HOW COMPARATIVE IS SEMANTICS? A UNIFIED PARAMETRIC THEORY OF BARE NOUNS AND PROPER NAMES*

Pages 22
Views 4

Please download to get full document.

View again

of 22
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Description
GIUSEPPE LONGOBARDI HOW COMPARATIVE IS SEMANTICS? A UNIFIED PARAMETRIC THEORY OF BARE NOUNS AND PROPER NAMES* One of the two central suggestions put forth in Longobardi (1991, 1994) was that Romance/English
Transcript
GIUSEPPE LONGOBARDI HOW COMPARATIVE IS SEMANTICS? A UNIFIED PARAMETRIC THEORY OF BARE NOUNS AND PROPER NAMES* One of the two central suggestions put forth in Longobardi (1991, 1994) was that Romance/English differences in the syntax of proper names were parametrically connected to supposed differences in the semantics of bare (plural and mass) common nouns (BNs). The present article will pursue this line of investigation, trying to make precise such meaning differences and to understand the reason for their apparently surprising parametric association with the syntax of proper names. It will be shown that in most Romance varieties BNs, unlike their English counterparts, distribute their existential and generic readings across all different contexts exactly like (Romance and English) overt indefinites. All the differences will be unified under the proposal that Romance BNs are nothing but a type of indefinites (variables, existentially or generically bound) in Kamp-Heim s DRT sense, while English BNs are rather systematically ambiguous between this quantificational interpretation and a referential (i.e. directly kind-denoting, much in the spirit of Carlson 1977a, b) one, providing for another type of generic reading. The analysis will therefore crucially exploit and empirically support Gerstner and Krifka s (1987) distinction between referential and quantificational genericity. On such grounds we will finally gain a conceptual understanding of the typological implication originally established in Longobardi (1991, 1994), thus confirming that the strategies of interpretation of nominals, whether proper or common nouns, are basically one and the same, though differently parametrized in different languages. This result, in turn, will shed some light on the question whether comparative semantics is possible and whether it can be singled out as a legitimate independent component of parametric theories of grammatical variation. 1. INTRODUCTION: VARIATION IN S YNTAX AND I NTERPRETATION At least two developments have characterized the progress of general linguistic theory for the last two decades: the emergence of the formal study of the syntax-semantics mapping and that of Principles&Parameters theories of grammatical variation, the latter perhaps the main achievement of the linguistic sciences after and along with the historical-comparative method. Grammatical variation (to be roughly understood as the overall linguistic diversity once we subtract Saussurean lexical arbitrariness) as investigated for the morphosyntactic components in parametric theories is supposed to be: * I am indebted to C. Boeckx for providing me with some bibliographical material and to G. Carlson, G. Chierchia, D. Delfitto, and an anonymous referee for discussion or comments on previous drafts of this paper. Natural Language Semantics 9: , Kluwer Academic Publishers. Printed in the Netherlands. 336 GIUSEPPE LONGOBARDI (1) a. finite b. discrete c. limited (with respect to the number of actual phenomenic points of contrast exhibited by surface variation across languages) In other words, there is assumed to be a finite set of parameters, each with a finite number of predefined values (in principle two), and under each parameter cluster together several surface differences. Now, in the light of the parallel development of theories of the mapping between syntax and semantics at least three questions arise naturally: (2) a. Does variation exist in the semantic component? b. Does it display the classical parametric properties mentioned above? c. Is it always/sometimes independent of morphosyntactic variation? Questions of a similar sort have been put on the research agenda most forcefully by Chierchia (1996). In short, one may ask whether comparative semantics is possible, which form it may take, and how independent it is of (comparative) syntax. The optimal case in point to investigate such problems should be provided, in principle, by instances of syntactic homonymy across languages; by this term let me understand cases in which what appears as roughly the same surface syntactic shape clearly corresponds to distinct logical representations in different languages. An ideal testing ground in this sense is the syntax-semantics mapping of determinerless nominals, for the combination of two reasons: the interpretation of such nominals has been a major focus of inquiry since (and thanks to) Carlson s (1977a, b) groundbreaking work and, at the same time, a good deal of comparative material has more recently been brought to light, especially in the Romance-Germanic domain. While many Germanic and Romance languages, in particular English and Italian, hardly display obvious structural contrasts