An interface analysis of verb subject inversion in Romance

AN INTERFACE ANALYSIS OF VERB SUBJECT INVERSION IN ROMANCE
MICHELLE SHEEHAN
0. Abstract
The idea that so-called ‘free inversion’ in Null Subject Languages (NSLs) results from
optional subject movement is incongruous with Chomsky’s (1995, 1998) ‘minimalist’ notion
of economy. Alexiadou & Anagstopoulou (A&A)(1998) evade this economy violation by
assuming the preverbal subjects are not the result of movement, but are base-generated in a
clitic left dislocated (CLLD) position in rich agreement NSLs. This assertion is the natural
consequence of their claim that the EPP, in rich agreement pro-drop languages, is satisfied by
verb movement to I, where V contains a pronominal inflection, which is the equivalent of a
pronoun. Preverbal lexical and pronominal subjects thus differ from their post-verbal
counterparts in that they are not generated as arguments of the verb, but are base-generated in
an A’-position and co-indexed with the pronominal INFL.
A&A’s (1998) analysis makes strong semantic predictions. A&A base their (semantic)
argument on the claim that preverbal subjects, like CLLDed objects display certain semantic
effects, which they do not share with A-moved arguments. They claim that arguments which
undergo A-movement preserve ambiguous scope, whereas CLLD arguments are always
unambiguous. Thus A&A claim that preverbal QPs in NSLs get only wide scope, whereas
preverbal QPs in non NSLs, like English, get ambiguous scope. They also assert that
preverbal indefinites get only a specific/partitive reading. I give similar evidence from the
interpretation of indefinites in embedded CPs in Spanish and Italian.
However, many aspects of A&A’s analysis remain unsatisfactory. Following Suñer
(2002), I give data from Spanish to contest A&A’s empirical claims. I also raise a number of
theoretical objections. I propose, moreover, that the semantics of indefinites does not
constitute proof that preverbal subjects are in an A’-position. Rather, I suggest, the
definiteness effects observed fall out from an adaptation of Diesing’s (1992) Mapping
Hypothesis, which takes syntactic scope to be parallel to semantic scope. I suggest that A&A
are correct in their claim that SVO order in Romance does not result from EPP-driven
movement. However, I differ from them by assuming that preverbal subjects are indeed
generated as true arguments inside vP, which then move to a preverbal position. I claim that
this movement is triggered by syntactic features of semantic import, which require the subject
to escape vP and avoid existential closure.
BP and the NSLs studied thus exhibit two different kinds of movement respectively.
The first is a “last resort”, driven by the EPP as an obligatory structural requirement
(Chomsky 1995). The second has “external motivation in terms of distinct kinds of semantic
interpretation and perhaps processing”(Chomsky 2001:3). I tentatively suggest that the
minimalist framework requires a further type of semantic feature in order to formalise the
kind of movement related to LF effects such as quantifier raising, scrambling or negative
scope.
1. Introduction
In this section I give a brief introduction to the Romance data and the problems that it
presents for minimalist syntax. Section 2 is a summary of A&A’s analysis and the various
predictions it makes. This is followed in section 3 by a consideration of the EPP setting in
Brazilian Portuguese, a partially pro-drop language. In Section 4, I raise some independent
objections to a CLLD analysis, based on work by Suñer (2002). In section 5 I introduce
Diesing’s Mapping Hypothesis (MH) and Adger’s (1996) Generalisation and show how,
together, they provide an alternative account of the data, whereby the movement of subjects in
NSLs is triggered by syntactic features of semantic import. Under this analysis the different
semantic effects fall out from economy. In Section 6 I draw some conclusions as to the wider
implications of these claims.
1.1. Free Inversion
It is widely observed that English, like most of Germanic, exhibits a kind of subject/verb (SV)
inversion, with an expletive in the structural subject position:
1) (a)
(b)
[TPThere T [vP entered a strange man]
[TP A strange man [vP entered]
In English this structure is limited to so-called unaccusative/ergative verbs, which assign a
theme -role and thus generate their subject as an internal argument. The same structure is
ungrammatical with an unergative verb, which generates the subject as an external argument:
2) (a)
(b)
*There ran a strange man
A strange man ran
The unaccusative structure is also rendered ungrammatical where the low subject is definite
(the definiteness restriction):
3) *There entered the man
This is only superficially similar to SV/VS alternation in Romance. Burzio (1986) claims that
that most matrix SV structures in Romance have an “essentially synonymous VS
counterpart”. This apparently ‘optional’ VS/SV alternation is often termed free inversion.
