Final Modals, Adverbs and Antisymmetry in Vietnamese

Abstract

Cet article etudie la syntaxe et la semantique des constructions du vietnamien ou l’element auxiliaire apparait a droite du syntagme verbal. A premiere vue, cette distribution represente une objection aux propositions universalistes de Kayne 1995 et Cinque 1998, objection qui a deja fait l’objet de deux analyses (Duffield 1998 et Simpson 1997, 1998). Le present article reexamine les faits soutenant ces precedentes propositions. Il montre que ces faits sont insuffisants pour justifier la complexite derivationnelle de l’une ou de l’autre interpretation, et propose un traitement plus simple, qui prend en consideration l’interpretation des conditions grammaticales et extragrammaticales regissant le placement de l’auxiliaire en vietnamien, tout en conciliant ces distributions avec les principes universels.

Full text

Érudit
est un consortium interuniversitaire sans but lucratif composé de l'Université de Montréal, l'Université Laval et l'Université du Québec à
Montréal. Il a pour mission la promotion et la valorisation de la recherche.
Érudit
offre des services d'édition numérique de documents
scientifiques depuis 1998.
Pour communiquer avec les responsables d'Érudit : info@erudit.org
Article
"Final Modals, Adverbs and Antisymmetry in Vietnamese"
Nigel Duffield
Revue québécoise de linguistique
, vol. 27, n° 2, 1999, p. 91-129.
Pour citer cet article, utiliser l'information suivante :
URI: http://id.erudit.org/iderudit/603177ar
DOI: 10.7202/603177ar
Note : les règles d'écriture des références bibliographiques peuvent varier selon les différents domaines du savoir.
Ce document est protégé par la loi sur le droit d'auteur. L'utilisation des services d'Érudit (y compris la reproduction) est assujettie à sa politique
d'utilisation que vous pouvez consulter à l'URI https://apropos.erudit.org/fr/usagers/politique-dutilisation/
Document téléchargé le 13 September 2016 07:45

Revue québécoise
de
linguistique, vol. 27,
n°
2,1999,
© RQL
(UQAM), Montréal
Reproduction interdite sans autorisation
de
l'éditeur
FINAL
MODALS,
ADVERBS
AND
ANTISYMMETRY
INVIETNAMESE*
Nigel Duffield
McGiII University
1.
Introduction
T
his paper is concerned with certain issues raised by the existence of
Vietnamese sentences having the form in (1) below, in which a modal
element, diroc. appears to the right of the main verb, often in clause-final posi-
tion.1 This 'final modal' phenomenon is not exclusive to Vietnamese: as Simp-
son 1997, 1998 discusses, in a number of areally and typologically related
languages, elements corresponding to English can are also placed to the right
of the main predicate complex.
(1) a. Ông Quang mua cái nhà duçfc.
PRN Quang buy CLS house can
"Quang can buy a house."
b.
Co yêu long chap nô'i vdi Thuc duçc.
PRN
weak heart get-married again with Thuc
CAN
"She was again scarcely able to resist marrying Thuc."
* Acknowledgement. This research - part of the McGiIl Vietnamese Grammar Project - was
supported
by an
FCAR Nouveau Chercheur Award (NC-1759) from
the
Government of Quebec.
I am extremely grateful to the following consultants: Le Tarn Hang, Julie Nguyen, Nguyen
Phuong Hong, and especially Tieu Thi 1
nanh
Xuân and Nhuy
Truong;
as well as to Hironobu
Hosoi
and
Ingrid Leung for invaluable research assistance.
I am
also grateful
to
Claire Lefebvre,
Ayumi Matsuo, Lisa Travis, and an anonymous reviewer, whose comments and suggestions
helped to improve this paper.
1
One of the goals of
the
Vietnamese
Grammar
Project is
to
compile
a
bibliography of accessible
books and articles on Vietnamese grammar. For a current list or to add relevant articles to the
archive, the reader is referred to http://ngd.lin^uistics.mcgill.caA^ietnamese.

92 HNAL MODALS, ADVERBS AND ANTISYMMETRY IN VIETNAMESE
Sentences of this type attract more than local interest because they seem to run
counter
to
two recent theoretical claims about Universal Grammar. Specifically,
these data appear to challenge Kayne's 1995 proposal that heads must always
precede their complements underlyingly, as well as recent claims by Cinque
1998 for a universally fixed hierarchy of markers of tense, aspect and modality.
Both of these general proposals are outlined in more detail in the next section.
The attractiveness of the larger theoretical claims has led researchers to
quite complex derivations for the Vietnamese sentences in (1) (or for their
structural equivalents in Thai and Cantonese). In order to reconcile the surface
data with universal
templates,
analyses have been put forward in which syntactic
material that is base-generated to the right of the modal is moved leftward,
resulting in a derived rightward position for the modal element. This is
represented in (2). Since what is raised is a phrasal constituent containing a
theta-role-assigning predicate, these analyses have been referred to as predicate
(-phrase)
raising analyses.2 As will be discussed here, the details of these
predicate raising analyses differ in important respects, crucially, in whether
one or more phrasal constituents
are
raised around the modal
element.
Therefore,
the derivation in (2) simply schematizes the common feature of these accounts.
(2) XP
YP X'
YP
t
J
Given this type of derivational analysis, both theoretical problems are
apparently simultaneously resolved: Kaynian Antisymmetry is satisfied, since
the modal element now appears to the left of its presumed complement
underlyingly; and Cinque's hierarchy is respected, since all modal elements
correctly
c-command
thematic material (the maximal
VP),
again underlyingly.
The purpose of this paper is to re-examine the evidence for this type of
approach to final modals. I review two recent analyses, that of Simpson 1997,
1998 and my own (Duffield 1998). The claim will be that, in the general case,
the empirical evidence fails to support any type of predicate-raising analysis
2
This
is to distinguish them from cases of simple verb-movement, in which only the head of
the
verb-phrase is moved leftward. The available evidence suggests that Vietnamese does not have
overt verb-movement out of the maximal VP, though there is some evidence for VP-internal
movement; see Duffield (in prep, a) for details.
Ông Quang mua cai nhà Xo
dirac

NIGELDUFFIELD 93
for sentences of this type, albeit a raising analysis is supported in certain
restricted contexts. If this is the correct conclusion, a different account is
required. The alternative proposed here seeks to reconcile the Vietnamese data
with universal hierarchies, but without recourse to syntactic movement: I ar-
gue that the apparently anomalous position of the modal is better explained by
the interaction of both formal (syntactic and phonological) and functional
(parsing) principles. This analysis is then shown to extend to a range of other
right-peripheral elements in South East Asian languages (cf. Cheng 1997).
The paper is organized as follows. Section 2 provides an overview of
basic clause structure and of
the
final modal phenomenon in Vietnamese. This
is followed in section
3
by a discussion of the two Universalist
claims,
those of
Kayne 1995 and Cinque 1998, for which the phenomenon poses an apparent
difficulty. Section 4 presents the two previous treatments of final modals, and
examines the strength of the evidence for either account. In the fifth section, an
alternative proposal is presented. In the conclusion, I briefly consider a number
of broader theoretical and methodological implications of this analysis.
2.
Vietnamese Clause Structure
In this section, the main distributional properties of Vietnamese clauses
are presented, to illustrate the contrast between the modal element diroc and
other grammatical elements, including other modals.
With the exception of the phenomenon under discussion here, Vietnamese
is a paradigm example of a strictly head-initial SVO language, displaying what
Hawkins 1990, 1995 refers to as strong 'cross-categorial harmony'. This is of
course the central background fact of the paper: if Vietnamese were generally
head-final, like Japanese for example, the appearance of modals following the
verb-phrase would be completely unremarkable.
Independently of any particular framework, the observations listed in (3)
below provide clear indication that Vietnamese is head-initial.
(3) a. Vietnamese has prepositions, rather than postpositions ([P NP]);
b.
Objects, both direct and indirect, follow the verb unless topicalized,
in which case they appear clause-initially (VO);
с Noun classifiers and quantifiers precede the noun they modify
([DNP]);
d. Possessor noun-phrases and attributive adjectives follow the noun
they modify (N GEN, [NAP]);

94 HNAL MODALS, ADVERBS AND ANTISYMMETRY IN VIETNAMESE
e. Relative clauses follow the noun they modify (N REL);
f. (Most) Degree Modifiers follow the Adverb/Adjective they modify
(ADEG);
g. Native (i.e., non-Sino-Vietnamese) compounds (NN, NA, VN) are
consistently left-headed.
In addition to these properties, other, more theory-dependent, facts are
indicative of the strict head-initiality of Vietnamese. In Duffield
(1998,
in prep,
b.),
I present an analysis of Vietnamese clause structure as sketched in (5)
below. In this analysis, the fixed distribution of tense, negation and topic
markers, relative to subject NPs and adverbials, is used in support of the
theoretical claims given in (4):
(4) a. Vietnamese matrix clauses involve at least three functional categories
above VP (CP (Topic Phrase), TP, AssertionP (NegP));
b.
Vietnamese subject arguments raise overtly to [Spec, TP];
с Vietnamese main verbs do not raise overtly out of the maximal VR
(5) TopP
së/dâ (không)
NEG/ASP°
VPl
CO
Since the analysis proposed in this paper here appeals directly to this type
of representation, it is worth briefly rehearsing some of the evidence for the
claims in (4) above.
2.1 Preverbal Functional Categories
Consider first the evidence for the three functional categories enumerated
in (4a): TopP, TenseP and AssertionP, respectively. The most direct kind of

NIGELDUFFIELD 95
evidence for these projections comes from the distribution of
Topic,
tense and
assertion morphemes, respectively. The task of finding such evidence is slightly
complicated by the fact that most of these markers -tense
markers,
in particular-
are optional in declarative clauses; however, the crucial point to observe is that
these elements exhibit a rigid ordering whenever they are phonetically realized
in the same clause.
The examples in (6) illustrate the distribution of topic marker thi- This
type of topicalization structure is extremely common in Vietnamese, especially
where there is a contrast, either implied or explicit as in
(6c).
As examples (6a)
and (6b) indicate, thi is generally omitted in non-contrastive contexts. Almost
any type of constituent may be topicalized, including dependent clauses. The
examples in (6d) and (6e) illustrate the conditional function of thi, while the
passage in (6f) shows its use in more neutral 'topic...comment' contexts. In
each case, exactly one topicalized constituent—underlined in (6)—precedes thi.
(6) a. (Con) toi (thi) (toi) hoàn-toàn tan-thành.
(as for) I TOP I completely approve
"As for me, I fully approve of it."
b.
Ngucri do (thi) toi không biét (anh áy).
person TOP I NEG knOW PRN DEM
"I don't know that person."
c. Co nói dirçrc tiéng Anh không?
you speak
CAN
language English
NEG
"Can you speak English?"
— ít thôi. Tiéng Pháp thi toi nói tôt làm, nhirng tiêng Anh
little only. lge. Fr.
TOP
I speak good very, but lge. Engl.
thi khó lâm.
TOP
hard very
"Just a little. I speak French very well, but English is very difficult."
d. (Nêu) со áy di Toronto thi toi không à Montreal.
(If)
PRN
DEMgo Toronto
TOP
I NEG Ье-at Montreal
"If she goes to Toronto, I won't stay in Montreal."
e. Nhirvâv thi toi không rành lâm.
as such TOP I NEG available very
"In that case, I'm not available."

