Locative Inversion in Vietnamese

Abstract

This paper has four primary objectives. First, we examine two Vietnamese constructions—the locative and existential constructions—in which a locative phrase appears clause-initially, preceding a verb-DP sequence. We demonstrate that, in the locative construction, the locative phrase is an argument of the locative verb, whereas, in the existential construction, it functions as a locative adjunct. Second, we show that this locative adjunct in the existential construction can undergo optional topicalization to the left periphery. More crucially, we argue that the fronting of the locative argument in the locative construction is neither Ā-topicalization to the left periphery nor A-movement targeting Spec,TP, supporting the featural approach to the A/Ā distinction. Third, by comparing the expletive nó with the fronting locative argument in the locative construction, we contend that both compete for the same specifier position of a hybrid A/Ā projection in the high TP domain. Fourth, to detail the movement of the locative argument in the locative construction, we posit a composite probe [D, δ] on the head of the A/ĀP to selectively move the locative argument to Spec,A/ĀP over the structurally higher theme DP.

1. Introduction

Locative phrases in Vietnamese are sometimes found sentence-initially preceding a verb-DP sequence, as exemplified by trên bàn ‘on the table’ in (1) and (2). The minimal difference between (1) and (2) lies in the choice of the main predicates: (1) contains a predicate of posture nằm ‘lie’, while (2) features an existential predicate có (lit. ‘have; exist’). For expository convenience, we refer to the former as locative constructions (LC) and the latter as existential constructions (EC).1

(1)	Trên	bàn	nằm	một	chồng	sách	(LC)
	on	table	lie	a	pile	book
	‘On the table lies a pile of books.’

(2)	Trên	bàn	có	một	chồng	sách	(LC)
	on	table	exist	a	pile	book
	‘On the table there exists a pile of books.’’

In this paper, we argue that the derivations of the sentence-initial locative phrases in (1) and (2) do not fall under a uniform analysis. In particular, we propose that the locative phrase in LC is a locative argument that raises a hybrid A/Ā functional projection in the high TP-domain extensively studied in Bošković (2024a), while that in EC is a locative adjunct.

The structure of this paper proceeds as follows. Section 2 examines the fundamental properties of LC and EC, focusing on the verb types and their syntactic relations with the locative phrases. We will see that the locative phrase in LC is an argument of the locative verb, while that in EC is a locative adjunct. Section 3 addresses the syntactic position(s) of the locative phrases in LC and EC. We show that the locative adjunct in EC may undergo optional topicalization to the left periphery. More importantly, we argue the fronting of the locative argument in LC is neither Ā-topicalization to the left periphery nor an A-movement targeting Spec,TP, under the traditional positional approach to A/Ā-distinction (Chomsky, 1981, p. 47; 1995, 2007, 2008; Mahajan, 1990; Déprez, 1989; Miyagawa, 2010, 2017). The goal of Section 4 is two-fold. First, based on the findings in Section 3, we show that the locative argument in LC competes with the expletive nó for the same specifier position of a hybrid A/Ā-projection in the high TP-domain (Bošković, 2024a). Second, we detail the derivational mechanics of the movement to this A/Ā-projection based on the composite probing mechanism in Coon and Bale (2014) and Van Urk (2015). Section 5 explores the theoretical implications of our analysis, including the Merge over Move principle (Chomsky, 1995) and Ā-opacity (Rezac, 2003). Section 6 concludes this paper.

2. Verb Classes and the EC/LC Distinction

An LC is typically composed of posture verbs (such as nằm ‘lie’, ngồi ‘sit’, and đứng ‘stand’), verbs of placement (such as đặt ‘place’, để ‘put, place’, and treo ‘hang’) and dwell verbs (such as ở and sống ‘dwell’) as exemplified by (3).

(3)	Trên	bàn	nằm/	để	một	chồng	sách.
	on	table	lie	put	a	pile	book
	‘On the table lies / is put a pile of books.’

These verbs must be used in their stative sense in LC, not their change-of-state sense, exhibiting “informational lightness” in the sense of Birner (1992). This property explains why LC is sometimes found accompanied by manner adverbs such as lăn lóc ‘in a disordered/scattered manner’ and lủng lẳng ‘in a dangling manner’ as in (4a) and (4b), respectively. These manner adverbs are incompatible with the dynamic/change-of-state interpretation of these verbs.2

(4)

a.

Trên

mặt

bàn

nằm

lăn lóc

mấy

vỏ

chai

bia

33.3

on

surface

table

lie

disorderly

several

shell

bottle

beer

33

‘On the table lie several empty “33” beer bottles in a disorderly manner.’

b.

Dọc theo

lan can

đò

treo

lủng lẳng

một số

áo phao.4

along

railing

boat

hang

danglingly

some

life.jacket

‘Along the railing of the boat hang several life jackets in a danglingly manner.’

Collectively referred to as locative verbs here, the verbs forming LC are intrinsically tied to a location, necessitating subcategorization for a locative phrase. The subcategorization requirement of the locative verbs can be jointly evidenced by the obligatory occurrence of the locative phrase in (5) and (6). First, (5) shows that an LC is grammatical only with the overt realization of the locative phrase. Next, although (6) indicates that the locative phrase can stay in situ within the VP hosting the locative verb when the theme DP moves to Spec,TP, its presence within VP is mandatory.

(5)	The clause-initial locative phrase is obligatory in LC:
	*(Trên	bàn)	nằm	một	chồng	sách.
	on	table	lie	a	pile	book

(6)	The locative phrase is a core argument of the locative verb:
	Một	chồng	sách	nằm	*(trên	bàn.)
	a	pile	book	lie	on	table
	‘A pile of books lies *(on the table).’

On the other hand, an EC typically goes with verbs of existence, which denote general existence and duration, such as có ‘exist’ and còn ‘remain’ as in (7).5 In principle, (7) allows two interpretations: it can be interpreted as a generic statement asserting the presence of water in general or as a contextualized statement asserting the presence of water in the specific speech context. Importantly, an EC remains grammatical without the overt presence of a clause-initial locative phrase, in contrast to (5). In addition, an EC allows both the theme DP and the locative phrase to appear post-verbally, as in (8). See the contrast between (8a) and its corresponding LC in (9).

(7)	The locative phrase is optional in EC:
	Có/	còn	nước.
	exist	remain	water
	‘There is/ remains water.’

(8)

EC allows a post-verbal locative phrase:

a.

Có

một

chồng

sách

trên

bàn.

exist

a

pile

book

on

table

‘There is a pile of books on the table.’

b.

Xuất hiện

băng

tuyết

tại

huyện

Bình Liêu,

Quảng Ninh.6

appear

ice

snow

at

district

B.L.

Q.N.

‘There appeared frost snow in Binh Lieu District of Quang Ninh Province.’

(9)	*Nằm	một	chồng	sách	trên	bàn.
	lie	a	pile	book	on	table
	Intended: ‘There lied a pile of books on the table.’

In view of the contrast between LC and EC with respect to the requirement of an overt locative phrase, we assume that the locative phrase is a core argument of the locative verb in LC (cf. Bresnan, 1994), while that in EC is a locative adjunct. Thus, we argue that the locative argument in LC is not base-generated in the clause-initial position as in (1). Rather, it is base-generated within VP as in (10). The derivation starting from the base structure in (10) may be followed by moving either the locative argument or the theme DP, resulting in (1) and (6), respectively.7^,8

(10)	The base structure of LC:
	[CP …… [vP v   [VP DP V LP ]]]

An alternative hypothesis assumes that the locative argument is base-generated at a position structurally higher than the theme DP in LC. This alternative poses a theoretical problem with respect to Baker’s (1997) Uniformity of Theta Assignment Hypothesis (UTAH): the locative phrase bears the oblique theta-role of Location, which should be lower than the Theme argument according to the UTAH’s hierarchy in (11) (cf. Wu, 2008).

