How automatic is convergence? Evidence from working memory

Abstract

Phonetic convergence has been argued to be automatic as well as socially motivated. Previous studies have shown effects of socially-driven factors on working memory (Beilock et al. 2007), suggesting that working memory might mediate the socially-driven and automatic components of convergence. We used Amazon's Mechanical Turk to test convergence toward a voice with lengthened VOT, with short-term and working memory (modified digit span) conditions as well as a control condition. Overall, participants converged in VOT while shadowing, but converged less when working memory was occupied. These results suggest that social factors affect phonetic convergence indirectly, through their effect on working memory load.