---
title: "Word Length Frequency and Distribution in English:"
description: "Constable had earlier proposed that writing metrical verse is a search through source text for runs of… 17 references, cited by 2 papers here."
authors: ["Hideaki Aoyama", "John Constable"]
year: 1998
venue: "arXiv:cmp-lg/9808004"
source: "https://arxiv.org/abs/cmp-lg/9808004"
doi: "10.48550/arxiv.cmp-lg/9808004"
arxiv: "cmp-lg/9808004"
licence: "arXiv — no licence stated (read only)"
summary_licence: "Elamur Cosmic License"
canonical: "https://research.elamur.ai/free/en/papers/aoyama-1998-word-length-frequency-distribution-english/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/aoyama-1998-word-length-frequency-distribution-english/)

[home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [1998](https://md.research.elamur.ai/free/en/papers/1998/) / Word Length Frequency and Distribution in English: Observations, Theory, and Implications for the Construction of Verse Lines

# Word Length Frequency and Distribution in English: Observations, Theory, and Implications for the Construction of Verse Lines

**Authors:** [Hideaki Aoyama](https://md.research.elamur.ai/free/en/authors/hideaki-aoyama/) and [John Constable](https://md.research.elamur.ai/free/en/authors/john-constable/) · **Year:** 1998 · **Venue:** arXiv:cmp-lg/9808004 · **Licence:** arXiv — no licence stated (read only)

**Links:** [Publisher's page](https://arxiv.org/abs/cmp-lg/9808004) · [PDF](https://arxiv.org/pdf/cmp-lg/9808004) · [DOI 10.48550/arxiv.cmp-lg/9808004](https://doi.org/10.48550/arxiv.cmp-lg/9808004) · [arXiv cmp-lg/9808004](https://arxiv.org/abs/cmp-lg/9808004) · [OpenAlex](https://openalex.org/W1675065936)

**Citations:** 1 citations (OpenAlex, 2026-10-14) · [2](https://md.research.elamur.ai/free/en/papers/aoyama-1998-word-length-frequency-distribution-english/cited-by/) in our library, of the 149,439 papers whose bibliographies we read

## Summary
Constable had earlier proposed that writing metrical verse is a search through source text for runs of complete words that add up to a target syllable count. He claimed that the number of such runs equals the word count divided by the mean syllables per word. Aoyama and Constable set out to prove this mathematically.

They used just under two million words of English prose, syllabified with a hand-built lexicon of 71,666 word-forms. For every target length n, they counted the word sequences whose syllables sum exactly to n.

The normalized count is nearly flat across n. This flatness follows if two things hold:
- word lengths are geometrically distributed, not following the adjusted Poisson preferred by the Göttingen Word Length Project;
- consecutive word lengths are roughly independent at the global level, though the fine structure shows real departures.

Generating functions tie the two observations together. The work matters for two reasons. It grounds a theory of verse composition in measurable statistics, and deviations from flatness give an empirical test for regular lineation.

## Key points
- **Corpus and lexicon:** about 2M words of prose, chosen for convenience. Syllable counts come from a single-annotator lexicon of 71,666 items, with fixed conventions for numbers, dates and abbreviations.
- **The flat constant:** the normalized frequency Q_n of word runs summing to n syllables averages q = 0.720316 over n = 1–30. It is close to constant, which matches the reciprocal of mean word length.
- **Methodology:** the authors reject fitting many-parameter curves to about seven data points, as the Göttingen work did with the Altmann Fitter. They prefer a simple global ansatz first and then a study of deviations from it.
- **Counting machinery:** a random-ordering hypothesis lets multi-word probabilities P_{n,k} be built from single-word probabilities p_n via generating functions. The counts treat the text as a circle (periodic boundary).
- **Verse test case:** Wordsworth's *Prelude* has 7,849 lines, 57,570 words and about 1.4 syllables per word. About 77.5% of its lines have ten syllables. Its Q_n shows peaks at 10 and every multiple of 10, plus a dip at n=2.

