1 Introduction

One of the core insights of modern syntactic theory is that the formation of syntactic dependencies is subject to intervention-based locality (also called “minimality”), informally defined as follows.

    1. (1)
    1. In the structure
    2. X … [ … Z … Y … ]
    3. a dependency between the syntactic objects X and Y cannot be established across an intervening (or “closer”) element Z where Z could have formed an equivalent dependency with X.
    4. (Modeled after Rizzi 2013: 172, (9))

Such intervention effects are instantiated by, among other things, superiority effects in multiple wh questions (on which see Kuno & Robinson 1972 and Chomsky 1973). For instance, an object bearing a [wh] feature cannot undergo Ā-movement over an intervening subject bearing a [wh] feature:

    1. (2)
    1. Wh subjects intervene for Ā-movement of wh objects
    1.  
    1. a.
    1.   Whoi did Joni claim ____i brought what?
    1.  
    1. b.
    1. *Whati did Joni claim who brought ____i?

We can account for this asymmetry by proposing that the interrogative C in wh questions requires its specifier to be filled by a constituent bearing [wh]. Adopting the “∙” notation for structure-building features from Heck & Müller 2007 and subsequent works, we can say that C bears a [∙wh∙] feature. The unacceptability of object wh movement in (2b), then, is attributable to the fact that C[∙wh∙] has entered into a dependency (in this case, an Ā-movement dependency) with an object bearing [wh] across an intervening subject bearing [wh], in violation of (1).

This raises the question as to what configurations give rise to intervention; that is, in what sense Z is closer than Y to X in (1). Closeness is commonly defined in terms of asymmetric c-command:1

    1. (3)
    1. “Closeness” as closest asymmetric c-command
    2. Given three syntactic objects X, Y, and Z, Z is closer to X than Y is iff:
    1.  
    1. a.
    1. X asymmetrically c-commands Z and
    1.  
    1. b.
    1. Z asymmetrically c-commands Y.

According to this definition of closeness, intervention effects are expected to be uniformly structure-dependent and never sensitive to linear order/precedence. In (4) and (5), the head X[∙F∙] will only attract Z[F], all else being equal. Structure-building features are crossed out once satisfied (or “checked”) by (Internal) Merge.2

    1. (4)
    1. (5)

The claim that closeness is structure-dependent and (in at least some cases)3 defined in terms of closest asymmetric c-command, as in (3), can be found in many proposed locality constraints, including (Feature-Based) Relativized Minimality (Fanselow 1990, Rizzi 1990, Fanselow 1991, Ferguson 1993, Ferguson & Groat 1994, Ferguson 1996, and Rizzi 2013), the Minimal Link Condition (Chomsky 1995b: chap. 4, Kitahara 1997), and Attract Closest (Pesetsky 2000: 15, (30)). For other similar principles, see Fiengo 1980: 130, (52), Chomsky 2000: 122, (40), Collins 2002: 51, (18), and Branan & Erlewine 2024: 2, (2). The claim can also be found in the path-based accounts of Hornstein 2009: 35–39 and Müller 2011: 20–21 (esp. (25) and fn. 9) and references cited in those works.

There is, however, another measure of closeness by which intervention could be calculated: in (2b), the intervener Z not only asymmetrically c-commands Y; it also linearly precedes Y. Contrary to conventional wisdom (see, e.g., Chomsky 2013: 39, Rizzi 2013: 173, and Chomsky et al. 2019: 233–235), I argue in this article that linear precedence plays a crucial role in determining intervention effects under syntactic movement. Intervention is not calculated solely over asymmetric c-command (or a similar metric for determining structural prominence).

In support of my claim, I report evidence for the following novel generalization.

    1. (6)
    1. Order dependence in English passivization
    2. When just one constituent can undergo passive A-movement in English, it is invariably the leftmost accessible passivizable constituent within vP.

Example (7) illustrates with a vP containing two PP internal arguments.

    1. (7)
    1. a.
    1.   Krishna was [vP depended [PP on ____] [PP for money]].
    1.  
    1. b.
    1. *Money was [vP depended [PP on Krishna] [PP for ____]].

This finding is arguably unexpected under analyses of intervention based exclusively on structural prominence (as implemented in (3), for example), given that there is clear empirical support for the availability of left-branching (also known as ascending) structures within the English VP/vP:

    1. (8)

If anything, structure-dependent approaches to intervention would lead us to expect that DPs further to the right in (8) should be privileged for extraction (cf. (5)). I show that this expectation is empirically falsified. Additionally, I show that (6) holds even when none of the potential goals c-commands the others. This demonstrates that the closeness metric in (3) is too permissive, since it wrongly predicts that any of the potential goals should be able to move in such cases.

Instead, I propose that the generalization in (6) is accounted for if, as has been recently argued by Ke 2019, Preminger 2019, Branan & Erlewine 2022, and Chow 2022, among others, intervention effects are derived from procedural aspects of the search algorithm that is initiated by particular features—called probes—on heads. Specifically, I argue that movement-triggering probes initiate a search procedure that proceeds iteratively from left to right in the probe’s search domain and that delves deeper into internally complex left daughters prior to considering right daughters. Search then halts when an accessible, matching goal is found. Because sister nodes are ordered for evaluation, the observed precedence-based intervention effects are accounted for. I therefore conclude that syntactic operations must be able to access information about linear precedence as well as structural height.

Moreover, I argue that conceivable alternative accounts of (6) that eschew the structure in (8) in favor of an obligatory right-branching (also known as descending) arrangement of arguments and adjuncts within vP (see Pesetsky 1995, Phillips 1996, Lechner 2003, Phillips 2003, R. Larson 2014, and R. Larson 2024)—

    1. (9)
    1. A strictly descending approach to English verb phrase structure (to be rejected)

—and that rely on purely height-based intervention (as in (3)) are not viable in light of three converging lines of evidence for the existence of ascending structures. The first two extend arguments from Janke & Neeleman 2012 concerning do so replacement and backward VP ellipsis, and the third comes from the anti-c-command condition on parasitic gap licensing (Taraldsen 1981, Chomsky 1982, and Engdahl 1983, among others).

The rest of the article is organized as follows. Section 2 develops a procedural analysis of intervention effects in movement dependencies that predicts precedence-based asymmetries in probing. Section 3 then turns to consider the structural geometry of the search domain of the passive movement probe. This section presents evidence from constituency and c-command tests for the availability of ascending and descending verb phrases. Section 4 argues that the predictions of the precedence-based analysis are borne out in passive A-movement: in both ascending and descending structures, only the leftmost accessible constituent within vP can be passivized; the relative structural prominence of other passivizable constituents is irrelevant. Section 5 considers and rejects various alternative accounts of the attested passivization asymmetries. I also argue in section 5 that well-known left–right asymmetries in binding and licensing fail to support a radically descending approach to English verb phrase structure. Section 6 considers the role that precedence might play in determining intervention in wh movement. Section 7 concludes by summarizing the main theoretical and empirical contributions of the article.

2 A search algorithm for Internal Merge

In this section, I develop a procedural analysis of intervention effects in movement dependencies, building on ideas in Branan & Erlewine 2022.

To begin, I adopt a probe–goal framework for analyzing how Merge features find their selectees. Merge features (e.g., [∙F∙]) are probes: they are lexically borne by heads, and they initiate a search for a matching goal. I adopt the following definition of matching, modeled closely after Zyman 2024: 31, (42).4

    1. (10)
    1. Definition of match
    2. For any selectional feature [∙F∙] and node N, N matches [∙F∙] iff [∙F∙] is keyed to
    1.  
    1. a.
    1. the categorial5 feature of N or
    1.  
    1. b.
    1. the lexical identity of N.

External Merge and Internal Merge are distinguished by the search space of the probe: with External Merge, the probe’s search space is the lexicon (or workspace or numeration), whereas with Internal Merge, I propose that the probe iteratively searches nodes in the c-command domain of the head that bears it (see (13) below).6

In order to derive intervention effects under movement, two things must be true of Merge-based probing. First, probing must halt once a matching goal is found.7 Otherwise, if no goal is found, the derivation crashes.8 Second, probing must proceed sequentially, node by node through the tree, and must not evaluate multiple nodes in parallel (contra Milway 2023: esp. 16–18 and Ke 2024: 859–860; see Richards 2019: 142 for related discussion). Additionally and quite importantly, I follow Branan & Erlewine 2022: 4 and Chow 2022: 3 in proposing that sisters are linearly ordered for evaluation—that is, linear precedence must be established during the syntactic derivation, prior to Internal Merge–based probing, and cannot be entirely postponed to the post-syntactic component (contra, e.g., Chomsky 1995b: sect. 4.8, Chomsky 2008: 139, Chomsky 2013: 36, Collins 2017, Chomsky 2020: 16, 27, 30–32, and Collins & Kayne 2023: 3, among many others).9

Two types of algorithms for searching trees, which differ in the order in which they evaluate nodes, have recently been discussed as candidates for probing in Atlamaz 2019, Ke 2019, Branan & Erlewine 2022, Krivochen 2023, Milway 2023, and Ke 2024: breadth-first search and depth-first search. A breadth-first search algorithm evaluates all nodes of a given depth in a tree (e.g., all daughters of node N1) before moving on to any of those nodes’ daughters (e.g., all granddaughters of N1). The algorithm then evaluates all nodes at the new depth, and so on. By contrast, a depth-first search algorithm evaluates (successive) daughter nodes in a straight path down the tree, without considering sisters. Assuming that sisters are linearly ordered for evaluation, let us propose that depth-first search evaluates the left daughter L1 of some node N1, followed by the left daughter L2 of L1, and so forth until a terminal node is reached. At that point, the depth-first search algorithm returns to the minimally dominating node that has an as yet unevaluated daughter and begins the procedure once more from there by evaluating that daughter. The following trees with probe [∙F∙] on head H illustrate the differences between these two algorithms; the nodes in the trees are numbered according to the order in which they are evaluated (see the works cited above for more in-depth discussion of these two types of search).10

    1. (11)
    1. Breadth-first, left-to-right search
    1. (12)
    1. Depth-first, left-to-right search

As these trees show, only a depth-first algorithm exhaustively11 searches an internally complex left sister (e.g., “2” in (12)) prior to probing into an internally complex right sister (e.g., “7” in (12)). In section 4, I will provide evidence from precedence-based intervention effects in English passivization that this ordering is necessary and hence that breadth-first search should be rejected (pace Ke 2019 and Ke 2024).

I propose a depth-first, left-to-right search algorithm for Merge-based probing in (13), taking inspiration from the algorithms in Atlamaz 2019: 88, (29), Preminger 2019: 24, (44), Chow 2022: 4, (4), and Deal, to appear 24. I assume that all derivable syntactic objects are binary-branching (Kayne 1984 and Collins & Stabler 2016: 57).12

    1. (13)
    1. A depth-first, left-to-right search algorithm for Merge-based probing
    1.  
    1. a.
    1. Let [∙F∙] be a syntactic probe on a head H, and let node N be H’s sister.
    1.  
    1. b.
    1. Mark N as visited and query: does N match [∙F∙]?
    1.  
    1.  
    1. i.
    1. If so, halt with the syntactic object of which N is the root as the goal.
    1.  
    1.  
    1. ii.
    1. Else, query: is N a non-terminal node?
    1.  
    1.  
    1.  
    1. •
    1. If so, mark N’s left daughter L as visited and query: does L match [∙F∙]?
    1.  
    1.  
    1.  
    1.  
    1. –
    1. If so, halt with the syntactic object of which L is the root as the goal.
    1.  
    1.  
    1.  
    1.  
    1. –
    1. Else, return to step (13bii), using L as the new “N.”
    1.  
    1.  
    1.  
    1. •
    1. Else, minimally backtrack in the reverse of the order in which nodes were visited until a visited non-terminal node is reached with an unvisited daughter node R and return to step (13b) using R as the new “N.” If there are no such nodes, halt with no goal.
    1.  
    1. c.
    1. query: was a goal found?
    1.  
    1.  
    1. i.
    1. If so, Merge(Hmax, “goal”).13
    1.  
    1.  
    1. ii.
    1. Else, abort the derivation.

To see how intervention effects follow from this probing algorithm, suppose that, as in (14) below, there are two nodes bearing the categorial feature [F] in the c-command domain of a head H bearing a probe [∙F∙] (i.e., in the search domain of the probe)—X[F] and Y[F]. In (14), X[F] is embedded in a complex left daughter of the starting node, whereas Y[F] is embedded in a complex right daughter of the starting node. According to the algorithm in (13), left daughters will be searched, node by node, until a terminal node is reached, at which point minimal backtracking will allow the algorithm to successively search through right daughters. Consequently, the probe’s search will encounter X[F] prior to Y[F].14 Branches of the tree that are traversed during search are in bold pink for salience.

    1. (14)
    1. Depth-first, left-to-right search predicts a precedence component in intervention with two potential goals

Because probing halts once a goal is reached and because probing operates from left to right in the tree, the algorithm in (13) thus predicts a precedence component in intervention.15

In Branan & Erlewine 2022: 5–8, esp. 7, it is noted that this prediction of depth-first, left-to-right search may be empirically supported by the existence of smuggling derivations (Collins 2005a and Collins 2005b), where extraction out of a fronted phrase in a left specifier preempts extraction of a phrase out of the specifier’s sister. In section 4, I adduce additional, novel evidence (which does not rely on the existence of smuggling derivations) showing that precedence-based intervention effects are indeed attested in passive A-movement in English, lending support to this general type of approach. However, in order to understand how the proposed search algorithm traverses the tree and to fully appreciate the role of precedence in this domain, I must first establish that English verb phrases can both ascend (i.e., branch left) and descend (i.e., branch right), since it is only with ascending verb phrases that the precedence-based and c-command-based definitions of “closeness” make different predictions (recall the discussion in section 1). This is the goal of the next section.

3 The structural geometry of the low, verbal domain in English

3.1 Structural assumptions

I follow much previous work in differentiating at least three heads in the low verbal domain: the lexical verb “V,” an argument-introducing head “v” that, among other things, introduces the external argument and determines the attachment site for by phrases (which I analyze as adjoined to the right of vP),16 and a voice head “Voice” that is the locus of diathetic alternations (see also Collins 2005b, Merchant 2013, Zyman 2017: esp. sect. 6.1, Roberts 2019: 395, Angelopoulos et al. 2020, Newman 2020: esp. 6–7, Zyman & Kalivoda 2020: sect. 3.1 (esp. 8), Collins 2024: chap. 8, and Hewett 2025). For the sake of explicitness, I take the clause-internal phase head to be Voice, at least in the passive (see Collins 2005b: 98). Furthermore, I assume that all internal arguments are base-generated as dependents of V and that low adjuncts (at least typically) adjoin to vP on the right:17

    1. (15)
    1. Structure of passive VoiceP

I abstract away from the presence and position of the passive auxiliary, which I take to be higher than VoiceP.

In Janke & Neeleman 2012 it is argued from constituency tests that both ascending and descending arrangements of arguments, as in (16) and (17) respectively, are possible in English VPs, though the availability of each structure depends on the categorial identities of the internal arguments in question.18

    1. (16)
    1. English ascending VP
    1. (17)
    1. English descending VP

The remainder of this section will defend this hypothesis by adducing evidence for both structures from do so replacement (section 3.2), VP ellipsis (section 3.3), and parasitic gap licensing (section 3.4).

3.2 Do so replacement reveals ascending and descending verb phrases

Particularly compelling evidence for the existence of both structures comes from do so replacement (see Lakoff & Ross 1966, Fu et al. 2001, Houser 2010, and Bruening 2019, among others). Comparing double object constructions, prepositional dative constructions, and constructions with two argument PPs (abbreviated here as “V–DP–DP,” “V–DP–PP,” and “V–PP–PP,” respectively), Janke & Neeleman 2012 shows that only in the latter two types of verb phrases can the verb and its linearly adjacent argument be replaced by do so (see also D’Elia 2016: 167–170):19

    1. (18)
    1.   V–DP–DP
    2. *If he gaveV [DP Mary] [DP anything], he did so [DP a woolen scarf].
    3.   (Janke & Neeleman 2012: 157, (13a))
    1. (19)
    1. V–DP–PP
    2. If he readV [DP a sonnet] [PP to anyone famous], he did so [PP to Salman Rushdie].
    3. (Janke & Neeleman 2012: 157, (14a))
    1. (20)
    1. V–PP–PP
    2. If they dependedV [PP on Congress] [PP for anything], they did so [PP for aid].

By contrast, do so cannot replace V and a non-adjacent argument, stranding V’s linearly adjacent argument:

    1. (21)
    1. V–DP–DP
    2. *If he gaveV [DP anyone] [DP a woolen scarf], he did so [DP Mary].
    1. (22)
    1. V–DP–PP
    2. *If he readV [DP anything] [PP to Salman Rushdie], he did so [DP a beautiful sonnet].
    1. (23)
    1. V–PP–PP
    2. *If they dependedV [PP on any government body] [PP for aid], they did so [PP on Congress].

Assuming that do so can only replace a constituent, these facts fall into place if V–DP–DP frames require a descending structure while V–DP–PP and V–PP–PP frames are compatible with an ascending structure:

    1. (24)
    1. (25)
    1. (26)

Only in V–DP–PP and V–PP–PP frames do the verb and its linearly adjacent argument form a constituent to the exclusion of the rightmost argument. Importantly, Janke & Neeleman 2012: 186–187 shows that the do so replacement facts cannot be explained by positing extraposition (or another form of Ā-movement) of the PP arguments in (19) and (20). This is because Ā-extraction cannot proceed out of do so (see also, e.g., Haddican 2007, Houser 2010: 21–22, and Bruening 2019):

    1. (27)
    2. (Adapted from Janke & Neeleman 2012: 187, (126c))
    1. (28)
    1. I know who Joni should read a sonnet to, but I don’t know who she shouldn’t(*do so).

