20di-raw

20di, PDF
Marcel den Dikken Teresa O’Neill

Clausal parataxis – and how!*

1. Introduction Carstens & Diercks (2013) present a meticulous and ingenious analysis of a peculiar agreement pattern in the Luhya languages (narrow Bantu), wherein ‘how’-questions obligatorily show agreement with the subject: (1)

a. b. c. d.

Ny-emba en-die? 1SG-sing 1SG-how ‘How do I sing?’ W-emba o-tie? 2SG-sing 2SG-how ‘How do you sing?’ Y-emba a-tie? 3SG-sing 3SG-how ‘How does she/he sing?’ Khw-emba khu-tie? 1PL-sing 1PL-how ‘How do we sing?’

[Lusaamia]

A similar pattern presents itself for the non-wh counterpart of ‘how’, i.e., ‘so/thus’ (Carstens & Diercks 2013:fnn. 2, 5): thus, compare (1b) with (1b′): (1)

bʹ.

W-emba o-rio. 2SG-sing 2SG-thus ‘You sing thus.’

For the analysis of these agreement patterns, Carstens & Diercks (2013) postulate an Agree relation between the uninterpretable φ-features of the vPadjoined manner adverbial and the matching features of the subject in its base position (SpecvP or lower). This Agree relation between the adverbial and the subject is independent of the probe‒goal relation that T is involved in. Its successful execution requires that the theory favour downward probing to upward agreement and Spec-Head agreement, and that it deny that only (phase) heads can probe. We will present an outlook on the agreeing ‘how’ and ‘thus’ facts that requires no modification of any of the standard assumptions regarding the

A version of this paper was presented by the first author at the third Össznyelvész conference, held at ELTE on 24 and 25 January 2020. Empirical material not otherwise attributed is the fruit of original fieldwork on Lubukusu and Kinande by the second author.

Clausal parataxis – and how! 2

Agree relation. It treats the manner adverbial as a predicate of a clause, linked paratactically to a preceding clause, with a silent pronoun (pro) coindexed with the subject of the preceding clause, as illustrated in (2). (2)

[Clause1 SUBJECTi VP] [Clause2 proi how/thus]

This simple alternative accounts for the data, and makes novel predictions which will be shown to be verified.

2. ‘How are you?’ Sentences of the type in (1b) are transparently concatenations of an ordinary declarative clause and the Luhya expression for ‘how are you?’. For Lubukusu, the language of illustration in most of Carstens & Diercks’ paper, let us illustrate this by juxtaposing (3a) to their example in (4b), reproduced here as (3b).1 (3)

O-riena? 2SG-how ‘How are you?’ b. W-a-ul-ile [oli ba-ba-ana b-oola] 2SG-PST-hear-PST that 2-2-children 2SG-arrived ‘How did you hear that the children arrived?’

o-rie(na)? 2SG-how

Our hypothesis is that what we have in (3b) is a paratactic juxtaposition of a declarative clause and a (zero) copular ‘how’-question whose pro-subject is coindexed with the subject of the preceding clause. Rendering this with English prose is distinctly awkward in the case of (3b): ‘You heard that the children arrived — how are/were you?’. But for examples such as How did I serve you?, English can mimic what we are proposing is the syntax of ‘how’questions in Luhya quite directly: I served you — how was I?.2 If this is the syntax of (3b), the fact that ‘how’ shows agreement with the subject is of course entirely straightforward, regardless of the specifics of the definition of Agree: o-rie(na) is a simple sentence with rie(na) as its predicate 1

The fact that in (3b) o-riena alternates with o-rie is a general fact about all wh-words (except naanu ‘who’) in Lubukusu. The distribution of the long form ending in -na is not fully understood, but seems to be intimately related to emphasis: see Carstens & Diercks 2013:182. This need not concern us in this paper. 2 While ours is the simplest possible approach to the structure of the ‘how’-question juxtaposed to the declarative clause, a reviewer mentions as a possible alternative an analysis of the juxtaposed ‘how’-question parallel to the first clause and undergoing ellipsis: ‘I served you — how did I serve you?’. This gives rise to the desired output if subject inflection on the verb can (indeed, must) survive ellipsis. Ensuring this is not a trivial matter. We will not try this here. The Even Yearbook 14 (2020), Department of English Linguistics, Eötvös Loránd University, Budapest ISSN 2061–490X, http://seas3.elte.hu/even, © 2020, Marcel den Dikken & Teresa O’Neill

Clausal parataxis – and how! 3

and a second-person singular pro as its subject, exactly as in (3a). It seems to us that this should be the null hypothesis for the syntax of ‘how’-questions in Luhya, and that more complex approaches (such as the one advocated by Carstens & Diercks, or the one mentioned in fn. 2, above) should be resorted to only if it can be shown that the null hypothesis is inoperable.

