Abstract

In Xhosa VSO clauses, subject agreement exhibits default features, objects cannot be pronominalized, a subject focus reading is obligatory, and experiencer verbs with two DP arguments are precluded. We argue that impoverished versions of T and v* in VSO clauses lack the probe features involved in subject agreement, EPP, object shift, and nominative/accusative valuation within Xhosa SVO sentences. Only an unusual focus-linked strategy can Case-license full DPs in VSO clauses, but this is incompatible with inherent Cases borne by arguments of experiencer verbs. We show that CPs and augmentless NPs appear in positions where DPs cannot surface because uCase is a feature of D. Given the striking evidence for abstract Case in Xhosa, we propose Case-friendly analyses for Bantu Case-theoretic anomalies that Xhosa shares.

1 Introduction

1.1 Overview

In the Xhosa language (narrow Bantu, S.40), canonical word order is Subject Verb (Object . . . ). The verb agrees with the preverbal subject in person, number, and noun class.1

(1)

  • a.

    I-ncwadi i-fik-ile.

    9-9letter 9SA-arrive-DISJ1

    ‘A letter arrived.’

  • b.

    A-ba-ntwana ba-fund-a i-si-Xhosa.

    2-2-children 2SA-learn-FV 7-7-Xhosa

    ‘The children study Xhosa.’

  • c.

    U-Themba u-fund-is-é a-ba-ntwana i-si-Xhosa.

    1-1Themba 1SA-learn-CAUS-CONJ1 2-2-child 7-7-Xhosa

    ‘Themba taught the children Xhosa.’

VS(O) constructions with default subject agreement (henceforth SA) are also possible. Henceforth, we refer to these by their traditional label, expletive constructions (ECs), though this should be understood as a matter of convention. We will argue that while SA features are indeed expletive in VSO clauses, there is nothing in the preverbal position where subjects typically appear in SVO sentences.

(2a) illustrates an unaccusative EC. As the glosses indicate, two readings (R1 and R2) are available for this sentence. It can function as a simple report or narration of a past event and is thus a felicitous answer to a ‘What happened?’ question like (2b). It can also convey subject focus and answer one of the subject questions in (2c–d).2

(2)

  • a.

    graphic

    17SA-arrive-CONJ1 9-9letter

    R1: ‘A letter arrived.’

    R2: ‘It was a letter that arrived.’

    [Lit.: (There) arrived a letter]

  • b.

    Kw-enzek-é ntoni namhlanje?

    17SA-happen-CONJ1 9what today

    ‘What happened today?’

    [Lit.: (There) happened what today]

  • c.

    Ku-fik-é ntoni?

    17SA-arrive-CONJ1 9what

    ‘What arrived?’

    [Lit.: (There) arrived what]

  • d.

    Y-i-ntoni e-fik-ile-yo?

    COP-9-what 9REL-arrive-DISJ1-RM

    ‘What is it that arrived?’

All intransitive verbs seem able to participate in Xhosa ECs, and some speakers find transitive expletive constructions (TECs) acceptable as well (a fact noted in Mali 1995, Mletshe 1995). But there are interesting asymmetries between TECs and intransitive ECs: TECs exhibit special properties and are subject to a number of constraints that we summarize in (3). Accounting for the full set of EC properties is the goal of our article.

(3) Properties of ECs in Xhosa

  • a.

    There are no definiteness effects for the postverbal arguments in an EC, whether transitive or intransitive.

  • b.

    The inverted subject of a TEC is obligatorily [+Focus]. In contrast, the subject focus interpretation is optional in intransitive ECs.

  • c.

    An internal argument in a TEC cannot be realized as a pronoun—never as the objectmarking variety and, for most speakers, not as an independent pronoun.3

  • d.

    In contrast, the external argument of a TEC or the sole argument of any intransitive EC may be an independent pronoun.

  • e.

    A verb with an experiencer argument cannot participate in a TEC unless

    • i.

      its internal argument is clausal, or

    • ii.

      its external argument is removed by passivization, or

    • iii.

      its arguments are both ‘‘augmentless’’ nouns used as question words or negative polarity items (an outer noun class prefix is omitted; see (4a–b)).

The ‘‘augmented’’/‘‘augmentless’’ distinction referred to in (3eiii) is illustrated in (4). Augmentless nominals can be recognized by their lack of a prefix vowel and by underlining.

(4)

  • a.

    ‘‘Augmented’’ or pre-prefixed nominals = full DPs and citation forms of wh-words

    graphic

  • b.

    ‘‘Augmentless’’ nominals (henceforth abbreviated [−A] and underlined) function as negative polarity items (NPIs) and wh-words in questions

    graphic

The asymmetries summarized in (3) are exemplified in (5)–(7).

(5) The focus asymmetry: In TECs but not intransitive ECs, the subject must be focused

graphic

(6) The pronominalization asymmetry: In TECs, only the subject can be pronominalized

graphic

(7) Experiencer verbs with two full DP arguments cannot participate in TECs

graphic

1.2 Sketch of the Analysis

We argue that Xhosa ECs are clauses in which T and v* are defective—unable to agree, raise DPs, or value Case (see (8a–b)). The properties in (3), (5), and (7) are indicative of noncanonical Case licensing. In contrast, SV(O) syntax involves robust Case, agreement, and A-movement (see (8c)). These findings are significant because the status of Case in Bantu languages is controversial (see Harford Perez 1985, Diercks 2012, Halpert 2012, Van der Wal 2012), as is the (non)universality of the EPP (see, e.g., Cable 2012). They also indicate a tight, unexpected dependency between features of v and T (see (9)).

(8)

graphic

(9) Profile of two opposing Xhosa clause types

  • a.

    SVO: [TP T[+EPP/Agr/Case] . . . [vP v*[+EPP/+Agr/Case] . . . ]]

  • b.

    VSO: [TP T[−EPP/Agr/Case] . . . [vP v*[−EPP/Agr/Case] . . . ]]

    In Xhosa, defective v*defective T.

The correlation of obligatory focus with transitivity in Xhosa ECs is the first indicator that Case is an issue in TECs, and the experiencer verb contrasts in (7) provide the second. Both patterns illustrate that problems arise if a TEC contains two augmented nominals, henceforth [+A]. We analyze the augment as D, and [+A] nominals as full DPs needing Case values because uCase is a feature of D.

Since the sole argument in any intransitive or passive EC can be a [+A] DP with a neutral, nonfocused interpretation, we propose that there is one purely structural Case optionally available for the highest argument in an EC (section 4.2.3 suggests this Case is valued by a particular choice of null C). In TECs, the inclusion of a second nominal creates a Case-theoretic challenge. We interpret the focus-transitivity connection as a strategy to surmount the absence of accusative: a middle field Focus head probes both subject and object,4 raising the former to its specifier (Spec) and giving both arguments Case values linked to [+/−Focus] features (see (10)).5

(10) The Focus head probes subject and object; linked Case and Focus values result

graphic

(7a) shows that DP arguments of experiencer verbs cannot be licensed via this strategy. In keeping with a common pattern in languages with overt Case systems, we propose that arguments of experiencer verbs bear inherent Cases. Adding the augment gives them an unvalued uCase feature as well. The inherent Cases are compatible with purely structural Cases; hence, experiencer verbs are licit in VS clauses if they have only one DP argument (see (7b–c)) and in SVO clauses (see (11)).

(11)

  • U-Sindiswa u-bon-é u-gqirha.

  • 1-1Sindiswa 1SA-see-CONJ1 1-1doctor

  • ‘Sindiswa saw the doctor.’

But the inherent Cases of an experiencer verb’s arguments are not compatible with a second semantically linked Case (see (12)).6

(12)

  • The Semantic Case Constraint

  • *DP bearing more than one semantically linked Case.

If an experiencer verb in a TEC has a [+A] object, the result is ill-formed since the only available Case value is linked to a Focus value (see (13)). (7c–d) show that TECs of experiencer verbs are possible if the object is a bare NP (= [−A]) or a CP. This follows if NPs and CPs lack uCase features, so structural Case licensing is inapplicable for them.

(13) An experiencer verb with a [+A] DP object cannot participate in a TEC. The only Case available for this DP would be Focus-linked, violating (12).

graphic

An implication is that Case values coming from the Xhosa Focus head straddle the line between structural and inherent: they have semantic associations, but are related to particular structural positions and compatible with a variety of thematic roles.

Turning to the restriction on pronoun use, we attribute this to a failure of object shift in TECs, stranding pronouns illicitly in VP-internal positions.

(14)

graphic

Summing up, we relate the absence of accusative and the unavailability of pronouns and object shift to defectivity of v* in Xhosa TECs (see Bobaljik and Branigan 2006 for a similar approach to ergative constructions in Chukchi). It is an unlikely coincidence that defective v* occurs in Xhosa clauses with low subjects and default SA. Absent evidence that probe features of T participate in EC syntax, we analyze T as lacking EPP, Case, and uφ (uninterpretable φ-features) in Xhosa ECs. Hence, T of an EC cannot agree, Case-license, or raise the subject, just as v* of an EC cannot shift objects or value accusative (following Preminger (2011), agreement failures do not cause a derivation to crash).7 In contrast, Xhosa SVO clauses have robust T and v*, that is, versions equipped with the standard probe and edge features.

While there is no evidence that Spec,TP is occupied in Xhosa ECs, we suggest that crosslinguistically, the use of expletive subjects might be triggered when T has an EPP feature but is unable to raise a thematic subject. Our idea is to extend to other ECs the last-resort-type approach common for ‘‘weather’’ verb constructions and English do-support.

1.3 Implications for Case in Bantu

This article makes a novel contribution to an important controversy over the status of structural Case in Bantu languages and hence its plausibility as a linguistic universal. Harford Perez (1985) and Diercks (2012) claim that Case is entirely absent in Bantu, and Halpert (2012) proposes that apparent Case theory violations in Zulu arise because [+A] nominals have intrinsic Case licensing, unlike [−A] NPIs. On the basis of the asymmetries summarized in (3), we argue that full DPs in Xhosa require Case licensing. But we show that they exhibit the same Case anomalies as Zulu, which include participating in hyperraising and occupying what appear to be Caseless positions. We approach these phenomena in ways compatible with the presence of abstract Case in Xhosa and conclude that the classic diagnostics yield misleading results. The pattern of facts strongly suggests that abstract Case is present in Bantu languages, but manifested in unexpected ways. Thus, contrasting properties of VSO and SVO clauses in Xhosa provide a novel window on the workings of agreement, A-movement, and Case.

1.4 Structure of the Article

This article is structured as follows. In section 1.5, we summarize our theoretical assumptions. In section 2, we describe in more detail the striking asymmetries that characterize Xhosa ECs. In section 3, we review analyses by Buell (2006) and Halpert (2012) of ECs in closely related Zulu, presenting insights they provide into the Xhosa facts and also several key questions that they cannot answer. In section 4, we flesh out our proposals in terms of defective T, defective v*, and a Focus Phrase between the two, whose head (for speakers who find TECs acceptable) values uCase. In section 5, we address apparent Case anomalies in Xhosa (and other Bantu languages that exhibit them), and in section 6, we provide conclusions and directions for future research.

1.5 Theoretical Background

Our article is written within the Minimalist theoretical framework of Chomsky 2000, 2001. We assume that syntactic objects are constructed bottom to top and that there is cyclic Transfer to the PF and LF interfaces triggered by the phase heads v* and C. Following Chomsky, we assume that agreement and structural Case are uninterpretable, unvalued features (uFs): uφ and uCase, respectively. When uφ is merged on a category α, it immediately probes its c-command domain to find a goal β that can provide values for α’s uFs. We assume a version of Chomsky’s (2000, 2001) ‘‘Activity Condition’’: that a participant in an Agree relation must bear a uF. Following Bošković (2011), we assume that Case-valuing heads have valued uCase features. Hence, the robust versions of T and v come from the lexicon with uNom and uAcc features, respectively, and confer these values on the unvalued uCase features of local DPs through the Agree relation.

2 Asymmetries in Xhosa ECs

2.1 The Empirical Puzzles

In this section, we lay out in detail the asymmetries that characterize Xhosa ECs. The analysis will be presented in sections 3 and 4.

2.2 Asymmetry 1: Focus and Transitivity in ECs

Subjects of all Xhosa ECs can be interpreted as focused. This is illustrated in (2a) (repeated here) for the unaccusative verb ‘arrive’ and in (15) for the unergative verb ‘sing’.

(2)

graphic

(15)

graphic

In contrast, the subject of a TEC has an invariant focus reading.

(16)

  • a.

    graphic

    17SA-buy-FV 2-2-women 10-10-flowers

    ‘It’s the women who buy flowers.’

  • b.

    Ku-bhaq-é u-Sindiswa i-mali.

    17SA-discover-CONJ1 1-1Sindiswa 9-9money

    ‘It was Sindiswa who discovered the money.’

That subject focus is obligatory in TECs is confirmed by two diagnostics. First, only an intransitive EC is a felicitous answer to a ‘What happened?’ question (see (17a–c)). Second, an indefinite can be subject of an SVO sentence (see (18a)) or subject of an intransitive EC (see (18b,d)) but is anomalous as subject of a TEC (see (18c)).8

(17)

  • a.

    Kw-enzek-é ntoni?

    17SA-happen-CONJ1 9what

    ‘What happened?’

    [Lit.: (There) happened what]

  • b.

    Ku-cul-é u-Sindiswa.

    17SA-sing-CONJ1 1-1Sindiswa

    ‘Sindiswa sang.’

    [Lit.: (There) sang Sindiswa]

  • c.

    graphic

    17SA-sing-CONJ1 1-1Sindiswa 6-6-songs

    ‘It was Sindiswa who sang songs.’

    [Lit.: (There) sang Sindiswa songs]

(18)

graphic

It is important to note that a sentence like (19) with [SV . . . ] word order does not have obligatory focus on the first postverbal constituent (though focus is optionally available postverbally in such clauses if the vP is otherwise empty; see section 4.4). Hence, it is clear that there is no general Xhosa requirement that an expression immediately following the verb be focused.9

(19)

  • U-mama u-nik-é a-ba-ntwana i-ncwadi.

  • 1-1mother 1SA-give-CONJ1 2-2-children 9-9book

  • ‘Mother gave the children a book.’

The examples in (17) and (18) make it very clear that Xhosa lacks the definiteness effect that characterizes ECs in Indo-European languages, as discussed in Milsark 1977, Safir 1987, 2009, Belletti 1988, and Vangsnes 2002, among many others; see (20a–c), adapted from Belletti 1988: 3, 4, and 14, and (21), from Vangsnes 2002:44.

