Revisiting Negative Particle Questions in Sixian Hakka

Abstract

This study investigates the syntactic structure of negative particle questions (NPQs), also known as VP-NEG questions, in Sixian Hakka (a variety of Chinese spoken in Taiwan). We revisit the existing literature on Hakka NPQs, pointing out unresolved issues in previous analyses. Drawing on previous analysis of VP-NEG questions in Middle Chinese, we argue that the negator mo in Hakka NPQs has grammaticalized into a disjunctive head, as mo and the predicate do not show agreement. This proposal not only accounts for the syntactic properties of NPQs in Sixian Hakka but also addresses potential problems found in previous studies.

1. Introduction

The term ‘negative particle question (NPQ)’ was first used in L. L.-S. Cheng et al. (1996) to refer to the questions formed with a sentence-final negation marker, as shown in (1), an example of an NPQ in Mandarin Chinese.

(1)	Hufei	kan-wan-le	nei-ben	shu	meiyou?1
	Hufei	read-finish-ASP	that-CL	book	not-have
	‘Did Hufei finish reading that book?’ (L. L.-S. Cheng et al., 1997, p. 65)

This construction is also attested in other varieties of Chinese (such as Cantonese, Southern Min, and Hakka) as well as in Southeast Asian languages (such as Khmer, Thai, and Vietnamese) (see L. L.-S. Cheng et al., 1996, 1997; Y.-J. J. Hsieh, 2013; Duffield, 2013; Liao & Lin, 2024; among many others).

In this work, we focus on NPQs in Sixian Hakka, specifically, on VP-mo questions.2 Mo ‘not have’ in Sixian Hakka is an existential negator; it can appear before a nominal or a verb phrase, as shown in (2).3 In addition, it can also occur in the sentence-final position and function like a question particle, as illustrated in (3).

(2)	a.		Amin		mo		qien.
			Amin		not.have		money
			‘Amin doesn’t have money.’
	b.		Amin		mo		oi		hi	hog-gau.
			Amin		not.have		want		go	school
			‘Amin doesn’t want to go to school.’

(3)		Amin		oi		hi		hog-gau			mo?
		Amin		want		go		school			not.have
		‘Does Amin want to go to school?’

Inspired by Aldridge’s (2011) analysis of VP-NEG questions in Middle Chinese, we propose that VP-mo questions in Sixian Hakka exhibit a disjunctive question structure, as shown in (4). We argue that the negator mo in Hakka NPQs originally headed the second vP in the disjunctive phrase and has grammaticalized into a disjunctive head.

(4)

This proposal is based on the similarities between VP-mo questions in Sixian Hakka and earlier VP-fou questions in Middle Chinese. Moreover, the proposed analysis accounts for the phenomena observed in VP-mo questions, including their incompatibility with negative predicates and lack of agreement between mo and the predicates.

The remainder of the paper is organized as follows: In Section 2, we review the literature on VP-mo questions, including R.-F. Chung (2000) and Y.-J. J. Hsieh (2013), and point out some unresolved problems. In Section 3, we introduce two possible approaches to NPQs. Section 4 presents our proposal and illustrates how it accounts for the VP-mo questions in Sixian Hakka. Section 5 is the conclusion.

2. Previous Literature on VP-mo Questions

NPQs in Sixian Hakka have not received as much attention as NPQs in other languages (L. L.-S. Cheng et al., 1996, 1997; M.-L. Hsieh, 2001; Li, 2006; R.-H. Huang, 2008; among many others). The relevant literature on VP-mo questions that we have found includes R.-F. Chung (2000) and Y.-J. J. Hsieh (2013), which we review below.

