1 Introduction

The discussion brought up in this paper is by no means exhaustive and has to be intended as an overview of an initial empirical evidence from which a more precise and extensive inquiry on the nature of the functional structure of the VP in this language could develop.

A crucial point in this brief syntactic overview is played by the position occupied by manner adverbs. It will be shown that manner adverbs are VP internal, mainly considering evidence coming from the distribution of floating quantifiers, and it will be shown that in Italian the VP-internal subject position (taken to be the highest specifier inside the VP) and the manner adverbs are the left and right edges of a syntactic space that contains possible landing sites for scrambled objects.

As it will be clear, the present analysis relies substantially on the approach to reordering put forth by Cinque (1999). However the framework in which the present paper develops is not accepted without some discussion. We show that alternative views of deriving some of the word orders found, when put forth, lead to wrong predictions (the case of extraposition of adverbial material is treated).

In the last section we introduce a domain in which the effects of object scrambling are visible, in that we explore constraints posed on the readings associated with indefinite objects.

2 The position of manner adverbs and the structure of VP

In (1) the property of "being intelligent" can be predicated both of the action and of the agent, so that simultaneously the way the question was answered and Giovanni in doing it were intelligent.

In the representation we give for the Italian VP the specifier hosting the adverbial element is located lower than the VP-internal subject. The position of the manner adverb in (2) will be justified in the course of the discussion. We take the adverbial element to be generated as a specifier of the light verb v (for the concept of v head cf. Larson 1988, Chomsky 1995) ².

In addition to manner adverbs proper we have manner adverbs of the kind of bene (good), male (bad, badly). They present a syntactic behavior and a distribution that justify the assumption of a different syntactic position. These adverbs, as (3) shows, can cliticize to the left of the past participle as ben and mal, respectively, a behavior that they share with arguments. In (3) and (4) the distribution of full and clitic forms with respect to the past participle is illustrated:

Additionally, note that bene and attentamente, thought both contributing a manner modification to the predicate are not in complementary distribution syntactically, as we would expect if they belonged to the same syntactic class (see (5) below). (6) shows that the relative order of the two classes is rigidly constrained in that manner adverbs proper always have to precede bene and male:

In the light of our data and of previous analyses for Italian (Lonzi 1991) we could propose that bene/male adverbs are a sort of V complements. In a VP structure like the one first proposed by Larson (1988) these complement adverbs could occupy the most internal position, the one usually reserved to PPs, or, in alternative a left branch, as long as it is projected lower than the specifier hosting manner adverbs proper ³.

The set of data considered up this point suggests that the relative order of the elements inside the VP is likely to be: manner adverbs proper > bene-manner adverbs > DP-objects.

The structure in (9) is a more detailed version of (2). Both in (2) and in (9) the VP-internal subject position is assumed to be the highest Spec in the VP projection. However we still do not have evidence to decide whether the manner adverb proper precedes or follows it, i.e. is internal or external to the VP. The position we assign manner adverbs proper in (9) is not justified yet. In the next section we will show some evidence coming from the domain of the distribution of floating quantifiers ⁴.

3 Head-movement and VP-shell movement

As it is the case for English, Italian too has a floating quantifier tutti+DP that shows all the relevant properties of the English quantifier all+DP, i.e. the DP can strand the quantifier part tutti and move to A-positions higher in the sentence, as shown in (10), where the (a) case shows the non-stranded variant and the (b) case the stranded one (for the position of floating quantifiers in Italian see also Belletti 1990 and Cinque 1999, among others).

In order to offer a complete picture of the Italian VP we will have to determine the position of the VP-subject in cases like 10(a)-(b) . One way of doing it is by testing the relative order of postverbal stranded quantifiers and manner adverbs proper, as in the following examples:

The data in 12(a) -(e) show that the position that is likely to be assigned to the subject inside the Italian VP is quite high, maybe the highest specifier position projected by the highest v head. As 12 (d) clearly shows the position must be at least higher than attentamente. The data in 12(b)-( c) show the stranded quantifier in its base position, with the past-participle at the right and at the left of the stranded quantifier, respectively.

