«SCOPE PARALLELISM AND THE RESOLUTION OF ELLIPSIS AT THE SYNTAX/SEMANTICS INTERFACE Gun ERRE Z- REXACH JAVIER In this paper several scope asymmetries ...»
SCOPE PARALLELISM AND THE RESOLUTION OF ELLIPSIS
AT THE SYNTAX/SEMANTICS INTERFACE
Gun ERRE Z- REXACH
In this paper several scope asymmetries in VP ellipsis constructions in
English and Spanish are studied. It is argued that an approach based on
Fox' (1995 a,b) Ellipsis Scope Generalization faces numerous conceptual
and empirical problems. Ellipsis resolution is conceived of as a phenomenon belonging to the conceptual-intentional pan of the computational system that is conditioned by the computation of the semantic features of quantifiers at LF. A semantic approach inspired in higher order unification theories of ellipsis is defended. This approach is compatible with the overall philosophy of minimalist grammar: the process of higher-order unification complies with the principle of inclusiveness (Chomsk-y. 19 or projection from the lexicon because the relevant semantic equations are set up and resolved at a discourse level, but they are determined by featuresensitive Logical Forms.
0. INTRODUCTION A widespread characteristic of natural languages is the ability to recover the content of missing fragments of a sentence from the preceding
material. Consider the following examples from English and Spanish:
(1) (a) Jill will come to the party but Jane won't I have bought many books recently. but I will only read some (b) Bill invited two candidates. I can't remember which ones (c) (d) Juan se quitó las gafas y María no lo hizo (e) Yo compré un libro y Pedro también (fl He leído tus libros y los de Pedro (g) Sé que desayunaron algo pero no sé el qué Universidad de Huelva 2009 52 JAVIER GUTI$RREZ-REXACH The above examples are instances of the class of phenomena commonly called ellipsis and illustrate several different sub-types, such as VP ellipsis, N' or N deletion and sluicing. Several syntactic analyses of these constructions have tried to identify the conditions under which ellipsis takes place and the elements that determine differences or similarities in interpretation (see Brucart  for an excellent analysis of ellipsis in Spanish). A standard syntactic solution within the Government and Binding paradigm would be to postulate the existence of empty categories in the elliptic segment and to study the licensing conditions of those empty elements. The emergence of the Minimalist Program (Chomsky, 1995, 1997, 1998 a,b) demands a stricter solution that does away with empty categories and X-bar related notions such as government or the Empty Category Principle. The elimination of empty categories follows Chomsky's (1995) "Inclusiveness Principle", which requires that only lexical elements drawn from the initial numeration be participants in grammatical derivations. This principle is obviously reminiscent of similar criteria in other frameworks, such as categorial grammar and unification-based formalisms where empty categories and representational notions such as the above mentioned ones are not present (see Gutiérrez-Rexach [19981 for a generalization of a formalized minimalist framework that takes into account some of these considerations).
In this paper, the syntactic and semantic properties of a group of scopal restrictions that arise in ellipsis constructions are studied. A potential explanation of these restrictions (Fox, 1995 a,b) is based on a global principle of Economy, which favors derivations that are less complex or consist of fewer derivational steps. Alternatively, it can be argued that the proper explanation bears heavily on semantic operations closely related to the general process of ellipsis resolution and that global Economy criteria have undesired consequences and should be dispensed with, as argued by Johnson and Lappin (1997) and Chomsky (1998b). An explanation of the latter type will be defended in this article, which is structured as follows: in the first section, the received view on the interactions between scope and VP ellipsis is described and, in section two, Fox's (1995a,b) Economy-based analysis is presented. In the next two sections, it is shown how this approach faces some empirical and theoretical problems and, finally, in the last sections an alternative analysis based on parallelism and higher-order unification is developed.
This analysis successfully accounts for the data discussed in the previous sections.
1. SCOPE ASYírLMETR1ES AND VP ELLIPSIS Sag (1976) and Williams (1977) noticed an interesting contrast that.
arises in the interaction between VP ellipsis and scope interpretation.
Sentence (2a) is ambiguous: the scopal order may be either the one corresponding to the surface linear order (SOME EVERY), or the quantifier in object position may scope over the quantifier in subject position (EVERY SOME). In a coordinate construction in which the second VP is elided, as in (2b), there is no ambiguity. The object wide scope reading is not available. The only possible reading is the one in which there is a unique boy who admires every teacher (SOME EVERY).
(2) (a) Some boy admires every teacher (b) Some boy admires every teacher and Mary does too Within May's (1985) theory of Logical Form, the two potential Logical Form representations (LFs) for (2a) are as in (3). The subject is generated in the specifier of IP and the object as a complement of the verb within the VP. In the subject wide scope reading, the subject quantifier raises to a position adjoined to IP and the object adjoins to VP.
Therefore, by the scope principle the subject scopes over the object. In the object wide scope reading, the subject adjoins to IP and the object adjoins to a position c-commanding the subject.
Let us now consider the scope disambiguation process involved in (2b). Under May's theory of LF, VP ellipsis resolution requires copying/ reconstructing the antecedent VP into the elided conjunct, so in the case where the LF corresponding to the antecedent clause is (3a), the constituent that undergoes the copying operation is íp every teacher;
f admires t, JI. The resulting LF for (2b) is well-formed, as shown in (4):
LF in which the reconstructed VP [, admires t1 Jhas an unbound trace, as in (5). The LF representation is ill-formed because the quantifier every teacher does not c-command the trace in the second conjunct, so it cannot bind it.
