Professional Documents
Culture Documents
Romance Languages
Ian Mackenzie
Unaccusative Verbs in Romance Languages
Also by Ian Mackenzie
List of Tables ix
Acknowledgement x
Sources of Historical Examples xi
1 The Ergative Analysis and the Unaccusative Hypothesis 1
1.1 The Ergative Analysis 1
1.2 The Ergative Analysis versus the Unaccusative
Hypothesis 5
1.3 General problems with the ‘deep-object’ hypothesis 8
1.4 Conclusion 15
2 Expletive Inversion 17
2.1 Introduction 17
2.2 Expletives and unergative subjects 18
2.3 Partitive case 20
2.4 A preliminary conclusion about expletive inversion 30
2.5 Distribution of French il 31
2.6 Conclusion 37
3 Partitive Cliticization 39
3.1 Introduction 39
3.2 Syntactic arguments 40
3.3 Partitive cliticization independent of unaccusative–
unergative distinction 55
3.4 The distribution of partitive ne/en 60
3.5 Conclusion 67
4 Bare Subjects 70
4.1 Introduction 70
4.2 Unergatives can have bare postverbal subjects 73
4.3 Constraints on bare subject distribution 78
4.4 Aspectually stative constructions 95
4.5 Modified and conjoined bare subjects 99
4.6 Conclusion 101
5 Perfect Auxiliary Selection 103
5.1 Introduction 103
5.2 Burzio’s theory 104
vii
viii Contents
8 Conclusion 182
Notes 187
References 212
Index 220
List of Tables
ix
Acknowledgement
The preparation of this book was fi nanced by a grant from the Research
Leave scheme of the Arts and Humanities Research Council.
x
Sources of Historical Examples
xi
xii Sources of Historical Examples
Not all of the relevant verbs exhibit all of the above properties (for
example, Spanish does not have a partitive clitic like Italian ne and
1
2 Unaccusative Verbs in Romance Languages
Notice that in (2) and (3) the subject Gianni is analysed as having
moved from its original position as the complement of arrivato, which
is now occupied by a trace t (bearing the same subscript as the moved
phrase). On the other hand, the indefinite postverbal subject in (1)
undergoes no movement and is thus structurally a direct object (albeit
one that does not have accusative case).3
In contrast to the above patterns, unergative subjects were assumed
to originate in the high preverbal subject position (spec-IP or spec-TP),
from which they could optionally be lowered to the same structural
position as that assigned in Belletti’s 1988 theory to postverbal definite
unaccusative subjects (see example (2) above).
In a recent updating of the above analysis, Belletti (1999, pp. 34–5)
has continued to argue that indefinite postverbal unaccusative subjects
originate (and remain) in the base position assigned to direct objects
The Ergative Analysis and the Unaccusative Hypothesis 3
The Ergative Analysis stems primarily from the work of Burzio (1981,
1986), although partly overlapping ideas can be found in the work
of previous authors (for example, Stowell 1978, Fiengo 1974). It
was originally inspired (at least in its post-1970s form) by the Unaccusa-
tive Hypothesis (Perlmutter 1978, p. 160; 1989, pp. 64–7) of Relational
Grammar (RG). Within the latter framework, grammatical functions
or relations are assigned numbers, with 1, for example, corresponding
to the subject and 2 corresponding to the direct object. Such numbers
were originally used to label arcs in diagrams that represented the
relations in question, but the ‘arc’ terminology has since fallen
into disuse – it will however be used here, given the need to refer to
certain seminal RG papers that are couched within the older
terminology.
6 Unaccusative Verbs in Romance Languages
As was mentioned earlier, the aim of this book is to question the utility
of the unaccusative–unergative distinction in terms of the analysis of
properties such as those enumerated as 1 to 5 in 1.1. Obviously, this
objective can only be advanced through consideration of the nature of
these properties on an individual basis, and that programme forms the
subject matter of the main chapters of this book. On the other hand,
certain general anomalies can be identified at this stage within the
basic assumption that unaccusative subjects are ‘deep’ direct objects (or
initial 2s) in a syntactic sense.
1.3.1 q -marking
The first such anomaly relates to a standard subject–object asymmetry
in q -marking. Since Chomsky (1986a, pp. 59–60), it has usually been
assumed that subject q -role is assigned not by the verb in isolation but
compositionally by the VP as a whole. For example, in (8) and (9) below
it is the verb plus its complement, and not merely the verb, that deter-
mines the q -role assigned to John (agent in (8), but not in (9)):
This restriction has been accounted for under the assumption that the
direct object position is not sufficiently ‘high’ in the clause structure
to c-command into the adjunct.16 In contrast, the preverbal subject in
(13) below occupies a high position and thus control into the adjunct
is unproblematic:
(13) The policei arrested two men [without PROi reading them their
rights].
would expect them not to have the ‘high’ control capability associated
with preverbal subjects.17 However, as noted by Chomsky (1995b,
p. 274) in connection with the Italian example below, postverbal unac-
cusative subjects behave like ordinary preverbal subjects in this
respect:18
Moreover older forms of French are identical in this respect, as the fif-
teenth-century example below illustrates:
(21) ung chapellet d’amour dont present a esté faict par une dame
a deux amoureux
‘a rosary of which a present was made by a lady to two lovers’
(Arr, 34) 21
Contreras accounted for the above disparity by claiming that the post-
verbal passive subject occupied a position that was not properly gov-
erned, hence the empty quantifier he believed to be present within the
subject phrase was not properly governed either (assuming, in addition,
that the empty quantifier was not governed from within the subject
phrase). The ungrammaticality of (23) thus resulted from a violation
of the Empty Category Principle.
Notice, however, that the same principle has been appealed to in
order to account for the distribution of partitive clitics such as Italian
The Ergative Analysis and the Unaccusative Hypothesis 13
ne (see, for example, Cinque 1990a, pp. 5–6; 1990b, p. 69), which sys-
tematically is possible with passives:
Thus the contrast between (22) and (23) remains mysterious. However,
under the assumption that passives and unaccusatives receive parallel
syntactic analyses, there should be no contrast at all.
14 Unaccusative Verbs in Romance Languages
It has often been assumed (following Rizzi 1982) that the COMP-
trace immunity illustrated by sentences such as (29) results from the
fact that the wh-extracted subject is in reality a postverbal subject, as
schematized in (30): 24
The comparison with English (27) indicates that the postverbal extrac-
tion position in (29) and (31) is analogous to an object position.
However, both unergative and unaccusative subjects can occupy this
position. This suggests that it is postverbal placement rather than an
association with an unaccusative verb that enables object-like behav-
iour in a subject. The implication of Chapters 2, 3 and 4 will be that
those object-like behaviour patterns that have been identified as unac-
cusative diagnostics do not in reality constitute an exception to this
principle.
1.4 Conclusion
2.1 Introduction
17
18 Unaccusative Verbs in Romance Languages
As was mentioned above, it has often been claimed that expletive in-
version is incompatible with unergative syntax. The basic theoretical
method for deriving this exclusion involves assuming that the expletive
and the verb’s subject (the would-be associate) are in competition for
the same structural position. In Burzio’s classic account (1986:119–77),
this strategy was implemented through the claim that overt expletives
were inserted into the high preverbal subject position at D-structure.
This immediately excluded unergatives, given the then prevalent
assumption that unergative subjects were base-generated in exactly that
position.
Obviously enough, however, Burzio’s account is insufficient under
the more recent VP-internal subject hypothesis, according to which
unergative subjects are base-generated at a position that is lower than
the preverbal subject position.1 Under any reasonable implementation
of this hypothesis, and given the surface postverbal position of the
associate, a would-be unergative associate would never need to occupy
the high position assigned to the expletive. A mechanism is of course
Expletive Inversion 19
The importance of the frame ‘verb . . . PP’ lay in the fact that Belletti
assumed it to diagnose direct object position. Thus the apparent
definiteness effects illustrated by (19), (21), (23) and (25) indicated, in
Belletti’s view, that while indefinite unaccusative/passive postverbal
subjects remained in direct object position (where they were assumed
to originate), their definite and universal counterparts could not. In
fact, these were analysed as having to move to a VP-adjoined position
where they were said to receive nominative case.6 This implied that the
22 Unaccusative Verbs in Romance Languages
However, the problem attaching to the cada phrases in (29) and (30)
can be reproduced when the verb is unergative, where partitive case is
ruled out on a priori grounds:
We find the same pattern with French and English unergatives in exple-
tive inversion. For example, voler ‘fly’ and voyager ‘travel’ are compatible
in principle with indefinite associates, as is shown by the earlier exam-
ples (9) and (12), repeated below:
Similarly, English vote is possible in cases such as (35), but not in cases
such as (36):
24 Unaccusative Verbs in Romance Languages
(35) There voted more women than men in the last election.
Proper name:
Definite description:
Demonstrative phrase:
(43) And when she was gone there appeared Apollo himself. (Alfred
Church)
(44) At that moment there arrived the officer of the guard carrying
his parcel. (Andrew Lang)
(45) Twenty-eight years later there occurred the Cape Ann
earthquake.
Thus the data from English and French provide no obvious grounds
for positing the existence of partitive case (in Belletti’s sense). More-
over, in the light of (27) and (28), reproduced below, the data from
Spanish point in the same direction:
From this perspective, the Italian data adduced by Belletti now look
somewhat isolated. The question that arises at this point is whether the
behaviour of definite postverbal subjects in Italian might be interpreted
in a way other than that indicated by Belletti.
With this question in mind, it is significant that Pinto (1997:49) has
attributed the deviancy of example (19), reproduced below, to the fact
that the definite description is insufficiently specific to secure unique-
ness of reference:15
Belletti (1988, p. 16) in fact recognized that there was no defi niteness
effect when the reference was to types of object that were clearly
unique, as in the example below, where there is a background assump-
tion that people usually have one and only one wallet:
Belletti claimed that, despite its normal indefinite import, partitive case
was licensed on the definite postverbal subject in the above type of
context, precisely because of the overtly unique reference (which she
termed the ‘uniqueness interpretation’). However, given that, as Russell
(1905) famously showed, the whole point of a singular definite descrip-
tion is that it has (or purports to have) unique reference, it is hard to
see on what grounds Belletti can allow that partitive case can be
licensed on uniquely referring definites and yet claim that it cannot be
licensed on definites in general.17 All felicitous singular definites are
uniquely referring and thus all such definites in principle fall under
some form of uniqueness interpretation.
Expletive Inversion 29
The above data indicate that focus does not systematically identify
definite postverbal subjects. Therefore there is no principled motivation
for assuming that definite postverbal subjects in Italian occupy the
specifier of a focus phrase. By the same token the Italian data, like the
English, French and Spanish data, do not appear to provide a principled
basis for assuming that indefinite and definite postverbal subjects
occupy distinct structural positions. We are thus left without a princi-
pled reason for assuming that the indefinites are licensed in a different
fashion from the definites. This finding in itself appears to deprive the
partitive case hypothesis of independent motivation. However, as just
noted, that hypothesis is additionally challenged by the felicitous
occurrence of definite postverbal subjects in syntactic contexts or
frames in which they are predicted to be infelicitous (or even ungram-
matical under the original theory) and also by the availability of Pinto’s
alternative explanation for the deviancies highlighted by Belletti.18
Given all of this, it does not seem unwarranted to conclude that
Belletti’s partitive case hypothesis is not motivated by the available
evidence.
30 Unaccusative Verbs in Romance Languages
With this conceptual obstacle removed, we are free to suppose that the
licensing mechanism (whatever this may be) that suffices for the uner-
gative associates in sentences such as (9) to (13), reproduced below,
represents the general case, that is, that it suffices also for unaccusative
and passive associates, definite or indefinite:
Here, the associate has the same pragmatic function as the postverbal
subject Miguel in the reply in the Spanish example below:
The above sentence can, and indeed is likely to, have the same (presen-
tational) information structure as the equivalent sentence involving
expletive il:
Thus what examples such as (40) appear to show is that expletive inver-
sion will tolerate narrow focus on the associate, but only if the relevant
verb or verbal construction has presentational capability to begin with.
From this perspective, the fact that narrow focus under expletive inver-
sion is attested primarily with passives (in French at least) results
from the coincidence of two circumstances, viz. (i) that expletive il +
passive has become an established presentational collocation and (ii)
that, notwithstanding this, postverbal passive subjects need not be
presentational.
Here, the verb naturally forms an informational unit with the clause-
final adjunct. Specifically, the verb and the adjunct are subsumed
within the comment component of a topic–comment structure. Given
that this possibility is systematically unavailable in the inverted coun-
terparts (41) and (42), it can be surmised that the principal effect of
expletive inversion (in this type of case) is to force an informational
division that segregates the verb and the associate from the clause-final
adjunct.
