Exploring the history of pronouns in South America with computer-assisted methods

Pronouns as a diagnostic feature of language relatedness have been widely explored in historical and comparative linguistics. In this article, we focus on South American pronouns, as a potential example of items with their own history passing between the boundaries of language families, what has been dubbed in the literature as ‘historical markers’. Historical markers are not a direct diagnostic of genealogical relatedness among languages, but account for phenomena beyond the grasp of the historical comparative method. Relatedness between pronoun systems can thus serve as suggestions for closer studies of genealogical relationships. How can we use computational methods to help us with this process? We collected pronouns for 121 South American languages, grouped them into classes and aligned the phonemes within each class (assisted by automatic methods). We then used Bayesian phylogenetic tree inference to model the birth and death of individual phonemes within cognate sets, rather than the typical practice of modelling whole cognate sets. The reliability of the splits found in our analysis was low above the level of language family, and validation on alternative data suggested that the analysis cannot be used to infer general genealogical relatedness among languages. However, many results aligned with existing theories, and the analysis as a whole provided a useful starting point for future analyses of historical relationships between the languages of South America. We show that using automated methods with evolutionary principles can support progress in historical linguistics research.