This article describes and analyzes possessor extraction (PE) in English, a restricted possibility in the colloquial language of some speakers. I argue that the complexities of this phenomenon reveal evidence for the Cyclic Linearization theory of Spell-Out (e.g., Fox and Pesetsky 2005), which constrains English PE by interacting with a (phase-level) requirement to keep the possessor and the possessive D adjacent at PF (e.g., Gavruseva and Thornton 2001). These factors prevent such PE from succeeding unless the possessum is stranded in a clause edge, among other restrictions. This analysis entails the nonphasehood of DP, clarifies the derivation of that’s-relatives, reveals some linearization constraints on stranding, and suggests that expletive there originates in vP.

1 Introduction

This article examines a case of possessor extraction (PE), the Ā-movement of a possessor from the possessed nominal phrase. For many English speakers, Ā-movement of possessors requires pied-piping of the containing possessum, as in (1).

(1)

  • Standard English possessum pied-piping

  • Mary is the author [CP [whose new book]k they said [CP ______k is good]].

Such pied-piping is often thought to be the only possibility for English. This view is challenged by examples like (2), which are the subject of this article. Example (2) is an equivalent of (1) available in the colloquial language of some speakers, in which the possessor extracts, stranding the Saxon genitive morpheme [’s] as well as the rest of the possessum in an embedded clause. As PE is not available to all speakers, this initial English PE example is appropriately marked with %, though I omit this from subsequent examples.

(2)

  • PE in English

  • %Mary is the author [CPwhok they said [CP [_____k’s new book] is good]].

In (2), [’s] becomes phonologically dependent on the verb said in the absence of the moved possessor. It is easy to see that this -s really must be a stranded Saxon genitive morpheme [’s]. The past tense and plural subject of the relative clause in (2), where PE is taking place, eliminate the possibility that this element is subject agreement. The fact that the possessor is female, since it corresponds to Mary, also eliminates the possibility that [’s] is a reduced resumptive his.

Though not all native speakers accept such PE examples, the present study of 34 speakers, mostly residents of the Boston area, resulted in 19 judging PE of this form to be acceptable.1 Those who accept the construction frequently note that it is distinctly colloquial, a fact that may contribute to its rarity in written form.2 Some speakers are aware that they are capable of English PE, while many others are surprised to notice its acceptability when it is brought to their attention. While PE has been well-established in various languages like Hungarian (3),3 the possibility of PE in English has received little attention.

(3)

  • Hungarian PE

  • Ki-nekk ismer-té-tek [a _____k vendég-é-

    graphic
    -t ]?

  • who-DAT know-PST-2PL [the guest-POSS-3SG-ACC]

  • ‘Whose guest did you know?’ [Lit.: ‘Whose did you know guest?’]

  • (Szabolcsi 1984:92, (14))

The only work I know to have considered the existence of PE in English is Gavruseva and Thornton 2001, discussed in section 1.1.4 While I argue that English PE involves true movement, we will see that its distribution is quite restricted, which plausibly contributes to the fact that it has not been widely noticed. An analysis of this construction’s restrictions and their consequences for syntactic theory is the focus of this article.

1.1 Background

The possibility of this form of PE contrasts with the known impossibility of extracting whose (which I take to be who + [’s]) or any possessor marked with [’s] in English.

(4)

  • No extraction of whose

  • *Mary is the author [CPwhosek they said [[_____knew book] is good]].

If a possessor DP is the specifier of a possessive D whose exponent is [’s] in English (e.g., Corver 1992, Chomsky 1995), then whose and elements like it are not syntactic constituents, and therefore their immobility is expected. However, the specifier of [’s] is surely a phrase, and as such is in principle extractable.

(5)

graphic

While for some English speakers the possessor DP in the specifier of [’s] can indeed be moved, as in (2), the marking % on (2) is a reminder that for many speakers such movement is not possible.

Various works attribute the typical illicitness of PE in English to a PF requirement that rejects movement that separates a possessor from the possessive D (e.g., Chomsky 1995, Radford 1997, Gavruseva 2000, Gavruseva and Thornton 2001). Indeed, Gavruseva (2000) argues that PF adjacency requirements of this variety play an important role in constraining PE crosslinguistically. In this article, I accept this general line of explanation for those English speakers who reject PE. However, I argue that the nature of PE in English does not indicate that the speakers who permit it lack such an adjacency requirement; rather, it shows that such speakers can satisfy this requirement more locally. This analysis accurately predicts a constrained form of PE in English, as we will see.

Gavruseva and Thornton’s (2001) investigation of PE focuses on whose-questions in child speech, where PE is quite frequent in long-distance questions. Gavruseva and Thornton argue that PE is possible in child English because children have not yet acquired the PF constraints that require pied-piping and consequently block PE. This perspective on the acquisition path leads us to expect a total lack of PE in a mature English grammar. However, in a control study with adults reported in the same work, Gavruseva and Thornton (p. 255) observed PE in adult speech. Of their adult data, 11% consists of PE of the form shown in (2).5 The authors suggest that these cases are the result of production errors. However, almost all such instances of PE gathered in their adult study were from two speakers, Cristy and Kath. Cristy produced PE about half as often as pied-piping, while Kath produced PE even more often. These speakers evidently have PE as a productive option. Indeed, in this article I argue that PE is a reality for some English speakers, though subject to restrictions, several of which Gavruseva and Thornton also identify. Explaining these constraints and their significance is the central task of this article.

1.2 Preview of Results

I argue that English PE obeys the following generalization, which subsumes an array of restrictions on this construction that I will demonstrate in section 2:

(6)

  • Generalization about English PE and clause edges

  • A possessor must reach the left linear edge of the local CP before extracting from the possessum DP.

One particularly clear instantiation of this generalization is evident with PE from nonsubjects, exemplified in (7a–b). Here we see that PE from an object possessum is unacceptable if this possessum is stranded in its base position (7a). Rather, the possessum must be displaced to the edge of the local CP (7b).

(7) Displacement of nonsubject possessum under PE

  • *Whok do they think [CP Mary read [_____k’s book]]?

  • Whok do they think [CP [_____k’s book]j Mary read tj ]?

I argue that the generalization in (6), which describes (7a) and related restrictions, emerges from two concepts: the Cyclic Linearization (CL) theory of Spell-Out (Fox and Pesetsky 2005a,b, Ko 2007, 2011, 2014, Podobryaev 2007, Sabbagh 2007, Fanselow and Lenertová 2011, Erlewine 2017, Davis to appear), and a version of the independently proposed requirement that the possessor and the possessive D [’s] be adjacent at PF. In particular, I argue that while English speakers who reject PE enforce adjacency between possessor and [’s] absolutely, speakers who permit PE can satisfy this requirement locally, as stated in (8).

(8)

  • Possessor-[’s] Adjacency (Local version, available to PE speakers)

  • [’s] must be adjacent to its associated possessor at the spell-out of the minimal phase (vP, CP) containing it.

According to CL, with which (8) interacts to yield the restricted distribution of English PE, successive-cyclic movement (and certain exceptions to it) is motivated by the information-preserving nature of Spell-Out—Order Preservation. This property of Spell-Out only allows syntactic derivations that do not generate contradictory linearization information; thus, as we will see, it yields a variety of constraints on movement.

The power of CL in predicting the details of English PE, a restricted and apparently infrequent construction, provides evidence that CL is an aspect of Universal Grammar (UG). Under this account, CL is a part of the grammar of both speakers who permit PE and those who do not. The difference between these two groups lies in how they enforce a PF condition. This understanding maintains a uniform syntax, variation being attributed to the PF interface.

1.3 Roadmap of the Article

Section 2 describes the facts about English PE, which section 3 argues is true extraction. Section 4 introduces the concepts used in section 5 to build an account of this phenomenon. Section 6 discusses this account’s consequences for syntactic theory, namely, the correctness of the CL approach and the nonphasehood of DP. Section 7 extends this account to a novel analysis of that’s-relatives (Seppänen and Kjellmer 1995, McDaniel, McKee, and Bernstein 1998, McDaniel et al. 2002), which supports Deal’s (2016) proposal that TP is a phase in relative clauses. Section 8 examines some predicted constraints on stranding in phase edges, leading to an argument that expletive there originates in vP (Biberauer and Richards 2005, Deal 2009). Section 9 discusses some residual considerations for this analysis of English PE, and is followed by the conclusion in section 10.

2 The Restricted Distribution of English PE

2.1 The Basic Pattern

This section describes the facts about English PE. This exposition involves some preliminary analysis, leading to the generalization in (6), which will be derived in section 5.

Gavruseva and Thornton’s (2001) study of PE in child speech focuses on questions, but English PE is possible in any Ā-movement context, as (9) shows.

(9)

  • Matrix question

    Whok do you think [[_____k’s kid] ate the most cake]?

  • Embedded question

    I can’t remember [whok I said [[_____k’s friend] is coming over]].

  • Relative clause

    This is the student [whok they suspect [[_____k’s answers] were copied]].

  • Free relative

    I’ll speak to [whok ever you suggest [[_____k’s idea] is the best]].

  • Cleft

    It’s Michelle [whok we heard [[_____k’s cat] is the cutest]].

  • Topic/Focus movement

    John’s life might be boring, but let me tell you about Jim. This guyk, I think [[_____k’s story] will intrigue you].

Most of the PE examples shown so far have involved extraction of who. Other possessors can be extracted also, as (10) shows. Extraction of phonologically larger possessors is often judged as more difficult to accept. For maximal clarity of judgments, most PE sentences reported here involve extraction of who.

(10) Extraction of other possessors

  • [Which student]k did he claim [[_____k’s idea] was stolen]?

  • [How many people]k do you think [[_____k’s papers] fell off the shelf]?

  • I went to the place [whichk she said [[_____k’s pastries] are tastiest]].6

Further examination of this construction reveals numerous restrictions. First, notice that all of the English PE examples shown so far have been multiclausal. This is not a coincidence. As (11) shows, English PE is not possible for clause-bound movement.

(11) PE unavailable for clause-bound movement

  • *Whok did you meet [_____k’s friend] yesterday?

  • *Whok should [_____k’s article] cause a big controversy?

  • *John is the one [whok you stole [_____k’s pie]].

Additionally, most of the English PE examples presented so far show extraction from a subject. As previewed in (7) and shown again in (12), nonsubject possessum DPs exited by PE must be displaced to the edge of the local CP.7

(12) Displacement of nonsubject possessum under PE

  • *Whok do they think [CP Sue found [_____k’s cat] today]?

  • Whok do they think [CP [_____k’s cat]j Sue found _____j today]?

  • *Mary is the person [whoj I heard [CP John ate [_____j’s food]]].

  • Mary is the person [whoj I heard [CP [_____j’s food]k John ate _____k]].

  • *Tell me [whoj they said [CP we should meet [_____j’s friend]]].

  • Tell me [whoj they said [CP [_____j’s friend]k we should meet _____k]].

Examples (12a–f) demonstrate the obligatory application of this displacement for PE from the object of a transitive clause. As (13)–(15) show, the same occurs with nonsubject possessum DPs in general. The examples in (13) and (14) show that this displacement holds for PE from either nonsubject argument of a ditransitive.

(13) PE from direct object of a ditransitive

  • Whoj do they think [[_____j’s book]k we should give Mary _____k]?

  • Whoj do they think [[_____j’s book]k we should give _____k to Mary]?

(14) PE from indirect object of a ditransitive8

  • Whoj do they think [[_____j’s kid]k we should give _____k the prize]?

  • Whoj do they think [[_____j’s kid]k we should give the prize to _____k]?

