What corpora tell us about the grammaticalisation of voice in get-constructions

Grammaticalisation studies are a potential meeting ground for theoretical linguistics and corpus-based approaches. The present paper uses evidence from the ARCHER corpus to investigate the relationship between frequency and the grammaticalisation of the get -passive. It seeks to verify whether the grammaticalisation of the passive function is reflected in an increase in the overall frequency of get , and whether grammaticalised patterns of get have spread at the expense of its lexical use. Givón/Yang (1993) have singled out several patterns that were involved in the grammaticalisation of the get -passive. Data from ARCHER are analysed to find out which of these patterns were frequent and which were of marginal importance. Finally, evidence from standard one-million-word corpora is used to test existing hypotheses on the maturation of the get -‘passive’. These data show that it has not lost its middle semantics and come closer to a true passive.