Whys and Wherefores: The Aetiology of the Left Periphery (With Reference to Vietnamese)

Abstract

This paper offers a detailed description of the left periphery of embedded clauses in Vietnamese. Five kinds of pre-subject constituent are considered in isolation, and in interaction with one another: subordinating conjunctions; embedded topics; fronted quantifier expressions; fronted adverbials, and the Vietnamese equivalent of English why (Italian perché). A systematic comparison is made with the functional sequence of Italian, proposed in the cartographic literature. Whilst largely consistent with the Italian pattern, our findings diverge in certain respects, especially in suggesting a modification of previous treatments of the *‘why-to’ constraint observed in English and a number of other varieties.

• Pour l’enfant, amoureux de cartes et d’estampes,
• L’univers est égal à son vaste appétit.
• Ah! que le monde est grand à la clarté des lampes!
• Aux yeux du souvenir que le monde est petit! …

•   To the child, enamoured of prints and maps
•   The universe has the size of his vast appetite.
•   How large the world seems by the light of a lamp!
•   How small it is, now, in memory’s sight!

•   Charles Baudelaire, Le Voyage [first stanza]
(Les Fleurs du Mal, 1861)

1. Introduction: Why Peripheral, Why?

In this paper I direct attention to two whys: the theoretical question “Why CP?” and a set of questions concerning the left periphery in Vietnamese, ultimately focusing on the contrastive behaviour of the lexical item tại sao (why, perchè), relative to all other wh-expressions. Whilst the former question is paramount, I will argue that an answer to it is best arrived at by understanding the latter contrasts: hence, the empirical focus of this paper is why~tại sao, and on its interaction with other elements of the left periphery (in English and Vietnamese, respectively).

Let us begin, though, with the conceptual questions facing any cartographic approach to word order, namely, why is CP where it is? Up there? On the left (or sometimes, right)1 periphery? Indeed, Why is CP considered ‘peripheral’ at all—part of another territory in the satirical map of The Derivation in Figure 1, posted more than a decade ago?

Designations matter greatly, both for analysis and theory construction. Although strictly speaking the term periphery denotes simply the edge or circumference of an object or region (compare the Parisian Périphérique), and is neutral between an inclusive (‘integral’) and an exclusive (‘adjacent but separate’) denotation, the latter—non-integral, interpretation—is the more salient in normal discourse; in the case of computer hardware, for example, peripherals are additional, non-essential components. In other domains of inquiry, essential components are never treated as ‘peripheral’, no matter their distance from the centre. Aeronautical engineers do not consider the nose cone or empennage to be ‘peripheral’ structures on an aircraft, even though the tailplane is not an integral part of the fuselage. No one—other than the thoracic surgeon, perhaps—conceives of the head as peripheral to human anatomy, despite its being non-integral to the torso; this is in obvious contrast to how we think about the limbs (which are more intricately attached). In syntactic theorizing, on the other hand, generativists have tended to treat CP as external to the root: indeed, in the literal chart reproduced as in Figure 1, the left periphery is drawn as another country, something best left alone. (Which it often was, until Rizzi (1997), except whenever subordinate clauses—and extractions therefrom—were at issue (Chomsky, 1977, 2000, 2005, 2008).)2

The original impulse to exclude CP no doubt stems from the earliest conceptions of re-write grammars, in which S was taken as the initial symbol (Chomsky, 1957, 1965):3 if your top-down theory of phrase structure (“S → NP VP”) has you start with what is now construed as a lower node (S = IP~TP), then everything to the left of this node inevitably becomes problematic (in a right-branching language; see Note 1). Subsequent Minimalist theorizing—by definition, proceeding from the bottom up—generally has had little to say about the derivation of root clauses beyond TP, except where features of a monolithic C are taken to trigger subject raising (Duffield, 2024), or in cases of A’-movement.4

It is worth noting, by contrast, that lower functional categories such as TP and NegP have always been treated as integral to S, even though they are also semantically peripheral to the core proposition: just like CP, TP is excluded from the core thematic domain (vP^max, cf. “Complete Functional Complex” (Chomsky, 1985)). The inclusion of TP presumably has to do with the fact that bare propositions ‘John love Mary’ and its permutations {SOV, VSO, VOS…} are grammatically unacceptable after the age of two—Tarzan movies, aphasic patients, and special learner populations aside; see Deprez and Pierce (1993), Guilfoyle and Noonan (1992), though cf. Chung and McCloskey (1987). In short, there is consensus that TP is required in root clauses, for grammatical well-formedness, and, since TP invariably dominates Aspect phrase, the latter projection is perforce also considered as integral—within, not without. Compare the location of ‘The United States of Tense and Aspect’ in Figure 1 above.5

Developing recent trends in cartography, the conclusion here will be that some CP components are required for syntactic well-formedness, even in root clauses (and even if these elements often go unpronounced). In other words, CP may be top-most, but at least some sections of it are non-peripheral. At the same time, other elements may in fact lie beyond the periphery. In both directions, then, Vietnamese data suggest that borders need to be redrawn.

The Cartographic Turn

Of course, syntactic cartography has hardly ignored CP: the clausal periphery has been grist to the cartographic mill for nearly 25 years (see Rizzi, 1997; Cinque, 1999), not to mention multiple articles in the sub-series The Cartography of Syntactic Structures (Oxford University Press). This prodigious output notwithstanding, researchers have focused almost exclusively on What (~Which)?, Where? and How? questions; specifically, “Which grammatical elements comprise the left periphery?”; “Where is a particular element located vis-à-vis other CP constituents?”; “How did that element come to occupy its position, by movement or base-generation (internal or external Merge, in Minimalist terms)?”

Rizzi and Bocci (2017) offer a masterful exposition of this approach, for Italian: the structure in (1a) summarizes their answers to the first two questions (What? and Where?), the how? question being addressed by postulating specific kinds of feature-driven movements, all constrained by the formal principle of Criterial Freezing (1b)—itself a variant of other previously entertained “economy” conditions on movement. Just as the greater part of Rizzi and Bocci’s (2017) paper is a descriptive commentary on these claims, using Italian data, so most of this paper is an examination of the corresponding Vietnamese facts.6

(1)	a.	[ Force [ Top* [ Int  [ Top* [ Foc [ Top* [ Mod [ Top* [ Qemb [ Fin [IP … ] ] ] ] ] ] ] ] ] ] ]  (From Rizzi & Bocci, 2017)
	b.	(42) Criterial freezing: A phrase meeting a criterion is frozen in place

Given that I will also consider how adverb placement interacts with the functional sequence in (1), it is useful to introduce Cinque’s Adverb Hierarchy here, reproduced in (2), which is simultaneously a hierarchy of functional heads and adverbial positions:

(2)	Moodspeech act > Mood evaluative > Moodevidential > Modepistemic > T (Past) > T (Future) > Mood (ir)realis > Modroot/Aspecthabitual/T (Anterior) > Aspectperfect > Aspectprogressive/ Aspectcompletive > Voice > V.

Now, this concern with What? Where? and How? is all to the good, since, at least from an inductive perspective, What? Where?, and How? questions are logically prior to Why?7 The problem is that cartographers—with the notable exception of Rizzi (2017) to which I will turn directly—are insufficiently concerned with what might be termed as ‘mid-level’ Why questions, even once the map is drawn.8

In the case of syntactic cartography, mid-level questions fall into two general categories: broad-span questions concerning general clausal architecture, and narrow-span questions that should explain cross-linguistic variation within a particular sub-domain. Broad-span Why-questions relate to the universal layering of syntactic domains diagrammed in (3): why, for example, CP elements are always projected higher than TP {CP₁ CP₂, CP₃…TP₁, TP₂, TP₃, }, rather than in inverse, or intercalated, orders {TP₁ TP₂, TP₃…CP_1, CP₂, CP₃, }, {CP₁ TP₂, CP₃…TP_1, CP₂, TP₃ }; why thematic subjects always raise out of the predicate phrase, even in VSO languages (McCloskey, 1996, 2017); why T is supervenient on ASP; conversely, why T and Neg vary in their relative order cross linguistically (Ouhalla, 1991); why almost all lexical categories appear below almost all functional categories (an exception being “Inner Aspect” projections, see above; see also (Phan, 2013, 2024, for discussion of Inner Aspect in Vietnamese.).9^,10

(3)

More specific, ‘narrow span’ questions include the following: given a set of morphemes in a particular language appearing at the left edge of the clause, why are certain elements supervenient on other elements? In the figure in (4), for example, why must τ c-command α, and α, γ within a given subdomain?

(4)

On such questions cartographers are largely silent.11 To my knowledge, the first discussion of the aetiology of CP at this level of explanation is found in the preliminary speculations in Rizzi (2017), who writes:12

“Why is it that we typically find certain properties of ordering and cooccurrence restrictions, rather than others?…[…] Two broadly defined candidates come to mind:13

i.

Certain properties could derive from requirements of the interface systems. For instance, it could be that functional head B may necessarily occur under functional head A (thus giving the linear order AB in head-initial languages and BA in head-final languages) because the opposite hierarchical order would yield a structure not properly interpretable…[…]…A special [my emphasis: NGD] case of the impact of interface requirements may be the ordering properties that follow from selectional requirements, e.g., the fact that the Force head in embedded clauses must be high enough to be accessible to higher selectors, which want to know if their complement is a declarative or a question, for instance (Rizzi, 1997);

ii.

When the functional heads occurring in specific orders trigger movement, the ordering may be a consequence of locality requirements. For instance, Abels (2012) has argued that almost all the ordering effects observed in the Italian left periphery may follow from the theory of locality based on a version of featural Relativized Minimality, along the lines developed in Starke (2001), Rizzi (2004): if A is a stronger island-creating element than B, then B will not be extractable from the domain of A, neither long-distance, nor locally…”

Setting aside the issue of selection—which applies only to a subset of subordinate clauses, and so is exactly a “special” (i.e., non-typical) case—neither of these responses is satisfactory. In the absence of a deeper explanation as to why “functional head B necessarily occurs under functional head A, alternatively, “why A is a stronger island-creating element than B [my emphases: NGD]”, purely formal explanations do little more than to restate the distributions they are supposed to explain (cf. appeal to ‘strong’ vs. ‘weak’ features in earlier treatments of verb movement, see, e.g., Chomsky (1989, 1993)).14

The alternative, functionalist proposal I have been promoting recently is cast in terms of four principles, two of which—Supervenience and ‘1-Arg’—offer a partial solution to the mid-level questions introduced above; see Duffield (2022). In this paper, I shall focus attention instead the core data question: “Wherefore is why different?”. And before that, we have some mapping to do.15

2. Vietnamese Clause Structure

As may be familiar, Vietnamese is distinguished from other, better known, MSEA languages in having a rich inventory of pre-subject constituents, whose relative distributions provide evidence of an articulated left periphery.16 While this super-structure is similar in many ways to that proposed by Rizzi and Bocci (2017), based on Italian, the Vietnamese facts presented below suggest significant points of divergence.

Notionally, we may distinguish five classes of pre-subject constituents, as follows:

i.

Clause initial conjunctions—including various kinds of subordinating conjunction {complementizer, relativizers}, topic markers, and linking elements. These elements have a fixed distribution relative to one another, and are never found lower in the clausal hierarchy (to the right of the subject): {rằng, liệu, (mà),17 (là)}, {thì, (là)}.

ii.

Left peripheral phrasal constituents that are amenable either to a movement or to a co-indexation analysis (Move or Merge): {topicalized XPs, NPs heading relative clauses (see note 1)}.

iii.

Speaker/subject-oriented adverbial phrases (e.g., quả thật, ‘indeed’).

iv.

Weak indefinites interpreted as universally quantified XPs (which in all probability have been moved to a pre-subject position, given the distribution of the same constituents in the general case);

v.

The wh-phrase why (tại sao), which—uniquely in this wh-in situ language—only ever appears to the left of the subject.

I consider each these classes in turn, subsuming the second category within the first: that is to say, I will assume that the overt phrasal constituents in (ii) occupy the ‘specifier’ position of their corresponding licensing heads (XP-thì, XP-mà, in the case of topicalization and relativization, respectively.)18

2.1. Clause-Initial (Pre-Peripheral?) Conjunctions

Although our core focus is with the last of these classes, it is best to start at the top, with the head elements that provide most direct evidence of a functional sequence on the left periphery. The examples in (5) illustrate the canonical distributions of all the Class (i) heads under consideration here:

(5)

a.