The examples in (15) show that the same holds for expletive associates. Such arguments are postverbal by default, though under PE they must end up at the edge of CP, as in (15c). This example is marked, but clearly improves on (15b), which lacks the needed displacement.9

(15) PE from expletive associate

  • Mary said [there were [someone’s books] on the table].

  • *Whoj did Mary say [there were [_____j’s books] on the table]?

  • ?Whoj did Mary say [[_____j’s books]k there were _____k on the table]?

Gavruseva and Thornton noticed the obligatory displacement of nonsubject possessum DPs under PE (specifically of objects) in their study as well, in both children and adults. They hypothesize that this is caused by the possessor’s movement pied-piping the possessum to the specifier of the embedded CP, and then stranding it there by subsequent extraction. This is the view defended here. If this analysis is correct (as section 9.1 argues in greater detail),10 then such stranding provides evidence that movement from CPs successive-cyclically passes through their edge, joining similar arguments from previous literature on stranding in mainstream English (Urban 1999), West Ulster English (McCloskey 2000, Henry 2012), Polish (Wiland 2010), and Wolof (Torrence 2019), among other languages. These works show elements that can be stranded in a CP edge, as well as in their base position. We have just seen, however, that English PE from a nonsubject possessum does not allow the possessum to be stranded in its base position (in contrast to PE in Hungarian, as shown in (3)). This fact suggests that English PE is more complex than typical cases of stranding.

2.2 The Possessor Extracts from DP via the Linear Edge of CP

We have seen that PE from nonsubject possessum DPs requires displacement of the possessum to the edge of the local CP. At first glance, this fact suggests that PE is only possible from the structurally highest DP in the clause. Before any Ā-movement, the structurally highest DP is whatever ends up in Spec,TP. If the DP exited by PE is not in Spec,TP, it consequently must be pied-piped to Spec,CP with the moving possessor prior to PE. This description is consistent with what has been shown so far.

If this were a sufficient description, however, PE of a postnominal possessor from a subject should be grammatical. In reality, such configurations are not accepted.

(16) PE unavailable for postnominal possessors

  • *Whok did you say [CP [a cookie recipe of _____k’s] is getting popular]?

  • *That’s the senator [whok they think [CP [a friend of _____k’s] got a huge bribe]].

We have already seen examples of PE from subjects, so there is no benefit to attributing the deviance of (16a–b) to the known fact that extraction from subjects is frequently degraded. Examples (16a–b) improve if [’s] is absent (to the extent that extraction from indefinite subjects is relatively tolerable), indicating that this morpheme’s requirements are influential in constraining PE, as this article argues.

Notice that in (16a–b), overt material within the possessum intervenes between the trace of PE at the right side of this DP, and the left edge of the local CP. The fact that PE is unacceptable in this situation suggests the following generalization, which I will show is correct:

(17)

graphic

This generalization is graphically depicted in (18).

(18)

graphic

Given (17), we expect the unacceptability of (16a–b), since here there is overt material separating the trace of PE from the left linear edge of CP. Additional facts about English PE shown in the remainder of this section have the same explanation.

Before we move on to the remaining facts, notice that the generalization in (17) is consistent with the displacement of nonsubject possessum DPs under PE that we saw earlier in this section: since this displacement brings the possessum to the CP edge, such configurations also fit (17). This generalization also clarifies the impossibility of PE for clausebound movement, shown in (11): if (17) holds, PE cannot become evident unless the possessor exits a possessum in an embedded CP edge.

2.2.1 Preposition Stranding and PE

PE is not possible from a DP within a PP; thus, that PP must be stranded prior to PE.

(19) Pied-piped P with displaced nonsubject blocks PE

  • Whoj do they think [[_____j’s house]k we should leave from _____k]?

  • *Whoj do they think [[from _____j’s house]k we should leave _____k]?

  • Whoj did they say [[_____j’s cat]k we should give the prize to _____k]?

  • *Whoj did they say [[to _____j’s cat]k we should give the prize _____k]?

As with other nonsubject DPs, PE from the DPs in (19) requires them to be displaced to the edge of the local CP, presumably via pied-piping with the possessor’s movement as discussed above. These DPs originate inside of PPs, and even though P may typically be pied-piped along with Ā-movement of its DP complement in English, in (19) only P-stranding is permitted. This is what we expect given (17): if P is pied-piped along with movement of the nonsubject possessum to the CP edge, then P illegally intervenes between the left edge of the embedded CP and the trace of PE in DP.

2.2.2 Complementizers and PE

The distribution of complementizers in PE derivations also fits the linear generalization in (17). Recall that in English, long-distance wh-movement of nonsubject DPs is compatible with an overt complementizer that in the embedded clause.

(20)

  • Overt complementizer that with nonsubject extraction

  • Whoj do they think [CP (that) Mary likes _____j ]?

Subject extraction, however, is not compatible with an overt complementizer that, a phenomenon well-known as the that-trace effect.11

(21)

  • Appearance of that-trace effect with subject extraction

  • Whok do they think [CP (*that) _____k likes Mary]?

PE from a subject is also incompatible with the complementizer that (22). This fact is interesting because here we have extraction from, but not (cross-clausal) movement of, the subject. Thus, this fact does not necessarily instantiate the that-trace effect.12

(22) Overt complementizer that disallowed with PE from subject

  • Whok do you think [(*that) [_____k’s friend] always laughs at bad jokes]?

  • Whok did they say [(*that) [_____k’s window] broke during the storm]?

Notice that in (22a–b), the presence of the complementizer that to the left of the possessum subject separates the trace of PE within this DP from the left linear edge of the local CP. Thus, the unacceptability of the complementizer here is expected, since it violates (17).

Nonsubject DPs exited by PE also cannot be preceded by an overt complementizer that (23). However, if the possessum in such configurations is stranded in Spec,CP as discussed earlier in this section, then a complementizer should be unable to precede it anyway, independent of the fact that a complementizer in this position would also violate (17). An overt complementizer that to the right of the possessum is also impossible here, as expected given the Doubly Filled Comp Filter (Chomsky and Lasnik 1977).

(23) Overt complementizer that disallowed with PE from a pied-piped nonsubject possessum

  • Whoj did you say [(*that) [_____j’s cat]k (*that) John saw _____k]?

  • Whoj does he think [(*that) [_____j’s cat]k (*that) he wants to take home _____k]?

2.2.3 Adverbs and PE

Typically, high adverbs may occur on either side of the English subject.

(24) Variable high adverb position

  • (Fortunately) Mary (fortunately) has money.

  • (Usually) John (usually) eats a burrito for lunch.

PE from a subject is incompatible with such an adverb to that subject’s left, but is possible with an adverb to its right.

(25)

  • An adverb cannot precede a subject exited by PE

  • Whok did you say [(*usually/*fortunately) [_____k’s friend] (usually/fortunately) has an extra taco]?

This is expected if (17) holds, since an adverb to the left of the possessum subject linearly intervenes between the trace of PE in this DP and the edge of CP. PE from a nonsubject, involving stranding in the CP edge, behaves the same.

(26)

  • An adverb cannot precede a nonsubject exited by PE

  • Whoj did you say [(*allegedly/*fortunately) [_____j’s cat]k (allegedly/fortunately) John found _____k]?

2.3 The Puzzle

This section has shown that English PE is subject to the generalization in (17), which describes the fact that no overt material may intervene between the trace of PE within DP, and the left linear edge of the local CP. I will argue that this generalization emerges from the interaction of CL and an adjacency requirement of [’s] that PE speakers can satisfy at a local (phase-bounded) level of the derivation.

2.3.1 Against a Selective [’s]

Some of the unacceptable examples of English PE shown in this section might, at first glance, be attributed to restrictions on what [’s] can cliticize to. Upon closer inspection, however, it becomes unclear how to state what the relevant restrictions would be. We have seen that [’s] can attach to verbs in PE contexts ((2), and many more) but not adverbs ((25), (26)) or functional heads like P ((16), (19)) or C ((22), (23)).13 These facts do not yield an obvious generalization about what [’s] may attach to in PE derivations. It is also not obvious why [’s] would be selective in PE contexts, even though in typical non-PE contexts it is not selective and can cliticize to adverbs and functional heads, among other elements (Zwicky 1987).

(27) Typical unselective cliticization of [’s]

  • [the person you’re talking to]’s jacket

  • [the man who left yesterday]’s book

It would also remain puzzling that [’s] can cliticize onto verbs in some PE contexts, but not in those like (7a) where an object possessum is stranded in its base position.

Given these issues, in this article I account for the facts about English PE without positing such restrictions on [’s]. With this hypothesis set aside, in the next section I make the case that English PE is truly movement, setting the stage for the core analysis previewed above.

3 English PE Is True Extraction

Recall that English PE only occurs in long-distance Ā-movement contexts, unlike PE in languages like Hungarian. This fact might be thought to show that English PE is an illusion created by a DP-internal parenthetical clause, between the possessor and [’s]. This parenthetical would make the resulting construction always appear multiclausal, since it contributes an additional verb to the surface string. For instance, the initial PE example (2) could involve true extraction in a biclausal context (28a), or a single clause with a parenthetical they said in the possessed DP (28b).

(28) String: Mary is the author who they said’s new book is good.

  • Extraction analysis

    Mary is the author [CP2 whok they said [CP1 [DP _____k’s new book] is good]].

  • Parenthetical analysis

    Mary is the author [CP [DP who (they said)’s new book] is good].

As (29) shows, parentheticals are not independently permitted in this DP-internal position (e.g., Emonds 1976), weakening the parenthetical analysis of PE.

(29) Parentheticals are not permitted between DP[Poss]and [’s]

  • [DP John (*in my opinion/*of course)’s idea] is funny.

  • [DP A friend of the teacher (*Mary said/*in fact)’s] visited yesterday.

Even if this argument were not conclusive on its own, other lines of evidence show that the parenthetical analysis of English PE is inferior to the movement analysis.

3.1 Failures of Parenthetical Subtraction

Parentheticals are optionally inserted into what are otherwise well-formed sentences. Therefore, if PE constructions in fact involve parentheticals, subtracting the content that is supposedly parenthetical should yield a licit configuration. This diagnostic reveals numerous PE configurations that cannot have been derived by use of a parenthetical.

Consider the PE question in (30). Subtracting the supposed parenthetical here yields an impossible string, whether or not the auxiliary do is counted as part of the parenthetical.14

(30) Failed parenthetical subtraction: Who do they think’s cat he saw?

  • Who do they think’s cat he saw? → *Whose cat he saw?

  • Who do they think’s cat he saw? → *Who do’s cat he saw?

Similar facts become evident when we consider the phenomenon of “free deletion in Comp” (Chomsky and Lasnik 1977), which can derive examples like (31), where the wh-operator responsible for relativization is optionally silent.

(31) Relativization with null operator

  • the one [whok/

    graphic
    k I like _____k]

  • the cat [whichk/

    graphic
    k I saw _____k]

Relativizing PE with no overt wh-operator is also possible (32).15 Removing the supposed parenthetical material from such sentences yields an unacceptable result.16

(32) PE of null relativizing wh-operator

  • the person [

    graphic
    WHkI said [[_____k’s cat] meows too much]]

    → *the person[’s cat meows too much]

  • the person [

    graphic
    WHkI said [[_____k’s cat]j you saw _____j]]

    → *the person[’s cat you saw]

  • the person [

    graphic
    WHkI think [[_____k’s friend]j you talked to _____j]]

    → *the person[’s friend you talked to]

A similar effect is apparent with control. In the PE sentence in (33), the subject of the supposed parenthetical controls into a before-adjunct. Removing the supposed parenthetical yields an unacceptable sentence where PRO is uncontrolled.