3. Beyond Bantu In English and Dutch rhetorical constructions of the type illustrated in (4) (for (4a), cf. https://en.wiktionary.org/wiki/and_how, https://www.merriam-webster.com/ dictionary/and%20how) and (5) (the latter featuring attested examples taken from the internet), we see overt coordination of how/hoe (with an exclamative function) and a clause that expresses a proposition which this wh-word strongly confirms the veracity of and makes out to hold to a high degree. (4)

a. b.

They launched an attack on Syria, and how! Ze hebben een aanval op Syrië uitgevoerd — en hoe! they have an attack on Syria executed and how

(5)

a. b. c. d.

The defense is going to use that to their advantage. And how! Bill O’Reilly gets some life into the panel, and how! He is well-educated too, really well-educated, and how! John Chen heeft BlackBerry gered, en hoe! John Chen has BlackBerry saved and how ‘John Chen saved BlackBerry, and how!’ Astrid is terug en hoe! Astrid is back and how ‘Astrid is back, and how!’ De Leip! Weken zijn van start gegaan… en hoe! the ‘Leip!’ weeks are from start gone and how ‘The ‘Leip!’ Weeks have started, and how!’

e. f.

The syntax of (4) and (5) is an instantiation of (2). In the Dutch and English examples, a conjunction particle (and/en) overtly links the two clauses; in Luhya (1) and (3b), clausal linking happens asyndetically.

4. Asyndeton vs overt linking Although the Luhya constructions discussed in Carstens & Diercks (2013) include no functional element signalling biclausality, the biclausal approach to agreeing ‘how’ raises the expectation that an overt linking element might occur in the Bantu languages as well, as in the Germanic ‘and how’ examples. The Even Yearbook 14 (2020), Department of English Linguistics, Eötvös Loránd University, Budapest ISSN 2061–490X, http://seas3.elte.hu/even, © 2020, Marcel den Dikken & Teresa O’Neill

Clausal parataxis – and how! 4

Indeed, Kinande uses an overt linking element in its ‘how’/‘thus’ constructions. In (6), the linker yo introduces agreeing ‘how’/‘thus’. The linker yo is the copular element that marks focus in certain specificational copular clauses in Kinande (Schneider-Zioga 2015): see (7). Indeed, in (6) the ‘how’/‘thus’ clause introduced by yo is in a specificational relation with the first clause. (On the structural parallel between specification and conjunction, see Den Dikken 2006; on the ‘:P’ label used in (6c), see Koster 2000.) (6)

a. b. c.

(7)

O-múlumé mw-á-lír-y’-enyamá y-á-ti? IV-man TNS-3SG-eat-meat LNK-3SG-how ‘How did the man eat the meat?’ Mó-n-a-l-íre e-nyamá yó nyi-tya. tns-1sg-tns-eat-tns iv-9meat lnk 1sg-thus ‘I ate the meat thus.’ [:P [Clause1 SUBJECTi VP] [:=yo [Clause2 proi how/thus]]

Marya yo Kambale alangira. Mary LNK Kambale saw ‘Mary (is the one that) Kambale saw.’

5. Word order In all the examples given so far, the ‘how’-expression appears in sentencefinal position. In Dutch it is also possible to place en hoe ‘and how’ in clausemedial position, following the finite verb, as a parenthetical: compare (8) with (4b). (8)

Ze hebben — en hoe! — een aanval op Syrië uitgevoerd. they have and how an attack on Syria executed

For Luhya, Carstens & Diercks (2013:sect. 2.3) point out that ‘how’ is likewise typically placed in clause-final position but can also be located immediately to the right of the finite verb. The clausal parataxis analysis accommodates this parallel between Luhya and Germanic straightforwardly: expressions that are paratactically construed with the main proposition are often used as parentheticals (cf. (9)). (9)

This claim is — (and) let me be perfectly clear about this — complete and utter nonsense.

Clausal parataxis – and how! 5

6. Verifying the clausal parataxis approach to Luhya agreeing ‘how’ So far, we have constructed a plausibility argument for the view that Luhya agreeing ‘how’-questions involve asyndetic parataxis of two clauses. Now let us ask how this clausal parataxis approach holds up in the face of the two main empirical puzzles presented by agreeing ‘how’-questions: the distribution of ‘alternative agreement’, and the variable agreement pattern of locative inversion constructions. It will turn out that in every respect, the parataxis analysis delivers the desired results. 6.1.

