Distribution of conjunctive and disjunctive forms in Xitsonga

The conjunctive and disjunctive forms in Xitsonga are examined with the purpose of presenting the distribution of these forms. While verbs in the conjunctive form are followed by some elements, the disjunctive is used when no element follows the verb. Xitsonga follows these basic patterns observed in other Bantu languages, but previous theories cannot explain all of the cases in which the conjunctive and disjunctive forms are distinguished. In previous work, three major approaches have been proposed: the constituency approach, the focus-based approach, and the information packaging approach. Xitsonga supports but also provides counterevidence to all of these approaches. This paper also re-examines the claim that the presence of conjunctive/disjunctive distinctions only exists in the present tense. Following Creissels (2014), we report that the conjunctive/disjunctive dichotomy is present in other tenses as well when prosodic patterns such as penultimate lengthening are further examined.


Introduction
The goal of this paper is to present a comprehensive picture of disjunctive and conjunctive forms in Xitsonga main clauses, which are an under-investigated phenomenon in Xitsonga.The paper will also report phonological patterns that distinguish these forms that are otherwise morphologically identical.
The disjunctive and conjunctive forms have been referred to by various names, as in (1).In descriptive work on Bantu languages, the forms are called "long form" and "short form", based on how the morpheme is realised.In Sharman (1956), whose work is concerned with prosody, the two forms were called "weak link" and "strong link", respectively.With reference to information structural status, Givón distinguishes the forms in terms of types of focus.
(1) Disjunctive (dj) vs. conjunctive (cj) forms (based on Hyman & Watters 1984:251 In Xitsonga, the distinction between the disjunctive (dj) and conjunctive (cj) forms appears in the present positive.In terms of distribution, the cj form cannot appear in the clause-final position, while the dj form can appear in this position.Unlike in Zulu, various types of syntactic constituents can satisfy conditions for the cj form in Xitsonga.In other tenses, the morphological distinction between cj and dj forms is neutralised.Recent work on Setswana (Creissels 2014) argues that tonal distinctions are maintained in other tenses that are not present positive.We follow Creissels' observation and will show that a similar distribution is also found in Xitsonga.
The dj form is expressed with the -amorpheme in Xitsonga.Due to the avoidance of vowel hiatus, there is vowel deletion (/ i + a / → [a]), glide formation (/ u + a / → [wa]), or vowel fusion (/ a + a / → [a] without lengthening).In personal pronouns, the cj forms and the dj forms are as shown in (3a) and (3b), respectively.
( Three major approaches aim to explain the distribution of cj versus dj forms in other Bantu languages.Van der Spuy (1993), Buell (2006), andHalpert (2016) propose a constituencybased approach, in which the dj form is argued to be realised when a verb lacks a complement.Drawing on Setswana, however, Creissels (1996) and subsequently Güldemann (2003) argue that the distribution of cj vs. dj forms is best to be analysed as a focus-driven phenomenon.
Updating the focus-based approach, Creissels (2014) argues for an information packaging approach that relies on information theory to explain the cj/dj distribution.The last approach will also be defended in this paper.
The rest of the paper will examine each of these approaches in sections 2, 3, and 4, respectively.The non-morphological distinction of cj/dj forms in other tenses, where penultimate lengthening and tone show a non-segmental distinction, will be presented in section 5.

The constituency-based approach
The constituency-based approach states that the cj form is used "when a verb is followed by complements within IP", while the dj form appears when "a verb is IP final" (van der Spuy 1993) (see also du Plessis, Nxumalo andVisser 1995 andBuell 2006).

Data in support of the constituency-based approach
Xitsonga has much data that support the constituency-based approach.As a main diagnostic, the cj form does not appear clause-finally (4b), but the dj form can appear in the clause-final position (5a).When the dj form is used with a nominal complement (5b), the clause is interpreted as if the speaker is contradicting an earlier claim or assertion.In the rest of this paper, the addition of 'do' before a verb will indicate such a usage.An example corresponding to (5b) in Setswana is ungrammatical unless there is an object marker.
( The presence of the object pronoun is not sufficient; the cj form is still ungrammatical.The ungrammaticality of (7a) suggests that the cj form must be the rightmost element within an IP.
( The cj form can also be used when a locative phrase is present.The verb is not IP-final in all of the examples in ( 9).The cj form (as well as the dj form) can be used with low adverbs such as ngopfu 'much', swo tala 'a lot', and kahle 'well', as in ( 10).This pattern is different from isiZulu and Setswana, where these adverbs cannot be used with the dj form.
(11) The presence of a temporal adverb ú dyá námú:ntlhà (cj) wá: dyà nàmù:ntlhà (dj) 'He eats/does eat today' When an object NP is coupled with a co-referenced object prefix, however, it does not make the cj form grammatical, as shown in (12a).This means that the verb is clause-final, and the object NP is outside of that clause.The presence of an object prefix indicates that the object NP is not in its canonical position (Cheng & Downing 2009, Yoneda 2011 a.o.).As such, nyama in (12a) is seen as right-dislocated, which makes the verb dya clause-final; thus, the cj form is not allowed.
( The presence of object agreement in ( 14) shows that the cj form is no longer grammatical, but the dj form is. Thus, the verb is clause-final.Note that penultimate lengthening patterns differ between (13b) and (14b).In (14b), the object prefix is lengthened, which furthermore suggests that the verb is clause-final.There is also a very short pause after the verb, which is not present in (13b).
( So far, the data in this section have shown that the distinction between cj and dj forms can be explained using the constituency-based approach.The following section will present data that are potentially problematic for this approach.