Now although this result is not achieved through the associate acquir-
ing prosodic prominence itself, there is nevertheless a similarity with
the presentational and narrow-focus (on the associate) structures,
because in all three cases prominence is withheld from the verb. In
other words, assuming focus to be essentially a reflex of accentuation,
the paradigm illustrated by (41) and (42) is simply a third instantiation
of the ‘unfocusable verb’ information structure identified at the begin-
ning of this section.
2.6 Conclusion
3.1 Introduction
39
40 Unaccusative Verbs in Romance Languages
As will be pointed out later, there is ample evidence that partitive cliti-
cization does not systematically distinguish between unaccusative and
unergative subjects. However, before proceeding to an empirical discus-
sion, it is worth considering whether there is any a priori motivation
for the assumption that postverbal unergative subject position will not
be a proper extraction site. Accordingly, I will review the various theo-
retical arguments that have attempted to link partitive cliticization to
the ‘deep-object’ analysis of unaccusative subjects. The arguments dis-
cussed below were all framed in regard to Italian ne, although it seems
that the assumption has generally been that analogous considerations
apply to Catalan en (see, for example, Picallo 1984, pp. 98–101).
cliticization and unaccusativity. They noted first of all that this opera-
tion was not possible in respect of ‘adverbial’ noun phrases, as is shown
by the ungrammaticality of the (b) examples below (from Belletti and
Rizzi 1981, pp. 126, 128):
VS. Within this newer model, Belletti assimilates (p. 37, n. 52) the appar-
ent failure of partitive cliticization from unergative postverbal subject
position to the class of effects that fall under the Condition on Extrac-
tion Domains (CED).4 The latter is a general condition restricting the
class of constituents from which elements may be extracted. It subsumes
more specific constraints such as the Adjunct Condition, which prohib-
its extraction from adjuncts, as illustrated below:
Examples such as those above show that the position occupied by (at
least some) definite postverbal subjects is a proper extraction site.
Presumably, then, such subjects cannot be in spec-FP. But if that is the
case, there seems to be no principled reason why unergative subjects
must be in spec-FP, given that in Belletti’s framework unergative post-
verbal subjects in effect fall together with defi nite postverbal subjects
(in that neither can be licensed by partitive/inherent case – see 2.3 in
this book).
Partitive Cliticization 43
In fact, native speakers other than linguistics experts are somewhat less
categorical in terms of distinguishing between sentences such as (19)
and (20). But even if we accept Longobardi’s evaluation, the contrast
in question appears not to be capable of being generalized to all verbs
in the unaccusative class. Thus the acceptability differential between
(19) and (20) appears to decrease if other unaccusatives such as
sbagliarsi ‘be mistaken’, cadere ‘fall’, arrivare ‘arrive’, lamentarsi
‘complain’ and so on, are substituted for scomparire in (20).
Moreover, the alleged sensitivity of postverbal wh-extraction to the
unaccusative–unergative distinction would be a peculiarly Italian phe-
nomenon. Spanish, for example, exhibits no comparable sensitivity,
and indeed cases exist in which, if anything, the data point in the
opposite direction from that claimed for Italian. For example, in the
pair of Spanish sentences below, wh-extraction from the postverbal
subject of unaccusative llegar ‘arrive’ is marginally less acceptable than
Partitive Cliticization 45
It turns out, then, that if we abstract away from the Italian data,
which themselves are not particularly conclusive, it is difficult to find
evidence consistently indicating that the unaccusative–unergative
distinction plays a role in CED effects.
To sum up: several problems can be pointed to within Belletti’s
updated thoery. First, there is the possibility of extracting ne from pre-
cisely the structural position that Belletti claims is not a possible extrac-
tion site. Secondly, the basis proposed by Belletti for assigning distinct
structural positions to unaccusative and unergative postverbal subjects
is empirically problematic, given that the criterion of narrow focus on
the postverbal subject does not systematically distinguish between
unaccusatives and unergatives. Finally, the evidence that other forms
46 Unaccusative Verbs in Romance Languages
Longobardi then argued (pp. 73–80) that unergative verbs did not
govern into the specifier of a postverbal subject, while unaccusatives
did. Accordingly, and given the further assumption (p. 73) that a trace
in the specifier of an NP required proper government by an external
head, ne-extraction (as well as wh-extraction) from a postverbal unerga-
tive subject NP was predicted to be ungrammatical, while such extrac-
tion from a postverbal unaccusative subject NP was predicted to be
grammatical.
Notice, however, that Longobardi’s theory requires acceptance of the
claim that an unaccusative–unergative asymmetry exists in terms of
the possibility of governing into the specifier of a postverbal subject.
In support of this position Longobardi adduced evidence relating to
wh-extraction and ne-extraction from NP subjects, as well as evidence
relating to successive cyclic extraction from infinitival clauses embed-
ded within sentential arguments. Now from the present perspective,
that is, in terms of determining whether Longobardi’s theory provides
a priori motivation for assuming postverbal unergative subject position
to be an impossible site for ne-extraction, the ne-extraction data them-
selves cannot be adduced without an obvious circularity. Moreover, as
noted in 3.2.2, the data pertaining to wh-extraction from NP/DP sub-
jects are somewhat inconclusive, and even confl icting if other Romance
languages are considered. This leaves us with just the evidence pertain-
ing to successive cyclic extraction from sentential arguments, discussed
by Longobardi in section 5 of his paper.
There he considered cases such as (33) below, in which a wh-phrase
(a cui) is extracted from a preverbal (infinitival) subject but the resul-
tant CED effect is much weaker than might be expected, with the
consequence that (33) was evaluated by Longobardi as being merely ‘?’,
as opposed to fully ungrammatical:
48 Unaccusative Verbs in Romance Languages
(33) Gianni, a cui temo che parlare sarà difficile (Longobardi 1991,
p. 74)
‘Gianni, to whom I fear that talking will be difficult’
(34) Gianni, con cui le è dispiaciuto che parlare non fosse facile
‘Gianni, with whom it displeased her that talking was not
easy’
(35) Gianni, con cui l’ha impressionata che parlare non fosse
possibile
‘Gianni, with whom it impressed her that talking was not
possible’
Clause (ii) of the above condition relates to the fact that the source of
ne must be postverbal. Clause (i) is the key clause for present purposes,
as it is this clause that alludes to the concept of ‘2hood’ (that is, direct
objecthood), by means of which Perlmutter distinguished between
unaccusatives and unergatives, with the former but not the latter
heading a 2-arc in the initial stratum (see the brief overview of Perl-
mutter’s Unaccuative Hypothesis in 1.2 of this book). Given this dis-
tinction, the above condition on partitive cliticization predicts that it
can occur with unaccusative postverbal subjects but not with unerga-
tive postverbal subjects.
However, it is far from clear that the concept of 2hood has any real
content. Consider first of all Perlmutter’s argument for assuming that
unaccusative subjects were 2s (that is, headed a 2-arc in some stratum),
which can be summarized as follows (based on Perlmutter 1989,
pp. 67–83):
1. The Italian intransitive verbs can be split into two classes depending
on whether they select auxiliary essere or avere in the perfect.
2. The subjects of verbs in the essere-selecting class pattern with transi-
tive direct objects in terms of (a) partitive cliticization and (b)
Partitive Cliticization 53
3.2.6 C-command
As mentioned in 1.1, Baltin (2001, p. 241) has argued that the distribu-
tion of partitive ne results from the need for it to c-command the trace
it leaves behind at the extraction site. He states that ‘the requirement
that ne c-command the nominal which it modifies is simply Fiengo’s
(1974, 1977) proper binding requirement on traces, assuming that ne
has moved out of the nominal.’ However, most modern theories assume
that the verb in null-subject languages like Italian raises to a very high
54 Unaccusative Verbs in Romance Languages
position in the sentence (T, for instance), and certainly one from which
a moved clitic could c-command its own trace in a postverbal unerga-
tive subject. Thus an analysis in terms of c-command would not, in
the present theoretical climate, secure the desired exclusion of ne-
extraction from postverbal unergative subject position.
In any case, Belletti and Rizzi (1981, p. 127) originally entertained
an explanation in terms of c-command (albeit under the looser ‘maximal
projection’ definition) and concluded that it accounted only for the
data relating to preverbal position. In the example below (from Belletti
and Rizzi 1981, p. 126), the phrase tre settimane is in VP, hence an
extracted clitic would c-command its own trace, but nevertheless the
corresponding sentence with ne is ungrammatical:
Finally, Catalan (43) and (44), reproduced below, can plausibly receive
a similar analysis to (48), given the implicit comparison in these two
cases between the quantity specified in the sequence in italics and a
quantity specified previously:
The fact that camminare (in the favoured sense) is an agentive activity
verb appears to be relevant, as does the absence of a context. As will
be argued in 3.4 below, partitive cliticization from a subject requires
the associated verb to have a presentational type of occurrence. More-
over, of all the semantic classes of verb, agentive activity terms are the
most difficult to reconcile with this type of use. Specifically, it seems
to be the case that presentational occurrences of agentive activity verbs
systematically involve a linguistic context that creates a weak existen-
tial interpretation of the verb–subject complex, and in so doing
backgrounds the verb’s usual agentive meaning, as in the examples
below:
Thus the evidence from French indicates that the possibility for parti-
tive cliticization from non-objects is a function of the presentational
capability of the relevant verb. An analysis of null subject languages
such as Italian and Catalan that assumes a parallel state of affairs to
the French situation (but with partitive cliticization from intransitive
subjects having a higher frequency in these languages due to the
greater productivity of null subject-style inversion in comparison to
62 Unaccusative Verbs in Romance Languages
What this suggests is that transitive direct objects are in essentially the
same situation, from the point of view of information structure, as
expletive associates, given that the effect of expletive inversion is to
withhold focus from the verb. If this is correct, the general transpar-
ency of (indefinite) direct objects to partitive cliticization falls under
the same generalization as that put forward in connection with intran-
sitive subjects, viz. that partitive cliticization diagnoses presentational
capability. In this way, the parallel between unaccusatives, passives and
66 Unaccusative Verbs in Romance Languages
Achievement/accomplishment terms:
(89) Alle 11.00 si sono schiantati due treni merci incanalati sullo
stesso binario.
‘At 11 a.m. two freight trains on the same line collided.’
State terms:
3.5 Conclusion
not least because not all postverbal unergative subjects are in fact
focused.
Longobardi argued that partitive cliticization involved successive
cyclic movement and that such movement was not possible in the case
of unergative subjects because unergative verbs did not govern into the
specifier of their subject. While this proposal obviates several of the
problems inherent in other theories, it has a major weakness, viz. that
it is difficult, if not impossible, to provide compelling independent
motivation for the assumption of an unergative–unaccusative asym-
metry in terms of the capability to govern into the specifier of a (post-
verbal) subject.
Of the remaining accounts, Burzio’s and Perlmutter’s theories are
ultimately stipulative and are thus immediately vulnerable to adverse
empirical data, while any attempt to explain the distribution of parti-
tive cliticization in terms of Fiengo’s proper binding requirement on
traces, as suggested by Baltin, requires the adoption of an outmoded
theoretical framework.
These findings, together with the fact that postverbal subjects of
certain unergative verbs routinely allow partitive cliticization, indicate
that the distribution of the latter phenomenon is unlikely to be syn-
tactically conditioned. In fact, it has been argued in this chapter that
the distributional patterns are determined by tendencies pertaining to
information structure. This is suggested initially by the fact that parti-
tive cliticization from non-objects in French is largely restricted to
expletive inversion, which is available only to verbs or constructions
that are capable of being used presentationally. On the basis of impor-
tant similarities between the French and Italian/Catalan situation, it
was argued in the text that partitive cliticization from the argument of
an intransitive verb is subject in all three languages to essentially the
same restriction in terms of information structure, the only significant
cross-linguistic disparity being that presentational inversion in the null
subject languages is significantly more productive than expletive inver-
sion in French.
Building on some of the ideas introduced in Chapter 2, it has also
been argued that there is a consonance between object q -role and the
capability for an argument to have a ‘presented’ function. This accounts
for the way in which both passive subjects and transitive direct objects
pattern with subjects of presentational unaccusatives (and also with
subjects of presentational unergatives).