Wh-subjects and alternative agreement

In Lubukusu (as in many other Bantu languages), constructions in which a third-person singular animate subject undergoes wh-movement see the usual subject agreement prefix on the verb obligatorily replaced with oo-: (10). When a ‘how’-question is added to a wh-question whose subject is an ootrigger, ‘how’ itself is not inflected with oo-; it appears with the regular aprefix instead: (11). (10)

a. b.

(11)

a. b.

Naliaka a-li mu-nju. Naliaka 3SG-be 18-house

Naanu a-li mu-nju?

1who 3SG-be 18-house ‘Who is in the house?’

b′.

Naanu oo-li mu-nju? 1who AAE-be 18-house ‘Who is in the house?’

Naanu oo-tekh-ile e-ngokho *o-/a-riena? 1who AAE-cook-PST 9-chicken *AAE-/3SG-how ‘Who cooked the chicken how?’ Ba-a-bona o-mu-ndu ow-a-tekha e-ngokho 2SG-PST-see 1-1-person AAE-PST-cook 9-chicken

o-/a-riena?
aae-/3sg-how

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

For Carstens & Diercks’s (2013) analysis of ‘how’-questions in Luhya, accounting for this calls for a technical exercise in uncoupling ‘how’-agreement from subject-verb agreement and from involvement with T. In the clausal parataxis approach, by contrast, nothing special needs to be said about the absence of the AAE on ‘how’ in (11): this falls out immediately from the fact that the subject of the ‘how’-clause is a non-wh pro linked to the wh-subject of the preceding clause: The Even Yearbook 14 (2020), Department of English Linguistics, Eötvös Loránd University, Budapest ISSN 2061–490X, http://seas3.elte.hu/even, © 2020, Marcel den Dikken & Teresa O’Neill

Clausal parataxis – and how! 6

(11a′)

Who cooked the chicken, and how was (s)he/*who?

Agreement on ‘how’ is controlled by the non-wh pronoun in the second clause. Support for this approach comes from Kinande, where, instead of using a multiple wh-question, speakers prefer to render the ‘who and how’ question by constructing two separate full clauses, where ‘how’ agrees with a non-wh pronominal subject. (12)

Iyóndi yó w-aly’á e-nyáma? A-alyá y-á-ti? 1who LNK AAE-eat 9-meat? 3SG-eat LNK-3SG-how? ‘Who ate the meat? How did he eat (it)?

The Kinande analogue of a Lubukusu multiple wh-question thus provides a fully spelled out source for the person agreement we observe in Lubukusu ‘how’ questions with operator subjects. 6.2.

Agreeing ‘how’ and locative inversion

A good portion of Carstens & Diercks’s (2013) paper is devoted to a discussion of the agreement behavior of locative inversion (LI) constructions in Luhya (on which, see Bresnan 1994, Bresnan & Kanerva 1989) — with particular reference to the agreement patterns they evince in ‘how’-questions. This discussion, though very interesting, ultimately lends no support for their approach to agreeing ‘how’. As a matter of fact, the behaviour of agreeing ‘how’ in the context of locative inversion rather supports a biclausal analysis. In LI-constructions, agreement between ‘how’ and the thematic subject, rather than the fronted locative, is pervasive in Lubukusu, while Kinande prefers agreement between ‘how’ and the fronted locative. The clausal parataxis analysis can exploit independently known parametric variation in the syntax of LI to capture this: whichever element — the thematic subject or the fronted locative — occupies the SpecTP position in the initial clause is picked to antecede the pro controlling agreement with ‘how’ in the second clause. In Carstens & Diercks’s (2013) approach, where the thematic subject agrees with ‘how’ from its base position in the vP domain, such variation is not predicted. Diercks (2011) establishes that Lubukusu has two types of locative inversion (distributed primarily based on the (un)ergativity of the verb): • •

One type gives rise to verb agreement with the fronted locative – as in (13a). (The indices in (13) mark the φ-feature agreement relations.) The other type has the verb agreeing with the postverbal thematic subject – as in (13b).

Clausal parataxis – and how! 7

(13)

a. b.

LOCi Vi DPk LOCi Vk DPk

The complex agreement pattern emerging in Lubukusu ‘how’-questions with LI is summarised in the table below (copied over from Carstens & Diercks 2013).

