Cohort Decay, Inventory Stability and the Retrospective Drift of English Verb Regularization

Creator: Kriger, Boris. Affiliations: Information Physics Institute, Gosport, Hampshire, United Kingdom; Institute of Integrative and Interdisciplinary Research, Toronto, Canada. ORCID: 0009-0001-0034-2903 Publication date: 2026-11-18 Description: The best-known quantitative result on English morphological change is the finding of Lieberman et al. (Nature, 2007). Of 177 Old English irregular verbs, 98 remain irregular today, and the half-life of irregularity scales as the square root of frequency. This result is usually read as evidence that English moves in one direction, toward regularity. This paper argues that the reading confuses two quantities: the decay of an anchored cohort and the change of an inventory. A cohort of forms fixed at a starting date can only shrink, whatever the language does, because forms that become irregular later are excluded by construction. This is a case of retrospective drift: a record conditioned on survival from a fixed origin shows apparent direction even when the system is stationary. The distinction is made formal, and the square-root frequency law is shown to be reproduced exactly by a birth–death process with no net direction. Both quantities are then measured on the forward record. The data are pronoun-anchored preterite bigrams (he/she + past form) from the Google Books 2019 corpora for 150 English pairs. American English, 1800–2019: the cohort lost 4.2% while the irregular inventory went from 120 to 122. Six pairs became majority-irregular (snuck, dove, lit, quit, fit) and six became regular. The verb predicted to regularize next: wed fell to an irregular share of 0.10 in the 1910s and then rose to 0.87. Millennial scale: post-Old English irregulars still in use put a lower bound of 63% on inventory retention, against 55% for the cohort. Rhyme: the form with the larger rhyme family gains share in 19 of 22 American and 22 of 25 British pairs (sign test p < 0.001). German control (de-2019, 39 pairs): the language has a stronger standardizing filter. It shows no births and no deaths, a threefold slowdown of change since 1820, and the rise and suppression of the analogical frug (peak 0.11 in the 1880s). The paper sets out a mechanism of births: audibility of the suffix, dialect and learner origin, amplification by song and rap, and joining a productive minority pattern. It relates the cross-linguistic fate of analogical preterites to cultural tightness. It also predicts that generative language models will act as a filter that freezes the written norm, while speech continues to change. The resulting gap between speech and writing is bounded by the rate at which transcribed speech is absorbed into model training. Code, raw Ngram series, analysis and simulation scripts, and the LaTeX source are included. Keywords: retrospective drift; survivorship bias; irregular verbs; regularization; language change; birth–death process; cohort analysis; Google Books Ngram; German strong verbs; cultural tightness; generative language models; emergence Language: English License: Creative Commons Attribution 4.0 International (CC BY 4.0) Related identifiers (relation “References”): https://doi.org/10.13140/RG.2.2.26532.64648 (The Retrospective Drift Theorem) https://doi.org/10.5281/zenodo.24278596 (Retrospective Drift in the Mineral Record: code and data pipeline) https://doi.org/10.5281/zenodo.24278667 (Crystal Polymorphs and Observability as a Persistence Filter) https://doi.org/10.1038/nature06137 (Lieberman et al. 2007) Subjects / communities (optional): Linguistics; Historical linguistics; Computational social science

Authors

Publication Details

Journal
Zenodo (CERN European Organization for Nuclear Research)
Published
2026-11-18
DOI
https://doi.org/10.5281/zenodo.24350138
Primary Topic
Syntax, Semantics, Linguistic Variation
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Cohort Decay, Inventory Stability and the Retrospective Drift of English Verb Regularization

Boris Kriger
Zenodo (CERN European Organization for Nuclear Research)
Syntax, Semantics, Linguistic Variation
article

Cohort Decay, Inventory Stability and the Retrospective Drift of English Verb Regularization

Boris Kriger
article en

Abstract

Creator: Kriger, Boris. Affiliations: Information Physics Institute, Gosport, Hampshire, United Kingdom; Institute of Integrative and Interdisciplinary Research, Toronto, Canada. ORCID: 0009-0001-0034-2903 Publication date: 2026-11-18 Description: The best-known quantitative result on English morphological change is the finding of Lieberman et al. (Nature, 2007). Of 177 Old English irregular verbs, 98 remain irregular today, and the half-life of irregularity scales as the square root of frequency. This result is usually read as evidence that English moves in one direction, toward regularity. This paper argues that the reading confuses two quantities: the decay of an anchored cohort and the change of an inventory. A cohort of forms fixed at a starting date can only shrink, whatever the language does, because forms that become irregular later are excluded by construction. This is a case of retrospective drift: a record conditioned on survival from a fixed origin shows apparent direction even when the system is stationary. The distinction is made formal, and the square-root frequency law is shown to be reproduced exactly by a birth–death process with no net direction. Both quantities are then measured on the forward record. The data are pronoun-anchored preterite bigrams (he/she + past form) from the Google Books 2019 corpora for 150 English pairs. American English, 1800–2019: the cohort lost 4.2% while the irregular inventory went from 120 to 122. Six pairs became majority-irregular (snuck, dove, lit, quit, fit) and six became regular. The verb predicted to regularize next: wed fell to an irregular share of 0.10 in the 1910s and then rose to 0.87. Millennial scale: post-Old English irregulars still in use put a lower bound of 63% on inventory retention, against 55% for the cohort. Rhyme: the form with the larger rhyme family gains share in 19 of 22 American and 22 of 25 British pairs (sign test p < 0.001). German control (de-2019, 39 pairs): the language has a stronger standardizing filter. It shows no births and no deaths, a threefold slowdown of change since 1820, and the rise and suppression of the analogical frug (peak 0.11 in the 1880s). The paper sets out a mechanism of births: audibility of the suffix, dialect and learner origin, amplification by song and rap, and joining a productive minority pattern. It relates the cross-linguistic fate of analogical preterites to cultural tightness. It also predicts that generative language models will act as a filter that freezes the written norm, while speech continues to change. The resulting gap between speech and writing is bounded by the rate at which transcribed speech is absorbed into model training. Code, raw Ngram series, analysis and simulation scripts, and the LaTeX source are included. Keywords: retrospective drift; survivorship bias; irregular verbs; regularization; language change; birth–death process; cohort analysis; Google Books Ngram; German strong verbs; cultural tightness; generative language models; emergence Language: English License: Creative Commons Attribution 4.0 International (CC BY 4.0) Related identifiers (relation “References”): https://doi.org/10.13140/RG.2.2.26532.64648 (The Retrospective Drift Theorem) https://doi.org/10.5281/zenodo.24278596 (Retrospective Drift in the Mineral Record: code and data pipeline) https://doi.org/10.5281/zenodo.24278667 (Crystal Polymorphs and Observability as a Persistence Filter) https://doi.org/10.1038/nature06137 (Lieberman et al. 2007) Subjects / communities (optional): Linguistics; Historical linguistics; Computational social science

Zenodo (CERN European Organization for Nuclear Research)
Syntax, Semantics, Linguistic Variation
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon