Abstract

This article investigates the syntax of the phrase-final focus particles kuphela and qha ‘only’ in Zulu and Xhosa (Nguni; Bantu). We show that kuphela’s and qha’s associations with a focused constituent respect the complex topography of information structure in Nguni and, like English only, a surface c-command requirement. However, unlike English only, the Zulu and Xhosa particles typically follow the focus associate they c-command, a fact that poses a serious challenge for Kayne’s (1994) antisymmetry theory. We demonstrate that the Nguni facts are incompatible with recent Linear Correspondence Axiom–inspired approaches to phrase-final particles in other languages and, after weighing the merits of several approaches, we conclude that kuphela is an adjunct and that syntax is only weakly antisymmetric: adjuncts are not subject to the LCA.

1 Introduction

1.1 Overview

In this article, we explore the syntax of the exclusive focus markers kuphela and qha ‘only’ in the Nguni languages Zulu and Xhosa.1Kuphela is used in both languages; qha is specific to Xhosa.

As the examples in (1) and (2) show, kuphela typically follows the focused constituent construed as its so-called associate; the same is true of qha (focus associates boldfaced). While there is a strong preference for these particles to associate with an adjacent phrase, this is not absolute, as (3) and (4) demonstrate.2

(1)

graphic

(2)

graphic

(3)

graphic

(4)

graphic

(1) shows kuphela and qha associating with focused DPs immediately to their left. (2) illustrates that kuphela can also be used adverbially, associating with an adjacent focused verb or VP; the same is true of qha (we have picked a VP idiom in (2b) to rule out a reading with object focus, which would otherwise be more prominent). For most speakers, the preferred interpretations of (3) are to associate kuphela with the adjacent direct object (3a), with the VP (3d), or even with the whole sentence (3e); however, focus association with a nonadjacent constituent, such as the indirect object (3b) or the verb (3c), is also possible for some speakers (on the lack of total agreement regarding (3a) and (3b) and the unacceptability of (3f), see section 4). (4) illustrates the same kinds of options for adverbial kuphela or qha in an impersonal passive construction with “expletive” class 17 subject agreement on the verb.

Languages where expressions meaning ‘only’ have been well-studied exhibit a requirement that such expressions c-command their associates (see, e.g., Aoun and Li 1993, Büring and Hartmann 2001, Erlewine 2014a,b, Tancredi 1990a). The fact that associates of kuphela and qha precede them therefore raises interesting theoretical issues connected with the antisymmetry hypothesis of Kayne (1994), which states that high-to-low relations map invariantly into left-to-right linear order, and with the related Final-over-Final Condition of Biberauer, Holmberg, and Roberts (2014) and Sheehan et al. (2017).

In the article, we will present a complex picture of associations for kuphela and qha. Certain positions in Xhosa and Zulu clauses are relatively focus-neutral in that they may but need not contain focused material. We will show that kuphela and qha must c-command the highest copy of an expression in such a position for association to succeed. After considering and rejecting as inadequate several alternative accounts of the facts consistent with antisymmetry in its strongest form, we conclude that syntax is only weakly antisymmetric in the sense of Takano (2003)—that is, at least some adjuncts fall outside of antisymmetry, and kuphela and qha are among these (assuming with Cardinaletti (2011) that some particles have the status of adjuncts).3 Only this conclusion is consistent with both crosslinguistic evidence for antisymmetry and the language-particular evidence that kuphela and qha c-command associates that precede them.

There are also syntactic positions in Xhosa and Zulu clauses that are [+focus], that is, restricted to focused material: clefts and the S of active [VSX] constructions, especially transitive expletive constructions. We found judgments on association at a distance to material in such positions to be quite unpredictable. Among speakers who accept these associations, some were entirely consistent in requiring surface c-command by kuphela or qha of the highest copy of an associate in a [+focus] position. Others judged associations in which kuphela or qha c-commands only a low copy to be marginal or well-formed. Occasionally, speakers even accepted associations to material in [+focus] positions wherein kuphela or qha c-commands no copy at all. Our impression is that the narrow focus reading characteristic of material in a cleft or [VS] construction is a major distractor in evaluating when and where exclusive focus readings are licit within the same utterance (see section 8 for some discussion). Associations to positions that can host either focused or nonfocused material (henceforth, focus-tolerant positions) are our primary concern because judgments are clearer on them. We found that they consistently require surface c-command, therefore providing some novel evidence on right-edge particles and the mapping of hierarchy to word order.

It has previously been recognized that right-edge particles appear to violate both antisymmetry theory and the related Final-over-Final Constraint/Condition (FOFC) of Holmberg (2000), Biberauer, Holmberg, and Roberts (2014), and Sheehan et al. (2017), among others. Our article contributes to a debate over why this is so. FOFC rules out head-final over head-initial configurations like *[XP [YP Y Complement] X] within specific domains with shared categorial features—extended projections in the sense of Grimshaw (1991). Within such domains, Biberauer, Holmberg, and Roberts (2014) propose that head-finality is due to an EPP-like feature ^. This is passed up the tree from head to head, deriving surface head-finality from universal head-complement order in the base.

Biberauer, Holmberg, and Roberts (2014) suggest that many particles are acategorial and hence outside of the domains in which FOFC applies. Therefore, though they are underlyingly initial heads, they may introduce ^ features independently of the heads below them and raise their complements to their specifiers (Specs), leading to the appearance of head-finality.4

Erlewine (2017) and Hsieh and Sybesma (2011) adopt a version of this approach, arguing for an underlyingly head-initial syntax for final particles in Mandarin. They propose that such particles are phase heads, triggering transfer of their complements (see (5a)). When phase interiors transfer, they become unanalyzable syntactic atoms, leading to symmetric, hence unlinearizable structures (5b). They must therefore raise to c-command the phase head in order to break symmetry, leading to surface head-finality, as in (5c).

(5)

graphic

Although kuphela and qha are phrase-final rather than exclusively sentence-final, such an approach might in principle be extended to them. We will show, however, that the proposals are incompatible with the pattern of judgments indicating that kuphela and qha must c-command the highest copy of an associate (with caveats noted above).

The same problem arises in connection with a proposal that Cardinaletti (2011) makes for right-edge modal particles in Italian: that they merge as Specs of functional categories and that the material to their left arrives at its surface position through remnant movement across them. Like the movement analyses of Erlewine (2017) and Hsieh and Sybesma (2011), this approach cannot be extended to word order involving right-edge kuphela and qha because it is incompatible with the surface c-command requirement.

For the sake of concreteness, we adopt as a working hypothesis the view that kuphela and qha adjoin to constituents of various categories. Examples we have introduced so far are consistent with adjunction to DP and vP (see (6a–b)). In section 5, we motivate adjunction to TP (6c). In sections 5, 7, and 8, we argue in detail for the superiority of (6) over other possible approaches.

(6)

  • [DP [DP u-Sipho] kuphela]

    AUG-1a.Sipho only

    ‘only Sipho’

  • [vP [vP hlab-a i-khefu] kuphela]

    stab-FVAUG-5.rest only

    ‘only taking a break’

    [TP [TP . . . ] kuphela]

It is worth noting that Xhosa speakers also consistently approve construals in which clause-medial kuphela precedes its associate, as in (7a) and (7bi). While there is less consistency on this point among Zulu speakers,5 we have found exemplars online, including (7c).6

(7)

graphic
graphic

As noted above, the theoretical goal of our article is to determine whether kuphela and qha must c-command their associates in surface syntax in Xhosa and Zulu. If they need not, then it is possible that the order [kuphela/qha > associate] is the underlying order, with the order [associate > kuphela] derived from it by raising of the associate. If, on the other hand, kuphela and qha must c-command their associates’ highest copy, then the existence of both orders entails that kuphela and qha may adjoin either on the left or on the right. We explore the c-command facts in detail in section 5 and their bearing on the underlying order in section 7.

Because our investigation uncovered no syntactic differences between kuphela and qha, we will use them interchangeably. Most of our examples feature kuphela because it is acceptable in both languages, and they focus on associations to preceding material for the same reason and because such associations constitute the challenge to antisymmetry theory that we wish to explore.

1.2 A Note on Speakers’ Judgments

At the request of an anonymous reviewer, we provide information on acceptance rates for key examples in footnotes. It is important for the figures we report to be interpreted in the context of restrictions relevant to them. The task of assembling evidence on c-command and associations at a distance is complex, for Zulu and Xhosa. Not all speakers accept associations at a distance, and for those who do, there is some variation in judgments across examples. Also, as we show in section 4, focused constituents are barred from many syntactic positions that would otherwise be very desirable to test. For each speaker we set out to interview, we first ascertained whether she or he accepted associations between adverbial kuphela/qha and a distant DP, in general. Then, with speakers who qualified by passing this first test, we checked whether they would accept focused material including adnominal kuphela/qha in a given position of interest, call it P. If they did, we next checked whether these speakers found associations at a distance between material in P and the adverbial kuphela/qha felicitous. Only then could we check for a surface c-command requirement with speakers who had not been excluded by negative judgments on the foregoing matters. Unanticipated difficulties with syntactic positions and long associations to them sent us back to the drawing board many times.

