[x rivista di grammatica generativa anno 1997 n.22 uni pressRIVISTA DI GRAMMATICA GENERATIVA Volume 22, anno 19967 Paolo ACQUAVIVA Deriving [+ strong] features from morphology: head movement p.3 Franco BENUCCI Sintassi del verbo e deittici cltici nelle iscri- zioni sudpicene p.37 Cristiano BROCCIAS Binding theory and its consequences for minimalist theory p.93DERIVING [+ STRONG] FEATURES FROM MORPHOLOGY: HEAD MOVEMENT Paolo ACQUAVIVA 1. Introduction and theoretical preliminaries 1.1, Morphological variation and the [+ strong] distinction Why and how do languages differ from one another? The question has always been a central concem for any theory of grammar aiming at explanatory adequacy, but a principled and unitary answer could not be attempted before the development of the so-called Principles and Parameters framework, ushered in by Chomsky’s Pisa lectures in 1979 and codified in his Lectures on Government and Binding (Chomsky 1981). The subsequent developments that culminated in the radically novel Minimalist Program (Chomsky 1995) privileged one line of thought among the various approaches to comparative grammar that had emerged within the Government-Binding model: namely, the leading insight, first extensively articulated by Huang (1982), that the same syntactic processes (say, Wh-movement) can be overt or non-overt across languages. In this perspective, a large part of Variation between grammatical systems is a misleading appearance caused by the different timing of essentially the same operations over essentially the same structures, The programmatic chapter 4 in Chomsky 1995 outlines a thory of grammar where differences between grammatical systems should eventually be analyzable in terms of a single formal concept: that of feature checking, whereby grammatically relevant “features” (the only primitive constituents of a linguistic representation) are introduced, moved and formally neutralized (or “checked”) in local domains, as steps in the construction of a well-formed syntactic representation. Feature checking is the unitary mechanism underlying syntax; it may or may not bring with it displacement of phonologically realized material (“overt movement”), depending on the [+ strong} nature of the features. In Minimalist terms, weak and strong features are the same type of object, except that strong features must be spelt out (by stipulation, a derivation must check a strong feature on category C before it can construct nodes not headed by C itself;Paolo Acquaviva Chomsky 1995: 233-234). The arbitrariness of this distinction has often been recognized, although there have been attempts to give it a morphological motivation (Pollock 1989, Vikner 1991, Roberts 1992, Sola 1996, Bobaljik 1995, 1997), But attention to the morphological legitimacy of the [2 strong] distinction should not obscure the existence of a distinct dimension of crosslngurstic variation: morphological features are grouped and organized differently across languages. That some languages but not others have a full-fledged case system is a morphological tact that must be leamed, just as it must be learned how many participles, for example, a given language has in its verbal system, or whether negation will be a particle, a verb, or even a noun (see Payne 1985). ‘This observation must be carefully distinguished from the observation that languages differ in morphological “richness”. As shown in (1), in a language like Finnish negation is a verbal auxiliary, but it is not morphologically richer than Fnglish forms like doesn’t or didn't. Actually, the negative auxiliary only expresses agreement and not tense, therefore doesn’t is richer because it expresses both, But doesn't and didn’t are not forms of a negative auxiliary, because negation is not expressed lexically on the stem of the verb do. There is a morphological difference, but it is not one of morphological richness. (1) a. Finnish: puhu-n e-npuhu e-n puhunat speak-Isg neg-Isg speak _ neg-1sg speak-pastpart.sg ‘speak’ ‘T.don’t speak" ‘I didn't speak’ b. English: doesn’t DO [age 3sg) [-past] didn’t DO [+past] Note that the clustering of morphosyntactic features is not the same thing as Saussurean arbitrariness of the sign. What is arbitrary is that the same concept, say, “day”, is mapped to the sounds day and giorno in different languages: but this is a matter of phonological realization and not of morphosyntactic features. In fact, Chomsky himself (1995: 235) explicitly recognizes this as a distinct dimension of crosslinguistic variation, in so far as he mentions language-specific “principles” covering “... aspects of phonology and morphology, choice of parametric functions, and whatever else may enter into language variation”. He also takes into account language-specific conditions on the realization of formal features (Chomsky 1995; 240-241: “Suppose, for example, that specific morphological properties of a language constrain the phonetic correlate of formal features ... It appears to be the case for some languages that “templatic” conditions are required for morphological structure.”). One could certainly object that more research will bring seemingly morphological variation within the grasp of a purely syntactic explanation, using no extra vocabulary than Feature Attraction and Checking Theory. But, as the cited remarks make clear, Chomsky himself does not embrace such a naive rejection of morphology popular among syntacticians (but rejected byDeriving [+s ong} features from morphology: head movement Baker (1988b)). His conclusion is rather that these aspects of grammar ultimately shed no light on the computational apparatus underlying UG (Chomsky 1995: 241: “Then the lexical entries will abstract from these [purely morphological] properties, presenting just the information that they do not determine”) ‘We have already mentioned the arbitrariness of the simple [+ strong] distinction, But the difference in strength has nothing to do with the clustering of features, cither in the abstract sense of morphosyntactic features relevant for syntax and interpretation, or in the purely morphological sense of features defining inflectional classes and ultimately phonological interpretation. Since any one feature, in Chomsky’s system, may be {+ strong] in any one language, the way features are clustered under a single terminal does not affect the choice between overt or covert movement, so that the two dimensions of variation I have indicated are not mutually related in any way. This is a significant weakness for a theory having the ambitious (hence interesting) aim of reducing all grammatically significant parametrization to feature strength. Indeed, the existence of morphological properties differing across languages and not reducible to phonology or syntax forces Chomsky to the characteristically cavalier assertion that such variation is not grammatically relevant, “.. morphology being a primary repository of exceptional aspects of particular languages ” (Chomsky 1995: 241). But denymg a problem does not solve it. 1.2, Outline of the argument In what follows I will treat both the arbitrariness of the (+ strong] distinction and the neglect for purely morphological variation as two genuine faults, and I will identify the culprit in an important assumption which Chomsky makes explicit in his section 4.2.2 (1995: 235-241): namely, that morphology is a catalogue of morphemes, supplemented if necessary by ancillary concepts which, however, are typically descriptive and in any case play no role in the workings of the computational system of human language. By morphology, then, Chomsky understands a sort of second-order phonology, a repository of information about “words” and smaller strmgs of phonological segments (affixes). This background assumption operates through two interrelated assumptions: that syntactic atoms are fully inflected and realized words, or vocabulary items, and that syntax is driven by formal properties of morphosyntactic features (Checking Theory). Because these features are the only relevant formal characterization of syntactic atoms, both overt and covert movement is triggered (and partly enacted) by them; for the same reason, additional information that may be morphologically significant (like the clustering of features) is irrelevant. Moreover, because syntax manipulates fully inflected and realized lexical items as opposed to abstract representations, the features in question cannot be reinterpreted as purely syntactic (as opposed to morphological); therefore, features are all that there is to know about lexical items, form the most abstract properties driving abstract movement to the properties of the spelt-out form. NoPaolo Acquaviva room is left for purely morphological information, unless in such a way that it is totally irrelevant for syntax. Hence, no room is left for a theory of lexicalization, that is an attempt to derive the locus of Spell-Out from morphological properties. The alternative I wish to propose requires that we drop the assumption that such features, qua morphosyntactic, have to do with the realization of vocabulary items. Understanding “morphological” as “reflecting properties of realized forms” is certainly not the only possible interpretation, and arguably a simplistic one. A host of researchers, in fact, have articulated the view that the clustering and manipulation of morphological features, on the one hand, and the rules and principles governing their realization, on the other hand, are not necessarily isomorphic, although obviously related (Beard 1987 1995; Halle and Marantz 1993; Noyer 1997); in a slogan, the third person singular of a non-past English verb and the element /z/ (with its alternants) are not the same thing. This general stance is generally known as the “Separation Hypothesis”: (2) The Separation Hypothesis: Rulles and generalizations about morphosyntactic feature bundles are separate from rules and generalizations about their realization. Syntactic terminals are feature bundles, not affixes or words. ‘The authors just cited defend different models of the role and organization of morphology in grammar, but they all share the view that the processes that construct morphosyntactic feature bundles are separate from the processes that map these feature bundles to phonology. The terminal nodes of syntactic structures are not affixes or lexical material, but just feature bundles with no phonologically relevant information. Processes of a different nature map them into words or affixes By drawing a clear line between the make-up of syntactic terminals (feature bundles) and their realization (or “lexicalization”),. the Separation Hypothesis can help us eliminate the stipulation that some features must be spelt out under a head just because the head is so marked (which is what the {+ strong] characterization comes down to). Instead of simply stating which terminal node must be lexicalized, I will propose that lexicalization of a feature bundle is enforced by a particular match between syntax and morphology. To this purpose, we need a second theoretical tool: the idea, clearly articulated and motivated by Aronoff (1994), that there is more to morphology than just its interfaces with syntax or phonology. I will refer to this concept as the “Autonomy of Morphology”: 1 Ih what follows I will use “realization”, “lexicalization” and “morphological interpretation” as synonyms. “Lexicalization” in this sense has nothing to do with “lexical” as opposed to “functional”Deriving [+ strong] features from morpholows head movement (3) Autonomy of Morphology: Morphologically relevant categories have a function within the morphological system independently of their “meaning”. According to Aronoff, morphologically significant units are autonomously definable linguistic categories that can, but need not, match syntactic or semantic units. The categories relevant to morphology do not coincide with those relevant to syntax: unlike Halle and Marantz 1993, Aronoff 1994 is not committed to a basically “item and arrangement” approach, but recognizes the role of notions such as stem forms or inflectional classes, which can only be construed as linguistic categories (as opposed to taxonomic constructs) within a morphological module. Obviously, morphological categories also include affixes, but crucially morphology is much more than the “syntax” of affixes. Viewed in this way, morphology is genuinely an independent module of grammar, with a conceptual vocabulary that cannot be reduced to that of syntax or phonology. “A match between syntax and morphology”, therefore, will be a match of a certain kind between, on the one hand, feature bundles that fill terminals in a completed syntactic structure, and on the other hand, a morphological component that does more than simply translate this information into instructions to phonology. A theory of lexicalization, I want to argue. 1s possible once we recognize that the relevant interface is not syntax-phonology, but rather syntax-morphology. But variation exists on both sides of this interface in the way morphosyntactic features are clustered (so, on the syntactic side) and m the way morphology organizes and realizes these clusters (so, on the morphological side). Since lexicalization is a function of the match, it will vary across languages as a function of both syntactic and morphological variation. 2. Reinterpreting weak features: extended projections 2.1. The status of abstract “movement” A theory of lexicalization requires in the first place that we reinterpret covert movement and the associate notion of {- strong] features. Abstract movement is supposed to have a number of peculiar properties. First, it is crosslinguistically more uniform than overt movement: a canonical illustration is the raising of verbs along the inflectional complex (which is not unanimously accepted: Hornstein 1995 points out that a phenomenon like antecedent-contained deletion presupposes that a copy of the verb, if not the verb itself, must be present in the VP at LF, to be copied into the ellipsis site; if verb movement out of VP were universal at LF, VP deletion should be equally unavailable or available to all languages, contrary to fact). Secondly, there is by now an ample literature on the apparent relaxation of some locality constraints on LF-movement and -pied piping (Nishigauchi 1990, Manzini 1992, Hornstein 1995, Brody 1995, among others) Finally, Chomsky 1995 seesPaolo Acquaviva abstract movement as the “pure” illustration of the process of feature transfer, which alone is directly triggered by the need for checking. Overt movement, with concomitant pied piping of additional features (including phonological ones), is taken to be a fundamentally accidental by-product of other properties, relative to phonetic interpretability. Note that this approach is consistent with, and indeed required by, the Minimalist assumption that features are the only objects of syntax, not categories, heads, words, or affixes. Before the introduction of Checking Theory and feature movement, the single tule “move any category” was the most natural way to unify all displacement ena, because the notion of category (head or XP) was equally well justified ible syntax or LF. Once the primitives of movement turn from categories to abstract features, however, the unity of covert and overt displacement must be re- justified: the latter involves more than the former, and this requires and explanation. In a Minimalist setting, the question is turned on its head: Spell-Out 1s more or less an irrelevant accident affecting some features, so the question is no longer 1f visible and invisible movement are the same thing, but why visible movement deviates from the simple feature transfer required by syntax. Displacement must be one and the same thing regardless of lexicalization, since it is uniformly triggered by a single reason, the need 10 check morphosyntactic features. Ina system like this, the only reason why abstract movement should be tiyzered by morphology is the desire to put all non-local relations into the same culcgory, that of “movement” (or “feature attraction”); and since overt movement is supposed to be triggered by morphology, so must its abstract counterpart, But if features are clearly distinguished from their realizations, contrary to Chomsky’s assumptions, perhaps we can find a reason for the displacement of abstract features that does not involve properties of their realization. 