How predictable is language change? A quantitative approach

While there is, at present, a fairly wide-spread pessimism about the predictability of language change, the historical trajectory of linguistic change is not haphazard. Like other social phenomena, language change is known to often follow an S-curve, with a slow onset, a rapid spread of an innovation in the middle, and a slow end phase, which can be mathematically modeled with the logistic equation.Historical linguists have shown that the logistic curve is a good approximation of many of the changes they study, but they are mostly not really interested in whether there is additional information in the residuals. The gold standard is to have an optimal fit between the curve and the data. This project takes an interest in where the changes deviate from the model. It will be investigated to what extent 26 changes in 19th and 20th century Dutch follow the parametrized logistic s-curve, in order to investigate under what circumstances the real course of change differs from the predicted one. More specifically, the regression parameter settings estimated on observed values from the period 1850-1949 in a register-stable acrolectal text corpus will be used to draw the future trajectory of the S-curve from 1950 to 1999. Then observed values from 1950 to 1999 will be compared to the predicted values, and for each of the 26 changes, mean standardized residuals will be calculated. The residual values will then be clustered, and it will be investigated whether the clusters distinguish different types of change.

Keywords:Historical linguistics, Dutch, Morphosyntax, Quantitative linguistics

Disciplines:Theory and methodology of literary studies