Tôi

nói

rằng

[ tôi

là

[cán bộ

ngọai giao

[mà

cần

liên hệ

với

sứ quán.]]]

PRN

say

??

I

COP

staff

foreign.affairs

REL

need

contact

with

embassy

‘I said that I was a diplomatic staff member who needed to contact the embassy.’

b.

Cô

gái

hỏi

liệu

[ cô

có thể

đi

đến

bữa tiệc

được

không. ]

PRN

girl

ask

??

PRN

Q? poss.

go

arrive

party

CAN

NEGQ

‘The girl asked if she could go to the party.’

c.

[Quyển

sách

[ mà

[ anh

thích

nhất ] ]

thì

bán

chạy.

CLF

book

REL

PRN

like

best

TOP

sell

run

‘The book that you like most, (it) is selling well.’

2.1.1. Conjunctions I: {rằng, liệu, là}

Let us take {rằng, liệu} first. As these occur as the most peripheral elements in any embedded sequence and signal the illocutionary force of the following clause, a reasonable assumption is that they are Force heads, contrasting only in selectional features (+IND, +INT): that is, it might be thought that rằng introduces declarative clauses, whereas liệu introduces interrogatives. However, though it is initially plausible to associate both elements with Force, the examples in (6) and (7) demonstrate that liệu and rằng co-occur. More surprising perhaps, both word orders are observed—whereas liệu precedes rằng in (6), the opposite order is found in (7)—even while the unacceptability of the (b) and (d) examples in each set shows that this is not a matter of free variation.

(6)

a.

Ði

tìm

lời đàp:

Liệu rằng

tia UV

có

làm

filler

biến dạng?

go

find

answer:

ray UV

ASR

make

filler

deform

‘Finding the answer: do UV rays cause filler deformation?’19

b.

Ði

tìm

lời đàp:

*Rằng liệu

tia UV

có

làm

filler

biến dạng?

go

find

answer:

ray UV

ASR

make

filler

deform

(as a)

c.

Liệu

rằng

khi

mất đi

em

có

còn

hối tiếc? [song lyric]

Q

COMP

when

lose

PRN

ASR

still

regret?

‘Will I regret it when I lose you?’

d.

*Rằng liệu

khi

mất đi

em

có

còn

hối tiếc?

COMP Q

when

lose

PRN

ASR

still

regret?

(as c)

(7)

a.

John

hỏi

rằng liệu

[tôi

có

muốn

hẹn hò

với

anh ấy

không]

John

ask

I

Q

want

go.out

with

PRN.DEM

NEGQ

‘John asked if I wanted to go on a date with him.’

b.

*John

hỏi

liệu rằng

[tôi

có

muốn

hẹn hò

với

anh ấy

không]

John

ask

I

Q

want

go.out

with

PRN.DEM

NEGQ

(as a)

c.

John

đã

hỏi mẹ

rằng liệu

[mẹ

có thể

đón

cậu bé

không]

John

ANT

ask mother

mother

mother

child

child

NEGQ

‘John asked his mother if she could pick him up.’

d.

*John

đã

hỏi mẹ

liệu rằng

mẹ

có

thể đón

cậu bé

không.

John

PAST

ask mother

mother

can

pick up

child

NEGQ

(as c)

The clue to this puzzle lies in the fact that both liệu and rằng also serve as matrix predicates in certain contexts. In the examples in (6), for example, liệu works in this fashion: in fact, there are no cases of {liệu…rằng} sequences where liệu itself is preceded by another matrix predicate. Considering this, it is plausible to analyse liệu in (6c) as extra-clausal: this is diagrammed in (8) (cf. French est-ce (que)):

(8)

As for rằng, etymologically this element is derived from a matrix verb, synonymous with nói (‘speak, say’) The diachronic source of rằng, as a matrix predicate, is illustrated by the proverb and folktale examples in (9) below:

(9)

a.

‘Chẳng

nói,

chẳng rằng!’

NEG

say,

NEG speak

‘Say nothing!’

b.

‘…Phú

ông

xin

đổi

ba

bò,

chín

trâu…’

…rich

man

ask

exchange

three

cows

nine

buffalo

‘The rich man asked to exchange three cows and nine buffaloes…’

…Bờm rằng:

Bờm

chẳng

lấy trâu…

…Bom say:

Bom

NEG

take buffalo

‘..(and) Bom said: “I (Bom) will not take the buffaloes.”’

Given its verbal origins, it is reasonable to analyse contemporary examples where nói rằng introduces direct speech, and is immediately followed by a colon or comma—examples such as those in (10)—in terms of a similar matrix analysis.

(10)

a.

Murakami

Haruki

nói

rằng:

“Tôi

“Tôi

bạn

biết

tôi

thích bạn …..”20

Murakami

Haruku

say

??:

“I

let

friend

know

I

like friend…

‘Haruki Murakami said: “I’m telling you that I like you…”’

b.

Tôi

thấy

lỗi

nói

rằng:

“Một

số

quy tắc

không

được

áp

dụng nhất”

I

see

error

say

??:

1

no.

rule

NEG

CAN

apply

correct most

‘I see an error saying “A number of rules are not most applicable”’21

Two variant possibilities present themselves: nói rằng might be analysed as a verbal doublet, diagrammed in (11a); alternatively, nói might function intransitively, with rằng heading a VP-adjunct, as in (11b). This would then be akin to the most likely analysis of the Early Modern English {…spoke, saying…} constructions in (12):

(11)	a.
	b.

(12)	a.	“…and he spoke, saying: These are the horns which have scattered Juda every man apart, and none of them lifted up his head …(Zechariah 1:21 KJV)”
	b.	But the king spoke, saying to Daniel, “Your God, whom you serve continually, He will deliver you…(Daniel 6:16 KJV)”

Whilst these extra-clausal analyses may deal with {liệu…rằng} order, as well as with some instances of ‘reporting’ rằng, it is nevertheless clear that in the general case this element lies within the periphery: that is to say, its etymology notwithstanding, rằng has been grammaticized. This is strongly implied by the fact that it appears following ‘non-reporting’ predicates—i.e., those that cannot take quoted speech as a complement. This set includes the factive predicates tin (‘believe’), nhận ra (‘realize’), bực tức (‘resent’), and tiếc (‘regret’), discussed in the section on embedded topicalization below.22

In addition, rằng appears in concessive clauses as part of the complex conjunction tuy rằng (=‘although’, cf. French bien que), where, once again, no verbal report is involved.

(13)

a.

Nó

không

nói

gì,

[ tuy rằng

[ nó

biết

rất rõ  ]].

PRN

NEG

say

what,

[ although

PRN

know

very clear

‘He didn’t say anything, even though he clearly knows (the answer).’

b.

[Tuy rằng

[nhớ

tên ]]  [

nhưng

lại

không

nhớ

ra

gương mặt ]

although

recall

name

but

again

NEG

recall

out

face

‘I remember the name, but not the face.’

So, if rằng is located within the complement clause in default contexts, lexicalising Force, then in examples such as those in (7) above, where rằng precedes liệu in non-reporting contexts, it follows that liệu must occupy a lower functional position within CP.

Within Rizzi & Bocci’s cartography, Int(errogative) presents itself as a prime candidate: liệu would then be equivalent to Italian se (cf. Rizzi & Bocci, ibid.). This analytic possibility is diagrammed in (14):

(14)	Positioning rằng and liệu (first pass)23

It is here that matters become interesting. For just as rằng and liệu are distinguished, so too are rằng and là, whenever the latter element behaves as a conjunction. In (5a) above we saw that là functions canonically as a copula, linking subjects and nominal predicates; see also the examples in (15). Crucially, in its copular function, là may not appear in construction with other kinds of non-verbal predicate, neither with adjectival (individual-level or stage-level) predicates (16), nor with prepositional predicates (17); examples taken from Phan and Duffield (2022).24

(15)

a.

John

là

người

đàn ông

đứng

ở

đằng

kia.

John

COP

person

gentleman

stand

LOC.COP

location

there

‘John is the person standing over there.’

b.

Người

đàn ông

đứng

ở

đằng

kia

là

John.

person

gentleman

stand

LOC

location

there

COP

John.

‘The person standing over there is John.’

(16)	a.	Anh ấy	(*là)		cao/thông minh/tốt.			*[AP]
		PRN.DEM	COP		tall/intelligent/nice
		‘He is tall/intelligent/nice.’
	b.	Cô ấy	(*là)	bận/		rất vội vàng/	rất vui.
		PRN.DEM	COP	busy/		very hurry/	very happy
		‘She is busy/in a great hurry/very happy.’

(17)

a.

Anh ấy

ở/(*là)

trên

tàu.

*[PP]

PRN.DEM

LOC.COP/COP

on

train

‘He is on the train.’

b.

Tiền

hoàn

thuế

của tôi

ở/(*là)

đâu?

money

complete

tax

POSS 1.SG

LOC.COP/COP

where

‘Where is my tax refund?’

c.

Nhìn xem!

Anh ấy

ở/(*là)

kia kìa.

look see

PRN.DEM

LOC.COP/COP

DEM3.DEM3

‘Look! He’s over there.’

Rằng can never replace là in its copular function. However, the opposite substitution does seem to be possible, since là alternates with rằng as a subordinating conjunction in a variety of contexts. The examples in (18) show that là readily introduces complement clauses, especially in spoken registers.

(18)

a.

Yui

đã

nói

là

đi

đến

nhà

của

bạn.25

Yui

PAST

say

CONJ

go

to

house

POSS

friend

‘Yui said (she was) going to her friend’s house.’

b.

Tôi

biết

là

chị

đang

yêu

một

ngừoi.

I

know

CONJ

she

PROG

love

1

person

‘I know that she is in love with someone.’

c.

Lan nói

là

(chị ấy)

thích

học

tiếng

Anh.

Lan say

CONJ

PRN.DEM

like

study

lge.

English

‘Lan said she liked learning English.’

However, just as was observed with liệu, no true substitution is involved: là once again occupies a functional position lower than rằng. This is evidenced by the examples in (19) below, which show that rằng and là may co-occur, abeit only in one order, namely, {rằng…là}. (Example (19c) shows that {rằng..là} collocations contrast directly with {*mà..l})} sequences: là cannot occur within relative clauses, nor may it substitute for mà as a relativizing head.)

(19)

a.

Phải

nói [

(*là)

[ rằng

[( là)

thế hệ

trẻ

của chúng.ta

rất tài năng.]]]26

must

say

??

COMP

??

generation

young

POSS PRN

very talented

‘(I) have to say that our young generation is very talented.’

b.

Bạn

có thể

nói

[ (*là)

[ rằng

[ là

mình

ổn,

nhưng…]]]27

friend

possible

say

??

COMP

??

self

fine,

but…

‘You can say that you’re fine, but…’

c.

Quyển

sách [

(*là)

[ mà

[ (*là)

anh

thích

nhất ]]]

thì

bán

chạy.

CLF

book

??

REL

??

PRN

like

best

TOP

sell

run

‘The book that you like most, (it’s) selling well.’

A further point to observe is whilst là may introduce interrogative clauses, it differs from liệu in having a “quasi-quotative”, rather than subordinating, function: that is to say, the question that follows là is interpreted as ‘direct-reported’, not indirect.28 In this respect, là functions similarly to colloquial English like in (20a), where inversion (SAI) signals the interrogative force of the following clause, whilst, at the same time, the sequence of tense rules and deictic shift indicate that this is not direct speech (cf. *20b vs. 20c):

(20)	a.	‘She asked me, like, What was I doing there?’	[direct reported]
	b.	%She asked me what was I doing. [*in standard varieties]	[indirect]
	c.	She asked me: “What are you doing here?”	[direct]

In short, là has a semantic status intermediate between rằng and liệu: whereas rằng is typically associated with propositional assertions, and liệu with alternative questions, là is apparently neutral between the two types of complement. This is suggested by the minimal contrasts in (21), involving không biết (‘not know’), which, like English know, can take either a declarative or interrogative complement: the corresponding examples in (22) suggest that là is equipotent:

(21)

a.