(20)

graphic

(21)

graphic

Xhosa also clearly lacks the restriction to unaccusatives that is common for ECs in more familiar languages. This, the focus-transitivity correlation, and the difference between Xhosa and Germanic ECs with respect to definiteness all require explanation. We argue in section 4 that these striking characteristics of Xhosa ECs are due to noncanonical, Focus-linked Case licensing necessitated by the absence of probe features on T and v of Xhosa ECs.

2.3 Asymmetry 2: Pronominalization in ECs

In Xhosa SVO clauses, an internal argument can be realized as a pronoun—either an independent pronoun (22a) or a clitic-like object marker (22b). In contrast, the internal argument in a TEC cannot be a pronoun of either variety (see (22c–d)).10 The subject of a TEC or the sole argument of any intransitive EC can, however, be pronominalized (see (23)).

(22)

graphic

(23)

graphic

2.4 Asymmetry 3: Experiencer Verb Restrictions

A verb with an experiencer argument cannot participate in a TEC unless (a) its external argument is removed by passivization; (b) its internal argument is a CP; or (c) both arguments of the verb are NPIs, lacking the augment vowel (see (7), repeated here).

(7)

graphic

Tables 1 and 2 summarize the verbs tested in Xhosa ECs, and their status. With very few exceptions, transitive experiencer verbs resist involvement in TECs.

Table 1.

Verbs with DP arguments that are acceptable in ECs

ukufika ‘arrive’ ukulumka ‘be.wise’ 
ukutsha ‘burn’ ukuxhuma ‘jump’ 
ukucula ‘sing’ ukuthetha ‘speak’ 
ukulila ‘cry’ ukutshaya ‘smoke’ (intrans. & trans.) 
ukuonwaba ‘be happy’ ukunuka ‘smell’ (intrans.) 
ukuqumba ‘be.sad’ ukuthenga ‘buy’ 
ukuzila ‘mourn’ (intrans.) ukuthengisa ‘sell’ (buy + CAUS
ukurhala ‘yearn’ ukutsiba ‘jump over’ 
ukubhala ‘write’ ukuoyika ‘fear’ 
ukulumkisa ‘warn’ ukubukela ‘watch’ 
ukufunda ‘learn/read’ ukufundisa ‘teach’ (learn + CAUS
ukupheka ‘cook’ ukurhalela ‘desire’ (with inanimate direct object) 
ukutyhola ‘accuse/blame’ ukukhumbula ‘remember’ (with inanimate direct object) 
ukubhaqa ‘discover’ ukubuhlungu ‘pain’ 
ukufika ‘arrive’ ukulumka ‘be.wise’ 
ukutsha ‘burn’ ukuxhuma ‘jump’ 
ukucula ‘sing’ ukuthetha ‘speak’ 
ukulila ‘cry’ ukutshaya ‘smoke’ (intrans. & trans.) 
ukuonwaba ‘be happy’ ukunuka ‘smell’ (intrans.) 
ukuqumba ‘be.sad’ ukuthenga ‘buy’ 
ukuzila ‘mourn’ (intrans.) ukuthengisa ‘sell’ (buy + CAUS
ukurhala ‘yearn’ ukutsiba ‘jump over’ 
ukubhala ‘write’ ukuoyika ‘fear’ 
ukulumkisa ‘warn’ ukubukela ‘watch’ 
ukufunda ‘learn/read’ ukufundisa ‘teach’ (learn + CAUS
ukupheka ‘cook’ ukurhalela ‘desire’ (with inanimate direct object) 
ukutyhola ‘accuse/blame’ ukukhumbula ‘remember’ (with inanimate direct object) 
ukubhaqa ‘discover’ ukubuhlungu ‘pain’ 
Table 2.

Verbs with DP arguments that cannot participate in active ECs

*ukuthusa ‘surprise’ *ukuthakazelela ‘appreciate’ 
*ukwazi ‘know’ *ukuthaza ‘make angry’ 
*ukucinga ‘think’ *ukuzilela ‘mourn’ (trans.) 
*ukucapukela ‘hate’ *ukunukisa ‘smell’ (trans.) 
*ukuthanda ‘like’ *ukubona ‘see’ 
*ukufuna ‘want’ *ukukhumbula ‘miss’ 
*ukuva ‘hear’ *ukurhalela ‘desire sexually’ (yearn + APPL; human direct object) 
*ukuthusa ‘surprise’ *ukuthakazelela ‘appreciate’ 
*ukwazi ‘know’ *ukuthaza ‘make angry’ 
*ukucinga ‘think’ *ukuzilela ‘mourn’ (trans.) 
*ukucapukela ‘hate’ *ukunukisa ‘smell’ (trans.) 
*ukuthanda ‘like’ *ukubona ‘see’ 
*ukufuna ‘want’ *ukukhumbula ‘miss’ 
*ukuva ‘hear’ *ukurhalela ‘desire sexually’ (yearn + APPL; human direct object) 

2.5 Summary

This section has demonstrated that TECs in Xhosa have properties that distinguish them from intransitive ECs, from SVO clauses, and from TECs in more familiar languages including absence of the definiteness effect, obligatory subject focus, a ban on object pronouns, and incompatibility with experiencer verbs having two full DP arguments.

The next section reviews existing analyses of Nguni ECs and shows that they provide insights into the structure, but on their own they cannot account for the pattern of facts we have described. Section 4 builds on these analyses to provide principled explanations.

3 Building a Structural Analysis

3.1 Introduction

Xhosa is a member of the Nguni subgroup, which includes the closely related Zulu language. Nguni ECs are addressed in several other works, including Mali 1995, Mletshe 1995, Buell 2006, and Halpert 2012. Sections 3.23.4 discuss some analytical contributions from these works that we will draw on in constructing our account. Sections 3.53.7 present three additional diagnostics for the analysis of Xhosa expletives. Section 3.5 shows that VSO order is only possible in a Xhosa EC, arguing that agreement with the thematic subject correlates with its raising to Spec, TP and conversely that in its absence the subject has not done so. Section 3.6 shows that auxiliary verbs must all precede the thematic subject in an EC, supporting a low subject position in TECs. Section 3.7 shows that ECs can occur in embedded clauses following an overt complementizer. This provides a final argument against a potential alternative structural analysis positioning the verb in C and the postverbal subject in Spec,TP. Section 3.8 summarizes the conclusions reached in the section regarding the structure of ECs, and the issues they leave to be explained.

3.2 Unaccusatives and the Conjoint/Disjoint Diagnostic (Buell 2006)

Xhosa ECs resemble those of Zulu in featuring invariant class 17 agreement and VS(O) order. Buell (2006) provides a diagnostic for the low position of the postverbal arguments in Zulu ECs that we will adopt in our analysis of Xhosa ECs. Like Zulu verbs, Xhosa verbs have some tense/ aspect alternations that correlate primarily with whether the verb is final in some minimal domain to be made specific below. These alternations are demonstrated in (24) for the optionally transitive verb funda ‘study’. In the linguistic literature, the inflected form of verbs in final position is generally referred to as the long or disjoint form, while the nonfinal form is referred to as short or conjoint.

(24)

graphic

While this kind of alternation has sometimes been attributed to the presence or absence of verb focus (Hyman and Watters 1984, Ndayiragije 1999), Buell (2006) argues against such an interpretation partly because either conjoint or disjoint morphology can appear in the answer to a ‘What happened?’ question depending on whether the verb has a complement; in this circumstance, the domain of focus is the whole sentence. We therefore adopt Buell’s proposal that the crucial factor is whether a constituent follows the verb within a local domain (see also Mali 1995, Van der Spuy 1993)—hence (25a–b). We analyze the domain as TP, but were the verb to raise to C and precede a subject in Spec,TP, we suspect that the conjoint form would still be required—hence the broad formulation in (26) (anticipating somewhat the pattern of tense morphology in ECs to be described below).

(25) Conjoint/Disjoint distribution (Buell 2006)

  • a.

    A disjoint verb form is final in its domain.

  • b.

    A conjoint verb form is nonfinal in its domain.

(26)

  • Condition on conjoint/disjoint forms

  • T in the conjoint form must c-command an expression with intrinsic φ-features; a disjoint form cannot.11

  • Buell (2006) demonstrates that the conjoint/disjoint alternation sheds light on inversion constructions in Zulu, and his observations extend to Xhosa. Consider (27a): the unaccusative subject is preverbal, leaving the verb final in its domain save for the subject’s unpronounced copy. Verbal morphology is accordingly disjoint. In contrast, the conjoint form is required in an EC like (27b). The appearance of the conjoint form in an unaccusative EC makes perfect sense if the theme subject remains in its base position so that the verb is nonfinal in its domain, ccommanding an overt DP. Thus, slightly adapting Buell’s proposals for these constructions, we arrive at (28a) and (28b), respectively. (Here and throughout, morphological structures and glosses in trees are simplified.)

(27)

  • a.

    I-ncwadi i-fik-ile/*é.

    9-9letter 9SA-arrive-DISJ1/*CONJ1

    ‘A letter arrived.’

  • b.

    Ku-fik-é/*ile i-ncwadi.

    17SA-arrive-CONJ1/*DISJ1 9-9letter

    ‘A letter arrived.’

(28)

graphic

On the other hand, when a postverbal subject controls SA, the verb takes the disjoint form (see (29a)). This follows from the reasonable assumption that the overt subject in such cases is not in situ but in a right-dislocated position. Building on Buell 2006 and Halpert 2012, we assume that doubling by SA or by an object clitic is a prerequisite for right-dislocation of argument DPs.12 (29b) illustrates two plausible hypotheses: one under which the right-dislocated subject in (29a) moves there via Spec,TP, and an alternative under which the structural subject is pro, doubled by the DP merged in right-adjoined position (both ideas come from Buell 2006). The choice between these options is tangential to present concerns, so we leave it aside.

(29)

graphic

3.3 Unergatives

Unergative verbs with preverbal subjects bear disjoint morphology, as expected. In an unergative EC, (26) correctly predicts the conjoint form since T c-commands the subject, as in (30c).

(30)

  • a.

    U-Sabelo u-ya-tshay-a.

    1-1Sabelo 1SA-DISJ2-smoke-FV

    ‘Sabelo smokes.’

  • b.

    Ku-tshay-a u-Sabelo

    17SA-smoke-FV 1-1Sabelo

    ‘(It’s) Sabelo (who) smokes.’

  • c.

    graphic

3.4 TECs in Halpert 2012 

We have noted that many Xhosa speakers accept TECs (see (31b)).

(31)

  • a.

    U-Sabelo u-tshay-a i-cuba.

    1-1Sabelo 1SA-smoke-FV 5-5tobacco

    ‘Sabelo smokes tobacco.’

  • b.

    Ku-tshay-a u-Sabelo i-cuba.

    17SA-smoke-FV 1-1Sabelo 5-5tobacco

    ‘It’s Sabelo who smokes tobacco.’

Halpert (2012) shows that TECs exist in Zulu and proposes that their arguments remain in situ. She posits a ‘‘downward’’ Case licenser between T and vP that she labels L (for simplicity, presumed roll-up of L into the verbal complex is not depicted in (32)).

(32)

graphic

3.5 VSO Order Entails Expletive Agreement

Halpert’s (2012) analysis is consistent with Buell’s (2006) in assuming that the VSO subject is unraised. In addition to the conjoint/disjoint facts, there is support for extending this conclusion to Xhosa in that VSO order is impossible if the verb agrees with the subject (see (33)). In this case, the only licit orders are SVO and VOS. The results confirm our proposal that SA correlates with subject raising to Spec,TP.

(33)

graphic

This contrast supports our claim that EPP and uφ probe features of T go together, while a defective T lacks both of these properties. A sentence like (33b) cannot be generated because T agrees with the subject but leaves it in situ. In (33a), T agrees with and raises the subject (whether pro or the right-adjoined DP; see the intransitive VS clause in (29a–b)).

3.6 Evidence from Auxiliaries

Although the above facts have argued against a right-adjoined position for the postverbal subject in an EC, they have not ruled out the possibility that the subject undergoes raising to the middle field of the clause as in the analysis of Icelandic in Bobaljik and Jonas 1996. (34) (from Vangsnes 2002) illustrates two positions for subjects within Icelandic ECs.

(34)

graphic

In contrast, the subject in a Xhosa EC must follow all auxiliaries. We demonstrate in (35) with the auxiliary phantse ‘almost’ and in (36) with the combination of soloko ‘often’ and a remote future auxiliary13 (for arguments that such multiply agreeing constructions are truly monoclausal in Bantu languages, see among others Slattery 1981, Carstens and Kinyalolo 1989, Kinyalolo 1991, Carstens 2001, and Halpert 2012).

(35)

  • a.

    U-Thandeka u-phantse w-aty-a i-papa.

    1-1Thandeka 1SA-almost 1SA-eat-FV 9-9polenta

    ‘Thandeka almost ate the polenta.’

  • b.

    Ku-phantse kw-aty-a u-Thandeka i-papa.

    17SA-almost 17-eat-FV 1-1Thandeka 9-9polenta

    ‘It was Thandeka who almost ate the polenta.’

  • c.

    *Ku-phantse u-Thandeka kw-aty-a/w-aty-a i-papa.

(36)

  • a.

    Wena u-be u-soloko u-fund-a lapha.

    2SG.IP 2SG.SA-RFUT 2SG.SA-often 2SG.SA-study-FV here

    ‘You will often study here.’

  • b.

    Ku-be ku-soloko ku-fund-a wena lapha.

    17SA-RFUT 17SA-often 17SA-study-FV 2SG.IP here

    ‘(It’s) you (who) will often study here.’

  • c.

    *Ku-be wena ku-soloko ku-fund-a lapha.

  • d.

    *Ku-be ku-soloko wena ku-fund-a lapha.

(35)–(36) help to flesh out the structure of Xhosa clauses. Anticipating evidence to be presented in section 4.2, we propose that a Focus Phrase (FocP) lies just above vP (see (37a)). In a clause like (36b), there are two freestanding auxiliaries and the subject can have a focused interpretation that we will tie to occupancy of Spec,FocP. The fact that the thematic verb nonetheless precedes the subject suggests that Vmain always raises beyond vP and Spec,FocP. Following proposals by Ngonyani (1999, 2006), Kinyalolo (2003), and Carstens (2005), we assume that the final vowel of the inflected verb is a midlevel functional head, Mood, to which the verb always adjoins. These assumptions are sketched in (37b–c) (_ (36b)).

(37)

graphic

3.7 ECs in Embedded Clauses

A question remains regarding the location of the highest inflected verb in a Xhosa EC. (38)–(39) show that ECs are possible in embedded clauses following overt complementizers. Putting together the full set of facts, we arrive at the representation in (40b) for (40a).