(33)

  • Adjunct control with PE

  • [Which author]jdid youksay [[_____j’s book] looked good] [before PROk ordering it]?

  • → *Which author’s book looked good [before PRO?? ordering it]?

To sum up, given that parenthetical subtraction fails in numerous PE configurations, I argue that this construction is not derived by insertion of a parenthetical.

3.2 PE Is Blocked by Nonbridge Verbs

If English PE is an illusion caused by a DP-internal parenthetical, then we expect the set of verbs that are licit in parentheticals and the set of verbs that can appear to be crossed by illusory PE to be the same. This is not the case. Consider whisper, which is productive in parenthetical clauses.

(34)

  • Productive parenthetical whisper

  • Mary (John whispered) wants (John whispered) a kitten (John whispered).

This verb is among the manner-of-speech verbs (mutter, stammer, mumble, groan, etc.) that are nonbridge verbs, whose complements are not transparent for extraction, though these verbs are productive in parentheticals (Zwicky 1971:225, (14)). As (35a) shows, pied-piping possessor movement from the complement of such a verb is unacceptable. An equivalent PE configuration is no better (35b).

(35) No extraction (PE or otherwise) from complement of nonbridge verbs

  • Pied-piping possessor movement

    the person [[whose cat]k I thought/said/*whispered/*groaned [_____k was cute]]

  • PE

    the person [whok I thought/said/*whispered/*groaned [[_____k’s cat] was cute]]

In contrast, verbs with transparent complements (think, say, claim, suspect, tell, believe, hear, etc.) are always acceptable with PE, as we have seen throughout the article.

3.3 PE Is Blocked by Islands

Finally, placing an island boundary between the extracted possessor and [’s] results in unacceptability, as expected if this construction involves genuine extraction. For more on islands, see section 9.1, which shows that PE is sensitive to weak islands as well.

(36) Adjunct Island Constraint

  • *Whok did you go home [because I said [[_____k’s friend] is ugly]]?

  • *[Which kid]k did you leave class [before I said [[_____k’s shoes] are ugly]]?

(37) Complex NP Constraint

  • *Whok did you hear [the rumor that they said [[_____k’s friend] will get hired]]?

  • *[Which teacher]k do you know a student [that said [[_____k’s class] is hard]]?

(38) Coordinate Structure Constraint

  • *Whok did [[she say [[_____k’s friend] is nice]], and [he say [Bill is mean]]]?

  • Whok did [[she say [[_____k’s cat] is nice]], and [he say [[_____k’s dog] is mean]]]?

Overall, the facts presented in this section indicate that English PE truly involves movement, and not a deceptive parenthetical construction.17

4 The Two Mechanisms That Constrain English PE

As previewed, this article argues that the restricted distribution of English PE emerges from the interaction of CL and an adjacency requirement of [’s] that PE speakers can satisfy locally. This section explains these concepts, and the next applies them.

4.1 Cyclic Linearization

Chomsky (2000, 2001, et seq.) and other authors argue that phrasal movement must successive-cyclically pass through the edge of certain constituents termed phases (including at least vP and CP). In this theory, the phase edge is an escape hatch because it is not subject to phase-level Spell-Out, which targets only the phase head’s complement, transferring it to the interfaces of PF and LF. After Spell-Out, the complement is inaccessible to further syntactic operations. Thus, material moving from a phase head’s complement must reach the edge of the phase before Spell-Out applies; otherwise, it will be trapped in the complement and unavailable for further movement.

In contrast, Fox and Pesetsky (2005a,b, et seq.) and other authors argue that Spell-Out applies to entire phasal constituents, edges included. Phases spell out as soon as they are done being built up by successive applications of Merge. This hypothesis requires that Spell-Out not make constituents impenetrable, because in this system, all movement from a phase is of material that has undergone Spell-Out within that phase. Thus, successive-cyclic movement through phase edges does not occur to prevent a moving phrase from being trapped by phase-level Spell-Out. Rather, Fox and Pesetsky argue that such movement is motivated by the information-preserving nature of Spell-Out, Order Preservation.

(39)

  • Order Preservation

  • Information about linearization, once established at the end of a given Spell-Out domain, is never deleted in the course of a derivation.

  • (Fox and Pesetsky 2005a:6)

If Order Preservation holds, it is not possible to revise previously established ordering information in order to save derivations that end up with a final linearization that is not coherent. Therefore, syntax must be able to construct configurations that end up with linearization information that is consistent across all phases in a given derivation, since if it does not, a crash at PF is unavoidable. Fox and Pesetsky argue that exiting a phase by moving out via its linear edge is one way to keep linearization information consistent within a derivation.

(40)

graphic

By moving via the linear edge of each phase passed, phase-exiting phrases are determined by PF to precede the rest of the content of each phase in question. This is ultimately consistent with a final representation where the moved material precedes all phases that it has exited.

If movement from a phase does not pass through the linear edge, hence crossing over some material in the phase on the way out, there is a way to salvage the derivation. Moving that crossed-over material into the next phase to a position preceding what previously crossed it, thus restoring the original order of those elements, keeps linearization coherent. For instance, (41a) is ungrammatical if it remains as is, because α. has non-successive-cyclically crossed over β on the way out of the phase. However, the derivation will not fail if, as (41b) shows, β moves to precede α in the next phase as it did in the first.

(41)

graphic

This schema is the essence of Fox and Pesetsky’s account of Holmberg’s Generalization.

In section 5, we will see that pressure to obey the scenarios in (40) and (41) restricts PE by interacting with the adjacency requirement of [’s], described below.

4.2 Adjacency of [’s] and the Possessor

As mentioned previously, several works argue that PE in English is typically prevented by a PF requirement mandating adjacency of the possessor and the possessive D [’s] (e.g., Chomsky 1995, Radford 1997, Gavruseva 2000, Gavruseva and Thornton 2001). Indeed, Gavruseva (2000) argues that a requirement of adjacency to possessive D is influential in blocking PE in many languages.

To describe an English grammar that does not allow PE, it will suffice to state the relevant requirement as follows:

(42) Possessor-[’s] Adjacency (Global version)

For any derivation containing [’s], [’s] must be linearly adjacent to its associated possessor at the final PF representation of that derivation.18

This constraint is phrased in such a way that it must be satisfied at the final PF representation generated by the derivation in question. A grammar with such a constraint will never permit separation of possessor and possessum, as is indeed the case for many English speakers.

However, it is necessary to say something different about the grammar of those speakers who do permit PE. I argue that (constrained) PE is an option for such speakers because they are able to satisfy adjacency to [’s] more locally. In particular, I argue that such speakers can satisfy this requirement in a phase-bounded way.

(43)

graphic

As we will see, after this locally evaluated requirement is satisfied, subsequent movement operations can break adjacency between the possessor and [’s]. Precisely because [’s] is not carried along into subsequent phases after successful PE, this adjacency requirement is not applicable to those later phases, and the possessor can move on freely. For more discussion about the nature of this adjacency requirement, see section 9.2 and footnote 34.

4.3 The Importance of Spelling Out Phase Edges

The fact that the CL theory includes phase edges in the Spell-Out domain of phases is critical for the account of English PE proposed here. This system allows phase-level Spell-Out and the PF adjacency requirement of [’s] in (43) to interact with successive-cyclic movement through phase edges. As we will see, this interaction results in satisfaction of the adjacency requirement of [’s] only under particular circumstances, as desired.

The needed interaction is not possible in Chomsky’s (2000, 2001) theory, where phase-level Spell-Out is limited to phase complements. To see why, consider that in a PE derivation, successive-cyclic Ā-movement brings the possessor to the edge of each phase being exited. For PE to actually occur, there will necessarily be a point in the derivation where the possessum is stranded in the complement of a phase to whose edge the possessor has extracted. In such a configuration, as schematized in (44), the extracted possessor and the possessum are separated by a Spell-Out domain (here, YP).

(44)

graphic

When Spell-Out applies in (44), the local adjacency requirement of [’s] will not be met. This is because the extracted possessor moved outside of the Spell-Out domain YP of this phase XP, before Spell-Out applied to YP. Thus, Spell-Out of YP finds [’s] nonadjacent to the possessor, and this derivation fails. This failure is avoided if the possessum is pied-piped along with movement of the possessor to the phase edge. However, when this happens, PE fails to occur. This issue arises at any point where a Spell-Out domain would separate the possessor and the possessum, leading this theory of phases to incorrectly predict a total lack of PE.19

5 Predicting the Facts

This section shows how the concepts explained above predict the details of PE in English, which obeys the following generalization:

(45)

graphic

I first discuss PE from subjects, then the more complex case of PE from nonsubjects. Following Legate (2003), Ko (2014), and references therein, I take all vPs to be phasal, a concept that will also be relevant to the discussion of expletives later on.

5.1 PE from Subjects

This section analyzes PE from subjects. We have seen that English PE is only possible for cross-clausal movement, a fact whose explanation will be clear by the end of the section. Given this constraint, the analysis of a PE sentence must begin with the vP of the embedded clause in which the possessor originates. In the case of PE from subjects, this part of the derivation is straightforward, since the interactions most important to this analysis do not fully emerge until the derivation of the embedded CP.

5.1.1 PE from Subjects: The Embedded-vP Level

The facts reported in this article show PE from both external-argument and internal-argument subjects. First, let us consider what occurs within the embedded vP of a derivation with PE from an external argument. If external arguments originate in Spec,vP, as in transitive and unergative contexts, then no successive-cyclic movement is necessary at this stage of the derivation. The in-situ external argument and its possessor simply start out at the linear edge of the vP phase, which they will soon exit.

(46)

graphic

Further, if movement of a phrase to the specifier of a head requires a probing feature on that head to find that phrase in its c-command domain (Chomsky 1995, 2001), then phrase-bounded specifier-to-specifier movement is not possible (Ko 2007, 2014). This is because a head does not c-command, and therefore cannot move, any of its specifiers or anything inside them. This is illustrated in (47), where we see that the head α. c-commands its complement κP and all that it contains, but not its specifier βP or any content thereof.

(47)

graphic

Thus, movement of βP or γP to a higher specifier of αP in (47) is not possible. In the same way, extraction of the possessor in (46) to a higher specifier of vP is not only unnecessary as far as CL is concerned, but impossible anyway.

Second, consider the vP derivation involved with PE from a subject that is an internal argument, as in unaccusative and passive contexts. Independent of PE, given that such a subject is externally merged as a complement of V, it must move to Spec,vP in order to maintain a coherent linearization under CL: because English V moves to v (e.g., Larson 1988, Chomsky 1995, Kratzer 1996), and because the subject ultimately precedes V from its final position in Spec,TP, it is necessary for the internal-argument subject to move to Spec,vP in order to precede V at the spell-out of vP. That A-movement of subjects indeed passes through vP edges is independently argued by Sauerland (2003). Notice that such movement of an internal-argument subject automatically brings a possessor contained within it to the linear edge of vP, as shown in (48).

(48)

graphic

In principle, it is also possible for the possessor to extract to the vP edge, with the internal-argument subject then moving to a lower specifier of vP below the extracted possessor via tucking-in (Richards 1997, 1999), as in (49). This string-vacuous PE satisfies the adjacency requirement of [’s] just as if the possessor had not exited DP.

(49)

graphic

Because the derivation in (48) accomplishes the same thing as (49) but with fewer movement operations (modulo the different final constituency), we might expect economy concerns to favor (48).20 However, nothing of substance changes for the account proposed here if the reality is (49), or if both derivations are available.