Finally, it was observed that all intransitives other than agentive
activity terms are likely to have a natural presentational capability and
Partitive Cliticization 69
4.1 Introduction
70
Bare Subjects 71
The above distribution has been claimed to result from the existence
of a proper or lexical government requirement on bare arguments,
under the assumption that they contain an empty determiner or quan-
tifier.3 The general principle from which this requirement is assumed
to derive is some form of the Empty Category Principle (ECP), a Gov-
ernment Binding construct requiring empty categories to have proper
governors.4 The original articulation of this view is due to Contreras
(1986), although Longobardi’s (1994) restatement within the frame-
work of the DP (determiner phrase) hypothesis has been highly influ-
ential. We can note in particular the conditions Longobardi assumed
(pp. 617–18) empty determiners to be subject to:
(a) They are restricted to plural or mass head nouns like several other
determiners.
(b) They are subject to a lexical government requirement like other
empty heads.
(c) They receive an indefinite interpretation corresponding to an exis-
tential quantifier unspecified for number and taking the narrowest
possible scope (default existential).
In addition, she noted that bare nouns could not appear as the small
clause subject in an ECM (Exceptional Case Marking) construction:
Given that it is the small clause predicate and not the matrix verb that
is responsible for q -marking the small clause subject, the ungrammati-
cality of (8) and (9) would follow, under Belletti’s analysis, from the
assumption that partitive case licensing requires a q -relation between
licenser and licensee (and indeed, this requirement is implicit in the
characterization of partitive case as ‘inherent’).
The final component of the jigsaw relates to the so-called ‘unerga-
tive–unaccusative alternations’ (Torrego 1989), as in (10) and (11)
below:
tion that unergatives switch to the unaccusative class when used with
a preverbal locative argument.
Torrego did not explicitly motivate, in syntactic terms, the possibility
of unergative-to-unaccusative switches. Conceivably, she may have rea-
soned that the locative argument (which she analysed as a ‘locative
subject’) was q -marked in the high preverbal subject position (spec-IP
or spec-TP) and thus could not co-occur with an unergative subject,
given that the latter, under 1980s assumptions, was presumed to be q -
marked in that very position.
The fact that relativization is possible with plural and mass bare nouns,
as well as with nouns that have an overt determiner, provides an
empirical link between each of these types of item. Longobardi’s obser-
vation in fact underlies one of the main empirical arguments in favour
of the so-called DP hypothesis (see Bernstein 2001, p. 543).
On the other hand, there is no compelling reason to assume that
some form of lexical government requirement would necessarily exclude
bare nouns from (postverbal) unergative subject position. Indeed the
usual assumption, pace Contreras (1986), has been that postverbal
unergative subjects are governed by the verb, and hence are lexically
governed. This follows from certain considerations pertaining to
the so-called COMP-trace effect, that is, a prohibition on wh-extraction
of the subject across an overt complementizer, as illustrated by
(13):
In fact, according to the example below from López Díaz (1996, p. 131),
an accusative clitic is possible also in Italian (pace Belletti):
Thus there are good reasons for rejecting the alleged link between
bare nouns and partitive case. From that perspective, we can regard
clitics like ne as being unspecified in terms of case, and the prohibition
on bare nouns occurring as ECM subjects, illustrated by examples (8)
and (9), can be attributed to the fact that the relevant ECM verbs are
individual-level predicates in Carlson’s (1977) sense. Bare arguments of
such predicates must have universal reference, which in principle is not
possible with (unmodified) bare nouns in Romance (see 4.4 below).
We find, then, that there is no compelling a priori motivation for the
assumption that postverbal unergative subjects cannot be bare. In addi-
tion it is quite clear, once all the descriptive facts are in, that unergative
verbs do admit bare postverbal subjects. I give a selection of examples
below. The unergative status of the verbs in question is indicated by a
number of facts: (i) their translation equivalents in Italian and French
are not assigned the ‘be’ perfect auxiliary (at least not in the senses
illustrated below), (ii) their meanings fall within the unergative spec-
trum on Sorace’s Auxiliary Selection Hierarchy (Sorace 2000), and (iii)
they cannot appear in participial absolutes.
In fact not just unergatives can appear with bare subjects but also
transitives:
If, in line with the view taken in 4.3.1, the position occupied by the
bare noun coches in the above sentence is inaccessible to quantification
(at least in the manner envisaged in schema (39)), it might seem sur-
prising that coches can serve as the antecedent for clitic los. This would
follow from the assumption that los in (44) is analogous to los in (45)
below:
Here the second conjunct amounts in effect to Pedro vende los coches
que Jorge repara ‘Pedro sells the cars that Jorge repairs’. Nevertheless,
as with (44), no genuine cross-reference into repara coches need be
assumed. The first conjunct pragmatically implicates the existence of
certain cars, and it is to these cars that los in the second conjunct
refers. Roughly speaking, then, sentence (47) has the logical structure
shown below, where the letter ‘p’ stands for the disposition sentence
Jorge repara coches, Fs are cars repaired by Jorge and Gs are cars sold by
Pedro:
Notice that the proposed analysis does not require any of the terms
within the first conjunct of sentence (47) to be exposed for cross-
referential purposes, whence the possibility of representing that fi rst
conjunct by the unanalysed sentence letter ‘p’ in schema (48). In other
words, as a consequence of the part played by pragmatics in determin-
ing the overall meaning of sentence (47), we can simultaneously treat
the second conjunct as a universal quantification about cars repaired
by Jorge and maintain that no direct referential connection exists
between that second conjunct and the bare noun coches in the first
conjunct.11
Therefore, scrutiny of the apparently adverse cases (44) and (47)
indicates that the possibility of a pronoun cross-referring into a verb +
bare noun complex is apparent rather than real. The Quinean analysis
adopted in 4.3.1 is thus undamaged by such cases.
An additional apparently adverse case is represented by event-
describing sentences that involve bare nouns. In his famous paper on
bare plurals, Carlson raises precisely this issue (1977, p. 429), in con-
nection with the two English sentences shown below:
According to Carlson, in sentences like (49) ‘it seems that a group really
is, in some sense, being set up and referred to. Otherwise, we would
simply have no understanding of why [(50)] gives us such a nice para-
Bare Subjects 83
The deviancy of sentence (54) can then be traced to the fact that unos/
unas represents the overt expression of a quantifier that calls for the
widest possible scope. Accordingly, were unas hormigas to receive its
customary analysis, the quantificational structure of sentence (54)
would be as in schema (56) below, where a is the hole in question, Fs
are ants, Gs are times within a certain three-hour period and Hs are
ordered triples consisting of a location, an individual that exited it and
the time at which that exit was made:
4.3.3.2 Locatives
An analogous situation exists in respect of sentences involving locative
phrases, as is illustrated by the paradigm below:
From the logical point of view, sentence (61) has the structure of a
standard Aristotelian universal affirmative (‘All Fs are Gs’), and so can
be represented by the schema below, with Fs understood as northern
provinces and Gs as places inhabited by wolves:
The reason for the unproblematic status of (65) is that unos lobos can
receive its customary quantificational analysis without the containing
sentence thereby coming to have a meaning that is pragmatically
deviant. Thus sentence (65) is unproblematically assigned the earlier
schema (39), reproduced below, with a understood as the forest in ques-
tion, Fs as wolves, and Gs as ordered pairs of places and individuals
that live in them:
Sentence (66) makes a statement about many African wars, viz. that
they involve children, whereas (67) says something about one or more
individual children, viz. that they are involved in many African wars.
The latter assertion is pragmatically deviant, whence the acceptability
differential between (66) and (67).
The deviancy of the above (b) sentences parallels that which would
attach to a null subject sentence such as Está roto ‘(It) is broken’ if it
were uttered in a context in which the identity of the subject could not
be determined. In both types of case we have a predicate but no iden-
tifiable argument, hence no assertion is made to which a truth value
could be assigned.
Now while the need for verb + bare subject complexes to be contex-
tualized applies regardless of whether the verb is unaccusative or uner-
Bare Subjects 93
gative, it may sometimes seem to apply less in the former case than in
the latter, and this circumstance may explain why a correlation has
so often been assumed to exist between bare subject compatibility
and unaccusativity (see also note 18 for an additional contributory
factor).
As a first step towards seeing how this appearance of an asymmetry
can arise, a contrast can be drawn between the argument structure
associated with agentive activity terms (prototypical unergative verbs)
and that associated with verbs of directed motion (prototypical unac-
cusatives). Outside of the marked case in which the subject is bare, the
use of an agentive activity term does not generally require the deter-
mination of any argument other than the subject and so we are not in
any particular way attuned to such a possibility. In contrast, verbs of
directed motion always involve a locative argument, which may be
explicit or implicit. In the latter case, we tacitly assume that the argu-
ment can be recovered from the context. For example, if we hear Pedro
entró ‘Pedro entered’, we are immediately attuned to the need for there
to be a place that Pedro entered, whereas if we hear Pedro cantó ‘Pedro
sang’ we are not immediately attuned to the need for there to be a place
that Pedro sang in. This in effect means that we envisage an appropriate
non-subject argument more readily for directed motion verbs than we
do for agentive activity verbs.
This latter circumstance, in the context of verb + bare subject com-
plexes, can result in an uneven appreciation of the deviancies resulting
from the non-satisfaction of the argument requirement alluded to at
the beginning of this section. For example, while both (80) and (81)
below are deviant if uttered completely out of context, the natural
assumption that a verb like ir ‘go’ must have a locative argument is
likely to ensure that (80) will elicit a more favourable acceptability
judgment than (81) if the two sentences are put to a linguistic
informant:
And morir ‘die’ is similar, although in the example below the implicit
argument is not determined deictically, and so (83) has the attenuated
deviancy that characterizes sentences such as (80):
The universal meaning attaching to the bare subjects ravens and gold
in (92) and (93) is demonstrated by the fact that they do not support
substitutions by hyperonyms (that is, nouns of which they are hypo-
nyms) and so are analogous to the corresponding universally quanti-
fied subjects all ravens and all gold. For example, bird is a hyperonym of
raven and (92) does not logically imply Birds are black (just as All ravens
are black does not logically imply All birds are black). Similarly, metal is
a hyperonym of gold and (93) does not logically imply Metal is yellow
(just as All gold is yellow does not logically imply All metal is yellow). In
contrast, the non-universal bare subjects in (94) and (95) do support
substitutions of this kind. Thus (94) logically implies People are shouting,
for example, and (95) implies for example, Liquid is leaking from the
cistern.
Bare Subjects 97
(96) Los cuervos son negros. (But not *Cuervos son negros.)
‘Ravens are black.’
(97) El oro es amarillo. (But not *Oro es amarillo.)
‘Gold is yellow.’
We can now account for the contrasts between (88) and (89) and
between (90) and (91), all of which are reproduced below:
The original examples (89) and (91) simply represent the bare subject
option.
Notice that, under this latter option, an argument must be deter-
mined for the logically unstructured verb–subject complex, in accor-
dance with the view outlined in 4.3.3.5. With some stative predicates,
it may be common for the argument to be determined by an implicit
reference to the here and now. Such was the case of faltar in example
(76), reproduced below:
The deviancy of the (b) sentences appears to stem from the fact that
the nouns hombres and espectadores are insufficiently informative or
100 Unaccusative Verbs in Romance Languages
4.6 Conclusion
5.1 Introduction
All the Romance languages have at some point in their history exhib-
ited a split perfect auxiliary system, with transitive verbs generally
taking a ‘have’ verb (normally descended from Latin habere) and some
intransitives taking a ‘be’ verb (normally related to Latin esse). For
convenience the letters ‘A’ and ‘E’ are used here to denote ‘have’ and
‘be’ verbs respectively.
In the standardized languages of the Iberian Peninsula (Spanish,
Catalan, Portuguese and Galician), the E perfect auxiliary has been
entirely displaced by the A auxiliary, although it survives in some
Catalan and Aragonese dialects. On the other hand, the E auxiliary
survives in standard French and is highly productive in standard
Italian.1 In both of these languages E is assigned uniformly to all
reflexives (including middle and impersonal reflexive constructions).
As regards non-reflexive verbs, E is assigned to a narrowly defined
group of French verbs but to a much larger class in Italian. Table 5.1
gives a reasonably complete list of the relevant French verbs,
while Table 5.2 gives illustrations for Italian. The semantic classification
given in the latter table is for convenience only and nothing depends
on it.
As far as perfect auxiliary selection is concerned, the Ergative Analy-
sis and Unaccusative Hypothesis are inspired primarily by the Italian
situation. Constructions that select perfect auxiliary E other than
passives, raising predicates (for example, parere ‘seem’ and risultare ‘turn
out’) and reflexives are assumed to be unaccusative. E assignment is
thus identified as the reflex of a syntactic property that is common to
all of these constructions.
