LINEBURG


<< . .

 4
( 19)



. . >>

familiar D elements. Clitic-doubling constructions may involve the it-
eration of the D head in a complex DP terminating with the lexical NP
restriction. In this way this notoriously recalcitrant domain can find a
natural account which is able to capture both the movement nature of
cliticization (Kayne 1975, Sportiche 1998) and the otherwise surprising
double occurrence of a single argument (see Belletti 1999, Uriagereka
1995, Torrego 1995, among other references).
We have already mentioned the idea that the functional struc-
ture of the clause is fundamentally uniform and much (and possi-
bly all) of the observed variation has to do with the degree of mor-
phological realization of the functional structure. This approach in
fact extends to the domain of verbal morphology the line of inquiry
that proved successful on Case morphology some twenty years ago:
apparently major differences in the functioning of Case systems were
amenable to basically uniform systems of Case assignment/checking,
with language-invariant syntactic consequences (i.e. the triggering of
movement in the passive, with unaccusative and raising verbs, etc.)
and with much of the variation reduced to the overt or covert morpho-
logical manifestation of Case (Vergnaud 1982). The emerging picture
then is one in which a fundamentally uniform syntax, except for a set
of parameters, is combined with systems of inflectional morphology
which allow variation (with an apparently large spectrum of possible
inflectional paradigms, ranging from very rich to extremely impov-
erished, and with the expression of parametric values for the syn-
tactic component: movement of phrases and heads must be overt or
covert, etc.).



28
Editors™ introduction

The empirical studies on the IP, CP, and DP uncovered extraordi-
narily rich functional structures which complete the lexical projections
of nouns and verbs. This discovery, started around the mid 1980s, has
given rise more recently to autonomous research projects, the “carto-
graphic” projects, whose aim is to draw maps as detailed as possible of
the syntactic configurations. The results of the cartographic research
in the late 1990s and in current work (see, for instance, the essays col-
lected in Cinque 2001, Belletti in prep., Rizzi in prep.), while leading
to syntactic representations much richer than those assumed a few
years back (with IP, CP, and DP identifying complex structural zones
rather than single layers), strongly support the view of the essential
uniformity of natural languages. On the one hand, they confirm the
fundamental invariance of the functional hierarchies with much more
realistic and fine-grained representations of syntactic configurations
than in previous work (of special prominence in this connection are
Cinque™s (1999) results on the clausal structure); on the other hand,
the complexity of the fine structures of clauses and phrases turns out to
be amenable to a single building block, the minimal structure arising
from the fundamental structure-building operation, “merge,” in the
system of Chomsky (1995a). The functional lexicon turns out to be
much richer that previously assumed, but the fundamental computa-
tions to string elements together are elementary and uniform across
categories and languages.
The discovery of the depth and width of cross-linguistic unifor-
mity made it possible to think of UG as a substantial component of
particular grammars, in fact by far the most fundamental component;
reciprocally, parametric models introduced the appropriate technical
language to enhance and deepen the discovery of cross-linguistic uni-
formity. So, the development of the models and the sharpening of the



29
On nature and language

empirical discovery that grounded them proceeded hand in hand in
the course of the last twenty years.


6 The Minimalist Program

6.1 Background
The Principles and Parameters approach provides a potential solution
to the logical problem of language acquisition, resolving at the same
time the tension between descriptive and explanatory adequacy: the
acquisition of very complex grammatical patterns can be traced back
to innate principles and a limited process of selection among options.
So, in a sense, the properties that are observed in a particular grammar
are explained, in that they are reduced to properties of UG and to a
limited residue. The next set of questions that arise concerns the very
form of UG: are UG properties amenable to a further explanation, or
has the explanation process somehow to stop there, at the current
state of our understanding? On the one hand, it is conceivable that
a deeper understanding of the physical substrate of UG may provide
further explanations for the existence of some of the properties of
UG: it could very well be that principles of structural organization and
interpretation of linguistic expressions have the shape we observe, and
not some other imaginable shape, because of some inherent necessity
of the computing hardware, the relevant brain structures. On the other
hand, a detailed exploration of the physical substrate is a distant goal
which awaits major advances in the brain sciences (not to speak of
the even more remote exploration of the embryological and genetic
factors involved), and may well require the introduction of entirely new
concepts. Major empirical discoveries and conceptual breakthroughs
may be necessary in order to connect and integrate the functional