96 HNAL MODALS, ADVERBS AND ANTISYMMETRY IN VIETNAMESE
f. Rung thi ram, dufrng di thi khó-khan.
forest
TOP
dense, routes-of-communication
TOP
difficult.
Chim rung, thuda thi nhfèu, con bóng nguói thi that là
bird forest TOP many, as-for shadow men
TOP
real
COP
hiém.
rare.
"The forest was dense, the routes of communication difficult. Forest
birds and wild animals were abundant, while the shadows of human beings
were truly rare." (Thompson 1987 : 244)
As will become important presently, topicalized constituents—[XP+thi]
phrases—are subject to two structural constraints: Io they always appear in
initial position preceding the subject;
2°
they are limited to one occurrence per
clause.3
With respect to TP, Vietnamese has at least three morphemes that mark
temporal (or aspectual) distinctions: the future morpheme se, the past-tense/
completive marker da, and the continuous morpheme dang.
AU
three items are
free morphemes and are generally in complementary distribution (se and dá
never co-occur).4
The contrasts in (7) and (8) below show that these three tense morphemes
share a fixed position in matrix clauses: they directly follow the subject NP
(which, as just noted, itself follows any topicalized constituent); they precede
modal elements other than dugç, sentential negation and (subject-oriented and
manner) adverbials, which themselves all precede the lexical verb.
(7) a. *Toi tiéng Pháp thi se nói tot lâm (cf. 6c)
I language French
TOP
FUT
speak good very
"I will speak French very well."
b.
*Tiêng Pháp thi se toi nói tot lâm.
lge.
French
TOP
FUT
I speak very well
"I will speak French very well."
3 In this respect, tłu exhibits the same rigid * second position effect' as the finite verb in verb-
second languages.
4 More detailed discussion of the semantics of these morphemes is provided in Duffield (in
prep.
b).
What is important for present purposes is that these are 'INFL-related elements', in the
sense that they always appear in a functional position higher than the maximal VP (and than the
proposed Assertion Phrase).

NIGELDUEFIELD 97
(8)
a. Toi (se) cân
than
(*se)
viet láthd
này.
I
FUT
carefully
FUT
write letter
DEM
"I will write this letter carefully."
b.
Anh ay (da) can
than
(*da) doc
quyén sách.
PRN
DEM PAST
carefully
PAST
read
CLS
book
"He read
the
book carefully."
с
Toi cho là
ngày
may
trôi
(se)
không
(*së) mua.
I think
COP
tomorrow
sky
FUT
NEG FUT
rain
"I think that
it
won't rain tomorrow."
d. Anh ay (dâ)
không
(*dâ) vê
Viêt
Nam.
PRN
DEM PAST NEG PAST
return Vietnam
"He
has not
returned
to
Vietnam."
(9)
a. Co áy se
không
gap em
duąc.
PRN
DEM
FUT
NEG
meet
PRN
can
"She will
not be
able
to
meet with
me."
b.
Co áy se
không phâi
gap em.
PRN
DEM FUT NEG
MUST meet
PRN
"She will
not
have
to
meet with
you."
c.
Co áy se
phâi không
gap em.
PRN
DEM FUT MOD NEG
meet
PRN
"She will have
to not
meet with
you."
d. *Co áy
không
se gap em.
PRN
DEM NEG FUT
meet
PRN
"She will
not
meet
you."
The sentences
in
(8a-c),
for
example,
show that tense must precede manner
adverbials,
as
well
as the
negation morpheme không: those
in (9)
illustrate
the
positioning
of
tense
and
negation morphemes relative
to
pre-verbal modal
elements other than
diroc:
here,
the
modal phai
(must).
Notice here,
in
(9b)
and
(9c),
that the negation morpheme may either precede or follow modal elements;
negation directly precedes whatever lexical constituent
it
negates. However,
negation always follows
any
tense element.
The next functional projection, immediately subjacent
to TP in (5),
here
labeled AssertionP (AsrP),
is the one
most relevant
to the
present analysis.
In

98 HNAL MODALS, ADVERBS AND ANTISYMMETRY IN VIETNAMESE
Duffield 1998, I propose that this functional projection is headed by the
Vietnamese morpheme со; the precise status of this element will prove crucial
in the alternative analysis of dime proposed
here.
Co appears in several different
types of Vietnamese sentences: in interrogative and emphatic contexts, and
also in ellipsis contexts. In fact, in many respects, со functions very similarly
to English ¿fo-support.5
(10) a. Horn qua anh không (có) den nhà chi.
yesterday
PRN
NEG
ASR
go house
PRN
"He didn't go to your house yesterday."
b.
Horn qua anh (со) dên nhà chi không?
yesterday PRN ASR go house
PRN
Q
"Did he go to your house yesterday?"
с Anh (có) di vê Viêt Nam không?
PRN
ASR
go return Vietnam Q
"Did he return to Vietnam already?"
d. Anh áy da di vê Viêt Nam, (со) phâi không?
PRN DEM PAST go return Vietnam ASR right Q
"He already returned to Vietnam didn't he?"
e. Co phâi anh áy dive Viet Nam không?
ASR
right
PRN
go return Vietnam Q
"Did he return to Vietnam already?"
The examples in (10b) and (10c) illustrate typical Yes-No questions in
Vietnamese, signalled by placing the negation element không sentence-finally,
with the assertion/interrogative marker со appearing in medial position. These
should be compared with the negative sentence in (10a). There are two points
to observe about these examples. First, with the exception of copular construc-
tions,
the со morpheme itself is always optional in Yes-No questions; the
presence of khôngis sufficient to signal the question. Second, there is nothing
inherently or implicitly negative about such expressions: in sentence-final, as
opposed to pre-verbal position, không functions as a pure question-marker (or
Q-marker, as indicated in the gloss). In what follows, I will argue that this is
because sentence-final không is licensed through c-command by the [±wh]
5 ££ may also function as a lexical verb corresponding to English 'have' indicating possession.
It is probably not coincidental that both of these unrelated languages should employ a functional
auxiliary element as a lexical verb meaning 'have/own'; see, for example, Kayne 1993.

NIGELDUFFIELD 99
features of AsrP. The examples in (lOd-e) show that со also appears in
Vietnamese tag-questions and clefts. Co also appears in non-interrogative
emphatic contexts, as illustrated in (11), and in elliptical responsive contexts in
(12),
used in place of the lexical verb.6 Notice that in every case in (11), со
appears as the lowest functional element preceding the lexical
verb:
the crucial
contrast here is in (lib), which shows that the sentential negation morpheme
must precede со, rather than vice versa. The contrasts in (llc-d) show further
that со also invariably follows the tense morphemes (se, da).
(11) a. Co X. со ànhô'i-lô.
PRN X. ASR eat-bribes
"Miss X. did take bribes."
b.
Co X. (*co) không (со) anhô'i-lô.
PRN X. ASR NEG ASR eat-bribes
"Miss X. did not take bribes."
c. Co X. (*có) da (có) anhô'i-lô.
PRN X. ASR PAST ASR eat-bribes
dl Co X. (*co) se (со) an hô'i-lô.
PRN X. ASR FUT ASR eat-bribes
(12) a. Chi со viét th- không?
PRN ASR write letter Q
"Did you write/Will you write the letter?"
b.
Anh со mua quyën sách này không?
PRN ASR buy CLS book DEM NEG
"Did you buy/Will you buy this book?"
с Co /không (со)!
"Idid!/Ididn't!"
The examples in (10-12) thus provide distributional as well as interpretive
evidence that со heads a functional projection immediately above VP. That
6 For some speakers,
ç& can
only be used
to
respond to [+telic]
verbs;
for states and activities the
predicate itself is required. Thus, s¿ is less than fully acceptable in contexts such as (a) below.
(a) Chi (có) bân không?
PRN ASR busy Q
"Are
you busy?"

100 HNAL MODALS, ADVERBS AND ANTISYMMETRY IN VIETNAMESE
this is a functional projection, as opposed to an adverbial position, for ins-
tance, is suggested by the fact that the interpretation of
со,
like that of dummy
do in English, is functionally determined: со is interpreted as a question marker
in interrogative contexts, but as an emphatic marker in declaratives. This type
of functional determination is arguably a diagnostic of functional, rather than
lexical, elements.
The idea that со heads a projection associated with the interpretation of
the whole sentence, rather than just the verb-phrase, is supported by the fact
that sentential negation, which in all other instances directly precedes whatever
it modifies, invariably precedes со rather than the verb, as in
(1
lb).
This idea is
further supported by the fact that со substitutes for the whole clause in
responsives (12c).
In addition to these language-specific facts, there is considerable cross-
linguistic as well as theory-internal evidence supporting the idea of some type
of Assertion Phrase or Polarity Phrase as a component of INFL. This is a
generalization of
the
'NegP' hypothesis to affirmative contexts that dates back
to Chomsky 1965 and Klima 1964; see also Pollock
1989,
Laka 1990, Zanuttini
1993.
Given these arguments, I assume that the negation marker and the emphatic
marker are jointly associated with a functional head Asrp between Tense and
VP.
The template in (13) summarizes the fixed linear order of the 'ENFL-related
elements' discussed thus far.
(13) topicalized XPs>Subject>Tense>Modals>Negation/Assertion>Verb
2.2 Postverbal Modal placement
The previous section has shown that with the exception of sentence-final
Q-markers, Vietnamese consistently places functional morphemes between the
subject NP and the left-edge of the
VP.
With this in mind, consider the following
examples in which the modal element diroc appears to the right of the verb.
Modal diroc is multifunctional in the sense of Baker, Lefebvre and Travis 1997,
in that it receives quite distinct interpretations as a function of its clausal distri-
bution.
(14) a. Toi kiâm viêc diiçrc.
I find work can
"I can find work."