R.-F. Chung (2000) provides a preliminary introduction to interrogative questions in Sixian Hakka, classifying VP-mo questions as particle questions formed with a sentence-final particle mo.4 Although Chung 2000 does not offer a syntactic analysis of VP-mo questions, he presents a set of examples that deserves our attention, as shown in (5). Chung suggests that a disjunctive question in Sixian Hakka (as in (5a)) can be reduced to a VP-mo question (as in (5c)).5

(5)	a.	Gi	iu	da	tien-fa	go-loi	ia	mo
		he	have	call	phone	come	or	not.have
		da	tien-fa	go-loi?
		call	phone	come
		‘Did he call or did he not call?’
	b.	Gi	iu	da	tien-fa	go-loi	ia	mo?
		he	have	call	phone	come	or	not.have
		‘Did he call or not?’
	c.	Gi	iu	da	tien-fa	go-loi	mo?
		he	have	call	phone	come	not.have
		‘Did he call?’ (R.-F. Chung, 2000, p. 160; glosses and translation ours)

Although the exact process of deletion is not specified in R.-F. Chung (2000), we assume this is what the author means: if the verb phrase following the negator mo in (5a) is deleted, the sentence becomes (5b); if the disjunctive head ia ‘or’ is subsequently removed from (5b), the result is (5c). However, this assumption encounters some issues. First, another set of examples provided by Chung in his work, as shown in (6), would cast doubt on a deletion analysis; it is unclear how examples (6a–b) transform into (6c), as the negator mo ‘not have’ does not appear in either (6a) or (6b) in the first place.

(6)	a.	Ng	oi	ten	ngai	siid fu	ia	m	ten	ngai	siid fu?
		you	want	with	I	suffer	or	not	with	I	suffer
		‘Do you like to suffer with me or do you not like to suffer with me?’
	b.	Ng	oi	ten	ngai	siid fu	ia	m?
		you	want	with	I	suffer	or	not
		‘Do you like to suffer with me or not?’
	c.	Ng	oi	ten	ngai	siid fu	mo?
		you	want	with	I	suffer	not.have
		‘Do you like to suffer with me?’ (R.-F. Chung, 2000, p. 159; glosses and translation ours)

Second, some VP-mo questions cannot be reconstructed, as shown in (7) and (8). If VP-mo questions were derived from disjunctive questions through deletion, it is puzzling as to why they cannot be reconstructed into disjunctive questions.

(7)	a.	Gi	zo-ded	hi	hog-gau	mo?
		he	can	go	school	not.have
		‘Can he go to school?’
	b.	*Gi	zo-ded	hi	hog-gau	ia	mo?
		he	can	go	school	or	not.have
	c.	*Gi	zo-ded	hi	hog-gau	ia	mo	zo-ded	hi
		he	can	go	school	or	not.have	can	go
		hog-gau?
		school

(8)	a.	Gi	voi	hi	hog-gau	mo?
		he	will	go	school	not.have
		‘Will he go to school?’
	b.	*Gi	voi	hi	hog-gau	ia	mo?
		he	will	go	school	or	not.have
	c.	*Gi	voi	hi	hog-gau	ia	mo	voi	hi	hog-gau?
		he	will	go	school	or	not.have	will	go	school

Now, let us turn to Y.-J. J. Hsieh (2013), which provides an early syntactic analysis of NPQs in Sixian Hakka. Hsieh points out the similarities between NPQs in Sixian Hakka and A-not-A questions in Mandarin, arguing that VP-mo questions in Sixian Hakka are a type of A-not-A question. An A-not-A question is a special type of question in Mandarin and other varieties of Chinese language characterized by the combination of both the affirmative form (A) and the negative form (not-A), as illustrated in (9).6 In (9), chi ‘eat’ represents A and bu chi ‘not eat’ represents not-A.

(9)	Ni	chi	bu	chi	niurou?
	you	eat	not	eat	beef
	‘Do you eat beef?’

Hsieh presents several similarities between NPQs in Sixian Hakka and A-not-A questions in Mandarin. First, neither A-not-A questions nor VP-mo questions can be answered with a simple positive/negative particle (i.e., yes or no), as shown in (10) and (11).

(10)	Mandarin A-not-A questions:
	A:	Xiaoming	ai	bu	ai	chi	qiao-ke-li?
		Xiaoming	love	not	love	eat	chocolate
		‘Does Xiaoming love to eat chocolate?’
	B:	Ai	/	Bu-ai	/	*Shi	/	*Bu-shi	/	*Dui.
		love		not-love		yes		no		correct
		‘Yes, Xiaoming does./No, Xiaoming does not.’ (Y.-J. J. Hsieh, 2013, pp. 27–28)

(11)	Hakka VP-mo questions:
	A:	Gi	voi	hi	hoggau	mo?
		he	will	go	school	not.have
		‘Will he go to school?’
	B:	Voi	/	M-voi	/	*He	/	*M-he	/	*Dui.
		will		not-will		yes		no		correct
		‘’Yes, he will./No, he will not.’ (Y.-J. J. Hsieh, 2013, p. 28)

Second, A-not-A questions and VP-mo questions are compatible with the adverb daodi ‘on earth; truly,’ which corresponds to dodi in Sixian Hakka; they are incompatible with the adverb nandao ‘actually,’ which corresponds to mosheng in Sixian Hakka. See (12) and (13).