103
104 Unaccusative Verbs in Romance Languages
aller go
apparaître appear
arriver arrive
descendre* go down
devenir become
entrer enter
intervenir intervene
monter* go up
mourir die
naître be born
partir depart
rentrer return home
rester stay
retourner return
revenir come back
sortir go out
survenir appear/happen
tomber fall
venir come
Thus the relation envisaged by Burzio for the type of case illustrated
by (5) and (6) is essentially analogous to the relation in (7) between
there and its associate a great commotion. The only difference is that the
108 Unaccusative Verbs in Romance Languages
From the present perspective, the above accounts can be ruled out
as evidence in favour of the ergative analysis of unaccusative verbs.
Belletti’s and Kayne’s accounts are essentially stipulative in that there
is no independent motivation for the assumed correlation between
unaccusative–unergative, on the one hand, and E versus A on the
other.11 Cocchi’s theory is less stipulative, but it immediately faces
problems of empirical adequacy.
Firstly, there is an analogue to the problem highlighted in section
5.2 for Burzio’s binding-related theory. Thus consider the occurrence
of perfect auxiliary E in examples such as the following:
(11) Sono entrati due ladri dalla finestra. (From Belletti 2001,
(34c))
‘Two thieves came in through the window.’
Inherent reflexives:
Ergative reflexives:
Transitive reflexives:
(20) *Se ne sono vestiti tre. (Reinhart and Siloni 2004, p. 172)18
‘Three of them have got dressed.’
Table 5.4 Revised taxonomy for French and Italian auxiliary assignment
mismatches
examples such as (24) below (= (49) from chapter 3) and (20) from
section 5.4 above (reproduced below):
(28) Questo segmento è cresciuto per quasi tutti gli anni’ 90.
‘This sector grew throughout almost the entire 90s.’
(33) Siamo saliti per un’ora e alla fine era come essere in paradiso.
‘We climbed for an hour and at the end it was like being in
heaven.’
This again falsifies the claim that activity terms always select A.
Conversely to the above cases, A can be assigned to non-activity
terms. A assignment to achievement or accomplishment terms is illus-
trated below:
Verbs in class (i) are [D+T]. Some of the verbs in class (ii) (for example,
morire ‘die’) are also [D+T], while others (e.g. cambiare ‘change’) are
[D−T]. Verbs in class (iii) are [St−D]. Verbs in class (iv) are [St].
Bentley and Eythórsson’s two generalizations about the semantic
distribution of E and A assignment (to intransitives) in Italian can then
be characterized as follows:
‘burst’, disparaître ‘disappear’, etc. This latter class is all the larger if
agentive verbs such as capituler ‘capitulate’, démissionner ‘resign’ and so
on are also deemed to be [D+T]. In addition, there is a counterpart to
the problem noted earlier for Van Valin’s account in relation to verbs
such as salire (see example (31)). Thus E is assigned equally to [−T] and
[+T] occurrences of verbs like descendre ‘go down’ and monter ‘go up’:
5.5.3 Subject-affectedness
The semantic accounts discussed in 5.5.1 and 5.5.2 focus on the nature
of the situation described by the verb. An alternative approach, in
which the focus is on the role of the subject, has been put forward by
Chierchia (2004 [originally written 1989]). Under the assumption that
auxiliaries were ‘property modifiers’28 and hence could be sensitive to
the meaning associated with what they modified, he argued (2004:46–
7) that the choice of Italian essere versus avere was sensitive to what he
termed ‘subject affectedness’, where the subjecting-affecting operations
were the following: P(assive), E(xpletive) and R(eflexive). Below, I
describe each of these operations in turn.
Note that, in order to make sense of Chierchia’s notation, it needs to
be understood that he takes non-linguistic items such as properties and
relations to be primitives within his system. Linguistic sentences are
then analysed as involving what he calls the ‘predication relation’,
represented by the symbol ∪, which takes a property as its input and
yields a propositional function as its output. For example, the sentence
‘John runs’ is analysed as in (51) below, where ‘run’ stands for the
property of running (Chierchia 2004, p. 26):
∪
(51) run(John)
∪
(52) [P(see)](x) ↔ ∃y∪ [see(x)](y)
∪
(55) [R(wash)](x) ↔ ∪ [wash(x)](x)
∪
(58) [E([affondare(la barca)](la barca))]⊥
However, the difficulty lies in the fact that there is no natural division
between cases that should be treated like affondare and cases that
should not. Moreover, it is doubtful that any such dividing line can
plausibly be made to coincide with the distribution of the perfect aux-
iliaries. To take a few examples, capitolare in sentence (65) is semanti-
cally like abdicare in the earlier sentence (40), reproduced below, and
unlike intransitive affondare, and yet the pattern of auxiliary assign-
ment completely ignores this state of affairs:
(40) Il re ha abdicato.
‘The king has resigned.’
Conversely, intransitive deviare, bollire and girare in (59), (61) and (63)
are like intransitive affondare and they are unlike abdicare in (40), and
yet auxiliary selection groups them with abdicare.
The implication of Chierchia’s theory is that the pattern of auxiliary
selection can be understood as determining the boundary between the
affondare paradigm and those intransitives that must receive a different
analysis. However, as observed above, an approach based on that
assumption will necessarily be circular.
The various problems identified for the accounts discussed in 5.5 appear
to be illustrative of an inescapable fact,33 viz. that the assignment of
130 Unaccusative Verbs in Romance Languages
5.6.1 E as a copula
As is well known, the paradigm of A perfect forms descends from a
Latin construction involving habere as a full lexical verb. For example,
in venenum praeparatum habebat ‘he had the poison ready’ the unit
venenum praeparatum is plausibly analysed as a small clause, on a par
with similar units in the following examples from modern Italian,
Spanish and French:
132 Unaccusative Verbs in Romance Languages
The A + small clause construction differs from the true perfect con-
struction in several ways. Most significantly from the present perspec-
tive, the participle in the small clause functions as a predicative adjective
and so, where extraneous factors do not prevent this, the small clause
can usually be expanded to a full copular construction in which the
tense of the copula replicates that of the A verb:
glance, however, this seems to be ruled out by the fact that periphrastic
forms like mentitus sum are assumed to have occupied the same place
in the deponent paradigm as synthetic forms like cucurri ‘I ran/have
run’ (from currere ‘run’) within the non-deponent paradigm. While
certain specific forms such as mortuus est may have been syntactically
ambiguous, as between the perfect structure (with the meaning ‘died/
has died’) and the copula–adjective construction (with the meaning ‘is
dead’), this duality cannot easily be generalized throughout the depo-
nent perfect. Accordingly the syntactic statuses of the forerunners of
the E and A perfects do not appear, in Classical Latin at least, to have
been analogous.
On the other hand, there are grounds for thinking that towards the
end of the Latin period the deponent perfect underwent a process of
syntactic reanalysis. With the demise of the passive/deponent mor-
phology and the migration of the deponent verbs to the various spoken
Latin conjugations, many of the deponent perfect forms were replaced
by non-standard formulations modelled on the synthetic perfect,
reflexes of which abound in the modern languages. 35 For example men-
titus sum ‘I (have) lied’, was replaced by a form constructed from the
root ment- together with the first person singular synthetic perfect
ending -i(v)i, giving rise ultimately to modern Italian mentii, Spanish
mentí, and so on. Such analogical forms were created also for those
deponent verbs (sequi ‘follow’, mori ‘die’, nasci ‘be born’, and so on)
which in Romance exhibit a deponent-style E perfect. For example,
when mori became assimilated to the -ir(e) conjugation, it both retained
the form corresponding to its periphrastic perfect, which surfaces in
the various Romance languages as è morto, est mort and so on, and also
acquired an analogical synthetic perfect paradigm constructed from
the stem mor- together with the relevant synthetic perfect endings. The
latter analogical paradigm surfaces in Romance as the synthetic pret-
erite (morì, murió, and so on).
Analogous remarks apply mutatis mutandis to the (larger) group of
verbs that are not descended from Classical Latin deponent verbs but
which in spoken Latin or Romance became attracted to the periphrastic
pattern. For example, spoken Latin ven + itum/utum (replacing Classical
ventum) sum yields Romance forms like sono venuto, suis venu and so
forth, but the paradigm of the synthetic perfect (Latin veni, and so on)
persisted into Romance: venni, vine, and so on.
Overall a pattern emerges, involving (i) the retention and generaliza-
tion of the synthetic perfect, and (ii) the reorganization of the member-
ship of the class of intransitives that entered into the periphrastic
134 Unaccusative Verbs in Romance Languages
(79) Fenduz en est mis olifans el gros, Caiuz en est li cristals e lis ors.
(R 2295–6)
‘My oliphant is split at the wide end, the crystal and gold
have/are fallen off.’
(81) L’altra fiata ch’i’ discesi qua giù nel basso inferno, questa roccia
non era ancor cascata. (I XII, 34–6)
‘The other time I came down here to lower Hell, that rock
had/was not yet fallen.’
(82) Los unos eran muertos e los otros fuydos (A [Paris ms.] 1415b)
‘Some were dead and the others had/were fled.’
(83) Anzi ‘mpediva tanto il mio cammino, chi’i’ fui per ritornar più
volte vòlto.
‘It so impeded my path that several times I turned to go back.’
(I I, 36)
(84) hasta que fue la alma de la carne partida (M 318b)
‘until the soul had separated from the flesh’
(85) Poi che mosso fue, intrai per lo cammino alto e silvestro. (I II,
142)
‘After he had moved, I entered along the high and wild
path.’
136 Unaccusative Verbs in Romance Languages
(86) Bis post Numoe regnum Ianus clausus fuit. (Dicc A27)
‘Twice after the reign of Numa the temple of Janus was closed
(that is, unopen).’
(87) María estuvo enferma cuatro veces el año pasado. (Mod. Sp.)
‘María was ill four times last year.’
(88) Quinci fuor quete le lanose gote. (I III, 97)
‘Whereupon the hairy cheeks were quiet.’
(89) E all’improvviso fu buio e freddo. (Mod. It.)
‘And suddenly it was dark and cold.’
(90) hasta que estuvo hundido en el barro (Mod. Sp.)
‘until it was sunk into the mud’
(94) Li reis Marsilie s’en fuit en Sarraguce, Suz un olive est descendut
en l’umbre.’ (R 2570–1)
‘King Marsile flees to Zaragoza, He dismounts under an olive
tree in the shade.’
Perfect Auxiliary Selection 137
(95) A icel jur venent a Sarraguce . . . Li amiralz est issut del calan.
(R 2645–7)
‘They arrive that day at Zaragoza. The emir disembarks from
the lighter.’
For Old Italian, compare mosso fue in the earlier example (85), repro-
duced below, with si mosse in (99) (see also the many examples in
Brambilla Ageno (1964, pp. 200–5):
(85) Poi che mosso fue, intrai per lo cammino alto e silvestro. (I II,
142)
‘After he had moved, I entered along the high and wild
path.’
(99) Allor si mosse, e io li tenni sietro. (I I, 136)
‘Then he moved and I followed him.’
138 Unaccusative Verbs in Romance Languages
Spanish:
Italian: 44
(106) Anzi avevo imparato la sottile arte di presentarmi alle cene già
cenata.
‘In fact I had learned the subtle art of appearing at dinners
already having dined.’
(107) O ci andate già mangiati, solo per vedere la partita . . . oppure
è meglio stare alla larga.
‘You either go there having already eaten, just to watch the
match . . . otherwise it’s best to stay away.’
English:
It is true that unergatives are in general rather less likely than unac-
cusatives to be available in an adjectival capacity. However, unaccusa-
tives can hardly be said to enjoy unfettered productivity in this respect.
Consider, for example, some of the complex semantic and syntactic
restrictions that apply in English:
I assume, then, that the past participle morphology does not block
the externalization of subject q -role and that, accordingy, the proposed
analysis of the E + intransitive past participle construction in early
Romance does not rule out any intransitive verbs on a priori grounds.
On the other hand, given that the earliest attested examples of the
construction almost invariably refer to the state resulting from the
event denoted by the verb (or the predicate as whole), it can be assumed
that access to the construction was limited to those verbs whose past
participles expressed a resultant state, a constraint which would obvi-
ously favour achievements and accomplishments. This point can be
Perfect Auxiliary Selection 141
The indirect perspective onto this ‘initial’ state of affairs that is afforded
by the data from the Middle Ages appears to confirm the basic conjec-
ture just outlined. This can be seen by considering the aspectual classes
of the verbs/predicates that appear in the medieval examples in 5.6.1
above.
