Language is a complex dynamical system that is shaped not just through biological evolution but by the way it is used in a social context. Sociolinguists have long understood that the structure of a society strongly affects the nature of the languages that emerge. Computational models of language evolution, however, generally neglect the effect of social structure by modelling extremely simple population dynamics. This study explores the coevolution of language and social structure using a simple, abstract model of language learning and a plausible mechanism for network growth, namely homophily. Evolved networks are found to possess the characteristic measures of social networks: assortative mixing, transitivity and prominent community structure. The effect of embedding language-learners in the network is found to be significant. This model may also provide a platform on which existing theories and computational models of language evolution can be evaluated.