2.2, Extended projections and categorial identification A ready instrument is provided by the concept of extended projection, due to Grimshaw 1991. How much of the functional architecture of the sentence is universally given and how much depends on language-specific inflectional categories remains an open question, but for our purposes the following assumptions will be made: heads with idiosyncratic, lexical content are dominated by a set of functional heads, roughly partitioning the sentence into lexical - inflectional - illocutionary information, (4) a, Extended projection “An extended projection ... is a unit consisting of a lexical head and its projection plus all the functional projections erected over the lexical projection”. (Grimshaw 1997: 376)Deriving [+ strong} features from morphology: head movement b, [illocutionary [quantificational [inflectional [lexical]]] (cf. Rizzi 1995 Cinque 1997) To provide the partitioning, these projections must be more than simple syntactic templates. They must be inherently interpreted according to very general categories: for example, the grammatical and semantic information loosely referred to as “tense” perhaps requires a finer articulation, but it certainly must be distinguished from the purely grammatical specification of pronominal features typically referred to as “agreement”. Whether UG specifies a full structure sufficient for all natural languages, as proposed by Cinque 1997, or only an essential articulation, is immaterial for our proposal. : Clausal projections, that is extended projections of verbs, are the best studied ones. In this case, the hierarchy in (4b) is generally taken to be instantiated by something like (5), where “Infl” stands for inflectional categories relevant for the language and Comp can accommodate Topic, Focus and illocutionary force indicators. The question whether the practice of including agreement among the beads is justified must be eventually answered, but I will leave the matter unsettled (Cf. Iatridou 1990, Speas 1991, Spencer 1992, and, in a Minimalist setting, Chomsky 1995); the same applies to the positioning of negative markers in Inflection (agreement and negation notoriously display a far greater positional variation across languages than other projections; see Cinque 1997). (8) [Comp" {Comp* .... [Inft! fn’ (VN The same approach should extend to nouns and adjectives, although here there is no such broad consensus. Roughly, we expect a pattern where inflectional projections are systematically inside projections relevant for the interpretation: so, Number and perhaps Agreement, if this is a head, should consistently be inside heads that express quantificational or referential properties, whether these are labelled D, Q, or Degree (cf. also Koopman 1994, with references). (6) [Q[D [Num [N]}]) (Deg [Al] Functional heads are not just there to host affixes. Inflectional heads jointly define a choice of morphosyntactic features for lexical heads, regardless of their actual realization (this partly contrasts with the more recent view of Grimshaw 1997), Projections above the inflectional level contribute information of a different kind, inherently relational; distributivity (following Beghelli and Stowell 1997), force as in imperatives and interrogatives, and discourse-related information like topic and focus. Instead of thinking that a verbal feature raises for abstract morphological requirements of the inflectional heads, we can more simply say that a lexical head must categorially identify all members of its extended projection:Paolo Acquaviva (1) Categoria! identification: A lexical head must categorially identify all members of its extended projection. In a purely syntactic sense, as opposed to morphological, all functional heads identified by a verb are light verbs, and all those above a noun are light nouns, As Grimshaw argued, selection effects are best viewed as restrictions on the co- occurrence of functional categories; for example, we do not expect an assertion projection to dominate a noun, but this does not mean that a complementizer head. “selects” verbal inflection in any syntactic sense. What Chomsky views as abstract feature movement has now nothing to do with morphology, in any sense: it is a formal relation expressing the fact that the information of the functional complex (say, tense) refers to the lexical head (say, the verb): so, ina way all verbs “raise” to the highest functional head of their extended projection. The metaphor of raising is misleading, however: we might as well say that a lexical head categorially identifies its functional shell, and that this is only possible up to the next lexical head. In fact, claiming that cutcgurial identification is brought about by movement would merely replace one motivation for abstract hea to-head movement with another (cf. again Koopman 1994), leaving intact the claim that “movement” as such is independent of lexicalization, A stronger claim is preferable: the lexical characterization expressed by a head X (or rather, by a projecting feature bundle X) remains visible for all those projections above X which lack a lexical characterization. Metaphorically, “lexical” and “non-lexical” information conveyed by features are on different planes: the information provided by pronominal agreement and by a determiner does not obliterate the information provided by a head noun, but becomes part of it, in the same way that a head X with feature F “projects” F to the next bar level in Chomsky 1995: here what I have just called “bar level” is in fact the simple marking that X and its complement form a unit sharing with X the information encoded as F. A bar level, in this sense, is totally void of additional information; a functional projection contributes new information, but of a different type that must be added to that of the lexical head. This view accords naturally with the intuition that the objects of semantic interpretation are whole “chains”, not just heads. We do not need to state this separately, once it is clear that functional projections express the syntactically contributed information relative to a lexical head — or probably to a simple structure of lexical heads, if we accept with Hale and Keyser (1993) and a host of other researchers that the building blocks of texical semantics behave like heads in a syntactic representation, 10Deriving [+t ong) features from morphology head movement 2.3. Visibility of morphological features ‘Negation poses an apparent problem that leads to an important clarification. If the functional projections above a lexical head specify syntactically contributed information, and if negation is among these projections, then being negative is one of the properties of an inflected verb. But we do not want to say that every inflected verbal form occurs both in affirmative and negative version. The point generalizes to interrogative, or relative, or emphatic marking. In order to avoid massive and unjustified overgeneration of lexical forms, we must make explicit in what ways our analysis, based on categoria! identification of an extended projection, contrasts with Chomsky-style approaches based on abstract movement. The latter holds that morphosyntactic specifications of lexical heads are checked (as in Checking Theory), or alternatively acquired, through movement across the inflectional complex, be it covert or overt. The question whether negation is one of the morphosyntactic specifications of a verb arises in this framework, but not in a framework where the relation between a lexical head and its functional projections is not accomplished through movement (more precisely, not through the same mechanism of displacement operative in overt movement). In this alternative view, the grammatical-interpretive connection between verb and inflection (and complementizer), noun and determiner (and quantifier), lexical and functional categories, has a different nature from the connection between a moved element and its trace, I have informally proposed above that functional heads are, by definition, accessory on a lexical one.2 ‘The abstract features present under each node collectively specify an interpretation for the whole extended projection: if the sentence is negative, or has past tense, or is a question, the node V does not need to carry these specifications, as long as we consider just abstract morphosyntactic features. This is stated in (8): (8) Visibility: Features expressed on the heads of an extended projection are all visible for the interpretation of the extended projection as a whole Itis assumed that the coarse partitioning provided by UG constrains the position where abstract features appear: for instance, if there is a Tense node, tense features will appear there by definition, not under V. The realization of such structures, or morphological interpretation, depends on the formatives available to the language; in some languages, inflected verbs might indeed carry affixes (or other forms of exponence) for negation, past tense or interrogativity. Following common practice, explicitly codified in Halle and 2 Recall that even in Chomsky’s (1995) “bare” model the distinction between lexical and non-lexical heads does not follow from any principle. iPaolo Acquaviva Marantz 1993, I will assume that a feature bundle is lexicalized by that inflected form that realizes (or “discharges”, in the more elaborate model of Noyer 1997) most of ts features.3 But by (8) it is not necessary to think that V raises to Tense whenever its form encodes information about grammatical tense: what determines the position of the verb in the tree and what is the relation between abstract features and verbal affixes are two distinct theoretical questions. An answer to the first one, namely which node (or nodes) in the extended chain is lexicalized, will be outlined in the following section. Granting the feasibility of such a theory of lexicalization, let us focus on the second one: our model allows a realized form to match its features not against those present under a single node, but against those of the whole extended projection, In other words, “interpretation” in (8) means both “semantic interpretation” and “morphological interpretation”. In this second sense. (8) may be read as (9): 4 (9) Morphological visibility: ‘A word form realizing feature F can be inserted under a node X not specified for F, provided that F appears in the extended projection which includes X. (8) and (9) state that the information provided by a whole extended projection is lexicalized under one or more of its heads (depending on the availability of free forms or bound affixes); any one terminal can in principle be spelt out. The next section will propose a way to constrain this freedom. Note that neither (8) nor (9) make any claim about the ordering of affixes in a realized form; they giant freedom 3° Although the issue of realization will not play an important role in what follows, some assumptions must be made explicit. In general terms, a realized form is associated with morphological features which must match those filling the syntactic terminal it lexicalizes. I will use the expression “word form” when referring to a form where all affixation processes are completed (just what it means to be “completed” should be made clear, but the argument can stand on its feet even without an analysis of cliticization and compounding). Otherwise, the simple term “form” is used: it is more neutral than “affix” ‘or “stem”, and simply refers to those instructions for phonology that constitute the output ‘of morphological operations on a feature bundle. Such a vague term purposely does not specify whether “word forms” aré directly the output of realization rules whose input is a feature bundle, or whether they are constructed by assembling smaller “forms” (affixes) It will be clear from the following section that } commit myself to the inclusion of “affixes” among the morphological categories, but this is compatible both with a realizational and with a combinatorial view of morphological interpretation (or “Spell- Out’) 4 Another necessary assumption is that, following Halle and Marantz, 1993, not all features in the input (under a terminal for Halle and Marantz, in an extended projection here) must firid a realization: voeabulary items may be underspecified. 12Deriving [strong] features from morpholoxy head movement to lexicalize under any node(s) of an extended projection, not freedom to lexicalize in just any way. 2.4. Negation and Do-support in English 2.4.1. Negation in English provides a convenient illustration. Consider the following, examples (0) a = [DP sn! [Neg [VP E-past] [neg] [¥] Bsg} John doesn’t like anyone E-past] Wy (3sg} [neg] b. (DP. Infl' [Neg (Vv pPy) Fpast] (meg) [¥] [neg] Bg] John likes no one [Vy] [neg} Epasi] (3sg] For reasons detailed in Acquaviva 1994, 1995, 1997a, I take negation to be abstractly represented on a node inside the inflectional complex whenever the whole sentence is negated, no matter whether the visible realization of negatton appears as an inflectional marker (10a) or lower down, as a negative DP (or rather QP: see Acquaviva 1995 for discussion)S. In any case, there is no need for the verb to be ‘The quantifier no one is inserted under a node marked [negative] in (10b) because its head is not itself part of the extended projection of V the whole direct object no one is a distinct extended projection, which must be marked [negatwve] given the fact that these quantifiers have a negative interpretation even in isolation This is not incompatible with the claim that negation is interpreted on a negative] inflectional head: in Acquaviva 1995, 1997b I defend the view that, in sentences like (10b), the head of io one and the abstract inflectional head Neg” are interpreted at LF as a single complex syntactic object. This proposal, however, has no consequences for morphosyntax. 13Paolo Acquaviva [negative] just because the whole extended projection is so marked, By the same reasoning, if we take seriously the idea that the objects of interpretation are extended projections, then there is no reason to insert the form likes under a node containing agreement features. In order to have the correct verbal form match the correct interpretation, all that is needed is that the information [3 person sg, nonpast] should be present on the extended projection, Where and how this piece of abstract morphosyntactic information is visibly encoded is a related but distinct question, whose answer depends on language-particular mechanisms of morphological realization. In the same way, where and how the negative feature is spelt out depends in part on the morphology of the language, a component that must be learned anyway. In English, evidence like that in 1! will teach the child that not is a particle, and that it marks simple negation in sentences and in non-sentential negation (I will briefly consider the reduced form n’t later on): (1) a, Idid not talk b. not very well ©. not even ten years ago, you could see those movies d. not in my life This entails, among other things, that the particle not does not “select” tense or any other verbal inflection: there is nothing inherently verbal in it. The child will also learn that all other expressions of negation, adverbs like never or indefinites like nobody carry with them extra lexical information; only not is the simple marker of bare negation. 2.4.2. This last observation has some implications for the analysis of do-support. Descriptively, do-support in negation is one facet of a general phenomenon: whenever the extended projection of the verb contains information not expressible on the verb itself, like negation, emphasis, future tense, interrogativity, and various modalities, finite inflection is lexicalized, If there is no modality to be lexicalized by an auxiliary, do appears, A description in these terms is made possible by the clear distinction between bundles of abstract features and their lexicalization (largely dependent on languaye-specific morphological properties). It contrasts starkly with common conceptions that view syntax as a mechanism that assembles words; however, it echoes some aspects of the analysis proposed by Wilder and Cavar (1994), who argue that the complex of lexical and inflectional features in English is in fact always expressed periphrastically, by means of an auxiliary that happens to be a null form in the present. Instead of following their Minimalist analysis, however, I will reject the idea of a form lexicalized as zero, and propose instead that the feature bundle filling Inflection in cases like'(10b) is not lexicalized, for reasons that will become apparent in the next section (when I will attempt to outline a theory of lexicalization). Granted that the feature bundle under Inflection does not need 14Deriving [strong] features from morphology head movement lexicalization in English, the question now is why it is obligatorily spelt out by do in yes-no questions and negation (plus, respectively, movement to C and noi), despite the fact that neither do nor other auxiliaries are inherently interrogative or negative. [ will focus just on negation to illustrate the main point, which is that do-support is more tightly linked to morphology than to syntax, 2.4.3. As is well known, in sentential negation do is ungrammatical with any negative expression except nor (12a,c,e), and not is only grammatical with do (12a\b): (12) a. John does not speak bb, * John not speaks ¢. * John does never speak d. John never speaks ¢. * John does speak to nobody f John speaks to nobody Recall that, under widely held assumptions, all negative sentences like those in (12) are marked by the feature [negative] (see (10) above and references there cited). If do-support were only sensitive to purely syntactic configurations, and if all negative sentences have a {negative] feature on Infl, then do should be uniformly present or absent in (12). Instead of stipulating with Bobaljik (1995) that only not, but not the adverb never, blocks adjacency between Infl and the verb, thus enforcing do as a default lexicalization of Infl to express inflectional features, we can exploit the unique morphosyntactic character of nor: this particle lexicalizes the negative feature, and nothing else. This circumstance, coupled with the observed mutual implication of do and not, suggests that do and not are two discrete lexicalizations of just one syntactic object, which is made up entirely of grammatical features.6 Elements like never or nobody are different because they do not consist just of the grammatical feature [negative]. Abstractly, they too are linked with inflection, and that is why the whole sentence is negative; but they also carry with them a certain amount of unpredictable lexical information, and therefore cannot take part in the realization of a grammatical formative. The main word in this last sentence is realization: the negative charactervzation of an English sentence is realized, or given visible substance, either through a lexical head (as in never/nobody), ot through Inflection itself. In the latter case the ingredients of a negative auxiliary are assembled: negation and inflectional features. Since the vocabulary of English has no truly negative auxiliary (a single inflected modal verb that expresses negation inherently by its very stem), the two ingredients appear under distinct nodes, and 6 Iris uncontroversial that modals and other auxiliaries are like do in consisting entirely of grammatical information, as opposed to lexical (cf. Beard 1995: 128-137) 15Paolo Acquaviva each formative also has other uses. But, crucially, they can lexicalize a set of grammatical features (split between two nodes) because both are exponents of grammatical features and nothing else. This essential distinction between not and never / nobody in connection with do-support could not be drawn distegarding the way features are clustered (that is, disregarding an important component of morphology). 