NIGELDUFFIELD 101
b.
Toi kiém duçrc viêc.
I find can work
"I can find work/I managed to find work."
These examples as well as those in (1) above show that duac invariably
follows the main verb when the intended interpretation of
this
modal is alethic
(abilitative) or, in some cases to be discussed below, epistemic. It should be
noted that when used deontically or as a passive auxiliary, duac appears pre-
verbally, as would otherwise be expected. The examples in (15) illustrate these
pre-verbal instances of
duac:
in (15a), duac corresponds to the deontic use of
can; in (15b), duac functions as a passive auxiliary; in (15c), duac is standardly
translated as 'receive, obtain, be benefitted by'. Although such pre-verbal uses
of duac are doubtless of some theoretical interest, they will be ignored
henceforth, since their analysis does not obviously call for any type of
movement.7
(15) a. Toi duąc kiém viêc.
I сап look-for work
"I am/was permitted to find work."
b.
Danchung duac chinhphû хау cho mot cái càu.8
people PASS government build for one CLS bridge
"The people had this bridge built for them by the government."
с Sáng này chi toi duorc thd.
morning this sister I receive letter
"My sister received a letter this morning."
d'. Chô này làm cho nguôita duąc mąnh-khoe.
place this make give person receive healthy
"This place makes one healthy."
Returning to the sentences in (14), it should be noted that while both
examples allow an abilitative interpretation for duac (which is the central focus
of this paper) the (b) example also allows a past tense achievement reading,
hence the alternative translation.
A
similar ambiguity is observed in English: 'I
7 Thomas 1988 provides the clearest statement of the generalizations upon which to build any
analysis of passive
duçrc:
see also Duong 1971, Clark 1971. For present purposes, the point to
observe is that (distributionally) passive
duqrc
presents no particular problem for
a
straightforward
SVO treatment.
8 Example (b) is from Duong 1971; examples (c) and (d) from Thompson 1987: 344.

102 FINAL MODALS, ADVERBS AND ANTISYMMETRY IN VIETNAMESE
could find work' can be interpreted as either 'was able to' or 'managed to'.
Elsewhere, I have suggested that for Vietnamese this ambiguity reflects a struc-
tural effect, that just in this position duac should be interpreted as an aspectual
element, in association with the functional category AspP within VP; see
Duffield 1998, in prep, a, cf. Travis
1991.
Whether or not this proves correct,
the fact remains that diroc retains an abilitative reading post-verbally, either
when it immediately precedes the object as in (14a), or in sentence-final posi-
tion in (14b); this constitutes a challenge to universalist assumptions regarding
headedness and modal placement.
3.
Universalist Challenges
Postverbal duac poses a potential challenge to two recent universalist
proposals regarding phrase-structure: the Antisymmetry proposals of Kayne
1995 and Cinque's 1998 claims regarding modal and adverbial placement.
3.1 Kayne 1995
Let us first consider Kayne 1995. Kaynian Antisymmetry imposes a strict
ordering requirement on heads, specifiers and complements universally, such
that phrasal constituents must always be head-initial underlyingly, with the
specifier (non-complement) position to the left of the head. To account for
languages such
as
Japanese, in which the surface position of the head
is
regularly
to the right of its apparent complement, it is suggested that in such cases the
complement phrase has undergone leftward raising to the specifier position of
some higher (functional) head; only in this way can Antisymmetry be satisfied.
Kayne's attempt to constrain phrase-structure representations in this
principled fashion is attractive both conceptually and empirically for languages
other than Vietnamese. In general, as we have seen, Vietnamese poses no chal-
lenge to Antisymmetry, being otherwise strictly head-initial. It
is
only rightward
duac that presents a possible difficulty.
An important consideration here is that duac really presents a challenge
only if it is analyzed as a head taking a VP (or IP) complement. Although this
is a standard assumption about such modals, and is implicit in both previous
treatments of duac discussed below, it is not a necessary one; indeed, I will
suggest in this paper that that assumption may be incorrect.

NIGELDUFFIELD 103
3.2 Cinque (1998)
Rightward duoc presents a different kind of challenge to Cinque's 1998
proposals concerning modal placement. Following von Wright 1951 and
Rescher 1968, Cinque 1998: 137-8 distinguishes several types of modal
interpretation conveyed by English can. Of
these,
three interpretations are re-
levant
to
the present
discussion:
epistemic, alethic (abilitative) and root (deontic).
The originality of Cinque's proposal lies not in the semantic distinctions drawn,
but rather in the claim that each of these senses is associated with a particular
structural position, and that these positions are strictly hierarchically ordered.
In Cinque's 1998 hierarchy, all modal elements occupy functional projections
higher than the minimal VP: epistemic modals are generated above TP, with
deontic and alethic modals positioned between TP and
AspP.
The hierarchical
positions relevant to this paper are schematized in the template in (16):
(16) mood^^^ Tense > mood„pERMissioN > aspectuals > VP
It should immediately be clear that Vietnamese distinguishes structurally
between deontic (permission) and alethic (abilitative) uses, with the former
appearing pre verbally as predicted -compare the sentences in (15)- and the
latter postverbally. It is, of course, the post-verbal, in particular the sentence-
final, positioning of alethic (and occasionally also epistemic) duoc that presents
some difficulty for this proposed hierarchy. If the positioning of modals is
indeed universally fixed, then it might seem as if sentences such as those in (1)
and
(14)
above must involve a rather complex derivation, moving other VP-related
material to the left of the modal in order to arrive at the observed surface word-
order. This is the tack taken by Simpson 1997, 1998.
In summary, if rightward duoc is analyzed either as a head or as a modal
adverb in its canonical position, it seems to challenge universalist assumptions.
Both previous treatments of rightward duorc have addressed this challenge in
terms of derivational complexity, essentially by moving material that 'should'
appear to the right of the modal to its left (in one or more steps). In this paper,
I will propose an alternative solution, one which largely preserves the theoretical
intuitions of both Kayne and Cinque as well
as
most of the empirical advantages,
but which considerably reduces derivational complexity in deriving modal pla-
cement in Vietnamese.
Before outlining that alternative solution, I review the two previous
treatments of the problem of rightward duoc.9 The claim will be that the
9 The two analyses were arrived at quite independently: at the time of writing, neither author
was aware of
the
other's work.

104 HNAL MODALS, ADVERBS AND ANTISYMMETRY IN VIETNAMESE
empirical facts fail to provide strong support for either of these accounts in the
general case, and that the derivationally simpler account offered here is to be
preferred.
4.
Predicate Raising and Postverbal
dxtqc
In this section, I review two recent accounts of post-verbal duac, Duffield
1998 and Simpson 1997. Although both authors pursue the same general
strategy, deriving sentence-final duac by preposing other material
to
some higher
position as in (2)
above,
the two analyses diverge in important
respects.
I
present
my previous analysis first.
In Duffield
1998,1
argued for
the
analysis of
rightward
duac as schematized
in (17) below. This analysis involves the raising of a single sentential consti-
tuent into a higher topic position; essentially, I treat the raised constituent as an
obligatory sentential subject of (epistemic) duac.
(17) TopP
ZP.
Top7
Tôikiêmviêc Top0 TP
(thi) T'
To MP
M'
M0 VP
duac t.
(18) *Duofc toi kiêm viêc
CAN
I find work
4T
can find work."
This derivation yields the correct word-order in certain instances, (14a)
for example; indeed, as I shall argue directly
below,
it may be the correct analysis
for some of these cases. Nevertheless, the analysis fails in at least three impor-
tant
respects.
First, there is no real conceptual motivation for the obligatoriness

NIGELDUFFIELD 105
of topic-raising; the ungrammaticality of (18) clearly shows this movement to
be obligatory. Second, while there are plausible reasons
to
believe that epistemic
diroc might trigger subject raising of this type, this is harder to maintain for
abilitative diroc: by standard assumptions, abilitative diroc should select for,
and assign a theta-role to, an animate or volitional subject (sentential subjects
should fail to qualify). Finally, this analysis fails to derive sentences such as
(14b),
which still permit an abilitative reading, but where other lexical material,
including object complements, appear to the right of the modal. For such ca-
ses,
which are by no means uncommon, it is difficult to maintain a sentential
subject account; on the other hand, an account in terms of verb-raising for just
these cases presents theoretical problems of its own, and can be excluded on
independent grounds.
It might seem then that this account should be rejected out of hand, were
it not for certain crucial sentences where it appears to be correct. Before
considering such
cases,
I will present Simpson's 1997,1998 alternative proposal,
since the evidence bears on that analysis as well.
The first obvious difference between the two approaches is that Simpson's
analysis is explicitly comparative: Simpson argues for predicate-raising in
several areally-related South East Asian languages, including Khmer,
Vietnamese, Thai and Cantonese, with most key examples being drawn from
the latter two varieties. While this more general approach potentially yields
more interesting insights, it also runs a greater risk that the generalizations
arrived at do not in fact correctly describe a particular language. I suggest that
this is the case here, that whether or not Simpson's analysis holds for Thai or
Cantonese (about which I have nothing to say), it does not for Vietnamese.
The main difference between the two accounts, however, is with respect
to derivational complexity. Simpson's proposal is schematized in (19). It
involves a combination of predicate-raising, subject raising and Control. The
derivation of a sentence such as (14a) would proceed as follows. The subject
toi would be generated in the specifier position of
the
projection headed by the
abilitative modal dugc, from which it receives its (abilitative) theta-role; toi is
then raised to [Spec,TP] by standard feature-checking/raising mechanisms. The
remaining material, including a PRO subject for the lexical verb, is generated
as a VP complement of the modal. Movement of some or all of this lexical
material to the Specifier of FocP is taken to be driven, at least historically, by
an interpretive requirement, namely, to de-focus indefinite non-focused material;
by hypothesis, the position to the right of the modal is a Focus position. Toi
then comes to be interpreted as the subject of the lexical verb through
(obligatory) Control of the lower PRO.

106 HNAL MODALS, ADVERBS AND ANTISYMMETRY IN VIETNAMESE
(19) TP
Spec
toi.
T°
Г
FocP
Spec Foe
PRO kiêm viêc Foc0 DeP
Spec
t.
De7
De0
VP
â\xgç
t.
Simpson's purpose in his paper is twofold. On the one hand, like Duffield
1998,
the goal is to find empirical evidence to support the proposed analysis,
which in turn would remove this apparent counterexample to a universalist
claim. Though for Simpson it is Cinque's rather than Kayne's proposal, which
is the principal theoretical concern. I will examine this evidence directly. On
the other
hand,
while proposing
a
predicate-raising analysis, Simpson explicitly
argues against the derivationally much simpler sentential subject raising analysis
in
(17).
Although I will ultimately reject any type of predicate-raising analysis
in the general
case,
(14a) for
example,
I would like to argue here that the analysis
in (17) is still correct in certain instances. For this reason, it is necessary to first
consider the evidence for and against sentential subject raising.
4.1 Arguments for and against a sentential subject analysis
4.1.1 Position of the topic marker
One of the main pieces of supporting evidence for the sentential subject
analysis in (17) is the prevalence in Vietnamese of sentential subjects more
generally. As the examples in (20) illustrate, Vietnamese sentential subjects are
not introduced by any subordinating complementizer: indeed, it
is
ungrammatical
to
place a complementizer
in
sentence-initial position.10 In spite of
this,
such cons-
tructions are highly frequent, and appear to be parsed without difficulty:
10 This shows, I believe, that the presence or absence of complementizers in sentential subject
contexts
is a matter
of grammar,
as argued in
Chomsky
and
Lasnik 1977, rather
than
of processing,
as Bever 1970 argued. See the discussion of
this
question in Newmeyer 1983.

NIGEL DUFFŒLD 107
(20) a. (*Rang) ho cirćri khúc khích làm chúng em then.
(Nguyen 1997: 222)
that PRN laugh giggle make PL
PRN
embarrassed
"(The fact that) they giggled embarrassed us."
b.
(*Ràng) nhà toi à trong hem thê' này ma anh tim ra ké
that house I be in alley like this
REL
you find show
giôi lam
skillful very
"(The fact that) my house is in an alley like this yet you found it
shows that you are pretty clever." (Huffman and Hai 1980)
с Ông Ba со ngû ngon không? (Dudng 1971)
PRN
Ba ASR sleep well Q
"Did Mr Ba sleep well?"
d. Ông Ba ngû со ngon không?
PRN
Ba sleep
ASR
well Q
"Did Mr Ba have a good sleep?" (lit. Mr. Ba sleeps is good)
By themselves, the examples in (20) show only that sentential subjects
are available in Vietnamese, and that they are superficially indistinguishable
from matrix clauses (until
the
point at which the matrix predicate
is
encountered.
However, the analysis in (17) above involves further raising from the clausal
subject position [Spec, TP] into a higher [Spec, TopP], at least in certain ins-
tances. Evidence for this additional raising is given by the position of
the
topic
morpheme thi; compare (5) and the examples in (6) above.
In general, the topic morpheme is only used to contrast possibilities
established by the previous discourse, or to mark the material preceding as
subordinate to the matrix predicate (for example as the antecedent of a
conditional, these functions being illustrated by the examples in (6)). Now, it
will be recalled from the earlier discussion that topic-phrases in Vietnamese
are subject to two structural constraints, namely that they are limited to one per
clause, and that they must appear in sentence-initial position.
Given these constraints, consider the sentences in (21)
below,
which contain
both lhi and duac. Example (21a) shows duçrc used in a quotative context;
here,
where the speaker is giving a grammaticality judgment, duac indicates
the acceptability of the preceding sentence. In example (21b), the presence of
thi simply indicates less certainty on the part of the speaker.

108 HNAL MODALS, ADVERBS AND ANTISYMMETRY IN VIETNAMESE
(21) a. Anh kiêm viêc à Montreal thi (nói) duąc...
PRN find work in Montreal
TOP
(say) can
"You can say: 'you'
11
find work in Montreal."'
b.
Anh kiêm viêc thi duąc.
PRN find work
TOP
can
"(I think) you can find work."
Notice that in both cases in (21) thi appears immediately before duflc.
following the contrasted constituent. This distribution is exactly what the
sentential subject analysis in (17) predicts. It is unclear, however, how this
could be derived by the alternative Control analysis in (19).
Further problems arise for the Control analysis when it is considered that
diroc itself can be modified by the future tense morpheme se. Notice that here
the competing analyses in (17) and (19) make quite different predictions with
respect to the possible positions for this element: whereas the Control analysis
only permits the tense morpheme to appear between the subject and the first
predicate—position
1
in (22a) below—the sentential subject analysis predicts
that this morpheme should occur either in position 1 as part of the sentential
subject, or in a second position immediately preceding dugc, as in (22b). This
latter option is blocked in the Control analysis since, on Simpson's account,
the VP-predicate is fronted to a specifier position [Spec,MP] situated below,
rather than above, TP.
(22) a. [TP Anhj[ Tl [MP [VP PRO kiêm
viêc].
t. duąc t. ]] ] ]
PRN find work
can
b.
[TopP [TP Anh Tl kiêm
viêc].
thi [TP
t.
[ T2 duąc ] ] ]
PRN find work
The grammatically of the sentences in (23), then, is clearly more consis-
tent with a sentential subject analysis than with the Control account.
(23) a. Anh kiêm viêc thi se duąc.
PRN find work TOP
FUT
сап
"You will certainly be able to get a
job."
b.
Em viê't láthd cho anh thi (?së) duąc.
PRN write letter for
PRN TOP FUT
can
"She will certainly be able to write a letter for you."

NIGELDUFFIELD 109
4.1.2 Selectional Restrictions
Simpson presents four arguments against treating the material preceding
duac as a raised sentential subject. First, he claims that the Thai correlate of
duac.
dai. imposes a selectional restriction on its subject, such that the subject
of a dai sentence must be [+animate]. To support this claim, Simpson cites the
unacceptability of dai with impersonal verbs.
(24) ??fon dok dai.
rain fall can
"It can rain."
Simpson interprets this restriction as implying theta-marking: dai is taken
to select for and to theta-mark the subject of its clause. If this were the case,
then by standard assumptions the subject NP could not simultaneously be the
subject of dai and
the
subject of the fronted predicate
phrase.11
Thus,
the example
in (24) would seem to preclude the raising analysis in (17), forcing instead a
control analysis, perhaps as in (19).
It is not clear, however, that examples such as (24) really warrant this
conclusion, since at least
in
Vietnamese duac does not seem
to
require
an
animate
subject in all instances. It is true that the direct Vietnamese equivalent of (24),
namely
(25a),
is just
as
anomalous
in
Vietnamese
as
in
Thai.
However, examples
(25b) and (25c) show that duac can in fact appear in sentences in which the
subject is inanimate or impersonal:
(25) a. ??Ngày mai trod mua duąc.
tomorrow sky rain can
"Tomorrow it can rain."
b.
?Ngày mai trci со thé mua duąc.
tomorrow sky perhaps rain can
"Tomorrow it could rain."
с Chuyên này со thé xày ra à moi ndi duąc.
story this perhaps happen at each place can
"Things like that can happen anywhere."
By introducing со thé ('perhaps, is possible') in (25b), grammaticality is
improved; sentence (25c), which contains an impersonal, inanimate subject
chuyên này. is fully grammatical for most speakers. This suggests that theta-
11 Naturally, the standard assumptions could be incorrect: see, for example, Hornstein 1999.

HO HNAL MODALS, ADVERBS AND ANTISYMMETRY IN VIETNAMESE
marking is not in fact the source of the problem in sentence (25a) either.12
Notice that the expression со thé, here glossed as 'perhaps', contains the same
AsrP morpheme ей that was already observed in emphatic and interrogative
contexts; see examples (11) and (12) above. That it is the same morpheme is
strongly supported by the fact that in the negative form of this expression
'perhaps not, is not possible' со thé is replaced by the negation morpheme
không: không thé. I will return to this shortly.
The correct interpretation of the ungrammaticality of (24) is, I think,
semantic rather than structural; or, at least structural in a different way from
that assumed by Simpson. Earlier, it was claimed that in sentence-final posi-
tion diroc is interpreted as either epistemic or abilitative, whereas in pre-verbal
position duac is interpreted deontically; compare again (14) vs. (15) above.
While this remains true, it is also correct that in normal deictic contexts, the
preferred reading of sentence-final duac is abilitative (or potential) rather than
epistemic.
This is no less true for non-deontic uses of English
can,
which only (easily)
allows epistemic readings in non-deictic generic contexts: compare the cases
in (26) and (27) below.
(26) a. ??It can snow tomorrow.
b.
It can snow at any time (in Montreal).
(27) a. Ayumi can show up tomorrow, ok: abilitative ??epistemic
b.
When preparing handouts, make
extras.
Some people can show up late.
?? "abilitative (less preferred) ok: epistemic
If anything, the preference for the abilitative reading in normal contexts is
stronger in Vietnamese than in English, unless, as in the grammatical examples
in (25), the epistemic reading is coerced by со thé. If this is the correct
interpretation of the restriction in (24), then such evidence does not by itself
preclude the raising analysis in (17).
What it does suggest, however, is that the epistemic reading of sentence-
final duac is only properly licensed whenever this other modal category is
present. Below, I will suggest that
this
licensing is structural rather than semantic,
and that it provides the clue to the problem of phrase-final elements in general.
12 This alternative explanation may also carry over to Simpson's original Thai example in
(20),
though more work needs to be done to establish this.