In 12(a) the quantifier has been stranded in an A-position, which lays between the base generation position and [Spec, AGRS], where the nominal part of the quantified nominal lands to check EPP and phi-features (Chomsky 1995). We can take this intermediate A-position to be the Specifier of TP, at least for the languages that project it (cf. Bobaljik and Jonas, 1996, among others).

In 12(e) the stranded quantifier finds itself in sentence final position. Data of the kind of 12(e) could be taken as evidence for the traditional view that in languages like Italian the postverbal subject is adjoined to the right of the VP (Kayne 1981, Belletti and Rizzi 1981, Rizzi 1982, among others). In what follows we will present evidence for a different approach. We will go along with an alternative view developed in work by Cinque (Cinque 1997, 1999). Following this approach, in a sentence like 12(e), the floating quantifier marks the base position of the subject analogously to 12(b) or 12(c).

If this is the case, it is clear that some kind of leftward movement of the VP-shell that finds itself to the right of the subject must be assumed.

Along the lines of Cinque (1997, 1999) it is possible to derive constraints on the relative order of various kinds of adverbs as a result of impossible VP-shell movement patterns. In 13(a) the adverb occurs sentence-finally. 13(b) and 13(c) represent two possible derivations for 13(a), as a result of the application of VP-shell movement past the manner adverb (see Cinque 1999) or via right extraposition of the adverb, respectively.

In 13(b) the sentence final position of the manner adverb is derived by means of leftward movement of the first VP-shell c-commanded by the adverb at its right (the trace ti marks this position) along the lines proposed by Cinque (1999) . Let us call this approach the VP-shell movement analysis. On the other hand, an analysis that allows for right-extraposition would deliver the representation in (c) where the adverb is extraposed and right - adjoined at the level of the relevant VP-shell or directly to VP.

Let us evaluate the extraposition analysis first. It presents problems both at the conceptual and at the empirical level. As already said, if we are to assume an analysis like 13(c) we will have to allow both for adverb movement and for right adjunction. Beyond the fact that the elimination of right adjunction and of adjunction in general simplifies syntactic theory to a great extent without limiting her explanatory power (see Kayne 1994), it remains that the general assumption that adverbs do not move represents a crucial fact for us, since it will enable us to test the syntactic position of DPs and other syntactic material accurately.

At the empirical level we find consistent data showing that the postulation of an extraposition rule makes the wrong predictions with respect to the relative order of adverbs. Consider for instance 14(a)-(b); if we were to follow the extraposition analysis we would be forced to assign 14(a)-(b) the derivation illustrated in 14(a') and 14(b'), respectively:

In a theory of adverbs allowing for right adjunction (see Ernst 1998, forthcoming) adverbs can be optionally right-adjoined under the condition that right-adjunction does not change the hierarchical level at which the same elements left-adjoin. This is required in order for adverbs to mantain their scope properties unaltered in the course of the derivation.

A second general requirement is that, if right-adjunction is a movement operation reordering adverbial constituents in the syntax, this as other movement operation will have to respect the cycle.

The problem with 14(a)-(b) is then clear: we expect, contrary to facts, 14(a) not to be a possible derivation from a base generated order of the kind illustrated in (9). Under the extraposition analysis, if both adverbs are extraposed we expect the manner adverb to be more embedded than the time adverb (which is attached higher). Note, moreover, that under the same analysis the ungrammatical 14(b) is predicted to be acceptable.

Given the shortcomings of a derivation allowing for extraposition, let us turn to the VP-shell movement analysis. As already said, in this approach the occurrence of the manner adverb at the right periphery of the sentence is not seen as an instance of movement of the adverb itself, but as a consequence of movement of part of the VP structure past the adverb to land in a position at its left.

For the moment we will not make any assumptions as to the nature of the landing site of this movement: for the purposes of the present discussion it will be sufficient to assume that the landing site is the specifier position of a functional projection aP, without any further specification for a, other than AGRS₀ < a < V₀. A movement rule in which portions of the VP are allowed to move derives 13(a) straightforwardly, as 13(b) shows⁵.