(13a) (13b) While (13b)-type LI behaves uniformly (only allowing ‘how’ to agree with the thematic subject),3 in (13a)-type LI, where the verb agrees with the fronted locative, there is speaker variation with respect to what ‘how’ agrees with: • Variety A still forces ‘how’ in (13a) to agree with the thematic subject, as in (13b)-type LI; • Variety B allows ‘how’ in (13a) to agree with the thematic subject or with the fronted locative. A monoclausal analysis of ‘how’-questions does not yield easily to a split agreement pattern of the type allowed in Variety B with (13b)-type LI: one would expect, a priori, that ‘how’ would pick the same constituent to agree with as does T. Carstens & Diercks (2013) present a detailed technical discussion of how the split agreement pattern allowed by Lubukusu-B speakers can come about in their approach to ‘how’-questions with LI. There is no space here to go over the details. But at any rate, the theoretical complications that this split agreement pattern calls for on their monoclausal approach to agreeing ‘how’3 This can be mimicked in English, as in (i): (i) a. Up the hill and down the valley ran the athlete; how was she? b.

Up the hill and down the valley ran the athlete; how were they?

Although the conjunction up the hill and down the valley can serve as the antecedent for they in a different context (I can’t choose between up the hill and down the valley: they both seem to be good options), (ib) is impossible: in English LI-constructions with a paratactic how-question, we always get the thematic subject to control agreement in the how-question.

Clausal parataxis – and how! 8

questions are unnecessary on a biclausal approach of the type that we advocate, which has the additional advantage of exporting beyond Lubukusu. To understand what is behind the agreement patterns in ‘how’-questions featuring LI, it is important to widen the empirical scope of the investigation, beyond Lubukusu, and to translate this bigger picture into an analytical perspective on the position occupied by the fronted locative in LIconstructions. It is particularly instructive to draw the facts of Kinande LI (not discussed by Carstens & Diercks 2013) into the picture. Kinande saliently differs from Lubukusu in the realm of LI in two ways. (i) Kinande LI always features agreement between the verb and the fronted locative, never with the thematic subject: (14). (14)

a. b.

O-ko-kitwá mó-kw-ákúmbagala e-ri-bwê. IV-17-hill TNS-17-roll IV-5-stone ‘Down the hill rolled a stone.’

O-ko-kitwâ mó-ry-ákúmbagala e-ri-bwê.

IV-17-hill TNS-5-roll IV-5-stone

(ii) In Kinande LI-constructions, ‘how’ or ‘thus’ always agrees with the fronted locative; agreement with the thematic subject is impossible: (15). (15)

a. b.

O-ko-kitwá mó-kw-ákúmbagala

IV-17-hill TNS-17-roll O-ko-kitwá mó-kw-ákúmbagala IV-17-hill TNS-17-roll ‘Down the hill rolled a stone thus.’

ri-tya e-ri-bwê. 5-thus IV-5-stone ku-tya e-ri-bwê. 17-thus IV-5-stone

The difference between Lubukusu (both varieties) and Kinande in the realm of ‘how’/‘thus’ agreement in LI-constructions (i.e., the inability of the thematic subject in Kinande to control agreement with the modifier, an agreement pattern that is always possible in Lubukusu), is apparently correlated with the fact that in Kinande LI-constructions the fronted locative systematically controls agreement with the verb (i.e., there is no split of the (13a/b) type in Kinande LI). The fact that Kinande fronted locatives always control agreement with the finite verb can easily be understood if in this language LI systematically involves A-movement of the locative to SpecTP, creating a configuration of the type in (15),4 where the locative is the closest element in an A-position 4

In the structures in (16)–(18), coindexation is used both for antecedence and agreement.

Clausal parataxis – and how! 9

with which the verb can agree. Being the highest element in an A-position in Clause 1, the locative in SpecTP in (15) also serves as the antecedent for the pro-subject of Clause 2.5 The locative’s φ-features thus get copied onto ‘how’, and all is as it should be, for Kinande. (16)

[Clause1 [TP [LOCi]A [Tʹ Ti … [θ-subjectk]A … ei]]] [Clause2 proi/*k how]