(11)	Agent > Theme > Goal > Obliques (manner, location, time)

We offer two more arguments for our proposal in (10) regarding the subcategorization requirement of the locative verbs forming LC. The first argument is related to the selectional restrictions of locative verbs in LC. If the locative phrase is an argument of the locative verb in LC, there should be a c-selection restriction on the kind of locative information that is permitted in LC. This prediction is borne out, as only locative phrases denoting a specified location via the use of a localizer are felicitous, while more general locations without a localizer are odd in LC, as in (12b). The contrast in (12) suggests that the locative verb in LC imposes a c-selection requirement on its locative argument. We assume that the c-selection requirement in (12) follows from the argument status of the locative phrase in LC.

(12)	The localizer is necessary in LC:
	a.	Trong	công viên		ngồi	một	cụ già.
		in	park		sit	one	old.person
		‘In the park sits an old person.’
	b.	??/*Công viên		ngồi	một	cụ già.
		park		sit	one	old.person
		Intended: ‘At the park sits an old person.’

By contrast, the locative information in EC can but does not need to be realized by a locative phrase, as shown in (13). Consequently, EC can express either a specified location as in (13a) or a general location in (13b) (which can be contextually determined).

(13)	The localizer is optional in EC:
	a.	Trong	công viên			có /	xuất hiện		một		cụ già.
		in	park			exist	appear		one		old.person
		‘In the park there is/ appears an old person.’
	b.	Công viên		có /	xuất hiện			một		cụ già.
		park		exist	appear			one		old.person
		‘At the park there is/ appears an old person.’

The second argument concerns the insertion of temporal adverbials or tense/aspectual markers in EC/LC. Specifically, note that although the locative phrase is optional in EC in (7), EC is most natural with some sort of overt temporal/locative anchoring.9 In the absence of a locative phrase, the anchoring of EC can be achieved via temporal adverbials or tense/aspectual markers, as in (14) and (15), respectively.

(14)

Anchoring of EC via temporal adverb(ial)s:

a.

Sáng

nay

xuất hiện

một

bệnh nhân.

morning

this

appear

one

patient

‘There appeared a patient this morning.’

b.

Hôm qua

rơi

nhiều

tuyết.

yesterday

fall

a.lot

snow

‘Yesterday fell a lot of snow.’

(15)

Anchoring of EC via tense/aspectual markers:

a.

Vừa

đến

một

bệnh nhân.

just

arrive

one

patient

‘A patient just arrived.’

b.

Đang

tan

nhiều

tuyết.

prog

melt

a.lot

snow

‘A lot of snow is melting.’

However, temporal adverb(ial)s and tense/aspectual markers cannot salvage an LC without a locative phrase, as evidenced by (16). Again, the ungrammaticality of (16) highlights the necessity of the overt presence of a locative phrase in LC.

(16)

a.

*Sáng

nay

nằm

một

bệnh nhân.

morning

this

lie

one

patient

Intended: ‘This morning lied a pile of books.’

b.

*Vừa

nằm

một

chồng

sách.

just

lie

one

pile

book

Intended: ‘Just lied a pile of books.’

In summary, the optional presence of a locative phrase in existential constructions (EC), as shown in examples (7), (14), and (15), supports its status as an adjunct. In contrast, the predicates forming an LC require a locative argument, as demonstrated by the grammatical occurrence of the locative phrase within the VP in (6), the c-selection requirement in (12), and the mandatory inclusion of the locative phrase in (16). This distinction highlights the subcategorization differences between EC and LC with regard to the locative phrases.

3. The Raising of the Locative Argument in LC

Given the conclusion established in the previous section, note that adding the locative phrase at the post-verbal position, as in (17), does not save (16).

(17)

a.

*Sáng

nay

nằm

một

chồng

sách

trên

bàn.

morning

this

lie

one

pile

book

on

table

Intended: ‘This morning lied a pile of books on the table.’

b.

Vừa

nằm

một

chồng

sách

trên

bàn.

just

lie

one

pile

book

on

Table

Intended: ‘Just lied a pile of books on the table.’

The ungrammaticality of (17) indicates that the locative argument of the locative verb has to raise out of the vP domain to construct a grammatical LC. Now, the question comes down to the landing site of the movement of the locative argument in LC. Two avenues can be envisaged to identify the nature of the position(s) targeted by the fronting locative argument in LC, as shown in (18): it raises either to the canonical subject position Spec,TP, or to a syntactic position other than Spec,TP.

(18)	Possible analyses of the movement of the locative phrase in LC:
	a.	It targets Spec,TP,
	b.	It raises to a position other than Spec,TP.

The possibility in (18a) encounters two analytical challenges. The first challenge concerns the minimality of the movement. In particular, suppose the raising to Spec,TP is triggered by the satisfaction of a D-feature on T in Vietnamese as in (19); the raising of the locative argument to Spec,TP across the theme DP runs afoul of the relativized minimality constraint (Rizzi, 1990). This is due to the fact that both the locative argument and the theme DP can grammatically occupy this position according to the hypothesis in (18a); however, the theme DP is situated closer to T than the locative argument. This raises the question of how the locative argument can move to Spec,TP, surpassing the theme DP, which typically occupies that position.

(19)

Second, if we assume that vP is a phase (Legate, 2003), the locative phrase may move to the edge of the vP phase to become a closer goal to T than the theme DP is, as in (20). However, given that this type of raising to the phase edge is typically considered as an instance of Ā-movement, it is unclear how a subsequent raising to Spec,TP (as a canonical instance of A-movement triggered by D/φ-features) in (20) circumvents the constraint on improper movement which prohibits an Ā-A sequence of derivation (May, 1979; Chomsky, 1981; Abels, 2007; Neeleman & Van De Koot, 2010; Williams, 2011).

(20)

Due to the minimality issue in (19) and the ban on improper movement in (20), we do not assume that the locative argument in LC raises to Spec,TP. In Section 4, we will see an additional empirical problem of the hypothesis in (18a): the raising of the theme DP and that of the locative argument in LC exhibit different semantic/pragmatic properties in terms of specificity.10

Given the challenges facing (18a), we pursue the possibility in (18b) to derive the fronting of the locative argument in LC. Our discussion of (18b) focuses on the relative order of the fronting locative argument and two elements: the complementizer in the left periphery and the expletive nó in the high TP-domain (Greco et al., 2017, 2018). We discuss the relative order with the complementizers là and rằng in this section. Note that the locative adjunct in EC may either precede or follow the complementizers là as in (21).

(21)

a.

Rõ ràng

là

trên

giường

có

một

bệnh nhân.

clearly

c

on

bed

exist

a

patient

‘Clearly on the bed there is a patient.’

b.

Trên

giường

rõ ràng

là

có

một

bệnh nhân.

on

bed

clearly

c

exist

a

patient

‘On the bed there clearly is a patient.’

Phan (2024) observes that the complementizer là can occur in main clauses only if it is immediately preceded by CP-level sentential mood and modal markers such as the evidential phrases rõ ràng ‘clearly’ and hình như ‘seemingly’. He proposes that là heads the Finiteness Phrase, marking the lowest bound of the Split-CP à la Rizzi (1997), and is licensed by these sentential mood/modal items. The grammaticality of the pair in (21) with respect to the relative order with là indicates that the locative adjunct in EC is adjoined TP-internally in (21a)11 and may undergo optional topicalization to the left periphery above the FinP in (21b). Note that the locative adjunct is not inherently a topic. The non-topical nature of the locative adjunct in EC can be evidenced by its occurrence in out-of-the-blue contexts as a felicitous reply to a global ‘what-happened’ question as in (22a), unlike the topicalized variation in (22b) (see Rizzi, 2004; Belletti & Rizzi, 2017).

(22)

Q:

Chuyện

gì

vậy?

matter

what

sfp

‘What happened?’

a.

Trên

bàn

có/

xuất hiện

một

chồng

sách.

on

table

exist

appear

a

pile

book

‘On the table there is/ appears a pile of books.’

b.

#Trên

bàn

thì

có/

xuất hiện

một

chồng

sách.

on

table

top

exist

appear

a

pile

book

Similarly, notice that the position of locative adjunct in relation to là has a crucial effect on the grammaticality of the insertion of the topic marker, as shown by (23). The insertion of the topic marker thì is allowed only when the locative adjunct precedes the complementizer, as in (23a), suggesting that the topic marker is available only for topic expressions moving to the left periphery.

(23)

The topic marker thì is available only for a topicalized phrase:

a.

Trên

giường

thì

có lẽ/

rõ ràng

là

có

một

bệnh nhân.

on

bed

top

perhaps

clearly

c

exist

a

patient

‘Perhaps/ clearly on the bed there is a patient.’

b.

*Có lẽ/

rõ ràng

là

trên

giường

thì

có

một

bệnh nhân.

perhaps

clearly

c

on

bed

top

exist

a

patient

On the other hand, the pair in (24) shows that an LC sounds degraded if the locative argument precedes the head of FinP and its licensers but not when it follows them, in contrast to (21).12 This suggests that the fronting of the locative argument in LC lands below CP and cannot keep moving to CP-level projections such as the Topic Phrase.

(24)

a.

Rõ ràng

là

trên

giường

nằm

một

bệnh nhân.

clearly

c

on

bed

lie

a

patient

‘Clearly on the bed lies a patient.’

b.

??Trên

giường

rõ ràng

là

nằm

một

bệnh nhân.

on

bed

clearly

c

lie

a

patient

Intended: ‘On the bed there clearly lies a patient.’

Our analysis of the locative phrases in LC and EC occupying different syntactic positions can be further supported by the insertion of the topic marker thì in Vietnamese. Note that the fronting locative argument in LC cannot be followed by the topic marker thì as in (25a), while the locative adjunct in EC is not subject to this constraint as in (25b).13 The contrast between (25a) and (25b) clearly indicates that the fronting of the locative argument in LC is not an instance of Ā-topicalization to the left periphery.14 Moreover, the ungrammaticality of (17) indicates the obligatory character of the fronting of the locative argument in LC, which is incompatible with the optionality of Ā-topicalization in general. This incompatibility makes an Ā-topicalization analysis of the fronting locative argument in LC difficult to maintain.

(25)	a.	*Trên	giường	thì		nằm	một	bệnh nhân.	(LC)
		on	bed	top		lie	a	patient
		Intended: ‘On the bed is lying a patient.’
	b.	Trên	giường	thì	có		một	bệnh nhân.	(EC)
		on	bed	top	exist		a	patient
		‘On the bed there is a patient.’

Another argument against the Ā-topicalization analysis of the fronting locative argument in LC concerns hyperraising. Several languages have been demonstrated to allow for A-movement from the subject position out of a finite CP, which is referred to as hyperraising since Ura (1994). Hyperraising can be demonstrated by Zulu in (26), where the subject DP Zinhle of the embedded predicate xova ”make” moves out of the finite CP complement across the matrix raising predicate bonakala “seem” (see also Ferreira, 2000, 2004, 2009; Rodrigues, 2004; Martins & Nunes, 2005, 2009, 2010; Nunes, 2008; Deal, 2017; Fong, 2019; Zyman, 2023).

(26)	Hyper-raising in Zulu: (Halpert, 2019, ex.3b)
	uZinhle1	u-bonakala	[CP	ukuthi  ___1	u-zo-xova	ujeqe].
	aug.1.Zinhle	1s-seem		that	1s-fut-make	aug.1steamed.bread
	‘It seems that Zinhle will make steamed bread.’

Lee and Yip (2024) argue that hyper-raising exists in Vietnamese in constructions involving (raising) attitude verbs such as nghe nói ‘hear’, as exemplified by (27b) (see their Section 3 for evidence for an A-movement analysis of (27b) and the finiteness of the CP complement).

(27)

Hyperraising in Vietnamese:

a.

Nghe nói

[CP

rằng/ là

cơn

mưa

này

sẽ

không

dừng.]

hear

c

cl

rain

this

fut

not

stop

‘It is heard that the rain will not stop.’

b.

Cơn

mưa

này

nghe nói

[CP

rằng/ là

sẽ

không

dừng.]

cl

rain

this

hear

c

fut

not

stop

‘It is heard that the rain will not stop.’ (adapted from Lee & Yip, 2024, p. 3)

What makes hyperraising relevant to our discussion is that LC is allowed to occur in the complement clause of nghe nói, as shown by (28a). More importantly, the locative argument is allowed to move out of the complement clause as in (28b).15

(28)

Hyperraising of the locative argument:

a.

Nghe nói

[CP

rằng/ là

[trên

giường] 1

nằm

một

bệnh nhân.

t1.]

hear

c

on

bed

lie

one

patient

‘It is heard that on the bed lies a patient.’

b.

[Trên

giường]1

nghe nói

[CP

rằng/ là

nằm

một

bệnh nhân.

t1.]

on

bed

hear

c

lie

one

patient

‘It is heard that on the bed lies a patient.’

Notice that the movement of the locative argument in (28b) cannot be analyzed as Ā-topicalization, as evidenced by the ungrammaticality of the insertion of thì (the marker of the Topic head in the left periphery) in (29).16

(29)	*[Trên	giường]1	thì	nghe nói	[CP	rằng/ là	nằm	một	bệnh nhân.	t1.]
	on	bed	top	hear		c	lie	one	patient
	Intended: ‘It is heard that on the bed lies a patient.’

The hyperraising of the locative argument lends further support to our analysis that the fronting of the locative argument in LC cannot be analyzed as Ā-topicalization. If the fronting of the locative argument in LC were Ā-topicalization, the subsequent A-movement of hyperraising in (28b) would constitute a violation of the ban on improper movement, which prohibits an Ā-A sequence of derivation (May, 1979; Chomsky, 1981; Abels, 2007; Neeleman & Van De Koot, 2010; Williams, 2011).17

Given the resistance of topic marking by thì shown in (25a), one may assume that the movement of the locative argument in LC can be analyzed as an A-movement. Aside from functioning as the grammatical subject in LC, the raising of the locative argument does not show a Weak Crossover effect (WCO) as in (30), indicating a typical feature of A-movement (see Culicover & Levine, 2001, pp. 289–290 for the same line of reasoning regarding the A-property of English locative inversion).18

(30)

[Trên

giường

của

mỗi

đứa

trẻ2]1

đều

để

[sách

của

nó2

t1].

on

bed

poss

each

cl

child

all

place

book

poss

he

‘On every child2’s bed are placed his2 books.’

Although the raising locative argument in LC resists overt topic marking by thì in (25a) and does not exhibit WCO19, it aligns with an Ā-topicalized phrase with respect to topic island effects and the specificity requirement of indefinite nominals. Let us begin with the topic island effects. It is well-known that locative inversion in English parallels Ā-topicalization in inducing island effects that block wh-movement, as illustrated in (31a) and (31b/c), respectively. This parallelism has been cited to support the Ā-properties of English locative inversion (Bresnan, 1994, p. 87; den Dikken, 2006, p. 100; Rizzi & Shlonsky, 2006, p. 344).

(31)	Topic island effects:
	a. *Wheni did he say that into the room walked Jack ti?
	b. *When1 did to Lee Robin give the pencil t1?
	c. *When1 did this book everyone read t1?