Hirschbühler (1982) showed that this type of approach incorrectly predicts that the quantifier in the object position must always have narrow scope. In the following example, the quantifier every building may scope over the subject quantifier yielding the most natural interpretation of the sentence: for every building x there is a different Canadian flag y in front of it (EVERY A).
(6) A Canadian flag is in front of every building and an American flag is too The May-style LF representation for (6) is predicted to be ill-formed, because the trace of every building would be ungoverned in the second conjunct after reconstruction.Therefore, May's explanation appears to be too restrictive. Cormack (1984) and Diesing (1992) related the contrasts in (2) to the presence of a proper noun in the second sentence. Diesing observed that the cases in which the object quantifier is forced to a narrow scope interpretation are those in which the overt NP in the correlate clause is "non-quantificational", i.e. a proper noun or a definite, as her examples in (7) illustrate.
(7) (a) Some bassoonist played every sonata, but Otto didn't (b) Every lawyer liked some decisions, but the doctor didn't Diesing also argued that when the NP in the second clause is quantificational, the expected scope interactions arise. The object wide scope interpretation of the second conjunct of the sentences in (8) is not blocked.
Summarizing, it seems evident that an account in terms of the classical GB theory of LF does not account for the whole array of semantic facts presented. In the next section a syntactic theory is presented that is cast within a minimalist framework. This theory has strong theoretical implications with respect to the role of Economy at the syntax/semantics interface and the determination of scopal relations in a minimalist grammar. After presenting Fox's (1995a; 1995b) account, I argue that there are important conceptual and empirical inadequacies in his theory, based on the Ellipsis Scope Generalization, and I present an alternative solution which dispenses with the necessity of global Economy in the determination of scopal differences. This solution is consistent with recent analyses of ellipsis that make use of higher order unification mechanisms.
2. ELLIPSIS AND ECONOMY
Fox (1995a, 1995b) presents a theory of scope interactions in elliptic constructions that attempts to derive their properties from general assumptions, of the Minimalist Program (Chomsky, 1995). More concretely, the scopal behavior of quantifiers follows from the
generalization in (9):
Ellipsis Scope Generalization (ESG):
(9) The relative scope of two quantifiers, one of which is in an antecedent VP of an ellipsis construction, may differ from the surface c- command relation only if the parallel difference will have semantic effects in the elided VP.
The generalization is true with respect to the examples that we have considered so far_ For instance, in sentence (2b) the scopal order of the quantifiers in the second conjunct, after reconstruction, does not yield a difference in truth conditions. The following LFs are truth-conditionally
Fox assumes, contra Cormack and Diesing, that proper names and definites are quantificational and that in the LFs above we have a genuine interaction of two quantifiers. Since the two LFs are equivalent,
the ESG predicts that the scopal order of the quantifiers in the source clause or antecedent VP is identical to their surface order, as in (3a).
Similarly, in (7b) only the scopal order EVERY SOME is allowed in the source clause, since the permutation of the scopal order of the definite in the reconstructed clause and any other quantifier yields truthconditionally equivalent LFs. Hirschbühler's example in (6) and the variant of sentence (2b), where the quantifier some girl is substituted for Mary, are predicted by the ESG to display a scope ambiguity in the source clause by the ESG, as in fact they do. In sentence (11a), after VPreconstruction, the quantifier some girl interacts with the quantifier every teacher yielding two possible orders, shown in (11b,c).
The two LFs above are trivially not equivalent,' and according to the ESG the scopal order of the quantifiers in the source clause may be different from the surface order. In other words, the object wide scope configuration is well-formed.
Fox brings in new empirical data from English to support the validity of the ESG. For instance, the combination of two universal quantifiers
yields two equivalent LFs:
Again, according to the ESG, only the scopal order of the quantifiers in the source clause which is identical to their surface order, i.e.
SUBJECT OBJECT, is allowed, since the OBJECT SUBJECT order of the quantifiers in the LF corresponding to the reconstructed clause (13b) would not have any semantic effect. It would not be semantically different from the scopal order that preserves the surface c-command- order.
Universidad de Huelva 2009
SCOPE PARALLELISM AND THE RESOLITION OF ELLIPSIS... 57Fox claims that the ESG follows from two assumptions independently needed: (i) Parallelism and (ii) Economy. Economy dictates that the object can move by QR over the subject only if the movement yields an interpretation which would be unavailable otherwise. Parallelism dictates that an operation applies in one conjunct if and only if a parallel instance of the same operation applies in the other conjunct. In the cases that we have been considering, Economy prevents an application of QR raising the object quantifier over the subject quantifier and yielding the LF OBJECT OBJECT scopal order if the resulting LF is truth-conditionally equivalent to an LF in which that operation has not applied. \'hen a proper name or a definite quantifier subject interact with any other quantifier, an application of QR is semantically inert and violates Economy.
When two universal quantifiers interact, any scope shifting operation is also semantically inert and, as a consequence, uneconomical. Parallelism prevents QR from applying in one conjunct without applying in the other conjunct. Therefore, if QR cannot apply in the reconstructed VP to avoid a violation of Economy it cannot apply in the source clause either to avoid a violation of Parallelism.
3. GLOBAL VS. LOCAL ECONOMY AND THE ORDERING ASYMMETRY PROBLEM