(38)

  • U-Sabelo u-cing-a [CP okokuba ku-fund-is-a u-Loyiso i-si-Xhosa].

  • 1-1Sabelo 1SA-think-FV that 17SA-learn-CAUS-FV 1-1Loyiso 7-7-Xhosa

  • ‘Sabelo thinks that it’s Loyiso who teaches Xhosa.’

(39)

  • U-Sipho u-buz-é [CP okokuba ngabe ku-bhal-é u-m-ntwana i-ncwadi na].

  • 1-1Sipho 1SA-ask-CONJ1 that whether 17SA-write-CONJ1 1-1-child 9-9letter Q

  • ‘Thandeka asked if it was the child who had written the letter.’

(40)

graphic

3.8 Summary

Buell (2006) provides useful diagnostics for determining that the arguments in Nguni ECs are not in a right-adjoined position; rather, they are clause-internal, c-commanded by the tense/aspect heads that precede them. Halpert (2012) extends this approach to Zulu TECs. We have shown in this section that both Buell’s and Halpert’s diagnostics are relevant to Xhosa, making correct predictions regarding the word order and morphology in ECs. The additional diagnostics of auxiliary placement, embedded ECs, and the correlation of VSO order with expletive SA also argue that the subject surfaces low in ECs and that the highest inflected verb of an EC is located lower than C.

4 Analysis in Detail

4.1 Introduction

In this section, we develop in detail our analytical approach to the properties of Xhosa ECs, addressing the questions we have already raised, summarized here:

(41) Issues in the analysis of Xhosa ECs

  • a.

    Focus is optional for subjects of intransitive verbs but obligatory for subjects of transitive verbs.

  • b.

    There is no definiteness effect constraining low, postverbal DPs in Xhosa ECs.

  • c.

    The subject of an EC can pronominalize but an object cannot.

  • d.

    An experiencer verb is illicit in a TEC unless (i) it is passivized, (ii) its internal argument is a CP, or (iii) its arguments are [−A] (augmentless) nominals.

Section 4.2 first argues that the focus reading for subjects in ECs indicates that they can raise into Spec,Foc. Given that subject focus is obligatory in TECs, we propose that this raising permits noncanonical Case licensing, in the grammars of speakers who accept TECs. Section 4.3 presents a Case-theoretic account of restrictions on transitive experiencer verbs in TECs. Section 4.4 applies proposals made by Cheng and Downing (2012) to derive an optional focus reading for experiencer subjects of intransitives and verbs with CP objects. Section 4.5 attributes the impossibility of object pronouns to a failure of object shift in TECs. Section 4.6 discusses the Subjectin- Situ Generalization (Alexiadou and Anagnostopoulou 2001, 2007), Chomsky’s (2013) labeling algorithm, and Richards’s (2010) Distinctness requirement, arguing that alternative approaches to Xhosa along these lines are less successful at deriving all the facts. Section 4.7 argues against the presence of an expletive pro in Spec,TP, given the combination of default SA with the peculiar syntax of objects in TECs. Under our analysis, the properties of T and v mirror each other in TECs. Section 4.8 shows that our analysis is compatible with several approaches to the definiteness effect in ECs of familiar languages. Section 4.9 concludes the section.

4.2 Deriving the Focus Interpretation

4.2.1 Intransitives

Consider the intransitive EC in (42a). As indicated, it can be the simple narration of a past event and a potential answer to ‘What happened?’ It can also have a subject focus reading. As previously mentioned and sketched out in (40b), we propose that this difference reduces to whether or not a middle field Focus head is present to raise the subject to its Spec (though see section 4.5 for a slightly more complex picture). Hence, the representations in (42b–c) correspond to the two readings.14

(42)

graphic

4.2.2 Obligatory Focus in TECs

The fact that subjects of TECs must have the focus reading strongly suggests that they cannot remain in their base positions; they are forced to raise to Spec,FocP. Thus, the only representation for (43a) is (43b).

(43)

graphic

Why should raising be obligatory in TECs but optional in intransitive ECs? The need for Case is a common factor underlying forced movement of arguments.15 But the optionality of the focus reading for subjects of intransitives argues that raising is not required in ECs to Case-license the subject. We accordingly propose that v* of Xhosa TECs cannot value accusative. There is just one structural Case in a Xhosa EC independent of the Focus head, and it goes to the highest argument (see (8b), repeated here). In the grammars of many speakers, this limitation cannot be surmounted and TECs are disallowed. But in the grammars of those who do accept TECs, we propose that raising the subject to Spec,FocP permits Case licensing of a second DP.16

(8)

graphic

A locality question arises since, for any potential licenser above FocP, the subject would intervene to block access to the direct object (see (44)). We accordingly propose that Focus can undergo Agree with both subject and object, with the result that the object is Case-licensed. Section 4.3.2 presents evidence that the Case value that the object so acquires has a semantic component, which we identify as linkage to a [−Focus] feature. It might be that only [−Focus] is linked to a Case value in this way, but in the interests of symmetry and avoiding stipulation, we suggest that both Focus values are linked to Case values (see (45a–b)).17

(44) A probe above FocP would not be expected to reach across the subject in Spec,Foc

graphic

(45)

  • a.

    The Focus head probes and raises the subject to its Spec, giving it [+Focus] and a linked Case value

    graphic

  • b.

    The Focus head probes the object, giving it [−Focus] and a linked Case value

    graphic

This analysis builds on Haegeman and Lohndal’s (2010) proposal for accomplishing multiple Agree relations serially, and on Halpert’s (2012) idea that Case licensing in closely related Zulu can happen late, across a position vacated by A-movement. Halpert draws a parallel that we adopt between this and a well-known Icelandic pattern of number agreement: an in-situ dative blocks T from agreeing in number with a nominative in Icelandic, but agreement succeeds across the same base position vacated by A-movement of the dative. In (46)–(48), we reproduce examples from Holmberg and Hróarsdóttir 2003:999–1000 and illustrations adapted from Halpert 2012: (46). Inspired by Halpert’s account, we propose that, as Icelandic T can probe and raise the dative and then agree with the lower nominative, so Xhosa Foc can raise and Case-value the subject, then probe the object as shown in (45a–b).

(46) Icelandic dative experiencers in situ block object agreement

graphic

(47) Icelandic raised dative experiencers do not block object agreement

graphic

(48) Icelandic dative raising feeds number agreement

graphic

Our account of Xhosa converges with Halpert’s approach to Zulu in assuming that v* cannot license accusative, but differs from Halpert’s approach in assuming this is true only in TECs. It differs also in proposing that [+A] nominals require Case, and in linking noncanonical Case licensing to the focus reading for postverbal subjects, verb class asymmetries, and pronominalization restrictions, all properties that our investigation is, to the best of our knowledge, the first to explore. Whether these phenomena also exist in Zulu is a question for future research (though see section 5 for some discussion of Case and DP positions in Zulu).

4.2.3 Case for a Single Argument without Focus

As noted above, assuming that Case is a major determinant of DP positions in Xhosa, it would seem that ‘‘downward’’ Case licensing is always available for a single postverbal argument in any EC (see (2a), (5a), (7b–c), and (42a)). Since this Case is not linked to any particular thematic role or semantic interpretation, we analyze it as a purely structural Case. It is possible that this Case is nominative, licensed downward by T, as attested in German and Icelandic (see Bobaljik and Wurmbrand 2005 on German, and the discussion of Icelandic in section 4.2.2). But this is not a necessary conclusion given that complementizers in some languages have the ability to independently confer a Case value. We reproduce Standard Arabic data from Melebari and Seely 2012 in (49) showing that while T values nominative, the C ʔanna values accusative. Given the apparent inertness of T’s and v*’s probe features in Xhosa ECs, we conjecture that there is a particular null C in Xhosa that, like ʔanna, can value Case on the argument it closest c-commands (see (50)).18 The hypothesis reflects and completes the symmetries between Xhosa T and v in TECs: neither one raises the argument that it typically raises in SVO clauses; neither agrees or values uCase features. Like T and v*, we assume the relevant C in Xhosa comes in varieties with and without Case valuation features.19

(49)

graphic

(50) A single purely structural Case in ECs comes perhaps from C, since T seems inert

graphic

A natural assumption would be that the subject in an EC always bears this single structural Case. But as we noted in section 4.2.2, experiencer verbs present evidence that the Case assigned by Focus to the direct object in a TEC has a semantic linkage, and we accordingly opt for a symmetrical approach under which the [+/−Focus] features are both linked to Case values. We turn to the evidence of experiencer predicates next.

4.3 Case and Experiencer Predicates

4.3.1 Prologue on [−A] NPIs

Contrasts in the distribution of [−A] and [+A] nominals play a significant role in the analysis of experiencer predicates in TECs that we develop in this section, and in the conclusions we draw from them regarding the workings of abstract Case in Xhosa.

Before we start on that topic, though, we need to introduce a constraint on [−A] NPI distribution that is constant across verb classes and that, in our view, has nothing to do with Case. This is a general ban on [−A] NPIs in the object position of TECs, illustrated in the contrast between (51a) and (51b). In the interests of providing complete paradigms, our presentation of the facts of nominal arguments of experiencer verbs in TECs will include such cases. So it is important to establish at the outset that they are part of a broader pattern and therefore fall under a broader solution rather than under our proposals for experiencer verbs. In what follows, we describe just enough of the facts and analysis from Carstens and Mletshe 2013 to make the pattern and the outline of an analysis clear.

(51)

graphic

In Xhosa, the ban on [−A] objects in TECs is overcome if there is a [−A] NPI subject in the same clause (see (52a)). In closely related Zulu, where an identical ban on [−A] direct objects in TECs is documented in Halpert 2012, a [−A] NPI object requires applied or causative morphology on the verb (see (52b), from Halpert 2012:106).

(52)

graphic

There are several further systematic constraints on the positions of [−A] NPIs. They cannot occupy preverbal subject position of an SVO clause, even embedded under negation (see (53a)). They cannot clitic-dislocate, and they cannot appear DP-internally if higher material in the DP is [+A]. In Carstens and Mletshe 2013, we account for the full pattern of facts by claiming that [−A] NPIs are like negative concord items in that they must raise to enter into a local licensing relation with a sentential operator, most commonly negation (interrogation and conditional also license them; details lie outside the scope of this article). The ill-formedness of (53a) is parallel to that of the French (53b) from Kayne 1981:318, attributed by Kayne to a that-trace effect prohibiting personne ‘no one’ from raising out of subject position into the matrix clause to interact with negation. The word order contrast in (54) illustrates that a comparable leftward movement requirement applies to West Flemish n-words (examples from Haegeman 1995:132).

(53)

graphic

(54) West Flemish n-words must move leftward in overt syntax

graphic

The ban on [−A] NPIs as objects in TECs arises because v* is defective and cannot accomplish object shift. We will show in section 4.5 that this conclusion has independent support in the ban on object pronouns in TECs.

(55) Like n-words, [−A] NPIs in Xhosa must move leftward to interact with a licensing operator; objects of TECs cannot do so because v* is defective in TECs

graphic

[−A] NPIs in the subject position of Xhosa and Zulu TECs satisfy the raising requirement by moving to Spec,FocP. The fact that a [−A] NPI object is acceptable in (52a) is attributed in Carstens and Mletshe 2013 to parasitic negation, a strategy documented in negative concord languages and likened to parasitic gaps ((56) and (57) taken from Haegeman 1995:220). Under parasitic negation, an n-word in an illicit location is rescued by a local n-word in a licit location.

(56)

graphic

(57)

  • a.

    *What did you file the papers [without reading _____]?

  • b.

    Which papers did you file _____ [without reading _____]?

Summing up, aspects of existing analyses of n-words account for several general restrictions on [−A] NPIs in terms of locality with negation and the raising requirement in (55).

With this background established, we are ready to approach the distribution of nominals and CPs in experiencer predicate TECs.

4.3.2 Nominals and CPs in Experiencer Predicate TECs

Recall that an experiencer verb with two [+A] arguments cannot participate in a TEC. As noted in section 1, neither the identity of the verbs nor their argument structure suffices to explain this restriction because it does not arise if (a) one of the arguments is removed by passivization, (b) one of the arguments is a CP, or (c) both of the arguments are [−A] NPIs (see (7), repeated here).

(7)

graphic

(58a) shows that if an experiencer verb in a TEC has two arguments, it is not acceptable for the subject alone to be [−A]. (58b) shows that, as expected under the account in section 4.3.1, it is not sufficient for just the object to be [−A] either. But if there is only one argument in the clause, or the second argument is a CP, either a [−A] or a [+A] subject is possible (see (58c–d)). (59) is a reminder that the ban on [+A] objects in TECs disappears when the experiencer verb is replaced with an agentive verb.

(58)

  • a.

    *A-ku-bon-anga m-ntu i-ntaka!

    NEG-17SA-see-NEG.PST1person 9-9bird

    [Intended: NOBODY saw a/the bird]

  • b.

    *A-ku-bon-anga i-ndoda nto.

    NEG-17SA-see-NEG.PST 9-9man 9thing

    [Intended: (It was) the man (who) didn’t see anything]

  • c.

    A-ku-bon-w-anga ntaka!

    NEG-17SA-see-PASS-NEG.PST9bird

    ‘NO BIRD was seen!’

  • d.

    A-ku-bon-anga m-ntu [CP ukuba u-m-ntwana u-ya-gul-a]!

    NEG-17SA-see-NEG.PST 1-person that 1-1-child 1SA-DISJ2-be.sick-FV

    ‘NOBODY saw that the child was sick!’

(59)

  • a.

    *A-ku-cing-anga m-ntu u-kutya!

    NEG-17SA-think-NEG.PST1-person 15-15food

    [Intended: NOBODY thought of (the) food]

  • b.

    A-ku-theng-anga m-ntu / u-Sabelo u-kutya!

    NEG-17SA-buy-NEG.PST 1-person / 1-1Sabelo 15-15food

    ‘NOBODY bought (the) food!’/‘SABELO didn’t buy (the) food!’

We summarize the distributional generalizations about experiencer ECs in table 3. The assumption that [−A] NPIs must satisfy the raising requirement in (55) plays a role in explaining h. of table 3 (see section 4.3.1 and the analysis to come in table 4). But it has nothing to say about key contrasts like that between c.–d. and e.–g. of table 3. The question these facts raise is, why should arguments of experiencer verbs in TECs exhibit any restrictions that arguments of agentive verbs in TECs do not share?

Table 3.