Thus, the verb and its linearly adjacent argument in V–DP–PP and V–PP–PP frames must form a constituent prior to Ā-movement. Janke & Neeleman’s arguments can be extended straightforwardly to VPs with CP internal arguments as well: V–DP–CP frames disallow stranding of any arguments under do so replacement, as (29) illustrates, while V–CP–PP frames allow stranding of PP but not of CP, as (30) illustrates.20,21

    1. (29)
    1. a.
    1. *Matt promisedV [DP Joni] [CP that we’d bring plates], but Ty did so [DP Celina] [CP that we’d bring cups].
    1.  
    1. b.
    1. *Matt promisedV [DP Joni] [CP that we’d bring plates], but Ty did so [CP that we’d bring cups].
    1.  
    1. c.
    1. *Matt promisedV [DP Joni] [CP that we’d bring plates], but Ty did so [DP Celina].
    1.  
    1. d.
    1.   Matt promisedV [DP Joni] [CP that we’d bring plates], and Ty did so too.
    1. (30)
    1. a.
    1. *Matt saidV [PP to Joni] [CP that we’d arrive early], but Ty did so [PP to Celina] [CP that we’d arrive late].
    1.  
    1. b.
    1.   Matt saidV [PP to Joni] [CP that we’d arrive early], but Ty did so [PP to Celina].
    1.  
    1. c.
    1. *Matt saidV [PP to Joni] [CP that we’d arrive early], but Ty did so [CP that we’d arrive late].
    1.  
    1. d.
    1.   Matt saidV [PP to Joni] [CP that we’d arrive early], and Ty did so too.

I propose that V–DP–CP frames require a descending structure whereas V–CP–PP frames allow an ascending structure, with CP being base-generated as the complement of V:22

    1. (31)
    1. (32)

Finally, as is well-known, do so can replace the verb and its argument(s), stranding rightward verb phrase adjuncts such as the locative adjunct in the following.

    1. (33)
    1. If the students workedV [PP on their project] [PP in any classroom], they did so [PP in 12B].

We can explain this if low adjuncts attach to vP on the right and if do so replaces VP:

    1. (34)

In summary, evidence from do so replacement shows that English verb phrases can ascend and descend.

3.3 Backward VP ellipsis reveals ascending and descending verb phrases

The second constituency test revealing the existence of both ascending and descending verb phrases is verb phrase ellipsis, though as I will show presently, care is needed to rule out potential confounds in the data.

Janke & Neeleman 2012: 157–158 builds on earlier observations in Pesetsky 1995, Phillips 1996, Lechner 2003, and Phillips 2003, among others, to argue that the verb and its linearly adjacent argument can be elided, stranding a rightmost argument, in V–DP–PP and V–PP–PP frames but not in V–DP–DP frames. Janke & Neeleman account for this contrast by proposing that, in ascending V–{DP/PP}–PP structures, ellipsis targets V′, stranding the rightmost PP argument. By contrast, if V–DP–DP frames only allow a descending structure (see (24)), then V and its linearly adjacent argument will never form a constituent, correctly excluding ellipsis.

Unfortunately, their analysis is confounded by the existence of pseudogapping (see also Lechner 2003: 187–189, and McInnerney 2025: sect. 5), which is commonly argued to require Ā-extraction of the remnant out of VP prior to ellipsis of VP or a larger constituent (see Gengel 2013 for discussion and references, and see Kuno 1981 for an important antecedent). For instance, Merchant 2013: 95 proposes that pseudogapping remnants land in a position at least as high as [spec, VoiceP]. If the elliptical examples discussed by Janke & Neeleman involve pseudogapping and if pseudogapping requires movement prior to deletion, then their key examples don’t necessarily reveal anything about base-generated VP-internal structure.23

To control for this confound, a reviewer suggests testing backward ellipsis, since VP ellipsis can apply backwards but pseudogapping cannot (see Sag 1976: 83, n. 22 and Johnson 1996: 83):

    1. (35)
    1. a.
    1.   VP ellipsis
    2.   Because John didn’t ∅, Mary interviewed Gingrich.
    1.  
    1. b.
    1.   Pseudogapping
    2. *Because John did ∅ Clinton, Mary interviewed Gingrich.
    1.  
    1. (Lasnik 1999: 202, (7), (8))

The following contrasts are therefore striking: backward VP ellipsis can only strand VP-final arguments in those verb frames that were shown, using do so replacement, to allow ascending structures—namely, in V–DP–PP, V–PP–PP, and V–CP–PP but not in V–DP–DP or V–DP–CP:

    1. (36)
    1. a.
    1. V–DP–PP
    2. Although he did [VP ⟨[V′ give [DP a scarf]]⟩ [PP to Mary]], he refused to give a scarf to his best friend.
    1.  
    1. b.
    1. V–PP–PP
    2. Although he does [VP ⟨[V′ depend [PP on his parents]]⟩ [PP for emotional support]], he doesn’t depend on his parents for money.
    1.  
    1. c.
    1. V–CP–PP
    2. Although he did [VP ⟨[V′ mention [CP that he was upset]]⟩ [PP to Mary]], he refused to mention that he was upset to his best friend.
    1. (37)
    1. a.
    1.     V–DP–DP
    2.   *Although he did ⟨give [DP Mary]⟩ [DP a scarf], he refused to give Mary a warm hat.
    1.  
    1. b.
    1.     V–DP–CP
    2. ??Although he did ⟨promise [DP Joni]⟩ [CP that we would bring plates], he refused to promise Joni that we would bring cups.

(Examples (36a) and (37a) are due to the aforementioned reviewer.) Additionally, as (38) shows, rightward verb phrase adjuncts can be stranded under backward VP ellipsis; this is expected if such adjuncts attach on the right somewhere above VP, for example, to vP.

    1. (38)
    1. Although she did [vP ⟨[VP eat [DP a sandwich]]⟩ [PP on the couch]], she refused to eat a sandwich on the living room floor.

Thus, backward ellipsis marches in lockstep with do so replacement to reveal the presence of ascending and descending verb phrases in English.

3.4 The anti-c-command condition on parasitic gap licensing reveals ascending and descending verb phrases

Finally, consider parasitic gap licensing. A parasitic gap (pg) is a gap that is only grammatical in the presence of another gap created by wh movement (= the “licensing gap”; Taraldsen 1981, Chomsky 1982, and Engdahl 1983). Moreover, parasitic gaps are subject to the following condition (see especially Taraldsen 1981: 493, Chomsky 1982: 40, (56), Engdahl 1983: 22, (64), Haegeman 1984, Safir 1987, and Ershova 2021).

    1. (39)
    1. The anti-c-command condition on parasitic gaps
    2. The licensing gap cannot c-command the parasitic gap.

The effects of this condition can be seen in (40). A direct object gap can license a parasitic gap contained in an adjunct to its right, while a subject gap fails to license a parasitic gap in the same position.

    1. (40)
    1. a.
    1.   Which spyi did John kill _____i [before anybody could speak to pgi]?
    1.  
    1. b.
    1. *Which spyi ____i killed John [before anybody could speak to pgi]?
    1.  
    1. (Safir 1987: 678, (1a, b))

We can make sense of this contrast if the before clause attaches to vP on the right (Nissenbaum 2000). Then the direct object gap will not c-command into the adjunct clause, while the subject gap will:

    1. (41)
    1. a.
    1.   Which spyi did John [vP [vP kill ____i] [before anybody could speak to pgi]]?
    1.  
    1. b.
    1. *Which spyi [TP ____i [vP [vP killed John] [before anybody could speak to pgi]]]?

Thus, vP level adjuncts must be able to attach in an ascending structure:

    1. (42)

But that is not the whole story. For many speakers, parasitic gaps can also be licensed inside internal arguments in V–DP–PP and V–PP–PP verb frames. As the following examples show, extraction of and sub-extraction out of V’s linearly adjacent argument can license a parasitic gap in the non-adjacent PP argument to its right (see also Arregi & Murphy 2022: 13, fn. 7).

    1. (43)
    1.   V–DP–PP
    2. %[DP Which dish]i did you only [VP [V′ serve _____i] [PP to people who had had a taste of pgi before]] at your dinner party?
    1. (44)
    1.   V–PP–PP (see also Richards 2014: 183, (60b))
    2. %[DP Who]i did you [VP [V′ talk [PP to _____i]] [PP about all the reasons you voted for pgi]]?

These facts are straightforwardly explained if the relevant VPs ascend, as the licensing gap will not c-command the parasitic gap to its right, in compliance with the anti-c-command condition:

    1. (45)
    1. (46)

By contrast, as first reported in Arregi & Murphy 2022: sect. 4.1, extraction of the first DP in a V–DP–DP frame fails to license a parasitic gap in the second:

    1. (47)
    1.   V–DP–DP
    2. *[Which employee]i did you promise to [vP give _____i [DP several reports that you wrote on pgi] tomorrow]?

Note that the unacceptability of (47) should not be entirely chalked up to a constraint on extracting indirect objects in the double object construction (on which see Jackendoff & Culicover 1971: 399 and the references cited in Citko et al. 2017: sect. 4.2): (47) is significantly degraded compared to the minimally different (48), which removes the problematic parasitic gap.24

    1. (48)
    1. ?[Which employee]i did you promise to [vP give ______i [DP several reports that you wrote on Terry] tomorrow]?

Instead, the unacceptability of (47) can be accounted for if V–DP–DP frames require a descending structure, in line with evidence from constituency tests. As shown in (49), the licensing gap left by Ā-movement of the indirect object will c-command the parasitic gap in violation of the anti-c-command condition.

    1. (49)

The distribution of parasitic gaps thus provides a third argument for the existence in English of both ascending and descending verb phrases, whose distribution is constrained by the categorial identity of the arguments involved.25

3.5 Section summary

This section has argued that three independent tests for constituency/c-command support the generalization that V–DP–DP and V–DP–CP verb phrases require a descending structure while V–DP–PP, V–PP–PP, and V–CP–PP verb phrases as well as verb phrases with rightward adjuncts permit ascending structures. With our understanding of the structure of the low verbal domain now firmly in place, we can turn to documenting and accounting for precedence-based intervention effects in passivization.

4 Passive asymmetries within vP: evidence for a precedence component in intervention

This section lays out evidence in support of the following novel generalization.26

    1. (50)
    1. Order dependence in English passivization
    2. When just one constituent can undergo passive A-movement in English, it is invariably the leftmost accessible passivizable constituent within vP.
    3. = (6)

I argue that attested precedence-based intervention effects in passive A-movement are accounted for with the depth-first, left-to-right search algorithm for Merge-based probing that I proposed in section 2. I set aside, until section 5, conceivable alternative accounts of (50) based solely on structural prominence. In section 4.1, I sketch my analysis of passive A-movement: a probe on Voice[Pass] initiates depth-first, left-to-right search and attracts the leftmost accessible goal within vP to its specifier. Sections 4.2 and 4.3 document precedence-based intervention effects in English passivization. Section 4.2 discusses intervention among multiple internal arguments of V arranged in a descending structure. Section 4.3 discusses intervention in ascending verb phrases, both with multiple internal arguments and with combinations of arguments and adjuncts. Ascending verb phrases are particularly important: it is with these structures that the predictions of precedence-based and height-based accounts of intervention can be teased apart. Only the predictions of the precedence-based account are borne out. Section 4.4 summarizes.

4.1 A sketch of passive A-movement

I analyze passive A-movement as follows. (Non-expletive) passive subjects obligatorily move from their vP-internal base positions to the specifier of Voice[Pass] (see also Bruening 2011). In most cases the passive subject will be a DP, which might suggest that Voice[Pass] bears the structure-building feature [∙D∙]. Such a feature would trigger movement of the closest accessible DP into [spec, Voice[Pass]P]. However, as I demonstrate below, CPs can also appear as passive subjects and can intervene for passive A-movement of an accessible DP. Consequently, the trigger for passive A-movement must be general enough to be matched by DPs and CPs alike. At the same time, it must be specific enough to ignore PPs in pseudopassives:

    1. (51)
    1. a.
    1.   [DP Krishna] was laughed [PP at ______].
    1.  
    1. b.
    1. *[PP At Krishna] was laughed ______.

To account for the discriminatory nature of this movement trigger, I propose that Voice[Pass] bears the disjunctive selectional feature [∙{D, C}∙].27 This feature can be matched (see (10)) by a node of category D or C. This proposal is inspired by recent work on Agree in the interaction/satisfaction framework of Deal 2015 and Deal 2023, which proposes that the conditions under which a probe stops searching can be disjunctively specified (see Roversi 2020, Scott 2021, and Scott 2023).

The [∙{D, C}∙] probe will initiate a depth-first, left-to-right search into the c-command domain of Voice[Pass] according to the algorithm in (13). Importantly, due to the ordered nature of the search procedure, the constituent that moves to [spec, Voice[Pass]P], checking [∙{D, C}∙], is predicted to always be the leftmost accessible matching constituent within vP.28 This is illustrated schematically in (52) with three accessible DPs in vP.

    1. (52)

My analysis predicts that the relative structural prominence of (non-dominating) DPs/CPs within the passive probe’s search domain will have no bearing on intervention effects (contra Ura 1996: chap. 5, McGinnis 1998, McGinnis 2001, Anagnostopoulou 2003, Hartman 2012, and Holmberg et al. 2019, among others). The next sections demonstrate that this prediction is borne out in English.29

4.2 Precedence-based intervention with multiple internal arguments in descending VPs

I begin with passivization asymmetries among multiple internal arguments in a descending VP. Recall from section 3 that constituency/c-command tests diagnose a descending VP structure in V–DP–DP frames and in V–DP–CP frames. I will show that only the leftmost accessible DP/CP within vP is capable of passivizing.

4.2.1 V–DP–DP

First, consider double object constructions in Standard English, exemplified with deny. In the active voice, the goal DP obligatorily precedes the theme DP, as (53a) illustrates, and in the passive, only the goal DP can appear as the passive subject, as shown by (53b) versus (53c).

    1. (53)
    1. V–DP–DP
    1.  
    1. a.
    1.   The bank denied {[DP Mike]} [DP a loan] {*[DP Mike]}.
    1.  
    1. b.
    1.   [DP Mike] was denied ______ [DP a loan] by the bank.
    1.  
    1. c.
    1. *[DP A loan] was denied [DP Mike] _____ by the bank.30

Thus, when the order of arguments is absolutely fixed in the active, we find a corresponding passive asymmetry: only the goal DP, which appears leftmost on the surface, can be passivized (see appendix A for discussion and analysis of theme passives in double object constructions in some languages).

We can account for this asymmetry as follows. The [∙{D, C}∙] probe on Voice[Pass] will initiate a depth-first, left-to-right search for a goal within vP, as shown in (54) (I abstract away from probing into complex heads for the sake of simplicity). This search will always reach a DP dominated by a left daughter node prior to reaching a DP dominated by a right daughter node. Since the goal (indirect object) DP occupies a left specifier of V, it will therefore be attracted to [spec, Voice[Pass]P], and because the probe never reaches the theme DP, we account for the attested intervention effect. In (54) and subsequent trees in this section, I only number nodes up to the root of the goal that satisfies the probe.

    1. (54)

Note, too, that the left-to-right nature of search means that the DP within the by phrase will never become the passive subject, despite being the most structurally prominent DP within vP,31 because the by phrase is adjoined on the right—a welcome result (*The bank was denied Mike a loan by). See section 4.3.3 for more discussion of passivization and high right adjuncts.

4.2.2 V–DP–CP

Next, consider verbs that select both a DP internal argument and a CP internal argument. I focus on CPs headed by that (abbreviated CPthat), but the relevant generalizations extend to many other kinds of CPs, including control infinitivals (abbreviated CPcontrol), finite and non-finite interrogative CPs headed by whether/if, and embedded finite and non-finite wh questions. The only licit constituent order in the active is V–DP–CP:

    1. (55)
    1. The principal assured {[DP Reuben]} [CP that Judah was not in trouble] {*[DP Reuben]}.

As with V–DP–DP frames, only the immediately post-verbal argument (i.e., DP) is able to passivize (see also Higgins 1973: 184–187 and Bach 1980: 326, (40), (42)):32

    1. (56)
    1. a.
    1.   [DP Reuben] was assured ______ [CP that Judah was not in trouble] by the principal.
    1.  
    1. b.
    1. * [CP That Judah was not in trouble] was assured [DP Reuben] ______ by the principal.

Note that we should not attribute this asymmetry to a failure of CPs to match the probe on Voice[Pass] that triggers passive A-movement. Verbs like guarantee that alternate between a ditransitive V–DP–CP frame and a transitive V–CP frame clearly show that CPs can appear as passive subjects in the absence of a DP intervener:

    1. (57)
    1. [CP That the judge would make a ruling quickly] wasn’t guaranteed (*[DP the defendant]) _____.

Furthermore, as we will see in sections 4.3.2 and 4.3.3, complement CPs headed by that actually intervene for pseudopassivization out of co-argument and adjunct PPs. All of these facts are accounted for if the probe on Voice[Pass] can be matched by either a DP or a CP. Thus, I propose that the extraction asymmetry in (56) manifests an intervention effect. The derivation in (58) illustrates.

    1. (58)

Thus, in V–DP–CP frames, as in V–DP–DP frames, we observe intervention in passive A-movement that is compatible with the linearity-based generalization argued for here: only the leftmost accessible passivizable constituent within vP can undergo A-movement.

4.3 Precedence-based intervention in ascending structures

In this section, I turn to consider intervention in ascending structures, where structural height and linear order cease to march in lockstep. First, I show that precedence-based intervention is attested with multiple internal arguments in ascending VPs, specifically in V–{DP/PP}–PP frames (section 4.3.1) and in V–CP–PP frames (section 4.3.2); recall from section 3 that tests for constituency/c-command consistently diagnose ascending structures in these cases. Second, I show that precedence-based intervention is observed whenever a vP level adjunct co-occurs with one or more arguments (section 4.3.3). In every case, our precedence-based generalization from (50) is substantiated. Importantly, this section demonstrates that the relative structural prominence of a DP or CP within vP does not affect its ability to undergo passive A-movement—a finding that runs counter to purely height-based accounts of intervention as in (3). I postpone considering non-intervention-based accounts of these passivization asymmetries until sections 5.2 and 5.3.

