Abstract
It has been argued that syntactically light and/or highly topical possessors tend to precede their respective possessa in discourse (O'Connor et al. 2013). Taking this as a functional cognitive pressure, we hypothesized that proper noun possessors will be more frequent in anteposition than generic possessors.
This was also supported by a comparatively higher proportion of antepositions in possessive pronouns, when we compared possessive adjectives, possessive pronouns, relational and qualitative adjectives sampled from diachronic corpora of Czech (Diakorp, Vokabulář webový) across the period of 1300-1850. Contrary to our expectations, we did not find a gradual decrease of postnominal possessive adjectives over time.
There was a marked increase in postpositions in the middle (1400- 1725) of three periods identified by variability-based neighbor clustering (Gries & Hilpert 2008). To test the generalization, possessive adjectives were coded for possessor referentiality, possessum alienability, and an array of syntactic variables.
The data was analyzed with generalized linear mixed effects modeling, revealing possessor status, possessum alienability, position relative to the verb, and syntactic function of the phrase as significant predictors of placement. While the model showed differences between proper and generic possessors, proper nouns are more strongly associated with postposition, contrary to the hypothesis.
Upon closer inspection, the increase of the postnominal pattern seems to be driven by two partially filled, highly frequent constructions with postnominal possessors. Thus, rather than a cognitive bias, we see the influence of both the historical and socio-cultural circumstances and the language internal mechanisms driven by frequency and similarity-based analogy, as argued for by usage-based and constructional approaches (e.g.
Bybee 2010).