Abstract

Some languages allow extraction of possessors from only a subset of nominals. I show that a juxtaposition of two proposals about Case and Agree (made by Rackowski and Richards (2005) and Bobaljik (2008)) correctly predicts these crosslinguistic restrictions on possessor extraction.

1 Introduction

This article deals with Ā-extraction of possessors and nonarguments from DP, where the path of extraction crosses a DP boundary. Many languages allow this sort of extraction, but for introductory purposes I will use Sidaama. As shown by the contrast between (1a) and (1b), the grammaticality of such extractions depends on the status of the DP boundary that is crossed. In (1a), we see that the possessor of a direct object may undergo Ā-extraction in Sidaama, but in (1b) we see that the possessor of an oblique may not.

(1) Nonoblique/Oblique contrast in Sidaama

  • a.

    [Bule [amá ___ ] la’-’ino ] manč-i

    Bule[NOM.F] mother.ACC see-3SG.F-P.PERF.3 person-NOM.MOD.M

    da-Ø-i.

    come-3SG.M-S.PERF.3SG.M

    ‘The man whose mother Bule saw came.’

  • b.

    *[Ise [miní-nni ___ ] ful-t-ino] mančó

    3SG.F.NOM house-GEN.M.ABL exit-3SG.F-P.PERF.1 person.ACC

    af-oo-mm-o.

    get.to.know-P.PERF.1-1SG-M

    ‘I know the person whose house she exited from.’

    (Kawachi 2007b:651, 667)

This asymmetry cannot be attributed to a more general inaccessibility of obliques to Ā-extraction processes, since, as we see in (2), Sidaama freely allows Ā-movement of obliques.

(2) Sidaama allows relativization of obliques

  • a.

    Ise hakk’iččó meesané-te-nni mur-t-anno.

    3SG.F.NOM tree(ACC) axe-GEN.F-INST cut-3SG.F-IMPF.3

    ‘She (habitually) cuts the tree with the axe.’

  • b.

    Isi [ise hakk’iččó ___ mur-t-anno ] meesané hiikk’-Ø-i.

    3SG.M.NOM 3SG.F.NOM tree(ACC) cut-3SG.F-IMPF.3 axe(ACC) break

    ‘He broke the axe with which she habitually cut the tree.’

    (Kawachi 2007b:639)

As I will show, such asymmetries are not specific to Sidaama. Rather, there is a general dispreference for extraction of nonarguments and possessors from DP when the DP in question is marked. I will show that English, Tagalog, Niuean, Inuktitut, Tzotzil, Modern Greek, and Kaqchikel all exhibit the same restriction that Sidaama does.

I will argue for a particular reason behind this crosslinguistic ban on these sorts of fillergap dependencies, given in (3).1

(3) Extraction generalization

Extraction of nonarguments from DP cannot take place from a phrase that is not targeted for Agree.

Furthermore, I will show that this ban can be derived straightforwardly from the juxtaposition of two independent proposals: a modified version of Rackowski and Richards’s (2005) proposal that Agree with a given domain is necessary to extract from that domain, and Bobaljik’s (2008) proposal that the possibility of Agree with a given domain is determined by the Case that the domain bears.

A tree illustrating the sequence of operations that must take place in derivations involving this sort of extraction is given in (4).

(4)

graphic

First, v probes for φ-features. It finds the DP that contains the Ā-mover and unlocks the DP for further probing (regardless of the presence or absence of an Ā-goal within the DP). Next, v probes for [wh]-features. The DP that has been unlocked no longer intervenes for this step of probing. v is thus able to Agree with the [wh]-bearing DP contained in the unlocked DP. As a result of this second Agree relationship, Ā-movement takes place. If the first step in (4) fails, the second and third steps will also fail. Crucially for the account, the Case that a DP bears determines whether or not it may be a goal for Agree. If that DP is not targeted for Agree, because it bears the wrong Case, it will not be able to be unlocked, so extraction from it cannot take place. Thereby, the generalization in (3) is derived.

In sections 2 and 3, I will provide arguments for the assumptions that my account is contingent on. I will first show that Agree with a DP is necessary for extraction of nonarguments from that DP to take place and give a version of Rackowski and Richards’s (2005) proposal to account for this. I will then show that Agree with a DP is contingent on the Case that DP bears and present Bobaljik’s (2008) account of these facts. In section 4, I will flesh out the account sketched in (4) and show how it handles languages that show the contrast in (1). In sections 5 and 6, I will show that the account makes a number of correct additional predictions. First, the account correctly predicts that extraction of nonarguments from DP does not make a distinction between subjects and objects in NOM-ACC languages, given a version of Case assignment that takes place in the syntax, following Preminger (2009) and Levin (2017). Second, the account correctly predicts that extraction of nonarguments or possessors from DP does make a distinction between subjects and objects in ERG-ABS languages; this distinction is determined by Bobaljik’s (2008) notion of accessibility.

2 Rackowski and Richards 2005 and the Relationship between Agree and Extraction

At the outset, I conjectured that extraction of nonarguments and possessors from DP requires Agree with the DP in question. In this section, I will give the first “moving part” involved in the account I propose. Rackowski and Richards (2005) propose that extraction that crosses a locality domain is contingent on Agree with the locality domain itself. I will present their original data, then show that it extends straightforwardly to a number of instances of extraction of a nonargument or possessor from DP. Then, I will reformulate their proposal.

2.1 Agree-Contingent Extraction

Rackowski and Richards (2005) argue that Agree with a given phrase determines whether or not that phrase may be extracted from, using data primarily from Tagalog. Tagalog instantiates an Austronesian voice system. Rackowski (2002) argues that, at least in Tagalog, this reflects that certain DPs are being targeted for Agree to the exclusion of others. The morphological reflexes of this Agree relationship can be seen in two places: the verbal complex reflects a certain form, determined by the Case of the DP targeted for Agree, and the DP targeted for Agree bears ang marking. For instance, as we see in (5), the agreement morphology of a matrix clause is variable.

(5) Variable matrix agreement in Tagalog

  • a.

    M-agsa-sabi ang kalabaw na masarap ang bulaklak.

    NOM-ASP-say ANG water.buffalo that delicious ANG flower

    ‘The water buffalo will say that the flower is delicious.’

  • b.

    Sa-sabih-in ng kalabaw na masarap ang bulaklak.

    ASP-say-ACCCS water.buffalo that delicious ANG flower

    ‘A/The water buffalo will say that the flower is delicious.’

    (Rackowski and Richards 2005:586)

In (5a), matrix verbal agreement is controlled by the nominal (ang) kalabaw, marked by m-. In (5b), matrix verbal agreement is controlled by the embedded clause, marked on the matrix verb by -in. Different agreement patterns in Tagalog also correlate with alternations in case morphology on local arguments.

Rackowski and Richards (2005) note that long-distance extraction is possible only if the clause that contains the gap is also the controller of agreement. The direct object, a clausal complement in (6), must control verbal agreement morphology.

(6) Matrix agreement determines extractability in Tagalog

  • a.

    Ang=kotse mo ay sinabi ni Pedro [na ninakaw ni Josie ___ ].

    ANG=car your INVPRFV.say.ACCCS Pedro that PRFV.ACC-steal CS Josie

    ‘Your car, Pedro said that Josie stole.’

  • b.

    *Ang=kotse mo ay nagsabi si Pedro [na ninakaw ni Josie ___ ].

    ANG=car your INVPRFV.say.NOMCS Pedro that PRFV.ACC-steal CS Josie

    ‘Your car, Pedro said that Josie stole.’

    (Kroeger 1993:218)

In (6a), the CP that contains the gap also controls agreement morphology in the matrix verb. In the unacceptable (6b), the CP that contains the gap does not control agreement morphology in the matrix verb. Here, we see that agreement clearly correlates with the possibility of extraction. To access the Ā-moved element in Tagalog, the clause containing the mover must be targeted for Agree. Henceforth, I will use the term unlocking to refer to this effect of Agree on a given phrase.

Interestingly, and not noted in Rackowski and Richards 2005, extraction of a nonargument or possessor from DP in Tagalog is subject to the same requirement. DPs, like CPs, must be unlocked in Tagalog if they are to contain a gap. As we see in (7a–b), Tagalog allows extraction from DP from agents and objects. However, as we see in (7c–d), this extraction is subject to a familiar constraint: the DP that contains the gap must control agreement morphology. In (7a–b), the gap-containing DPs have been targeted by Agree and are unlocked, allowing extraction of their possessors. In contrast, the gap-containing DPs in (7c–d) have not been targeted by Agree and are not unlocked. As a result, extraction of their possessors is blocked.

(7) Tagalog possessor extraction requires the DP containing the gap to control agreement

  • a.

    Sino ang b-in-ili ni Juan ang ___ kotse?

    who ANGPRFV.buy.ACCCS Juan ANG car

    ‘Whose car did Juan buy?’ (Lit.: Whose did Juan buy ___ car?)

  • b.

    Sino ang b-um-ili ang ___ nanay ng kotse?

    who ANGPRFV.buy.NOMANG mother CS car

    ‘Whose mother bought the car?’

  • c.

    *Sino ang b-um-ili si Juan ng ___ kotse?

    who ANGPRFV.buy.NOMANG Juan CS car

    ‘Whose car did Juan buy?’ (Lit.: Whose did Juan buy ___ car?)

  • d.

    *Sino ang b-in-ili ng ___ nanay ang kotse?

    who ANGPRFV.buy.ACCCS mother ANG car

    ‘Whose mother bought the car?’