(12)	Mandarin A-not-A questions:
	Ta	daodi	/ *nandao	yao	bu	yao	lai	gen	women	chi
	he	truly	actually	want	not	want	come	with	us	eat
	wancan?
	dinner
	‘Is he coming to eat with us or not?’ (Y.-J. J. Hsieh, 2013, p. 28)

(13)	Hakka VP-mo questions:
	Gi	dodi	/ *mosheng	voi	hi	hog-gau	mo?
	he	truly	actually	will	go	school	not.have
	‘Is he going to the school or not?’ (Y.-J. J. Hsieh, 2013, p. 29)

Third, it is possible to add a specialized sentence-final particle to both types of questions: Mandarin ne in (14) and Hakka no in (15), which might be roughly paraphrased into ‘so,’ function as attitudinal markers expressing the speaker’s concern about the situation being inquired into and/or their interest in the addressees’ responses to questions (cf. Paul, 2014; Pan, 2015; J.-M. Cheng, 2007; Fan, 2025).

(14)	Mandarin A-not-A questions:
	Ni	xihuan	bu	xihuan	zhe-ge	nu-sheng	ne?
	you	like	not	like	this-CL	girl	PRT
	‘So, do you like this girl?’ (Y.-J. J. Hsieh, 2013, p. 29, quoted with miner modification)

(15)	Hakka VP-mo questions:
	Gi	zung-i	ia-ge	se-moi-e	mo	no?
	he	like	this-CL	girl	not.have	PRT
	‘So, do you like this girl?’ (Y.-J. J. Hsieh, 2013, p. 30, quoted with miner modification)

Fourth, the A in A-not-A questions and the VP in VP-NEG questions can be a verb, an adjective, a modal, an adverb, or a preposition, as shown in (16) and (17).

(16)	Mandarin A-not-A questions:
	a.	Ni	he	bu	he	jiu?		[verb]
		you	drink	not	drink	wine
		‘Do you drink?’ (Y.-J. J. Hsieh, 2013, p. 30)
	b.	Zhe-ge	nu-sheng	piao-liang	bu	piao-liang?		[adjective]
		this-CL	girl	beautiful	not	beautiful
		‘Is this girl beautiful?’ (Y.-J. J. Hsieh, 2013, p. 31)
	c.	Ta	neng	bu	neng	he	jiu?	[modal]
		he	can	not	can	drink	wine
		‘Can he drink wine?’ (Y.-J. J. Hsieh, 2013, p. 31)
	d.	Ta	chang	bu	chang	mai	yifu?	[adverb]
		she	often	not	often	buy	clothes
		‘Does she often buy new clothes?’ (Y.-J. J. Hsieh, 2013, p. 32)
	e.	Ta	jin-tian	zai	bu	zai	jia?	[preposition]
		he	today	at	not	at	home
		‘Is he at home today?’ (Y.-J. J. Hsieh, 2013, p. 32)

(17)	Hakka VP-mo questions:
	a.	Ngi	siid	jiu	mo?		[verb]
		you	drink	wine	not.have
		‘Do you drink?’ (Y.-J. J. Hsieh, 2013, p. 31)
	b.	Ia-ge	se-moi-e	jiang	mo?		[adjective]
		this-CL	girl	beautiful	not.have
		‘Is this girl beautiful?’ (Y.-J. J. Hsieh, 2013, p. 31)
	c.	Gi	zo-ded	siid	jiu	mo?	[modal]
		he	can	drink	wine	not.have
		‘Can he drink wine?’ (Y.-J. J. Hsieh, 2013, p. 31)
	d.	Gi	jiab	mai	sam-fu	mo?	[adverb]
		she	often	buy	clothes	not.have
		‘Does she often buy new clothes?’ (Y.-J. J. Hsieh, 2013, p. 32)
	e.	Gi	gim-bu-ngid	ti	vug-ka	mo?	[preposition]
		he	today	at	home	not.have
		‘Is he at home today?’ (Y.-J. J. Hsieh, 2013, p. 32)