NIGELDUFFIELD 111
4.1.3 Extraction Asymmetries
A second argument that Simpson presents against the sentential subject
analysis has to do with extraction asymmetries. Simpson observes for Thai
that, whereas relativization or topicalization out of a sentential subject is
generally ungrammatical, no such restriction is observed in dai sentences; com-
pare (28) vs. (29) (Simpson's (28-30)).13
(28) a. *phuu-chaaiO. thii [ loonkhop t.] mai dii kokhuu...
man REL she associates-with
NEG
good be-namely...
"The man who that she associates with is bad is...(e.g. John)"
b.
*sing-law-nan-na.
[
khaw phuut t. ] may dii
things-group-that-TOP he speak not good
"Those things, [that she says t. ] are bad."
(29) phuu-chaai O. thii [loon khop t. ] mai dai ko khuu
man REL she see
NEG
can is
"The man who she may not date is ... (John)"
As Simpson points out, the grammaticality of a sentence such as (29) is
unexpected if it is structurally parallel to those in (28), as a sentential subject
analysis would have it. The bracketing in (29) indicates the domain of the
sentential subject, if such an analysis were to be applied. The sentences in (30)
and (31) reveal an identical contrast in Vietnamese.
(30) a. *Ngirod dàn ông O. ma t. thích u6ng bia không tô't là Ông James.
person man REL like drink beer
NEG
good
COP
PRN
James
"*That man there., that t. likes beer is not good, is James."
b.
Chuyên do. thi [со áy nói t. ] không t6t.
matter
DEM
TOP
PRN
DEM
say NEG good
"It's not good that she says those things."
(31) Ngirôti dàn ông. ma [со áy hen t.] không diroc là Ông James,
person man REL PRN
DEM
date NEG CAN
COP
PRN
James
"The man that she cannot date is James."
Although such evidence does argue against treating the material preceding
duoc as a sentential subject, it is not clear that the Control analysis offered by
13 These examples are taken directly from Simpson's paper. The minor discrepancies between
the glosses and translations do not appear to be material to the main argument.

112 HNAL MODALS, ADVERBS AND ANTISYMMETRY IN VIETNAMESE
Simpson really fares much better in this regard. The most that can be said is
that these contrasts are weakly consistent with a control analysis such as (19),
since they would also be consistent with the null hypothesis about dai structu-
res,
namely that no predicate raising of any kind is involved in their derivation.
Precisely the same comment applies to Simpson's third argument against
(17),
which has to do with relative scope (once again in Thai constructions
with dai). Simpson points out that in sentences which involve two modal
elements (sentence-final dai plus a pre-verbal modal auxiliary) it is the other
modal element that takes scope over dai, rather than vice versa. If relative
scope
is
strictly determined
by
c-command, this result is quite unexpected under
a sentential subject analysis, since
the
pre-verbal modal, being contained within
the sentential subject, should be unable to c-command
dai.
That is to say, rela-
tive scope should be reversed in such cases on a sentential subject analysis,
contrary to fact. Simpson's example (32) is reproduced in (32) below.
(32) khun doong phoo phaasa thai dai nit-nooi.14
you must suffice speak Thai can a little
"You must be able to speak a little Thai."
Here once again there is no crosslinguistic data conflict: the examples in
(33) show that Vietnamese and Thai exhibit the same relative scope effects.
(33) a. Anh phâi nói tiêng Viêt mot it duac.15
you must speak lge. Vietnamese a little
CAN
"You must be able to speak a little Vietnamese."
b.
Co ay nên hoc à Montreal duac.
PRN
DEM
should study in Montreal
CAN
"She should be able to study in Montreal."
Naturally, one could question the necessary assumption that the relative
scope of modals is determined purely by (s-structure) c-command. If we accept
this assumption, then Simpson is clearly correct
to
suggest that these data cannot
be handled by the analysis in (17). However, it does not follow from this that
(17) is wrong in all cases; as was argued above, something like the sentential
subject analysis in (17) is required to handle examples such as (21) and (23).
14 Simpson does not say how
it
is possible for nit-nooi to modify Thai while appearing after
dai.
15
Two speakers consulted found
both
of these examples marginal with duac in
final
position: if
duac is placed before the object noun-phrase in (33a), the sentence improves considerably.
However, this is not possible in (33b), since here the only reading for duac if it is immediately
postverbal is that of completion/result: this is incompatible with the meaning of
the
modal nên.
For further discussion of
this
point see below, as well as Duffield in prep. a.

NIGELDUFFIELD 113
More importantly, to the extent that arguments from relative scope apply, they
turn out to damage the control analysis in (19) at least as much as the sentential
subject account in
(17).
As will be shown directly, Vietnamese facts concerning
the relative scope of negation cast serious doubt on the viability of any sort of
predicate-raising of duac in the general case.
To summarize, the evidence presented in this section appears somewhat
contradictory. On the one hand, there is reasonably direct evidence from the
distribution of tense morphemes and topic markers that at least some instances
of clause-final duac involve sentential subjects, as I proposed previously.
Conversely, there are contexts (those involving extraction and/or relative scope
of modals) where the sentential subject analysis makes the wrong predictions,
and where Simpson's analysis in (19) is more consistent with the facts.
In almost all of these latter contexts, however, the facts are equally con-
sistent with a much simpler
analysis,
namely, one in which there is no predicate-
raising at all, where clause-final duac really is clause- or phrase-final at every
point in the derivation. In fact, there are only two pieces of empirical evidence
(to which I now turn) that potentially favour Simpson's control analysis over
the null hypothesis (that is, no movement): all other arguments are driven by
theoretical commitment.
4.1.4 Scope and Negative Polarity
Simpson's main empirical arguments against sentential subjects and in
favour of the Control analysis both involve relative scope, but work in opposite
directions. In (32) and (33) above, we saw that modal elements following the
subject NP take scope over clause-final duac. something which should not
happen if the modal were contained within a sentential subject. Just the opposite
is claimed to obtain in contexts of pre-verbal negation in
Thai.
Simpson claims
that in a sentence such as (34), the Thai negation morpheme mai takes scope
over the underlined material, excluding
dai.
This yields the interpretation
[CAN
[you
NOT
go]],
rather than
[NOT
[you
CAN
go]],
which would be expected if mai
behaved like the modal verb in (33) above.
(34) khun mai pai кар khaw dai. (Simpson 1998: ex. (23))
you
NEG
go with him can
"You can (choose) not (to) go with him."
This argument would indeed constitute good evidence for a predicate-
raising account if the data were more robust. Once again, I cannot address the
Thai data directly; for Vietnamese, however, I have been unable to find any

114
FINAL
MODALS,
ADVERBS AND ANTISYMMETRY IN VIETNAMESE
native speaker with judgments comparable to those for (34). As the examples
in (35) show, all those consulted uniformly interpret equivalent Vietnamese
structures as if negation takes scope over final duac.
(35) a. Anh không viê't lathdduąc.
PRN NEG write letter
CAN
"You cannot write the letter."
b.
Co áy không an thit dircfc.
PRNDEM NEG eat meat can
"She cannot eat meat."
с Toi không di duąc.
PRN NEG
go сап
"Fm unable to go." (because of circumstances that make it impossible)
This is not to say that the relative position of không has no effect on
interpretation. The examples in (36) demonstrate the fact that không can either
appear preverbally
(the
unmarked
case),
or postverbally immediately preceding
duac:
in the latter case, the reading is more emphatic, and is normally
accompanied by some reason clause.16 However, the point here is that relative
scope is not reversed: không is consistently interpreted as taking scope over
duac.
Notice also that
in
(36c) diroc intervenes between the verb and
the
(generic)
object NP.
(36) a. Toi không an thit duac.
I NEG eat meat
CAN
"I can't eat meat." (due to unwillingness, lack of money, etc.)
b.
Toi an thit không duąc.
I eat meat
NEG
CAN
"I (really) can't eat meat."
(physically incapable, perhaps strict vegetarian)
с Toi không an duąc thit.
I NEG eat
CAN
meat
"I can't eat meat."
16 Simpson 1998: 6 implies that this latter position is the only possible position for sentential
negation in
Thai;
indeed, this counts as one of his main pieces of evidence that clause-final dai
is outside the Thai VP. This is clearly not true of
Vietnamese:
the preverbal order in (36a) is
unmarked and fully natural, while the order in (36b) is only used in emphatic contexts.

NIGELDUFFIELD 115
This point is reinforced by the fact that pre-verbal không can license
negative polarity items following duac. This is demonstrated by the examples
in (37) as follows. First, the examples in (37a-d) show that the indefinites ai
('who') and gi ('what') must be interpreted as wh-expressions unless they are
c-commanded by negation. Second, the interpretation of (37e-f) show that
không has scope over duac. Finally, the interpretation of ai and gi as negative
polarity items in (37g-h) shows that không also has scope over material
following duac.
(37) a. Anh viét lá thd cho ai.
PRN write letter for who
"Who are you writing a letter for?"
(*you are writing a letter for someone)
b.
Anh làm gi cho anh ay.
PRN do what for
PRN
DEM
"What are you doing for him?"
(*you are doing something for him).
с Anh không viét lá thd cho ai.
PRN
NEG write letter for who
"You are not writing a letter for anyone."
(*Who are you not writing a letter for?)
d. Anh không làm gi (hét) cho anh áy.
PRN
NEG do what at all for
PRN
DEM
"You are not doing anything for him."
(*What are you not doing for him?)
e. Anh không viét lá thd cho ai duçrc.
PRN
NEG write letter for who can
"You cannot write a letter for anyone."
(*You are able not to write a letter for anyone.)
f. Anh không làm gi câ cho anh áy duac.
PRN
NEG
do what at all for
PRN
DEM
CAN
"You cannot do anything (at all) for him."
g. Anh không viét lá thd duąc cho ai câ .
PRN
NEG write letter CAN for what at all
"You cannot write a letter for anyone."