in the process of interpretation of overtly determined argument nominals (e.g. every man, the man, a man, two men,... and their literal correspondents), the situation is much more intriguing and problematic in the case of overtly determinerless nominals. In fact, while one of the two main types of argument determinerless nominals, proper names, seems to display a crosslinguistic constant semantics, but a variable syntax, quite the opposite is likely to be the case for the other type, namely bare (mass/plural) common nouns (henceforth BNs). In this article I will try to provide some preliminary answers to the general questions above, analyzing the relation between the syntax and the semantics of precisely these two types of entities. HOW COMPARATIVE IS SEMANTICS? BARE N OUNS AND P ROPER N AMES In the Romance and Germanic languages, nouns may normally occur in singular or plural form and, if morphologically singular, are subject to either a mass or a count interpretation. In most of these languages these three types of head nouns may all occur superficially determinerless (not overtly introduced by any member of a class of mutually exclusive items called determiners: usually a definite or indefinite article, a quantifier, or a demonstrative) in argument function, but according to their semantic behavior within argument phrases they seem to fall into two pretheoretically well distinguishable classes, exemplified in (3) and (4) below, respectively: 1 1) determinerless arguments with singular non-mass interpreted head nouns appear to denote a definite (roughly, unique in the domain of discourse), specific (i.e. particular) entity, tend to be rigid designators in Kripke s (1980) sense, and to always assume the widest possible scope (no logical operator or intensional context may take them within its scope); 2) determinerless arguments with mass or plural head nouns never denote a definite specific entity and in certain instances cannot assume scope over any logical operator (Carlson 1977a, b): (3) Ho incontrato Maria. I met Maria. (4) a. Bevo sempre vino. I always drink wine. b. Ho mangiato patate. I ate potatoes. The set of head nouns entering the first type of argument is essentially lexically defined (though in a non-trivial way; cf. Longobardi 1996) and is roughly coextensive with the traditional class of proper names; the second type may virtually concern all nouns, including in particular traditional common nouns. Assuming Carlson s (1977a) ontological partition of individual entities into objects and kinds, let us suppose that the head nouns entering type (1) phrases are by themselves, i.e. as part of their lexical meaning, object- 1 For present purposes I will class among determiners also cardinals and certain quantity expressions, such as molto much, poco little, or abbastanza enough. For subtler distributional distinctions among determiners cf. Szabolcsi (1994), Longobardi (2000b). Note that throughout the paper, English translations of the Italian examples are often word-by-word glosses, thus occasionally ungrammatical or infelicitous. 338 GIUSEPPE LONGOBARDI naming items (thus, are capable of referring to objects), whereas those only entering type (2) phrases are kind-naming items. I will say, further, that all nominal arguments (i.e. potentially phrasal constituents, not just heads) denote entities taken, in principle, from the Carlsonian ontology presupposed by natural language, therefore again objects and kinds; I will furthermore assume that, essentially by virtual conceptual necessity, they may be of two types: quantificational and referential. Quantificational arguments denote via the intervention of a variable, while the referential ones denote as constants just on the grounds of the lexical referring potential of the noun. The kind-naming meaning typical of common nouns will be taken as always able to provide a predicate used in the restrictor of a variable, while the object-naming meaning of proper names cannot do so, except under special selective conditions (cf. Longobardi, forthcoming, for some discussion). Thus, nominal arguments formed by an overt determiner and a common noun will be said to denote objects through an essentially quantificational structure, in which the lexically kind-naming meaning of the head noun is used as a predicative restriction (expressing membership in the extension of the kind) for the variable bound by the determiner (its selective operator; cf. Longobardi 1994). In the case of determinerless proper names (type (1)), on the contrary, the object-denoting effect of the argument will be regarded as a direct consequence of the lexical object-naming nature of the noun itself: therefore the nominal argument as a whole will be a(n) (object)-referential expression. If so, determined common nouns and determinerless proper names represent the two extreme or pure cases of variable (quantificational) and constant (referential) interpretation for an argument. Much of the following discussion will address the problem of how, instead, determinerless arguments formed on mass/plural nouns (the subtypes of (2)) behave within this system. Notice, first of all, that these two subtypes behave alike from essentially all relevant viewpoints, except precisely for the mass/plural distinction; this is why I will systematically refer to them under the collective term bare nouns (BNs). 