Linguistics Vn
-
Upload
mytienganh -
Category
Documents
-
view
219 -
download
0
Transcript of Linguistics Vn
-
8/13/2019 Linguistics Vn
1/81
Page 1
Abstract
In this paper, some data are presented from Vietnamese that provide significant
empirical support for the theoretical claims articulated in Klein (1998, in press): first,
that finiteness should be understood as a composite of tense and assertion, and that
assertion may be realized independently of tense marking; second, that the assertion
operator so realized has only partial scope over elements of the clause, so that fronted
elements may evade this scopal influence. Vietnamese is of special interest because it
expresses assertion quite independently of tense or aspect: it differs in this regard
from most Indo-European languages, as well as from other isolating East Asian
languages. The formal analysis of these data involves two further, quite unorthodox,
claims: that assertion is syntactically projected in a low functional projection
immediately above vP; here, the clear parallelisms between Vietnamese assertion
markers, and Englishdo
-support suggest that the latter is not a language-specificmechanism as usually supposed, but the reflex of a more universal rule. The second
claim developed here is that in Vietnamese the displacement of certain constituents is
explained by their requirement to come within alternatively, to evade the scope
of this assertion operator. That is, syntactic movement may be driven by
considerations other than purely formal feature-checking.
-
8/13/2019 Linguistics Vn
2/81
Page 2
Aspects of Vietnamese Clausal Structure: Separating
Tense from Assertion
1 Introduction*This paper has both descriptive and theoretical goals. Its descriptive purpose is to
present some theoretically relevant data from Vietnamese, a language that has
received scant attention in contemporary formal linguistics. In this context, the focus
is on certain formal features of Vietnamese, which distinguish it from some other
more familiar languages of the region. The theoretical focus is on the structural
expression of functional categories; specifically, on the expression of tense and
assertion. Since Vietnamese is almost completely devoid of inflectional morphology,
it should be an excellent candidate for the extremely spare approach to functional
categories proposed in Bare Phrase Structure and assumed in much subsequent work
in Minimalist Grammar (Chomsky 1995a, 1995b, 1998). I will show here that,
despite its morphological impoverishment, Vietnamese provides evidence for at least
two distinct functional categories, including Tense Phrase and Assertion Phrase.
The focus is principally on the syntactic expression of finiteness. Following the
conceptual lead of Klein (1998, in press), finiteness is construed as the merger of two
more basic notions: TENSEand ASSERTION. Vietnamese will be shown to represent
these notions separately, morphologically and syntactically; as such, it provides
particularly clear empirical confirmation of Kleins theoretical approach. In this
respect, this language contrasts not only with many Indo-European languages, where
such notions tend to be conflated, but also with many other South East Asian
languages, where one or other notion fails to be reliably structurally expressed.
-
8/13/2019 Linguistics Vn
3/81
Page 3
Theoretically, the paper relates and integrates insights from three distinct
approaches. First, as just mentioned, I adopt the leading conceptual-semantic idea of
Klein (1998, in press) concerning finiteness, namely, that it is decomposable into
more basic notions of temporality (tense and mood) and assertion. Kleins position is
articulated in some more detail directly below. Second, regarding the syntactic
implementation of this idea, I incorporate mechanisms originally due to Chomsky
(1957), intended to handle the related phenomena of emphasis, negation and do-
support in a transformational description of English. While some of these ideas have
been revised and extended in more recent versions of generative theory up to
Minimalism, others, including the treatment of contrastive intonation, have fallen into
neglect. In this paper, the Vietnamese data serve to resurrect two aspects of that
earlier work. The first of these is that the base position of assertion, and of modality
in general, is very low in the functional structure of the clause below Tense,
immediately above the lexical vP rather than in the extended CP or left periphery,
as is more commonly assumed.1 This claim is schematized in (1) below:
(1) TP4
4T AsrP
4Asr vP
Assertion/Modality = c
The second idea to be re-adopted from Chomsky (1957) is that English do-support
is a reflex of a core transformational operation, rather than being a peripheral
language-specific strategy to support stray tense and agreement morphology. (Again,
I take the latter to represent more current generative assumptions: see, for example,
Chomsky & Lasnik 1977; Pollock 1989; Chomsky 1989; Bobaljik 1995).
-
8/13/2019 Linguistics Vn
4/81
Page 4
Finally, the paper seeks to integrate a multifunctional approach to the varying
interpretations of functional elements in different positions.2 Following Lefebvre &
Massam (1988) and Travis, Bobaljik & Lefebvre (1998), I construe a multifunctional
functional category (MFC) as:
one that is inherently underspecified with the unspecified properties of the host
head...[where]...syntax can provide additional information not available in the
lexical entry of the item. The lexical entry encode[s] the INTERSECTIONof the
uses of the item...[d]ifferent senses [of a multifunctional item] follow from the
different head positions in which it occurs (Travis, Bobaljik & Lefebvre (1998:3,
emphasis in the original).
On this view, the syntactic position of a functional category influences in some
cases, fully determines the interpretation of that element. As an example of this
type of multifunctionality, consider the alternations in (2), involving the Vietnamese
element ai:
(2) a. Anh quen ai ? = IN-SITU WHPRN know ai
Who(m) do you know?
b. Toi khong quen ai. = NPI
I NEG know AI
I dont know anyone.
-
8/13/2019 Linguistics Vn
5/81
Page 5
c. Ai co ay cung quen. =UNIV. QUANTIFIER
ai PRN DEM also know
She knows everybody.
In (2a), aifunctions as a WH-element; the sentence is obligatorily interpreted as an
interrogative. Sentence (2b), which is otherwise identical to (2a), contains the
sentential negation element khng: here, aiis necessarily interpreted as a negative
polarity item (!English anyone). Where aiis fronted from its base-position and in
construction with cung, in (2c), it is interpreted as a universal quantifier.
This particular set of alternations finds direct parallels in varieties of Chinese that
have previously been analyzed by other researchers, including Li, (1992), Tsai (1994)
and Cheng (1997). I return to these analyses presently. Note that the principal
contribution of the present paper is not to demonstrate that the interpretation of MFCs
depends on their position something that Li (1992) and others have already shown
well for Chinese but rather to use these varying interpretations to pinpoint more
precisely the (low) position of the operators that have scope over these elements.
Thanks to the presence and distribution of the assertion marker cin Vietnamese, this
is a much more tractable task than in Chinese, and it is here that the main contribution
lies.
A less familiar alternation is illustrated in (3); to my knowledge, this is peculiar to
Vietnamese. This concerns the WH-adverbial element bao gi (when), whose
temporal interpretation varies according to its syntactic distribution: in its canonical,
post-verbal position (3a), bao gi is assigned a past-tense interpretation; in pre-
-
8/13/2019 Linguistics Vn
6/81
Page 6
verbal. or sentence-initial position, bao giis interpreted as referring to the present or
future (3b):
(3) a. Co ay i My bao gi?PRN DEM go America when
When did she leave for America?
b. Bao gi co ay i My?
when PRN DEM go America
When will she go to America?
It will be claimed that as with the alternations in (2), those in (3) should be
explained in terms of syntactic scope: the varying interpretations of bao gi are
determined by being within or outside the scope of some sentential operator.
While this general idea has certainly been advanced before, there is some originality
in the idea that the relevant operator in this case is an assertion feature projected
comparatively low in the phrase-structure, and perhaps less originally, but more
controversially from a formal perspective in the idea that MFCs may move for
negative reasons in order to evade scope. That is, MFCs may move to avoid being
assigned a particular interpretation rather than purely for feature-checking reasons.
A variant of this notion of scope evasion constitutes one of Kleins arguments for
structurally represented assertion, to which we turn directly.3
The rest of the paper is structured as follows. In section 2, I briefly outline Kleins
theoretical proposals for the separation of tense and assertion, as well as for the
(abstract) structural representation of assertion in different clause-types. Following
this, in section 3, I set out the varying distributions of the Vietnamese morpheme c,
-
8/13/2019 Linguistics Vn
7/81
Page 7
and provide an analysis of this element as a relatively pure instantiation of assertion
divorced from tense. In this presentation, explicit parallels will be drawn between c
and English do-support. Having set out the main analysis in section 3, section 4
revisits Chomskys original analysis: I also provide some additional supporting
evidence for low modality, this time from the Vietnamese imperative system. In
section 5, I explore some further consequences of the main analysis, and develop a
hypothesis about scope evasion that directly accounts for the bao gi alternation (just
introduced). Section 6 concludes the paper.
2 Finiteness: Separating Tense and Assertion in IE Languages (Klein 1998, in press)Klein (1998, in press) henceforth K. presents a number of arguments
motivating the decomposition of finiteness into two meaning components, separating
tense under a particular construal of this notion from what is termed assertion
(also, assertion-markedness).
Four of the arguments presented by K. are especially relevant to the present
discussion, having obvious exponents in Vietnamese: three of these, discussed in
sections 2.1-2.3 below, motivate a purely interpretive splitting of the notion of
finiteness into tense and assertion; the fourth argument, in 2.5, supports the claim that
assertion is not only an autonomous semantic notion (distinct from tense) but that it is
also structurally represented in a sentence-medial position. (In section 2.4, I briefly
sketch K.s own analysis of how finiteness might be syntactically represented).