4.3.1 V–{DP/PP}–PP

Precedence-based intervention is observable in ascending VPs with two internal arguments when the second is a PP and the first is either a DP or a PP. In the active voice, the neutral order of arguments is V–{DP/PP}–PP, as (59a) and (60a) illustrate. Crucially, as with multiple arguments in descending VPs (see section 4.2), passivization only targets the leftmost accessible DP within vP, as in (59b) and (60b); pseudopassivization out of the second PP is impossible over the intervening DP/PP, as (59c) and (60c) illustrate (see also Van Riemsdijk 1978: 219, Hornstein & Weinberg 1981: 65, (38), Stowell 1981: 438–439, 451, and Blight 2000: 85–86; see appendix A for an analysis of apparent exceptions to this generalization in Swedish and (idiolects of) English).33

    1. (59)
    1. V–DP–PP
    1.  
    1. a.
    1.   They attributed [DP that idea] [PP to Joni].
    1.  
    1. b.
    1.   [DP That idea] was attributed _____ [PP to Joni] by them.
    1.  
    1. c.
    1. *Joni was attributed [DP that idea] [PP to _____] by them.
    1. (60)
    1. V–PP–PP34
    1.  
    1. a.
    1.   The kids depended [PP on Krishna] [PP for money].
    1.  
    1. b.
    1.   Krishna was depended [PP on _____] [PP for money] by the kids.
    1.  
    1. c.
    1. *Money was depended [PP on Krishna] [PP for ____] by the kids.

Such intervention effects are derived straightforwardly from depth-first, left-to-right search. The tree in (61) illustrates for a V–PP–PP frame.35

    1. (61)

By contrast, these asymmetries are unexpected under purely structural approaches to intervention. Neither [DP Krishna] nor [DP money] in (61) asymmetrically c-commands the other within VP prior to A-movement, so neither should qualify as “closer” to the probe on Voice[Pass] given a definition of closeness like (3).36 Furthermore, path-based definitions of closeness as in Hornstein 2009: 37–39 that measure distance in terms of the number of maximal projections dominating both the extraction site and the landing site lead us to expect either DP to be passivizable, since they are equidistant from Voice[Pass].37

As observed in Postal 1986: 225–226 and Pesetsky 1995: 275, passive asymmetries in V–{DP/PP}–PP frames persist even when the immediately post-verbal argument undergoes extraposition to the right:

    1. (62)
    1. *Jonik was attributed ______i [PP to ______k] [DP every single idea that made the company money]i.
    1. (63)
    1. * Moneyk was depended _____i [PP for ______k] [PP on every funding source that was available]i.

We can understand why extraposition of an argument A to the right of an argument B fails to feed pseudopassivization out of B if extraposition necessarily follows (passive) A-movement within VoiceP. Determining why exactly this ordering relation should hold is ultimately a task for future work, but I will make one concrete suggestion here. Adopting the traditional analysis of extraposition/heavy shift of XP as rightward Ā-movement of XP (e.g., Ross 1967), I propose that XP attaches to the right of VoiceP—the clause-internal phase (see also Bruening 2018; see Overfelt 2015: 77–82 for additional discussion of the landing site of heavy shift). If this movement is driven by a feature on Voice—call it [∙Ā∙]—then we can posit a strict ordering of features on Voice[Pass]: Voice[Pass][∙{D, C}∙ ≺ ∙Ā∙] (where “≺” is to be read as “precedes”). Ideally, such an ordering should not be extrinsically stipulated but should follow from independent principles of grammar. One possibility is that A-movement features such as [∙{D, C}∙] on a head X are invariably ordered before Ā-features such as [∙Ā∙] on X, though I will leave exploring this and other possibilities to future work. I offer a similar explanation to account for intervention effects with verbs selecting a PP argument and a CP argument in section 4.3.2.

It should be noted at this juncture that, unlike depend, there are some verbs that freely allow reordering of their multiple PP internal arguments. With such verbs, pseudopassivization out of either PP is possible, though crucially only when that PP appears to the left of its PP co-argument(s). The verb talk is illustrative in this regard:38

    1. (64)
    1. a.
    1. The witnesses were talked {[PP to _____]} [PP about the crime scene] {*[PP to _____]} by the investigator.
    1.  
    1. b.
    1. The crime scene was talked {[PP about _____]} [PP to the witnesses] {*[PP about ____]} by the investigator.

We can account for the difference between these two classes of verbs as follows. Depend type verbs only permit one base-generated order of internal arguments (i.e., [VP [V′ depend on X] for Y]), so we only expect pseudopassivization to be possible out of the PP generated as the complement of V. On the other hand, I propose that talk type verbs are compatible with both base-generated orders of internal arguments (i.e., [VP [V′ talk to X] about Y] or [VP [V′ talk about Y] to X]). As a consequence, we expect variation such that the DP that can undergo passive A-movement corresponds to the DP that is generated leftmost within vP.

4.3.2 V–CP–PP

Turning now to verbs that select a PP internal argument and a CP internal argument, the facts are somewhat complicated by a frequent preference for CPs to be extraposed. Such extraposition obscures the underlying order of constituents within VP, which I propose is V–CP–PP.39 I will show that, if the underlying order is indeed V–CP–PP (a hypothesis that is independently supported by the constituency facts reviewed in section 3), then the depth-first, left-to-right probing algorithm argued for here successfully accounts for the precedence-based intervention effects observed in these clauses. As in section 4.2.2, I restrict my attention to verbs selecting CPs headed by that though similar patterns can be shown to hold with numerous other kinds of CPs (see footnote 42 for discussion of control CPs).

Under neutral prosody, the preferred order of constituents in the active voice is V–PP–CPthat. For me and the other speakers I have consulted (see also Bruening 2018: 364, fn. 2), the order V–CPthat–PP is acceptable (pace Emonds 1970: 97–99, Stowell 1981: 161, (91b), and Postal 1986: 226, (63b)), if marked; though it is strongly dispreferred if CP is considerably prosodically heavier than PP. The following example illustrates the two possible orders.

    1. (65)
    1. The doctor suggested {[PP to Vandy]} [CP that Russ should see a specialist] {[PP to Vandy]}.

Despite the preference for a V–PP–CPthat order, only CPthat can appear as the passive subject (see also Higgins 1973: 184–185):40

    1. (66)
    1. [CP That Russ should see a specialist] was suggested _____ [PP to Vandy] by the doctor.

The DP complement of P cannot pseudopassivize, regardless of the position of the stranded preposition relative to CPthat (see Alexander & Kunz 1964: 26, Higgins 1973: 184–185, (83), (84), Emonds 1976: 124, fn. 7, Bach 1980: 326, (41), (43), Postal 1986: 225, (60), Pesetsky 1995: 275, (670), (671), and Drummond 2011: 171, (316)):

    1. (67)
    1. *Vandy was suggested {[PP to ______]} [CP that Russ should see a specialist] {[PP to _____]}.

This is despite the fact that, with verbs that allow CPthat to be omitted, pseudopassivization is fine in CPthat’s absence:

    1. (68)
    1. The gorilla was signaled [PP to ______] by Anthony (*[CP that dinner time was near]).
    2. (Postal 1986: 225, (61d))

Thus, I contend that the failed pseudopassivization in (67) should be attributed to intervention by CPthat and not to a lexical incompatibility of these verbs with pseudopassivization.

Note additionally that stranding P to the immediate left of CPthat as in (67) is not degraded due to the *P–CPthat effect familiar from Ā-movement (e.g., Kuno 1973: 382—which attributes the observation to Judith Aissen—Langendoen & Pullum 1977, Stowell 1981: 207–208, Moulton 2015: 322–324, and Lebowski 2021). In some idiolects, the presence of an intervening adverbial can improve Ā-movement that strands P to the left of CPthat (as observed by Bruening 2018: 381–392; see also Rauber 2019: 1–2 and Zyman 2022: 151, fn. 26), but there is no comparable amelioration contributed by adverbials in A-movement (thanks to Erik Zyman for pointing this out to me); compare the acceptable P-stranding wh question in (69) with the unacceptable P-stranding passive in (70).

    1. (69)
    1.   Who did the doctor suggest [PP to _____] on Tuesday [CP that Russ should see a specialist]?
    1. (70)
    1. *Vandy was suggested [PP to ____] on Tuesday [CP that Russ should see a specialist].

I thus propose to account for passive asymmetries in V–CP–PP frames by holding that CPthat is base-generated as the complement of V and PP is base-generated in a right specifier of V, as shown in (71). Depth-first, left-to-right search in passive A-movement will therefore always reach the CP internal argument first, explaining the attested intervention effects.41

    1. (71)

The order V–PP–CPthat observed in the active is derived through extraposition of CPthat, which I analyze as rightward movement to VoiceP (see Bruening 2018: 365). Because extraposition must follow passive A-movement (see section 4.3.1), we account for the failure of pseudopassivization in (67) and (70).42

4.3.3 Passive asymmetries among internal arguments and rightward adjuncts

Finally, precedence-based intervention is observable with passive asymmetries obtaining between internal arguments and vP level adjuncts. Consider verbs that optionally select a DP, PP, or CP internal argument. When just this internal argument is present, it can (pseudo)passivize:

    1. (72)
    1. [DP A sandwich] was eaten _____ by every guest.
    1. (73)
    1. No group project was worked [PP on ______] by the students.
    1. (74)
    1. [CP That the world is flat] was written ______ by the conspiracy theorist.

When the optional internal argument is omitted, the same verbs permit pseudopassivization out of a locative adjunct PP:

    1. (75)
    1. This couch was eaten [PP on ____] by every guest.
    1. (76)
    1. No classroom was worked [PP in ____] by the students.
    1. (77)
    1. This notebook was written [PP in ____] by the conspiracy theorist.

However, if both the internal argument and the adjunct PP are present, only the DP (contained in the) internal argument can passivize; pseudopassivization out of the adjunct PP is impossible:

    1. (78)
    1. a.
    1.   [DP A sandwich] was eaten ____ [PP on this couch] by every guest.
    1.  
    1. b.
    1. * This couch was eaten {[PP on ____]} [DP a sandwich] {[PP on ____]} by every guest.
    1. (79)
    1. a.
    1.   No group project was worked {[PP on ____]} [PP in this classroom] {*[PP on ____]} by the students.
    1.  
    1. b.
    1. * No classroom was worked {[PP in ____]} [PP on this group project] {[PP in ____]} by the students.
    1. (80)
    1. a.
    1.   [CP That the world is flat] has never been written ____ [PP in this notebook] by anybody before.
    1.  
    1. b.
    1. * This notebook has never been written {[PP in ____]} [CP that the world is flat] {[PP in ____]} by anybody before.

These asymmetries are predicted by my account of intervention based on left-to-right probing, under the assumption that locative adjuncts attach above internal arguments on the right (I omit an explicit derivation for reasons of space).

4.4 Section summary

Let us take stock. This section has marshaled empirical evidence in favor of the following novel generalization.

    1. (81)
    1. Order dependence in English passivization
    2. When just one constituent can undergo passive A-movement in English, it is invariably the leftmost accessible passivizable constituent within vP.
    3. = (6, 50)

I proposed an analysis of precedence-based intervention effects that relied on procedural aspects of the search algorithm triggered by Merge features: probing exhaustively searches through a complex left daughter of a node N prior to evaluating N’s right daughter. Because the search procedure halts when an accessible, matching goal is found, we account for attested left–right asymmetries in passivization. A crucial claim of this analysis is that the relative structural prominence of two or more accessible goals within the search domain of the probe makes no difference for intervention effects, which are determined via linear order (and dominance). My analysis is therefore compatible with the evidence laid out in section 3 for ascending and descending arrangements of arguments and adjuncts.

We can, in fact, strengthen the argument for precedence-based intervention. Consider (82). Backward ellipsis strands V’s rightmost argument PP, indicating that the stranded PP is higher than its co-argument, and yet the passive probe systematically picks out the leftmost argument for movement.

    1. (82)
    1. Although they have been [VP ⟨[depended on ____]⟩ [PP for money]], they have never been depended on for emotional support.

I therefore conclude that there is robust evidence for countenancing a linear component in passive intervention effects.43 In the next section, I consider and reject alternative accounts of left-to-right intervention that do not refer to linear precedence.

5 Against alternative analyses of precedence-based intervention

While the discussion up to this point has focused on deriving left-to-right intervention via a search procedure that is sensitive to linear precedence, there are other conceivable accounts of the asymmetries documented in section 4 that ascribe no role to precedence in the syntax. In this section, I discuss three such alternatives and argue in every case that the precedence-based approach is to be favored on empirical grounds. The alternative accounts to be considered are: those that combine traditional height-based intervention with uniformly descending verb phrases (section 5.1), those that derive failed pseudopassivization through a V–P adjacency requirement on “re-analysis” (section 5.2), and those that derive failed pseudopassivization from the (adjunct) islandhood of certain PPs (section 5.3).

5.1 Left-to-right intervention should not be re-analyzed as height-based intervention with radically descending vP structure

A first alternative would be to abandon the precedence-based view of closeness in favor of a more traditional metric for closeness based on c-command, like the following.

    1. (83)
    1. “Closeness” as closest asymmetric c-command
    2. Given three syntactic objects X, Y, and Z, Z is closer to X than Y is iff:
    1.  
    1. a.
    1. X asymmetrically c-commands Z and
    1.  
    1. b.
    1. Z asymmetrically c-commands Y.
    1.  
    1. = (3)

On this approach, we could derive the generalization in (81) by proposing that arguments and adjuncts uniformly descend within vP. In other words, under purely structure-dependent approaches to intervention, it must be the case that “rightward is downward” (Pesetsky 1995: 161).

While such an analysis arguably faces challenges with the constituency facts discussed in section 3, it boasts provisional support from well-known binding and licensing asymmetries first reported in Barss & Lasnik 1986, R. Larson 1988, and R. Larson 1990 and discussed extensively in subsequent literature (see especially Pesetsky 1995, Phillips 1996, Lechner 2003, Phillips 2003, R. Larson 2014, and R. Larson 2024). Let us consider four such asymmetries here.44 In the domains of variable binding, the each … the other construction, negative polarity item (NPI) licensing, and Condition C, binding and licensing seem to only go rightward within vP:

    1. (84)
    1. Variable binding goes rightward in vP
    1.  
    1. a.
    1.   Which snack did you [vP give to every boyi despite himi not liking it]?
    1.  
    1. b.
    1. * Which snack did you [vP give to himi despite every boyi not liking it]?
    1. (85)
    1. Each … the other licensing goes rightward in vP
    1.  
    1. a.
    1.   What did you [vP give to each contestant together with the other’s drawing of Pepper]?
    1.  
    1. b.
    1. * What did you [vP give to the other together with each contestant’s drawing of Pepper]?
    1. (86)
    1. NPI licensing goes rightward in vP
    1.  
    1. a.
    1.   Joni [vP described that painting to no children at any lecture on it].
    1.  
    1. b.
    1. * Joni [vP described that painting to any children at no lecture on it].
    1. (87)
    1. Condition C goes rightward in vP
    1.  
    1. a.
    1.   What did you [vP give to Matti by throwing it at himi]?
    1.  
    1. b.
    1. * What did you [vP give to himi by throwing it at Matti]?

In all grammatical examples, the binder/licensor is (contained in) an argument while the bindee/licensee is contained in an adjunct.45 Under the assumption that these relations require the binder/licensor to c-command the bindee/licensee, we seem to have evidence in favor of a radically descending structure like (88), where (traditional) adjuncts can be generated within the c-command domain of internal arguments.46

    1. (88)

Additionally, because the binders/licensors in (84–87) are complements of prepositions, it must be assumed that PP nodes “do not count” for determining command relations (see the “everything-but-PP-command” of Pesetsky 1995: 173, (453)).

Consequently, if left-to-right order within vP unambiguously maps onto a descending structure as in (88), then the order dependence of English A-movement (81) can be chalked up to height-based intervention: the leftmost DP within vP will also be the structurally most prominent, ostensibly accounting for the facts.

However, previous work has noted that the uniformly descending structures seemingly required by binding and licensing asymmetries are not supported by constituency tests, which consistently point to the availability of ascending structures even in clauses where binding goes rightward, yielding so-called Pesetsky paradoxes (see, in addition to the works cited above (84), Landau 2007, Janke & Neeleman 2012, Bruening 2014b, and Neeleman & Van de Koot 2022). Rather than rehearse the arguments of the prior literature here, I will adduce novel evidence from sentences combining parasitic gap licensing and left-to-right binding within vP that illustrates that the relevant binding and licensing relations do not require c-command. Crucially, without the binding/licensing argument for uniformly descending structures, there are no independent reasons to define “closeness” in terms of asymmetric c-command, as in (83). I therefore conclude that the precedence component in intervention cannot be reduced to structural height, as in (88).

Consider (89–92), which differ from (84–87) only in the addition of a parasitic gap; in all cases, parasitic gaps are licensed in vP level adjuncts at the same time that binding and licensing proceed rightward into the adjunct.47

    1. (89)
    1. Parasitic gap licensing with simultaneous left-to-right variable binding in vP
    1.  
    1. a.
    1.   Which snackk did you [vP give ____k to every boyi despite himi not liking pgk]?
    1.  
    1. b.
    1. * Which snackk did you [vP give ____k to himi despite every boyi not liking pgk]?
    1. (90)
    1. Parasitic gap licensing with simultaneous left-to-right each … the other licensing in vP
    1.  
    1. a.
    1.   Whatk did you [vP give ____k to each contestant together with the other’s drawing of pgk]?
    1.  
    1. b.
    1. * Whatk did you [vP give ____k to the other together with each contestant’s drawing of pgk]?
    1. (91)
    1. Parasitic gap licensing with simultaneous left-to-right NPI licensing in vP
    1.  
    1. a.
    1.   Whatk did Joni [vP describe ____k to no children at any lecture on pgk]?
    1.  
    1. b.
    1. * Whatk did Joni [vP describe ____k to any children at no lecture on pgk]?
    1. (92)
    1. Parasitic gap licensing with simultaneous left-to-right Condition C effects in vP48
    1.  
    1. a.
    1. *Whatk did you [vP give ____k to himi by throwing pgk at Matti]?
    1.  
    1. b.
    1.   Whatk did you [vP give ____k to Matti by throwing pgk at himi]?

