Abstract

In Lubukusu and Lusaamia, the wh-expression ‘how’ agrees in φ-features with the subject of its clause. We show that agreement on ‘how’ is not always identical to subject agreement on the verb: the two diverge in certain locative inversion and subject extraction environments. On the basis of these facts, we argue that ‘how’ is a vP adjunct with downward-probing uφ independent of the uφ that underlies subject agreement. We also explore locality paradoxes that arise in connection with agreeing ‘how’ in locative inversion constructions. These present challenges to the traditional notion of equidistance from a probe as an explanation for inversion, show that operators may have ‘‘active’’ φ-features even while they are Ā-opaque, and offer insight into the mechanisms making locative inversion possible.

1 Introduction

1.1 The Empirical Issue

In some varieties of Luyia (narrow Bantu, J.30), the wh-word ‘how’ agrees in person, number, and noun class with the subject of its clause. We illustrate with data from Lusaamia (Uganda/Western Kenya) and Lubukusu (Western Kenya).1

(1)

  • a.

    graphic

    1SG.SA-sing 1SG-how

    ‘How do I sing?’

  • b.

    W-emba o-tie?

    2SG.SA-sing 2SG-how

    ‘How do you sing?’

  • c.

    Y-emba a-tie?

    3SG.SA-sing 3SG-how

    ‘How does she/he sing?’

  • d.

    Khw-emba khu-tie?

    1PL.SA-sing 1PL-how

    ‘How do we sing?’

  • e.

    graphic

    ‘How did the trees fall?’

  • f.

    Si-tanda si-funikhe si-rie(na)?

    7-bed 7SA-broke 7-how

    ‘How did the bed break?’

In this article, we describe and analyze the distribution and properties of Luyia agreeing ‘how’, relying mainly on data from Lubukusu. We show that agreement on ‘how’ is usually but not always identical to subject agreement (SA), diverging crucially in the features it exhibits in certain locative inversion and subject extraction contexts. The contrasts argue that ‘how’ has an agreement relationship with the subject independent of the relation that yields agreement on the verb. We analyze agreeing ‘how’ as an XP adjunct to vP and demonstrate that only a downward-probing Agree analysis (Chomsky 2000, 2001) captures all of the ‘how’ agreement facts. This runs counter to proposals by Baker (2008) and Diercks (2011b) that agreement in most Bantu languages involves systematic upward probing as a matter of crosslinguistic parametric choice. The facts of agreeing ‘how’ also argue against the view that agreement always spells out a specifier-head (Spec-head) relation as proposed by Koopman (2000, 2006) and others. They are incompatible with Chomsky’s (2007, 2008) claim that all probe features are introduced on phase heads and surface on phase head complements, and they run counter to the common assumption that only heads can probe. Building on proposals of Epstein and Seely (2006), Bošković (2007, 2011), and Carstens (2012), we argue that unvalued uF of X can probe XP’s c-command domain.

Traditionally, agreement has often been viewed as a reflex of the relationship between a predicate and its arguments, or a head and its modifiers. In Minimalism, the parties participating in an agreement relationship are instead determined largely through syntactic architecture and locality. Agreeing ‘how’ strongly supports the latter approach.

1.2 Structure of the Article

This article consists of eight major sections. Section 1.3 summarizes our theoretical assumptions. Section 2 describes the properties and distribution of agreeing ‘how’, lays out several potential approaches, and sketches the analysis the article will argue for. Section 3 addresses two of the potential alternative accounts and shows how they fail. Section 4 presents some crucial diagnostic evidence relating to noncanonical subjects: inverted locatives and subject operators. Largely on the basis of these facts, section 5 develops the analysis of agreeing ‘how’ as a downward-probing vP adjunct, showing that ‘how’ must have independent uφ that agrees with the subject in its base position. Section 5 also argues against analyzing the agreement of ‘how’ as arising through control of a pro subject within an interrogative predicate phrase headed by ‘how’. Section 6 lays out the reasons why upward-probing, Spec-head agreement and the feature inheritance theory of Chomsky (2007, 2008) and Richards (2007) fail to explain the properties of ‘how’, and it sketches a theory of agreement compatible with the facts. Section 6 also argues that agreeing ‘how’ is one among many cases in grammar where unvalued uF of X can probe XP’s c-command domain. Section 7 explores some apparent locality paradoxes connected with the analysis of ‘how’, including socalled Ā-opacity, based on the ability of the locative to move across the thematic subject and vice versa. Section 8 summarizes and concludes.

1.3 Theoretical Assumptions

The article is written within the Minimalist theoretical framework of Chomsky (2000, 2001). In particular, we assume that syntactic objects are constructed from bottom to top by the Merge operation, and that Transfer to the conceptual-intentional and sensorimotor interfaces takes place cyclically. We follow Chomsky in taking Transfer to occur when a phase head v* or C is merged. We also assume with Chomsky that agreement and Case are uninterpretable, unvalued features (uFs): uφ and uCase, respectively. When uφ is merged on some category α, it immediately probes its c-command domain to find a goal β—an expression that can provide values for α’s unvalued features. We assume a version of the ‘‘activity requirement’’ of Chomsky 2001: that is, a goal must have a uF itself. In Indo-European A-relations, the feature satisfying this requirement is usually a DP’s uCase.

As noted in section 1.1, we depart from more recent Minimalist ideas in several significant respects. Chomsky (2007, 2008) proposes that agreement is universally restricted to one occurrence per phase, for reasons rooted in the conceptual-intentional interface. The data in (1) are a preliminary indicator of the difficulties that Bantu languages raise for this view, particularly given evidence to be presented in section 4 that agreement on ‘how’ is independent of SA on the verb. In section 6.5, we argue for a principled and predictive account of the differences between agreement in English and agreement in Bantu languages. On specific problems that ‘how’ raises for Chomsky’s (2007, 2008) proposals, see section 6.4.

2 Overview of Agreeing ‘How’

2.1 Introduction

Despite its novelty from a crosslinguistic standpoint, agreeing ‘how’ is unremarkable in many language-internal respects, sharing several key properties with other Luyia wh-phrases.2 This section describes the morphosyntactic properties of agreeing ‘how’ with a focus on Lubukusu, and presents a brief preview of our analysis.

2.2 Long and Short Forms

All wh-words apart from naanu ‘who’ have both short and long forms in Lubukusu, the long forms bearing the suffix -na. Long forms are obligatory when a wh-phrase is clefted—perhaps, as Justine Sikuku ( pers. comm.) suggests, because clefted material is emphasized and the short forms cannot bear emphasis (see section 2.7 for discussion and examples of clefts). The suffix is optional in all nonemphatic contexts that we are aware of. What is important for our purposes is not the nature or function of the suffixes; rather, it is that ‘how’ patterns with the other wh-phrases in this respect (see (2)).

(2)

graphic

2.3 Two Locations

Although agreeing ‘how’ is usually clause-final, like other nonsubject wh-phrases in many Bantu languages it has an alternative position immediately after the verb. (3) demonstrates this with Lusaamia data.

(3)

graphic

Similarly, if a verb has a clausal complement, ‘how’ can either precede or follow, as the Lubukusu examples (4a–b) show.

(4)

  • a.

    graphic

    2SG.SA-PST-hear-PST 2SG-how that 2-2-children 2SA-arrived

    ‘How did you hear that the children arrived?’

  • b.

    graphic

    2SG.SA-PST-hear-PST that 2-2-children 2SA-arrived 2SG-how

    ‘How did you hear that the children arrived?’

The fact that agreeing ‘how’ has access to the immediate postverbal position aligns it with other Luyia wh-phrases. (5) and (6) (taken from Wasike 2007:356) illustrate the same alternation between final and immediately postverbal positions for ‘where’ and ‘when’ (here and henceforth, examples are Lubukusu unless otherwise indicated).

(5)

  • a.

    graphic

    2-2-children 2SA-PST-receive 10-report where

    ‘Where did the children receive information?’

  • b.

    graphic

    2-2-children 2SA-PST-receive where 10-report

    ‘Where did the children receive information?’

(6)

  • a.

    graphic

    2-2-children 2SA-PST-receive 10-report when

    ‘When did the children receive information?’

  • b.

    graphic

    2-2-children 2SA-PST-receive when 10-report

    ‘When did the children receive information?’

The alternative position poses intriguing syntactic questions, but it is available to all nonsubject wh-phrases, not to ‘how’ alone. Since the focus of this article is the special properties of agreeing ‘how’, we will not explore this alternation here, and henceforth we restrict examples to final ‘how’.3

2.4 ‘How’ Agreement Correlates with Scope

The agreement features of ‘how’ can serve to disambiguate its scope. (7a–b) are otherwise identical, but the differing φ-features on ‘how’ suffice to distinguish two possible interpretations.

(7)

  • a.

    graphic

    2-2-children 2SA-PST-hear that 3SG.SA-stole 7-book 2-how

    ‘[How did children hear] that she/he stole a book?’

  • b.

    graphic

    2-2-children 2SA-PST-hear that 3SG.SA-stole 7-book 3SG-how

    ‘The children heard that [she/he stole a book by what means]?’

The scope of ‘how’ is ambiguous if ‘how’ follows clauses whose subjects have identical φ-features. Hence, an example like (8) allows two construals.4

(8)

  • graphic

  • Nafula 3SG.SA-PST-hear that Wafula 3SG.SA-stole 7-book 3SG-how

  • a.

    ‘[How did Nafula hear] that Wafula stole a book?’

  • b.

    ‘Nafula heard that [Wafula stole a book how]?’

‘How’ located in the main clause cannot have lower-clause construal, or agree with an embedded subject. In (9), though the subject of each clause is third person singular and ‘how’ is marked third person singular, its scope can only be the matrix clause. In (10), the two subjects have different φ-features and ‘how’ is located in the matrix clause, agreeing with the embedded subject. The result is simply unacceptable.

(9)

  • graphic

  • Nafula 3SG.SA-PST-hear 3SG-how that Wafula 3SG.SA-stole 7-book

  • OK: ‘[How did Nafula hear] that Wafula stole the book?’

  • #‘Nafula heard that [Wafula stole the book by what means]?’

(10)

  • graphic

  • 2SG.SA-PST-hear-PST 2-how that 2-2-children 2SA-arrived

  • ‘How did you hear that the children arrived?’

2.5 A Wh-Adnominal Homophone

Agreeing ‘how’ has a wh-adnominal homophone usually translated as ‘what kind’ (see (11) and (12)). The homophone also lacks intrinsic φ-features, instead acquiring φ-values through agreement (see section 3 for discussion on this point).

(11)

  • Ka-ma-ki ka-rie ka-katikh-e?

  • 6-6-eggs 6-how 6SA-break-PST

  • ‘What kind of eggs broke?’

(12)

  • Ku-mu-nyu ku-rie ku-kho-kuya?

  • 3-3-soup 3-how 3SA-PRES-cook

  • ‘What kind of soup is cooking?’

It seems unlikely that the identity between agreeing ‘how’ and ‘what kind’ is accidental. In fact, question words meaning ‘how’ and ‘what kind’ are also homophonous in Polish, German, and Russian, as are the adverbial modifier ‘thus’ and adnominal ‘such’. Given these similarities, Landman and Morzycki (2003) argue that the two types of expressions are parallel in that adverbials denote kinds of events, while adnominals denote kinds of individuals. Landman and Morzycki propose that ‘thus’ and ‘such’ are functionally identical in being kind-anaphoric: the adverbial, manner modifier ‘thus’ in a sentence like She danced thus is anaphoric to a contextually salient kind of event, in their view, whereas the ‘such’ adnominal modifier in an expression like such a dog is anaphoric to kinds of individuals. ‘How’ and ‘what kind’ can be viewed as the whcounterparts to these two parallel kind-anaphors. Landman and Morzycki’s analysis explains why an identical wh-modifier can be used both adnominally and adverbially in the languages they examine. We assume that this approach can be extended to explain the dual uses of Lubukusu’s agreeing -rie in adnominal and adverbial wh-questions.5

2.6 φ-Features of Wh-Operators: Why Does ‘How’ Agree?

A natural question about -rie(na) ‘how’ is why it must agree. Most expressions in narrow Bantu languages either have intrinsic φ-features or acquire them in the form of agreement (for discussion and references on the latter issue, see section 6.5). But most Luyia wh-words differ from ‘how’ and ‘what kind’ in being lexically specified for noun class.

(13)

graphic

An explanation for this difference between ‘how/what kind’ on the one hand and ‘who’, ‘what’, and ‘where’ on the other is readily available in the fact that answers to the latter name an individual or location—items that, following Chomsky’s (1981) categorial typology, are typically [+N]. All nominal expressions have intrinsic φ-features in Bantu languages, and it is in keeping with this generalization that wh-words have noun class features like those of their canonical answer types (‘who’ is answered with nouns referring to humans, so is class 1/2; etc.). Answers to ‘when’ questions are usually nominal expressions like ‘today’ or ‘last week’; hence, the corresponding question word is [+N] and has class features. Reason questions are expressed with the complex wh-phrase sikila siina ‘reason what’ (13e), class 7 by nature of sikila’s and siina’s class 7 membership. The fact that manner adverbs, adjectives, and their wh-counterparts ‘how’ and ‘what kind’ are nonnominal and lack noun class features is likely attributable to the relationship of these expressions to kinds rather than entities or individuals (see section 2.5), and perhaps to the fact that manners and the like are less countable than entities, individuals, or even reasons.

Summing up, most wh-phrases are entity-oriented [+N] expressions, so they have class features; on the other hand, manner and adnominal adjuncts are kind-oriented and hence nonnominal. Their lack of intrinsic φ-features is therefore not surprising. Though this does not predict that they will have to agree,6 it opens up the possibility for them to do so, and Lubukusu and Lusaamia exploit this possibility.

2.7 Clefting Wh-Phrases: Only ‘How’ Cannot Cleft

One other point of contrast between ‘how’ and other Lubukusu wh-phrases is that only ‘how’ cannot cleft. Following Wasike (2007), we propose below that this is a direct consequence of the fact that its φ-features are not intrinsic, but the result of agreement.7

Wasike (2007:361–362) shows that most wh-phrases in Lubukusu can appear left-peripherally in cleft constructions as an alternative to surfacing in situ. Clefts involve an agreeing complementizer, as shown in (14a–b) for siina ‘what’ and waaena ‘when’.8

(14)

  • a.

    graphic

    ‘What did Nangila cook for Wafula?’

  • b.

    graphic

    ‘Where is it that Nafula is going?’

‘How’ cannot be clefted, however, whether agreement on the cleft is class 1 to match the features of the subject that ‘how’ agrees with or default class 16 (15b). Wasike also notes that manner adverbs cannot be clefted either (15c); nor can prepositional phases (15d).

(15)

  • a.

    graphic

    Nafula 3SG.SA-PRES-walk 3SG-how

    ‘How is Nafula walking?’

  • b.

    graphic

    3SG/16-be 3SG-how COMP-3SG/16 Nafula 3SG.SA-PRES-walk

    ‘How is it that Nafula is walking?’

  • c.

    graphic

    ‘It is quickly that Wafula drank beer.’

  • d.

    graphic

    3SG/16-be with Nekesa COMP-3SG/16 Wekesa 3SG.SA-PST-speak

    ‘It is with Nekesa that Wekesa spoke.’

Wasike proposes that only expressions with interpretable φ-features can be clefted in Lubukusu. Following Carstens (2010b, 2011) and Bošković (to appear), we analyze the grammatical gender features of nouns as valued but uninterpretable; we also assume with Carstens (1991) that Bantu noun class is a composite of gender and number features (see footnote 1). We accordingly adopt a modified version of Wasike’s proposal under which only expressions with intrinsic φ-features can be clefted in Lubukusu. This we take to be a consequence of the broader claim in (16), adapted from Carstens 2010b (here and henceforth, iφ = intrinsic φ-features).9

(16)

  • No Agree with agreement

  • Only iφ can value uφ.