2 2 It is in fact hardly felicitous to talk about the relevant properties of these items by referring to them just as bare plurals, as has often been the case in the literature, precisely because this obscures the fact that plurals and mass nouns behave alike in all important respects (on this point also cf. Chierchia 1996; Carlson 1999). For the whole question see Delfitto (to appear). The term bare nouns, used in the intended sense (i.e., excluding bare count singulars), is meant to suggest that plurals and mass terms form a crosslinguistic natural class, parametrically well distinguished from bare singulars, though perhaps subject to the same general interpretive conditions, along lines discussed in some more detail in Crisma (1997, 1999). HOW COMPARATIVE IS SEMANTICS? 339 While determinerless proper names instantiate quite different surface structures in English and Romance, BNs can be shown to essentially display the same superficial syntax in the two language types (cf. Longobardi 1994 and Section 5 below). 3. ROMANCE B ARE N OUNS AS I NDEFINITES 3.1. The Two Readings As in all West European languages, BNs in Romance are subject to just two fundamental types of interpretation: the existential one (henceforth Ex ) and the generic one 3 (henceforth symbolized as Gen, to be taken as a mere epiphenomenal label for very different types of genericity 4 ). Three descriptive proposals have so far been made as to the interpretation of Italian BNs (some other Romance languages, perhaps all except for Portuguese, appear to be at best equally restrictive; therefore, as a working hypothesis, I will tentatively use Romance throughout the paper as referring to the whole family, with the important proviso that we already know there to be exceptions, at least in some Portuguese varieties: cf. e.g. Munn and Schmitt 1999a, b): 5 Casalegno (1987): only Ex, unlike English Longobardi (1994): Ex, sometimes Gen (but only with I(ndividual)-level predicates, thus with a distribution unlike the English one) Chierchia (1996): Ex, Gen, distributed essentially as in English In this work it will be shown that, taken literally, all these proposals are 3 For the hardly definable semantics of such readings cf. the various essays collected in and referred to in Carlson and Pelletier (1995). Here we take generic lato sensu to also cover the readings identified by Condoravdi (1994) and termed functional. 4 One of the aims of the present research is precisely that of contributing to tell apart the various sources of genericity in Romance and Germanic, which until very recently have too often been gathered under a single heterogeneous and spurious category, rather reminiscent of other temporary entities in the history of the field, e.g. Proto-Indoeuropean /a/ before Brugmann and Saussure or, perhaps, autonomous phonemic representations. 5 Also cf. Dobrovie-Sorin and Laca (1996) and Benedicto (1997) for some recent discussion. I will analyze the interpretation of Romance BNs as methodologically independent of the peculiar constraints on their distribution, which are rather widely discussed in the literature (e.g. cf. Contreras 1986; Delfitto and Schroten 1992) and can anyway be circumvented by adding some adjectival or relative modification to the BN. For the syntax of Romance BNs cf. also Lois (1986), Torrego (1989), Longobardi (1994), among others. 340 GIUSEPPE LONGOBARDI inadequate; it will be argued, instead, for the following descriptive generalizations: (5) a. The distribution of the interpretations of Italian BNs = that of Italian overt indefinites = that of English overt indefinites in the same environments. b. The distribution of the interpretations of Italian BNs that of English BNs. c. Italian BNs (unlike English BNs) are generic only in independent quantificational environments. d. It is Italian overt definites that can be generic in all (pragmatically appropriate) environments, like English BNs. In the next three sections I consider the distribution of the two readings for BNs with the main potentially relevant types of predicates (all the following paradigms abstract away from the subkind or taxonomic reading 6 ) S(tage)-Level Predicates With a S(tage)-level predicate there are three main subcases to consider: a) episodic sentences; b) characterizing sentences (in the sense of Carlson and Pelletier 1995, i.e. gnomic); c) episodic sentences with a generalizing adverb. (6) a. Elefanti di colore bianco hanno creato in passato grande curiosità. Ex White-colored elephants raised a lot of curiosity in the past. b. Elefanti di colore bianco possono creare grande curiosità. Gen/?Ex White-colored elephants may raise a lot of curiosity. c. Elefanti di colore bianco hanno creato sempre/spesso in passato grande curiosità. Gen/?Ex White-colored elephants always/often raised a lot of curiosity in the past. As can seen, the generic reading of a subject BN with S-level predicates is possible (pace Casalegno 1987 and Longobardi 1994), but appears to 6 Following Carlson and Pelletier (1995: Introduction), by taxonomic I refer to readings where the denotation of the nominal ranges over kinds (subkinds) rather than objects, as e.g. in Many prehistoric animals have become extinct. HOW COMPARATIVE IS SEMANTICS? 