On the one hand, in order to satisfy the anti-c-command condition, the direct object trace within vP must not c-command the parasitic gap–containing adjunct; this requires an ascending structure. On the other hand, if the relevant binding and licensing phenomena obtain only under c-command, then a descending structure is required. This is a constituency paradox:49

    1. (93)

Similar paradoxes can be constructed with tritransitive verbs like trade and sell that select one DP internal argument and two PP internal arguments, though I omit the data here for space.

In order to resolve the constituency paradox, we must therefore abandon one of our premises: either binding and licensing do not diagnose c-command, or parasitic gap licensing does not diagnose the absence of c-command.50 While I am not aware of an alternative account of the anti-c-command condition on parasitic gap licensing, there are alternative analyses of binding that do not rely on c-command. For instance, Bruening 2014b (building on earlier proposals by Langacker 1969, Jackendoff 1972, and Lasnik 1976) argues that binding, as defined in (94), is constrained by the conjunction of two separate relations: linear precedence and phase command, the latter as defined in (95).

    1. (94)
    1. Binding
    2. A binds B iff A and B are co-indexed and A precedes and phase-commands B.
    3. (Bruening 2014b: 344, (4); formatting adjusted)
    1. (95)
    1. Phase command
    2. X phase-commands Y iff there is no Z P, ZP a phasal node, such that ZP dominates X but does not dominate Y.
    3. (Bruening 2014b: 343, (2); formatting adjusted)
    1. (96)
    1. Phasal nodes
    2. CP, VoiceP, DP51

By adopting Bruening’s definition of binding, we can account for the data in (89–92) without abandoning the anti-c-command condition on parasitic gap licensing.52

Consider (89a) once more, which combines parasitic gap licensing with rightward binding into an adjunct. The parasitic gap–containing adjunct must attach above the direct object licensing gap to comply with the anti-c-command condition, and it must attach to the right of it. I assume for simplicity that the adjunct attaches to vP on the right. In order for the QP every boy to bind the pronominal variable him in the adjunct, every boy must precede and phase-command him. Because PP is not a phasal node (see (96)), the minimal phasal node dominating every boy is VoiceP, and this node also dominates him. Therefore, every boy precedes and phase-commands him and can bind him despite the lack of c-command:

    1. (97)
    1. Which snackk did you [VoiceP [vP give ____k [PP to every boyi]] [PP despite himi not liking pgk]]?

Thus, the precede-and-phase-command approach accounts for binding and licensing asymmetries straightforwardly without adopting a radically descending approach to English verb phrase structure. Indeed, as we have seen, there is solid empirical evidence from various constituency/c-command tests that arguments and adjuncts can ascend, even in sentences where binding/licensing proceed from left to right. I therefore conclude that intervention effects in (A-) movement dependencies cannot be accounted for with a height-based definition of closeness (83), since the leftmost accessible passivizable constituent within vP is not always highest. Instead, intervention must be determined linearly (at least in part).53

5.2 Failed pseudopassivization is not due to a lack of V–P adjacency

Many of the illicit instances of passive A-movement discussed in section 4 involve pseudopassivization. While I argued there that the unacceptability of these examples results from intervention, a reviewer wonders whether this unacceptability might have a different source. Consider the claim from prior work, stated in (98), that pseudopassivization requires the stranded P to be adjacent to its local verb (e.g., Van Riemsdijk 1978: 218–223, Hornstein & Weinberg 1981: 85–86, Bresnan 1982: 54, Pesetsky 1995: 275, Blight 2000, Drummond & Kush 2015: 430–432, Richards 2017, and McInnerney 2021: 321–322; see Baltin & Postal 1996: 127 for many more references).54

    1. (98)
    1. V–P adjacency hypothesis for pseudopassivization (to be rejected)
    2. Pseudopassivization requires V–P adjacency.

The requirement for V–P adjacency is often taken to follow from a mechanism of “re-analysis,” which is generally understood to involve (possibly lexically determined) incorporation of P into V. Putative evidence in favor of this hypothesis comes from examples like (99), which show that certain elements like temporal adverbials cannot intervene between V and P under pseudopassivization.

    1. (99)
    1. The movie was talked(*today) about ____.
    2. (Richards 2017: 313, (1a))

If (98) could be maintained, it would offer an alternative explanation for a significant portion of the data in section 4.3. However, I will show presently that (98) is much too restrictive, incorrectly ruling out a range of acceptable examples in English where V and P are non-adjacent.55 Consequently, I conclude that there is no general requirement for V–P adjacency in pseudopassivization.

First, as a number of previous authors have observed, pseudopassivization actually is possible across (potentially phrasal) low manner and degree adverbials in English, as in (100) (see also Stowell 1981: 449, Takami 1992: 99, Culicover & Jackendoff 2005: 208, Postal 2010: 201, Findlay 2016: 258–259, McInnerney 2022b: 379, (19), and Therrien 2023: 27, fn. 17).

    1. (100)
    1. This door has been leaned heavily [PP on ____] once too often.
    2. (Truswell 2008: 168)

Second, a stranded P can be immediately preceded by modifiers like right (see also Tseng 2006: 149, (23), Postal 2010: 200, (5.6h), and Findlay 2016: 258, (8a)):

    1. (101)
    1. The ditch was jumped [PP right over ____] by the running child.
    2. (Therrien 2023: 27, (30b))

Not only is P not adjacent to V in (101), but, if right modification indicates that P projects phrasal structure (see Den Dikken 1995: sects. 2.2.2, 2.4.6 and Janke & Neeleman 2012: sect. 5), then such examples clearly demonstrate that P has not incorporated into or been re-analyzed with the verb.

Third, Baltin & Postal 1996 presents a battery of forceful arguments against “re-analysis” analyses of English pseudopassives. One of Baltin & Postal’s most convincing arguments comes from the recognition that stranded Ps can be coordinated under across-the-board A-movement (see Postal 2004: 261 and Postal 2010: 200–201 for many additional examples):

    1. (102)
    1. The bridge was flown over and then, but only then, under.
    2. (Baltin & Postal 1996: 130, (9a, c))

The non-initial P in (102) is non-adjacent to V, in violation of (98). Moreover, as with right modification, the presence of the coordinator and indicates that there is phrasal structure hierarchically intervening between P and V, ruling out a “re-analysis” analysis.

In summary, there is no case for a general V–P adjacency requirement on pseudopassivization in English nor for a corresponding “re-analysis” analysis of pseudopassives. I therefore reject the hypothesis in (98) and maintain that an intervention-based account of the failed pseudopassives in section 4 is empirically superior. The unacceptability of pseudopassivization across temporal adverbs as in (99) is explained on my analysis if such adverbs are necessarily generated higher on the right than argument PPs. This is because extraposition of PP would be necessary to derive the V–temp.adv–PP order and, as shown in section 4.3, extraposition counter-feeds (pseudo)passivization.

5.3 Failed pseudopassivization is not due to PP (adjunct) islandhood

The final alternative account that I will consider attempts to derive the relevant asymmetries not from relative locality constraints (i.e., intervention) but from absolute ones. Specifically, some previous work has suggested that pseudopassivization is possible out of argument PPs but not out of adjunct PPs (e.g., Baker 1985: sect. 5.3.3, Kageyama & Ura 2002, Alsina 2009, and Honda 2012: sect. 4.4.3). If adjuncts are generally opaque for extraction, due to their being islands (see Truswell 2011: sect. 1.2 for critical discussion and references), then we might try to explain the unacceptability of pseudopassives like (103) as the result of an adjunct island violation.

    1. (103)
    1. * Money was depended [PP on Krishna] [PP for ____] by the kids.
    2.   = (60c)

Nevertheless, the islandhood account of illicit pseudopassivization fails for at least three reasons. First, all of the PPs characterized as arguments in section 4.3 behave like arguments when subjected to argument/adjunct diagnostics (for an overview, see Zyman 2022 and Zyman 2023: sect. 2).56 vP pseudoclefting is representative in this regard: in a vP pseudocleft, adjuncts but not arguments can immediately follow do (i.e., can be stranded; Zyman 2022: 134–136; see also Hedberg & DeArmond 2009). Crucially, the relevant PPs behave like arguments, despite their failure to launch pseudopassivization when preceded by an accessible intervener. I illustrate this in (104) with the PP arguments of depend, but similar results hold for the other PP arguments discussed in section 4.3.

    1. (104)
    1. V–PP–PP (compare (60))
    1.  
    1. a.
    1. *What the kids did [PP on Krishna] was [vP depend [PP for money]].
    1.  
    1. b.
    1. *What the kids did [PP for money] was [vP depend [PP on Krishna]].57

Second, section 4.3.3 showed that pseudopassivization is possible out of low adjuncts in the absence of a preceding intervener. Example (105) illustrates, and example (106) demonstrates that the on PP is an adjunct through vP pseudoclefting.

    1. (105)
    1. This couch was eaten [PP on ____] by every guest.
    2. = (75)
    1. (106)
    1. What every guest did [PP on this couch] was [vP eat].

Third, many of the same PPs that do not allow pseudopassivization, such as the rightmost argument PP in (103), are transparent for wh extraction, as (107) illustrates, and hence cannot be absolute islands.

    1. (107)
    1. Whatk was Krishnai depended [PP on ____i] [PP for ____k]?

Thus, the constraints on pseudopassivization documented in section 4.3 cannot be accounted for by appealing to the alleged adjuncthood of the relevant PPs.

5.4 Section summary

I have argued that three alternative explanations of the left-to-right intervention effects documented in section 4 fail to account for the full range of data and/or require untenable assumptions. In every case, I have contended that the intervention-based approach developed in section 2 provides a better account of the facts. Section 5.1 also marshaled novel evidence, in the form of a constituency paradox, that the binding and licensing asymmetries commonly taken to support uniformly descending verb phrases in English do not, in fact, require c-command—a conclusion recently defended by Bruening 2014b. This finding contributes to our general understanding of what diagnoses constituency and why.

6 Precedence-based intervention outside of passives

I now turn to intervention outside of passives. A clear prediction of my analysis, which holds that left-to-right58 search is initiated by all Internal Merge–triggering features, is that precedence-based intervention ought to also be attested in other forms of movement, such as wh movement.59 Contrasts like (2), which illustrate classical superiority effects, can be explained by either a precedence-based or height-based approach to intervention, since subjects precede objects in addition to c-commanding them. More interesting, then, are the contrasts in (108), which suggest that linear order may indeed be the right metric for determining closeness in wh questions (see also Fiengo 1980: 122–123 and Jackendoff 1990: 433 on superiority effects in other ditransitive verb phrases).60,61

    1. (108)
    1. a.
    1. i    Whoi did you talk [PP to ____i] [PP about what]?
    2. ii *Whati did you talk [PP to who] [PP about ____i]?
    3. (McInnerney 2022b: 285, (147); see also Jackendoff 1990: 433, (25c, d) and R. Larson 2014: 317, (48d))
    1.  
    1. b.
    1. i    Whoi did you give the cookies [PP to ____i] [PP on what day]?
    2. ii *What dayi did you give the cookies [PP to who] [PP on ____i]?
    3.        (Cf. What dayi did you give the cookies [PP to Joni] [PP on ____i]?)

These contrasts are expected on a precedence-based approach to intervention but are surprising on (and arguably unaccounted for by) a height-based approach.

Of course, a complete account of intervention in wh movement would need to take into account all of the variables reported to bear on the presence versus absence of superiority effects, including D-linking (Pesetsky 1987) and, in languages like German, whether or not the wh phrases are clausemates (e.g., Haider 1983, Grewendorf 1988, Haider 1993, Grohmann 1997, and Wiltschko 1997).62 This is not the place to give such an account. Nonetheless, I will mention one open question for a precedence-based account of superiority effects and sketch how it might be addressed.63

Consider the following pair (see Kuno & Robinson 1972: 474).

    1. (109)
    1. a.
    1. What did you eat where?
    1.  
    1. b.
    1. Where did you eat what?

If where is base-generated in a clause-final position in both examples, then (109b) would ostensibly attest to a lack of left-to-right intervention. By contrast, a height-based approach to intervention could account for these facts by positing variation in whether locatives are generated in a descending structure (yielding (109a)) or an ascending one (yielding (109b)).

While the assumption that where is moving from a position to the right of what in (109b) is certainly plausible (see Lechner 2004: 220–222 for an analysis along these lines), it is by no means necessary. In fact, there is suggestive evidence that where does not move from a clause-final position in (109b). Compare (109) with (110).

    1. (110)
    1. a.
    1. Whati did you eat ____i near where?
    1.  
    1. b.
    1. *Wherei did you eat what near ____i?

Example (110b) uses P-stranding to disambiguate the launching site of movement for where; strikingly, the superiority effect re-emerges in this example.64 Returning to (109b), a precedence-based approach could account for the absence of an intervention effect by proposing that where moves from a position to the left of what (plausibly in a descending structure; see Fiengo 1980: 124 for a similar account of these facts).65 I therefore provisionally conclude that the basic facts of superiority in English are compatible with a precedence-based approach to intervention, though I must leave a fuller investigation for future research.

7 Conclusion

This article reported and provided an account of the following novel generalization.

    1. (111)
    1. Order dependence in English passivization
    2. When just one constituent can undergo passive A-movement in English, it is invariably the leftmost accessible passivizable constituent within vP.
    3. = (6, 50, 81)

I argued that (111) finds a simple explanation in procedural aspects of probing. Building on Branan & Erlewine 2022 and Chow 2022, among others, I proposed a depth-first, left-to-right algorithm for search triggered by Merge features: the algorithm exhaustively searches through an internally complex left daughter of a node N prior to evaluating N’s right daughter. Because search halts when a matching goal is found, the leftmost among multiple potential goals in the probe’s search domain will always be privileged for extraction. Unlike most previous approaches to intervention effects, my analysis ascribes no role to c-command, though dominance does play a role due to search being top-down. By not making assumptions about the relative structural prominence of multiple potential goals within the probe’s search space, my analysis is compatible with both ascending and descending structures. Crucially, I showed there to be strong independent evidence for ascending structures, despite apparent evidence from binding and licensing to the contrary. Radically descending approaches to English verb phrase structure, as well as a number of other alternative accounts of the passive intervention facts, were argued to be empirically inadequate. The discovery in (111) therefore provides new and independent support for countenancing precedence in the syntax. It remains to be seen whether my proposed top-down left-to-right search algorithm might also be deployed to facilitate parsing, for instance in a filler-driven parsing account of filler–gap dependencies (see Momma & Phillips 2018: sect. 4.3.1 and references therein). Future work will hopefully elucidate this potential connection between parser and grammar.

Supplementary material

A file containing the appendices can be downloaded at https://doi.org/10.16995/star.17274.s1.

Acknowledgments

Thanks, for their feedback, to Karlos Arregi; Ruth Kramer; Zach Lebowski; Jason Merchant; Andy Murphy; Ad Neeleman; Kutay Serova; Erik Zyman; audiences at Georgetown University, New York University, Yale University, and the LASER Workshop at the University of Göttingen; and three anonymous reviewers.

Competing interests

The author declares that they have no competing interests.