1.3 Structure of the Article

Section 2 provides a little background on the interpretation of focus and expressions meaning ‘only’. Section 3 reviews the relevant notions of antisymmetry theory. Section 4 surveys the topography of focus in Nguni and its relevance to kuphela and qha. Section 5 presents evidence that the two particles must c-command the head of their associate’s chain in surface syntax, if the associate is in a focus-tolerant location. Section 6 details the reasons why we do not reject antisymmetry theory, despite its incompatibility with the syntax of kuphela and qha. Section 7 provides arguments against the approach in Erlewine 2017 and Hsieh and Sybesma 2011. Section 8 discusses associations of kuphela and qha with material in [+focus] positions. Section 9 concludes.

2 ‘Only’ as an Alternative-Sensitive Particle

2.1 Focus and ‘Only’

The semantics of focus is typically analyzed in terms of alternatives that are introduced into the discourse by a focused constituent. For example, in Rooth’s (1985, 1992) influential theory of Alternative Semantics, every node is assumed to have, in addition to its ordinary semantic value, a focus semantic value, which is derived by replacing the ordinary meaning of the focused constituent with contextually plausible alternatives. To illustrate, if the ordinary semantic value of a focused DP like Mary is the individual Mary, then its focus semantic value is the set of individuals who are potential alternatives to Mary.7

(8) [Mary]F:

  • Ordinary semantic value: the individual Mary

  • Focus semantic value: the set of alternative individuals {Mary8, Sue, Bill, . . . }

The focus semantic value of the sentence in (9), which includes the focused DP Mary, is the set of propositions of the form “John likes y”, where y is an element from the set in (8b).

(9) John likes [Mary]F.

  • Ordinary semantic value: the proposition “John likes Mary”

  • Focus semantic value: the set of alternative propositions {“John likes Mary”, “John likes Sue”, “John likes Bill”, . . . }

The Nguni focus markers kuphela and qha are focus-sensitive (or alternative-sensitive) particles comparable to English only. Focus/Alternative-sensitive means that the semantic contribution made by these elements depends on the alternatives introduced by the focus; they associate with the focus (Büring and Hartmann 2001, Erlewine 2014b, Jackendoff 1972, König 1991, Krifka 2006, Rooth 1985, 1992, among many others). Exclusive focus markers like only universally quantify over the alternatives introduced by their focus associate: the sentences in (10), with either adnominal or adverbial only, are true if every y from the set of alternatives in (8b) for which “John likes y” is true is identical to Mary (in other words, (10a) and (10b) are true if “John likes Mary” is true and all other propositions in (9b) are false).

(10)

  • John likes only [Mary]F. (adnominal only)

  • John only likes [Mary]F. (adverbial only)

Note that while adnominal only in (10a) is adjacent to the focus, adverbial only in (10b) can also associate with the focused object, and (10a) and (10b) have the same truth conditions.

2.2 Association at a Distance and the C-Command Requirement

As already illustrated by (10b), the English focus adverb only can associate “at a distance,” giving rise to multiple association options such as those in (11) (see Jackendoff 1972).

(11)

  • John only [gave]F his daughter a new bicycle.

  • John only gave [his]F daughter a new bicycle.

  • John only gave his [daughter]F a new bicycle.

  • John only gave his daughter a [new]F bicycle.

  • John only gave his daughter a new [bicycle]F.

But there is a crucial constraint on what only can associate with: only must c-command its associate. Tancredi (1990a) formulates this requirement as the Principle of Lexical Association (PLA) in (12).

(12)

  • Principle of Lexical Association

  • An operator like only must be associated with a lexical constituent in its c-command domain.

  • (Tancredi 1990a:30)

In English, the c-command requirement holds in surface syntax: lower copies of a moved expression do not suffice to permit that expression to associate with only (Aoun and Li 1993, Erlewine 2014a,b, Tancredi 1990b). (13)–(15) illustrate this. The lower copy in Spec,vP does not permit a subject in Spec,TP to serve as only’s associate in (13). Nor does the copy of an Ā-moved expression in (14a), unlike the in-situ wh-phrase of an echo question (14b) or the unraised infinitival subject in (15a). (≠ indicates an unavailable reading; # marks an infelicitous continuation.)

(13)

graphic

(14)

graphic

(15)

  • There only seems to be a man in the room. ✓There doesn’t seem to be a woman.

  • A man only seems to be in the room. #There doesn’t seem to be a woman.

However, there is evidence of crosslinguistic variation on this important point. As Erlewine (2014b) notes, Barbiers (1995) and Jacobs (1983) report that German and Dutch have expressions meaning ‘only’ that can associate through reconstruction, unlike English only. In (16a) and (16b), the expressions twee boeken ‘two books’ and jedes Buch ‘every book’, which include the focused element, have been topicalized. As a result, the exclusive focus markers no longer c-command the highest copy of their focus associates.

(16)

graphic

Given this point of contrast, we propose that there are weak and strong versions of the PLA as shown in (17).

(17)

  • Strong PLA

    If only associates with α and there are multiple copies of α in the representation, only must c-command the highest copy of α.

    (Erlewine 2014b:115)

  • Weak PLA

    Reconstruction permits association of operators meaning ‘only’; thus, c-command of a copy suffices (German and Dutch).

One of the tasks of this article is to determine whether a version of the PLA holds in Xhosa and Zulu and if so, which: weak or strong? This will be crucial to assessing the compatibility of kuphela and qha with Kayne’s (1994) antisymmetry theory, which we review next.

3 Antisymmetry, ‘Only’, and the Principle of Lexical Association

3.1 Antisymmetry Theory

Kayne (1994) has proposed the Linear Correspondence Axiom (LCA): that is, that hierarchy maps invariantly into linear order. For expository convenience, we adopt the formulation in (18).

(18) Linear Correspondence Axiom A lexical item α precedes a lexical item β iff

Under the LCA, underlying Spec–head–complement order is universal. Apparent deviations from this pattern are taken to be the result of movement (see Kayne 1994 and Cinque 2005 for extensive discussion). We illustrate in (19).

(19)

graphic

3.2 Kuphela/Qha, Antisymmetry, and the Principle of Lexical Association

As noted, kuphela typically follows the associate in Zulu (see (20)–(22)) and this order is very common in Xhosa too, raising LCA-related issues.

(20)

graphic

(21)

graphic

(22)

graphic

In (20), kuphela associates with the indirect object DP of a ditransitive verb. (21) is an expletive construction with VSX word order, and kuphela associates with the focused postverbal subject in this example. In (22), the direct object DP imifino ‘vegetables’ has been right-dislocated (as indicated by the corresponding object marker of class 4 that is attached to the verb9). Dislocation constructions such as (22) can express contrastive verb focus, and in this case kuphela can follow, and be associated with, the focused verb.

If kuphela/qha are subject to the Strong PLA—that is, if they need to c-command their associates in surface syntax—then antisymmetry theory predicts that they must precede their associates. But in examples like (20)–(22), the associate precedes kuphela/qha. Under LCA assumptions, this order entails that the associate asymmetrically c-commands kuphela/qha and the Strong PLA accordingly cannot be met. As a consequence, if there is evidence that the Strong PLA governs associations in Xhosa and Zulu, a reassessment of antisymmetry is called for—either rejection or weakening of the idea that high to low maps invariantly from left to right.

If, on the other hand, kuphela/qha are only subject to the Weak PLA in the associations in question, and thus only need to c-command a copy, no antisymmetry problem arises: we can assume that the associate precedes kuphela/qha by virtue of raising across it.

Another possibility is that Nguni will present Strong PLA effects that turn out to be only apparent, capturable under some alternative, antisymmetry-friendly approach which maintains that only the Weak PLA holds in Nguni.

Last but not least, it might be that kuphela/qha need not c-command the associate at all; this remains to be established here.

(23) summarizes the analytical options that we have identified with respect to kuphela/qha and antisymmetry theory.

(23) Analytical possibilities for kuphela/qha vis-à-vis the LCA

  • Option 1: Kuphela/Qha need not c-command their associates.

  • Option 2: Kuphela/Qha can associate with a lower copy of a moved expression in Xhosa and Zulu; thus, the Weak PLA is correct for these languages, like German and Dutch.

  • Option 3: The Strong PLA is correct for Zulu and Xhosa, and therefore the LCA is wrong. Syntax is not antisymmetric.

