The next of Seminário CLUL is scheduled for the 27th of November at 4 pm in FLUL, in room B112.B. 

Marcel den Dikken 

Título: On the derivation and density of successive-cyclic Ā filler–gap dependencies

Resumo: 

1 Â-dependencies between a gap in a subordinate clause and a filler in a higher clause have, throughout the generative tradition, been modelled in terms of movement of the filler from the position of the gap up to its spell-out site. This movement is standardly taken to proceed successive cyclically. The most densely successive-cyclic movement derivations for long-distance  filler–gap dependencies are in Chomsky (1986) and Chomsky (2001), with intermediate stopovers on the edges of all VP/vPs and CPs. The morphosyntactic reflexes of long-distance  filler–gap dependencies advocate for terminal movement into the left periphery of the subordinate clause linked to a proleptic filler–gap dependency footed in the matrix VP and headed in CP.

2 In recent work, Chomsky (2024) advances an analysis of long wh-dependencies according to which the only movement step proceeds from the è-position (ec1) to the local phase-edge (ec2), within the confines of the subordinate clause. CQ in the matrix clause gains access to the features of the moved element relevant for interpretation at the interfaces, which are copied over onto CQ and the wh-phrase is externalized in the specifier of CQ. The analysis in (1), which delivers three syntactic manifestations of the displaced embedded object (one in the spell-out position, one in the è-position, and one on the edge of the phase local to the è-position), includes ingredients of Chomsky’s analysis of obligatory control. This paper aims to show that, although as they stand Chomsky’s accounts of control and long-distance Â-dependencies are untenable, the latter can be modified into a bona fide representative of the typology of cross-clausal  filler–gap dependencies.

(1) [CP wh CQ ... V [CP C ... [vP ec2 [vP ... V ec1]]]]

3 The West Ulster English examples in (2) (Henry 2012, amplifying McCloskey 2000) demonstrate with the help of the floating quantifier all that a nominal wh-expression serving as the object of a subordinate clause can be represented in the object position (2a) and on the edge of the embedded vP (2b) but not in SpecCP (2c). This is in line with Chomsky’s (1). However, problematic for this analysis are the facts in (2d,e), which show that the wh-phrase can float a quantifier in two positions in the matrix clause. For (2d,e), (1) provides no immediate explanation.

(2) a. what did she say yesterday [CP that he bought all in town]? (West Ulster English)
b. what did she say yesterday [CP that he all bought in town]?
c. *what did she say yesterday [CP all that he bought in town]?
d. what did she say all yesterday [CP that he bought in town]?
e. what did she all say yesterday [CP that he bought in town]?

4 The hypothesis that the copy on the edge of the embedded vP phase (wh2) is the product of terminal object shift (an anti-focus device), in conjunction with the assumption of a feature copying relation between ec2 and wh, presents an information-structural problem for (1). After terminal object shift to the edge of vP, the chain (ec1, ec2) in (1) is assigned the information-structural profile of a topic when the subordinate vP phase is sent off to the interfaces. Feature copying between ec2 (a topic) and wh (a focus) would then lead to an information-structural clash. Also problematic for the feature copying approach is the fact that the physical wh-expression upstairs can have a case feature distinct from the one expected for the wh-dependency downstairs — (3).
(3) kit gondolsz, hogy ec eljött? (Hungarian)
who.ACC think.2SG.INDEF that NOM PV.came.3SG
‘who do you think came?’

5 In the wake of Salzmann (2006) and den Dikken (2018), this paper argues for a syntax of (3) in which the overt wh-expression is a proleptic object originating in the matrix clause, linked to a silent topic in the subordinate clause: prolepsis cum resumptive topic drop. This is schematized in (4): kit ‘who.ACC’ originates in the matrix clause, gets assigned ACC case there, and is resumed in the subordinate clause by an anaphoric topic, droppable thanks to its recoverability.

