The regularity of sound correspondences in computer-assisted language comparison
Year: 2026
Type: dissertation
Abstract: This dissertation addresses new computer-assisted methodology for historical linguistics. These methods are directly linked to the comparative method, a set of nineteenth-century principles that combine methodological rigor with an implicit theory of how languages evolve. The computer-assisted make it possible to automize and formalize those principles. Understanding language history offers insight into human history more broadly, particularly in regions where other historical records are not available. The central problem motivating this work is that the empirical groundwork for such historical work, the systematic comparison of language data, has traditionally been carried out manually, in labor-intensive processes, while computational and traditional approaches have often been perceived as contradictory, partly because earlier quantitative methods lacked the scrutiny of manual comparative practice. The implementation of computer-assisted methods further faciliates the comparison of the cross-linguistic diversity in data, since generalizable theories of language change and cognition require evidence from diverse languages rather than a narrow selection of well-documented ones. To address these issues, the dissertation pursues two research questions. First, how the regularity of sound correspondences, the core principle of the comparative method, can be implemented computationally, and what this teaches us about regularity in language change generally. Second, how quantitative analyses of cross-linguistic datasets, using methods related to typology and machine learning, can illuminate broader processes of language evolution. These questions are explored through five studies organized into three parts. The first lays theoretical groundwork, introducing the comparative method, computer-assisted language comparison, and standardized data formats, concluding with a study on constructing standardized comparative wordlists. The second part contains two case studies examining the regularity of correspondence patterns from qualitative and quantitative perspectives, addressing the first research question by showing how this principle can be operationalized computationally. The third part turns to the second research question, presenting analyses extending beyond the comparative method into quantitative typology and machine learning, testing and generating hypotheses about evolutionary processes in language, including language affiliation. Together, these studies show that computational tools, grounded in established comparative principles, can scale up traditional workflows to large datasets, strengthening the empirical foundation of historical linguistics rather than working against it. The thesis demonstrates how formalized implementations of regularity and correspondence patterns can complement manual practice while enabling connections to adjacent fields through interoperable data standards, which in turn facilitate the assembly of large, diverse datasets needed to test hypotheses about language change at scale. In conclusion, the dissertation argues that combining computational rigor with the theoretical insights of the comparative method advances both practical workflows and theoretical understanding of language evolution. The outlook emphasizes building further on standardized data infrastructures and computational models, ultimately contributing to a more comprehensive, empirically grounded theory of how languages change, and how this connects to the broader history of human societies.
Author: Blum, Frederic
Language: English
Subfield: Language and Linguistics
Field: Arts and Humanities
Domain: Social Sciences
Open Access status: green
OpenAlex ID: W7172581699