How gradual change progresses

Posted on November 27, 2013 by

UPDATE: Rescheduled due to strike. New details below.

On Wednesday 4th December, Hendrik de Smet (KU Leuven) will be giving a talk at the LEL Research Seminar on How gradual change progresses: the expansion of –ing-clauses with begin through time and across individuals. The abstract is below.

The talk is at 9.00am in Ellen Wilkinson A2.16. The evening before (Tue 3rd), we’ll be meeting for drinks and dinner; contact George if you’d like to join.

Hendrik, a rising star in the historical-linguistics world, is in town for less than 24 hours, so come along to listen to his talk and say hello!


This paper examines the relation between diachronic gradualness and the divergent behavior of individual language users in synchronic populations. The goal is to obtain a clearer picture of how the steps in a process of gradual grammatical change relate to one another. The specific focus is on the spread of –ing-clauses, at the expense of to-infinitives, following the aspectual verb begin. Using historical corpus data from the Corpus of Historical American English, the development is shown to be diachronically gradual in at least two respects. It affected contexts with agentive subjects faster and earlier than contexts with non-agentive subjects, and it affected horror æqui contexts (i.e. following the form to begin) faster than other contexts. Next, it is examined how these two diachronic gradualness effects play out in the behavior of individual language users in two synchronic populations. Using the New York Times Annotated Corpus the usage is analyzed of 83 New York Times journalists, writing around the end of the twentieth century. Using the Corpus of English Novels, a similar analysis is conducted for 25 novelists writing around the end of the nineteenth century. It is shown that individual variation faithfully matches the diachronic progression of change, in that an individual’s degree of advancement in one context of change is correlated to their advancement in other contexts. Moreover, the correlations are found to be non-linear, which implies they are compatible with the operation of thresholds in gradual changes, such that one step in a gradual development cannot be initiated unless a preceding step has reached a particular level. As such, patterns of individual behavior support a cascade model of gradual grammatical change.