Carstens proposes (16) on the basis of the fact that although Romance determiners agree with nouns in gender, when uφ of Romance T probes DP it does not acquire a gender value from the gender agreement on D. Carstens argues that SA on T includes gender features only in languages where N systematically adjoins to D. This morphological amalgamation leads D(P) to inherit N’s gender feature, and gender is therefore accessible to value uφ of T (see section 6.5 for details). Carstens suggests that (16) is due to the fact that agreement/Case values are context-sensitive information about how a uF should be pronounced in a particular location, but the rationale is not crucial here. What is important is that (16) predicts the inability of ‘how’ to cleft, since Lubukusu clefts must agree with the clefted expression.

As our exploration of ‘how’ proceeds, it will become clear that T probes the subject in its base position, ignoring valued uφ on ‘how’. In addition to accounting for the inability of ‘how’ to cleft, (16) explains why ‘how’ does not intervene to block Agree (T, SU).

2.8 Summary and Analytical Preview

We have shown that agreeing ‘how’ is a wh-expression with unvalued φ-features (uφ). Possible analyses consistent with the facts presented so far include at least the following:10

(17) Potential hypotheses for agreeing ‘how’

  • a.

    Agreeing ‘how’ is a sort of floating wh-modifier of the subject like English all in The boys have all left.

  • b.

    Agreeing ‘how’ is an interrogative version of a subject-oriented depictive secondary predicate.

  • c.

    Agreeing ‘how’ receives its φ-features by downward spreading from T.

  • d.

    Agreeing ‘how’ is a predicate with a pro subject bound by the higher subject.

  • e.

    Agreeing ‘how’ is a head in the functional structure of the clause and receives its φ-values through Spec-head agreement.

  • f.

    Agreeing ‘how’ is a wh-adjunct XP comparable to English ‘how’ but with uφ that probes the subject from a position adjoined to either (i) vP or (ii) TP.

Because responding to these potential analyses requires extensive diagnostic evidence, full arguments against (17a–d) are distributed throughout the article. In section 3, we will provide evidence for rejecting (17a) and (17b) (flawed also by their incompatibility with our analysis of ‘how’ as an event-kind interrogative). In section 4, we present some highly significant patterns connected with noncanonical subjects, specifically agreement on ‘how’ with subject operators and inverted locatives. These facts lead us, in section 5, to reject the direct dependency between T and ‘how’ suggested in (17c), the controlled pro analysis in (17d), and the TP-adjunct hypothesis in (17fii). We argue in section 5 that (17fi), in which ‘how’ is a downward probing vP-level wh-adjunct, accounts for all the facts (see (18)).11 In section 6, we provide evidence against the only remaining alternative hypothesis (17e).

(18)

graphic

3 Two Failed Analyses

3.1 Introduction

In this section, we address the first two hypotheses of (17), demonstrating that ‘how’ is not a floating modifier or a secondary predicate of any sort.

3.2 Agreeing ‘How’ Is Not a Floating Modifier

3.2.1 The Hypothesis

One conceivable approach to agreeing ‘how’ might be to analyze it along the lines of Sportiche’s (1988) theory of floating quantifiers. That is, ‘how’ and the subject could be hypothesized to originate as a single constituent XP in which ‘how’ is the subject’s modifier. The subject raises out of XP and moves leftward, stranding ‘how’ as shown in (19).

(19) A hypothetical floating modifier approach to ‘how’

  • a.

    . . . [XP SU uφ-how] . . .

  • b.

    SU . . . [XPSU uφ-how] . . .

While this analysis would make the agreement relationship between ‘how’ and the subject easy to explain, it is inconsistent with the restriction to a ‘what kind’ interpretation for DP-internal ‘how’ (see (11) and (12)). We will also show in section 4 that ‘how’ agrees with an in-situ subject of a locative inversion construction—a context where ‘‘floating’’ could not have taken place, casting further doubt on this as a possible analysis. Expletive and infinitive constructions provide additional evidence that we turn to next.

3.2.2 Expletives and Infinitives

In Lubukusu [expletive . . . CP] constructions, the verb bears class 6 SA (see (20a)). (20b) shows that ‘how’ can agree with the expletive in these contexts.

(20)

  • a.

    Ka-nyalikhana khu-khu-pila lu-simu.

    6SA-be.possible 15-2SG.OA-hit 11-phone

    ‘It is possible to call you.’

  • b.

    Ka-nyalikhana khu-khu-pila lu-simu ka-rie?

    6SA-be.possible 15-2SG.OA-hit 11-phone 6-how

    ‘How is it possible to call you?’

Alternatively, ‘how’ can agree with the infinitival clause. Infinitives in Bantu languages often have nominal properties; they comprise a separate noun class, class 15, and can control agreement on modifiers. (21a–b) illustrate agreement of ‘how’ in class 15.12

(21)

  • a.

    Ka-nyalikhana khu-khu-pila lu-simu khu-rie?

    6SA-be.possible 15-2SG.OA-hit 11-phone 15-how

    ‘How is it possible to call you?’

  • b.

    Khu-khu-pila lu-simu khu-nyalikhana khu-rie?

    15-2SG.OA-hit 11-phone 15SA-be.possible 15-how

    ‘How is it possible to call you?’

(21a–b) raise questions about the nature of infinitives in Bantu languages that are outside the scope of this article (but see footnote 12). For present purposes, it suffices to say that these examples and the expletive agreement examples in (20) seem quite anomalous for a discontinuousmodifier approach to agreeing ‘how’.

3.3 Agreeing ‘How’ Is Not a Subject-Oriented Depictive

The analysis represented in (18) is not far removed from recent treatments of subject-oriented depictive secondary predicates like drunk in Mary likes to attend church drunk, or naked in John danced naked: Pylkkӓnen (2002, 2008) proposes that subject-oriented depictives merge at the vP level, and Irimia (2005) uses Agree to account for agreement on depictive secondary predicates in Armenian, Slovenian, and Albanian. One might therefore consider analyzing -rie ‘how’ as a wh-depictive secondary predicate.13

The ability of ‘how’ to agree with an expletive or an infinitive (see section 3.2) is a first reason for skepticism regarding such an analysis. In this section, we present additional evidence against the approach by showing that the interpretation of ‘how’ is not subject-oriented, that the uses of ‘how’ are consistent in active and passive sentences, and that ‘how’ cannot agree with and modify an object in the way that secondary predicates typically can.

First, the most natural answers to agreeing ‘how’ questions are not subject-oriented depictives. This is apparent in (22), where responses that came to mind for speakers we interviewed describe manners, instruments, and even properties of the direct object. While not precluded, a subject-oriented depictive answer is judged unexpected and somewhat anomalous. This result is unexpected if ‘how’ is analyzed as a wh-version of a subject-oriented secondary predicate.

(22)

  • a.

    graphic

    ‘How did he eat the meat?’

  • b.

    graphic

    slowly

    ‘Slowly.’

  • c.

    graphic

    with 7-spoon

    ‘With a spoon.’

  • d.

    graphic

    9-raw

    ‘Raw.’

  • e.

    graphic

    #‘He ate it (while he was) drunk /tired.’

One might perhaps consider treating manners and instruments as extended properties of agents in an event and hence not incompatible with the subject-oriented depictive analysis (though (22d–e) would constitute anomalies to be explained). This brings us to our second reason for rejecting an analysis of agreeing ‘how’ as a secondary predicate: its uses and interpretation are indifferent to whether a sentence is active or passive, and this is uncharacteristic of secondary predication. ‘How’ is natural in both active and passive clauses, agreeing with the surface subject in either case, and consistently eliciting the same kinds of answers (see (23)).14

(23)

  • a.

    graphic

    ‘How was the meat eaten?’

  • b.

    graphic

    hastily

    ‘Hastily.’

  • c.

    graphic

    with 7-spoon

    ‘With a spoon.’

  • d.

    graphic

    9-raw

    ‘Raw.’

Thus, the interpretation of ‘how’ does not covary with changes in the item it agrees with—there seems to be no link between the properties of the surface subject and the meaning or function of ‘how’. This contrasts sharply with the behavior of secondary predicates.

Our third argument is based on Pylkkänen’s (2002, 2008) observation that a depictive cannot be predicated of the implicit argument in a passive, a generalization that holds up for Lubukusu secondary predicates.15 (24a) and (24b) show that a Lubukusu postverbal depictive can be predicated of an active subject or the derived subject of a passive, as such a depictive can in English. (25) demonstrates that a depictive cannot modify the implicit agent of a passive.

(24)

  • a.

    E-sok-ile mu-nyanja si-chula.16

    1SG.SA-swim-PST 18-9lake 7-naked

    ‘I swam in the lake naked.’

  • b.

    E-nyama e-li-l-we e-mbisi.

    9-meat 9SA-eat-PST-PASS 9-raw

    ‘The meat was eaten raw.’

(25)

  • *E-barua ey-andik-we si-chula.

  • 9-letter 9SA-write-PASS 7-naked

  • *‘The letter was written naked.’

In contrast, we saw in (23) that ‘how’ is fine in a passive sentence, ranging over the same interpretations as in an active sentence even though it agrees with the thematic object.

The comparisons among (22), (23), and (25) strongly argue that ‘how’ is not a subjectoriented depictive secondary predicate. Not only are object-oriented answers possible when ‘how’ agrees with the subject of a transitive clause; the ability of agreeing ‘how’ to interrogate such matters as the manner or instrument in an event does not covary with whether the logical subject is explicit or implicit, or with whether the structural subject is agent or theme. These facts are consistent with our analysis of ‘how’ as an ordinary manner wh-phrase, not with an analysis of ‘how’ as a wh-version of a subject-oriented depictive.

A final reason for rejecting an analysis of ‘how’ as a wh-depictive secondary predicate is that it is unable to modify and agree with an object and retain the ‘how’ meaning, in contrast to the general behavior of secondary predicates. While the word -rie(na) may in fact agree with an object, it is easily demonstrated that when it does so, its location is DP-internal, as in (11) and (12), and not that of a depictive secondary predicate at all. This is clear from the fact that the ‘what kind’ meaning is its only interpretation in this context (see (26))—the ‘how’ reading is lost. As Pylkkänen (2002:26) notes, depictive secondary predicates describe a state that holds of one of the arguments of the verb during the event described by the verb. For this reason, individuallevel adjectives sound strange as depictives (an observation for which Pylkkänen cites Geuder 2000); thus, He entered the room annoyed is fine, unlike #He entered the room tall. Given this, the restriction to ‘what kind’ readings for object-agreeing -rie(na) is inconsistent with a whsecondary predicate analysis.

(26)

  • Ba-khal-ile lu-karatasi lu-riena?

  • 2SA-cut-PST 11-paper 11-how

  • ‘What kind of paper did they cut?’ (i.e., letter or legal size)

  • #‘How did they cut the paper?’ (i.e., into circles or triangles)

In contrast to cases like (26), an object-oriented answer to a -rie(na) question is readily available when -rie(na) agrees with the subject, as shown in (27).17

(27)

  • a.

    Ba-khal-ile lu-karatasi ba-riena?

    2SA-cut-PST 11-paper 2-how

    ‘How did they cut the paper?’

  • b.

    graphic

    18-8-circle / 18-8-piece

    ‘Into circles/pieces.’

Moreover, when -rie(na) is construed with an object, it must generally be adjacent to it, as demonstrated in (28).18 This is consistent with an analysis of -rie(na) as a DP-internal wh-phrase, but incompatible with an object-oriented depictive secondary predicate analysis.

(28)

  • *Ba-khal-ile lu-karatasi [nende ka-ma-kasi] lu-riena?

  • 2SA-cut-PST 11-paper with 6-6-scissors 11-how

  • ?*‘What kind of paper did they cut with the scissors?’ (i.e., letter or legal size)

  • *‘How did they cut the paper with the scissors?’ (i.e., into circles or triangles)

To sum up this section: -rie(na) cannot function as a wh-object depictive, and in its use forming ‘how’ questions -rie(na) does not pattern like a wh-subject depictive secondary predicate either; the most natural kinds of responses to -rie(na) questions are not subject depictives, and the content of responses is consistent even when the subject is changed from an agent to a theme by passivization. For all of these reasons, we will not pursue a wh-secondary predicate analysis of -rie(na) and will henceforth refer to -rie(na) solely as agreeing ‘how’, avoiding for the most part further discussion of its DP-internal usage.

3.4 Summary and Preview

In this section, we argued against two logical possibilities for analyzing agreeing ‘how’: (17a) as a floating modifier and (17b) as a wh-subject depictive secondary predicate. We will show in the next section that in operator and inversion constructions, agreement on ‘how’ differs from SA on T. The facts lead us to reject (17c), the hypothesis that agreement on ‘how’ is copied or spread from T, and (17d), the hypothesis that ‘how’ is a predicate agreeing with a pro subject controlled by the overt subject of the main clause. Arguments against analysis of ‘how’ as a functional head (hypothesis (17e)) are presented in section 6.

4 Mismatches between ‘How’ and T with Noncanonical Subjects

4.1 Introduction

As noted in (17c), a plausible approach to Lubukusu agreeing ‘how’ might rely on some mechanism of feature sharing between ‘how’ and T. We will argue against this possibility on the basis of three constructions with noncanonical subjects: subject extractions and two varieties of locative inversion.19 After briefly describing each construction, we illustrate the pattern of agreement that occurs when the ‘how’ question word is added. We demonstrate that in locative inversion and subject extraction contexts, the agreement features on ‘how’ do not typically match those of SA on the verb. These patterns demonstrate that any sort of feature-sharing relationship between T and ‘how’ is empirically insufficient, leading us to conclude in section 5 that ‘how’ has independent uφ probing downward for valuation by the subject in its base position (Spec,vP)—hypothesis (17fi), as illustrated in (18).

4.2 Operator Subjects

When a third person singular animate subject is questioned or relativized in Lubukusu, a special agreement form [o-] appears on the verb in place of the usual [a-] (the special form is glossed AAE = alternative agreement effect). These facts are illustrated in (29), from Wasike 2007: 235–236.

(29)

  • a.

    Naliaka a-li mu-nju.

    Naliaka 3SG.SA-be 18-house

    ‘Naliaka is in the house.’

  • b.

    Naanu oo-li mu-nju?

    1who AAE-be 18-house

    ‘Who is in the house?’

  • c.

    *Naanu a-li mu-nju?

    1who 3SG.SA-be 18-house

    ‘Who is in the house?’

Kinyalolo (1991) demonstrates that the crucial property of AAE in Kilega is an absence of person features (agreement in noun class is not affected). Henderson (2009b, to appear) and Diercks (2010) show that in Bemba, Luganda, and Lubukusu as well, the alternative agreement lacks person features.20 They argue that syntactic strategies to avoid or repair extraction from Spec,TP give rise to this effect, relating it to that-trace and que-qui phenomena of English and French, respectively (see Rizzi and Shlonsky 2007). What is important for our purposes is not the details of these analyses but the fact that ‘how’ can only bear [a-] agreement (see (30a–b)). Thus, there is a mismatch between SA and ‘how’ agreement in these cases, unexpected if the latter were contingent on the former or related to Spec,TP.s

(30)

  • a.

    Naanu oo-tekh-ile e-ngokho *o-/a-riena?

    1who AAE-cook-PST 9-chicken *AAE-/3SG-how

    ‘Who cooked the chicken how?’

  • b.

    Ba-a-bona o-mu-ndu ow-a-tekha e-ngokho *o-/a-riena?

    2SA-PST-see 1-1-person AAE-PST-cook 9-chicken *AAE-/3SG-how

    ‘They saw [the person who cooked the chicken how]?’

4.3 ‘How’ Agreement in Locative Inversion Constructions

4.3.1 Introduction to Luyia Locatives

Like many other Bantu languages, Lubukusu has three locative noun classes. In Lubukusu, these are primarily expressed in the morphemes a-, khu-, or mu- added to a noun, replacing the so-called pre-prefix but leaving intact the inner prefix of the noun’s intrinsic class (see (31), from Mutonyi 2000:26–27). As (32) demonstrates, locativized nouns can appear in argument positions and control agreement in the locative noun classes; thus, they pattern like DPs.

(31)

  • a.

    ku-mu-lyaango

    3-3-door

    ‘(a/the) door’

  • b.

    a-mu-lyaango

    16-3-door

    ‘near the door’

  • c.

    khu-mu-lyaango

    17-3-door

    ‘on the door’

  • d.

    mu-mu-lyaango

    18-3-door

    ‘in the door’

(32)

  • Mu-nju mu-unya.