  • Option 4: An antisymmetric analysis can capture apparent Strong PLA effects differently.

  • Option 5: The Strong PLA constrains associations in Zulu and Xhosa. Syntax is only weakly antisymmetric in that it allows rightward adjunction (Carstens 2008, 2017, Takano 2003). Kuphela/Qha are adjuncts and can c-command an associate to the left.

In what follows, we will describe in detail the distributional constraints on kuphela/qha, showing that they reflect two factors: (a) the topography of [+focus], focus-tolerant, and antifocus positions in Nguni clauses, and (b) the Strong PLA (though with caveats mentioned in section 1 and discussed in section 8). Associations with material in focus-tolerant positions require surface c-command by kuphela/qha and are thus inconsistent with Options 1 and 2. We will argue that Option 3 must be rejected on the basis of strong crosslinguistic arguments for underlying Spec–head–complement order. As for Option 4, we will consider an antisymmetry-friendly approach to final particles proposed by Erlewine (2017) and Hsieh and Sybesma (2011) and show that it is not viable for kuphela/qha. We conclude by adopting Option 5 for these associations.

4 Capturing the Distribution of Kuphela/Qha

4.1 The Topography of Focus and Antifocus

Given the classes of expressions that can appear in particular clausal positions in Xhosa and Zulu, previous studies distinguish [+focus], antifocus, and focus-tolerant locations (see, e.g., Adams 2010, Buell 2008, Carstens and Mletshe 2015, 2016, Cheng and Downing 2009, Sabel and Zeller 2006, Zeller 2008, 2015). This constrains the interpretation of kuphela and qha in ways that do not come up for English only because the associates of kuphela and qha are restricted to positions where foci are licit.

In this section, we review and illustrate the topography of focus uncovered in the above-cited works and show how it constrains the distribution of the associates of kuphela and qha. We follow these works in exemplifying focal properties of each position by means of the distribution of (a) phrases modified by kuphela and (b) wh-phrases, as these are generally recognized as [+focus] expressions. This established, we can describe the ways in which associations with kuphela and qha are further constrained by a surface c-command requirement.

4.2 [+Focus] Positions

Studies of [VSO] constructions in Nguni languages have proposed that the verb raises across the subject, which either fails to raise at all, as in Halpert’s (2015) analysis of Zulu, or raises very locally, to Spec of a FocusP atop vP (Carstens and Mletshe 2015, 2016). (24a) exemplifies the focus interpretation characteristic of postverbal subjects in active [VSO] constructions in Xhosa (Carstens and Mletshe 2015, 2016), and (24b) does the same for clefts (Sabel and Zeller 2006). (25) shows that these are not felicitous answers to a ‘What happened?’ question, which requires an all-new, sentence-focus answer whereas the subject cleft and [VSO] constructions convey subject focus. The examples in (26) and (27) show that wh-phrases and expressions modified by kuphela appear freely in these two [+focus] positions (while (24)–(27) are Xhosa, the Zulu facts pattern alike).

(24)

graphic
graphic

(25)

graphic

(26)

graphic

(27)

graphic

4.3 Focus-Tolerant Positions

Material inside vP of an [SVO(O)] construction may but need not include expressions interpreted as focused. There is a preference for such items to appear in the immediately postverbal position (see, e.g., Buell 2009, Cheng and Downing 2009), but this is not absolute, as (28b) and (29b) illustrate.

(28)

graphic
graphic

(29)

graphic

[SVOO] double object constructions with clause-final kuphela, such as (28b) or (3b) (repeated below), are sometimes judged as ambiguous, allowing for kuphela to associate with various constituents of the vP (for details on these judgments, see footnote 2).

(3)

graphic

Either argument in an impersonal passive of a ditransitive verb can be focused.

(30)

graphic

(31)

graphic

The subject of an intransitive expletive construction also may but need not have a focus interpretation (these data are from Xhosa, but Zulu patterns the same way; Zeller 2008).

(32)

graphic
graphic

Some speakers of both languages permit preverbal subjects in subjunctives or relative clauses to contain focused material.10

(33)

graphic

4.4 Antifocus Positions

4.4.1 Preverbal Subject Position of Indicatives

Example (34a) illustrates that a preverbal subject of an indicative clause cannot be modified by kuphela. The intended meaning can instead be expressed with a cleft (see (24b) and (34b)) or a [VS] construction (see (24a)). (35a–c) show that wh-phrases share this distributional pattern. Sabel and Zeller (2006) and Zeller (2008) accordingly propose that the preverbal subject position is antifocus in Zulu. Carstens and Mletshe (2016) report that the facts are the same in Xhosa and adopt the antifocus account.

(34)

graphic

(35)

graphic
graphic

As noted in section 4.3, some Zulu and Xhosa speakers permit focused material in the preverbal subject position of a subjunctive or relative, but this is not universal (see footnote 10). For other speakers, the prohibition illustrated in (34) and (35) holds across clause types, and thus (36a–b) are unacceptable.

(36)

graphic

4.4.2 Dislocated Expressions Are Antifocus

Kuphela and qha cannot be associated with a dislocated expression; nor can wh-phrases be clitic-doubled, indicating that they cannot be dislocated either. Following Buell (2008) and Cheng and Downing (2009), Zeller (2015) proposes that dislocated material in Nguni is antifocus (as in other languages; on information structure effects of clitic doubling in Greek and Spanish, see Anagnostopoulou 1994, Ordóñez 1997, Schneider-Zioga 1994).

(37)

graphic

(38)

graphic

(39)

graphic

4.4.3 Xin [VSX] Is Antifocus

Carstens and Mletshe (2015, 2016) show that in active expletive constructions with [VSX(Y)] word order, X is generally resistant to focus in Zulu and Xhosa.11

(40)

graphic

(41)

graphic

(42)

graphic

(40)–(42) illustrate that an element X cannot be focused when it directly follows the postverbal subject in an expletive construction with [VSXY] order.12 This contrasts sharply with the licit focus of Xin [SVXY] constructions (compare with (28a) and (29a)).

4.4.4 Summary and Discussion

Table 1 summarizes the way [+focus] material is distributed in Xhosa and Zulu clauses.

Table 1

Topography of focus in Zulu and Xhosa

[+focus]: kuphela or qha acceptable
  • Clefted material

  • Sin [VSO] constructions

 
Focus-tolerant: kuphela or qha may associate
  • VP-internal material of active SV construction: S [VPV(X)(Y) . . . ]

  • VP-internal material of (impersonal) passive: [VPVPASS(X)(Y) . . . ]

  • S of an intransitive expletive construction: [VS . . . ]

  • S of [SV . . . ] in a subjunctive or relative clause, for some speakers

 
Antifocus: kuphela or qha not possible; associations banned
  • Preverbal subject position is antifocus—either of indicatives only or, for some speakers, across the board:

    • S of [SVO], all clause types, for some speakers, or

    • S of [IndicSVO] for others

  • Dislocated expressions are antifocus: [[S (cl+)V] . . . DP] and [DP [S (cl+)V . . . ]]

  • X in active expletive constructions [VSXY] is antifocus (especially if there is an overt Y)

 
[+focus]: kuphela or qha acceptable
  • Clefted material

  • Sin [VSO] constructions

 
Focus-tolerant: kuphela or qha may associate
  • VP-internal material of active SV construction: S [VPV(X)(Y) . . . ]

  • VP-internal material of (impersonal) passive: [VPVPASS(X)(Y) . . . ]

  • S of an intransitive expletive construction: [VS . . . ]

  • S of [SV . . . ] in a subjunctive or relative clause, for some speakers

 
Antifocus: kuphela or qha not possible; associations banned
  • Preverbal subject position is antifocus—either of indicatives only or, for some speakers, across the board:

    • S of [SVO], all clause types, for some speakers, or

    • S of [IndicSVO] for others

  • Dislocated expressions are antifocus: [[S (cl+)V] . . . DP] and [DP [S (cl+)V . . . ]]

  • X in active expletive constructions [VSXY] is antifocus (especially if there is an overt Y)

 

One logical possibility is that this focus topography in Nguni plays the roles that both focal stress and c-command play in a language like English: thus, Nguni focus-sensitive particles are freely associated with any expression in a [+focus] or focus-tolerant position, though perhaps subject to the independent locality constraints of phase-based syntax. Another possibility is that both the focus topography and a version of the PLA are involved. Our task is to determine the facts and their implications.

Since the preverbal subject position of indicatives is antifocus (as discussed in section 4.3.1), the simple test of c-command in (13) (repeated here) is inapplicable in Xhosa and Zulu. Because wh-question words are typically in situ, a translation of (14a) also does not provide a useful test.