(117) El ladrón cató ora que el religioso fuese desviado. (Cal p. 138)
‘The thief waited until the monk had/was turned away.’
Elsewhere, the period at which the verb in question entered the lan-
guage is likely to be relevant. Many of the achievement/accomplish-
ment terms that select A in either Italian or French or both (for example,
words meaning ‘vote’, ‘abdicate’, ‘explode’) are modern or early modern
borrowings (either from written Latin or from other European lan-
guages). These obviously had no ‘inherited’ perfect auxiliary assign-
ment and instead were assimilated to the prevailing principles governing
such assignment. However, assuming that such principles were by then
already self-conflicting, as in the modern period, any such assimilation
is unlikely to have followed a consistent pattern. One striking case is
that of Italian esplodere and French exploser, which are eighteenth/nine-
teenth century derivations from words meaning ‘explosion’, themselves
post-Renaissance calques modelled on Latin explosio. Despite both
being obvious achievements (hence obvious candidates for E assign-
ment), esplodere is assigned E and exploser is assigned A. The reason for
this disparity is simply that by the time exploser was incorporated into
the French lexicon, E assignment in that language had long since
become a purely conventionalized affair and the E-selecting class of
verbs was effectively closed. Attempting to capture the auxiliary assign-
144 Unaccusative Verbs in Romance Languages
Convenire:
Mancare:
Fallire:
The reverse implication can hold, but only under the perfect of result
(see beginning of 5.6). Thus (128) is implied by (129) qua perfect of
result but not qua experiential perfect. In the latter case, (129) indicates
merely that there is at least one occasion in the past when I have under-
stood the theory, and nothing is implied about the present moment
(compare I have understood the theory [that is, previously] but have since
forgotten).47
Discussions of verbs that can be both achievement terms and state
terms tend to focus on psychological verbs (understand, recognize, see
and so on) but the same duality of use applies to verbs like remain, for
example. This latter verb clearly has a state sense in sentences such as
(130) below:
(132) ?When the others were released, he remained in prison for six
months.
In order for there not to be any deviancy, the duration phrase must
construe with another verb, as in (133) below:
this is the case with some of the verbs that actually mean ‘remain’.
Thus Italian rimanere has an achievement occurrence in (134) and a
stative occurrence in (135):
(134) Molti emigrati sono poi tornati in Italia ma milioni sono rimasti
all’estero.
‘Many emigrants have since returned to Italy but millions have
remained abroad.’
Note that not all ‘remain’ verbs can be assimilated to this pattern.
Spanish quedar, for example, is more robustly an achievement term.
Thus while it translates rimanere as used in (134) – han quedado en el
extranjero – it cannot do so when the use is as in (135)/(136): *quedan
en el extranjero.
Other verbs that fit the ‘remain’ pattern include Italian mancare ‘be
missing’ and avanzare ‘be left over’:
the perfect of result, the stative (a) sentences below are implied by the
corresponding (b) sentences, and so the relation is analogous to that
which obtains between the (a) and (b) sentences for mancare and
avanzare:
This follows from the nature of the perfect of result, together with the
achievement–state implicational relation described earlier. With am-
biguous state–achievement verbs, the effect of the perfect of result is
to indicate that a state exists of the type designated by the verb under
its stative meaning, but only the achievement sense of the verb in ques-
tion brings such a state about.
It is widely assumed that, at first, the Romance perfect was primarily
a perfect of result.49 Other functions, for example, those associated with
the experiential perfect and the perfect of recent past, were typically
effected using the preterite, as in the earlier example (78), reproduced
below, and in the examples from note 36, reproduced below as (145) to
(147):
early freezing into place of E assignment may have been assisted by the
previously common ‘cease’ type of meaning, which is attested across
the Romance languages:
The above ‘occur’ acceptation has always existed, with the stative ‘be
necessary’ acceptation perhaps developing from the formerly common
collocation of occorrere (in the ‘occur’ sense) with the deverbal noun
bisogno, as in sequences such as the example below:
(a) Bastare. This verb and its cognates in Spanish, Catalan and so on
were commonly used with achievement meanings that approximated
more to those of the etymon, viz. late Latin bastare ‘carry/endure’. As
regards Italian bastare, Corominas (1980–91) gives ‘durar’ [‘last’], ‘dar
de sí’ [‘last out’] and ‘alcanzar’ [‘reach’] as typical former acceptations.
The ‘last’ and ‘reach’ meanings would appear to be illustrated in the
examples below:
(154) Per grande spazio bastò il rovinìo delle pietre che cadevano giù.
(F 957, 29)
‘The destruction caused by the falling stones reached far and
wide.’
(155) In molte parti [la neve ghiacciata] bastò nella città piu di tre
mesi. (V 4, 65)
‘In many parts [the frozen snow] lasted more than three
months in the city.’
152 Unaccusative Verbs in Romance Languages
Note that the phrase piu di tre mesi in (155) is a durational complement
as opposed to a durational adjunct. In other words it is not asserted
that some state designated by [la neve ghiacciata] bastò endured for
more than three months. Rather, bastò . . . piu di tre mesi forms a
single achievement term indicating that a certain period, viz. that
during which snow was on the ground, exceeded a determinate
measure.
(b) Importare. In older forms of Italian this was more commonly used
in the achievement senses of ‘cost’ or ‘attain’:
(c) Stare. In addition to the ‘remain’ and ‘stand (up)’ meanings that
persist into modern Italian, stare formerly had a variety of meanings
related to the concepts of becoming still and not succumbing to exter-
nal actions:
The ‘stop’ sense is also illustrated in cases such as the example below,
where presumably si stanno is an intransitive reflexive and stea is
transitive:
(d) Esistere and appartenere. A similar account to that put forward for
bastare, importare and stare may be available for esistere ‘exist’ and
appartenere ‘belong to’, although early examples are somewhat incon-
clusive. The first of these verbs is a relatively late borrowing from Latin
and therefore is likely to have been used initially with the meaning
assigned to its Latin model exsistere. Now Latin exsistere was primarily
an achievement term meaning ‘appear/arise’, as in the example
below:
Thus it is unlikely that Italian esistere (or indeed its cognates in other
languages) was always a state term. As regards appartenere, this, like its
Old French cognate apartenir, was formerly widely used in the achieve-
ment sense of ‘come to be close/linked to’. Plausibly this achievement
acceptation could have been significant in conditioning E assignment
at an early period.
While turn out is in may respects similar to the verb appear, the two
differ crucially in terms of their compatibility with duration phrases:
154 Unaccusative Verbs in Romance Languages
(164) For a few weeks it appeared/*turned out that that they were
innocent.
Assuming that duration phrases are possible with state terms but not
with achievement or accomplishment terms, turn out must belong to
one of the latter classes. Presumably it is not an accomplishment term
(compare *It took five years to turn out that . . .) and although one cannot
say, for example, *At that moment it turned out . . ., the event implied in
a formulation such as It later turned out that . . . can only be punctual.
Now an achievement term like turn out can also be used in habitual
sentences such as (165) below:
But this is not in itself significant, given that all achievement and
accomplishment terms become compatible with duration phrases when
used with habitual meaning:
(168) For several decades she wrote a book every six months.
Several of the Italian E-selecting verbs are like English turn out.
In particular spettare ‘fall to’ patterns in a similar way. Standard
examples such as (169) below can be regarded as involving a kind of
historic present occurrence of the achievement sense illustrated in
(170):
We can assume, then, that in the favoured sense both risultare and
spettare are achievement terms. With the weakened meaning of ‘be’,
however, risultare becomes fully compatible with duration phrases,
implying a stative aspectual classification:
Thus the actual historical process whereby the reflexive pronoun was
substituted for the passive morphology can to a large extent be regarded
as a superficial phenomenon that did not alter the nature of the
construction.
The Latin ‘intransitive passive’ or middle voice formed its perfect in
an identical fashion to the true passive, that is, it used a periphrastic
construction involving esse and the past participle of the relevant verb.
This pattern persisted even when the passive morphology gave way to
the reflexive pronoun. However, as mentioned in relation to the earlier
Perfect Auxiliary Selection 157
examples (96) to (101), it was not until centuries later that the reflexive
pronoun began to be used with the past participle. Thus the history
of the perfect of the intransitive reflexives can be assimilated to that
of the non-reflexive E-selecting intransitives discussed in 5.6.1, with
apparently the same initial semantic constraint being in force (cap-
ability to denote a resultant state). Numerous examples from the
early modern corpus confirm this parallel semanticism. Brambilla
Ageno (1964, pp. 201–9) provides a copious supply of instances from
old Italian:
(178) la nebbia che era levata in quel padule (Verb: levarsi) (Trec 48,
18)
‘the mist that had risen up in that marsh’
And many similar examples are forthcoming from Old French and Old
Spanish too.
In some cases, purely intransitive verbs also became reflexive in
Romance or late Latin, giving rise to the Romance ‘inherent reflexives’.
As with the verbs just considered, the associated semanticism appears
to have been identical to that of the non-reflexive E-selecting
intransitives:
5.6.6 Impersonal si
According to Burzio (1986, p. 43), the clitic si has an impersonal occur-
rence in any of the following construction types:
Under either view, the construction illustrated by (183) and (184) was
at the very least uncommon until the end of the Middle Ages, and so
attraction to the highly productive intransitive reflexive pattern of
perfect formation is entirely unsurprising.
5.7 Conclusion
We saw in the first part of this chapter that the mere fact that some
intransitives select perfect auxiliary E, and in so doing pattern with
passives and reflexives, does not in itself motivate the assumption that
the subjects of such verbs are deep objects. Burzio (and to a lesser extent
Cocchi) attempted to identify a property that was common to all of
these constructions under the assumption that unaccusative subjects
were generated in object position, but the blindness of auxiliary selec-
tion to whether the subject is preverbal or postverbal constitutes an
insuperable obstacle to that enterprise. The remaining theories are
essentially stipulative, in that they do no more than state – albeit
within a certain theoretical architecture – that unaccusative verbs
select E and unergatives select A.
Possibly the stipulative nature of the latter theories could be over-
looked if other unaccusative diagnostics defined classes that were iso-
160 Unaccusative Verbs in Romance Languages
In contrast, Italian and French, which both retain the E auxiliary, show
participle agreement in the perfect in broadly similar circumstances.
These are as follows:
162
Past Participle Agreement 163
(15) Voilà les sottises que Jean n’aurait jamais faites. (Belletti 2001,
2.2)
‘These are the stupid things that Jean would never have
done.’
participle agrees with its argument.13 This entails that the participle
will agree with the object of the finite verb when this is A and with
the subject of the finite verb when this is E (the latter may occupy either
a high preverbal position or remain in the small clause). The A type of
agreement is illustrated by collocatas in (24) above, while the E type is
illustrated by migrati in the sequence below, extracted from a medieval
Latin example (Menéndez Pidal 1926, p. 33):
Presumably, when the direct object is ‘high’ this obscures the re-
analysis of the A verb as perfect auxiliary, thus braking the historical
process towards the complete elimination of A-type agreement. In
languages like Spanish and Portuguese this process reached completion
at a comparatively early period. For Italian, French, Catalan and so on,
it has to be assumed that the high object pattern of agreement was
sufficiently robust to withstand the general eliminatory pressure. In
the modern languages, maintenance of vestigial A-type agreement is
assisted by standardization. Note in particular the normative French
rule of ‘the preceding direct object’, which according to Togeby (1974,
p. 189) can be attributed to a single individual, the sixteenth century
poet Marot, himself apparently inspired by Italian formulations such
as Dio noi a fatti ‘God has made us’.15
Thus the rather untidy modern situation concerning past participle
agreement can receive a natural explanation in terms of the operation
of long-term processes of linguistic change upon an initially simple
system. Where these processes have reached a natural terminus, as in
Spanish and Portuguese, participle agreement has reacquired the
transparency of the early Romance system. Where these processes are
only partially complete, as in most of the other Romance languages,
participle agreement becomes inconsistent and often optional, both
classic symptoms of a system in transition.
6.5 Conclusion
172
Participial Absolutes 173
Given the general recognition that structures such as (5) to (9) are in
fact possible, the subject q role-suppression analysis of perfective past
participles immediately looks suspect.
It looks even more so in the light of two other types of data. Firstly,
it appears that transitive verbs in Italian can have active (that is, non-
passive) occurrences in participial clauses, as indicated by the overt
accusative case on me in the example below:
(12) Anzi avevo imparato la sottile arte di presentarmi alle cene già
cenata.
‘In fact I had learned the subtle art of appearing at dinners
already having dined.’