Người mẹ

không

biết

*rằng/liệu

bọn trẻ

đang

ngủ

hay thức.

mother

NEG

know

C/INT

children

PROG

sleep

or awake

‘The mother doesn’t know whether the kids are sleeping or awake.’

(= the mother is unsure about whether her kids are asleep.)

b.

Người mẹ

không

biết

[ rằng/*liệu

[ bọn trẻ

có

đang

ngủ.]]

mother

NEG

know

C/INT

children

ASR

PROG

sleep

‘The mother doesn’t know that the kids are in fact sleeping.’

(= the kids are in fact asleep: the mother doesn’t know that fact.)

c.

Cô ấy

không biết

[ *rằng/liệu

[ tôi

có

đang

vui vẻ

PRN.DEM

NEG know

C/INT

I

Q

PROG

happy

tại

bữa tiệc

không ]]

at

party

NEGQ

‘She doesn’t know whether I am having fun at the party.’

(= it is unclear to her whether I am having fun.)

d.

Cô ấy     không biết

[ rằng/*liệu

[ tôi

đang

vui vẻ

tại

bữa tiệc]]

PRN.DEM NEG know

C/INT

I

PROG

happy

at

party

‘She doesn’t know that I am having fun at the party.’

(= I am having fun, indeed, but she doesn’t know that fact.)

(22)

a.

Người mẹ

không

biết

là

bọn trẻ

đang

ngủ

hay

thức.

mother

NEG

know

??

children

PROG

sleep

or

awake

‘The mother doesn’t know whether the kids are sleeping or awake.’

(= the mother is unsure about whether her kids are asleep.)

b.

Người mẹ

không

biết

là

bọn trẻ

có

đang

ngủ.]

mother

NEG

know

??

children

ASR

PROG

sleep

‘The mother doesn’t know that the kids are sleeping.’

(i.e., the kids are in fact asleep: the mother doesn’t know that fact)

c.

Cô ấy

không

biết

là

tôi

đang vui vẻ

tại

bữa tiệc

PRN

NEG

know

??

I

PROG happy

at

party

hay

không.29

or

NEG

‘She doesn’t know whether I am having fun at the party or not.’

(= it is unclear to her whether I am having fun.)

d.

Cô ấy

không

biết

là

tôi

đang

vui vẻ

tại

bữa tiệc.

PRN.DEM

NEG

know

??

I

PROG

happy

at

party

‘She doesn’t know that I am having fun at the party’

(I am having fun, indeed, but she doesn’t know that fact.)

Moreover, it turns out that conjunction là has a larger set of distributional privileges than either rằng or liệu. Another significant difference is that là introduces the complement of certain epistemic predicates such as those in (23), contexts where both rằng and liệu are excluded. It is likely no coincidence that many of these are raising predicates: the raised counterparts are illustrated in (24). This is exactly what one would expect if là occupies a lower node than Force, which seems to be the projection that blocks A-movement in English.

(23)

a.

?Có

nên

*rằng/là

đàn ông

là

trụ cột

gia đình?30

ASR

MOD

men

COP

pillar

family

‘Should the man be the family breadwinner?’

b.

Có

phải

*rằng/là

cái

chết

là

chấm hết?

ASR

right

CLF

death

COP

end final

‘Is Death really the end?’

c.

Chắc

*rằng/là

anh ấy

bḷ

tắc

đường

nȇn

đến mưộn.

Likely

PRN.DEM

AUX

delay

road

MOD

come late

‘Maybe he was stuck in traffic, so he was late.’

d.

Nhà hàng

kia

đông

quá!

Chắc

*rằng/là

đồ ăn

ngon

lắm!

restaurant

DEM3

full

very

likely

food

delicious

very

‘That restaurant is so crowded! The food must be very good.’

(24)

a.

Đàn ông

có

nên

là

đàn ông là

trụ cột

gia đình?31

man

ASR

MOD

COP

pillar

family

‘Should men be the family breadwinners?’

b.

Cái

chết

ASR

phải

là

cái chết   là

chấm

hết?

CLF

die

ASR

MOD

COP

end

final

‘Is death the end?’

c.

Anh ấy

chắc

là

bḷ

tắc

đường

nȇn

đến

mưộn.

PRN.DEM

likely

CONJ

AUX

delay

road

MOD

come

late

‘Maybe he was stuck in traffic, so he was late.’

d.

Nhà hàng

kia

đông

quá!

Đồ ăn

chắc

là

ngon lắm!

restaurant

DEM3

full

very

food

likely

CONJ

delicious very

‘That restaurant is so crowded! The food must be very good.’

The VN conjunction là differs in subtle but crucial respects from its putative Italian counterpart di. Comparing the Italian examples in (25), involving rare (‘seem’), it can be seen that whereas raising is allowed out of bare non-finite complements (25a), di blocks A-extraction (25b) just as extraction is blocked from the corresponding finite complements introduced by che (compare 25c vs. 25d). A further difference is that raising is obligatory in Italian (and English) non-finite structures, as shown by the unacceptability of (25e), presumably for Case reasons. Notice that Italian di only appears with the Control use (of sembrare (25f)).32

(25)

a.

ma (loro) sembrano

[ usare

la violenza

per altre cose ].

but they seem-3.PL

use-INF

DET violence

for other things

‘…but they seem to use violence for other things.’

b.

*ma (loro) sembrano

[di

usare

la violenza

per altre cose ].

but they seem-3.PL

di

use-INF

DET violence

for other things

(as a)

c.

ma sembra

[che

(loro)

usino

la violenza

per altre cose].

but seem-3.SG

that

they

use.SUBJNC.3.PL

DET violence

for other things

‘…but it seems that they use violence for other things.’

d.

*ma (loro)

sembrano

[ che

la violenza

per altre cose].

but

seem-3.PL

that

use.SUBJNC.3.PL

DET violence

for other things

‘…but they seem that (they) use violence for other things.’

e.

*ma sembra [ (*loro)

di (*loro)

usare

la violenza

per altre cose ].

but seem-3.sg them

di them

use-INF

DET violence

for other things

‘[intended]…but they seem to use violence for other things.’

f.

Mi

sembra

di impazzire.

me.DAT.

seem-3SG

di go.crazy

‘It seems like I am going crazy.’

By contrast, as we have already seen, conjunction là is compatible with either finite or non-finite complements. This implies that, unlike di, which lexicalizes Fin (in Rizzi & Bocci’s cartography, là occupies an intermediate position between Force and Fin. Once again, Int would seem to be the most promising candidate:

(26)

An additional reason not to identify là with di/Fin is that—as we shall see in the next section—là invariably precedes topicalized constituents in Vietnamese, unlike Italian di: as Rizzi & Bocci point out, di follows the topic constituent; this is illustrated in (27) (Rizzi & Bocci’s examples (1) and (2)).

(27)	a.	Ho deciso che, la macchina, la comprerò quest’anno.
		‘I decided that, the car, I will buy it this year.’
	b.	Ho deciso, la macchina, di comprarla quest’anno.
		‘I decided, the car, of to buy it this year.’

Since the order in (27b) FORCE-TOP-là in Vietnamese is unacceptable, and since—as will be shown—the Vietnamese left periphery only seems to admit the lower Top position, it follows that là must occupy a relatively high position, higher than Fin. Once again, then, Int presents itself.

However, it should be clear that this yields a cartographic clash: the tree in (14) that derives {rằng…liệu} sequences is the same as the one that describes {rằng…là} order in (26). This leads to the prediction that liệu and là should be in complementary distribution, whenever they both function as conjunctions (i.e., following rằng). Unfortunately, this prediction is immediately disconfirmed by the examples in (28) below: these demonstrate that the sequence { rằng…liệu..là} is perfectly acceptable (in any context where {rằng…liệu} is also good).

(28)

a.

John

hỏi

rằng

liệu

(là)

[tôi

có

muốn

hẹn hò

với

anh ấy

không]

John

ask

COMP

INT

??

I

Q

want

go.out

with

PRN.DEM

NEGQ

‘John asked if I wanted to go on a date with him.’

b.

John

đã

hỏi

mẹ

rằng

liệu

(là)

mẹ

có thể

John

PAST

ask

mother

COMP

INT

??

mother

Q can

đón

cậu bé

không]

pick up

child

NEGQ

‘John asked his mother if she could pick him up.’33

Dismissing the ad hoc solution of an iterative Int node, which would in any case fail to explain the impossibility of *{là..liệu} order, some reanalysis is necessary. The simplest solution would be to treat liệu as a specifier element in Int, with là as its head, as in (29), so preserving the Rizzi/Bocci hierarchy of C-related heads.34

(29)

Naturally, other analyses are possible, given the unspecified semantics of the criterial head Int, and the lexically underspecified nature of Vietnamese là. Considering topicalization more closely allows greater precision, though no certainty. The proliferation of Top positions in Rizzi & Bocci’s structure is unhelpful in this endeavour, as is the surprising claim that in certain dialects Italian che may lexicalise any one of several functional heads:

‘In fact,…[…]…elements like che are typically versatile, and can occur in different positions in the clausal spine: in the dialects under consideration, they occur in the highest position as declarative force markers, also in a lower position, lower than the wh-element in indirect questions…

(Rizzi & Bocci, 2017, fn 32).

At all events, the data allow us to distinguish “high” from “mid-high” in the left periphery of Vietnamese clauses. Having now devoted more space than the average reader may regard as tolerable to the highest levels of the left periphery, we can inch our way further down to consider the “mid-high” region of CP.

2.1.2. Conjunctions II: {thì, (mà), là}

The “topic” marker thì, along with the distinguished phrase that typically precedes it, has been an object of theoretical interest for several decades—likely much longer within traditional scholarship. Notable studies in English include Thompson (1965/1987), Cao (1992), Clark (1992), and Tran (2016, 2024). Nguyen (2021) provides an excellent summary; she also contributes some useful new data, particularly in respect of clause-initial thì.35 The main reason that thì has attracted such interest is its indeterminate nature—construed more positively, its multifunctionality, as the examples below indicate, thì serves many different functions, including as an “Aboutness Topic” marker (30), a Contrastive Topic marker (31), and a clausal conjunction (32); in addition, as both Tran (2016) and Nguyen (2021) point out, thì frequently appears utterance-initially as a discourse linker (33)—for Nguyen, this is its primary function.

(30)

a.

Cô ấy

thì

gặp

anh ấy.

[subject topic]

PRN.DEM

TOP

meet

PRN.DEM

‘Speaking of her, she meets him.’

b.

Anh ấy

thì

cô ấy

gặp.

[object topic]

PRN.DEM

TOP

PRN.DEM

meet

‘Him, she will meet.’

c.

Hồi

trước

thì

cô   ấy

gặp

anh   ấy.

[adjunct topic]

time

before

TOP

PRN.DEM

meet

PRN.DEM

‘In the old days, she would meet him.’

(31)

a.

Ở

miền Bắc

có

bốn

mùa

nhưng …

[contrast topic]

LOC

North

EXIST

four

season

but…

…ở

miền Nam

thì

chỉ

có

hai mùa.

LOC

South

TOP

only

have

two season

‘In the North, there are four seasons, but in the South, only two.’

b.

Mày

thì

gầy,

nó

thì

béo.

(Clark, 1992: [5])

you

TOP

slim,

PRN

TOP

fat

‘You are slim, and he is fat.’

(32)

a.

[Anh

mà

đến ]

thì

[chị  ấy

rất

vui lòng.] [clausal linker]

PRN

CONJ

come

CONJ

PRN.DEM

very

happy

‘She’ll be very happy, (if) you come.’

b.

[Anh

không

thấy

ai ]

thì

[không

cần

ở lại

một mình.]

PRN

NEG

see

wh

CONJ

NEG

need

stay

oneself

‘(If) you don’t see anyone, you don’t have to stay there by yourself.’

c.

(Nếu)

trời

sập

thì

cô   ấy

gặp

anh ấy.

if

sky

fall

CONJ

PRN.DEM

meet

PRN.DEM

‘If the sky falls, she will meet him.’

