The syntax of Welsh “direct object mutation” revisited

Like all the Celtic languages, Welsh displays a set of initial consonantal mutations. Most instances of mutation have a lexical or morphosyntactic trigger, but a number of the contexts for Soft Mutation are generally considered to be triggered under specific syntactic conditions. This paper investigates the precise syntactic environment in which the process applies. The idea that a major function of this mutation is to mark accusative case on the direct object of a finite verb has a long history, encompassing both traditional grammars and also work within the generative linguistic tradition. Recently, this proposal for ‘direct object mutation’ has been revived and extended, within a Minimalist framework, by Ian Roberts. Roberts argues that the mutation is linked with the raising of a finite verb, and marks (abstract) accusative Case. Here, a range of data is considered which has not previously been discussed in connection with syntactic Soft Mutation. On the basis of this, I argue that the trigger for mutation proposed by Roberts is unworkable, and that the Case hypothesis is both too weak and also too strong. An existing analysis of syntactic Soft Mutation, the XP Trigger Hypothesis, not only captures the generalizations discussed by Roberts, but also correctly predicts the occurrence of all other instances of the syntactic mutation, unifying these into a single environment.