4) (a)
(b)
Juan
llegó
Juan
arrived:3s
Llegó
Juan
Arrived:3s
Juan
‘Juan arrived.’
[Spanish (like Italian)]
Example (4) uses an unaccusative verb, but free inversion is also observed with unergative
and transitive verbs:
5) (a)
(b)
Juan
Juan
canta
sings:3s
Canta
mucho
Sings:3s
much
‘Juan sings a lot.’
mucho
much
[Spanish (like Italian)]
Juan
Juan
Post-verbal subjects is structures such as (4)&(5) do not display the definiteness restriction.
1.2. Optional Movement or Null Expletives: Minimalist Concerns
The label free inversion implies that the subject in (4)&(5) can move optionally to a preverbal
position, or remain in-situ. This is, however, incongruous with the general concept of
economy in minimalism, whereby:
6) “Simpler operations are preferred to more complex ones, so that Merge or Agree (or
their combination) preempt Move, which is a ‘last resort’, chosen when nothing else is
possible.” (Chomsky 1998)
Movement of the subject to a preverbal position can, therefore, only be triggered where its
failure to move will lead to a crash in the derivation. This is formalised in the concept of
uninterpretable features, which are not interpretable at LF and so must be deleted in the
course of the derivation.
The most infamous of these movement-triggering uninterpretable features is the EPP,
which Chomsky has claimed to be a universal property of natural languages (NLs): the
requirement that every sentence have a subject (Chomsky 1982:10). In formal terms this is
often conceived of as an uninterpretable [D] feature on T, which requires a D element to
merge in Ts specifier. The EPP provides a simple explanation for expletive structures like the
following:
7) It seems that John is ready
8) John seems to be ready
In (7), the embedded subject John cannot move out of a Case position and so and expletive
‘it’ is inserted to satisfy the EPP [D] feature on T.
Assuming (perhaps naively) that Chomsky is right, and that all languages do indeed
share this kind of EPP feature, we need to account for the fact that VS order in NSLs does not
crash the derivation. How is the EPP satisfied in these instances? Rather than assuming that
movement is optional we could account for the data by assuming that VS and SV orders result
from different numerations. The VS numeration would include a null expletive, which is the
basic equivalent to the overt expletive ‘there’ in English. However there are at least two
empirical reasons to reject this idea. As we have seen, expletive VS structures in English are
limited to unaccusative verbs and display definiteness restrictions. Free inversion in
Romance shows no such syntactic or semantic constraint.
This empirical mismatch gives weight to obvious theoretical objections against an
analysis of free inversion based on a null expletive. A null expletive that has no (obvious) LF
interpretation is potentially not interpretable at either interface (PF or LF). In a framework in
which the basic premise is to avoid unnecessary theoretical baggage, and remain minimalist,
this kind of purely syntactic entity is obviously highly objectionable.
BP, however, clearly patterns differently to NSLs. It displays definiteness restrictions
(Kato 2000), and licenses VS order only with unaccusative verbs (Britto 1999):
9) (a)
(b)
Os pássaros
The birds
cantam
sing:3pl
[Brazilian Portuguese]
*Cantam
os passaros
sing:3pl
the birds
‘The birds sing/are singing’
10) Tinha chegado
muitas cartas
Had arrived
many letters
‘There had arrived many letters.’
11) ???Tinha
chegado
o homem
Had
arrived
the man
‘There had arrived the man.’
There is therefore evidence to suggest that BP does license a null expletive, which is in some
way visible at LF, with the same distribution as ‘there’. Let us assume then that BP licenses
VS order through the merging of a null expletive to satisfy the EPP. Whatever the reason that
unergatives fail to license expVS structures in English is also the explanation of the BP facts.
I leave this matter open, assuming only that it is the result of the differing argument structures
of the two verb types.