Notes

  1. Some relevant definitions:
      1. (i)
      1. Constituent command (c-command)
      2. X c-commands Y iff:
      1.  
      1. a.
      1. neither X nor Y dominates the other and
      1.  
      1. b.
      1. the first branching node that dominates X dominates Y.
      1.  
      1. (Adapted from Reinhart 1976: 32, (36))
      1. (ii)
      1. Asymmetric c-command
      2. X asymmetrically c-commands Y iff X c-commands Y and Y does not c-command X.
    ⮭
  2. For explicit proposals as to how a selectional feature specified on a head can be satisfied under merge of the selectee with a non-minimal projection of that head, see Adger 2003: 109–110, Merchant 2019: 326, and Zyman 2024: 28–44. ⮭
  3. Other kinds of intervention-based locality have been argued for, including A-over-A locality, which is sensitive to irreflexive dominance rather than asymmetric c-command; see Chomsky 1973, Hornstein 2009: sect. 2.5.1, Müller 2011, Halpert 2019, Preminger 2019: sect. 7.3, and Müller 2021: sect. 12.4.1.1, among others, and see Zyman 2023: sect. 7 and references therein for arguments against an A-over-A Condition. ⮭
  4. In (10), what matches the Merge feature [∙F∙ ] is the root R of a syntactic object X, whereas according to Zyman 2024’s definition of match, what matches [∙F∙] is the syntactic object X itself. ⮭
  5. If the [wh] feature on a wh DP is not a categorial feature per se and if the definition in (10) also constrains Internal Merge–based selection for wh phrases in wh movement (see section 6), then this definition will likely need to be modified. Alternatively, if C[+wh] doesn’t probe for a DP bearing a [wh] feature but rather probes for a QP, in the terms of Cable 2007 and Cable 2010, then the current definition will suffice. ⮭
  6. This precludes upward Merge-based probing (under the assumption that structure-building features do not project from the heads that bear them), ruling out movement of a constituent from the specifier of a phrase to another specifier of the same phrase. On the absence of phrase-bounded specifier-to-specifier movement, see Abels 2003: 104, 105, fn. 50, 114, Ko 2007, Grohmann 2011, Ko 2011, Ko 2014, Davis 2020: sect. 4.1, and Bondarenko & Davis 2021 and works cited in those references. Also, see Branan & Erlewine 2022: sect. 4.2 on alternative starting nodes for the search algorithm. ⮭
  7. A number of authors have attributed this property to what Noam Chomsky has called “minimal search” (e.g., Chomsky 2004: 109, 113, Chomsky 2005: 14, Chomsky 2013: 39, 43, Chomsky 2015: 6, Chomsky 2021: 18, and Berwick et al. 2013: 41–42; see also Epstein et al. 2014). However, “minimal search” does not seem to have been formally defined until very recently. See, for instance, Atlamaz 2019, Ke 2019, Aycock 2020, Epstein et al. 2020: 3–6, Branan & Erlewine 2022: 2, Chow 2022, Ke 2023, Krivochen 2023, Milway 2023: 11–18, and Ke 2024 for some exemplars, and see B. Larson 2015 for an alternative definition and conceptualization. Because there is no current consensus on the definition of minimal search, I avoid using it here. ⮭
  8. I adopt this perspective for Merge-based probing only: selectional features must be checked/satisfied on pain of a crash (see Merchant 2014: 1 and Zyman 2024; pace Longenbaugh 2019: chap. 5 and Newman 2021), whereas Agree has been argued to be fallible (Preminger 2014). ⮭
  9. For other work suggesting that precedence is part of the syntax, see Kayne 1994: esp. 48–49, Williams 1994, Saito & Fukui 1998, Fukui 2001, Kayne 2011, Zwart 2011, Bruening 2014b, Al Khalaf 2015, Bruening 2018, Bruening & Al Khalaf 2020, and Kayne 2022. One possibility is that every application of Merge(A,B) yields an ordered pair ⟨A,B⟩ rather than an unordered set {A,B}. ⮭
  10. Note that the root node is not evaluated for match as it does not fall within the search space of the probe, which is the c-command domain of H; see also footnote 6. ⮭
  11. Save for portions of the structure opaque to probing (e.g., spellout domains that have already been spelled out and are thus syntactically inaccessible); see Branan & Erlewine 2022: sect. 4.1 for discussion. ⮭
  12. Precedence-based intervention could also be derived using a static representational definition of closeness along the lines of (i) paired with the definition of intervention in (1).
      1. (i)
      1. “Closeness” based on precedence or irreflexive dominance
      2. Given three syntactic objects X, Y, and Z, Z is closer to X than Y is iff:
      1.  
      1. a.
      1. X asymmetrically c-commands Z and
      1.  
      1. b.
      1. Z precedes or irreflexively dominates Y.
    I believe that the algorithmic definition in (13) has a slight advantage over this representational alternative in predicting that a relation other than c-command is what determines intervention. Specifically, on my procedural account, the search algorithm must be sensitive to some asymmetric relation between sisters/daughters in order to determine the order of evaluation. C-command will not suffice because sisters c-command each other. Linear precedence (in addition to irreflexive dominance), on the other hand, is the obvious candidate for such an asymmetric relation (but see Atlamaz 2019: 89–91 for a different approach). On representational accounts, however, the need for a non-c-command-based asymmetry between sisters does not follow—at least, not as naturally. Still, a reviewer cautions that the procedural account based on (13) requires admitting diacritics and counter-cyclicity into syntactic theory and therefore might be disfavored for independent reasons. In the end, if the representational alternative turns out to be optimal, none of my arguments in favor of precedence-based intervention or of my claims about the structure of English verb phrases will change. ⮭
  13. See Zyman 2024 on the definition of Merge. ⮭
  14. In (14), I label nodes in the order in which they would be evaluated by the search algorithm were it not to halt. Strictly speaking, however, once X[F] (i.e., node “7”) is reached, none of the subsequently numbered nodes will be searched. ⮭
  15. As observed in Branan & Erlewine 2022: 6, fn. 9, breadth-first search as in (11) also predicts a precedence effect among potential goals at the same depth level, though not among goals at different depths. ⮭
  16. My “v” thus corresponds to Kratzer 1996’s “Voice.” ⮭
  17. Not all aspects of the proposed structure in (15) are crucial. All that is necessary on my account is that the relevant arguments and adjuncts are introduced within the search domain of the movement-triggering probe on Voice[Pass], namely vP. For instance, some or all (internal) arguments might be introduced by low functional heads, as in Borer 2005a, Borer 2005b, Pylkkänen 2008, Bowers 2010, Lohndal 2012: esp. 121–128, and De Belder & Van Craenenbroeck 2015: 647–648. Some adjuncts might also attach as low as VP (e.g., Bowers 2002: 187–189). ⮭
  18. See Janke & Neeleman 2012 for an account of the distribution of these structures based on Case licensing and economy, and see Abels & Grabska 2022 for arguments in favor of ascending and descending ditransitive verb phrases in Polish. ⮭
  19. There is well-known inter-speaker variation in this domain. See Zyman 2022: 133–134 for evidence that arguments (including PPs) cannot be stranded under do so replacement in many idiolects, in contrast to the behavior of adjuncts. ⮭
  20. For related observations about the unstrandability of CPs under do so replacement, see Johnson 2004: 44, (126) and Mita 2009: 132, (29) on monotransitive and tritransitive verbs, respectively. ⮭
  21. The pattern of judgments reported in (30) for V–CP–PP verbs extends to verbs selecting a PP not headed by to, such as conclude from X that Y (data omitted for space). This finding militates against an alternative analysis of (30b) according to which do so selects a “patient-like to-phrase,” as proposed by McInnerney 2022b: 136–140. Even if do so can select such a to phrase in many idiolects, as McInnerney argues, the alternative analysis fails to account for the parallel pattern of judgments with verb phrases lacking a to phrase. See additionally example (iii) in footnote 42 for evidence that do so can replace (certain) verbs and their control CP complements, stranding an of PP argument. ⮭
  22. The unmarked surface order V–PP–CP arises due to rightward movement of CP (see Bruening 2018 and section 4.3.2). ⮭
  23. Similarly, arguments for ascending VPs on the basis of verb phrase preposing (e.g., Janke & Neeleman 2012: 157–158) are confounded by the possibility of remnant movement. Concretely, the acceptability of (i) demonstrates that V and its linearly adjacent DP argument can form a constituent to the exclusion of the stranded PP argument in a V–DP–PP frame (though see R. Larson 2024 for an alternative approach that denies that the preposed string forms a constituent):
      1. (i)
      1. I wanted to giveV [DP the book] [PP to someone interesting], and [give the book] I did [PP to Mary].
      2. (Slightly adapted from Janke & Neeleman 2012: 184, (120a))
    What is not clear from just this example is whether (i) reflects base-generated VP constituency or instead constituency derived through rightward movement of the stranded PP argument prior to preposing. On the former analysis, the preposed constituent give the book would be V′, whereas on the latter analysis, it could be as large as vP or even VoiceP, depending on where extraposition lands. Indeed, if existing arguments for the preposed constituent being vP (or larger) are on the right track (see Thoms & Walkden 2019 and McInnerney 2022b: 166–171), then the interaction between verb phrase preposing and stranding arguably does not shed light on the structure of VP. ⮭
  24. For other work noting the acceptable (if marginal) status of indirect object Ā-movement in the double object construction in (standard) English, see Bolinger 1971: 138–139 and Barss & Lasnik 1986: 348, and see Holmberg et al. 2019: 681 and references there on the acceptability of such sentences in varieties of British English. ⮭
  25. As is well-known, parasitic gaps in complement CPs are also licensed by Ā-movement of clausemate internal argument DPs (i.e., in V–DP–CP frames):
      1. (i)
      1. [DP Who]i did you tell ____i [CP that we were going to vote for pgi]?
      2. (Engdahl 1983: 11, (18))
    If the arguments for a descending analysis of V–DP–CP frames in sections 3.2 and 3.3 are on the right track, then (i) would seem to constitute a violation of the anti-c-command condition, as originally observed by Contreras 1984. To resolve this tension, I adopt Safir 1987’s re-analysis of examples like (i) as involving string-vacuous extraposition of CP (which I propose lands CP at the right edge of vP; see also Arregi & Murphy 2022: sect. 6.2.3):
      1. (ii)
    Post-CP-extraposition, the licensing gap in [spec, VP] (see (31)) will not c-command a parasitic gap contained in CP, satisfying (39). ⮭
  26. For earlier work that ascribed a non-trivial role to (immediate) linear precedence in passivization, see McCawley 1998: 88–90 and Emonds 2017: 54, n. 34. ⮭
  27. See Hewett 2025 for other possible feature specifications for Voice[Pass]. For completeness, I will mention three alternative ways to account for the parallelism between DPs and CPs without invoking disjunctive selectional features. The first would involve analyzing sentential subjects as essentially nominal in nature (e.g., Lees 1960, Rosenbaum 1967, Ross 1967, Delahunty 1983, Davies & Dubinsky 1998, Davies & Dubinsky 2001, Han 2005a, Han 2005b, Davies & Dubinsky 2009, Takahashi 2010, and Hartman 2012). Specifically, we might propose that CPs can be encased in a DP shell whose head is null. On such an approach, [∙D∙] would be a sufficient trigger for movement to [spec, Voice[Pass]P]. However, for such an approach to work, CPs would always need to be encased in a DP shell, since CPs are obligatory (not optional) interveners for passive A-movement (see section 4.3.2, especially footnote 42). It remains to be seen whether the conjecture that CPs are always encased in a DP shell is justifiable. A second alternative would be to propose that DP and CP share a super-categorial feature in the sense of Bruening & Al Khalaf 2020 (see also Sag et al. 1985)—one that is crucially not shared with PPs or VPs, among others. The intervention facts could then be accounted for by positing a selectional feature for this super-category on Voice[Pass]. A third option would be to propose that DP and CP are both potential goals for ϕ-Agree relations with Voice[Pass], under the assumption that Agree is a pre-condition on movement (e.g., Chomsky 2000 and Chomsky 2001). However, see Halpert 2019: 154 for evidence that CPs are not satisfactory goals for ϕ-agreement in English. I leave deciding between these and other alternatives for future work. ⮭
  28. A constituent contained in a domain that is opaque to the [∙{D, C}∙] probe on Voice[Pass] (e.g., a DP in the spellout domain of a lower phase head) will never be considered by the probe and hence will not count as an intervener (see also footnote 11). Furthermore, as Erik Zyman (personal communication) points out to me, because the search algorithm proposed in (13) is sequential and evaluates mother nodes before their daughters, my analysis correctly excludes passives as in (ic) (which are unacceptable even in idiolects like Zyman’s that allow possessor extraction under all available forms of Ā-movement, on which see Davis 2021).
      1. (i)
      1. a.
      1.   Someone keyed [DP1 [DP2 Katie]’s car].
      1.  
      1. b.
      1.   [DP1 [DP2 Katie]’s car] was keyed ____DP1.
      1.  
      1. c.
      1. *[DP2 Katie] was keyed [DP1 ____DP2’s car].
    The search initiated by [∙{D, C}∙] on Voice[Pass] will encounter the node DP1 prior to encountering the node DP2 because the former dominates the latter, thereby accounting for the (A-over-A) intervention effect. As a reviewer observes, there remain other types of A-movement that, at least superficially, appear to violate A-over-A minimality, like possessor raising in languages like Nez Perce (see Deal 2013 and Deal 2017). Note, however, that much hangs on what exactly the trigger for movement is in possessor raising. If it is [∙D∙] or the like, then an A-over-A effect is expected à la (ic). But if the trigger is more specific, then an A-over-A violation could be avoided. The intervention-like constraints identified for possessor raising in Nez Perce ditransitives by Deal 2013: 403–405 seem especially enlightening in this regard, though I must leave a fuller exploration of locality in possessor raising for future work. ⮭
  29. As a reviewer notes, data like (i) suggest that surface subjects in English show quite a bit of categorial flexibility: they can evidently not only be DPs and CPs but also, for example, PPs.
      1. (i)
      1. A.
      1. What route shall we take?
      1.  
      1. B.
      1. [PP Through the woods] takes the least time.
    Similarly, the inverted PP in locative inversion constructions has been argued to display properties associated with canonical subjecthood; see Diercks 2017 for an overview. If—as seems likely—at least some of these subjects cannot be re-analyzed as being DPs, then it may be that the probe on T triggering movement into T’s specifier is even less picky about the category of the goal it is searching for. Importantly, such a state of affairs does not present any issues for my analysis, which locates the movement-triggering probe responsible for passive intervention effects on Voice[Pass]; (i) does not contain Voice[Pass]. ⮭
  30. See footnote 33 for arguments that the unacceptability of (53c) and similar examples cannot be entirely attributed to violations of the Case Filter. ⮭
  31. At least according to closeness metrics that calculate the length of a path according to the cardinality of nodes (or branches) on that path (see Müller 1998: 130 for a proposal along these lines). The cardinality of nodes/branches on the path from the starting vP node to the root node of the DP within the by phrase is one less than the cardinality of nodes/branches on the path from the starting vP node to the root node of the indirect object DP. A similar issue arises if, as suggested by Müller 2011: 20–21, fn. 9, path length is determined by the number of intervening categories (in the sense of May 1985: 56–57 and Chomsky 1986: 7), as opposed to segments. This is because, under the segment theory of adjunction, adjuncts are not dominated by the categories to which they are adjoined; hence, the DP within the by phrase adjoined to vP is separated from (the probe on) Voice[Pass] by one fewer category than either internal argument DP within VP, and the by phrase DP is therefore expected to intervene, contrary to fact. Finally, if structural prominence were evaluated solely with respect to asymmetric c-command, the DP complement of by and the indirect object DP should count as equally close to the probe, and we would still fail to predict intervention. ⮭
  32. Other verbs displaying a similar pattern are enumerated by Alexander & Kunz 1964: 24–25, including: advise, bet, caution, convince, forewarn, guarantee, inform, notify, persuade, promise, remind, show, teach, tell, and warn. ⮭
  33. A skeptical reader might object that (59c) is independently ruled out by the Case Filter: under the assumption that passive verbs do not assign accusative Case (Chomsky 1981, Jaeggli 1986, and Baker et al. 1989), the unmoved DP will not be assigned Case, and the derivation will crash. While this is certainly true, the objection still fails to address why a DP structurally closer to the probe (e.g., in the higher argumental PP or even in the PP by phrase) can be bypassed by the passive probe—that is, why doesn’t the derivation of a passive clause always crash in such examples? The precedence-based approach provides an answer: within the probe’s search domain, only precedence (and irreflexive dominance), not asymmetric c-command, matters for intervention. Note too that this objection fails to explain why there is intervention in V–PP–PP frames, where the assignment of structural accusative Case is never at stake. ⮭
  34. Other V–PP–PP verbs that behave similarly are apologize to X for Y and appeal to X for Y. ⮭
  35. I remain agnostic as to whether V raises to v in ascending structures. Nothing in my analysis hinges on this point. ⮭
  36. On the other hand, if PP nodes “do not count” for evaluating command relations (see, e.g., the “everything-but-PP-command” of Pesetsky 1995: 173, (453)), then the expectation would be that intervention in passivization should proceed from right to left, contrary to fact. ⮭
  37. Concretely—and assuming in line with Bare Phrase Structure (Chomsky 1995a) that the two PP nodes in (61) are differentiable by their lexical identities (i.e., on vs. for)—we have Path(Voice[Pass], Krishna) = {VoiceP, vP, VP, onP} and Path(Voice[Pass], money) = {VoiceP, vP, VP, forP}. Thus, in a Hornstein-style analysis where comparison of path lengths requires one path to form a subset of the other, the two DPs are erroneously predicted to be equally passivizable. Additionally, if “closeness” were measured by counting the nodes (or branches) intervening between the probe and the accessible goals (e.g., Müller 1998: 130), we would again expect right-to-left intervention effects, contrary to fact. ⮭
  38. Another verb allowing pseudopassivization out of either of its PP arguments is abscond with X from Y. ⮭
  39. For important predecessors to my claim that CP is the inner argument of V, see Stowell 1981: 161, Pesetsky 1995: 273–275, and Bruening 2018. ⮭
  40. Alexander & Kunz 1964: 28–29 cites numerous other verbs occurring in the V–CPthat–PP frame that behave similarly to those discussed in the main text, including: acknowledge, admit, announce, assert, communicate, confess, convey, declare, demonstrate, disclose, divulge, explain, hint, indicate, point out, relate, report, reveal, say, signal, and whisper. ⮭
  41. For concreteness, I assume that CPs undergo leftward A-movement from their base-generated positions within VP. But this assumption is not necessary; my analysis is also compatible with the alternative view that CP subjects are base-generated topics associated with a null nominal operator, where it is the operator and not the CP itself that undergoes A-movement (see Koster 1978 and Alrenga 2005). Even if argumental CPthat is incapable of undergoing leftward movement, as this approach maintains (see also Bruening 2018), it nonetheless obligatorily intervenes for pseudopassivization out of a co-argument PP, indicating that the probe on Voice[Pass] halts when it reaches the leftmost DP or CP. On this approach, then, intervention by CP would constitute a case of defective intervention (e.g., McGinnis 1998, Chomsky 2000, and Hartman 2012). ⮭
  42. Passive intervention actually reveals two classes of verbs selecting a PP and an infinitival, non-interrogative control CP (i.e., CPcontrol) for many speakers. With verbs like plead (call these “class 1 verbs”), only the DP complement of P can passivize; this is shown in (i). With verbs like beg, ask, and demand (call these “class 2 verbs”), on the other hand, only CPcontrol can passivize; this is shown in (ii).
      1. (i)
      1. a.
      1.   The clerk was pleaded [PP with ____] [CP to keep the store open a bit longer].
      1.  
      1. b.
      1. * [CP To keep the store open a bit longer] was pleaded [PP with the clerk] ____.
      1. (ii)
      1. a.
      1. *The clerk was begged/asked/demanded [PP of ____] [CP to keep the store open a bit longer].
      1.  
      1. b.
      1.   [CP To keep the store open a bit longer] was begged/asked/demanded ____ [PP of the clerk].
    Other class 1 verbs are call [PP on X] [CP to … ], appeal [PP to X] [CP to … ], and depend [PP on X] [CP to … ]. Another class 2 verb is suggest [PP to X] [CP to … ]. I account for this difference as follows. Class 1 verbs select PP as their complement and CPcontrol as a right specifier: [VP [V′ plead PPwith] CPcontrol]. Class 2 verbs project a structure parallel to verbs selecting a that CP, with CPcontrol as a complement and PP in a right specifier: [VP [V′ beg CPcontrol] PPof]. Evidence from constituency tests supports this distinction: do so replacement identifies V + CPcontrol as a constituent with class 2 verbs like beg, as in (iii), but not with class 1 verbs like plead, as in (iv).
      1. (iii)
      1.   If Matt beggedV/askedV/demandedV [PP of anyone] [CP to keep the store open a bit longer], he did so [PP of the clerk].
      1. (iv)
      1. *If Matt pleadedV [PP with anyone] [CP to keep the store open a bit longer], he did so [PP with the clerk].
    That said, one might be tempted to relate this asymmetry in CPcontrol passivization to another contrast—namely, that only class 2 verbs allow CPcontrol to be replaced by a DP:
      1. (v)
      1. *The customer pleaded [DP something unreasonable] [PP with the clerk].
      1. (vi)
      1.   The customer begged/asked/demanded [DP something unreasonable] [PP of the clerk].
    Previous authors have noted that a CP that has been moved leftward in English must always relate to a gap that could have been occupied by a DP in the active voice (e.g., Bruening 2018: 366). Therefore, an initially plausible account of the contrasts in (i) and (ii) might attempt to correlate the status of CPcontrol as an intervener for passive movement with the ability of that CPcontrol to undergo leftward movement. On such an analysis, CPcontrol with class 1 verbs would never constitute a licit goal for the probe on Voice[Pass]. This account is immediately challenged by the following contrast. CPthat arguments of boast cannot undergo leftward (A-) movement, as (vii) illustrates (this is in line with the fact that boast forbids DP internal arguments), yet they still intervene for pseudopassivization out of co-argument PPs, as (viii) illustrates. (These examples are due to a reviewer.)
      1. (vii)
      1. *[CP That she had won a Pulitzer] was boasted ____ [PP to me] by Mary.
      1. (viii)
      1. *I was boasted [PP to ____] by Mary [CP that she had won a Pulitzer].
    Thus, CPs can defectively intervene for passive A-movement (see also footnote 41). I conclude that the contrasts in (i) and (ii) result from intervention and do not simply reflect the (in)ability of different constituents to undergo leftward movement. ⮭
  43. A reviewer suggests another way to combine tests for ascending structure and left-to-right intervention. Barbiers 1995 observes that only in its scalar and temporal uses must immediately c-command its semantic argument. Neeleman & Payne 2020 utilizes the strict locality of scalar/temporal only to diagnose the presence of ascending structures in Dutch and English. If we accept Neeleman & Payne’s arguments, then, in (i) (under the indicated interpretation), only must immediately c-command the rightmost PP argument, providing additional evidence for ascending structures in passive sentences.
      1. (i)
      1. Mary was only [VP [V′ apologized [PP to ____]] [PP for three of the insults hurled at her]].
      2. ‘Mary was apologized to for three insults, and three is a low number in this context.’
    ⮭
  44. The first three were discussed originally in Barss & Lasnik 1986 in the context of multiple internal arguments in the double object construction. Barss & Lasnik also discussed reciprocal binding, which I leave aside for space. Condition C violations induced by a direct object binding into an adjunct clause to its right were discussed in R. Larson 1990 622, citing observations made by Contreras 1984. For additional discussion of direct objects binding rightward into adjuncts, see Lasnik & Saito 1991. ⮭
  45. See Bowers 2010: 33–37 for the related observation that arguments can bind into passive by phrases. ⮭
  46. The structure in (88) is inspired by Larsonian VP shells and shows the result of successive head movement of V, but invoking the cascade structures of Pesetsky 1995 would achieve a similar result. ⮭
  47. While the (a) examples in (89–91) may be marked for some speakers (Erik Zyman reports that he is one such speaker), all the speakers I have consulted report an extremely sharp contrast in acceptability between the (a) and (b) examples. ⮭
  48. See Arregi & Murphy 2022: 34, fn. 16 for discussion of a contrast similar to the one reported in (92) using an embedded complement clause instead of a vP adjunct. ⮭
  49. A defender of the radically descending analysis in (88) might try to account for (89–92) by positing extraposition of a constituent—presumably a kind of VP—containing just the to PP and the adjunct to its right (and a trace of V). Such extraposition would move the parasitic gap out of the c-command domain of the licensing gap, and, assuming that PP nodes do not count for binding, the DP complement of P could bind into the adjunct to its right:
      1. (i)
    Not only is such an analysis unlikely given that predicates that undergo Ā-movement obligatorily reconstruct (see, e.g., Huang 1993 and Heycock 1995), since such reconstruction would re-introduce the pathological c-command relation between the licensing and parasitic gaps, but also, such an analysis is not supported by constituency tests. For example, the alleged “PPargument + PPadjunct” constituent cannot be preposed around though:
      1. (ii)
      1. *But [VP? to every boyi despite himi not liking it]m though I will [VoiceP [vP give a healthy snack ____m]], I wish there were a better way to promote good dietary choices.
    This is despite the fact that the extraposition analysis in (i) holds that the putative constituent can be moved rightwards. ⮭
  50. Alternatively, we could reject the hypothesis that strings correspond to a single constituent structure, as Pesetsky 1995 does. See Janke & Neeleman 2012: 181–182 for conceptual reasons not to pursue such an approach. ⮭
  51. I diverge from Bruening 2014b in two regards: (i) I take the clause-internal phase, at least in the case of the passive, to be VoiceP rather than vP, and (ii) I propose that the maximal nominal projection, which forms a phase, is DP, not NP. ⮭
  52. For other accounts of binding and licensing phenomena that countenance a role for precedence, see Barss & Lasnik 1986: 352, Jackendoff 1990, Ernst 1994, Shan & Barker 2006, Barker 2012, Janke & Neeleman 2012: sect. 7, and Neeleman & Van de Koot 2022. ⮭
  53. See appendix B for arguments against an alternative account of the constituency paradoxes based on derivational ordering. ⮭
  54. McInnerney 2022b: 376, fn. 5 diverges from this consensus in proposing that it is the phrase headed by the stranded P that must be right-adjacent to the verb. However, even this proposal is empirically falsified by data like (100). ⮭
  55. Similarly, V–P adjacency is not required for (expletive and non-expletive) pseudopassives in Norwegian (e.g., Christensen 1986: 156, (64), (65) and Hestvik 1986: 184–185). ⮭
  56. Though see McInnerney 2022a, McInnerney 2022b, McInnerney 2023, and McInnerney 2025 for the proposal that traditional argument/adjunct diagnostics do not provide sufficient evidence for a syntactic differentiation between arguments and adjuncts. ⮭
  57. This example is acceptable on an irrelevant parse where the for PP is a purposive adjunct. The “*” diacritic only applies to the argument interpretation of this for PP. ⮭
  58. My analysis, on the face of it, leads us to expect that, in languages with dominant VOS or OVS word order, O should be privileged for (classical or hyper-) raising to subject or object (assuming that independent constraints do not rule out extraction of O; see Rackowski & Travis 2000: 124 and Chung 2017 for pertinent discussion). If it turns out that, in such languages, S is actually privileged for extraction, my analysis could nonetheless be maintained by positing cross-linguistic parameterization of the direction of search: either left-to-right or right-to-left. I will leave this important question for future work. ⮭
  59. Similarly, a reviewer points out that my analysis predicts left-to-right intervention effects in raising-to-subject across languages, all other things being equal. Yet, it is well-known that languages seem to differ in this regard: in English, experiencer PPs can intervene between the verb and the infinitival under raising to subject, but in languages like French and Italian, they cannot. An English example is in (i), and a French example is in (ii); for Italian, see Rizzi 1986.
      1. (i)
      1. John seems [PP to Mary] [____ to have talent].
      1. (ii)
      1. Jean
      2. Jean
      1. semble
      2. seems
      1. (??[PP
      2.  
      1. à
      2. to
      1. Marie])
      2. Marie
      1. [____
      2.  
      1. avoir
      2. to.have
      1. du
      2. of
      1. talent].
      2. talent
      1. ‘Jean seems (??to Marie) to have talent.’
      2. (French; McGinnis 1998: 90, (55a, b))
    I can see at least two ways of addressing this issue. On the one hand, an intervention-based approach could be reconciled with these facts by positing variation in how raising verbs project their arguments. For instance, the underlying constituent order with raising verbs could be subject to cross-linguistic variation. Suppose that in English it is V–TP–PP (with extraposition yielding V–____i–PP–TPi) but that in French/Italian it is V–PP–TP. This would predict an intervention effect of the experiencer PP in French/Italian but not in English (assuming that extraposition of TP follows raising). On the other hand, see Bruening 2014a for arguments that the degradation seen in (59) has nothing to do with intervention. ⮭
  60. Additionally, Erik Zyman (personal communication) reports the following series of judgments to me that seem to provide additional strong support for my conjecture that intervention effects in wh movement are (or can be) determined by linear precedence rather than asymmetric c-command (thanks also to Kutay Serova for valuable discussion).
      1. (i)
      1. a.
      1.     You introduced [DP the inventor of the Xbox] [PP to [DP the promoter of the Wii]].
      1.  
      1. b.
      1.     Wait, you introduced [DP the inventor of what] [PP to [DP the promoter of what]]?!
      1.  
      1. c.
      1.     What did you introduce [DP the inventor of ____] [PP to [DP the promoter of what]]?
      1.  
      1. d.
      1. ?*What did you introduce [DP the inventor of what] [PP to [DP the promoter of ____]]?
    ⮭
  61. See Bruening 2001: 264 for related empirical observations from English that deserve further scrutiny in this context. Heck & Müller 2000, Heck & Müller 2003, and Müller 2011: chap. 3, sect. 4 also discuss the possibility of intervention without c-command in German wh questions. Their examples involve a left–right asymmetry that would seem well-suited for a precedence-based analysis. However, their examples crucially differ from those discussed in the main text by (mostly) involving defective intervention: the intervening wh phrase cannot undergo extraction due to being inside an island. A reviewer points out to me that the present analysis could be made to account for these data if we abandoned the hypothesis that certain domains (e.g., islands and spellout domains that have already been spelled out) are opaque to Merge-based probing (see footnote 11). While such an approach is certainly conceivable (see, e.g., Pesetsky 2021: 12, (20) for the proposal that phase complements are penetrable to probing but impenetrable to movement), it arguably undergenerates, as it predicts that sentences like (ia), where the leftmost wh phrase prior to wh movement is inside a subject island, should be unacceptable, contrary to fact (see Lasnik & Saito 1992: 188, n. 17, Stroik 1995: 244, and Hornstein 2009: 35, (27a) for additional examples; compare Heck 2008: 62, (102a, b), which cites examples parallel to (ia) as being unacceptable).
      1. (i)
      1. a.
      1.   Who did [pictures of who] please ____?
      1.  
      1. b.
      1. *Who did [pictures of ____] please who?
      1.  
      1. (Fiengo et al. 1988: 89, (21))
    On the other hand, these data are accounted for if (wh) probing is constrained by islands/spellout domains. Unfortunately, this leaves the German data discussed by Heck & Müller unaccounted for. ⮭
  62. The lack of superiority effect with clausemate wh phrases is often attributed to the possibility of A-scrambling of the further wh phrase over the closer one prior to wh movement to [spec, CP]. See Abels & Dayal 2023: 442 for an attempt to also derive the D-linking effect from (covert) scrambling. ⮭
  63. Another puzzle for all intervention-based analyses of superiority effects comes from examples like (i), discussed by Lasnik & Saito 1992: 118–121 (see Williams 1994: 194, Reinhart 2006: 26–34, Stroik 2009: 80–88, and Frank & Kotek 2022: 5–8 for related discussion).
      1. (i)
      1. A:
      1.   Who1 wants to know what2 who3 bought ____2?
      1.  
      1. B:
      1. *Joni does. (I.e., Joni wants to know what2 who3 bought ____2.)
      1.  
      1. B′:
      1.   Joni wants to know what Matt bought, Eli wants to know what Krishna bought, …
    Question (iA) appears to involve a Superiority violation in the embedded clause (with what2 moving to the embedded [spec, CP] over the clausemate subject who3), and yet it is acceptable, in contrast to the following.
      1. (ii)
      1. *Joni wants to know what2 who3 bought ____2.
    Importantly, (iA) is only acceptable if who3 takes matrix scope, yielding a multiple wh question with who1 (though see Chomsky 1995b: 387, n. 69, Kim 2001: 82–83, n. 5, and Reinhart 2006: 33–34 for empirical wrinkles). Thus, (iA) can be felicitously answered by (iB′) but not by (63B), as the latter would require who3 to take embedded scope with what2, triggering a superiority effect. Faced with these facts, some authors have concluded that superiority effects (and perhaps minimality effects in general) must be sensitive to interpretation—a Superiority-violating derivation is allowed if and only if it has a meaning distinct from that of its Superiority-compliant counterpart (see Fox 2012 and Kotek 2019: 47–49 for proposals along these lines). As a reviewer observes, the search algorithm in (13) makes no reference to interpretation and therefore would need to be modified to derive the alleged meaning-sensitive superiority effects. Note, however, that the inclusion of trans-derivational constraints and comparisons adds significant complexity to the grammar. Consequently, such constraints should be avoided if possible (see Müller 2021: sect. 12.2 on what makes a constraint “good” or “bad”). A purely syntactic account of Superiority-violating wh questions like (i) therefore remains a desideratum. Whatever the correct analysis is, it will not affect my analysis of intervention in passive A-movement, since interpretation arguably plays no role in that domain. ⮭
  64. While (i) may be a bit marked, it is nonetheless acceptable and, importantly, much improved over (110b).
      1. (i)
      1. ? Wherei did you eat the cookies near ____i?
    ⮭
  65. To explain why locatives cannot overtly appear between the verb and direct object (*You ate there the cookies), we could adopt a surface-oriented Case Adjacency principle along the lines of Janke & Neeleman 2012: 153, (2) (which develops ideas originally from Stowell 1981). Janke & Neeleman contend that Case Adjacency is a PF condition and show that lower copies of movement do not intervene for calculating adjacency between V and the direct object that it assigns accusative Case to (sect. 6.3). ⮭