2.1. The contrastive intonation argumentThe first argument concerns contrastive intonation on English verb forms. K.
points out an important distinction between auxiliary verbs including copular verbs
and do on the one hand, vs. lexical verbs on the other: whereas contrastive
-
8/13/2019 Linguistics Vn
8/81
Page 8
intonation on finite auxiliaries can function either to assert or deny the VALIDITYof a
prior claim or to contrast the TIME about which the assertion is made, the former
function is not available to lexical verbs, which can contrast only the relative time
value (or the semantic content of the lexical verb).4 This contrast is illustrated in (4)
and (5) below (K.s examples (1) and (2)):
(4) a. The book was on the table.b. The book is on the table No, the book WASon the table.
c. The book was not on the table. No, thats wrong, the book WASon the table.
(5) a. John LOVEDMary.b. John LOVEDMary, but he doesnt love her any longer.
c. John LOVEDMary, but he didnt ADOREher.
The availability of this validity function in (4c) provides prima facie evidence for
Kleins claim that:
the finite element wascarries at least two distinct meaning components: 1.
the tense component: it marks past, in contrast to present or future; 2. it
marks the claim the fact that the situation described by the utterance
indeed obtains, in contrast to the opposite claim. (K. in press,pn-pending).
K. goes on to observe that it is possible to mark the same function in sentences
containing lexical verbs, but ONLYby supplying do-support, as in (6):
-
8/13/2019 Linguistics Vn
9/81
Page 9
(6) The idea that he didnt love her is plainly wrong: John DIDlove Mary.
The contrast between the interpretations available to stressed lexical verbs vs. those
available to stressed do indicates that do-support really DOEShave a function other
than as a host for tense and agreement inflection (as usually supposed); in other
words, that do is neither a dummy host in morpho-syntactic terms, nor a true
pleonastic, semantically. As will be shown in the next main section, Vietnamese c
signals this pure validity function even less ambiguously.
2.2 (Non-relational) Marking of Topic Time
K.s second argument for tense being distinct from assertion concerns the
interpretation of English sentences containing simple past forms (as opposed to those
involving the present perfect). In question-answer pairs such as those in (7), K.
claims that the simple past simply marks the time-span for which an assertion is
made, rather than the relationship, contrastive or otherwise, between the utterance
time and the event time (which is the standard RELATIONALconstrual of tense). Thus,
(7ii), for example, is well-formed as an answer, even though the dog is (presumably)
still dead at the utterance time:
(7) Why didnt Mika/the dog come to the park this morning?i. Mika was sick.
ii. The dog was dead.
Regarding such sentences, K. writes:
What is meant by the simple past is the fact that at some particular time span
in the past, Mika was sick, and the dog was dead. An assertion is made only
-
8/13/2019 Linguistics Vn
10/81
Page 10
about this time in the past, and it is simply left open whether the state
obtaining then also obtains later or earlier. IT IS NOT THE TRUTH OF HIS BEING
SICK OR DEAD THAT IS CRUCIAL BUT THE FACT WHETHER (sic) SOMETHING IS
ASSERTED ABOUT SOME TIME. Such a time span for which an assertion is
made I will call Topic Time (TT) and it is the function of tense to mark
whether TT precedes, contains or follows the time of utterance. The time of
the situation itself may precede, contain or follow TT. I think it is this
relation between TT and the time of the situation which is traditionally called
Aspect (Klein, in press,pn-pending, emphasis in original).
That is to say, the English simple past signals two functions: it indicates tense
the temporal relation between Topic Time and Utterance Time and it
simultaneously marks the (non-relational) assertion that some state-of-affairs or event
obtains at the Topic Time in question. On this construal, the wellformedness of
sentence (7.ii) follows from the fact that the use of the simple past in contrast to
the more aspectual present perfect involves no claim whatsoever about the
relationship between Topic Time and the time of the situation; in other words, no
claim is made about the dogs health outside of the Topic Time.
Although the CONCEPTUAL distinction here is relatively clear, the obviousEMPIRICALproblem is that, in the English past tense forms, these two functions are
morphologically syncretic. As we shall see directly especially in the discussion of
responsive constructions Vietnamese c realizes the assertion function of
finiteness directly, unencumbered by any requirement to mark tense.
-
8/13/2019 Linguistics Vn
11/81
Page 11
2.3 Tense without assertion: non-declarative sentences
K. observes that non-declarative sentences, such as yes-no questions, imperatives,
and norm-creating sentences, can be tense-marked (in his sense) without expressing
an assertion.5 Of such sentences, K. remarks:
In all of these cases, there is a sentence baseit gives a description of what
should be made true (imperatives), is to be decided whether it is true (yes-no-
questions), or ought to hold (norm-creating sentences). There is also a
counterpart to the topic time. In the imperative, this topic time must be
after the utterance time. There is no difference for questions. Norm-creating
sentences also hold for the future; sometimes, they explicitly specify the
beginning time. So, the crucial difference seems to rely on the notion of
assertion (K. in press,pn-pending).
Now, if K. is correct in claiming that non-declaratives such as imperatives can betense-marked without expressing any assertion, then tense and assertion must perforce
be separable. However, what is perhaps most important in this discussion and
most relevant to Vietnamese is the idea that assertion as a structural notion is
intimately linked to illocutionary role; in other words, that assertion is simply one
formal value of a more general multi-valued structural operator. K. (1998) states this
idea quite explicitly:
It is plausible to assume that tense only marks that some arbitrary time span,
for which we keep the term TT, is placed somewhere on the time axis, and
that either ASS, OR , DEPENDING ON THE PARTICULAR ILLOCUTION, SOME
OTHER 'MODALITY MARKER assigns a special function to this time span. So,
-
8/13/2019 Linguistics Vn
12/81
Page 12
TT can be the time span for which a claim is made, but it can also be the time
span, at which some obligation is put into force (on in whichever way we
want to analyse the role of the imperative) (K. 1998, emphasis mine: NGD).
Thus, what is crucial here is not only the claim that tense and assertion are
separable components of finiteness, but also that ASS is only one possible value of a
more complex modality element, whose feature values determine illocutionary force.
Precisely these properties are clearly instantiated by Vietnamese c.
2.4 Kleins structural proposal
Thus far, the concern has been to motivate two semantic-conceptual ideas
concerning finite clauses: first, that finiteness FIN, in K.s terminology is
composed of two more fundamental elements, tense and assertion; second, that
assertion is only one value of a multi-valued semantic operator determining the
illocutionary force of an utterance. The question now is how these semantic notionsare structurally realized, and how they interact with other sentential material to yield
well-formed, interpretable utterances.
K. (1998, in press) is quite explicit that FIN is always syntactically realized in the
logical structure of the sentence (although the reader should be aware that Kleins
idea of structural realization is somewhat more abstract than the surface
representations to be presented in the remainder of this paper). For K., finite
sentences have, abstractly, a rigid tripartite structure. First, there is the TOPIC
COMPONENT. Minimally, this involves a Topic Time (and topic place), the time (and
place) to which the assertion, or whichever the illocutionary role is, is confined. The
topic component may also involve other constituents, including a TOPIC ENTITY,
typically realized by the grammatical subject. The second component of the tripartite
-
8/13/2019 Linguistics Vn
13/81
Page 13
structure is what K. terms the SENTENCE-BASE notated as INF* in K. (1998)
which conveys the content of the assertion: minimally, this consists of a non-finite
predicate and an appropriate filling of its arguments (K in press: pn). Finally, there
is the in IE languages composite finiteness element (FIN*) that relates the
sentence-base to the topic component. (The asterisk notation denotes the level of
representation of these elements: by hypothesis in all utterances, TOP, FIN and INF
will be represented at a level of logical structure corresponding to LF in Minimalist
terms (as FIN*, INF* and TOP*); however, all three components need not be
projected in the surface (PF) representation in every language.) K.s schematisation
of this tripartite structure is reproduced in (8) below:
(8) UTTERANCETOPIC COMPONENT FIN SENTENCE BASE
topic topic topic (topic Vs and arguments
time place world entity)
As we shall see directly, in Vietnamese as in some constructions in German
this tripartite structure is reflected directly on the surface, with the bonus that
Vietnamese overtly splits finiteness into its constituent parts (T and Asr).
2.5 On the partial scope of Finiteness
One potential consequence of this structural arrangement is that (the sub-
constituents of the finiteness operator) FIN* should only have partial scope over other
sentential material; specifically, they should have scope over material to their right,
but not to their left.6
-
8/13/2019 Linguistics Vn
14/81
-
8/13/2019 Linguistics Vn
15/81
Page 15
components is directly represented in Vietnamese phrase-structure, as diagrammed
in (11):
(11) Top TOPIC44
Top TP1 4th 4
T AsrP1 4se 4
ASR vP1 4c 4
vFOCUS
It should be noted in passing that the partial scope of assertion and of other
modality operators is directly predicted if assertion is projected low in the
structure, as schematized in (11); this interpretive effect would be much more
surprising if assertion were a property of CP, in which case no amount of leftward
movement would be sufficient to evade scope. Later, it will be shown that dependent
elements in Vietnamese need only be fronted to the immediate left of AsrP to evade
scope, providing further evidence for this comparatively low position.
