Torrego, Esther, (1998), The Dependencies of Objects. Linguistic
Inquiry Monograph 34. Cambridge, Massachusetts: the MIT Press. 197
Reviewed by Alex Alsina, Universitat Pompeu Fabra (Barcelona).
This book is a study of overtly case-marked objects in Spanish that pays
attention to similar phenomena in other languages, particularly other
Romance languages. It is well-known that direct objects, or accusative
objects, in Spanish, when expressed as full NPs, or DPs (to use the
author's term), alternate between two forms depending on various
semantic and syntactic properties: DPs withouth any overt case
morphology, and overtly case-marked DPs. The overt case marker in
question is the preposition "a", which also marks dative objects. The
study adopts the Minimalist Program, outlined by Chomsky (1992, 1994,
1995), as a framework for developing an analysis for the various
phenomena under investigation.
Chapter 1 outlines the theoretical framework of the book and sketches
the content of the subsequent chapters.
Chapter 2 presents the various generalizations that characterize Spanish
accusative objects marked by the dative preposition. Among these
generalizations are correlations between overt case-marking and
specificity, between overt case-marking and telicity, between overt
case-marking and an interpretation of the subject as an agent or a
cause, etc. The common element in the explanation of all of these
generalizations is the idea that the marked accusative object raises
overtly to a position outside VP, to the specifier position of the
functional category v, which takes the VP as its complement. For
example, the two following facts are observed: marked accusative
requires (or may require) a telic interpretation of the event which is
absent with unmarked accusative objects, and an affected interpretation
of the (animate) object requires overt case-marking on the object. It
is suggested that objects moved outside the VP have a delimiting role
for their predicates and, at the same time, the position to which
objects moved, the spec of vP, is the locus of affectedness. Thus, two
different properties of overt case-marking would be made to follow from
the idea that overtly case-marked objects in Spanish undergo overt
movement to spec of vP. Likewise, the observation that overt
case-marking on the object requires the subject to be interpreted as an
agent or as a cause is linked to the idea that an overtly case-marked
object is in spec-vP and to the hypothesis that the causative or agent
role is assigned to the v-VP configuration.
Torrego proposes that there are two types of marked accusatives, one
with structural Case and the other with inherent Case. Marked
accusative is inherent with causative and agentive verbs and those with
affected objects, and is structural with other verbs. This theoretical
difference is appealed to in order to account for an asymmetry with
respect to extraction out of marked accusative objects: such extractions
are ill-formed with the former class of verbs, but not with the latter.
The presence of the extra structure created by the inherent case marker
is a crucial element in the explanation of this contrast.
Chapter 3 examines variation in causative structures in Spanish and
argues that this variation is largely conditioned by Case assignment.
Much of the discussion is based on a causative structure in which the
causee appears between the causative verb "hacer" 'make' and the
infinitive. One of the goals of the chapter is to explain the
restrictions on the "faire-par" causative in several Spanish dialects.
In these dialects, the "faire-par" causative is only fully acceptable in
the following situations: (1) when an object clitic is present, or (2)
when a reflexive clitic is present, and that, otherwise they are best
with verbs identified as "constructive accomplishments" such as 'build'
or with case-marked objects.
Chapter 4 analyzes constructions with dative objects, both with
ditransitive verbs and with unaccusatives. Restrictions regarding
constructions with two objects with identical case-marking are examined.
The chapter also deals with contrasts between the presence and absence
of doubling clitics. A comparison is made with the double object
constructions in English. An analysis is presented of the contrast in
raising constructions with 'seem' between those without the dative
argument (acceptable) and those with it (unacceptable). Conclusions are
briefly summarized at the end.
This book is a mine of linguistic puzzles concerning objects, causative
constructions, clitics, and related phenomena in Spanish. It is of
interest to anyone concerned with these matters, regardless of what
analysis and what theoretical framework one may wish to adopt. This
book is also of interest to linguists who want to find out how the
Minimalist Program can be applied to a particular empirical domain. The
author has done a great job of synthesizing the data and presenting
generalizations that need to be explained. Despite these positive
aspects of the book, I experienced serious frustrations while reading it
having to do with the data and with the analysis.
The problem with the data, which is a minor one compared to the other
one, has to do with the fact that many of the data presented are alien
to the varieties of Spanish that I speak or am familiar with.