References

Abels, Klaus. 2003. Successive cyclicity, anti-locality, and adposition stranding. Doctoral thesis. University of Connecticut.

Abels, Klaus & Dayal, Veneeta. 2023. On the syntax of multiple sluicing and what it tells us about wh-scope taking. Linguistic Inquiry 54.3.429–477.

Abels, Klaus & Grabska, Dagmara. 2022. On the distribution of scope ambiguities in Polish. Glossa 7.1.  http://doi.org/10.16995/glossa.8170.

Adger, David. 2003. Core syntax: a minimalist approach. Oxford University Press.

Alexander, D. & Kunz, W. J. 1964. Some classes of verbs in English. Indiana University Linguistics Research Project.

Al Khalaf, Eman. 2015. Coordination and linear order. Doctoral thesis. University of Delaware.

Alrenga, Peter. 2005. A sentential subject asymmetry in English and its implications for complement selection. Syntax 8.3.175–207.

Alsina, Alex. 2009. The prepositional passive as structure-sharing. In: Butt, Miriam & King, Tracy Holloway (editors). Proceedings of the LFG09 Conference. CSLI Publications. 44–64.

Anagnostopoulou, Elena. 2003. The syntax of ditransitives: evidence from clitics. De Gruyter.

Angelopoulos, Nikos & Collins, Chris & Terzi, Arhonto. 2020. Greek and English passives, and the role of by-phrases. Glossa 5.1.90.

Arregi, Karlos & Murphy, Andrew. 2022. Argument-internal parasitic gaps. Unpublished paper. Work conducted at: University of Chicago.

Atlamaz, Ümit. 2019. Agreement, case, and nominal licensing. Doctoral thesis. Rutgers University.

Aycock, Seth. 2020. A third-factor account of locality: explaining impenetrability and intervention effects with minimal search. Cambridge Occasional Papers in Linguistics 12.1.1–30.

Bach, Emmon W. 1980. In defense of passive. Linguistics and Philosophy 3.3.297–341.

Baker, Mark. 1985. Incorporation: a theory of grammatical function changing. University of Chicago Press.

Baker, Mark & Johnson, Kyle & Roberts, Ian. 1989. Passive arguments raised. Linguistic Inquiry 20.2.219–251.

Baltin, Mark & Postal, Paul M. 1996. More on reanalysis hypotheses. Linguistic Inquiry 27.1.127–145.

Barbiers, Sjef. 1995. The syntax of interpretation. Holland Academic Graphics.

Barker, Chris. 2012. Quantificational binding does not require c-command. Linguistic Inquiry 43.4.614–633.

Barss, Andrew & Lasnik, Howard. 1986. A note on anaphora and double objects. Linguistic Inquiry 17.2.347–354.

Berwick, Robert C. & Chomsky, Noam & Piattelli-Palmarini, Massimo. 2013. Poverty of the stimulus stands: why recent challenges fail. In: Piattelli-Palmarini, Massimo & Berwick, Robert C. (editors). Rich languages from poor inputs. Oxford University Press. 19–42.

Blight, Ralph. 2000. Pseudo-passives and adjacency. In: Veselinova, Ljuba & Robinson, Susan & Antieau, Lamont (editors). Proceedings of the 9th Student Conference in Linguistics (SCIL 9). MIT Working Papers in Linguistics. 75–91.

Bolinger, Dwight. 1971. The phrasal verb in English. Harvard University Press.

Bondarenko, Tatiana & Davis, Colin. 2021. Sub-extraction asymmetries and linearization in Russian. Journal of Slavic Linguistics 29, extra issue containing proceedings of 28th Formal Approaches to Slavic Linguistics workshop. https://ojs.ung.si/index.php/JSL/article/view/130.