The same reasoning can be applied to demonstrate the Ā-properties of the fronted locative phrase in LCs. Since Vietnamese is a wh-in-situ language, Ā-focalization, as shown in (32a), is employed to construct the relevant argument. Notably, the Ā-topicalization of the locative adjunct in ECs blocks Ā-focalization, as illustrated in (32b).

(32)	a.	Là	[sách ngôn ngữ học]1	Tí  nghĩ  [CP rằng có   [một chồng t1] trên  bàn].
		foc	book linguistics	Tí  think        c      exist a      pile          on    table
		Literal: ‘It is linguistics books that Tí thinks there is a pile of on the table.’
	b.	*Là	[sách ngôn ngữ học]1	Tí  nghĩ [CP rằng [trên bàn]2  có   [một chồng t1] t2].
		foc	book linguistics	Tí  think       c       on     table   exist a     pile
		Intended: ‘It is linguistics books that Tí thinks on the table is a pile of.’

Crucially, the fronting of the locative argument in LCs similarly induces a topic island effect, as shown in (33). The parallelism between (32b) and (33) indicates that the fronting of the locative phrase in LCs and ECS, though targeting different structural positions, exhibits the same Ā-property in terms of the topic island effects.

(33)	The fronting locative argument in LC creates a topic island:
	*Là	[sách	ngôn ngữ học]1	Tí	nghĩ	[CP rằng [trên bàn]2	nằm	[một chồng t1] t2].
	foc	book	linguistics	Tí	think	c       on     table	lie	a       pile

Next, we turn to the argument concerning the specific reading for indefinite nominals. Note that nominal phrases that feature the indefinite article một ‘a/an’ followed by a classifier-noun sequence allow both a specific and a nonspecific/generic interpretation (see also Phan & Chierchia, 2022; Enç, 1991). For instance, the subject một bệnh nhân ‘a patient’ in (34) is allowed either a specific reading or a nonspecific/generic reading. The two possible interpretations of an indefinite NP with một can be further demonstrated by the two possible continuations in (35a) and (35b) (see Fodor & Sag, 1982, p. 355; Von Heusinger, 2002, p. 245).

(34)	Một		bệnh nhân	nằm	trên	một	cái		giường.
	a		patient	lie	on	A	cl		bed
	a.	‘A specific patient lies on a bed.’						(specific)
	b	‘A/any patient lies on a bed.’						(nonspecific/generic)

(35)

Một

sinh viên

lớp

Cú pháp

gian lận

trong

giờ

thi.

a

student

class

syntax

cheat

in

time

test

‘A student in Syntax cheated on the exam.’

a.

Tên

sinh viên

đó

là

Tí.

(specific)

name

student

that

be

Tí

‘The student’s name is Tí.’

b.

Chúng tôi

đang

điều tra

xem

đó

là

ai.

(nonspecific)

we

prog

investigate

see

that

be

who

‘We are trying to figure out who it was.’

Notice that the specificity ambiguity of indefinite nominals disappears under Ā-topicalization as in (36) because topics display definiteness/specificity effects (Enç, 1991; Erteschik-Shir, 1997, a.o.).20 Similarly, Phan and Lander (2015, fn.3) suggest that topicality is a function of specificity (cf. Cresti, 1995; Portner, 2002) and that the topic particle thì is one of the specificity markers in Vietnamese. Thus, the topicalized indefinite nominal marked by thì in (36) receives only the specific reading.

(36)	A topicalized indefinite nominal must receive a specific reading:
	Context: Professor A is telling about his three favorite students…
	Một	sinh viên	thì	rất	giỏi	cú pháp.
	a	student	top	very	good	syntax
	‘As for one student, s/he is really good at syntax.’ (specific/*nonspecific/*generic)

The specificity ambiguity of indefinite nominals is carried over to locative phrases containing indefinite NPs, as demonstrated by (37). When the locative phrase containing an indefinite NP stays within vP as in (37), it patterns with the indefinite NPs in (34)/(35) in displaying ambiguity with respect to specificity.

(37)	Một		bệnh nhân	nằm	trên	một	cái		giường.
	a		patient	lie	on	a	cl		bed
	a.	‘A patient lies on a specific bed.’						(specific)
	b	‘A patient lies on a/any bed.’						(nonspecific/generic)

Importantly, the ambiguity in (37) disappears when the locative phrase is fronted as in (38): the fronted LP is necessarily interpreted as being specific, relating to a specific bed in the speaker’s mind. This specific spatial coordinate helps anchor the eventuality reported in the LC to a specific context.

(38)	Trên		một	cái	giường	nằm	một		bệnh nhân.
	on		a	cl	bed	lie	a		patient
	a.	‘On a specific bed lies a patient.’						(specific)
	b	#‘On a/any bed lies a patient.’						(*nonspecific/*generic)

Now, we encounter a dilemma in the analysis of the trigger of the movement of the locative argument in LC. On the one hand, the fronting locative argument in LC patterns with a topicalized phrase in terms of the topic island effects and the specificity requirement of indefinite nominals. On the other hand, the fronting locative argument resists the topic marker thì and does not show WCO. How do we capture the fact that the movement of the locative argument, though not a case of Ā-topicalization to the left periphery, exhibits the specificity requirement seen in Ā-topicalization? What makes this problem trickier is the contrast between (34) and (38). The indefinite subject NP in the categorical sentence in (34) allows either a nonspecific/generic or a specific reading, while the fronted locative argument in (38) permits only a specific interpretation. This contrast indicates that the movement of the indefinite subject NP in (34) and that of the locative argument in LC in (38) are driven by two distinct motivations and target different syntactic positions such that only the former retains the specificity ambiguity of an indefinite NP.

Summing up, the ungrammaticality of (25a), the contrast in (23)/(24), and the grammatical hyperraising of the locative argument in (28b) corroborate our analysis that the fronting locative argument in LC stays below the CP level (more precisely, the FinP headed by the complementizer là), whereas the locative adjunct in EC may undergo optional topicalization to the left periphery. Thus, the fronting locative argument does not target Spec,TP, and it cannot be analyzed as Ā-topicalization. Now the question is where exactly in the high TP-domain the fronting locative argument in LC moves to. In the next section, we propose that the locative argument in LCs moves to the specifier position of a hybrid A/Ā-projection in the high TP-domain extensively investigated by Bošković (2024a). Our proposal will be empirically corroborated by the similarities of the locative argument in LCs and the non-referential expletive nó (lit. ‘it’) in Vietnamese.

4. Proposal

4.1. Movement to a Hybrid A/Ā-Projection Between TP and CP

Bošković (2024a) claims that there are not just one but three distinct structural positions for subjects across languages. Quirky subjects such as Icelandic mér ‘me’ and Þeim ‘them’ in (39) are in Spec,TP. Agreeing/nominative subjects such as Amy in (40) merge in a higher Spec,AgrsP, and non-nominal subjects and locally Ā-moved subjects such as who in (41) move further to the Spec of an A/Ā projection above AgrsP.21 The relative position of these three different projections hosting different types of subjects is shown in (42).

(39)	Mér/	Þeim	er	kalt.	(Bošković, 2024a, p. 26)
	meDAT	themDAT	is	cold

(40)	I think Amy left. (Bošković, 2024a, p. 8)

(41)	I wonder who left. (Bošković, 2024a, p. 8)

(42)	[A/ĀP wh-moved subject [AgrsP Mary [TP quirky subjects…
	(adapted from Bošković, 2024a, p. 26)

Bošković (2024a) refers to the highest projection as the A/ĀP because it resides on the border of the traditional A and Ā field, with its Spec (where English focalized or wh-moved subjects are located) exhibiting mixed A/Ā-properties. Adopting this line of analysis of multiple landing sites of subject movement, we suggest that the locative argument in LC is moved to the specifier of the A/ĀP, a “subject-dedicated position” which Bošković (2024a, p. 23, fn.12) views as not having “the same featural content in all its instantiations.” Note that Bošković (2024a) argues that Spec,A/ĀP is the dedicated position for non-nominal subjects, making it the ideal landing site for quirky PP subjects (with a null P) in Serbo–Croatian (see Bošković, 2024b and citations therein) and locative PPs in English locative inversion.