This concludes the consideration of the embedded vP phase for derivations with PE from a subject. So far, the adjacency requirement of [’s] has not been critically relevant, since the context where its influence is most evident has not yet arisen. This requirement is much more influential on the derivation of the CP phase, as we will see next.

5.1.2 PE from Subjects: The Embedded-CP Level

After the completion of vP, I assume that upon external Merge of T, the subject undergoes A-movement to Spec,TP. When C is merged, the opportunity for PE arrives. In section 2, I showed that at this stage of the derivation, various restrictions hold. In short, as (50) shows again, nothing can intervene between the trace of PE in the possessum DP and the linear edge of the embedded CP.

(50) CP edge restrictions on PE from subjects

  • Whok did you say [(*often) [_____k’s friend] (often) has money]?

  • Whok do they think [(*that) [_____k’s name] (*that) is Mary]?

Before I analyze the unacceptable variants of the examples in (50), let me establish why these examples are acceptable when this problematic material in the left edge of the embedded CP is absent.

If no such material is present in the embedded CP edge, then after A-movement of the possessed subject to Spec,TP, the possessor it contains is already at the linear edge of CP. The possessor could string-vacuously extract to Spec,CP from here, though such movement is unnecessary.

(51)

graphic

The linear order established at the spell-out of this CP satisfies the adjacency requirement of [’s], which is linearly adjacent to the possessor at PF whether or not the possessor string-vacuously extracts at this point. CL is also satisfied here, since the extracting possessor has reached the linear edge of the embedded CP either way.

Next, the possessor can extract into the matrix vP (and then onward), stranding the possessum and the [’s] it contains in the lower CP, as in (52). When the matrix vP spells out, [’s] is not present within that phase to enforce its adjacency requirement. This is because [’s] has been stranded in a lower phase that has already undergone Spell-Out (as signified by the box around the embedded CP in (52)), at which point the adjacency requirement of [’s] was locally satisfied. As a result, extraction of the possessor succeeds.

(52)

graphic

Critical to this analysis is the hypothesis that the adjacency requirement under discussion is a property of the bound morpheme [’s] only, not of the possessor.

Next, let us examine a PE derivation where there is problematic material in the edge of the CP exited by PE. Consider (53), where the embedded CP contains an adverb in the left periphery. CL motivates the possessor to exit this CP via its linear edge. Therefore, the possessor must move to a position within CP to the left of that high adverb, in order to facilitate later movement. Notice that if the possessor moves around the adverb within CP in this way, thereby stranding the possessum subject in Spec,TP, then the adverb consequently intervenes between the moved possessor and [’s].

(53)

graphic

While later movement of the possessor from CP in (53) would satisfy CL, there is a problem. When this embedded CP undergoes Spell-Out, PF will find the possessor and [’s] nonadjacent because of the intervening adverb. Therefore, this CP will be deviant at PF. However, there is a way to avoid this problem: pied-piping the possessum with the possessor’s movement over the adverb to the edge of CP, as in (54).

(54)

graphic

This pied-piping movement satisfies CL as well as the adjacency requirement of [’s] within this embedded CP. After (54), the possessor can successfully extract into the matrix vP. Such a derivation ends up with the high adverb to the right of the stranded possessum subject. As we saw in (50a), this is indeed the only way for such an adverb to occur in a CP exited by PE.21

Derivations that have an overt complementizer that instead of a high adverb, such as (50b), will be identical to the above adverb scenario. If CP contains an overt complementizer, the possessor must move to its left, pied-piping the possessum with it to Spec,CP in order to maintain adjacency with [’s]. The eventual stranding of the possessum in Spec,CP will result in deletion of the complementizer due to the Doubly Filled Comp Filter, something we independently know to hold in English. Thus, as (50b) shows, an overt complementizer on either side of a subject that PE has exited is impossible.

This concludes the analysis of PE from subjects. The next section analyzes the properties of PE from nonsubject DPs, which is governed by the same principles.

5.2 PE from Nonsubjects

5.2.1 PE from Nonsubjects: The Embedded-vP Level

Recall that PE from a nonsubject DP requires that DP to be pied-piped to the edge of the local CP in English. Consequently, in contrast to more typical PE in a language like Hungarian, English PE cannot strand a nonsubject possessum in its base position.

(55) Nonsubject exited by PE must be stranded in a CP edge

  • *The person [whoj you think [John ate [_____j’s food]] is Mary].

  • The person [whoj you think [[_____j’s food]k John ate _____k] is Mary].

To begin understanding why this is so, let us first examine such derivations at the embedded-vP level. PE from any nonsubject DP will work in essentially the same way.

In (56), we see a transitive vP in which PE has exited a direct object, stranding it in situ in the complement of V. This derivation would be satisfactory for CL, since the moving possessor has reached the linear edge of this phase. However, Spell-Out of this structure will find it to be in violation, since the in-situ subject (here EA, the external argument) and V, both underlined in (56), intervene between the moved possessor and [’s].

(56)

graphic

Satisfying CL as well as [’s] requires pied-piping the object possessum to the vP edge along with the possessor’s movement, instead of stranding it. Consequently, the possessor reaches the phase edge while keeping [’s] adjacent, as in (57).

(57)

graphic

It is now clear why base position stranding of a nonsubject exited by PE is impossible. Such stranding violates the adjacency requirement of [’s] at the spell-out of vP.

At this point, the nonsubject possessum has been pied-piped into the edge of vP, but it cannot remain here, as demonstrated in (58).

(58) Nonsubject possessum cannot be stranded in Spec,vP by PE

  • *The person [whoj I think [CP they [vP [_____j’s food]k ate _____k]]] is Mary.

  • *[Which student]j did you say [CP she [vP [_____j’s book]k found _____k]]?

As the contrast in (55) and similar examples has shown, the nonsubject possessum must be pied-piped to the edge of the embedded CP. This fact is now puzzling, since the analysis so far provides no reason why the pied-piped possessum should not be able to remain in Spec,vP, where the adjacency requirement of [’s] was met. To see why the concepts defended here in fact predict that the nonsubject possessum must exit Spec,vP, we must consider the next phase of the derivation.

5.2.2 PE from Nonsubjects: The Embedded-CP Level

Upon the merger of T, the subject undergoes A-movement to Spec,TP from its external Merge site in Spec,vP. This movement carries the subject across the nonsubject possessive DP that in (57) underwent Ā-movement to an outer Spec,vP, due to pied-piping of the possessum along with the possessor’s movement. The result is the structure in (59).

(59)

graphic

Recall that CL motivates elements exiting a phase to pass through that phase’s linear edge. For this reason, in principle A-movement of the subject might in fact stop off in a higher Spec,vP, above the moved possessive DP, as in (60). However, such a derivation requires movement of the subject from one specifier of vP to another. Such a phrase-bounded specifier-to-specifier movement is not possible, as discussed in section 5.1.

(60)

graphic

Thus, we expect the only possibility to be the derivation in (59), where the subject reaches Spec,TP by non-successive-cyclically moving across the possessive DP occupying the linear edge of vP. The derivation in (60) would end up being problematic for linearization anyway: here, the subject would precede the possessor at the spell-out of vP, but the possessor will ultimately extract and move to Spec,CP, and thus precede the subject. Since in this situation the relative order of possessor and subject is not consistent across both phases, a linearization contradiction will arise. In contrast, the derivation in (59) avoids this linearization problem, since here the possessor precedes the subject within vP, just as will be the case post-PE.

As mentioned in section 4.1, CL makes a prediction about how to permit non-successive-cyclic phase exits, which do not pass through the linear edge of the phase. In these scenarios, the material crossed over by movement from a nonedge position in the phase must also move into the next phase, to a position preceding what previously crossed it. This movement ensures the coherence of the linearization information of both phases involved, as (61) shows once more.

(61)

graphic

Given this prediction, if the Ā-moved possessive DP in Spec,vP must be non-successive-cyclically crossed by A-movement of the subject as in (59), then we expect that the possessum will be unable to remain in Spec,vP. Rather, it must move to a position that precedes the subject within the next phase. This is precisely what is accomplished by continuing to pied-pipe the possessum along with successive-cyclic Ā-movement of the possessor to the edge of the embedded CP, which, as shown in (62), maintains a coherent linearization.

(62)

graphic

The present account thus correctly predicts that nonsubject possessum DPs must be pied-piped to the embedded Spec,CP when exited by PE. While in principle nonsubject possessum stranding in Spec,vP should be licit, the interaction of CL with A-movement of the subject forces additional pied-piping to CP.

After pied-piping the nonsubject possessum to the edge of the embedded CP (62), the possessor can move on freely, as we saw in (52). Nothing forces further pied-piping, as the facts have shown. The possessor extracts on into the matrix clause, with [’s] remaining stranded below in the lower CP phase.

We have just seen a successful derivation of PE from a nonsubject, but this is not the end of the story. Recall that, just as for PE from subjects, PE from nonsubjects involves a restriction on the content of the embedded CP. That is, no material may intervene between the trace of PE in the possessum DP and the embedded CP edge.

(63)CP edge restriction on PE from nonsubjects

  • *Overt complementizer on either side of pied-piped nonsubject

    Whoj did you say [CP (*that) [_____j’s cat]k (*that) John saw _____k]?

  • *Adverb to left of pied-piped nonsubject

    Whoj did you say [CP (*allegedly) [_____j’s cat]k (allegedly) John stole _____k]?

  • *Pied-piped P to left of pied-piped nonsubject

    Whoj do you think [CP (*from) [_____j’s house]k we should leave (from) _____k]?

The concepts under discussion yield a familiar explanation for this constraint. To satisfy CL, the possessor must move to a Spec,CP above any peripheral material separating it from the linear edge of CP, before exiting it. In scenarios like those in (63), this would require the possessor to move to a higher Spec,CP from within the nonsubject possessum that it previously pied-piped into CP (62). This would thus be extraction from one specifier to another within the same CP, which is independently ruled out (see section 5.1.1), in addition to the fact that such movement over the problematic peripheral material would render the possessor and [’s] illegally nonadjacent within this CP (see section 5.1.2). Since no movement within the CP edge is possible, the only option is for the possessor to non-successive-cyclically cross over any problematic peripheral material in CP when moving from the pied-piped nonsubject and into the matrix vP, thus violating CL.

5.3 Why English PE Requires Cross-Clausal Movement

Recall that English PE is impossible for Ā-movement that is not cross-clausal.

(64) PE unavailable for clause-internal movement

  • *Whok will [_____k’s article] be published next year?

  • *Whok did they criticize [_____k’s article]?

Given that nonsubject possessum DPs must be pied-piped to a CP edge prior to PE as just discussed, there is no chance for the possessor to extract from a nonsubject if the derivation contains only one clause. In this case, there is no opportunity for the possessor to break away after pied-piping the possessum to the local CP edge, since at this point the derivation ends. Thus, more than one clause is needed for PE from nonsubjects to occur.

For similar reasons, PE from a subject possessum cannot become evident in a single-clause derivation. In such a scenario, any material between the extracted possessor and [’s] that would diagnose the occurrence of that PE violates the adjacency requirement of [’s] within CP, as (65) illustrates with T-to-C movement.

(65)

  • Diagnosing clause-internal PE from subject violates adjacency

  • *[CPWhok will(C-Tj) [TP [DP _____k’s kid] tj win the contest]]?

Thus, the concepts defended in this article correctly derive the fact that English PE cannot become apparent unless movement of the possessor crosses a clause boundary. Consequently, such PE cannot be surface-evident with clause-bound movement.

