49 local areas, the last produced by spatial clustering (the US alone has 9 local groups).
- **Features:** constructions from Computational Construction Grammar (Dunn 2017, 2022), grouped into macro- and micro-clusters through similarity inside the grammar itself, independently of the dialect model.
- **Distributed signal:** in an unmasking experiment, accuracy stays fairly high even after the top 25% of constructions are removed.
- **Shifting similarity:** dialect closeness depends on the feature subset. For example, New Zealand English can look closest to Australian English in phrasal verbs but closest to UK English in dative phrases.

## Interesting ideas
- Fixing the same keyword distribution in every sample keeps topic from masquerading as the target signal, so a classifier is forced to rely on structure.
- Similarity between categories is derived from a classifier's mistakes: classes that get confused more often are treated as closer, which turns the error matrix into a measurement tool.
- Unmasking, meaning repeatedly removing the strongest features and retraining, shows whether a signal is concentrated in a few features or spread widely.
- Running the same task on the whole feature system and on isolated sub-clusters shows when the whole carries information that no single part does.
- Repeating an analysis at several nested granularities (region, country, local area) reveals whether conclusions hold across scales.

## For Mova

How these ideas could serve Mova, our small open language model — its files, stages and tests:

- **I1** (learn): build evaluation sets for parser or reader comparisons where topic words are balanced across conditions, so gains cannot come from topical shortcuts.
- **I2** (tree): compare treebanks or UD dialects by how often a classifier on parse features confuses them, to decide which conversions are most needed.
- **I3** (learn): use it as a negative-control check on self-induced repair rules or silver-data models, to detect reliance on a handful of brittle cues.
- **I4** (frames): evaluate role assignment with all VerbNet classes and with single class families, to see whether errors come from interactions between classes rather than from one family.
- **I5** (answer): report reader accuracy per course, per topic and per individual text, so improvements are confirmed at every level instead of only in the aggregate.

*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

- [University of Canterbury](https://md.research.elamur.ai/free/en/universities/university-of-canterbury/) — University, Christchurch, New Zealand · 5 works · 274 citations

📚 Shelves (4): [🎭 Arts and Humanities](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/) · 17,450 works · [🗣️ Language and Linguistics](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/language-and-linguistics/) · 17,296 works · [💭 Psychology](https://md.research.elamur.ai/free/en/shelves/psychology/) · 1,649 works · [💭 Developmental and Educational Psychology](https://md.research.elamur.ai/free/en/shelves/psychology/developmental-and-educational-psychology/) · 523 works

## Cite

Cite as: Jonathan Dunn. 2023. Syntactic Variation Across the Grammar: Modelling a Complex Adaptive System. arXiv:2309.11869.

The paper itself, not this page. BibTeX from the library's record of the work.

```bibtex
@misc{dunn2023syntactic,
  title = {{Syntactic Variation Across the Grammar: Modelling a Complex Adaptive System}},
  author = {Jonathan Dunn},
  year = {2023},
  doi = {10.48550/arxiv.2309.11869},
  eprint = {2309.11869},
  archivePrefix = {arXiv},
  url = {https://doi.org/10.48550/arxiv.2309.11869}
}
```

- [← previous Rosetta Stone at KSAA-RD Shared Task: A Hop From Language Modeling To Word–Definition Alignment](https://md.research.elamur.ai/free/en/papers/elbakry-2023-rosetta-stone-ksaa-rd-shared/)
- [next → Towards more Human-like Language Models based on Contextualizer Pretraining Strategy](https://md.research.elamur.ai/free/en/papers/xiao-2023-more-human-like-language-models/)
- [↑ back to papers of 2023](https://md.research.elamur.ai/free/en/papers/2023/page/14/)

## References

19 references, 6 of them with a card of ours. ↗ leads to the work.