That the locative argument in Vietnamese LC is not canonically nominal, given that it can be headed by either a localizer (43) or a preposition (44), is well aligned with the characteristics of this A/Ā subject-dedicated position shown in Bošković (2024a, 2024b).22

(43)

a.

Trên

bàn

nằm

một

chồng

sách.

(=(1))

on

table

lie

a

pile

book

‘On the table lies a pile of books.’

b.

[LP

trên

[NP

bàn…

(44)

a.

Ở

trên

bàn

nằm

một

chồng

sách.

at

on

table

lie

a

pile

book

‘On the table lies a pile of books.’

b.

[PP   ở

[LP

trên

[NP

bàn…

Therefore, in this paper, we propose that the fronting locative argument in LCs targets the specifier position of the hybrid A/Ā-projection proposed by Bošković (2024a, 2024b). In the next section, we discuss the similarities between the locative argument in LCs and the expletive nó in Vietnamese to corroborate this proposal.

4.2. The Similarities of the Fronting Locative Argument and the Expletive Nó

We propose that the occurrence of the non-referential expletive nó (lit. ‘it’) in Vietnamese, as discussed in Greco et al. (2017, 2018), sheds light on the question of the syntactic position occupied by the locative argument in LC. We find that LC is incompatible with the expletive nó, regardless of the word order of the locative argument and the expletive, as evidenced by (45).

(45)	The complementary distribution of LC and the expletive nó:
	a.	*Trên	bàn	nó	nằm	một	chồng	sách.
		on	table	expl	lie	a	pile	book
		Intended: ‘There lies a pile of books on the table.
	b.	*Nó	trên	bàn	nằm	một	chồng	sách.
		expl	on	table	lie	a	pile	book

By contrast, Greco et al. (2017, 2018) find that the locative adjunct in EC precedes the expletive as in (46a). Reversing the word order of the locative adjunct and the expletive, as in (46b), results in an ungrammatical sentence.23

(46)	The locative adjunct in EC precedes the expletive: (adapted from Greco et al., 2018, p. 44)
	a.	Trên	bàn	(thì)	nó	sẽ	không	có	cái	bút	nào.
		on	table	top	expl	fut	neg	exist	cl	pen	any
		‘On the table, there will be no pens.’
	b.	*Nó	trên	bàn	(thì)	sẽ	không	có	cái	bút	nào.
		expl	on	table	top	fut	neg	exist	cl	pen	any
		Intended: ‘On the table, there will be no pens.’

The complementary distribution of LC and the expletive shown in (45) indicates that Bruening’s (2010) null expletive analysis of English locative inversion cannot be extended to the derivation of Vietnamese LC as depicted in (47) (see also Postal, 1977, 2004; Maruta, 1985; Coopmans, 1989; Hoekstra & Mulder, 1990). The null expletive analysis suggests that Spec,TP is occupied by a null expletive, with the locative phrase undergoing Ā-topicalization. The pattern we see in LC refutes both of these two hypotheses. For one thing, the fronting locative argument in LC resists marking by thì, representing the Topic head in the left periphery (see (25a)). For another, the postulation of a null expletive in Vietnamese is implausible in light of the incompatibility of LC and the expletive nó (see (45)).

(47)	A null expletive analysis of LC: (with the null expletive indicated by pro)
	[TopP	Trên	bàn	[TP	pro	nằm	một	chồng	sách]].
		on	table		expl	lie	a	pile	book
	‘There lies a pile of books on the table.’

The pattern in (45) is expected if the fronting locative argument and the expletive compete for the same syntactic position in the derivation of LC, so the obligatory fronting of the locative argument in the derivation of LC is blocked, resulting in the ungrammaticality of the pair in (45). A further piece of evidence supporting this line of analysis concerns Ā-topicalization. Specifically, Greco et al. (2018) observe that nó cannot undergo topicalization and precede the topic marker thì, as shown in (48). This ban on topicalization patterns with the locative argument in LC, as shown in (25a), repeated here in (49).

(48)	Nó cannot undergo topicalization: (Greco et al., 2018, p. 41)
	a.	*Nó	thì	ngã	thằng		bé.
		expl	top	fall	cl		little
		‘As for him, the boy fell.’
	b.	*Nó	thì	chết		cá	tao.
		expl	top	die		fish	mine
		‘As for him/her, my fish died.’

(49)	The locative argument in LC cannot undergo topicalization:
	*Trên	giường	thì	nằm	một	bệnh nhân.	(=(25a))
	on	bed	top	lie	a	patient
	Intended: ‘On the bed is lying a patient.’

The similarity between nó and the locative argument in LC goes beyond their syntactic distribution. Greco et al. (2017, 2018) find that the existential and thetic sentences featuring the expletive nó show a specificity effect. Concretely, the contrast in (50) shows that the presence of preverbal nó necessarily restricts an existential sentence to a contextual interpretation where the relevant eventuality is tied to some specific spatial/temporal coordinates.

(50)	The specificity requirement of nó: (Greco et al., 2017, p. 78)
	a.	Không		có	ma.
		neg		exist	ghost
		Generic: ‘Ghosts do not exist.’ Contextual: ‘There are no ghosts speaking of a certain place/time.’
	b.	Nó	không		có		ma.
		expl	neg		exist		ghost
		#Generic: ‘Ghosts do not exist.’ Contextual: ‘There are no ghosts speaking of a certain place/time.’

To account for the ban on the generic reading of existential sentences in the presence of nó, Greco et al. (2018, p. 49) state that the referent of the postverbal DP is necessarily linked to a certain set of entities that was previously determined, and the addition of nó obligatorily “narrows the possible denotations to the values that can be assigned through a retrievable relation ranging over familiar entities, either present in the discourse or in the speaker’s mind.” As pointed out by a reviewer, this discourse property of nó resonates with Bolinger’s (1973, 1977) view of ambient expletives or situation pronouns that have special referential content. That is, ambient expletives such as nó have a deixis ad oculos, “a reference to the immediate situation” (Bolinger, 1973, p. 263). Using Milosavljević and Milosavljević’s (2022) wording, nó may be understood as having a discourse-topical referent, referring back to an epistemically specific Topic Situation.

Therefore, both the fronting locative argument in LC and the (ambient) expletive nó in existential/thetic sentences function to narrow down the domain of validity of the associated eventuality to a specific context. This is aligned with Birner’s (1992, p. iii) conception of inversion (with locative inversion as an instance) as having the discourse function of introducing “relatively unfamiliar information to the prior context via the clause-initial placement of information that is relatively familiar in the discourse” (see also Levin & Rappaport Hovav, 1995). In the case of LC, the more familiar information is encoded by the fronting locative phrase, while the less familiar information is conveyed by the postverbal DP. It is the latter that receives presentational focus (Bresnan, 1994 and citations therein) and is (re)introduced into the domain of discourse.24 Using Enç’s (1991, p. 24) language, specificity in LC involves a (weak) link to a previously established referent, with the denotation of the locative phrase “being a subset of or standing in some recoverable relation to a familiar object” that is already prominent in the discourse. In short, both the expletive nó and the locative argument in LC communicate a discourse-related meaning of specificity in that it denotes the speaker’s ability to identify the specific eventuality reported.