3 Separating Tense and Assertion in VietnameseHaving set out a particular interpretation of Kleins proposals about finiteness, we
may now consider some Vietnamese data that support these.
For the purposes of this analysis, I make three other assumptions concerning
Vietnamese phrase-structure, as set out in (12) below. None of these is especially
controversial; all three are supported by rather obvious distributional evidence.
-
8/13/2019 Linguistics Vn
16/81
Page 16
(12)a. Subject arguments raise overtly to [Spec, TP]b. Lexical verbs do not raise overtly out of the maximal VP (vP).
c.
In matrix clauses, there is a distinguished initial position for topicalised
constituents (TopP); this projection is headed by the topic morpheme th.7
Given these assumptions, the data in the following sections can be shown to
provide clear evidence for the syntactic separation of tense and assertion.
3.1. Tense-marking
In order to show that tense and assertion are separated in Vietnamese, ideally BOTH
should be structurally represented; however, in more traditional treatments of
Vietnamese grammar, it is often denied that Vietnamese has tense at all. This is made
quite explicit in Nguyen c Dans assertion [T]rong tieng Viet khong co pham
tru th(There is no tense in Vietnamese.) (Nguyen c Dan 1998: 116).
Presumably, what is meant here is that tense-marking is almost always optional in
Vietnamese; this contrasts with its obligatory presence in independent clauses in IE
languages. Otherwise, the claim is plainly false, since there exist morphemes whose
sole function is to carry temporal information of one kind or another, and which have
the distribution of tense or (grammatical) aspect morphemes in more familiar
languages. A representative subset of these items is listed in (13); the sentences in
(14) and (15) illustrate the distribution of these elements with respect to temporal and
manner adverbials, respectively, showing that they are distributed to the right of the
former and to the left of the latter:
-
8/13/2019 Linguistics Vn
17/81
Page 17
(13)a. se; sap, sap sa future; near futureb. a ; va, mi, va mi, mi va past; recent past
c. ang progressive
(14)a.*Anh Lai a hom qua giup toi.PRN Lai ANT yesterday help me
Anh Lai helped me yesterday. (00/10)8
b. ?Anh Lai hom qua a giup toi.
PRN Lai yesterday ANT help me
Anh Lai helped me yesterday. (07/10)
(15)a. Toi se can than viet la th nay.I FUT carefully write letter DEM
I will write the letter carefully. (26/27)
b. *Toi can than se viet la-th nay.
I carefully FUT write letter DEM
I will write the letter carefully. (05/27)
The examples in (14) and (15) show exactly the distributions that would be
expected if these tense morphemes occupy the T node, with lexical verbs remaining in
vP in accordance with (12b) above and if adverbial expressions are immediately
left-adjoined to the syntactic projection that they modify.9
-
8/13/2019 Linguistics Vn
18/81
Page 18
While the principal focus of this paper is on assertion for a more complete
discussion of tense and aspect, see Duffield (in preparation) it is nevertheless
worth pointing out that these temporal elements do not have exactly the same values
as their English equivalents. In particular, the morpheme a, which is usually glossed
as past tense, appears to denote a more aspectual relationship between the Topic
Time and a preceding situation time, either of which may be independent of the
utterance time. Thus for example, in contrast to the English past tense, amay be
used in future-perfect contexts, such as that in (16):
(16)a. Ngay mai, neu anh en luc bay gi sang th toi a i hoc roi.tomorrow, if PRN go time seven hours a.m. TOP I ANT go.study already
If you come at seven a.m. tomorrow, I will have already gone to study.
b. en cuoi nam nay, toi a ra trng.
until end year DEM, PRN ANT graduate
I shall have graduated by the end of this year.
Such complications notwithstanding, it remains the case that these morphemes
have the same syntactic distribution as tense morphemes in IE languages, but are
apparently without the assertion component associated with IE finite auxiliaries. It isto the expression of this component that we now turn.
3.2. Distribution and function of c: parallels with English do-support
Setting aside its function as a morpho-syntactic host for tense and agreement
something that is clearly irrelevant for Vietnamese English do-support is observed
in five main contexts (or three, if negation and emphasis, and responsive and ellipsis
-
8/13/2019 Linguistics Vn
19/81
Page 19
constructions, respectively, are interpreted as different reflexes of common underlying
constructions):
(17) a.i dooccurs (in construction with lexical verbs) in negative clauses.a.ii dooccurs (in construction with lexical verbs) in emphatic clauses.
b.i dooccurs (in construction with lexical verbs in responsives.
b.ii. dooccurs (in construction with lexical verbs) in ellipsis constructions.
c. dooccurs (in construction with lexical verbs) in direct questions.
As we shall see directly, this way of describing English do-support characterizes
almost perfectly the distribution of Vietnamese c, the only difference being the
(phonetic) optionality of the latter in most contexts. When it ISrealized, however,
Vietnamese cfunctions just like English do-support, except that obviously it doesnt
support any bound morphology. Each of these contexts is presented in turn.
3.2.1. Negative Environments
The most common marker of lexical or sentential negation in Vietnamese is khng
(except when it appears in final position, where it indicates a question); other negative
elements with near-parallel distribution and function include (more literary) chang,
cha(no, not) and cha(not yet). As the examples in (17) illustrate, in contexts of
CONSTITUENT NEGATIONkhngmust immediately precede whatever constituent it
modifies in order to be interpreted as a negator of that constituent. Compare
especially (18c) and (18d), where the reversed scope of khngwith respect to the
adverbial hon tonradically alters the interpretation of the sentence:
-
8/13/2019 Linguistics Vn
20/81
Page 20
(18)a. Mon an nay khngngon.meal eat DEM NEG tasty
That dish is not tasty.
b. Mon an nay ngon khng?
meal eat DEM tasty NEG
Is that dish tasty? *That dish is not tasty.
c. Co ay hoan toan khong tan thanh.
PRN DEM completely NEG approve
She totally disapproved.
d. Co ay khong hoan toan tan thanh.
PRN DEM NEG completely approve
She didnt totally approve.
The close relationship between the syntactic position of khng and its scopal
interpretation is brought out clearly by the interpretation of indefinite subjects in (19):
where immediately preceded by khng, the subject argument ai is necessarily
interpreted as a polarity item (19a), otherwise it is treated as a WH-element (19b): cf.
(2) above, see also Li (1992), Tsai (1994).10
(19)a. Khong ai thay anh.NEG AI see PRN
No-one saw you.
-
8/13/2019 Linguistics Vn
21/81
Page 21
b. Ai khong thay anh?
AI NEG see PRN
Who didnt see you?
Given this rigid distribution for CONSTITUENTnegation, it is interesting to see
where khngappears in cases of SENTENTIALnegation. As the examples in (20) make
clear, khngobligatorily occurs immediately to the left of c: other positions either
to the left of the tense element (20c), or to the right of cbefore or after the lexical
verb (20d, e) are wholly ungrammatical:11
(20)a. Hom qua anh ay a khong co en nha ch.yesterday PRN DEM ANT NEG CO go house PRN
He didnt go to your house yesterday.
b. Hom qua anh ay a co en nha ch khong?
yesterday PRN DEM ANT CO go house PRN NEG
Did he go to your house yesterday?
*He didnt go to your house yesterday?
c. *Hom qua anh ay khong a co en nha ch.
d. *Hom qua anh ay a co khong en nha ch.
e. *Hom qua anh ay a co en khong nha ch.
This order is so fixed that, in colloquial spoken Vietnamese, khngand care
often treated as a fused element khng-ceven though cis always optional in terms
of phonetic realization, strictly speaking (Pham Hoa, personal communication).
-
8/13/2019 Linguistics Vn
22/81
Page 22
Now, given the parsimonious assumption that sentential negation is covered by the
same generalization as applies to constituent negation, the distribution of khng in
(20) provides prima facie evidence for one of the central claims of this paper, namely,
that the assertion phrase headed by c, rather than the higher tense projection, is the
head of the clause. This is diagrammed in (21); compare the structure in (11) above:12
(21) 5T AsrP
1 4(se/a) khng 4
Asr vP
1 4[+NEG] 41 vco
3.2.2. Emphatic assertion contexts
In addition to negative contexts, c may optionally appear in any declarative
clause. In this respect, it mirrors the use of do-support in earlier varieties of Standard
English (including Early Modern English) and in some current non-standard varieties:
see Ellegrd (1953); Visser (1963-1973); Warner (1993); also Schtze (2003). More
typically, however, cis realized in emphatic contexts (in conjunction with khng, in
negative emphatic contexts, and often also in conjunction with other emphatic
particles such as m). This emphatic function is exemplified by the sentences in (22):
(22)a. Anh (khong) co mua sach!PRN NEG CO buy book
He DID(did NOT) buy the book!
-
8/13/2019 Linguistics Vn
23/81
Page 23
b. Ch (khong) co i lam!
PRN NEG CO go work
She IS(is NOT) going to work!
c. Toi co gi th mi ong ay ma! (N:153)
I CO send letter invite PRN DEM EMPH
I DIDsend him an invitation.