2.4.4. These are morphological facts about the vocabulary entties of English; a child Iearning Finnish wil be led by the evidence to posit a negative auxiliary. In other words, what is lexicalized as a cluster of grammatical features in Finnish is split between two distinct terminals in English. Other Germanic languages express the same information (simple negative sentences) with arguably the same structure as English, but without do-support. This ayain suggests that the obligatory insertion of do under an inflectional node appropriately marked is an accwent of the English morphology and vocabulary in tandem with syntax: if English did not have a semantically empty, purely functional auxiliary, then the “complex negative auxiliary” made up of do and nof could not arise; but the history of the language led to the cmergence of such a dummy element, and therefore finiteness and negation have their inflectional realization (do ... not) on a par with finiteness and various modalities. ‘Our discussion of neganon was started by the question whether verbs carry the information that they ovcur in a negative or affirmative clause; the answer has been that they do not, because being negative is a property of the whole extended projection. Having distinguished abstract featural characterization from morphological realization, we then argued that the distribution of do-support is largely determined by properties of the realization: English realizes the feature {negative] on Inflection either by morphologically negative lexical elements or by separately spelling out inflectional and negative features, lexicalized as do (or other auxiliaries) and nor respectively. Now, we can briefly consider how this approach clarifies the question whether negation can really be regarded as an inflectional affix of the verb. The Separation Hypothesis helps us to disentangle two aspects of the question: as an abstract feature, negation is indeed like an affix because 1t is part of the accessory information about a lexical item provided by its extended projection In this sense, “affix” must mean something like “piece of information added to the core lexical information” — a rather imprecise usage of this word. But the realizations of this information are not necessarily bound affixes, where “affixes” means “forms available to a language for making abstract information visible” — the common meaning of the word. So, in agglutinating languages like Turkish negation is in fact a verbal affix (13a); even in English, Zwicky and Pullum have pointed out that the reduced form n’s behaves like an affix because it induces allomorphy on the stem of the auxiliary. As shown in (13b), the stem [du] becomes {dou-], and [wil] becomes [wou-] 16Deriving [+ strony) features from morphology: head movement (13) a. gel-di-m gel-ne-di-n Turkish (Beard 1995: 58) come-past-Isg__come-neg-past-Isg “Teame* ‘Ididn’t come’ b.do[du] -» don’t [dount] will [wil] -> won't [wount] (Zwicky and Pullum 1983) But being a bound affix is a property of single forms, Negation may perfectly well take a non-affixal form, just like the category of [future tense] may be realized by an auxiliary like will or by a bound inflectional affix. English shows that the same language may have both an affixal variety, n't, and a free form nor. 3. Reinterpreting strong features: the role of autonomous morphology 3.1. Outline of the argument The morphological mechanisms by which feature bundles access phonology have already played a role in the foregoing discussion, but have not been focussed on. This section addresses the central question of why access to the component of morphology feeding phonology is restricted to certain syntactic terminals, but not others, The outline of the argument is the following: syntactic terminals are interpreted by feature bundles, not words or affixes; these abstract representations are manipulated by morphology, which eventually provides instructions for the phonological spell-out component; if the features under a syntactic head happen to identify a discrete morphological category, and this in turn identifies a range of phonological realizations, then the syntactic head is lexicalized. The key of this argument is that morphology recognizes its own units and significant categories: it is when morphological and syntactic unit coincide that lexicalization is enforced. (14) Principle of lexicalization If the feature bundle F under the head X° is matched by a unit of autonomous morphology, then F is lexicalized under X°, (15) A unit of autonomous morphology is any category (inflectional class, stem form, affix, ...) by reference to which morphological generalizations can be made. (Cf. Aronoff’s (1994) notion of “morphome”) Instead of specifying what degree of morphological “richness” makes movement mandatory, (14) states under what conditions a node must be lexicalized. As we will see later, this formulation has empirical consequences in cases where (14) enforces lexicalization of more than one head in an extended chain. Another consequence of this formulation is that multiple exponence in the same “chain” is not ruled out in principle, not just for distinct features (as in auxiliary-verb constructions), but also 7Paolo Acquaviva sn cases of multiple exponence of the same features. However, I will not explore this line of analysis here, limiting myself to note that the representational approach underlying (14) is better equipped than pure feature-transfer to handle multiple exponence (cf. the discussion by Holmberg (1997) in a Minimalist setting). 3.2. Morphology-driven lexicalization Aronoff (1994) exemplifies the autonomy of morphology by considering the English formative /z/, which marks plural i nouns and third person singular agiecment in non-past finite verbs. As illustrated in 16, this affix defines a range of alternants in a phonologically predictable way, but the rules are stated over the affix and not over the segment // (16) agr3sgofV ——pluralofN auxiliary genitive kick{s} kick{s] Dick{s] gone! Dick{s] arrival taglz] taglz] Meg{z] gone! Meg{7] arrival rowlz] rowlz] Joe{z] gone! Joe{z} arrival missliz] miss[iz] Ross{iz] gone! _—_-Rossfiz} arrival Aronoff points out that the same attix 15 also the exponent of reduced auxiliaries is, has and, in colloquial registers, does, beside marking zenitive as a phrasal cliti. The morphophonological alternations it defines are almost the same, with minor differences; so the affix must be recognized as a unit in the morphology of English, although it is put to use to express different and unrelated “meaning”, or morphosyntactic features. In other words, different teature bundles are mapped to a single unit of morphology. What is phonologially interpreted is the unit of morphology, not directly a feature bundle.7 Now recall that English main verbs are syntactically lower than Italian or French, ones (Pollock 1989), Then we have a correlation: on the one hand lack of overt movement by the verb, on the other hand a morphological marker (/2/) that does not uniguely mark the function [3sg nonpast]. Consider now another example of the same correlation, namely that lexicalization in a high position goes hand in hand with particular morphological status of the relevant marker, and conversely. French and Italian infinitives of main verbs are functionally near-equivalents; | will assume that ignoring spell-out they have the same feature bundles. However, thanks to the 7 Still within the English verbal system, past is either marked by unpredictable stem forms, or by a dental affix which is for all intents and purposes undistinguishable from the dental affix of past participle; as Aronoff (1994) again notes, the notion of past participle itself is ptimarily a morphological category, which expresses quite unrelated meanings such as perfect and passive, 18Deriving [+ strong] features from morphology: head movement work of Pollock (1989) and Belletti (1990, 1994) we know that Italian infinitives “raise higher” than French ones, as 17-18 exemplify: (17) a. Pourne pas manger. French b. * Pourne manger pas. “For not to eat’, ¢. [Infl PRO ne [Negp pas [In [VP manger }]]} (18) a. * Per non pid mangiare. Italian b. Per non mangiare pit. “For not to eat again’. ©. [Inf PRO non mangiare (NegP pit {Infl (Pt I]]] (Pollock 1989: 412) Here again we can see a morphological difference. The category [infinitive] ‘must certainly be encoded for both languages in the morphosyntactic representation constructed by syntax. It is irrelevant just what features define this morphosyntactic category. But the realization is different: the endings of French infinitives, listed in (19), are all vaguely similar for historical reasons, but there is no specific infinitival affix recognizable as such by morphonological idiosyncrast (19) chanter, vouloir, prendre, conclure, mourir “to sing, to want, to take, to conclude, to die” Chanter ‘to sing’ has the same form in the infinitive, past participle and present indicative 2nd person plural; prendre ‘to take’ or vivre to live’ have the same endings as the adjectives rendre ‘tender’ and ivre ‘drunk’, and their final /t/ cannot in any case be considered an infinitive affix because it docs not induce any morphologically conditioned effects. Italian, by contrast, has a single marker consisting of the thematic vowel plus Ire/, plus a few nearly identical cases like condurre (cf. Vogel 1993): (20) cantare, volere, prendere, concludere, morire, condurre “to sing, to want, to take, to conclude, to die, to lead” This ending is recognizable as an affix by its sensitivity to morphonologi.al rules, like English /2/ in this respect. If it combines with a clitic, the final vowel must be deleted: ‘to love you’ is [amarti), not (amare ti]. (21) amare + ti amar.ti *amareti) “to love + you’ However, the omophonous amare, feminine plural form of the adjective amaro ‘bitter’ cannot undergo the same rule; here the proncunciation must be [amare ti]: 19Paolo Acquaviva (22) a, le mandorle amare ti piacciono (amare ti.piacciono) b. * te mandorle amarti piacciono _(tamar.ti piacciono) ‘you like bitter almonds’ Syntactically, this means that a clitic attaches to a verb but not to an adjective, a familiar generalization, But this syntactic information must be available at a level where phonological material is being combined: in other words, there is such a thing as an infinitive (and therefore verbal) affix. Like the English /z/, this affix is more than a phonological spell-out; it must be recognized as a unit of Italian morphology {a “unit of the form”); unlike the English /z/, however, here the unit of the form matches a unit of the meaning, that is, the morphosyntactic category [infinitive] is realized by a purely morphological category [infinitive]. It is no accident that Italian infinitives are lexicatized under « higher node than French ones, for the same reason why English verbs do not raise despite being marked by /z/. This regularity points to a principle of lexicalization like (14), here repeated, with (15) as ancillary definition (4) Principle of lexicalization If the feature bundle F under the head X° is matched by a unit of autonomous morphology. then F is lexicalized under X°. (15) A unit of autonomous morphology is any category (inflectional class, stem form, affix, ...) by reference to which morphological generalizations can be made, (Cf. Aronoff’s (1994) notion of “morphome”) 3.3. The syntax-morphology interface In this simple form, (14) attempts to derive the locus of spell-out among the heads of an extended projection. In accordance with Distributed Morphology and with the syntactic model proposed by Brody (1995), it presupposes that complete syntactic representations are interpreted by an autonomous morphological component. The units and categories of this component must be learned anyway: the morphonological properties of English affixal /2/ or Italian infinitival ending /Vre/ must be learned, just like the fact that English expresses future tense by means of an auxiliary while Romance encodes future within verbal inflection. What motivates a verb to appear under Tense is not just the fact that its Tense morphology is “rich” or realized as a discrete affix; rather, a tensed verb appears under Tense if the feature [tense] in the extended projection of V is a sufficient input to define a verbal affix, or possibly a set of realization rules; in other words, a verb raises to Tense in a language if Tense corresponds in this language to one side of a linguistic sign: a mapping of sound and meaning. And, by definition, this must be interpreted 20Deriving [strong] features from morphology: head movement phonologically. English /z/ is not a sign, because it is compatible with a range of unrelated meanings; it is only a unit of the form. The feature [infinitive] in French (or the corresponding feature bundle) is not the “meaning” part of a sign; in that language infinitive verbs are signs, but not infinitive affixes, If there was no autonomous morphological component, then all syntactic terminals would be directly mapped to phonology. Knowing the grammar and the vocabulary of a language entails knowing, among other things, what features are clustered under syntactic terminals and how and to what extent available formatives match them. Knowledge of morphology, in particular, entails knowing how the information scattered along extended projections is reorganized, expanded and reduced by available forms. When, despite this reorganization, the map from a feature bundle to a “unit of the form” is still direct, then also the map with phonology will be direct, In effect, a syntactic terminal will be spelt out whenever this is possible: that is, whenever its features can be read directly as instructions to phonology. It is worth emphasizing that a proposal along these lines is incompatible with the view of the syntax-morphology interface sketched out in Chomsky 1995. Whether a terminal will be spelt out depends here on what features are clustered under it and on the way the morphological system is organized in that language. This presupposes that the sound-meaning mapping is not a by-product of syntactic operations (encoded in the form of [+ strong] features), but is performed by an autonomous morphological module sufficiently structured to recognize its own significant categories. A global consideration of the morphological system as a whole will tell the speaker whether a given exponent (an affix or a mule of exponence) is a sign, something that by itself establishes a match between the “sound” corresponding to a form (or set of forms) and the “meaning” corresponding toa feature bundle under a terminal. Lexemes, of course, tolerate a certain degree of synonymity, but within grammatically encoded information a formative compatible with more than one feature bundles (unrelated) cannot be regarded as a sign. In such acase, the whole inflected word will be a sign, but not the formative. it follows that even a clearly identifiable affix unambiguously signalling a grammatical feature F will not necessarily license lexicalization under the inflectional node hosting F abstractly: what is crucial is the status of the affix within the whole morphological system, Such a global consideration of the whole morphological system is independently necessary for any analysis attempting to give the (+ strong] distinction a morphological motivation, Consider the variety of the exponence of [past] in English verbs: (23) a. watch, watched = -/v b. drink, drank ablaut (root revowelling) ©. put, put ablaut (no change) 21Paolo Acquaviva Any conclusion drawn from the exponence of {past] in English must take into account that inflected verbs “raise” or stay in sifu as a whole class, not depending on the morphological make-up of single forms: otherwise, if syntactic properties were directly triggered by the simple presence of discrete affixes, we would expect a difference at least between put and the unambiguously past drank (neither of which can legitimately be considered a lexically marked exception).8 Thus, consideration for morphological properties automatically leads to considering the system as a whole, even for approaches that do not otherwise envisage changes in Chomsky’s (1995) framework. But in that framework any connection between syntax and morphology proper is problematic, beyond the expression of morphosyntactic features, Such features do not determine any aspect of their exponence; so much so, that the very same features characterize in the very same way purely abstract forms. As mentioned above, Chomsky explicitly claims that language-specific morphological characterizations are irrelevant for the computational system simply called “grammar”. It is hard 10 avoid the consequence that this must include information such as the presence of unambiguous affixes, especially if this entails {as we have seen) a global consideration for the whole system of exponence in a language Therefore, I believe, the lack of any morphological motivation for the [+ strong) distinction in Chomsky’s analysis is no accidental gap: a morphological motivation cannot exist, because “grammar” is blind to (or better, does not speak the same language as) the morphological system. 