Date created: | Last Updated:

Category: Project

Description: To describe language change, studies in diachronic corpus linguistics frequently rely on quantitative analysis of changes in the distribution of one or more linguistic elements. However, this method typically underestimates the degree of multicollinearity of contextual variables; change in one distributional dimension is rarely independent of change in other distributional dimensions. This paper proposes to use Latent Class Analysis (LCA) as a solution to the problem. We develop a data-driven description of a linguistic element's meanings in historical corpus data using the systematic co-occurrence of a number of contextual proxies. An analysis of the diachronic distribution of the predicted probability for each token to express these latent meanings allows us to objectively and replicatively map meaning change. As a case study, we examine changes in the preferred usage contexts of the Spanish intransitive ser 'be' + past participle (PtcP) construction between the 14th and the 18th century. Our findings confirm the hypothesis that ser + PtcP underwent functional specialization as its overall usage frequency declined and give preliminary evidence for the relevance of social factors for its conservation.