1. Barbaresi, A. (2018). Computationally efficient discrimination between language varieties with large feature vectors and regularized classifiers. Proceedings of the Fifth Workshop on NLP for Similar Languages, Varieties and Dialects , 164–171 Beckner, C., Ellis, N., Blythe, R., Holland, J., Bybee, J., Ke, J., et al. (2009). Language Is a Complex Adaptive System: Position Paper. Language Learning 59, 1–26 Belinkov, Y. and Glass, J. (2016). [A Character-level Convolutional Neural Network for Distinguishing Similar Languages and Dialects](https://md.research.elamur.ai/free/en/papers/belinkov-2016-character-level-convolutional-neural-network/). In Proceedings of the Third Workshop on NLP for Similar Languages, Varieties and Dialects (Association for Computational Linguistics), 145–152 Buchstaller, I. (2008). The localization of global linguistic variants. English World-Wide. A Journal of Varieties of English 29, 15–44. doi:10.1075/eww.29.1.03buc [↗](https://doi.org/10.1075/eww.29.1.03buc)
2. Bybee, J. (2007). Frequency of Use and the Organization of Language (Oxford University Press) Calle-Martin, J. and Romero-Barranco, J. (2017). Third person present tense markers in some varieties of English. English World-Wide 38, 77–103 Campello, R., Moulavi, D., Zimek, A., and Sandler, J. (2015). Hierarchical density estimates for data clustering, visualization, and outlier detection. ACM Transactions on Knowledge Discovery from Data 10. doi:https://doi.org/10.1145/2733381 [↗](https://doi.org/10.1145/2733381)
3. Campello, R. J. G. B., Moulavi, D., and Sander, J. (2013). Density-based clustering based on hierarchical density estimates. In Advances in Knowledge Discovery and Data Mining (Springer Berlin Heidelberg). 160–172. doi:10.1007/978-3-642-37456-2 14 Deshors, S. C. and Götz, S. (2020). Common ground across globalized english varieties: A multivariate exploration of mental predicates in world englishes. Corpus Linguistics and Linguistic Theory 16, 1–28. doi:doi:10.1515/cllt-2016-0052 [↗](https://doi.org/10.1007/978-3-642-37456-2)
4. Donoso, G., Sánchez, D., and Sanchez, D. (2017). Dialectometric analysis of language variation in Twitter. In Proceedings of the Fourth Workshop on NLP for Similar Languages, Varieties and Dialects (VarDial) (Valencia, Spain: Association for Computational Linguistics), vol. 4, 16–25. doi:10.18653/v1/W17-1202 [↗](https://doi.org/10.18653/v1/w17-1202)
5. Doumen, J., Beuls, K., and Van Eecke, P. (2023). Modelling language acquisition through syntactico-semantic pattern finding. In Findings of the Association for Computational Linguistics: EACL 2023 (Dubrovnik, Croatia: Association for Computational Linguistics), 1347–1357 Dunn, J. (2017). Computational learning of construction grammars. Language & Cognition 9, 254–292 Dunn, J. (2018a). Finding variants for construction-based dialectometry: A corpus-based approach to regional cxgs. Cognitive Linguistics 29, 275–311 Dunn, J. (2018b). [Modeling the Complexity and Descriptive Adequacy of Construction Grammars](https://md.research.elamur.ai/free/en/papers/dunn-2018-modeling-complexity-descriptive-adequacy-construction/). In Proceedings of the Society for Computation in Linguistics , 81–90 Dunn, J. (2019a). Frequency vs. Association for Constraint Selection in Usage-Based Construction Grammar. In Proceedings of the Workshop on Cognitive Modeling and Computational Linguistics (Association for Computational Linguistics) Dunn, J. (2019b). Global Syntactic Variation in Seven Languages: Toward a Computational Dialectology. Frontiers in Artificial Intelligence 2, 15. doi:10.3389/frai.2019.00015 [↗](https://doi.org/10.3389/frai.2019.00015)
6. Dunn, J. (2019c). [Modeling Global Syntactic Variation in English Using Dialect Classification](https://md.research.elamur.ai/free/en/papers/dunn-2019-modeling-global-syntactic-variation-english/). In Proceedings of the Sixth Workshop on NLP for Similar Languages, Varieties and Dialects (Ann Arbor, Michigan: Association for Computational Linguistics), 42–53. doi:10.18653/v1/W19-1405 [↗](https://doi.org/10.18653/v1/w19-1405)
