Micro to Macro: Modelling Language Variation Across Time

Language evolution and biological evolution have been courting each other since Darwin. Darwin was interested in language evolution because change is rapid enough to be visible, unlike biological evolution. Linguistics has borrowed methods from biology to understand how languages change on a macro (phylogenetics) and, more recently, micro (population genetics) levels.

The mapping from the micro to macro is a fundamental puzzle in biology for understanding how speciation works. For linguistics, it is about how population-level language use leads to new languages, such as linguistic diversification.

Variation is a fundamental ingredient in all evolutionary systems, including language. All fields of linguistics that deal with language evolution from population-level variation and change (sociolinguistics) to linguistic diversification (historical linguistics) have the (non)-use of language features at their core.

Modelling the relationship between micro-level processes (such as variation among individuals in a population) and macro-level outcomes (language diversity) has proven difficult. Still, it is key to understanding how languages emerge. In order to link the micro and macro in language evolution, a number of step changes have been required.

This talk steps through a number of recent developments in language evolution which bridge these identified gaps. Professor Felicity Meakins introduces BayesVarbrul, developed by Hua (2022), which models changes in the frequency of linguistic variants across generations and regions, and the effects of social factors on the uptake and loss of these variants across space and time. It is designed for datasets with multiple variables and multiple speakers from different generations and speech communities.

An early version of BayesVarbrul was first applied to a situation of language change in northern Australia where an intergenerational shift from Gurindji to Gurindji Kriol is underway. But the Gurindji Kriol dataset did not allow exploration of both time and space because the data was collected only in one community and therefore lacks a regional dimension. This paper introduces a new dataset of Shawi speakers from Peru which has the necessary regional dimension, as well as a generational dimension.

This dataset is used to model the emergence of a new dialect of Shawi across 168 speakers of different ages and regions, paying particular attention to variation in word order at the individual-speaker and regional levels.

Felicity Meakins then returns to the Gurindji Kriol dataset, describing an experimental approach that has enabled her to develop a metric for the (non)-association between social salience and its effect on the flow of linguistic variants over time.

Date

Friday 7 March 2025

Time

1 to 2 pm

Location

Online via Zoom

Speaker

Professor Felicity Meakins
University of Queensland