30
Editors™ introduction

modeling and the study of the computation at the cellular level, as is
stressed in the third chapter of this book. Is there any avenue to pursue
in the meantime? Here the minimalist questions come into play.
That language may be economically designed is suggested by
various kinds of considerations. Much work in the structuralist tradi-
tion already suggested that the organization of linguistic inventories
obeys certain economy principles (see Williams (1997) for a recent dis-
cussion in terms of the Blocking Principle of the Saussurean idea that
“dans la langue il n™y a que des diff erences” [in langue there are only
´
differences]). Within the tradition of generative grammar, attempts
to provide an evaluation measure to select among competing analy-
ses were systematically based on the notion of simplicity, with most
highly valued solutions being those involving the minimum of com-
plexity (smallest number of elements, smallest number of rules). Direct
reflexes of these ideas are also found in the study of performance, with
attempts to define complexity metrics based on the number of compu-
tational operations to be performed (as in the “Derivational Theory of
Complexity”; see Fodor, Bever, and Garrett (1974) for critical discus-
sion). Principles such as the Avoid Pronoun Principle also implied the
choice of the most elementary form compatible with well-formedness
(in particular, null pronouns must be preferred to overt pronouns when
available), an idea that has connections to the Gricean approach to the
successful conversational use of linguistic structures. The “Avoid Pro-
noun” idea was later generalized, giving rise to principles of structural
economy (e.g. Cardinaletti and Starke 1999, Giorgi and Pianesi 1997,
Rizzi 1997b) with the effect of enforcing the choice of the minimal
structure compatible with well-formedness. As of the mid 1980s, prin-
ciples of representational and derivational economy came to the fore
of syntactic theory. (See also the introductions to the concepts and
techniques of minimalist syntax in Radford 1997, Uriagereka 1998.)


31
On nature and language

6.2 Representational and derivational economy
As for the first kind, an important role was acquired by the principle
of Full Interpretation, according to which at the interface levels every
element must be licensed by an interpretation. So, if computational
processes involve the presence of uninterpretable elements on some
level of representation, they must have disappeared by Logical Form
(LF). For instance, expletive elements like there, necessary to express the
obligatory subject position in constructions such as (34a), do not have
a referential content, and presumably don™t receive any interpretation
at all at Logical Form, hence they must disappear before this level
is reached, under FI. One classical approach to this problem is the
hypothesis that the expletive is replaced by the contentive subject at LF,
another instance of covert movement, yielding an LF representation
like (34b), which respects Full Interpretation (but see Williams (1984),
Moro (1990) for a different analysis).

(34) a. There came a man
b. A man came

This analysis immediately accounts for the fact that the relation
between the expletive and the contentive subject is local in the same
sense in which argumental chains are (e.g. the relation between the
surface subject of a passive sentence and its “trace,” the empty object
position in which the surface subject is semantically interpreted: “John
was fired ”; by and large, both relations must obey locality con-
straints like Relativized Minimality, on which see below): the same
kind of configuration holds at LF in both cases (Chomsky (1986a),
based on observations in Burzio (1986)).
At the derivational level, economy was expressed by a principle
stating that movement is a last-resort operation: there is no “free,” truly


32
Editors™ introduction

optional movement, every extension of a chain must be motivated by
some computational need. A good intuitive illustration of this idea
is provided by the movement of the verb to the inflectional system,
which is motivated by the need for the verb to pick up affixes of tense,
agreement, etc. which do not constitute independent words: so, certain
kinds of movement are motivated by the need to express the structure
as a sequence of well-formed and pronounceable words. This kind
of connection between movement and morphological requirements
also helps explain certain diachronic generalizations: English lost verb
movement to the inflectional system concomitantly or shortly after a
radical weakening of the inflectional paradigm (Roberts (1993) and
much related work). Many factors of complexity must be taken into
account, but a basic correlation between inflectional richness and verb
movement appears to hold quite robustly, at least in Romance and
Germanic (see also Vikner (1997) and references quoted there).
A direct illustration of movement as last resort is provided by the
pattern of past participle agreement in Romance. The past participle
does not agree with an unmoved direct object, e.g. in French, but it
does if the object has been moved, say, in a relative construction:

(35) Jean a mis(— e) la voiture dans le garage
Jean has put(— Agr) the car in the garage
(36) La voiture que Jean a mise dans le garage
The car that Jean has put+Agr in the garage

Following Kayne™s (1989) classical theory of participial agreement,
we may assume that agreement is triggered when the object passes
through a position structurally close to the past participle (technically,
the specifier of an agreement head associated to the participle). So,
the relevant representation must be something like (37), with t, t™ the


33
On nature and language

“traces” of movement (on this notion, see below). Now, the point is
that the direct object can transit through this position, but not remain
there: (38), with the object expressed in the pre-participial position, is
ungrammatical:

(37) La voiture que Jean a t™ mise t dans le garage
The car that Jean has put+Agr in the garage

(38) Jean a la voiture mise t dans le garage
Jean has the car put in the garage

Why is it so? Like every nominal expression, the direct object must
receive a Case, and presumably it receives accusative Case in its canon-
ical object position (or anyhow in a position lower than the participial
verb). So, it has no reason to move further, and (38) is ruled out because
of the “useless” movement step. On the other hand, in (37) the object,
as a relative pronoun, must move further to the left periphery of the
clause, so it can licitly pass through the position which triggers past
participle agreement.
The movement-as-last resort approach implies that there is no
truly optional movement. This has made it necessary to reanalyze ap-
parent cases of optionality, often leading to the discovery of subtle in-
terpretive differences. For instance, the so called “subject inversion” of
Italian and other Null Subject Languages, previously analyzed as a fully
optional process, turns out to involve a necessary focal interpretation
of the subject in postverbal position, or a topic interpretation signaled
by an intonational pause and destressing (Belletti 2001):
(39) a. Maria me lo ha detto
Maria said it to me
b. Me lo ha detto Maria
Said it to me Maria(+Foc)


34
Editors™ introduction

c. Me lo ha detto, Maria
Said it to me, Maria(+Top)

6.3 Uninterpretable features
Derivational and representational economy met in the idea that
syntactic movement is always triggered by the goal of eliminating unin-
terpretable elements and properties. A specification typically consid-
ered uninterpretable is structural Case (nominative and accusative): an
element bearing nominative Case in English can bear any thematic role
(Agent, Benefactive, Experiencer, Patient/Theme), and even no role at
all, as in (40e):

(40) a. He invited Mary
b. He got the prize
c. He saw Mary
d. She was invited/seen by John
e. There was a snowstorm

Accusative is equally blind to interpretive thematic properties:

(41) a. I expected [him to invite Mary]
b. I expected [him to get the prize]
c. I expected [him to see Mary]
d. I expected [her to be invited/seen by John]
e. I expected [there to be a snowstorm]

Other types of Case, inherent Cases, are linked to specific thematic
interpretations: in languages with rich Case systems, an argument
marked with locative Case designates a location, etc., but nominative
and accusative appear to be thematically blind: in this sense, they are
considered uninterpretable (languages allowing subjects with oblique
Case, so called quirky subjects, apparently allow nominals sharing


35
On nature and language

both types of Case properties: Zaenen, Maling, and Thrainsson (1985),
Bobaljik and Jonas (1996), Jonas (1996), Bobaljik (1995), Sigurdsson
(2000) and references quoted there).
Another feature specification which is considered uninter-
pretable is the grammatical specification of person, number, and gen-
der (and other analogous specifications such as the class specification
in Bantu languages) which appears on predicates, e.g. in the following
Italian example:

(42) La ragazza ` e stata vista
The girl FS3P hasS3P beenFS seenFS

The specification of gender, number, and (by default) person on the
noun phrase la ragazza in (42) has an obvious interpretive import, but
this specification in the predicate (reiterated on the inflected aspectual
auxiliary, on the passive auxiliary and on the passive past participle
in Italian) is redundant and, as such, is considered non-interpretable:
external systems interpreting linguistic structures will certainly want
to know if the sentence is talking about one girl or many girls, but
the reiteration of this information on the predicate does not seem to
add anything of interpretive relevance. In fact, predicates not reiterat-
ing the feature specification of the subject in non-finite structures,
or in morphologically more impoverished languages, are perfectly
interpretable. (In some cases, an agreement specification seems to
have consequences for interpretation, as has been argued for the par-
ticipial agreement in French discussed above (Obenauer 1994, D eprez ´
1998), but this may be an indirect effect of the theory of reconstruction
(Rizzi 2001b).)
Movement is seen, in this system, as a way of eliminating
uninterpretable features. For instance, movement of a direct ob-
ject to the subject position has the effect of placing it in a local


36
Editors™ introduction

environment in which its uninterpretable Case feature can be checked
off by the agreeing inflectional head, an operation which, simultane-
ously, checks off the uninterpretable agreement features on the inflec-
tional head. The checking off of a feature amounts to its elimination
from the derivational path leading to the computation of a Logical
Form. So, movement is last resort in that it must be motivated by
the goal of eliminating uninterpretable features, an elimination which
in turn makes it possible to satisfy Full Interpretation at the interface
representations (Chomsky 1995a). An element undergoing movement
must have an inner motivation to move, an uninterpretable feature
specification to eliminate. For instance, going back to the French
construction (36), the object cannot move because it has its unin-
terpretable accusative Case already checked in its base position (or,
anyhow, in a position lower than the participial specifier), and it has
no other feature which would make it “active,” i.e. available for further
movement. In case the object must undergo further movement, e.g.
to the relative complementizer, as in (37), it will have whatever unin-
terpretable features are involved in left-peripheral movement in this
system (Grewendorf 2001), which will make it a suitable candidate for
movement.
In spite of its teleological flavor, the principle of movement
as last resort can be implemented in a very elementary way, taking
only local decisions and not requiring computationally complex pro-
cedures such as transderivational comparisons, look-ahead, and the
like (on local economy, see Collins (1997)). A distinct but related case
of the limitation on movement imposed by economy considerations is
the proposal (Chomsky 1995a, 2000a, 2001a) that Merge, the funda-
mental structure-building operation, preempts movement whenever
both operations are applicable to satisfy computational needs. A case
that illustrates this point is the peculiar distribution of nominals in


37
On nature and language

expletive constructions. The expletive construction (43a) suggests that
the nominal a man is first introduced in the structure as the subject of
the locative in the garden, functioning as a predicate: if no expletive is
selected, the nominal is moved to the subject position of the copula,
otherwise the expletive is inserted:

(43) a. There is [a man in the garden]
b. A man is [t in the garden]

Now, in a more complex structure involving a higher raising verb, such
as (44a), the following peculiar constraint emerges: either no expletive
is selected, and the nominal moves all the way to the subject position of
the higher raising verb (as in (44b)), or an expletive is inserted already
in the embedded clause and then is raised (as in (44c)); the a priori
remaining option (44d), with the nominal moved to the embedded
subject position and the expletive inserted in main subject position, is
excluded:

(44) a. seems [ to be [a man in the garden]]]
b. A man seems [ t™ to be [ t in the garden]]]
c. There seems [ t to be [a man in the garden]]]

d. There seems [a man to be [t in the garden]]]

It appears that, if an expletive is selected, it must be inserted as soon
as possible: the structure (44d), involving partial movement of the
nominal and then expletive insertion, is excluded. Why is it so? A simple
explanation of this paradigm is provided by the assumption that merge
is less costly than movement, so that, in (44a), if an expletive has been
selected, the option of merging it as the subject of be will preempt the
option of moving a man to that position (Chomsky (2000a); see also
Belletti (1988), Lasnik (1992), for alternative analyses in terms of the



38
Editors™ introduction

Case requirements of the nominal, and Moro (1990) for an analysis
based on the idea that the expletive is a pro-predicate).



6.4 Locality
The study of locality is an independent, important research direction
in modern formal linguistics which points to the role of economy in
language design. If there is no upper bound to the length and depth of
linguistic expressions, as a consequence of the recursive nature of natu-

<< . .

 4
( 19)



. . >>

Copyright Design by: Sunlight webdesign