Licit and illicit arguments in experiencer ECs

StatusSubjectObject
a. ✓ [+A]  
b. ✓ [−A]  
c. ✓ [−A] CP 
d. ✓ [+A] CP 
e. ✓ [−A] [−A] 
f. x [+A] [+A] 
g. x [−A] [+A] 
h. x [+A] [−A] 
StatusSubjectObject
a. ✓ [+A]  
b. ✓ [−A]  
c. ✓ [−A] CP 
d. ✓ [+A] CP 
e. ✓ [−A] [−A] 
f. x [+A] [+A] 
g. x [−A] [+A] 
h. x [+A] [−A] 
Table 4.

Licit and illicit arguments of experiencer verbs in ECs, with explanations

StatusSubjectObjectExplanation
a. ✓ [+A]  [+A] SU obtains a structural Case value from C. 
b. ✓ [−A]  [−A] SU has no uCase needing a value; raising to Spec,Foc satisfies the raising requirement in (55). 
c. ✓ [−A] [−A] [−A] SU and OB have no uCase features. SU meets (55) by raising to Spec,Foc, and OB satisfies (55) parasitically. 
d. ✓ [−A] CP [−A] SU and CP OB have no uCase to require values. SU meets (55) by raising to Spec,Foc. 
e. ✓ [+A] CP [+A] SU gets Case from C. CP OB needs no uCase. 
f. x [+A] [+A] FocP strategy is ruled out by (63). [+A] SU can get Case in situ from C but [+A] OB’s uCase cannot be valued. 
g. x [−A] [+A] [−A] SU has no uCase and can meet (55) by raising to Spec,Foc, but [+A] OB either lacks a needed Case value or gets one from Foc that is illicit under (63).a 
h. x [+A] [−A] [+A] SU gets Case from C. [−A] OB cannot satisfy (55). 
a

This account might lead one to expect that the structural Case licenser that we identified in section 4.2.3 as C could Case-value a [+A] object across a [−A] subject. We assume that (defective) intervention prevents this.

 
StatusSubjectObjectExplanation
a. ✓ [+A]  [+A] SU obtains a structural Case value from C. 
b. ✓ [−A]  [−A] SU has no uCase needing a value; raising to Spec,Foc satisfies the raising requirement in (55). 
c. ✓ [−A] [−A] [−A] SU and OB have no uCase features. SU meets (55) by raising to Spec,Foc, and OB satisfies (55) parasitically. 
d. ✓ [−A] CP [−A] SU and CP OB have no uCase to require values. SU meets (55) by raising to Spec,Foc. 
e. ✓ [+A] CP [+A] SU gets Case from C. CP OB needs no uCase. 
f. x [+A] [+A] FocP strategy is ruled out by (63). [+A] SU can get Case in situ from C but [+A] OB’s uCase cannot be valued. 
g. x [−A] [+A] [−A] SU has no uCase and can meet (55) by raising to Spec,Foc, but [+A] OB either lacks a needed Case value or gets one from Foc that is illicit under (63).a 
h. x [+A] [−A] [+A] SU gets Case from C. [−A] OB cannot satisfy (55). 
a

This account might lead one to expect that the structural Case licenser that we identified in section 4.2.3 as C could Case-value a [+A] object across a [−A] subject. We assume that (defective) intervention prevents this.

 

Only Case theory seems to have the potential to address these curious patterns. Many languages with overt Case morphology mark the arguments of experiencer predicates with inherent Cases. Bhatt (2003:1) shows that in Marathi, the Case of experiencer subjects is dative, while in Bhojpuri it is genitive.

(60)

  • a.

    graphic

    she-DAT anger came

    ‘She got angry.’

  • b.

    graphic

    I-GEN.OBL this not find

    ‘I didn’t find it.’

It has also been demonstrated that in Ukrainian, experiencer predicates are barred from participation in a kind of TEC (see (61a) vs. (61b)). Lavine (2010) argues convincingly that the restriction underlying (61a) is Case-theoretic in origin.

(61)

  • a.

    graphic

    Ivan.ACC was surprised lightning.INSTR

    ‘Ivan was surprised by lightning.’

  • b.

    Kulju rozirvano evjaxom.

    balloon.ACC pierced nail.INSTR

    ‘The balloon was pierced by a nail.’

Following these precedents, we propose that arguments of experiencer predicates in Xhosa bear special inherent, hence semantic, Cases. Since experiencer subjects routinely raise to Spec,TP and value SA in SVO clauses (see (62)), we assume the inherent Cases are compatible with purely structural Cases.20 In (62), the [+A] DPs obtain nominative and accusative Case values from T and v*, respectively.

(62)

  • U-m-fazi u-bon-é i-ntaka.

  • 1-1-woman 1SA-see-CONJ1 9-9bird

  • ‘The/A woman saw the/a bird.’

Suppose further that all [+A] nominals have uCase features, because augments are Ds and every D has uCase. Then adding the augment to an inherently Cased argument of an experiencer verb entails that it will have to obtain an additional Case value. But the Case values that the Xhosa Focus head confers are linked to binary [+/−Focus] features as proposed in section 4.2.2. The [+Focus]/Case feature goes to the raised argument, and the [−Focus]/Case feature to the internal argument. Experiencer arguments, already marked with the special semantic Cases, cannot bear a second such feature (see (63)).

(63)

  • The Semantic Case Constraint

  • *DP bearing more than one semantically linked Case.

Under these assumptions, [+A] arguments of experiencer verbs must be local to a purely structural Case valuer: T and v* of SVO clauses and C of TECs. [−A] NPIs are immune to Case clashes because they have no uCase features that might obtain values incompatible with their inherent Cases. And Case problems do not arise for CP arguments, assuming with Stowell (1981) that they do not (necessarily) have uCase. Table 4 summarizes the proposals so far regarding arguments of experiencer verbs in ECs.

4.3.3 Independent Evidence That CPs and [−A] NPIs Do Not Require Case

The fact that Xhosa CPs can appear in the locations for which there is independent evidence of Case-theoretic difficulties is powerful language-internal evidence for Stowell’s (1981) proposal that CPs do not require Case. There is further evidence that Xhosa CPs need not obtain Case values: unlike nominals, they can appear vP-externally without accompanying object marking on the verb.21 Recall from section 3.2 that disjoint verbal morphology on the verb signals that it is final in vP; hence, anything that follows it lies outside. A reasonable assumption is that the obligatory object marker doubling a vP-external DP object in cases like (64) licenses it by linking it to a Case position.22

(64)

  • U-gqirha u-*(m)-bon-ile ]vP u-m-ntwana.

  • 1-doctor 1SA-*(1OM)-see-DISJ1 1-1-child

  • ‘The doctor saw the child.’

In contrast, CPs can follow a verb bearing either conjoint or disjoint morphology, without doubling by a clitic (see (65)). Their distribution thus suggests that CPs can but need not be in or linked with Case positions in Xhosa, which suggests that they can but need not bear uCase features.

(65)

graphic

We noted in section 1 that while intransitive ECs are very widespread and well-accepted in Xhosa, TECs are not universally accepted. Thus, a significant number of speakers find a sentence like (66a) quite degraded. But for these conservative speakers, TECs seem to be much improved if the internal argument is a CP as in (66b).

(66) Speakers conservative about TECs accept them if the object is a CP

graphic

The conservative dialect also supports our claim that [−A] NPIs do not require Case licensing. Speakers who reject TECs with [+A] arguments find them much improved if both arguments are [−A] NPIs (see the contrast between (67) and (68)).

(67)

  • Conservative dialect: TECs with [_A] arguments are rejected

  • *Ku-theng-a a-ba-fazi i-i-ntyatyambo.

  • 17SA-buy-FV 2-2-women 10-10-flowers

  • ‘It’s the women who buy flowers.’

(68) Conservative speakers judge TECs much improved with [−A] NPI arguments

graphic

Thus, the distinctions we have made regarding the Case needs of [+A], [−A], and CP arguments have independent support from facts unrelated to experiencer verbs.

4.3.4 Interim Summary and Discussion

This section has so far made the following proposals based on subject focus and the distribution of nominals and CPs in TECs:

(69) Implications of the distribution of arguments in Xhosa ECs

  • a.

    All [+A] nominals in Xhosa have uCase features, requiring valuation.

  • b.

    In contrast, [−A] nominals and CPs do not need Case licensing.

  • c.

    T and v* of TECs are defective, leading to a paucity of structural Cases in TECs.

  • d.

    There is one purely structural Case available to the highest EC argument from a particular (optional) choice of C.

  • e.

    A middle field Focus head can confer two Case values linked to [+/−Focus] features.

  • f.

    This does not work for arguments of experiencer verbs in TECs, which have inherent, thematically linked Cases and cannot bear an additional semantically linked Case.

  • g.

    Hence, experiencer verbs in TECs can only have CP objects or [−A] objects, the latter subject to an important proviso in h.

  • h.

    A general requirement that [−A] NPIs raise leftward precludes them from many positions including object in a Xhosa TEC, since v* of TECs has no edge feature (however, an NPI object can be parasitically licensed by a cooccurring NPI subject).

Two distinctive properties of Xhosa TECs—the Case/Focus linkage and the restrictions on DP arguments of experiencer verbs—are reminiscent of the Russian genitive of negation. When negation is added to a Russian clause, theme arguments that would otherwise be accusative or nominative may be realized as genitive, with some attendant semantic effects (the genitive has an NPI-like interpretation). Bailyn (1997) argues that negation licenses genitive under syntactic locality (see (70), taken from Bailyn 1997:85; see also Babby 1980, Pesetsky 1982, Neidle 1983, Freidin and Sprouse 1991 among others for discussion and a variety of approaches).

(70)

  • a.

    graphic

    Sasha.NOM buys books.ACC

    ‘Sasha is buying books.’

  • b.

    Saša ne pokupaet knig.

    Sasha.NOMNEG buys books.GEN.NEG

    ‘Sasha doesn’t buy (any) books.’

  • c.

    Saša ne pokupaet knigi.

    Sasha.NOMNEG buys books.ACC

    ‘Sasha doesn’t buy books.’

Like the Focus-linked Cases we have proposed for Xhosa, the Russian genitive of negation is a noncanonical Case correlating with the syntactic environment and the semantic interpretation of its bearer. Strikingly, inherently Cased expressions cannot bear the genitive of negation (see (71), taken from Pesetsky 1982:65).

(71)

  • a.

    graphic

    I NEG help no girls.DAT

    ‘I don’t help any girls.’

  • b.

    *Ja ne pomagaju nikakix devušek

    I NEG help no girls.F.GEN

    ‘I don’t help any girls.’

While details lie outside the scope of this article, we take the facts to be strong support for the Semantic Case Constraint in (63) and for the view that a noncanonical structural licenser may link a Case value with both a position and an interpretation.

The remainder of section 4 completes our analysis of TEC syntax by explaining an additional focus fact and the pronoun restriction, considering some alternative analyses, and addressing the absence of definiteness effects and expletive pro subjects in our account.

4.4 In-Situ Focus for a Single Postverbal Expression

The analysis developed above explains a ban on experiencer verbs in TECs: the Case values available from Focus are incompatible with the inherent Cases that arguments of experiencer verbs bear. This cannot be the whole story, however. Subjects of intransitive experiencer verbs can have focus readings in ECs (see (72)). And there is a second, subject focus reading available for a sentence like (7c) ( previously unmentioned for expository reasons) in which the object of the experiencer verb is a CP (see (73) and the contrast between (74b) and (74c)). If full DP experiencer subjects cannot raise to Spec,Foc, these readings must have some other origin.

(72) Experiencer subjects of intransitive verbs can have focus in ECs

  • a.

    Ku-qumb-a u-Sabelo.

    17SA-be.sad-FV 1-1Sabelo

    ‘Sabelo is sad.’ or ‘It’s Sabelo who is sad.’

  • b.

    Ku-zil-a u-Nomsa.

    17SA-mourn-FV 1-1Nomsa

    ‘Nomsa is mourning.’ or ‘It’s Nomsa who is mourning.’

(73)

graphic

(74)

  • a.

    U-Loyiso u-cing-a u-kutya.

    1-1Loyiso 1SA-think-FV 15-15food

    ‘Loyiso thinks of food.’

  • b.

    *Ku-cing-a u-Loyiso u-kutya.

    17SA-think-FV 1-1Loyiso 15-15food

    ‘Loyiso thinks of food.’ or ‘It’s Loyiso who thinks of food.’

  • c.

    Ku-cing-a u-Sabelo ukuba u-mhlaba u-ngqukuv-a.

    17SA-think-FV 1-1Sabelo that 3-3world 3SA-be.round-FV

    ‘Sabelo thinks that the world is round.’ or ‘It’s Sabelo who thinks that the world is round.’

Independently motivated proposals by Cheng and Downing (2012) for Zulu pave the way for an explanation of these facts. Cheng and Downing show that neutral word order in Zulu is [SVO . . . ] (see (75)), but nonsubject focused items appear in an immediately postverbal position (IAV = immediately after verb). For example, a wh-word like ngani ‘how’ and its answer must be IAV (see (76)). ((75) and (76) are from Cheng and Downing 2012:248; glosses adapted; our boldface on object markers.) Crucially, Cheng and Downing argue that this kind of focus is derived not by leftward movement of the IAV constituent but by vacating the vP of everything else and assigning a focus interpretation to what remains alone in a domain of phrasal prominence (see Buell 2007 for arguments along similar lines against leftward movement into a Spec,FocP as the mechanism for obtaining postverbal focus readings). Among their evidence is the fact that the direct object in a case like (76) must be doubled by a pronominal object marker on the verb. On the basis of this phenomenon and patterns of phonological phrasing (indicated by parentheses), Cheng and Downing argue that everything following the IAV expression is right-adjoined. The IAV item is in situ within an otherwise empty vP. While an in-depth investigation lies outside the scope of this article, Xhosa seems to exhibit quite similar phenomena, as shown in (77).

(75)

  • Neutral word order, in Zulu as in Xhosa: S-V-O-XP

  • graphic

  • 1PL.SA-carry 6-6-pumpkin with-1a.basket

  • ‘We are carrying the pumpkins in a basket.’

(76) Focused or questioned nonsubjects are immediately postverbal; all else is vP-external (internal arguments are obligatorily doubled by object markers on the verb)

graphic

(77)

graphic

While Cheng and Downing take a prosodic approach to deriving the Zulu facts, given the evidence for a FocP in Xhosa we suggest the following syntactic account. Suppose that, absent anything in its Spec, the Xhosa Focus head can confer a [+Focus] feature upon its complement, vP (on Focus as an assignable feature, see Horvath 1986 and Tuller 1992). Whatever material the vP contains is accordingly interpreted as focused—but for reasons that lie outside the scope of our article, this is always restricted to a single postverbal expression (see Cheng and Downing 2012 for relevant discussion and (78a), which for simplicity does not depict roll-up of Foc into the verbal complex). Locality might make this downward Focus feature assignment the only possibility in SVO clauses: if an XP moves to Spec,FocP, the subject perhaps cannot raise across it to Spec,TP (see (78b)).