3.4. Syntactic head raising What the principle of texicalization claims is that under certain circumstances an inflectional head must be spelt out, not that it attracts the verb (or any lexical head). In this respect it differs significantly from any attempt to motivate “raising” of a hhead. In the consistently representational perspective here assumed heads do not 8 ‘The formal category [past] in English and German characterize verbal forms which are not notionally past, and are traditionally included in the subjunctive mood: (i) If John drank a bit less, he would be almost charming. Gi) Selbst wenn ich ihm ein neues Auto kauflc, wiire er ri “Even if | bought him a new car, he wouldn't be happy" But this is not the reason why the “past” suffix is not a sign. Other langnuges, Swedish or Dutch, only express subjunctive analytically; yet, the presence of strong verbs as a well-established class in verbal morphology entails that even in these cases the past suffix is not the sign for past within verbal inflection, although it is within single forms. tzuftieden. 9 ‘This is not the only reason why Chomsky's “morphosyntactic” features cannot possibly exhaust morphology. As Halle and Marantz (1993: 115, 135) argue, the morphological component must be able to add or copy certain features according to language-specific requirements. 22Deriving [strong] features from morphology: head movement “raise” within their extended projection: rather, independently available pieces of knowledge about syntax and morphology, including by hypothesis (14), conspire to enforce lexicalization under one on more nodes, Knowledge of morphology also involves knowing purely phonological information, such as whether a grammatical formative is a prefix, suffix, clitic or a free form; is not a new observation that what is an auxiliary in a language may be an affix in another. Because of this independently available knowledge, a speaker who already knows that (14) forces the node Tense to be lexicalized, and who also knows that the exponent for tense is a verbal suffix, will spell out Tense by combining the verbal stem to the suffix. But by the Separation Hypothesis, this information is not available at the point where a syntactic representation is completed: all that is represented are feature bundles, This leads to an important refinement. Consider the exponence of tense and subject agreement in an inflected verb, as in 24: the abstract information about these two categories is represented on nodes we may call Tense and Agreement: (24) a. Vo Tos Agr Icelandic késtu du m ‘we threw? (past, | pl) b. Vo Tas Agr Ttalian getta va mo “we threw’ (imperfective past, | pl) c. [AgrP Agr’ [rp T° [vp V° JI] Speakers of Icelandic or Italian must know that both kinds of information are expressed by the morphology of finite verbs. Assume that tense and agreement affixes are morphological units that match the features under their respective syntactic heads. Then we have a problem: on the one hand, (14) demands Iexicalization of both nodes; on the other hand, this cannot happen because there is no encoding in morphology for just Verb-plus-tense or Verb-plus-agreement. The resulting structure, although not syntactically ill-formed, would be filtered out by morphology at the point where abstract feature bundles match the vocabulary resources available to the language (this obviously leaves open the possibility that in different languages the very same construction may be well-formed). ‘The answer I would like to suggest is that these are the cases that motivate traditional head movement, in this case of (the feature bundle under) Tense onto (the feature bundle under) Agreement: 23Paolo Acquaviva (25) AgrP Agr oN Agr® TP S\N Agr? aon ae t 7 \p Purely syntactic displacement of features must be retained, along with the severe restrictions that make this an explanatory tool, but its nature is strictly syntactic and not morphological. “Movement” should not be invoked just to account for the position of a visible formative, but for more complex reasons that can only be gleaned by making reference to the positioning of abstract feature bundles. In this case the placement of the node Tense under the node Agreement ensures that both are Jexicalized, in compliance to the universal principle of lexicalization, while at the same time the lexicalization of both results in spelling out a single word form: the two affixes plus a verbal base, whose necessity only emerges at the morphological level where Tense and Agreement are introduced as bound forms. The net result is that principle (14) in fact ensures that a terminal that can be lexicalized under node X will be spelt out under X or higher, but not lower. We thus derive on principled grounds an important insight due to Brody (1995), whose model underlies many aspects of the present proposal: (26) Transparency (Brody 1995): The contentive category in the chain must be in the highest position licensed by morphology. Thus reinterpreted, movement is still triggered by morphological reasons, but it only affects the abstract feature bundles syntax is assumed to operate on, not affixes. Its nature is “abstract”, but it- brings about visible displacement of realized categories. It may be possible to dissociate it from the traditional idea of syntactic movement by imposing different constraints, but I see no reasons for doing so, and many for not doing so. In fact, the assumption that the constraints on it are those familiar from much syntactic literature (c-command, locality, and hierarchical nesting of the nodes involved) allows us to retain the wealth of empirical results that followed the pioneering work of Baker 1988a: syntax is not the same thing as morphology (see in particular Baker 1998b and Noyer 1997), but constraints on syntactic operations have important morphological consequences. There is another 24Deriving [sa ong] features from morphology: head movement desirable consequence in accepting that a feature bundle may move onto another feature bundle: this presupposes a full-fledged syntactic structure, which is necessary to account for the distribution of adverbs and other elements assumed to fill specifier positions. This marks a clear difference between the approach I am defending here, which is still essentially syntactic, and attempts to explain inflection only in lexical and word-structural terms (cf. for instance Sells 1995). ‘The principle of lexicalization (14) therefore should not be construed as a claim that syntax does not require a theory of movement, or that visible movement results entirely from the principles governing morphological spell-out (as advocated by Beard (1995: 356-375). But the movement allowed, far from being a residue forced by recalcitrant data, is organically integrated in the general approach, because it is limited to abstract movement of the features of one head onto another, triggered by (and obviously compatible with) the tension between the principle of le\tcalization and language-specific morphology. It is worthwhile stressing why this solution does not weaken the analysis, and does not make it indistinguishable from more standard ones. The main difference is that the morphological requirements of affixes here do have some role to play in head movement, but movement as a whole, not even just overt movement, is not based on that. The principle enforcing lexicalization is independent of affixes, and in this way avoids well-known paradoxes like having to claim affixal or clitic-like properties for null heads. A second major difference is that the approach here proposed makes a sharp and principled distinction between affixes that enforce lexicalization, and affixes like English /z/which do not. Only the former are units of the form biuniquely mapped to units of the meaning. As far as I can see, no comparable result is achieved in the analyses of Beard (1995) and Brody (1995) 4. Beyond inflection: movement to C 4.1.4 different type of movement The relation between verb and complementizer (short for “highest heads in the verbal extended projection”) is not of the same type as the relation between a lexical base and its inflectional accessories. In languages like Germanic and Romance, the grammatical information expressed by the verb and its auxiliaries is independent from the interrogative, relative of hypothetical characterization of the whole clause; the former information is grouped in the inflectional complex, the latter in the complementizer region (negation being a conspicuous exception). This distinction is a conceptual one {as opposed to “formally encoded in the grammar”), between information that makes up a predicate and illocutionary information about the predicate thus formed; to what extent either type is encoded by affixes or free forms is a separate and secondary question, The primary one, which will be addressed 25Paolo Acquaviva here, is how to account for instances of visible “movement” of V or auxiliaries to C within the framework we ate developing. ‘A comparison with “movement” to inflectional positions will highlight the problematic nature of V- (or Aux-) to-C movement, illustrated in (27)-(29): (27) a. Do [1p you t speak English] ? b. Sprechen {1p Sie t Englisch t]? (28) a. Had (1p you t told me before], ... b, Hattest ([p Du es mir t frither gesagt t J, (29) Ich weiB [yp tt nichts t } ‘Iknow nothing’. First, neither English nor German expresses through verbal affixes any of the features plausibly in C, so that the most straightforward way to replicate the effects of head raising is precluded (see (14) above and the ensuing discussion). Second, if raising to C is certainly not triggered by affixes, it appears to be construction- dependent: it signals interrogative or conditional value of the clause, but in other clause types the very same forms remain within IP. This constrasts starkly with movement of V to an inflectional position; it is as if a particular choice of tense or mood (or negation) was expressed by raising V past some adverb, with no further marking on V’ or elsewhere. Thitd, raising to C is not coextensive with the relevant characterization of C: if the latter position is filled, V or Aux stays in situ and no do- support is triggered: (30) a. I wonder if [1p you speak English]. b. Ich frage mich, ob [1p Sie Englisch sprechen} (31) a. If[yp you had told me before], .. b, Wenn [1p Du es mir friher gesagt hattest], .. (32) Ich wei, da [yp ich t nichts wei] “I know that I know nothing’. These well-known facts suffice 10 discard any variant of the idea that V / Aux raises to C in order to somehow achieve morphological well-formedness, the way V does to Inflection in some languages. Conversely, C cannot be said to attract V / ‘Aux the same way that Infl attracts V: an optionality like that displayed by the synonymous (28) and (31) has no parallel in the inflectional domain (it is as if future tense was altematively expressed either by a free morpheme or by repositioning the 10 See Reuland 1990: 143-146 for a critical view on the purported role of inflection in C in verb-second languages. 26Deriving [+ strong] features from morphology: head movement unchanged verb).! Notice in passing that for the same reasons it must be wrong that a functional head FUNCT hosts a lexical one LEX (i.e. LEX raises to FUNCT) unless FUNCT is already filled by a free morpheme.!2 Such a “generalized affixation” approach cannot explain why V / Aun does not appear in C whenever the latter is not filled by a free morpheme, like in Cnglistt declarative clauses; nor can it explain why (28) are at all possible, given that the vocabulary of the language has free forms available to spell out © that should constantly block (in the morphological sense) a raised V / Aun (again, no such optionality is attested in the inflectional domain). The nub of the problem is that movement from the inflectional region to C cannot be made to follow from morphosyntactic requirements on the raising element, no matter how the analysis is phrased. The question why this holds for movement to C and not to Infl requires a principled answer; | will tum to this issue later. The other possible approach consists in making use of properties of the target element, namely C. Wilder and Cavar (1994) pursue this line, anticipating Chomsky’s (1995) reformulation of movement as “feature attraction”, where attractors must be functional categories. From this perspective the differences between an approach in terms of feature checking and one in terms of a theory of lexicalization stand out more clearly. The former must primarily identify a feature shared by the choices of C triggering raising and by the raised clement; the latter must demonstrate that the raised element biuniquely matches the feature(s) under C —- apparently a hopeless task, given that the raised element is a verb or an auxiliary. Checking Theory seems beticr equipped to analyse the facts, but this is simply because the unconstrained assumption of features, their position and especially their [4 strong] value is compatible with many, if not all, states of atiairs, In fact, a 11 Wilder and Cavar (1994) highlight the peculiarities of construction-dependent movement in an early Minimalist setting (earlier than chapter 41 n Chomsky 1995), arguing that @ strong feature on a head X may be checked by a category ¥ even if that feature is weak on Y. Because Y only raises overtly to check X’s strong feature, it may fail to raise if X°s feature is already checked: thus, Y's raising depends on the construction. They propose that [+operator] C in English requires checking of a strong [finiteness] feature, which is performed by raising an auxiliary (which, they argue, is always present in English even if null in simple present and past). In verb- second languages [+finiteness] is strong even on [-operator} C. In addition, embedded [+operator] C always has @ weak value for [+finiteness). My analysis makes ample use of Wilder and Cavar's views, but attempts to dispense with the stipulations necessary in a Minimalist analysis 12 Cf. Beard (1995: 153): “... a more integrated theory of Aux may be developed, one in Which Infl categories are raised until the raising process reaches a node which the MS- component invests with a free morpheme.” Beard’s proposal concems the inflectional domain only, however, while in the text I am eriticizing the general application of this, simplistic view 27Paolo Acquaviva slightly more articulated theory of vocabulary insertion is all that is needed to derive Infl-to-C from the principle of lexicalization To illustrate this, | will now sketch an analysis of English auxiliary inversion in intenogative and conditional main clauses. Additions and qualifications will no doubt be necessary in order to extend it to subordinate clauses and to verb-second phenomena; but the purpose of the following illustration is to show that the scope of a morphosyntactic explanation goes well beyond the realm of inflection, and to that purpose it will suffice. 4.2. Input features Some premisses are needed to begin with. First, those choices of C that trigger movement are marked by a contentful feature: call it (Q] for questions and [IF] for conditionals. Second, the feature bundles filling the inflectional node(s) are not marked for [Q] and [IF]. These two assumptions are not stipulative: the first is the mere extension to C of what we are taking for granted throughout (for instance, that tense features are under the Tense node), and the second is the null hypothesis, in absence of any reason for thinking that English has interrogative or conditional auxiliaries.13 Now the tricky fact to be accounted for is that auxiliaries are exponents for [Q] and [IF] (if and only if they move). Let us restate this descriptive observation and assume, as a third premiss, that the various auxiliaries realize not only their respective bundle of grammatical features (call them [INFLCAN]. (INFLMust). [INFLHAVE], etc.), but also [INFL] + (QJ and [INFL] + [IF], the latter excluding do. This may seem a costly move, and requires some justification. | am not claiming that auxiliaries optionally “mean” Q or IF; that would hardly be an explanation, and in addition would incorrectly predict that intetrogative or conditional clauses could come about through the simple presence of auxiliaries, regardless of raising to C. The assumption that inflectional nodes (in English) do not host [Q] or [IF] avoids just such circularity ("Aux can express Q because it can mean Q”). I am claiming instead that the language learner has cnough positive evidence to take auxiliaries as output forms for both inputs [INFL] and [INFL] + [Q] /{IF}, just like put may or may not realize {+past], or like in general a form may realize a feature bundle despite being underspecified for some features, so long as 13 Both assumptions retain their plausibility within Checking Theory, especially the first one: their naturalness is not theory-internal. The second one, that interrogative or conditional features are not under Inflection, can be supported by the following argument (so it is not really an assumption): in informal registers questions, including. yes-no questions, may fai to employ an inverted auxiliary: (i) % You did it yourself? But no register admits something like (ii), which would be expected if the interrogative characterization for the auxiliary was in principle independent from raising to Gi) *You did do it yourself? 