In Lubukusu, there is variation both in the syntax of LI and in the way in which ‘how’ agrees in LI-constructions. The fact that the thematic subject agrees with the verb in Lubukusu LI-constructions of type (13b) cannot be derived from a syntax in which the locative A-moves to SpecTP and becomes the subject of the clause. Instead, the locative in (13b) occupies an Ā-position in the left periphery of the clause. From this Ā-position, the locative itself cannot agree with the verb of its clause, and it cannot identify the pro-subject of the paratactic ‘how’ clause either, on the plausible assumption that (except in the case of pro qua bound-variable pronoun, which we are not dealing with here) pro can only be identified by an antecedent in an A-position. The thematic subject in (16) is the highest element in an A-position in Clause 1, so in Lubukusu LI-constructions of type (13b), the thematic subject systematically controls agreement with ‘how’, as desired.6 (17)

[Clause1 LOCi]Ā [TP Tk … [θ-subjectk]A … ei] [Clause2 prok/*i how]

What now remains to be accounted for is the fact that for (13a)-type LIconstructions, there is partial variation among Lubukusu speakers on what ‘how’ agrees with: while speakers are unanimous in allowing ‘how’ to agree with the thematic subject, speakers of Variety B accept agreement between ‘how’ and the fronted locative as an alternative. This variation rests on how exactly the verb establishes agreement with the fronted locative in an LIconstruction of type (13a). There are two ways, in principle, in which such agreement can come about. One is via the Kinande strategy of raising the locative physically into SpecTP, as in (16). This strategy, as we have seen, 5

Though pro-subject of Clause 2 of clausal parataxis structures can never achieve the ideal of having a strictly c-commanding antecedent, what the grammar will settle for is an antecedent in an A-position that comes as close to c-commanding pro as possible. The locative in SpecTP of Clause 1 in (16) ‘almost c-commands’ (Hornstein 1995:108–10) pro in Clause 2. 6 The SpecTP position of LI-constructions of type (13b) remains unprojected, keeping the thematic subject within the complement of T. Projecting SpecTP and raising the thematic subject into it in the structure in (16) would be grammatical but would not result in a locative inversion construction: we would then be dealing a case of ‘plain’ PP topicalisation. Since the discussion in this section concerns itself with the behaviour of agreeing ‘how’ in Luhya locative inversion constructions, we will set this derivation aside. The Even Yearbook 14 (2020), Department of English Linguistics, Eötvös Loránd University, Budapest ISSN 2061–490X, http://seas3.elte.hu/even, © 2020, Marcel den Dikken & Teresa O’Neill

Clausal parataxis – and how! 10

leads to ‘how’ agreeing with the locative. This is something that Lubukusu-B speakers find agreeable for type (13a) LI. But it is not the only strategy that these speakers have at their disposal for forming (13a)-type LI. In addition, speakers of Variety B also allow a derivation (which is the only one that Lubukusu-A speakers accept for (13a)) that is more similar to that of LI-constructions of type (13b) (given in (17)). In the syntax in (18), the fronted locative sits in an Ā-position in the left periphery, just as in (13b)-type LI; but the twist that is added is that the clause-peripheral locative is now associated with a silent expletive in SpecTP.7 (18)

[Clause1 LOCi]Ā [TP [eci]A [Tʹ Ti … [θ-subjectk]A … ei]] [Clause2 prok/*i how]

The silent category in SpecTP (see Buell 2007 for Bantu, and also Den Dikken 2006 for some cases of LI in Germanic: cf. the discussion of (21), below) is coindexed with the physical locative, and agrees with the verb of its clause, thereby yielding the agreement pattern in (13a). But as ec is an expletive, it cannot identify the pro in the subject position of the paratactic second clause of the ‘how’-question. Since the overt locative, occupying an Ā-position, cannot identify this pro either, the only possible identifier for pro in is the notional subject.8 For (13a)-type LI constructions analysed as in (19), therefore, we derive the peculiar split between verb agreement and ‘how’-agreement evinced by Lubukusu speakers of both varieties. For Kinande speakers, (18) does not come into the picture: in this language, it is always the physical locative itself (not a silent proform linked to it) that takes the SpecTP position. Concomitantly, in Kinande agreeing ‘how’ constructions featuring LI, we only find agreement between ‘how’ and the locative. For their (13a)-type LI, Lubukusu-B speakers agree with Kinande speakers in taking the locative to have an association with SpecTP. They can establish this association exactly as in Kinande (i.e., via (16)); but alternatively, cognizant of the fact that the fronted locative in LI-constructions of the type in (13b) feature the locative in an Ā-position (see (17)), they can also resort to the syntax in (18). This syntax shares with (16) the fact that the locative has an association with SpecTP (albeit, in the case of (18), an indirect one, established via ec); but it shares with (17) the fact that the fronted locative physically finds itself in the Ā-periphery. Lubukusu-A does not bother with (16) at all: for speakers of this variety, the locative is always in an 7 8

Being a pro-drop language, Lubukusu has no overt expletive to insert in SpecTP. The thematic subject of (13b)-type LI-constructions can actually end up in SpecTP at LF, as a result of expletive replacement. This may manoeuvre the thematic subject into a high A-position from which it can ‘almost c-command’ pro.