(33)	A:	[Gia đình ấy			có	nhà		to]	[discourse linker]
		family DEM2			have	house		big
		‘That family has a big house.’
	B:	Thì	[ cô	ghen tị			à?].
		PTL	PRN	jealous			PTL
		‘Are you jealous of them?’

For our purposes, the most pressing question about thì, regardless of label, is its distribution and interaction with respect to other left-peripheral elements, particularly in embedded contexts.

The first point to establish here is whether Vietnamese allows embedded topicalization (ET) in the first place; after that, we can determine to which topicalization class it belongs, within the typology of Miyagawa (2017), to be discussed directly.

It has long been known, since Hooper and Thompson (1973)’s response to Emonds (1969)’s original proposal, that English ET is restricted by the semantic class of the matrix predicate. In Hooper & Thompson’s taxonomy, reproduced in (34), ET is excluded from C & D class predicates. This illustrated by the contrast between the examples in (35) and those in (36):

(34)

(35)	a.	I exclaimed that this book, I will never read. (Class A)
	b.	I think that this book, he read thoroughly. (Class B)
	c.	I found out that this book, no one is willing to read for the test (Class E)

(36)	a.	*It’s likely that this book, everyone will read for the test. (Class C)
	b.	*He was surprised that this book, I had not read. (Class D)

In the same paper, Hooper and Thompson (1973, p. 485) observe that ET is disallowed in non-finite clauses, irrespective of semantic class; compare the examples in (37). They propose that this is because infinitival complements are reduced clauses, lacking a Topic position, something that will become crucial shortly.

(37)	a.	My friends said [TOP the more liberal candidates], they had always supported.]
	b.	My friends tend to support [the more liberal candidates]. [Miyagawa, 2017: [24a]]
	c.	*My friends tend [TOP the more liberal candidates] to support. [Miyagawa, 2017: [24b]]

From a cartographic standpoint (Rizzi, 1997, 2001; Haegeman, 2006, 2010), ‘reduction’ means truncation: by hypothesis, non-finite clauses are truncated structures, where the upper regions of the CP domain are deleted—or ”spliced” in the terminology of Shlonsky and Soare (2011):

(38)	ForceP  >  IntP  >  TopP … > Fin(ite)P

Miyagawa (2017) relates these two restrictions on ET, by proposing the following universal structural constraint:

(39)	Topic Projection: The topic projection TopP is allowed for the complement of A, B, and E, but not for complement of classes C and D.

The constraint in (39) means that all unacceptable instances of ET can be ruled out cartographically: either the TopP is deleted, in the case of infinitivals, or it is unavailable in the first place (a function of semantically driven c-selection).36

However, the significance of Miyagawa’s work lies less in his formalization of ET than in his proposal that this universal constraint on ET is parameterized in ways that are immediately relevant to our cartographic project.37 Developing his (2010) work, Miyagawa (2017) distinguishes two language types (cf. Li and Thompson’s (1976): “agreement-based” languages such as English, and “discourse-configurational” languages, such as Japanese. Miyagawa shows that Japanese-type languages deviate from English-type languages by exhibiting a clearer morphosyntactic and syntactic division between different kinds of topic (Aboutness, Contrastive, and Familiar Topics).

Regarding the universality of Hooper & Thompson’s taxonomy, Miyagawa points out that, with respect to ‘Aboutness Topics’, the English pattern {{A, B, E}; {*C,*D}} is exactly replicated in Japanese: he also shows that the permitted vs. disallowed complements are further distinguished by combining with different (head-final) complementizers (to vs. koto), indicative of different functional projections.38

Compare first the examples in (35) and (36) above with those in (40) and (41), from Miyagawa (2017, pp. 12–13). Notice that these examples involve simple “Aboutness Topics”:

(40)

a.

Class A:

Hanako

wa

[ sono

hon

wa

kodomo ga

yonda

to ]

itta

Hanako

TOP

that

book

TOP

child NOM

read

C

said

‘Hanako said that, as for that book, her child read it.’

b.

Class B:

Hanako

wa

[ sono

hon

wa

kodomo ga

yonda

to ]

sinziteiru.

Hanako

TOP

that

book

TOP

child NOM

read

C

believe

‘Hanako believes that, as for that book, her child read (it).’

c.

Class E:

Hanako

wa

[ Taroo

wa

kanozyo

ga

suki da

to ]

kizuita.

Hanako

TOP

Taroo

TOP

she

NOM

like COP

C

realize

‘Hanako realized that. as for Taroo, he likes her.’

(41)

a.

Class C:

*Hanako

wa

[ sono hon wa

kodomo ga

yonda

koto ]

o

hiteishita.

Hanako

TOP

that book TOP

child NOM

read

C

ACC

denied

‘Hanako denied that, as for that book, her child read (it).’

b.

Class D:

*Hanako

wa

[ sono hon wa

zibun ga

yonda

koto ]

o

kookai shita.

Hanako

TOP

that book TOP

self NOM

read

C

ACC

regretted

‘Hanako regretted that, as for that book, she herself read (it).’39

Having established that Japanese patterns with English in the default case (‘Aboutness Topics’), Miyagawa then draws on work by Bianchi and Frascarelli (2010), demonstrating how the two languages diverge with respect to Contrastive Topics (Contrastive Focus) (see Kuno, 1976; Büring, 2003):40 this is where the main empirical contribution lies. Miyagawa claims that Japanese diverges from English in allowing CTs in all environments—including Class {C, D} contexts, where ATs are blocked. Now compare the sentences in (41) with those in (42) below:41

(42)

a.

Hanako

wa

[ sono hon WA

kodomo ga

yonda

koto ]

o

hiteishita

Hanako

TOP

that book CONTR.TOP

child NOM

read

C

ACC

denied

‘Hanako denied that as for that book, her child read (it).’  (but not this book)

b.

Hanako

wa

[ sono hon WA

zibun ga

yonda

koto ]

o

kookai shita.

Hanako

TOP

that book CONTR.TOP

self NOM

read

C

ACC

regretted

‘Hanako regretted that as for that book, she read (it).’ (but not this one)

The implication is clear: if the unacceptability of the examples in (41) is due to the unavailability of TopP, then the Contrastive Topics in (42) must occupy a distinct pre-subject position. Rizzi & Bocci’s cartography offers a direct candidate, namely, Foc.

…

(1)	[ Force [ Top* [ Int  [ Top* [ Foc [ Top* [ Mod [Top* [Qemb [Fin [IP … ] ] ] ] ] ] ] ] ] ] ]
	(From Rizzi & Bocci, 2017)

The preceding discussion immediately raises the question of which class Vietnamese belongs to. Intuitively, since Vietnamese is normally classified as a “topic-prominent” language,42 we would expect it to pattern with Japanese. Nevertheless, as was illustrated by the examples (30)–(33) above, thì is multifunctional in a way that makes it harder to tie it to one functional position.

The examples in (43)–(45) below make it clear that Vietnamese closely tracks the Japanese patterns: whereas no obvious difference is found between AT and CT topics with {A, B, E} predicates in (43), a contrast is found in the *{C,D} sets (44)–(45)—the same contrast as found in Japanese. There is some variation in judgments according to the predicate, but where a contrast is observed, as in (44b) for example, it always favours the CT alternant, just as Miyagawa’s analysis predicts.

(43)

a.

Class A:

AT: Mary

nói

rằng

những cuốn sách đó

thì

cô ấy sẽ đọc trong hôm nay.

CT: Mary

nói

rằng

những cuốn sách đó

thì

cô ấy sẽ đọc,

không phải những cuốn này. (nói = ‘say’)

b.

Class B:

AT: Mary tin rằng

những cuốn sách đó

thì cô ấy có thể đọc

trong hôm

nay.

CT: Mary tin rằng

những cuốn sách đó

thì cô ấy có thể đọc,

không

phải

những cuốn này. (tin = ‘believe’)

c.

Class E:

AT: Mary nhận ra rằng những cuốn sách đó thì cô ấy có thể

đọc trong hôm nay.

CT: Mary nhận ra rằng những cuốn sách đó thì cô ấy có thể đọc,

không phải những cuốn này. (nhận ra = ‘realize’)

(44)	a.	Class C (1)
		AT:	?Mary phủ nhận rằng những cuốn sách đó thì cô sẽ đọc trong hôm nay.
			Mary deny COMP PL CLS book DEM TOP PRI FUT read in today
			*‘Mary denied that those books, she could read today.’
		CT:	?Mary phủ nhận rằng những cuốn sách đó thì cô sẽ đọc,
			Mary deny COMP PL CLS book DEM2 TOP PRN FUT read,
				không	phải	những cuốn này.
				not	correct	PL CLS DEM1
			*‘Mary denied that those books, she could read today, but not these.’
	b.	Class C(2)
		AT:	*Không thể có chuyện những cuốn sách đó thì John sẽ đọc trước cuối tuần.
			NEG fact ASR story PL CLS book DEM2 TOP John FUT read before end week.
			‘*It’s impossible that those books, John will read by the end of the week.’
		CT:	Không thể có chuyện những cuốn sách đó thì John đọc,
			NEG fact ASR story PL CLS book DEM2 TOP John FUT read,
				không	phải	những cuốn này.
				NOT	correct	PL CLS DEM1
			‘It’s impossible that those books, John read, but not these.’

(45)

a.

Class D (1)

AT:

*Mary bực tức

rằng

những

cuốn sách đó

thì

John đọc

Mary resent

COMP.

PL

CLF book DEM2

TOP

John read

trong

kỳ nghỉ.

in

vacation

*‘Mary resents that those books, John read while on vacation.’

CT:

Mary bực tức

rằng

những

cuốn sách đó

thì

John đọc,

Mary resent

COMP

PL

CLF book DEM2

TOP

John read,

không

phải

những cuốn này.

NEG

correct

PL CLS DEM1

*‘Mary resents that those books, John read, but not these.’

b.

Class D (2)

AT:

?Tôi

tiếc

rằng

những

cuốn sách đó

thì

John đọc

I

regret

COMP

PL

CLS book DEM2

TOP

John read,

mà

không

tham.khảo

ý.kiến tôi.

CONJ

NEG

consult

opinion I

?*‘I regret that those books, John read without consulting me.’

CT:

Tôi

tiếc

rằng

những

cuốn sách đó

thì

John đọc,

I

regret

COMP

PL

CLF book DEM2

TOP

John read,

không

phải

những cuốn này.

NEG

correct

PL CLS DEM1

?*‘I regret that those books, John read, but not these.’

A crucial point to bear in mind is that these are judgments of only two native speakers: follow-up investigation is clearly necessary.43 Nevertheless, assuming the patterns to be valid, the Rizzi and Bocci (2017) hierarchy offers a direct cartographic mapping: Contrastive Topics (CT) can be mapped to Foc across all predicate classes, whilst—in those constructions where TopP is allowed—AT constituents could in principle map to either the higher or lower TopP in (46), building on (21) above:

(46)

This diagram not only fits the data well, it also allows us to go further, to examine whether Vietnamese makes use of the higher or lower position for Aboutness Topics.44 For, unlike Italian, which allows for double topicalization in embedded clauses—compare (47) below—ET_AT in Vietnamese is restricted to one position, that is, only one single topicalization is allowed. The question is which: AT₁ or AT₂, in (46)?

(47)	a.	Mi domando, a mio figlio, se, la macchina, gliela compreremo quest’anno.
		‘I wonder, to my son, if, the car, we will buy it to him this year’
		(Rizzi & Bocci, 2017: [9])
	b.

The chief diagnostic here is the presumed Int head là: if là can follow thì, this would provide evidence of the availability of the higher TopP position (AT₁); conversely, if là is unable to appear before the AT in ET contexts, we could exclude the lower position (AT₂) as a possible Topicalization site. Finally, if rằng and là can co-occur preceding the AT-thì, and the other orders were excluded, this removes the possibility that Vietnamese disposes of a higher topic position.