28Deriving [strong] features from morphology: head movement no conflict arises (cf. Halle and Marantz 1993), We have already considered another such case: namely do, which was argued to be (in descriptive terms) the spell-out of Infl when the latter contains extra information not expressed by the verb. In that case do (or any modal verb) and not / n’t jointly express the information corresponding to a negative auxiliary. Likewise, we are assuming, do is employed to spell out the interrogative feature (Q], although it does not “mean” Q; and other modals are employed to spell out [IF] even if they do not “mean” IF To sum up to this point: (Q] and [IF] are always under C and never under Infl, but auxiliaries may lexicalize {INFL] + [Q] and (just formally past auxiliaries like had, should and were) {INFL] + [IF].14 In other words, [INFL] + [Q] / [IF] is a possible input to a realization rule, but the input feature cluster is scattered between Cand Inflection instead of being under one node: (33) a. agreement) [tense] > do/ does / did (Q) b, {agreement} {tense} > have /has / had, must, can, .. [modality] (Q) (uF) If the argument stopped here, “raising” to C could not be derived. By (8) and especially (9) above an auxiliary may be inserted under Inflection and realize both UINFL] and {Q) (JF), although the latter is under C: (9) A word form realizing feature F can be inserted under a node X not specified for F, provided that F appears in the extended projection which includes X. Because auxiliaries do not spell out [Q] / [IF] at all, least of all by units of autonomous morphology, they would not be forced by (14) to be lexicalized any higher than Inflection. And, by (9), an auxiliary in Inflection can lexicalize [Q] / [IF] on C. This is where a readjustment is needed, 4.3. Lexical and functional heads Conditions (8) and (9) were motivated by cases like English verbal inflection: a form like goes contains a formative marking [3 sg], but there 1s no other reason why 14 | will not address the question whether “subjunctive” is a genuine morphological category of the English verbal system. 29Paolo Acquaviva that feature should in fact be under V rather than Inflection (Agreement) in abstract syntactic structure. Besides, the reinterpretation of weak feature cheching as categorial identification forces us to regard the whole extended projection as a wnt, where the information on one node is interpretively accessible from another node. The whole analysis is simplified if “accessible” means “morphologically accessible”, as in (9). But in so doing we are considering the relation of a lexical head to the set of functional heads that complement it. For semantic and morphological interpretation, a lexical category “is” the lexical head plus its functional accessories. Should we extend this to the relation between any two heads in an extended chain, including that between two functional heads? Raising of auxiliaries to C, and in general construction-dependent movement, is a good 1eason not to do it. The particular trait of such head-to-head relations ("the nub of the problem” alluded to above) is that both heads are functional. This marks a clear difference from the relation of a lexical head to its morphological complements, whether they are free or bound affixes. This difference is stated in (34):15 (34) a. The morphological realization of a lexical head LEX can refer to features which are not under LEX (provided they are in its extended projection); b. The morphological realization of a functional head FUNCT cannot refer to features which are not under FUNCT, (33) and (34) now derive the correct pattern. A feature [Q] under C cannot be directly lexicalized by a free or bound affix (since English has no question particles), so it would remain without exponence without a modal or the default auxiliary do. These forms have [Q] among their input features, but as a language- specific fact about English moxphosyntax [Q] and inflectional features are not clustered together under the node that hosts auxiliaries, or Inflection. However, the feature bundle under Inflection can be transferred to C, as another instance of the purely syntactic head movement we argued to raise Tense onto Agreement (recall throughout that the label “Inflection” is a convenient way to name all inflectional features grouped under one single terminal, whether or not this is the result of previous head movement). By moving to C, the raised inflectional features join [Q] under the same terminal. The feature bundle under C is realized by the corresponding inflected auxiliary, including do if no modality is featurally encoded in the input. The realization schema (33) implies that (QJ can be considered realized if the auxiliary is spelt out, even though no affix in the auxiliary corresponds to that feature, As noted, [Qj could not be realized unless the bundle of inflectional features (“Inflection”) raised to C: because of (34), a word form realizing a functional head is only sensitive to the features under that head, in contrast to what happens with 15 Recall that the formal distinction between lexical and functional categories is a necessary condition for reinterpreting abstract feature checking as eategorieal identification. 30Deriving [+ strong] features from morphology: head movement lexical heads. The same applies to [IF], except that do is not available as default lexicalizer. In addition, the free form if is available in the vocabulary. But the conditions for inserting if are not more restrictive than those for inserting an auxiliary: in both cascs, the only triggering feature is [IF], so that optionality is correctly predicted. If however the auxiliary is chosen, there will be other ‘morphological constraints (formally past form, unavailability of do); but these do not qualify as more specific conditions for insertion, because C will uniformly be just [IF]. 4.4, Recapitulation To summarize, some core aspects of auxiliary movement to C can be derived without reference to checking of [+ strong] features, by paying attention to the features that the “raising” auxiliary realize. The guiding idea has been that inflected forms may be the realization for a set of features scattered along several syntactic heads. The relation between a lexical head and its functional satellites can be seen as “inflection”, and it can be assumed that all morphological information is simultaneously available. This does not seem justified, however, for the relation between two functional heads in the same extended projection. Following this intuition, I have proposed that a word form lexicalizing a functional head cannot realize features represented under different heads. Construction-dependent movement may then be viewed as movement of one abstract feature bundle onto a higher one, which is then lexicalized by a form matching the resulting complex of features. In the case of movement of Inflection to C, the auxiliaries that lexicalize Inflection are compatible with certain features of C (features which are not represented under Inflection). Raising of Inflection then allows C to be lexicalized just in case it is marked by such features, despite the fact that auxiliaries are not, strictly speaking, lexicalizers for C. This analysis implies that inflected forms prescnt in the vocabulary of a language need not correspond to syntactic atoms, that is, to those feature bundles that are manipulated by syntax. It follows that the clustering of features under one head, before syntactic movement, is not given, but must be deduced by the learner on the basis of existing word forms and of syntactic word order. Such a strong interpretation of the Separation Hypothesis suggests that purely morphological reanalysis or oscillation in use may have dramatic repercussions on the grammar deduced by speakers of a younger generation, with interesting prospects for historical analysis, However, I will not be pursue this suggestion any further here. 5. Conclusion: phrasal movement Having found a way to incorporate head-movement and some of its empirical results, the obvious development would be to extend the approach to movement of 31Paolo Acquaviva phrases into specifiers, like Wh-movement or movement to subject position. This topic would deserve another paper for itself, but a few considerations can be added here by way of conclusion. First and foremost, it is not at all clear that phrasal movement into specifier is a homogeneous class, as it should be if its trigger were always just the need to check a strong feature. The traditional view of A-movement has been recently called into question by Manzini and Roussou 1997, who claim that it is better understood as a relation between a category and a thematic / aspectual feature, with no movement involved. Independently of this argument, A-movement of DPs into subject position is a little strange among other cases of movement to specifier: it is so common crosslinguistically, and so semantically unmotivated, that the need to have a subject has traditionally been regarded as an independent principle, the Extended Projection Principle. It is certainly possible to reformulate it as the need to check a strong D feature on Inflection, as in Chomsky 1995, but this leads us to expect that VSO languages should not in principle be any rarer than languages where the subject raises overtly, just as languages do not show a clear preference for avert or covert Wh-movement. But VSO languages are comparatively rare. Movement into the specifier of operator heads is relatively more homogeneous and better understood, Necessary overt specifier-head configurations for operators (the Wh / Neg / Focus / Affect Criterion; see Brody 1995 for references) lend themselves to be reinterpreted as morphosyntactic requirements on the head of the raised XP (cf. Simpson 1995). Notions such as operator or topic are defined over a syntactic structure, and are represented by appropriate features in functional projections. Again, the clustering of features and the availability of affixes or free morphemes from the vocabulary should determine whether the projection is made visible by spell-out of its head, its specifier, or neither (more rarely, both). Even so, the various conditions on the realization of operator phrases converge, but do not overlap completely: to simply assume that movement of a Wh-phrase checks an interrogative feature on C has nothing to say on why in English exactly one Wh- phrase must raise, no more and no less, but in Bulgarian all of them. An analysis in terms of feature checking is ideally suited to explain a yes-no distinction, between raising and non-raising, but the facts are more complex than that (see again Simpson 1995). Adopting multiple specifiers as suggested in Chomsky 1995 gives us the way to represent the data in a Minimalist framework, but does not in itself offer a principled explanation. Problems are even more acute when we turn to cross- and intra-linguistic asymmetries in the raising of negative categories, as detailed in Acquaviva 1995, 1997a. In general, one thing that is clear is that morphology is involved: raised phrases may or may not be unambiguously marked for the morphosyntactic feature they are supposed to check, like Wh-phrases, and languages differ as to the availability of question, negation or focus particles. But the question whether phrasal and head 32Deriving [+strong] features from morphology: head movement movement are both entirely reducible to the same mechanism of feature checking is too broad to be addressed here. In so far as the two are driven by the same principles, the facts and reflections I have presented here suggest that for phrasal movement as well as for head movement the locus of spell-out is predictable if we consider both syntax and autonomous morphology, not just one component or the other. The proposal is that a simple distinction of strong and weak features is at best insufficient, and at worst misguided. If the critical view | have articulated is on the right track, I see three advantages accruing to a revision of the Minimalist along the indicated lines: one, syntax is more uniform and less stipulative across languages; two, the morphological problems of Checking Theory are ameliorated; and three, advances in morphology lead to advances in syntax “by itself”. References Acquaviva, Paolo, 1994. The representation of operator-variable dependencies in sentential negation, Studia Linguistica 48/2: 91-132. Acquaviva, Paolo. 1995, Operator composition and the derivation of Negative Concord. Geneva Generative Papers 3/2: 72-104. Acquaviva, Paolo. 1997a, The Logical Form of Negation. Garland, New York. Acquaviva, Paolo, 1997b. The LF of Negative Concord. In Empirical Issues in Formal Syntax and Semantics, ed. Francis Corblin, Danielle Godard and Jean-Marie Marandin, Bertin: Peter Lang. Aronoff, Mark. 1994. Morphology by Itself: Stems and Inflectional Classes. Cambridge, Mass.: MIT Press. Baker, Mark. 1988a. Incorporation: A Theory of Grammatical Function Changing. Chicago: University of Chicago Press. Baker, Mark. 1988. Morphology and Syntax: An Interlocking Dependence. In Morphology and Modularity. In Honour of Henk Schultink, ed. Martin Everaert, Arnold Evers, Riny Huybregts and Mieke Trommelen, 9-32. Publications in Language Sciences 29, Berlin: Peter Lang, Beard, Robert. 1987. Morpheme Order in a Lexeme-Morpheme Based Morphology. Lingua 72: TH-1I6. Beard, Robert. 1995. Lexeme-Morpheme Base Morphology. Albany: State University of New York Press. Reghellt, Filippo and Timothy Stowell. 1997. Distributivity and Negation. In Ways of Scope Taking, ed, Anna Szabolesi, 71-107. Dordrecht: Kluwer. Belletti, Adriana, 1990. Generalized verb movement. Torino: Rosenberg and Sellier. Belletti, Adriana. 1994. Verb positions: Evidence from Italian. In Verb Movement, ed. David Lightfoot and Norbert Hornstein, 19-40, Cambridge: Cambridge University Press. Bobaljik, Jonathan David. 1995. Morphosyntax: The Syntax of Verbal Inflection, PhD diss., MIT, Cambridge, MA. 33Paolo Acquaviva Bobaljik, Jonathan David. 1997. If the head Fits... on the morphological determination of | Germanic syntax. Linguistics 35: 1029-1085. Brody, Michael. 1995. Lexico-Logical Form: A Radical Minimalist Theory. Cambridge, Mass.: MIT Press. Chomsky, Noam, 1981. Lectures on Government and Binding. Dordrecht: Foris. Chomsky, Noam. 1995. The Minimalist Program. Cambridge, Mass.: MIT Press. Cinque, Guglielmo. 1997. Adverbs and Functional Heads: A Crosslinguistic Perspective. Ms., University of Venice. To appear with Oxford University Press. Grimshaw, Jane. 1991. Extended Projection. Ms., Brandeis University, Waltham Mas Grimshaw, Jane. 1997. Projection, Heads, and Optimality. Linguistic Inquiry 28 373-422, Hale, Kenneth and Samuel Jay Keyser. 1993. On Argument Structure and the Lexical Expression of Syntactic Relations. In The View from Building 20: Essays in Linguistics in Honor of Sylvain Bromberger, ed. Kenneth Hale and Samuel Jay Keyser, 52-109. Cambridge, Mass.: MIT Press Halle, Morris and Alec Marantz, 1993. Distributed Morphology and the Pieces of Inflection. In The View from Building 20: Essays in Linguistics in Honor of Sylvain Bromberger, ed. Kenneth Hale and Samuel Jay Keyser, 111-176. Cambridge, Mass.: MIT Press, Holmberg, Anders. 1997. Introduction: Some Reflections on Movement. In Clits, Pronouns and Movement, ed. James R. Black and Virginia Motapanyane, 9-22 (Current Issues in Linguistic Theory 140). Amsterdam / Philadelphia: Benjamins. Homstein, Norbert. 1995, Logical Form: from GB to Minimalism. Oxford: Blackwell, Huang, C.-T. James. 1982, Logical Relations in Chinese and the Theory of Grammar. PhD diss., MIT, Cambridge, MA. Iatridou, Sabine. 1990. About Agr(P). Linguistic Inquiry 21: 551-571. Koopman, Hilda, 1994, Licensing heads. In Verb Movement, ed. David Lightfoot and Norbert Hornstein, 261-296. Cambridge: Cambridge University Press. Mancini, Maria Rita. 192. Locality: A Theory and its Empirical Consequences. Cambridge, Mass.: MIT Press. Manzini, Maria Rita and Anna Roussou. 1997. A Minimalist Theory of A-Movement and Control. Ms. Universita di Firenze / University of Wales at Bangor. Nishigauchi, Taisuke. 1990. Quantification in the Theory of Grammar. Dordrecht: Kluwer. Noyer, Rolf, 1997. Features, Positions, and Affixes in Autonomous Morphological Structure. New York: Garland Press. Payne, John R. 1985. Negation. In Language Typology and Syntactic Description, vol. 1: Clause Structure, ed. Timothy Shopen, Cambridge: Cambndge University Press. Pollock, Jean-Yves. 1989. Verb Movement, Universal Grammar and the Structure of IP, Linguistic Inquiry 20: 365-424 Reuland, Bric. 1990, Head Movement and the Relation Between Morphology and Syntax. Yearbook of Morphology 3: 129-161. Rizzi, Luigi. 1995. The fine structure of the left periphery. Ms. University of Geneva. Roberts, Ian. 1992. Verbs and Diachronie Syntax. Dordrecht: Kluwer. 34Deriving [+strong] features from morphology: head movement Sells, Peter. 1995, Korean and Japanese Morphology from a Lexical Perspective. Linguistic Inquiry 26: 217-325, Simpson, Andrew. 1995, Wh-movement, licensing and the locality of feature-checking. PhD. diss., University of London, School of Oriental and African Studies. Sold, Jaume. 1996 Morphology and word order in the Germanic languages. In Minimal Ideas Svniactic Studies in the Minimalist Framework, ed. Werner Abraham, Samuel David Epstein, Hoskuldur Thrdinsson and C. Jan-Wouter Zwart, 218-251 (Linguistik 4kiuel! 12) Amsterdam / Philadelphia: Benjamins Speas, Margaret 1991, Functional heads and inflectional morphemes. The Linguistic Review 8 380-417, Spencer, Andrew. 1992. Nominal inflection and the nature of functional categories. Journal of Linguistics 28: 313-341. Vikner, Sten. 1991. Verb movement and the licensing of NP-positions in the Germanic languages. PhD diss., University of Geneva, Vogel, Irene. 1993. Verbs in Italian Morphology. Yearbook of Morphology 6: 219-254. Wilder, Chris and Damir Cavar. 1994. Word order variation, verb movement, and economy principles. Studia Linguistica 48/1: 46-86. Zwicky, Amold and Geoffrey Pullum. 