5.4 A Final Prediction

This concludes the analysis of the constraints on English PE. In closing, consider one final prediction. In a configuration with multiple recursively embedded CPs, we expect PE to allow stranding of the possessum in the edge of any CP within the possessor’s movement path, provided that the constraints discussed in this section are met. This prediction is correct.

(66) PE with multiple embedding: Possessum stranding in any CP edge22

  • Whok do you think [CP ([_____k’s cat]) he said [CP ([_____k’s cat]) is cute]]?

  • Whok do you think [CP ([_____k’s cat]j) he said [CP (?[_____k’s cat]j ) they saw _____j]]?

In the next section, I discuss some general consequences of the analysis presented here, before moving on to several extensions.

6 General Consequences

6.1 In Support of CL

The CL theory is vital to the account provided here, because it includes phase edges in Spell-Out domains. This feature of CL allows successive-cyclic movement through phase edges to interact with the adjacency requirement of [’s], which is enforced at Spell-Out. I have shown that CL and this requirement together derive some otherwise puzzling facts about English PE. These facts are intricate, and the construction in which they hold does not appear to be very frequent (as perhaps indicated by the fact that it is nearly undocumented). Given this, the complex restrictions on English PE are unlikely to be a set of memorized quirks. Rather, these details should emerge from more general grammatical principles.

I have argued that these facts emerge from just two principles, one language-specific and one general. I propose that CL is an aspect of UG, automatically possessed by all speakers. Having CL intrinsically, the only thing English speakers must learn in order to determine whether their grammar bans PE, or permits it in the restricted form described here, is the point in the derivation when the adjacency requirement of [’s] may be satisfied. Thus, if correct, this analysis entails that CL is a part of UG.

An important detail captured by this account is the fact that nonsubject possessum DPs must be pied-piped as far as CP under PE. I have argued that stranding of nonsubject possessum DPs in Spec,vP should be possible in principle, though in reality it is not (58). We have seen that under CL, the crossing-over of this position by A-movement of the subject is predicted to require that it be emptied, thus forcing the possessive DP to be pied-piped further. This is an instantiation of a general prediction of CL explained in section 4.1, that if an element in a phase is crossed by something non-successive-cyclically moving from that phase, then that crossed-over material must also move, to a position above what crossed it.

(67)

graphic

In the approach to phases in Chomsky 2000, 2001, there is no reason why movement of a lower specifier across a higher one of the same phase should require movement of the higher one as well, though this is an automatic consequence of the CL approach.23 In section 8, I will examine further facts that also support this prediction of CL.

6.2 The Nonphasehood of DP

While it is often assumed that DP is a phase, the account I have given of English PE is not compatible with such a claim. I have proposed that the adjacency requirement of [’s] can be satisfied in its local phase, thus allowing the possessor to separate from [’s] provided that [’s] remains within a phase where its adjacency requirement was met at Spell-Out. If DP were a phase in and of itself, this requirement of [’s] would be immediately satisfied within DP. In this situation, there would be no requirement to ever pied-pipe under PE, which incorrectly predicts the possibility of allowing a nonsubject possessum exited by PE to remain in situ in VP.

The strongest conclusion to draw from this result is that the English DP is not a phase. This conclusion superficially agrees with Matushansky’s (2005) argument that the phasehood of DP is ambiguous. The present proposals are compatible with there being a nominal phase boundary below the possessor and [’s] (perhaps nP, if categorizing heads are phases, following Embick and Marantz (2008), among others), but not above them.24

Like the account developed here, Sabbagh’s (2007) and Zyman’s (2019) must exclude DP from the set of phases. Such research may be compatible with a theory in which the English DP is a phase for LF but not PF (entailing that DP is not an independent domain for linearization), a possibility that arises if a given phase need not necessarily transfer its contents to both interfaces. See Citko 2014 for an overview of works that make use of such nonsimultaneous transfer.

A phenomenon that may provide an independent diagnostic for DP phasehood is the stranding of adjuncts like exactly and precisely under wh-movement (e.g., Urban 1999). Such adjuncts can be stranded in their base position or in an intermediate CP edge.25

(68)

  • Exactly-stranding

  • Whatk (exactly/precisely) did you say _____k (exactly/precisely) that she wants _____k (exactly/precisely)?

If DP is a phase, and thus a constituent that movement must exit via the edge of, then such stranding should be possible in the edge of DP. However, as Zyman (2019) observes, this is not the case (69). This result is consistent with a nonphasal DP.

(69)

  • No exactly-stranding in the edge of DP

  • Whatk (exactly/precisely) did you write [DP _____k (*exactly/*precisely) a book about _____k]?

7 A PE Analysis of That’s-Relatives

In this section, I examine an additional configuration that can be analyzed as involving PE, which I will call the that’s-relative (Seppänen and Kjellmer 1995, McDaniel, McKee, and Bernstein 1998, McDaniel et al. 2002). This relativization strategy, available to many (though not all) English speakers, can prima facie be described as relativization of a possessor that is accomplished by replacing the typical relative pronoun who with that.

(70) That’s-relatives

  • This is the girl [that’s hat is red].

  • There’s the house [that’s door is blue].

  • This is the dog [that’s food the cat ate].

  • That is the bike [that’s wheel the kids broke].

With the addition of Deal’s (2016) proposal that TP is a phase in relative clauses, the concepts defended so far in this article lead to an account of this construction.

On the basis of facts about word order, complementizer choice, and case attraction in Nez Perce, Deal (2016) argues that relativization involves successive-cyclic Ā-movement to Spec,CP via an intermediate step of movement through an outer specifier of a phasal TP. She argues that this conclusion also explains anti-that-trace effects (Bresnan 1972) in English,26 in the context of Pesetsky and Torrego’s (2001) account of that-trace phenomena. While I do not discuss Deal’s analysis in detail here for space reasons, this concept sheds new light on the derivation of that’s-relatives, when the possibility of PE in English is considered.

At first glance, it is conceivable that that’s-relatives use a special instance of that which is a relative pronoun/relativizing wh-operator. The exact position occupied by that in this construction is not obvious from the surface string, but if that really can be a relative pronoun, then it should be able to take the place of the typical wh-operator when relativization involves a pied-piped P, for instance. This turns out to be impossible.

(71) That is not a productive relative pronoun

  • Mary is the student [[about whose/*that’s book] I wrote a review].

  • I saw a man [[on whose/*that’s face] there were many tattoos].

  • That is the person [[with whom/*that] I traveled].

  • This is the office [[to which/*that] I sent a package].

We expect this result if that is in fact never a relative pronoun. In the absence of such a possibility, the natural hypothesis is that the that of that’s-relatives is simply a complementizer, as usual. If this is so, it suggests that that’s-relatives must actually have a gap in possessor position, from which a relativizing wh-operator moves to Spec,CP.27

(72)

graphic

Such an analysis might be regarded as ad hoc under standard assumptions about the impossibility of PE in English, but this is not so in the present context.

In this article, I have shown that English PE can occur with all types of Ā-movement, including relativization, provided that such movement strands the possessum at the edge of an embedded CP. In section 5, I argued that this is so because CL and the adjacency requirement of [’s] together prevent PE from occurring until the possessum can be stranded in a phase edge.

As we’ve seen, independent factors prevent the possessum from occupying the vP edge; thus, it always surfaces in an embedded CP edge in the PE derivations shown so far.

(73) Typical PE in relative clauses with possessum stranding in CP edge

  • Mary is the author [whok they said [CP [_____k’s new book] is good]].

  • The person [whoj I said [CP [_____j’s food]k John ate _____k] is Mary].

If TP is a phase in relative clauses, then such clauses should allow a second possibility for PE derivations: instead of the possessum following the moving possessor until reaching a CP edge as in (73), the possessum could also be stranded in the edge of the phasal TP, below C. Such a derivation produces a that’s-relative. In the case of a that’s-relative with a gap inside a subject possessum, the possessor wh-operator simply exits the subject, which remains in Spec,TP, as already diagrammed in (72). In a that’s-relative with a gap in a nonsubject possessum, this possessum is pied-piped above the subject to an outer specifier at the edge of the TP phase, with the wh-operator subsequently extracting to Spec,CP.

(74)

graphic

Notice that in that’s-relatives, the relativizing wh-operator that moves from possessor position is null. Since an overt complementizer that is present in such configurations, we indeed expect the wh-operator to be null, given the Doubly Filled Comp Filter. If the Doubly Filled Comp Filter is resolved in the opposite way in such a structure, realizing the possessor operator but keeping the complementizer silent, then the possessor and the possessum will simply appear linearly adjacent as they do in typical relative clauses.28

(75) Alternative resolution of the Doubly Filled Comp Filter

  • the girl [who that’s hat is red] (= whose hat is red)

  • the cat [who that’s tail the kids pulled] (= whose tail the kids pulled)

To conclude this section, note that this analysis of that’s-relatives maintains the phase-bounded adjacency requirement of [’s] that I have argued generally constrains English PE (43). This requirement’s interaction with CL forces PE to strand the possessum in the linear edge of a phase crossed by the possessor’s movement. Ultimately, this is the CP edge in the cases examined earlier in this article, since the vP edge is unavailable. If TP in relative clauses is also a phase, then in this context the TP edge provides another position where the possessum can be stranded.

8 On Stranding in the vP Edge

The analysis of English PE I have argued for leads to some general predictions about how CL will govern stranding in the vP edge, which I examine in this section. See Davis to appear for a more thorough crosslinguistic consideration of stranding in phase edges.

8.1 Predicting the Distribution of Stranding in Spec,vP

McCloskey (2000) shows that in West Ulster English, wh-movement can strand the post-wh quantifier all either in its base position or at the edge of an intermediate CP.29

(76)

  • Spec,CP all-stranding

  • Whatk (all) did he say [CP _____k (all) that he wanted _____k (all)]?

  • (McCloskey 2000:61, (8))

McCloskey argues that such intermediate wh-associated all-stranding provides evidence for successive-cyclic Ā-movement through CP edges. However, he notes that such stranding in specifiers of vP is not possible.30 This is a puzzle if both vP and CP are phases. McCloskey’s analysis of West Ulster English argues that V moves to a head above vP; thus, his examples demonstrating this fact attempt stranding after V, as in (77).

(77)

  • Illicit all-stranding in Spec,vP

  • Whatk did he tellj [vP _____k (*all) _____j his friends [CP _____k (all) that he wanted k]]?

  • (McCloskey 2000:63, (14e))

The concepts defended here predict this fact, for two reasons.

First, recall that CL requires a phrase Ā-moving from vP to stop in its most linearly peripheral position, which must be a specifier above the subject in situ in a lower Spec,vP. The subject later A-moves to Spec,TP across that outer Spec,vP, presumably non-successive-cyclically as argued in section 5.2.2. There is no problem with this derivation, as long as the Ā-moved material in the outer Spec,vP moves along to Spec,CP. However, if Ā-movement were to strand all (or anything else) in that Spec,vP, movement of the subject across that stranded material is predicted by CL to cause a linearization problem.

(78)

graphic

The second reason why wh-associated all-stranding is banned in Spec,vP relates to head movement. Recall that McCloskey (2000) argues that V moves from vP in West Ulster English. If heads can only move to other heads, there is no position V can move to that precedes the specifiers of vP within this phase. Therefore, movement of V from vP will necessarily non-successive-cyclically cross the edge of vP, preventing stranding there.