Potentially problematic data for the constituency-based approach
In the future tense and the perfect tense, there is no morphological contrast between the cj form and the dj form, as shown in (18b,d,f,h).If the verb were clause-final, we would not expect the ungrammaticality of these examples.This contrast in the distribution of the dj form has also been noted in Baumbach (1987:221).We will revisit this issue in section 5. (

Summary
So far, we have seen the cj and dj forms in Xitsonga that support the constituency-based approach.In Xitsonga, the cj form cannot be used when a verb is in the clause-final position.
The cj form can be used regardless of whether the following constituent is an argument or an adjunct.In non-present tenses, the cj form must be used, even when the verb is in the clausefinal position.
The basic patterns of cj versus dj will be extended in the following section, where the distinction between the cj form and the dj form is taken to be focus-driven.Two competing hypotheses will be compared.

3.
Focus-based approaches 3.1 Two hypotheses Buell (2006) presents two hypotheses from earlier studies on the cj/dj distinction.The Postverbal Term Focus hypothesis is based on a study on Setswana (Creissels 1996).Under this hypothesis, the element following a cj form is in focus, while the element following a dj form is not in focus.As such, if the cj form is not followed by any other elements, the absence of focused elements is the source of ungrammaticality (see also Ndayiragije (1999) on Rundi and Sabel and Zeller (2006) on isiZulu wh-questions).
A competing focus-based approach is the Verb Focus hypothesis proposed in Güldemann (2003) and based on a study of isiZulu.Under this hypothesis, it is the verb itself that is in focus when the verb is in its dj form, while the verb is not in focus in its cj form.
The following subsections will examine Xitsonga counterparts to the isiZulu datasets explored in Buell (2006).As shown in Buell (2006), Xitsonga has both examples that support the focusbased approaches and examples that raise questions about these approaches.

3.2
The Postverbal Term Focus hypothesis

In support of the Postverbal Term Focus hypothesis
The response (R) to the wh-question (Q) in (20a) shows an example of postverbal focus.The postverbal adverb ehandle 'outside' is focused, and only the cj form is allowed (see also van der Wal (2014) for morphosyntactic encoding of focus in Makhuwa and other Bantu languages).The dj form is not a possible response to the wh-question (20b).
'He eats OUTSIDE' (cj) b. #wá: dyà éhà:ndlé (not a possible response to Q) (dj) It is also the case that when the postverbal subject is the logical subject of the verb, the dj form is not allowed, as shown in (21b). (

Potential problems for the Postverbal Term Focus hypothesis
As in Buell's (2006) work on isiZulu, when other contexts are examined, Xitsonga also has examples that are problematic for the Postverbal Term Focus hypothesis.In (23a), the neutralcontext question does not require an answer with focus.Even so, the cj form is required as a response (23b).The dj form (23c) is not a salient response.Resumptive pronouns are required in Xitsonga relative clauses.In (24c), the resumptive pronoun rona is used.A resumptive pronoun is not a focused element, but it is the cj form that is required in relative clauses.Moreover, the dj form is ungrammatical.In this sense, Xitsonga patterns with isiZulu (Buell 2006:18).

3.3
The Verb Focus hypothesis

In support of the Verb Focus hypothesis
Under the Verb Focus hypothesis, it is the verb itself that is argued to be in focus when the verb is in its dj form, while the verb is not in focus in its cj form (i.e.elements other than the verb are in focus).As shown in ( 26), the dj form has two meanings: one focused; one non-focused.
(26) The dj form with a verbal focus a. ú tlángá éhà:ndlé 'He plays outside' (cj) b. wá tlá:ngà éhà:ndlé 'He plays/does play outside' In Xitsonga, kahle 'well' can co-occur with the dj form, which suggests that the preceding verb is focused (27b).In isiZulu, sentences with kahle 'well' must have the cj form (cf. Buell 2006:18-19).Contrastive focus on the verb is also possible when sentences occur with kahle.While the cj form in (28a) means that the person sings better than he dances, in (28b), there is an additional focal meaning that emphasises the event of singing itself.

A potential problem for the Verb Focus hypothesis
Buell ( 2006) also identifies a potential problem for the Verb Focus hypothesis.If the dj form signifies the verb focus, the cj form should not be allowed in verb focus cases.In (29), a relative resumptive pronoun is used, which suggests that the focused element in the sentence must be the verb nyíkílé 'gave'.Contrary to the prediction of the Verb Focus hypothesis, the cj form is used in (29a).Moreover, the dj form is ungrammatical.

