--- title: "Computational Analysis of Predicational Structures in English" description: "To test this, he counts every predication in the Brown Corpus, about one million words of American English in 500 samples and… 2 references." authors: ["Henry Kucera"] year: 1980 venue: "COLING 1980 Volume 1: The 8th International Conference on Computational Linguistics" source: "https://aclanthology.org/C80-1006/" licence: "CC BY-NC-SA 3.0 (ACL Anthology, before 2016)" licence_url: "https://creativecommons.org/licenses/by-nc-sa/3.0/" summary_licence: "Elamur Cosmic License" canonical: "https://research.elamur.ai/free/en/papers/kucera-1980-computational-analysis-predicational-structures-english/" --- [Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/kucera-1980-computational-analysis-predicational-structures-english/) [home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [1980](https://md.research.elamur.ai/free/en/papers/1980/) / Computational Analysis of Predicational Structures in English # Computational Analysis of Predicational Structures in English **Authors:** Henry Kucera · **Year:** 1980 · **Venue:** [COLING 1980 Volume 1: The 8th International Conference on Computational Linguistics](https://md.research.elamur.ai/free/en/venues/coling/) ([COLING 1980](https://md.research.elamur.ai/free/en/venues/coling/1980/)) · **Licence:** [CC BY-NC-SA 3.0 (ACL Anthology, before 2016)](https://creativecommons.org/licenses/by-nc-sa/3.0/) **Links:** [Publisher's page](https://aclanthology.org/C80-1006/) · [PDF](https://aclanthology.org/C80-1006.pdf) · [OpenAlex](https://openalex.org/W1998469843) **Citations:** 8 citations (OpenAlex, 2026-10-14) ## Summary Kucera asks whether sentence length is a good stand-in for syntactic complexity. To test this, he counts every predication in the Brown Corpus, about one million words of American English in 500 samples and 15 genres. Complexity is measured as how many predications a sentence contains. Both finite and non-finite verbal groups count: infinitival complements, gerunds and participles. The verbal groups are retrieved by an algorithm that relies on the corpus's semi-automatic grammatical tagging, which uses 82 categories. Informative prose (non-fiction) has much longer sentences than imaginative prose (fiction). The gap in predications per sentence is far smaller and not consistent across genres. Words per predication split the two groups cleanly, which points to heavier noun phrases in non-fiction. Non-fiction also uses relatively more non-finite predications. The study matters because it shows that length is a poor proxy for clause structure, and it identifies noun-phrase processing as the main burden when parsing informational text. ## Key points - **Data:** - The Brown Corpus has 54,724 sentences, excluding headings, with a mean length of 18.49 words. - It contains 1,011,965 tokens and is tagged with 82 categories that distinguish modals and the verbs be, have and do. - Headlines were excluded because of their verb-dropping grammar. - Verbless sentences (1,869, or 3.4%) were kept in the statistics. - **Informative vs. imaginative prose:** | Measure | Informative | Imaginative | |---|---|---| | Words per sentence | 21.12 | 13.55 | | Predications per sentence | 2.80 | 2.38 | | Words per predication | 7.54 | 5.69 | The corpus averages are 18.49 words per sentence, 2.65 predications per sentence and 6.97 words per predication. - **Genre exceptions:** - Skills and hobbies falls below the corpus mean in predications per sentence. - Humor is well above it, at 2.84. - For words per predication, every informative genre is above the corpus mean and every imaginative genre is below it. - **Finite vs. non-finite:** Of 145,287 predications, 70.54% are finite and 29.46% non-finite. Non-finite predications make up 31.91% in informative prose and 24.04% in imaginative prose. - **Long verbal groups are rare:** - The five-element finite passive (may have been being considered) never occurs. - The four-element active (may have been considering) occurs 8 times. - The four-element passive (may have been considered) occurs 68 times. - Among non-finite groups, the three-element passive (to have been considered) occurs 22 times, and its active counterpart never occurs. ## Interesting ideas - Measuring complexity by counting predications rather than words separates sentences that look alike in length but differ in clause structure. A seven-word sentence can hold three predications. - Words per predication works as a cheap indirect gauge of noun-phrase heaviness, since what is not verbal is mostly nominal. - Non-finite verbal