(13)

graphic

(14)

graphic

Recall also from section 4.2 that association of kuphela or qha with nonadjacent, focusable material is considerably less accessible to most speakers than association with an immediate neighbor, giving rise to variations in judgment patterns like those in (43). All long-distance associations are rejected by some speakers (see footnote 2), complicating the investigation of c-command effects (see also the judgments reported in (4), discussed in sections 1 and 4.2).

(43)

graphic

Nonetheless, those speakers who accept long-distance associations provide several kinds of evidence that associations of kuphela and qha to material in focus-tolerant positions are subject to the strong PLA in Xhosa and Zulu.

5 Evidence on the Principle of Lexical Association in Zulu and Xhosa

5.1 Introduction

In this section, we present evidence that association of kuphela or qha to material in a focus-tolerant position is sensitive only to the intended associate’s highest copy. We first point out a strong implication in this direction from the failure of associations when the subject lands in the antifocus preverbal subject position, [IndicSVO]. We then add evidence from associations to subjects of subjunctives and relative clauses.

5.2 Evidence from the Preverbal Subject Restriction

Recall that the preverbal subject of an indicative cannot contain kuphela or qha, nor can adverbial kuphela or qha associate with a preverbal subject (see (34a) and (3f), repeated here).

(34)

graphic

(3)

graphic

The infelicity of reading (3f) argues that lower copies in a movement chain are not available for association with kuphela/qha. Low subjects, including agentive ones, are always focusable, as (21) shows (repeated here). While (21) arguably involves adnominal kuphela, (44) shows that a nonadjacent adverbial kuphela/qha can also associate with a low subject.

(21)

graphic

(44)

graphic

As previously noted, Carstens and Mletshe (2015, 2016) attribute postverbal subject focus to movement into a low Spec,FocusP, while other authors have argued that a sole vP-internal expression may be interpreted as focused (Cheng and Downing 2012, Halpert 2015). What is important for present purposes is that kuphela/qha freely associate with low subjects. Hence, if adverbial kuphela/qha could associate with a low copy in a movement chain, (3f) would be expected to be licit. We illustrate in (45): FP = vP or FocusP, whichever functional category houses low subjects.

(45)

graphic

The unacceptability of readings like (3f) is thus a strong indication that associations are computed only in relation to the highest copy of a movement chain; hence, the Strong PLA is the condition relevant to Zulu and Xhosa kuphela. If this copy is in an antifocus position, the association fails.

It is important to acknowledge that so far, we have presented no real evidence that associations are based upon kuphela c-commanding the associate and not vice versa. Thus, it might conceivably be the case that (3f) is illicit because in the intended reading, kuphela is c-commanded by an associate in an antifocus position. The following sections, on subjects of subjunctives and relative clauses, make it clear that associations with material in focus-tolerant positions are dependent upon kuphela c-commanding the associate and not vice versa.

5.3 Long-Distance Associations with Focus-Tolerant Preverbal Subject Positions

5.3.1 Focus in the Subject Position of Subjunctives

Recall that some speakers allow focused material in the preverbal subject position of a subjunctive.13 Inclusion of the complementizer ukuthi in the Zulu example (46) shows that this is not a raising-to-object construction; the subject surfaces within the embedded clause.

(46)

graphic

Pietraszko (2017) reports comparable facts for Ndebele, another Nguni language, and proposes that subjunctives lack some structure that is present in indicatives. Indicative subjects may not be focused because they surface in Spec of a TopP projection that is absent in subjunctives. Subjunctive subjects surface in Spec,TP, as shown in (47a) vs. (47b). Arguing for this difference in size are telltale contrasts in negation and agreement in Nguni. Pietraszko argues that part of the structure present in indicatives (see (48a)) is a ΣP, the locus of a negation marker a- that precedes subject agreement (see (48b)) in indicatives. In subjunctives, where this structural layer is lacking, negation is expressed by a morpheme nga- that follows subject agreement. Pietraszko proposes that nga- heads a NegP located between TP and vP.

(47)

graphic

(48)

  • A-ngi-phek-i.

    NEG1-1SG.SM-cook-FV

    ‘I don’t cook.’

  • . . . ngi-nga-phek-i

    1SG.SM-NEG2-cook-FV

    ‘ . . . that I not cook’

A systematic difference between indicatives and subjunctives in subject agreement morphology is illustrated in (49a–b). The u- agreement marker accompanying third person subjects of noun class 1/1a in indicatives is analyzed by Pietraszko as uφ of Top, whereas a- subject agreement for this class in subjunctives is uφ of T as shown in (47a–b) (see Schneider-Zioga 2002, 2007 for similar conclusions regarding subject position and clause size in Kinande).

(49)

graphic

5.3.2 Kuphela at a Distance Associating with the Subjunctive Subject

We found that some Xhosa and Zulu speakers who accept adnominal kuphela/qha in preverbal subject position of a subjunctive also tolerate long-distance associations between such a preverbal subject and adverbial kuphela/qha, as in (50).14 We propose that such associations are possible because kuphela/qha may adjoin to TP, as shown in (51).

(50)

graphic

(51) . . . [CP that [TP[TP Thandeka sing] kuphela]]

The possibility of this association at a distance makes possible a revealing test: if a vP-level modifier follows kuphela, restricting kuphela to a lower adjunction site, what readings will be available? We take nonselected locatives to be vP-level modifiers in Nguni. Example (52) from Xhosa shows that association to the subject is impossible when a locative follows kuphela.15

(52)

graphic

Thus, speakers who otherwise accept long-distance associations to subjunctive subjects in Spec,TP rule them out when kuphela is unambiguously located in the vP, since vP excludes the highest copy of the associate. The relevance of the PLA to kuphela is given support comparable to the evidence of (13) (repeated here) for English.

(13)

graphic

5.3.3 Preverbal Subjects in Relative Clauses

Recall that the preverbal subject position of object relative clauses is also focus-tolerant, for some speakers of Nguni languages (see the Xhosa example (33b), repeated here). Pietraszko (2017) attributes this property of Ndebele to movement-facilitating structure deletion (Pesetsky’s (2019) exfoliation), eliminating the obstacles of phasal IP and TopP to object extraction.

(33)

graphic

(53)

  • Ordinary: [CP C [ΣP Σ [TopP Subj Top [TP T [NegP Neg [vP . . . ]]]]]]

  • Reduced: [CP ObjwhC [TP Subj Tφ [NegP Neg [vP . . .

    graphic
    Obj
    graphic
    . . . ]]]]

As was true of subjunctives, long-distance association is acceptable to some speakers, if kuphela is final (see (54), from Zulu), suggesting that it may TP-adjoin as in (51). For these speakers, association fails when a vP-level modifier follows kuphela, arguably forcing it to adjoin lower (see (55)).16

(54)

graphic

(55)

graphic

Summing up, speakers who accept long-distance associations to relative clause subjects in Spec,TP rule them out when kuphela is unambiguously vP-internal. Like the subjunctive facts, these data support the relevance of the PLA to kuphela.

5.4 Interim Conclusions

While section 4 demonstrates that the distribution of kuphela/qha is constrained by the [+/−focus] values of particular clausal positions, this section has shown that there is more to the picture. The ban on long-distance associations to subjects of indicatives provides important confirmation that the strong version of the PLA holds in Xhosa and Zulu (see (45), repeated here).

(45)

graphic

Further, evidence from clause-final adjuncts has shown that kuphela/qha must c-command their associates in the focus-tolerant subject positions of subjunctives and relative clauses. We accordingly rule out the following two options of (23):

  • Option 1: Kuphela/Qha need not c-command the associate. ✗

  • Option 2: Kuphela/Qha can associate with a lower copy of a moved expression in Xhosa and Zulu; thus, the Weak PLA is correct for these languages. ✗

Since kuphela/qha appear to the right of their associates, their syntax therefore conflicts with the antisymmetric view that X precedes Y if and only if X or a category that contains it asymmetrically c-commands Y.

6 The Antisymmetry Question

How should the conflict between the linear order facts of kuphela/qha be reconciled with the LCA? One possibility is that the LCA is simply wrong—our Option 3 from (23).

  • Option 3: The strong PLA is correct for Zulu and Xhosa, and therefore the LCA is wrong. Syntax is not antisymmetric.

But there is much compelling crosslinguistic evidence supporting universal Spec–head–complement order.

As Kayne (1994) points out, the absence of successive-cyclic wh-movement to the right is one indicator. Despite the diversity of wh-question strategies in the languages of the world, nothing along the lines of (56) is attested.

(56) *[CP [TP Mary [VP [CP [TP Calvin what bought] that what] thinks] PRES] CWHwhat]?

[Intended: ‘What does Mary think that Calvin bought?’]