Naturally, I cannot claim to know all varieties of Spanish, so I should
not be surprised to discover data that do not belong to the Spanish that
I know. Some examples follow. Torrego presents a contrast in pp. 58-59
between "A quienes (les) vieron?" and "A quienes (*los) vieron?", but
both forms are equally ill-formed for me when the clitic (the form in
parentheses) is included. The example "La guerra los hizo subir" in
page 101 is claimed to allow only an animate reading for the clitic
"los", although for me it also allows an inanimate reading, but then
this fact is claimed to hold only in Spanish dialects where the
pre-infinitival causee with "hacer" and an intransitive verb is
attested, and I find this causative structure highly marginal (examples
such as "La maestra hizo a los alumnos subir"). And the construction
exemplified by sentences such as "Me hizo pedir yo," which is analyzed
in this book, is also alien to me.
The main problem with these numerous discrepancies about the data is
that it makes it hard to evaluate when a difference in the
interpretation of the data is due to a difference in the data (that may
not be obvious) or to a possible misanalysis of what actually is the
same data. To illustrate this, consider the argument given in support
of the claim that the structure embedded under "hacer" in a sentence
like "La guerra hizo subir los precios" is a single unit (p. 102).
Torrego claims that standard constituency tests indicate that the
sequence "subir los precios" in this sentence is a constituent and
provides examples such as the following in support of this claim:
(38) a. Lo que la guerra hizo es subir los precios.
what the war caused is to rise the prices
d. Que ha hecho la guerra sino subir los precios?
what has the war caused except rising the prices?
For me the translation of these sentences is not as indicated in the
glosses, but rather as "What the war did is to raise the prices." and
"What has the war done but raise the prices?" for (38a) and (38d)
respectively. I would tend to believe that Torrego is playing on the
polysemy of "hacer" and "subir". "Hacer" can be either the causative
verb meaning 'cause' or 'make' or the generic transitive verb
translatable as 'do', and "subir" can be either intransitive, equivalent
to 'rise', or transitive, equivalent to 'raise'. For the examples in
(38) to be relevant to Torrego's claim, "hacer" must be causative and
"subir" intransitive. What makes these exemples acceptable (even though
not perfect) for me is that "subir" can be taken to be a transtive verb,
whose unexpressed subject is coreferential with "la guerra". But, under
this interpretation, "hacer" would not be functioning as the causative
verb, but as the counterpart of 'do'. What shows conclusively that, in
my variety of Spanish, this interpretation is right is that, if, instead
of "subir", we use a verb that only has an intransitive use, we get an
unacceptable result. "Caer" 'fall' is such a verb: even though the
sequence "caer los precios" can appear following the causative verb
"hacer", as in "La guerra hizo caer los precios," it cannot appear as a
focused or topicalized constituent of this verb, as we see in (I).
(I) a. *Lo que la guerra hizo es caer los precios.
what the war did/caused is to fall the prices
b. *Que ha hecho la guerra sino caer los precios?
what has the war done/caused except falling the prices?
>From this I would conclude that Torrego's data in (38) do not provide
any evidence for the claim that a sequence such as "subir los precios"
in a causative construction is a constituent, because those examples do
not include the causative verb "hacer". Of course, I can only reach
this conclusion for the variety of Spanish that I am familiar with,
where (I) is ungrammatical. But, for all I know, the examples in (I)
could be grammatical in the variety of Spanish analyzed by Torrego, in
which case my conclusion would not apply and the examples in (38) would
constitute relevant evidence for the claim under investigation.
My second difficulty with this book is in the analysis (or analyses)
provided. I have trouble understanding many of the analyses given: in
many cases, I fail to derive the predictions that should follow from the
analysis, I cannot tell what would constitute a counterexample to the
analysis, I cannot figure out what role a given assumption plays in the
analysis, etc. This could well be due to a deficiency on my part, but
other readers might experience the same difficulties. Next, I will
point out some problem cases.
Chapter 2 of the book starts out with a list of six generalizations
about marked accusative (pp. 14-16): (1) marked accusative can cooccur
with doubling clitics in some dialects (it is a necessary condition for
clitic doubling), (2) marked accusative objects are interpreted as
specific, (3) telicity requires marked accusative (with animate
objects), (4) the subject in a clause with a marked accusative object is
an agent or a cause, (5) marked accusative is restricted to animates,
and (6) affected animate objects require marked accusative. The
indications given all along are that these generalizations will follow
from a unified analysis. The main idea underlying this analysis is that
an object with marked accusative raises to the specifier position of the
functional category v, which takes a VP as its complement. Ideally, an
overt movement operation of this kind should have some testable effects
on word order and all of the properties of the construction should
follow straightforwardly from this operation. However, as Torrego
notes, there is no visible effect of object raising on word order, so
that the motivation for the analysis rests entirely on how well it
explains the generalizations listed above. And these generalizations
are explained by attributing the various properties that correlate with
overt case-marking to the spec of v-VP position on a one-by-one basis.