With respect to emphasis, c or rather, the syntactic projection hosting c
shows two further parallels with do-support. The first of these interactions involves
the multifunctional WH-element au. In post-verbal argument positions, the latter
element is obligatorily interpreted as a locative expression (with verbs that select
locative arguments). Elsewhere, however, au functions as an emphatic negative
particle. In this latter function, auhas a quite different distribution, appearing either
preceding the assertion marker c, or in sentence-final position, following c or
khng. Nguyen nh-Hoa(1990: 59) provides the examples in (23) below.
Note that the order in (23a) and (23d) only permits the emphatic interpretation for
au: in its interrogative function, aumust always appear in a post-verbal argument
position. Notice also that emphatic au is always negative, even when khngis
unexpressed, as in (23a) and (23b):13
(23)a. Ong au co en!PRN DAU CO come
He did NOTshow up! (*Where did he show up?)
-
8/13/2019 Linguistics Vn
24/81
Page 24
b. Ong co en au!
PRN CO come DAU
He did NOTshow up!14
c. Ong khong en nha au!
PRN NEG come house DAU
He is NOTgoing to your house! (*Where did he come to the house?)
d. Ba au co phai la ngi Hanh-thien!
PRN DAU CO right COP person H.-T.
She is NOTa native of Hanh Thien, I tell you!
The phonetic optionality of khngin contexts that are nevertheless obligatorily
interpreted as negative suggests that emphatic aumodifies, and is in construction
with, a syntactic projection with negative features, rather than with a particular lexical
item. This claim is diagrammed in (24):
(24) 5 T AsrP
1 4(se/a) au 4
khng Asr vP
1 4[+NEG][+EMPH]
[-WH]1(co)
Such an analysis is supported by the fact that preverbal au and (sentential
negation) khng are in complementary distribution, illustrated in (25) below. This
indicates EITHERthat they compete for the specifier position of this projection, as
-
8/13/2019 Linguistics Vn
25/81
Page 25
shown in (24), OR that they are both head-adjoining adjuncts in the sense of Travis
(1988), attaching directly to the head of AsrP:
(25)a. Ong (*khong) au co en!PRN NEG DAU CO come
He did notshow up!
b. Ong (*au) khong co en!
PRN DAU NEG CO come
He did notshow up!
The analysis in (25) is further supported by some other facts about emphatic stress
in Vietnamese. Nguyen nh-Hoa (1990: 60) reports that, especially in womens
speech, pre-verbal WH-elements have a negative function if they are heavily stressed.
This is shown by the minimal contrast in (26):15
(26)a. AInoi.ai speak
No-one spoke.
b. Ai noi?
ai speak
Who spoke?
On closer examination, however, an alternative generalization here is that, for
negative emphasis, stress is placed on the first overt element to the left of the verb,
-
8/13/2019 Linguistics Vn
26/81
Page 26
since either khngor c, rather than ai, attracts stress whenever either of these other
elements are realized:
(27)a. Ai Cnoi!AI CO speak
No-one spoke/Who did speak?
b. Ai KHNGnoi!
ai NEG speak
No-one spoke/Who did not speak?
c. ?AIkhngnoi?
d. ?AI cnoi?
This distributional pattern finds immediate parallel with emphatic contexts
involving doin English, where once again it is the element to the immediate left of
the lexical verb that attracts emphatic stress:
(28)a. She DIDsay that!b. She did NOTsay that!
c. She did SOsay that! (*She DIDso say that!)
d. She did TOOsay that! (*She DIDtoo say that!)
In both English and Vietnamese, the distribution of emphatic stress can be
captured by the structural analysis in (24), where stress falls on the first phonetically
realized element to the left of the head of AsrP, if the head itself is not realized. On
this analysis, the only syntactic difference between English and Vietnamese would be
-
8/13/2019 Linguistics Vn
27/81
Page 27
that doraises to T to pick up Tense and Agreement features, whereas cremains in
situ. This difference is diagrammed in (29):
(29) 5 T AsrP1 4
VN. a khng 4Eng. didj not Asr vP
1 4[+NEG]
[+EMPH][-WH]1
VN. coEng. tj
As discussed in section 4.1 below, this analysis is essentially a minor revision of
Chomskys original (1957) proposal for handling emphatic assertion.16
3.2.3. Default Past Tense reading
The claim that cis a pure assertion marker immediately accounts for an additional
property of c mentioned in more several traditional sources namely, what may
be termed its default past tense reading. Ngo Nh Bnh, for example, notes
explicitly that c tends to be used to emphasize the fact that an action definitely
takes/took place... (Ngo Nh Bnh 1999: 176). Since one is generally only sure
about asserting past events, and since the optional tense markers seand aare nearly
always omitted in the spoken language, it is natural that the default interpretation of
assertion cis as a past tense marker since both pick out Topic Time in different
ways.17
-
8/13/2019 Linguistics Vn
28/81
Page 28
This is further evidenced by the fact that Vietnamese speakers often translate
sentences containing cusing English past tenses (unless another temporal reading is
forced by the context, or by the presence of an overt future morpheme). This default
past property of cforms part of the explanation of the instances of scope evasion
presented in section five below.
Notice that these two properties of c its association with specific Topic Times
and its default past tense interpretation highlight the special sense of assertion
proposed in K. (1998): assertion does not simply mean the speakers support for a
particular proposition otherwise ccould as well be used with generic statements
rather, it involves the linking of propositional material (K.s INF*) to a particular
Topic Time.
3.2.4. Vietnamese responsive constructions
The next parallelism between Vietnamese cand English doinvolves constraints
on responsive constructions. In common with many other languages, Vietnamese has
no specific words directly corresponding to English Yes and No. Rather,
Vietnamese speakers give assent to a prior Yes-No-question either by responding with
cor by repeating the predicate; speakers show dissent either by responding with
khng(c) (or again, with khng plus the repeated predicate). This is illustrated in
(30):
-
8/13/2019 Linguistics Vn
29/81
Page 29
(30) Anh (co) mua sach khong?PRN (CO) buy book NEG
Did you buy the book?
a. Co Yes, I did!
b. Khong (co)! No, I didnt!
What is of interest here are the constraints determining whether speakers employ
khng(c) or choose to repeat the predicate: it is these that would appear to expose c
as a pure assertion marker in K.s sense. Consider first the typical question-answer
pairs in (31a) and (31b), involving an active and a stative predicate, respectively.
(31)a. Ch Phng co mua nha khong?Phuong CO buy house NEG
'Did Phuong buy a house?'
A: Da, co! (10/10) A:??Da, mua! (02/10)
b. Co ay (co) ep khong?
PRN DEM (CO) beautiful NEG
'Is she beautiful?'
A:??Da, co! (03/10) A: Da, ep! (08/10)
In principle, a speaker can assent to the question involving the active predicate in
(31a) either by repeating the predicate or by using c; however, the latter is clearly
preferred where the question is interpreted as asking about a specific past event (see
below). In (31b), by contrast, the judgment is reversed: in response to the question
-
8/13/2019 Linguistics Vn
30/81
Page 30
about an individual-level property is she beautiful? the preferred response is ep
(beautiful) rather than c.
(32)a. Ch (co) ban khong?PRN CO busy NEG
Are you busy?
A: ?Da, co! (16/20) A: Da, ban! (18/20)
b. Ong (co) nh toi khong?
PRN DEM remember I NEG
Do you remember me?
A: ?Da co! A: Nh ch!
c. Em (co) hieu khong?
PRN CO understand NEG
Do you understand?
A: ?Da co! (14/21) A: Da, hieu! (21/21)
d. Anh am co li khong?
PRN Dam CO lazy khong
Is Dam lazy?
A: ??Da, co! (03/10) A: Da, li! (09/10)
As is further illustrated by the sentences in (32), this contrast is representative of a
more general split between typical uses of active vs.stative predicates, with active
predicates preferring cin the responsive clause, and statives preferring repetition of
the predicate in question. (These examples also indicate that it is more common to
-
8/13/2019 Linguistics Vn
31/81
Page 31
omit cin the question if a stative predicate is involved, a point we will return to
directly).18
However, non-typical uses such as those in (33)-(35) show that the correlation
between predicate-type and responsive is not necessary, and that the assertion
property of cis actually independent of predicate-type. So, in the generic questions
in (33), which make no reference to specific events, the preferred responsive is
predicate repetition (even though an active predicate is involved). This pattern
minimally contrasts with the sentences in (34): there, a specific event is referred to,
and cis once again strongly preferred over predicate repetition.
(33)a. Ngi Nhat (co) an ca khong?person Japan CO eat fish NEG
Do Japanese people eat fish?
A: ?Da, co. A: Da, an.
b. Anh (co) sa may.anh khong?
PRN CO repair camera NEG
Do you repair cameras?
A: ?Da, co! A: Da, sa.
(34)a. Tuan roi trong nha hang cua anh ngi Nhat o co an ca khong?last week in restaurant POSS PRN people Japanese CO eat fish NEG
Last week in your restaurant did Japanese people eat fish?
A: Da, co. A: *Da, an.
-
8/13/2019 Linguistics Vn
32/81
Page 32
b. Hom qua khi may.anh cua chung toi b h, anh co sa khong?Yesterday when camera POSS PL I PASS break, PRN CO fix NEG
Yesterday when our cameras broke did you fix the cameras?