(79)

graphic

This analysis predicts that stranding in Spec,vP is possible only when what is stranded is not later crossed by non-successive-cyclic movement from vP. One case in which this prediction may be borne out comes from Ko (2011). Ko argues that object scrambling in Korean can strand a numeral quantifier in Spec,vP, above an in-situ subject (80). Since the subject does not move from vP here, we indeed predict stranding in Spec,vP to be permitted. V also cannot have moved leftward across Spec,vP here, since Korean is a head-final language.

(80)

  • Stranding of numeral quantifier in Spec,vP in Korean

  • Kong-ulk amato [vP [_____ksey-kay]j haksayng-tul-i _____j patassulkesita].

  • ball-ACC probably 3-thing student-PL-NOM received

  • ‘The students probably received three balls.’

  • (Ko 2011:736, (24))

The same string is possible in Japanese, which has the same basic syntactic properties as Korean.

(81)

  • Stranding of numeral quantifier in Spec,vP in Japanese

  • Ringo-ok osoraku/tabun [vP [_____ksan-ko]j John-ga umaku _____j nusun-da].

  • apple-ACC probably 3-thing John-NOM well steal-PST

  • ‘John probably skillfully stole 3 apples.’

  • (Takashi Morita, pers. comm.)

Convergent evidence for the concepts under discussion comes from constraints on the stranding of adjuncts like exactly and precisely under wh-movement. Such adjuncts can be stranded in their base position or in a CP edge, as shown in (68). The present account of English PE predicts that A-movement of the subject will prevent such adjuncts from being stranded in Spec, vP. This prediction appears accurate: at best, examples like (82) have an odd reading construing the stranded element as an adverb of vP/VP, rather than as a modifier of DP.

(82)

  • *Exactly-stranding in Spec,vP

  • [CP Whatk did you [vP _____k (*exactly/*precisely) eat _____k]]?

A similar stranding pattern can be found with other DP adjuncts of quantity/degree, like to the nearest pound, which can be stranded in its base position or at a CP edge.

(83)

  • Quantity adjunct stranding

  • Tell me [CP [how much flour]k (to the nearest pound) you said [CP _____k (to the nearest pound) that the bakery wants ____k (to the nearest pound)]].

This adjunct cannot be stranded in Spec,vP, as (84) shows. Since this adjunct cannot be construed as a modifier of vP/VP, this judgment is clearer than that of (82).

(84)

  • *Quantity adjunct stranding in Spec,vP

  • [How much flour]k (to the nearest pound) did the bakery [vP _____k (*to the nearest pound) ask for _____k (to the nearest pound)]?

Example (84) shows this fact in a transitive sentence, but the same restriction holds in passive (85) and unaccusative (86) derivations.

(85)

  • *Quantity adjunct stranding in Spec,vP: Passive31

  • [How many boats]k (to the nearest hundred) has the Micronesian navy [vP _____k (*to the nearest hundred) been provided with _____k (to the nearest hundred)]?

(86)

  • *Quantity adjunct stranding in Spec,vP: Unaccusative

  • [How many firefighters]k (to the nearest dozen) did the house [vP _____k (*to the nearest dozen) burn down despite the efforts of _____k (to the nearest dozen)]?

This is expected, if CL requires internal-argument subjects to pass through the vP edge to precede V (see section 5.1.1). From that position, A-movement to Spec,TP crosses anything stranded in the periphery of vP by Ā-movement, causing a linearization problem.

8.2 Diagnosing the Origin of Expletive There

As demonstrated in section 2, expletive associates exited by PE must be pied-piped to the edge of the local CP, just as all nonsubject DPs must.

(87)

graphic

This possessum stranding can be used as a diagnostic for the derivational history of the expletive. Several works argue that expletive there is externally merged in Spec,vP (e.g., Biberauer and Richards 2005, Deal 2009) before A-moving to Spec,TP. If this is so, we expect A-movement of the expletive to result in a crossing effect that makes it impossible for PE to strand the expletive associate in Spec,vP.

Example (87b) is ambiguous between stranding in situ or in Spec,vP, since copular V in English moves to T, unless T is filled by an auxiliary. For this reason, (88) adds an auxiliary in order to keep V low and disambiguate the position of stranding. Here, we see that the expletive associate exited by PE cannot remain in a position corresponding to Spec,vP, as expected if the expletive moved from vP.

(88)

  • Expletive associate must be stranded in Spec,CP (not Spec,vP) under PE

  • Whok do you think [CP (? _____k’s friends) there have [vP (* _____k’s friends) been a lot of stories told to (* _____k’s friends)]]?

This conclusion is supported by the fact that adjunct stranding in Spec,vP under wh-movement, as discussed in section 8.1, is also not possible in expletive constructions.

(89) No adjunct stranding in Spec,vP in expletive constructions

  • [How many kids]k have there [vP _____k (*exactly) been _____k in the office today]?

  • [How many kilos of gold]k (to the nearest hundred) have there [vP _____k (*to the nearest hundred) been _____k consumed in the production of fancy pens]?

These facts only stand as evidence for A-movement of the expletive if vP is a phase in expletive constructions. If it is not, there would be no reason to expect successive-cyclic movement through, or stranding in, this phrase’s edge. Thus, evidence for vP phasehood in this environment is necessary. Nissenbaum (2000) argues that parasitic gaps in clausal adjuncts are licensed by successive-cyclic movement through Spec,vP. If such a parasitic gap can be licensed in a given environment, we can infer that successive-cyclic movement through Spec,vP occurred. Legate (2003) uses this logic to diagnose the phasehood of verb phrases in several contexts, but she does not perform this test in expletive constructions. When we do, I argue that we find successful parasitic gap licensing.

(90) Parasitic gaps licensed by Ā-movement in expletive constructions

  • Whok was there a big rumor about _____k [after the police arrested PG]?

  • Whok was there a party for _____k [before the boss promoted PG]?

To the extent that this constitutes evidence for a vP phase in expletive constructions, the facts in (88)–(89) indicate movement of the expletive from vP.

9 Final Considerations for the Analysis of English PE

Before concluding the article, in this section I discuss two additional considerations for the examination of English PE. First, I further defend this article’s analysis of nonsubject displacement under PE as derived by pied-piping along with successive-cyclic Ā-movement of the possessor, by showing that such displacement is not the result of independent topic/focus fronting prior to extraction. Second, I offer some final comments on the requirement of local adjacency between the possessor and [’s] that this article has argued for.

9.1 Against a Topic/Focus Movement Analysis of Possessum Displacement

In section 2, following Gavruseva and Thornton (2001), I adopted the hypothesis that the displacement of nonsubjects to the embedded CP edge under PE is caused by pied-piping along with Ā-movement of the possessor. Alternatively, we might hypothesize that this displacement is caused by the nonsubject possessum undergoing a step of topic/focus movement into the left periphery of the embedded CP prior to PE, since such movement is independently possible for many English speakers. Since it is only vital to the analysis proposed here that the possessum reaches the embedded CP edge, either of these accounts could achieve correct results. Thus, this choice does not affect the article’s central findings. However, there are reasons to believe that a topic/focus movement analysis is not correct.

For speakers with PE who independently accept topic/focus movement, the semantic/pragmatic effect of such movement is absent in PE examples. For instance, consider (91a), where the fronted phrase John’s cake is associated with a particular information structure: either setting up John’s cake as the topic of the clause, or focusing on John’s cake in contrast to other alternatives. Speakers who accept (91a) and allow PE report that such information structure is absent from PE sentences with a displaced nonsubject, like (91b). Rather, such PE examples have a neutral information structure (modulo questionhood), just like their full pied-piping equivalents do (91c).32 Furthermore, some speakers I have interviewed do not readily accept topic/focus movement in the spoken language (and thus do not accept examples like (91a)) but nevertheless accept PE examples like (91b). For such speakers, it is especially clear that PE and topic/focus movement are unrelated.

(91)

  • Topic/Focus Movement

    [John’s cake]k, you said they really liked _____k.

  • PE from nonsubject

    Whoj did you say [[_____j’s cake]k they really liked _____k]?

  • Pied-piping equivalent of (91b)

    [Whose cake]k did you say [they really liked _____k]?

If the topic/focus movement analysis is incorrect, the relevant sentences with PE from a nonsubject should be accepted even with matrix verbs that ban embedded topicalization. Hooper and Thompson (1973) show that nonfactive/nonassertive verbs like doubt and deny ban embedded topicalization (among other “root transformations” like VP-fronting). In fact, such matrix verbs degrade PE, as (92) shows by comparing a deviant sentence with PE from a nonsubject (92a) with its pied-piping equivalent (92b).

(92) PE is blocked by nonfactive verbs

  • *Whok do you deny/doubt [[_____k’s cat]j they abused _____j ]?

  • [Whose cat]k do you deny/doubt [they abused _____k]?

However, if the unavailability of embedded topic/focus movement were the problem in (92a), then such an example should be permitted when the matrix verb is factive, like realize, know, or see—embedding verbs that Hooper and Thompson show allow embedded topicalization. Importantly, PE with such factive embedding verbs also fails.

(93) PE is blocked by factive verbs

  • *Whok did you know/realize/see [[_____k’s cat]j they abused _____j ]?

  • [Whose cat]k did you know/realize/see [they abused _____k]?

Both nonassertive verbs like doubt and deny in (92) and the factive verbs in (93) are classified as weak island inducers (see Szabolcsi and Lohndal 2017 for a recent overview). This set also includes wh-islands caused by whether, which interfere with PE as well.

(94) PE is sensitive to the whether island

  • *Whoi do you wonder [whether we said [[ i’s friend]k we like _____k]]?

  • [Whose friend]k do you wonder [whether we said we like _____k]?

NPI (negative polarity item) licensers like few/only also induce weak islands and degrade PE ((95a) vs. (95b)). A similar effect occurs with basic negation ((95c) vs. (95d)). While the violations in (95) are not absolute, these contrasts are evident for many speakers.33

(95) PE is degraded by NPI licensers

  • [Whose art]k have few people/only they said [_____k is interesting]?

  • *??%Whok have few people/only they said [[_____k’s art] is interesting]?

  • That’s the author [[whose work]k I didn’t say [_____k is any good]].

  • ?%That’s the author [whok I didn’t say [[_____k’s work] is any good]].

Overall, the hypothesis that PE (from nonsubjects) depends on topic/focus movement fails to predict the actual set of circumstances where English PE is degraded, which fits the set of weak islands. English PE is in essence an instance of left-branch extraction (Ross 1967), a type of extraction that is known to be sensitive to weak islands in several other Western European languages. For instance, combien-split in French (Obenauer 1984/1985) and wat voor/wat aan-split in Dutch (Corver 1990, Beermann 1997, Honcoop 1998) are subject to the same constraint. Thus, the weak island sensitivity of English PE falls within the purview of known constraints on extraction of this variety.

9.2 On the Adjacency Requirement of [’s]

Following works proposing that PE in mainstream English is banned because the possessive D [’s] must be adjacent to the possessor at PF, I hypothesized that the constrained nature of English PE emerges when speakers allow such a requirement to be enforced within a structurally smaller domain. Namely, I hypothesized that while this constraint is globally enforced in English grammars without PE (42), those speakers with PE can satisfy this constraint at the minimal phase level (43).

If such adjacency constraints can generally be locally evaluated in this way, we expect to see phenomena in other languages whose distribution is analogous to that of English PE: that is, we expect to find circumstances where something like a bound morpheme can only be stranded after some movement. At the time of writing, I have not found additional facts that obviously fit this description. If such patterns are not clearly attested, it may be that the English pattern studied here is atypical and therefore perhaps unstable. If this is so, we may expect subsequent generations to rearrange, or fail to acquire, the patterns reported here.

