Investigating the vowel systems of contact languages using a population of artificial agents

Abstract

Contact languages are often described as having very simple phonological systems, with a small itinerary of vowels. This study analyses whether a computational architecture implemented to investigate the emergence of vowel systems (de Boer 2001) can be used to model these findings from the contact linguistics literature. Simulations were run to generate emergent vowel spaces (much like those of de Boer 2001) which were then used to generate contact simulations. Two kinds of contact simulations were explored. A pidgin simulation in which agents were taken from two different simulations, and a creole simulation featuring agents from two separate simulations along with a number of blank agents (agents with no vowels in their vowel spaces). Simulations in which vowels systems are generated from scratch provide similar results to those reported by de Boer (2001), but results from the contact simulations do not explain the patterns we encounter of real contact language vowel systems. “Pidgin” contact systems were discovered to contain a high number of vowel clusters and the creole systems fewer. This contradicts the language literature. Reasons for these findings are explained on the basis of shortcomings in the model and the complexity in modelling contact languages.