Abstract
This chapter uses data from the Old Bailey Corpus to study the demise of the English be-perfect between the 1720s and 1910s. The corpus provides ample evidence on the development in the period that saw the transition from be to have as perfect auxiliary in constructions with mutative verbs (i.e., intransitive verbs referring to change of state or place); it also makes it possible to gauge the relative importance that social factors (such as speaker sex, speaker role or socio-economic background) may have played in the loss of one of the perfect variants. The stages for the development are derived from the data in a bottom-up approach using Variability-based Neighbour Clustering (VNC). In a second step, random forests and conditional inference trees are fit to the data for each stage. The former provide information on the overall relative importance of predictor variables and the latter on interaction of predictor variables. It turns out that while, overall, social predictors play a far less important role in the process of loss than the predictor lexical verb, they do show significant interaction with this language-internal variable. The speech-based socio-historical data thus add important details to previous studies on the loss of the be-perfect, not least by lending support to the fact that this change happened largely below the level of speakers’ awareness.
Hundt, Marianne