1983. Cliticization vs. Inflection: English n’t. Languages9: 502-513, 35SINTASSI DEL VERBO E DEITTICI CLITICI NELLE ISCRIZIONI SUDPICENE Franco BENUCCI 0. In Benucci (1996) abbiamo studiato la sintassi del Verbo nell’antico Umbro, quale ci @ documentato soprattutto dalle Tavole Iguvine € da alcune altre iscrizioni ‘minor’, riconoscendovi una fase di trapasso dalla sintassi indoeuropea ereditaria a Verbo finale a quella pitt moderna caratterizzata dalla ben nota fenomenologia di Verb Second. Questo lavoro nasce sulla scia del precedente ¢, rivolgendo lo sguardo alla sin- tassi verbale delle iscrizioni ‘sudpicene’, gli fornisce una pitt ampia prospettiva dia- cronica ed una conferma di validita, pur indagando in modo particolare le specificita con cui tale corpus realizza le stesse modalita sintattiche. Cid, naturalmente, pur senza addentrarci nel dibattito sulla reciproca posizione genealogica del Sudpiceno e dell’Umbro anche in relazione alle supposte comuni origini sabine (per il quale rin- viamo agli autori citati sotto), assumendo come valida l’ipotesi generale che vede nelle iscrizioni ‘sudpicene’ una documentazione di quel!”Italico unitario anteriore da cui affiorarono “in veste autonoma, dopo lo iato culturale [determinato da un] pe- riodo di sconquasso ¢ riassetto generale, grosso modo in corrispondenza del V sec. a.C.” (Prosdocimi (1977a: 140)), le varie etnic e lingue italiche, tra cui gli umbri € 1'Umbro come lingua. 1. Il corpus delle iscrizioni ‘sudpicenc’ (edito da Marinetti (1985), cui ci rifacciamo per le sigle designative delle iscrizioni e per il loro inquadramento ed interpretazio- ne generale, pur accogliendo in alcuni punti, che verrano segnalati, le proposte puntali di altri autori) & costituito da 23 iscrizioni di vario tipo, risalenti per lo pitt al VV sec. a.C,, redatte in un alfabeto encorio molto caratteristico, la cui difficile de- cifrazione & stata fino ad anni recenti il principale ostacolo all’interpretazione delle epigrafi. Per la maggior parte si tratta di iscrizioni su cippi o stele monumentali, al- cune delle quali con rappresentazioni antropomorte in rilievo (intera figura umana 37Franco Benucci (TE2) o volti stilizzati (TES, TE6)).! A questi si aggiungono alcuni manufatti ‘art stici’ (Pisside da Campovalano TE4, Bracciale di Chieti CH1, Guerriero di Cape- strano AQ2) e due elmi di IV-III sec. a.C., ritrovati rispettivamente a Bologna (BO1) ed a Canosa (BA1) Sia per le loro caratteristiche alfabetiche ed epigrafiche che per quelle figurative, le iscrizioni sudpicene denotano una facies culturale antica molto caratterizzata, che copre, pur con variazioni al suo interno, la vasta area corrispondente alle odiemne Marche meridionali (MC, AP), Abruzzo (TE, CH, AQ) ¢ provincia di Rieti.2 Carat- teristiche alfabetiche ¢ linguistiche ed elementi del contenuto delle iscrizioni, con il reiterato (presumibile) richiamo alla sabinita dei mandanti e dei fruitori delle epigra- fi (in TES-7), la menzione della coppia istituzionale ocar-tota (in TE7, v. anche TES, CHI ¢ RI) ¢ lesplicito richiamo detl’agro umbro (in CH2, cui fanno riscon- tro nelle Tavole Iguvine (Vb: 8-18) gli agri Tlatio e Casito del Pico Marzio), come pure la stessa area di ritrovamento delle iscrizioni, indicano I’esistenza in fase arcai ca di un’unita culturale, tinguistica e probabilmente politico-istituzionale tra la Sa- bina storica (RI) ¢ la fascia costiera medio-adriatica (il che conferma le notizic stori- che delle fonti antiche) e di tegami, certo diversi ¢ pil dilniti, con l'Umbria, fornen- do spunti del massimo interesse sia dal punto di vista storico, culturale ed istituzio- nale che da quello linguistico (storico e classificatorio). Non ci soffermeremo oltre su questi aspetti, per essere materia che esula dall’ambito immediato di questo tavo- 1 Riteniamo vada invece ridimensionato laspetto generale antropomorfo (figura umana di profilo), riconosciuto da Marinetti (1985) in MC2, AP3 ¢ APS. Cid ¢ dovuto all’attuale sistemazione museale delle stele, del tutto fuori terra o addirittura rovesciate (MC2): ri- pristinando la collocazione originaria, e tenuto conto della piit rigorosa tipologia (e cor- relata terminologia) monumentale definita in Mezzena (1988: 321-2, secondo cui “come stele antropomorfe si debbono intendere solo quei monumenti mobili nei quali siano pre- senti ¢ ben riconoscibili i tratti umani- sagome, particolari anatomici, abbigliamenti, attri- buti”), Ie tre stele in parola andranno piuttosto attribuite alla classe dei “menhir, stele grezze e aniconiche”, oft. Marinett (1985: Figg, 2,7 11)). 2 Gili elmi di Bologna e di Canosa, per loro natura oggetti trasportabili, non sono significativi per I'aspetto areale della ‘sudpicenita’ culturate, ma lo sono per quello della sua proiezio- ne Nel tempo, collocandosi almeno un paio di secoli dopo il resto del corpus, in un’epoca in cui I’uso alfabetico locale sembrerebbe altrimenti dissolto e per la quale non vi sono nora altri affioramenti epigrafici di orizzonte sudpiceno: ancora qualche decennio pitt tar- di si inizieanno a produrre in zona le ben note iscrizioni marrucine, vestine e peligne, di orizzonte culturale locale ¢ redatte in alfabeto latino, segno di un nuovo assestamento sto- rico, sociale e culturale dopo il definitive tramonto di una fase unitaria all’insegna della sabinita, di cui fe nostee iscrizioni sono testimonianza 38Sintassi del verbo e deittici clitici nelle iscrizioni sudpicene ro, trattata col dovuto approfondimento da numerosi autori (cff., in anni recenti, M: rinetti (1981 ¢ 1985), Prosdocimi (1977, 1987 ¢ 1993), Meiser (1987), Adicgo (1990 © 1992), ecc.): ci bastava richiamare qui il fatto per quanto di utile alla nostra indagine sintattica esso potra dare, Tra le 23 iscrizioni, 6 non presentano forme verbali (MC2, AP4, APS, TE3, BO], BAL) ed altre 4 presentano Verbi isolati (APG: ?}heries(? ‘(se) vorrai’) o in contesti frammentari (TE6: Jnis safintim nerf persukant pl *...dei-Sabini 1-principi invocano...", AQ3: Joitis botiediin haligati{ *...nel boviedo si-haligatcyet ’) 0 co- munque sintatticamente non perspicui (TE4: apies esum ‘di-Appio sono") } Le 13 isctizioni rimanenti, in cui le forme verbali presenti sono inserite in un contesto epi- grafico integro o sintatticamente integrabile con sufficiente certezza, attestano un totale di 30 strutture frasali (19 certe): 22 (13) principali e 8 (6) subordinate, 2. Se analizziamo tali frasi dal punto di vista della collocazione lineare del Verbo, possiamo riconoscere due modelli fondamentali, presenti sia nelle frasi principali che nelle subordinate. In 8 frasi principali (6 certe) e 5 subordinate (4 certe) il verbo appare in posizione finale’, come é attendibile a priori in una antica lingua indocu- ropea: 3 Indecidibile tra Verbo finale ¢ Verbo secondo, in quanto secondo di due costituenti lessicali in frase principale. 4 In due casi, il Verbo occupa in realta la penultima posizione della frase (i.b) interpretata secondo Marinetti ep., cfr. anche Meiser (1987: 115) e Adiego (1992: 76-7) (i) a. pupanis nir mefiin veiat vepeti (MCI: 2-4) il-piceno principe in-mezzo sta, nel-sepolcro b. Jminoh homanah molkiljjh kladieh katiieis pugloh praistait pom{ (AQI: 2-5) dalla-terra jl-monumento a-Gavio di-Gavio al-figlio si-erge, che (XY fece?) Coerentemente con quanto proposto in Benucci (1996: 108-15), analizziamo questi esempi come casi di estraposizione ed aggiunzione a destra di un (sotto-)costituente, in ragione della sua natura di chiarimento importante pet la comprensione dell’epigrafe ((i.a) dove vepeti & da interpretarsi allora come afterthought, cft. sez. 7.2.) 0 della sua ‘pesantezza’ lessicale e strutturale (i.b). In modo analogo saranno allora da analizzare altri due esempi, in cui i] contesto Jacunoso la- scia vedere una posizione ‘seri-finale’ del Verbo, ma non permette una completa inter- pretazione del dettata ((ii.b) in trascrizione basata su Adiego (1990a), Meiser (1987: 119- 20), Marinetti (| 98:1), con nostra integrazione ed interpretazione): Gi) a. Yojir imefion | Jerut atehdl (AP3: 1-2) il-principe nel-mezzo Xette, 72? 39Franco Benucci (1) a. miireis maro{n|im |ujeliam uelaimes staties qora kduiit (CHI: 4) Mureio dei-magistrati decisionali (e) Velaimo Statio la-stele offrironoS b. udifas| fitas estas amgenas dfilkdeintem atim- jplepie[ (AP3: 2-3) Vodieno del-giuramento di-codesta gens il-magistrato-dicente X avri- purificatos b. es]meh lufaniom ombriien akren postiknam putih kniiskem diinoh defia iflfift ] ti- tii tefeh (CH2) a-questo volentieri netl’umbro agro T'ultima che (io-}abbia-conosciuto, (co- meMdono Defia offti al-genio (e) a-te Per fufaniom cft. Osco loufir ‘oppure’ (Ve.2).con regolare sviluppo fonetico *bh > f dalla stessa radice di Latino libet/lubet (in Osco con lo stesso sviluppo semantico di Latino vel, Umbro her’); per 1fffift (if fit ?) cfr. Osco aflukad ‘porti, diriga’ (Ve.6), qui con pre- verbo corrispondente a Latino ob “verso” nel valore di oblatus, La stessa analisi di estraposizione e aggiunzione a destra di un costiruente ‘pesante’ potra es- sere applicata ad un quinto caso, in subordinata, in cui il Verbo appare incastrato tra due sintagmi wh, e cid in allemativa al considerare tale altestazione come un caso di gapping in coordinazione: (iii) povaisis pidaitipas fitiasom (TES: 1-2) chiungue-(tu-}sia (e) di-qualunque delle-stirpi (< ‘chiunque e di qualunque stirpe tu sia’, piuttosto che ‘chiunque mu sia e di qualunque stinpe (tu sia)') Per motivi di verosimiglianza sintattica (si veda ad esempio il caso delle lingue germaniche moderne come Tedesco e Olandese), nel caso delle subordinate calcoliamo infati tra le occorrenze di Verbo finale anche le frasi composte di un unico costituente (I'operatore di complementazione) oltre al Verbo, come putide pepie| (TEL: 4-5) *(qu)alora avra- purificato’ e putih kntiskem (CH2) ‘che (j0-)abbia-conosciuto’. 5 Seguiamo qui l’interpretazione di Marinetti (1985), che vede in questa frase la formula sottoscrittoria dei magistrati che hanno eretto la stele (contenente per il resto una prescri- zione di carattere sacrale). Per kduiti come Verbo di “dare” con preverbo ke- “her, fuori’, ff. Marinetti (1984: 41), Prosdocimi/Marinetti (1989: 290), Prosdocimi (1995: 280). Se- condo Rix (1994), che anatizza invece la forma verbale come corrispondente 2 Latino clueo, Vintera frase va interpretata come dedica al destinatario della stele, non nominato nel testo precedente: ‘del grande tra i magistrati [decisionali}, di V. S. il monumento sono chiamato’. Qualora fosse questa I’interpretazione pit corretta (ma ci pare meno perspicua, anche alla luce del sostanziale accordo tra gli studiosi circa I’interpretazione complessiva delle stele di Crecchio quale lex sacra), essa comporterebbe I’esclusione della frase in (1) dal novero di quelle rilevanti per !'analisi della sintassi verbale, in quanto essa verrebbe ad essere composta da soli due costituenti (cff. n. 3) 6 Rix (1994), pur non offrendo un quadro interpretativo altemativo né un significato specifi- co, analiza dlifkdeiniem come derivato dalla radice *klei- ‘chinarsi, appoggiarsi’ (il de- clinante?). 40Sintassi del verbo e deittici clitici nelle iscrizioni sudpicene (2) poioiieta iokipedu pdufem okjrjikam enet (CHI: 2-3) da-qualsiasi-parte per-il-territorio(-sacro?) la-via dell’arce si-inizi? Inaltre 14 frasi matrici (7 certe) ed in 3 subordinate (2-certe, v. n. 9 10) il Ver- bo appare invece in ‘seconda’ posizione, preceduto da un altro costituente della fra- se, non necessariamente il Soggetto: (3) a. apacs qupat jelsmin (MCI: il-pater giace qui ». postin viam videtas tetis tokam (TE2: 1-2) lungo la-via vede(6)te di-Teti la-tomba -2) Qualora il costituente preverbale sia diverso dal Soggetto, questo, se é lessicale, appare in posizione immediatamente postverbale:® 7 Ci sembra in questo caso che vi sia accordo sostanziale (ma non etimologico!) tra Pinterpretazione di Marinetti (1985), adottata nella glossa, e quella di Rix (1994), che ve- de in pdufem un derivato della radice *pleu ‘scorrere' (il percorso ?) Si & reso |’ablativo iokipedu con ‘per il territorio (sacro ?)” sulla base di Marinetti (1985: 112- 3) che riconosce in -pedu “un potenziale termine relativo a ‘terreno, suolo” (eft. Umbro er'om)": a nostra volta abbiamo interpretato ioki- come relativo alla potenziale sacralita dell’area circostante I’arce sulla base di un confronto con termini umbri, € cid con due possibilita. Da un lato TI IIb: 23 e Ill: 28 attestano un iukula (mersuva) ‘formule (di ti- to)’, collegato alla radice IR *iet- ‘parlare’ (eff. Latino iocus ‘discorso scherzoso"), per cui lo iokiped potrebbe essere inteso come ‘territorio éffato’ cio’ consacrato per mezzo della parola (a scopo augurale-auspicatorio, cft, TI Vla: 8-16 in cui sia lo spazio “ che a prospezione augurale sono definiti e delimitati in rapporto ai confini cittadini (eft. Prosdocimi (1978: 746-8). Alternativamente, tenuto conto della palatalizzazione di /- mostrata da CHI (cl, ‘eperen vs. vepeten di MCI, TE2 < lepetin di MC2 ‘lapide, sepol- fo’ cft. Umbro (TT parsim) \aper’- ‘lapide, pietra’), si potrebbe porte ioki- in parallelo con Umibro (TI passim) uocs-ivuku- “luco, bosco sacro’ ed intendere lo iokiped come ‘ter- ritorio di pertinenza del luco’. In entrambi i casi, si ricava una sacralita (per lo meno ge- nerica) del territorio che circondava (da tutti j lati) 'arce di Creechio (eff. per la seconda ipotesi la persistenza di toponimi come Monteluco PG, Passo del Lucomagno TI, Piedilu- co TR “ai piedi di un colle ripido € brullo” (DEI s.v), identificato da Grant (Class. Journ 18: 220-4) con i Vacunae nemora di Plinio n.h. 111.109, nonché l’attestazione antica di numerosi luci romani ¢ latini in colle € la verosimile identificazione di Costantini (1970: 53-4) dell’iguvino vuku Coredier in uno dei colli alle spalle di Gubbio, il Monte Anscia- no, sulla base di documenti notarili medievali in cui tale monte € denominato Mons Go- redius) 8 Per un caso eccezionale in cui il Soggetto non é immediatamente postverbale, v. sotto n. 16 (eff. anche n. 15). alFranco Benucci (4) ma kupri koram opsiit aninis rakinevii pomp Ji (AQ2) me bella statua fece(-fare) Anino per-Rakineve Pomp(eo?) Come si nota dagli esempi in (3) e (4), scelti tra quelli epigraficamente sicuri, abbiamo a che fare in questi casi con la fenomenologia classica delle lingue a Verb Second (V2). A differenza di quanto avviene nelle linge moderne piit note in cui ricorre tale fenomenologia sintattica (ad es. Tedesco e Olandese), ma al pari di aleu- ne altre lingue germaniche modeme (Istandese e Yiddish: per i dettagli cfr. Benucci (1996) con bibliografia), la fenomenologia V2 ricorre nelle iscriziom sudpicene an- che in contesti incassati: nelle subordinate, oltre al sintagma relativo, appare in posi- zione preverbale anche un ulteriore costituente:? (5) imih puih pépinum estukf apai B3) a-questi ai-quali dei-Piceni costi i-patres eressero colle-loro mani il-cippo! adstaith stais manus meitimam (AP2: A2- 9 La presenza di un secondo costituente preverbale oltre all"elemento wh é normale nelle co- struzioni a V2 incassato (eff. Benucei (1996: Cap. 3)) e si riscontra in due su tre delle fra- si sudpicene di questo genere: quella in (5) ¢ CHI: 1 di cui ci occuperemo sotto a sez. 7.1. In (5) il costituente preverbale & il Soggetto lessicale, mentre in CHI: 1 @ il Predicato dell Oggetto relativizzato, mentre il Soggetto non é espresso: & naturale assumere in tal caso la presenza di un Soggetto non-lessicale pro, regolarmente collocato in posizione postverbale, come nello schema: (wh Pred V pro]. L’analisi con un Soggetto pro, questa volta collocato in posizione preverbale, sara allora da estendere alla terza subordinata a V2, in cui non compaiono né il Soggetto né il normale costituente lessicale preverbale, Tale analisi ripristina la regolarita sintattica, ma impone allora di interpretare it costi- tuente in lacuna a[ non come Soggetto di un Verbo intransitivo (cosi Marinetti (1985: 134-5)), ma come Oggetto (alritis) dello stesso Verbo in valore transitivo (possibilita ammessa ed attestata per i Verbi della classe di sta- (cfr. Marinetti (1985: 74-5)), allora con Soggetto pro riferito ai cittadint erettori delle goras = aritis, riformulando unipotesi scartata da Marinetti (1985: 134)), oppure come Locativo riferito alla molteplicita di go- vas erette (alkren (in valore comune, non territoriale, come & l'ager emps et termnas di Ass.1 (Rocea (1996: 40-2)) e l'ager del paterfamilias catoniano), 0 altro), allora con Ver- bo intransitivo (inaccusativo) e Soggetto pro coreferente ai monumenti eretti) (®)_Jips tosta tefet posmi praistaint al (TE7: 4-3) V2 non sembra sulficientemente probante, data Pesiguita del campione c quindi I"alta probabilita di distorsioni statistiche, per con- clusioni certe su questo punto 11 propone invece una diversa interpretazione, che vede in ognuna delle due facce di AP2 una frase compiuta ¢ indipendente dall’altra, come indicato in (i) (integriamo tra quadre la possibile resa di due termini lasciati non tradotti da Adiego (1992); per imih ‘profon- damente’ eft. RII (Marinetti (1985: 149)): (i matereih pateresh qolofitir qupirih aritih imih push // pipdnum estuk? apaitis adstaith suis manus meitimtim alla-madre (e) ak-padre [sia-dato-onore] bene, con-arte, [profondamente] (e) piamente. dei- Piceni qui i-paires stabilirono colle-loro mani il-dono Si noti che dal punto di vista statistico generale ({rasi a V# vs. V2) questa interpretazione al- ternativa non cambia nulla rispetto all'interpretazione di Marinetti (1985) (‘della madre del padre sia cclebrats la bella opera d'arte, a questi ai quali..."), trattandosi in ogni caso di due frasi a V2: nelinterpretazione di Adiego (1992) sparirebbe perd una incassata a ‘V2 e aumenterebbero corrispondentemente {e frasi principali, Per la presenza (in entram- be le interpretazioni) del Locativo estufk nell'ambito del costituente piiptinum apatiis v sotto a sezz. 6.1, 7.48, 11 Essa conferma comunque la priorita del Verbo nel cambio sintattico, contro quanto soste~ nuto da Konneker (1972, 1975), che assegna invece ta priorita ai costituenti nominali, ed 43Franco Benucci Lo stesso dicasi per le apparenti differenziazioni areali, che danno il Centro (TE: 8 frasi principali a V2 su 8, 4 su 4 certe) ad uno stadio pitt avanzato che il Nord (AP, MC: 3 principali a V2 su 7,2 su 5 certe) ed il Sud (CH, AQ, RE: 4 su 7, 1 su certe), il che corrisponderebbe peraltro ad un’ aspettativa ‘bartoliana’!2 (anche in considerazione del fatto che é da Penna S. Andrea TE che sembra pro- manare la ‘sabinita’ sudpicena (cfr. Marinetti (1985: 32-40)) e quindi, presunti- vamente, la base della koiné linguistica e culturale) ¢ sarebbe tendenzialmente coerente con la situazione riscontrata in Umbro (area