    (Nakamura 1996:86)

For extraction of a nonargument or possessor from DP to take place, it must be preceded by unlocking. I will assume, following Rackowski (2002) and Matushansky (2002), that DP is a phase. Following Van Urk (2015), I will assume that v may bear both φ- and [wh]-features, which may probe sequentially. An annoted tree schematizing the structure in question and the series of necessary operations is given in (4).

2.2 More Motivation for the Extension of the Analysis

I have just proposed that φ-agreement is necessary for unlocking, and that unlocking is necessary for extraction of a nonargument or possessor to take place from DP. When φ-agreement fails, unlocking does not take place, and extraction of a nonargument or possessor cannot take place from DP. Thus, we expect these sorts of extractions to be generally contingent on the successful establishment of an Agree relationship for φ-features. Here, I will examine two unrelated languages, Northern Ostyak and Blackfoot, which show clearly that this is the case. Both languages allow extraction of possessors or quantificational elements from subjects, which trigger obligatory agreement. In addition, both languages have optional agreement with direct objects, but, crucially, extraction from direct objects is possible only when they control agreement.

(8) shows an instance of extraction from DP taking place from a subject in Northern Ostyak. The subject in Northern Ostyak obligatorily controls agreement morphology in the verbal complex.

(8) Extraction from NOM DP in Northern Ostyak

  • Imi ijolti [lik-dl ___ ] et-dl nawdrnin pela.

  • woman always anger-3SG come-T.3SGS frog to

  • ‘This woman, her anger always comes to the frog.’

  • (Nikolaeva 1999a:346)

We expect (8) to be grammatical. The subject has been targeted for Agree and is thus unlocked. Therefore, Ā-movement out of the subject is expected to be good.

Blackfoot shows the same possibility of extraction from an obligatorily agreeing subject.

(9) Extraction from NOM DP in Blackfoot

  • a.

    A-ipaawanii-yi om-iksi p’kssii-iksi.

    IMPF-fly.AI-3PLDEM-PL bird-PL

    ‘Those birds are flying.’

  • b.

    Om-iksi a-ipaawanii-yi p’kssii-iksi.

    DEM-PLIMPF-fly.AI-3PL bird-PL

    ‘Those birds are flying.’

    (Bliss 2012:4)

In (9a), we see that the subject controls verbal agreement morphology. In (9b), we see that a demonstrative construed with the subject may be extracted to the front of the sentence. The subject has been targeted for Agree and thus unlocked. As a result, Ā-movement out of the subject is expected to be allowed.

As noted above, Northern Ostyak and Blackfoot both optionally allow agreement with direct objects. There are two paradigms of agreement morphology in Northern Ostyak, one that crossreferences the subject and another that cross-references both the subject and the direct object. In (10a), only the φ-features of the subject are realized as agreement morphology. In (10b), the φ-features of both the subject and the direct object are realized in the agreement morphology.2

(10) Optional object agreement in Northern Ostyak

  • a.

    Ma tǎm kǎlan wel-s-әm.

    I this reindeer kill-T-1SG

    ‘I killed this reindeer.’

  • b.

    Ma tǎm kǎlan wel-s-e:m.

    I this reindeer kill-T-1SG.3OBJ

    ‘I killed this reindeer.’

    (Nikolaeva 1999b:64)

Blackfoot also has a process of optional agreement with the direct object. There are a number of different verbal paradigms in Blackfoot. When the “animate intransitive” paradigm is instantiated, as in (11a), only the φ-features of the external argument are reflected in the verbal complex.3 In contrast, when the “transitive animate” paradigm is instantiated, as in (11b), the φ-features of the external argument as well as the direct object are reflected.

(11) Optional object agreement in Blackfoot

  • a.

    Nit-ii-yaapi nato’kami piitaa-iksi.

    1-IC-see.AI two eagle-PL

    ‘I saw two eagles.’

  • b.

    Nit-ii-ino-a-yi nato’kami piitaa-iksi.

    1-IC-see.TA-DIR-3PL two eagle-PL

    ‘I saw two eagles.’

    (Bliss 2012:9)

We have just seen that both Northern Ostyak and Blackfoot have optional processes of agreement with direct objects. Both languages also allow extraction of structurally high elements in DP. I have shown that these sorts of extraction are contingent on unlocking of the phrase that contains the relevant gap, which requires Agree for φ-features with the gap-containing phrase. As a result, we expect that the normally optional process of agreement with direct objects should be obligatory if the direct object contains a gap. As we see in (12)–(13), this prediction is borne out.

(12) Extraction from DP requires object agreement in Northern Ostyak

  • a.

    Juwan motta [xot-әl ___ ] kǎśalә-s-e:m.

    John before house-3SG see-T-1SG.3OBJ

    ‘I saw John’s house before.’

  • b.

    *Juwan motta [xot-әl ___ ] kǎśalә-s-ә:m.

    John before house-3SG see-T-1SG

    ‘I saw John’s house before.’

    (Nikolaeva 1999b:67)

(13) Extraction from DP requires object agreement in Blackfoot

  • a.

    Nato’kami nit-ii-ino-a-yi piitaa-iksi.

    two 1-IC-see.TA-DIR-3PL eagle-PL

    ‘I saw two eagles.’

  • b.

    *Nato’kami nit-ii-yaapi piitaa-iksi.

    two 1-IC-see.AI eagle-PL

    ‘I saw two eagles.’

    (Bliss 2012:9)

In (12a) and (13a), we see that extraction from the object DP is grammatical if that DP controls verbal agreement morphology. In contrast, in (12b) and (13b), we see that extraction from the object DP becomes ungrammatical if that DP does not control verbal agreement morphology. This is the same process that we saw in Tagalog, where an otherwise optional agreement controller must control agreement when it contains a gap. If Agree does not occur, the DP is not unlocked, so extraction of a nonargument or possessor from that DP cannot occur. In other words, what we have seen in this section is that extraction of nonarguments or possessors from DP is subject to a particular restriction: namely, this sort of extraction is possible only if the DP that extraction takes place from is subject to Agree. This provides a piece of the solution to the puzzle of extraction from obliques: presumably, if oblique DPs cannot be targeted for Agree, then it should not be possible for nonarguments or possessors to undergo extraction from them. However, so far we do not have an independently motivated mechanism that would prevent oblique DPs from being targeted for Agree. In the following section, I will introduce such a mechanism.

3 Bobaljik 2008 and the Relationship between Case and Agree

In the previous section, I showed that Agree is a precondition for extraction of nonarguments from DP. A DP must be the goal of Agree for a nonargument to be extracted from it. In this section, I will give data from Bobaljik 2008 showing that only a subset of DPs can be licit goals for Agree, determined by a language-invariant Accessibility Hierarchy. When juxtaposed with unlocking, in the Rackowski and Richards (2005) sense, this makes an interesting prediction: extraction of nonarguments from DP should be grammatical only if the DP is accessible for Agree in that language.

Bobaljik (2008) argues that Agree is sensitive to Case and that the ability of a goal to serve as a possible goal for Agree depends on the Case that it bears. He divides Case into three categories: unmarked Case, dependent Case, and lexical Case. Unmarked Case maps to absolutive (ABS) and nominative (NOM) cases; it is always available, and, in many languages, it is morphologically the least marked case form. Dependent Case maps to ergative (ERG) and accusative (ACC) cases—it is not always available, and it is assigned only when there are two or more nominals in a sufficiently local c-command relationship. DPs marked with dependent Case require the presence of a sufficiently local DP bearing unmarked Case. Finally, lexical Case is assigned idiosyncratically by certain verbs, prepositions, or other functional heads such as applicatives, to their sisters.

In addition, Bobaljik (2008) argues that Case interacts with Agree, proposing that there is an implicational hierarchical ordering of Case, shown in (14). In a given language, a DP may or may not control verbal agreement morphology depending on the Case that it bears, independent of the position that it occupies in the syntactic structure.

(14) Accessibility Hierarchy

Unmarked Case > Dependent Case > Lexical Case

Languages differ on how much of the hierarchy is accessible. Crucially, since the Accessibility Hierarchy is implicational, we expect only certain patterns of agreement crosslinguistically. Possible ranges of accessibility are given in (15).

(15)

graphic

The expectation is that only these patterns of accessibility exist, and this expectation appears to be borne out: in a survey of 100 languages, Bobaljik (2008) finds no language that has patterns of agreement other than the four in (15). For instance, there are no languages in the survey that allow agreement only with ergative DPs, or with absolutive and dative DPs to the exclusion of ergative DPs.

Icelandic is argued to allow only DPs with unmarked Case to control agreement. As shown in (16a), this holds even when the DP bearing unmarked Case is not in subject position.

(16)

graphic

Bobaljik (2008) attributes this to the Accessibility Hierarchy in (16b). What this shows is that Icelandic only allows DPs with unmarked Case to be targeted for Agree. Although the subject in (16a) is structurally higher than the object, it is nonetheless the object that controls agreement morphology. Only the object in (16a) is accessible, since it is unmarked, and thus it controls agreement morphology.

In languages where the higher of two DPs in a transitive sentence receives dependent Case, we expect agreement to target the direct object in at least some languages. Tsez is such a language, as we see in (17).4

(17)

graphic

In (17a), we see that an ABS-marked DP controls agreement morphology when it is the only DP in the clause. Furthermore, as demonstrated by (17b–c), agreement consistently targets the DP bearing ABS Case. ERG and DAT DPs are not accessible in Tsez, so they are not licit goals for Agree in Tsez. Only the object in (17b–c) is accessible; therefore, only the object can control agreement morphology in Tsez.