## Interesting ideas
- Use the simplest global model with far fewer parameters than data points, and treat residuals from it as the real signal; this avoids overfitting a handful of bins with a flexible curve.
- Test sequential independence by comparing observed counts of multi-item runs with counts predicted from single-item probabilities alone; any gap reveals local order.
- A flat statistic that breaks into periodic peaks reveals hidden structure such as fixed line length, even when layout is lost.
- A lexicon grown incrementally, where unknown forms trigger human input and one annotator ensures consistency, gives clean, explainable features.
- Shorter words create more ways to fill a fixed syllable budget, so a length constraint pushes word choice toward short words.

## For Mova

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

- **I1** (learn): Evaluations of parser or tagger statistics could first fit a minimal baseline model and then report deviations, rather than tuning many-parameter fits that explain nothing.
- **I2** (learn): The same check could act as a negative control on silver data, flagging corpora where token-level patterns differ from an independence baseline because of templating or duplication.
- **I3** (tokens): The tokenizer could use syllable-run peaks to detect verse or lineated text that was flattened into prose, and segment or tag it differently before parsing.
- **I4** (tokens): Syllable counts or stress patterns could be stored in a curated lexicon inside crate en, with unknown words queued for review instead of guessed silently.
- **I5** (back): When generating text under rhythm or length constraints, the generator could prefer lower-syllable synonyms to widen its candidate pool, and it could explain that choice.

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

📚 Shelves (3): [🎭 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 · [📚 Literature and Literary Theory](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/literature-and-literary-theory/) · 541 works

🧭 Topics (1): [English](https://md.research.elamur.ai/free/en/topics/english/) · 2,379 works

## Cite

Cite as: Hideaki Aoyama and John Constable. 1998. Word Length Frequency and Distribution in English: Observations, Theory, and Implications for the Construction of Verse Lines. arXiv:cmp-lg/9808004.

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

```bibtex
@misc{aoyama1998word,
  title = {{Word Length Frequency and Distribution in English: Observations, Theory, and Implications for the Construction of Verse Lines}},
  author = {Hideaki Aoyama and John Constable},
  year = {1998},
  doi = {10.48550/arxiv.cmp-lg/9808004},
  eprint = {cmp-lg/9808004},
  archivePrefix = {arXiv},
  url = {https://doi.org/10.48550/arxiv.cmp-lg/9808004}
}
```

- [← previous Rationality, Cooperation and Conversational Implicature](https://md.research.elamur.ai/free/en/papers/lee-1998-rationality-cooperation-conversational-implicature/)
- [next → Paradigmatic Treatment of Arabic Morphology](https://md.research.elamur.ai/free/en/papers/smets-1998-paradigmatic-treatment-arabic-morphology/)
- [↑ back to papers of 1998](https://md.research.elamur.ai/free/en/papers/1998/)