marginale e di ristagno so- cio-culturale, oltre che probabile ‘ramo collaterale’ della stessa famiglia linguisti- ca: cfr. Marinetti (1985: 40-3), Prosdocimi (1987 ¢ 1993: 130), Adiego (1990), ecc.), dove il trapasso completo alla sintassi V2 & sostenibile su base statistica solo per le iserizioni (Tavole Iguvine e ‘minori’) di I sec. a.C. (cfr. Benucci (1996: cap. 5)).13 a conferma di quanto osservato in Benucci (1996: 153-4) per !"Umbro. Si confrontino i dati in (6) con la netta predominanza delle sequenze AN rispetto a quetle NA (5 casi con- tro 2 in contesti epigraficamente sicuri) ¢ delle sequenze GenN rispetto a NGen (8 casi contro 4) nelle iscrizioni sudpicene. Particolarmente significative appare il caso di safinas tutas “della-sabina citti’ (TES: 4-5), rispetto a toutai maroucai ‘della-citta marrucina’ det bronzo di Rapino (Ve.218), promanante dalla stessa macro-area (in effetti dalta provincia di Chieti), ma circa 2 secoli pili tardi: il cambio sintattico sia era net frattempo esteso a Sud, raggiungendo anche l'ambito nominale e perfino quello det lessico istituzionale, per sua natura conservative, 12 Adiego (1992: 78-9 e passim) evidenzia perd una situazione speculare per quanto riguarda Tevoluzione fonetica come riflessa nelle iscrizioni ‘sudpicene’, in cui I’arca centrale ri- sulta pitt conservatrice di quelle periferiche. Per queste ultime si potrebbe quindi ipotizza- re una posizione simile a quella del Portoghese modemo nell’ambito delle lingue roman- ze; forte evoluzione fonologica a fronte di un altrettanto forte conservativismo sintattico, 13 Nessun tipo di valutazione diacronica é invece possil per la relativa omogeneita di datazione del corpus (VI-V sec. a.C.) che per !’inaffidabilita detl’unico criterio di datazione reciproca che potrebbe essere applicato in mancanza di contesto archeologi- co, cioé ta variazione grafematica manifestata dalle varie iscrizioni, per la quale non sono eschudibili infiuenze delle diverse tradizioni locali piuttosto che dell’epoca di in- cisione (cfr. Marinetti (1985: 61-2)). Appare tuttavia significative che |’iscrizione det “Guerriero di Capestrano’ (AQ2), unanimemente considerata dagli studiosi come la pity antica del corpus proprio sulla base delte “caratteristiche grafiche di notevole arcai- smo” (cfr. Marinetti (1984: n, 12 e 1985: 57, 239), Adiego (1992: 84)) malgrado le oscillazioni della datazione archeologica proposta dai diversi autori (cfr. Marinetti (1985: 62 n.18), presenti gia una costruzione a V2.Sintassi del verbo e deittici clitici nelle iscrizioni sudpicene 3. La polarizzazione sintattica V#/V2 ¢ mdieata con certezza dalle frasi integre corroborata da quelle strutturalmente integrabili, tra le quali si riscontrano le se- guenti sequenze a V2 (per quelle a Verbo finale v. n. 4, per un caso di V2 in subor- dinata v. n, 10) (F = Frase subordinata): (7) a. apinis qupat af (AP3: 1) il-pater giace... (SV Loc?) b, e suhuh suaipis ehueli de{ (TEI: 2-3) tra i-suoi se-qualcuno vorra dia... (FV... c. brimeidinais epe[ _o}pstq qoras qdufeniii (TET: 1-2) per-l’attacco purific... // ..fece-fare) le-stele per-il-dominus!4 (AvvV S?, VOL) 4, fe}rtur brimeqhii alintion okrei safinal (TE7: 6-5) ... &-offerto per-il-guerriero degli-Alinti nell’arce. la-Sabina (cit Log, S...)15 14 Interpretazione secondo Marinetti (1985: 137-9 ¢ c.p.). Rix (1994) interpreta qdufenii come Nome proprio: ‘per-Clufennio’, 15 Adottiamo qui la prima delle ipotesi interpretative avanzate da Marinetti (1985: 135-9), scartata peraltro a favore dell’altra che, utilizzando tutto il materiale epigrafico di que- sta sezione in un’unica frase, la rende cosi: ‘é offerto per il guerriero l'alinto (= mo- numento) nell'arce della Sabina (citti)', postulando una -s di Genitivo in inizio di la- cuna, Tale soluzione altemativa ci sembra infatti non solo pitt costosa per la necessaria Integrazione di -s ¢ la ‘riqualificazione’ del monumento, gia definito gora, come alin- tio, ma anche improbabile su base sintattica, Considerando infatti Jo schema sequen- ziale, si avrebbe: ... VISL.or con il Soggetto lessicale non immediatamente postverbale, rma separato dal Verbo dall’Ogectto Indiretto (beneficiario), contro la norma delle costruzio- nia V2 e senza che se ne poss dare un’analisi per aggiunzione che non sia del tutto ad hoc (ft. n. 16). L'analisi adotiara nel testo, che fa di safina V’inizio de] Soggetto di una seconda frase in lacuna, presenta inoltre il vantaggio di fornire il nome dell’arce in cui si riconosce la safina toita, altrimenti inattestato nelle nostre iscrizioni, mantenendo per entrambi i sintagmi Pordine AN, ancora canonico a quest'epoca (vs. i pill tardivi ocar tarincri, toula marouca, cff.n. 1). Resta naturalmente l’incognita del valore lessicale di alintiom ¢ della ragione per cui un arce dovesse avere quatificazione al plurale; se I'ipotesi non fosse troppo azzardata, sarebbe allettante vedere negli alinti- il corrispondente sudpiceno dei Dioscuri/"”Avaxss, allora naturalmente plurale e ‘di ambito giovio’ come le qualifiche degli arci italici documentati (cfr. Prosdocimi (1977: 38-9)). Con ulteriore azzardo, si potrebbe vedere nell’arce dioscu- ride il corrispondente istituzionale delle toiitaih (probabile plurale) di RII, cio’ il refe~ ente divino (0 lipostatizzazione) del rapporte *gemellare’ tra Cures (Fara Sabina, sito di 45Franco Benucei e. Ims im fitias im // Jededa imid{ (RII: AS, C) .. Xamo questo della-stirpe... // ..diede dallimo/questi (2)... (..VO. VLocl0...) La situazione sintattica cosi delineata, i cui & spia macroscopica Ia tendenziale asimmetria tra principali e subordinate, é confermata da una serie di fenomeni col- laterali, documentati anche da aicune delle iscrizioni mutile ¢ non direttamente uti- lizzabili, che permettono un maggior dettaglio di analisi strutturale. Oltre ai gia citati casi di estraposizione ed aggiunzione a destra da frasi a V# (cfr. n. 4), sono rilevanti i casi in cui, in una frase subordinata, il Complementatore o sintagma wh di apertura appare preceduto da un ulteriore costituente estratto dalla subordinata stessa e che possiamo considerare focalizzato: (8) a, e sihith suaipis ehueli def (TEI; 2-3) tra i-suoi se-qualeuno vorra, dia... b. safinas tutas trebegies titti praistaklasa posmui (TES: 3-4) della-Sabina citta Vedile al-genio, monumento-questo a-cui (8, con copula @) c. Jah selah pimpih[ (AQ3: 1) -~ del-villaggio chiunque. RII) ¢ Penna S. Andrea (sito di TES-7), adombrato in Marinetti (1985: 32-40, efi. anche 147-8). Dal punto di vista lessicale, alintion richiama cvidentemente il Greco ‘adwvSa “giro, (av)volgo’, corradicale di Latino volvo (< IE *wel/ule-): data Virregolarith fonctica ([#¥] € di solito conservato in Italico) si dovra pensare ad un grecismo (cfr. Prosdocimi (1978: 1049-88), da considerare come ‘descrittivo’, basato cioé sull’iconografia classica dei Dioscuri greci ‘awolti’ nel chitone 0 nella clamide: una variante ‘tecnica’ del tipico paradigma di prestito Warter und Sachen. Come fonte dell’itradiazione si possono consi- derare in questa ipotesi gli scultori greci stanziati o transitanti al seguito dei marmi a Nu- mana (emporio greco fin dal VII sec. aC. ¢ dalla fine del VI sec. colonia attica) o in uno dei numerosi scali secondari in territorio piceno a Sud del Concro, siti alle foci dei fiumi (tra cui Castram Novum presso il Vomano, fiume di riferimento di Teramo ¢ Penna S. Andrea) lungo importante rotta adriatica occidentale che univa la (Magna) Grecia alla Daunia, al Piceno ed ai territori venetici, veicolo di diffusione ¢ scambio reciproco di beni materiali ¢ di moduli artistici significativi (cfr. Luni (1988: 347-8), Balestrazzi (1989: 179-80, 185)). La datazione di TE7 al V sec. a.C. sembra rendere praticabile questa ipote- si interpretativa, dato che essa precede di circa 3 sccoli le iscrizioni Ve.202 (peligna), Ve.224 (marsa) € Poc.20 (osca da Pietrabbondante), in cui i Dioscuri italici risultano ormai inseriti nel modulo teonimico greco ed etrusco (ma non romano!), con la designa- zone (iouies) puk(ejle-“figli di Giove’ (cfr. Prosdocimi (1978: 842-3) 46Sintassi del verbo e deittici clitici nelle iscrizioni sudpicene Si conferma cosi la situazione strutturale e diacronica proposta per l’Umbro (eff. Benucci (1996), che trova riscontro in precisi paralleli de! Latino (cfr. Salvi (1993 ¢ 1996)) € del Germanico pitt antico (cfr. Raschella (1974/75), Kiparsky (1995)): ta struttura frasale vera ¢ propria (per breviti, designamo con IP il complesso di proie- zioni funzionali flessive, ed assumiamo che avessero testa finale, corrispondendo cosi alla posizione finale del Verbo nelle frasi del tipo di (1) e (2)) era preceduta da una duplice proiezione di tipo complementativo a testa mediana (CP > AgrCP, per un totale di 4 posizioni: 2 Specificatori ¢ 2 teste) dove avevano luogo i fenomeni sintattici qui rilevanti (focalizzazione in Spec-CP; complementazione in C°, movi- mento wh in Spec-CP o Spec-AgrCP). Schematizziamo in (9) la struttura proposta per Ia sezione superiore delle frasi in (8), esemplificate con (8.a): % cp. VN Spee @ ce AgrCP YN Spec Agr oN Agi IP aS © suai [pis 4] ,_ ehueli siihith Una struttura come (9), disponibile in tutte le frasi sudpicene (e umbre), indipen- dentemente dall’ essere esse selezionate, aggiunte o matrici, nonché dalla collocazione Gel Verbo, gioca un ruolo cruciale nella fase i passaggio dalla sintassi V# a quella V2. Assumiamo infatti che la testa della proiezione di complementazione pitt incassata (AgrCP) sia caratterizzata, come suggerisce I'etichetta, da un forte tratto di accordo che, nelle opportune condizioni sintattiche (cfr. Benucci (1996: 143-52)), attira a sé il ‘Verbo determinando, nei normali contesti matrice ed assieme al parallelo proceso di topicalizzazione di un qualsiasi costituemte in Speo-AgrCP, le caratteristiche costruzio- ni a V2, con il Verbo appunto in ‘seconda posizione strutturale’. La stessa struttura rende anche perfettamente conto della collocazione immediatamente postverbale del 47Franco Benucci Soggetto nelle frasi a V2, qualora non sia esso stesso L’elemento topicalizzato: esso si ‘rover’ allora semplicemente nella sua posizione canonica, Spec-IP:16 16 Un unico esempio del nostro corpus sembra far eccezione a questa generalizzazione struttu- male, ed é la frase centrale di TES, costruzione a V2 in cui il Soggetto appere si in posizione postverbale ma non a contatto con il Verbo, essendone invece separato da un Avverbio Sub- ject-oriented: panivii ‘chiaramente, apertamente, manifestamente’ (dalla stessa radice di Lat. ‘pando -ere ‘aprire, manifestare, spieyare”). Per la miglior comprensione dell’esempio, gio- va ricordare che TES & una stele figuratya, che riporta alla sommita la figura stilizzata di un volto umano (0 antropomorfo), da not intesa come icona del mostrum (‘spirito’) supremo (cfr. sez. 7.3 en. 34): (i mifglim mefistrii neminei praistait panivad meitims (TES: 2-3) dei-monstra per-Vinfetiore a-nessuno sietge chiaramente il-cippo Dato il valore semantico e pragmatico-contestuale dell” Avverbio, che deve avere portata sul Soggetto della frase (cioé su Spec-IP) per definime la modalita di svolgimento dell’azione, sembra naturale collocarlo sopra IP, come aggiunto 0 come Specificatore di una proiezione opzionale, ma naturalmente sotto alle proiezioni di complementazione, a pari di quanto avviene in Italiano: (Mi ha detto che) manifestamente Gianni dava vegni di impazienza (con intonazione neutra, senza pause né picchi), Si tratta dunque di un falso controesempio, proprio per lz natura dell’elemento che interviene in (i) tra Verbo e Sog- getto. Questa analisi, basata su Marinetti (1985: 117-30 e cp.), & evidentemente contearia all’interpretazione di Adiego (1992: 33) che, tralasciando le prime 3 parole di (i) ed uunendo il resto alla frase qui riportata in (8.b), ne fa un’unica proposizione a Verbo ini- ziale: ‘si erge panivit (Avv.) come dono dell’edile della citta Sabina per Tito Postumo ‘questo monumento’. E’ evidente che una tale interpretazione é del tutto inverosimile dal punto di vista sintattico, anche indipendentemente dalla specifica analisi qui proposta, proprio per la collocazione iniziale del Verbo, che non ha nessun parallelo nel corpus € non avrebbe alcuna collocazione naturale nel quadro della polarizzazione sintatica deli neata (V2/V#): si noti che nemmeno frasi dichiaratamente imperative come quelle di CHI riportate in (15) ¢ (16) comportano il Verbo in posizione iniziale. 4BSintassi del verbo ¢ deittici clitici nelle iscrizioni sudpicene Spec r ma kupri aninis —_rakinevii koramj opsit | pomp. ty (4) puih —pupinum | siiais manus adstaiih meitimim 4 (5) Jn Bemucci (1996: 143-52) si & visto che cruciale per Vattivazione strutturate di AgrCP, e successivamente di AgrC® come attrattore del Verbo, & la presenza di cle- menti focalizzati, come nelle frasi in (8). Abbiamo infatti assunto che la struttura in (9) ¢ (10) sia interamente disponibile non solo per le frasi subordinate, ma anche per le principali: non stupira quindi di trovare anche nelle iscrizioni sudpicene almeno un ca- 80 di costruzione a V2 in frase principale con un ulteriore elemento preverbale:!7 (11) Sidom safinds estuf eSelsit tiom povaisis pidaitipas fitiasom (TES: 1-2) come-nemico, i-Sabini costi respingono te, chiunque-(tu-)sia (€) di-qualunque delle-stirpi Nella nostra analisi, il predicato didom va interpretato come elemento pragmat camente focalizzato ¢ collocato quindi in Spec-CP, mantenendo per il resto la nor- male struttura derivata delle frasi a V2 con Soggetto topicalizzato ¢ Verbo in AgiC* La struttura proposta é schematizzata in (12): 17 Trascuriamo per ora la presenza, pure in posizione preverbale, del Locativo estuf, su cui tomeremo subito sotto. 49Franco Benucci (12) Ln~w - Spec c Vio e AgrCP Vo Spec AgrC” JN AgrC? IP SN Spec r | | | US Sidomj — safinus; eSelsit fj [tiom §] fy 4. Questo quadro sintattico nitidamente configurato sembra in qualche modo turbato da 4 frasi, finora considerate come casi di V2, in cui il Verbo é preceduto (all’interno della proiezione rilevante, AgrCP) da due costituenti anziché da uno solo come ci si aspetterebbe in una configurazione di V2 canonico. L?identificazione di AgrCP come proiezione che contiene, oltre al Verbo ed al co- stituente topicalizzato, anche il costituente ‘supplementare’, deriva dal fatto che esso appare sempre incastrato tra il Verbo stesso (AgrC°) ¢ l’elemento topicalizzato (Spec-AgrCP). Questi casi vanno quindi tenuti distinti da quelli in cui il Verbo & pu- Te preceduto da due costituenti, i! primo dei quali (elemento focalizzato 0 wh) va pero collocato in Spee-CP (v. sopra), Si noti che in 3 casi su 4 (13.2,¢,d) i costi- tuente ‘supplementare’ coesiste con un elemento di Spec-CP, portando cosi a tre il numero dej costituenti preverbali; cid non ostante continueremo a sostenere che si tratta di frasi con sintassi V2 (riportiamo qui solo le parti di iscrizioni rilevanti): (13) a. imih puth pipinum estufk apaitis adstaiiih siiais manus meitinim (AP2: A2-B3) a-questi, ai-quali dei-Piceni costi i-patres eressero colle-loro mani il-cippo b, alics esmen vepses vepeten (TE2: 2-4) Alio qui riposa net-sepolero ¢. Sidom safinis estuf e¥elsit tiom povaisis (TES: 1-2) come-nemico, i-Sabini costi respingono te, chiunque-(tu-)sia 50Sintassi del verbo e deittici clitici nelle iscrizioni sudpicene d. deiktam hf--]Ipas pimoftorim esmenadstaeoms (CHI: 1) la-dichiarazione di-h..[pa, che-come-stele qui-innalzammo!8 In tutti i casi considerati, il costituente preverbale ‘supplementare’, evidenziato, @ un deittico Locativo: esmen 0 estuf(k). Per estuf{k) (2 attestazioni) cid esaurisce la casistica, ma per esmen i 2 Casi, attestati secondo lo schema generale (wh-)X-esmen- V-Y, si oppongono ad altre 4 attestazioni in cui fo stesso deittico compare in altre posizioni: finale di frase in 2 casi (di cui uno a V2 (14.a) ed uno senza Verbo (co- pula zero 0 Verbo in lacuna?) (14,b)), terzo costituente (almeno, date le lacune) in altri 2 casi, entrambi a V# (14.¢, d): (14) a, apaes qupat jelsmin piptinis nir mefiin veiat vepeti (MC1) il-pater giace qui. il-Piceno principe in-mezzo giace, nel-sepolero b. apais pomp{ J pufnles lepetin esmin (MC2) il-pater X (@/giace) Y nel-sepolero qui ¢. Jraicimim titi anaium aidaqum esmin idiins uv{_ Jpeiéit| (API) il-monumento al-genio degh-anaio audaci qui V'odieno X dedicd d. Jorom iorkes iepeten esmen ektisim raeliom rufrasim [...] bie (CHI: 2-3) degli-X agli-iork nel-sepolcro qui l’ektisi dei-raeli Rufrasio [...] saluta! 5. In 4 casi su 6 (13.b, 14.a, b, d), indipendentemente dalla sua collocazione nella frase, esmen é compresente nella stessa iscrizione (ma non sempre nello stesso ordi- ne, né necessariamente a contatto, né sempre nella stessa frase) con un Locativo les- sicale vepeten ‘nel-sepolero’ (e varianti); in entrambi i casi (13.a, c) estuf(k) ricorre nella stessa iscrizione con ['Oggetto (nominale 0 pronominale) del Verbo, all’ Accusativo (non a contatto: meitimtim ‘cippo’, tiom ‘tc’). Riteniamo di poter escludere Pinterpretazione, adombrata in Marinetti (1981: 130) e messa in atto in Marinetti (1985: 106, 110, 142-5, 153) e Untermann (1990), di esmen ... vepeten/vepeten esmen come ‘in-questo sepolero’ ¢ di estufk ... mei- timiim come ‘codesto cippo’, con esmen/estufk dimostrativi,!9 sulla base del con- fronto morfologico e sintattico con i corradicali e certamente dimostrativi estas amgenas ‘di-codesta gens’ di AP3 ¢ esmik uepetil ‘in-questo sepolero’ di RU: A.20 18 Vedi sopra n. 9. 19 Per estuf’... tiom tale interpretazione ¢ automaticamente esclusa per !’evidente nonsenso di ‘una designazione come ‘codesto te”. 