116 HNAL
MODALS,
ADVERBS
AND
ANTISYMMETRY
IN
VIETNAMESE
h. Anh không làm diroc gi câ cho anh áy.
PRN NEG do CAN what at all for him
"You cannot do anything (at all) for him."
The Vietnamese facts in (35-37) are not necessarily fatal to Simpson's
proposed structure: there are at least two ways in which they might be
accommodated. First, one might claim that Thai and Vietnamese negation differ
precisely on where negation attaches: Thai sentences such as (34) would have
the structure of
(38a)
below, with the negation morpheme mai within the raised
VP;
the corresponding Vietnamese sentence, for example (35a), would have the
structure of
(38b),
in which negation is outside of
the
raised
VP
and above FocP.
(38) a. [TP khun. [ MP [VP PRO mai pai кар
khaw].
t dai t. ] ] ] ]
PRN NEG gO With him CAN
b.
[TP Anh. [ không [MP[VP PRO kiem
viêc].
t duoç t.] ] ] ]
PRN NEG find WOrk CAN
However, this move seems ad hoc at best, especially since it is just these
relative scope facts that constitute the remaining piece of empirical evidence
for the type of predicate raising that Simpson proposes.
4.1.5 On Pre-supposition and De-Focusing
As an alternative, it might be suggested that sentences such as (36c) and
(37g-h) are derived by stranding generic or indefinite phrases (thit, cho ai and
gi,
respectively) in situ, and raising all other material to the higher [Spec, FocP]
position. While such a move is technically feasible, and close to what Simpson
proposes, it seems to run directly counter
to
the de-focusing principle proposed
to motivate raising in other contexts. Simpson 1998: 22 writes :
Considering at least
the northern
dialects of
Vietnamese
and
the
position of
the
object,
one seems to find
the
similar pattern which appears in Thai and
Middle
Chinese.
There seems to be a
heavy preference for
indefinite non-focused
objects to precede the potential modal
and for focused DP s to follow it.
In support of this claim, Simpson cites the examples 81-82 reproduced in (39)
and (40) below:
(39) a. Toi lai xe dirąc.
I drive car can
"I can drive cars."
b.
?toi lai dirge xe. (cf. 36c)

NIGELDUFFIELD 117
(40) a. Ông áy nói moi tiéng duçc.
PRN
DEM
speak every lge. CAN
"He can speak all languages."
b.
??Ong ay nói diroc moi tiêng.
Simpson observes in a footnote (1988: fn. 12) that this contrast holds for
speakers of Northern and Central dialects of Vietnamese, but not for Southern
Saigonese speakers. Most of the speakers that I have consulted, who come
from all three regions, do seem to detect a contrast. However, the explanation
of this contrast seems to have more to do with heaviness in the sense of Heavy
NP-Shift (HNPS) than with either definiteness or focus.
This is suggested by the fact that the judgements for (39) and (40) are
reversed if the indefinite phrase is modified by an attributive adjective-phrase
or relative clause: compare (41) and (42), respectively.
(41) a. ??Toi lai xe dâttfên (va) со may ląnh diroc.
I drive cars expensive (and) have a.c. CAN
"I can drive fast cars with airconditioning."
b.
Toi lái diroc xe dât tfên (va) со may ląnh.
I drive can cars expensive (and) have a.c.
"I can drive fast cars with airconditioning."
(42) a. ??Ong áy nói moi tiêng tôicùngbiét duçc.17
PRN
DEM
speak all lge. I also know
CAN
He can speak every language that I know."
b.
??Ong ay nói moi tiéng ma ngucri ta nói à bên Nam duąc
PRN
DEM
speak every lge.
REL
people speak in side South
CAN.
"He can speak every language that people speak in the South."
c. Ông ây nói dime moi tiéng toi cüng biét.
PRN
DEM
speak
CAN
every lge. I also know
"He can speak every language that I know."
d. Ông áy nói duąc moi tiéng ma ngircti ta nói à bên Nam.
PRN
DEM
speak
CAN
every lge.
REL
people speak in side South
"He can speak every language that people speak in the South."
17 This
sentence is acceptable in the possible (though irrelevant) case where diroc is analyzed as
modifying the relative clause 'He speaks every language that
I
can know.' Similarly for (41b).

118 HNAL MODALS, ADVERBS AND ANTISYMMETRY IN VIETNAMESE
Even if Simpson's hypothesis were correct, it would still be necessary to
account for
the
grammars of those speakers who fail to show
a
contrast between
(39) and (40). Moreover, it is clear that 'heaviness' itself calls for further
explanation, though whether HNPS is a processing rule of some kind (see
especially Hawkins 1995) or a reflex of grammar (following Stowell 1981,
Larson 1988) is something that still needs to be resolved.
In fact, I believe that an analysis of certain instances of post-object duorc
in terms of a combination of verb-movement and object shift, something like
(43),
is warranted for Vietnamese, but I do not think that it holds for the
abilitative duac under consideration here (see Duffield 1998, in prep, a, for
discussion of aspectual duac). Under such an analysis, the lexical verb would
move from its base position
(VV)
through
Asp0
to v; object complements would
move from [Spec,VP2] to [Spec,AspP].
(43) VPl
TM(toi) V7
v
lái Spec
(xe)
Toi lái xe duac.
I drive car can
"I managed to drive a car.'
AspP
Asp0
diroc t
Asp
Spec
(xe này)
VP2
<V
V
Toi lái duac xe này.
I drive can car
DEM
"I managed to drive this car."
Notice that I present the aspectual interpretation of duflcfor these senten-
ces (compare (39a) above); given what has been discussed thus far, I suppose
that abilitative duac does not appear in this structure. If it did, though, we
would still be left with one of the theoretical problems with which we began.
The position of duae in (42) would no longer constitute a problem for Kayne,

NIGEL DUFFlELD 119
since this structure would conform with Antisymmetry requirements, but for
Cinque 1998 this would still present a real difficulty, since abilitative (alethic)
dirge would now be appearing in a VP-internal position (albeit not in the mini-
mal VP).
It seems, then, that in general the facts from the relative scope of negation
in Vietnamese speak against any predicate raising analysis, rather than providing
clinching evidence in
its
favor. One might still wonder, though, how the 'reversed
scope' readings illustrated in (35) above, which Simpson obtained for Thai
and from some of his Vietnamese consultants, can be explained. I have
presented evidence showing that in the general case in Vietnamese, normal
scope obtains: không is interpreted as taking scope overdirac. If this is correct,
then the explanation for the reversed scope readings cannot be structural in the
way suggested by either the Control analysis in (19) or the sentential subject
analysis in (17); that is to say, không cannot generally be contained within
some raised phrasal constituent.
As an alternative, it can be suggested that reversed scope readings in
Vietnamese are obtained in precisely the same way as in the English contrasts
in (44)
below:
namely, negation can either
be
construed
as
sentential (the default
case),
where it takes scope over the modal, as in (43a); or as constituent (VP-)
negation, in which case the modal element
is
interpreted as having wider scope,
as in (43b). In these examples, the following clause provides further indication
of the scope of negation.18
(44) a. He can't (always) eat (when he wants to).
b.
He can (always) not eat (if he doesn't want to).
On the basis of the foregoing discussion, I conclude that at least for
Vietnamese, there are no empirical arguments that support the Control analysis
in (19) over the null hypothesis in the general case. Furthermore, I conclude
that to the extent that there is any predicate-raising going on in Vietnamese, the
simpler sentential subject account provides a better analysis of the relevant
facts.
Of course, granting this conclusion returns us to the problem that inspired
the predicate-raising analyses in the first instance; namely, that ifditflc really is
subject to the principles of adverb placement found in other
languages,
it would
seem to present a challenge to the idea that adverb placement is universally
fixed. Since this proposal has considerable theoretical and empirical support,
we need to find a way out of this bind.
18 It is admittedly unclear whether such an explanation would carry over to Thai, where the
judgements as reported appear to be more categorical than in Vietnamese: more work would
need to be done to determine the constraints on (and the relationship between) constituent and
sentential negation in that language.

120 FINAL MODALS, ADVERBS AND ANTISYMMETRY IN VIETNAMESE
4.1.6 Discussion
Before sketching an alternative analysis, I can briefly summarize the paper
thus far. In a number of areally-related South East Asian languages, a modal
element corresponding to English can unexpectedly appears in post-verbal
position, usually clause-finally. The distribution of this element is unexpected
because of two universalist assumptions about phrase-structure. First, that heads
uniformly appear to the right of their complements (Kayne 1995); second, that
modal, temporal and aspectual elements are universally realized in a fixed
hierarchical order (Cinque 1998). Specifically, the relevant claim is that all
modals (but especially epistemic and alethic modals) necessarily c-command
the VP: given the first assumption, this requires that modals should appear to
the left of the VP underlyingly.
In two previous treatments of this rightward-modal problem, appeal is
made to some form of predicate-raising to derive the marked surface order.
However, both analyses fail in certain crucial respects. The sentential subject
analysis of Duffield
1998
fails to explain the following facts: Io in the unmarked
case,
the rightward modal is interpreted as abilitative, rather than epistemic;
2° the supposed sentential subject freely allows extraction (in contrast to other
sentential subjects where extraction or relativization is blocked); and 3° other
pre-verbal modal elements take scope over the final modal, (which should not
happen if these modals are contained within a sentential subject).
Simpson's 1998 analysis was shown to be unsatisfactory, for Vietnamese
at least, for a different set of
reasons.
These include the following: Io in certain
contexts, the final modal can be interpreted epistemically; 2° the final modal
can appear in certain constructions with its own (immediately preceding) tense
morpheme; 3° the main piece of scope-related evidence for predicate-raising
makes the wrong predictions for Vietnamese with respect to relative scope and
to the licensing of NPIs following a final modal;
4°
the semantic motivation for
predicate raising that Simpson proposes makes incorrect predictions, when
compared to a processing account in terms of heaviness.
5. An alternative Approach to licensing rightward elements
5.1 Underlying Hierarchies and Surface Representations
The way out of the universalist problem, I suggest, is to question the
assumption that universal conditions such as Antisymmetry or constraints on