Other languages, such as Nepali, are argued to allow more-marked DPs to control verbal agreement morphology. As we see in (18a–b), a DP marked either NOM/unmarked Case or ERG/dependent Case may control verbal agreement morphology in Nepali. In both instances, the object also bears unmarked Case, but, as proposed by Bobaljik (2008), the subject is structurally closer to the probe and so controls agreement.

(18)

graphic

However, as in Icelandic, a dative subject, which bears lexical Case, may not be targeted for Agree, as we see in (18c). Instead, the object of the clause controls agreement. This is attributed to Nepali instantiating the Accessibility Hierarchy in (18d). What this shows is that, given two equally accessible DPs, only the structurally higher of the two controls agreement morphology.5 However, when the higher DP is inaccessible, a structurally lower DP may control agreement morphology, so long as it bears an accessible Case.

In this section, I have introduced Bobaljik’s (2008) notion of accessibility. As I will show in the following sections, this interacts with unlocking through Agree in a number of interesting ways. However, the question arises as to how much of the Accessibility Hierarchy is accessible in a given language. None of the surveyed languages appear to allow agreement with DPs bearing lexical Case. As we will see in sections 56, some languages surveyed allow Agree to target DPs bearing both dependent and unmarked Case, whereas some appear to target only DPs bearing unmarked Case.

4 Assembling the Pieces and the Puzzle of Obliques

In this section, I summarize what I have proposed so far, explore some of the predictions of the theory, and flesh out the core proposal slightly. In section 2 I showed, following Rackowski and Richards (2005), that extraction of nonarguments or possessors from DP is contingent on Agree, and in section 3 I showed, following Bobaljik (2008), that Agree with a DP is contingent on the Case that a given DP bears. When these two proposals are juxtaposed, an interesting prediction results: extraction of nonarguments or possessors from DP should be possible only if the DP in question is Case-accessible. To see this, consider the steps necessary for extraction of a nonargument or possessor from DP to take place, given in (19). If the first step fails, then so too will the second and third. Since Agree for φ-features is sensitive to the Case of potential goals, the Case of the DP in (19) must be such that the DP is a licit goal for Agree in a given language. Otherwise, extraction will be ungrammatical, since the DP cannot be targeted for Agree for φ-features and therefore cannot be unlocked for further probing.6

(19)

graphic

In the cases we are looking at, unlocking through Agree appears to be necessary, rather than simply an option. This means that the edge of DP is not accessible for Ā-movement in the examples we are examining. To capture this, I will assume an antilocality constraint like that of Murasugi and Saito (1994), Bošković (1997), Ishii (1997), Saito and Murasugi (1999), Abels (2003), Grohmann (2003), and Erlewine (2014), which blocks movement to the edge of a locality domain in these cases. I formulate the constraint that I assume to hold in (20).

(20)

  • General antilocality constraint

  • Ā-movement from a nonargument position cannot target the most local phase edge.

This antilocality constraint blocks movement from a high position in a phase to the edge of that phase.7 Therefore, extraction from higher positions in a phase must utilize unlocking, but this may not be the case for extraction from lower positions. Evidence for this comes from asymmetries in long-distance extraction from English gerunds. As shown in (21a), extraction of the subject of a gerund is impossible if the gerund itself is the complement of a preposition. This contrasts with (21b), which shows that extraction of the subject of a gerund is allowed if the gerund is the complement of a verb.

(21) Extraction from a low position is not Case-sensitive in English

  • a.

    *How much headway did you talk about ___ being made?

  • b.

    How much headway did you talk about John making ___ ?

    (based on Kayne 1981)

I assume that P, such as about in (21a–b), is not a probe. I propose that the asymmetry demonstrated in (21) is the result of extraction like that in (21a) being unable to reach the edge of the locality domain, causing unlocking to be mandatory, contrasting with extractions like (21b), where movement through the edge of the clause is in principle possible.

We are now in a position to explain the puzzle presented at the start of this article. In Bobaljik’s (2008) system, the complements of prepositions are marked with lexical Case and are therefore the least accessible.8 In (22), we see that extraction from DP is in principle possible from a DP that bears unmarked Case.9 This is a clear consequence of the proposal I have put forth. Such nominals have been targeted for Agree10 and as a consequence are unlocked, allowing the nonargument or possessor to be targeted by a [wh]-probe and thereby undergo movement.

(22) Extraction of nonarguments from DP is in principle possible

  • a.

    How much headway did you discuss [___ being made]?

  • b.

    Tinos [i afiksi ___ ] prokalese haos?

    whose the arrival.NOM created chaos.ACC

    ‘Whose arrival created chaos?’

    (Modern Greek; Ntelitheos 2002:6)

  • c.

    Achoj r-ixjayil x-Ø-a-ch’ey [ru-tz’i’ ___ ] rat?

    whose A3SG-wife PRFV-B3SG-A2SG-hit A3SG-dog you

    ‘Whose wife’s dog did you hit?’

    (Kaqchikel; Imanishi 2015:14)11

  • d.

    Buch’u iav-il-be [s-tot ___ ]?

    who A2-see-IOA3-father

    ‘Whose father did you see?’

    (Tzotzil; Aissen 1996:456)

  • e.

    Kua fia-mohome oti tuai [e tau tagata ___ na ].

    PERF want-sleep all PERF ABS PL person that

    ‘All of those people have gotten sleepy.’

    (Niuean; Seiter 1980)

  • f.

    Arna-up [anguti-it ___ ] taku-laur-ta-ngit atuniit.

    woman-ERG man-ABS.PL see-PST-IND-3SGA:3PLU each

    ‘The woman saw each of the men.’

    (Inuktitut; Beach 2003:8)

Interestingly, as we see in (23), such languages display a pattern identical to Sidaama with respect to the grammaticality of extraction of nonarguments or possessors from DP. Such extraction is possible when the DP is high on Bobaljik’s (2008) hierarchy, but impossible when the DP is marked with lexical Case.

(23) Extraction from DP is consistently ungrammatical from the complement of prepositions

  • a.

    *How much headway did you talk about [ ___ being made]?

    (based on Kayne 1981)

  • b.

    *Tinos anisihi i kivernisi jia [ti dieksagogi ___ ]?

    whose worries the government.NOM for the holding.fall.ACC

    ‘Holding what does the government worry about?’

    (Modern Greek; Ntelitheos 2002:6)

  • c.

    *Achoj ru-chaq’araxel ru-xb’al x-a-loq’-o’n pa [ru-k’ayij ___ ]?

    whose A3SG-younger A3SG-brother PRFV-B3SG-A2SG-shop at A3SG-store

    Intended: ‘Whose younger brother’s store did you shop at?’

    (Kaqchikel; Imanishi 2015:6)

  • d.

    *Buch’u ch-a-bat [ta s-na ___ ]?

    who ICP-B2-go PA3-house

    ‘Whose house are you going to?’

    (Tzotzil; Aissen 1996:469)

  • e.

    *Ne tutala oti a au [ke he [tau ___ momotua]].

    PST talk all ABS 1SG to LOC PL elders

    ‘I talked to all of the elders.’

    (Niuean; Seiter 1980:67)

  • f.

    *[Anguti-nut ___ ] aannia-na-laur-tuq atuniit.

    man-DAT.PL be.sick-CAUS-PST-IND(3SG) (*each)

    ‘It made each of these men sick.’

    (Inuktitut; Beach 2003:11)

This too is a direct consequence of the system I have proposed. These languages all allow Agree to target DPs with unmarked Case, but do not allow Agree to target DPs with lexical Case. Some of the languages surveyed—namely, Modern Greek, Kaqchikel, Tzotzil, and Niuean—simply do not appear to allow obliques in the canonical subject position or a place in which they could in principle control agreement morphology, but also do not display overt agreement with obliques. However, English, Inuktitut, and Sidaama all allow obliques either in the canonical subject position or in some other position in the clause where they could potentially control agreement morphology. As we see in (24), such obliques fail to control agreement morphology, even though they occupy a position from which agreement morphology may normally be controlled.

(24)

graphic

Example (24a) shows a locative inversion sentence in English.12 Here, a PP occupies the canonical subject position. However, the plural DP contained within that PP is unable to control agreement morphology. Rather, the postverbal DP is the controller of agreement. In (24b), we have a possessor- raising construction in Sidaama. Here, an oblique DP is structurally higher than a nonoblique DP. However, the oblique DP does not control agreement morphology; rather, the possessum, which bears NOM Case, controls agreement. Finally, in (24c), we have an oblique experiencer construction in Inuktitut. Here, the paradigm of φ-agreement utilized is that which normally cross-references the φ-features of only one argument. Although there are two possible goals for agreement in (24c), the verb inflects as though there were only one. Furthermore, we see that the plural number of the oblique experiencer is not expressed in the morphological paradigm that is used, further showing that the oblique DP is not an agreement controller. I assume that all of the languages surveyed in this section are alike, in that they do not allow Agree to target DPs with lexical Case. Since DPs bearing lexical Case may not be targeted for Agree, they cannot be unlocked, and therefore nonarguments and possessors may not be extracted from them.

One might now wonder if extraction of nonarguments from DP is simply not possible from the complement of a preposition, rather than being determined by Case, as I have proposed.13 Niuean provides a clear argument that this restriction is tied to Case, rather than to position or grammatical role. As we see in (25a–b), it is not possible to extract a quantifier from the complement of a preposition that has been assigned LOC Case. In contrast, however, it is possible to extract a quantifier from the complement of a preposition that has been assigned ABS Case, demonstrated by (25c–d). The only relevant difference between (25a–b) and (25c–d) is the Case of the DP that extraction takes place from, demonstrating clearly the influence of Case on extraction of nonarguments from DP.

(25) The Case of the oblique argument matters

  • a.