341 depend on the presence of a DP-external operator of generality, such as the habitual verbal aspect responsible for the characterizing meaning of (6b) or a quantificational adverb (presumably binding the subject à la Lewis 1975) as in (6c). In a way, the subjects of such predicates seem to acquire genericity through generalization from an indefinite series of singular events I(ndividual)-Level Predicates I-level predicates are supposed to be always characterizing by their lexical meaning (cf. Carlson and Pelletier 1995: Introduction, and especially Chierchia 1995); thus a subject BN should normally be able to be Gen and only Gen. This appears correct: (7) Cani da guardia di grosse dimensioni sono più efficienti. Gen Watch dogs of large size are more efficient. On such grounds, Chierchia (1995) suggested that genericity here is the direct product of a lexical property of I-level predicates. However, serious doubts can be cast on this idea, for the situation is more complicated: first, as was the case in (6a) above, the availability of the generic reading becomes much more degraded as soon as the present (habitual) tense of (7) is replaced by a tense implying an episodic interpretation; therefore, the presence of a characterizing (habitual) aspect appears to be crucial to license genericity with I-level predicates as well. Second, I-level predicates are at least split in two with respect to such phenomena (all judgments through (10) are given w.r.t. Gen, since the Ex readings are awkward anyway: (8) a. Stati di grandi dimensioni sono pericolosi. States of large size are dangerous. b.??stati di grandi dimensioni sono prosperi. States of large size are prosperous. (9) a. Cani da guardia di grosse dimensioni sono più efficient/ aggressivi. Watchdogs of large size are more efficient/aggressive. b.??cani da guardia di grosse dimensioni sono più pelosi/neri. Watchdogs of large size are more hairy/black. (10) a. Uccelli di zone paludose sono ghiotti di insetti. Birds from marshy areas are greedy for insects. b.??uccelli di zone paludose sono scuri/intelligenti. Birds from marshy areas are dark/intelligent. 342 GIUSEPPE LONGOBARDI Let us call the two subclasses A and B, respectively. What is the rationale for this split in the class of I-level predicates? A tentative conjecture is the following: notice that class A predicates are somewhat more eventive than those of class B, which are more stative : if so, we may hypothesize that the present tense here instantiates a formal super-category of durational or imperfective aspect, superficially neutralizing the semantic distinction eventive/stative; but only the imperfective aspect associated with an eventive aktionsart would count as habitual. Thus, the latter aspect would never be selected by stative class B predicates; but only an habitual aspect, the one semantically selected just by class A, would provide a real generic operator. If so, even bare subjects of I-level predicates would acquire genericity through generalization from singular events. 7 Not surprisingly, an explicit adverb of generality like usually can restore full acceptability of Gen in all the (b) examples. Therefore it is likely that the factors licensing the generic reading of BNs with S-level and I-level predicates are just the same (an aspectual operator of habituality, essentially as suggested by Delfitto 1997, or an adverb of generality) and have actually nothing to do with the distinction between these two predicate types. 8 To apply a test suggested in Carlson and Pelletier (1995: Introduction), note further that in all cases considered the generic flavor seems a property of the sentence, not of the subject DP, in other respects as well: for it is retained even if the BN is replaced by a definite specific singular or a proper name, which are certainly object- and not kind-denoting expressions. The descriptive discoveries made so far about the readings of BN subjects in the context of the various predicates can be summed up, for just practical purposes, as follows: (11) S-level a. episodic Ex b. characterizing (habitual aspect) Ex/Gen I-level c. lexically characterizing A (habitual aspect) Gen d. lexically characterizing B (non-habitual aspect) The existence of case (11b) is what provides for the occurrence of true ambiguities, as noticed by Diesing (1992) for English, exemplified below: 7 I am indebted to P. Bertinetto for discussion of some terminological issues relative to these distinctions. 8 These conclusions reinforce, then, those advanced by Higginbotham and Ramchand (1997) ab
Advertisements
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks
SAVE OUR EARTH

We need your sign to support Project to invent "SMART AND CONTROLLABLE REFLECTIVE BALLOONS" to cover the Sun and Save Our Earth.

More details...

Sign Now!

We are very appreciated for your Prompt Action!

x