Accusative pronominal clitics ju and je in Bosnian, Croatian and Serbian: a contrastive corpus study of standard and colloquial varieties

Empirical studies on variation in clitic use are almost entirely lacking. However, a careful contrastive analysis of both BCS grammar books and language advisor handbooks indicates differences in the usage of clitics not only between BCS standard varieties but also within their colloquial varieties. This paper deals specifically with variation in usage of the third-person feminine accusative CL ju and je ‘her’. According to the previous literature, the usage of the CL variant ju is more limited in Bosnian and Serbian than in Croatian standard (cf. Silić & Pranjković 2007; Mrazović & Vukadinović 2009; Ridjanović 2012; Piper & Klajn 2014).

To test the range of this variation, we turn not only to traditionally compiled reference corpora with language material that should reflect standard BCS varieties (Santos 1998; Utvić 2011; Čermák & Rosen 2012; Brozović Rončević et al. 2018) but also to {bs,hr,sr}WaC, three massive web corpora (Ljubešić & Klubička 2014) available via a unified, functional interface NoSketchEngine, with morphosyntactic annotation based on the common tagset. We manually annotated 3.916 data points with CL variants ju and je. This study shows how language corpora can give us valuable insights into the range of variation not only in the BCS standard varieties but also in the real language use, i.e., their colloquial varieties through the frequency of forms and patterns which are partially discarded by the BCS normativists.

The generalized linear regression model revealed statistically significant differences in the usage of the CL ju between the standard Croatian variety on the one hand, and standard Bosnian and Serbian on the other. It also showed that BCS colloquial varieties differ significantly concerning the distribution of the CL variant ju. Furthermore, our data showed that the distribution of the CL variant ju is significantly different in standard and colloquial varieties of analyzed languages. Finally, based on our generalized linear regression model, we showed that the -(j)u ending of the CL host has a significant impact on the choice between the CL variants ju and je: it decreases the probability of the ju usage.