The deciding data are given in (48) below, with the unacceptability of (48c) weighing strongly in favour of the third option:

(48)	a.	*Mary nói	[AT1 những cuốn sách đó ]	thì	là	cô ấy	sẽ	đọc	trong
		Mary say	PL CLF book DEM2	TOP	INT	PRN	FUT	read	in
		hôm nay.
		today
		‘Mary says that those books, she’ll read by the end of the day.’
	b.	*Mary nói	rằng	[AT1 những cuốn sách đó ]	thì	là	cô ấy	sẽ	đọc…
		Mary say	C	PL CLF book DEM2	TOP	INT	PRN	FUT	read…
		(as a)
	c.	Mary nói	rằng	là	[AT2 những cuốn sách đó ]	thì	cô ấy	sẽ	đọc…
		Mary say	C	INT	PL CLF book DEM2	TOP	PRN	FUT	read…
		(as a)

These examples clearly imply that the higher AT position (AT1) is unavailable in Vietnamese, and that the tree in (46) should be pruned to that in (49). If so, the AT/CT contrast in (43–45) then would vary only according to the availability of the lower (non-recursive) TopP:45^,46

(49)	AT in lower TopP only (cf. 40a):

To be sure, one can find examples in which là immediately follows thì, such as those in (50), but here thì is more plausibly analysed as an utterance-initial conjunction—indeed intonation suggests that thì.là may be a compound conjunction in such cases:47

(50)

a.

“không

phải

anh

thì.là

anh

khác…”48

NEG

right

PRN

CONJ-CONJ

you

different…

‘If not you, then someone else…’

b.

“ừ,

thì.là

em

có nỗi

buồn

thật đẹp…” [song lyric]49

well,

CONJ-CONJ

PRN

have

sadness

beautiful

‘Well, then I (?you) have a beautiful sadness.’

Before moving on to investigate positions below TopP, there is one further complication to examine, namely, a context in which a topicalized phrase is followed by non-copular là. Consider the minimal contrast in (51): if the sentential subject in (51a) counts as an embedded topic (AT), then so too should the apparently identical clause in (51b), where là replaces thì, yielding a shift in interpretation (irrealis to realis/finite):

(51)

a.

[Nó

không

học

toán]

thì

tốt.

PRN

NEG

study

maths

??

nice

‘It would be nice for him not to study mathematics.’

b.

[Nó

không

học

toán]

là

tốt.

PRN

NEG

study

maths

??

nice

‘It’s nice that he didn’t study mathematics.’

The reason that this presents a problem is that là cannot be a copula here since—as noted previously—copular là is prevented from appearing with adjectival predicates.50 But if là is a coordinating conjunction, then it should not be found following any topicalized constituent, given (49).

Fortunately for our ongoing analysis, the superficial minimal contrast in (51) belies a significant structural difference, namely, that (51a-thì) is non-finite, whereas (51b-là) is finite (its own clause). This is demonstrated by the fact that the tense morpheme đã is excluded from (52a) below, while being fully acceptable in (52b). In other words, the string [Nó không học toán] has a different analysis in the two cases: in (51a), it really is a topicalized subject argument (≈ English small clause, or gerund); in (51b), by contrast, the same string is analysed as a separate clause:51

(52)

a.

[Nó

(*đã)

không

học

toán]

thì

tốt.

PRN

PAST

not

study

maths

TOP

nice

‘It’s (a) nice (thing) that he didn’t study mathematics.’

b.

[Nó

(đã)

không

học

toán]

là

tốt.

PRN

PAST

not

study

maths

CONJ

nice

‘It’s (a) nice (thing) that he didn’t study mathematics.’

This not only explains how conjunction là can follow [Nó không học toán] in (51a): it also accounts for the fact that neither rằng (nor là) can introduce sentential subjects in Vietnamese; compare the examples in (53) below (examples from Duffield, 2013; Phan & Duffield, 2022). Both restrictions follow if sentential topics cannot be full clauses, unlike in English.

I postpone to future work discussion of the clause following là.

(53)	a.	(*rằng) họ		cười khúc khích]			làm		chúng em	thẹn.
		COMP PRN		laugh giggle			make		PL PRN	embarrassed
		‘[*(that) they giggled] embarrassed us.’
	b.	(*là)	cô ấy		rời đi sớm	làm		tôi		ngạc nhiên.
		CONJ	PRN.DEM		leave early	make		I		surprise
		‘[*(that) she left early] surprised me.’

Having thus established the upper-to-middle sections of the Vietnamese CP, we may turn attention to the remaining region of the left periphery—from the AT topic head to the subject position, schematized in (54) below—to consider three kinds of pre-subject phrase: speaker-oriented adverbials; weak indefinites interpreted as universal quantifiers; and the adjunct wh-expression tại sao (‘why’).

(54)

2.2. Left Peripheral Adverbials

Even before we come to weak (universal) indefinites and why questions, sentences involving subject-oriented adverbials (specifically, the adverbial phrases quả thật ‘indeed’, and nhất định ‘absolutely’) provide evidence that the (AT) topic position is located significantly higher in the left periphery than the canonical subject position.

Consider the paradigm in (55) and (56), adapted from Phan and Duffield (2022). At first glance, the alternation in (55) appears to indicate that quả thật may either follow or precede the subject argument. However, given the optional pronunciation of the topic marker thì, it is necessary to ensure that there is true variability here: that (55a) is not a case of Top-ADV-(subject) order, such that (55b) represents the base order. The examples in (56) show clearly that this is not the case: first, the unacceptability of (56b) establishes that adverbials cannot appear to the left of an explicit topic; more crucially, the alternations in (57), which contain both an explicit topic [những cuốn sách đó] and an explicit subject [cô ấy]—compare (44)–(45) above—allow the adverbial to precede or follow the subject; indeed, the latter order seems to be preferred (57b):

(55)

a.

Cô ấy

quả thật

đã

nói

với

anh ấy.

PRN.DEM

indeed

ANT

talk

with

PRN.DEM

‘She indeed talked to him.’

b.

?Quả thật

cô ấy

đã

nói

với

anh ấy.

indeed

PRN.DEM

ANT

talk

with

PRN.DEM

‘Indeed, she talked to him.’

(56)

a.

Cô ấy

thì

quả thật (pro)

đã

nói

với

anh ấy.]

PRN.DEM

TOP

indeed

ANT

talk

with

PRN.DEM

‘Speaking of her, she indeed talked to him.’

b.

*Quả thật

[ cô.ấy

thì

[ (pro)

đã

nói

với

anh  ấy ] ]

indeed

PRN

TOP

ANT

talk

with

PRN.DEM

intended (a)

(57)

a.

?[Những

cuốn

sách

đó]

thì

nhất định

cô ấy

phẚi

đọc.

PL

CLF

book

DEM

TOP

absolutely

PRN.DEM

must

read

‘(?)That book absolutely she had to read.’

b.

[Những

cuốn

sách

đó]

thì

cô ấy

nhất định

phẚi đọc.

PL

CLF

book

DEM2

TOP

PRN.DEM

absolutely

must read

‘That book, she absolutely had to read.’

From these data, it is plausible to conclude that adverbials are moved from a post-subject, pre-verbal slot to a pre-subject position, to the right of the Topic, as diagrammed in (58), consistent with the Rizzi/Bocci hierarchy:52

(58)

Here, I am co-opting Rizzi & Bocci’s Mod position for the moved adverbial phrases. This is in the letter of Rizzi & Bocci’s cartography, but perhaps not in the spirit, since in their discussion Rizzi & Bocci exemplify Mod with fronted adverbials such as rapidamente (‘rapidly’) (59a), elements that can also appear post-verbally, as in (59b). By contrast, quả thật is only ever found pre-verbally: this is shown by the unacceptability of the examples in (59c,d) below (see Phan & Duffield, 2022):

(59)

a.

Rapidamente, Gianni trovò la soluzione.

‘Rapidly, Gianni found the solution.’

b.

Giahái ha trovato rapidamente la soluzione.  (Rizzi & Bocci, 2017: 15a)

‘Gianni found rapidly the solution.’

c.

*Cô ấy

đã

nói

quả thật

với

anh ấy.

PRN.DEM

ANT

talk

indeed

with

PRN.DEM

‘She indeed talked to him.’

d.

*Cô ấy

đã

nói

với

anh ấy

quả thật.

PRN.DEM

ANT

talk

with

PRN.DEM

indeed

(as 58c).

Still, whatever the correct label may be, the distribution of LP-adverbials such as quả thật provides a useful diagnostic of the separation of topic and subject positions in Vietnamese.

2.3. “Quantifier-Raising”

Let us now consider fronted weak indefinites interpreted as universal quantifiers (“QP_∀”).53 As discussed in Duffield (2007, 2014), Vietnamese disposes of a subset of weak indefinite expressions that must appear in front of the pre-verbal ‘adverbial operator’ cũng, even where these phrases are interpreted as thematic objects, or as adjuncts that would normally appear post-verbally. In the examples in (60), the displaced indefinite appears to the right of the subject, yielding S-O_QP-cũng-V order in (60a,b)), S-Adjunct_QP-cũng-V, in (60c,d):

(60)

a.

Anh ấy

từ

nào

*(cũng)

nhớ.

[√S-OQP-cung-V order]

PRN.DEM

word

WH

OP

recall

‘He remembers every word.’

b.

Cô   ấy

ai

[*(cũng)

quen].

PRN.DEM

WH

OP

know

‘She knows everybody.’

c.

Anh ấy

bao giờ

[*(cũng)

đến

muộn].[√S-ADJNQP-cung-V order]

PRN.DEM

WH time

OP

come

late

‘He is always late.’

d.

Tôi

ngày nào

[*(cũng)

tập thể thao]

I

day WH

OP

do exercise

‘I do exercise every day.’

However, QP_∀-S-V order is also found, as shown by the corresponding examples in (61): both thematic objects and QP_∀-adjuncts may also precede the subject:

(61)

a.

Từ

nào

[anh.ấy

[cũng

nhớ].]

[√OQP-S-cung-V order]

word

WH

PRN.DEM

OP

recall

‘He remembers every word.’

b.

Ai

[cô ấy

[cũng

quen]].

WH

PRN

OP

know

‘She knows everybody.’

c.

Bao

giờ

[anh  ấy

[cũng

đến

muộn ]].

[√ADJQP-S-cung-V order]]

WH

time

PRN.DEM

OP

come

late

‘He is always late.’

d.

Ngày

nào

[tôi

[cũng

tập thể thao ]].

day

WH

I

OP

do exercises

‘I do exercises every day.’

A point to observe is that this QP_∀-fronting is obligatory in the case of underspecified weak indefinites (ai (‘who/anyone’), nào (‘which/anything’, etc.,): expressions that would otherwise be interpreted as wh-questions if left in situ.54 This state-of-affairs contrasts with regular, fully specified, universal quantifiers (VN. mọi~mỗi) such as those in (62), where overt QP_∀ fronting is impossible, and where cũng, if present, only has its non-operator interpretation (=‘also’):

(62)

a.

*Anh  ấy

[cũng

nhớ

từ

nào].

[*S-cung-V OQP]

PRN DEM

OP

recall

word

WH

‘He remembers every word.’

(ok: ‘Which word does he remember?’)

b.

Anh ấy

[(cũng)

nhớ

mọi

từ].

[√S-cung-V-OQP (mọi)]

PRN DEM

also

recall

each

word

‘He (also) remembers every word.’

c.

*Anh ấy

mọi

từ

[(cũng)

nhớ].

[*S-OQP-V: mọi]

PRN DEM

each

word

also

remember

*‘He every word remembers.’

d.

Tôi

[tập

thể thao

mỗi ngày].

[√S-V-XQP: mỗi]

I

do

exercise

every day

‘I do my exercises every day.’

e.

??Tôi

mỗi

ngày

[tập

thể thao].

[??S-XQP-V: mỗi]

I

every

day

do

exercise

??‘I every day do my exercises.’

In Duffield (2007), these patterns are accounted for in terms of ‘scope evasion’: QP_∀-movement is driven by the functional need to avoid being interpreted as wh-expressions if they remain in-situ (alternatively, as negative polarity items (NPI), when in the scope of the negative operator không). In that paper, little attention was paid to the landing-sites involved, the issue of position being less important than that of motivation. Here, the converse is true: to establish the fine structure requires that we determine the precise position of the moved weak indefinite, in both paradigms ((60) and (61)). It is also necessary to determine the position of the subject argument in the two contexts: the optionality of the topic marker thì means that nothing can be taken for granted with respect to starting position.55

The most obvious move is to suppose that QP_∀-raising involves raising to Foc, in the truncated (single topic position) version of Rizzi & Bocci’s cartography, schematized in (63):

(63)	∀

Following the logic of the previous discussion of adverb placement, such an analysis could allow us to derive both SOV(60) and OSV(61)) order, provided that: (i), the S in S-O_QP∀-V is topicalized; equally, (ii), the S in O_QP∀-S-V must not be (topicalized).