    Ne tutala a au ke he [tau momotua oti].

    PST talk ABS 1SG to LOCPL elders all

    ‘I talked to all of the elders.’

  • b.

    *Ne tutala oti a au ke he [tau momotua ___ ].

    PST talk all ABS 1SG to LOCPL elders

    ‘I talked to all of the elders.’

  • c.

    To ta e ia e fale [aki e tau mena gahua oti na ].

    FUT build ERG 3SGABS house INSTABSPL thing work all that

    ‘He will build the house with all those tools.’

  • d.

    To ta oti e ia e fale [aki e tau mena gahua ___ na ].

    FUT build all ERG 3SGABS house INSTABSPL thing work that

    ‘He will build the house with all those tools.’

    (Niuean; Seiter 1980:67, 251)

So far, in this section I have shown how the proposal I have put forth accounts for the oblique/nonoblique asymmetry introduced in section 1. Extraction of nonarguments or possessors from a DP is contingent on Agree with that DP, and oblique arguments, generally, bear lexical Case. In all of the languages surveyed, it is not possible to Agree with DPs bearing lexical Case, and therefore it is not possible to extract nonarguments or possessors from oblique DPs. This raises the important question of DPs that bear dependent Case, as the account I have put forth makes predictions about such DPs. More specifically, we expect that extraction of nonarguments or possessors from DPs that bear dependent Case should be grammatical if the language allows Agree with such DPs, and ungrammatical if the language does not allow such Agree relationships. The following two sections explore this particular ramification of the proposal.

5 NOM-ACC Languages Are Always Symmetric

In this section, I will show that the proposal, as outlined so far, apparently predicts incorrectly with respect to languages with a NOM-ACC Case alignment, but that it actually captures the attested facts given a view of configurational Case assignment that takes place during the narrow syntactic derivation, consistent with Preminger 2009 and Levin 2017.14 The argumentation presented so far does not rely on a particular model of Case assignment, and it is for the most part compatible with both a configurational account like that of Bobaljik (2008) and an account where Case is assigned as a reflex of an Agree relationship. However, the data in this section can only be accounted for under a configurational approach, lending further credence to configurational theories of Case.

As we saw in the previous sections, English, Modern Greek, Sidaama, and Northern Ostyak all display a NOM-ACC Case alignment. As expected, they all allow extraction of nonarguments and possessors from DP from NOM arguments,15 and Northern Ostyak also allows extraction of nonarguments or possessors from DP from an ACC-marked argument, so long as that argument has been targeted for Agree. However, as we see in (26), English, Modern Greek, and Sidaama all appear to allow extraction of nonarguments or possessors from ACC-marked DPs.

(26)

  • a.

    How much headway did you discuss [ ___ being made]?

  • b.

    Tinos empothise i kivernisi [tin afiksi ___ ]?

    whose prevented the government.NOM the arrival.ACC

    ‘Whose arrival did the government prevent?’

    (Modern Greek; Ntelitheos 2002:6)

  • c.

    [Bule [amá ___ ] la’-’-ino’ ] manč-i

    Bule[NOM.F] mother.ACC see-3SG.F-P.PERF.3 person-NOM.MOD.M

    da-Ø-i.

    come-3SG.M-S.PERF.3SG.M

    ‘The man whose mother Bule saw came.’

    (Sidaama; Kawachi 2007b:651)

The lack of an asymmetry between NOM- and ACC-marked DPs appears to extend across languages and is reflected typologically in a striking fact: no NOM-ACC languages that allow extraction from DP disallow extraction from DP from an ACC-marked DP. In addition to Modern Greek, English, Sidaama, and Northern Ostyak, the following languages conform to this pattern:16

(27)

  • a.

    Haga ɨɨ [ ___ kaadzi] pisapi?

    who.GEN 2SG.NOM car like.DUR

    ‘Whose car do you like?’

    (Northern Paiute; Toosarvandani 2014:801)

  • b.

    Phikaso-ga [ ___ chosanghwa-lul] kuli-n (ku) yeca

    Picasso-NOM portrait-ACC painted-NOM the woman

    ‘the woman who Picasso painted a portrait of ’

    (Korean; Choe 2009:27)

  • c.

    A fiúk-nak tegnap láttam [ ___ az any-já-t ].

    the boys-DAT yesterday saw.I the mother-POSS-ACC

    ‘The boys, I saw the mother of yesterday.’

    (Hungarian; É . Kiss 2014:60)

  • d.

    Bi a:nta [gada-i [ ___ zugdi-we-ni ]] b’agdi-e-mi.

    me woman buy-PRP house-ACC-3SG meet-PST-1SG

    ‘I met the woman whose house you bought.’

    (Udihe; Nikolaeva and Tolskaya 2001:679)

  • e.

    [pә [ ___ χotɕha] χui ] -saŋ khmәnnә laoʂi yi

    1SG book borrow.PRFV.NMLZ person 1SG.GEN teacher COP2.SUBJ

    ‘The person whose book I borrowed is my teacher.’

    (Bao’an Tu; Fried 2010:158)

The lack of such an asymmetry is surprising, but can be accounted for straightforwardly given a particular view of configurationally assigned Case.17 Preminger (2011, 2014) and Levin (2017) follow Bobaljik (2008) and ultimately Marantz (1991) in assuming that ergative and accusative Case are not assigned to DPs as the result of an Agree relationship between those DPs and some functional head. Rather, ERG and ACC Case are assigned to DPs that enter into a particular structural configuration with other DPs, with no functional head mediating Case assignment. Preminger (2011) differs from Marantz (1991) and Bobaljik (2008) in assuming that the process of Case assignment takes place in the narrow syntax. Lexical Case is assigned to a DP when it is a sister to certain Ps and Vs. Dependent Case is assigned to a DP that is in an asymmetric ccommand relationship with another DP. When the “lower” of the two is assigned Case, the language has a NOM-ACC case system. When the “higher” of the two is assigned Case, the language has an ERG-ABS case system. Finally, unmarked Case is not assigned at all.18 Instead, it merely represents the failure of a DP to have entered into either of the Case-assigning configurations.

This system leads us to expect an asymmetry between NOM-ACC languages and ERG-ABS languages. Namely, ACC-marked nominals should behave as though they were NOM for portions of the derivation. In contrast, ERG-marked nominals will not behave in this way. Recall that NOM is simply the absence of Case—a DP with NOM Case has not been the sister of a Case-assigning P, nor has it been asymmetrically c-commanded by another DP. In this approach to Case, a direct object will be marked with ACConly once it has entered into an asymmetric c-command relationship with another DP, as in (29); prior to that, it will bear NOM. Until that point in the derivation, a direct object will not bear Case, and it will be accessible for agreement, as illustrated in (28).19

(28)

graphic

(29)

graphic

As a result of this, an Agree relationship, which will result in the DP in question being unlocked, may be established with a direct object like that in (28), so long as that Agree relationship is established prior to Merge of the external argument, or if DPs bearing dependent Case are accessible for Agree, as they are in some languages. Crucially, this Agree relationship may be established even in languages that do not allow Agree with DPs bearing dependent Case.20 In other words, we expect the lack of an otherwise expected asymmetry between DPs with unmarked Case and DPs with dependent Case in NOM-ACC languages, and indeed such an asymmetry is not attested in any of the surveyed languages.

In this section, I dealt with one-half of the question of extraction of nonarguments or possessors from DPs with dependent Case. In languages with a NOM-ACC Case alignment, there appears to be no difference between unmarked Case and dependent Case with respect to the grammaticality of extraction of nonarguments or possessors from DP. I showed that an approach to Case in which Case is assigned configurationally throughout the derivation correctly captures the lack of an asymmetry in this respect between DPs bearing unmarked Case and DPs bearing dependent Case, in combination with the proposal I have put forth in sections 24. In the following section, I turn to the question of extraction of nonarguments or possessors from DPs with dependent Case in languages with an ERG-ABS Case alignment, given what has been said so far.

6 ERG-ABS Languages Are Not Always Symmetric

In this section, I examine extraction asymmetries in ERG-ABS languages. Extraction from DP is contingent on unlocking, which requires Agree, the acceptability of a given DP as a goal for Agree being determined by the Case that the DP bears. We have seen that this accounts for the ban on extraction from obliques: in the surveyed languages, obliques may not be targeted for Agree, and as a result, they cannot be extracted from. We expect this to be true of nonobliques as well: if a nonoblique DP cannot be targeted for Agree in a given language, then the language should not allow extraction from DP. If a language allows ERG-marked DPs to be targeted for Agree, then the proposal made here predicts that it should also allow extraction from ERG-marked DPs. Conversely, if a language does not allow ERG-marked DPs to be targeted for Agree, then it also should not allow extraction from ERG-marked DPs, since extraction from DP is contingent on unlocking, which is contingent on Agree.

In the previous section, I argued that extraction of nonarguments or possessors from DPs bearing either NOM or ACC is equally acceptable, given a particular view of Case assignment and the proposal put forth in sections 24. For some portion of the derivation, ACC-marked DPs will lack Case and therefore will be equivalent to NOM-marked DPs for the purposes of Case accessibility. This theory makes a different prediction about ERG-marked DPs. Such DPs will be assigned Case immediately after they are merged in Spec,vP, as schematized in (30)–(31). This is because these DPs will meet the condition for ERG to be assigned immediately after being merged into the clausal spine.

(30)

graphic

(31)

graphic

In other words, we expect there to be languages that allow extraction of nonarguments or possessors from DPs with ERG Case and languages that do not. This will be determined by the Accessibility Hierarchy: if a language allows Agree with ERG-marked DPs, extraction of nonarguments or possessors from ERG-marked DPs will be allowed. Conversely, if a language does not allow Agree with ERG-marked DPs, extraction of nonarguments or possessors from ERG-marked DPs will not be allowed. In this section, I will show that this is the case.