(78)

  • a.

    The Focus feature assigned downward to vP gives a focus reading to its sole contents

  • b.

    Hypothesis: Subjects cannot raise across material in Spec,Foc, so focus in SVO clauses relies on the downward feature-assignment strategy

    graphic

Assuming with Cheng and Downing that a single item in an otherwise vacant vP can be focused, the postverbal subject in any intransitive EC can obtain a focus reading without moving to Spec, Foc, and hence without being directly assigned the complex Case-Focus feature that is incompatible with full DP experiencer subjects. As for CP objects of experiencer verbs, we have already argued that they can extrapose string-vacuously (see (65b) and (79)), and unlike right-adjoined DP arguments, CPs need not be doubled by an object marker (see, e.g., Stowell 1981, Richards and Van Urk 2015 on the tendency of finite CPs to extrapose).

(79) CP direct objects extrapose string-vacuously, leaving the experiencer in the vP that obtains the Focus feature by downward assignment

graphic

Summing up, we can neatly explain the ban on experiencer verbs in TECs in terms of a Case-Focus connection. Adopting for Xhosa a version of Cheng and Downing’s (2012) proposal of in-situ focus for a single expression in Zulu vPs accounts for the residue: subjects of intransitive verbs or verbs with one CP argument can acquire focus without Case.

4.5 The Pronominalization Asymmetry

As previously noted, pronominal objects are not possible in a Xhosa TEC, unlike in an SVO clause (see (22) and (23), repeated here). Only the highest DP in a TEC can be a pronoun.23

(22)

graphic

graphic

(23)

graphic

Many languages require pronouns to undergo object shift out of VP. Diesing and Jelinek (1995) and Diesing (1997) tie this to the unambiguous definiteness of pronouns. They argue from contrasts like the one in (80), from Diesing 1997:376, that there are interpretive differences associated with object shift in languages that allow two positions for objects: their base position and an object shift position outside VP. Diesing and Jelinek conclude from such interpretive contrasts that VP is the domain of existential closure, where definites do not belong (see also Heim’s (1982) Novelty Condition, requiring that VP-internal material be discourse-novel—something never true of pronouns). Then they present data from German, Icelandic, Arabic, and English demonstrating that even if full DP objects shift optionally, object pronouns must do so obligatorily (see (81)–(82)), as stated in (83).

(80)

graphic

(81)

graphic

(82)

  • a.

    Bert looked the reference up.

  • b.

    Bert looked up the reference.

  • c.

    Bert looked it up.

  • d.

    *Bert looked up it.

(83) Pronouns must vacate VP. (Diesing 1992, 1997, Diesing and Jelinek 1995)

Assuming that Spec,vP is the canonical object shift position (Chomsky 1995, 2001), our proposal that v* is defective in Xhosa ECs accounts for the ban on object pronouns in Xhosa TECs. Defective v* has no edge feature to shift pronouns.24

(84)

graphic

If object shift to Spec,vP is not available, Spec,Foc is the closest potential landing site for object pronouns to raise to. But in a TEC, Spec,Foc must be occupied by the external argument (which in any case intervenes to block closest c-command between Focus and the object). Hence, object pronouns are predicted to be illicit.25

4.6 The Subject-in-Situ Generalization, Distinctness, and Labeling

Alexiadou and Anagnostopoulou (2002, 2007) (henceforth A&A) argue that Case licensing is impossible when two DPs are left vP-internal.

(85)

  • The Subject-in-Situ Generalization (SSG) (A&A 2007:31)

  • By Spell-Out, vP can contain only one argument with a structural Case feature.

The Xhosa TEC facts described in this article parallel to some extent those that the SSG aims to capture. This section explores the possibility of a unified account.

A&A argue that the reason for SSG effects is that v adjoins to T before Case is checked, burying one of the potential Case-licensing features illicitly within a complex head. Quite a number of problems accompany this approach. As A&A (2007:50, (46)) acknowledge, under a cyclic, derivationalist view of syntax there should be no obstacle to probing by v* prior to its incorporation with T. Furthermore, Bruening (2013) presents evidence that in key English constructions the relevant subject is not actually in situ, and that in-situ subjects of transitives do not always yield ungrammaticality.26 Adding to the complexity of the picture, Baker and Collins (2006) show that a similar constraint in Kinande rules out the cooccurrence of two VP-internal nominal expressions even when one of them does not require Case licensing (hence, it applies even when one or both of them are [−A]). Thus, while there do seem to be recurring challenges across languages associated with multiple low ( postverbal) arguments, an overarching explanation has been an elusive goal.

Another approach to SSG phenomena attributes them to linearization issues. According to Richards (2010:5), two syntactic objects whose category labels are the same cannot be successfully linearized (see (86)). The proposal avoids Case-theoretic problems and, as it is not restricted to vP, it is not subject to Bruening’s (2013) criticisms.

(86)

  • Distinctness

  • If a linearization statement 〈α, α〉 is generated, the derivation crashes.

Chomsky (2013) also presents an alternative interpretation of SSG phenomena. He argues that syntactic objects obtain labels by an algorithm that applies at the phase level. If the labeling algorithm encounters a configuration of the form [XP YP], then labeling fails, and either XP or YP has to move. This drives raising of external arguments out of vP, where the external argument = XP and v′ = YP.

Let us suspend our proposal that v* is defective in a Xhosa TEC and see whether either the labeling or the distinctness hypothesis can provide a good alternative account of the facts. Suppose it is not Case valuation but one of these two processes that fails if the subject stays in situ in a Xhosa TEC. Raising the subject to Spec,FocP takes it outside the vP phase, so the grammar need not attempt to linearize it with the object. Similarly under the labeling approach, raising the subject to Spec,FocP eliminates the structural obstacle to correct labeling of vP.

Note, first, that while raising of subjects to Spec,Foc makes Xhosa TECs comply with the distinctness and labeling hypotheses, these approaches cannot eliminate the role that abstract Case plays in the account of Xhosa TECs. This is because the experiencer verb asymmetries still require reference to Case theory for a solution.

Note, second, that analyses along these lines have nothing to say about the inability of object pronouns to occur in TECs where the subject has raised to Spec,FocP. If v* is not defective and the subject raises to Spec,Foc, it is not clear why object pronouns cannot raise to Spec,vP, the canonical object shift position, as shown in (87) (Chomsky 1995, 2001).

(87)

  • If v* is robust and the subject raises to Spec,Foc, object shift should be possible

  • [FocP SU Foc [vP OB [vP 〈SU〉 [v′ vrobust [VP V 〈OB〉]]]]]

We might still address this under the hypothesis that the object shift position for Xhosa independent pronouns is in a location like XP of (88): lower than TP (hence postverbal in SVO clauses), but higher than FocP. Hence, raising the subject to Spec,FocP would not place it high enough to avoid intervention problems for pronoun shift.

(88)

graphic

This approach to Xhosa object shift would be language-particular and stipulated: the Mapping Principle approach underlying Diesing 1997 and Diesing and Jelinek 1995 only requires raising of pronouns out of VP, and a low, vP-level landing site for object shift has been well-motivated in a range of languages. While certain languages including varieties of Swedish optionally raise pronouns higher, this ‘‘long object shift’’ is able to cross low subjects (see (89), from Holmberg 1999:15), in contrast to the scenario sketched in (88).

(89)

  • graphic

  • therefore gave me Marit not any present

  • ‘Therefore Marit did not give me any present.’

Summing up, we can provide an alternative account by replacing the hypothesis of defective v* with three ingredients: (a) linearization/labeling problems; (b) the assumption that object pronoun shift cannot target Spec,vP in Xhosa (even when the subject raises to Spec,Foc, eliminating any potential intervention account of this); and (c) consequently, restriction of object pronoun shift to a nonstandard target above the low FocP. The alternatives do not eliminate the role of inherent Case or FocP in the account, since these are still needed to explain the ban on experiencer verbs in TECs.

We claim that, taken together, the components of the alternatives constitute more complex and hence less successful analyses than the one we have pursued here. Analyzing both T and v* of Xhosa TECs as defective, we capture the full constellation of TEC properties, motivating raising to Spec,Foc for the subject, predicting the impossibility of object pronouns, and tying these phenomena to the absence of agreement in TEC clauses.

We would also like to suggest that our approach to Xhosa might offer insight into SSG-type problems in other languages. Expletives in sentences with ‘‘weather’’ verbs are routinely taken to be a last resort, rescue mechanism when nothing thematic is available to occupy Spec,TP. Suppose that, in line with our analysis of Xhosa, this is true wherever expletives are used: they fill Spec,TP when T is unable to raise the subject, just as English do fills T when it is unable to raise a verb. Transitivity restrictions on ECs frequently arise, crosslinguistically, because in a dependency reminiscent of Chomsky’s (2007, 2008) Feature Inheritance proposal, clauses with defective T often have defective v* as well.

It lies outside the scope of this article to develop a fuller picture, making concrete the extension of these ideas to languages besides Xhosa. We leave this to future research.

4.7 Against an Expletive Pro Subject

We have provided evidence that v* of TECs is defective: it cannot enter an Agree relation to shift pronouns or value accusative Case. It is striking that this kind of v* is found in a clause with default SA and in-situ subjects. This seems to us unlikely to be a coincidence. We accordingly propose that T of Xhosa ECs is defective as well. The lack of agreement with the overt subject is treated as paralleling the absence of accusative Case and object pronominalization; thus, the properties of T and v* mirror each other in this construction.

A common approach to ECs in null subject languages has been to posit an expletive pro subject in Spec,TP. It is possible that Xhosa defective T retains the EPP feature, satisfied by a null expletive because T is unable to raise the subject. But there is no positive evidence for expletive pro in Xhosa. We tentatively conclude that Xhosa Spec,TP can be completely empty, as argued by Bobaljik and Wurmbrand (2005) for German and Cable (2012) for Luo.

4.8 On the Absence of Definiteness Effects

Our approach to Xhosa ECs is compatible with several approaches to the definiteness effect, including that it is Case-related and/or a function of an expletive . . . associate chain (see, e.g., Belletti 1988, Safir 1987, 2009). Since we are assuming that there is no expletive, there can be no such chain in Xhosa, and since subjects in Xhosa ECs have a noncanonical Case valuation strategy, Case-related approaches to the definiteness effect are generally likely to be compatible with the analysis.

Other researchers have argued that subjects of ECs must be indefinite because definites must vacate vP. As noted in section 4.5, Diesing (1992, 1997) and Diesing and Jelinek (1995) conclude from interpretive contrasts like that in (80) (repeated here) that VP is the domain of existential closure, where definites do not belong. The alternative approach to the definiteness effect simply analyzes this domain as a little bigger.

(80)

graphic

While we have argued that some subjects of intransitive ECs occupy a vP-internal position, it is possible, given the intricacies of the tense/aspect system, that the lowest subject position in Xhosa ECs might in fact be in a functional category a little higher than Spec,vP (and similarly for definite direct objects in TECs). Our data thus do not clearly determine which approach to the definiteness effect is preferable, for languages that have it. We leave resolution of this question to future research.

4.9 Interim Conclusions

This section has argued that the obligatory focus reading for subjects of TECs indicates that they must raise to Spec of a middle field FocP—a movement impossible for subjects of experiencer verbs because they bear an inherent Case. The facts support the conclusion that abstract Case is a force regulating DP positions in Xhosa syntax. A second, in-situ focus strategy based upon Cheng and Downing 2012 neutralizes the distinction between experiencers and other subjects of intransitive verbs and verbs with CP direct objects.

We have also argued that objects cannot be pronominalized in TECs because pronouns must raise out of VP.

Our analysis attributes both Case and pronoun-raising problems to the nature of v* in TECs, claiming that v* and T in TECs both lack the probe features involved in A-relations in SVO clauses.

Finally, we have argued that under our analysis, the absence of definiteness effects in Xhosa TECs is unsurprising.

5 Addressing Case Anomalies

5.1 The Case for No Case

Xhosa exhibits a set of Case-theoretic anomalies exemplified in (90)–(94). Comparable facts led Halpert (2012) to propose that full DPs are intrinsically Case-licensed in Zulu. Other researchers encountering similar phenomena recurring in Bantu languages have argued that Case is absent altogether in the family (Harford Perez 1985, Diercks 2012).27

(90) Licit in-situ subjects of passives

graphic

(91) Multiple SA in monoclausal constructions28

  • a.

    U-Sipho u-phantse w-a-tya nge-cephe.

    1-1Sipho 1SA-almost 1SA-PST-eat with-5spoon

    ‘Sipho almost ate with a spoon.’

  • b.

    Wena u-be u-soloko u-fund-a lapha.

    2SG.IP 2SG.SA-RFUT 2SG.SA-often 2SG.SA-study-FV here

    ‘You will often study here.’

(92)

  • Raising to object out of agreeing clauses

  • Ndi-funa u-Nomahlubi [okokuba a-phek-e a-ma-qanda].

  • 1SG.SA-want 1-1Nomahlubi that 1SA-cook-SUBJ 6-6-eggs

  • ‘I want Nomahlubi to cook eggs.’

  • [Lit.: I want Nomahlubi that she cook eggs]

(93) Subject raising from finite clauses preserving idiomatic readings and feeding passive

  • a.

    U-Hili u-bonakala [okokuba u-phum-ile e-ngcongolwe-ni].

    1-1Hili 1SA-seem that 1SA-exit-PSTLOC-10weeds-LOC

    ‘The secret seems to have come out.’

    [Lit.: The troll seems that exited the weeds]

  • b.

    U-Nomsa u-khol-w-a [okokuba u-phum-ile].

    1-1Nomsa 1SA-believe-PASS-FV that 1SA-depart-DISJ1

    ‘Nomsa is believed to have left.’

    [Lit.: Nomsa is believed that left]

(94) Postverbal subjects are licit when something else occupies Spec,TP and controls SA29

graphic

In previous sections of this article, we have provided numerous arguments that abstract Case is present in Xhosa and a key determinant of DP positions. If we are correct, Case is responsible for (a) the obligatory subject focus in TECs, (b) the ban on TECs with experiencer subjects, and (c) the unacceptability of TECs to the numerous speakers who accept only intransitive ECs. It follows that the phenomena in (90)–(94) must be given explanations consistent with this assessment.

