According to recent research, the get-passive, e.g. get arrested, rather than the be-passive, was arrested, has been increasing in frequency since 1850 (Hundt 2001). While some researchers argue that the two variants remain differentiated by semantic nuances like adversativity and agent animacy (e.g. Quirk et al. 1985: 167–71), others assume they are interchangeable and vary according to social factors (Weiner & Labov 1983). Recent corpus-based studies (Allen 2022; Fehringer 2022) tested linguistic and social factors, finding that both play a role. In this article, we aim to contribute new insights by analyzing get vs be in a large corpus of vernacular English from the late nineteenth to the twentieth century in Ontario, Canada. Using a combination of mixed-effects logistic regression and decision tree analysis, we find significant effects of animacy, explicit agent, adversativity, speaker gender, level of education and year of birth. The results show that the get-passive is increasing in apparent time. Moreover, we discover a prevailing effect of animacy that reveals the nature of the reorganization taking place in the system across the twentieth century. We conclude that get emerged as a change from below but is gradually losing its stigma and continues to advance into the grammar of English.