A: Da, co. A: *Da, sa.
c. Bay gi anh co hoc khong?
now PRN CO study NEG
Are you studying now?
A: Da, co ! (20/22) A:??Da, hoc! (09/22)
Conversely in (35), typically stative predicates such as vui ve(cheerful) and t
tin(confident), which normally require repetition in the responsive clause, prefer c
as a response when embedded under a raising predicate such as trng (appear) that
refers to a specific time.
(35)a. Ho co trong vui ve khong?PRN CO appear cheerful khong
Did they appear cheerful?
A: Da, co. (10/16) A:??Da, vui ve (07/16)
-
8/13/2019 Linguistics Vn
33/81
Page 33
b. Ho co trong t tin khong?
PRN CO seem confident khong
Did they seem confident?
A: Da, co. A;?Da, t tin
To capture this set of contrasts across the two predicate types, the best
generalization seems to be that cis used to associate a predicate with a Topic Time
where there is one: that is to say, it is a pure assertion marker in K.s sense. Wherethere is no relevant Topic Time, as in typical stative clauses and with active predicates
in generic contexts, cis dispreferred.19
3.2.5. Ellipsis Contexts
Just as cmirrors the assertion component of English doin responsives, so it also
functions like doin ellipsis contexts more generally. This is of course to be expected
if responsives are simply a special case of VP-ellipsis; nevertheless, it is still of
interest given that VP-ellipsis is often treated as a language-particular property
peculiar to English do. Consider the examples in (36), showing ellipsis of the verb in
affirmative and negative contexts, respectively.
(36)a. Anh khong co mua xe, nhng ho co.PRN NEG CO buy car, but PRN co
He didnt buy a car, but they did.
b. Cau Charles co mua mu mi, nhng toi khong.
PRN Charles CO buy hat new, but I khong
Charles bought a new hat, but I didnt.
-
8/13/2019 Linguistics Vn
34/81
Page 34
c. Chng ti c lm bi nhng ho khng.
PL I CO do homework but PRN NEG.
We did our homework, but they didn't.'
The main point to observe here if the analysis proposed in (29) is correct is
that English and Vietnamese differ with respect to the strong head licensing VP-
ellipsis (in the sense of Lobeck (1995): in English, it is Tense or strong Negation,
whereas in Vietnamese, only Assertion licenses ellipsis. This difference is
diagrammed in (37), where Neg is construed as one value of AsrP:
English ellipsis head (affirmative contexts)
(37) 5 T AsrP
1 4 VN ellipsis headVN. a khng 4
Eng. didj not Asr vP1 4
[+NEG][+EMPH]
[-WH]1VN. coEng. tj
That Vietnamese T is not strong enough to license ellipsis is evidenced by the
ungrammaticality of the examples in (38); here, neither anor secan support ellipsis:
-
8/13/2019 Linguistics Vn
35/81
Page 35
(38)a. *Anh a khong mua xe, nhng ho sePRN ANT khong buy car, but PRN FUT
He didnt buy a car, but they will.
b. *Cau Charles a khong mua mu mi, nhng toi a.
PRN Charles ANT khong buy hat new, but I ANT
Charles didnt buy a new hat, but I did.
It seems reasonable to think that this difference is related to the claim, expressed in
(29), that the morpheme carrying assertion moves to T in English, but remains in situ
in Vietnamese: the common generalization would then be that VP-ellipsis is licensed
by whatever head hosts assertion features at Spell-Out.20
3.2.6. Questionformation
The last obvious parallelism between English do-support and Vietnamese c is
observed in Yes-No questions. As has already been shown, Vietnamese Y-N
questions are formed using cplus the negation element khng in sentence-final
position. As in the other contexts, cneed not always be overtly realized; I assume,
however, that the corresponding syntactic projection is always expressed.21
(39)a. Anh (co) mua sach khong?PRN CO buy book khong
Did you buy the book?
b. Chieu nay ch (co) i lam khong?
evening DEM PRN CO go work khong
Are you going to work this evening?
-
8/13/2019 Linguistics Vn
36/81
Page 36
c. Anh ay co phai la ngi viet khong?
PRN DEM CO right COP person viet khong
Is that person Vietnamese?
In analysing these interrogative constructions, two related questions arise: namely,
the proper analysis of cand that of sentence-final khng. Whether the presence and
position of c in Vietnamese Yes-No questions is surprising or predictable depends
on one's basic assumptions about the syntax of question formation more generally. It
is surprising if one adopts the (currently standard) assumption that WH-features are
an inherent property of the highest functional projection of the clause, namely CP.
On this view, in English question-formation the movement of dofrom T to C to pick
up or check features is accidental, and essentially unrelated to the base-position of
wh-features. Such an assumption is supported by languages such as Modern Irish, as
exemplified in (40), where the presence of an overt +WH morpheme already in Cappears to preclude any similar T-C movement. If, universally, +WH features are an
inherent property of CP, one would expect Vietnamese to show a morphological
reflex of such features in a much higherposition than that occupied by c.
(40)a. A-r ith s an t-ll?Q-PAST eat PRN DET apple
Did he eat the apple?
b. An bhfuil Cit anseo?
Q BE C. here
'Is Cit here?'
-
8/13/2019 Linguistics Vn
37/81
Page 37
Conversely, if one construes assertion as one of a broader set of modality features
as K. and others have proposed (see above) then the occurrence of cin this
medial position in Y-N-questions is completely natural. Notice that since Chomsky's
(1957) proposal, a series of generative analyses have treated WH as an underlying
INFL feature, rather than an inherent property of C. Such a view is expressed in
work by Rizzi (1990, 1996), Haegeman & Zanuttini (1991); see also Haegeman
(1995), and especially Laka (1990).
Here, I develop a revised version of Rizzis (1996) proposal that [wh] features are
base-generated in INFL and only subsequently moved to C; see Noonan (1989) for a
related idea, see also Aoun & Li (1993). For Rizzi, morpho-syntax provides the
motivation for this claim: referring to work by Chung (1982), Clements (1984),
Georgopoulos (1985) and Hak (1990), among others, Rizzi notes that several
languages exhibit special verbal morphology for interrogatives. By locating +WH
features in the main inflection (Rizzi 1996: 66), Rizzi accounts for the distribution
of this morphology and provides formal motivation for I-to-C movement, as well as
for its restriction to root clauses (by hypothesis, lexically-selected C nodes have
inherent WH features, and thus do not need to inherit them through movement).
Rizzis (1996) analysis comes close to the same structural association between
assertion and modality (more broadly construed) as the one advanced here. He
writes:
It is natural to assume that such a position [for +wh features] can be the main
inflection (or one of the main inflectional heads, if some version of the Split
Infl hypothesis is adopted, as in Pollock (1989)), THE HEAD THAT ALSO
CONTAINS THE INDEPENDENT TENSE SPECIFICATION OF THE WHOLE SENTENCE.
-
8/13/2019 Linguistics Vn
38/81
Page 38
I would like to propose that among the other autonomously licensed
specifications, the main inflection can also be specified as [+WH] (Rizzi
1996: 66, emphasis mine: NGD)
While endorsing Rizzis general proposal that WH features are generated in a
sentence-medial position, I suggest that the more specific claim namely, that the
WH feature is associated with the TENSE projection is incorrect. Instead, the
Vietnamese facts, where features of assertion, emphasis and modality are
systematically dissociated from tense features, all suggest an analysis that places WH
features one projection LOWER(in AsrP):
(41) 5 T AsrP
44Asr vP
1 4[NEG][EMPH][WH]1
VN. coEng. tj
That modality features, including WH features, are dissociated from Tense is
generally obscured in less isolating languages by the fact of verb-movement; in
Vietnamese, by contrast, the two always remain separate.
If this is the correct analysis of the position of cin Vietnamese interrogatives, the
remaining issue is the proper treatment of sentence-final khng. Three analyses
suggest themselves immediately: first, that khng is a sentential tag, roughly
equivalent to English isnt it/she/etc; second, that khngis in the head of a rightward
-
8/13/2019 Linguistics Vn
39/81
Page 39
C; finally, that khngis right-adjoined to vP within the syntactic scope of +WH c.
These options are schematized in (42) below:
(42)a. [[ .........TP........], khng]
b. CP5TP C
4 1T khng
3T AsrP
c. 5 T AsrP
44Asr vP1 4
[-NEG][EMPH][+WH]
1VN. co
Eng. tj
There are several reasons to reject the former alternatives and to adopt the third
(42c). With respect to the tag analysis, two points should be noted. First, in
Vietnamese Y-N questions, khngfails to show the comma intonation pattern typical
of sentential tags; instead, it is fully prosodically integrated into the vP. More
significantly perhaps, Vietnamese has an alternative tag strategy for forming Y-N
questions: this latter strategy, illustrated in (43) below, employs exclusively positive
rather than negative tags, and displays exactly the expected intonational pattern.
Taken together, these facts cast doubt on the idea that bare sentence-final khng
should be analyzed as a sentence-tag.
-
8/13/2019 Linguistics Vn
40/81
Page 40
(43)a. Anh ay khong (co) en, (co) phai khong?PRN DEM NEG CO come, CO right NEG
He didnt come, did he?