5.2 Explaining the Anomalies

5.2.1 In-Situ Subjects of Passives

As Diercks (2012) points out, impersonal passive constructions are found in languages that clearly do have abstract Case, including German (see Bobaljik and Wurmbrand 2005 for related discussion). The hypothesis of downward-probing Agree provides the theoretical apparatus for explaining this possibility. Our analysis has followed that of Halpert (2012) in assuming that in-situ subjects of impersonal constructions can acquire Case values under closest c-command from a higher Case licenser (though we have applied this mechanism to the licensing of full DPs in Xhosa, unlike in Halpert’s analysis). This means that (90) is not problematic for the hypothesis that full DPs require Case.

5.2.2 Multiple SA

The account of multiple SA plays an explanatory role in several of the key Case anomalies, so we present it in some detail. Carstens (2010) observes that iterating SA (for which we adopt her term hyperagreement) is characteristic not just of Bantu but also of Semitic languages (see (95), from Bahloul and Harbert 1992:16). Since Semitic languages have clear morphological Case distinctions (see (96)), they show that an absence of abstract Case is not a necessary condition for hyperagreement. ((96) is adapted from Fassi Fehri 1993:219.)

(95) Hyperagreement in Arabic

graphic

(96) Case in Arabic

graphic

For Carstens (2010, 2011), the key factor in hyperagreement is the grammatical gender feature of nouns, made accessible to all clause-level probes in Semitic and Bantu by N-to-D raising and adjunction (see (97) and (98)). Word order evidence for Xhosa N-to-D raising is presented in (99). Absent N-to-D raising, Carstens argues that D’s person feature blocks access to nominal gender by clause-level probes (apart from one like a Romance participle, insensitive to person as a lexical property). (100) illustrates this intervention effect. (Num = Number, head of a functional category between DP and NP. Not shown is DP-internal Quantifier Raising of Num, which Carstens argues makes it broadly available for agreement and concord. (98) is from Fassi Fehri 1993:215.)

(97) Bantu N-to-D adjunction leads to SA in gender when T agrees with DP

graphic

(98) Arabic nouns adjoin to D

graphic

(99) Xhosa nouns adjoin to D

graphic

(100) Absent N-to-D raising, person intervenes to block probing of gender by T

graphic

In Carstens’s view, grammatical gender is a meaningless formal feature (uGen) and hence satisfies Chomsky’s (2000, 2001) ‘‘Activity Condition’’ (see (101)) as abstract Case does. But unlike uCase, nominal uGen comes from the lexicon with a value that is not affected by Agree relations. The reusability of gender as an activity feature is demonstrated in the very common phenomenon of DP-internal concord on multiple items (see (102)). Carstens accordingly advocates a view of ‘‘deactivation’’ effects in which valuation of structural uCase makes its bearer unable to enter further Agree relations because multiple values for uCase are impossible to pronounce (see also Nevins 2005).30 No such problem arises for the reuse of nominal gender to satisfy the Activity Condition; hence, iterating DP-internal concord is possible in Romance, Bantu, and Semitic but not English. Bantu and Semitic have SA that functions like concord because N-to-D raising gives DPs the uGen activity feature that does not obtain a value through Agree, unlike a DP’s uCase.31

(101)

  • The Activity Condition

  • Each participant in an Agree relation must have an unchecked uninterpretable feature (uF).

(102) Concord: Iterating Agree with N, based on reusability of uGen (not Case)

graphic

One might still expect that multiple aspectuals could agree in English, since only the highest would occupy T and hence value the subjects’ uCase. Carstens (2010, 2011) rules this out with a matching constraint on Agree that she calls the Strong Activity Condition (see (103)), which prevents a pure uφ probe from interacting with a DP whose sole activity feature is uCase. Hence, English T but not lower Asps can agree because T has a Case feature of its own (uNom) to match uCase of the subject DP (see (104)). Bantu aspectuals can agree because their uφ features are matched in kind by the uninterpretable nominal φ-feature uGen (see (105)). While Carstens (2010, 2011) assumes with Diercks (2012) that Case is absent in Bantu, we assume it is present but irrelevant to Agree relations with heads that have no Case values to confer.

(103)

  • The Strong Activity Condition

  • Probe and goal in a licit Agree relation have matching uFs, one of which can value the other.

(104) English: Only T can agree, since T’s uNom and SU’s uCase match up

  • a.

    *Jessie has is skating

  • b.

    graphic

(105) Bantu: Involvement of uGen on N enables any head to agree because nominal uGen is a variety of uφ; hence, probe and goal meet the matching requirement in (103)

graphic

5.2.3 Raising to Object

Raising to object in Xhosa takes place only out of subjunctives. In an exploration of Zulu raising, Zeller (2006) notes that subjunctives are more transparent than other clause types across languages and, in raising constructions, perhaps do not value Case on their subjects. Assuming this is true, raising to object is not a strong challenge to the claim that abstract Case is functional in Xhosa. The fact that the raised DP is agreed with in both clauses (see (92)) is no different from the other forms of hyperagreement discussed above.

5.2.4 Hyperraising

To a large extent, the derivation of hyperraising follows from the hyperagreement mechanics sketched in section 5.2.2: one DP is goal in serial Agree relations, yielding multiple instances of SA. We assume this does not violate the single Case constraint for the rather pedestrian reason that the DP involved is redundantly valued as nominative twice (note that this would not be the case for raising to object of the subject of a tensed clause).

More interesting is the question of how hyperraising gets around Chomsky’s (2000:108) Phase Impenetrability Condition, reproduced in (106).

(106)

  • The Phase Impenetrability Condition

  • In a phase α with head H, the domain of H is not accessible to operations outside α; only H and its edge are accessible to such operations.

Carstens and Diercks (2013b,c) show that several strategies underlie raising out of tensed clauses in Bantu languages. The most relevant for our purposes is clausal complementation with a nonphasal CP. Carstens and Diercks argue that given the articulated left periphery laid out in Rizzi 1997, 1999, it is not unexpected that some CPs would be nonphasal; Carstens (2012) argues that the CP-level phase head is Rizzi’s Int(errogative) (see (107)) and that clausal complements transparent to raising in Luyia are bare FinPs. This is likely the case where the Xhosa embedded okokuba ‘that’ clause is concerned in (93a–b).

(107)

graphic

5.2.5 Case for the Postverbal Subject of Inversion Constructions

It has often been noted that postverbal subjects in the unusual inversion constructions found in various Bantu languages have a focused interpretation (see, e.g., Ndayiragije 1999). This seems to be the case in the Xhosa example (94), repeated here.32

(94) Postverbal subjects are licit when something else occupies Spec,TP and controls SA

graphic

We have argued at length that there is a low Spec,FocP above vP and that it is a Case position. We have also provided evidence that there is one purely structural ‘‘downward’’ Case licenser available for the highest postverbal argument in a VS construction even when that argument is not agreed with or focused. While full details lie outside the scope of this article, we are confident that between them the two mechanisms can account for the licitness of postverbal subjects in inversion constructions such as these.

5.3 Summary and Analysis of Class 15

While Xhosa shares with other Bantu languages a set of properties that look rather anomalous from the standpoint of Case theory, TECs present strong evidence that Case is a factor in Xhosa grammar. This section has acknowledged the significance of these anomalies and suggested some ways of analyzing them consistent with Case theory. It is worth noting that the Xhosa Case anomalies seem to be just like those of Zulu reported in Halpert 2012, weakening the motivation for supposing that [+A] nominals in Zulu do not need Case-licensing.

A full understanding of Case in Bantu must explain why apparent infinitives (= class 15, ku + V forms) in many Bantu languages can have preverbal subjects—a fact that Harford Perez (1985) and Diercks (2012) point out is unexpected if abstract Case is present. The construction is absent in Xhosa, so we illustrate with Swahili data from Carstens 1991:184, 185.

(108)

graphic

(109)

graphic

Among the questions that arise is whether the preverbal subjects need Case licensing, and/or whether we should take the absence of this construction in Xhosa as evidence that the languages that allow it are significantly different in being languages without Case.

Some class 15 forms control agreement and have adjectival modifiers and genitive arguments (see Baker, Safir, and Sikuku 2012 for discussion). Carstens (1991) argues that the class includes derived nominals, two kinds of gerunds, and true infinitives (example from Carstens 1991:185).

(110)

graphic

Carstens (1991) proposes that only the gerunds allow overt subjects. Her evidence is based on Stowell’s (1982) observation that infinitives have an unrealized, future-like tense interpretation with respect to the tense of a matrix clause, whereas the tense construal of a gerund is determined by the semantics of the governing verb. This difference is clear in the contrast between (111a) and (111b), from Stowell 1982:563.

(111)

  • a.

    Jenny remembered [PRO to bring the wine].

  • b.

    Jenny remembered [PRO bringing the wine].

On the basis of such contrasts, Stowell argues that infinitives have tense features; gerunds do not.

Carstens reports that in Swahili, class 15 forms with null subjects are systematically ambiguous between a gerund-like interpretation and the unrealized tense interpretation characteristic of infinitives. ((112) is from Carstens 1991:183.)

(112)

graphic

But in (113), where the class 15 verb form has an overt subject, the only available interpretation is one in which singing took place in the past, as is consistent with the interpretation of a gerund. An unrealized interpretation for the class 15 form is not possible. Carstens argues that (114) is unacceptable for the same reason: the only reasonable interpretation would involve the unrealized/ future reading characteristic of infinitives. But with an overt subject, the class 15 constituent is a gerund. ((113) and (114) are from Carstens 1991:185.)

(113)

graphic

(114)

graphic

To account for the absence of overt subjects in Swahili’s true infinitives, Carstens argues that they are all CPs whose subjects cannot get exceptional Case marking from the matrix verb.

This proposal does not answer the question of how preverbal subjects in examples like (108), (109), and (111) obtain Case licensing, but it does reduce it to a familiar problem. The source of accusative Case for the subject of English so-called acc-ing gerunds (Him writing a book was surprising) has always been puzzling, especially since an overt subject alternates freely with PRO here. Whatever the explanation is for this curious property of gerunds, perhaps it extends to the preverbal subjects in Bantu class 15, and their absence in Xhosa is thus due to the failure of a particular strategy for ‘‘exceptional’’ Case.

6 Conclusions and Directions for Future Research

In this article, we have argued for the existence of a radically defective v* in Xhosa, which pairs with a defective T to yield clauses with no agreement, no subject or object raising, and Case valuation through exceptional strategies. Xhosa VSO clauses thus afford novel insights into the interconnected workings of agreement, A-movement, and Case.

The phenomena suggest strong linkages in Xhosa among the so-called probe features involved in A-relations as stated in (9), repeated here.

(9) Profile of two opposing Xhosa clause types

  • a.

    SVO: [TP T[+EPP/Agr/Case] . . . [vP v*[+EPP/+Agr/Case] . . . ]]

  • b.

    VSO: [TP T[−EPP/Agr/Case] . . . [vP v*[−EPP/Agr/Case] . . . ]]

    In Xhosa, defective v*defective T.

In fact, the evidence of constructions with auxiliaries suggests that in VSO clauses, all the usual probe features are absent. Recall from section 3.6 that every auxiliary or verbal element in a TEC bears default agreement, and subjects can only surface to their right (see (36), repeated here). This, we showed, contrasts with the Icelandic (34). (115) and (116) add demonstrations that no auxiliary or verbal element can agree or raise the subject when any other in the clause fails to do so.

(36)

graphic

(34)

graphic

(115) Agreeing and nonagreeing auxiliaries and verbs cannot mix in a clause

  • a.

    *Wena u-be ku-soloko u-fund-a lapha.

    2SG.IP 2SG.SA-RFUT 17SA-often 2SG.SA-study-FV here

    ‘You will often study here.’

  • b.

    *Ku-be u-soloko ku-fund-a wena lapha.

    17SA-RFUT 1SA-often 17SA-study-FV 2SG.IP here

  • c.

    *Ku-be ku-soloko u-fund-a wena lapha.

    17SA-RFUT 17SA-often 1SA-study-FV 2SG.IP here

(116) Combining agreement ‘‘mixing’’ with intermediate subject raising also does not work

  • a.

    *Ku-be wena u-soloko u-/ku-fund-a lapha.

  • b.

    *Ku-be ku-soloko wena u-fund-a lapha.

Where the subject has the focused reading, the impossibility of raising and agreement on higher heads may reduce to something like criterial freezing of the subject in Spec,FocP (see Rizzi and Shlonsky 2007). But when the subject focus reading is absent, assuming Case can be valued ‘‘downward’’ ( perhaps, as we have suggested, by a C), questions remain. And a basis for the dependencies among flavors of T and v is not obvious: why should defective T not cooccur with robust v*? This aspect of Xhosa clausal properties bears a resemblance to cooccurrence patterns involving English C and T. Because English finite, agreeing T is always found within CPs, Chomsky (2007, 2008) argues that robust T obtains its features from C through a process of Feature Inheritance. If we apply the same reasoning to the cooccurrence patterns exhibited by the two varieties of Xhosa T and v, we might be led to the somewhat surprising conclusion that v gets some or all of its features from T in Xhosa. Regrettably, it lies outside the scope of this article to do justice to this issue. We leave it to future research.

It is also outside the scope of this or any article to generalize about all of Bantu in a conclusive way. But our exploration of Xhosa ECs has convinced us that abstract Case can manifest itself quite differently across languages. The motivation for rejecting Case in Bantu and hence as a universal seems significantly weaker as we close this investigation.

Notes

Pilot questionnaire work for two Afranaph projects (www.africananaphora.rutgers.edu) formed the foundation for the research in this article. Our thanks to Ken Safir for Afranaph support, and to the University of Missouri South African Education Program, which funded our collaboration. This article owes a large intellectual debt to Claire Halpert’s (2012) work on transitive expletive constructions and argument licensing in Zulu, which inspired our investigation and influenced the analysis in crucial respects. Thanks to Sabelo Sawula, Sindiswa Silo, Noluthando Xolilizwe, and Stella Zondi for assistance with Xhosa data, and to the Xhosa Language Department at University of the Western Cape for its hospitality.

1 In glosses, SA = subject agreement; OM = object marker; Arabic numerals = noun classes (number + gender) unless followed by SG or PL, in which case they are person features. DISJ1-2 are tense morphemes on verbs that are final in their domains; CONJ1-2 are tense morphemes on verbs that are nonfinal in their domains. ACC = accusative; APPL = applicative; CAUS = causative; COP = copula; D = dual; DAT = dative; EXPL = expletive; F = feminine; FV = final vowel; GEN = genitive; INSTR = instrumental; IP = independent pronoun that can be conjoined and dislocated; LOC = locative; NEG = negative; NOM = nominative; OBL = oblique; PASS = passive; POSS = possessive; PRES = present; PST = past; Q = question; RFUT = remote future; RM = relative marker; SUBJ = subjunctive; wh.AGR = wh.agreement. Some but not all nouns have two separable class prefixes. We indicate class on prefixes and, if there is only one prefix, on the noun root. Negative polarity items and most in-situ wh-words in questions lack the outer prefix.