2. Alexiadou & Anagstopoulou (1998) Parameterising Agr
We have assumed that (restricted) VS order in English and BP is the result of an expletive
construction. Assuming that no such structure is available in NSLs, and that movement, in a
minimalist framework, cannot be optional, we are no closer to an account of free inversion in
Romance NSLs. We have also failed to explain how the EPP is satisfied in sentences of
NSLs with VS order.
A&A provide a possible account of free inversion based on the claim that verbal
morphology in a rich agreement language “includes a nominal element ([+D, +interpretable
phi-features, potentially +Case])”. This means that ‘agreement’ morphology in languages
like Greek and Spanish have “exactly the same status as pronouns in the English
paradigm”(A&A 1998:516). Under their analysis, the pronominal Agr absorbs the subject
theta-role and precludes the need for a null or overt referential pronoun or lexical subject in
vP. Pronouns and lexical subject can be generated in vP and in such cases Agr has the same
status as object clitics in clitic-doubling structures such as (12):
12) Lo
vio
a
CLacc:him
saw:3s
A
‘I saw (him) Juan yesterday.’
Juan
Juan
ayer
yesteday
A&A also construct their theory so as to forgo the need for the (theoretically objectionable)
category of a null expletive. Thus VSO word order is claimed to be an obvious by-product of
the following parameter regarding EPP satisfaction:
13) Parameterised mode of EPP checking: Move/Merge XP vs. Move/Merge X0
Languages such as English, which lack a D feature in their agreement morphology, require
move/Merge XP to satisfy the EPP, as can be observed in expletive constructions. In rich
agreement languages, on the other hand, V movement to I satisfies the EPP. This means that,
according to A&A’s assumptions, spec IP never projects a specifier position as it lacks the
motivating feature (Chomsky 2001) and so no expletive pro can, or need, be merged.
An obvious corollary of this analysis is that preverbal subjects in null subject
languages cannot be in Argument position, as spec IP is not projected. A&A thus make a
strong distinction between post-verbal subjects, which remain vP-internal, and preverbal
subjects, which cannot be the result of EPP-driven movement and so, they assume, must be
base-generated in a CLLD position.
2.2. Implications for the semantics of lexical subjects
A&A show that, in Greek quantifier phrases (QPs) and indefinites in preverbal subject
position have unambiguous scope, as do CLLDed objects. The following is taken from A&A
1998:
14) Kapjo pedi to
eksetase
kathe kathigitis
Some child cl-ACC
examined
every professor
‘There was some child that every professor examined.’
[wide scope only]
15) Kapios
fititis
sitihiothetise kathe arthro
Some
student
filed
every article
‘There was some student that filed every article.’
[wide scope only]
16) Sitihiothetise kapios fititis
kathe arthro
[ambiguous]
Filed
some student
every article
(a)‘There was some student that filed every article.’
(b)‘Every article was filed by some student (though not necessarily the same student).’
They claim that the same is also true of other rich agreement NSLs and conclude that in all
such languages preverbal subjects are always CLLDed.
They also give evidence, again from Greek, of the interpretational effects of
‘indefinites’, which can potentially have either a ‘weak’ existential, or ‘strong’
specific/partitive reading, in de Hoop’s (1992) terms. Again preverbal subjects (18) appear to
behave like CLLDed objects (17):
17) ?Enan anthropo
ton
heretise
i Maria
One person
CL-acc
greeted
Maria
‘Mary greeted one of the people.’
[strong/presup.]
18) Ena pedhi diavase
to
Paramithi
horis Onoma
A
child read
the
fairy-tale
without a title
‘One (specific) child read the fairy tale without a title.’
[specific/strong]
19) Diavase
ena
pedhi to
Paramithi
horis Onoma
read
a
child the
fairy-tale
without a title
‘The fairy tale with a title was read by some child.’
[existential/weak]
The CLLDed object in (17) can only get a strong presuppositional interpretation, as does the
preverbal subject in (18). Compare these facts with English, which displays an ambiguous
reading with preverbal subjects:
20) A child read the fairy tale without a title
‘The fairy tale with a title was read by some child.’
‘One (specific) child read the fairy tale without a title.’
[existential/weak]
[specific/strong]
Again A&A take this as evidence that preverbal subjects are CLLDed in NSLs like Greek,
and by extension Spanish and Italian.