Summing up, the complementary distribution of the locative argument in LC and nó, their incompatibility with topicalization, and their similarity in terms of the specificity requirement jointly suggest that the locative argument and nó target the same syntactic position.25 Thus, the question comes down to the syntactic position occupied by the expletive nó. Greco et al. (2017, 2018) argue that syntactically, the expletive nó functions as a subject and occupies the highest position in the inflectional layer, although it still merges below CP. However, Greco et al. (2017, 2018) remain agnostic about the precise merge position of the expletive, mentioning three possibilities: nó could merge either in Kiss’s (1996) RefP (a specialized projection hosting subjects endowed with a [+specific] feature), in a projection that also attracts referential subjects in categorical sentences (Cardinaletti, 2004) or in a dedicated SubjP for subjects à la Rizzi and Shlonsky (2006).

In this paper, we propose that these three possible positions of nó assumed by Greco et al. (2017, 2018) can be subsumed under a special subject position exhibiting mixed A/Ā-properties located between TP and CP investigated in Bošković (2024a).26 Based on this assumption, we argue that the raising locative argument in LC and the merger of the expletive nó both target the specifier of this hybrid A/Ā-projection. We spell out the derivational details of the movement to this A/Ā-projection based on the composite probing mechanism in Coon and Bale (2014) and Van Urk (2015).

4.3. The Derivation of LCs

One important implication of adopting Bošković’s (2024a) A/ĀP in the high TP-domain is that we need to abandon the traditional positional approach to A/Ā-distinction (Chomsky, 1981, p. 47; 1995, 2007, 2008; Déprez, 1989; Mahajan, 1990). According to the positional approach, A-positions include thematic positions and the subject position (i.e., spec,TP), while Ā-positions encompass positions in CP and the edge of a phase. However, the A/ĀP in the high TP domain does not fall into either of these two categories.

Therefore, to spell out the derivational details of the movement of the locative argument in LC, we adopt the featural approach to A/Ā-distinction in (51).

(51)	The featural approach to A/Ā-distinction:
	(Obata & Epstein, 2011; Van Urk, 2015; Fong, 2019)
	A and Ā-properties derive from properties of the attracting feature: A-positions are crea ted by the valuation of D/φ-features, whereas Ā-positions by that of δ-features (e.g., wh, topic focus).

Considering the A-properties (such as case, φ-agreement, and anaphoric binding) of Ā-topicalization in Dinka, Van Urk posits that the distinction between A and Ā is not an intrinsic characteristic of a syntactic position. Instead, A and Ā properties emerge as derivative by-products of the specific features that trigger movement and/or establish the position. In particular, A-positions are generated through the valuation of D/φ-features, whereas Ā-positions result from the valuation of δ-features (for example, wh, Focus, Topic, etc.). In this paper, D-feature is used in Vietnamese to be responsible for A-properties under the featural approach to A/Ā-distinction, in accordance with Preminger’s (2019) No-Null-Agreement Generalization against postulating φ-agreement in languages that lack exponence of φ-features systematically throughout an entire paradigm (e.g., Vietnamese). One important theoretical implication of (51) is that an operation would exhibit a hybrid of A/Ā-properties if it is driven by both A- and Ā-features. Van Urk proposes that A/Ā operations can be realized through Coon and Bale’s (2014) composite probing mechanism, which allows A-features and Ā-features to jointly probe from a single functional head, ultimately converging on a single goal that possesses both matching A-features and Ā-features.

As a consequence, operations guided by A/Ā-composite probing are expected to display a hybrid mixture of A/Ā-properties. This prediction aligns with the fronting of the locative argument in LC. Thus, we propose that Spec,A/ĀP hosts the locative argument in LC, and the head of A/ĀP is endowed with a composite A/Ā-probe [D, δ]. As for the δ-feature of this composite probe, we follow Bošković’s (2024a) characterization of Spec,A/ĀP as not having “the same featural content in all its instantiations”, and assume that the δ-feature can be instantiated by one of the following Ā-features: [uTopic], [uFocus] and [uwh]. In other words, the fronting locative argument in LC can deliver topicality, carry focus, or function as a wh-interrogative. As for topicality, recall that the fronting locative argument in LC exhibits a paradoxical pattern: it patterns with Ā-topicalization with respect to the specificity requirement as in (38), yet it resists the topic marker thì as shown in (25). We posit that this empirical dilemma can be reconciled by assuming that thì realizes the Topic head in the left periphery, and the specificity requirement of the fronting locative argument follows from the Topic feature on the head of the A/ĀP. Next, the compatibility of the locative argument with focus/wh-interrogative can be evidenced by (52) and (53): the locative argument may carry contrastive/identificational focus as in (52) or function as an interrogative wh-expression as in (53).

(52)	Locative arguments in LC are able to encode contrastive/identificational focus:
	Là	trên	bàn	để	sách	ngôn ngữ học	(chứ	không	phải	trên	ghế).
	be	on	table	place	book	linguistics	conj	neg	be	on	chair
	‘On the table are placed linguistic books (, not on the chair).’

(53)

Locative arguments in LC can function as an interrogative wh-expression:

Trên

cái

bàn

nào

để

sách

ngôn ngữ học?

on

cl

table

which

place

book

linguistics

‘On which table are linguistic books placed?’

The non-uniform pattern of the discourse function of the fronting locative argument is reminiscent of Van Urk’s (2015, p. 107) hypothesis that Ā-probes may be flat or relativized (Rizzi, 1997; Abels, 2012).27 Only a relativized Ā-probe is specified for what type of Ā-goal (topic, focus, or wh) is able to satisfy the probe. On the other hand, a flat Ā-probe is non-selective and free from such restrictions on the type of Ā-goal it requires. Van Urk demonstrates that the Ā-probe on the composite probe associated with C in Dinka is flat and non-selective, as the agreement correlation linked to topicalization also applies to relativization. Furthermore, topicalization and relativization exhibit complementary distribution in Dinka; that is, the flat and non-selective Ā-probe can be satisfied by either a topic or a relativized operator but not by both simultaneously. Following this line of analysis, we assume that the Ā-probe on the composite probe on the head of A/ĀP in Vietnamese is also a flat and non-selective probe.

With the hypothesis regarding composite probing on the head of the A/ĀP in place, we show the derivation of the fronting of the locative argument in LC. Note that the fundamental problem of the fronting of the locative argument within the generative framework is the featural basis that gives rise to the syntactic primacy of the locative argument to move over the DP to Spec,A/ĀP in (54).

(54)

A/Ā[uD, uδ]

……

[vP  v  [VP

DP

V

LP]]

In order for the locative argument to be considered a valid target for the composite probe, it must possess both a D-feature and a δ-feature. In terms of the featural composition of locative phrases, Grimshaw (1991) and van Riemsdijk (1990, 1998) argue that locative PPs are more referential compared to other types of PPs, categorizing them as extended nominal categories that carry an incomplete set of φ-features.

Building on the work of Grimshaw (1991) and van Riemsdijk (1990, 1998), Doggett (2004, pp. 34–35) observes that locative PPs differ from other prepositional phrases in their capacity to move to Spec,TP. This distinction is illustrated by the ungrammatical movement of the PP object in double object unaccusative verbs, as shown in examples (55) and (56).28

(55)	a.	The play appealed to Mary.
	b.	*To Mary appealed the play.

(56)	a.	The grade mattered to John.
	b.	*To John mattered the grade. (adapted from Pesetsky, 1995, p. 115)

Considering the difference between locative inversion and the examples in (55) and (56), Doggett (2004) argues that locative PPs must possess the ability to establish some form of agreement with T, which serves as the featural foundation for their movement to Spec,TP. In contrast, general prepositional phrases lack this capacity. Although Vietnamese lacks overt manifestation of φ-features/agreement, we follow the above-mentioned characterization of locative PPs to assume that the locative argument in LC bears a defective D-feature, à la Grimshaw (1991), van Riemsdijk (1990, 1998), and Doggett (2004) (see also Huang et al., 2009 on the analysis of localizers as a deficient sub-category of nouns in Mandarin Chinese). With the D-feature and an appropriate δ-feature compatible with the intended pragmatic function (i.e., topicality, focus, or wh-interrogative) of a locative construction, the locative argument is an eligible goal for the composite probe on the head of the A/ĀP as depicted in (57).