To begin, let us consider Tsez. As we saw in section 3, Tsez only targets ABS-marked DPs for φ-agreement. In addition, as we see in (32), Tsez freely allows relativization on both ergatives and absolutives and therefore does not display the ergative/absolutive extraction asymmetry common to languages that are syntactically ergative.21

(32) Tsez is not syntactically ergative

  • a.

    Už-ä kaɣat kid-be-r teƛ-si/teƛ-xo.

    boy-ERG letter.ABS.II girl-OS-LAT give-PST.WIT/give-PRS

    ‘The boy gave/gives a letter to the girl.’

  • b.

    [ ___ kaɣat kid-be-r teƛ-si/teƛ-xo] uži

    letter.ABS.II girl-OS-LAT give-PST.WIT/give-PRS boy.ABS.I

    ‘the boy that gave a letter to the girl’

  • c.

    [už-ä ___ kid-be-r-teƛ-si/teƛ-xo ] kaɣat

    boy-ERG girl-OS-LAT give-PST.WIT/give-PRS letter.ABS.II

    ‘the letter that the boy gave to the girl’

    (Polinsky 2015:266)

If the analysis presented in this article is on the right track, then we expect relativization of possessors from ERG-marked DPs to be ungrammatical in Tsez, even though Tsez allows relativization of ERG-marked DPs. This case is comparable to the contrast in Sidaama (see examples (1)–(2)) between oblique arguments, which can relativize, and possessors of oblique arguments, which cannot relativize, even though possessor relativization is allowed in the language. In other words, the analysis in this article predicts that the set of things that can be extracted and the set of things that can be extracted from may be disjoint. As we see in (33), this expectation is borne out. (33a) simply shows that an ABS-marked DP controls agreement in an intransitive clause. In (33b), we see that possessor relativization from an ABS-marked DP is grammatical in Tsez. ABS-marked DPs in Tsez are able to be targeted for Agree, as demonstrated in (32a), and thus are unlocked. So, we expect extraction from ABS-marked DPs to be grammatical in Tsez, and indeed this is the case, as we see in (33b). (33c) shows that ERG-marked DPs do not control agreement in transitive clauses. In (33d), we see that possessor relativization from an ERG-marked DP is not grammatical in Tsez, in contrast to what we see in (33b). ERG-marked DPs are not able to be targeted for Agree in Tsez, so they cannot be unlocked. So, we expect extraction from ERG-marked DPs not to be grammatical in Tsez.

(33) Extraction from an ERG-marked DP is ungrammatical in Tsez

  • a.

    Uži-s babiw Ø-exu-s.

    boy-GEN father.ABS.II-die-PST.WIT

    ‘The boy’s father died.’

  • b.

    [[babiw ___ ] Ø-äxu-ru ] uži

    father.ABS.II-die-PST.PTCP boy.ABS.I

    ‘the boy whose father died’

  • c.

    Uži-z yay-ä di ħan-si.

    boy-GEN dog-ERG 1SG.ABS bite-PST.WIT

    ‘The boy’s dog has bitten me.’

  • d.

    ??[[yay-ä ] di ħän-ru ] uži

    dog-ERG 1SG.ABS bite-PST.PTCP boy.ABS.I

    ‘the boy whose dog bit me’

    (Polinsky 2015:268, 269, 270)

Tzotzil patterns like Tsez in allowing extraction from ABS-marked DPs but not from ERG-marked DPs or from the complements of prepositions. Tzotzil exhibits head rather than DP marking in its Case system; I follow the Mayanist literature in associating B-class morphemes with the absolutive and A-class morphemes with the ergative.22

(34) Extraction from DP is ungrammatical from ERG-marked DPs and complements of P in Tzotzil

  • a.

    Buch’u i-cham [x-ch’amal ___ ]?

    who CP-die A3-child

    ‘Whose child died?’

  • b.

    Buch’u iav-il-be [s-tot ___ ]?

    who A2-see-IOA3-father

    ‘Whose father did you see?’

  • c.

    *Buch’u y-elk’an chij [x-ch’amal ___ ]?

    who A3-steal sheep A3-child

    ‘Whose child stole sheep?’

  • d.

    *Buch’u ch-a-bat [ta s-na ___ ]?

    who ICP-B2-go PA3-house

    ‘Whose house are you going to?’

    (Aissen 1996:456, 460, 469)

Tzotzil exhibits another type of extraction from DP, which is subject to the same constraint.

(35)

  • Quantifier dislocation in Tzotzil

  • 7ep 7i-laj [ti Pinedae ___ ].

  • lots CP-die the Pinedists

  • ‘Many of the Pinedists died.’

  • (Aissen 1987:253)

I have assumed that such processes of dislocation involve the same sort of dependency formation that is necessary for Ā-movement of possessors. Thus, quantifier dislocation is contingent on unlocking in the same way that possessor extraction is. As the contrast in (36) illustrates when considered alongside (35), a dislocated quantifier may not be construed with an ERG-marked DP.23

(36) Quantifier dislocation is Case-sensitive in Tzotzil

  • a.

    7ep 7i-s-k’el-ik k’in li tzebetik -e.

    lots CP-A3-look-PL fiesta the girls CL

    ‘The girls saw many fiestas.’

    (not: Many girls saw the fiesta.)

  • b.

    7ep 7i-s-ti7-ik kaxlan li viniketik -e.

    lots CP-A3-eat-PL chicken the men CL

    ‘The men ate plenty of chicken.’

    (not: Many men ate chicken.)

    (Aissen 1987:253)

We see that quantifier dislocation in Tzotzil, exemplified by (35)–(36), is only possible from ABS-marked DPs. Tzotzil therefore instantiates the same Accessibility Hierarchy as Tsez and Kaqchikel. In such languages, only ABS-marked DPs may be targeted for Agree, so only ABS-marked DPs may be unlocked.

Kaqchikel patterns the same way as Tsez and Tzotzil. Kaqchikel, like Tzotzil, exhibits head rather than DP marking in its Case system. Extraction from DP is only possible from DPs bearing unmarked Case. As we saw in (22c) (repeated here in (37a)), Kaqchikel allows possessor extraction from ABS-marked DPs.

(37) Possessor extraction in Kaqchikel

  • a.

    Achoj r-ixjayil x-Ø-a-ch’ey [ru-tz’i’ ___ ] rat?

    whose A3SG-wife PRFV-B3SG-A2SG-hit A3SG-dog you

    ‘Whose wife’s dog did you hit?’

  • b.

    Achoj r-ixjayil x-Ø-tzaq [ru-nupq’a ___ ]?

    whose A3SG-wife PRFV-B3SG-fall A3SG-ring

    ‘Whose wife’s ring fell?’

  • c.

    Achoj r-ixjayil x-Ø-xajo’ [ru-te ___ ]?

    whose A3-wife PRFV-B3SG-dance A3SG-mother

    ‘Whose wife’s mother danced?’

    (Imanishi 2015:14)

In Kaqchikel, B-class morphemes cross-reference the φ-features of ABS-bearing DPs. Clearly, this suggests that such DPs have been targeted for Agree. As we have seen before, extraction from DP in Kaqchikel is sensitive to Case. Complements of prepositions, which bear lexical Case, cannot contain a gap, as we see in (38a).

(38) Extraction from DP is ungrammatical from ERG-marked DPs and complements of P: Case-sensitive

  • a.

    *Achoj ru-chaq’araxel ru-xb’al x-a-loq’-o’n pa [ru-k’ayij ___ ]?

    whose A3SG-younger A3SG-brother PRFV-B3SG-A2SG-shop at A3SG-store

    Intended: ‘Whose younger brother’s store did you shop at?’

  • b.

    *Achoj r-ixjayil x-a-r-’a/x-a-b’a-o’ rat [ru-tz’i’ ___ ]?

    whose A3SG-wife PRFV-B2-A3SG-bite/PRFV-B2SG-bite-AF you A3SG-dog

    Intended: ‘Whose wife’s dog bit you?’

    (Imanishi 2015:6, 14)

Interestingly, ERG-marked DPs also cannot contain a gap, as we see in (38b). Possessor extraction in Kaqchikel requires unlocking through Agree. In the cases in (38), Agree must have failed as in the Tsez cases—so something must be said about the putative ergative agreement morphology in (38b). Preminger (2014:210–211) argues for independent reasons that Kichean languages such as Kaqchikel have an accessibility hierarchy like that in (39).

(39)

graphic

While Preminger argues that absolutive agreement morphology is the result of a number of probegoal relationships, he argues that ergative agreement morphology is the result of a process of clitic doubling. Similarly, Imanishi (2014) proposes that ERG-marked nominals in Kaqchikel cannot be in an Agree relationship, to account for the broader distribution of ERG-marked clitics in the language. Crucially for the account presented here, the ergative clitic-doubling process in Kichean does not involve Agree. As a result of the hierarchy in (39), only ABS-marked DPs may be targeted for Agree and unlocked. ABS-marked DPs are indeed targeted for Agree and thus unlocked. ERG-marked DPs are not targeted for Agree, so they cannot be unlocked.