Independently collected data from Spanish and Italian patterns in the same way, as
regards the interpretation of indefinites. Consider the following from Italian, which also holds
for Spanish:1
21) Mi
domando
se
cantano
sempre
alcuni uccelli.
1s-refl ask
if
sing
always
some birds
(a)‘I wonder if there are always some birds singing at any time.’ [existential/weak]
(b)*‘I wonder if some (specific) birds sing all of the time.’
[presup./strong]
22) Mi
domando
se
alcuni uccelli cantano
sempre.
1s-refl ask-1s-pres if
some birds sing
always
(a)*‘I wonder if there are always some birds singing at any time.’ [existential/weak]
(b)‘I wonder if some (specific) birds sing all of the time.’
[presup./strong]
In (21) the postverbal subject gets only an existential reading, as does the Greek equivalent. In
preverbal position (22) the same indefinite gets an unambiguously strong reading. Evidence
does suggest, then, that there are semantic differences between preverbal and postverbal
subjects in NSLs, as A&A’s analysis predicts. Note, however, that post-verbal indefinite
subjects are also unambiguous, always getting a weak existential reading. A&A attribute this
to the fact that DPs inside vP are subject to existential closure, as Diesing (1992) has claimed.
We will return to this idea below.
1
Data collected from native speakers with the manipulated semantic renderings shown here.
3. The EPP parameter setting of Brazilian Portuguese
In BP null referential subjects are ungrammatical in matrix CPs, as Duarte (2000) has shown:2
23) *Comprou
um
carro novo
Bought
a
new car
‘He/you/we bought a new car.’
This fact, coupled with the ungrammaticality of free inversion appears to suggest that BP is
not a NSL of the Spanish/Italian type. We have already given evidence that the licensing of
VS word order involves an interpretable null expletive. All this points to the fact that BP has
the same EPP parameter setting as English: move/merge XP. I leave it open how the EPP is
satisfied in sentences like (24)&(25), which require an overt ‘it’ expletive in English:
24) choveu
a
noite
rained
the
night
‘It rained the whole night long,’
inteira.
whole
25) parece
que
o João
passou
seems
that the João
passed
‘It seems that John passed by here.’
por aqui.
by here
Chomsky (1981:323-325) has argued that the subject in sentences like (24) is actually quasiargumental, and therefore interpretable at LF.3 This could potentially be a null quasiargument equivalent to the English it. (24), on the other hand, is more problematic as the verb
parecer (to seem) does not assign an external -role. However, the fact that the subject in
English is it rather than the ‘pure’ expletive there suggests that it might be an extraposition
pronoun which is interpretable at LF. It seems that (24) & (25) might be accounted for
unproblematically if we assume a null quasi-argumental pronoun. I leave this matter open to
future research. A&A’s analysis only takes account of the EPP parameter setting with respect
to referential arguments.
3.1. The Semantics of Preverbal Subjects
As we would predict, BP patterns differently from the Romance NSLs. In BP, preverbal
indefinite subjects get ambiguous readings:
26) Eu
me
pergunto
se alguns pássaros cantam
sempre
I
1s-refl ask-1s-pres if some birds
sing
always
(a) ‘I wonder if there are always some birds singing at any time.’ [existential/weak]
2
Note however, that BP does allow discourse pro-drop of the type exhibited by English, where the 1s pronoun
can be dropped in the highest specifier postion.
3
This is supported by the following data from English, where ‘it’ can control a PRO in an adjunct clause:
1) It often clears up right after [PRO snowing heavily]
(b)‘I wonder if some (specific) birds sing all of the time.’
[presup./strong]
This contrast is explained if we assume that preverbal subjects in BP are the result of Amovement to satisfy the EPP. In A&A’s terms the moved subject preserves its ambiguous
reading.
A slight complication stems from the fact that BP also allows ambiguous readings
with post-verbal indefinites in embedded CPs:
27) Eu
me
pergunto
se cantam
sempre
alguns pássaros.
I
1s-refl ask
if sing
always
some birds
(a) ‘I wonder if there are always some birds singing at any time.’ [existential/weak]
(b)‘I wonder if some (specific) birds sing all of the time.’
