This dissertation investigates preposition stranding observed in Colloquial Welsh using a generative grammar approach. Welsh traditionally disallows P-stranding. In Literary Welsh, prepositional relatives require a resumptive pronoun which is licensed by the rich agreement of P, and wh-questions require pied-piping of the entire PP. However, the use of uninflected stranded preposition can be found colloquially, both in relatives and wh-questions. This dissertation proposes an account of the different syntactic behaviour between the two varieties, based on the notion of PF feature checking proposed by Ackema and Neeleman (2004). I claim that if PF checking takes place between a P head and its DP complement, the DP will be unable to move out of the complement position. This means that P-stranding is impossible. On the other hand, if PF checking does not hold between P and its complement, P-stranding is possible. Chapter 1 provides theoretical assumptions and background of previous research on Pstranding. Chapter 2 describes wh-constructions in Welsh and investigates the (un)availability of Pstranding in the literary and colloquial varieties. Chapter 3 considers formal properties of Welsh resumptives. Following Willis (2011), I assume that Welsh wh-dependencies with resumptive pronouns obey successive cyclicity. Chapter 4 presents the PF feature checking approach to P-stranding. I assume that the crucial difference between the two varieties is that P in Literary Welsh possesses AGRfeatures, but in Colloquial Welsh does not. I claim that the availability of PF checking regulates the availability of P-stranding. Chapter 5 investigates P-stranding generalizations, observed in Abels (2003), on pseudopassives, clitics, verbal particles and sluicing. Chapter 6 discusses diachronic implications of the occurrence of P-stranding. I consider how the P-stranding option came into Welsh grammar, using the notion of bilingual mode developed by Grosjean (2001).