BCS pronominal clitic variants ju and j...

Studies of Bosnian, Croatian and Serbian (BCS) clitics (CLs) are mainly characterized by either formal-theoretical  or  normativist  descriptions  of  their morphology  and  syntax.  At  the  same time,  studies  on  variation  in  language use  based  on  big  data  are  extremely  rare.  However, careful  and  contrastive analysis  of  BCS  grammar  books  and  language  usage  guides  reveals various differences in the usage of CLs not only between standard BCS varieties, but also within their  colloquial  idioms.  This  paper  deals  specifically  with  variation in  the  usage  of  the  third person singular feminine accusative CLs ju and je ‘her’.  

Although the usage of the pronominal CL ju is still widespread not only in Croatian but also in Bosnian and Serbian dialects, standard Bosnian and Serbian limit it to contexts of suppletion triggered by its combination with the verbal CL je ‘is’ and its negated form nije. In contrast, many Croatian grammarians and authors of  language usage guides allow a much wider usage of the CL ju and some put it on the same footing as its variant je or even prefer it over the pronominal CL je. 

Considering  the  state  of  the  art  presented  in  the  literature,  we address  the  following research questions:  

RQ1: Does the distribution of pronominal CL variants ju and je in corpora representing BCS standard varieties confirm the expected differences between these language varieties? 

RQ2:  Does  the  distribution  of  pronominal  CL  variants  ju  and  je  in  corpora representing colloquial BCS varieties show any similarities to their distribution in standard varieties? 

RQ3: Does the distribution of pronominal CL variants ju and je within one BCS variety show different patterns in the standard and colloquial registers? 

To test the range of this variation we turned not only to traditionally compiled reference corpora with language material which should reflect standard BCS varieties, but also to {bs,hr,sr}WaC, three massive web corpora, and extracted random samples of size 400 per pronominal CL (ju vs je) for each variety. In the next step we manually annotated words preceding and following the pronominal CLs ju and je to determine the impact of context on 
CL choice.

The data not only confirm the expected differences between the three standards, but also reveal differences between standard and colloquial idioms of the varieties. We expect more interesting insights from logistic regression analysis after the annotation process is completed.