Borer, Hagit. 2005a. In name only. Volume 1 of Structuring sense. Oxford University Press.

Borer, Hagit. 2005b. The normal course of events. Volume 2 of Structuring sense. Oxford University Press.

Bowers, John. 2002. Transitivity. Linguistic Inquiry 33.2.183–224.

Bowers, John. 2010. Arguments as relations. MIT Press.

Branan, Kenyon & Erlewine, Michael Yoshitaka. 2022. Locality and (minimal) search. Unpublished paper. Work conducted at: ZAS/University of Göttingen and National University of Singapore.

Branan, Kenyon & Erlewine, Michael Yoshitaka. 2024. Ā-probing for the closest DP. Linguistic Inquiry 55.2.375–401.

Bresnan, Joan. 1982. The passive in lexical theory. In: Bresnan, Joan (editor). The mental representation of grammatical relations. MIT Press. 3–86.

Bruening, Benjamin. 2001. QR obeys Superiority: frozen scope and ACD. Linguistic Inquiry 32.2.233–273.

Bruening, Benjamin. 2011. Pseudopassives, expletive passives, and locative inversion. Unpublished paper. Work conducted at: University of Delaware.

Bruening, Benjamin. 2014a. Defects of defective intervention. Linguistic Inquiry 45.4.707–719.

Bruening, Benjamin. 2014b. Precede-and-command revisited. Language 90.2.342–388.

Bruening, Benjamin. 2018. CPs move rightward, not leftward. Syntax 21.4.362–401.

Bruening, Benjamin. 2019. Passive do so. Natural Language and Linguistic Theory 37.1.1–49.

Bruening, Benjamin & Al Khalaf, Eman. 2020. Category mismatches in coordination revisited. Linguistic Inquiry 51.1.1–36.

Cable, Seth. 2007. The grammar of Q: Q-particles and the nature of wh-fronting, as revealed by the wh-questions of Tlingit. Doctoral thesis. Massachusetts Institute of Technology.

Cable, Seth. 2010. The grammar of Q: Q-particles, wh-movement, and pied-piping. Oxford University Press.

Chomsky, Noam. 1973. Conditions on transformations. In: Anderson, Stephen A. & Kiparsky, Paul (editors). A festschrift for Morris Halle. Holt, Rinehart and Wilson. 232–286.

Chomsky, Noam. 1981. Lectures on government and binding: the Pisa lectures. Foris Publications.

Chomsky, Noam. 1982. Some concepts and consequences of the theory of government and binding. MIT Press.

Chomsky, Noam. 1986. Barriers. MIT Press.

Chomsky, Noam. 1995a. Bare phrase structure. In: Webelhuth, Gert (editor). Government and Binding Theory and the Minimalist Program. Blackwell Publishers. 383–439.

Chomsky, Noam. 1995b. The Minimalist Program. MIT Press.

Chomsky, Noam. 2000. Minimalist inquiries: the framework. In: Martin, Roger & Michaels, David & Uriagereka, Juan (editors). Step by step: essays on minimalist syntax in honor of Howard Lasnik. MIT Press. 89–155.

Chomsky, Noam. 2001. Derivation by phase. In: Kenstowicz, Michael (editor). Ken Hale: a life in language. MIT Press. 1–52.

Chomsky, Noam. 2004. Beyond explanatory adequacy. In: Belletti, Adriana (editor). Structures and beyond. Volume 3 of The cartography of syntactic structures. Oxford University Press. 104–131.

Chomsky, Noam. 2005. Three factors in language design. Linguistic Inquiry 36.1.1–22.

Chomsky, Noam. 2008. On phases. In: Freidin, Robert & Otero, Carlos P. & Zubizarreta, Maria Luisa (editors). Foundational Issues in Linguistic Theory. MIT Press. 133–166.

Chomsky, Noam. 2013. Problems of projection. Lingua 130.33–49.

Chomsky, Noam. 2015. Problems of projection: extensions. In: Di Domenico, Elisa & Hamann, Cornelia & Matteini, Simona (editors). Structures, strategies and beyond: studies in honour of Adriana Belletti. John Benjamins Publishing Company. 3–16.

Chomsky, Noam. 2020. The UCLA lectures. Transcribed and edited by Freidin, Robert. https://ling.auf.net/lingbuzz/005485.

Chomsky, Noam. 2021. Minimalism: where are we now, and where can we hope to go. Gengo Kenkyu 160.1–41.

Chomsky, Noam & Gallego, Ángel J. & Ott, Dennis. 2019. Generative grammar and the faculty of language: insights, questions, and challenges. Catalan Journal of Linguistics, special issue. 229–261.

Chow, Keng Ji. 2022. A novel algorithm for minimal search. Snippets 42.2.3–5.

Christensen, Kirsti Koch. 1986. Complex passives, reanalysis, and word formation. Nordic Journal of Linguistics 9.135–162.

Chung, Sandra. 2017. VOS languages: some of their properties. In: Everaert, Martin & van Riemsdijk, Henk C. (editors). The Wiley Blackwell companion to syntax, second edition. Wiley Blackwell.  http://doi.org/10.1002/9781118358733.wbsyncom069.

Citko, Barbara & Emonds, Joseph Embley & Whitney, Rosemarie. 2017. Double object constructions. In: Everaert, Martin & van Riemsdijk, Henk C. (editors). The Wiley Blackwell companion to syntax, second edition. Wiley Blackwell.  http://doi.org/10.1002/9781118358733.wbsyncom109.

Collins, Chris. 2002. Eliminating labels. In: Epstein, Samuel David & Seely, T. Daniel (editors). Derivation and explanation in the Minimalist Program. Blackwell Publishers. 42–64.

Collins, Chris. 2005a. A smuggling approach to raising in English. Linguistic Inquiry 36.2.289–298.

Collins, Chris. 2005b. A smuggling approach to the passive in English. Syntax 8.2.81–120.

Collins, Chris. 2017. Merge(X,Y) = {X,Y}. In: Bauke, Leah & Blümel, Andreas (editors). Labels and roots. De Gruyter. 47–68.

Collins, Chris. 2024. Principles of argument structure: a Merge-based approach. MIT Press.

Collins, Chris & Kayne, Richard S. 2023. Towards a theory of morphology as syntax. Studies in Chinese Linguistics 44.1.1–32.

Collins, Chris & Stabler, Edward. 2016. A formalization of minimalist syntax. Syntax 19.1.43–78.

Contreras, Heles. 1984. A note on parasitic gaps. Linguistic Inquiry 15.4.698–701.

Culicover, Peter W. & Jackendoff, Ray. 2005. Simpler syntax. Oxford University Press.

Davies, William D. & Dubinsky, Stanley. 1998. Sentential subjects as complex NPs: new reasons for an old account of subjacency. In: Gruber, M. Catherine & Higgins, Derrick & Olson, Kenneth S. & Wysocki, Tamara (editors). CLS 34: the main session (the proceedings from the main session of the Chicago Linguistic Society’s thirty-fourth meeting). Chicago Linguistic Society. 83–94.

Davies, William D. & Dubinsky, Stanley. 2001. Functional architecture and the distribution of subject properties. In: Davies, William D. & Dubinsky, Stanley (editors). Objects and other subjects: grammatical functions, functional categories, and configurationality. Kluwer Academic Publishers. 247–279.

Davies, William D. & Dubinsky, Stanley. 2009. On the existence (and distribution) of sentential subjects. In: Gerdts, Donna B. & Moore, John C. & Polinsky, Maria (editors). Hypothesis A/hypothesis B: linguistic explorations in honor of David M. Perlmutter. MIT Press. 111–128.

Davis, Colin P. 2020. The linear limitations of syntactic derivations. Doctoral thesis. Massachusetts Institute of Technology.

Davis, Colin P. 2021. Possessor extraction in colloquial English: evidence for successive cyclicity and cyclic linearization. Linguistic Inquiry 52.2.291–332.

Deal, Amy Rose. 2013. Possessor raising. Linguistic Inquiry 44.3.391–432.

Deal, Amy Rose. 2015. Interaction and satisfaction in ϕ-agreement. In: Bui, Thuy & Özyıldız, Deniz (editors). NELS 45: proceedings of the forty-fifth annual meeting of the North East Linguistic Society. GLSA Publications. 179–192.

Deal, Amy Rose. 2017. External possession and possessor raising. In: Everaert, Martin & van Riemsdijk, Henk C. (editors). The Wiley Blackwell companion to syntax, second edition. Wiley Blackwell.  http://doi.org/10.1002/9781118358733.wbsyncom047.

Deal, Amy Rose. 2023. Interaction, satisfaction, and the PCC. Linguistic Inquiry 55.1.39–94.

Deal, Amy Rose. To appear. Current models of Agree. In: Crippen, James & Dechaine, Rose-Marie & Keupdjio, Hermann (editors). Move and Agree: towards a formal typology. John Benjamins Publishing Company. https://ling.auf.net/lingbuzz/006504.

De Belder, Marijke & van Craenenbroeck, Jeroen. 2015. How to merge a root. Linguistic Inquiry 46.4.625–655.

Delahunty, Gerald P. 1983. But sentential subjects do exist. Linguistic Analysis 12.4.379–398.

D’Elia, Samuel C. 2016. The spray/load and dative alternations: aligning VP structure and contextual effects. Doctoral thesis. University of Kent.

den Dikken, Marcel. 1995. Particles: on the syntax of verb-particle, triadic, and causative constructions. Oxford University Press.

Diercks, Michael. 2017. Locative inversion. In: Everaert, Martin & van Riemsdijk, Henk C. (editors). The Wiley Blackwell companion to syntax, second edition. Wiley Blackwell.  http://doi.org/10.1002/9781118358733.wbsyncom082.

Drummond, Alex. 2011. Binding phenomena within a reductionist theory of grammatical dependencies. Doctoral thesis. University of Maryland, College Park.

Drummond, Alex & Kush, Dave. 2015. “Reanalysis” is raising to object. Syntax 18.4.425–463.

Emonds, Joseph E. 1970. Root and structure-preserving transformations. Doctoral thesis. Massachusetts Institute of Technology.

Emonds, Joseph E. 1976. A transformational approach to English syntax: root, structure-preserving, and local transformations. Academic Press.

Emonds, Joseph E. 2017. Adjectival and analytic passives. In: Everaert, Martin & van Riemsdijk, Henk C. (editors). The Wiley Blackwell companion to syntax, second edition. Wiley Blackwell.  http://doi.org/10.1002/9781118358733.wbsyncom087.

Engdahl, Elisabet. 1983. Parasitic gaps. Linguistics and Philosophy 6.5–34.

Epstein, Samuel David & Kitahara, Hisatsugu & Seely, T. Daniel. 2014. Labeling by minimal search: Implications for successive-cyclic A-movement and the conception of the postulate “phase.” Linguistic Inquiry 45.3.463–481.

Epstein, Samuel David & Kitahara, Hisatsugu & Seely, T. Daniel. 2020. Unifying labeling under minimal search in “single-” and “multiple-specifier” configurations. In: Nitschke, Remo & Romero Diaz, Damian Y. & De la Cruz Sánchez, Gabriela & Powell, John W. W. (editors). Proceedings of ALC 13 (Coyote Papers 22). University of Arizona Linguistics Circle. 1–11.

Ernst, Thomas. 1994. M-command and precedence. Linguistic Inquiry 25.2.327–335.

Ershova, Ksenia. 2021. Diagnosing clause structure in a polysynthetic language: wh-agreement and parasitic gaps in West Circassian. Linguistic Inquiry 52.1.1–38.

Fanselow, Gisbert. 1990. On the nature of proper government and syntactic barriers. In: Abraham, Werner & Kosmeijer, Wim & Reuland, Eric (editors). Issues in Germanic syntax. De Gruyter. 33–48.

Fanselow, Gisbert. 1991. Minimale Syntax [Minimal syntax]. Postdoctoral (habilitation) thesis. University of Passau.

Ferguson, Kenneth Scott. 1993. Notes on the shortest move metric and object checking. Harvard Working Papers in Linguistics 3.65–80.

Ferguson, Kenneth Scott. 1996. A feature-relativized shortest move requirement. Doctoral thesis. Harvard University.

Ferguson, Kenneth Scott & Groat, Erich. 1994. Defining “shortest move.” Unpublished paper. Work conducted at: Harvard University.

Fiengo, Robert. 1980. Surface structure: the interface of autonomous components. Harvard University Press.

Fiengo, Robert & Huang, C.-T. James & Lasnik, Howard & Reinhart, Tanya. 1988. The syntax of wh-in-situ. In: Borer, Hagit (editor). WCCFL 7: the proceedings of the Seventh West Coast Conference on Formal Linguistics. CSLI Publications. 81–98.

Findlay, Jamie Y. 2016. The prepositional passive in Lexical Functional Grammar. In: Arnold, Doug & Butt, Miriam & Crysmann, Berthold & King, Tracy Holloway & Müller, Stefan (editors). Proceedings of the Joint 2016 Conference on Head-Driven Phrase Structure Grammar and Lexical Functional Grammar. CSLI Publications. 255–275.

Fox, Danny. 2012. Possible ramifications for superiority. Handout. Massachusetts Institute of Technology. http://lingphil.mit.edu/papers/fox/Superiority.pdf.

Frank, Robert & Kotek, Hadas. 2022. Arguments for top-down derivations in syntax. Proceedings of the Linguistic Society of America 7.1.5264.

Fu, Jingqi & Roeper, Thomas & Borer, Hagit. 2001. The VP within process nominals: evidence from adverbs and the VP anaphor do-so. Natural Language and Linguistic Theory 19.3.549–582.

Fukui, Naoki. 2001. Phrase structure. In: Baltin, Mark & Collins, Chris (editors). The handbook of contemporary syntactic theory. Blackwell Publishers. 374–406.

Gengel, Kirsten. 2013. Pseudogapping and ellipsis. Oxford University Press.

Grewendorf, Günther. 1988. Aspekte der deutschen Syntaxe: eine Rektions-Bindungs-Analyse [Aspects of German syntax: a Government and Binding analysis]. Gunter Narr Verlag.

Grohmann, Kleanthes K. 1997. German Superiority. Groninger Arbeiten zur Germanistischen Linguistik 40.97–107.

Grohmann, Kleanthes K. 2011. Anti-locality: too-close relations in grammar. In: Boeckx, Cedric (editor). Oxford handbook of linguistic minimalism. Oxford University Press. 260–290.

Haddican, Bill. 2007. The structural deficiency of verbal pro-forms. Linguistic Inquiry 38.3.539–547.

Haegeman, Liliane. 1984. Parasitic gaps and adverbial clauses. Journal of Linguistics 20.2.229–232.

Haider, Hubert. 1983. Connectedness effects in German. Groninger Arbeiten zur Germanistischen Linguistik 23.82–119.

Haider, Hubert. 1993. Deutsche Syntax—generativ [German syntax—generative]. Gunter Narr Verlag.

Halpert, Claire. 2019. Raising, unphased. Natural Language and Linguistic Theory 37.1.123–165.

Han, Hye Jin. 2005a. A DP/NP-shell for subject CPs. In: Cover, Rebecca T. & Kim, Yuni (editors). Proceedings of the 31st annual meeting of the Berkeley Linguistics Society: general session and parasession on prosodic variation and change. Berkeley Linguistics Society. 133–143.

Han, Hye Jin. 2005b. A nominal-shell for CPs: beyond subject CPs. In: Edwards, Rodney L. & Midtlyng, Patrick J. & Sprague, Colin L. & Stensrud, Kjersti G. (editors). CLS 41: the main session (proceedings from the main session of the forty-first meeting of the Chicago Linguistic Society). Chicago Linguistic Society. 95–109.

Hartman, Jeremy. 2012. Varieties of clausal complementation. Doctoral thesis. Massachusetts Institute of Technology.

Heck, Fabian. 2008. On pied-piping: wh-movement and beyond. De Gruyter.

Heck, Fabian & Müller, Gereon. 2000. Successive cyclicity, long-distance superiority, and local optimization. In: Billerey, Roger & Lillehaugen, Brooke Danielle (editors). Proceedings of the 19th West Coast Conference on Formal Linguistics. Cascadilla Press. 218–231.

Heck, Fabian & Müller, Gereon. 2003. Derivational optimization of wh-movement. Linguistic Analysis 33.1–2.97–148.

Heck, Fabian & Müller, Gereon. 2007. Extremely local optimization. In: Bainbridge, Erin & Agbayani, Brian (editors). Proceedings of the thirty-fourth Western Conference on Linguistics. California State University. 170–182.

Hedberg, Nancy & DeArmond, Richard C. 2009. On complements and adjuncts. Snippets 19.11–12.

Hestvik, Arild. 1986. Case theory and Norwegian impersonal constructions: subjects-object alternations in active and passive verbs. Nordic Journal of Linguistics 9.2.181–197.

Hewett, Matthew. 2025. The (lack of) Case for A-movement. Linguistic Inquiry, early access.  http://doi.org/10.1162/ling.a.59.

Heycock, Caroline. 1995. Asymmetries in reconstruction. Linguistic Inquiry 26.4.547–570.

Higgins, Francis R. 1973. On J. Emonds’s analysis of extraposition. In: Kimball, John (editor). Syntax and Semantics 2. Academic Press. 149–195.

Holmberg, Anders & Sheehan, Michelle & van der Wal, Jenneke. 2019. Movement from the double object construction is not fully symmetrical. Linguistic Inquiry 50.4.677–722.

Honda, Takahiro. 2012. The syntax of passive constructions. Doctoral thesis. Osaka University.

Hornstein, Norbert. 2009. A theory of syntax: minimal operations and universal grammar. Cambridge University Press.

Hornstein, Norbert & Weinberg, Amy. 1981. Case theory and preposition stranding. Linguistic Inquiry 12.1.55–91.

Houser, Michael John. 2010. The syntax and semantics of do so anaphora. Doctoral thesis. University of California, Berkeley.

Huang, C.-T. James. 1993. Reconstruction and the structure of VP: some theoretical consequences. Linguistic Inquiry 24.1.103–138.