Kayne also points out that while there are numerous verb-second (V2) languages, there are no verb-penultimate languages. If V2 German, exemplified in (57), had a mirror-image counterpart in the languages of the world, it would exhibit word order patterns like (58). But nothing of the kind has been identified, to our knowledge.

(57)

graphic

(58)

  • Hypothetical V-penultimate language, nonoccurring

  • OK: S O Time V Place

  • OK: S Time Place V O

  • OK: S O Place V Time

  • *non-V-penultimate orders

The West African language Ịjọ also provides persuasive evidence for underlying Spec–head–complement order. Muysken (1988) observes that though there are head-final languages with serial verb constructions (SVCs), they do not have the mirror-image order [O2 V2 O1 V1] of head-initial SVCs. Instead, only the local ordering of verbs and objects is reversed, from [V1 O1 V2 O2] to [O1 V1 O2 V2]. On the basis of a comparison between SVCs in head-final Ịjọ and SVCs in head-initial languages (see (59)), Carstens (2002) proposes that this typological gap is due to antisymmetry: head-finality in SVCs results from V-movement or object shift around the verb, as shown in (60c).

(59)

graphic

(60)

graphic
graphic

FOFC effects provide some additional arguments for underlying left-headedness. Holmberg (2000), Biberauer, Holmberg, and Roberts (2014), and Sheehan et al. (2017), among others, point out that the patterns [Aux VO], [OV Aux], and [Aux OV] are common, the latter a case of mixed-headedness. But the logically possible mixed-headedness pattern [VO Aux] is very rare. Biberauer, Holmberg, and Roberts (2014) propose that this is because head-initial syntax is underlying. An EPP-type feature ^ must be passed up the tree from head to head to induce surface head-finality in any category. If V has the feature and raises its object, it can also pass the feature to Aux, which will raise VP. If V does not, then Aux has no source of this feature (see (61) vs. (62)). Consequently, Aux can follow VP only if VP is also head-final.17

(61)

graphic

(62)

graphic

We conclude that antisymmetry captures important crosslinguistic generalizations and that this greatly outweighs its incompatibility with Zulu and Xhosa ‘only’. The question that remains is how to reconcile these two results.

7 An Antisymmetric Approach to Kuphela and Why It Fails

7.1 Introduction

This section explores a potential means of explaining apparent Strong PLA effects in Xhosa and Zulu without relying on surface c-command, thus addressing Option 4 of (23).

  • Option 4: An antisymmetric analysis can capture apparent Strong PLA effects differently.

The starting point is a proposal made by Erlewine (2017) and Hsieh and Sybesma (2011): sentence-final particles are underlyingly initial phase heads whose complements raise after Transfer. We will show that this approach to kuphela and qha is untenable, leaving us with Option 5.

7.2 Final Particles as Underlyingly Initial Phase Heads

As noted in section 1, Erlewine (2017) and Hsieh and Sybesma (2011) propose that clause-final particles in Mandarin are actually phase heads that precede their complements. As phase heads, they trigger Transfer of their complements (also known as phase interiors). Transferred material is a syntactic atom, so the result is a symmetric and hence unlinearizable representation

graphic
H, α
graphic
(e.g., Kayne 1994, Moro 2000, Richards 2010). The complement α must raise to break symmetry (see Moro 2000).18 We illustrate in (63) how this would work for a Xhosa DP selected by a hypothetical adnominal phase head kuphela or qha.

(63)

graphic

Such an analysis takes as underlying the alternative order

graphic
kuphela/qha associate
graphic
that is accepted in Xhosa and by some Zulu speakers as well (see (7a), repeated here).

(7)

graphic

This approach maintains the attractive assumption that the LCA applies uniformly in syntax. But it is not obvious how to reconcile it with the evidence of the Strong PLA for Xhosa and Zulu kuphela/qha, because in the configuration it yields, the focus particle no longer c-commands the highest copy of the focus. The same question comes up with respect to the Mandarin final particle eryi ‘only’ that Erlewine (2017) discusses, since he claims it must find an associate within its scope, which we take to be its c-command domain.

A possible answer to this objection might be provided by the following hypothesis about the way grammar operates. It could be the case that the PLA ceases to apply to a focus associate once it undergoes Transfer, within a unit that excludes ‘only’. Perhaps, once the constituent containing the associate becomes a syntactic atom, the PLA ceases to be relevant because the height of the associate relative to the focus particle (in the case at hand, kuphela or qha) is no longer visible.

This hypothesis can be tested in English, where the Strong PLA is well-established. The prediction is that sensitivity to surface c-command ends when an associate is contained within a phase interior. However, (64) shows that this prediction is not borne out. The shaded material is the TP complement to the embedded phase-head complementizer that. By assumption, this TP is transferred to Spell-Out prior to construction of the higher clause that dominates it, and forms a syntactic atom after Transfer. According to the hypothesis that the Strong PLA no longer holds in this case, only should be able to associate with a focused constituent inside the raised TP because it c-commands this constituent prior to atomization and cannot “see” its location afterward. But in reality, if the embedded CP is raised into that higher clause by passivization, its contents are unavailable for association with the adverbial only, as the unacceptable continuation not books in (64b) illustrates.

(64)

  • It is only believed [CP that Julie likes movies]

    . . . ✓not proven/✓ not Sue/✓not books.

  • [CP That Julie likes movies] is only believed [CP that Julie likes movies]

    . . . ✓not proven/✗not Sue/✗not books.

See also Chomsky, Gallego, and Ott to appear on the continued availability of transferred material in syntactic relations and for a proposal that the Phase Impenetrability Condition permits this so long as the transferred material is not altered.

We are unable to replicate this test in Zulu or Xhosa, where CP subjects are unacceptable (see Halpert 2015) and the content of the preverbal subject position is antifocus. This applies not only to the whole of a simple subject, as in our foregoing examples, but also to the subparts of more complex subjects (compare (65a–b) and (66a–b)). Therefore, no such test based on raising a constituent to subject position can succeed.19

(65)

graphic
graphic

(66)

graphic

The implication of the English evidence is clear, however: associations to transferred material are not “frozen”; movement alters them. This is unlikely to be a language-particular fact.

We conclude that the conflict between leftward associations and antisymmetry cannot be resolved by an analysis of final focus-sensitive particles as phrase-initial phase heads that obligatorily trigger Spell-Out and raising to their Specs, contra Erlewine (2017).

7.3 Summary

At this point, we have explored and rejected four of the five possibilities presented in (23) (repeated here) for explaining the phrase-final position of kuphela/qha. There are no remaining options consistent with a strict version of Kayne’s (1994) LCA.

(23) Analytical possibilities for kuphela/qha vis-à-vis the LCA

  • Option 1: Kuphela/Qha need not c-command their associates. ✗

  • Option 2: Kuphela/Qha can associate with a lower copy of a moved expression in Xhosa and Zulu; thus, the Weak PLA is correct for these languages, like German and Dutch. ✗

  • Option 3: The Strong PLA is correct for Zulu and Xhosa, and therefore the LCA is wrong. Syntax is not antisymmetric. ✗

  • Option 4: An antisymmetric analysis can capture apparent Strong PLA effects differently. ✗

This leaves only Option 5.

  • Option 5: The Strong PLA constrains associations in Zulu and Xhosa. Syntax is only weakly antisymmetric in that it allows rightward adjunction (Carstens 2008, 2017, Takano 2003). Kuphela/Qha are adjuncts and can c-command an associate to the left.

That adjuncts are exempt from the LCA was proposed independently by Takano (2003) to account for the ability of material inside adjuncts to license negative polarity items on their left, as in (67). Takano notes that while the judgments are delicate, these examples are at least much better than *Anyone saw nothing.

(67)

  • John paints pictures at all well only rarely.

  • Jay tells jokes with any gusto only occasionally.

Relatedly, Carstens (2008, 2017) presents arguments that postnominal numerals, AP modifiers, and demonstratives in Swahili and Shona (also Bantu) are base-generated in high, right-adjoined positions—an additional source of evidence that adjuncts are not universally attached to the left of constituents that they c-command.

We briefly consider some recent arguments against immunity of adjuncts to the LCA in section 9.

8 Associations to [+Focus] Positions

We turn now to some intriguing deviations from the patterns we have described. As noted in section 1, judgments vary on associations to (a) clefted expressions and (b) the postverbal subjects of transitive expletive constructions. Material in these two positions has systematically [+focus] interpretations, as we have demonstrated (see (24a–b), repeated here).

(24)

graphic

The argument begins with hyperraising out of tensed clauses, which is quite productive in Zulu and Xhosa (see Zulu examples (68a–b)). (69) shows that hyperraising preserves idiomatic readings, a standard diagnostic for true raising constructions.

