Optimization in the Evolution of Language

What drives languages to evolve and change? A common answer points to the optimization of communication, but this is difficult to square with the historical evidence of language change (Labov 1994). For instance, a most common pattern of change merges two vowels or consonants into a single category, which renders previously distinct words into homophonic ones thereby disrupting linguistic communication. By contrast, merger reversals, which disambiguate words and ought to facilitate communication, are exceedingly rare (Labov 2011).

We suggest that the notion of optimization in language evolution can still be maintained, though the metric of evaluation relies on the specific organization of language processing and learning. We review evidence from speech and language perception, specifically how humans resolve ambiguities caused by homophony. It is possible to provide quantitative “fitness” measures of competing linguistic systems, which in turn determine the changes in their frequencies of usage over time. This approach correctly predicts the requisite demographic conditions under which the so-called “cot-caught” merger spreads from East Massachusetts, where the two words are pronounced the same, to Rhode Island, where the two words were traditionally distinct.