2 We use clefts to translate subject focus readings in affirmative clauses. This is for convenience and is based on rough functional equivalency, not analogous structural properties; the ECs we so translate are not in fact clefts. We indicate focus of negative polarity item subjects with uppercase.

3Mali (1995) also notes this restriction. Relatedly, reflexives (which seem to occupy the same morphological slot as object markers) are impossible in Xhosa TECs, though the reciprocal ending is licit. In contrast, Buell (2005) reports that reflexive markers are possible in Zulu ECs. We leave these mysteries to future research.

4 This proposal is based on Hiraiwa’s (2001) Multiple Agree but takes a serial approach to it, on which see Haegeman and Lohndal 2010. See also section 4.2.2 and references therein on Icelandic T first raising a dative and then agreeing with a nominative object, the relevance of which is pointed out in Halpert 2012.

5 If v* of a TEC induces phasal Transfer, this analysis is incompatible with the Phase Impenetrability Condition as formulated in Chomsky 2000 though perhaps not with the version in Chomsky 2001. Since nothing, not even an agent subject, can be extracted from a TEC, it is not obvious how to settle this question; we leave it aside.

6 See section 4.3.4 for a comparison with the Russian genitive of negation—a Case similar for its semantic linkage and its incompatibility with inherent Cases.

7 See section 4.6 for discussion of Alexiadou and Anagnostopoulou’s (2001) Subject-in-Situ Generalization and some related alternatives: distinctness for linearization purposes (Richards 2010) and the labeling algorithm of Chomsky 2013.

8Buell (2005) reports that in Zulu a TEC can answer a ‘What happened?’ question, and he gives one example. The contrast is potentially interesting, though two Zulu speakers we consulted did not share the judgment.

9 Thanks to an anonymous reviewer for raising this question. Other VS constructions also have subject focus readings; see the discussion of (94) in section 5.2.5, and see section 4.4 on focus in SVO clauses.

10 It is worth noting that independent pronoun objects are not systematically incompatible with expletive subject agreement and VS order; they are possible in an impersonal passive like (i). In contrast, pronominalizing any object in a TEC, transitive or ditransitive, is unacceptable. In terms of our article’s proposals, this difference suggests that the weak v of passives is different from the defective v* of TECs. Details lie outside the scope of the article.

(i)

  • Ku-nik-w-é u-Sindiswa zona.

  • 17SA-give-PASS-PST 1-1Sindiswa 10them

  • ‘Sindiswa was given them.’ (i.e., flowers)

11 A similar approach was independently proposed in Halpert 2012. Buell (2006) points out that the Zulu adverb kahle ‘well’ triggers the conjoint form, as do focused adverbs like phandle ‘outside’ (assume that they occupy Spec,Foc). Kahle doubles as an adjective and ka- is historically noun class morphology; phandle descends from locative class 16. Thus, they are plausibly viewed as having φ-features. See section 4.3.2 on CP complements.

12 Diagnostics in Cheng and Downing 2012 and Halpert 2012 suggest that the Zulu object marker is a clitic, and Xhosa patterns the same way. In contrast, since subject marking iterates in clauses with certain auxiliaries (see (36)), we analyze it as true agreement.

13 Following Slattery’s (1981) analysis of Zulu, we analyze these items as auxiliaries rather than adverbs. They appear between the subject and the inflected verb, where categorially unambiguous adverbs are illicit, and they bear SA as is common for clause-medial auxiliaries in Bantu languages.

14 As an anonymous reviewer points out, the possibility of a low FocP in Nguni and other Bantu languages has been much discussed. See section 4.4 for facts connected with postverbal focus in SV(O) clauses that lead Buell (2007) and Cheng and Downing (2012) to argue against a low FocP in Zulu (and see Hyman and Polinsky 2007 for arguments against FocP in Aghem). In contrast, Ndayiragije (1999), Van der Wal (2006), and Riedel (2009) argue for low FocPs in Kirundi, Makhuwa, and Sambaa, respectively. The facts of focus in Xhosa TECs provide novel evidence hitherto undiscussed for any language we know of in relation to this debate.

15 See section 4.6 for consideration of other possibilities including labeling problems (Chomsky 2013) and linearization problems (Richards 2010). The ban on TECs of experiencer verbs with two arguments is something these proposals seem to offer no insight into, as is the ban on pronominal subjects (see section 4.5). Under the Case approach, these are a package; under labeling or distinctness approaches, they are coincidental.

16 As an anonymous reviewer notes, this entails that the one structural Case available to the highest argument in ECs must be optional; on this, see section 4.2.3. Our proposal that Focus can value Case is intended to be languageparticular. We are not claiming that Focus can do this generally in other languages (though see section 5.2.5 for a conjecture that Focus might have a Case involvement in other Bantu languages where inverted subjects have obligatory focus readings). Even in the grammars of Xhosa speakers who do not accept TECs, we assume Focus cannot value Case. Thanks to an anonymous reviewer for raising these points. For reasons of length, we ignore the issue of Case in TECs that include applied and causative morphemes.

17 An anonymous reviewer asks, if Focus can value uCase, what the implications are for passive clauses. Xhosa allows both SV and VS impersonal passives (the latter illustrated in (i)), in which the internal argument surfaces low and may but need not have a focused reading.

(i)

  • Ku-theng-w-a i-i-ntyatyambo.

  • 17SA-buy-PASS-FV 10-10-flowers

  • ‘Flowers are bought./It’s flowers that are bought.’

The facts are suggestive of the same mechanisms that function in (T)ECs to value the highest argument’s Case: Focus or ‘‘downward’’ structural. Exploring how these work lies outside the scope of this article.

18 There is no φ-agreement reflex of this hypothetical Case valuation by a null C, but agreement and Case seem often not to correlate in Bantu languages. That is, in any treatment of Bantu that assumes there is abstract Case, it must be divorced to some extent from agreement, which can track inverted material (see Baker 2003, Carstens 2005, Henderson 2006).

19 See also Cable 2012 for arguments that in languages where subject raising to Spec,TP is optional, the best conclusion is that EPP is an optional feature of T. Another possibility is that the [+Focus] feature assigned to the subject in Spec,Foc is not linked to a Case value, unlike the Case/[−Focus] feature, though then the desired linkage could only be stipulated, as pointed out in section 4.2.2. Thanks to an anonymous reviewer for raising these questions.

20 An anonymous reviewer points out that Icelandic DPs bearing inherent dative cannot value SA. The compatibility of inherent and structural Cases would seem to be an area of crosslinguistic variation. See Jelinek 1984 for a proposal that certain structural and inherent Cases licitly combine in Warlpiri [clitic . . . DP] chains, and Halpert 2012 for arguments that agreement with inherent Cases is a logical possibility (though we differ with Halpert’s analysis of [+A] nominals as inherently Cased, we agree with the spirit of the proposal). And see Woolford 2006 for arguments that there are different kinds of inherent Case.

21 See Halpert 2012 for arguments that Zulu clauses headed by ukuthi ‘that’ differ in this respect from CPs headed by sengathi ‘as if ’, and for evidence that Zulu CPs can but need not be agreed with. The full set of distributional contrasts for different kinds of CPs and nominals is intricate and very interesting but doing justice to it would take us too far afield in this article.

22 In contrast, if the role of clitic doubling were construed as necessary for linkage to a θ-position, why CPs do not require it would be mysterious (as will be illustrated in (65)).

23 An anonymous reviewer asks how pronominalization works in double object constructions. Neither object can pronominalize in a TEC. In an SVO clause, either object can be a clitic pronoun, and both objects can be independent pronouns simultaneously. Double object constructions are ‘‘symmetrical’’ in Xhosa in that either the direct object or the indirect object can passivize or pronominalize. We assume with McGinnis (2001) that the Appl head raises the direct object out of VP to the level of the indirect object (= outer Spec,ApplP), satisfying the pronoun-raising requirement. Further details would take us too far afield, but see Carstens and Mletshe 2013.

24 An anonymous reviewer suggests that only clitics and nonconjoined, unstressed weak pronouns should have to raise, and that clitics do so for morphological reasons. This is at odds with well-established semantic accounts of the pressure for pronoun raising discussed above (and see Diesing and Jelinek 1995 for a proposal that object clitics are a grammaticized means of meeting the semantic pressure). Conjoining or stressing Xhosa independent object pronouns does not make them felicitous in TECs, though in some languages these factors eliminate pronoun shift. But there is a precedent in Spanish, where Suñer (2000) shows that even independent conjoined or stressed pronouns are subject to the raising requirement. Suñer argues that, on a language-particular basis, LF raising may suffice if overt pronoun shift is not available, and that in some languages conjunction or stress makes pronouns too ‘‘heavy.’’ But by assumption, all pronouns and definite DPs raise out of VP by LF. In Spanish, systematic clitic doubling of all object pronouns allows overt satisfaction of the requirement in a uniform way. Further details lie outside the scope of this article.

25Richards (2007) claims that object shift is never possible in TECs because expletives merge in the object shift position, making it unavailable. Höskuldur Thráinsson ( pers. comm.) reports that object shift is possible in Icelandic TECs, including for weak pronouns. Pronoun objects are also licit in Xhosa impersonal passives of ditransitives, though details lie outside the scope of this article; we leave investigation for future research.

26Bruening (2013) argues that the crucial factor in determining whether low subjects are grammatical in English is whether the subject precedes the inflected verb. His proposals cannot be extended to Xhosa as they are not compatible with the Xhosa solution of raising one DP to (postverbal) Spec,Foc. Discussing this in the framework of Bruening’s assumed top-to-bottom structure-building approach would lead us far afield, so we leave it aside.

27Van der Wal (2012) takes a more nuanced view on the Case issue, pointing out that in some Bantu languages SA tracks the logical subject, and proposing that such Bantu languages have Case, in contrast to the ones that Diercks discusses. Our claim is that even languages that exhibit ‘‘no Case’’ profiles may nonetheless have Case.

28 For arguments that constructions like (91a–b) are truly monoclausal, see references cited in section 3.6.

29 This construction was first documented in Zulu by Zeller (2011), who names it instrument inversion.

30 Under the conclusions reached in section 4.3.2, an inherent Case and a structural Case are compatible. This is consistent with Carstens’s approach because inherent Cases, like the gender features of nouns, are not altered by participation in Agree relations.

31 Because Romance T cannot obtain a gender value by agreeing with D, Carstens (2010) argues that there is no ‘‘Agree with agreement’’; see also Carstens and Diercks 2013a for a further argument from Luyia clefting that only intrinsic φ can value uφ.

32 Percy Buthelezi ( pers. comm.) reports that this construction in Zulu would be suited to answer ‘Where is the spoon?’ This gives a kind of topic status to the instrument; as in Xhosa, any focus would have to be the postverbal DP.