Participial Absolutes 175
In this type of case, Belletti assumed a covert PRO subject in the spec-
AgrP position, with the past participle in the Agr position (after verb
raising), and the object in its original position as complement of the
verb.
The CP paradigm is instantiated by unaccusative clauses, as in (15)
below:
The statement in (22) is also true, though this time true necessarily,
for precise theoretical reasons. Consider the case of a verb which
takes the direct object but does not assign Case to it. This verb will
have to fail to assign q -role to the subject position, since the only
two possibilities for such a direct object to receive Case will be:
(i) that it be linked with a non-argument subject; 8 (ii) that it move
into subject position. Both possibilities require -q s. (Burzio 1986,
p. 184)
Notice that the argumentation refers only to verbs that take a direct
object (understood as the complement of the verb). Now strictly speak-
ing such verbs are transitive. Unergatives, in contrast, are intransitive.
Therefore, on the face of it, the above line of reasoning does not affect
the position of unergatives. Nevertheless, much has been made of the
possibility of unergatives appearing with ‘expletive’ objects (He slept the
sleep of the just and so forth), which has been claimed to show that
unergatives can appear with direct objects. However, if such items
really are direct objects, then logically the verb must be deemed to have
a transitive occurrence rather than an unergative occurrence and the
situation would then be analogous to that of verbs that can be either
transitive or unaccusative (for example, John opened the door versus The
door opened). On the other hand, if expletive objects are not really direct
objects, then the possibility of assigning accusative case does not arise.
Under either interpretation of the data, there is no obvious motive for
assuming that verbs that have an unergative argument structure must
be capable of licensing accusative case.
178 Unaccusative Verbs in Romance Languages
The answer to this appears to be that they do, although the possibility
is confined to a subset of the unergative class (examples below from
Italian and Spanish): 9,10
Notice that the PRO type of construction illustrated by (5) to (9) does
not appear to be subject in principle to any overriding aspectual require-
ment. For in this case, the occurrence of activity verbs is unproblematic
(examples from Spanish and Italian):
7.4 Conclusion
182
Conclusion 183
187
188 Notes
14. The brother-in-law is the postverbal indefi nite nominal in the ‘impersonal’
construction illustrated by example (4).
15. Spanish pasar is standardly classified as unaccusative (see, for example,
Torrego 1989, Garrido 1996). Note in particular that (i) it is routinely
cited as the type of verb that has bare subject capability (the main
Spanish-internal diagnostic for unaccusativity), (ii) it occurs in participial
absolutes (Pasados los botes a la otra ribera . . . ‘With the boats having
passed to the other bank . . .’), (iii) it falls within the unaccusative semantic
spectrum on Sorace’s cross-linguistic Auxiliary Selection Hierarchy (see
Sorace 2000), and (iv) its Italian equivalent is assigned perfect auxiliary
‘be’.
16. Note, though, that in Lasnik’s view (1999:186), accusative objects acquire
the necessary ‘height’ to c-command into an adjunct by raising overtly to
an object agreement projection for case-checking purposes. However,
Lasnik then acknowledges that the failure of a transitive direct object
to control into an adjunct (as in my example (12)) is ‘mysterious’
(1999:189).
17. Notice that, unlike accusative direct objects (see note 16), postverbal unac-
cusative subjects are not widely assumed to have to raise out of their base
position for case-checking purposes (not when they are indefinite, at least).
For example, in Belletti’s partitive/inherent case framework, it is assumed
that indefi nite postverbal unaccusative subjects have their case licensed in
their base position (see Belletti 1999: 34). The same applies if postverbal
unaccusative subjects are assumed to have nominative case, because they
then naturally fall within the scope of the ‘long-distance agreement’ ana-
Notes 189
(i) Sono rimasti nel paese dei profughi ungheresi senza ottenere permessi
di lavoro.
‘Some Hungarian refugees remained in the country without obtaining
work permits.’
Presumably the deviancy here results from something specific to this par-
ticular example, rather than from a general prohibition on the configura-
tion in question. Possibly the problem stems from the information structure
implied by the placement of dei profughi ungheresi after the prepositional
complement nel paese. Belletti (1999:37) analyses ‘reordered’ structures of
this sort as having a topicalized VP and a focalized subject. Conceivably
the relevant adjunct control is degraded under a marked information struc-
ture of this kind.
20. Marandin (2001) observes a parallel control disparity, internal to French,
between the il construction and certain ‘stylistic’ inversion constructions
not involving il:
(i) *Il est entré deux hommes avinés sans frapper/en riant.
‘Two drunk men entered without knocking/while laughing.’
(ii) *Il est entré sans frapper/en riant deux hommes avinés.
‘Two drunk men entered without knocking/while laughing.’
(iii) Alors sont entrés sans frapper/en riant deux hommes avinés.
‘Then two drunk men entered without knocking/while laughing.’
Marandin’s conclusions are rather different from those drawn here. In addi-
tion, he assumes (erroneously) that the ‘stylistic’ inversion illustrated by
(iii) is limited to unaccusative verbs. (See Legendre and Sorace 2003 for a
demonstration of this error.)
21. This example is cited by Martin and Wilmet (1980: 201).
22. Burzio 1986 (152–3) attempted to account for the French–Italian disparity
in terms of his binding-related analysis of perfect auxiliary selection (see
1.1 above). He argued that the relevant binding relation was weaker in the
case of passives than in that of unaccusatives, because with passives
it had to cross a small-clause boundary (assuming a small-clause analysis
of passives). Accordingly, some degree of variation might be expected.
However, it will be shown in 5.2 that Burzio’s attribution of ‘be’-selection
to the existence of a determinate binding relation cannot be accepted, for
190 Notes
2 Expletive Inversion
1. Where exactly depends on the structure envisaged for VP. Typically the VP
of earlier accounts is now assumed to have at least two layers of structure,
for example, a low VP and higher ‘shell’ VP, designated as ‘vp’. The subject
(i.e. of unergatives and transitives) is then assumed to be base-generated or
‘merged’ in spec-vp.
2. Note that the very same linear sequence ‘expletive + unergative verb +
argument’ is routinely countenanced under the supposition that postverbal
subjects in general (regardless of whether the verb is unergative or unaccu-
sative) co-occur with a null expletive fi lling the preverbal subject position.
This assumption was once widespread in Romance linguistics and remains
popular (see, for example, Belletti 1999: 11).
3. This example is cited in Legendre and Sorace (2003).
4. I assume the verbs in these examples are unergative because (i) they select
perfect auxiliary A (as do their Italian equivalents), (ii) they cannot occur
in participial absolute constructions, (iii) they appear within the unergative
semantic spectrum on Sorace’s cross-linguistic auxiliary selection hierarchy
Sorace (2000).
5. For a similar account, specifically in relation to the French il construction,
see Arteaga and Herschensohn (2004:9–10). They assume that in the lexical
array for the sentence Il arrive des jeunes filles ‘There arrive some young girls’
the associate is marked for partitive case and the expletive for nominative
case. The case of the associate is licensed directly by the verb while that of
the expletive is checked and deleted in spec-TP/IP. Any other combination
of cases would cause the derivation to crash.
6. For example, È arrivato Gianni ‘Gianni has arrived’ was assigned the follow-
ing analysis (p. 17):
Notice that the postverbal subject Gianni has moved from its original posi-
tion as the complement of arrivato, which is now occupied by the empty
category e. The empty NP in the preverbal subject position is a null exple-
tive of the kind alluded to in note 2 above.
Notes 191
13. In this particular case, in addition to remedying the deficiency in the asso-
ciate I have deleted the word here and inserted the PP at this point. This is
for stylistic reasons only. The sequence there arose the long-dreaded storm has
a somewhat literary flavour, which would be at odds with the rather banal
locative here. Presumably this additional amendment does not invalidate
the basic point, viz. that there arose is not incompatible with defi nite
associates.
14. Like the second sentence in (40), this presupposes prior mention of the
referent of the associate. Here, though, unlike in (40) the associate is not
in (narrow) focus, or at least this is unlikely. More plausibly, the clause-fi nal
locative can be expected to be focused or, alternatively, the sentence may
receive wide focus. Given this latter possibility, the remark made at the
beginning of note 11 above may require some qualification.
15. Pinto’s analysis of (19) parallels the earlier conclusions regarding French/
English (14) to (16). It would also be applicable to a ‘defi niteness effect’ that
192 Notes
is sometimes alleged for the Spanish faltar construction, as in (i) and (ii)
below, where (ii) is alleged to be deviant:
To the extent that there is a problem with (ii), it is that, out of context, it
is impossible to identify the referent of el alumno. Nevertheless, faltar does
not impose any prohibition on definite subjects per se, as is shown by the
unproblematic examples below:
16. The apparent acceptability of sentence (52) may appear to be at odds with
Belletti’s claim that the earlier sentence (26), repeated below, is marginally
deviant:
Unis and en 1890 in (41) and (42). These cases are discussed at the end of
this section.
20. In fact, this is truer of Italian than French, given that many of the proto-
typical presentational verbs select perfect auxiliary avoir in French and so
are potentially classifiable as unergative: manquer ‘be absent’, exister ‘exist’,
disparaître ‘disappear’, surgir ‘emerge’, falloir ‘be lacking’ etc. However, if
these French verbs are deemed to be unergative, then the empirical basis
for the claim that the il construction exhibits a bias towards unaccusatives
becomes severely degraded.
21. Lonzi (1986:106) and others have assumed that the presentational type
of information structure corresponds to the type of syntax assigned to
unaccusatives under the Ergative Analysis (see 1.1 in this book). However,
I can fi nd no theoretical motivation for this assumption. Moreover, as
noted below, unergative verbs routinely enter into presentational
information structures. These presentational unergatives can only be
accommodated within a model that identifies presentational information
structure with ergative/unaccusative syntax under the assumption that
they ‘switch’ to the unaccusative class. However, the positing of such taxo-
nomic switches merely to accommodate adverse data is methodologically
a somewhat suspect strategy. Normal scientific practice calls for a revision
of the theory in such cases, rather than a reclassification of thedata.
22. For an interesting study of the relationship between objecthood and pre-
sentational information structure in a variety of languages, see Lambrecht
(2000).
3 Partitive Cliticization
1. Extraction from a preverbal subject is rare (except in the case of the French
en-avant construction). This has been attributed to Fiengo’s (1974, 1977)
proper binding requirement on traces. On this point, see Belletti and Rizzi
(1981: 120).
2. Note, however, that extraction from a postverbal subject of telefonare
appears to be possible when the postverbal subject is not followed by a
subcategorized PP. The example below is from Lonzi (1986: 113):
20. Van Valin (1993) and Bentley (2004a) assume a Dowty-style decomposi-
tional treatment of aspectual classes as described (briefly) in 5.5.1 of this
book. Within that framework, they argue that there is a direct link between
partitive cliticization and aspectual class. In the approach adopted here the
principle consideration is information structure, with the aspectual link
arising only as a corollary of that.
21. In this respect, one is perhaps tempted to agree with Lambrecht (2000),
when he suggests (641) that Italian ne has received an ‘undue’ amount of
attention in the Generative literature.
196 Notes
4 Bare Subjects
1. There are other diagnostics such as compatibility with certain participial
constructions, but these alone would defi ne a smaller membership. These
constructions are discussed in Chapter 7.
2. Bare postverbal subjects also occur preverbally in literary or journalistic
styles when conjoined or modified:
the habitual nature of sentences like (37) and render the bare noun position
accessible to quantification. I leave the issue open, however.
10. A partly analogous asymmetry is apparent among complements of ser ‘be’,
as is illustrated by (i) and (ii) below:
The pragmatics of the above sentences are rather different. Roughly speak-
ing, sentence (i) indicates what the persons implicitly referred to are, while
sentence (ii) indicates who they are. Thus son in (i) functions as a genuine
copula, simply linking the implied subject to the predicate médicos, which
in turn ascribes a property to the subject. In contrast, son in (ii) can plau-
sibly be analysed as denoting the identity relation, with the overall sen-
tence asserting an identity between the subject and certain individuals that
are doctors. Accordingly, while (i) can be analysed as a conjunction corre-
sponding to the schema ‘Fa & Fb . . .’, where Fs are doctors and a, b and so
forth are the persons implicitly referred to, (ii) is more transparently rep-
resented in the manner shown below (assuming that the letters ‘a’, ‘b’ and
so on must designate distinct individuals):
11. An analysis that is in some respects similar to that proposed here for cases
such as (47) is put forward by McNally (2004:125–6) in connection with
sentences such as (i) below:
(i) Mucha gente que tiene perros los abandona durante las vacaciones.
‘Many people that have dogs abandon them during the holidays.’