Fifth, both types of questions are non-biased. In a context where the speaker notices that his grandmother is very nice to a girl and suspects that she likes the girl, the speaker cannot use A-not-A questions (18a) or VP-mo questions (19a) to ask for confirmation.7

(18)	Mandarin A-not-A questions:
	a.	#Nainai	xihuan	bu	xihuan	zhe-ge	nu-sheng?
		grandma	like	not	like	this-CL	girl
		‘Does grandma like this girl?’
	b.	Nainai	xihuan	zhe-ge	nu-sheng	hou?
		grandma	like	this-CL	girl	PRT
		‘Grandma likes this girl, right?’

(19)	Hakka VP-mo questions:
	a.	#Apo	zhong-i	ya-ge	se-moi-e	mo?
		grandma	like	this-CL	girl	not.have
		‘Does grandma like this girl?’ (Y.-J. J. Hsieh, 2013, p. 33)
	b.	Apo	zhong-i	ya-ge	se-moi-e	ho?
		grandma	like	this-CL	girl	PRT
		‘Grandma likes this girl, right?’ (Y.-J. J. Hsieh, 2013, p. 33)

Based on these similarities, Hsieh argues that VP-mo questions are a type of A-not-A question and proposes an analysis inspired by C.-T. J. Huang’s (1991) analysis of A-not-A questions. C.-T. J. Huang (1991) proposes that A-not-A questions in Mandarin have an abstract Q under the INFL position, as illustrated in (20).8 The abstract Q is realized by a process of reduplication and the insertion of negation (bu ‘not’ in (20)).

(20)	a.	Ni	xihuan	bu	xihuan	zhe-ben		shu?
		you	like	not	like	this-CL		book
		‘Do you like this book?’
	b.						(C.-T. J. Huang, 1991, p. 316)

Drawing on C.-T. J. Huang’s (1991) analysis, Hsieh 2013 proposes that the sentence-final mo is an overt realization of the interrogative INFL. Hsieh argues that the negator mo is base-generated in Neg⁰, and then raises to INFL, and subsequently to C⁰ to form a question. The surface order of the NPQ is then derived after the remnant TP raises to Spec, CP, as illustrated in (21).

(21)	a.	Amin	oi	hi	hog-gau	mo?
		Amin	want	go	school	not.have
		‘Does Amin want to go to school?’
	b.

This analysis explains why a negative predicate is not allowed in VP-NEG questions, as shown in (22). It is argued that, because the sentence-final mo moves out of the predicate from the position of Neg⁰, an additional negator is not allowed within the same predicate.

(22)	*Apo	m	zhong-i	ya-ge	se-moi-e	mo?
	grandma	not	like	this-CL	girl	not.have
	Intended reading: ‘Doesn’t grandma like this girl?’ (Y.-J. J. Hsieh, 2013, p. 56)

However, this analysis has its own weakness. If the negator mo is base-generated in Neg⁰, its reconstruction would be expected to be acceptable. However, mo in the example sentences in (23) cannot be placed in the preverbal position, as shown in (24). This indicates that it is unlikely that the sentence-final mo in NPQs is base-generated in the head of NegP in a preverbal position.

(23)	a.	Amin	zo-ded	hi	hog-gau	mo?
		Amin	can	go	school	not.have
		‘Can Amin go to school?’
	b.	Amin	voi	hi	hog-gau	mo?
		Amin	will	go	school	not.have
		‘Will Amin go to school?’
	c.	Amin	iu	hi	hog-gau	mo?
		Amin	have	go	school	not.have
		‘Did Amin go to school?’

(24)	a.	*Amin	mo	zo-ded	hi	hog-gau.
		Amin	not.have	can	go	school
		Intended reading: ‘Amin cannot go to school.’
	b.	*Amin	mo	voi	hi	hog-gau.
		Amin	not.have	will	go	school
		Intended reading: ‘Amin will not go to school.’
	c.	*Amin	mo	iu	hi	hog-gau.
		Amin	not.have	have	go	school
		Intended reading: ‘Amin did not go to school.’