[presup./strong]
This is not problematic, however when we consider that BP exhibits optional VS inversion (Ito-C movement) in questions, both matrix and embedded. The following data illustrates this
free variation heard in everyday speech:4
28) De
onde você é?
From where you are
‘Were are you from”’
29) Quanto
custa uma ida
How-much costs a
go
‘How much is a return ticket?’
y
and
volta?
return
I therefore assume that in both (26) and (27) the subject is in spec IP. In (26) the verb
remains in I. However, in (27) the verb has raised to a position in the split CP system below
INT (the position of se-if) (Rizzi 1995, 1999). In both cases the A-moved subject in spec IP
gets an ambiguous interpretation.
4. Problems with a CLLD analysis
There are however independent reasons to object to A&A’s analysis. There are many reasons
to believe that preverbal subjects in Spanish and Italian are not always the result of CLLD.
Suñer (2002) gives evidence from scope, ad sensum agreement and reconstruction to suggest
that in many cases preverbal subject in Spanish must be in an A-position. I also suggest that
an analysis in which all preverbal subjects are in an A’-position co-referential to a referential
pronoun is semantically anomalous. I give evidence from BP in support of this claim.
4.1. Preverbal Subjects in Spanish
Suñer (2002) gives evidence to show that Spanish does not behave like Greek in terms of
scope. Preverbal subjects in Spanish can get ambiguous scope, unlike their Greek
equivalents:
4
Note that I-to-C is more common with certain wh-words. This kind of variation is less problematic that the
stable SV/VS order in affirmative CPs in NSLs, purely because it seems to be a change in progress, and it
involves head movement, rather than XP movement.
30) Algún estudiante
sacó prestado
cada libro
Some student
took lent
each book
(a) ‘Each book was borrowed by some student.’
[each>>some]
(b) ‘Some (specific) student borrowed each book.’
[some>>each]
Whereas in the Greek equivalent only wide scope is available for the preverbal subject, the
more normal reading for (30) is (a) where it gets narrow scope. Suñer does note, however
that both readings are available. This is expected as (b) is one possible way of satisfying the
truth conditions of (a). It this sense (b) is a subset of the intension of (a).
Suñer (2002), following Bosque (1999), also shows that true CLLD subjects in
Spanish allow optional ad sensum agreement:
31) El
jurado,
María nos
aseguró
que
The jury:ms
María us
assured:3s
that
‘The jury, Maria assured us that they felt pressured.’
estaban presionados
were:3pl pressured:mpl
In (31) the predicate ‘be pressured’ has 3pl agreement. Its subject ‘the jury’ is grammatically
singular but notionally plural. Agreement therefore reflects semantic number rather than
syntactic number, indicating that the subject is not in a spec-head relation with I. This ad
sensum agreement is not possible with all preverbal subjects:
32) El jurado
*estaban presionados /estaba presionado
The jury
*were pressured
was pressured
‘The jury felt pressured.’
This leads Suñer to conclude that preverbal subjects cannot always be the result of CLLD in
Spanish.
4.2. Non-Referential Preverbal Subjects
As A&A acknowledge (A&A 1998:508), their analysis predicts that elements, which typically
cannot be left dislocated should never appear preverbally. They give the following evidence
from Italian to support their claim that QPs, contrary to popular belief can actually be CLLD:
33) (a)
(b)
Qualcuno
lo
troveranno
Someone
him will-find:3pl
‘They will find some specific person.’
Qualcuno
(*lo) troveranno
Someone
him will-find:3pl
‘They will find someone.’
[specific:strong]
[existential:weak]
I propose that the fact that the clitic is obligatorily absent in (33b) is due to the fact that
qualcuno has a non-referential reading. This strongly suggests that (33b) is not an instance of
CLLD, where the object is generated in an A’-position, but is the result of object movement to
a position in the CP layer. There is in fact no evidence to suggest that in (33b) the object is
CLLD. A&A gloss over this distinction.
I maintain that, at least in Romance, CLLD is impossible with non-referential
subjects/objects.5 This is clearly illustrated by BP, which forms CLLD structures with overt
co-referential pronominal subjects. Britto (2000) shows that only referential subjects can be
CLLD in BP:
34) [O
Instituto de F.]i
elei
manda os piores professores, (...)
the
Institute of F.
it
sends the worst professors...