The rightward complementizer analysis in (42b) is somewhat harder to reject,
based as it is on a treatment by Cheng (1997) of similar sentence-final particles in
Mandarin Chinese. However, whereas this analysis is plausible for Mandarin, other
facts distinguishing the two languages suggest that it is not directly transferable to
Vietnamese. The first is a narrow typological fact: in contrast to Mandarin, which
displays more mixed word-order properties, Vietnamese is otherwise strictly head-
initial; if khngwere analyzed as occupying a rightward C position, it would be the
only syntactic head in the language projected to the right of its phrasal complement.22
More importantly, and again in contrast to Mandarin, Vietnamese has what appear
to be LEFTWARDcomplementizers in other syntactic environments. This set includes
the -WH complementizer rangin (44a) and the +WH complementizer neuin (44b),
corresponding to English that and if, respectively. Assuming that these latter
elements are indeed complementizers in C, the case for sentence-final khngas a
rightward complementizer would appear to be further reduced.
(44)a. Anh a noi (rang) co ta khong tin *(rang)PRN ANT say (that) PRN khong believe (that)
He said that she didnt believe (him).
-
8/13/2019 Linguistics Vn
41/81
Page 41
b. (Neu) ngi ta mang hang-hoa en tan nha *(neu), ...
(if) people bring goods go visit house (if)
If people bring goods to your house...
The alternative analysis proposed here is that given in (43c): sentence-final khng
is right-adjoined to vP, and derives its interpretation in virtue of being in the syntactic
scope of the assertion head specified with +WH features (the head that is optionally
phonetically realized by c).
3.3. Interim summary
The Vietnamese data presented in the preceding sections illustrate the strong
formal and functional parallelism between Vietnamese c and English do-support: all
of the functions served by one are served by the other, with the obvious exception that
English dohosts inflection, whereas cdoes not. The functional parallelism supports
the claim, made earlier, that English do-support does much more than rescue stranded
affixes it is no mere dummy or pleonastic element and that it is a reflex of a
much more general syntactic means of expressing assertion and related modality
features.
As important as the functional parallelism, however, are the formal distributional
parallels: the position of c in Vietnamese clauses sentence-medial and
syntactically separated from tense provides evidence for an independently-
projected assertion (modality) head rather low in the functional structure of the clause.
The evidence laid out above, from the position of sentential negation and of emphatic
negative and assertive modifiers, as well as the from the ellipsis and responsive
systems, all points to this head (AsrP) as the head of the clause, at least in
-
8/13/2019 Linguistics Vn
42/81
Page 42
Vietnamese. The analysis diagrammed in (29) above shows that such an analysis may
also be applied to English, modulo the effects of verb movement.
Moreover, the additional interpretive facts that have been presented, in particular,
the default past interpretation of c , and the typical interactions between the
realization of cin responsives and the type of lexical predicate involved, suggest that
the features of this head are intimately bound up with the syntax of what is termed
event representation in other formal analyses. I suggest that it is not coincidental
that AsrP occupies precisely the hierarchical position at which other researchers have
postulated some type of Event Phrase: see Travis (1994, 2000), Ritter and Rosen
(2000), and other contributors to Tenny & Pustejovsky (2000), amongst others; see
also Davidson (1967); Parsons (1990); and Kratzer (1996). Note that the main
difference between many of these accounts and that proposed here is that in the
former the head of the phrase typically serves to express an event, and is thus
dependent on the lexical semantics of the predicate, whereas on the present analysis
cserves to assert a Topic Time, which need not in principle make reference to any
specific EVENT: it could as well refer to a STATEat a particular time, as in (35).
Both syntactically and semantically, then, Vietnamese cand English auxiliary do
show rather striking parallels in those contexts where they share the clause with
another predicate. In fact, the only obvious point of divergence between these two
functional elements aside from morpho-syntactic support is in their
interpretation as main predicates: as a lexical predicate, Vietnamese c is translated as
have, possessor be in existential contexts (see below) rather than do. Even
here, though, both elements belong to the class of light verbs; see Jespersen (1965);
cf. Kearns (1989) and Grimshaw (1990).
-
8/13/2019 Linguistics Vn
43/81
Page 43
The data presented thus far therefore provide clear empirical support for Kleins
proposal, that Assertion is syntactically expressed, and that it can, in principle, be
dissociated from Tense: in Vietnamese, these two components of finiteness ARE
clearly separated, morphologically and syntactically.
In addition to motivating a low functional projection for modality immediately
above vP, the Vietnamese facts also provide evidence for the other leading idea in
Klein (1988), namely, that constituents can be placed inside or outside the scope of
this assertion head, and that this has immediate consequences for clausal
interpretation. In section five below, this notion of scope is shown to account for the
distribution and interpretation of several independent multifunctional elements in
Vietnamese: the analyses presented there further reinforce the idea that the relevant
scopal operator is projected in this low position below TP, and the subject position
in [Spec, TP] rather than in the highest functional projection (CP).
Before turning to these scope facts, however, it is worth drawing attention to the
similarities between the structural analysis proposed here for c and Chomskys
(1957) original analysis of the constructions involving do-support in English. Given
the parallelisms that have been reviewed here, it should not be surprising that
Chomskys analysis of do-support fits Vietnamese so well. The point here is that
subsequent developments in generative theory have so tended to marginalize English
do-support that such parallels in analysis become much less obvious.
4 Motivating AsrThe general idea that assertion, negation and modality are closely associated
syntactically is of course a classic notion in generative grammar: it pre-dates the split
between generative vs. interpretive semantics, informing both traditions,23 and
-
8/13/2019 Linguistics Vn
44/81
Page 44
provides the conceptual basis for the GB distinction between lexical and functional
categories, a distinction that drove nearly fifteen years of syntactic research.24
4.1. Chomsky (1957)
In Syntactic Structures, negation, modality (question formation), and emphasis are
analyzed in terms of three transformational rules: Tnot, Tq, and TA, respectively; cf.
Katz & Postal (1964), Baker (1970). The first rule derives negative sentences by
introducing the lexical item NOTinto a position immediately to the right of the base
position of verbal affixes (anachronistically, following Infl). Given the extrinsicordering of these rules, this results in the non-adjacency of Af and the main verb,
which in turn triggers do-support (to carry the affixes). The question formation rule,
Tqalso operates on the affixal position, interchanging this with the first element of
the sentence, that is, the subject. As with Tnot, this has the effect of triggering do-
support, since the subject now intervenes between the affix and its verbal host.
Finally, to handle emphatic constructions, Chomsky:
set[s] up a morpheme A of contrastive stress to which the following
morphophonemic rule applies:
(45) ..V..+A --> "V, where " indicates extra heavy stress.TAimposes the same structural analysis of strings as does Tnotand adds A
to these strings in exactly the position where Tnotaddsnotor n't. (Chomsky
1957:65).
In Syntactic Structures, Chomsky was insistent that the three rules were intimately
related: indeed, it was the feeding/bleeding relationship among them that provided the
-
8/13/2019 Linguistics Vn
45/81
Page 45
motivation for a transformational approach. The quotation above relates the negation
and emphasis transformations (Tnot and TA); a few paragraphs earlier, Chomsky
explicitly relates Tnotand Tq:
The crucial fact about the question transformation Tqis that almost nothing
must be added to the grammar in order to describe it. Since both the
subdivision of the sentence that it imposes and the rule for appearance of do
were required independently for negation, we need only describe the
inversion effected by Tqin extending the grammar to account for yes-or-no
questions. Putting it differently, transformational analysis brings out the fact
that negatives and interrogatives have fundamentally the same structure
(Chomsky 1957: 64-65).
There are several points to notice immediately about this analysis. First, though it
provides a parallel treatment of negation and assertion, it is just that, parallel: given
a strong transformational rule system (including ordered lexical insertion rules), it is
essentially an accident, albeit a convenient one, that the two rules Tnotand TAhappen
to apply in the same structural context. Furthermore, the analysis treats as accidental
the fact that the rule targets a functional category, as opposed, say, to a main verb.
Second, unlike Tq, Tnot and TA do not in fact target any constituent directly;
instead, they insert new material into space to the right of the Infl constituent (C, in
Chomsky 1957). This property, taken in conjunction with the fact that it was
obviously not a meaning-preserving operation, led to the abandonment of Tnot in
later, more representational, versions of the theory, especially those that assumed
-
8/13/2019 Linguistics Vn
46/81
Page 46
some version of the Projection Principle (Chomsky (1981). Once negation came to be
viewed as necessarily a d-structure property, parallels with TA were lost: in
subsequent treatments of do-support, emphatic do is either ignored entirely or
relegated to a footnote; see, for example, Chomsky (1989). As a result, most of the
extensive research on the NegP hypothesis following Pollock (1989) disregards the
fact that the same position that modulates negation also modulates emphasis (or
assertion).