The negator mo is not compatible with the predicates in (24), namely zo-ded ‘can,’ voi ‘will,’ and iu ‘have.’ To negate zo-ded ‘can,’ one must say zo-m-ded ‘cannot,’ inserting the negator m ‘not’ between zo and ded. To negate the modal verb voi ‘will,’ the negator m ‘not’ must be used instead of mo. To express the negation of iu ‘have,’ mo ‘not have’ as the negated counterpart of iu is used instead. In other words, there is no agreement between these verbs and the negator mo, which is not considered in Hsieh’s Neg-to-C movement approach. If the sentence-final mo is not base-generated in the head of NegP within the predicate, its exact syntactic position remains unclear. In the next section, we introduce two alternative analyses of NPQs from other languages and examine whether they can be applied to VP-mo questions in Sixian Hakka.

3. Two Analyses of NPQs in Other Languages

3.1. Agreement Languages and Non-agreement Languages

L. L.-S. Cheng et al. (1997) discuss NPQs in Mandarin, Cantonese, and Taiwan Southern Min. They point out that different languages adopt different strategies for forming NPQs. Languages that exhibit agreement patterns are classified as agreement languages, while those without such patterns are classified as non-agreement languages. In agreement languages such as Mandarin, aspect markers and verb types that are compatible with the negator also remain compatible when the negator appears in the sentence-final position as a negative particle in NPQs, as shown in (25) and (26). The negator meiyou ‘not have’ is compatible with the aspect marker guo but incompatible with the modal verb hui ‘will,’ as exemplified in (25). These agreement patterns are maintained when the negator meiyou appears in the sentence-final position to form an NPQ, as shown in (26).9

(25)	a.	Hufei	meiyou	qu-guo.
		Hufei	not.have	go-EXP
		‘Hufei has not been (there).’
	b.	*Hufei	meiyou	hui	qu.
		Hufei	not.have	will	go
		Intended reading: ‘Hufei will not go.’ (L. L.-S. Cheng et al., 1997, p. 67)

(26)	a.	Ta	qu-guo	meiyou?
		he	go-EXP	not.have
		‘Has he been there?’
	b	*Ta	hui	qu	meiyou?
		he	will	go	not.have
		Intended reading: ‘Will he go?’ (L. L.-S. Cheng et al., 1997, p. 74)

In non-agreement languages such as Cantonese and Taiwan Southern Min, agreement between negation and an aspect or verb is not maintained. Cantonese examples are given below. The negator mei ‘not have’ in Cantonese is incompatible with the modal verb hoyi ‘can’ and the perfective aspect marker zo. However, in NPQs, the sentence-final particle mei is compatible with both.

(27)	a.	*Keoi	mei	hoyi	lei.
		he	not.have	can	come
		Intended reading: ‘He will not come.’
	b.	*Keoi	mei	lei-zo.
		he	not.have	come-PERF
		Intended reading: ‘He has not come yet.’ (L. L.-S. Cheng et al., 1997, p. 69)

(28)	a.	Ngo	hoyi	ceot-heoi	mei?
		I	can	go.out	not.have
		‘Can I go out?’
	b.	Keoi	sik-zo	fan	mei?
		he	eat-PERF	rice	not.have
		‘Has he eaten?’ (L. L.-S. Cheng et al., 1997, p. 75)

L. L.-S. Cheng et al. (1997) further argue that NPQs in agreement languages, such as Mandarin, involve Neg-to-C movement, while those in non-agreement languages, such as Cantonese and Taiwan Southern Min, do not. In the latter case, the sentence-final negative particle is base-generated in C⁰.

According to L. L.-S. Cheng et al.’s (1997) classification, Sixian Hakka should be categorized as a non-agreement language, because agreement between verbs and negation is not maintained in NPQs, as shown in (23) and (24). If this analysis is on the right track, mo in VP-mo questions would be base-generated in the C⁰ position. In that case, the negative predicates should be compatible with mo. However, this is not borne out. VP-mo questions do not allow negative predicates, as evidenced by (22).