groups are treated as full predications with their own understood participants. In the paper's example, an infinitive chain implies that the asked person is the one who dances. - Imperatives with and without an overt subject are placed in one class, and quasi-auxiliaries are kept apart from true auxiliaries following Joos. These are explicit, consistent counting conventions for borderline verb forms. - The paper compares theoretically possible auxiliary chains against what is actually attested and finds that the longest forms never occur. ## For Mova How these ideas could serve Mova, our small open language model — its files, stages and tests: - **I1** (learn): Evaluation sets could be stratified by predication count, so that parser and reader scores show where multi-clause sentences break, independently of length. - **I2** (tree): Tracking this ratio per document could flag NP-heavy texts, such as learned or legal ones, where attachment errors inside noun phrases are likely and deserve targeted checks. - **I3** (frames): Each infinitive, gerund and participle should yield its own frame with a recovered controller or implicit subject, not be folded into the main verb. - **I4** (tokens): Documenting a fixed policy for constructions like going to and used to, and for subjectless imperatives, would keep the tagger, lemmatizer and downstream counts stable and explainable. - **I5** (back): The generator could prefer attested chain lengths and treat five-element passives as a red flag, while the parser could use the same distribution as a sanity check. *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 - [Brown University](https://md.research.elamur.ai/free/en/universities/brown-university/) — University, Providence, United States · 80 works · 26,036 citations 📚 Shelves (5): [🖥️ Computer Science](https://md.research.elamur.ai/free/en/shelves/computer-science/) · 18,614 works · [🤖 Artificial Intelligence](https://md.research.elamur.ai/free/en/shelves/computer-science/artificial-intelligence/) · 17,965 works · [🎭 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: Henry Kucera. 1980. Computational Analysis of Predicational Structures in English. COLING 1980 Volume 1: The 8th International Conference on Computational Linguistics. The paper itself, not this page. BibTeX from the library's record of the work. ```bibtex @inproceedings{kucera1980computational, title = {{Computational Analysis of Predicational Structures in English}}, author = {Henry Kucera}, booktitle = {COLING 1980 Volume 1: The 8th International Conference on Computational Linguistics}, year = {1980}, doi = {10.3115/990174.990181}, url = {https://doi.org/10.3115/990174.990181} } ``` - [← previous Slot Grammars](https://md.research.elamur.ai/free/en/papers/mccord-1980-slot-grammars/) - [next → Computer-Aided Grammatical Tagging of Spoken English](https://md.research.elamur.ai/free/en/papers/svartvik-1980-computer-aided-grammatical-tagging-spoken/) - [↑ back to papers of 1980](https://md.research.elamur.ai/free/en/papers/1980/) ## References 2 references, 0 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet. 1. 1. The list of tags and the rationale for the grammatical annotation system is given in W.N. Francis and H. Ku~era, Manual of Information to Accompany a Standard Corpus of Present-Day American English (Department of Linguistics, Brown University, Providence, 1979). Much of the computer programming required to produce the final form of the tagged Corpus, as well as the retrieval of the predications, was done by Andrew Mackie whose imaginative assistance is gratefully acknowledged. 2. Cf. Mary L. Marckworth and Laura M. [find](https://scholar.google.com/scholar?q=1%2E+The+list+of+tags+and+the+rationale+for+the+grammatical+annotation+system+is+given+in+W%2EN%2E+Francis+and+H%2E+Ku%7Eera%2C+Manual+of+Information+to+Accompany+a+Standard+Corpus+of+Present%2DDay+American+English) 2. Bell, "Sentence-Length Distribution in v the Corpus," in Henry Kucera and W. Nelson Francis, Computational Analysis of Present-Day American English (Brown i v e r s i t y Press, Providence, 1967). 3. Cf., for example, Noam Chomsky and Howard Lasnik, "Filters and Controls," Linguistic Inquir Z, Vol. 8, No. 3 (1977). 4. Martin Joos, The English Verb (The University of Wisconsin Press, Madison 1964). 37 • [find](https://scholar.google.com/scholar?q=Bell%2C+%22Sentence%2DLength+Distribution+in+v+the+Corpus%2C%22+in+Henry+Kucera+and+W%2E+Nelson+Francis%2C+Computational+Analysis+of+Present%2DDay+American+English+%28Brown+i+v+e+r+s+i+t+y+Press%2C+Providence%2C+1967%29%2E+3%2E+) *Updated 2 November 2026*