Erik Elgh: Language Phylogeny from Beginning to End: A holistic approach to model and data with case studies on Torricelli and Germanic languages

Abstract

Borrowing, especially between closely related dialects, has long been recognized as an impediment to linguistic subgrouping. The ‘waves’ of features spreading after initial divergence obscure the patterns of that divergence. This complicates the process of inferring language genealogy. To tackle this issue, in this thesis I present a new model for linguistic genealogy and divergence: the Dialect Chain Tree, or DCT. The DCT divides linguistic genealogy into two components. The first is a backbone tree that describes the primary split pattern of uniform dialects. The second is a set of dialect links that describes the structure of mutual intelligibility and contact between closely related dialects in the early stages of divergence. I also develop the model further, into the Dialect Chain Tree with Segments, or DCTS. The DCTS is more explicit concerning the structure and position of the dialect links in relation to the backbone tree, facilitating analysis by Maximum Parsimony when an appropriate algorithm is implemented. Furthermore, I show the process of preparing data for phylogenetic analysis with the DCTS, from data collection by fieldwork and the establishment of regular sound correspondences, to the coding of a character matrix, the input for computational analysis. This data is then used in a case study on a subgroup of Torricelli languages, using regular trees and Maximum Parsimony. I also perform a case study on Germanic languages, using an already existing dataset. Finally, I use a tree for the entire Torricelli family to reconstruct a word for dog, *NVmPVt(V), for the ancestor of a large portion of the Torricelli languages. The case studies and the paper on reconstruction show the need for a more complex framework like the DCT(S) to accurately represent linguistic genealogy.