Before presenting my PF feature checking analysis to Welsh P-stranding, I will first show Willis’ (2011) formal account on Welsh wh-dependencies. He claims that Welsh wh- dependencies in both movement and resumptive structures obey successive cyclicity. As already shown in 3.4, Willis provides empirical evidence that the specifiers of CP and vP serve as escape hatches in Welsh, and argues that a wh-operator can be freely extracted as long as it moves through these escape hatches. I will consider his analysis in some depth since he also deals with P-stranding.
Willis assumes that wh-movement is triggered by an uninterpretable wh-feature that possesses an EPP-feature on the head of the wh-clause. That is, C exists in two versions, a non-wh version (realised in Welsh as y(r), in Irish as go, in English as that etc.) and a wh- version. The wh-version bears an uninterpretable wh-feature with an EPP-feature:
(21) CEPP[uWH:___ ]
This uninterpretable wh-feature searches for an interpretable wh-feature. Such an interpretable feature is a wh-expression in wh-questions, or a null operator in relative clauses. This is illustrated by a wh-expression pwy ‘who’:
(22) pwy D[WH:+]
An uninterpretable wh-feature on a C head triggers movement of a wh-expression to Spec-CP, and the uninterpretable feature is valued (or checked) by the interpretable feature of wh-
119 expression. In (23), the uninterpretable wh-feature on C is valued by the moved subject wh- expression pwy ‘who’ (lower copies are indicated in angle brackets).
(23) Pwy sy <pwy> ’n gwybod yr ateb? [CP [WH:+] CEPP[uWH:+] … ]
who be.PRES.REL PROG know the answer
‘Who knows the answer?’ (Willis 2011: 209)
As we saw above, Willis argues that wh-movement is derived through Spec-vP as well as Spec-CP in a successive cyclic manner. Since movement across v (from object position) and C (across clause boundaries) occurs in Welsh, these heads must be capable of attracting a wh- operator to their specifiers. Therefore, Willis assumes that a v head also bears an uninterpretable wh-feature just like a C head.
(24) vEPP[uWH:___ ]
If the wh-versions of both C and v are selected, wh-movement proceeds via Spec-vP and Spec-CP positions, as illustrated in (25). The object clitic ’i is not associated with resumption for Willis as we saw in 3.4. The wh-expression pwy first moves to Spec-vP and then to Spec- CP in order to value the wh-feature on v and C heads.
(25) Pwy mae ’r heddlu wedi <pwy> ’i ddal <pwy>? [CP [WH:+] CEPP[uWH:+] … [vP vEPP[uWH:+] … ]]
who be.PRES.3S the police PERF 3MS catch
‘Who have the police caught?’ (Willis 2011: 210)
We finally look at prepositional relatives which make use of the resumptive strategy. Willis points out that even resumptive wh-dependencies show some successive-cyclic effects, since morphological alternations that suggest cyclic movement via CP and vP can be observed even when a resumptive pronoun is used. We saw in (47) in the previous chapter that the tense restriction of the embedded clause is voided in wh-environments. In (26) below, the embedded verb is in the present tense mae rather than the non-finite form bod in a prepositional relative with an overt resumptive in a variable position.
120 (26) y llyfr mae pawb yn dweud mae Mair yn sôn amdano fe
the book be.PRES.2S everyone PROG say be.PRES.3S Mair PROG talk about.3MS it
(Willis 2011: 215)
We also saw in (49) that the object clitic may precede a non-finite verb in the main clause in long-distance wh-dependencies. In (27), soft mutation on main verb with or without an object clitic is grammatical in long-distance prepositional relatives formed using a resumptive pronoun.
(27) y llyfr roedd pawb yn (ei) feddwl oedd Mair yn sôn amdano fe the book be.IMPE.2S everyone PROG 3MS think be.PRES.3S Mair PROG talk about.3MS it
‘the book that everyone thought that Mair was talking about’ (Willis 2011: 215)
According to Willis, the above data suggest that the resumptive structure shows the cyclic nature as in the movement structure.
To account for these facts, Willis suggests that a wh-operator is inserted from the lexicon into the specifier of P, then that operator moves though specifers of v and C following successive cyclicity. He assumes that a P head that takes resumptive pronouns bears a feature that blocks movement but allows insertion (Merge) of a wh-operator. According to Willis, such a feature is an uninterpretable wh-feature lacking an EPP-feature. The lexical entry of a P head is expressed as follows.
(28) P [uWH:___ ] (Willis 2011: 215)
Willis claims that this uninterpretable wh-feature on P is valued by a null operator Op inserted from the lexicon. Then, uninterpretable wh-features on v and C is valued by that moved null operator. The configuration of a prepositional relative is illustrated below.
(29) y bobl [CP werthodd Ieuan [vP y ceffyl iddyn nhw
[CP CEPP[uWH:_] … [vP vEPP[uWH:_] … [PP Op[WH:+] P[uWH:_] [DP [them]]]]]
the people sell.PAST.3S Ieuan the horse to.3P them
121 McCloskey (2002, 2006) observes that resumptive pronouns are simply ordinary pronouns. Following this observation, Willis assumes that resumptive pronouns bear no wh-feature. The pronoun nhw ‘they’ is introduced to satisfy the argument structure of the preposition i ‘to’. The uninterpretable wh-feature on P cannot be valued by any element within the PP. Therefore, Willis proposes that the null operator which bears the wh-feature, Op [WH:+], is merged into Spec-PP from the lexicon. Then, the unvalued wh-feature is valued by the null operator through the specifier of v and C. This null operator movement is illustrated in (30) below. This configuration after the derivation contains no uninterpretable features:
(30) y bobl [CP werthod Ieuan [vP y ceffyl iddyn nhw
[CP Op[uWH:+] CEPP[uWH:+]..[vP <Op> vEPP[uWH:+]..[PP <Op>P [uWH:+] [DP]]]]
the people sell.PAST.3S Ieuan the horse to.3P them
Willis (2011: 217) also very briefly considers P-stranding that allows extraction of the complement in Colloquial Welsh. He suggests that speakers who allows P-stranding either possess a wh-version of some functional projection at the left edge of P by creating an escape hatch for movement from within PP, or else P is not a phase head for them. Willis states that the escape hatch analysis seems to be more preferable since it keeps with the idea that the locus of cross-linguistic and dialect variation is the lexicon. Although Willis’ analysis nicely captures behaviour of Welsh wh-dependencies on both empirical and theoretical grounds, his treatment of P-stranding is rather speculative. Willis seems to suggest that there is a functional head within PP which may bear an uninterpretable wh-feature with an EPP-feature in the P-stranding structure. However, he does not specify what kind of functional projection it is, and what makes this difference between the colloquial and literary varieties.
Although Willis’ analysis nicely captures Welsh data on wh-dependencies, his treatment of P-stranding is still not conclusive. In the next section, I will present the PF checking account on Welsh P-stranding.