Summary
An examination of focus-based approaches in light of Xitsonga data shows that some but not all distributional facts of the cj/dj forms can be accounted for.As Buell (2006) points out, the examples involving resumptive pronouns as well as the examples in non-focused contexts are non-trivial cases that require further explanation.

Information packaging
In a recent study, Creissels (2014:10) proposes that the cj/dj distinction can be best accounted for by information packaging theory.The cj form is used whenever the comment or verb phrase (VP) includes at least one element other than the verb itself, which implies that a cj verb form is followed by at least one phrase forming part of the comment, since the VP is strictly headinitial.The dj form is used whenever the comment/VP includes no element other than the verb itself, which implies that a dj verb form can only be followed by extraposed phrases that do not form part of the comment.This proposal is a departure from the structure-based approaches (section 2) and the focus-based approaches (section 3).

4.2
In support of the information packaging approach Creissels (2014: 8) suggests a diagnostic for separating the cj form from the dj form. 2 The naconstruction has two meanings: 'with' and 'as well'.The distribution is shown in (30).The 'with' meaning of na can be used with both the cj and the dj form, whereas the 'as well' meaning can only be used with the dj form.There is also a difference in penultimate lengthening.The main verb shows penultimate lengthening in the dj form, but not in the cj form (see section 5 for more discussion). (

A potential problem for the information packaging approach
The information packaging approach predicts that only the cj form should be used with adverbs such as ngopfu 'much', swo tala 'a lot', and kahle 'well', because these adverbs are part of the VP and they cannot be topicalised (á la Creissels 2014:12).In Xitsonga, the dj form appears with these adverbs as well, which requires some updates to the information packaging approach. (

The cj/dj distinction in other tenses: the role of prosody
As shown earlier, the dj form is not compatible with the perfect or the future tense.We maintain that the cj/dj distinction is still present, and we use penultimate lengthening as evidence of this distinction.In van der Spuy (1993:348-349), it has been reported that penultimate lengthening can be used to distinguish the cj and dj forms.Recall that van der Spuy's main proposal was based on syntactic constituency.
(35) Penultimate lengthening and the cj/dj forms a.The penultimate vowel of the dj form is obligatorily lengthened.b.The penultimate vowel of the cj form may not be lengthened.
Creissels (2014) also reports tonal distinctions between cj and dj in some tenses in Setswana.
In Xitsonga, penultimate lengthening also plays an important role in the distinction between the cj and the dj forms, but tonal distinctions play a relatively minor role.Examples in this section will use the Creissels test ('with him' vs. 'too') to determine the presence of the distinction between the cj and dj forms.

Perfect and past
There is no morphological distinction between the cj form and the dj form in the perfect and past tenses.If we adopt the Creissels' test with na, the penultimate lengthening of the verb, however, distinguishes the cj form from the dj form: penultimate lengthening in the dj form (36b,d) vs. no such lengthening in the cj form (36a,c).The present progressive in (44) shows the difference in penultimate lengthening as well.Due to the tonal spreading in Xitsonga, the tone of the verb's final syllable differs in the cj and the dj forms.The continuative present in ( 45) and ( 46) shows the same distribution. (

Present negative
Unlike the present positive tense, the present negative does not have a morphological dj form.The absence of the dj form does not mean that there is no distinction between the cj form and the dj form.The dj form has an additional penultimate lengthening on the verb (47b,d,f), while the cj form does not have such lengthening (47a,c,e).

Figure 1 :
Figure 1: Pitch track of examples (40a) and (40b) Examples of conjunctive and disjunctive forms in Xitsonga, isiZulu, Siswati, and Setswana are shown in (2).Descriptively speaking, the disjunctive form has a morpheme [(j)a] in all these languages, while the conjunctive form does not have this morpheme.
When the object NP is right-dislocated from its canonical position, an object prefix is obligatory.As shown in (15a,b), both cj and dj forms are grammatical, which suggests that the verb is not in the clause-final position.This is consistent with other examples(8)(9)(10)(11)in which adjuncts also license the appearance of the cj form.The pattern in (15a) contrasts with (7a), in which the cj form only has an object prefix in the absence of an NP complement.Moreover, the sentence in (15c) has the intended reading of a right-dislocated object NP in the absence of an adjunct.In such a case, the sentence is ungrammatical.The requirement of an adjunct occurring after the verb (compare (15a) with (15c)) further suggests that the verb must not be in the clausefinal position when it is in the cj form.In yes-no questions, the presence of the question morpheme blocks the cj form, which suggests that the verb is in the clause-final position.
18) No morphological contrast in the dj form in the future and perfect tense In the future tense, temporal adverbs can be fronted and only the cj form is allowed(19a,b).The morphological contrast is neutralised and the dj form causes ungrammaticality, as in (19c).The same pattern is found in past tense sentences (19d-f).Note that this pattern is the opposite of the pattern observed in the present tense (19g-i).
The distribution of wh-words such as yini 'what' and rini 'when' is restricted to the postverbal position (22).As reported in Setswana(Creissels 2014:7), the use of wh-words requires the cj form of the verb.Xitsonga examples so far support the Postverbal Term Focus hypothesis.The cj form is required in examples where the element after the verb is focused.