  • 18-house 18SA-stink

  • ‘The inside of the house stinks.’

On the basis of similar facts in Shona and Chicheŵa, Myers (1987) and Bresnan and Mchombo (1995) propose that the cognate morphemes pa, ku, and mu are nouns. Because they do not trigger ‘of’ insertion like other nouns and fail to meet a two-mora minimum size requirement from which only functional categories are exempt in Chicheŵa, Carstens (1997) argues instead that Chicheŵa locative DPs are headed by null nouns: [N e]cl16 ‘surface’, [N e]cl17 ‘vicinity’, and [N e]cl18 ‘inside’. Assuming there is a DP headed by a silent locative noun that contains and dominates the DP headed by the ‘‘locativized’’ noun allows Carstens to explain concord and modification phenomena exemplified in (33) (Carstens 1997:384–385). Carstens analyzes Chicheŵa pa, ku, and mu as gender-particular Case markers or prepositions selected by the silent locative nominals (see (34)).21

We have found that a Lubukusu locativized noun can have modifiers that agree in its intrinsic class or in its locative class, just as in Chicheŵa—evidence that the complement of a-, khu-, or mu- is at least an NP and perhaps a DP. The superordinate locative DP always controls SA in the locative class. Hence, we adopt Carstens’s analysis of Chicheŵa for Lubukusu. The structure of (35a) is as shown in (35b). (N-to-D movements proposed in Carstens’s account are omitted for simplicity’s sake.)

(33)

  • a.

    graphic

    ‘There are flies on every door.’

  • b.

    Pa-li n-chenche pa chi-seko pariponse.

    16SA-be 10-flies 16LOC 7-door 16every

    ‘There are flies all over the door.’ (i.e., on every surface of it)

(34)

graphic

(35)

  • a.

    graphic

    18-9-house 9-small 18SA-stink

    ‘The inside of the small house stinks.’

  • b.

    graphic

We turn now to the facts of locative inversion. We will show that, like subject extraction, locative inversion exhibits mismatches between agreement on ‘how’ and agreement on T.

4.3.2 Two Types of Locative Inversion

Two types of locative inversion (LI) are found in Lubukusu. Both involve a postverbal clitic agreeing with the fronted locative phrase (on which, see section 7), but their properties differ in other crucial respects. In one variety, which Diercks (2011b) calls repeated agreement LI (RALI), subject agreement (SA) reflects the features of the fronted locative phrase (henceforth DPloc; and see (37)). Only unaccusative verbs can participate in this construction. In the other variety, which Diercks calls disjoint agreement LI (DALI), SA is with the postverbal thematic subject (SU) (see (38)). Both unaccusative and unergative verbs can participate in DALI. (Henceforth, for clarity we underline the class prefix of the thematic subject and agreement with it, and we boldface the locative prefix on DPloc and locative agreement with it.)

(36)

  • graphic

  • ‘A tree fell in the forest.’

(37)

  • graphic

  • ‘In the forest fell a tree.’

(38)

  • a.

    graphic

    ‘In the forest fell a tree.’

  • b.

    graphic

    ‘Into the hole jumped the rabbit.’

Diercks shows that LI is possible only in clauses where the verb’s meaning includes or is suggestive of directionality, as with ‘fall’ and ‘jump into’, or is fairly intimately tied to a location, as is arguably the case for ‘grow’ referring to trees and plants, entailing as it does that they send roots into the ground. This restriction is illustrated by the unacceptability of (39).

(39)

  • *Mw-i-duka mw-/ka-a-chekha-moo-mw-ana.

  • 18-9-store 18SA/3SG.SA-PST-laugh-18LOC 1-1-child

  • ‘In the store laughed a child.’

On the basis of facts like (39), Diercks argues that inverted locatives are selected arguments, merged as sister to V. Apart from this, he demonstrates that the constructions have the divergent structures in (40) and (41).

(40)

  • Repeated agreement LI (RALI): DPlocraises to Spec,TP; thematic subject in situ

  • graphic

(41) Disjoint agreement LI (DALI): Thematic subject to Spec,TP; DPlocto Spec,CP

graphic

The crucial contrast is thus that in RALI, DPloc undergoes A-movement to Spec,TP, whereas DALI involves raising of the subject to Spec,TP and raising of DPloc to Spec,CP. The features of SA on the verb are a helpful indicator of this difference.

While we do not want to devote too much space to the details of LI, it is worth presenting the evidence for the representations in (40) and (41) since LI plays an important role in the analysis of agreeing ‘how’ in sections to follow. All of the data and arguments below are drawn from Diercks 2011b.

Our first concern is to motivate the contrasting positions of the fronted locative phrases in the two constructions. The fact that SA agrees with DPloc in RALI but not DALI constructions provides an initial argument in support of the divergent structures in (40) and (41). Additional, crucial evidence for the two structures comes from morphosyntactic properties of extraction. Subject extraction triggers C-agreement and disallows a freestanding complementizer, as seen in the subject relative clause in (42). In contrast, nonsubject extraction, illustrated by the object relative clause in (43), requires the freestanding complementizer and disallows C-agreement. Thus, extraction provides a diagnostic for whether an Ā-moving expression is a structural subject or not.

(42)

  • Subject relative clause

  • ba-ba-andu (*ni-bo) ba-ba-a-kula ka-ma-tunda

  • 2-2-people (*COMP-2) 2CA-2SA-PST-buy 6-6-fruit

  • ‘the people who bought the fruit’

(43)

  • Object relative clause

  • ka-ma-tunda *(ni-ko) ba-ba-andu ba-a-kula

  • 6-6-fruit *(COMP-6) 2-2-people 2SA-PST-buy

  • ‘the fruit which the people bought’

The positions posited for the fronted locatives in (40) versus (41) for RALI and DALI, respectively, make different predictions for the morphosyntax of extraction. When a fronted locative is extracted, in RALI constructions C-agreement should occur because the locative is in canonical subject position. In contrast, if the locative is relativized or questioned in a DALI construction, the freestanding complementizer should occur rather than C-agreement, because the fronted locative is a nonsubject. As shown in (44) and (45), respectively, these are precisely the facts.

(44)

  • RALI relative clause

  • graphic

  • ‘the forest in which fell a tree’

(45)

  • graphic

  • ‘the forest in which fell a tree’

Diercks (2010, 2011b) demonstrates that these same morphosyntactic asymmetries consistently differentiate subject and nonsubject movements in subject-to-subject raising, clefts, and wh-extraction; when applied to LI constructions, they always support the conclusion that DPloc occupies the canonical subject position, Spec,TP, only in RALI. Given that in DALI, DPloc is in a left edge position (≠ Spec,TP), an Āmovement approach is attractive. This approach permits a simple account of why the verb agrees with the logical subject in DALI constructions: while uφ of T probes DPloc in a RALI construction, DPloc interacts instead with C in a DALI construction and uφ of T probes the thematic subject. The expression that values uφ of T generally raises to Spec,TP in Lubukusu; hence, the order [DPloc-V-SU] suggests that the verb raises across Spec,TP to C.

We turn now to evidence supporting the claim that in DALI, the order [DPloc-V-SU] results from V-to-T-to-C movement. While Diercks (2011b) draws on several sorts of evidence, for the sake of brevity we restrict discussion to the location of the manner adverb bwangu ‘quickly’. Diercks argues that bwangu is a vP-level adjunct because it can never precede the inflected verb. In a clause where no inversion has occurred, bwangu can immediately follow the verb, in which case Diercks argues that it is left-adjoined to vP. It can also surface clause-finally, in which case it is right-adjoined under Diercks’s account (see (46)).22

(46)

  • a.

    Ku-mu-saala (*) kw-a-kwa (✓) mu-mu-siiru o-mwo (bwangu).

    3-3-tree 3SA-PST-fall 18-3-forest DEM-18 (quickly)

    ‘A tree fell in this forest quickly.’

  • b.

    [TP DPloc (*Adv) fall-T [vP (✓Adv) tv [VPtV DPloc] (✓Adv)]]

Taken together with the structures in (40) and (41), this analysis of bwangu makes the following prediction: in DALI constructions, the adverb should only be able to occur clause-finally, because the postverbal subject and the inflected verb have moved leftward past even the adverb’s leftadjoined position, to Spec,TP and C, respectively. In contrast, assuming the postverbal subject in RALI is in situ within the VP, then the immediately postverbal adverb position is predicted to be available. (47) and (48) show that the prediction is borne out.

(47)

  • a.

    Mu-mu-siiru o-mwo (*) kw-a-kwa-mo (*) ku-mu-saala (✓bwangu). DALI

    18-3-forest DEM-18 3SA-PST-fall-18LOC 3-3-tree (quickly)

    ‘In this forest fell a tree quickly.’

  • b.

    [CP DPloc fall-T-C [TP SU tT [vP (✓Adv) [tv [VPtSUtVtloc (✓dv)]]]]]

(48)

  • a.

    Mu-mu-siiru o-mwo (*) mw-a-kwa-mo (✓) ku-mu-saala (bwangu). RALI

    18-3-forest DEM-18 18SA-PST-fall-18LOC 3-3-tree (quickly)

    ‘In this forest fell a tree quickly.’

  • b.

    [TP DPloc fall-T [vP (✓Adv) tv [VP SU tVtloc (✓Adv)]]]

An anonymous reviewer points out that the V-to-T-to-C movement component of this analysis runs counter to Julien’s (2002) influential proposal that verbs in Shona raise only to the low middle field of the clause, immediately above vP. In Julien’s account, verbs stop at the point of the highest verbal suffix, and verbal prefixes concatenate with the verb at PF. Assuming with Kayne (1994) that head movement left-adjoins to its host, the analysis explains why suffixes show mirror-image ordering in relation to the hypothetical inflectional projections of the clause, while prefixes exhibit ‘‘direct’’ ordering. The structure in (49) replicates Julien’s analysis of Shona verbs.

(49) [FinP SU SA- [TP Tns- [MoodP V-v-Mood [vP . . . ]]]]

Assuming that the Lubukusu diagnostics discussed above and in Diercks 2010, 2011b are valid, DALI constructions raise an important challenge for a Julien-style account of the inflected verb in Lubukusu: the diagnostics place the subject in Spec,TP and, absent head movement of the verb over Spec,TP, there is no obvious means of accounting for DALI’s LOC-V-SU word order. While it is conceivable that this could be derived by pronunciation of the lower, vP-internal copy of the subject, independent motivation for this approach is lacking; and the adverb facts noted above become mysterious. Thus, the explanatory benefits of the Julien-style account with respect to morpheme ordering must be weighed against significant word order complications that it introduces in DALI constructions. In contrast, under the V-to-T-to-C movement analysis that we adopt, head movement accounts straightforwardly for the syntactic facts of word order including verb, subject, and adverb positions. Morpheme order on the verb must be attributed to morphological mechanisms (contra Julien 2002).

It is important to note, however, that the best analysis of the verb position in DALI constructions is a question independent of the location of DPloc and, given the potential analysis of lowercopy subject Spell-Out, it is even independent of the question of whether the subject raises to Spec,TP. Our analysis assumes that DALI is an Ā-movement type of LI construction, while RALI is an A-movement type of LI construction landing in Spec,TP. This is all that is crucial to interpreting the evidence of RALI and DALI vis-à-vis the properties of agreeing ‘how’. For concreteness and expository convenience, we will abstract away from the pros and cons of verb movement, assuming it for the reasons given in Diercks 2011b, summarized above. Should the alternative in terms of lower-copy subject pronunciation prove correct, the analysis of agreeing ‘how’ in LI constructions should not be affected.

Turning back to agreeing ‘how’, as in the subject operator constructions discussed in section 4.3.1, there are two logical possibilities when ‘how’ is added to an LI construction: the features on ‘how’ might match the features of SA and hence the contents of Spec,TP, or they might fail to match, agreeing with the expression that remains in vP. In section 4.3.3, we describe the pattern of facts. In section 5, we provide a comprehensive analysis in terms of downward probing by vP-adjoined ‘how’.

4.3.3 Agreeing ‘How’ in Locative Constructions

In noninverted sentences involving locatives, ‘how’ can only agree with the preverbal subject, as is consistent with the data reported so far.

(50)

  • graphic

  • ‘How did a tree fall in the forest?’

In inversion constructions, however, judgments diverge slightly among speakers. Of our three main consultants, speakers 1 and 2 accept agreement on ‘how’ only with the postverbal thematic subject. Agreement with the class 18 preverbal DPloc is strongly rejected even in (51a), where SA on the verb is class 18.23

(51) Locative inversion + agreeing ‘how’ variety A: ‘How’ can agree only with the thematic subject (speakers 1 and 2).

  • a.

    graphic

    ‘How did a tree fall in the forest?’ (Lit.: In the forest fell a tree how?) DALI

  • b.

    graphic

    ‘How did a tree fall in the forest?’ (Lit.: In the forest fell a tree how?)

For speaker 3, the two constructions differ. In his judgments, in DALI ‘how’ must agree with the thematic subject (52b), but in RALI ‘how’ can agree with either the subject or DPloc in Spec,TP (52a).

(52) Locative inversion + agreeing ‘how’ variety B: ‘How’ can agree with the preposed locative or the thematic subject in the RALI construction, but only with the thematic subject in the DALI construction (speaker 3).

  • a.

    graphic

    ‘How did a tree fall in the forest?’ (Lit.: In the forest fell a tree how?)

  • b.

    graphic

    ‘How did a tree fall in the forest?’ (Lit.: In the forest fell a tree how?)

The properties of ‘how’ agreement in LI constructions are summarized in (53).

(53)

graphic

The pattern, then, is that ‘how’ can agree with the postverbal thematic subject in all LI constructions for all speakers, and ‘how’ cannot agree with the fronted DPloc in DALI constructions for any speakers. The sole point of variation is in RALI, where the fronted DPloc triggers SA. Variety B allows ‘how’ to agree with DPloc as an option; variety A does not.

5 Analysis of Agreeing ‘How’

5.1 Our Proposal

We have established that in two distinct circumstances, ‘how’ agreement diverges from SA:

  1. When a subject is extracted, ‘how’ agrees with it in person, number, and gender, while the verb agrees with it only in number and gender.

  2. When a locative phrase occupies Spec,TP, all speakers accept agreement with the thematic, in-situ subject. For two of the three speakers we consulted, this is the only licit option.

To account for the mismatches between agreement on ‘how’ and SA on T, we propose that ‘how’ has its own uφ and probes the subject independently. In line with its position and the location of modifiers that are typical answers to ‘how’ questions, we analyze agreeing ‘how’ as a right adjunct to vP. The closest c-command locality constraint on Agree ensures that the uφ features of ‘how’ will be valued by the thematic subject (see (18), repeated here).

(18)

graphic

5.2 Analysis of Agreement with Operator Subjects

As shown in section 4.2, ‘how’ bears canonical [a-] SA in operator constructions, while T bears the special [o-] agreement lacking person features; we represent this schematically in (54).24

(54)

graphic

In section 4.2, we noted that Henderson (2009b, to appear) and Diercks (2009, 2010) analyze T’s special agreement in operator constructions as related to that-trace and que-qui effects, and as a means of avoiding or repairing subject extraction from Spec,TP (see Rizzi and Shlonsky 2007). The mismatch in features between ‘how’ and T is therefore important evidence that the two probe independently.

We conclude that agreement on ‘how’ is not parasitic on T’s features (contra hypothesis (17c)). The facts also strongly suggest that ‘how’ does not probe Spec,TP contra hypothesis (17fii), as it is questionable whether the subject ever occupies this position (see Diercks 2010). Rather, the facts argue that ‘how’ probes and agrees with the subject in its base position.

5.3 Analysis of ‘How’ in Expletive Constructions

A clear prediction of the analysis in (18) is that ‘how’ will agree with an expression other than a thematic subject, provided that expression is the most local goal for Agree. The availability of expletive agreement on ‘how’ strongly supports this conclusion.25 We propose that when an expletive pro is merged in Spec,vP of a ‘how’ question, it is probed by uφ of ‘how’ under closest c-command (see (20b), repeated here, and (55)).