7. Dunn, J. (2020). Mapping languages: the Corpus of Global Language Use. Language Resources and Evaluation 54, 999–1018. doi:10.1007/s10579-020-09489-2 [↗](https://doi.org/10.1007/s10579-020-09489-2)
8. Dunn, J. (2022). Exposure and emergence in usage-based grammar: Computational experiments in 35 languages. [↗](https://doi.org/10.1515/cog-2021-0106)
9. Cognitive Linguistics Dunn, J. (2023a). Exploring the constructicon: Linguistic analysis of a computational CxG. In [Proceedings of the First International Workshop on Construction Grammars and NLP (CxGs+NLP, GURT/SyntaxFest 2023)](https://md.research.elamur.ai/free/en/papers/anon-2023-proceedings-first-international-workshop-construction/) (Washington, D.C.: Association for Computational Linguistics), 1–11 Dunn, J. (2023b). Variation and instability in dialect-based embedding spaces. In Tenth Workshop on NLP for Similar Languages, Varieties and Dialects (VarDial 2023) (Dubrovnik, Croatia: Association for Computational Linguistics), 67–77 Dunn, J. (In Revision). Computational Construction Grammar: A Usage-Based Approach (Cambridge University Press) Dunn, J. and Nijhof, W. (2022). Language identification for austronesian languages. In Proceedings of the 13th International Conference on Language Resources and Evaluation (European Language Resources Association), 6530–6539 Dunn, J. and Nini, A. (2021). Production vs Perception: The Role of Individuality in Usage-Based Grammar Induction. In Proceedings of the Workshop on Cognitive Modeling and Computational Linguistics (Association for Computational Linguistics), 149–159 [↗](https://doi.org/10.48550/arxiv.2301.12642)
10. Dunn, J. and Tayyar Madabushi, H. (2021). [Learned Construction Grammars Converge Across Registers Given Increased Exposure](https://md.research.elamur.ai/free/en/papers/dunn-2021-learned-construction-grammars-converge-across/). In Conference on Natural Language Learning (Association for Computational Linguistics) Dunn, J. and Wong, S. (2022). Stability of syntactic dialect classification over space and time. In Proceedings of the 29th International Conference on Computational Linguistics (Gyeongju, Republic of Korea: International Committee on Computational Linguistics), 26–36 Eisenstein, J., O’Connor, B., Smith, N., and Xing, E. (2010). A latent variable model for geographic lexical variation. In Proceedings of the Conference on Empirical Methods in Natural Language Processing (287: Association for Computational Linguistics), 221–227 Eisenstein, J., O’Connor, B., Smith, N., and Xing, E. (2014). Diffusion of lexical change in social media. PloSOne 10, 1371 Gamallo, P., Pichel, J., Algeria, I., and Agirrezabal, M. (2016). Comparing two Basic Methods for Discriminating Between Similar Languages and Varieties. Proceedings of the Third Workshop on NLP for Similar Languages, Varieties and Dialects , 170–177 Goldberg, A. (2006). Constructions at work: The nature of generalization in language (Oxford: Oxford University Press) Grafmiller, J. and Szmrecsanyi, B. (2018). Mapping out particle placement in Englishes around the world A study in comparative sociolinguistic analysis. Language Variation and Change 30, 385–412 Grieve, J. (2012). A statistical analysis of regional variation in adverb position in a corpus of written Standard American English. Corpus Linguistics and Linguistic Theory 8, 39–72 Grieve, J. (2016). Regional variation in written American English (Cambridge, UK: Cambridge University Press) Grieve, J., Montgomery, C., Nini, A., Murakami, A., and Guo, D. (2019). Mapping Lexical Dialect Variation in British English Using Twitter. Frontiers in Artificial Intelligence 2, 11. doi:10.3389/frai. 2019.00011 Kachru, B. (1990). The Alchemy of English The spread, functions, and models of non-native Englishes (Urbana-Champaign: University of Illinois Press) Koppel, M., Schler, J., and Bonchek-Dokow, E. (2007). Measuring differentiability: Unmasking pseudonymous authors. Journal of Machine Learning Research 8, 1261–1276 Kroon, M., Medvedeva, M., and Plank, B. (2018). When Simple n-gram Models Outperform Syntactic Approaches Discriminating between Dutch and Flemish. Proceedings of the Fifth Workshop on NLP for Similar Languages, Varieties and Dialects , 225–244 Laitinen, M., Fatemi, M., and Lundberg, J. (2020). Size Matters: Digital Social Networks and Language Change. Frontiers in Artificial Intelligence 3, 1–15. doi:10.3389/frai.2020.00046 [↗](https://doi.org/10.3389/frai)