(57)

Finally, to move the locative argument to Spec,A/ĀP from within the vP phase (Legate, 2003), we adopt the weak version of Chomsky’s (2001) Phase Impenetrability Condition in (58). In effect, one derivational consequence of (58) is that vP becomes a phase only when the next phase head up (say, C) enters the derivation, thereby delaying the opaqueness of the VP complement of v. Thus, the locative argument DP within VP remains visible to the composite probe on A/Ā and is able to move to Spec,A/ĀP as in (59) because C has not entered the derivation.

(58)	Phase Impenetrability Condition: (Chomsky, 2001, p. 14)
	The domain of H is not accessible to operations at ZP; only H and its edge are accessible to such operations (where H is a phase head, and ZP is the smallest phase dominating the HP phase).

(59)

We want to end this section with two clarifications. The first clarification concerns whether the intervening DP induces a minimality problem for the raising of the locative argument to Spec,A/ĀP. It is important to note that neither the DP nor the locative argument fully satisfies the requirements of the composite probe. The theme DP is missing the δ-feature29, while the locative argument is defective in its D-feature. Consequently, it is not immediately clear why the locative argument in (59) should be considered the optimal goal that aligns with the composite probe.

Coon and Bale (2014) offer a formal calculus of minimality for composite probing, specifying that the ranking of potential goals for a composite probe is determined by multiplying the ranks assigned by each sub-probe, as outlined in (60). Crucially, if either sub-probe assigns a rank of 0 to a potential goal, the overall rank becomes 0 due to the multiplicative nature of the calculation. This means that failure to match one sub-probe results in failure to meet the requirements of the composite probe as a whole. Consequently, the DP is excluded as a valid goal for the composite D+δ probe on the head of A/ĀP because it lacks the δ-feature.

(60)	Conjoined Match of Rank n:
	A syntactic object σ is a match of rank n for a fused probe ρ composed of α and β, where i × j = n and where i is σ’s rank with respect to α and j is σ’s rank with respect to β.

Second, Bošković (2024a) views the locus of the EPP as not a fixed and unique position; instead, syntactic context determines the highest projection in an EPP domain, which is the position where the EPP is satisfied. Specifically, he states that the lower projection in the EPP domain, say Agrs seeking a nominal goal in (61a), is unable to trigger the movement of the wh-phrase because the wh-phrase is a QP, rather than a nominal to meet the probing requirement of Agrs, according to Cable (2010). As a consequence, the A/Ā head and Agrs in (61b) “undergo feature sharing in the sense of Frampton and Gutmann (2000), where two instances of an unvalued feature become one—when one of them gets valued they both do” (Bošković, 2024a, p. 51), so the A/Ā head, with its Ā-feature, is able to agree with the wh-expression and trigger the wh-movement.

(61)	Bošković’s (2024a, pp. 51–52) local computation of the contextuality of the EPP effect:
	a.  Agrs [vP  [QP who] …]
	b.  A/Ā [AgrsP  Agrs  [vP  [QP who] …]]

However, we do not see how this line of analysis can be extended to derive the contextuality of the EPP effect in the derivation of LC. In particular, when the T head enters the derivation of LC in (62a), the theme DP is clearly an eligible goal for the D-probe on T. Thus, the theme DP should undergo movement to merge with T, yielding a sentence similar to (62b). That is, Bošković’s (2024a, pp. 51–52) deduction of the contextuality of the EPP effect is unable to derive the fronting of the locative argument in LC such as (62c): the A/Ā head and T in (62) do not “undergo feature sharing in the sense of Frampton and Gutmann (2000)” because the theme DP satisfies the D-probe on T in (62b). Even if we allow the A/Ā head to attract the locative argument based on its Ā-probe, the resulting sentence in (62d) is ungrammatical.

(62)

a.

T[uD]

[VP

v  [VP   DP   V

LP[def-D, δ]]]

b.

[DP

Một

chồng

sách]

nằm

[LP

trên

bàn.]

a

pile

book

lie

on

table

‘A pile of books lies on the table.’

c.

[LP

Trên

bàn]

nằm

[DP

một

chồng

sách].

on

table

lie

a

pile

book

‘On the bed lies a pile of books.’

d.

*[LP

Trên

bàn]

[DP

một

chồng

sách]

nằm.

on

table

a

pile

book

lie

Therefore, although we adopt Bošković’s (2024a) contextual approach to the EPP effect in the derivation of LC, we do not follow his deduction outlined in (61), at least not for the derivation of LC. To derive the absence of the EPP effect at the TP level in the derivation of LC as in (63), we posit that the alternation between (62b) and (62c) boils down to the numeration choice for the derivation built based on a locative verb such as nằm ‘lie’. If the A/Ā head is not selected in the numeration, TP would be the highest projection in the EPP domain. As a result, the theme DP undergoes movement to merge with T, yielding (62b). The locative argument will raise to derive LC, such as (62c) if the A/Ā head is selected to enter the derivation.

(63)

5. Theoretical Implications

In this section, we discuss two theoretical implications of our proposal of deriving Vietnamese LC based on Bošković’s (2024a) A/ĀP and the composite probing mechanism in Van Urk (2015) and Coon and Bale (2014): the Merge over Move preference and the Ā-opacity of the locative adjunct to the composite probe on A/Ā.

5.1. Merge over Move

Chomsky (1995) argues for the Merge over Move preference in (64), where satisfying a requirement by Merge (external merge) is preferred over satisfying it by Move (internal merge) because the former is “simpler”.

(64)	Merge over Move (MoM):
	If, at some point in the derivation, Merge (for lexical insertion) and Move can both apply, then Merge always applies first.

For instance, Chomsky (1995) explains the contrast between (65b) and (65c) in terms of MoM: both merging the expletive there and moving a man can satisfy the EPP on the embedded T, but the latter option is ruled out because Merge preempts Move.

(65)	a.	[TP	A man1	seems	[TP	t1	to be	t1	outside]].
	b.	[TP	There1	seems	[TP	t1	to be	a man	outside]].
	c.	*[TP	There	seems	[TP	a man1	to be	t1	outside]].

However, this preference has been questioned because it is not easy to maintain if Move is just a sub-class of Merge (e.g., Castillo et al., 2009; Epstein & Seely, 1999, 2006; Bošković, 2002, 2007). Specifically, Chomsky (2000, 2001, et seq.) recasts Merge and Move as external Merge and internal Merge, respectively, treating them as two sub-classes of Merge, an elementary and generalized syntactic operation. The consequence of this conceptual unification is that the contrast between (65b) and (65c) can no longer be explained in terms of MoM.30 In fact, the considerations discussed in Chomsky (2019) would lead to preferring Move (internal merge) to Merge (external merge). That is, the opposite of (64) should be the preferred derivational choice when both Move and Merge are eligible operations to fulfill a syntactic requirement. Indeed, based on the evidence from reflexes of successive-cyclic movement in Wolof and Duala, Georgi (2014) argues that MoM is not a universal principle, and Move can precede Merge.

The derivation of LC offers additional evidence for the view that the application of movement could preempt the merger of an expletive. Specifically, at the derivational stage of LC in (66), the composite probe on the head of A/ĀP seeks an appropriate goal carrying matching A- and Ā-features. Assuming that the expletive nó is endowed with both a D-feature and a δ-feature (because of its specificity requirement), either moving the locative argument or merging the expletive nó may satisfy the composite probe on the head of A/ĀP.