(20)

  • b.

    Ka-nyalikhana khu-khu-pila lu-simu ka-rie?

    6SA-be.possible 15-2SG.OA-hit 11-phone 6-how

    ‘How is it possible to call you?’

(55)

graphic

Merging of expletives to Spec,vP is proposed by Bowers (2002) (and see Radford 2009). Thus, the facts of ‘how’ agreement lend support to these independently motivated analyses.

5.4 Accounting for the Locative Inversion Facts

5.4.1 The Basics

We showed in section 4.3 that there are two Lubukusu LI constructions, RALI and DALI, and that speakers’ judgments diverge regarding which expression ‘how’ agrees with in RALI. In this section, we will present an account of these facts, building on the analysis of LI in Diercks 2011b.

Recall from section 4.3.2 that inverting locatives are merged as sisters to V (see Diercks 2010, 2011b). Given this, the LI constructions of (56) are represented as in (57).26

(56)

  • a.

    graphic

    ‘In the rocks grows a tree.’

  • b.

    graphic

    ‘Into the hole jumped the rabbit.’

(57)

In combination with our analysis of ‘how’ as a vP adjunct, Diercks’s proposals predict that the closest DP to ‘how’ is the thematic subject in both LI constructions. This accounts for the fact that agreement on ‘how’ is always with the thematic subject in Lubukusu variety A.

(58)

  • a.

    Agreeing ‘how’: Unaccusative verb with selected locative

    graphic

  • b.

    Agreeing ‘how’: Unergative verb with selected locative

    graphic

5.4.2 RALI in Variety B

Recall however that in RALI, speaker 3 accepts agreement on ‘how’ with either the postverbal subject or the fronted DPloc in Spec,TP (see (52), repeated here). We propose that in variety B there is a structural ambiguity in unaccusative constructions with selected locatives: either DPloc or the theme argument can merge as sister to the verb (59a–b).

(52) Locative inversion + agreeing ‘how’ variety B: ‘How’ can agree with the preposed locative or the thematic subject in the RALI construction, but only with the thematic subject in the DALI construction (speaker 3).

  • a.

    graphic

    ‘How did a tree fall in the forest?’ (Lit.: In the forest fell a tree how?)

  • b.

    graphic

    ‘How did a tree fall in the forest?’ (Lit.: In the forest fell a tree how?)

(59)

graphic

As a result, either of the two expressions may be more local to ‘how’ (see (60a–b)).

(60)

graphic

5.4.3 DALI and an Issue in Variety B

We showed in section 4.3.3 that for all the speakers consulted, ‘how’ can only agree with the thematic subject in an unergative LI, not with DPloc (see (61)). (58b) sketched out why this is so, and (62) demonstrates in greater detail: a locative selected by V will necessarily merge lower than the agentive subject in Spec,vP. This pattern of facts is predicted even under the assumption that the inverting DPloc must adjoin to the unergative vP to escape the vP phase (Chomsky 2001). It is well-established that Merge takes precedence over Move operations (Chomsky 1995, 2001), so ‘how’ will merge before DPloc adjoins, ensuring that the unergative subject is the most local goal for Agree in its c-command domain.

(61)

  • graphic

  • ‘How did the rabbit jump into the hole this morning?’ (Lit.: Into the hole jumped the rabbit this morning how?)

(62) Locative escaping phasal vP adjoins to it after ‘how’ merges, leaving agreement of ‘how’ with the thematic subject as the only option

graphic

What we have not explained so far is why the alternative base for unaccusative LI in variety B (see (60b), repeated here) can result in agreement on ‘how’ in RALI (where DPloc occupies Spec,TP), but not in DALI, where the thematic subject raises to Spec,TP and DPloc occupies Spec,CP. Agreement between ‘how’ and DPloc is uniformly unacceptable in this case (see (52b), repeated here).

(60)

graphic

(52)

  • b.

    graphic

    ‘How did a tree fall in the forest?’ (Lit.: In the forest fell a tree how?)

Given our proposal that ‘how’ agrees with the highest expression in vP, this state of affairs is exactly what standard locality constraints on A-movement would lead one to expect: DPloc intervenes to block raising of the thematic subject to Spec,TP in (60b). We will show in section 7 that locatives have a special means of transiting out of VP across the Merge position of subjects in LI constructions, but not vice versa (a preview of this account is provided in section 5.4.5). Hence, the only option for continuing (60b) is raising DPloc to Spec,TP—a RALI construction. Our findings argue strongly that expressions inside VP are not equidistant from probes outside it, though this has been a common approach to inversion phenomena over the years (see footnote 40 and references therein). We return to this issue in sections 7 and 8.

5.4.4 The Argument against (17d)

We are now in a position to consider hypothesis (17d), repeated here.

(17) d. Agreeing ‘how’ is a predicate with a pro subject bound by the higher subject.

An anonymous reviewer suggests that since ‘how’ follows the main clause, it should always occupy a position c-commanded by the subject. Agreement could therefore be based on a control or binding relationship between the main clause subject and a pro subject within an XP whose predicate, ‘how’, is a kind of interrogative verb/predicate. We represent a version of this hypothesis in (63), where ‘how’ is identified as V2 heading VP2, and the constituent surrounding ‘how’ is labeled HowP. So that HowP will fall within the c-command domain of both in-situ and exsitu subject positions, we represent it as adjoined to VP.

(63) [TP SUI T [vPSUI v [[VP1 . . . V1 . . . ] [HowPproI [VP2 howV2]]]]]

The fatal argument against this hypothesis is that it cannot account for the facts of RALI, the A-movement LI construction. Recall that RALI is possible only with verbs whose meanings entail directionality or another strong locative component; hence, the inverting locatives are analyzed as selected by the verb. RALI is also restricted to unaccusatives, whose subjects are arguments selected by the verb as well. RALI constructions move DPloc to Spec,TP, leaving the thematic subjects VP-internal. We reproduce in (64) (see (51a), (52a)) the crucial data showing that all three speakers consulted allow ‘how’ to agree with the postverbal in-situ subject; speaker 3 also accepts agreement with the inverted locative as in (65) (see (52a)).

(64)

  • Mu-mu-siiru mw-a-kwa-moku-mu-saala ku-rie?

  • 18-3-forest 18SA-PST-fall-18LOC 3-3-tree 3-how

  • ‘How did a tree fall in the forest?’ (Lit.: In the forest fell a tree how?)

(65)

  • Mu-mu-siiru mw-a-kwa-moku-mu-saala mu-rie?

  • 18-3-forest 18SA-PST-fall-18LOC 3-3-tree 3-how

  • ‘How did a tree fall in the forest?’ (Lit.: In the forest fell a tree how?)

The structures that we adopted for RALI in (59) are repeated below. The problem that arises in relation to the proposed analysis in (63) is that HowP is not low enough for its pro subject to be c-commanded by the VP-internal subject, and there is no modification of the representations under which this could be the case, particularly given that ‘how’ can never agree with a direct object.

(59)

graphic

The only conceivable way to implement a controlled/bound pro analysis, based as it is on ccommand by the main clause agreement trigger, would be to assume (contrary to (59)) that HowP is complement to the verb and that the verb’s theme and locative arguments are higher up in the structure. But there is no evidence to support a selectional relationship between the verb and ‘how’, whereas evidence is robust that the verb selects the locative and the unaccusative subject. An even more severe problem arises if we nonetheless assume that ‘how’ can be merged lowest in VP so that an in-situ unaccusative subject can c-command it: on such an analysis, there is no accounting for the fact that ‘how’ cannot agree with the direct object in a transitive clause. We will accordingly not pursue this possibility further here.

5.4.5 Summary and Remarks

We have demonstrated in section 5 that agreeing ‘how’ is a whmanner adjunct bearing uφ that probes the c-command domain of ‘how’ independently of T. The ‘how’ agreement patterns in RALI and DALI constructions of both Lubukusu varieties are accounted for by this proposal. We will argue in section 7 that the next derivational step in these constructions is to merge the locative clitic that always surfaces on the verb in LIs. In our analysis, the clitic heads a projection that Diercks (2011b) dubs AgrLP. It probes for and raises DPloc and is thus crucial to allowing movement of the locative over the subject DP.

(66)

graphic

There is more to be said about the properties of AgrL and its role in the two LI constructions. In sections 7 and 8, we address this and several complex locality questions raised by LI with ‘how’. But first, we turn to implications of agreeing ‘how’ for agreement theory.

6 Luyia ‘How’ and Agreement Theory

6.1 Introduction

In this section, we consider some agreement-theoretic issues connected with agreeing ‘how’. We will argue that the ‘how’ facts contradict the claims of Baker (2008) and Diercks (2011b) that agreement in Bantu languages probes upward as a matter of parametric choice. We will show that the ‘how’ facts are also inconsistent with the Spec-head agreement hypothesis. Moreover, we will argue that they are incompatible with Chomsky’s (2007, 2008) and Richards’s (2007) claim that probe features are only introduced on phase heads. Finally, we will argue that proposals by Carstens (2010b, 2011) account for all the agreement facts.

6.2 Against an Upward Agree Account

Baker (2008) and Diercks (2011b) argue that the directionality of Agree is parameterized, searching upward in most Bantu languages rather than probing the c-command domain. Major difficulties arise for an upward Agree account in RALI ‘how’ questions, however, where ‘how’ agrees with the in-situ unaccusative subject (see (51a), repeated here).

(51)

  • a.

    graphic

    ‘How did a tree fall in the forest?’ (Lit.: In the forest fell a tree how?)

For upward Agree to work in a case like (51a), ‘how’ would have to be lower than the unaccusative subject. But independent evidence regarding the position of the thematic subject in LI reported by Diercks (2011b:710–714) argues that it remains in situ within the VP, as discussed in section 4.3.2. And as we noted in section 5.4.4, if ‘how’ can be positioned lower in the clause than the in-situ unaccusative subject, it becomes difficult to explain why ‘how’ cannot agree upward with the direct object of a transitive clause. If we were to assume, contrary to the evidence, that the unaccusative subject raises to some intermediate position, it is not obvious where this position would be, why the subject would move, why a direct object could not also move there, and how upward agreement with the raised DPloc would be avoided. A downward Agree account avoids all of these problems.27

6.3 Against a Spec-Head Agreement Account

An alternative account of agreeing ‘how’ might suppose that ‘how’ heads a functional projection in the clause and the subject passes through its Spec (see Wasike 2007 for a proposal along these lines). As we will show here, however, word order in ‘how’ constructions is very difficult to reconcile with analysis of ‘how’ as a head.

Consider (67), an illustration of the hypothesis that ‘how’ agrees with an expression that transits through a Spec,HowP. (67) is based on (1f), repeated here.

(1)

  • f.

    Si-tanda si-funikhe si-rie(na)?

    7-bed 7SA-broke 7-how

    ‘How did the bed break?’

(67)

graphic

We now present three word order arguments against this approach.

First, Lubukusu is a strictly head-initial language, and the position of ‘how’ is always postverbal. Raising of the subject DP in (67) goes a little way toward deriving the surface word order, but there is no obvious means of moving the verb to the left of ‘how’. If ‘how’ is a head, it should block verb raising, leaving the verb to its right.28

Second, recall that in transitive clauses the direct object often precedes ‘how’ (see (68)). This is inconsistent with the analysis of ‘how’ as a head, given Lubukusu’s strict left-headedness. As (69) illustrates, the HowP analysis of agreeing ‘how’ predicts that not only the verb but also the direct object will appear to the right of ‘how’, contrary to fact.

(68)

  • Ba-ba-ana ba-kha-kule bi-tabu ba-rie(na)?

  • 2-2-children 2SA-FUT-buy 8-books 2-how

  • ‘How will the children buy books?’

(69) *[TP the children [T′ FUT [HowPt [How′ how [vPt buy books]]]]]

There is one possible means of deriving surface word order under an analysis of ‘how’ as head of HowP: pied-piping the entire vP containing the verb and its object to Spec,HowP, as in (70).

(70) [TP the children [T′ FUT [HowP[vPt buy books] [How′ how tvP]]]]

This brings us to our third word order argument against analyzing ‘how’ as a head. Recall that like all other wh-phrases, ‘how’ has an alternative immediately postverbal position (see section 2.3 and the contrast between non-wh (72a) and the object question in (72b)). This is difficult to reconcile with a pied-piping/‘how’-as-head account of the consistently postverbal position of ‘how’. If we adopt the assumption that ‘how’ is a head, raising the goal that values its uφ to Spec,HowP, the only way of deriving the alternative word order would seem to be positing an optional movement of the direct object out of the vP before the vP raises (see (71)). A different derivation would be needed to account for a postverbal wh-phrase like ‘what books’ in (72b), which is not plausibly a head. To attribute a systematic alternation that all wh-phrases participate in to such disparate sources without independent motivation seems unwarranted.29

(71)

  • [TP the children [T′ FUT [HowP[vPtSU buy tOB] [How′ how tvP]]]] [DP books]

(72)

  • a.

    Ba-khasi ba-we-le ba-ba-ana bi-tabu.

    2-women 2SA-give-PST 2-2-children 8-books

    ‘The women gave the children books.’

  • b.

    Ba-khasi ba-we-le [bi-tabu si] ba-ba-ana?

    2-women 2SA-give-PST 8-books what 2-2-children

    ‘What books did the women give the children?’

Thus, analyzing ‘how’ as a head leads to highly unsatisfactory results, defeating the last remaining alternative analysis (17e). We conclude that downward probing is the best tool for analyzing agreeing ‘how’. This being the case, it cannot be that all agreement in natural language is Spec-head agreement (contra Koopman 2000, 2006), or (as argued above) that Lubukusu is parametrically committed to upward Agree (contra Baker 2008, Diercks 2011b).

6.4 Against Phase Heads as the Only Sources of uφ

6.4.1 General Problems

Properties of agreement in Bantu languages raise difficulties for the feature inheritance approach to agreement proposed by Chomsky (2007, 2008) and Richards (2007). Richards writes, ‘‘It thus follows from the SMT [Strong Minimalist Thesis] that uninterpretable (unvalued) features can only be a property of phase heads, that is, those heads that trigger Spell-Out/Transfer’’ (Richards 2007:567). Phase heads hand down probe features to the heads of their complements because ‘‘value and transfer of uFs must happen together’’; otherwise, there would be no way to distinguish valued uFs from interpretable features at the conceptual-intentional interface (Richards 2007:566).

It is well-known that clauses in many Bantu languages can include multiple SA, and that in the same clause C can agree with a wh-phrase (see (73), from Wasike 2007:342). The feature inheritance approach requires that phasal and nonphasal heads interleave, and as a consequence we should see alternations of agreeing and nonagreeing heads. In studies of many Bantu languages, the predicted interleaving is not attested.

(73)

  • Siina ni-syo a-kha-be ne-a-khola?

  • 7what COMP-7 3SG.SA-FUT-be NE-3SG.SA-do

  • ‘What will she/he be doing?’

Building on the analysis of Carstens (2001, 2005), (73) can be represented schematically as follows (lines indicate agreement relationships between an aspectual affix and the subject, between T and the subject, and between C and the operator):

(74)

graphic

There is no independent evidence that Lubukusu has phase heads that are absent in English, intervening between C and T and between T and Asp. Given problems of this kind, Carstens (2010b) argues that Chomsky’s and Richards’s conclusions must be abandoned. Epstein, Kitahara, and Seely (2010) provide compelling arguments from English against the Chomsky/Richards approach (among them the presence of valued uCase at the phase edge on whom in Whom do they like? and valued uφ on v* in They like him). Epstein, Kitahara, and Seely propose that uFs are recognizable as such throughout the derivation and simply ignored at the conceptual-intentional interface. As long as valuation precedes Transfer to PF so that uFs can be pronounced, they are licit on any head. This proposal has much to recommend it, and we adopt it here.