Laka (1990, 1994) and Haegeman (1995) are notable exceptions to this; however,
these authors place the relevant functional head labeled "and [Pol] by Laka and
Haegeman, respectively above, rather than below, Tense. Only much more
recently, in Cormack & Smith (2002), has it been proposed for English that assertion
and negation are united in a low functional head: the proposals made here for
Vietnamese appear to be quite consistent with Cormack and Smiths analysis for
English, except that the present proposal would also include WH-features in the samehead.25
Of the three rules, only Tqis still considered a transformational operation: it is of
course the direct precursor of T-C movement. What is interesting to observe,
however, is that in Syntactic Structures the other two transformations were logically
prior to Tq. As the quotation above makes clear, Tq is an extension to the
transformational system set up for Tnot and TA. Without such a system as
importantly, without the associated extrinsic rule ordering Tqalone is insufficient
to derive all the properties of yes-no questions. This is brought out forcefully by the
sentences in (46), which involve both emphasis and T-C movement (small capitals
indicates emphasis):
-
8/13/2019 Linguistics Vn
47/81
Page 47
(46) She DIDNT say that whales were fish.a. Well, what DIDshe say, then?
b. #Well, what did she SAY, then?
Ignoring (phrasal) wh-movement momentarily, it is relatively straightforward to
derive (44a) under the (1957) analysis. First, the morpheme A is inserted. This
interrupts the adjacency between Af and V, blocking Affix-hopping, and triggering
do-support. TA Chomskys Rule (45) then applies, yielding emphatic do.
Finally Tqis applied to the output of the other rules.
Two points bear mention. First, the fact that examples (46a) and (46b) have
distinct interpretations again suggests that the morpheme A of contrastive stress
does more than affect the morpho-phonemics of the element it attaches to,
correspondingly, that auxiliary dois more than simply a dummy host for stray affixes;
if this were not so, (46a) and (46b) should be synonymous.
Second, the derivation only works if the transformational rules are extrinsically
ordered, since alternative orders of application yield the wrong results. Crucially, and
perhaps unexpectedly, do-support must be triggered BETWEENthe insertion of the
morpheme A and the application of TA, and TA must precede Tq; otherwise, it is
impossible for A to be carried along by Tq
. This extrinsic ordering is contrary to
Chomskys own proposal, which is for do-support to apply last (1957:63).
Of course, this difficulty does not arise if Asr, negation and other modality features
occupy the same distinguished base projection, and if the head of this projection is the
initial target of English Aux-to-Comp; that is to say, if Aux-to-Comp is really Asr-T-
C, rather than T-C movement (in current terms), as proposed in (29) above.26The fact
that such an architecture is required to account for cand related markers of emphasis
-
8/13/2019 Linguistics Vn
48/81
Page 48
and assertion in Vietnamese would seem to provide independent support for such a
claim.
4.2 Additional Evidence for Low Modality: Vietnamese imperatives
Vietnamese imperative constructions provide one further piece of evidence in
favour of a low base position for modality features. As the examples in (47) from
Ngo Nh Bnh (1999)illustrate, imperative morphemes are predominantly placed to
the immediate left of the verb: where the subject is overtly realized, these morphemes
always appear like c to the right of the subject. Note especially the
ungrammaticality of the clause-initial placement that would be expected if
imperatives occupied the left-peripheral CP projection (47d-f):27
(47)a. Cac anh HAY oc bai nay.PL PRN IMP read lesson this
Read this text!
b. Anh C hoi.
PRN IMP ask
Go ahead. Ask.
c. (Anh) NG noi to!
PRN NEG.IMP talk loud
Dont speak loudly!
-
8/13/2019 Linguistics Vn
49/81
Page 49
d. *HAYcac anh oc bai nay.
e. *Canh hoi.
f. *NGanh noi to!
5 Scope and Scope EvasionThe data presented thus far have provided evidence in support of the analysis of
Vietnamese clause structure diagrammed in (11) repeated below for convenience
in which modality features, including features of assertion, negation, emphasis and
WH, are all associated with a sentence-medial, functional category projected below
TP. By hypothesis, it is the features in the head of this projection that determine the
interpretation of multifunctional elements in positions lower than this head:
multifunctional elements placed ABOVE AsrP are outside of the scope of these
features.
(11) Top TOPIC4
4Top TP1 4th 4
T AsrP1 4se 4
ASR vP1 4c 4
vFOCUS
If this analysis is correct, then multifunctional elements are expected to display
two types of exceptional behavior: first, MFCS that would normally appear above
AsrP in terms of argument structure specifically, subject arguments might be
displaced so as to come within the scope of the Asr head; conversely, MFCSnormally
-
8/13/2019 Linguistics Vn
50/81
Page 50
projected to the right of Asr might raise to evade scope. As the following data shows,
both of these predictions are borne out.
5.1. Multifunctional elements in subject position: moving INTOthe scope of assertion
The first context to consider is the clausal subject position, which falls outside of
the scope of Asr, and where multifunctional elements nevertheless appear to receive
varying interpretations. Some relevant cases are those in (48):
(48)a. Toi khong thay AI.I NEG see AI
I dont/didnt see anyone.
b. Khong ai thay anh.
NEG AI see PRN
No-one sees/saw you.
c. Ai khong thay anh?
AI NEG see PRN
Who doesnt/didnt see you?
The sentences in (48) present two opposing challenges to the analysis developed
thus far. Consider first the MFCSthat are interpreted as negative polarity items, in
(48a) and (48b). The position of khngin (46a) is unproblematic if it is in [Spec,
Asr], as assumed in (11). In (48b), however, khngDOESpose a problem since it
appears to be in a pre-subject position; on the current analysis, khngshould not have
any negative force in this position, nor should it cause subject aito be treated as a
negative polarity item.
-
8/13/2019 Linguistics Vn
51/81
Page 51
However, things may not be as they seem. A consideration of Vietnamese
existential clauses suggests an alternative analysis of (48b) that is perfectly consistent
with the structure in (11). Given the discussion thus far, it should not be surprising
that Vietnamese existentials are expressed using the multifunctional element c; in
this usage, c corresponds to the English copular be. Some examples are given in
(48):
(49)a. Co tin-tc quan trong (ma) lam moi ngi xuc dong.CO news important REL make every person excited
There is such important news as to make everyone excited.
b. Khong co nhieu xe-la hoat-ong na.
NEG CO much train operate more
Theres not much train travel anymore.
What is immediately relevant about these constructions is that, just as in English
there-constructions, the subject (there-associate) obligatorily occurs in a lower
position following c; this order contrasts directly with all the other cases we have
seen involving a subject NP plus (emphatic or negative) c.28
If it is now assumed that c occupies the same position in all of these
constructions, namely, under Asr, it allows for an alternative analysis for (48b) in
which the subject ai occupies a position LOWER than the normal position of the
subject at Spell-Out, and where (negative) khng preceding this subject occupies
[Spec, Asr], as shown in (50):29
-
8/13/2019 Linguistics Vn
52/81
Page 52
(50) 4T AsrP
4khng 4
Asr vP
1 3[+NEG] ai thay anh[-EMPH][-WH]1(c)
Such a re-analysis makes two straightforward predictions, both of which are borne
out. First, c should optionally appear between (negative) khngand the subject ai,
as in (51); by contrast, if khngwere really an inherent constituent negation, negating
aiin subject position, then cwould appear in preverbal position FOLLOWINGai, as in
the ungrammatical (51b).
(51)a. Khong co ai thay anh.NEG CO AI see PRN
No-one saw you, there is no-one who saw you.
b. *Khong ai co thay anh.
c. Khong co ai ma thay anh.
d. Khong co co gai thong-minh nao lay ngi co ta khong phuc.
NEG co PRN girl smart which take person PRN DEM NEG admire
No smart girl gets married to a man whom she doesn't admire.
On the analysis in (50), sentences such as (48b) are really covert existentials,
parallel to those in (49). As such, the second prediction is that the clausal material
following the NP aishould be analyzed as a relative clause, with a phonetically empty
subject. This latter prediction is borne out by the fact that (48b) is understood as
-
8/13/2019 Linguistics Vn
53/81
Page 53
wholly synonymous with (51c), in which the (optional) relative clause-marker m
appears between the subject/associate aiand the predicate thay.
Therefore, there is good reason to think that khng can be given a uniform
treatment as an underspecified multifunctional item within the scope of Asr in its
Specifier position even where khngapparently precedes indefinite subjects.
The remaining, and opposing, challenge is (48a), where ai receives a +WH
interpretation even though it is outside the scope of a Asr[+WH] head. Unlike the NPI
case just discussed, there is no possibility here of analyzing these as post-verbal
subjects; all of the available distributional evidence, including c and khng
placement, indicates that in this example aioccupies the canonical subject position
[Spec, T], above Asr.
Here, the most reasonable alternative analysis is that the +WH feature of Asr itself
raises (covertly) to C, functioning as a +WH operator in C, and thus restoring the
scope configuration required to license a +WH subject. This analysis, diagrammed in
(52) below, is entirely consistent with standard treatments of wh-movement in WH-in
situ languages.
(52) CP4C TP
[+WH] 4ai 4
T AsrP4
khng 4Asr vP1 4
[+NEG] ai thay anh[-EMPH][+WH]
1(co)
-
8/13/2019 Linguistics Vn
54/81
Page 54
If these proposals are correct, then the appearance of multifunctional elements
apparently in subject position does not constitute a real challenge to the main idea
behind (11), namely, that the interpretation of preverbal multifunctional elements
including c, aig, au, and khng(chang) is determined exclusively by their
scopal relationship to the syntactic head Asr, and by the values of the feature-set of
Asr.
