Revisiting the population vs phoneme-inventory correlation

Abstract

Relationships between phonological systems and extralinguistic factors is controversial. Using a sample of 961 phoneme inventories from the PHOIBLE database, genetic and speaker population data from Ethnologue and WALS, and a hierarchical linear model that accounts for genetic relatedness at family and genus levels, we show that speaker population accounts for little to no variation in most aspects of the phonological system (number of phonemes, vowels, obstruents, etc). We argue that spurious correlations in previous studies resulted from failure to control for language relatedness, samples skewed by small size or over-representation of certain families, and/or case-based reasoning lacking statistical rigor.