Clausal parataxis – and how! 11

Ā-position in the left periphery; but there still is a difference between (13a) and (13b) for these speakers, coming down to whether or not there is an association of the locative with SpecTP. For agreement between the locative and the verb, (16) and (18) both lead to (13a); but for agreement with ‘how’, the two derivations have different outputs: (16) gives rise to ‘how’ agreeing with the locative (systematically in Kinande, and optionally in Lubukusu-B) while (18) delivers agreement between ‘how’ and the thematic subject (disallowed in Kinande, and grammatical for all Lubukusu speakers). The variation seen in Bantu LI with respect to whether the locative fronts SpecTP, associates with a null expletive in SpecTP from a higher Ā-position, or simply sits in the Ā-periphery finds a close match in Germanic. That English and other Germanic languages can front predicates to SpecTP is well known from the existence of copular inversion (19a): that my best friend is in an A-position and not in the left periphery is clear from the grammaticality of copular inversion in ECM contexts, as shown in (19b). (19)

a. b.

My best friend is John. I believe my best friend to be John.

In English locative inversion constructions such as (20a), on the other hand, the fronted locative is unquestionably not in SpecTP: the ungrammaticality of (20b) would be difficult to account for if it were. The locative in (20a) is in the Ā-periphery, not associated with a silent expletive in SpecTP: the postverbal thematic subject here is definite, and from expletive there (which is the natural overt counterpart to a putative silent expletive in SpecTP) we know that it only accepts indefinite associates. Thus, it is likely that the syntax of (20a) mimics (17). (20)

a. b.

On this wall (*there) hangs the Mona Lisa.

I believe on this wall to hang the Mona Lisa.

When the postverbal thematic subject is indefinite, an overt expletive becomes possible — though by no means obligatory. The version of (21a) without overt there allows, in principle, for two different parses, one as in (17) and the other featuring a silent expletive in SpecTP. This latter parse is parallel to (18): the locative itself is still clearly in an Ā-position (see (21b)), but there is room for a silent expletive in SpecTP.

Clausal parataxis – and how! 12

(21)

a. b.

On this wall (there) hangs a priceless painting.

I believe on this wall to hang a priceless painting.

Finally, in so-called ‘beheaded locative inversion’ constructions such as (22a), a locative predicate is arguably raised to SpecTP (though the head (to) of the PP is stranded in clause-internal position; see Den Dikken 2006 for discussion of this and other predicate inversion constructions), as in (16). Unlike in (20) and (21), the Ā-periphery is not active in ‘beheaded locative inversion’, as is evident from the grammaticality of embedding in ECM contexts (22b). (22)

a. b.

This question has been paid too little attention to. I believe this question to have been paid too little attention to.

So we see that the parataxis approach to agreeing ‘how’-questions can exploit independently known parametric variation in the syntax of locative inversion to capture the variation in the agreement target for ‘how’. Agreement in the LI-clause depends on whether the fronted locative occupies or is associated with the SpecTP position; agreement with ‘how’ depends on what is the highest meaningful element in an A-position in the LI-clause. Since in general, the ‘how’/‘thus’ facts of Kinande are quite similar to Lubukusu, and since the two languages are closely related, the fact that the parataxis approach makes it possible to paint a precise and accurate picture of the variation regarding agreement with ‘how’ in LI-constructions, both within Lubukusu and between Lubukusu and Kinande, is an important asset of this analysis. Carstens & Diercks’s (2013) monoclausal approach, by contrast, cannot be extended to Kinande. For Carstens & Diercks, the preference should always be for the thematic subject of an LI-construction to control agreement with ‘how’ from its base position. Because it cannot exclude agreement between ‘how’ and the thematic subject agreement in LI constructions, the fact that such agreement is not an option in Kinande stands out as a problem for this monoclausal analysis.