Rather than considering these elements massively ambiguous, the present proposal
makes it possible to reduce their lexical specifications to a minimum, explaining their
ambiguity directly in terms of their distribution. For example, aiinterpreted variously
as who, no-one and anyone, may be minimally specified with a single lexical
feature [PERSON]; au(where, nowhere, anywhere, at all) might be specified as
[LOCATION], assuming that at all is a covert locative; see footnote 13 above. The
most significant reduction is offered for c , which as we have seen, is variously
interpreted as an emphatic, negative, interrogative, and existential marker, and to
express possession (have). Given the present proposal, c can be minimally
specified as the lightest of light verbs, a pure assertion marker.
As was already noted in the introduction, it is not original to claim that MFCs are
lexically underspecified, or that they owe their interpretation to the features of some
c-commanding operator: these ideas have already been articulated and developed by
linguists working on Chinese, especially Li (1992), Tsai (1994) and Cheng (1997);
also Aoun & Li (1993). However, as should be clear from the preceding sections,
Vietnamese provides something extra: on the one hand, a clearer distinction between
Tense and Assertion; on the other, clearer evidence for a very low position for the
Assertion operator.
-
8/13/2019 Linguistics Vn
55/81
Page 55
5.2. Evading Scope
Having considered the case where an element is exceptionally moved into the
scope of Asr, we turn finally to examine two contexts, previewed in the introduction,
where multifunctional elements are moved leftward from their canonical position to
evade the scope of a particular head. Note that this movement is different from
standard instances of scope-related movement QR, and the like in that the
movement is not driven by a property of the raised constituent, but by properties of
the configuration it is moving out of: here, an element is moving, not to check formal
features or to TAKEscope, but to EVADEit.
5.2.1. Universal readings
The most striking case of scope evasion is provided by the multifunctional
elements ai, g, and no(which, any), when functioning as universal quantifiers.
As the preceding sections have shown, these elements are construed in other contexts
either as negative polarity items when in the scope of Asr[+NEG], or as [+WH ]
variables in the scope of a WH-operator (in C, by 51)). In conjunction with the
morpheme cung, standardly translated as also, however, these elements are
obligatorily interpreted as universal quantifiers. The examples in (53) below show ai,
noin subject position, outside the scope of Asr (assuming the analysis in (11)): the
examples (53a,b) show the interrogative function of ai,no, while those in (53c, d)
show their universal function.30
(53)a. Ai biet co ay au? ai know PRN DEM be where
Who knows where she is?
-
8/13/2019 Linguistics Vn
56/81
Page 56
b. Sinh-vien nao biet giao s ay.
student which know teacher DEM
Which student knows that professor?
c. Ai cung biet co ay au?
who also knows PRN DEM be where
Everyone knows where she is.
d. Sinh vien nao cung biet giao s ay.
student which also know teacher DEM
Every student knows that professor.
At a morphological level, there are obvious parallels with some North East Asian
languages, which also form universal expressions by combining WH-elements with a
morpheme meaning also (compare, for example, Japanese dare-mo, Korean nookoo-
to). This morphological parallelism might lead one to suppose that in Vietnamese ai
cungforms a syntactic constituent also, contrary to the proposed analysis in (11).
However, there is clear evidence that this is not the case. When positions other than
the subject position are considered, a striking pattern is observed: if any argument
other than the subject is expressed with universal ai, g, no, etc., then that argument
must be fronted to a position preceding the Asr head, typically preceding the subject.
In such cases, cung remains in the same immediately pre-verbal position. The
examples in (53), from Ngo Nh Bnh(1999), illustrate fronting of direct objects and
temporal adjuncts. What is of interest here is that the quantified expression must not
be left in its canonical object position, nor in any position following preverbal cung,
as evidenced by the ungrammaticality of the sentences in (55):31
-
8/13/2019 Linguistics Vn
57/81
Page 57
(54)a. T nao anh ay cung nh.word which PRN DEM also remember
He remembers every word.
b. Ai co ay cung quen.
ai PRN DEM also know
She knows everybody.
c. Bao gi anh ay cung en muon.
what time PRN DEM also come late
He is always late.
d. Ngay nao toi cung tap the thao.
day which I also practise exercise
I do exercises every day.
(55)a. *Anh ay cung nh t nao.b. *Co ay cung quen ai.
c. *Anh ay cung en muon bao gi.
d. *Toi cung tap the thao ngay nao.
Prima facie,the ungrammaticality of the examples in (55) could be taken as a sign
that Quantifier Raising (QR) in the sense of May (1977, 1985), and subsequent
work is an obligatorily overt operation in Vietnamese. If this were correct, these
examples would not speak directly to the issue of scope evasion, whatever their
interest might be from the point of view of comparative syntax. However, an
examination of INHERENTLY SPECIFIEDuniversal quantifiers shows that overt QR is
-
8/13/2019 Linguistics Vn
58/81
Page 58
not obligatory in Vietnamese. If underspecified elements such as aiand no are
replaced with the inherently specified moi (every), as in the examples in (56), no
such movement is observed; indeed, cungmay not even co-occur with such elements.
In other words, fronting is exclusively restricted to underspecified elements whose
canonical position is below Asr.
(56)a. Ai cung nh moi t.ai also remember every word
Everyone remembers every word.
b. Vai thay-giao biet moi sinh vien.
Some teachers know every student
Some teachers know every student.
c. Vai co gai cung tap the thao moi ngay.
some PRN girl also practise exercise every day
Some girls also do exercises every day.
The notion of scope evasion permits a reasonably direct explanation of this
restriction. On the analysis in (11), underspecified elements within the scope of Asr
must receive one of three interpretations depending on the values of the feature-set of
Asr: in the case of ai, for example, different specifications of Asr yield who [+WH],
no-one [+NEG] and the indefinite someone [-NEG, -WH]. If these interpretations are
obligatory WITHINthe scope of Asr, then the only way for these elements function as
universal quantifiers is if they are moved to a syntactic position beyond the scope of
Asr. I suggest that this is what obtains here: that underspecified elements are fronted
in order to avoid an unwanted interpretation.32
-
8/13/2019 Linguistics Vn
59/81
Page 59
Finally, consider the position of fronted objects and adjuncts in the grammatical
sentences in (57): in contrast to the pre-subject position in the sentences in (54), these
examples show that MFCS may be fronted to a position immediately following the
subject position, but still preceding and outside the scope of the assertion
head.33 Note that Vietnamese is generally a rigidly configurational language: other
than the very free movement of topicalized constituents to the designated clause-
initial topic position, there is no scrambling in Vietnamese, so that in contrast to
the MFCSin (57) inherently specified objects must not be placed in this position.
(57)a. Anh ay t nao cung nh.PRN DEM word which also remember
He remembers every word.
b. Co ay ai cung quen.
PRN DEM ai also know
She knows everybody.
The sentences in (57) thus provide further confirmation for the claim that the
active syntactic head (Asr) occupies a comparatively low position in Vietnamese
phrase-structure: if TP, rather than AsrP, were the head to be evaded, these MFCs
could only be adjoined in the higher clause-initial position (53) in order to be
interpreted as universal quantifiers.
5.2.2. Tense alternations
As already prefaced in the introduction, the second instance of apparent scope
evasion in Vietnamese is observed with certain +WH temporal adjuncts (bao gi, luc
-
8/13/2019 Linguistics Vn
60/81
Page 60
nao, khi nao); that is to say, with expressions corresponding to English when, what
time, etc. In Vietnamese, such expressions occur in one of two syntactic positions:
following vP-internal arguments and clause-initially. In both positions, the
expressions mean when. There is a crucial interpretive distinction, however: in
initial position, bao giis obligatorily interpreted with future time reference, whereas
in its (more canonical) sentence-final position, bao giis interpreted as referring to
past time. This minimal contrast is illustrated in (58):
(58)a. Bao gi co ay i My?when PRN DEM go America
When will she go to America?
b. Co ay i My bao gi?
PRN DEM go America when
When did she leave for America?
c. Luc nao anh xem cuon phim o?
when PRN watch CLS film DEM
When will you watch that film?
d. Anh xem cuon phim o luc nao?
PRN watch CLS film DEM when
When did you watch that film?
This pattern is noted in all standard Vietnamese reference texts and teaching
guides.34 This temporal contrast is apparently grammatical, as opposed to being a
(pragmatically determined) preference rule however that might be described. This
-
8/13/2019 Linguistics Vn
61/81
Page 61
is made clear by the fact that each order is compatible with only one or other tense
morpheme, whenever these are explicitly realized: clause-initial temporal expressions
are incompatible with the pre-verbal past tense morpheme a, but possible with the
future morpheme se; the opposite judgments hold of the same expressions in clause-
final position. Compare (58) with the examples in (59) below.
(59)a. *Bao gi co ay a i My?when PRN DEM ANT go America
When did she go to America? (07/25)
b. Bao gi co ay se i My?
when PRN DEM FUT go America
When will she go to America? (25/25)
c. Co ay a i My bao gi?
PRN DEM ANT go America when
Whe