(68)

graphic

(69)

graphic

Recall that the preverbal subject position of an indicative cannot host focused material, so we did not use subject-to-subject hyperraising as a test of kuphela/qha’s sensitivity to low copies. The results are ill-formed in both languages (see (70) from Zulu), but shed no new light since they are consistent with the results of raising subjects to preverbal position in a single clause and therefore support the generalization in (45) (repeated below)—that the preverbal subject restriction for adverbial kuphela/qha supports the Strong PLA.

(70)

graphic

(45)

graphic

But hyperraising can feed clefting, so in principle, this combination of operations provides a further test of the Strong PLA.

A few speakers we consulted disallowed any association in which the word order is inconsistent with kuphela or qha c-commanding the highest copy of the [+focus] associate, as in the following Xhosa examples, where qha is sandwiched into the lower clause by an embedded locative (71) or time adjunct (72):20

(71)

graphic

(72)

graphic

But other speakers were more liberal in their judgments on these cases, permitting at least some associations where kuphela or qha c-commands a low copy of the associate surfacing in a [+focus] position, at least some of the time, as in (71b) and (72c). Variations of this kind occurred somewhat unpredictably across individuals, across structurally comparable examples, and across sessions with the same individual. On a 1–5 scale, speakers tended to give examples that violate the Strong PLA a middling rating of 3—thus, they are not fully acceptable but not entirely excluded either. However, judgments were quite variable and occasionally more positive than 3.21

Last but not least, in a few instances, some Xhosa and Zulu speakers accepted associations in which kuphela and qha c-command no copy of an associate in a [+focus] position; see (73) and (74).

(73)

graphic

(74)

graphic
graphic

Curiously, if kuphela or qha is located in an island, judgments remain about the same.

(75)

graphic

(76)

graphic

It seems clear to us that occupying a [+focus] position has a strong effect in these cases. We are not sure whether to attribute this phenomenon entirely to difficulty in isolating two foci in a single clause, or whether a second strategy exists for association at a distance with a [+focus] position. Recall from section 4.4.4 our conjecture that as a logical possibility, the clausal topography of focus in Nguni might play the roles that both focal stress and c-command play in a language like English. The pattern of judgments on association to material in focus-tolerant positions did not bear this out as a general conclusion, but for speakers who accept associations without c- command to material in [+focus] positions like (73)–(76), the possibility seems worth considering. On the other hand, the fact that acceptance is only middling needs to be explained if such an alternative strategy exists. We leave further exploration to future research.

9 Conclusion

Association of the particles kuphela and qha is constrained by the topography of Nguni focus, by a preference for string adjacency consistent with a strategy of constituent marking, and by a need to c-command an associate in surface syntax, modulo some anomalies associated with clefted material and [+focus] postverbal subjects of transitive expletive constructions.

The fact that kuphela and qha are generally to the right of their associates raises subtle analytical challenges. On close examination, the facts are not compatible with a strict version of the LCA.

We propose that kuphela and qha are adjunct particles, and that at least some adjuncts fall outside the purview of the LCA (see Carstens 2008, 2017, Takano 2003).

The claim that adjuncts may be exempt from antisymmetry is controversial. Sheehan (2017) considers this possibility in relation to the order [VO Adv], putting it aside in favor of a derivation consistent with the LCA and FOFC: she argues for independent movements of the verb and its object across the adverb, from a position to its right. As she points out, Cinque (1999) analyzes adverbs as Specs of functional categories; hence, in his view they always merge to the left of heads.

The association requirements of adverbial and adnominal kuphela/qha are important tests of the generality of these claims. The results we have described argue against adopting them as universals.

These results also argue against Bruening’s (2014) proposal that effects which are standardly analyzed in terms of c-command are in fact due to precedence plus a domain restriction, phase command, under which a bindee must be included in the same phase as its binder. In (77a), the pronoun she illicitly binds the R-expression Bernice because it precedes it and no phasal node separates them.

(77)

  • *Shei likes Bernicei’s friends.

  • Heri mother likes Bernicei’s friends.

    (Bruening 2014:344)

(78)

  • Binding Principle C

    An R-expression may not be bound.

  • A binds B iff A and B are coindexed and A precedes and phase-commands B.

  • Phase-command

    X phase-commands Y iff there is no ZP, ZP a phasal node, such that ZP dominates X but does not dominate Y.

    (Bruening 2014:344)

Clearly, the association facts of kuphela and qha are at odds with Bruening’s proposal to recast c-command in terms of precedence.

We leave a number of questions to future research, among them an explanation for the apparent exemption of adjuncts/adjunct particles like kuphela and qha from the LCA, and the important and intriguing task of clarifying how associations work when the associate is in a [+focus] position and surface c-command is lacking, for speakers who allow this.

Notes

1 Zulu (or isiZulu) and Xhosa (isiXhosa) are Bantu languages spoken primarily in South Africa. They belong to the Nguni group, which also includes (si)Swati and (isi)Ndebele. While the Nguni languages show a modest degree of lexical and grammatical variation, they are mutually intelligible and sometimes considered varieties of one language.

2 Glosses: Arabic numbers = noun classes (number + gender) unless followed by SG or PL, in which case they indicate person features. APPL = applicative; ASS = associative marker; AUG = augment; CAUS = causative; COP = copulative prefix; DEM = demonstrative; DJ = disjoint verb form; FUT = future tense; FV = final vowel; LOC = locative marker; NEG = negation; OM = object marker; PASS = passive; PL = plural; POSS = possessive marker; PRO = pronoun; PST = past tense; REL = relative marker; SG = singular; SJ = subjunctive; SM = subject marker. Examples are not marked for tone.

To collect data, we constructed a set of Xhosa and Zulu sentences with the help of native speakers. These and subsequent speakers explored potential ambiguities connected with kuphela/qha, and the allowable meanings for variations in which they were repositioned elsewhere in the clause. We provide information on acceptance rates for examples in footnotes.

In our study, eleven out of fourteen Xhosa speakers accepted associations at a distance such as (4b), as did three out of five Zulu speakers for (3b–c). Associations with nonadjacent material are greatly facilitated by the addition of disambiguating continuations compatible with them, such as hayi izinja ‘not the dogs’ for (3b) and hayi uMary ‘not Mary’ for (4b). Regarding (3b), although there are reports in the Nguni syntax literature that O2 of [SV O1 O2] cannot be focused, we did not find this; see section 4.3 for details.

3 The analysis contrasts with Cardinaletti’s (2011) approach to deriving the location of final particles in Italian; see below (5).

4 Though not compatible with this mechanics, kuphela and qha seem consistent with the FOFC generalization that headedness within extended projections is harmonic, absent evidence that they form part of such domains. See Biberauer 2017 for relevant in-depth discussion of this issue regarding particles in a variety of languages.

5 All Xhosa speakers seem to accept examples along the lines of (7a) and (7bi); three out of seven Zulu speakers in our study accepted them.

7 In the English examples, we follow the standard convention (going back to Jackendoff 1972) and mark the syntactic focus by means of square brackets and the focus feature F, which mediates between the semantics of focus and its prosodic realization. In languages such as English, focused constituents are prosodically prominent and marked by a pitch accent on the main stress-bearing syllable (compare Her husband likes [the MEATballs]F with [Her HUSband]Flikes the meatballs).

In contrast, we have not adopted F-marking for focused material in our Nguni examples (which we mark with boldface instead), because focus is not correlated with prosodic prominence in Nguni and only influences prosodic phrasing indirectly, through its syntactic position (Downing 2010). There also seem to be no prosodic cues to disambiguate sentences with multiple possible focus readings in Zulu and Xhosa. Speakers listening to recordings of ambiguous sentences involving kuphela did not reliably identify intended readings even when the recorded speech was their own. On the link between syntactic position and focus, see section 4.

8 Note that in Rooth’s (1985, 1992) theory, the ordinary semantic value of an expression is always an element of its focus semantic value.

9 See Adams 2010, Buell 2005, Cheng and Downing 2009, Halpert 2015, Van der Spuy 1993, Zeller 2015, and others for ample evidence that object-marked DPs in Zulu are always dislocated to a VP-external position.

10 This possibility came to our attention through Pietraszko 2017, where it is shown that subjunctive and relative clause subject positions are focus-tolerant in Zimbabwean Ndebele (another Nguni language, very close to Zulu). Pietraszko attributes this pattern to subjunctives’ and relative clauses’ having a smaller clause size than that of indicatives; we discuss this in section 5.3.1 (see also Schneider-Zioga 2002, 2007 on Kinande). (Thanks to Karlos Arregi for making us aware of Pietraszko’s work.)