For example, it is proposed that the spec of v-VP has a specific
interpretation, that the presence of an object in spec of v-VP may shift
the aspect of the predicate to a telic situation, that the presence of
an object in spec of v-VP gives an agentive interpretation to the
subject, etc. One wonders why this analysis should be preferable to one
that attributes all of these properties directly to overt case-marking:
in both cases, a one-by-one stipulation of those properties is required.
It is not clear what the advantage is in attributing these properties
to a special phrase structure position for which very weak evidence is
In pp. 23 ff, Torrego claims that there are two types of marked
accusative: marked accusative of non-affected objects is structural
Case, whereas marked acusative of affected objects is inherent Case.
However, this claim is faced with a major problem that Torrego notes,
but then simply disregards. Inherent (or quirky) Case has the generally
accepted property that it is preserved under passivization: an argument
with inherent Case doesn't lose or change its case marking when its
clause is passivized. Genitive and dative case in Icelandic are a good
example of this, and dative case in Spanish is also a good example of
case marking that does not change from active to passive. However, the
marked accusative case of affected objects in Spanish is not preserved
(see relevant examples in p. 28). Torrego notes that, in contrast with
Spanish, some Hindi dialects preserve the accusative "-ko" marking in
passives, citing Mohanan 1990. Torrego then goes on to say (p. 29): "If
the marked accusative of "affected" objects were inherent, overt marking
in passivization could be preserved (just as the marking of datives and
genitives is preserved in Icelandic passives). It remains to be seen
whether relevant data from the dialects of Hindi confirm this
expectation." But the relevant data from the dialects of Hindi does
confirm this expectation, as the examples from Mohanan 1990 (also
Mohanan 1994) show. If we agree that we identify inherent Case because
its case marking is preserved under passivization, then we must conclude
that marked accusative case is inherent in some dialects of Hindi,
because its case marking does not change under passivization, but is
structural in the other dialects of Hindi, where it does change under
passivization. And we are forced to conclude that it is structural in
Spanish because it is not preserved under passivization. To ignore this
evidence and still conclude that marked accusative case can be inherent
Case in Spanish is to devoid the distinction between structural Case and
inherent Case of the conceptual and empirical substance that it is
standardly asociated with. And Torrego does not explicitly reject this
The agentivity effect of clauses with marked accusative is explained as
follows: following a suggestion of Chomsky's (1995), the causative or
the agent role can be understood as the interpretation assigned to the
v-VP configuration; since marked accusative is assigned to objects that
raise to spec of vP, the presence of marked accusative implies the
existence of the v-VP configuration. Therefore, whenever there is
marked accusative, there must be an agentive or causative
interpretation. But what happens with verbs that allow an alternation
between marked and unmarked accusative on the object? This should mean
that such verbs should alternate between a v-VP configuration and some
other configuration, but this second option is not spelled out. Also,
we have no way of knowing what consequences the presence or absence of
the v-VP configuration has other than on overt case marking and on the
corelation between overt case marking and other properties such as
In pp. 34 ff., Torrego tries to relate the idea that marked accusative
on affected objects is lexical quirky Case to Grimshaw's (1990) theory
of quirky Case-marking. In Grimshaw's theory, arguments at argument
structure are ranked by prominence in two tiers: the aspectual tier and
the thematic tier. Adopting this theory, Torrego proposes (p. 36) that
"an argument (subject or object) can be lexically or quirky marked when
it has thematic or aspectual prominence relative to the other argument."
It is unclear how one is to interpret this proposal. Prominence is a
comparative property: A is more or less prominent than B. To say that A
has prominence relative to B is either meaningless or would be true of
any pair of arguments in an argument structure: it is necessarily true
of any pair of arguments A and B of an argument structure that A is in a
prominence relation to B. Torrego goes on to state: "The two arguments
of an agentive telic transitive verb are both aspectually prominent."
But this should be true of all transitive verbs. Torrego goes on to
conclude: "This approach covers the fact that the object of only
agentive and caustive verbs can or must be lexically or quirky
Case-marked, and also covers the fact that the overall phenomenon of
marked accusative Case happens with verbs that have agentive subjects."