6.4.2 Agreeing ‘How’ and the Origin of Probe Features

The facts of agreeing ‘how’ reinforce these general conclusions. Since ‘how’ is adjoined to vP, it cannot undergo transfer with the VP complement of v; it is part of the higher CP phase. Thus, under the Chomsky/Richards view that only phase heads introduce probe features, it would follow that uφ of T, Asp, and ‘how’ all originate on C. The mismatches between SA on T/Asp and agreement on ‘how’ can only be accommodated under feature inheritance theory if C’s features can be duplicated, distributed to the various agreement bearers including ‘how’, and allowed to agree with different expressions (see (75)).

(75)

graphic

For the feature inheritance approach to work in Lusaamia and Lubukusu, therefore, it seems there can be no constraints placed on the number of heads over which copies of C’s uφ can be distributed (since additional agreeing aspectuals could multiply the items sharing copies of C’s uφ bundle to 3 or 4).30 There would also need to be no constraint prohibiting the features of copies from diverging: once C’s uφ is distributed, each copy would have to be entirely independent in terms of ultimate feature values. There can accordingly be no way to verify whether or not a given instance of agreement has any relation to C, since it can appear on any expression in the phase that lacks iφ, and agree with whatever DP is local to it. Thus, we seem to arrive at a system that cannot be falsified by any empirical facts of agreement. We consider this a fatal flaw and put the approach aside.

6.5 An Agreement Theory That Works for Agreeing ‘How’

The tools to explain why Bantu languages, but not English, should allow an expression like ‘how’ to agree are available in Chomsky 2001, particularly in the ‘‘activity’’ requirement, which we interpret as in (76) from Carstens 2011:721 (based on discussion in Chomsky 2001:6).

(76)

  • The activity requirement

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

While the sole uF activating goals in English A-relations is uCase, Carstens (2010b, 2011) proposes that the grammatical gender of nouns (a component of noun class; see footnote 1) is a meaningless formal feature, hence uninterpretable (see also Zamparelli 2008, Bošković, to appear); as such, it makes its bearer ‘‘active’’ as a goal for Agree. This property is clearly demonstrated in the familiar phenomenon of concord within noun phrases. It is a well-established crosslinguistic pattern that if nouns have grammatical gender (henceforth uGen), many modifiers may agree with them (see (77), from Carstens 2011:728).31

(77)

  • a.

    graphic

    ‘the green house’

  • b.

    graphic

    ‘my small shoe’

What concord also makes clear is that when a noun’s uGen is goal in an Agree relation, no ‘‘deactivation’’ effect occurs. In other words, the same noun can value concord on a number of items inside the DP, as shown in (77). In this respect, concord contrasts sharply with English SA. On the basis of facts like (78), Chomsky (2001:6) writes, ‘‘Once the Case value is determined, N no longer enters into agreement relations and is ‘frozen in place.’’’

(78) *He seems ______ has left.

To account for this contrast between uCase and uGen as active goal features, Carstens (2010b, 2011) proposes that deactivation accompanies valuation of a DP’s uCase because further Agree relations have the potential to tamper illicitly with the value determined in the initial Agree, leading to a PF crash based on unclarity regarding how uCase is to be pronounced. This problem does not arise for iterating Agree where N’s uGen is the active goal feature, because N’s uGen enters the syntax with a value rather than acquiring it via Agree (see Pesetsky and Torrego 2007 on the logical independence of interpretability and value, and Bošković, to appear, for similar ideas on gender in Serbo-Croatian, though somewhat different in implementation).

To return to the issue of the abundant clause-level agreement in Bantu languages: Carstens points out that in all cases, it shares with concord the inclusion of grammatical gender features. This in turn follows, in Carstens’s proposals, from the fact that nouns in Bantu languages raise and adjoin to D, making N’s uGen uniformly accessible to clause-level probes like T and C. Because of N-to-D adjunction, Bantu languages contrast with Romance languages, which also have uGen; its effects as an activity feature in Romance are seen only inside the noun phrase because Romance lacks N-to-D adjunction (compare (79) with (80)–(81)).32 The sole exception to this pattern of Romance agreement is gender agreement on participles, which, significantly, are lexically insensitive to D’s person feature. Carstens argues that [person] intervenes, blocking access to N’s gender for any probe outside the DP with sensitivity to [person], absent N-to-D adjunction.33 Example (80), taken from Carstens 2011:733, represents the analysis of Bantu DPs first presented in Carstens 1991.

(79)

  • a.

    graphic

    ‘my house’

  • b.

    [DP la [FP mia casa [NPtN]]]

(80)

  • a.

    graphic

    ‘my nice house’

  • b.

    [DP nyumba + D [FP yangu tF [NP nzuri [NPtN]]]]

(81)

  • Bantu N-to-D adjunction yields inclusion of gender in SA when uφ of T probes DP

  • graphic

The proposal that nominal gender is never deactivated explains a broad range of constructions in Bantu languages where on standard assumptions a DP should cease to agree and move, but in fact does not.34 Carstens refers to these phenomena as hyperagreement and hyperactivity. Diercks (2012) adds the proposal that some or perhaps all Bantu languages lack uCase altogether, so uGen is the sole activity feature.

With uGen providing inexhaustible ‘‘activity,’’ any DP in a Bantu language with the relevant properties (noun class and N-to-D adjunction) can serve as goal in an infinite series of Agree relations. Thus, a subject can value agreement on ‘how’, on an aspectual, on T, and if extracted, on C. Similarly, fronted DPloc can value the features of the locative clitic and move on to a second Agree relation with T and/or C, as we will show in section 7.

Summing up, a principled explanation for differing crosslinguistic patterns of agreement is not available in Chomsky’s (2007, 2008) framework, but Chomsky’s (2001) approach yields a nicely predictive system once the roles of grammatical gender and N-to-D adjunction are taken into account. Under this approach to agreement, the cooccurrence of ‘how’ agreement and SA with a single expression is not problematic at all.

6.6 On the Status of ‘How’ and Probing by uF of XP

We assume that as a vP adjunct, ‘how’ is an XP—perhaps an adverbial phrase like many manner adjuncts (very quickly, forcefully, etc.).35 The question arises whether this is inconsistent with the proposal that the c-command domain of its uφ is vP.

There is certainly a recent tradition of assuming that uFs must obtain values from within the c-command domain of the head that bears them, and that if this does not happen, the derivation cannot continue. Chomsky (2000:132) writes, ‘‘Properties of the probe . . . must be satisfied before new elements of the lexical subarray are accessed to drive further operations.’’ A logical corollary of this statement is that only (features of ) heads can probe.

But there is also quite a bit of evidence that this view is not on the right track. Consider the status of uCase on a DP. Since uCase is an unvalued, uninterpretable feature, it meets the definition of a probe in Chomsky’s system. It should therefore probe its c-command domain and, if a match is not available there, the above quotation from Chomsky 2000 leads to the expectation that the derivation should crash, yet it does not. According to Carstens (2012), this fact gives rise to a stipulation that uCase has the status of a ‘‘goal’’ feature, but in reality there is no principled basis for the probe-goal distinction among uFs.

Epstein and Seely (2006), Bošković (2007, 2011), and Carstens (2012) accordingly argue that uCase features are in fact probes. Epstein and Seely (2006) and Bošković (2007, 2011) argue that A-movement exists so that a DP can c-command a source of Case valuation. Carstens (2012) adds that the defining property of ‘‘goal’’ features is that they are probes that find nothing relevant in their c-command domains at first Merge. (We illustrate this for uCase of D in (82) and (83). Following Bošković (2011) and Carstens (2012), an abstract Case ‘‘assigner’’ has uninterpretable but valued Case features; thus, T has uNom.)36

(82)

graphic

(83)

graphic

See also Bobaljik and Wurmbrand 2005 for important evidence that nominative valuation works this way in some but not all syntactic contexts in German, based on several options for obtaining what Carstens (2012) calls delayed valuation.

Carstens (2012) argues that any uF of a head X that cannot obtain a value in the complement of X is inherited by XP, because the features of X form XP’s label. uF of XP can therefore probe XP’s c-command domain and, so long as uF is valued before phasal transfer, the derivation can converge. Carstens argues that this is how uφ of adjectives within APs is valued by N.

(84)

  • a.

    graphic

    ‘the very pretty girl’

  • b.

    graphic

    ‘a faster car’

(85)

  • a.

    graphic

  • b.

    graphic

Returning to the analysis of agreeing ‘how’, we conclude that it makes up the sole content of an adjunct XP ( just as English how is generally taken to be the sole content of an adjunct XP). uφ of ‘how’ is valued by probing the c-command domain of this XP, hence probing vP.

7 Inversion, Locality, and Agreeing ‘How’

7.1 Introduction

In the preceding sections, we have analyzed agreeing ‘how’ as a vP adjunct with uφ. We have shown that this proposal nicely accounts for agreement phenomena associated with several kinds of noncanonical subjects including expletives, fronted locatives, and operators. Several potential locality problems associated with LI constructions are readily explained by assuming two Merge options for selected locatives in unaccusative constructions (see discussion of (58) and (59)) and by adopting Merge-over-Move (see Chomsky 1995, 2001 and discussion of (62)). Some additional locality puzzles arise, however, in connection with agreeing ‘how’ and LI.

First, it is interesting to note that the thematic subject does not systematically block raising of DPloc: this is a classic conundrum of inversion constructions often addressed in terms of material in VP being equidistant from outside probes (see Chomsky 1995, Ura 1996, Collins 1997). But the fact that agreement on ‘how’ is subject to strict locality rules out equidistance of arguments in VP as an explanatory mechanism. We will argue that DPloc is probed and raised to Spec of the locative clitic that appears on the verb whenever a locative is fronted. Because the clitic is sensitive only to locatives, the subject is irrelevant to its search.

This proposal explains why the subject does not block raising of DPloc, but it leads to a second locality question: if DPloc transits through an intermediate position, we need to explain why it does not block raising of the subject to Spec,TP in the DALI construction. We relate this to Ā-opacity effects in a range of other languages.

7.2 Why the Subject Does Not Block Raising of the Locative

Recall our proposal that the expression valuing the uφ of ‘how’ does so because it is the highest DP in vP and hence most local to ‘how’. When ‘how’ and T agree with different expressions, a locality puzzle therefore arises: surely the expression closest to ‘how’ is closest to T as well (see (60a), repeated here). Why then, when ‘how’ agrees with the logical subject in a RALI construction like (51a) (repeated here), is it possible for T to probe and raise DPloc?

(60)

graphic

(51)

  • a.

    graphic

    ‘How did a tree fall in the forest?’ (Lit.: In the forest fell a tree how?)

A likely explanation for this phenomenon lies in the special locative clitic that always and only agrees with locatives when left-dislocated (86b), raised to Spec,CP (86d), or occupying Spec,TP (86c); it never agrees with in-situ locatives (86a) (see Diercks 2011a,b).

(86)

  • a.

    graphic

    ‘A tree fell in the forest.’

  • b.

    graphic

    ‘In the forest, a tree fell.’

  • c.

    graphic

    ‘In the forest fell a tree.’

  • d.

    graphic

    ‘In the forest fell a tree.’

We adopt Diercks’s (2011b) proposal that this clitic heads a syntactic projection in the middle field of the clause,37 which he dubs AgrLP. The clitic probes for and agrees with DPloc. Because its uφ is sensitive only to locatives, intervening nonlocative DPs are irrelevant to its search (see Rizzi 1990 and Preminger 2011 on effects of this kind, and discussion in section 7.3.4). We depart from Diercks’s account in proposing that the clitic has an edge feature38 that raises DPloc to Spec,AgrLP, where it is closest to uφ of T.39 Note that DPloc is not deactivated or ‘‘frozen in place’’ after entering into this Agree relation with the locative clitic; it remains active to agree with T for the reasons sketched in section 6.5.

(87) RALI, variety A: ‘How’ agrees with the subject; T and AgrL agree with DPloc

  • a.

    graphic

  • b.

    graphic

  • c.

    graphic

But now a different locality question arises: if fronted locatives move to Spec,AgrLP, how can T ever agree with the thematic subject in an LI construction, as is the case in DALI?40

7.3 Why the Locative in Spec,AgrLP Does Not Block Raising of the Subject

7.3.1 The Problem

In the preceding section, we explained the absence of an intervention effect for T probing a locative phrase that begins lower in the structure than the subject of the unaccusative RALI construction: we argued that the special locative clitic heads a projection midway between T and vP, and that it probes and raises all and only extracted or inverted locatives to an intermediate position between T and the thematic subject. Once DPloc is in that position, T can probe and raise it to Spec,TP, yielding the construction we call RALI.

(88)

  • a.

    T [AgrLP DPloc AgrL [vP v [VP SU . . . DPloc]]] →

  • b.

    [TP DPloc T [AgrLPDPloc AgrL [vP v [VP SU . . . ]]]]

Under these assumptions, both unaccusative and unergative DALI constructions raise a new locality puzzle. Recall that in DALI constructions, T and ‘how’ agree with the thematic subject, but the locative clitic agrees with the inverted locative phrase. The pattern is exemplified for an unaccusative and an unergative in (38a–b), repeated here.

(38)

  • a.

    graphic

    ‘In the forest fell a tree.’

  • b.

    graphic

    ‘Into the hole jumped the rabbit.’

Recall also that DALI constructions have the ( partial) structures in (41a–b), repeated here, where the landing site of DPloc is Spec,CP and that of the logical subject is Spec,TP.

(41) Disjoint agreement LI (DALI): Thematic subject to Spec,TP; DPloc to Spec,CP

graphic

Updating these representations to reflect the hypothesized intermediate positions of the operator DPloc including Spec,AgrLP yields (89) for unergative and (90) for unaccusative DALI constructions (in which head movement of V to v to C is omitted to conserve space). Notice that the surface position of SU is separated from its base position by the intermediate occurrences of DPloc.41

(89) OK [CP DPloc C [TP SU T [AgrLPDPloc AgrL [vPDPloc [SU v [VP V DPloc]]]]]]

(90) OK [CP DPloc C [TP SU T [AgrLPDPloc AgrL [vP v [VPSU V DPloc]]]]]

In a DALI ‘how’ question, then, an intriguing locality puzzle arises in connection with SA and subject raising to Spec,TP: why doesn’t DPloc block Agree (T, SU)?

This issue may seem at first glance to arise as a consequence of our proposal that the locative clitic probes and raises DPloc to the intermediate Spec,AgrLP. But in fact the problem exists independently of that. Let us consider in closer detail the intermediate stage in the derivation of an unergative DALI construction before AgrL is even merged, as depicted in (91) (head movements again omitted). Assuming that phasal v raises DPloc to its Spec en route to Spec,CP, DPloc constitutes an expression with iφ features closer to any external probe than the subject. Even if AgrLP were not present, the question would arise as to how and why T ignores DPloc.

(91)

graphic

Even at this point in the derivation, then, before AgrL is involved, strict locality might lead us to expect DPloc to intervene and prevent T from raising or agreeing with the logical subject.

The problem described here is not by any means a novel one restricted to the analysis of Lubukusu locative constructions. (91) is no different in principle from the configuration that arises in wh-questioning of an English nonsubject. We illustrate with a case of object extraction in (92) (transferred material is shaded); see in particular the intermediate stage (92b). The thematic subject can (and in fact must) be raised to Spec,TP across an intervening operator bound for Spec,CP in a wh-question of this kind. Explaining such phenomena in a way consistent with other evidence of strict locality in movement and agreement is a recurring issue in Minimalist syntactic theory.

(92)

graphic

7.3.2 Toward a solution

It has often been argued in relation to this issue that a chain including both A- and Ā-positions is ‘‘improper’’ (for helpful recent discussion and review of the literature on this question, see Obata and Epstein 2011; see also Svenonius 2000, Rezac 2003, Chomsky 2008, among many others). We express this idea in (93) with a constraint that relates the invisibility of the locative operators in (89)–(91) to the object operator’s nonintervention in (92) when T is probing—a state of affairs that Rezac (2003) aptly refers to as Ā-opacity.42 We discuss some approaches to deriving (93) in section 7.4.43

(93)

  • Prohibition on Mixed Chains

  • Expressions in Ā-positions are inaccessible to A-relations.