Nine out of thirteen Xhosa speakers approved focused material in the subject position of subjunctives, as did five out of seven Zulu speakers. Most of the tests we did for relative clauses were conducted with Zulu speakers, five out of seven of whom accepted focused material in the subject position of object relative clauses (four of these speakers were the same ones who accepted focused material in the subject position of subjunctives). One out of four Xhosa speakers tested accepted focused material in the subject position of relative clauses.

11 As for why this might be so, Carstens and Mletshe (2015) propose that a low Focus head marks the higher argument [+focus] and the lower [−focus] (the latter might be better understood as covert focus-background marking; see Grubic 2015, Zimmermann 2016 on this). Carstens and Mletshe (2016:796) report a suggestion by Julie Anne Legate that an expression following a [+focus] constituent might be destressed and that this yields the [−focus] interpretation. We leave this interesting question aside. The effect is strongest and most consistent when there is overt material Y following X of [VSXY], as in the ditransitive expletive constructions of (40)–(42). Carstens and Mletshe (2016) report that in a monotransitive VSO construction, many speakers accept a wh-phrase as O. Other focused material—negative concord items, strict negative polarity items, and associates of kuphela—is generally judged degraded there, however.

12Carstens and Mletshe (2016) report that Y of [VSXY] is focus-tolerant. Since other sentence-final focus-tolerant positions are represented in the discussion, in the interests of brevity we ignore this case here.

13 See footnote 10 for speaker numbers.

14 Six of the nine Xhosa speakers who permit focused material in subjunctive clause subject position also permit long-distance association of kuphela to it, with the restriction exemplified in (52). Of the five Zulu speakers who permit focused material in subjunctive clause subject position, three accept long-distance associations to the subject, with the same requirement of surface c-command.

15 An anonymous reviewer suggests that ekhitshini ‘in the kitchen’ might be TP-adjoined, as locatives are versatile in their attachment. The results we report argue that this is not the case, since ekhitshini blocks kuphela from associating with the subject in Spec,TP. It might alternatively be argued that the modifiers we have tested are not vP-level but VP-level, adjoining lower than the subject’s base position—hence, that these results are uninformative about high vs. low copies. But when kuphela immediately follows a verb, that verb must appear in the so-called disjoint form, which is a standard diagnostic of the Nguni syntax literature that a postverbal modifier is outside VP (see, e.g., Buell 2005, Zeller 2012). Anticipating our ultimate conclusion that right-adjunction is possible and in the interests of brevity, it is also worth noting that this discussion does not engage with the antisymmetric idea that such a right-hand locative modifier has been raised across by the clause that it follows.

16 Of the five Zulu speakers in our study who accept focused material in the subject position of relative clauses, three accept associations to it at a distance—providing there is surface c-command. The sole Xhosa speaker (of four tested) who accepts focused material in the subject position of a relative clause also accepts association at a distance to the relative clause subject, with the same requirement of surface c-command.

17 As noted in section 1, Biberauer, Holmberg, and Roberts (2014) suggest that particles can introduce their own ^ features because they are acategorial and hence not part of the domains to which FOFC applies. See section 7 for an attempt to apply this approach to kuphela/qha and to reconcile it with the Strong PLA effects we have described (and see Erlewine 2017 for arguments against acategoricity as a generalization about final particles).

18Erlewine (2017) suggests that many phase heads are silent, obscuring this general pattern.

19 Another potential obstacle to successful testing on this point is uncertainty regarding the acceptability of kuphela associating to a subpart of a nonadjacent constituent. This is an area that we have not explored, and given the delicacy of judgments on all long-distance associations, we anticipate complexity. Otherwise, a test might conceivably be constructed by asking speakers who (a) find the preverbal subject positions of subjunctives and relatives to be focus-tolerant and (b) allow associations at a distance to their contents whether (c) a postverbal kuphela can associate with contents of a transferred subpart of a passivized DP in subjunctive subject position, such as ‘[I asked that [the rumor [that Mary’s brothers like me]] be disproved

graphic
DP
graphic
kuphela . . . ]’, while (d) positioning an adjunct after kuphela to ensure that what is tested is the persistence of associations to transferred material after it moves (given the evidence from sections 5.3.2 and 5.3.3 that clause-final kuphela scopes over the preverbal subject). This is a sizable collection of conditions to be addressed in exploring the question. We leave it to future research to determine whether associations to subparts of nonadjacent expressions are acceptable to the relevant speakers and if so, the status of examples like these.

20 As noted in section 1.1, some speakers also permit association with a following expression—an option more widely available in Xhosa than in Zulu, so we have abstracted away from it. It does not correlate in any way with strictness or permissiveness regarding construals with a raised expression.

21 Only three out of eleven Xhosa speakers who accepted some associations at a distance consistently rejected associations to a clefted constituent unless kuphela surface-c-commanded it. Three more did so in all but one or two cases, which appeared to be random inconsistencies. One speaker was quite unpredictable, and two accepted the majority of associations of kuphela to clefted expressions whether there was any c-command or not. Out of eight Zulu speakers, five accepted long-distance associations to clefted material. Only two of the eight gave consistent judgments indicative of a surface c-command requirement. The eighth speaker was uncertain.

Acknowledgments

For helpful feedback and discussion, thanks to our audiences at the University of Chicago; the University of Illinois at Chicago; the 4th South African Microlinguistics Workshop at Rhodes University in Grahamstown; the 47th and 48th Annual Conference on African Linguistics at Indiana University, Bloomington, and University of Illinois Urbana-Champaign, respectively; the 31st Comparative Germanic Syntax Workshop at the University of Stellenbosch; and the Bantu 6 Conference at the University of Helsinki. Special thanks to Loyiso Mletshe for much helpful input, and to the many Zulu and Xhosa speakers who assisted us with judgments. This work is based on research supported in part by Southern Illinois University and by the National Research Foundation of South Africa. Any opinion, finding, and conclusion or recommendation expressed in this material is that of the authors. SIU and the NRF do not accept any liability in this regard.