NIGELDUFFIELD 121
adverb placement apply to all surface elements in a given sentence. This
assumption seems in any case suspect under a minimalist approach, where
surface outputs are in some sense accidental or contingent properties of the
particular derivation (Chomsky 1995). Surface distributions are contingent in
the sense that particular word-orders at Spell-Out are primarily determined by
the abstract syntactic features of the elements in the numeration, rather than by
underlying universal hierarchies. So, for example, whether one observes SOV,
SVO,
or VSO word-order in a language is taken to be determined by the PF-
requirements of verbs or DPs in a given numeration. Implicit in most current
minimalist discussions is the idea that only this subset of lexical items (the
extended projections of theta-assigning heads and their associated (DP) argu-
ments) bears the sorts of features that force movement. Again by implication,
the distribution of lexical items without such features is determined by other
factors.19
Adopting such a view, however, does not imply abandoning the idea of
universal hierarchies, although it does imply that surface representations may
not always reliably reflect such hierarchies. Standardly, it is assumed that
universal hierarchies are respected underlyingly even if the surface exponents
of these categories appear in unexpected positions. To cite a simple example
from English, tense morphemes (the surface exponents of T0) are observed
clearly within the VP at Spell-Out. This does not force the conclusion that TP
is VP-internal functional category; rather, the general conclusion is that the
relevant features of tensed verbs are checked after, rather than before Spell-Out.
The closest relevant analogy to Cinque's adverbial hierarchy
are
the various
thematic hierarchies that have been proposed, among which Baker's 1985,1988
UTAH (Uniformity of Theta Assignment Hypothesis)
is
probably
the
best known
(cf. also Rosen 1984, Grimshaw 1990). In almost every thematic hierarchy, the
highest thematic relation is that of
AGENT:
there seems to be broad consensus
that elements interpreted as bearing agent theta-roles are projected relatively
high or merged relatively late, under a derivational approach.20 In spite of
this,
elements obligatorily interpreted as agents in passive constructions, namely
agentive
by-phrases,
typically appear
to
the
right
of all other
arguments;
indeed,
they usually appear clause-finally.
19 Such factors may be grammatical (the
c-command
relation is the clearest instance of this) or
have other sources: parsability and iconicity also appear to be likely candidates.
20 This is true independently of whether thematic roles are considered as primitives (Rosen
1984) or derivative of phrase-structural or conceptual configurations (Jackendoff
1990,
Baker 1995).

122 FINAL MODALS, ADVERBS AND ANTISYMMETRY IN VIETNAMESE
If one considered only the surface word-order of long passives, as in
analysis (46a) below, by-phrases should pose a clear empirical challenge to
UTAH, in just the same way
as diroc
apparently challenges the Cinque hierarchy.
However, the standard view of by-phrases, which I suppose to be correct, is
that they are not real counterexamples to the thematic hierarchy. In an analysis
such as that of Baker, Johnson and Roberts 1989, long passives are analyzed as
respecting UTAH in that the element which receives the theta-role, and which
licenses the adjunct by-phrase (namely the passive morpheme en) correctly
c-commands other positions to which theta-roles are assigned. That is to say
(46b),
rather than
(46a),
is
the correct thematic description. Under this description,
the fact that the by-phrase itself violates UTAH is irrelevant: what is important
is the position of the licensing head.
(45) UTAH Baker 1988; cf. Grimshaw 1990, Rosen 1984
(46) The reckless driver, was giv-en a life sentence t. by the harsh judge.
a. G T 'A
b.
ATG
cf. Baker, Johnson and Roberts 1989
Consequently, the surface position of by-phrases is determined not by
hierarchical constraints, but rather through the interaction of other grammatical
and extragrammatical principles, including processing constraints: see Frazier
1987,
Frazier and Fodor 1978, Hawkins 1995. As the examples in (47) are
intended to show, by-phrases appear in a variety of (postverbal) positions
depending on relative heaviness.
(47) a. ?The reckless driver was accused of conduct likely to endanger life
by the state prosecutor,
b.
The reckless driver was accused by the state prosecutor of conduct
likely to endanger life,
с The defendant was given a life sentence yesterday by the judge/by
the judge yesterday.
d. ??The defendant was given a life sentence the day before yesterday
in Montreal by the judge.
e. The defendant was given a life sentence the day before yesterday in
Montreal by the presiding judge specially appointed to this case.

NIGELDUFFIELD 123
Precisely the same arguments can be applied to rightward duac in
Vietnamese. All that is required is that we find evidence of a licensing head in
the position predicted by the Cinque hierarchy: for the epistemic reading of
duac.
this head should appear above
TP;
for the (more usual) abilitative reading,
it should appear between TP and VP
(48) epistemic modals > tense > deontic modals > alethic modals> aspectuals> VP
It turns out that Vietnamese provides exactly the right type of element
with just the right distribution. This is the complex modal element со thé that
we have seen in a number of examples up to now. Co thé appears in two pre-
verbal positions, preceding the subject, or immediately preceding the verb. In
the former position, it is interpreted only epistemically; in the latter position,
the preferred interpretation is abilitative, although the epistemic reading is also
available in certain contexts. Crucially, for every sentence in this paper that
contains clause-final duac. except for the sentential subjects contexts, there is
a completely synonymous sentence
with
pre-verbal
со
thé added. Representative
examples are repeated below:
(V) a. Ông Quang со thé mua cái nhà (duąc).
(14') a. Toi со thé kiém viêc (duac).
(33') a. Anh со thé phâi nói tiéng Viêt mot it (duac).
(350 a. Anh không thé viêt la thd (duąc).
(360 a. Toi không thé an thit (duąc).
(370 g. Anh không thé viét la thd (duąc) cho ai cà.
(390 a. Toi со thé lai xe (duąc).
Whenever со thé is present in this position, duac is licensed, but wholly
redundant; Vietnamese speakers prefer to omit it in just
the
same way as English
speakers prefer to omit redundant by-phrases, such as 'by someone'.
If we now assume that the abilitative/alethic modal in Vietnamese is со
thé and that duac is simply parasitic on the position occupied by со
thé,
just as
by-phrases are parasitic on the head bearing the Agent theta-role, we solve the
Cinque hierarchy problem without resort to predicate raising.
5.2 Some Empirical Advantages
5.2.1 Cinque's 1988 hierarchy reconsidered
Aside from being conceptually and technically much simpler, this analysis
has a number of empirical advantages: by shifting the focus from duąc to its

124 FINAL MODALS, ADVERBS AND ANTISYMMETRY IN VIETNAMESE
licensing head со thé. Vietnamese turns out to directly support Cinque's
hierarchy, rather than challenging it. First of
all,
as
just mentioned, со thé can
appear in pre-subject position. As Cinque's hierarchy predicts, the only possi-
ble interpretation in this position is epistemic:
(49) Co thé anh áy dê'n (dirąc).
CAN PRN DEM СОШе (CAN)
"It is possible that he will come." *He is able to come.
Second, со thé actually appears in two postsubject positions, rather than
one.
In sentences containing со thé and another modal element phâi. со thé
may either appear above or (marginally) below the deontic modal
phái.
This is
illustrated in (5Oa) and (5Ob), respectively. If со thé appears higher than the
modal, as in (50a), only the epistemic reading is available; conversely, placing
со thé below phâi. as in (50b), yields only the abilitative reading. If we now
further assume, as seems plausible, that epistemic со thé functions as a (subject)
raising predicate, we can assimilate sentences (50a) with its unraised variants:
(49) and (50c). In this way, Vietnamese can be shown to exhibit precisely the
split in modal functions predicted by Cinque's hierarchy.
(50) a. Anh áy со thé phâi dén.
PRN DEM CAN DEM СОШе
"It is possible that he must come." *He must be able to come,
b.
?Anh áy phài со thé dên.
PRN DEM MUST CAN СОШе
"He must be able to come." *It is possible that he must come.
с Co thé anh áy phâi dén.
can PRN
DEM
must come
"It is possible that he must come." *He must be able to come.
5.2.2 Possible extensions
A second advantage of this indirect licensing approach to final duflc is
that the same licensing mechanism is independently required for other phrase-
final and clause-final elements in Vietnamese. These include the interrogative
use of the negation marker không. which as we saw in (10), is used to signal
Yes/No questions; (1Ob)-(IOc) are repeated below for
convenience.
The natural
assumption in this case is that final không is syntactically licensed by the pre-
verbal assertion head
со;
more precisely, không is licensed by the
[+WH]
feature

NIGELDUFFIELD 125
of this head (cf. Rizzi
1990,
Roberts
1993).21
The alternative, implicitly adopted
by Cheng 1997 and others for languages such as Chinese, is to assume that the
Q-marker itself is a rightward C0 head; as was the case with duoc, this is a
highly unintuitive conclusion, especially given the clause-initial distribution
of other complementizer elements in Vietnamese.
(lO')b.
Horn qua anh (со) dén nhà chi không?
yesterday
PRN
ASR
go house
PRN
Q
"Did he go to your house yesterday?"
с Anh (co) di vê Viêt Nam không?
PRN
ASR
go return Vietnam Q
"Did he return to Vietnam already?"
5.2.3 Additional non-syntactic factors
Finally, by setting aside purely syntactic explanations for the distribution
of final
duoc.
it becomes easier
to
take account of other non-syntactic properties
that may better explain its distribution. Although more work needs to be done
to determine this, my intuition is that the distribution of duoc is determined by
two factors. The first appears to be a universal processing constraint that prefers
to maximize immediate constituent recognition domains by re-arranging and
extraposing heavy material to the right of lighter elements (Hawkins 1995).
According to this principle, an optional adverbial such as duoc will tend to
appear on the right periphery following obligatory arguments, unless these
latter elements are so heavy that they extend the constituent recognition domain
for the clause.
The second constraint governing the distribution of diroc is language-
particular, and has to do with tone and tonality. Northern Vietnamese has six
tones (including the neutral
tone):
high (a), low (à), low rising (à), high broken
(a),
and low broken
(ą).
Dirac is an example of a low broken tone word (nâng).
In general, it seems that a declarative sentence is considered 'better balanced'
if two phonetic conditions are met. The first is that declaratives should end
with a low or low broken tone: the majority of sentence-final particles (ma. à.
a, etc.) bear one of these two lexical tones. The second condition is that a
sequence of high or high broken tones followed by a low broken tone sounds
better than when the low broken tone interrupts the sequence. Of
course,
these
21
The analysis
can also be
extended
to
final
question
tags:
phâi không.
isn 'i it? and nhé
(Duffield
in
prep.
b).