McNally observes that while the bare plural perros ‘licenses a subsequent
pronoun’, this licensing ‘cannot be direct, as the BP does not denote the
antecedent for the pronoun’. She then infers that the antecedent must be
determined via ‘accommodation’, which in her view is facilitated by ‘the
descriptive content of the BP and the fact that [the containing sentence
entails] the existence of an entity that could support the anaphora’
(126).
12. An alternative way of analysing sequences such as wanted to find squirrels is
to say that there is a quantifier but it must have narrow scope (in the sense
of Russell 1905). Alternatively, some authors state that bare nouns can only
have a ‘weak’ reading, as opposed to the ‘strong’ reading represented here
by schemata such as (39). Compare, for example, Laca (1996:254).
13. Garrido (1996) makes effectively the same point when he refutes Masullo’s
(1996) claim that Spanish (i) below can have the same meaning as English
(ii):
Of the Spanish sentence, Garrido writes: ‘It is not that a group of partridges
is represented, rather the property of hunting partridges is applied to a situa-
tion, with the result that in the situation there must exist entities consisting
in hunted partridges.’ (Garrido 1996: 315; my translation.)
14. It is worth pointing out at this stage that I am extending Quine’s approach
into areas on which Quine himself does not appear to have given any clear
pronouncement. Quine’s (1960:175) assessment of verb plus bare noun
combinations as being ‘relative to canonical notation, just not structure’ is
expressed in relation to the dispositional case. I do not claim that he would
endorse the generalization of this position that is being proposed here.
15. In Latin, of course, matters were different, given that there bare nouns
routinely had quantificational meaning. The modern situation results from
a gradual diachronic change that has had partially different results in each
of the various Romance languages. Modern French, for example, completely
disallows bare subjects and objects, while modern Spanish systematically
rejects the use of semantically unmotivated determiners. On the other
hand, the system embodied in modern Italian is in some respects hybrid,
given that it allows bare subjects and objects in the kinds of context in
which they are allowed in Spanish, and yet, like French, it also allows
semantically unmotivated occurrences of the partitive article, as mentioned
in note 8 above.
16. Carlson (1977:422) noted an analogous contrast in cases such as the
following:
The deviancy of the (b) sentences stems from the assignment of wide scope
to a rabbit and a fly, which implies that the same rabbit or fly was repeatedly
discovered/killed.
17. Notice also that the occurrence of unos in (57) has nothing to do with any
alleged incompatibility between vivir and bare subjects. Thus unos can be
deleted from (57) without damaging grammaticality or acceptability,
although the quantificational structure changes accordingly:
naturally to iterative assertions of this sort (on this point, see Lyons 1977:
712). In practice, then, achievement verbs are likely to have a built-in
affinity for bare subjects. Now this circumstance, given that most of the
intransitive achievement verbs are classified as unaccusative in Romance
languages (for example, those meaning ‘die’, ‘arrive’, ‘enter’, ‘depart’ and so
on), may well contribute to the impression that unaccusatives in general
are somehow, in virtue of their unaccusativity, better suited to bare subjects
than unergatives are. See also the discussion in the latter part of 4.3.3.5
below.
19. The formulation Viven lobos en aquel bosque would also be possible, but the
meaning would be subtly different. With bare lobos as the subject, the
structure of the sentence is given by the basic predicational schema ‘Fa’,
where a is the forest in question and Fs are places inhabited by wolves.
Compare the discussion of the later examples (68) and (69) and also note
21.
20. The assignment here of a single question mark here, rather than a double
question mark, is intended to indicate a lower level of deviancy than in the
previous examples. Indeed, at the relevant stylistic level, (69) will be wholly
acceptable to many speakers.
21. The earlier sentence (65), reproduced below, presumably carries an analo-
gous implicature to (69) and yet is not deviant:
The absence of deviancy in (65) appears to result from the fact that no
anomaly is created if the speaker is implicated to be referring to certain
wolves. For example, the speaker may live in an area where wolves are nearly
extinct and might then almost always be thinking of specific wolves when
he or she uses the phrase unos lobos.
22. Interestingly, the occurrence of English some in the translations of the (b)
sentences may also be slightly deviant, but only if it is unstressed (that is,
sm, as in Milsark 1974). In the light of the present analysis, this suggests
two things. First, that stressed some is often not referential in Donnellan’s
sense and, secondly, that Spanish unos/unas equates to English sm rather
than to (stressed) some. The latter in fact has a better Spanish equivalent
in algunos/algunas. For example, the type of context illustrated by the sen-
tence below calls for algunas rather than unas, and the English translation
must then involve some rather than sm:
Defined in that way, the thetic–categorical distinction cuts across the dis-
tinction between complexes with bare subjects and those with determined
subjects. For example, both faltan cuadros in (i) below and faltan unos
cuadros in (ii) are thetic (in the modern sense at least), but whereas
faltan cuadros is a single predicate from the logical point of view, faltan unos
cuadros is a quantification (‘There are some pictures that are missing’):
Note also that while the class of thetic sentences is usually assumed to
include the class of presentational sentences, many of the prototypical
presentational verbs may in fact be deviant when used with bare subjects
(unless there is independently an imperative for the use of the bare noun,
as per the principle stated at the beginning of 4.3.3):
Both the type of data illustrated by (i) and (ii) and that illustrated by (iii)
and (iv) indicate that verb + bare subject complexes exhibit a property over
and above theticity (understood as a type of information structure). Accord-
ing to the view adopted here, this additional property consists in the logi-
cally unstructured nature of these complexes.
24. The view espoused here should not be confused with the kind of theory
put forward by Kratzer (1995), who argues that stage-level predicates sys-
tematically have an argument position corresponding to an event or a
spatio-temporal location. The independence of the two approaches can be
seen by considering a sentence such as (i) below:
Here the verb gritó is a stage-level predicate and so, according to Kratzer’s
theory, has a place in its argument structure for an eventive or locational
argument. On the other hand, in terms of the analysis developed here, the
subject and the verb do not form a logically simple unit, given that that
the presence of the determiner un forces a division into a predicative
element (gritó) and a bound variable which is the argument of that element.
According to the view put forward here, then, sentence (i) is a complete
predication as it stands and thus does not require the provision of any
additional argument.
25. This particular English translation could be understood with intervene as a
(dispositional) individual-level predicate and teachers as having universal
Notes 201
Notice, incidentally, that the type of case illustrated by (ii) calls for a dis-
continuous intonation, with a rather noticeable separation of the predicate
from the subject.
28. In fact, because modification of the bare subject enables the possibility of
a universal interpretation (see note 27), gustar can have a bare modified
subject:
What appears to be the case here is that the adverb continuamente naturally
attracts the accent away from the espectadores position, which is thus ren-
dered immune to the informational richness requirement (assuming the
latter to be a reflex of accentuation). The correct generalization for bare
subjects, then, is that they must be informationally rich unless another
item is present that can absorb the postverbal accent.
4. The trace here occupies direct object position under Burzio’s particular
definition, viz. ‘an A-position governed by the verb’ (1986:56).
5. Another seminal work is Perlmutter 1989, which contains the following
auxiliary selection rule for Italian: ‘If there is a nominal heading both a 1-
arc with tail b and an Object arc with tail b, then clause b requires the
perfect auxiliary essere. Otherwise it requires avere.’ (Perlmutter 1989: 82)
Note, however, that like the other Relational Grammar formulations given
as (5) and (7) in 1.2, this rule is essentially a descriptive generalization,
because the crucial concepts (1hood [that is, subjecthood] and objecthood)
are assumed to be undefi ned primitives. Thus the formulation ‘a nominal
Notes 203
heading both a 1-arc . . . and an Object arc’ is in effect a label for any subject
that is also analysed as an object (in some stratum). Accordingly, Perlmut-
ter’s rule simply asserts a parallel between passives, unaccusatives, raising
verbs and reflexives, given the RG analysis of these constructions as having
subjects that are also objects. This analysis may or may not be accepted,
but the auxiliary selection rule does nothing more than group the relevant
constructions together.
6. In fact, if Belletti’s (1988, 1999) analysis is adopted, this is only true in the
case of indefinite postverbal subjects (see 2.3).
7. Alternatively, under Belletti’s partitive/inherent case hypothesis, (partitive)
case is directly licensed on the associate by the verb (see 2.3).
8. For example, in Chomsky (1995b) [Chapter 3], it is recast as the require-
ment that the strong D feature of the T(ense) position be checked. In lan-
guages like English, this requirement is satisfied by the occurrence of a
(structural) subject in the high preverbal subject position, that is, spec-TP
(or spec-IP under older analyses).
9. Somewhat differently, but with the same overall effect, Rizzi (1982) required
that binding relations must not involve q -dependency, which ruled out the
relation between the preverbal subject position and a postverbal subject,
given that he assumed postverbal subjects were q -marked from the subject
position.
10. Cocchi in fact refers to this as an object agreement projection (1994:99–
100). The crucial point from the present perspective is that the agreement
projection in question (whatever its exact nature may be) is implicated in
determining the agreement form of the past participle.
11. The type of account envisaged in note 7 to Chapter 1 (advanced primarily
for Germanic languages) is stipulative in a similar way.
12. It is true that the past participle exhibits overt agreement in respect of the
subject due ladri, which, according to the agreement-by-movement analysis
of past participle agreement discussed in 6.2, implies that movement has
taken place. However this simply illustrates a general problem with the
agreement-by-movement analysis itself (see 6.3 for further discussion).
13. The equivalent problem is solved in Burzio’s account (1986:58) under
the assumption that an empty category in subject position binds the
impersonal clitic si, thus creating a subject–clitic binding relation which,
for Burzio’s purposes, has the same status as the relation illustrated by
the earlier example (1). Centineo (1996:235) takes issue with this
solution, observing that the c-command relation is incorrect (the empty
category, in effect the trace, c-commands the antecedent rather than vice
versa).
14. By the term ‘transitive reflexive’ I mean a reflexive construction in which
the verb has a transitive occurrence and the reflexive clitic bears a q -role
(other than that assigned to the subject). In contrast, the clitic in an intran-
sitive reflexive construction has no q -role and the verb is not deemed to
have a transitive occurrence. In fact, intransitive reflexives are typically
regarded as a subtype within the overall unaccusative class. Burzio (1986)
divided intransitive reflexives into ‘ergative reflexives’ and ‘inherent
reflexives’, where the former alternate with transitives (for example, rom-
persi ‘become broken’ versus rompere ‘break (something)’) but the latter do
204 Notes
18. In fact, despite assigning ‘*’ to this sentence, Reinhart and Siloni note that
at least some speakers fi nd it acceptable. As was implied by the discussion
in Chapter 3, the view adopted in this book is that the partitive cliticization
diagnostic detects pragmatic failures rather than outright grammaticality
failures, and so a degree of variation in speaker judgments is only to be
expected.
19. For this type of diagnostic, see note 4 in Chapter 7 of this book.
20. This can be seen by considering the aspectual classes of the verbs in Table
5.2.
Notes 205
21. The term ‘degree achievement’ (Dowty 1979) is sometimes used in connec-
tion with some of the verbs in this category.
22. In fact it also diagnoses stativity, but that possibility can presumably be
ruled out given the dynamic nature of the processes described.
23. Compatibility with a time-span phrase introduced by a word meaning ‘in’
is a standard diagnostic for accomplishment status. Compare, for example,
Smith wrote a book in two months (accomplishment) as opposed to Smith was
writing a book for two months (activity).
24. This point is discussed in Rosen (1984).
25. In various places (for example, Centineo 1996, Rohlfs 1949) it is asserted
that verbs like suonare and nevicare select only A when used as activity
terms. However, scrutiny of authentic contemporary sources reveals E
assignment to be possible in this use too.
26. For example, compare (44) to the sentence below:
27. More specifically, this is their generalization for European French. For
Canadian French they acknowledge a degree of lexical conditioning.
28. In fact this phraseology is slightly problematic (in my view), given that
properties presumably must be extra-linguistic objects and so, strictly
speaking, cannot enter into linguistic relations such as modification. I leave
this detail aside, however.
29. In a theory such as Belletti’s (1988/1999), a definite subject such as la barca
would receive the same analysis as a postverbal unergative subject, that is,
it would presumably not be an ‘internal’ argument in Chierchia’s sense.
30. This leaves the case of impersonal si, which Chierchia also claims falls
within the scope of this theory (47–50). For simplicity, I leave this issue out
of the present discussion.
31. Presumably, something similar could be said about r, for example, that it
substitutes a copy of the object for the subject.
