This section provides a preview of the coming chapters.
Chapter 2 describes Welsh wh-constructions, dealing with wh-questions and relative clauses. We will first look at the distribution of a trace and a resumptive pronoun in various positions in Welsh wh-constructions, mainly based on Borsley et al. (2007). We will then focus on prepositional wh-constructions. The default pattern in Literary Welsh is that relative clauses require a resumptive pronoun which is licensed by the rich agreement of preposition
38 (see1.3.2 above) and wh-questions require pied-piping of an entire PP. On the other hand, P- stranding without agreement on the preposition is widely observed in Colloquial Welsh in both relative clauses and wh-questions (Willis 2000). The following generalization can be drawn from the above observation. A preposition in Literary Welsh is followed by its pronominal complement, i.e. a resumptive pronoun in relatives and a wh-expression in interrogatives, whereas a preposition in Colloquial Welsh is followed by a trace left by movement. Native speakers’ acceptability judgement tests also show that the P-stranding option is available colloquially.
Chapter 3 presents the typology of resumption developed by McCloskey (2006) and Asudeh (2010) and considers where Welsh resumptives may fit in this system. Three types of resumptive pronouns will be first introduced: processor resumptives, syntactically active resumptives and syntactically inactive resumptives. McCloskey and Asudeh suggest that the syntactic behaviour of resumption varies from language to language. Syntactically active resumptives do not display general properties of movement such as island-sensitivity and weak crossover violation, whereas syntactically inactive resumptives display them. The data of island-sensitivity and weak crossover effects are not very clear. However, Willis (2011) shows empirical evidence that Welsh wh-dependencies in both movement and resumptive structures obey successive cyclicity. This suggests that there is movement with resumptive pronouns.
Chapter 4 presents an account of the different syntactic behaviour on the availability of P-stranding between Literary Welsh and Colloquial Welsh. As we saw in 1.2, my account is based on the notion of PF feature checking in Ackema and Neeleman (2004). I assume that the crucial difference between the two varieties is that a P head in Literary Welsh possesses AGR-features on person, number and gender, but in Colloquial Welsh does not. In relatives, we can see this contrast from the presence/absence of inflectional morphology on P. In Literary Welsh, given a P head possesses AGR-features, PF checking takes place between P and its complement to check AGR-features in the same prosodic domain. At the same time, an EPP-feature on v and C head requires their specifier positions to be filled for syntactic checking (see 1.3 above). However, since PF checking already holds between P and its DP complement in Literary Welsh, the DP in the complement position no longer moves to the Spec-CP position to check the EPP feature with the C head (see 1.2.3). Adopting Willis (2011), I will assume that a null wh-operator is inserted into the specifier of P from the lexicon, and then the EPP requirement is satisfied by movement of that operator to Spec-CP
39 via Spec-vP. In Colloquial Welsh, PF feature checking does not hold between P and its complement due to the lack of AGR-features on P. As a consequence, a wh-operator can move out from the complement position of P, then it moves following successive cyclicity. For wh-questions in Literary Welsh, I continue to assume that the P head bears AGR- features despite its appearance in a bare form. This is because Welsh wh-expressions are in fact non-pronominal in terms of morphological agreement (Borsley 2009). Therefore, the PF feature checking takes place between P and its complement. This disallows movement of a
wh-word to Spec-CP on its own. Rather, the elements in the PF checking relation move
together to satisfy the EPP requirement on C, which is simply a phrasal movement of PP. I will also consider the consequences of this analysis and advantages of the feature checking approach.
Chapter 5 investigates the four P-stranding generalizations expressed in Abels (2003a). First, all languages that allow P-stranding under passive (i.e. pseudo-passive) also allow P- stranding under wh-constructions. Second, languages that disallow P-stranding do not allow clitic pronouns as the complement of P. Third, all languages that allow P-stranding also have verbal particles (i.e. phrasal verbs) (Stowell 1982). Fourth, a language allows P-stranding under sluicing only if it allows P-stranding under wh-question (Merchant 2001). I will first check whether these generalizations hold in Welsh. Then, I will consider these phenomena making use of the PF feature checking analysis developed in chapter 4.
Chapter 6 discusses the occurrence of P-stranding in Colloquial Welsh from a diachronic point of view. Borsley et al. (2007: 116) suggest that the appearance of P-stranding is “a twentieth-century innovation from language contact, modelled on preposition stranding as found in English”. This seems to be the case if we consider the very extensive contact with English; virtually all Welsh speakers are bilingual in Welsh and English (see 1.1 above). I will suggest that the occurrence of P-stranding in colloquial Welsh is due to the activation of English syntactic knowledge in terms of processing while Welsh-English bilinguals are speaking Welsh, using the notion of ‘bilingual mode’ developed by Grosjean (2001). Following an acquisition-based model of language change (Andersen 1973, among others), I assume that language change takes place if a younger group acquires a grammar which is slightly different from the one of the older group. In my case study, this means that language change takes place if children acquire the P-stranding option in their mental grammar. I will argue that synchronically Welsh speakers have two grammars in their mind, i.e. Literary Welsh and Colloquial Welsh. Adopting the competing-grammars analysis in Kroch (1989), I
40 will suggest that the P-stranding and non-P-stranding options associated with the two registers are in competition diachronically. The competing-grammars analysis predicts that the non-P-stranding option will be replaced by the P-stranding option over the period of time. In chapter 7, I review my claims in this dissertation, and then consider remaining problems for future research.
41 CHAPTER 2