So far, neither the Neg-to-C movement approach nor the base-generation approach for non-agreement languages can fully account for Hakka VP-mo questions. This raises the question of whether a third type exists—distinct from both agreement and non-agreement languages. We now turn to Aldridge’s account of NPQs in Middle Chinese to explore whether a historical perspective might offer a possible solution.

3.2. Analysis of VP-NEG Questions in Middle Chinese

Aldridge (2011) examines VP-NEG questions observed in Middle Chinese, especially VP-wu and VP-fou questions. See (29) for some examples, where the negators wu ‘not have’ and fou ‘not be’ can appear in the sentence-final position and function similarly to a question particle.

(29)	a.	秋	寒	有	酒	無？
		Qiu	han	you	jiu	wu?
		autumn	cold	have	liquor	not.have
		‘In the autumn cold, is there any liquor?’
		(Bai Juyi, 9th c., from Aldridge, 2011, p. 2)
	b.	尊	者	能	食	粗	惡	食
		Zun	zhe	neng	shi	cu	e	shi
		respect	DET	can	eat	coarse	bad	food
		不？
		fou?
		not.be
		‘Oh great one, can you eat inferior food?’
		(Zabao Zangjing 50, ca. 5th c., from Aldridge, 2011, p. 9)

Aldridge 2011 investigates the syntactic structure of VP-NEG questions in Middle Chinese and proposes a reanalysis process (i.e., a grammaticalization path) of the negators into a Q element. The analysis of VP-fou questions in Middle Chinese is illustrated in (30). It is proposed that VP-fou questions in the earlier stage (before the 5th century) involve a disjunctive construction composed of a positive predicate and a negative predicate. The head of the disjunctive phrase (&P) bears two uninterpretable features, [uNeg] and [uQ].

(30)

(Aldridge, 2011, p. 4)

Aldridge argues that at this stage, fou was part of the vP complement of &P before raising to attach to & in a head-to-head manner. The [uNeg] feature is valued after fou moves to &; the [uQ] feature on the disjunctive head is valued through an Agree relation with the [Q] feature on C. See (31).

(31)

(Aldridge, 2011, p. 17)

At this stage, fou was still undergoing grammaticalization and had not been fully grammaticalized into a question marker. The supporting facts are as follows: Negative predicates occurring before sentence-final fou are not attested, which may suggest that fou was base-generated in a position inherently endowed with [Neg] feature. However, the lack of agreement between predicates and fou supports the idea that fou occupied C⁰. For instance, the predicate headed by you ‘have,’ which cannot be negated by fou, can still be followed by sentence-final fou. Thus, Aldridge suggests that fou at this time was still base-generated in a lower position.

When fou moved to &, it acquired the feature [uQ], and fou was eventually grammaticalized to be base-generated as the head of CP (or ForceP in the sense of Rizzi, 1997). By the 5th century, the grammaticalization of fou ‘not be’ was complete, with fou functioning as a full-fledged Q particle base-generated in C, as we illustrate in (32). The complement of CP then moves to Spec, CP, deriving the surface order.

(32)

This argument is based on the following observations: First, fou could occur with a wide range of predicates. Fou and the predicates do not need to show agreement. Second, fou is compatible with a negative predicate, which indicates that it cannot be base-generated in a position endowed with the [Neg] feature. See (33).

(33)	無	諸	惡	不？
	Wu	zhu	e	fou?
	not.have	DET.PL	evil	not.be
	‘Are (you) free of the various irritations?’
	(Zabao Zangjing 73, ca. 5th c., from Aldridge, 2011, p. 9)

The analysis of VP-fou questions across different stages provides us with a new perspective on VP-mo questions in Sixian Hakka, especially regarding the role of grammaticalization, which is often overlooked in syntactic research. In the next section, we present our analysis by adopting Aldridge’s disjunctive approach to VP-mo questions and provide supporting evidence.

4. Proposal

Drawing inspiration from Aldridge’s 2011 disjunctive analysis of VP-fou questions in Middle Chinese, we propose that VP-mo questions in Sixian Hakka share a similar structure with earlier VP-fou questions, as shown in (34). The disjunctive phrase &P has two uninterpretable features: [uNeg] and [uQ].