I have now established that quantifier dislocation is sensitive to Case in the same way that possessor extraction is, as it too requires unlocking. In contrast to what we have seen in the languages examined so far, we expect quantifier dislocation to be grammatical from ERG-marked DPs in languages where ERG-marked DPs are accessible. In other words, the ungrammaticality of extraction from ERG-marked DPs in the languages surveyed so far does not reduce to a general ban on extraction from ERG-marked DPs.24 Now, I will show that this is the case in two languages that we have already considered: Inuktitut and Niuean. First, I will reintroduce facts about Niuean and Inuktitut which suggest that these languages allow Agree with ERG-marked DPs. Then, I will show that quantifier dislocation is, as expected, grammatical from these DPs.

Let us begin by looking at Inuktitut. Inuktitut instantiates an ERG-ABS case system, as we see in (40).

(40) Ergative case system in Inuktitut with ERG/ABS agreement

  • a.

    Jaani itir-tuq.

    Jaani.ABS enter-IND(3SG)

    ‘Jaani entered.’

  • b.

    Jaani-up nanuq qukir-ta-nga.

    Jaani-ERG bear.ABS shoot-IND-3SGA:3SGU

    ‘Jaani shot the bear.’

    (Beach 2003:3)

In (40), we see that the subjects of transitive clauses receive additional case marking and that direct objects do not. Furthermore, we see that verbal agreement morphology in transitive clauses reflects φ-features of both subject and object, suggesting that Inuktitut instantiates an accessibility hierarchy like that in (41).

(41)

graphic

If a DP can be targeted for Agree, it can be unlocked. Inuktitut allows both ERG- and ABS-marked DPs to be targeted for Agree. As a result, we expect ERG-marked subjects in Inuktitut to be unlocked for extraction, in contrast with ERG-marked subjects in Kaqchikel, which may not be targeted for Agree, by hypothesis. A further argument that morphology reflecting the φ-features of ERG-marked DPs in Inuktitut involves Agree, in contrast with morphology in Kichean and Basque, is provided by Compton (2014). Compton (2014) notes that ergative agreement morphology in Inuktitut is not tense-aspect-mood (TAM) invariant, in contrast with ergative agreement morphology in Kichean and Basque. If TAM invariance is a property of clitic doubling, following Nevins (2011), and TAM variance is a property of true Agree, then we can be sure that Inuktitut ergative agreement morphology is the result of a true Agree relationship.

If ERG-marked nominals are in fact targeted for Agree in Inuktitut, then we expect them to be able to contain a gap. As we see in (42), this expectation is borne out. Inuktitut allows extraction from both subject and direct object. In both (42a) and (42b), the quantificational phrase atuniit is dislocated from the DP it modifies, appearing to the right of the relevant DP, at the edge of the clause.

(42) Extraction from DP is possible from both subject and object in Inuktitut

  • a.

    [Anguti-it ___ ] arnaq taku-laur-ta-ngat atuniit.

    man-ERG.PL woman(ABS) see-PST-IND-3PLA:3SGU each

    ‘The men each saw the woman.’

  • b.

    Arna-up [anguti-it ___ ] taku-laur-ta-ngit atuniit.

    woman-ERG man-ABS.PL see-PST-IND-3SGA:3PLU each

    ‘The woman saw each of the men.’

    (Beach 2003:8)

In both (42a) and (42b), the relevant DP may be targeted for Agree, and as a result, those DPs are unlocked, allowing extraction. This accounts for the fact that atuniit may dislocate from ABS-marked DPs as well as ERG-marked DPs in (42).

We will now see that Niuean behaves identically to Inuktitut. As will be evident, Niuean has a slightly more complicated Case system. However, it also shows ERG-ABS parity in allowing extraction. Crucially for the argument at hand, Niuean, like Inuktitut, also allows ERG-marked subjects to control verbal agreement morphology.

Subjects of intransitive verbs are marked for Case in Niuean, but the form of Case is not consistent. Specifically, there is allomorphy between the ergative and absolutive case marker paradigms; I will give a thorough description of the two paradigms, so I can show that it is the type of Case on a nominal that determines whether or not it is transparent for extraction, not the morphological expression of that Case.

(43) Different surface forms for unmarked Case in Niuean

  • a.

    Ne tohitohi a Sione.

    PST writing ABS Sione

    ‘Sione was writing.’

    (Massam 2006:28)

  • b.

    Kua egaega e kau kauvehe.

    PERF rosy ABSPL cheek

    ‘The cheeks are rosy.’

    (Sperlich 1997:55)

If the noun is a name, as in (43a), or a pronoun, unmarked Case is realized as a. If the noun is a common noun, as in (43b), unmarked Case is realized as e. Plural subjects of intransitive verbs may control a reduplicative process of agreement, demonstrated in (44).25

(44) Reduplicative agreement with ABS in Niuean

  • a.

    Kua mohe a ia he fale.

    PERF sleep ABS he in house

    ‘He has slept in the house.’

  • b.

    Kua momohe a lautolu he fale.

    PERF sleep.PLABS they in house

    ‘They have slept in the house.’

    (Seiter 1980:63)

This can be taken as evidence that ABS-marked/unmarked DPs are accessible in Niuean. Since ABS-marked DPs may be targeted for Agree, we expect them to be unlocked for extraction in Niuean.

Subjects of transitive clauses are marked for Case in Niuean, as we see in (45). This Case marking is distinct from the Case marking that subjects of intransitive clauses receive. However, the form of Case is not consistent, and it is inconsistent in the same way that ABS Case is in Niuean: the form of the Case marker makes a distinction between pronouns and names on the one hand and common nouns on the other.

(45) Different surface forms for dependent Case in Niuean

  • a.

    Koe tele e Sione a Sefa.

    PRS kick ERG Sione ABS Sefa

    ‘Sione is kicking Sefa.’

  • b.

    Ne kai he pusi ia e moa.

    PST eat ERG cat that ABS bird

    ‘That cat ate the chicken.’

    (Seiter 1980:29)

If the subject is a name, as in (45a), or a pronoun, dependent Case is realized as e. If the subject is a common noun, as in (45b), dependent Case is realized as he. Furthermore, the direct objects in (45) bear the same case morphology as the subjects of intransitive clauses. A plural ergative subject may also control reduplicative agreement, as in (46).

(46) Reduplicative agreement with ERG in Niuean

  • a.

    Kua mohe e ia e timeni.

    PERF sleep ERG he ABS floor

    ‘He has slept on the floor.’

  • b.

    Kua momohe e lautolu e timeni.

    PERF sleep ERG they ABS floor

    ‘They have slept on the floor.’

    (Seiter 1980:63)

Both ERG- and ABS-marked DPs may control agreement morphology in Niuean. This suggests that Niuean instantiates the accessibility hierarchy in (47).

(47)

graphic

So, we expect that extraction from DP should be grammatical from both ERG- and ABS-marked DPs, as in Inuktitut. I will now provide evidence that such an operation exists and that it patterns as we expect.

Niuean has a quantifier word oti, which Seiter (1980) translates as ‘all’. As we see in (48), it may appear internal to a nominal, and its distribution is unconstrained by Case.26

(48) The presence of oti is unconstrained by Case in Niuean

  • a.

    Kua fia-mohome tuai [e tau tagata oti na ].

    PERF want-sleep PERFABSPL person all that

    ‘All of those people have gotten sleepy.’

  • b.

    Moua e maua mo Sione [e tau mata afi oti].

    get ERG 1PL.DU.EX with Sione ABSPL piece fire all

    ‘Sione and I have already won all the matches.’

  • c.

    Kua tele tuai [e lautolu oti] a au.

    PERF kick PERFERG 3PL all ABS 1SG

    ‘All of them have kicked me.’

    (Seiter 1980:65, 66)

We have seen that ERG- and ABS-marked DPs may control agreement morphology in Niuean, showing that both are accessible, and both may be unlocked through Agree. Thus, we expect that extraction from both sorts of DP should be grammatical. As we see in (49), this expectation is borne out. Extraction from DP is possible from ERG-marked DPs and ABS-marked DPs. Oti may be extracted from ABS-marked subjects and objects, as well as ERG-marked subjects. More generally, we see that extraction of nonarguments or possessors from ERG-marked DPs is not limited to Inuktitut.

(49) ABS/ERG parity in extraction from DP in Niuean

  • a.

    Kua fia-mohome oti tuai [e tau tagata ___ na ].

    PERF want-sleep all PERFABSPL person that

    ‘All of those people have gotten sleepy.’

  • b.

    Moua oti e maua mo Sione [e tau mata afi ___ ].

    get all ERG 1PL.DU.EX with Sione ABSPL piece fire

    ‘Sione and I have already won all the matches.’

  • c.

    Kua tele oti tuai [e lautolu ___ ] a au.

    PERF kick all PERFERG 3PLABS 1SG

    ‘All of them have kicked me.’

    (Seiter 1980:66, 67)

In this section, I showed that the theory of extraction I have presented predicts that extraction of nonarguments or possessors from DP should be banned from DPs bearing dependent Case in languages where Agree with these DPs is not allowed and in which dependent Case is assigned upward—in other words, ERG-ABS languages. This prediction is confirmed: the languages that allow extraction from ERG-marked DPs also appear to allow Agree with ERG-marked DPs, and the languages that do not allow extraction from ERG-marked DPs appear to not allow Agree with ERG-marked DPs. I also showed that the ban on extraction from ERG-marked DPs does not correlate with syntactic ergativity. Tsez, for instance, allows extraction ofERG-marked DPs, but does not allow extraction fromERG-marked DPs.