Our formulation in (93) is intended to avoid problems of lookahead since it refers to Ā-positions only. We follow Chomsky (2008) in defining an Ā-position as a position created by the edge feature of a phase head, and we assume that there are phasal and nonphasal versions of AgrL. The phasal AgrL feeds Ā-movement of DPloc to Spec,CP in DALI constructions because it permits T to probe the subject.44 Spec of the nonphasal AgrL is not Ā-opaque (T cannot probe over it); hence, nonphasal AgrL can only feed RALI constructions.

(93) is also formulated to be compatible with participation of operators in A-relations prior to extraction,45 which is important given that they can value SA in a question like Who likes coffee? or acquire a Case value as in Whom did you see?.

Lubukusu LI constructions provide some novel evidence of the visibility of in-situ operators for A-relations. Recall from section 5.4.3 that T cannot probe the subject when DPloc is merged higher than the subject. The relevant configuration is the unaccusative VP in (60b), proposed because for speaker 3, ‘how’ could agree either with the subject or with DPloc in RALI constructions. To account for this speaker’s pattern of judgments, we proposed that in his dialect, unaccusative VPs have the structural ambiguity shown in (60a–b), repeated here.

(60)

graphic

Now notice that in the hypothetical derivation of an unacceptable DALI continuation from (60b), DPloc would surface in Spec,CP; hence, it would be an operator in an Ā-position (see (94)).

(94)

  • Unacceptable unaccusative DALI construction based on (60b)

  • *[CP DPloc C [TP SU T [AgrLPDPloc AgrL [vP v [VPDPloc V SU]]]]]

The contrast between the impossible derivation for DALI in (94) and the licit ones in (89) and (90) (repeated here) argue that there is no A-opacity effect for the DPloc operator in its base position. Only its intermediate occurrences are successfully ignored by T when T probes.

(89) OK [CP DPloc C [TP SU T [AgrLPDPloc AgrL [vPDPloc [SU v [VP V DPloc]]]]]]

(90) OK [CP DPloc C [TP SU T [AgrLPDPloc AgrL [vP v [VPSU V DPloc]]]]]

It has been noted elsewhere that operators in their landing sites can be invisible to A-probing, like their intermediate occurrences are here. Consider (95), from Svenonius 2000:260. Svenonius provides a number of arguments that the clause-medial negative expression in the Icelandic (95) is an Ā-operator. It is therefore significant that the subject can move to Spec,TP across it, indicating that the relationship (T, SU) succeeds though the null hypothesis is for the negative expression’s _-features to be relevant to T’s search.46

(95)

  • graphic

  • ‘The boys had thrown no rocks at the cars.’

Together, the facts of (89), (90), (94), and (95) confirm that while an operator in its base position is visible in A-relations, once it has moved this visibility ceases. Thus, the formulation in (93) (repeated here) predicts all the facts.

(93)

  • Prohibition on Mixed Chains

  • Expressions in Ā-positions are inaccessible to A-relations.

When T probes in (94), it finds an occurrence of DPloc in its Merge position, higher than the subject. This occurrence is not in an Ā-position. The desired result is obtained: (60b) cannot feed a DALI construction because T cannot probe the subject across DPloc in the VP of (60b). In all cases, however, T will ignore intermediate occurrences of DPloc in an Ā-position. The one instance where T does not ignore an intermediate occurrence of DPloc is in RALI constructions, where T is able to probe DPloc instead of the thematic subject precisely because DPloc has moved to Spec of nonphasal AgrL (as discussed in section 7.2). Thus, nonphasal AgrL feeds RALI because its Spec is not an Ā-position and therefore the raised DPloc can be probed by T, but phasal AgrL only feeds DALI, as a DPloc raised to Spec,AgrL in that case is subject to Ā-opacity.

As pointed out in footnote 43, the Prohibition on Mixed Chains in (93) also allows us to explain why ‘how’ and AgrL are not sensitive to A/Ā differences in their goal. AgrL and ‘how’ probe expressions in their base positions, where Ā-opacity is not a factor.

7.3.3 Summary and Remarks

We have argued that apparent locality paradoxes in LI constructions have two primary sources. First, A-movement of DPloc across the thematic subject is mediated by the locative clitic. The subject is irrelevant to its search because the clitic is sensitive only to locative material. We argued that the clitic has both uφ and an edge feature (we argued in section 6 that uφ probe features are not properties of phase heads alone, contra Chomsky 2008).

The second apparent locality paradox we considered was the ability of T to probe the subject across the intervening DPloc in DALI constructions. We argued that this is because in DALI, DPloc is an operator and hence the positions it moves through are Ā-positions; A-probing typically ignores the contents of Ā-positions. The structure in (91) showed that the initial locality issue in unergative DALI constructions is essentially the same as those of object operator constructions more generally in that both of the potential goals for Agree relations are at the edge of the vP phase, as shown in (92). But if we are correct about the role of AgrL in locative inversions, then DALI constructions in Lubukusu provide some novel evidence that A-probing can reach over intervening operators even across a significant structural divide. (96) represents DALI with an unaccusative verb (such as in (38a), repeated here), where the subject is internal to VP, and T nonetheless successfully probes and raises it across the locative in Spec,AgrL.

(38)

  • a.

    graphic

    ‘In the forest fell a tree.’

(96)

graphic

DALI constructions in Lubukusu thus suggest that Ā-opacity is not attributable to local or parochial properties such as the hypothesis of equidistance among multiple Specs of a single head (Ura 1994, 1996).

The Prohibition on Mixed Chains in (93) also provides insight into why unergative verbs cannot participate in RALI constructions. Since ergative vP is phasal, DPloc cannot escape it without first moving to an outer Spec,vP. Given that this is an Ā-position, T will never be able to probe DPloc there. Hence, unergatives are restricted to DALI constructions.

7.4 Approaching Ā-Opacity

Locality paradoxes like (89)–(92), and hence Ā-opacity effects more generally, are independent of the analysis of agreeing ‘how’, and so in principle a variety of approaches to them might well be compatible with the analysis of ‘how’ presented in this article. It is worth noting, however, that the Luyia phenomena present some special challenges.

Two prominent analyses of Ā-opacity propose that it arises because the φ-features of an operator (henceforth OP and ; recall from discussion of (16) that i = intrinsic) cease to be visible before the point where T probes. Rezac (2003) argues that an Agree relation encapsulates the goal in a KP shell of functional structure. Once features have been encapsulated, nothing further can probe them. Obata and Epstein (2011) argue that Ā-opacity arises for an English object OP because of ‘‘feature splitting’’: in Agree (v, OB), uCase and iφ features move to v, leaving OP with just the Q-feature relevant in Ā-relations (see (92a), repeated here, and (97)).

(92) a.

graphic

(97) Obata and Epstein's (2011) feature splitting

  • a.

    Little v probes the object who

    graphic

  • b.

    At the phase edge, English who no longer has oφ uCase

    [vP whoQ [vP SU v [VPVwho]]]

Luyia raises novel problems for both of these approaches because it exhibits Ā-opacity, but its operators control agreement. Hence, iφ of OPs must be syntactically visible. This is illustrated in (99), the schematic representation of (98), where a locative OP first values uφ of the locative clitic and subsequently uφ of the complementizer in a cleft (subscripted numbers are noun classes).47 T is able to probe the subject across the operator nonetheless, as (99) demonstrates.

(98)

  • graphic

  • ‘It was in the house that the children broke the stick.’

(99)

  • In (98), OP’s φ-features are active to value agreement on AgrL and C, but T successfully probes and raises the subject nonetheless

  • graphic

(100) demonstrates the same sort of phenomenon for an object operator: it values agreement on C of the cleft. This shows that OP’s iφ is visible and active. Yet Ā-opacity obtains nonetheless: T agrees with the thematic subject, raising it to Spec and adopting its φ-values in SA (see (101)).

(100)

  • graphic

  • ‘It was the stick that the children broke.’

(101)

  • In (100), the object OP retains itsφ -features to value agreement on C, but T successfully probes and raises the subject nonetheless

  • graphic

We conclude that Ā-opacity cannot be attributed to inaccessibility of OP’s iφ, at least in Luyia; assuming that a unitary approach is desirable, the solution must be sought elsewhere.

It seems to us that a promising factor in approaching opacity effects is the selectivity of probe features that this investigation has uncovered. Compare ‘how’ and AgrL, both of which probe with uφ. As we have shown, uφ of AgrL ignores iφ of a nonlocative DP, whereas ‘how’ must agree with the closest DP to it. The fact that T can ignore operators in Ā-positions seems to us to align it with AgrL: the conclusion suggested is that T’s uφ probe is not a ‘‘pure’’ one like that of ‘how’, but a more selective one like that of AgrL.

In some Inverse Case Filter analyses, T has a probe feature that can be thought of as uNom linked to its uφ (see (102)). This seems to us one possible angle on Ā-opacity.48

(102)

  • T ignores OP, looking for unvalued uCase

  • graphic

Something along these lines seems to have potential to explain the selectivity with which T probes, ignoring the intervening operator’s occurrences in Ā-positions.49 Much about the role of Case in Bantu languages is unclear, however, and we are unable to do the topic justice here for reasons of space (see footnote 26 on the restriction of Luyia LI to intransitives, and on the other hand Diercks 2012 and references cited therein for arguments that Case plays no role in many Bantu languages). Adapting an idea from Rizzi and Shlonsky 2007, an alternative might be to suppose that T’s probing involves a feature connected, at least historically, with discourse or information packaging (topic/comment; theme/rheme structure), and incompatible with the focus properties of operators encoded in their Q-feature. Bundled with T’s uφ, the relevant feature (labeled uD in (103) for its possible discourse origins) causes T to ignore the operator-in-transit when seeking a goal.50

(103)

  • uD of T leads it to ignore OP because of its incompatible Q specification

  • graphic

Both (102) and (103) lead to questions outside this article’s scope about the impossibility of T probing/raising the subject across a locative merged higher than the subject in the VP (see discussion of (60b) in sections 5.4.3 and 7.3.3).

We leave it to future research to find the best explanation for Ā-opacity. We conclude only that it is unexpected for OPs with active iφ on the analyses discussed, and that sensitivities of T’s probe features seem a promising direction to look for insight.51

8 Summary and Conclusions

In this article, we have described the properties of agreeing ‘how’ in the Luyia language Lubukusu. We have shown that it is neither a floating modifier nor a wh-subject depictive. It questions kinds of events and hence is nonreferential and [−N]. In Bantu languages, expressions lacking intrinsic φ-features generally acquire them via Agree, and this is true of Luyia ‘how’. Relying on evidence from expletive subjects, subject questions, and locative inversion constructions, we have argued that agreeing ‘how’ is merged as a vP adjunct. It bears uφ that probes the subject in its base position independently of T’s uφ. This proposal accounts for the fact that ‘how’ and T can agree with different expressions in inversion constructions, or in different features with a subject operator.

This conclusion has important consequences for various aspects of syntactic architecture. First, it argues that downward probing is a better tool for analyzing agreement in Lubukusu than are upward agreement and Spec-head agreement, neither of which can adequately account for the facts of agreeing ‘how’. Second, it poses a strong challenge to the feature inheritance model, which claims that all probes originate on phase heads. Like many other facts in Bantu languages, the evidence of ‘how’ suggests that probe features are licit on any expression, providing they obtain values before Transfer to PF for Spell-Out.

This article also makes an original contribution to the study of inversion phenomena, and locative inversion in particular. Looking at ‘how’ questions in LI sentences, we showed that the two expressions in an LI construction are not actually equidistant from probes outside of VP (see footnote 40). A pure uφ probe like ‘how’ identifies the highest DP in the VP. For the lower DP to invert, undergoing A-movement to Spec,TP, a special strategy is required to get it across the higher one; in Lubukusu, this strategy comes in the form of raising to Spec,AgrL.

The interaction of ‘how’ and inversion provides some novel evidence on the topic of Ā-opacity since it shows that locative (and other) operators in Lubukusu have and retain φ-features throughout the derivation but are nonetheless opaque to probing by T. We have made a tentative suggestion relating this to the contrasting behavior of ‘how’ and AgrL as uφ probes, arguing that even probes sensitive to the same basic sorts of features can differ in selecting different subsets of them (and perhaps selecting them in combination with other features), with consequences in terms of what constitutes an intervening expression in a relation otherwise constrained by closest c-command.

Agreeing ‘how’ is one among many systematic cases that have been reported in the Bantu syntax literature where a single DP participates in multiple Agree relations: alongside multiple agreement in compound tense constructions and operator constructions, syntactic theory must recognize that a DP can value agreement on ‘how’ and the locative clitic. The phenomenon is well-established (see, e.g., Carstens 2001, Henderson 2007) and is related to very wide-reaching claims about the nature of DP-licensing in human language (see Carstens 2010b, 2011, Diercks 2012, and the summary in section 6.5).

Our analysis also adds to the many existing arguments for a low base position for clausal subjects (see Kratzer 1996 and many others).

Finally, the position and agreement properties of ‘how’ provide an interesting and compelling argument that height and ‘‘leftness’’ do not correlate, contra Kayne 1994 and Cinque 2005. In particular, the facts of agreement on ‘how’ strongly argue that it is merged higher than vP material to its left. This being the case, it cannot be true that syntactic hierarchy maps consistently and universally into left-to-right linear order.

Notes

The basic Lubukusu ‘how’ facts were first described in Wasike’s (2007) dissertation. Thanks to Dennis Odalloh for Lusaamia data and to Justine Sikuku, Lillian Jivetti, and Aggrey Wanyonyi for Lubukusu data. For helpful discussion and comments on this material, our thanks to Daniel Seely, Juvsénal Ndayiragije, Meredith Landman, Justine Sikuku, Brent Henderson, members of our audience at the 4th International Conference on Bantu Languages (2011), and two anonymous LI reviewers.

1 We use X agrees with Y to mean that intrinsic features of Y are reflected on X—not the converse.

In glosses, SA = subject agreement, and cardinal numerals (1–3) denote person features when they are accompanied by a number specification (SG = singular, PL = plural); thus, 2SG.SA = second person singular subject agreement. Arabic numerals 1–17 indicate noun classes; hence, 2SA = subject agreement for noun class 2. We gloss agreement with a singular human as 3SG.SA but agreement with plural humans as 2SA ( = class 2 subject agreement) because there is evidence for person features in the former but not the latter (see, e.g., Bokamba 1976, Kinyalolo 1991, Henderson 2009a,b, to appear, Diercks 2010). This will become relevant in sections 4.2 and 5.2. Other glossing conventions are AAE = alternative agreement effect, APPL = applicative, CA = complementizer agreement, COMP = complementizer, DEM = demonstrative, FEM = feminine, FUT = future, FV = final vowel (identified only when it follows verbal suffixes like the applicative), LOC = locative, OA = object agreement, PASS = passive, PRES = present, PST = past.

We follow Carstens 1991 in analyzing noun class as number and gender. (1a) ny- vs. en- and (1c) y- vs. a- are phonologically conditioned variants. Tone marking is omitted for lack of a guiding analysis or confidence in our transcriptions.

We use the Greek character φ to represent person, number, and gender features. For expository convenience, we treat uφ of category X as singular, and see section 6.5 for an analysis of agreement in Bantu and Indo-European languages as the product of one undifferentiated uφ probe per agreeing category (not per feature).

Lubukusu’s classification by Guthrie (1948) as E31 was revised to J.E31c by Maho (2008) and J.30 by Lewis (2009).

2Wasike (2007) reports that agreeing ‘how’ occurs in a variety of interlacustrine Bantu languages, including Runyoro (Uganda), Luganda (Uganda), and Kinyarwanda (Rwanda). He notes that Taylor (1985) discusses a similar pattern for Nkore-Kiga (Uganda). We are unaware of other agreeing adjuncts apart from these except apparently an agreeing ‘thus’ in some Luyia varieties; see footnote 5.

