A match made in heaven: Using parallel corpora and multinomial logistic regression to analyze the expression of possession in Old Spanish

Abstract

AbstractThis study applies multinomial regression analysis to a parallel corpus of Spanish medieval translations of the Bible in order to study the different factors that condition variation in the expression of possession in Old Spanish. Our methodology allows us to determine the degree to which less frequent possessive constructions (ART + POSS, as inla su casa‘the his house’; GEN, as inla casa de él‘the house of him’; and ART/BARE, as inla casa‘the house’) can be considered competitors to the dominant POSS construction (as insu casa‘his house’) as a function of usage context differences. In comparison to the POSS construction, the ART + POSS construction usually expresses pragmatic functions such as reverence, the GEN construction is typically used to disambiguate a reference, and the ART/BARE construction is bound to contexts in which the possessor is highly accessible. Crucially, the analysis also sheds light on historical changes in the balance between structural and contextual constraints on the use of these different variants. Whereas in the 13th century, structural and stylistic constraints are almost equally important, the importance of structural constraints diminishes in the 15th century. The study thus illustrates how in reductive processes of language change, variation due to structural constraints yields to stylistic variation.