Phys. Rev. X (in the press); preprint at http://arxiv.org/abs/1703.00533

We're all familiar with the idea that two people from the same country, living a short distance apart, might speak very different dialects. Intuitively, you'd expect that history and geography have each played a role in shaping the boundary between their dialects, but is it possible to formalize this hunch?

To address this question, James Burridge resorted to statistical mechanics, deriving macroscopic equations for language evolution. In his theory, speakers adopt the language variant that they hear the most: this local alignment, like spins in a ferromagnet, leads to the emergence of domains bound by domain walls — or 'isoglosses' in the linguist's lingo. Taking into account the varying population density, the boundaries between domains spontaneously bend around relevant geographical features, reflecting the stronger influence of cities and the linguistic barrier posed by mountains or other scarcely populated areas.

The agreement between the model's predictions and real data is a reminder of the efficacy of coarse-grained descriptions based on simple principles — in this case, surface tension minimization — to capture the salient features of complex systems.