3 See Buell 2009 and Cheng and Downing 2012 for exploration of an alternation in Zulu that looks quite similar. Their analyses show that a Zulu position immediately after the verb (IAV) is associated with true questions and focus. Some preliminary work with native speakers of Lusaamia revealed that an echoic reading was strongly favored for nonsubject wh-elements not occupying the position immediately after the verb, and a true question interpretation was strongly favored for nonsubject wh-elements in the position immediately after the verb as reported for Zulu by Cheng and Downing. However, this intuition faded under continued questioning for reasons not clear to us. More work with additional speakers is therefore needed on nonsubject wh-phrases and focused material in Luyia languages to determine whether the position after the verb has the same properties as in Zulu. Because this is not a question about ‘how’ in particular but about nonsubject questions and foci in general, we leave it to future research. See Cheng and Downing 2012 for arguments that wh-expressions in the Zulu IAV position are in situ and post-IAV wh-expressions are rightdislocated. On the other hand, see Van der Wal 2009 for a movement analysis of a similar-looking position in Makhuwa.

4 This sentence is taken from Wasike 2007:360. Wasike claims that the right-edge position of ‘how’ makes its scope unambiguous. In contrast, the three speakers whom we have consulted consistently find this and comparable examples to be compatible with both matrix and embedded construals.

5 Toward the end of our research on agreeing ‘how’, we learned from David Odden and Michael Marlo ( pers. comm.) that some varieties of Luyia have an agreeing manner adverb meaning ‘thus’, strengthening the resemblance between Luyia languages and the languages discussed by Landman and Morzycki (2003). Space considerations prevent an investigation of agreeing ‘thus’ in this article.

6 Thanks to an anonymous reviewer for pointing this out.

7 The wh-phrase ‘what kind’ can be part of a DP that clefts. C of the cleft agrees with i_ of the whole DP.

8 As noted in section 2.2, only the long forms of wh-phrases can appear in clefts. Before noninterrogative clefted expressions an agreeing form of ‘be’ is required, but before clefted wh-phrases it is optional. We include it consistently to facilitate comparisons among the interrogative and noninterrogative examples.

9 We pin the Luyia problem on agreement because PPs and adverbials can cleft in languages where clefts do not involve agreement: It was [in the morning] that Sue found the letter.

10 This list mainly comprises suggestions made by colleagues, audience members, and anonymous reviewers. We would credit each individual for his or her idea if recollection served us well enough. But given the vagaries of our memories, we hope a general thank-you will suffice.

11 We adopt the assumption that verbs in Bantu languages generally raise into the middle field of the clause. See Julien 2002, Carstens 2005, and Ngonyani 2006, among others, for proposals as to its precise location in various languages. In section 4.3.2, we present a proposal from Diercks 2011b that Lubukusu V raises first to T and then to C in a particular variety of locative inversion construction.

12 Though in general we expect a scope difference to correlate with differences in agreement as in (20b) versus (21a), it’s not clear that this extends to cases where the matrix subject is an expletive. One speaker suggested that class 15 agreement (21a–b) would be most appropriate if a lot of people might want to call. Juvénal Ndayiragije ( pers. comm.) makes the plausible suggestion that this would be consistent with an arbitrary PRO reading for the subject of the infinitive. We leave analysis of class 15 ‘how’ agreement for future research. See Carstens 1991 for arguments that class 15 includes true infinitives as well as counterparts to the so-called acc-ing and poss-ing gerunds of English, possibilities that would take us far afield to explore in relation to ‘how’.

13 Thanks to Juvénal Ndayiragije ( pers. comm.) for suggesting this possibility to us.

14 An anonymous reviewer asks whether our analysis in combination with (23) entails that the logical object raised in a passive is the highest DP within vP. The answer is ‘‘maybe.’’ Since the uφ of ‘how’ cannot be valued by the DP contained within a by-phrase, perhaps the latter adjoins above ‘how’: [vP[vP[vP Vpass DP] ‘how’] by-phrase]. Alternatively, however, it could simply be that by-phrases are opaque and noninterveners for agreement for reasons to be determined. We leave this to future research.

15 Lubukusu is a ‘‘high’’ applicative (APPL) language in Pylkkänen’s terms, allowing APPL on intransitive verbs and lacking the rigid transfer-of-possession semantics that she argues holds in ‘‘low’’ applicative languages like English. Contrary to the predictions of her analysis, Lubukusu does not allow depictive secondary predication of indirect objects (IOs) apart from clausal-looking ones like ‘(while he was) drunk/tired’, a class of expressions that can refer to English IOs too (I gave John a book while he was drunk vs. *I gave John a book naked—ruled out on construal of naked with John). For this reason, IO evidence is not useful in relation to our investigation of ‘how’. In contrast, a question with -rie inside the IO is fine, such as ‘What kind of children did you buy food for?’.

16 In (24), -chula ‘naked’ has invariant class 7 agreement morphology when used adverbially. Embisi ‘raw’, in contrast, agrees with the class 9 subject. One speaker we consulted accepted (25). Some depictives involve more internal structure: for example, ne-a-mel-ile ‘(while he was) drunk’ from (22e). These differences do not affect the analysis of ‘how’.

17 Justine Sikuku ( pers. comm.) finds object-oriented answers more natural than instrument-type answers, because a more specific strategy exists for questioning an instrument: namely, adding an applicative morpheme to the verb and using ‘what’ to formulate a question like ‘What did they cut the paper with?’.

18 We have found a class of exceptions to this generalization, in which the verb is ‘want’ or a perception verb like ‘see’ and a time expression intervenes licitly between the apparent direct object and ‘how’ agreeing in the object’s features. We suggest that this is because ‘want’ and the like can take a small clause complement headed by a zero copula. This is consistent with the general availability of a zero copula in Lubukusu, and with the standard properties of ‘want’-type verbs crosslinguistically.

(i)

  • W-enya ka-ma-ki asubuhi ka-rie?

  • 2SG.SA-want 6-6-eggs morning 6-how

  • ‘How do you want your eggs in the morning?’ (i.e., fried or scrambled)

  • Our analysis: You want [your eggs in the morning (to be) how]

19 Thanks to Dan Seely for suggesting that we explore this topic.

20 The leveling of first and second person distinctions under subject extraction is the clearest evidence for this analysis of AAE; see data presented in Kinyalolo 1991, Diercks 2009, 2010, Henderson 2009a,b, to appear.

21Bresnan and Mchombo (1995) demonstrate that the locative morphemes pa, ku, and mu in Chicheŵa are independent words; hence, it is possible to conjoin them as shown in (i), from Bresnan and Mchombo 1995:206. This makes the Chicheŵ a two-mora size requirement relevant to them under Bresnan and Mchombo’s analysis of them as nouns. But assuming with Carstens (1997) that the locative morphemes are functional categories, whether or not their cognates in Lubukusu are morphologically dependent does not affect the analytical possibilities.

(i)

graphic

18 and 16 6basket

‘in and on baskets’

22 [✓] and [*] mark where the adverb can and cannot occur, respectively.

23 Since ‘tree’ and ‘how’ are string-adjacent and ‘how’ agrees with ‘tree’, (51) and (52) can also mean ‘What kind of tree fell in the forest?’. We will ignore such systematically available but irrelevant readings.

24 Not depicted here is agreement of C with an operator, which results in what appears to be doubling of the SA morpheme as shown in (42) and (44) (see Diercks 2010, 2011b on the analysis of this as C-agreement). The features of this agreement are also restricted to gender and number; person is precluded.

25 Thanks to an anonymous reviewer for comments clarifying this line of argumentation.

26 This analysis sets aside questions of how to constrain (i) LI to selected locatives, (ii) RALI to unaccusatives, and (iii) DALI to intransitives. We present a movement-theoretic account of (ii) in section 7.3.4. See also Belletti 1988 and Alexiadou and Anagnostopoulou 2001 for Case-theoretic ideas relevant to (ii)–(iii), and Diercks 2011b for an alternative approach without Case. Details are beyond the scope of this article.

27 Frequent pairing of an edge feature with agreement explains its apparent upward orientation in many constructions; see Collins 2004, Carstens 2005. An anonymous reviewer makes the interesting suggestion that adjuncts and functional heads might differ systematically in that although both can have uφ, the former lack edge features. This seems a promising idea but we leave pursuit of it to future research.

28 An additional linear order problem for treating ‘how’ as a head in the middle field of the clause arises in relation to syntactic structure associated with the clitic in LI constructions, because like the verb it must precede ‘how’ in linear order (see our preview of this analysis in section 5.4.5, and details in section 7).

29 On the basis of careful comparison, Buell (2011) argues that Zulu ‘why’ in the position immediately after the verb (IAV) merits a special account different from that of other IAV wh- and focused phrases. Accordingly, he argues that the IAV position of wh-phrases in Zulu may not be syntactic in origin; rather, it may result from some late-level Spell-Out effect. An anonymous reviewer suggests that if this is the case, the argument here is weakened. Given the oftdescribed correlation of true question/focus interpretation and IAV position, we are skeptical; but even if Buell’s conclusion is borne out, we think the other arguments given in section 6.3 suffice to rule the Spec-head approach out. Footnote 28 points out yet another word order problem that would arise upon consideration of the locative clitic’s position (see section 7), though we will not discuss this here for reasons of space.

30 Once the agreeing locative clitic is taken into account, this problem is exacerbated; see section 7.

31 We assume that concord is simply agreement under closest c-command with the features available in the noun phrase (see Carstens 2000, 2010b, 2011 and Baker 2008 on the relationship of concord to Minimalist theory).

32Cinque (2005) rejects N-raising and head movement generally. But see Carstens 2010a for an N-raising analysis of DP-internal word order variation; Matushansky 2006 for a narrow-syntax approach to head movement that does not violate the Extension Condition; and Roberts 2010 for arguments that the Extension Condition need not hold and that narrow-syntax head movement in the traditional sense does exist. While Bošković (2008) argues that many languages without articles lack DP projections, the syntactic correlates to this in the languages he considers are not replicated in the Bantu languages that Carstens has studied (including Lubukusu and Lusaamia), suggesting that although these Bantu languages lack articles they nonetheless have DPs. See Carstens 2010a on this question.

33 Carstens points out that number agreement outside DP is also unexpected on the common assumption that number heads a functional projection in the DP’s middle field (see, e.g., Carstens 1991, Ritter 1992). She argues that in a featural version of Quantifier Raising (QR), number always raises to D to take scope over DP; hence, interpretable number features are generally accessible to value uφ of clause-level probes.

34 A few additional technicalities connected with constraining iterating agreement lie outside the scope of this brief sketch; see Carstens 2010b, 2011 for details.

35 Cinque (1999, 2005) argues that adverbs and adjectives are heads, and when an adverb is to the right of the VP or an adjective is to the right of NP, the word order indicates raising of VP/NP to a Spec higher than the modifier. See Carstens 2011 for arguments against this approach to adjectives. We will not pursue the matter further here for reasons of space.

36 The copy theory of movement yields two copies of DP bearing uF, which apparently inherit the value acquired by the higher member of the chain.

37 In a compound tense construction (not illustrated here for reasons of space), the locative clitic attaches to the highest auxiliary. We therefore locate the projection that the clitic heads as sister to T in (87c).

38 Alternatively, we might suppose that even unaccusative v has an edge feature (see Legate 2003) raising DPloc to an outer Spec,vP from whence DPloc can either A-move to Spec,TP or Ā-move to Spec,CP. Then the obligatory presence of the locative clitic would be viewed as coincidental.

39 Probing by AgrL feeds both A- and Ā-movement of DPloc. We address this in section 7.3.2 (and see footnote 44).

40 As noted in sections 5.4.3 and 7.1, a number of previous analyses of LI assume that it exists in part because a locative and an unaccusative subject in VP are ‘‘equidistant’’ from T (see Collins 1997, Culicover and Levine 2001, Rezac 2006, Diercks 2011b; see Ura 1994, 1996 and Chomsky 1995 on equidistance). Since for many speakers ‘how’ agrees only with the subject, this analysis will not work for Lubukusu; instead, such data motivate our analysis based on properties of AgrL. Our approach may assist in the analysis of inversion phenomena generally (see Zeller 2012 for similar ideas on other inversion constructions): while a head like AgrL is overtly realized in Lubukusu as the locative clitic, it might be phonologically null but syntactically active in other languages with similar LI properties.

41 It is conceivable that AgrL’s edge feature is used only in A-movement, so Ā-movement does not pass through its Spec. This seems stipulative (and inverts Minimalist approaches to movement in a curious way). We will not pursue the idea here partly for reasons of space and partly, anticipating the discussion of (91) and (92), because avoiding Ā- movement through Spec,AgrL seems not to alter the outcome substantially.

42 As Rezac (2003) and Obata and Epstein (2011) note, not all languages exhibitA¯ -opacity. Under Carstens’s (2005) analysis, Kilega is a Bantu language that does not. We note a potential conceptual connection between (93) and the alternative agreement effect (AAE) described for subject operators in sections 4.2 and 5.2. But AAE does not correlate with Ā-opacity: it is found in Kilega and Luyia both. Further consideration lies outside the scope of this article.

43 Note that Agree (‘how’, DP) probes DP in situ, as does Agree (AgrL, DPloc). Hence, Ā-opacity is not expected in these cases.

44 This definition of an Ā-position entails adopting Chomsky’s (2001:14) version of the Phase Impenetrability Condition (see (i), a slight restatement of Chomsky’s (11)). Otherwise, in DALI constructions like (38)/(89)–(90), phasal AgrL will trigger cyclic transfer of vP before T is merged, hence removing the in-situ subject before T can probe it. An alternative is to avoid the assumption that AgrL is ever a phase head, defining an Ā-position as any position that an expression with a Q ( = operator) feature moves to. This is compatible with Chomsky’s (2000:108) version of the Phase Impenetrability Condition in (ii). Thanks to an anonymous reviewer for bringing this issue to our attention.

(i) Given strong phase HP contained within strong phase ZP, the domain of H is inaccessible to operations at ZP; only H and its edge are accessible to such operations.

(ii) In 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.

45 Under the feature inheritance model of Chomsky (2008), the entire chain is visible at the crucial point in (92b) because probing waits until C is merged and gives uF to T. We have already detailed reasons in section 6.4 for rejecting, for Luyia, Chomsky’s proposal to introduce all probe features on phase heads. We conclude that for Luyia at least, the requirement of uniform chains in (93) cannot be derived in the way proposed in Chomsky 2008.

46 See Jayaseelan 2001 for discussion of several similar Ā-opacity problems.

47 In an attempt to explain why Bantu languages that appear to be monoclausal [OP Agr-C [SU SA-T . . . ]] have agreement with OPs but Ā-opacity effects nonetheless, Carstens (2005) argued that wh-constructions in such languages must in reality be biclausal clefts formed with featureless null OPs raised to Spec,CP. Lubukusu’s ability to have clausemedial agreement of ‘how’ and the locative clitic with a locative OP seems to rule this approach out.

48 See Bošković 1997, 2002, Martin 1999, and Duguine 2010 for arguments in favor of the Inverse Case Filter. See Bošković 2007, among other works, for arguments against it.

49 An anonymous reviewer suggests that our highly feature-specific approach to locality might be incompatible with defective intervention effects. In fact, well-known cases of defective intervention are quite feature-sensitive: German and Icelandic T seems able to value a VP-internal nominative across an intervening dative, though it cannot agree across the dative (see, e.g., Holmberg and Hróarsdóttir 2003).

50 An anonymous reviewer points out a possible connection between this and Miyagawa’s (2010) proposal that the head of the phrase that houses subjects (αP for Miyagawa) may inherit topic/focus features from C.

51 We noted in footnote 42 that Kilega lacks Ā-opacity; its operators value uφ of T and raise through Spec,TP (and the same is true for any AspPs present; see Carstens 2005, Obata and Epstein 2011). Kilega T thus seems a good candidate for a pure uφ probe, like ‘how’. Obata and Epstein propose that feature splitting works differently in Kilega than in English, leaving iφ features of operators active. But the problematic prediction remains that agreement with OP and absence of Ā-opacity should go hand in hand. The contrast between Kilega and Luyia is perhaps the strongest argument against concluding that operators simply do not ever need to stop in an outer Spec,vP, in which case the expectation of Ā-opacity effects at this derivational point would not arise.

