Category clustering: A probabilistic bias in the morphology of verbal agreement marking: Supplementary Material

Abstract

A minor limitation of our Chintang prefix bigram analysis is the small number of tokens for the ma-, mai-1NS.P prefixes in combination with other prefixes. We have only 14 observations of ma-, mai-combining with SUBJ prefixes, and just five observations combined with mai-NEG. 1 However this poverty of data in itself points to an interesting pattern: the number of ma-, mai-1NS.P + mai-NEG tokens turns out to be unexpectedly small, when we compare this with overall corpus counts. The ma-, mai-prefixes, which encode exclusive and inclusive versions of 1NS.P respectively, are associated with speakers from the village of Mulgaũ, while kha-1NS.P, making no clusivity distinction, is associated with speakers from the village of Sambugaũ. However this is not a rigidly maintained dialectal distinction, and the corpus includes a substantial number of speakers using both forms (N=72). Both variants are relatively recent innovations, drawing on distinct etymological sources (kha-< *khəәl 'all'; ma-, mai-< *rak-mi 'person') to provide a strategy for speaker effacement under the influence of Maithili politeness strategies (Bickel & Gaenszle 2015).