Convergence is an oft-used notion in contact linguistics and historical linguistics. Yet it is problematic as an explanatory account for the changes it represents. In this study, we model one specific case of convergence (Duoxu, an endangered Tibeto-Burman language with 9 last speakers) to contribute to a more systematic understanding of the mechanisms underlying this phenomenon. The goals are (1) to address the role of some linguistic and social factors assumed to have an effect on the process of convergence, and (2) to test the following explanations of empirical observations related to phonological convergence: (a) the loss of phonological segments in a language that has undergone convergence is correlated with the relative frequency and markedness of these segments in the combined bilingual repertoire, and (b) widespread bilingualism is a prerequisite for convergence. The results of our agent-based simulation affirm the importance of frequency and markedness of phonological segments in the process of convergence. At the same time, they suggest that the explanation related to widespread bilingualism may not be valid. Our study suggests computer simulations as a promising tool for investigation of complex cases of language change in contact settings. Convergence, agent-based modeling, bilingualism, phonology, markedness, social factors