7. Two notes on the interpretation of agreeing ‘how’-questions 7.1.

Depictive secondary predication

Carstens & Diercks (2013) briefly entertain an analysis that is superficially similar to ours, featuring ‘how’ as a depictive secondary predicate in a vPadjoined small clause with a null subject. They reject this analysis, primarily on account of the fact that although the agreeing ‘how’-question in (23) supports quite a wide range of interpretations, the subject-oriented depictive interpretation in (23d) is strikingly unacceptable. If ‘how’ were a depictive The Even Yearbook 14 (2020), Department of English Linguistics, Eötvös Loránd University, Budapest ISSN 2061–490X, http://seas3.elte.hu/even, © 2020, Marcel den Dikken & Teresa O’Neill

Clausal parataxis – and how! 13

predicated of a PRO controlled by the subject, then the opposite pattern would be expected, with the subject-oriented answer being preferred over the others. (23) a. b. c. d.

A-li-le e-nyama a-riena? (Carstens & Diercks: (22)) 3SG-eat-PST 9-meat 3SG-how ‘How did he eat the meat?’ Kalaa. Manner answer slowly ‘Slowly.’ Nende si-chiko. Instrument answer with 7-spoon ‘With a spoon.’ E-mbisi. Object-depictive answer 9-raw ‘Raw.’

A-li-le ne-a-mel-ile. Subject-depictive answer 3SG-eat-PST NE-3SG-be.drunk-PST

‘He ate it (while he was) drunk.’

In addition, unlike bona fide depictive secondary predicates, ‘how’ cannot be predicated of a PRO controlled by the object and retain the same interpretation. When ‘how’ agrees with the object, rather than eliciting an objectoriented depictive or resultative answer, it loses its ‘how’ interpretation and is instead interpreted as ‘what kind’: (24). Carstens & Diercks take these facts as evidence that ‘how’ is not a depictive secondary predicate. (24)

Ba-khal-ile lu-karatasi lu-riena? (Carstens & Diercks: (26)) 2SG-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)

At first blush, these facts may seem equally incompatible with our proposed alternative, where ‘how’ is the predicate of a reduced clause whose subject is anteceded by the main clause subject. However, the paratactic model escapes the most fatal problems of a depictive approach by establishing a much looser syntactic relationship between the main clause subject and the paratactic clause subject. The subject of our proposed paratactic clause is pro, rather than the controlled PRO of a depictive. Instead of treating the interpretive relationship between ‘how’ and the main clause subject as secondary predication, the present analysis models the interpretive relationship as asyndetic specification between two conjuncts: the main clause and the interrogative clause. As such, the range of interpretations of the ‘how’ question is The Even Yearbook 14 (2020), Department of English Linguistics, Eötvös Loránd University, Budapest ISSN 2061–490X, http://seas3.elte.hu/even, © 2020, Marcel den Dikken & Teresa O’Neill

Clausal parataxis – and how! 14

broad, and, as Carstens & Diercks observe, does not covary with the argument structure of the main clause. Neither Carstens & Diercks’s proposal — adjunction of ‘how’ to vP — nor ours can perfectly capture the restriction on subject-oriented answers. Here, we can only offer a suggestion that may lead us towards a solution. Adjectives form a closed lexical class in Luhya, and depictives have a particularly restricted distribution. In the unacceptable subject-oriented example in (23e) above, the depictive is not rendered by a plain adjective, but rather, by a ‘while’-clause. We observe that the overt paraphrase a subject-oriented ‘while’ answer to a ‘how’ question is similarly unacceptable in English: (26)

How did she eat the meat?

While she was drunk.

The unavailability of the subject-oriented answer therefore finds a potential explanation that is entirely independent of (and therefore does not adjudicate on) the difference between Carstens & Diercks’s (2013) monoclausal approach to agreeing ‘how’ and our biclausal analysis. 7.2.

Scope of adverbial modification

A novel prediction of the biclausal approach to Luhya ‘how’-questions is that it should be possible to modify the main clause and the ‘how’-clause separately. This prediction is tested in (27) with the declarative counterpart of ‘how’: (27)

Naliaka yesi a-tekh-ile wakana a-rio. Naliaka certainly 3SG-cook-PST perhaps 3SG-thus ‘Naliaka certainly cooked — perhaps thus.’