‘The institute of F. sends the worse proffessors.’
[referential]
35) [Toda pessoa que
assiste uma peça]i elai
tem uma opinião (...)
every person who attends a
play he
has an opinion (...)
‘Everyone who goes to see a play has an opinion.’
[referential]
36) *[Dois homens]i
elesi querem
falar
Two men
they want:3pl
speak:inf
‘There are two men who want to speak wit you.’
contigo
with-you
[non-referential]
37) *[Ninguém]i elei
gosta de
Nobody
he
likes of
‘Nobody likes to cry.’
[non-referential]
chorar
cry:inf
Examples (36)-(37) involve non-referential subjects and thus CLLD is rendered
ungrammatical. This stands to reason, as a robustly non-referential subject cannot be coreferential with a referential pronoun.
If INFL in NSLs has exactly the same syntactic status as a referential pronoun, as
A&A claim in their analysis, then the same restrictions should apply. The simple fact that
Spanish and Italian allow the robustly non-referential QPs in preverbal subject position means
that not all preverbal subjects can be CLLD:
38) Ninguno
quiere ser
político
Nobody
wants be:inf politician
‘Nobody wants to be a politician.’
5. An alternative analysis
A&A take the unambiguous semantics of SV order as proof of their claim that preverbal
subjects are always CLLD in NSLs. My data supports their empirical claims but I would like
to show, as they acknowledge in footnote 16, that this data can also be explained using
Diesing’s (1992) Mapping Hypothesis. Given the objections raised to a CLLD analysis, this
alternative analysis is claimed to have better theoretical and empirical motivation.
5
Note that in Barbosa’s (2000) analysis of preverbal subjects in Romance, she claims that QPs cannot be
CLLDed. She gives evidence to support this claim from the positioning of adverbs, but her claim results from
the idea that non-referential QPs cannot be co-referential with a referential pronoun.
5.1. Diesing’s (1992) Mapping Hypothesis
Deising (1992) claims that semantic scope corresponds directly with syntactic scope. Her
Mapping Hypothesis (MH) claims that syntactic position at LF is mapped directly to semantic
scope in the following way:
39) Mapping Hypothesis (MH) (Diesing 1992)
Material from vP is mapped into the nuclear scope.
Material from IP is mapped into a restrictive clause.
Material that is mapped to nuclear scope gets a ‘weak’ existential reading, and material
mapped to restrictive scope gets a ‘strong’ presuppositional interpretation. In the case of
indefinites, semantic interpretation will depend upon syntactic position. DPs that remain in
vP will get a weak reading. DPs that move into IP will get a strong presuppositional reading.
Diesing derives this generalisation from ‘scrambling’ in embedded CPs in German:
40) weil ja doch [vP zwei cellisten in diesem Hotel
since indeed [VP two cellists in this hotel
‘…since two cellists stayed in this hotel.’
abgestigen
sind]
stayed
are]
[weak/existential]6
41) weil [IPzwei cellisten ja doch in diesem Hotel
since[IP two cellists indeed
in this hotel
‘…since two of the cellists stayed in this hotel.’
abgestigen sind]
stayed
are]
[strong/partitive]
Diesing assumes that the adverb ja doch adjoins to vP and therefore marks the vP/IP
boundary. In (40) the subject is therefore taken to be vP internal and therefore gets a
weak/existential reading. In (41), on the other, hand it has scrambled to spec IP and so gets a
strong/partitive reading. Languages such as English pose a problem for the MH as IP internal
subjects can get an optionally existential reading:
42) Three cats are in the garden
(a) There are three cats in the garden
(b) Three of the cats are in the garden (the rest are on the roof)
In order to make her generalisation universal, Diesing proposes a Lowering Parameter (LP).
This means that, in languages such as English, IP-internal subject can be lowered at LF and
mapped to nuclear scope to get an existential reading.
I would like to suggest, following Adger (1996), that recent developments in
minimalist theory render this LP superfluous.
5.2. Adger’s (1997) Generalisation and Economy
6
Note that de Hoop (1992) challenges this, asserting that the subject can get ambiguous readings.