## References

17 references, 0 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. Attridge, D. (1982). The Rhythms of English Poetry. Longman, London. [find](https://scholar.google.com/scholar?q=Attridge%2C+D%2E+%281982%29%2E+The+Rhythms+of+English+Poetry%2E+Longman%2C+London%2E)
2. Attridge, D. (1995). Poetic Rhythm: An introduction. Cambridge University Press, Cambridge. [find](https://scholar.google.com/scholar?q=Attridge%2C+D%2E+%281995%29%2E+Poetic+Rhythm%3A+An+introduction%2E+Cambridge+University+Press%2C+Cambridge%2E)
3. Becker, C. (1996). Word Lengths in the Letters of the Chilean Author Gabriela Mistral. Journal of Quantitative Linguistics 3.128-131. [↗](https://doi.org/10.1080/09296179608599622)
4. Best, K.-H. and Altmannn, G. (1996). Project Report. Journal Quantitative Linguistics 3.85-88. [find](https://scholar.google.com/scholar?q=Best%2C+K%2E%2DH%2E+and+Altmannn%2C+G%2E+%281996%29%2E+Project+Report%2E+Journal+Quantitative+Linguistics+3%2E85%2D88%2E)
5. Constable, J. (1997). Verse Form; A Pilot Study in the Epidemiology of Representations. Human Nature 8.171-203. [find](https://scholar.google.com/scholar?q=Constable%2C+J%2E+%281997%29%2E+Verse+Form%3B+A+Pilot+Study+in+the+Epidemiology+of+Representations%2E+Human+Nature+8%2E171%2D203%2E)
6. Dittrich, H. (1996). Word Length Frequency in the Letters of G. E. Lessing. Journal of Quantitative Linguistics 3.26-264. [↗](https://doi.org/10.1080/09296179608599633)
7. Frischen, J. (1996). Word Length Analysis of Jane Austen’s Letters. Journal of Quantitative Linguistics 3.80-84. [↗](https://doi.org/10.1080/09296179608590066)
8. Li, W. (1992). Random Texts Exhibit Zipf’s-Law-Like Word Frequency Distribution. IEEE Transactions on Information Theory 38.1842-1845. [↗](https://doi.org/10.1109/18.165464)
9. Mandelbrot, B. (1961). On the Theory of Word Frequencies and on Related Markovian Models of Discourse. In Structure of Language and its Mathematical Aspects. Proceedings of Symposia in Applied Mathematics. 12.190-219. American Mathematical Society. [↗](https://doi.org/10.1090/psapm/012/9970)
10. Riedemann, H. (1996). Word Length Distribution in English Press Texts. Journal of Quantitative Linguistics 3.265-271. [↗](https://doi.org/10.1080/09296179608599634)
11. Röttger, W. (1996). Distribution of Word Length in Ciceronian Letters. Journal of Quantitative Linguistics 3.68-72. [↗](https://doi.org/10.1080/09296179608590064)
12. Wimmer, G., Köhler, R., Grotjahn, R., and Altmann, G. (1996). Towards a Theory of Word Length Distribution. Journal of Quantitative Linguistics, 1.98. [find](https://scholar.google.com/scholar?q=Wimmer%2C+G%2E%2C+Ko%CC%88hler%2C+R%2E%2C+Grotjahn%2C+R%2E%2C+and+Altmann%2C+G%2E+%281996%29%2E+Towards+a+Theory+of+Word+Length+Distribution%2E+Journal+of+Quantitative+Linguistics%2C+1%2E98%2E)
13. Wimmer, G. and Altmann, G. (1994). The theory of word length: some results and generalizations. Glottometrika, 15. [find](https://scholar.google.com/scholar?q=Wimmer%2C+G%2E+and+Altmann%2C+G%2E+%281994%29%2E+The+theory+of+word+length%3A+some+results+and+generalizations%2E+Glottometrika%2C+15%2E)
14. Wimsatt, W. K., and Beardsley, M. C. (1959). The Concept of Meter: An exercise in abstraction. PMLA 74.585-598. [↗](https://doi.org/10.2307/460509)
15. Ziegler, A. (1996). Word Length Distribution in Brazilian-Portuguese Texts. Journal of Quantitative Linguistics 3.73-79. [↗](https://doi.org/10.1080/09296179608590065)
16. Zipf, G. K. (1965). The Psychobiology of Language. M.I.T. Press, Cambridge Mass. [find](https://scholar.google.com/scholar?q=Zipf%2C+G%2E+K%2E+%281965%29%2E+The+Psychobiology+of+Language%2E+M%2EI%2ET%2E+Press%2C+Cambridge+Mass%2E)
17. Zuse, M. (1996). Distribution of Word Length in Early Modern English: Letters of Sir Philip Sidney. Journal of Quantitative Linguistics 3.272-6. Author Bunyan, John Eliot, George Frankau, Gilbert Title(s) Pilgrim’s Progress Middlemarch Woman of the Horizon (section; first 67 pages 10 chapters) Vicar of Wakefield The Altar of the Dead; The Ambassadors; The American; The Aspern Papers; Confidence; Daisy Miller; Death of the Lion; The Europeans; The Figure in the Carpet; The Golden Bowl; An International Episode; Portrait of a Lady; Roderick Hudson; Sacred Fount; Turn of the Screw; Watch and Ward; Washington Square Rewards and Fairies; The Jungle Book History of Britain; Colasterion; Martin Bucer Total Goldsmith, Oliver James, Henry Kipling, Rudyard Milton, John Table 1: Content of the corpus; authors, titles of the sources and number of words from each author Words 52,504 317,827 24,597 63,096 1,285,041 115,602 119,009 1,977,676 n L n,1 L n,2 L n,3 1,433,426 371,500 1,025,719 122,179 558,679 733,202 40,314 246,132 611,737 9,048 99,647 348,154 1,082 33,891 169,801 9,790 73,374 2,983 27,502 [↗](https://doi.org/10.1080/09296179608599635)

*Updated 1 November 2026*