Looking back on two centuries of back-formation research

The notion of back-formation was brought up in German linguistics in the first half of the nineteenth century and has never since ceased to be discussed controversially. One early account equated back-formation with the resolution of a proportional formula, with the only difference that the unknown ( x) was a′ ( a : b = x : b′) and not b′ ( a : b = a′ : x), as in ordinary derivation. This account has been criticized because in some cases at least no exact analogue can be found for such proportional formulae, but it could well be that approximate analogues are sufficient for speakers. From early on, reanalysis has also been said to be a necessary ingredient of back-formation, but it turns out that, although this is often the case, it is not a necessary condition. In the present paper, the position is endorsed which sees back-formation as a two-step process consisting of two elementary cognitive processes, namely similarity-based categorization and abduction.