These expectations are only partially borne out, as evidenced by the examples in (64). On one hand, it is clear—as it was with adverbial expressions—that QP_∀ cannot raise past the Topic position. In (64a), where the (Aboutness) topic is a non-argument (câu chuyện này), which effectively precludes the subject topicalization, the raised QP intervenes between the Topic and Subject position, as predicted by (63); the unacceptability of (64b) also implies that the weak indefinite cannot move beyond the topic position, again as predicted.56 What is more, the acceptability of TOP_SUBJ-O_QP∀-V in (64c) offers further support for QP_∀-movement to an intermediate position between TOP and Subject:

(64)

a.

Câu chuyện

này

thì

từ

nào

anh ấy

cũng

nhớ.

CLS story

DEM1

TOP

word

WH

PRN.DEM2

ALSO

remember

‘As for this story, he remembers every word (of it).’

b.

*Từ

nào

[câu chuyện

này

thì

anh ấy

cũng

nhớ.]

word

WH

CLS story

DEM1

TOP

PRN.DEM2

ALSO

remember

‘As for this story, he remembers every word (of it).’

c.

“Câu chuyện này?

Anh ấy

thì

từ

nào

cũng

nhớ.”

CLS story DEM1

PRN.DEM2

TOP

word

WH

ALSO

remember

‘This story? He’s the one that remembers every word (of it).’

d.

Câu chuyện

này

thì

anh ấy

từ

nào

cũng

nhớ.

CLS story

DEM1

TOP

PRN.DEM2

word

WH

ALSO

recall

‘As for this story, he remembers every word (of it).’

All would be well for the analysis in (63), but for the acceptability of (64d): here, the topic position is filled, as in (64a), meaning that the subject cannot be topicalized, yet the QP_∀ can still appear to its right {…TOP…SUBJ…QP_∀…}. The acceptability of this word order is obviously inconsistent with the idea that QP_∀-movement in Vietnamese involves raising to a dedicated Foc position, unless certain assumptions are revised.57

A more serious problem for the dedicated position analysis in (63) arises when we consider the interaction between fronted QPs and evaluative adverbs. The analysis of fronted adverbs in (58) predicts the acceptability of {TOP-QP-(ADVP)-Subj} order: the same analysis, prevents adverbials from appearing between the Top and the QP_∀ in Foc. Unfortunately, the data show the reverse: the predicted order, (65a) is marginal at best, while the excluded order (65b) is fully acceptable.

(65)

a.

??[AT

Câu chuyện này]

thì

từ

nào

quả thật

anh ấy

cũng

nhớ.

CLF story DEM1

TOP

word

WH

indeed

PRN.DEM

ALSO

recall

‘As for this story, he remembers every word (of it).’

b.

[AT

Câu chuyện này]

thì

quả thật

từ

nào

anh ấy

cũng

nhớ.

CLF story DEM1

TOP

indeed

word

WH

PRN.DEM2

ALSO

recall

‘As for this story, he remembers every word (of it).’

In short, the analysis in (63) faces challenges from two directions: on one hand, the acceptability of (64d) shows that the raised QP_∀ can appear to the right of the subject, well below the hypothesized Foc node; conversely, the acceptability contrast in (65a) reveals that fronted QPs preferentially follow the position of fronted adverbials, to the left of the subject, but well below the hypothesized Foc landing-site. Whilst all these alternative orders could be derived by ad hoc stipulation, the more principled solution would be to revert to the original (2007) analysis, which freely adjoined QP to any projection within IP outside the scope (c-command) of cũng. This would correctly derive the variable word-orders, but would necessitate giving up on Foc as a unique landing-site for QP movement.

This leaves us with just one pre-subject constituent to place, and which is the raison d’être of this whole endeavour: tại sao (‘why’), a wh–expression that diverges from other wh-phrases across a wide range of languages—also, within constructions in those varieties. As we shall see, sentences involving tại sao constitute a Pandora’s box of interesting data, potentially jeopardizing the results obtained thus far, as well as casting doubt on the universality of the Rizzi/Bocci hierarchy.

2.4. Where(fore) Why?

2.4.1. English Why

Let us begin with English. The observation that why is anomalous vis-à-vis other English wh-expressions had been noted previously, especially in the child language acquisition literature: see de Villiers (1991) et seq.; cf. Rizzi (1990). However, until the advent of cartography and an articulated CP, it was not considered theoretically interesting, for want of mechanisms available to capture the distributional differences.

The following minimal contrasts describe the asymmetric lay of the land with respect to why (vs. other wh-phrases). First, the examples in (66) show that in a suitable discourse context why or why (not)—uniquely among wh-phrases—can merge with bare predicate phrases (VPs, PPs, DPs, ADVPs), though not with bare vPs (thematic phrases containing a subject (67).58 The examples in (68) show that other wh-expressions are either excluded from the same environments, or else highly restricted (unless the sentence is parsed as a sluicing context):

(66)	a.	Why (not) [risk it]?
	b.	Why (not) [talk to a therapist]?
	c.	Why [ fix the car]?
	d.	Why [ him]? Why [ not her]?
	e.	Why [only yesterday]? Why not [ two weeks ago]?
	f.	Why [with his brother]? Why not [with his partner]?59

(67)	a.	*Why (not) he/him go there?
	b.	*Why she/her take the money?
	c.	*Why (not) they/them came?

(68)	a.	*When/*who (not) [risk it]?
	b.	*When (not) talk to a therapist?
	c.	*How fix the car?
	d.	*Where him? Where not her?
	e.	*Who only yesterday? *Who not two weeks ago?
	f.	*Where with his brother? *How not with his partner?

Perhaps unsurprisingly, these bare why-XP fragments may not appear as embedded complements (69a,b), other than in quotative (direct speech) contexts (69c):

(69)	a.	*She wondered [why her]. cf. (“Why her?,” she wondered)
	b.	*She knew [why with his brother].
	c.	She said: “Why not risk it?”

In contrast to the bare predicate examples in (66)–(68), where it is privileged over other wh-expressions, why is uniquely excluded from non-finite complement clauses (indirect questions), such as those in (70): this is despite being just as acceptable as other fronted wh-expressions in the corresponding finite contexts. Compare the examples in (70) with those in (71):

(70)	a.	She knew what/when/how/*why to tell him.
	b.	She wondered who to talk to/how best to talk to him/*why to stay any longer.

(71)	a.	She knew what/when/how/why she should tell him.
	b.	She wondered who to talk to/how best she should talk to him/why she should stay any longer.

This asymmetry obtains even in Sluicing contexts, where the elided portion is no less silent: compare (72a) vs. (72b).

(72)	a.	He was told where [ to meet Jane], and when…*but not why [ to meet Jane ].
	b.	He was told where [ he should meet Jane], and when… but not why [ he should meet Jane ].

These contrasts suggest that semantic interpretability is not at issue here: the constraint seems to be purely grammatical. This intuition is further supported by the contrast between the Sluicing contexts in (72) above, and the Tough-Construction complements in (73): in the latter context, once dissociated from the non-finite complement, [why … [ + non-finite clause]] is just as acceptable as any other embedded wh-expression.60

(73)	a.	She knew [ what/why [ it was best/tough [ to tell him ]]]
	b.	He wondered [ who [ it might be easier [ to live with: Jane or Amy? ]]]
	c.	He wondered [ why [ it might be easier [ to live with Jane. ]]]

Why once again diverges from other wh-adjuncts when it comes to LD wh-movement (that is, at the maximal extension of the clause). For many speakers—see Aoun et al. (1987)—why exhibits ‘that-trace’ effects on a par with wh-subjects, such that the presence of the embedded complementizer in (74a) blocks the ‘downstairs’ construal of why, in what is otherwise an ambiguous string (74b). No similar constraint applies to other wh-adjuncts: if anything, the downstairs reading is favoured elsewhere, regardless of the presence of that. (And, in (74d), where the adjunct is semantically selected by put, the lower reading is the only one available:)

(74)	a.	When/How did Justin say (that) he had finished the painting?
	b.	Why did Justin say he had finished the painting early?
	c.	??Why did Justin say that he had finished the painting early?
	d.	Where did Justin say that he had put the painting?

Moreover, for speakers that share the judgments in (74), *why-to effects can be ameliorated, as shown in (75)—just like that-trace effects in (76)—by intervening phrasal material: see Culicover (1993), Douglas (2017); Sobin (1987, 2002); Salzmann et al. (2013); also Duffield (2018):

(75)	a.	She told you [ why *(not) to write about this problem]
	b.	She told you [ why *(on no account) to write about this problem].

(76)	a.	Lee forgot which dishes Leslie had said that *(under normal circumstances) should be put on the table.
	b.	Which kinds of drugs did you say that *(without proper testing) had been released on the market?

The connection with wh-subjects is probably not coincidental: notice that exactly these arguments pattern with why with respect to non-finite complements; compare (77a) with (70a) above, also with its finite, and non-subject counterparts (77b, c):

(77)	a.	*She wondered [who to tell Mary the news].
	b.	She wondered [who had told Mary the news].
	c.	She wondered [who (PRO) to tell the news to].

The traditional explanation of the contrast in (77) has it that the non-finite clause lacks the relevant features—standardly, Case features—to license an overt subject. The cartographic alternative proposed by Shlonsky and Soare (2011) extends this deficiency account to the why cases in (70), except that what is missing are not merely the features but the structural projections containing them: non-finite clauses are truncated versions of their finite counterparts. This is diagrammed in (78), from Shlonsky and Soare (2011): by hypothesis, the licit non-finite clauses involve splicing/deletion of all upper CP projections:

(78)	Infinitival clauses are spliced at WhP:	(Shlonsky & Soare, 2011: [11])
	ForceP > IntP > TopP > FocP > WhP > FinP

On the explanation given by Shlonsky and Soare (2011), the specific unacceptability of ‘*why-to’ follows from the special licensing conditions on why: whereas regular wh-expressions in English are licensed through movement to WhP—hence, splicing of the higher nodes has no effect on well-formedness—why, again by hypothesis, exceptionally requires movement to IntP for feature licensing. This, the authors claim, explains not only the distributional *why-to constraint, but also the ambiguity of (79b): why may not be able to appear in the lower clause, but it does allow a “downstairs” reading (long construal), for some speakers, at least.61

(79)	a.	*She asked me why to resign.
	b.	Why did you ask her to resign?	(Shlonsky & Soare 2011: [12])
		i. What is the reason x, such that for x, you asked her to resign e.g.,: Because I didn’t want to just tell her. (short construal) ii. What is the reason x, such that you asked her to resign for that reason x? e.g., I asked her to resign because of her health, not because of her intelligence (long construal)

If Shlonsky & Soare’s analysis is correct, then why is not only licensed higher in CP than regular wh-expressions: it is also initially merged higher than these other expressions. Shlonsky & Soare’s analysis of why aligns exactly with Rizzi & Bocci’s treatment of Italian perché. The latter authors write:

‘These considerations led to the postulation of an independent position Int(errogative), hosting se in the head position, also wh-elements like perché (‘why’) and other reason adverbials in the specifier position, as they can also be surrounded by topics and can co-occur with a following focus position (see Rizzi, 2001 and the revision in Shlonsky & Soare, 2011), both in main and embedded questions:

	(11)	Italian
		A Gianni, perché, la macchina, gliela volete regalare?
		‘To Gianni, why, the car, you want to give it to him?

	(12)	Perché LA MACCHINA/∗LA MACCHINA perché gli volete regalare,
				e non la moto?
		‘Why THE CAR/∗THE CAR why you want to give to him,
			and not the motorbike?’

A topic can also occur between Int and Foc:

	(13)	Perché, a Gianni, LA MACCHINA gli volete regalare, e non la moto?
		‘Why, to Gianni THE CAR you want to give to him, and not the motorbike?’