(4) [CP wh CQ ... V ec3 [CP C ... [ec2=TOPIC [COMMENT ... ec1 ...]]]]
The dependency between the wh-expression in the matrix and the silent topic chain in the embedded clause involves neither copying nor movement. That no copying is involved is clear from the case and information-structural mismatch between the upstairs wh-expression and the silent element downstairs: (3). That no movement out of the lower clause into the higher clause is taking place is shown by the absence of wh-island effects, illustrated by sentences such as (5).

(5) kit kérdeztél, hogy mikor jött el? (Hungarian)
who.ACC ask.PST.2SG.INDEF that when came.3SG PV
(lit.) ‘who did you ask when came?’

6 In (4), the matrix wh-operator originates in the VP of the matrix verb, as an object, and gets the case expected of this object (ACC in (3)). This entails that the subordinate clause in (4) is not an argument of the verb. The matrix wh-phrase and the CP are in a predication relation, with the CP serving as the predicate of the matrix wh-phrase. CP is capable of serving as a predicate in (4) thanks to the fact that it contains an Â-dependency (between ec2 and ec1) to which no range is being assigned, as a result of which the CP that contains this Â-dependency is not fully saturated, allowing it to be predicated of the wh-constituent in the matrix clause. This is
directly parallel to what we find in tough-movement constructions (this book is tough to read).

7 The derivation in (4) sheds light on the severity of wh-island effects. Ross’s (1967) cline in (6) tracks the interaction of subordinate topicalization and question formation: (7). The extent to which relativization is possible out of an embedded question introduced by a range of different wh-elements parallels the extent to which it is possible to perform topic fronting inside an embedded question introduced by that same range of wh-elements. (4) directly predicts this parallel.
(6) he told me about a book which I can’t figure out {why/?whether/??when} he read
(7) I can’t figure out {why/?whether/??when} this book he read (while that one he sold)
The syntax in (4) is also explains why the English translations of Hungarian (5) are ungrammatical.This is because of a combination of two independent facts: (a) vacuous topicalization of the subject is not allowed (Lasnik & Saito 1992), and (b) English (unlike Hungarian) is not a pro-drop language, therefore the silent subject-topic can neither topicalize nor remain in situ, in SpecTP.

8 Because vacuous topicalization of subjects is not an option, (4) is unavailable in non-prodrop languages for subject dependencies unless topicalization of the subject into the high left periphery is non-vacuous. This explains why insertion of a high adverbial lifts the ban on a subject wh-dependency across an overt complementizer (Culicover 1993): for (8a) a derivation à la (4) is not available but for (8b) it is because the subject can be topicalized non-vacuously.
(8) a. *who do you think that is the mayor of this town?
b. who do you think that for all intents and purposes is the mayor of this town?

9 Chomsky’s (2024) (1) and my (4) both eschew successive-cyclic movement via SpecCP and establish an identificational relationship between a wh-element originating in the matrix clause and a silent topic in the subordinate clause; but (1) and (4) differ with respect to the base position of the matrix wh-operator, the nature of the relation between it and the silent topic, the relationship between the matrix wh-operator and the subordinate CP, and the position of the silent topic. By its nature, (4) can apply only to long-distance Â-dependencies involving arguments. Alongside (4), the typology of Â-dependencies includes several other derivations (incl. fell-swoop movement, à la Rackowski & Richards 2005, and wh-scope marking), which will also be laid out in the paper. But no long-distance Â-dependency is successive cyclic in the traditional sense of featuring an intermediate copy in the highest specifier position of the subordinate clause.

Chomsky (1986) Barriers • Chomsky (2001) Derivation by phase • Chomsky (2024) The miracle creed and SMT • den Dikken (2018) Dependency and directionality • Lasnik & Saito (1992) Move á • Henry (2012) Phase edges, quantifier float and the nature of (micro-)variation • Rackowski & Richards (2005) Phase edge and extraction • Salzmann (2006) Resumptive prolepsis.