The availability of independent modification by a sentential adverb like wakana ‘perhaps’ is expected if ‘thus’ occupies a separate clause. If agreeing ‘thus’ and ‘how’ were adverbial adjuncts, as in Carstens & Diercks’s (2013) approach, we would not expect a sentential modifier to be acceptable: as a vP/VP adjunct, ‘how’/‘thus’ should always fall in the scope of any sentenceor speaker-oriented adverbs, turning (27) into a contradiction (with yesi ‘certainly’ and wakana ‘perhaps’ clashing head-on; cf. English *Naliaka cooked certainly perhaps thus). The grammaticality of (27) thus provides further support for the clausal parataxis approach to agreeing ‘how’/‘thus’. Independent sentential modification and independent illocutionary force are hallmarks of parenthesis (Quirk et al. 1985), analysable as clausal parataxis (Koster 2000; De Vries 2009; Kluck 2011; Heringa 2012; Ott & De

Clausal parataxis – and how! 15

Vries 2016). Agreeing ‘how’/‘thus’ thus fits under the larger umbrella of paratactic constructions, along with Germanic exclamative ‘and how’ (section 3).

8. Conclusion A clause that is paratactically juxtaposed to another can provide information modifying the proposition expressed by the other clause. The paratactic manner clause can be agentive and feature an adverbial modifier (see (28a)), but it can also be copular, with the modifier as the predicate of the subject (as in (28b)). In both cases the manner clause can be a question, yielding (28a′,b′). (28)

a. a′. b. b′.

He danced at the party — He did great./He did so wildly. He danced at the party — How did he do? He danced at the party — He was great/wild. He danced at the party — How was he?

The main contention of this paper is that sentences in Bantu in which ‘how’ is inflected for the φ-features of the subject feature ‘how’ as the predicate of a silent pronoun linked to the occupant of the SpecTP position in the antecedent clause. Viewed this way, Bantu agreeing ‘how’ is syntactically very much like English parataxis constructions of the type in (28b′). An approach to agreeing ‘how’ along these lines procures an account of the φ-agreement facts (including the complex variation found in ‘how’questions featuring locative inversion, not just within Lubukusu but between Lubukusu and Kinande as well) without requiring any modifications to the technicalities of the Agree relation. Moreover, this analysis makes novel predictions (regarding the possibility of an overt element linking agreeing ‘how’ to the preceding clause (section 4) and adverbial modification (section 7.2)) that are borne out by the facts. References Bresnan, Joan. 1994. Locative inversion and the architecture of universal grammar. Language 70.1:72–131. https://doi.org/10.2307/416741 Bresnan, Joan, and Jonni M. Kanerva. 1989. Locative inversion in Chicheŵa: a case study of factorization in grammar. Linguistic Inquiry 21.1:1–50. https://www.jstor.org/stable/4178612 Buell, Leston. 2007. Semantic and formal locatives: Implications for the Bantu locative inversion typology. SOAS Working Papers in Linguistics 15:105–120. https://www.researchgate.net/publication/251621048

Clausal parataxis – and how! 16 Carstens, Vicki, and Michael Diercks. 2013. Agreeing How? Implications for theories of agreement and locality. Linguistic Inquiry 44.2:179–237. https://doi.org/10.1162/LING_a_00125 Diercks, Michael. 2011. The morphosyntax of Lubukusu locative inversion and the parameterization of Agree. Lingua 121:702–720. https://doi.org/10.1016/j.lingua.2010.11.003 Dikken, Marcel den. 2006. Relators and linkers: The syntax of predication, predicate inversion, and copulas. Cambridge, MA: MIT Press. https://doi.org/10.7551/mitpress/5873.001.0001 Heringa, Herman. 2012. Appositional constructions. Utrecht: LOT. https://www.rug.nl/research/portal/files/2449766/Thesis_appositional_constructi_1.pdf Hornstein, Norbert. 1995. Logical Form. Oxford: Blackwell. Kluck, Marlies. 2011. Sentence amalgamation. Utrecht: LOT. https://www.lotpublications.nl/Documents/285_fulltext.pdf Koster, Jan. 2000. Extraposition as parallel construal. Ms., University of Groningen. Marten, Lutz. 2006. Locative inversion in Otjiherero: More on morpho-syntactic variation in Bantu. ZAS Papers in Linguistics 43:97–122. https://www.researchgate.net/publication/242733558 Schneider-Zioga, Patricia. 2015. The linker in Kinande re-examined. Selected Proceedings of the 44rd Annual Conference on African Linguistics (ACAL). https://www.researchgate.net/publication/265600408 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. https://ecommons.cornell.edu/bitstream/handle/1813/4029/wasike%20dissertation.pdf? sequence=1&isAllowed=y Corresponding author: Marcel den Dikken Department of English Linguistics, Eötvös Loránd University & Research Institute for Linguistics marcel.den.dikken@nytud.hu