32. In essence, the domain of a function is the set of items that can serve as
the input to the function (i.e. the set of possible arguments for which the
function will return a value).
33. See Perlmutter (1989) and Centineo (1996) for reviews of other earlier, now
largely discredited accounts.
34. In fact, given that the syntactic analysis of auxiliary selection does not
appear to follow from any general theoretical principle (see 5.2 and 5.3
above), the attribution of E selection to a determinate but non-verifiable
syntax is to all intents devoid of explanatory power.
35. The corollary of this process in the passive voice was the well-known rein-
terpretation of the tense value of the auxiliary, with for example factus est
giving way to factus fuit as the locus of the meaning ‘was done/has been
done’.
36. Even in the later Middle Ages, the synthetic perfect/preterite was still the
principle vehicle for the majority of the functions that in the modern lan-
guages are associated with the perfect:
206 Notes
37. This would apply also to E + past participle qua ‘perfect’ passive – see note
56 below.
38. Presumably, this is an instance of Comrie’s ‘experiential perfect’ (see begin-
ning of 5.6).
39. Here I take for granted the traditional assumption that true perfect
au xiliaries have a quasi-morphological function. On this point, see Bentley
and Eythórsson (2003) section 5.2.
40. Thus the quasi-equivalence between the periphrastic constructions shown
in (91) to (93) and the corresponding synthetic preterite forms (se fue,
naquimes, cadde) is analogous to that which exists between, for example,
modern Spanish estuvo cerrado ‘was closed’ (that is, ‘became closed’) and
the corresponding ‘action’ passive fue cerrado ‘was closed’. This near equiva-
lence can be accounted for in terms of one construction indicating the
inception of a state (estuvo cerrado) and the other the occurrence of an event
that initiates the same state (fue cerrado).
41. ‘The pronoun is not omitted but has not yet been introduced.’
42. In fact, assuming that the Latin past participle originated as an adjective,
the Classical Latin situation probably represents the outcome of a process
of syntactic reanalysis also. Such cyclical developments are not uncommon
historically.
43. This follows because unergatives are assumed to assign subject q -role to
their subject. If the past participle morphology blocks the externalization
of subject q -role, an unergative past participle will fail to assign any q -role
either to a matrix subject in a predicative construction (for example, *The
customer was complained) or to the modified noun in an attributive con-
struction (for example, *a coughed patient).
44. In addition to the two examples in the text, consider also the example with
convenuti in note 53 below. In the latter case, the participle presumably
comes from the ‘agree’ sense of convenire, to which perfect auxiliary A is
normally assigned (by hypothesis, this assignment indicates an unergative
occurrence).
45. Monter ‘go up’ and descendre ‘go down’ are notable exceptions.
46. As noted in many places, an analogous alternation persists in modern
Italian:
(i) Luisa ha corso nel parco per un’ora. (Centineo 1996: 251)
‘Luisa ran in the park for an hour.’
(ii) Luisa è corsa a casa in un’ora. (Centineo 1996: 251)
‘Lucy ran home in an hour.’
47. The exact status of the implication in question depends on the particular
analysis of the perfect. If the perfect of result and the experiential perfect
Notes 207
are assigned the status of different meanings, then the implication has a
semantic status. On the other hand, if the usages in question do not involve
different meanings, the implication presumably is a pragmatic
implicature.
48. Note also that the achievement potential of mancare is rendered rather
explicit when the verb is used as a euphemism for morire:
49. In fact, given the discussion in 5.6.1, many apparent instances of the
perfect of result in the Middle Ages might in reality be instances of a
copula–adjective construction. The ‘resultant state’ meaning is the same
under both analyses, however.
50. In fact, in early modern Italian, there are fairly frequent attestations with
auxiliary avere of at least some of the verbs in question, but never when the
perfect is a perfect of result.
51. Possibly bisognare ‘be necessary’ could be assimilated to the general model
discussed in this section, given the achievement sense illustrated by the
(somewhat archaic) examples below:
54. Translation: ‘verbs expressing an action in which the subject did not par-
ticipate but which the subject simply underwent, and whose agent remained
completely indeterminate’.
55. Alternatively, it might be classified as an instance of the ‘middle voice’.
56. Notice that this circumstance implies that E + past participle was three-way
ambiguous, as between a strictly intransitive meaning, a reflexive intransi-
tive meaning, and a passive meaning.
208 Notes
Ne ho viste/*visto molte.
‘I have seen many of them.’
7 Participial Absolutes
1. Typically this will produce a control-type of structure, with an item outside
the participial clause controlling into that clause.
2 Here the participial clause una vegada jugat presumably could be classified
as involving an instance of ‘arbitrary’ PRO, analogously to the participial
clause in the following example from Belletti (1992:43):
(i) Il re ha abdicato.
‘The king has abdicated.’
(ii) L’erba ha attecchito.
‘The grass has taken root.’
(iii) Il latte ha bollito.
‘The milk has boiled.’
(iv) Il lago ha tracimato.
‘The lake has overflowed.’
As far as the Spanish examples are concerned, I classify the relevant verbs
as unergative on the grounds that (i) their Italian and French equivalents
are assigned avere/avoir and (ii) they fall within the unergative semantic
spectrum on Sorace’s (2000) auxiliary selection hierarchy. Given the unreli-
ability of the bare subject diagnostic discussed in Chapter 4, there is in fact
no principled Spanish-internal basis for classifying verbs as unaccusative
or unergative.
10. In addition to the examples in the text, it is also worth noting cases such
as the following:
212
References 213
López Díaz, Enrique. 1996. ‘Aspectos de los SSNN desnudos en español’ Cuad-
ernos de lingüística 3, 119–35.
Loporcaro, Michele. 2003. ‘The Unaccusative Hypothesis and participial abso-
lutes in Italian: Perlmutter’s generalization revised’ Rivista di Linguistica 15/2,
199–263.
——2001. ‘La selezione dell’ausiliare nei dialetti italiani: dati e teorie’ in F.
Albano Leoni, R. Sornicola, E. Stenta Krosbakken, and C. Stromboli (eds) Dati
empirici e teorie linguistiche, Atti del XXXIII Congresso della Società di Linguistica
Italiana, Napoli, 28–30 ottobre 1999, 455–76. Roma: Bulzoni.
——1998. Sintassi comparata dell’accordo participiale romanzo. Torino: Rosenberg
& Sellier.
Luca, Giuseppe. De (ed.) 1954. Prosatori minori del Trecento. Milano–Napoli:
Ricciardi.
Lyons, John. 1977. Semantics. Cambridge: Cambridge University Press.
——1968. Introduction to Theoretical Linguistics. Cambridge: Cambridge Univer-
sity Press.
Marandin, Jean-Marie. 2001. ‘Unaccusative inversion in French’ in Y. d’Hulst,
J. Rooryck and J. Schroten (eds) Romance Languages and Linguistic Theory 1999:
selected papers from ‘Going Romance’ 1999, Leiden, 9–11 December 1999, 195–
222, Amsterdam: John Benjamins.
Martin, R. and Wilmet, M. 1980. Syntaxe du moyen français. Bordeaux:
SOBODI.
Masullo, Pascual. 1996. ‘Los sintagmas nominales sin determinante: una pro-
puesta incorporacionista’ in I. Bosque (ed.) El sustantivo sin determinación. La
ausencia de determinante en la lengua española, 169–200. Madrid: Visor.
——1992. ‘Incorporation and case theory in Spanish: a crosslinguistic perspec-
tive’. Ph.D. dissertation. University of Washington, Seattle.
McNally, Louise. 2004. ‘Bare plurals in Spanish are interpreted as properties’
Catalan Journal of Linguistics 3, 115–33.
——1998. ‘Existential sentences without existential quantification’ Linguistics
and Philosophy 21, 353–92.
Mendikoetxea, Amaya. 1999. ‘Construcciones inacusativas y pasivas’ in I.
Bosque and V. Demonte (eds) Gramática descriptiva de la lengua española,
1574–629 (vol 2). Madrid: Espasa Calpe.
Menéndez Pidal, Ramón. 1926. Orígenes del español, estado lingüístico de la Penín-
sula ibérica hasta el siglo XI. Madrid: Hernando.
Milanesi, Gaetano (ed.) 1864. Dell’arte del vetro per musaico. Tre trattatelli dei secoli
XIV e XV. Bologna: Romagnoli.
Milsark, Gary. 1974. ‘Existential sentences in English’. Ph.D. dissertation. Mas-
sachusetts Institute of Technology, Cambridge, MA.
Palmer, Frank R. 1986. Mood and Modality. Cambridge University Press:
Cambridge.
Perlmutter, David M. 1989. ‘Multiattachment and the unaccusative hypothesis:
the Perfect Auxiliary in Italian.’ Probus 1/1: 63–119.
——1983. ‘Personal vs. impersonal constructions’ Natural Language and Linguis-
tic Theory 1, 141–200.
——1978. ‘Impersonal passives and the unaccusative hypothesis’ Proceedings of
the Fourth Annual Meeting of the Berkeley Linguistics Society, 157–89. Berkeley,
California.
218 References
Picallo, M. Carme. 1984. ‘The Infl node and the null subject parameter’ Linguis-
tic Inquiry 15, 75–102.
Pinto, Manuela. 1997. ‘Licensing and interpretation of inverted subjects in
Italian’. Ph.D. dissertation, Utrechts Instituut voor Linguistiek, Universiteit
Utrecht. (Page references are to the on-line version at http://www.library.
uu.nl/digiarchief/dip/diss/01751455/inhoud.htm)
Pirson, Jules. 1906. ‘Mulomedicna Chironis. La syntaxe du verbe’ Festschrift zum
12. Neuphilologentag, 390–431.
Posner, Rebecca. 1996. The Romance Languages Cambridge: Cambridge
University.
Pountain, Christopher. 1985. ‘Copulas, verbs of possession and auxiliaries in
Old Spanish: the evidence for structurally interdependent changes’ Bulletin
of Hispanic Studies 62, 337–55.
Quine, Willard Van Orman. 1960. Word and Object. Cambridge, MA.: MIT
Press.
Radford, Andrew. 1997. Syntactic Theory and the Structure of English: a minimalist
approach. Cambridge: Cambridge University Press.
Reichenbach, Hans. 1947. Elements of Symbolic Logic. New York: Macmillan.
Reinhart, Tanya and Siloni, Tal. 2004. ‘Against the unaccusative analysis of
reflexives’ in A. Alexiadou, E. Anagnostopoulou and M. Everaert (eds) The
Unaccusativity Puzzle: explorations of the syntax–lexicon interface, 159–80.
Oxford: Oxford University Press.
Richards, Norvin. 1998 ‘The principle of minimal compliance’ Linguistic Inquiry
29:4, 599–629.
Richards, Marc and Biberauer, Theresa. 2005. ‘Explaining Expl’ in M. den
Dikken and C. Tortora (eds) The Function of Function Words and Functional
Categories, Amsterdam: John Benjamins.
Rizzi, Luigi. 1982. Issues in Italian Syntax. Dordrecht: Foris.
Rohlfs, Gerhard. 1949. Historische grammatik der italienschen Sprache und ihrer
Mundarten. Bern: A Francke.
Rosen, Carol. 1984. ‘The interface between semantic roles and initial grammati-
cal relations’ in D. Perlmutter and C. Rosen (eds) Studies in Relational Grammar
2, 38–77. Chicago: Chicago University Press.
Russell, Bertrand. 1905. ‘On denoting’ Mind 14, 479–93.
Russinovich Solé, Yolanda. 1990. ‘Valores aspectuales en el español’ Hispanic
Linguistics 4:1, 57–86.
Saccon, Graziella. 1992. ‘VP-internal arguments and locative subjects’ in Pro-
ceedings of the Twenty-second Annual Meeting of the North-Eastern Linguistics
Society, 383–97. Amherst: Graduate Linguistic Student Association of the
University of Massachusetts, Amherst.
Schiaffini, A. (ed.) 1926. Testi fiorentini del Dugento e dei primi secoli del Trecento.
Firenze: Sansoni.
Schroten, Jan. 2002. ‘Sobre la ausencia de determinante y su interpretación’ in
R. Bok-Bennema (ed.) La oración y sus constituyentes: Estudios de sintaxis genera-
tiva Foro Hispánico 21, 71–86. Amsterdam: Rodopi.
Sorace, Antonella. 2000. ‘Gradients in auxiliary selection with intransitive
verbs’ Language 76: 859–90.
Stowell, Timothy. 1978. ‘What was there before there was There’ in D. Farkas,
W. M. Jacobsen and K. W. Todrys (eds) Papers from the 14th Regional Meeting
References 219
220
Index 221