7 Conclusion

In this article, I proposed a theory to account for crosslinguistic asymmetries in extraction of nonarguments and possessors from DPs. I showed that the juxtaposition of proposals by Rackowski and Richards (2005) and Bobaljik (2008) correctly predicts a variety of extraction asymmetries. Rackowski and Richards (2005) propose that Agree with a phase is a precondition of extraction from that phase, while Bobaljik (2008) proposes that Agree with a DP is contingent on the Case that DP bears. As a result of this, extraction of a nonargument from a DP is preconditioned on the Case of that DP, since Case determines whether or not the DP in question may be targeted for Agree. I noted that the model apparently predicts incorrectly for NOM-ACC languages and suggested that adopting a derivational approach to dependent Case like that put forth by Preminger (2011, 2014) and Levin (2017) will cause the model to make correct predictions. I then showed that this amended version of the proposal correctly predicts extraction of a nonargument from an ergative DP to be contingent on whether or not the ergative DP may be targeted for Agree.

Abbreviations

1 = first person

2 = second person

3 = third person

1PL.DU.EX = first person dual excluding addressee

3PLA:3SGU = third person plural agent; third person singular undergoer

3SGA:3PLU = third person singular agent; third person plural undergoer

3SGA:3SGU = third person singular agent; third person singular undergoer

3SGS = third person singular subject

A = A-class morpheme

ABL = ablative

ABS = absolutive

ACC = accusative

AF = agent focus

AI = animate intransitive

ANG = ang

ASP = aspect

B = B-class morpheme

CAUS = causative

CL = clitic

COP = copula

CP = completive aspect

CS = case

DAT = dative

DEM = demonstrative

DIR = direct

DUR = durative

ERG = ergative

EVID = evidential

F = feminine

FUT = future

GEN = genitive

HON = honorific

IC = initial change

ICP = incompletive aspect

IMPF = imperfective

IND = indicative

INST = instrumental

IO = indirect object

LAT = lative

LOC = locative

M = masculine

MOD = modified

NMLZ = nominalization

NOM = nominative

NONPST = nonpast

OBJ = object

OBL = oblique

OS = oblique stem

P = preposition

PERF = perfect

PL = plural

POSS = possessive

P.PERF = past perfect

PRFV = perfective

PRP = present participle

PRS = present

PST = past

PTCP = participle

SG = singular

S.PERF = simple perfect

SUBJ = subject

T = tense

TA = transitive animate

WIT = witness

Notes

1 I do not argue that this is the only restriction on Ā-extraction from DP more generally. For instance, as we see in (i), English displays a subject/object asymmetry in Ā-extraction of arguments of nominals.

  • (i)

    • a.

      What did Mary hear rumors of ___?

    • b.

      *What did rumors of ___ bother Mary?

I will put aside the question of (i) in this article and assume that the contrast is caused by some other factor.

2Nikolaeva (1999b) states that sentences like (10b) involve the object being more “topical” than the object in sentences like (10a).

3 The term animate intransitive is inherited from the Algonquian literature and appears to be somewhat of a misnomer for Blackfoot, since (11a–b) clearly show that “animate intransitive” verbs may have direct objects.

4 We see clearly that mother in (17b) does not control agreement morphology, since it is a class II noun (Polinsky 2015).

5 This is, of course, not universally true. Some languages allow both subjects and objects to control agreement morphology simultaneously. Other languages preferentially Agree with nominals bearing first or second person, regardless of argument structure. I set such questions aside for further research.

6 As noted by a reviewer, this account presupposes that movement cannot proceed through the edge of a locked DP. Rackowski and Richards (2005) argue that long-distance Ā-extraction requires an Agree relationship with the clause that extraction takes place from, even in languages where this Agree relationship is not expressed overtly. I will assume something slightly weaker than Rackowski and Richards (2005): that an unlocking Agree relationship may allow successive-cyclic movement to bypass the CP edge. However, I do not rule out movement through the edge of CP as an alternative option, also allowable at least in principle. Such movement to Spec,CP must be blocked in Tagalog to maintain the account given by Rackowski and Richards (2005). One possible analysis could invoke Chomsky’s (2008) notion of “feature transmission.” Languages like Tagalog would transmit both their φ- and [wh]-features to T, thereby blocking movement of [wh]-phrases to Spec,CP.

7 It is not essential to the analysis that possessors be nonarguments. So long as nonarguments in the DP and arguments that are structurally high in the DP, such as possessors, class together with respect to the antilocality constraint, they will require the DP that contains them to be unlocked if they are to undergo extraction. An analysis along these lines will predict that subjects which undergo movement to a comparably high position will require the clause that contains them to undergo unlocking if they are to be extracted.

8 While all of the languages surveyed in this article neither allow Agree with nor allow extraction from DPs with lexical Case, two anonymous reviewers correctly note that there are languages that do allow Agree with such DPs. We expect these DPs to allow extraction from DPs bearing lexical Case. Unfortunately, I have not been able to locate in the literature a language that clearly allows Agree with DPs bearing lexical Case that have paradigms of extraction like the ones this article is concerned with. I leave finding these paradigms as a topic of further research.

9 Modulo the English example, which I will provide an explanation for in section 5.

10 I follow Rackowski and Richards (2005) in allowing a certain level of abstraction with respect to Agree. In other words, I will assume that Agree here need not result in overt agreement morphology appearing in the verbal domain.

11Imanishi (2015) notes that extraction of possessors from Kaqchikel is subject to a curious but, I believe, independent requirement: the possessor may undergo Ā-extraction only if it consists of two or more phonological words. I leave an explanation of this requirement to further work.

12 It is important here to disambiguate the theoretical notion of lexical Case and “lexical Case” as it has been referred to in the literature. The former is assigned idiosyncratically to nominals in sisterhood relationships with certain V0s and P0s; the latter is associated with a particular verb semantics.

13 One anonymous reviewer notes that the impossibility of extracting nonarguments from PPs could be attributed to the phasehood of the PP at hand. As we see in (25a–d), such an account cannot be maintained, since the possibility of extracting nonarguments from PPs appears to be contingent on the Case that the preposition assigns. Thanks to another anonymous reviewer for pointing out the importance of this particular contrast.

14 As a result of the nature of the incorrect prediction, I will not provide particular accessibility hierarchies for the languages surveyed in this section. Crucially, given the approach to Case assignment that I assume in this section, we will not expect there to be a difference between languages that allow Agree only with NOM-bearing DPs and languages that allow Agree only with NOM- and ACC-bearing DPs.

15 Modulo English, which displays a more general restriction on extraction from NOM subjects. As mentioned in footnote 1, I assume that this restriction is caused by some factor independent of the theory developed in this article.

16 Examples (27a–e) are taken mostly from grammars where Case assignment could be seen. In addition, a resumptive strategy for extraction in the language is mentioned in either the grammar or the analysis, but this strategy is not taken in any of (27a–e).

17 A reviewer raises the question of how dependent Case approaches might account for classic Case-licensing effects. In this regard, Baker (2015) proposes that dependent Case might have a licensing function, and Richards (2010, 2016) accounts for many classic Case-licensing effects without making recourse to Case as a licenser at all.

18 A reviewer points out that this theory collapses Marantz’s (1991) default case with unmarked case, but that, as Schütze (1997) demonstrates, the two do not always correspond.

19 Under this system, we expect to find languages that instantiate “Case stacking,” where more than one Case is assigned to a given DP. See Levin 2017 for evidence that such languages do exist, as well as a configurational account of Case compatible with what I propose in this section.

20 An anonymous reviewer notes that the proposal in this section appears to undermine a prediction of Bobaljik’s (2008) analysis: that there should be no languages that allow Agree with ACC-marked DPs to the exclusion of NOM-marked DPs. One way to rescue this prediction would be to assume that what has been referred to as Agree actually consists of (at least) two operations, Match and Copy. Match takes place in the syntax and feeds later Copy operations, which take place in the morphology. It is Copy that would be responsible for the valuation and realization of agreement morphology on probing heads. Suppose now that both operations are sensitive to the Accessibility Hierarchy, and that it is Match that is responsible for unlocking a DP so that a nonargument or possessor may be extracted from it. Given these assumptions, we may keep the proposal put forth in this section while maintaining the prediction of Bobaljik’s (2008) analysis. Match would take place in the syntax and therefore would not discriminate between NOM and ACC for the reasons laid out above. Copy would take place in the morphology, after the entire syntactic structure had been built, and therefore would discriminate between NOM and ACC, as it could never apply at points in the derivation like that in (29).

21 No inflection morphology appears on the verb teƛ ‘give’ in (32). This appears to be a lexically idiosyncratic property of the verb in question—in all examples involving this verb given in Polinsky 2015, the verb consistently fails to display agreement morphology.

22 A-class morphology also appears on the possessum in Tzotzil (and Kaqchikel), reflecting the φ-features of the possessor.

23 Unfortunately, no judgments are given for dislocation of quantifiers from oblique arguments in Aissen 1987.

24 This does not seem reducible to a ban on extraction from nominals in Spec,TP. For instance, Aissen (1996) proposes that nothing occupies Spec,TP in Tzotzil; the subject and object remain within v/VP, which displays an ERG/ABS asymmetry. Longenbaugh and Polinsky (2017) propose the same for Niuean, which does not display the asymmetry.

25 Plural objects, to my knowledge, do not control this process of agreement.

26 We will see shortly that oti may appear within a DP bearing oblique case, as well.

Acknowledgments

Thanks go first to David Pesetsky and one particularly helpful anonymous reviewer. Earlier stages of this work have been presented at NELS 45, and I also thank that audience for their helpful suggestions.

A list of abbreviations used in the glosses appears at the end of the article.