(66)		A/Ā[uD, uδ] …   [VP v [VP DP V  LP[def-D, δ]]]
	a.	Move: Moving the locative argument
	b.	Merge: Lexical insertion of the expletive nó

If MoM is an effective principle in Vietnamese, (66b) would be predicted to be the next step in the derivation. However, this prediction is not borne out, as evidenced by the ungrammaticality of (67). Therefore, we agree with Georgi’s (2014) conclusion that MoM is not a universal principle, and Move may preempt Merge, as demonstrated by the choice of (66a) over (66b) in the derivation of LC.

(67)	*Nó	nằm	một	chồng	sách	trên	bàn.
	expl	lie	a	pile	book	on	table
	Intended: ‘There lies a pile of books on the table.’

5.2. Ā-Opacity

Interestingly, merging the expletive at Spec,A/ĀP is a grammatical step for the derivation of EC, as demonstrated in (68), in sharp contrast to (67).

(68)	Nó	có/	xuất hiện	một	chồng	sách	trên	bàn.
	expl	exist	appear	a	pile	book	on	table
	‘There is/ appears a pile of books on the table.’

Given that the minimal difference between (67) and (68) lies in the argument/adjunct status of the locative phrase trên bàn ‘on table’, the contrast between LC and EC with respect to the grammaticality of inserting the expletive at Spec,A/ĀP suggests the ineligibility of the locative adjunct to serve as a goal for the composite probe on the head of A/ĀP. Concretely, assuming that the locative adjunct is right-adjoined to VP in the derivation of EC (69a) is not a possible option at this stage of the derivation of EC.

(69)		A/Ā[uD, uδ]  …  [VP    v   [VP DP V] LP[def-D, δ]]]
	a.	Move: Moving the locative adjunct
	b.	Merge: Lexical insertion of the expletive nó

In addition, recall that the fronting locative adjunct in EC has to precede the expletive, as shown by the contrast in (37), repeated below as (70). Therefore, it is not the case that the locative adjunct is immobile in the presence of the expletive. It just cannot move to Spec,A/ĀP.

(70)

The locative adjunct in EC precedes the expletive: (adapted from Greco et al., 2018, p. 44)

a.

Trên

bàn

(thì)

nó

sẽ

không

có

cái

bút

nào.

on

table

top

expl

fut

neg

exist

cl

pen

any

‘On the table, there will be no pens.’

b.

*Nó

trên

bàn

(thì)

sẽ

không

có

cái

bút

nào.

expl

on

table

top

fut

neg

exist

cl

pen

any

Intended: ‘On the table, there will be no pens.’

We want to point out that the opaqueness of the locative adjunct to the composite probe on the head of the A/ĀP is similar to the phenomenon of Ā-opacity discussed in Rezac (2003). Ā-opacity can be instantiated by the Icelandic sentence in (71), where the Ā-moved object quantifier engu grjóti ‘no rock’ appears opaque (i.e., invisible) as a (closer) goal for the φ-probe on T (see also Jayaseelan, 2001; Obata & Epstein, 2011). In short, the issue of Ā-opacity pertains to why Ā-interveners do not block A-agreement relations.

(71)	Ā-opacity in Icelandic: (Svenonius, 2000; cited in Rezac, 2003, p. 178)
	Strákarnir2	höfðu	[engu	grjóti]1	[vP  t2	[VP	hent  t1	í	bílana]].
	the.boys	had	no	rock			throw	in	the.cars
	‘The boys had thrown no rocks at the cars.’

Importantly, this Ā-opacity effect is by no means a language-particular problem encountered only in specific constructions because it is also evident in other general derivations in other languages. For example, Obata and Epstein (2011, p. 130) note that the Ā-opacity effect also manifests itself in the derivation of wh-object movement similar to (72). In their own words, “when T (does) φ-agrees with the subject John, the intervening DP what cars bearing ‘‘plural’’ can be mysteriously skipped. As in Icelandic data, elements at Ā-positions do not function as interveners of T-agreement”.

(72)

Similarly, the derivation of internal object topicalization in Chinese, such as (73), also exemplifies the Ā-opacity effect (Shyu, 1995; Ting, 1995; Paul, 2005; Kuo, 2009; Chen, 2023). In particular, why does the closer object DP not cause an interference problem for the D-probe on T to find and attract the lower subject DP, as shown in (74)?

(73)	Internal topicalization:
	Lisi	zhe-ben	shu1	zaojiu	kan-guo	t1	le
	Lisi	this-cl	book	already	read-exp		sfp
	‘As for this book, Lisi has already read it.’

(74)

Now, notice the parallelism between the Ā-opacity effect in (68)–(71) and the opacity of the locative adjunct to the composite probe on A/Ā in EC in (66)/(67). An A-probe (whether it is contained in a composite probe) skips an appropriate goal when the goal is at an Ā-position. Although we do not have a principled account for the Ā-opacity effect, a possible line of analysis is to attribute the opacity to the ban on improper movement, excluding the derivational sequence of the Ā-A movement (Chomsky, 1973; May, 1979; Chomsky, 1981; Abels, 2007; Neeleman & Van De Koot, 2010; Williams, 2011). Pursuing this line of analysis goes beyond the scope of this paper, so we leave it to future studies.31

6. Conclusions

The hybrid A/Ā properties of English locative inversion present a notable analytical challenge within the generative framework. For instance, Bresnan (1994) argues that English locative inversion cannot be adequately derived within Chomsky’s generative framework, which posits that all human languages utilize universal syntactic operations on underlying structures (see also Postal, 2004). Two major analytical challenges posed by English locative inversion are (i) why the fronting locative phrase does not behave as a canonical subject (e.g., lack of agreement; see Bruening, 2010 for further discussion), and (ii) how the movement of the locative phrase, if analyzed as an instance of A-movement targeting Spec,TP, is able to move across the structurally higher theme DP, which canonically does so. To address these puzzles, some researchers have proposed a null expletive analysis, suggesting that Spec,TP is occupied by a null expletive, with the locative phrase undergoing Ā-topicalization (e.g., Postal, 1977, 2004; Maruta, 1985; Coopmans, 1989; Hoekstra & Mulder, 1990; Bruening, 2010).

In this paper, we demonstrate that the null expletive analysis cannot be extended to account for the derivation of LC in Vietnamese. Specifically, the fronting locative argument in LC resists marking by thì, representing the Topic head in the left periphery (see (25a)). Moreover, the postulation of a null expletive in Vietnamese is implausible in light of the incompatibility of LC and the expletive nó (see (45)). After establishing that the null expletive analysis is unsuitable for LC in Vietnamese (see Diercks, 2017 for cross-linguistic considerations), we explore parallels between the fronting locative argument and the expletive nó with regard to their shared specificity requirement (see Section 4.2). These findings lead us to a dilemma: the fronting locative argument in LC is not a topic in the left periphery, yet it patterns with the expletive nó regarding the specificity requirement, a property that is aligned with Ā-topicalization. We suggest a solution based on three assumptions. First, we adopt Bošković’s (2024a) proposal of three distinct subject positions between TP and CP across languages and argue that both the fronting locative argument and nó target the specifier position of the A/ĀP, the uppermost of these subject positions. Second, to clarify the derivational process of locative inversion in LC, we adopt the featural characterization of the A/Ā-distinction (Obata & Epstein, 2011; Van Urk, 2015; Fong, 2019) and assume a composite probe [D, δ] on the head of the A/ĀP. Finally, we show that Coon and Bale’s (2014) formal calculus of minimality enables the composite probe to target and move the locative argument to Spec,A/ĀP over the structurally higher theme DP, achieving the desired derivation.