Adger (1996) proposes a more theoretically appealing explanation of the mechanics of
Diesing’s LP, based on a consideration of the composite operations making up move.
According to Chomsky (1995), move consists of copy and merge. Where an element moves, a
trace, which is still available to the semantics, remains in its base-position. Thus the DP’s
syntactic information occupies two positions, one in vP and one in IP. At LF either the trace
or the copy is deleted and the remaining DP is mapped to semantic scope via the MH.
Adger (1996) takes this explanation further to show that ‘lowering from IP’ at LF
(Diesing’s LP) is not parametrically determined but rather is the direct result of economy
concerns. Adger’s (1997) generalisation is as follows:
43) “Where obligatory movement occurs from vP to IP the interpretation of the subject
will be ambiguous between existential and quantificational readings, whereas when
this movement is optional there will be no such ambiguity.”
The explanation for this falls out from economy. The composite operation copy-mergedelete-map is used to raise a subject from vP to IP and then to ensure it is mapped to
restrictive scope. Now consider the composite operation needed to map an IP-raised subject
to nuclear scope: copy-merge-delete-map. Consider how uneconomical this operation is
compared to the simple operation by which a vP-internal subject is mapped directly to nuclear
scope: map. Recall Chomsky’s (1998) assertion:
44) “Good design conditions would lead us to expect that simpler operations are preferred
to more complex ones.”
Theoretically then, a one-stage mapping will be preferred over a 4-stage operation. It might
be, however, as is the case in English, that the subject is obliged to move for independent
reasons. Where this is true and movement is obligatory, clearly the syntax has no option but
to use the less economical mapping process (copy-merge-delete-map) for a nuclear-scope
mapping. For this reason subjects such as the one in example (42) are ambiguous as the
subject has moved obligatorily to satisfy the EPP and so either copy can be mapped at LF. I
assume that this is also the case with preverbal subjects in BP.
In NSLs, on the other hand, the EPP is satisfied by verb movement. This means that
any DP movement to a preverbal position is not obligatory. For this reason I propose that an
ambiguous DP can only move where it does so for interpretational effects. An indefinite
moves to preverbal position to get a strong interpretation at LF. A weak indefinite cannot
move, however, as the one-stage mapping (map) is available and so the less economical fourstage operation is not available.
5.3. Syntactic Features of Semantic Import
I claim therefore, that the unambiguously ‘strong’ interpretation of preverbal subjects in
Spanish and Italian is the result of movement of a vP internal subject. This movement is an
instance of movement motivated by a syntactic feature of semantic import. Chomsky (2001)
states that movement appears to be an imperfection in the system except that:
45) “It has (at least plausible) external motivation in terms of distinct kinds of semantic
interpretation and perhaps processing”. (Chomsky 2001)
As the feature which triggers movement is responsible for a difference in semantic
interpretation, economy is not challenged. The two alternative structures represent distinct
semantic information.
6. Conclusion
I have given evidence that suggests that, at least in Spanish and Italian, preverbal subjects are
not always CLLD. I have shown, however, that A&A’s analysis does not necessarily require
all preverbal subjects to be in an A’-position. An alternative analysis has been posited in
which preverbal subjects are the result of movement from a vP-internal position. This
movement is triggered by a syntactic feature of semantic import which is required to
disambiguate an indefinite. The result is that an in-situ post-verbal indefinite gets only a
weak/existential reading, whereas a moved preverbal indefinite gets only a
strong/presuppositional interpretation. This has been shown to be derived from economy
concerns.
Movement to what appears to be the same preverbal position results in ambiguity in
languages such as BP, where the subject moves to satisfy the EPP. In this case movement is
obligatory and so an in-situ (more economical) mapping to nuclear scope and an existential
reading, is not available in the grammar. For this reason indefinites in preverbal position in
BP get ambiguous weak/strong readings.
This analysis raises many new questions, and leaves many of them unsolved.