I fail to see how these predictions are derived.
Regarding the animacy restriction on overtly Case-marked objects,
Torrego proposes that the term "actor" is more appropriate to
characterize the class of nominals that appear in marked accusative than
the term "animate". The motivation for this terminological change is
found in the existence of some expressions denoting inanimate entities
that can take overt Case-marking, as in the following example (p. 55):
(45) El acido afecta (a) los metales.
'Acid affects metals.'
(I suspect that the presence of marked accusative in this example has
more to do with the verb than with the semantics of the object.)
Torrego does not propose a definition of "actor", but refers the reader
to Jackendoff's (1983) characterization of "actor" as the character in
an action-type event that performs the action. Given this
characterization, it is hard to see how "los metales" 'the metals' in
example (45) can be interpreted as the argument that performs the action
of the event; what would "el acido" 'the acid' be, then? Torrego goes
on to say: "The distinction between "animates" (or "actor"-like
nominals) and "inanimates" does not appear to be semantic in nature."
This is an odd claim to make, given that the notion of "actor," which is
proposed to replace "animate," is given a semantic characterization.
The motivation for this claim, however, is found in the fact that
certain expressions, specifically proper names of humans, can be used to
refer to anything, from a boat to a magazine to a can of beer, and
always require overt Case morphology, as in the following example:
(46) Esconde *(a) Barbara.
It is suggested that a purely morphological property underlies overt
Case-marking, although no indication is given as to what this
morphological property might be. Possibly the most robust
generalization about overt Case-marking of objects in Spanish is that it
is restricted to "animates", even though there are problem cases such as
example (45) and "personified" expressions such as (46). Because of
these problems, Torrego gives no explanation for the correlation between
Case-marking and animate interpretation and between absence of
Case-marking and inanimate interpretation that we see in pairs such as
(II) and (III):
(II) a. Esconde a este. ('Hide this one (male animate entity).')
b. Esconde este. ('Hide this one (inanimate entity).')
(III) a. Veremos a otra. ('We shall see another one (female animate
a. Veremos otra. ('We shall see another one (inanimate entity).')
Torrego often appeals to intuition in the exposition of the analysis and
does not work out the details of the analysis. The following remark,
very reminiscent of Chomsky's style, illustrates this point (p. 139, end
of 3rd par.): "Whatever the details of this intuitive account are, the
general approach seems quite plausible." If one does not work out the
details of an account, it is not possible to derive predictions from it,
and therefore the theory is unfalsifiable.
This work by Torrego is part of a move towards deriving explanations
from very general and simple principles that interact in complex ways
with each other. This move is a welcome one, but we should be careful
that it does not land us in vagueness and imprecision. There is a style
of explanation in which appeal is made to principles, often unstated or
imprecisely stated principles, from which predictions are claimed to
follow, but in which the derivation of predictions (the logical steps
from which predictions follow) is not worked out and it is not possible
to know what a prediction of the theory really is and what a
counterexample to the theory might be. There is a danger that a reader
of Torrego's "The Dependencies of Objects" might have the impression
that the explanations in this book are of this style. To prevent this
impression, Torrego should have made a greater effort in terms of
exposition (even if it means leading the reader by the hand) to show how
the various predictions are derived. For that, it is necessary to state
the crucial principles explicitly, as numbered items so the reader can
refer to them easily, to make the assumptions about representations also
explicit, and to give step-by-step derivations of the various
Chomsky, Noam. 1992. A Minimalist Program for Linguistic Theory. MIT
Working Papers in Linguistics, no. 1.
Chomsky, Noam. 1994. Bare Phrase Structure. MIT Occasional Papers in
Linguistics, no. 5.
Chomsky, Noam. 1995. The Minimalist Program. MIT Press.
Grimshaw, Jane. 1990. Argument Structure. MIT Press.
Jackendoff, Ray. 1983. Semantics and Cognition. MIT Press.
Mohanan, Tara. 1990. Arguments in Hindi. Ph.D. dissertation, Stanford
Mohanan, Tara. 1994. Argument Structure in Hindi. Stanford: CSLI
Alex Alsina, Professor titular d'universitat, Faculty of Translation and
Interpretation, Universitat Pompeu Fabra, Barcelona, Spain. Ph.D.,
Stanford University, 1993. Research interests include syntax,
morphology, morphosyntax, argument structure, linguistic theory.
Facultat de Traduccio i Interpretacio
Universitat Pompeu Fabra
La Rambla, 30-32