20 Cui vanno aggiunti gli ipotetici |es|mei ulepeti|, proposto da Sironen (1990) come integra- Zione di RII: B, ed jesimeh ... diinok (in rapporto di predicazione, il che sprega I"identita 51Franco Benucci Dal punto di vista della sintassi di posizione, notiamo infatti una prima differen- za tra l’ordine DN (a contatto) dei casi in cui il deittico ha sicuro valore dimostrativo ¢ Vordine ND 0 D...N (senza contatto) manifestato dai deittici in (13) ¢ (14). Ancora pitt significativa appare la differenza tra i due tipi di occorrenze dal punto di vista morfosintattico: i dimostrativi puri appaiono infatti in accordo morfologico con il nominale retto nei Casi primari (doppia desinenza di Genitivo (-s) in AP3 di Dati Vo (-ei) in RII: B (eff. n. 20), eff. anche casi come Umbro esmei stahmei stahmitei “in questo spazio stabilito’ (T1 Via: 5, 18) in accordo di Locativo), ma privi di po- sposizione locale (-i[n), che appare solo sul nominale retto, nei Casi locali secondari (RU: A); al contrario, tutte le attestazioni di esm- in (13) e (14) prescntano la po- sposizione -en/in, eventualmente coesistente con quella che determina in senso lo- calc (inessivo) il termine per ‘sepolero’: segno questo che i deittici esmen/estuf{k) stanno qui in valore locativo ‘qui/costi’, a titolo proprio (come nei casi in cui com- paiono da soli) 0 in qualche modo correlati al Locativo lessicale compresente (¥. sotto),2! come nell"ipotesi principale di Marinetti (1981; 148-53).22 di Caso morfologico accompagnata da distanza lineare, v. appresso in testo), da noi pro- posto sopra an. 4 (es. (ii.b)) come integrazione di CH2. 21 Assumiamo qui che la rianalisi di -en come ‘segnacaso’, cioé desinenza in accardo su tutti i ‘membri del sintagma nominale, documentata in Osco (Hurtin kerriiin nel Bronzo di Agnone) e in Umbro (Tavole Iguvine: ocrem fisiem, toteme(n) tiouinem (Vib: 26, 46), vapefem avieklufe(n) (Ib: 14), ecc.), in valore locativo (con specializeazione inessiva/ilativa e supe- ressiva/sullativa, eff, Prosdocimi (1969/70: 39-41), Nocentini (1992), Benucci (1997), abbia avuto Iuogo in epoca successiva a quella di cui alle scrzion “sudpicene”. Cid & abbastanza owio nel caso delle Tavole Iguvine e del Bronzo ch Agnone, risalenti al HILT sec. aC., ma ud diventare un importante indizio di datazione per 1! Bracexale di Chieti (CH2), dove pure é attestato, caso unico ne! corpus sudpiceno, ombriien akren con duplice desinenza locativa Ora, dato che la sequenza iepeten esmen, da noi interpretata come un caso di Locativo lessi- cale + deitico locative (‘nel sepolcro qui’) piuttosto che come Nome + dimostrativo ("¥in ‘questo sepolcro’), appare tra !’altro in CHI, iscrizione considerata da Marinetti (1981: n. 96) ‘come “almeno apparentemente rappresentaftiva della fase pit antica del sudpiceno”, e co- ‘munque una delle poche per cui & stata proposta una datazione, cid lascia praticamente in- tatto lo spazio temporale entro cui collocare i} Bracciale CH2. A. La Regina, scopritore del Bracciale stesso, ha proposto, in occasione di alcune relazioni congressuali, una datazione archeologica al V sec. a.C.: per prudenza possiamo assumere la seconda meta di tale secolo, il che pud essere confacente per i tempi di rianatisi di en come desinenza di Caso Locativo ¢ permette di superare la datazione alla prima meta dello stesso secolo proposta dallo stesso La Regina per la stele TES, in cui compare I’altro deittico locativo estuf 22 Per quanto riguarda estuf(k), !’analisi come Locativo & sostenuta anche da Adiego (1992: 101) ¢ Rix (1994: 107). L’analisi sviluppata nel testo sembra potersi estendere alla se- 52Sintassi del verbo e deittici clitici nelle iscrizioni sudpicene 5.1, LYidentificazione di esmen/estuffk) come deittici locativi piuttosto che come dimo- strativi permette innanzitutto di evitare Vinterpretazione strutturale delle sequenze D...N (in (13.a-c) ¢ (14.2) come casi di scrambling stilistico (ricerca di sequenze allit- teranti) da uno stesso sintagma, che si imporrebbe invece se si trattasse di dimostrativi, come presuppone (ed esemplifica) Marinetti (1981: 130 e 153-8, 1985: 85-8), ¢ che sembra pera in qualche misura contraria alle esigenze sintattiche (v. sotto sez. 6.4). In particolare, I’interpretazione contrastiva di tali deittici rispetto al Locativo les- sicale eventualmente compresente nell’iscrizione, & possibile sia come effettivo contrasto ‘topografico’ tra i due (qui vs. altrove) sia come rafforzamento pragmatico di tale Locativo (‘proprio qui’) rispetto ad altro riferimento spaziale contestualmente dato 0 prevedibile su base fattuale e/o di presupposizioni comuni al committen- te/esecutore ed al fruitore delle iscrizioni. L’interpretazione contrastiva rispetto ad una prevedibiliti presuppositiva & evi- dentemente possibile indipendentemente dalla realizzazione lessicale del secondo Locativo o di altro riferimento spaziale: essa potra infatti essere data dalla colloca- zione fisica del supporto dell’iscrizione, che ne fissa ovviamente nella reaita le co- ordinate spaziali espresse con ‘qui/costi’, e/o dalla sua tipologia monumentale, che rinvia ad una valenza presupposizionale, cio’ culturale. 6. Una possibile direzione di ricerca per la soluzione del problema d’ordine delie parole nelle attestazioni in cui il deittico compare come costituente “supplementare’ in posizione preverbale in costruzioni altrimenti analizzabili come casi di V2 (ess. (13)) viene dall’interpretazione ‘forte’ det suggerimento di Marinetti (1985: 82, n. 43), secondo eui Ia giunzione grafica di esmen+adstaeoms in CHI: | (13.4) @ do- vuta a ‘ragioni di atonia/cliticiti’. Malgrado la sua prudenza nell'impegnarsi nella “valenza sintattica’ del concetto di cliticita e la sua corretta distinzione da quello di atonia, possiamo assumere come ipotesi di lavoro che in questo caso, ed in tutti quelli in cui compare come costituente preverbale ‘supplementare’, il deitt quenza esmak totitaih di RI: Vordine sequenziale deporrebbe a favore di un‘interpretazione come sintagma unico, analizzabile con Marinetti (1985: 148) allo Strumentale plurale: ‘con queste citté’; verrebbe perd cosi_a mancare Iattesa identita di desinenza tra i due clementi ed acquista quindi valore l'ipotesi alternativa, seguita anche dalla stessa Marinetti (1985: 148), che vede in tale sequenza un accostamento casuale do- vato alla frammentarieté del testo, da interpretarsi allora con un deittico locativo in valore proprio (Strumentale/Ablativo?: ‘per di qui’), indipendente dal nominale (Dativo 0 Stru- mentale plurale: ‘colle/slla2/alle citti") che segue. La lacunositi dell'iscrizione e Vincertezza della morfonologia ricavabile dalla grafia non consentono perd certezze né sul piano strettamente linguistico né su quello interpretativo ed istituzionale (cfr. n. 15 per ‘un’ipotesi su totitaih come plurale).Franco Benucci effetti un Clitico sintattico (e solo di riflesso fonologico). Prima di procedere nella formalizzazione dell’analisi, verificheremo le potenziali carattcristiche di questa cli- ticita e quindi la verosimiglianza dell’ipotesi stessa. 6.1. In tutti gli esempi rilevanti, il deittico (di cui si suppone la cliticita) appare in seconda posizione lineare all’interno della proiezione pertinente (AgrCP), prece- dendo cosi il Verbo che nelle costruzioni V2 occupa la seconda posizione strutturale (AgrC?), Cid & del tutto conforme alle aspettative per una lingua Indoeuropea antica, trattandosi della ¢.d. posizione Wackemagel. In particolare, si ricade nel caso pitt antico della ‘Legge di Wackernagel’, in cui ‘seconda posizione’ significa ‘dopo la prima parola’ (contrariamente a quanto av- viene nelle lingue (slave) modeme che ancora collocano i loro elementi clitici se- condo tale ‘Legge’, in cui ‘seconda posizione’ significa ‘dopo il primo costituen- te’).23 Cid é neutrale per gli esempi tratti da TE2, TES ¢ CHI (13.b-d), ma cruciale per AP2 (13.a), dove estufk cade all’intemo del Sintagma Nominale Soggetto, ¢ pre- cisamente dopo la prima parola: pupunum estufk apaius, provocandone la tmesi € venendo a sua volta separato dal Verbo da oui dipende.24 Assumeremo quindi che tutti gli esempi in (13) sono potenziali casi di collocazione del deittico clitico ‘dopo la prima parola della frase’ (minimale = AgrCP). 23 La ‘Legge di Wackernagel’ risulla ogg: applicata da Sloveno, Serbo-Croato, Ceco ¢ Slovacco standard. In Polacco, Sorabo € Slovacco orientale essa é residuale ed é stata recentemente sostituita dalla collocazione adverbale degli elementi clitici (cfr. Benacchio/Renzi (1987: 9- 12 en. 12) con bibliografia). Si noti che in Scrbo-Croato le due apzioni di collocazione dei Clitici possono alternare, ma !opzione “dopo !a prima parola’ sembra essere recessiva (sen- tita come arcaizzante e letteraria) rispetto a quella ‘dopo il primo sintagma’ (cfr. Benac- chio/Renzi (1987: 10 ¢ n. 10), Browne (1975: 113-4) c limitata probabilmente ai casi di ‘en- fasi’ (contrastiva?) sul costituente che viene spezzato dal Clitico (cfr. Fontana (1997: n. 12): (_&. Moja mladja sestra ée dogi u wtorak mia minore sorella FUT. venire il martedi b. Moja ée mladja sestra doci u utorak mia Fut. minore sorela venire it martedi (!) (mia sorella minore verri martedi) La stessa alternanza sussisteva in Slavo ecclesiastico antico e nelle altre lingue slave citate, ma si & persa nel corso dei secoli: per esempio in Ceco la possibilita di collocare i Clitici dopo la seconda parola della frase venne meno in epoca umanistica (XVI sec.), cf, Toman (1986: 126). 24 Cid sussiste anche se si adottasse per iscrizione AP2 analisi di Adiegn (1992) discussa an, 10. 54Sintassi del verbo e deittici clitici nelle iscrizioni sudpicene 6.2. La natura sintattica della cliticita dei deittici locativi sudpiceni (che porta solo riflessi fonologici (atonia) non sempre evidenziati dalla grafia) ci sembra confermata dal contrasto in CHI: | (13.4) tra la collocazione sintattica di esmen (in seconda po- sizione = dopo la prima parola di AgrCP: oftorim) e la sua giunzione grafica (cioe fonologica: prima della seconda parola libera, il Verbo adstacoms).25 Nel caso specifico, si potrebbe pensare ad una proclisi fonologica generalizzata (almeno in questa iscrizione, cfr. pim-oftorim), o pitt probabilmente all’impossibilita per oftorim di portare due clitici fonologici (di cui il secondo anche sintattico) da ‘dire zioni” opposte (*pim-oftorim-esmen adstacoms > pim-oftorim esmen-adstacoms)26 25 Per un'opportuna distinzione dei due livelli di clisia, cfr. da ultimo Fontana (1997). Con- trasti analoghi, ma speculari (enclisi fonologica e proclisi sintattica) si avevano nelle lin- ‘gue romanze antiche, come effetto della Legge Tobler/Mussafia combinata con i contesti fonetici (i) Eu te fiz, nu’m cognouist (Passion Clermont 67) io ti feci, non-mi riconoscesti 26 Casi come le frasi interroyative Francesi (¢ Romanze) del tipo di Ol M'as-tu vue?, in cui MAusiliare as porta due Clitns fonologici (il primo anche sintattico) da “direzioni’ opposte sembrano deporre contro 1a verosimiglianza della restrizione proposta. Fatte salve le diffe- renze interlinguistiche e diacroniche, sembra tuttavia di poter distinguere i due casi su basi teotiche (cfr. Poletto (1990): in Francese, il Clitico (sintattico e fonologico) Ogeetto occupa una posizione interna alla testa flessiva complessa che ospita il Verbo anche a prescindere dall'inversione interrogativa e dalla presenza del Clitico Suggetto (eff. Tu las vue, Jean I'a \ue). ¢ come tale non ostacola la struttura di aggiunzione fonologica che interessa il Clitico Sougetto in Spec-IP (a destra di C° nelle interrogative, a sinistra di I? nelle dichiarative), co- me nelle strutture seguenti (le freccie indicano l’aggiunzione fonologica): @ a P b. CP rN 1 \ Spec 1° Spec C u/s \ otf \ yy \ c P P vp se iV LS PR Spec” Vastyywe = /\ RuL\ Fast, vue Per quanto riguanda (13.4), al contrario, si sosterra che il Clitico locativo esmen vi ricorre in una struttura di aggiunzione (sintattica) a Spec-AgrCP, che ospita anche offorim; in prima istan- Za, possiamo assumere che st si aggiungesse (da destra) allo stesso costituente anche dal unto di vista fonologico, cid ostacolerebbe la necessaria aggiunzione fonologica (da sini- stra) del relative pim (che non dispone di altro costituente interno alla frase a cui aggiungersi fonologicamente): esso é quindi obbligato ad aggiungersi fonologicamente (da sinistra) al oeFranco Benucci Questa seconda ipotesi sembra pitt coerente con la clisi fonologica (e grafica) genera- lizzata mostrata dai sintagmi wh (p- < *k-) dell’intero corpus sudpiceno, indipen: dentemente dal ‘lato’ in cui é disponibile un supporto (purché interno alla frase (estese = CP)): off. suai-pis ‘se qualeuno’ (TE1) vs. po-ioteta ‘da qualunque parte’ (TES). Cio suggerisce una ulteriore differenza tra la cliticita sintattica ¢ quella pura- mente fonologica: l'ambito di riferimento dei Clitici sintattici, sia per il ‘compute delle posizioni” che per l’appoggio fonologico & la proiezione AgrCP, che nelle co- struzioni a V2 ospita anche il Verbo: gli eventuali clementi collocati in CP (elementi relativi o focalizzati) non contano per il ‘computo delle posizioni’ e non sono utiliz- zabili come supporto fonologico, Casi analoghi ricorrono in Latino (Salvi (1993 € 1996)), Vedico (Banti (1980)), Serbo-Croato (Browne (1975)), ecc.). I clitici solo fonologict possono invece trovare supporto anche su parole collocate in altre proie- Zion (ssar-pis da AgrCP a CP, pim-oftorim da CP ad AgrCP (cf. sez. 8), povai-sis da IP a CP (si assume che il vero Clitico fonologico sia qui il Verbo, v. fine)). L’osservazione che negli esempi in (13) la proiezione di riferimento dei Clitici ¢ la stessa che ospita il Verbo, confrontata con la situazione delle lingue (ad es. ro- manze modeme ¢ slave non-Wackemagel) non a V2, dove sia il Verbo che i Clitici stanno in IP, si presta a due possibili anatisi. Una prima ipotesi, che va nel senso delle analisi dei Clitici pronominali romanzi e germanici di Kayne (1991) e Cardi- naletti/Roberts (1991), parte dal fatto che sia IP che AgrCP sono proiezioni (+Agr}: questo sarebbe dunque il tratto rilevante per la collocazione dei Clitici. La pertinen- za del tratto di accordo sembra abbastanza naturale per i Clitici pronominali (e per le forme di Ausiliati clitici delle lingue slave), ma lascia qualche perplessiti nel caso dei nostri deittici clitici (avverbiali). La seconda ipotesi si basa invece sul fatto che in tali lingue i Clitici (pronominali ed ausiliari) sono sempre associati alla stessa (macro-)proiezione in cui si trova il Verbo: si potrebbe supporre che cid avvenga perché essi ne sono argomenti (0 ri- spettivamente specificatori, inteso in senso lato, non strettamente strutturale). Si noti infatti che anche i nostri deithct clitici locativi cooccorrono con Verbi di stato/moto Cinnalzarelerigere, riposare‘giacere, respingere’), che plausibilmente selezionano un argomento locativo (nei cast in cui esmen non é clitico (ess. (14)) pud cooccorrere come aggiunto anche con Verbi che non lo selezionano: ‘salutare, dedicare’). Lallargamento del confronto a lingue come quelle germaniche a V2 (asimmetri- co), quelle slave Wackernagel, ¢ quelle romanze piis arcaiche (che, come il Polacco modemo, presentavano una duplice possibilita dt collocazione dei Clitici (fenomeno detto di ‘interpolazione’, cff. Ramsden (1963), Rivero (1997), Fontana (1997))), in cui ‘Verbo in AgrC?. L’analisi sintattica sviluppata a sez. 8 comporteri la precisazione di quanto qui schizzato (cfr. sez. 8.3), 56Sintassi del verbo e deittici clitici nelle iscrizioni sudpicene la posizione dei Clitici (in una proiezione corrispondente, nei nostri termini, ad AgrCP) @ indipendente da quella del Verbo (in IP 0 mobile tra questa e AgrCP), conferma pero la pertinenza esclusiva del tratto di accordo per la collocazione dei Clitici. Nel nostro caso si pud assumere, per fugare le perplessiti di cui sopra, che tale tratto sia rilevante per la realizzazione della deissi a 3 termini (cft. Marinetti (1985: 67-74 e passim)) del tipo di quella dell’ Umbro, del Latino ¢ del Toscano (romanzo) (esmen (esuflesme, hic) come qui/questo/questi = EGO, estuf(k) (effesto, istic/iste) come costi/codesto/costui = TU; (ulu/are, illic/lle) come (co)la/quello/eolui = WLLE)27 6.3. Se si considerano esmen/estuf(k) negli esempi in (13) come Clitici, essi non contano come costituenti per il computo della posizione verbale ¢ viene quindi me- no apparente turbativa alla sintassi V2: si avrebbe cosi una situazione paragonabile agli apparenti casi di V3 in Antico Alto Tedesco e Old English, dove uno dei costi- tuenti preverbali é il pronome Soggetto e va considerato Clitico, restituendo cosi una normale sintassi V2 (cfr. Tomasells (1995)).. Assumiamo che dove i deittici sono clitici il loro contenuto @ ‘dato’, sia in quanto contestualmente specificato © perché presupposto e prevedibile sulla base della ‘conoscenza del mondo’ e delle convenzioni sociali della civilta ‘sudpicena’ Tenteremo in seguito (sez. 7) di dimostrare questa che é la condizione stessa di sus- sitenza della nostra ipotesi Come accennato precedentemente, non in tutte le sue occorrenze esmen é analiz- zabile come Clitico, dato che, a parita di forma fonetica, esso ricorre in posizione finale 0 come terzo costituente (4* o 54 parola) della sua frase. Si tratta di un pro- blema solo apparente, perché non necessariamente i Clitici delle antiche lingue IE dovevano avere forma ridotta o comunque diversa dalle corrispondenti forme libere cosi in Latino (cf. Salvi (1993: 1-3 e 1996: 10-1)) ed in buona parte del paradigzma pronominale del Vedico (tra cui segnatamente il Locativo asmin, a cui esmen set bbra essere etimologicamente collegato, cfr. Marinetti (1981: 150), Banti (1980: 15, 18)), i pronomi mantenevano la stessa forma indipendentemente dal loro uso libero 0 clitico, differenziati solo per I’assenza o presenza di tonicita.28 ‘Assumeremo quindi semplicemente che quando esmen non é Clitico esso & usato come forma libera/tonica e non é quindi vincolato ad alcuna posizione fissa nella 27 Da un punto di vista strutturale si pud pensare ad un morfema di accordo (@/-s ) cui la base avverbiale si incorpora prima della salita ad AgrCP. II deittico locativo sudpiceno di 3 persona non é attestato. 28 Tali forme, in uso tonico, fornivano inoltre la base per Ja formazione dei dimostrativi (per il Latino v. sopra sez, 6.2., per il Vedico eff. Banti (1980: 19)), al pari di quanto avviene in Sudpiceno (cfr. sez. 5). 37