126 HNAL MODALS, ADVERBS AND ANTISYMMETRY IN VIETNAMESE
are only preference rules which depend on the particular lexical items chosen;
in many instances, there is no choice but to use a disprefered sequence.
Nevertheless, where the choice exists, it seems to affect the position of dirge,
as the rankings in (51) illustrate; in the unmarked case, they favor duoc on the
right periphery.
(51) a. Toi không nói duąc tiêng Pháp -» Toi không nói tiêng Pháp duąc.
H LB H H H H H LB
I NEG speak
CAN
lge. French
"I cannot speak French."
b.
Toi không an duąc cá
—>
Toi không an ca duąc.
N LB H NH LB
"I cannot eat fish."
c. Toi không an duąc thit = Toi không an thit duąc.
N LB LB N LB LB
I NEG eat can meat
"I can't eat meat."
6. Conclusion
In this paper, I have examined the empirical evidence for a syntactic
approach to a particular clause-final modal element in Vietnamese. The con-
clusion of this investigation is that a purely syntactic approach receives little
empirical support in Vietnamese, though it is still possible that it may be correct
for Thai or for earlier stages of the language, about which I have little to say.
The alternative account proposed here remains for the most part syntactic, but
explains the distribution and interpretation of duoc representationally in terms
of c-command, rather than through appeal to complex movement. In addition,
it is suggested that general processing factors, as well as language-specific
phonological factors may contribute to
a
complete account of this phenomenon.
It seems that
this
is
a case
where excessive prior commitment
to
a particular
theoretical position -rather than helping researchers to make sense of the data-
has led to unwarranted complexity in grammatical description. This is hardly
an isolated case, I think, for it is always tempting, and often invaluable, to be

NIGEL
DUFFIELD 127
directed by a strong and predictive theoretical framework. The trick, as usual,
is not to resist temptation, but to avoid (literal) seduction.22
References
ADGER, David 1997 "VSO Order and Weak Pronouns in Goidelic Celtic". Canadian
Journal of Linguistics 42 9-29.
AOUN,
Joseph and Y-H. AUDREY
LI
1993
"WH-elements
in
situ:
syntax or
LF'.
Linguistic
Inquiry 24-2: 199-238.
BAKER, Mark 1985 Incorporation: A Theory of Grammatical
Function
Changing.
Ph.
D.
dissertation,
MIT.
Revised version
1988
published
by
University of Chicago Press.
BAKER, Mark 1995 Thematic Roles and
Syntactic
Structure.
Ms. McGiIl University.
BAKER, Mark, Kyle Johnson, and Ian Roberts 1989 "Passive Arguments Raised". Lin-
guistic Inquiry
20:
219-251.
BAKER, Mark, C. LEFEBVRE, and Lisa TRAVIS 1997 ms. (FCAR proposal/application),
McGiIl University.
BEVER, Thomas. 1970 "The cognitive basis for linguistic structures". In Hayes, John,
ed. 1970.
Cognition
and
the development
of
language,
277-360.
New
York,
Wiley.
CHENG, Lisa
L.
1997 On the
Typology
ofWH-Questions. Outstanding dissertations in
Linguistics series. New York, Garland Publishing.
CHOMSKY, Noam
1965
Aspects of The
Theory
of Syntax. Cambridge, MIT Press.
CHOMSKY, Noam 1995
The
Minimalist
Program.
Cambridge, MIT Press.
CHOMSKY,
Noam, and Howard LASNIK.
1977
"Filters and
Control".
Linguistic Inquiry 8:
425-504.
CINQUE, Guglielmo 1998 Adverbs and functional
heads:
a cross-linguistic perspective.
New York, Oxford.
CLARK, Mary Beth 1971 "Passive and Ergative in Vietnamese". University of Hawaii
Working
Papers in Linguistics 3-8: 103-117.
DudNG, Thanh Bînh 1911 A tagmemic comparison of the structure of English and
Vietnamese
sentences. The Hague, Paris, Mouton.
DUFFIELD, Nigel 1995 "Are
You
Right? On Pronoun-Postposing and other Problems of ,
Irish Word Order". In Raul Aronovich, William Byrne, Susanne Preuss, and
Martha Senturia (eds.), Proceedings of
the
Thirteenth West Coast Conference
on Linguistics, pp. 221-236. Stanford Linguistics Association.
DUFFIELD, Nigel 1998 "Auxiliary placement and interpretation in Vietnamese". In M.
Catherine Gruber, Derrick
Higgins,
Kenneth
E.
Olson and Tamra Wysocki (eds.)
Proceedings of
the
34th Regional Meeting of
the
Chicago Linguistic Society,
p.
95-109.
22 There is a
personal irony
in this
conclusion:
I have been
driven
by
theoretical commitment to
Antisymmetry not only with respect to Vietnamese (in Duffield 1998) but also with respect to
pronoun-postposing
in Irish
(Duffield
1995);
in the latter
case,
Adger
1997 provides the
necessary
corrective.

128 HNAL
MODALS,
ADVERBS AND
ANTISYMMETRY
IN
VIETNAMESE
DUFFEELD,
Nigel (in
prep,
a) Vietnamese functional
categories.
Ms.
McGiIl University.
DUFFIELD, Nigel (in
prep,
b)
Tense
and
Scope
Relations in
Vietnamese
Grammar. Ms.
McGiIl University.
FRAZIER, Lynne 1987 "Sentence
Processing:
A tutorial
review".
In Max Coltheart (éd.),
Attention and
Performance
XII. Hillsdale, NJ: Lawrence Erlbaum Associates.
FRAZIER, Lynne, and Janet
D.
FODOR 1978 "The sausage machine: a two-stage parsing".
Cognition 13: 291-325.
GRIMSHAW, Jane 1979 "Complement selection and the lexicon". Linguistic Inquiry 10:
279-326.
GRIMSHAW, Jane 1990 Argument
Structure.
Cambridge, MIT Press.
HAWKINS,
John
1990
"A parsing theory of word-order uni
versais".
Linguistic Inquiry
21:
223-260.
HAWKINS, John 1995 A Performance Theory of Order and Constituency. Cambridge,
Cambridge University Press.
HORNSTEIN,
Norbert 1999 "Movement and Control". Linguistic Inquiry
30-1:
69-96.
HUFFMAN,
Franklin
E.
and Tran
Trong
HAI
1980
Intermediate Spoken
Vietnamese.
Ithaca,
NY, Cornell South East Asia Program.
JACKENDOFF, Ray 1990 Semantic
Structures.
Cambridge, MIT Press.
KAYNE, Richard 1993 "Towards a modular theory of auxiliary selection". Studia
Lingüistica
47:
3-31.
KAYNE, Richard 1995
The
Antisymmetry of Syntax. Cambridge, MIT Press.
KLIMA, Edward 1964 "Negation in
English".
In
J.
Fodor
and
J.
Katz
(eds.),
The Structure
of Language. Englewood Cliffs, Prentice Hall.
LAKA, Itziar
1990
Negation in
syntax:
on the nature of functional categories and
pro-
jections. Ph.D. dissertation, MIT.
LARSON, Richard 1988 "On the double object construction". Linguistic Inquiry 19:
335-391.
NEWMEYER, Frederick 1983 Grammatical
Theory:
Its Limits and Its Possibilities. Chi-
cago,
University of Chicago Press.
NGUYÔN Dïnh Hôa 1997
Vietnamese.
Amsterdam/Philadelphia: John Benjamins.
POLLOCK, J.-Y. 1989 "Verb Movement, Universal Grammar and the Structure of IP".
Linguistic Inquiry 20:365-424.
RESCHER, Nicholas 1968
Topics
in Philosophical
Logic.
Dordrecht, Reidel.
RIZZI, Luigi
1990
"Speculations on Verb-Second".
In
Nespor,
Marina
et
al.
(eds.) Gram-
mar in
Progress:
A Festschrift for Henk van
Riemsdijk,
Dordrecht, Foris.
ROBERTS, Ian 1993
Verbs
and
Diachronic
Syntax: A Comparative History of English
and
French.
Dordrecht, Kluwer Academic Publishers.
ROSEN, Carol 1984 "The interface between semantic roles and initial grammatical rela-
tions".
In David Perlmutter and Carol Rosen (eds.), Studies
in
Relational Gram-
mar 2. Chicago, University of Chicago Press.
Ross,
John R. 1969 "Auxiliaries as main verbs". In W. Todd (éd.), Studies in Philo-
sophical Linguistics. Great Expectations, Evanston, Illinois.

NIGEL
DUFFIELD 129
SIMPSON, Andrew 1997 "VP-final modals and Pied Piping in South East Asian". In
Proceedings of NELS 28.
SIMPSON, Andrew 1998 Focus, Presupposition and Light Predicate Raising in South
East Asian, ms. London, SOAS.
STOWELL, Timothy 1981 Origins of Phrase
Structure.
Ph.D. dissertation, MIT.
THOMAS, Margaret 1988 "Submissive Passives in Vietnamese". In L. MacLeod,
G. Larson, and D. Brentani (eds.), Proceedings of the 24th Annual Regional
Meeting of the Chicago Linguistic Society.
THOMPSON, Lawrence
C.
1987 A
Vietnamese reference
grammar (Previously published
as Mon-Khmer Studies XIII-XIV éd.). Honolulu, University of Hawaiii Press.
TRAVIS, Lisa 1991 "Derived Objects, Inner Aspect and the Structure of VP". Paper
delivered at
NELS,
University of Delaware.
VON
WRIGHT, George
1951
An essay in modal
logic.
Amsterdam, North Holland
ZANUTTINI, Raffaela
1993 Re-examining negative
clauses.
Ms., Georgetown University.