(34)	a.	Amin	oi	hi	hog-gau	mo?
		Amin	want	go	school	not.have
		‘Does Amin want to go to school?’
	b.

We propose that the negator mo in Sixian Hakka originally headed the second vP of a disjunctive construction and would move to the head of a disjunctive phrase to value the feature [uNeg]. Through grammaticalization, mo has become the head of the disjunctive phrase, but it has not further grammaticalized into a question particle base-generated in C.10 The feature [uQ] is valued through an Agree relation with the [Q] feature on C. In addition, following Aldridge’s (2011) analysis of VP-NEG questions, we assume that the VP-mo questions in Sixian Hakka are derived by deletion of the second VP in the disjunctive construction and stranding of the negator in v.

This proposal is supported by the following facts: First, VP-mo questions, similarly to earlier-stage VP-fou questions, cannot occur with negative predicates; see (35). They also show a lack of agreement between the negator and predicates; see (36).

(35)	a.	*Apo	m-zung-i	ya-ge	se-moi-e	mo?
		grandma	not-like	this-CL	girl	not.have
		Intended reading: ‘Doesn’t grandma like this girl?’ (Y.-J. J. Hsieh, 2013, p. 56)
	b.	*Amin	mo	hi	hog-gau	mo?
		Amin	not.have	go	school	not.have
		Intended reading: ‘Didn’t Amin go to school?’

(36)	a.	Amin	zo-ded	/	voi	/	iu	hi	hog-gau	mo?
		Amin	can		will		have	go	school	not.have
		‘Can/Will/Did Amin go to school?’
	b.	*Amin	zo-ded	/	voi	/	iu	hi	hog-gau	ia
		Amin	can		will		have	go	school	or
		mo	zo-ded	/	voi	/	iu	hi	hog-gau?
		not.have	can		will		have	go	school
		Intended reading: ‘Can/Will/Did Amin go to school, or not?’

In addition, mo ‘not have’ can stand alone, as shown in (37). This suggests that mo functions as a head rather than an adjunct in syntactic terms, and is capable of undergoing head movement. This is quite similar to the status of fou in Middle Chinese, as exemplified in (38).

(37)	Context: Jane wishes that someone would drive her to the party (she would rather not go if she had to take the bus) …
	Mo,	ngai	moi	hi	ho	le.
	not.have	I	do.not	go	good	SFP
	‘If it’s not the case, it is better for me not to go.’

(38)	順	則	進，	否	則	退。
	Shun	ze	jin,	fou	ze	tui.
	accept	then	proceed	not.be	then	hold.back
	‘If (your opinion) is accepted, then proceed; if that is not the case, then hold back.’ (Yanzi Chunqiu 3.14, from Aldridge, 2011, p. 11)

Moreover, mo in VP-mo questions does not further move to C⁰, in contrast to fou in VP-fou questions at a later stage of its diachronic development. Notably, VP-mo questions can be embedded—unlike questions with the Q-particle ma in Mandarin, in which the sentence-final particle ma occupies Force⁰ (Paul, 2014; Pan, 2015, 2019).

(39)	a.	Amin	m-di	[gi	iu	ti	ia	mo].
		Amin	not-know	he	have	at	here	not.have
		‘Amin doesn’t know whether he is here.’
	b.	*Wo	bu-zhidao	[ta	zai	ma].
		I	not-know	he	at	Q
		‘I don’t know wether he is here.’ (Aldridge, 2011, p. 8)

The proposed account is capable of addressing the problems that remain in the previous literature.11 On the other hand, these observations of Sixian Hakka appear to provide an answer to the question raised in Section 2: whether there is a third type of NPQ, distinct from those found in agreement and non-agreement languages. We suggest that more than two types of NPQs may exist cross-linguistically.12

5. Conclusions

This work points out potential issues in the previous literature on VP-mo questions in Sixian Hakka and provides a syntactic analysis that addresses these issues. Adopting Aldridge’s 2011 disjunctive construction analysis, we propose that VP-mo questions share a similar syntactic structure with VP-fou questions in Middle Chinese prior to the 5th century, and that the negator mo in VP-mo questions has grammaticalized into a disjunctive head.