Quantitative Language Evolution: Case Studies in Finnish Dialects and Uralic Languages

This dissertation focuses on quantitative analysis techniques that relate to language evolution. Over the last few decades an increasing number of linguists have acknowledged the benefits that quantitative analysis approaches can bring to the study of language data, such as increased objectivity, ease of replication as well as increased ability to handle large volumes of data. In addition to a more widespread acceptance of quantitative approaches in general, this has also brought analysis techniques from other research fields to linguistics. This includes methods from evolutionary biology, such as phylogenetic techniques, used to study historical relationships between different species, and population genetic techniques, which explore the relationships between the populations of a species. In the field of linguistics phylogenetic techniques can be used to quantitatively study related languages, while population genetic techniques are useful for studying more closely-related languages or varieties of the same language.

The present work focuses on phylogenetic and population genetic analysis techniques through four research articles that explore the applicability of these approaches in linguistic research, and also proposes new quantitative techniques to the toolkits of historical linguistics and dialectology. It also brings evolutionary research to both Uralic languages and Finnish dialects. The research is part of the work done by the multidisciplinary BEDLAN research initiative and its follow-up projects.

The language data used in the research comes from a dataset of Uralic basic vocabulary, compiled as part of the BEDLAN research initiative and its follow-up projects, and made publicly available as part of this dissertation’s work. This language family has not previously been explored using phylogenetic techniques, but is well-researched from a traditional standpoint, making it a good test subject for these techniques. In addition to the basic vocabulary data, the dissertation also uses dialect data from the digitized version of Lauri Kettunen’s Dialect Atlas of Finnish, error-checked and revised by the BEDLAN project and released by Kotus. Similarly to the Uralic language family, Finnish dialects also represent a good testing ground for new population genetic analyses due to the large amount of traditional research done on them. In addition to real-life datasets, the dissertation’s research also employs simulated language data as part of its methodological exploration.

The four research articles show that evolutionary techniques work well for modeling relationships both between languages and within dialects, based on how the overall results compare to the literature. The articles also bring new techniques from evolutionary biology to linguistics, including population genetic clustering as a way of inferring dialect areas and a phylogenetic metric called TIGER values as a way of estimating how treelike a linguistic dataset is. The research that makes up this work has also laid groundwork for multidisciplinary research on languages as part of the overarching study of human history.