The integration of Int thus gave rise to the following map:

(14)

[Force [Top∗ [Int [Top∗ [Foc [Top∗ [Fin [IP …]]]]]]]] ‘

2.4.2. Vietnamese Why

The previous discussion raises two kinds of questions: where is why/perché, initially? and why is it (there)? Vietnamese, as a wh-in situ language, should at least prove helpful in addressing the first question. Perhaps not surprisingly, the answer is more complicated than one might have wished, having come this far.

Before considering our final data sets, it is useful to summarize our interim progress, climbing down from the top of the left periphery. First, we have distinguished two head positions in the highest regions of CP: Force, typically lexicalized by the declarative complementizer rằng, and Int, lexicalized by the force-neutral linker là, just in case it functions as a conjunction; in interrogative (alternative question) contexts liẹu may appear as a specifier of Int.

Second, we have established that Aboutness Topics invariably appear below Int, apparently ruling out the higher AT position, which is supposed to be available in Italian. We saw that Vietnamese also supports a functional distinction between Aboutness Topics and Contrastive Topics, as proposed by Miyagawa (2009, 2017). However, the evidence does not support any positional contrast (any more than is found in Japanese, in fact): hence, the data implies only one Topic position, which may be filled either by an Aboutness Topic or a Contrastive Topic, depending on context. This speaks against the availability of a distinct Foc node in Vietnamese, contra Rizzi & Bocci.

Finally, the case for a distinct Foc projection is further eroded when we consider how fronted QPs (weak indefinites with a universal interpretation) interact with the three other classes of left-peripheral elements. On one hand, the data clearly show that QPs move to a region below Force and Int and the dual service Top positions, but above the thematic subject position (making certain allowances for example (64d), see note 57). On the other hand, the more flexible interaction of QP with quả thật—specifically, the marginal acceptability of (65a)—casts additional doubt on the idea that QPs are moved to a dedicated position. It seems then, as proposed in Duffield (2007), that these weak indefinites are repelled, not attracted.

The dIagram in (80) summarizes the resulting functional sequence for Vietnamese. (It should be clear that some of these elements cannot grammatically co-occur in any particular sentence: the object is to fix their relative positions in one tree.)

(80)

So now, finally, we come to the ultimate (where?) question, namely, where is why; more specifically, where is Vietnamese why (tại sao)? The first point to establish is that tại sao, like English why and Italian perché, only ever appears to the left of the subject. This is made clear by the paradigm in (81):

(81)

a.

Tại sao

người ta

không

trồng

nấm

trên

đất…?62

why

people

NEG

grow

mushrooms

in

soil

‘Why do people not grow mushrooms in soil…?’

b.

*Người ta

tại sao

không

trồng

nấm

trên

đất…?

people

why

NEG

grow

mushrooms

in

soil

‘Why do people not grow mushrooms in soil…?’

c.

*Người ta

không

trồng

nấm

trên

đất

tại sao?

people

NEG

grow

mushrooms

in

soil

why

‘Why do people not grow mushrooms in soil…?’

Second, we need to show that tại sao-questions can be embedded. The examples in (82) demonstrate that they can be moreover, that tại sao-complements can be introduced by the complementizer rằng:

(82)

a.

Anh

đã

nhiều lần

hỏi em

[rằng

[tại sao … ]]

[song lyric]

PRN

ANT

many times

ask PRN

C

why

‘I have asked you why, so many times.’

b.

Có nhiều

bạn    hỏi

[ rằng

[tại sao

thầy

lại quay

được

các

Video TiếngHàn]]63

ASR many

friends ask

C

why

teacher

upload

CAN

CLF

video lge. Korean

‘I have many friends ask why my teacher can make Korean videos…’

c.

[DP

câu hỏi

[rằng

[tại sao

bỗng nhiên

thế giới

lại

có thang máy]]]64

CLF ask

C

why

suddenly

world

again

have elevator

‘…the question as to why the world suddenly has elevators…’

Taken together, the two example sets (81)/(82) provide clear evidence that tại sao occurs in our crucial region of interest, between Force (rằng), and the subject position. The question is, exactly where? Examining its position relative to the other left-peripheral elements we have encountered to this point, may allow us to pinpoint the location of reason phrases {tại sao, why, perché} prior to movement; we can also investigate whether Vietnamese manifests the *why-to effect modelled by Shlonsky and Soare (2011).

As noted earlier, cartographic treatments of why/perché analyse the peripheral position of this element in Italian and English as the result of high external Merge to a Reason Phrase above the subject position, followed by further leftward movement: both Rizzi and Shlonsky & Soare assume that why/perché is attracted post-Merge to the {spec, Int} for formal criterial (checking) reasons. Recall that this is necessary to explain the unavailability of *why to. However, since WHY is required to move further than Int in both English and Italian, to satisfy selection—(2) diagrammed in (83)—(She doesn’t know (*that) why she came/Non sa (*che) perché è venuta’), the intermediate movement (1) is hard to verify independently. However, as the examples in (82) show, wh-movement never satisfies c-selection in Vietnamese: hence, wherever tại sao is located, it is lower than the Force head. But is it as high as Int? Or does it remain in situ, once merged?

(83)	ForceP … IntP > TopP >  WhP > ReasonP > … FinP  (Shlonsky & Soare, 2011: [35])

The examples in (82) leave open the possibility that overt movement to Int has taken place. It is not entirely clear what we should expect to find here. On one hand, given that Vietnamese is a wh-in situ language, we might not expect to see overt movement of tại sao; on the other, we have seen that universal quantifiers do undergo obligatory A’-movement for interpretive reasons, and so we might suppose that tại sao could exhibit similar behaviour.

Before considering the final data sets, it is useful to step back from the formal explanation of the position of why, to consider the broader why-question, namely, why is why in the left periphery? (More broadly, why is the left periphery peripheral?65

Intuitively, the final answer is an interpretive one: unlike other wh-expressions, which probe gaps in incomplete propositions, why is interpreted as having semantic scope over a complete proposition, including specifications of tense and polarity: it ‘makes sense’ that this semantic scope is reflected in a high syntactic position.66

Moving down the tree in (80), the first point to establish is whether tại sao appears to the left of Int, which would imply overt movement, or to its right. The paradigm in (84) clearly supports the latter option: tại sao cannot appear between rằng and là.

(84)

a.

Cô ấy

không biết

(rằng)

là

[tại sao [

anh ấy

nhớ

mọi

từ ]]

PRN

NEG know

C

INT

why

PRN

recall

every

word

‘She doesn’t know why he remembers every word.’

b.

*Cô ấy

không biết

(rằng)

tại sao

[ là

[ anh ấy

nhớ

mọi

từ ]]

PRN

NEG know

C

why

INT.

PRN

recall

every

word

(as a)

c.

*Cô ấy

không

biết

(rằng)

là

[anh ấy

tại sao

nhớ

mọi

từ]

PRN.DEM

NEG

know

C

INT

PRN.DEM

why

recall

every

word

(as a)67

d.

*Cô ấy

không biết

(rằng)

là

[ anh ấy

nhớ

mọi

từ ]

tại sao.

PRN.DEM

NEG know

C

INT

PRN.DEM

recall

every

word

why

(as a)

It is worth noting immediately that the unacceptability of (84b) is not due to any simple collocational restriction on {tại sao…là} sequences: we will consider grammatical instances of this directly.68

Moving down the tree, the next question concerns the position of tại sao relative to the other left-peripheral elements diagrammed in (80) above: Topic constituents, POV-adverbials and (universal) QPs, respectively. The simplest hypothesis—on either formal or functionalist (semantic) motivation—would place tại sao in a dedicated projection having scope over all these elements, as in (85a), below. A plausible alternative might be that tại sao would sit between the topic position and the other nodes, as in (85b), on the grounds that tại sao would still have scope over all elements interpreted within the proposition; this would be true whether topicalization is derived via movement or co-indexation with a null pronominal. Whichever of these options is allowed, however, we would not expect tại sao to be able to appear to the right of the other two pre-subject constituents (ADVP, QP).

(85)	a.
	b.

We will again consider these projections in turn. The alternations in (86)69 reveal that both affirmative hypotheses are borne out, with respect of topic constituents: tại sao may appear either to the left (86a) or—marginally, to the right (86b)—of the topic:70

(86)	Cô ấy không biết…
	a.	…tại sao	[những	cuốn sách đó]		thì cô ấy có thể đọc,
		why	PL	CLF book DEM2		TOP PRN possible read,
							không	phải	những cuốn này.
							NEG	correct	CLF DEM1
		‘I don’t know why he was able to read those books, but not these ones.’
	b.	…?[những cuốn sách đó]			thì tại sao cô ấy có thể đọc,
		PL CLF book DEM2			TOP why PRN.DEM possible read,
							hông	phải	những cuốn này.
							EG	correct	CLF DEM1
		(as a.)

Since we have established that tại sao does not raise to Int, a formal approach to tại sao placement should keep it in its base position [ReasonP]: in (83), this position is unequivocally below the Topic position, closer to that of the surface Subject. Yet, this formal expectation is contradicted by the alternation in (86)—particularly given the preference for the (a) alternant, where tại sao precedes the Topic. This is unexpected for two reasons: not only is there no plausible probe between Int and Top, even if there were, standard feature-driven movement should then prohibit optionality, ruling out (86b). Consider (87). (Alternatively, if [ReasonP] were projected exceptionally high in Vietnamese, to capture (86a) without recourse to movement, (86b) should still be ruled out, given the ban on lowering).71

(87)

In this case, as was also true of QP placement earlier, the semantic supervenience approach makes better predictions than any appeal to feature-driven movement, or to a fixed functional sequence.

The same is true of adverbials. In a previous section we saw that adverbials such as nhất định can appear either to the right of the subject—their presumed base position—or in a position to the left of the subject, below the Topic phrase: following Rizzi & Bocci’s cartography in (1), we analysed this higher adverbial position as Mod. If we now insert this node into Shlonsky & Soare’s functional sequence in (83), we obtain the sequence in (88):

(88)	ForceP … IntP > TopP > ModP > WhP > ReasonP > … FinP  (Shlonsky & Soare, 2011: [35])

The formal prediction is that if tại sao remains in situ in (88), then it should follow higher adverbs {TOP-ADV-tại sao-SUBJ}, there being no plausible position between TopP and ModP for it to occupy. Once again, the facts are clear: the predicted order (89a) is unacceptable, while the underivable order (89b) is (largely) unexceptional. The most preferred order, (89c) is where the adverb remains in its base position, so this sentence does not speak directly to the question at hand.

(89)

*Mary hỏi rằng…

a.

…*[những

cuốn sách đó]

thì

nhất định

tại sao

[cô ấy

phẚi

đọc]

PL

CLF book DEM2

TOP

absolutely

why

PRN

must

read

‘Mary asked why that book she absolutely had to read.’

b.

…?[những

cuốn sách đó]

thì

tại sao

nhất định

[cô ấy

phẚi

đọc]

PL

CLF book DEM2

TOP

why

absolutely

PRN

must

read

(as a)

c.

…[những

cuốn sách đó]

thì

(tại sao)

[cô ấy nhất định phẚi đọc.]

PL

CLF book DEM2

TOP

why

PRN absolutely must read

(as a)

For completeness, let us consider the position of tại sao relative to moved QP_∀s. Since we have just seen that tại sao invariably appears to the left of the higher adverbs, and since we observed previously in (65) that moved QP_∀ appear to the right of the same adverbs—apart from the highly marginal (65c) above—it should be logically impossible for a moved QP_∀ to appear to the left of tại sao, at least given normal assumptions about how syntax works. Fortunately for syntax, {TOP-QP_∀-tại sao} is just as unacceptable as it should be, as revealed by the examples in (90) below. Here, the paradigm aligns exactly with that in (89): example (90a) shows that {QP_∀-tại sao} order is unacceptable; example (90b) shows the acceptable left-peripheral order {tại sao-QP_∀}, while (90c) shows that QP_∀, like pre-verbal adverbs, can appear within FinP (unlike tại sao, which is restricted to the higher CP-domain).

(90)

a.

*Cô ấy

không biết

[từ

nào

[ tại sao

[anh  ấy

[ cũng

nhớ ]]]

PRN.DEM

NEG know

word

WH

why

PRN.DEM

OP

recall

‘She doesn’t know why he remembers every word.’.

b.