References

References
Adams,
Nikki
.
2010
.
The Zulu ditransitive verb phrase
.
Doctoral dissertation, University of Chicago, Chicago, IL
.
Anagnostopoulou,
Elena
.
1994
. On the representation of clitic doubling in Modern Greek. In
EUROTYP working papers, Theme Group Clitics: Clitic doubling and clitic groups
, ed. by
Lars
Hellan
and
Henk
van Riemsdijk
,
1
66
.
Strasbourg
:
ESF Eurotype
.
Aoun,
Joseph
, and
Yen-hui Audrey
Li
.
1993
.
Wh-elements in situ: Syntax or LF ?
Linguistic Inquiry
24
:
199
238
.
Barbiers,
Sjef
.
1995
.
The syntax of interpretation
.
Doctoral dissertation, Leiden University
.
Biberauer,
Theresa
.
2017
. The Final-over-Final Condition and particles. In
The Final-over-Final Condition: A syntactic universal
, ed. by
Michelle
Sheehan
,
Theresa
Biberauer
,
Ian
Roberts
, and
Anders
Holmberg
,
97
120
.
Cambridge, MA
:
MIT Press
.
Biberauer,
Theresa
,
Anders
Holmberg
, and
Ian
Roberts
.
2014
.
A syntactic universal and its consequences
.
Linguistic Inquiry
45
:
169
225
.
Bruening,
Benjamin
.
2014
.
Precede and command revisited
.
Language
90
:
342
388
.
Buell,
Leston
.
2005
.
Issues in Zulu verbal morphosyntax
.
Doctoral dissertation, UCLA, Los Angeles, CA
.
Buell,
Leston
.
2008
. VP-internal DPs and right-dislocation in Zulu. In
Linguistics in the Netherlands 25
, ed. by
Marjo
van Koppen
and
Bert
Botma
,
37
49
.
Amsterdam
:
John Benjamins
.
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
McLaughlin
, and
Eric
Potsdam
,
166
172
.
Somerville, MA
:
Cascadilla Proceedings Project
.
Büring,
Daniel
, and
Katharina
Hartmann
.
2001
.
The syntax and semantics of focus-sensitive particles in German
.
Natural Language and Linguistic Theory
19
:
229
281
.
Cardinaletti,
Anna
.
2011
.
German and Italian modal particles and clause structure
.
The Linguistic Review
28
:
493
531
.
Carstens,
Vicki
.
2002
.
Antisymmetry and word order in serial verb constructions
.
Language
78
:
3
50
.
Carstens,
Vicki
.
2008
. DP in Bantu and Romance. In
The Bantu-Romance connection
, ed. by
Katherine
Demuth
and
Cécile
De Cat
,
131
166
.
Amsterdam
:
John Benjamins
.
Carstens,
Vicki
.
2017
. Noun-to-determiner movement. In
The Blackwell companion to syntax
, ed. by
Martin
Everaert
and
Henk
van Riemsdijk
,
2758
2783
. 2nd ed.
Hoboken, NJ
:
Wiley-Blackwell
.
Carstens,
Vicki
, and
Loyiso
Mletshe
.
2015
.
Radical defectivity: Implications of Xhosa expletive constructions
.
Linguistic Inquiry
46
:
187
242
.
Carstens,
Vicki
, and
Loyiso
Mletshe
.
2016
.
Negative concord and nominal licensing in Zulu and Xhosa
.
Natural Language and Linguistic Theory
34
:
761
804
.
Cheng,
Lisa Lai-Shen
, and
Laura
Downing
.
2009
.
Where’s the topic in Zulu?
The Linguistic Review
26
:
207
238
.
Cheng,
Lisa Lai-Shen
, and
Laura
Downing
.
2012
. Against FocusP: Arguments from Zulu. In
Contrasts and positions in information structure
, ed. by
Ivona
Kučerová
and
Ad
Neeleman
,
247
266
.
Cambridge
:
Cambridge University Press
.
Chomsky,
Noam
,
Ángel J.
Gallego
, and
Dennis
Ott
.
To appear
.
Generative grammar and the faculty of language: Insights, questions, and challenges
.
Catalan Journal of Linguistics
. https://ling.auf.net/lingbuzz/003507.
Cinque,
Guglielmo
.
1999
.
Adverbs and functional heads: A cross-linguistic perspective
.
New York
:
Oxford University Press
.
Cinque,
Guglielmo
.
2005
.
Deriving Greenberg’s Universal 20 and its exceptions
.
Linguistic Inquiry
36
:
315
333
.
Downing,
Laura
.
2010
.
An edge-based approach to the alignment of syntactic phases and prosodic phrases
.
Transactions of the Philological Society
108
:
352
369
.
Erlewine,
Michael Yoshitaka
.
2014a
. Explaining leftward focus association with even but not only. In
Proceedings of Sinn und Bedeutung 18
, ed. by
Urtzi
Etxeberria
,
Anamaria
Fălăuş
,
Aritz
Irurtzun
, and
Bryan
Leferman
,
128
145
.
Bayonne and Vitoria-Gasteiz
:
University of the Basque Country
.
Erlewine,
Michael Yoshitaka
.
2014b
.
Movement out of focus
.
Doctoral dissertation, MIT, Cambridge, MA
.
Erlewine,
Michael Yoshitaka
.
2017
.
Low sentence-final particles in Mandarin Chinese and the Final-over-Final Constraint
.
Journal of East Asian Linguistics
26
:
37
75
.
Grimshaw,
Jane
.
1991
.
Extended projection
.
Ms., Brandeis University, Waltham, MA
.
Grubic,
Mira
.
2015
.
Focus and alternative sensitivity in Ngamo (West Chadic)
.
Doctoral dissertation, University of Potsdam
.
Halpert,
Claire
.
2015
.
Argument licensing and agreement
.
Oxford
:
Oxford University Press
.
Holmberg,
Anders
.
2000
. Deriving OV order in Finnish. In
The derivation of VO and OV
, ed. by
Peter
Svenonius
,
123
152
.
Amsterdam
:
John Benjamins
.
Hornstein,
Norbert
,
Jairo
Nunes
, and
Kleanthes K.
Grohmann
.
2005
.
Understanding Minimalism
.
Cambridge
:
Cambridge University Press
.
Hsieh,
Feng-fan
, and
Rint P. E.
Sybesma
.
2011
.
On the linearization of Chinese sentence-final particles: Max spell out and why CP moves
.
Korean Journal of Chinese Language and Literature
49
:
53
90
.
Jackendoff,
Ray
.
1972
.
Semantic interpretation in generative grammar
.
Cambridge, MA
:
MIT Press
.
Jacobs,
Joachim
.
1983
.
Fokus und Skalen
.
Tübingen
:
Niemeyer
.
Kayne,
Richard
.
1994
.
The antisymmetry of syntax
.
Cambridge, MA
:
MIT Press
.
König,
Ekkehard
.
1991
.
The meaning of focus particles
.
London
:
Routledge
.
Krifka,
Manfred
.
2006
. Association with focus phrases. In
The architecture of focus
, ed. by
Valerie
Molnar
and
Susanne
Winkler
,
105
136
.
Berlin
:
Mouton de Gruyter
.
Moro,
Andrea
.
2000
.
Dynamic antisymmetry
.
Cambridge, MA
:
MIT Press
.
Muysken,
Pieter
.
1988
. Parameters for serial verbs. In
Niger-Congo syntax and semantics 1
, ed. by
Victor
Manfredi
,
65
75
.
Boston
:
Boston University, African Studies Center
.
Ordóñez,
Francisco
.
1997
.
Word order and clause structure in Spanish and other Romance languages
.
Doctoral dissertation, City University of New York
.
Pesetsky,
David
.
2019
.
Exfoliation: Towards a derivational theory of clause size
. .
Pietraszko,
Joanna (Asia)
.
2017
.
Clause size and transparency in Ndebele
.
Poster presented at the 91st annual meeting of the Linguistic Society of America, Austin, TX
.
Richards,
Norvin
.
2010
.
Uttering trees
.
Cambridge, MA
:
MIT Press
.
Rooth,
Mats
.
1985
.
Association with focus
.
Doctoral dissertation, University of Massachusetts, Amherst
.
Rooth,
Mats
.
1992
.
A theory of focus interpretation
.
Natural Language Semantics
1
:
75
116
.
Sabel,
Joachim
, and
Jochen
Zeller
.
2006
. Wh-question formation in Nguni. In
Selected proceedings of the 35th Annual Conference on African Linguistics: African languages and linguistics in broad perspectives
, ed. by
John
Mugane
,
John P.
Hutchison
, and
Dee A.
Worman
,
271
283
.
Somerville, MA
:
Cascadilla Proceedings Project
.
Schneider-Zioga,
Patricia
.
1994
.
The syntax of clitic doubling in Modern Greek
.
Doctoral dissertation, University of Southern California, Los Angeles
.
Schneider-Zioga,
Patricia
.
2002
. The case of anti-agreement. In
Proceedings of the eighth meeting of the Austronesian Formal Linguistics Association
, ed. by
Andrea
Rackowski
and
Norvin
Richards
,
325
339
.
MIT Working Papers in Linguistics 44
.
Cambridge, MA
:
MIT, MIT Working Papers in Linguistics
.
Schneider-Zioga,
Patricia
.
2007
.
Anti-agreement, anti-locality, and minimality: The syntax of dislocated subjects
.
Natural Language and Linguistic Theory
25
:
403
446
.
Sheehan,
Michelle
.
2017
. The Final-over-Final Condition and adverbs. In
The Final-over-Final Condition: A syntactic universal
, ed. by
Michelle
Sheehan
,
Theresa
Biberauer
,
Ian
Roberts
, and
Anders
Holmberg
,
97
120
.
Cambridge, MA
:
MIT Press
.
Sheehan,
Michelle
,
Theresa
Biberauer
,
Ian
Roberts
, and
Anders
Holmberg
.
2017
.
The Final-over-Final Condition: A syntactic universal
.
Cambridge, MA
:
MIT Press
.
Takano,
Yuji
.
2003
.
How antisymmetric is syntax?
Linguistic Inquiry
34
:
516
526
.
Tancredi,
Christopher D
.
1990a
.
Not only EVEN, but even ONLY
.
Ms., MIT, Cambridge, MA
.
Tancredi,
Christopher D
.
1990b
. Syntactic association with focus. In
Proceedings from the First Meeting of the Formal Linguistic Society of Mid-America
, ed. by
Denis
Meyer
,
Satoshi
Tomioka
, and
Leyla
Zidani-Eroglu
,
289
303
.
Madison
:
University of Wisconsin
.
Van der Spuy,
Andrew
.
1993
.
Dislocated noun phrases in Nguni
.
Lingua
90
:
335
355
.
Zeller,
Jochen
.
2008
.
The subject marker in Bantu as an antifocus marker
.
Stellenbosch Papers in Linguistics
38
:
221
254
.
Zeller,
Jochen
.
2012
.
Object marking in isiZulu
.
Southern African Linguistics and Applied Language Studies
30
:
219
235
.
Zeller,
Jochen
.
2015
.
Argument prominence and agreement: Explaining an unexpected object asymmetry in Zulu
.
Lingua
156
:
17
39
.
Zimmermann,
Malte
.
2016
.
Cross-linguistic variability (and uniformity) in focus-background partitioning
.
Conference presentation at Language Variation in Action, Koninklijke Nederlands Academie van de Wetenschapen, Amsterdam. https://www.ling.uni-potsdam.de/~zimmermann/papers/MZ2016-backgroundingAmsterdam.pdf (last accessed 18 January 2019)
.