11. Langacker, R. (2008). Cognitive Grammar: A Basic Introduction (Oxford: Oxford University Press) Larsson, M. M. (2023). Modelling incipient probabilistic grammar change in real time: the grammaticalisation of possessive pronouns in european spanish locative adverbial constructions. Corpus Linguistics and Linguistic Theory 19, 177–206. doi:doi:10.1515/cllt-2021-0030 [↗](https://doi.org/10.1515/cllt-2021-0030)
12. Li, Y., Szmrecsanyi, B., and Zhang, W. (2023). The theme-recipient alternation in chinese: tracking syntactic variation across seven centuries. Corpus Linguistics and Linguistic Theory 19, 207–235. doi:doi:10.1515/cllt-2021-0048 [↗](https://doi.org/10.1515/cllt-2021-0048)
13. Lucy, L. and Bamman, D. (2021). [Characterizing english variation across social media communities with bert](https://md.research.elamur.ai/free/en/papers/lucy-2021-characterizing-english-variation-across-social/). Transactions of the Association for Computational Linguistics 9, 538–556. doi:10.1162/ tacl a 00383 [↗](https://doi.org/10.1162/tacl_a_00383)
14. McInnes, L. and Healy, J. (2017). Accelerated hierarchical density clustering. In IEEE International Conference on Data Mining Workshops (ICDMW) (IEEE), 33–42. doi:https://doi.org/10.1109/ICDMW. 2017.12 Mocanu, D., Baronchelli, A., Perra, N., Gonçalves, B., Zhang, Q., Vespignani, A., et al. (2013). The Twitter of Babel: Mapping World Languages through Microblogging Platforms. PLOSOne 10, 1371 Nevens, J., Doumen, J., Van Eecke, P., and Beuls, K. (2022). Language acquisition through intention reading and pattern finding. In Proceedings of the 29th International Conference on Computational Linguistics (Gyeongju, Republic of Korea: International Committee on Computational Linguistics), 15–25 Rahimi, A., Cohn, T., and Baldwin, T. (2017). A neural model for user geolocation and lexical dialectology. ACL 2017 - 55th Annual Meeting of the Association for Computational Linguistics, Proceedings of the Conference (Long Papers) 2, 209–216. doi:10.18653/v1/P17-2033 [↗](https://doi.org/10.1109/icdmw)
15. Rautionaho, P. and Hundt, M. (2022). Primed progressives? predicting aspectual choice in world englishes. Corpus Linguistics and Linguistic Theory 18, 599–625. doi:doi:10.1515/cllt-2021-0012 [↗](https://doi.org/10.1515/cllt-2021-0012)
16. Schilk, M. and Schaub, S. (2016). Noun phrase complexity across varieties of English: Focus on syntactic function and text type. English World-Wide 37, 58–85 Schneider, G., Hundt, M., and Schreier, D. (2020). Pluralized non-count nouns across englishes: A corpus-linguistic approach to variety types. Corpus Linguistics and Linguistic Theory 16, 515–546. doi:doi:10.1515/cllt-2018-0068 [↗](https://doi.org/10.1515/cllt-2018-0068)
17. Szmrecsanyi, B., Grafmiller, J., and Rosseel, L. (2019). Variation-Based Distance and Similarity Modeling: A Case Study in World Englishes. Frontiers in Artificial Intelligence 2, 23. doi:10.3389/frai.2019.00023 [↗](https://doi.org/10.3389/frai.2019.00023)
18. Xu, H., Jiang, M., Lin, J., and Huang, C.-R. (2022). Light verb variations and varieties of mandarin chinese: Comparable corpus driven approaches to grammatical variations. Corpus Linguistics and Linguistic Theory 18, 145–173. doi:doi:10.1515/cllt-2019-0049 [↗](https://doi.org/10.1515/cllt-2019-0049)
19. Zampieri, M., Nakov, P., and Scherrer, Y. (2020). Natural language processing for similar languages, varieties, and dialects: A survey. Natural Language Engineering 26, 595–612. doi:10.1017/ S1351324920000492 [↗](https://doi.org/10.1017/s1351324920000492)

*Updated 1 November 2026*