References

Alexiadou,
Artemis
, and
Elena
Anagnostopoulou
.
2001
.
The subject-in-situ generalization and the role of Case in driving computations
.
Linguistic Inquiry
32
:
193
231
.
Alexiadou,
Artemis
, and
Elena
Anagnostopoulou
.
2002
.
Raising without infinitives and the nature of agreement
. In
Dimensions of movement: From features to remnants
, ed. by
Artemis
Alexiadou
,
Elena
Anagnostopoulou
,
Sjef
Barbiers
, and
Hans-Martin
Gärtner
,
17
30
.
Amsterdam
:
John Benjamins
.
Alexiadou,
Artemis
, and
Elena
Anagnostopoulou
.
2007
.
The subject-in-situ generalization revisited
. In
Interfaces + recursion = language?
, ed. by
Uli
Sauerland
and
Hans-Martin
Gärtner
,
31
60
.
Berlin
:
Mouton de Gruyter
.
Babby,
Leonard
.
1980
.
Existential sentences and negation in Russian
.
Ann Arbor, MI
:
Karoma Publishers
.
Bahloul,
Maher
, and
Wayne
Harbert
.
1992
.
Agreement asymmetries in Arabic
. In
Proceedings of the Eleventh West Coast Conference on Formal Linguistics
, ed. by
Jonathan
Mead
,
15
32
.
Stanford, CA
:
CSLI Publications
.
Bailyn,
John F
.
1997
.
Genitive of negation is obligatory
. In
Formal Approaches to Slavic Linguistics 4: The Cornell Meeting
, ed. by
Wayles
Browne
,
Ewa
Dornsich
,
Natasha
Kondrashova
, and
Draga
Zec
,
84
114
.
Ann Arbor
:
Michigan Slavic Publications
.
Baker,
Mark
.
2003
.
Agreement, dislocation, and partial configurationality
. In
Formal approaches to function in grammar
, ed. by
Andrew
Carnie
,
Heidi
Harley
,
Eloise
Jelinek
, and
MaryAnn
Willie
,
107
132
.
Amsterdam
:
John Benjamins
.
Baker,
Mark
, and
Chris
Collins
.
2006
.
Linkers and the internal structure of vP
.
Natural Language and Linguistic Theory
24
:
307
354
.
Baker,
Mark
,
Ken
Safir
, and
Justine
Sikuku
.
2012
.
Categories of clausal constituents in Lubukusu and limits to their selection
.
Ms., Rutgers University, New Brunswick, NJ
.
Belletti,
Adriana
.
1988
.
The Case of unaccusatives
.
Linguistic Inquiry
19
:
1
34
.
Bhatt,
Rajesh
.
2003
.
Experiencer subjects
.
Unpublished course materials. Available at http://web.mit.edu/rbhatt/www/24.956/dative.pdf
.
Bobaljik,
Jonathan David
, and
Philip
Branigan
.
2006
.
Eccentric agreement and multiple Case-checking
. In
Ergativity: Emerging issues
, ed. by
Alana
Johns
,
Diane
Massam
, and
Juvénal
Ndayiragije
,
47
77
.
Dordrecht
:
Springer
.
Bobaljik,
Jonathan David
, and
Dianne
Jonas
.
1996
.
Subject positions and the roles of TP
.
Linguistic Inquiry
27
:
195
236
.
Bobaljik,
Jonathan David
, and
Susi
Wurmbrand
.
2005
.
The domain of agreement
.
Natural Language and Linguistic Theory
23
:
809
865
.
Bošković,
Željko
.
2011
.
Last Resort with Move and Agree in derivations and representations
. In
The Oxford handbook of linguistic Minimalism
, ed. by
Cedric
Boeckx
,
327
353
.
Oxford
:
Oxford University Press
.
Bruening,
Benjamin
.
2013
.
Restrictions on inversion in English: Not the subject-in-situ generalization, but linear order
.
Ms., University of Delaware, Newark
.
Buell,
Leston
.
2005
.
Issues in Zulu morphosyntax
.
Doctoral dissertation, UCLA, Los Angeles, CA
.
Buell,
Leston
.
2006
.
The Zulu conjoint/disjoint verb alternation: Focus or constituency?
ZAS Papers in Linguistics
43
:
9
30
.
Buell,
Leston
.
2007
.
Evaluating the immediate postverbal position as a focus position in Zulu
. In
Selected proceedings from the 38th Annual Conference on African Linguistics
, ed. by
Masangu
Matondo
,
Fiona
McLaughlin
, and
Eric
Potsdam
,
166
172
.
Somerville, MA
:
Cascadilla Press
.
Cable,
Seth
.
2012
.
The optionality of movement and EPP in Dholuo
.
Natural Language and Linguistic Theory
30
:
651
697
.
Carstens,
Vicki
.
1991
.
The morphology and syntax of DPs in Kiswahili
.
Doctoral dissertation, UCLA, Los Angeles, CA
.
Carstens,
Vicki
.
2001
.
Multiple agreement and Case deletion: Against phi-(in)completeness
.
Syntax
4
:
147
163
.
Carstens,
Vicki
.
2005
.
Agree and EPP in Bantu
.
Natural Language and Linguistic Theory
23
:
219
279
.
Carstens,
Vicki
.
2010
.
Implications of grammatical gender for the theory of uninterpretable features
. In
Exploring crash-proof grammars
, ed. by
Michael T.
Putnam
,
31
57
.
Amsterdam
:
John Benjamins
.
Carstens,
Vicki
.
2011
.
Hyperactivity and hyperagreement in Bantu
.
Lingua
121
:
721
741
.
Carstens,
Vicki
.
2012
.
Delayed valuation
.
Available at ling.auf.net/lingbuzz/001432. To appear in Syntax
.
Carstens,
Vicki
, and
Michael
Diercks
.
2013a
.
Agreeing how?
Linguistic Inquiry
44
:
179
237
.
Carstens,
Vicki
, and
Michael
Diercks
.
2013b
.
The great escape: Raising out of finite clauses in Bantu languages
.
Paper presented at the 31st West Coast Conference on Formal Linguistics
.
Carstens,
Vicki
, and
Michael
Diercks
.
2013c
.
Parameterizing Case and activity: Hyper-raising in Bantu
. In
NELS 40
, ed. by
Seda
Kan
,
Claire
Moore-Cantwell
, and
Robert
Staubs
,
99
118
.
Amherst
:
University of Massachusetts, Graduate Linguistic Student Association
.
Carstens,
Vicki
, and
Kasangati K. W.
Kinyalolo
.
1989
.
On IP structure: Tense, aspect, and agreement
.
Ms., UCLA, Los Angeles, CA
.
Carstens,
Vicki
, and
Loyiso
Mletshe
.
2013
.
N-words in disguise: A negative concord analysis of augmentless NPIs in Xhosa and Zulu
.
Available at ling.auf.net/lingbuzz/001938
.
Cheng,
Lisa Lai-Shen
, and
Laura
Downing
.
2012
.
Against FocusP: Arguments from Zulu
. In
Contrasts and positions in information structure
, ed. by
Ivona
Kučerová
and
Ad
Neeleman
,
247
266
.
Cambridge
:
Cambridge University Press
.
Chomsky,
Noam
.
1995
.
The Minimalist Program
.
Cambridge, MA
:
MIT Press
.
Chomsky,
Noam
.
2000
.
Minimalist inquiries: The framework
. In
Step by step: Essays in linguistics in honor of Howard Lasnik
, ed. by
Roger
Martin
,
David
Michaels
, and
Juan
Uriagereka
,
89
155
.
Cambridge, MA
:
MIT Press
.
Chomsky,
Noam
.
2001
.
Derivation by phase
. In
Ken Hale: A life in language
, ed. by
Michael
Kenstowicz
,
1
52
.
Cambridge, MA
:
MIT Press
.
Chomsky,
Noam
.
2007
.
Approaching UG from below
. In
Interfaces + recursion = language?
, ed. by
Uli
Sauerland
and
Hans-Martin
Gärtner
,
1
29
.
Berlin
:
Mouton de Gruyter
.
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
.
Chomsky,
Noam
.
2013
.
Problems of projection
.
Lingua
130
:
33
49
.
Diercks,
Michael
.
2012
.
Parameterizing Case: Evidence from Bantu
.
Syntax
15
:
253
286
.
Diesing,
Molly
.
1992
.
Indefinites
.
Cambridge, MA
:
MIT Press
.
Diesing,
Molly
.
1997
.
Yiddish VP order and the typology of movement in Germanic
.
Natural Language and Linguistic Theory
15
:
369
427
.
Diesing,
Molly
, and
Eloise
Jelinek
.
1995
.
Distributing arguments
.
Natural Language Semantics
3
:
123
176
.
Fassi Fehri,
Abdelkader
.
1993
.
Issues in the structure of Arabic clauses and words
.
Dordrecht
:
Kluwer
.
Freidin,
Robert
, and
Rex A.
Sprouse
.
1991
.
Lexical Case phenomena
. In
Principles and parameters in comparative grammar
, ed. by
Robert
Freidin
,
392
416
.
Cambridge, MA
:
MIT Press
.
Halpert,
Claire
.
2012
.
Argument licensing and agreement in Zulu
.
Doctoral dissertation, MIT, Cambridge, MA
.
Haegeman,
Liliane
.
1995
.
The syntax of negation
.
Cambridge
:
Cambridge University Press
.
Haegeman,
Liliane
, and
Terje
Lohndal
.
2010
.
Negative concord and (multiple) Agree: A case study of West Flemish
.
Linguistic Inquiry
41
:
181
211
.
Harford Perez,
Carolyn
.
1985
.
Aspects of complementation in three Bantu languages
.
Doctoral dissertation, University of Wisconsin, Madison
.
Heim,
Irene
.
1982
.
The semantics of definite and indefinite noun phrases
.
Doctoral dissertation, University of Massachusetts, Amherst
.
Henderson,
Brent
.
2006
.
Multiple agreement and inversion in Bantu
.
Syntax
9
:
275
287
.
Hiraiwa,
Ken
.
2001
.
Multiple Agree and the Defective Intervention Constraint in Japanese
. In
Proceedings of the 1st HUMIT Student Conference in Language Research
, ed. by
Ora
Matushansky et al.
,
67
80
.
MIT Working Papers in Linguistics 40. Cambridge, MA
:
MIT, MIT Working Papers in Linguistics
.
Holmberg,
Anders
.
1999
.
Remarks on Holmberg’s Generalization
.
Studia Linguistica
53
:
1
39
.
Holmberg,
Anders
, and
Thorbjörg
Hróarsdóttir
.
2003
.
Agreement and movement in Icelandic raising constructions
.
Lingua
113
:
997
1019
.
Horvath,
Julia
.
1986
.
FOCUS in the theory of grammar and the syntax of Hungarian
.
Dordrecht
:
Foris
.
Hyman,
Larry M.
, and
Maria
Polinsky
.
2007
.
Focus in Aghem
.
Ms., University of California, Berkeley, and Harvard University, Cambridge, MA
.
Hyman,
Larry M.
, and
J. R.
Watters
.
1984
.
Auxiliary focus
.
Studies in African Linguistics
15
:
233
273
.
Jelinek,
Eloise
.
1984
.
Empty categories, Case, and configurationality
.
Natural Language and Linguistic Theory
2
:
39
76
.
Kayne,
Richard
.
1981
.
Two notes on the NIC
. In
Theory of markedness in generative grammar
, ed. by
Adriana
Belletti
,
Luciana
Brandi
, and
Luigi
Rizzi
,
317
346
.
Pisa
:
Scuola Normale Superiore
.
Kinyalolo,
Kasangati K. W
.
1991
.
Syntactic dependencies and the Spec-head agreement hypothesis in Kilega
.
Doctoral dissertation, UCLA, Los Angeles, CA
.
Kinyalolo,
Kasangati K. W
.
2003
.
Limiting the scope of V-movement in Kilega
.
Handout of a talk given at Stony Brook, New York
.
Lavine,
James
.
2010
.
Case and events in Ukrainian experiencer predicates
. In
Formal Approaches to Slavic Linguistics 18
, ed. by
Wayles
Browne
,
Adam
Cooper
,
Alison
Fisher
,
Esra
Kesici
,
Nikola
Predolac
, and
Draga
Zec
,
285
300
.
Ann Arbor
:
Michigan Slavic Publications
.
Mali,
Zoliswa Olga
.
1995
.
The existential sentences in Xhosa
.
Master’s thesis, University of Stellenbosch
.
McGinnis,
Martha
.
2001
.
Variation in the phase structure of applicatives
.
Linguistic Variation Yearbook
1
:
105
146
.
Melebari,
Aláa
, and
T. Daniel
Seely
.
2012
.
Agreement without ‘exception’ in Standard Arabic?
Paper presented at the Illinois Symposium on Semitic Linguistics, University of Illinois, Urbana-Champaign
.
Milsark,
Gary
.
1977
.
Towards an explanation of certain peculiarities of the existential construction in English
.
Linguistic Analysis
3
:
1
29
.
Mletshe,
Loyiso
.
1995
.
The subject in Xhosa
.
Master’s thesis, University of Stellenbosch
.
Ndayiragije,
Juvénal
.
1999
.
Checking economy
.
Linguistic Inquiry
30
:
399
444
.
Neidle,
Carol
.
1983
.
The role of Case in Russian syntax
.
Doctoral dissertation, MIT, Cambridge, MA
.
Nevins,
Andrew
.
2005
.
Derivations without the Activity Condition
. In
Perspectives on phases
, ed. by
Martha
McGinnis
and
Norvin
Richards
,
283
306
.
MIT Working Papers in Linguistics 49
.
Cambridge, MA
:
MIT, MIT Working Papers in Linguistics
.
Ngonyani,
Deo
.
1999
.
X0-movement in Kiswahili relative clauses
.
Linguistic Analysis
29
(
1–2
):
137
159
.
Ngonyani,
Deo
.
2006
.
Attract F and verbal morphology in Kiswahili
.
The Linguistic Review
23
:
35
66
.
Pesetsky,
David
.
1982
.
Paths and categories
.
Doctoral dissertation, MIT, Cambridge, MA
.
Preminger,
Omer
.
2011
.
Agreement as a fallible operation
.
Doctoral dissertation, MIT, Cambridge, MA
.
Richards,
Marc
.
2007
.
Object shift, phases, and transitive expletive constructions in Germanic
.
Linguistic Variation Yearbook
6
:
139
159
.
Richards,
Norvin
.
2010
.
Uttering trees
.
Cambridge, MA
:
MIT Press
.
Richards,
Norvin
, and
Coppe
van Urk
.
2015
.
Two components of long-distance extraction: Successive cyclicity in Dinka
.
Linguistic Inquiry
46
:
113
155
.
Riedel,
Kristina
.
2009
.
The syntax of object marking in Sambaa
.
Doctoral dissertation, Leiden University
.
Rizzi,
Luigi
.
1997
.
The fine structure of the left periphery
. In
Elements of grammar
, ed. by
Liliane
Haegeman
,
281
337
.
Dordrecht
:
Kluwer
.
Rizzi,
Luigi
.
1999
.
On the position ‘‘Int(errogative)’’ in the left periphery of the clause
.
Ms., University of Siena
.
Rizzi,
Luigi
, and
Ur
Shlonsky
.
2007
.
Strategies for subject extraction
. In
Interfaces + recursion = language?
, ed. by
Uli
Sauerland
and
Hans-Martin
Gärtner
,
115
160
.
Berlin
:
Mouton de Gruyter
.
Safir,
Kenneth
.
1987
.
What explains the definiteness effect?
In
The representation of (in)definiteness
, ed. by
Eric J.
Reuland
and
Alice G. B.
ter Meulen
,
71
97
.
Cambridge, MA
:
MIT Press
.
Safir,
Kenneth
.
2009
.
Syntactic chains
.
Cambridge
:
Cambridge University Press
.
Slattery,
H
.
1981
.
Auxiliary verbs in Zulu
.
Grahamstown, South Africa
:
Rhodes University, Department of African Languages
.
Stowell,
Tim
.
1981
.
Origins of phrase structure
.
Doctoral dissertation, MIT, Cambridge, MA
.
Stowell,
Tim
.
1982
.
The tense of infinitives
.
Linguistic Inquiry
13
:
561
570
.
Suñer,
Margarita
.
2000
.
Object-shift: Comparing a Romance language to Germanic
.
Probus
12
:
261
289
.
Tuller,
Laurice
.
1992
.
The syntax of postverbal focus constructions in Chaddic
.
Natural Language and Linguistic Theory
10
:
303
334
.
Van der Spuy,
Andrew
.
1993
.
Dislocated noun phrases in Nguni
.
Lingua
90
:
335
355
.
Van der Wal,
Guenever Johanna
.
2006
.
An empty Immediate After Verb position in Makhuwa
.
ZAS Papers in Linguistics
43
:
233
256
.
Van der Wal,
Guenever Johanna
.
2012
.
Parameterizing Case: Other evidence from Bantu
.
Paper presented at the meeting of the Linguistic Society of Great Britain, University of Salford, Manchester
.
Vangsnes,
Øystein Alexander
.
2002
.
Icelandic expletive constructions and the distribution of subject types
. In
Subjects, expletives, and the Extended Projection Principle
, ed. by
Peter
Svenonius
,
43
70
.
Oxford
:
Oxford University Press
.
Woolford,
Ellen
.
2006
.
Lexical Case, inherent Case, and argument structure
.
Linguistic Inquiry
37
:
111
130
.
Zeller,
Jochen
.
2006
.
Raising out of finite CP in Nguni: The case of fanele
.
Southern African Linguistics and Applied Language Studies
24
:
255
275
.
Zeller,
Jochen
.
2011
.
Instrument inversion in Zulu
. In
African languages in context (selected proceedings of the 42nd Annual Conference on African Linguistics)
, ed. by
Michael R.
Marlo
,
Nikki B.
Adams
,
Christopher R.
Green
,
Michelle
Morrison
, and
Tristan M.
Purvis
,
134
148
.
Somerville, MA
:
Cascadilla Proceedings Project
.