Obligatorily strong QP subjects appear to be licensed in preverbal and post-verbal position:
46) Me
pregunto
si todos
los pájaros
1s-refl ask:1s
if all
the birds
47) .Me pregunto
si cantan
siempre
1s-refl ask:1s
if sing:3pl
always
‘I wonder if all the birds sing all of the time.’
cantan
siempre.
sing:3pl
always
todos los pájaros.
all
the birds
This is certainly not predicted by the MH. Diesing (1992) assumes that all QPs raise out of
vP by LF in order to be mapped to restrictive scope. Because they are never ambiguous, I do
not see the semantic mapping of QPs as directly problematic. What is problematic is the
apparent availability of free inversion with QP subjects, if it is the case that no interpretational
effects obtain in (46)-(47). It is possible that the subject in (46), though quantificational is
referential enough to be CLLD. Clearly this requires further research.
This in turn raises a further question: what is the nature of the feature which triggers
the movement that in turn gives rise to interpretational effects? Is it interpretable or
uninterpretable. It could be uninterpretable and checked off in the course of the derivation.
Under this analysis the feature would be a purely syntactic tool which moves a DP to the
required position in the CP for the MH to apply. This assumes the MH to be a process at LF.
It could be argued, however that the MH is too representational to be minimalist. If this is so
then it might be that the features that trigger movement are also, themselves interpretable at
the LF interface. Under this analysis the movement and interpretation would result from the
same feature but would not be directly linked to one another.
7. References
Adger, D. (1996) Economy and optionality: Interpretations of subjects in Italian. Probus 8.
117.
Alexiadou, A. & Anagnostopoulou, E. (1995) SVO and EPP in null subject languages and
Germanic. FAS Papers in Linguistics 4.1-21.
Alexiadou, A. & Anagnostopoulou, E. (1998). Parametrizing AGR: Word Order, Vmovement and EPP-checking. Natural Language and Linguistic Science 16. 491-539.
Barbosa, P. (2000) Clitics -a Window into the Null Subject Property. In Costa, J. (ed.),
Portuguese Syntax-New Comparative Studies, Oxford University Press, Oxford.
Britto, H. (1999) Syntactic Codification of Categorical and thetic judgments in Brazilian
Portuguese. In Kato, M & Negrão, E (eds.) Brazilian Portuguese and the Null Subject
Parameter. Vervuert, Germany.
Burzio, L. (1986) Italian Syntax, Dodrecht, Reidel.
Chomsky, N. (1981) Lectures on Government and Binding, Foris, Dordrecht.
Chomsky, N. (1982) Some Concepts and Consequences of the Theory of Government and
Binding, MIT Press, Cambridge, Mass.
Chomsky, N. (1995) The Minimalist Program. MIT Press, Cambridge, Mass.
Chomsky, N. (2001) Derivation by Phase. In Kenstowicz, M (ed.), Ken Hale: A Life in
Language. MIT Press, Cambridge, Mass.
De Hoop, H. (1992) Case Configuration and the Noun Phrase, Rijksuniversiteit, Groningen.
Diesing, M. (1992) Indefinites, MIT Press, Cambridge, Mass.
Duarte, E. (2000) The loss of the ‘avoid pronoun principle in Brazilian Portuguese. In Kato,
M & Negrão, E (eds.) Brazilian Portuguese and the Null Subject Parameter. Vervuert,
Germany.
Figueiredo Silva, M. (1996) A Posição do Sujeito no Português do Brasil: Frases Finitas e
Infinitivas. Campinas. SP, Ed. da UNICAMP.
Kato, M. (2000) The Partial Pro-drop Nature and the Restricted VS Order in Brazilian
Portuguese. In Kato, M & Negrão, E (eds.) Brazilian Portuguese and the Null Subject
Parameter. Vervuert, Germany
Rizzi, L. (1995) The Fine Structure of the Left Periphery. In Haegeman, L. (ed.) Elements of
Grammar. Kluwer Academic Publishers, Dordrecht.
Rizzi, L. (1999) On the Position “Int(errogative)” in the Left Periphery of the Clause. ms.
Università di Siena.
Suñer, M. (2002) The Lexical Preverbal Subject in a Romance Null Subject Language: Where
art Thou? In Nuñez-Cedeño, R, López, L & Cameron, R (eds.) A Romance
Perspective on Language Knowledge and Use. John Benjamins, Amsterdam.
Michelle Sheehan
Department of Linguistics
Durham University
Elvet Riverside Building
New Elvet
Durham
[email protected]