a1 Dartmouth College
In this study, we present the first agent-based simulation of vowel chain shifts across large communities, providing a parsimonious reinterpretation of Labov's (2007) notions of transmission, diffusion, and incrementation. Labov determined that parent-to-child transmission faithfully reproduces structural patterns such as the Northern Cities Shift (NCS), but adult-to-adult diffusion does not. NCS is transmitted faithfully to new generations of U.S. Inland North children. But St. Louis speakers, depending only on adult-adult contact, only attain an incomplete, unsystematic version. Labov (2007) attributed the difference to children's superior language-learning ability; transmission and diffusion are categorically different processes in that approach. By contrast, our multiagent simulation suggests that such transmission/diffusion effects can be derived by simple density of interactions and simple exemplar learning; we also find that incrementation is a natural outcome of this model. Unlike Labov (2007), this model does not require a dichotomy between transmission and diffusion. While dichotomous assumptions about child versus adult learning may be necessary in other contexts, our results suggest that the NCS effects in Labov (2007) may be explained economically in terms of simple density of interactions between speakers. Our results also provide an agent-based perspective supporting and explicating the notion of speech community.
We would like to thank Dennis Preston, Zsuzsanna Fagyal, the audience at New Ways of Analyzing Variation 40, William Labov, Gillian Sankoff, Naomi Nagy, Erik Thomas, Miriam Meyerhoff, Lindsay Whaley, Ioana Chitoran, David Peterson, Tim Pulju, Francois Pellegrino, John L. Stanford, Kenny Baclawski, Dan Rockmore, Victoria Smith, Sravana Reddy, and the reviewers of this journal. Brad Svenson assisted in programming. The project was supported by a grant from the Neukom Institute for Computational Science, Dartmouth College, and the William and Constance Burke Research Award.