Summarizing our findings we can see that movement reordering of constituents base generated in the VP must involve movement of portion of the VP (VP-shells).

At this point the natural question would be whether it is possible to analyse movement involving the verbal head as an instance of VP-shell movement. In other words whether it is possible to dispense with head-movement entirely, at least for the case of the past participle. The answer is negative. Consider (15):15) a) I ragazzi hanno ripetuto tutti la lezione attentamente [The boys have repeated all the lesson carefully] 'The boys have all repeated the lesson carefully' b) I ragazzi hanno risposto tutti alla maestra diligentemente [The boys have answered all to the teacher diligently] 'The boys have all answered to the teacher diligently'What we called the VP-shell movement analysis, taken alone, is not adequate as a possible approach to 15(a) or 15(b). Let us take for example 15(a) (the analysis of 15(b) goes along the same argument) and try to derive its linear order as a result of the application of the VP-shell movement rule. As always, the underlying word order will be the one in (9). The output of the application of VP-shell movement will be (16) ⁶:

First of all we observe that, in this case, VP-shell movement contributes no formal advantage with respect to head movement, since it is vacuous.

Secondly, and more importantly, if Cinque (1999) is correct in claiming that VP-shell movement has effects on the PF level in that it triggers focus of the stranded part of the VP, then an analysis for 15(a) on the lines of (16) makes the wrong predictions, since (17) is not acceptable, since tutti la lezione attentamente cannot be focused.

Finally, it is clear that the VP-analysis alone cannot do without independent head movement of the verbal head anyway, since the past participle has to move at least inside the VP-shell as (16) shows.

As a consequence, for cases like (15) we will have to assume the availability of past participle head movement independently of VP-shell movement. The overall picture will be that in Italian the verbal head can undergo at least two independent types of movement, head movement and movement along other material in the VP-shell. As already said it seems that Italian must allow for both of the two possibilities independently and that none of the two can be reduced to the other.

4 Argumental XP-movement ?

The aim of the discussion in this section is twofold: from the one side I would like to show that the commonly shared view that the order of the verbal arguments in the Italian VP is free is not completely correct and that this is an apparent effect due to independent factors. Secondly, I would like to focus on movement phenomena involving argumental XPs.

Let us start by consider again 15 (a) and (b), here repeated as 19 (a)-(b).

In both 19(a) and 19(b) the past participle has moved past the position where the subject is base generated. Additionally an object DP, a DO and an IO respectively, moved from their base positions to land in a position located higher than the manner adverb proper, as the position of diligentemente witnesses. In both cases in (19) we are forced to posit the existence of at least two different positions for objects, i.e. that the position they surface in is different from the position where they are generated, in other words that some kind of movement operation must have taken place ⁷.

This fact would not be so clear if we allowed adverbs to move, and, in particular, right-extrapose. However, we already investigated the question whether manner adverbs can be right-extraposed and found quite convincing evidence of the contrary. So, building up our discussion from those observations it is clear that no kind of extraposition for the adverb diligentemente can come into play here in order to mantain the hypothesis that the IO and DO objects are occupying their base positions in (19). In this case the representations for 19(a)-(b) will be 20 (a)-(b):

Now consider (20) in the light of the structure in (2). If our premises are correct, then 20 (a)-(b) are telling us something new with respect to the structure we assumed in (2). What is new is namely the existence of a "syntactic space" between the VP-internal subject position and the manner adverb proper.

Given this evidence, the next step will be to check what kind of syntactic material can occurr in it and if there are constraint with respect to order, for instance.

In (20) we see that not only IO and DO can appear in this position independently but they can also cooccur:

However the relative order DO-IO cannot be inverted, as the marginality of (22) shows:

The fact that (22) is marginal and not completely ungrammatical could depend on the possibility of 'deaccenting' diligentemente (and in general almost any adverb) when it occurs in final position (cf. Cinque 1997, 1999 and Pili 1999). These readings are not relevant here. A means of avoiding deaccenting effects could be to check constituent order in a moved VP-shell. In this case, once we filter out undesired effects the contrast between DO-IO and IO-DO orders becomes sharper, as 23(a)-(b) witness:

Another question is whether other constituents, for example non-arguments, can appear in this part of the sentence. In 24(a) and (b) we see that this is not the case:

Summarizing, it seems that the syntactic space comprehended between the position of the VP-subject and the manner adverb proper is a space that can contain only argumental projections and in a constrained order in that the direct object has to precede the indirect object. We arrive to a VP representation for Italian structured as the following one:

5 Predictions on the interpretation of indefinite objects

Indefinite noun phrases as attested by a by now very rich literature (Kamp 1981, Heim 1982, Kratzer 1989, Diesing 1992, de Hoop 1992, Chierchia 1995 and many others) are characterized by the possibility of allowing more than one reading: in particular, they can be assigned (at least) a specific, a generic or an existential reading. We will leave the generic readings aside for the moment and focus on the alternation between specific and existential readings.

In the sentence Mario greeted a woman the specific reading for the indefinite object would be paraphrased as a woman chosen from a set which is already given in the context. Under an existential reading the same DP would be paraphrased as a woman (or other) where 'the woman' is not present in the representation of our discourse, and introduces new information modifying the context by means of a new index (cf. Enç 1991 among others).

Studies that tried to relate the syntactic occurrence of indefinites with their interpretation showed that the object position, and in particular the direct object position seems to be linked to the assignment of an existential interpretation in a large number of languages.

What we will try to do here is to check first of all if this holds for Italian and secondly how the assignment of such readings could be constrained syntactically.

The first thing we note is that in Italian the direct object can often be assigned both readings (similar phenomena in Dutch led de Hoop 1992 to minimize the role of syntactic constraints for the interpretation of this kind of nominals). Consider the following data:

Now consider a case of object movement, as 29(a), whose derivation is given in 29(b). As we saw, given that the order in 29(b) is derivable only by application of VP-shell movement, the stranded part including the floating quantifier and the moved object must be in focus:

In 29(a), as illustrated, the existential reading for the direct object disappears. Why should it be so? Let us try to defend the hypothesis that the specific reading of the indefinite depends crucially on the fact that the object has been moved from its base position to the specifier of VP5. (cf. 25).

Up to this point we did not discuss the nature of the projections I named VP4 and VP5 but we could assume that they are comparable to agreement projections.

If this is the case, then Italian behaves analogously to many other languages with respect to the assignment of specific readings for indefinites, i.e. even in Italian specificity phenomena are linked to movement of the relevant DP from its base position to an agreement projection at its left (cf. German and Catalan scrambling data in Meinunger 1996 as well as scrambling Dutch data in de Hoop 1992).

For space reasons I will not introduce a detailed comparison with other languages that seem to behave similarly with respect to this phenomenon but before I end I would like to show that a structural constraint seems to be responsible for the two following facts: 1) in (29) we have no ambiguity with respect to interpretation: the indefinite cannot be read existentially, and 2) sentences like (28) are ambiguous.

Consider again the derivation in 29(b). Moreover, assume that existential readings for indefinites objects are assigned to them in their base position. If VP-shell movement is a kind of movement that is related to the PF level, then it is quite improbable that it can be reconstructed, reconstruction being a LF operation. If reconstruction of the VP-shell does not take place, then the indefinite sitting in the specifier of VP5 will not be in a configuration in which it can c-command the trace in the object base position. Hence it will not be possible to reconstruct the object movement as well, and the existential interpretation for this object will be lost, so to say. Consider now the case of (28). Two of the possible representations this sentence can be given are illustrated in 30(a)-(b):

As (30) makes clear, the ambiguity in the interpretation of the indefinite stems from a structural ambiguity, i.e. from the fact that (30) can be derived, at least, in two different ways. The case of 30(a) is analogous to (29), since here too VP-shell movement blocks reconstruction and consequently the recovery of the existential reading. In the case of 30(b) only the existential reading comes into play, since the object is not moved.

6 Conclusions

In the final section we show that the availability of landing sites for movement inside the VP predicts restrictions with respect to the readings assigned to indefinite DPs along the lines of an analysis allowing for movement of VP-shells, as suggested in Cinque (1997, 1999).

