Introduction

The technical term S-curve goes back at least to 1839 (OED, as two words), and in the context of linguistic change certainly to 1954 (Altmann et al. 1983: 105). The importance of S-curves is now generally taken for granted in historical linguistics. Devitt, for example, asserts that ‘[t]he S-curve pattern of diffusion occurred throughout [her] data’ (1989: 75). Here is a more recent example concerning early Modern English morphosyntax (Nevalainen 2000: 339):

In Nevalainen's paper, many graphs need just five or six data-points to reveal a jagged but nevertheless recognisably S-shaped form. I myself have long assumed ‘that the time course of the propagation of a language change typically follows an S-curve’, as Croft expresses it (2000: 183). In this chapter I wish to question why this might be so, what it means and why perhaps, in the end, it is actually rather unexpected. In touching on explanations of language change and notably the shape that changes take, there is a risk (or one might well say, the hope) of echoing some of Roger Lass's published work: there can be few topics in historical linguistics that he has not profitably addressed at some time or other.