Provided that the arguments presented here for the adjacency requirement of [’s] are successful, understanding it more deeply should indeed be a topic for further research. My hope is that the insights of this article (the impact of CL being the main one) will be preserved independently of what concepts are ultimately brought to bear on this requirement’s explanation.34

10 Conclusion

In this article, I have described and analyzed the complexities of PE in English, a little-studied possibility for some speakers. I have argued that English PE provides evidence for CL, which constrains English PE via its interaction with a phase-level version of an independently proposed PF adjacency requirement of [’s]. Additionally, I have argued that this analysis entails the non-phasehood of DP, sheds light on the derivation of that’s-relatives (and the phasehood of TP in relative clauses), reveals some linearization constraints on stranding, and suggests that expletive there originates in vP.

Notes

1 My consultants are mostly American, though my set of speakers who accept PE includes one Australian, one British, and two Canadian individuals. There is no clear generalization about the age/origin/background of speakers who accept PE.

The sentences reported here are the aggregate of notes taken from interviews during 2016–2018. Since these PE sentences are most suited to colloquial language, I typically interviewed individuals verbally face-to-face. Generally, I constructed some sample PE sentences and asked that they be judged, though individuals capable of the phenomenon often readily volunteered more.

2 English PE can be found in informal writing, however, as the following examples retrieved from the Internet show:

3 In addition to the Finno-Ugric Hungarian, PE is attested in Chamorro (Austronesian; Chung 1991), the Mayan languages Tzotzil (Aissen 1996) and Chol (Coon 2009), as well as much of the Slavic family (e.g., Ross 1967, Bošković 2005). Romance and Germanic languages permit some extraction of postnominal/PP possessors (Gavruseva 2000).

4 The closest thing to a mention of the construction in other work that I am aware of (thanks to a reviewer’s comment) comes from Heck (2009), who in footnote 64 credits Andrew McIntyre for the observation of a person who I think’s reputation would be better if they stopped dribbling incessantly. Heck assumes that this phrase is formed by placing a parenthetical in DP, an analysis that I argue against in section 3.

5 A very small percentage of their data consists of other unusual extraction configurations (e.g., movement of whose), which I take to be genuine errors.

6 The possibility of examples like (10c) is interesting in light of the fact that which cannot occur in possessor position in nonextraction circumstances (*This is the shopwhich’sprices are lowest). Whatever the source of this unacceptability is, it is apparently avoided by movement. While there is preliminary evidence that the same amelioration by extraction is possible for other typically unacceptable wh-possessors (*where’s/*what’s/*when’s), the study of this effect must wait for future work.

7 I use the term nonsubject to refer to all DPs whose base position prior to Ā-movement is not Spec,TP, but a lower, postverbal position.

8 Though Ā-movement of indirect objects is independently ruled out for some speakers.

9 A reviewer notes that (15c) may be degraded because the nature of a phrase’s possessor influences its ability to be an expletive associate: who is not independently a licit there-associate (i), and a phrase containing who may be dispreferred as a there-associate for this reason. As far as I know, this is true of wh-phrases generally, meaning that examples like (15c) cannot be formed without some degradation.

  • Which teacher said there is a student/*who in the room?

10 Specifically, section 9.1 argues that this displacement of nonsubjects is not caused by something like topic/focus movement into the left periphery of the embedded clause prior to PE, which is the most plausible alternative analysis that I am aware of.

11 At least in mainstream English, since (as a reviewer points out) there are other English varieties that allow configurations that would incur a that-trace violation in standard English (Sobin 1987).

12 This fact is compatible with accounts of the that-trace effect as a linear filter on an overt C that adjacent to a trace (Bresnan 1972, Chomsky and Lasnik 1977). I do not aim to say anything about complementizer-trace effects here, since the account I present predicts (22a–b) automatically.

13 PE in ditransitives can also yield strings where [’s] is adjacent to a pronoun. A lexical noun in the same position is less acceptable, but I suspect that this is attributable to a garden path effect.

  • Whok did they tell me/you/??your friend [[_____k’s name] is Bill?

14 Though it ought to be counted, as it is required for a parenthetical in a matrix question.

  • Whose book, (do you think/*you think), did Mary buy?

15 These PE examples with a null relative operator behave the same as PE examples with an overt operator, including, as we see in (32b–c), the property of pied-piping nonsubject possessum DPs to the local CP edge. The analysis to come in section 5 attributes this pied-piping to the interaction of CL and a requirement of adjacency between [’s] and the possessor, but such adjacency is not obviously relevant when, as in these examples, the possessor is covert. I hypothesize that the possessor here was originally overt as usual, before being removed from the surface string via the mechanism responsible for free deletion in Comp.

16 Though the postsubtraction strings in (32) are acceptable on an irrelevant structurally distinct parse.

17 Given that some English speakers can extract possessors via Ā-movement, it would also be reasonable to expect such speakers to permit possessor raising via A-movement, though this is not in fact possible (i). I hypothesize that because English possessors are always Case-licensed DP-internally, A-movement of possessors from DP is not an option, since English lacks phenomena like hyperraising.

  • *Johnk was washed [_____k’s hands].

18 I define adjacency as a relation between two elements α and β, whereby α. and β form a linear string with nothing intervening between them. This notion of adjacency is not a primitive of CL, which is concerned only with relative order, for which intervening material is irrelevant. I suggest that adjacency of this sort is enforced by the PF requirements of certain elements, intuitively what are termed “bound morphemes.”

19 A reviewer notes that one might defend Chomsky’s (2000, 2001) phase theory here by saying that if X in (44) is null (like a null C), the adjacency requirement would be satisfied upon the spell-out of the next phase. The reviewer points out, however, that this idea could only apply where PE is string-vacuous, and since non-string-vacuous PE does occur, such an approach does not achieve the right results.

20 Furthermore, the derivation described for (49) may represent an instance of chain interleaving, which, following Collins (1994), is independently ruled out. If this is so, then on these grounds as well, the derivation in (48) is superior.

21 This result could also have been achieved by adjoining the adverb linearly to the right rather than to the left of the subject, but we see here that even if the adverb originates to the left of the subject, the derivation can converge.

22 As (66b) shows, PE from a nonsubject that strands the possessum in the lower Spec,CP is somewhat degraded. This is a puzzle that I will leave aside for now.

23Bošković (2016) offers a modification of phase theory in which, modulo traces, it is only possible to access the outermost constituent of a phase that has multiple specifiers. This approach makes similar predictions to CL for movement from multispecifier phases, but because it does not obviously capture the fact that the distribution of English PE is fundamentally constrained by concerns of linear order, I do not explore it here.

24 This hypothesis is evocative of Chomsky’s (2000, 2001) approach to phases, in which the edge is not subject to Spell-Out. However, as discussed in section 4.3, this theory is generally incompatible with the results reached here.

25Zyman (2019) argues that adjuncts like exactly and precisely actually cannot be stranded VP-internally, despite appearances, though this does not affect the present analysis.

26 The anti-that-trace effect describes the necessity of the complementizer in subject relatives (The woman *(that) likes cats). This effect is also present in it-clefts (It’s Mary *(that) likes cats). While uniting cleft clauses with relative clauses is a debatable issue, under Deal’s (2016) analysis this fact suggests a TP phase in cleft clauses also. The presence of a TP phase in cleft clauses accurately predicts the availability of a that’s-relative-like structure in clefts (It’s Johnthat’sfamily loves dogs) under the analysis proposed in this section.

27McDaniel et al. (2002) consider several possible analyses of this construction and conclude that modern English is converging on an analysis involving complementizer that plus [’s], precisely as I argue in this section.

28 It is possible to have a nonsubject relative clause where C and the relativizing wh-phrase in its specifier are both silent (the cake [which thatI ate]), but this possibility does not seem to extend to that’s-relatives (*the person [whothat’s friend I like]). I must leave this puzzle aside for the time being.

29 A reviewer notes that if such stranding is really in Spec,CP, the fact that an overt C can cooccur with it is surprising, given something like the Doubly Filled Comp Filter. It is conceivable that the Doubly Filled Comp Filter only applies when a full phrase is sitting in Spec,CP, rather than something like a remnant quantifier. The same might be said of examples like (68). A different reviewer points out that the Doubly Filled Comp Filter fails to apply in the matrix CP of a wh-question, where C is filled by an auxiliary (WhatwillJohn buy?). Evidently, the Doubly Filled Comp Filter is only applicable to the complementizer that, and not in all situations. These facts indeed raise intriguing questions about the nature of the Doubly Filled Comp Filter, but are tangential to the analysis of the PE facts that this article focuses on.

30Henry (2012) shows that there is more variance in West Ulster English all-stranding than McCloskey (2000) reports. I will leave varieties other than the one studied by McCloskey aside for now.

31 I assume that the edge of the clause-internal phase in passives is to the left of the passive auxiliary be, which, following Harwood (2015), is merged in v.

32 The prosody typical of topic/focus movement is also absent from examples like (91b).

33 Examples (95b,d) are best with focus on the intervener. Focus, among other semantic and pragmatic factors, is known to circumvent the (weak) islandhood of negation and related elements (e.g., Kroch 1989, Kuno and Takami 1997).

34 Adjacency between [’s] and the possessor it selects can be motivated by Richards’s (2016) Contiguity Theory, which hypothesizes that syntactic dependencies like agreement and selection require linear adjacency of the elements involved in many cases, due to prosodic requirements. Importantly, Richards’s theory allows such adjacency to be satisfied derivationally and later broken. While this concept is reminiscent of the present conclusion that speakers with PE permit extraction of the possessor after earlier satisfaction of adjacency, it does not explain why such eventual breaking of adjacency is available only to some speakers.

It is possible that grammars differ in how the relevant adjacency requirement is prioritized. This hypothesis can be implemented using the alignment constraints of Optimality Theory phonology (e.g., McCarthy and Prince 1993). In Optimality Theory, all constraints are in principle violable, depending on how they are ranked relative to each other. For speakers without PE, the hypothesis would thus be that adjacency of the possessor and [’s] is enforced by a highly ranked constraint, which is never violated. In contrast, speakers with PE possess a colloquial grammar in which this adjacency constraint is violable, though not utterly, since it must be satisfied for a minimal part of the derivation. Allowing adjacency to be broken only after some minimal satisfaction is also evocative of Richards’s (1998) Principle of Minimal Compliance. However, if this adjacency requirement is in principle violable depending on language-internal constraint ranking, we predict a third sort of English grammar: one in which this adjacency constraint is prioritized so low that it asserts no influence. There is preliminary evidence for such a variety. In a highly colloquial register, an anonymous reviewer tolerates PE with no pied-piping, stranding an object in situ.

    • ?So tell me, [which girl]/whok didja/’dja meet [_____k’s friend] yesterday?

    • b. ?So tell me, [which girl]/whok d’you think he met [_____k’s friend] yesterday?

Thanks to Adam Albright, Jonathan Bobaljik, Tanya Bondarenko, Željko Bošković, Kenyon Branan, Noam Chomsky, Justin Colley, Stephen Crain, Michel DeGraff, Danny Fox, Edward Flemming, Peter Grishin, Martin Hackl, Heidi Harley, Sabine Iatridou, Roni Katzir, Loes Koring, Nick Longenbaugh, Travis Major, Dana McDaniel, Andrew McInnerney, Takashi Morita, Patrick Niedzielski, David Pesetsky, Norvin Richards, Martin Salzmann, Michelle Sheehan, Juliet Stanton, Abdul-Razak Sulemana, Gary Thoms, Michelle Yuan, Stanislao Zompì, Erik Zyman, and one especially dedicated reviewer, as well as audiences at MIT, NELS 49, and the 93rd annual LSA meeting. This project would not exist without Loes Koring and George Oscar Bluth II.

