Phylogeny in Phonology: How Tai Sound Systems Encode Their Past

Abstract

The study of sound change is foundational to traditional historical linguistics, particularly the linguistic comparative method. It is well established that the phonology of modern languages encodes useful data for studying the history of those languages, and their genetic relationships to one another. However, phonology has typically been the means to the end, enabling the comparative method, and coding of a comparative lexicon for cognacy. Once coded, the particular sounds involved no longer factor into the analysis. This study examines whether the phoneme inventories and phonotactic profiles of a set of languages themselves contain phylogenetic signal detectable using established statistical tests D statistic (Fritz & Purvis 2010), K (Blomberg et al 2003), and NeighborNet delta score (Holland et al 2002) and Q-residual (Gray et al 2010). This study adds to the growing body of work on the use of phonological traits in computational phylogenetics for linguistics. Using data from 20 Tai lects from the Kra-Dai language family, this study confirms and extends previous findings. This includes detection of strong phylogenetic signal in phoneme frequency and biphone transition probabilities, but also relatively strong phylogenetic signal detected in even coarse-grained phoneme and biphone presence/absence, which previous work was unable to do.

Phylogenetic signal in phonotactics

“…However, the pilot study results should be treated with caution -particularly the failure to reject the D = 0 null hypothesis in the case of Ngumpin-Yapa -due to the small sample sizes (10 languages for Ngumpin-Yapa, 7 for Yolngu), well below the minimum of 50 taxa recommended by Fritz & Purvis (2010). Dockum (2018) performs the same analysis using biphone characters from 20 Tai language varieties of the Kra-Dai family. In contrast to Macklin-Cordes & Round (2015), Dockum finds some evidence of phylogenetic signal in the Tai data and suggests perhaps the earlier result was due to insufficient variation in that particular language sample rather than a limitation of binary biphone characters per se.…”

Phylogenetic signal in phonotactics

“…However, the pilot study results should be treated with caution-particularly the failure to reject the D = 0 null hypothesis in the case of Ngumpin-Yapa-due to the small sample sizes (10 languages for Ngumpin-Yapa, 7 for Yolngu), well below the minimum of 50 taxa recommended by Fritz & Purvis (2010). Dockum (2018) performs the same analysis using biphone characters from 20 Tai lects of the Kra-Dai family.…”