?Cô ấy

không biết

[ tại sao

[ từ

nào

[anh  ấy

[cũng

nhớ ]]]

PRN.DEM

NEG know

why

word

WH

PRN.DEM

OP

recall

‘She doesn’t know why he remembers every word.’.

c.

Cô ấy

không biết

[ tại sao

[ anh  ấy

từ

nào

[cũng

nhớ ]]]

PRN.DEM

NEG know

why

PRN.DEM

word

WH

OP

recall

‘She doesn’t know why he remembers every word.’

These results are theoretically significant, in as much as they reinforce the conclusion that tại sao is merged/adjoined relatively high in the left periphery: below Top, but above all other pre-subject positions. The implication, for those that subscribe to a cartographic account of the Left Periphery, must be either that ReasonP is ordered differently in Vietnamese relative to other CP projections, so that the functional sequence is subject to parametrization, or—taking a universalist stance—that ReasonP is generally projected significantly higher than supposed by Rizzi and Shlonsky & Soare, something like (91) perhaps:

(91)	ForceP  >  IntP  >  TopP  >  ReasonP  >  ModP  >  WhP  > … FinP
	(cf. Shlonsky & Soare, 2011: [35])

Although I remain unconvinced of the viability of this approach, especially given the variability of tại sao placement with respect to Topics and Adverbials, it could still derive the finite vs. non-finite asymmetry (*why-to), with which we started this section: if ReasonP was projected higher than WhP, the *why-to constraint could be derived by splicing the top three nodes (92a), as originally proposed (see 78), assuming covert movement to Int in Vietnamese; alternatively, the splicing could be more radical, entirely removing ReasonP and ModP (as in 92b):

(92)	a.	ForceP >  IntP > TopP > ReasonP >  ModP > WhP >  … FinP
	b.	ForceP >  IntP > TopP > ReasonP >  ModP > WhP >  … FinP

Note that the latter analysis obviates the need to assume raising to Int (whether overt or covert movement): *why-to would be blocked ab initio, for want of an initial merger site. (The alternate possibilities in (92) could also explain the inter-speaker variation in the judgments in (79) above: speakers that accept long-distance readings would have the splicing in (92a); those for whom long-distance readings are blocked would have the full splice in (92b).)

Which brings us, almost, to the end of our descriptive journey. The one question remaining is whether tại sao displays the same finiteness restriction as observed in English and Italian, despite its not raising overtly to Int, even in finite clauses. If not, all bets are off, so to speak: Vietnamese is simply different from English and Italian with respect to why complements. On the other hand, if Vietnamese does exhibit the same constraint, lack of overt wh-movement notwithstanding, then some further consideration must be given, either to better accommodate the distributional facts of Vietnamese within a revised cartography, or to find alternative, non-feature driven, explanations for the constraint. The choice, I suggest, depends as much on one’s theoretical aesthetics as on any empirical argument.72

The key contrasts are to be found in the comparison between the examples in (93) and (94); compare the English examples in (70) and (71) above.

(93)

a.

Chị ấy

biết

chị

nên

nói

cái

gì.

PRN.DEM

know

PRN

should

say

CLF

what

‘She knew what she should say.’

b.

Chị ấy

biết

chị

nên

đi đâu.

PRN.DEM

know

PRN

should

go where

‘She knew where she should go.’

c.

Chị ấy

không

biết

chị

nên

giải thích vấn đề

như thế nào.

PRN.DEM

NEG

know

PRN

should

explain problem

as how

‘She doesn’t know how she should explain the problem.’

d.

Chị ấy

tự hỏi

chị

nên

rời đi

khi nào.

PRN.DEM

wonder

PRN

should

leave

when

‘She wonders when she should leave.’

e.

Chị ấy

tự hỏi

tại sao

chị

nên

rời đi.

PRN.DEM

wonder

why

PRN

should

leave

‘She wonders why she should leave.’

(94)

a.

Chị ấy

biết

nói

cái

gì.

PRN.DEM

know

say

CLF

what

‘She knew what to say.’

b.

?Chị ấy

biết

chị

nên

đi đâu.

PRN.DEM

know

PRN

should

go where

‘She knew where to go.’

c.

Chị ấy

không

biết

giải thích

vấn đề

như thế nào.

PRN.DEM

NEG

know

explain

problem

as how

‘She doesn’t know how to explain the problem.’

d.

?Chị ấy

tự hỏi

rời đi

khi nào.

PRN.DEM

wonder

leave

when

‘She wonders when to leave.’

e.

?Chị ấy

tự hỏi

tại sao

rời đi.

PRN.DEM

wonder

why

leave

‘*She wonders why to leave.’

Although once again not as clear as one might wish, the results are consistent with formal expectations: whereas all of the finite options in (93) are grammatically acceptable, the pattern of results among the non-finite complements (94)—especially the contrast between (94a) and (94c) on one hand, and (94e, tại sao) on the other—is consistent with the splicing analysis in (92b), which would prevent tại sao from being merged, whilst allowing other wh-elements.

Nevertheless, there remains the two-fold problem: first, that (94b) and (94d) are less than fully acceptable; second, that (94e) is more acceptable (for the speakers consulted) than it should be, given the theory. On the first count, it is worth noting, with respect to (94b), that other licit instances of đi đâu (‘to go where’) are attested, such as the example in (95a) below. As for the marginality of khi nào in (94d), other when-adjuncts are possible with non-finite complements ((95b); furthermore, khi nào becomes more acceptable if fronted (95c):

(95)

a.

Cậu Abdullah

đồng ý

dẫn tôi đi….và

chỉ cho tôi biết

đi đâu.

PRN Abdullah

agree

take me go and

just let me know

go where

‘Abdullah agreed to take me…and to let me to know where to go.’

b.

Chị

không

biết

[ rời đi

lúc

nào ]

PRN

NEG

know

leave

time

WH

‘She doesn’t know when to leave.’

c.

Chị

không

biết

[ khi

nào

[ rời đi ]]

PRN

NEG

know

time

WH

leave

‘She doesn’t know when to leave.’

As for the marginal acceptability of (94e), I have no clear explanation other than to note that even some English speakers find ‘*why-to’ sentences acceptable. Possibly, then, this could be a sampling error; alternatively, these speakers may exceptionally allow the shorter splicing in (92a). These wrinkles notwithstanding, the general pattern is in line with formal expectations.

2.5. Summary

Aiming for a better understanding of the fine structure of the left periphery in Vietnamese, we have examined ordering constraints and interactions among five different kinds of constituent appearing to the left of the core proposition: (i) complementizers/relativizers {rằng, liệu, (mà), (là)}: (ii) topicalized expressions, optionally marked by thì; (iii) fronted adverbials appearing in the left periphery; (iv) fronted weak indefinites QP_∀ (arguments and adjuncts), with universal interpretations; (v) the wh-expression tại sao (why). Except for QPs, all these elements appear not only to the left of the proposition, but also to the left of the non-topicalized subject: that is to say, they appear above FinP. Hence, details of their interaction and relative orderings provide important clues to the fine structure of the Left Periphery.

With respect to complementizers, the rigid ordering of rằng vs. the multifunctional element là illustrated in (19) above provides evidence of at least two positions in the highest layer of the clause:

(96)	ForceP  >  IntP  >

As for embedded topicalization, our investigation reveals an important difference between Italian, with its proliferation of Topic positions (above and below Int) and Vietnamese, having a unique lower slot below Int:

(97)	ForceP  >  IntP  >  TopP  >

What is more, despite displaying the same functional contrast between Aboutness Topics and Contrastive Topics (43–45) as found in Japanese, there is no evidence for any positional distinction between the two topic types; in fact, there is no direct evidence from any source for a distinct focus position (Foc). Pending further study, therefore, I suppose that this position is not a component of the Vietnamese left periphery.

Moving further down the clausal spine, the marginal possibility of placing tại sao below the topic in embedded clauses (86b) is consistent with the existence of a dedicated ReasonP projection—from which tại sao typically raises to have scope over the Topic constituent (86a), as in (98). Alternatively, tại sao may be a pure adjunct, as it is in bare non-integrated structures in English (Why not go? Why say that?), its position determined by interpretive scope requirements; in that case there is no need to posit a dedicated Merger site.

(98)	ForceP >  IntP >   TopP >  (ReasonP) >

Finally, the interactions between adverbials and fronted quantifier expressions (QP_∀) reveal an important pre-subject/post-subject distinction: whereas either order is acceptable where both constituents appear to the right of the subject, fronted adverbs must appear to the left of universal QPs when both appear in the left periphery (65). This is consistent with Rizzi & Bocci’s proposal of a distinct Mod position for fronted adverbials (albeit no special interpretive effect applies to this position in Vietnamese).73

(99)	ForceP >  IntP  >  TopP  >  (ReasonP)  >  ModP  >  (WhP)  > … Fin

3. Conclusions (Nusquam Sunt Dracones)

As implied in the introduction, cartography is a curiously unsatisfying product of exploration: whether considered literally or figuratively, a tremendous amount of legwork is required only to demystify uncharted territory, especially in the higher ground; this is true whether one is hiking up (in typical physical surveys, before the age of drones) or climbing down (in syntactic ones). Mapping is only a means, not an end in itsIlf.

It is nonetheless necessary work, since one can Inly ask interesting why-questions (of the broad range kind) once one has a clear picture of the lay of the land: where what is, or—more pertinently—where why? is. Hence, though much of the resulting chart of the Vietnamese left periphery may be unsensational—in contrast to a famous medieval map of uncharted territory,74 there are no dragons here—what has been uncovered allows us to pose more useful research questions. These include, but are not limited to, the following:

• If, as seems to be the case for Vietnamese, the availability of constituent orders within the left periphery is determined more by semantic considerations—scope and scope evasion—than by the featural properties of dedicated positions, is this also true of other languages such as Italian, or are we dealing with a formal typological contrast? (Consider, for example, the difference between Contrastive and Aboutness Topics, the variable positioning”of a’verbs and weak indefinite (universal) QPs, the pre-/post-Topic positioning of tại sao: in each case, observed alternations are more consistent with functional-interpretive demands than with a fixed cartography);
• Vietnamese also challenges the idea of a sharp delineation between main and subordinate clauses. At least in some contexts what would be unambiguous subordinators in English and Italian have a much looser linking function; other constructions—passives, causatives, other serial verb structures—exhibit similar main~subordinate indeterminacies. Once again, this raises the question of whether such differences imply a typological contrast between hypotactic and paratactic languages (as suggested in Clark, 1992), or whether they simply urge a more flexible and nuanced view of functional categories;
• There also remain unanswered questions concerning the position of (non-topicalized) subjects. One specific issue is whether canonical subjects occupy a position within ‘IP’ or in the intermediate zone of the low left periphery; alternatively, whether the region conventionally labelled ‘IP’ may be more highly articulated, to include a lower FocP (cf. Belletti, 2004);
• More narrowly, a conceptual puzzle remains about the “*Why-to” constraint. Splicing of non-finite clauses (Shlonsky & Soare, 2011), Truncation (Hooper & Thompson, 1973), Exfoliation (Pesetsky, 2019/2021) typically operates to facilitate extraction and binding, thus increasing effability: in the case of why questions, however, splicing reduces effability. Whether considered from a formal or functional standpoint, this is a theoretical mystery. Why should this be, even in as permeable a language as Vietnamese?

So, whilst the fact-finding may be nearly complete, for now, the real discoveries still lie in front of us, once we begin to engage with the theory. And at that point, it may well turn out that the map is wrong. Maybe we should just head out, sans savoir pourquoi…

Naturally, Baudelaire said it better:

• …Mais les vrais voyageurs sont ceux-là seuls qui partent
• Pour partir; coeurs légers, semblables aux ballons,
• De leur fatalité jamais ils ne s’écartent,
• Et, sans savoir pourquoi, disent toujours: Allons!

•   …But real travellers are just those for whom departure
•   Is its own reward; [who leave], hearts light as air
•   Not to evade their fate/[but] always declaring
•   – without knowing why—‘Let’s go!’

•   Charles Baudelaire, Le Voyage [final stanza]
(Les Fleurs du Mal, 1861)