References

Aissen,
Judith
.
1996
.
Pied-piping, abstract agreement, and functional projections in Tzotzil
.
Natural Language and Linguistic Theory
14
:
447
491
.
Beermann,
Dorothee A
.
1997
.
Syntactic discontinuity and predicate formation
.
Doctoral dissertation, Tilburg University
.
Biberauer,
Theresa
, and
Marc
Richards
.
2005
. Explaining Expl. In
The function of function words and functional categories
, ed. by
Marcel
den Dikken
and
Christina
Tortora
,
115
154
.
Amsterdam
:
John Benjamins
.
Bošković,
Željko
.
2005
.
On the locality of left branch extraction and the structure of NP
.
Studia Linguistica
59
:
1
45
.
Bošković,
Željko
.
2016
.
Getting really edgy: On the edge of the edge
.
Linguistic Inquiry
47
:
1
33
.
Bresnan,
Joan
.
1972
.
Theory of complementation in English syntax
.
Doctoral dissertation, MIT, Cambridge, MA
.
Chomsky,
Noam
.
1995
.
The Minimalist Program
.
Cambridge, MA
:
MIT Press
.
Chomsky,
Noam
.
2000
. Minimalist inquiries. In
Step by step: Essays on Minimalist syntax in honor of Howard Lasnik
, ed. by
Roger
Martin
,
David
Michaels
, and
Juan
Uriagereka
,
89
115
.
Cambridge, MA
:
MIT Press
.
Chomsky,
Noam
.
2001
. Derivation by phase. In
Ken Hale: A life in language
, ed. by
Michael
Kenstowicz
,
1
52
.
Cambridge, MA
:
MIT Press
.
Chomsky,
Noam
, and
Howard
Lasnik
,
1977
.
Filters and control
.
Linguistic Inquiry
8
:
425
504
.
Chung,
Sandra
.
1991
.
Functional heads and proper government in Chamorro
.
Lingua
85
:
85
134
.
Citko,
Barbara
.
2014
.
Phase theory: An introduction
.
Cambridge
:
Cambridge University Press
.
Collins,
Chris
.
1994
.
Economy of derivation and the Generalized Proper Binding Condition
.
Linguistic Inquiry
25
:
45
61
.
Coon,
Jessica
.
2009
.
Interrogative possessors and the problem with pied-piping in Chol
.
Linguistic Inquiry
40
:
165
175
.
Corver,
Norbert
.
1990
.
The syntax of left branch extractions
.
Doctoral dissertation, Tilburg University
.
Corver,
Norbert
.
1992
. Left branch extraction. In
NELS 22
, ed. by
Kimberley
Broderick
,
67
84
.
Amherst
:
University of Massachusetts, Graduate Linguistic Student Association
.
Davis,
Colin
.
To appear
.
Crossing and stranding at edges
.
Glossa
.
Deal,
Amy Rose
.
2009
.
The origin and content of expletives: Evidence from ‘selection’
.
Syntax
12
:
285
323
.
Deal,
Amy Rose
.
2016
.
Cyclicity and connectivity in Nez Perce relative clauses
.
Linguistic Inquiry
47
:
427
470
.
Embick,
David
, and
Alec
Marantz
.
2008
.
Architecture and blocking
.
Linguistic Inquiry
39
:
1
53
.
Emonds,
Joseph E
.
1976
.
A transformational approach to English syntax: Root, structure preserving, and local transformations
.
New York
:
Academic Press
.
Erlewine,
Michael Yoshitaka
.
2017
. Why the null complementizer is special in complementizer-trace effects. In
A pesky set: Papers for David Pesetsky
, ed. by
Claire
Halpert
,
Hadas
Kotek
, and
Coppe
van Urk
,
232
288
.
MIT Working Papers in Linguistics 80
.
Cambridge, MA
:
MIT, MIT Working Papers in Linguistics
.
Fanselow,
Gisbert
, and
Denisa
Lenertová
.
2011
.
Left peripheral focus: Mismatches between syntax and information structure
.
Natural Language and Linguistic Theory
29
:
169
209
.
Fox,
Danny
, and
David
Pesetsky
.
2005a
.
Cyclic linearization of syntactic structure
.
Theoretical Linguistics
31
:
1
45
.
Fox,
Danny
, and
David
Pesetsky
.
2005b
.
Cyclic Linearization and its interaction with other aspects of grammar: A reply
.
Theoretical Linguistics
31
:
235
262
.
Gavruseva,
Elena
.
2000
.
On the syntax of possessor extraction
.
Lingua
110
:
743
772
.
Gavruseva,
Elena
, and
Rosalind
Thornton
.
2001
.
Getting it right: Acquisition of whose-questions in child English
.
Language Acquisition
9
:
229
267
.
Harwood,
William
.
2015
.
Being progressive is just a phase: Celebrating the uniqueness of progressive aspect under a phase-based analysis
.
Natural Language and Linguistic Theory
33
:
523
573
.
Heck,
Fabian
.
2009
.
On certain properties of pied-piping
.
Linguistic Inquiry
40
:
75
111
.
Henry,
Allison
.
2012
.
Phase edges, quantifier float, and the nature of (micro-)variation
.
Iberia
4
:
23
39
.
Honcoop,
Martinus F
.
1998
.
Dynamic excursions on weak islands
.
Doctoral dissertation, Leiden University
.
Hooper,
Joan B.
, and
Sandra A.
Thompson
.
1973
.
On the applicability of root transformations
.
Linguistic Inquiry
4
:
465
497
.
Ko,
Heejeong
.
2007
.
Asymmetries in scrambling and cyclic linearization
.
Linguistic Inquiry
38
:
49
83
.
Ko,
Heejeong
.
2011
.
Predication and edge effects
.
Natural Language and Linguistic Theory
29
:
725
778
.
Ko,
Heejeong
.
2014
.
Edges in syntax: Scrambling and cyclic linearization
.
Oxford
:
Oxford University Press
.
Kratzer,
Angelika
.
1996
. Severing the external argument from its verb. In
Phrase structure and the lexicon
, ed. by
Johan
Rooryck
and
Laurie
Zaring
,
109
137
.
Dordrecht
:
Kluwer
.
Kroch,
Anthony
.
1989
.
Amount quantification, referentiality and long wh-movement
.
Ms., University of Pennsylvania, Philadelphia
.
Kuno,
Susumu
, and
Ken-ichi
Takami
.
1997
.
Remarks on negative islands
.
Linguistic Inquiry
28
:
553
576
.
Larson,
Richard
.
1988
.
On the double object construction
.
Linguistic Inquiry
19
:
335
391
.
Legate,
Julie Anne
.
2003
.
Some interface properties of the phase
.
Linguistic Inquiry
34
:
506
516
.
Matushansky,
Ora
.
2005
. Going through a phase. In
Perspectives on phases
, ed. by
Martha
McGinnis
and
Norvin
Richards
,
157
181
.
MIT Working Papers in Linguistics 49
.
Cambridge, MA
:
MIT, MIT Working Papers in Linguistics
.
McCarthy,
John J
., and
Alan S.
Prince
.
1993
. Generalized alignment. In
Yearbook of morphology
, ed. by
Geert
Booij
and
Jaap
van Marle
,
79
153
.
Dordrecht
:
Kluwer
.
McCloskey,
James
.
2000
.
Quantifier float and wh-movement in an Irish English
.
Linguistic Inquiry
31
:
57
84
.
McDaniel,
Dana
,
Cecile
McKee
, and
Judy B.
Bernstein
.
1998
.
How children’s relatives solve a problem for Minimalism
.
Language
74
:
308
334
.
McDaniel,
Dana
,
Helen E.
Stickney
,
Sadie
Fowler
, and
Cecile
McKee
.
2002
.
What’s thats?
English Studies
83
:
53
69
.
Nissenbaum,
Jon
.
2000
.
Investigations of covert phrase movement
.
Doctoral dissertation, MIT, Cambridge, MA
.
Obenauer,
Hans-Georg
.
1984/1985
.
On the identification of empty categories
.
The Linguistic Review
4
:
153
202
.
Pesetsky,
David
, and
Esther
Torrego
.
2001
. T-to-C movement: Causes and consequences. In
Ken Hale: A life in language
, ed. by
Michael
Kenstowicz
,
335
426
.
Cambridge, MA
:
MIT Press
.
Podobryaev,
Sasha
.
2007
. ‘Postposition stranding’ and related phenomena in Russian. In
Studies in formal Slavic phonology, morphology, syntax, semantics and information structure
, ed. by
Uwe
Junghanns
,
Denisa
Lenertová
, and
Peter
Biskup
,
197
208
.
Frankfurt am Main
:
Peter Lang
.
Radford,
Andrew
.
1997
.
Syntax: A Minimalist introduction
.
Cambridge
:
Cambridge University Press
.
Richards,
Norvin
.
1997
.
What moves where when in which language?
Doctoral dissertation, MIT, Cambridge, MA
.
Richards,
Norvin
.
1998
.
The Principle of Minimal Compliance
.
Linguistic Inquiry
29
:
599
629
.
Richards,
Norvin
.
1999
. Featural cyclicity and ordering of multiple specifiers. In
Working Minimalism
, ed. by
Samuel David
Epstein
and
Norbert
Hornstein
,
127
158
.
Cambridge, MA
:
MIT Press
.
Richards,
Norvin
.
2016
.
Contiguity Theory
.
Cambridge, MA
:
MIT Press
.
Ross,
John
.
1967
.
Constraints on variables in syntax
.
Doctoral dissertation, MIT, Cambridge, MA
.
Sabbagh,
Joseph
.
2007
.
Ordering and linearizing rightward movement
.
Natural Language and Linguistic Theory
25
:
349
401
.
Sauerland,
Uli
.
2003
.
Intermediate adjunction with A-movement
.
Linguistic Inquiry
34
:
308
312
.
Seppӓnen,
Aimo
, and
Göran
Kjellmer
.
1995
.
The dog that’s leg was run over: On the genitive of the relative pronoun
.
English Studies
4
:
389
400
.
Sobin,
Nicholas
.
1987
.
The variable status of comp-trace phenomena
.
Natural Language and Linguistic Theory
5
:
33
60
.
Szabolcsi,
Anna
.
1984
.
The possessor that ran away from home
.
The Linguistic Review
3
:
89
102
.
Szabolcsi,
Anna
, and
Terje
Lohndal
.
2017
. Strong vs. weak islands. In
The Wiley Blackwell companion to syntax
, ed. by
Martin
Everaert
and
Henk
van Riemsdijk
. 2nd ed.
Hoboken, NJ
:
John Wiley
.
Torrence,
Harold
.
2019
.
Left edge agreeing elements in Wolof wh-questions
.
Ms., UCLA, Los Angeles, CA
.
Urban,
Emily
.
1999
.
Exactly stranding
.
Ms., University of California, Santa Cruz
.
Wiland,
Bartosz
.
2010
.
Overt evidence from left-branch extraction in Polish for punctuated paths
.
Linguistic Inquiry
41
:
335
347
.
Zwicky,
Arnold M
.
1971
.
In a manner of speaking
.
Linguistic Inquiry
2
:
223
233
.
Zwicky,
Arnold M
.
1987
.
Suppressing the Zs
.
Journal of Linguistics
23
:
133
148
.
Zyman,
Erik
.
2019
.
Phase constrained obligatory late adjunction
.
Ms., University of Chicago, Chicago, IL
.