Notes

References

Belletti, Adriana (1990): Generalized Verb Movement. Turin, Rosenberg and Sellier.

Belletti, Adriana / Rizzi, Luigi (1981): "The syntaxy of 'ne'. Some theoretical implications." In: The Lingustic Review 1: .

Belletti, Adriana / Shlonsky, Ur (1995): "The Order of Verbal Complements in Italian and in Hebrew." In: Natural Language and Linguistic Theory 13: .

Bobaljik, Jonathan / Jonas, Dianne (1996) "Subject Positions and the Role of TP." In: Linguistic Inquiry 27: .

Cinque, Guglielmo (1993): "A Null Theory of Phrase Structure and Compound Stress." In: Linguistic Inquiry 24: .

Cinque, Guglielmo (1997): Adverbs and the Universal Hierarchy of Functional Projections., Ms. University of Venice. Published (1999), Oxford University Press.

Cinque, Guglielmo (1999): Adverbs and functional heads: a cross-linguistic perspective. Oxford Universtiy Press.

Chierchia, Gennaro (1992) "Anaphora and Dynamic Binding." In: Linguistics and Philosophy 15: .

Chomsky, Noam (1995): The Minimalist Program. MIT Press, Cambridge, Mass.

Chomsky, Noam (1998): Minimalist Inquires. Ms. MIT

Collins, Chris / Thráisson, Höskuldur (1996): "VP-internal Structure and Object Shift in Icelandic." In: Linguistic Inquiry 27:

De Hoop, Helen (1992): Case Configuration and Noun Phrase Interpretation. Doctoral dissertation: University of Groningen.

Diesing, Molly (1992): Indefinites. MIT Press, Cambridge Mass.

Enç, Mürvet (1991): "The Semantics of Specificity." In: Linguistic Inquiry 22: 1-25.

Ernst, Thomas (1998): Talk delivered at the DGfS conference. Halle/Saale (Germany).

Ernst, Thomas (forthcoming): The Syntax of Adjuncts. Cambridge Univ. Press.

Frey, Werner / Pittner, Karin (1998): "Zur Positionierung der Adverbiale im deutschen Mittelfeld." In: Linguistische Berichte 176: .

Haider, Hubert (1998): Adverbs and the Syntax-Semantics Interface. Ms. University of Salzburg.

Heim, Irene (1982): The Semantics of Definite and Indefinite Noun Phrases. Doctoral dissertation. UMass., Amherst.

Kamp, Hans (1981): "A Theory of Ttruth and Semantic Representation." In: Groenedijk, J.A.G. / Janssen, T.M.V. / Stokhof, M.B.J (eds.): Formal methods in the study of language. Mathematical Centre Tracts 135, University of Amsterdam. Reprinted in 'Truth, information and interpretation'. GRASS 2. Dordrecht.

Kayne, Richard (1994): The Antisymmetry of Syntax. MIT Press, Cambridge, Mass.

Kratzer, Angelika (1989): "Stage-Level and Individual-Level Predicates." In: Papers on Quantification, NSF Grant Report, Dept. Of Linguistics, UMass, Amherst.

Kratzer, Angelika (1994): The Event Argument and the Semantics of Voice. Ms. U.Mass.

Laenzlinger, Christopher (1998): Comparative Studies on Word Order Variation. Amsterdam/Philadelphia.

Larson, Richard (1988): "On the Double Object Construction." In: Linguistic Inquiry 19: .

Lonzi, Lidia (1991): "Il Sintagma Avverbiale." In: Renzi, Lorenzo & G. Salvi (eds.) (1991): Grande Grammatica Italiana di Consultazione II: .

Meinunger, Andre (1996): Discourse dependent DP (De-) Placement. Doctoral dissertation. University of Potsdam.

Pili, Diana (1999): Deaccented Constituents and the Structure of the Left Periphery. Ms. Università di Verona.

Rizzi, Luigi (1982): Issues in Italian Syntax. Dordrecht.

Sportiche, Dominique (1988): "A Theory of Floated Quantifiers and its Corollaries for Constituent Structure." In: Linguistic Inquiry. 19: .