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
.
Baker
,
Mark
.
2008
.
The syntax of agreement and concord
.
Cambridge
:
Cambridge University Press
.
Belletti
,
Adriana
.
1988
.
The Case of unaccusatives
.
Linguistic Inquiry
19
:
1
34
.
Bobaljik
,
Jonathan David
, and
Susi
Wurmbrand
.
2005
.
The domain of agreement
.
Natural Language and Linguistic Theory
23
:
809
865
.
Bokamba
,
Eyamba
.
1976
.
Question formation in some Bantu languages
.
Doctoral dissertation, Indiana University, Bloomington
.
Bošković
,
Željko
.
1997
.
The syntax of nonfinite complementation
.
Cambridge, MA
:
MIT Press
.
Bošković
,
Željko
.
2002
.
A-movement and the EPP
.
Syntax
5
:
167
218
.
Bošković
,
Željko
.
2007
.
On the locality and motivation of Move and Agree: An even more minimal theory
.
Linguistic Inquiry
38
:
589
644
.
Bošković
,
Željko
.
2008
.
What will you have, DP or NP?
In
NELS 37
, ed. by
Emily
Elfner
and
Martin
Walkow
,
1:101
114
.
Amherst
:
University of Massachusetts, Graduate Linguistic Student Association
.
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
.
Bošković
,
Željko
.
To appear
.
On valued uninterpretable features
. In
NELS 39
, ed. by
Suzi
Lima
,
Kevin
Mullin
, and
Brian
Smith
.
Amherst
:
University of Massachusetts, Graduate Linguistic Student Association
.
Bowers
,
John
.
2002
.
Transitivity
.
Linguistic Inquiry
33
:
183
224
.
Bresnan
,
Joan
, and
Sam
Mchombo
.
1995
.
The Lexical Integrity Principle: Evidence from Bantu
.
Natural Language and Linguistic Theory
13
:
181
254
.
Buell
,
Leston
.
2009
.
Evaluating the immediate postverbal position as a focus position in Zulu
. In
Selected proceedings of the 38th Annual Conference on African Linguistics
, ed. by
Masangu
Matondo
,
Fiona
Mc Laughlin
, and
Eric
Potsdam
,
166
172
.
Somerville, MA
:
Cascadilla Press
.
Buell
,
Leston
.
2011
.
Zulu ngani ‘why’: Postverbal and yet in CP
.
Lingua
121
:
805
821
.
Carstens
,
Vicki
.
1991
.
The syntax and morphology of Determiner Phrases in Kiswahili
.
Doctoral dissertation, UCLA, Los Angeles, CA
.
Carstens
,
Vicki
.
1997
.
Empty nouns in Bantu locatives
.
The Linguistic Review
14
:
361
410
.
Carstens
,
Vicki
.
2000
.
Concord in Minimalist theory
.
Linguistic Inquiry
31
:
319
355
.
Carstens
,
Vicki
.
2001
.
Multiple agreement and Case deletion
.
Syntax
4
:
147
163
.
Carstens
,
Vicki
.
2005
.
Agree and EPP in Bantu
.
Natural Language and Linguistic Theory
23
:
219
279
.
Carstens
,
Vicki
.
2010a
.
Head-movement in Bantu DPs
.
Paper presented at the 41st annual meeting of the North East Linguistic Society; manuscript in progress
.
Carstens
,
Vicki
.
2010b
.
Implications of grammatical gender for the theory of uninterpretable features
. In
Exploring crash-proof grammars
, ed. by
Michael
Putnam
,
31
57
.
Amsterdam
:
John Benjamins
.
Carstens
,
Vicki
.
2011
.
Hyperactivity and hyperagreement in Bantu
.
Lingua
121
:
721
741
.
Carstens
,
Vicki
.
2012
.
Delayed valuation
.
Ms., University of Missouri, Columbia
.
Cheng
,
Lisa Lai-Shen
, and
Laura J.
Downing
.
2012
.
Against Focus P: Arguments from Zulu
. In
Contrasts and positions in Information Structure
, ed. by
Ivona
Kuáerová
and
Ad
Neeleman
,
247
266
.
Cambridge
:
Cambridge University Press
.
Chomsky
,
Noam
.
1981
.
Lectures on government and binding
.
Dordrecht
:
Foris
.
Chomsky
,
Noam
.
1995
.
The Minimalist Program
.
Cambridge, MA
:
MIT Press
.
Chomsky
,
Noam
.
2000
.
Minimalist inquiries: The framework
. In
Step by step: Essays on Minimalist syntax 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
, ed. by
Robert
Freidin
,
Carlos P.
Otero
, and
Maria Luisa
Zubizarreta
,
133
166
.
Cambridge, MA
:
MIT Press
.
Cinque
,
Guglielmo
.
1994
.
On the evidence for partial N-movement in the Romance DP
. In
Paths towards Universal Grammar: Studies in honor of Richard S. Kayne
, ed. by
Guglielmo
Cinque
,
Jan
Koster
,
Jean-Yves
Pollock
,
Luigi
Rizzi
, and
Raffaella
Zanuttini
,
85
110
.
Washington, DC
:
Georgetown University Press
.
Cinque
,
Guglielmo
.
1999
.
Adverbs and functional heads
.
New York
:
Oxford University Press
.
Cinque
,
Guglielmo
.
2005
.
Deriving Greenberg’s Universal 20 and its exceptions
.
Linguistic Inquiry
36
:
315
332
.
Collins
,
Chris
.
1997
.
Local economy
.
Cambridge, MA
:
MIT Press
.
Collins
,
Chris
.
2004
.
The agreement parameter
. In
Triggers
, ed. by
Anne
Breitbarth
and
Henk
van Riemsdijk
,
115
136
.
Berlin
:
Mouton de Gruyter
.
Culicover
,
Peter
, and
Robert
Levine
.
2001
.
Stylistic inversion in English: A reconsideration
.
Natural Language and Linguistic Theory
19
:
283
310
.
Diercks
,
Michael
.
2009
.
Subject extraction and (so-called) anti-agreement effects in Bukusu: A criterial freezing approach
.
Paper presented at the 45th annual meeting of the Chicago Linguistic Society. University of Chicago, 16–18 April
.
Diercks
,
Michael
.
2010
.
Agreement with subjects in Lubukusu
.
Doctoral dissertation, Georgetown University, Washington, DC
.
Diercks
,
Michael
.
2011a
.
Incorporating location in argument structure: The Lubukusu locative clitic
. In
Selected proceedings of the 40th Annual Conference on African Linguistics: African languages and linguistics today
, ed. by
Eyamba G.
Bokamba
,
Ryan K.
Shosted
, and
Bezza Tesfaw
Ayalew
,
65
79
.
Somerville, MA
:
Cascadilla Proceedings Project
.
Diercks
,
Michael
.
2011b
.
The morphosyntax of Lubukusu locative inversion and the parameterization of Agree
.
Lingua
121
:
702
720
.
Diercks
,
Michael
.
2012
.
Parameterizing Case: Evidence from Bantu
.
Syntax
15
:
253
286
.
Duguine
,
Maia
.
2010
.
The Inverse Case Filter and the nature of Case
.
Paper presented at the 5th Brussels Conference on Generative Linguistics
.
Epstein
,
Samuel David
,
Hisatsugu
Kitahara
, and
T. Daniel
Seely
.
2010
.
Uninterpretable features: What are they, and what do they do?
In
Exploring crash-proof grammars
, ed. by
Michael
Putnam
,
125
142
.
Amsterdam
:
John Benjamins
.
Epstein
,
Samuel David
, and
T. Daniel
Seely
.
2006
.
Derivations in Minimalism
.
Cambridge
:
Cambridge University Press
.
Geuder
,
Wilhelm
.
2000
.
Oriented adverbs: Issues in the lexical semantics of event adverbs
.
Doctoral dissertation, Universität Tübingen
.
Guthrie
,
Malcolm
.
1948
.
The classification of the Bantu languages
.
London
:
Oxford University Press
.
Henderson
,
Brent
.
2007
.
Multiple agreement and inversion in Bantu
.
Syntax
9
:
275
289
.
Henderson
,
Brent
.
2009a
.
Anti-agreement and [person] in Bantu
. In
Selected proceedings of the 38th Annual Conference on African Linguistics: Linguistic theory and African language documentation
, ed. by
Masangu
Matondo
,
Fiona
Mc Laughlin
, and
Eric
Potsdam
,
173
181
.
Somerville, MA
:
Cascadilla Press
.
Henderson
,
Brent
.
2009b
.
Agreement and person in anti-agreement
. In
CLS 45. Vol. 2, Locality in language
, ed. by
Ryan
Bochnak
,
Nassira
Nicola
,
Peet
Klecha
,
Jasmin
Urban
,
Alice
Lemieux
, and
Christina
Weaver
,
89
102
.
Chicago
:
University of Chicago, Chicago Linguistic Society
.
Henderson
,
Brent
.
To appear
.
Agreement and person in anti-agreement
.
Natural Language and Linguistic Theory
.
Holmberg
,
Anders
, and
Þorbjörg
Hróarsdóttir
.
2003
.
Agreement and movement in Icelandic raising constructions
.
Lingua
113
:
997
1019
.
Irimia
,
Monica-Alexandrina
.
2005
.
Types of secondary predication
. In
Niagara Linguistics Society 2005 conference proceedings
, ed. by
Kevin
Heffernan
and
Eugenia
Suh
,
20
29
.
Toronto Working Papers in Linguistics 25
. .
Jayaseelan
,
K. A
.
2001
.
IP-internal topic and focus phrases
.
Studia Linguistica
55
:
39
75
.
Julien
,
Marit
.
2002
.
Syntactic heads and word formation
.
Oxford
:
Oxford University Press
.
Kayne
,
Richard
.
1994
.
The antisymmetry of syntax
.
Cambridge, MA
:
MIT Press
.
Kinyalolo
,
Kasangati K. W
.
1991
.
Syntactic dependencies and the Spec-head agreement hypothesis in Kilega
.
Doctoral dissertation, UCLA, Los Angeles, CA
.
Koopman
,
Hilda
.
2000
.
The syntax of specifiers and heads
.
London
:
Routledge
.
Koopman
,
Hilda
.
2006
.
Agreement configurations: In defense of ‘‘Spec-head.’’
In
Agreement systems
, ed. by
Cedric
Boeckx
,
159
199
.
Amsterdam
:
John Benjamins
.
Kratzer
,
Angelika
.
1996
.
Severing the external argument from its verb
. In
Phrase structure and the lexicon
, ed. by
Johan
Rooryck
and
Laurie
Zaring
,
109
137
.
Dordrecht
:
Kluwer
.
Landman
,
Meredith
, and
Marcin
Morzycki
.
2003
.
Event-kinds and the representation of manner
. In
Proceedings of the 28th Western Conference on Linguistics
, ed. by
Nancy Mae
Antrim
,
Grant
Goodall
,
Martha
Schulte-Nafeh
, and
Vida
Samiian
,
1
12
.
Fresno
:
California State University
.
Legate
,
Julie Anne
.
2003
.
Some interface properties of the phase
.
Linguistic Inquiry
34
:
506
516
.
Lewis
,
M. Paul
, ed.
2009
.
Ethnologue: Languages of the world
. 16th ed.
Dallas, TX
:
SIL International
. .
Maho
,
Jouni
.
2008
.
NUGL online: The web version of the New Updated Guthrie List, referential classification of the Bantu languages
.
Version dated 25 March 2008. Available at http://goto.glocalnet.net/maho/bantusurvey.html
.
Martin
,
Roger
.
1999
.
Case, the Extended Projection Principle, and Minimalism
. In
Working Minimalism
, ed. by
Samuel David
Epstein
and
Norbert
Hornstein
,
1
25
.
Cambridge, MA
:
MIT Press
.
Matushansky
,
Ora
.
2006
.
Head movement in linguistic theory
.
Linguistic Inquiry
37
:
69
109
.
Miyagawa
,
Shigeru
.
2010
.
Why Agree? Why Move? Unifying agreement-based and discourse-configurational languages
.
Cambridge, MA
:
MIT Press
.
Mutonyi
,
Nasiombe
.
2000
.
Aspects of Bukusu morphology and phonology
.
Doctoral dissertation, The Ohio State University, Columbus
.
Myers
,
Scott
.
1987
.
Tone and the structure of words in Shona
.
Doctoral dissertation, University of Massachusetts, Amherst
.
Ngonyani
,
Deo
.
2006
.
Attract F and verbal morphology in Kiswahili
.
The Linguistic Review
23
:
37
68
.
Obata
,
Miki
, and
Samuel David
Epstein
.
2011
.
Feature-splitting internal Merge: Improper movement, intervention, and the A/A′ distinction
.
Syntax
14
:
122
147
.
Pesetsky
,
David
, and
Esther
Torrego
.
2007
.
The syntax of valuation
. In
Phrasal and clausal architecture
, ed. by
Simin
Karimi
,
Vida
Samiian
, and
Wendy K.
Wilkins
,
262
294
.
Amsterdam
:
John Benjamins
.
Preminger
,
Omer
.
2011
.
Agreement as a fallible operation
.
Doctoral dissertation, MIT, Cambridge, MA
.
Pylkkänen
,
Liina
.
2002
.
Introducing arguments
.
Doctoral dissertation, MIT, Cambridge, MA
.
Pylkkänen
,
Liina
.
2008
.
Introducing arguments
.
Cambridge, MA
:
MIT Press
.
Radford
,
Andrew
.
2009
.
Analyzing English sentences: A Minimalist approach
.
Cambridge
:
Cambridge University Press
.
Rezac
,
Milan
.
2003
.
The fine structure of cyclic Agree
.
Syntax
6
:
156
182
.
Rezac
,
Milan
.
2006
.
The interaction of Th/Ex and locative inversion
.
Linguistic Inquiry
37
:
685
697
.
Richards
,
Marc
.
2007
.
On feature inheritance: An argument from the Phase Impenetrability Condition
.
Linguistic Inquiry
38
:
563
572
.
Ritter
,
Elizabeth
.
1992
.
Cross-linguistic evidence for number phrase
.
Canadian Journal of Linguistics
37
:
197
218
.
Rizzi
,
Luigi
.
1990
.
Relativized Minimality
.
Cambridge, MA
:
MIT Press
.
Rizzi
,
Luigi
, and
Ur
Shlonsky
.
2007
.
Strategies for subject extraction
. In
Interfaces + recursion = language?
, ed. by
Uli
Sauerland
and
Häns-Martin
Gärtner
,
115
160
.
Berlin
:
Mouton de Gruyter
.
Roberts
,
Ian
.
2010
.
Agreement and head movement: Clitics, incorporation, and defective goals
.
Cambridge, MA
:
MIT Press
.
Sportiche
,
Dominique
.
1988
.
A theory of floating quantifiers and its corollaries for constituent structure
.
Linguistic Inquiry
19
:
425
451
.
Svenonius
,
Peter
.
2000
.
Quantifier movement in Icelandic
. In
The derivation of VO and OV
, ed. by
Peter
Svenonius
,
255
292
.
Amsterdam
:
John Benjamins
.
Taylor
,
Charles
.
1985
.
Nkore-Kiga
.
London
:
Croom Helm
.
Ura
,
Hiroyuki
.
1994
.
Varieties of raising and the feature-based theory of movement. MIT Occasional Papers in Linguistics 7
.
Cambridge, MA
:
MIT, MIT Working Papers in Linguistics
.
Ura
,
Hiroyuki
.
1996
.
Multiple feature-checking: Atheory of grammatical function splitting
.
Doctoral dissertation, MIT, Cambridge, MA
.
Wal
,
Guenever Johanna van der
.
2009
.
Word order and information structure in Makhuwa-Enahara
.
Doctoral dissertation, Leiden University
.
Wasike
,
Aggrey
.
2007
.
The left periphery, wh-in-situ and A-bar movement in Lubukusu and other Bantu languages
.
Doctoral dissertation, Cornell University, Ithaca, NY
.
Zamparelli
,
Roberto
.
2008
.
On the interpretability of φ-features
. In
The Bantu-Romance connection
, ed. by
Ce´cile
De Cat
and
Katherine
Demuth
,
167
199
.
Amsterdam
:
John Benjamins
.
Zeller
,
Jochen
.
2012
.
Instrument inversion in Zulu
. In
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
.