This dataset contains corpus-based frequency data for an analysis of key verbs in published academic writing. The data are from the Corpus of Contemporary American English (COCA; Davies 2008-) and cover a period of 30 years (1990-2019). The section ‘academic’, which contains research articles from peer-reviewed journals, represents the target variety, and the reference variety is fictional writing as represented in the ‘fiction’ section (which contains short stories, plays, movie scripts, and the first chapter of novels). The total number of text files is 26,137 (academic) and 25,992 (fiction). To reduce computational expense for our methodological simulation study, we restrict our attention to verb lemmas whose whole-(sub)corpus normalized frequency exceeds 10 pmw in the academic section of COCA. The data therefore contain frequency information on only 700 verb lemmas.