Reproducibility bundle for the study "Owning Our Own Language: Managerial Convergence in Arts and Humanities Research-Impact Writing." The study analyses 1,546 REF 2021 impact case studies from six Units of Assessment (Art & Design, Performing Arts, Media Studies, Computer Science, Physics, Business & Management) and measures how each is written. A word-frequency classifier identifies a case study's subject 82% of the time; but once subject-specific vocabulary is removed, art and design impact narratives argue in the same register as Business and Management. Two further reference corpora (contemporary art writing and museum annual reports) are scored against validated psycholinguistic norms (Lancaster Sensorimotor Norms; concreteness; NRC) to trace a register gradient from concrete, sensory arts writing toward the abstract register of assessment.This item contains: the filtered corpus, all analysis code (feature extraction, classification, topic modelling, within-corpus and rewrite tests, and the reference-corpus register gradient), intermediate result tables, and all figures (PDF and PNG). Every step is deterministic (fixed random seed) and reproduces exactly. See README.md for full instructions, data provenance and licensing.Data provenance: the REF 2021 impact case studies are Crown Copyright, released under the Open Government Licence v3.0 and re-published by Research England under CC BY 4.0. Reference corpora are rebuilt by the released scripts rather than redistributed (museum accounts are OGL documents from GOV.UK; art-writing texts remain under publishers' copyright, with only derived per-window features shared; psycholinguistic norms are downloaded at run time from their sources).