References

Abels,
Klaus
.
2003
.
Successive cyclicity, anti-locality, and adposition stranding
.
Doctoral dissertation, University of Connecticut, Storrs
.
Aissen,
Judith
.
1987
.
Tzotzil clause structure
.
Dordrecht
:
Springer
.
Aissen,
Judith
.
1996
.
Pied-piping, abstract agreement, and functional projections in Tzotzil
.
Natural Language and Linguistic Theory
14
:
447
491
.
Baker,
Mark C
.
2015
.
Case: Its principles and its parameters
.
Cambridge
:
Cambridge University Press
.
Beach,
Matthew
.
2003
. Asymmetries between passivization and antipassivization in the Tarramiutut subdialect of Inuktitut. In
Proceedings of the LFG 03 Conference
, ed. by
Miriam
Butt
and
Tracy
Holloway King
,
1
21
.
Stanford, CA
:
CSLI Publications
.
Bickel,
Balthasar
, and
Yogendra P.
Yādava
.
2000
.
A fresh look at grammatical relations in Indo-Aryan
.
Lingua
110
:
343
373
.
Bliss,
Heather
.
2012
. A split DP analysis of Blackfoot nominal expressions. In
Proceedings of the 2012 annual conference of the Canadian Linguistic Association
, ed. by
Paula
Caxaj
. http://homes.chass.utoronto.ca/~cla-acl/actes2012/actes2012.html.
Bobaljik,
Jonathan David
.
2008
. Where’s phi? Agreement as a post-syntactic operation. In
Phi-theory: Phi features across interfaces and modules
, ed. by
David
Adger
,
Daniel
Harbour
, and
Susana
Béjar
,
295
328
.
Oxford
:
Oxford University Press
.
Bošković,
Željko
.
1997
.
The syntax of nonfinite complementation: An economy approach
.
Cambridge, MA
:
MIT Press
.
Choe,
Hyon Sook
.
2009
. On Left-Branch Extraction (LBE) and Left-Branch Condition (LBC) effects. In
Proceedings from the Annual Meeting of the Chicago Linguistic Society 45
, ed. by
Ryan
Bochnak
,
Nassira
Nicola
,
Peter
Klecha
,
Jasmin
Urban
,
Alice
Lemieux
, and
Christina
Weaver
,
2
:
27
42
.
Chicago
:
University of Chicago, Chicago Linguistic Society
.
Chomsky,
Noam
.
2008
. On phases. In
Foundational issues in linguistic theory: Essays in honor of Jean-Roger Vergnaud
, ed. by
Robert
Freidin
,
Carlos P.
Otero
, and
Maria Luisa
Zubizarreta
,
133
166
.
Cambridge, MA
:
MIT Press
.
Compton,
Richard
.
2014
.
An argument for genuine object phi-agreement in Inuit
.
Poster presented at GLOW 37. Brussels, 2–4 April
.
Erlewine,
Michael Yoshitaka
.
2014
. Anti-locality and Kaqchikel agent focus. In
WCCFL 31: Proceedings of the 31st West Coast Conference on Formal Linguistics
, ed. by
Robert E.
Santana-LaBarge
,
150
159
.
Somerville, MA
:
Cascadilla Proceedings Project
.
Fried,
Robert
.
2010
.
A grammar of Bao’an Tu, a Mongolic language of Northwest China
.
Doctoral dissertation, State University of New York, Buffalo
.
Grohmann,
Kleanthes K
.
2003
.
Prolific domains: On the anti-locality of movement dependencies
.
Amsterdam
:
John Benjamins
.
Hrafnbjargarson,
Gunnar Hrafn
.
2002
.
Subject case and agreement in the history of English and Scandinavian
.
Guest lecture series of the Graduiertenkolleg, Universität Stuttgart
.
Imanishi,
Yusuke
.
2014
.
Default ergative
.
Doctoral dissertation, MIT, Cambridge, MA
.
Imanishi,
Yusuke
.
2015
.
Null resumption in Kaqchikel and its theoretical implications
.
Ms., MIT, Cambridge, MA
.
Ishii,
Toru
.
1997
.
An asymmetry in the composition of phrase structure and its consequences
.
Doctoral dissertation, University of California, Irvine
.
Kawachi,
Kazuhiro
.
2007a
. Dative external possessor constructions in Sidaama. In
Proceedings of the 33rd annual meeting of the Berkeley Linguistics Society
, ed. by
Thera
Crane
et al
,
258
269
.
Berkeley
:
University of California, Berkeley Linguistics Society
.
Kawachi,
Kazuhiro
.
2007b
.
A grammar of Sidaama (Sidamo), a Cushitic language of Ethiopia
.
Doctoral dissertation, State University of New York
.
Kayne,
Richard S
.
1981
. Two notes on the NIC. In
Theory of markedness in generative grammar: Proceedings of the 1979 GLOW conference
, ed. by
Adriana
Belletti
,
Luciana
Brandi
, and
Luigi
Rizzi
,
317
346
.
Pisa
:
Scuola Normale Superiore
.
É. Kiss,
Katalin
.
2014
.
Ways of licensing Hungarian external possessors
.
Acta Linguistica Hungarica
61
:
45
68
.
Kroeger,
Paul
.
1993
.
Phrase structure and grammatical relations in Tagalog
.
Stanford, CA
:
CSLI Publications
.
Levin,
Theodore
.
2017
.
Successive-cyclic case assignment: Korean nominative-nominative case-stacking
.
Natural Language and Linguistic Theory
35
:
1
52
.
Longenbaugh,
Nicholas
, and
Maria
Polinsky
.
2017
.
Equidistance returns
. http://ling.auf.net/lingbuzz/003262/current.pdf.
Marantz,
Alec
.
1991
. Case and licensing. In
Proceedings of ESCOL 91
, ed. by
Germán F.
Westphal
,
Benjamin
Ao
, and
Hee-rahk
Chae
,
234
253
.
Columbus
:
Ohio State University, Department of Linguistics
.
Massam,
Diane
.
2006
. Neither absolutive nor ergative is nominative or accusative. In
Ergativity
, ed. by
Alana
Johns
,
Diane
Massam
, and
Juvenal
Ndayiragije
,
27
46
.
Dordrecht
:
Springer
.
Matushansky,
Ora
.
2002
.
Movement of degree/Degree of movement
.
Doctoral dissertation, MIT, Cambridge, MA
.
Murasugi,
Keiko
, and
Mamoru
Saito
.
1994
. Adjunction and cyclicity. In
Proceedings of the 13th West Coast Conference on Formal Linguistics
, ed. by
Raúl
Aranovich
,
William
Byrne
,
Susanne
Preuss
, and
Martha
Senturia
,
302
317
.
Stanford, CA
:
CSLI Publications
.
Nakamura,
Masanori
.
1996
.
Economy of chain formation
.
Doctoral dissertation, McGill University, Montreal
.
Nevins,
Andrew
.
2011
.
Prospects and challenges for a clitic analysis of (A)SL agreement
.
Theoretical Linguistics
37
:
173
187
.
Nikolaeva,
Irina
.
1999a
.
Object agreement, grammatical relations, and information structure
.
Studies in Language
23
:
331
376
.
Nikolaeva,
Irina
.
1999b
.
Ostyak
.
Munich
:
Lincom Europa
.
Nikolaeva,
Irina
, and
Maria
Tolskaya
.
2001
.
A grammar of Udihe
.
Berlin
:
Walter de Gruyter
.
Ntelitheos,
Dimitris
.
2002
.
Possessor extraction in Greek and the left periphery of DP
.
Ms., UCLA, Los Angeles
.
Polinsky,
Maria
.
2015
.
Tsez syntax: A description
.
Ms., Harvard University, Cambridge, MA
.
Polinsky,
Maria
, and
Eric
Potsdam
.
2001
.
Long-distance agreement and topic in Tsez
.
Natural Language and Linguistic Theory
19
:
583
646
.
Preminger,
Omer
.
2009
.
Breaking agreements: Distinguishing agreement and clitic doubling by their failures
.
Linguistic Inquiry
40
:
619
666
.
Preminger,
Omer
.
2011
.
Agreement as a fallible operation
.
Doctoral dissertation, MIT, Cambridge, MA
.
Preminger,
Omer
.
2014
.
Agreement and its failures
.
Cambridge, MA
:
MIT Press
.
Rackowski,
Andrea
.
2002
.
The structure of Tagalog: Specificity, voice, and the distribution of arguments
.
Doctoral dissertation, MIT, Cambridge, MA
.
Rackowski,
Andrea
, and
Norvin
Richards
.
2005
.
Phase edge and extraction: A Tagalog case study
.
Linguistic Inquiry
36
:
565
599
.
Richards,
Norvin
.
2010
.
Uttering trees
.
Cambridge, MA
:
MIT Press
.
Richards,
Norvin
.
2016
.
Contiguity Theory
.
Cambridge, MA
:
MIT Press
.
Saito,
Mamoru
, and
Keiko
Murasugi
.
1999
. Subject predication within IP and DP. In
Beyond principles and parameters
, ed. by
Kyle
Johnson
and
Ian
Roberts
,
167
188
.
Berlin
:
Kluwer
.
Schütze,
Carson
.
1997
.
INFL in child and adult language: Agreement, case and licensing
.
Doctoral dissertation, MIT, Cambridge, MA
.
Seiter,
William J
.
1980
.
Studies in Niuean syntax
.
New York
:
Garland
.
Sperlich,
Wolfgang
, ed.
1997
.
Tohi vagahau Niue: Niue language dictionary
.
Honolulu
:
University of Hawai’i Press
.
Toosarvandani,
Maziar
.
2014
.
Two types of deverbal nominalization in Northern Paiute
.
Language
90
:
786
833
.
van Urk,
Coppe
.
2015
.
A uniform syntax for phrasal movement: A case study of Dinka Bor
.
Doctoral dissertation, MIT, Cambridge, MA
.