Jackendoff, Ray S. 1972. Semantic interpretation in generative grammar. MIT Press.

Jackendoff, Ray S. 1990. On Larson’s treatment of the double object construction. Linguistic Inquiry 21.3.427–456.

Jackendoff, Ray S. & Culicover, Peter. 1971. A reconsideration of dative movements. Foundations of Language 7.397–412.

Jaeggli, Osvaldo. 1986. Passive. Linguistic Inquiry 17.4.587–622.

Janke, Vikki & Neeleman, Ad. 2012. Ascending and descending VPs in English. Linguistic Inquiry 43.1.151–190.

Johnson, Kyle. 1996. When verb phrases go missing. Glot International 2.5.75–103.

Johnson, Kyle. 2004. Introduction to transformational grammar. Unpublished paper. Work conducted at: University of Massachusetts, Amherst.

Kageyama, Taro & Ura, Hiroyuki. 2002. Peculiar passives as individual-level predicates. Gengo Kenkyu 122.181–199.

Kayne, Richard S. 1984. Connectedness and binary branching. Foris Publications.

Kayne, Richard S. 1994. The antisymmetry of syntax. MIT Press.

Kayne, Richard S. 2011. Why are there no directionality parameters? In: Washburn, Mary Byram & McKinney-Bock, Katherine & Varis, Erika & Sawyer, Ann & Tomaszewicz, Barbara (editors). Proceedings of the 28th West Coast Conference on Formal Linguistics. Cascadilla Proceedings Project. 1–23.

Kayne, Richard S. 2022. Antisymmetry and externalization. Studies in Chinese Linguistics 43.1.1–20.

Ke, Alan Hezao. 2019. The syntax, semantics and processing of agreement and binding grammatical illusions. Doctoral thesis. University of Michigan.

Ke, Alan Hezao. 2023. Search downward: Minimal Search-based Agree. Glossa 8.1.  http://doi.org/10.16995/glossa.6526.

Ke, Alan Hezao. 2024. Can Agree and Labeling be reduced to Minimal Search? Linguistic Inquiry 55.4.849–870.  http://doi.org/10.1162/ling_a_00481.

Kim, Soowon. 2001. Chain composition and uniformity. Natural Language and Linguistic Theory 19.1.67–107.

Kitahara, Hisatsugu. 1997. Elementary operations and optimal derivations. MIT Press.

Ko, Heejeong. 2007. Asymmetries in scrambling and cyclic linearization. Linguistic Inquiry 38.1.49–83.

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 University Press.

Koster, Jan. 1978. Why subject sentences don’t exist. In: Keyser, Samuel Jay (editor). Recent transformational studies in European languages. MIT Press. 53–64.

Kotek, Hadas. 2019. Composing questions. MIT Press.

Kratzer, Angelika. 1996. Severing the external argument from its verb. In: Rooryck, Johan & Zaring, Laurie (editors). Phrase structure and the lexicon. Kluwer Academic Publishers. 109–137.

Krivochen, Diego Gabriel. 2023. The search for Minimal Search: a graph-theoretic approach. Biolinguistics 17.e9793.

Kuno, Susumu. 1973. Constraints on internal clauses and sentential subjects. Linguistic Inquiry 4.3.363–385.

Kuno, Susumu. 1981. The syntax of comparative clauses. In: Hendrick, Roberta A. & Masek, Carrie S. & Miller, Mary Frances (editors). Papers from the seventeenth regional meeting of the Chicago Linguistic Society. Chicago Linguistic Society. 136–155.

Kuno, Susumu & Robinson, Jane J. 1972. Multiple wh questions. Linguistic Inquiry 3.4.463–487.

Lakoff, George & Ross, John Robert. 1966. Criterion for verb phrase constituency. In: Oettinger, Anthony G. (editor). Mathematical linguistics and automatic translation: report no. NSF-17 to the National Science Foundation. Computation Laboratory, Harvard University. 2.1–11.

Landau, Idan. 2007. Constraints on partial VP-fronting. Syntax 10.2.127–164.

Langacker, Ronald. 1969. On pronominalization and the chain of command. In: Reibel, David A. & Schane, Sanford A. (editors). Modern studies in English. Prentice-Hall. 160–186.

Langendoen, D. Terrence & Pullum, Geoffrey K. 1977. Preposition stranding in English: a problem and a mystery. In: Fox, Samuel E. & Beach, Woodford A. & Philosoph, Shulamith (editors). CLS book of squibs/cumulative index 1968-1977. Chicago Linguistic Society. 64–65.

Larson, Bradley. 2015. Minimal search as a restriction on Merge. Lingua 156.57–69.

Larson, Richard K. 1988. On the double object construction. Linguistic Inquiry 19.3.335–391.

Larson, Richard K. 1990. Double objects revisited: Reply to Jackendoff. Linguistic Inquiry 21.589–632.

Larson, Richard K. 2014. On shell structure. Routledge.

Larson, Richard K. 2024. VP-preposing and constituency “paradox.” Linguistic Inquiry 55.4.659–695.  http://doi.org/10.1162/ling_a_00485.

Lasnik, Howard. 1976. Remarks on coreference. Linguistic Analysis 2.1.1–22.

Lasnik, Howard. 1999. On feature strength: three minimalist approaches to overt movement. Linguistic Inquiry 30.2.197–217.

Lasnik, Howard & Saito, Mamoru. 1991. On the subject of infinitives. In: Dobrin, Lise M. & Nichols, Lynn & Rodriguez, Rosa M. (editors). CLS 27: papers from the 27th regional meeting of the Chicago Linguistic Society. Chicago Linguistic Society. 1.324–343.

Lasnik, Howard & Saito, Mamoru. 1992. Move α: conditions on its application and output. MIT Press.

Lebowski, Zach. 2021. CED effects in ascending and descending structures: evidence from extraction asymmetries. Unpublished paper. Work conducted at: University of Chicago.

Lechner, Winfried. 2003. Phrase structure paradoxes, movement and ellipsis. In: Schwabe, Kerstin & Winkler, Susanne (editors). The interfaces: deriving and interpreting omitted structures. John Benjamins Publishing Company. 177–203.

Lechner, Winfried. 2004. Extending and reducing the MLC. In: Stepanov, Arthur & Fanselow, Gisbert & Vogel, Ralf (editors). Minimality effects in syntax. De Gruyter. 205–240.

Lees, Robert B. 1960. The grammar of English nominalizations. Mouton.

Lohndal, Terje. 2012. Without specifiers: phrase structure and events. Doctoral thesis. University of Maryland, College Park.

Longenbaugh, Nicholas. 2019. On expletives and the agreement-movement correlation. Doctoral thesis. Massachusetts Institute of Technology.

May, Robert. 1985. Logical form: its structure and derivation. MIT Press.

McCawley, James D. 1998. The syntactic phenomena of English, second edition. University of Chicago Press.

McGinnis, Martha. 1998. Locality in A-movement. Doctoral thesis. Massachusetts Institute of Technology.

McGinnis, Martha. 2001. Variation in the phase structure of applicatives. Linguistic Variation Yearbook 1.1.105–146.

McInnerney, Andrew. 2021. Reevaluating the reanalysis analysis of prepositional passives. In: Hewett, Matthew & Kasper, Corinne & Kim, Sanghee & Kurtz, Naomi (editors). CLS 56: proceedings of the fifty-sixth annual meeting of the Chicago Linguistic Society. Chicago Linguistic Society. 321–333.

McInnerney, Andrew. 2022a. Against the argument/adjunct distinction. University of Pennsylvania Working Papers in Linguistics 28.1.1–10.

McInnerney, Andrew. 2022b. The argument/adjunct distinction and the structure of prepositional phrases. Doctoral thesis. University of Michigan.

McInnerney, Andrew. 2023. The argument/adjunct distinction does not condition islandhood of PPs in English. Linguistic Inquiry, early access.  http://doi.org/10.1162/ling_a_00511.

McInnerney, Andrew. 2025. Separability of dependents from VP in English: beyond the argument/adjunct distinction. Journal of Linguistics 61.1.137–167.  http://doi.org/10.1017/S0022226724000069.

Merchant, Jason. 2013. Voice and ellipsis. Linguistic Inquiry 44.1.77–108.

Merchant, Jason. 2014. Some definitions. Unpublished paper. Work conducted at: University of Chicago.

Merchant, Jason. 2019. Roots don’t select, categorial heads do: lexical-selection of PPs may vary by category. Linguistic Review 36.3.325–341.

Milway, Daniel. 2023. A formalization of Agree as a derivational operation. Biolinguistics 17.e9877.  http://doi.org/10.5964/bioling.9877.

Mita, Ryohei. 2009. On tritransitive verbs. In: Askedal, John Ole & Roberts, Ian & Matsushita, Tomonori & Hasegawa, Hiroshi (editors). Germanic languages and linguistic universals. John Benjamins Publishing Company. 121–142.

Momma, Shota & Phillips, Colin. 2018. The relationship between parsing and generation. Annual Review of Linguistics 4.1.233–254.

Moulton, Keir. 2015. CPs: copies and compositionality. Linguistic Inquiry 46.2.305–342.

Müller, Gereon. 1998. Incomplete category fronting: a derivational approach to remnant movement in German. Kluwer Academic Publishers.

Müller, Gereon. 2011. Constraints on displacement: a phase-based approach. John Benjamins Publishing Company.

Müller, Gereon. 2021. Constraints on grammatical dependencies. In: Allott, Nicholas & Lohndal, Terje & Rey, Georges (editors). A companion to Chomsky. Wiley Blackwell. 190–209.

Neeleman, Ad & Payne, Amanda. 2020. On matrix-clause intervention in accusative-and-infinitive constructions. Syntax 23.1.1–41.

Neeleman, Ad & van de Koot, Hans. 2022. PP-extraposition and precedence. Linguistic Inquiry 53.1.182–198.

Newman, Elise. 2020. Facilitator effects in middles and more. Glossa 5.1.62.

Newman, Elise. 2021. The (in)distinction between wh-movement and c-selection. Doctoral thesis. Massachusetts Institute of Technology.

Nissenbaum, Jonathan W. 2000. Investigations of covert phrase movement. Doctoral thesis. Massachusetts Institute of Technology.

Overfelt, Jason. 2015. Rightward movement: a study in locality. Doctoral thesis. University of Massachusetts Amherst.

Pesetsky, David. 1987. Wh-in-situ: movement and unselective binding. In: Reuland, Eric J. & ter Meulen, Alice G. B. (editors). The representation of (in)definiteness. MIT Press. 98–129.

Pesetsky, David. 1995. Zero syntax: experiencers and cascades. MIT Press.

Pesetsky, David. 2000. Phrasal movement and its kin. MIT Press.

Pesetsky, David. 2021. Exfoliation: towards a derivational theory of clause size; version 3.1. Unpublished paper. Work conducted at: Massachusetts Institute of Technology. https://ling.auf.net/lingbuzz/004440.

Phillips, Colin. 1996. Order and structure. Doctoral thesis. Massachusetts Institute of Technology.

Phillips, Colin. 2003. Linear order and constituency. Linguistic Inquiry 34.1.37–90.

Postal, Paul M. 1986. Studies of passive clauses. State University of New York Press.

Postal, Paul M. 2004. Skeptical linguistic essays. Oxford University Press.

Postal, Paul M. 2010. Edge-based clausal syntax: a study of (mostly) English object structure. MIT Press.

Preminger, Omer. 2014. Agreement and its failures. MIT Press.

Preminger, Omer. 2019. What the PCC tells us about “abstract” agreement, head movement, and locality. Glossa 4.1.13.

Pylkkänen, Liina. 2008. Introducing arguments. MIT Press.

Rackowski, Andrea & Travis, Lisa. 2000. V-initial languages: X or XP movement and adverbial placement. In: Carnie, Andrew & Guilfoyle, Eithne (editors). The Syntax of verb initial languages. Oxford University Press. 117–141.

Rauber, Ross. 2019. (To me) it seems (to me) (like a mess) (to me) (like a mess). Unpublished paper. Work conducted at: University of Chicago.

Reinhart, Tanya. 1976. The syntactic domain of anaphora. Doctoral thesis. Massachusetts Institute of Technology.

Reinhart, Tanya. 2006. Interface strategies: optimal and costly computations. MIT Press.

Richards, Marc. 2019. Problems of “Problems of projection”: Breaking a conceptual tie. Catalan Journal of Linguistics, special issue. 139–152.

Richards, Norvin. 2014. A-bar movement. In: Carnie, Andrew & Sato, Yosuke & Siddiqi, Daniel (editors). The Routledge handbook of syntax. Routledge. 167–191.

Richards, Norvin. 2017. Multiple case assignment and the English pseudo-passive. In: LaCara, Nicholas & Moulton, Keir & Tessier, Anne-Michelle (editors). A schrift to fest Kyle Johnson. Linguistics Open Access Publications. 313–320.

Rizzi, Luigi. 1986. On chain formation. In: Borer, Hagit (editor). The syntax of pronominal clitics. Academic Press. 65–95.

Rizzi, Luigi. 1990. Relativized Minimality. MIT Press.

Rizzi, Luigi. 2013. Locality. Lingua 130.169–186.

Roberts, Ian. 2019. Parameter hierarchies and Universal Grammar. Oxford University Press.

Rosenbaum, Peter. 1967. The grammar of English predicate complement constructions. MIT Press.

Ross, John Robert. 1967. Constraints on variables in syntax. Doctoral thesis. Massachusetts Institute of Technology.

Roversi, Giovanni. 2020. How to satisfy probes: person/number hierarchy effects in Äiwoo. In: Asatryan, Mariam & Song, Yixiao & Whitmal, Ayana (editors). NELS 50: proceedings of the fiftieth annual meeting of the North East Linguistic Society. GLSA Publications. 3.99–112.

Safir, Ken. 1987. The anti-c-command condition on parasitic gaps. Linguistic Inquiry 18.4.678–683.

Sag, Ivan A. 1976. Deletion and logical form. Doctoral thesis. Massachusetts Institute of Technology.

Sag, Ivan A. & Gazdar, Gerald & Wasow, Thomas & Weisler, Steven. 1985. Coordination and how to distinguish categories. Natural Language and Linguistic Theory 3.2.117–171.

Saito, Mamoru & Fukui, Naoki. 1998. Order in phrase structure and movement. Linguistic Inquiry 29.3.439–474.

Scott, Tessa. 2021. Formalizing two types of mixed A/Ā-movement. Unpublished paper. Work conducted at: University of California, Berkeley. https://lingbuzz.net/lingbuzz/005874.

Scott, Tessa. 2023. Pronouns and agreement in San Juan Atitán Mam. Doctoral thesis. University of California, Berkeley.

Shan, Chung-chieh & Barker, Chris. 2006. Explaining crossover and superiority as left-to-right evaluation. Linguistics and Philosophy 29.1.91–134.

Stowell, Timothy Angus. 1981. Origins of phrase structure. Doctoral thesis. Massachusetts Institute of Technology.

Stroik, Thomas S. 1995. Some remarks on superiority effects. Lingua 95.4.239–258.

Stroik, Thomas S. 2009. Locality in minimalist syntax. MIT Press.

Takahashi, Shoichi. 2010. The hidden side of clausal complements. Natural Language and Linguistic Theory 28.2.343–380.

Takami, Ken-ichi. 1992. Preposition stranding: from syntactic to functional analyses. De Gruyter.

Taraldsen, Knut Tarald. 1981. The theoretical interpretation of a class of marked extractions. In: Belletti, Adriana & Brandi, Luciana & Rizzi, Luigi (editors). Theory of markedness in generative grammar: proceedings of the 1979 GLOW conference. Scuola Normale Superiore di Pisa. 475–516.

Therrien, Ray. 2023. Anti-locality and preposition stranding in a variety of Ontario French. Doctoral thesis. University of Ottawa.

Thoms, Gary & Walkden, George. 2019. vP-fronting with and without remnant movement. Journal of Linguistics 55.1.161–214.

Truswell, Robert. 2008. Preposition stranding, passivization, and extraction from adjuncts in Germanic. Linguistic Variation Yearbook 8.131–178.

Truswell, Robert. 2011. Events, phrases, and questions. Oxford University Press.

Tseng, Jesse. 2006. English prepositional passives in HPSG. In: Wintner, Shuly (editor). Proceedings of FG 2006: the 11th conference on Formal Grammar. Center for the Study of Language and Information. 147–159.

Ura, Hiroyuki. 1996. Multiple feature-checking: a theory of grammatical function splitting. Doctoral thesis. Massachusetts Institute of Technology.

van Riemsdijk, Henk. 1978. A case study in syntactic markedness: the binding nature of prepositional phrases. Peter de Ridder Press.

Williams, Edwin. 1994. Thematic structure in syntax. MIT Press.

Wiltschko, Martina. 1997. D-linking, scrambling and superiority in German. Groninger Arbeiten zur Germanistischen Linguistik 41.107–142.

Zwart, Jan-Wouter. 2011. Structure and order: asymmetric Merge. In: Boeckx, Cedric (editor). The Oxford handbook of linguistic minimalism. Oxford University Press. 96–118.

Zyman, Erik. 2017. P’urhepecha hyperraising to object: an argument for purely altruistic movement. Proceedings of the Linguistic Society of America 2.53.

Zyman, Erik. 2022. Proleptic PPs are arguments: consequences for the argument/adjunct distinction and for selectional switch. Linguistic Review 39.1.129–158.

Zyman, Erik. 2023. On the symmetry between Merge and Adjoin. Unpublished paper. Work conducted at: University of Chicago. https://drive.google.com/file/d/1Bgc6k3_tS-bmFo58Oko2U1i_fmOOrWKA/view.

Zyman, Erik. 2024. On the definition of Merge. Syntax, early access.  http://doi.org/10.1111/synt.12287.

Zyman, Erik & Kalivoda, Nick. 2020. XP-and X0-movement in the Latin verb: evidence from mirroring and anti-mirroring. Glossa 5.1.20.