Restructuring Verbs in Italian HPSG Grammar

Restructuring Verbs in Italian HPSG Grammar Paola Monachesi  Tilburg University 1 Introduction Restructuring verbs (Rizzi 1982) exhibit certain prop...
Author: Mary Mills
1 downloads 1 Views 99KB Size
Restructuring Verbs in Italian HPSG Grammar Paola Monachesi  Tilburg University

1 Introduction Restructuring verbs (Rizzi 1982) exhibit certain properties: they trigger clitic climbing, long NP-Movement and auxiliary change. The aim of this paper is to propose a mechanism which is able to give a unitary account of this class of verbs. The analysis will focus mainly on the first two properties and will be carried out within the framework of Head-Driven Phrase Structure Grammar (HPSG) (Pollard & Sag 1987, Pollard & Sag 1993). Given the nonlocal character of clitic climbing and long NP-Movement, it seems appropriate to propose an approach in terms of nonlocal features and the Nonlocal Feature Principle, which is the mechanism used within HPSG to deal with long distance dependencies. It will be shown that this mechanism doesn’t account for clitic climbing in an adequate way; it doesn’t naturally capture the fact that only a specific class of verbs trigger it. Furthermore it doesn’t account for long NP-Movement in an elegant way. An alternative analysis in terms of argument composition will be adopted; it is based on the idea that the arguments of a verb governed by a restructuring verb can be raised to become arguments of the governor. The effect achieved is similar to a lexical analysis of functional composition as developed within the tradition of Categorial Grammar. This approach is able to give a unified account of clitic climbing and long NP-Movement and it can be easily extended to handle similar phenomena in other languages. The structure of the paper is as follows. Section 2 introduces the basic facts about restructuring verbs in Italian. Section 3 gives an outline of the middle si construction in Italian since it interacts in a crucial way with restructuring verbs. Section 4 shows that a treatment in terms of nonlocal features and the Nonlocal  I would like to thank Ivan Sag, Anne Abeill´e and Andreas Kathol for comments and suggestions. This material was also presented at Tilburg University, Birmingham University and at Edinburgh University. I would like to thank the audiences for their comments. I would also like to thank Jerry Sadock for presenting the paper at the conference. Author’s address: Tilburg University - ITK, Postbus 90153, 5000 LE Tilburg, Nederland. E-mail: [email protected]

Feature Principle does not provide a satisfactory solution. In section 5, an analysis in terms of argument composition will be proposed. It will be shown that the mechanism is adequate to handle the relevant data in an elegant way. Section 6 contains some concluding remarks.

2 Data There are three classes of verbs which trigger clitic climbing, long NP-Movement and auxiliary change: (1)



modal verbs (e.g. potere, dovere, volere, sapere)

 temporal aspectual verbs (e.g. cominciare, finire, continuare)  pure motion verbs (e.g. venire, andare, tornare) The verbs in (1) will be referred to as restructuring verbs, after Rizzi(1982) who proposed a restructuring rule in order to account for the phenomena triggered by them. Restructuring verbs trigger clitic climbing, namely a clitic which originates as a dependent of a complement verb can climb and attach to the restructuring verb: (2) Martina lo vuole leggere Martina cl.(acc) wants to read ‘Martina wants to read it’ Clitic climbing is optional with these trigger verbs; the following sentence where the clitic stays within the lower clause is also grammatical: (3) Martina vuole leggerlo Martina wants to read cl.(acc) ‘Martina wants to read it’ Restructuring verbs also have the property of triggering long NP-Movement. If the clitic si, which acts as a passivizing element, is attached to the restructuring verb, the direct object of the verb in the embedded clause can become the subject of the restructuring verb: (4) Queste case si vogliono vendere a caro prezzo These houses SI want to sell at high price ‘These houses are wanted to be sold at a high price’ In this example queste case originates as direct object of vendere and becomes the subject of si vogliono. It is possible to have several restructuring verbs embedded one beneath the other, in this case the clitic can climb more than one clause:1 1

The clitic can also attach to the lower verb or to the intervening verb.

(5) Martina lo vuole poter leggere Martina cl.(acc) wants can read ‘Martina wants to be able to read it’ A similar situation can be found in the long NP-Movement configuration, if more than one restructuring verb is present: (6) Queste case si vogliono poter vendere a caro prezzo These houses SI want can sell at high price ‘These houses are wanted to be able to be sold at a high price’ In this case queste case, which is the object of vendere, becomes the subject of si vogliono. Clitic climbing in Italian has an interesting property. If there are two clitics which semantically belong to the same verb, they cannot be separated; they must climb together. Therefore the following sentence where the clitics are split is not grammatical: (7)

* Piero mi voleva spedirlo Piero cl.(dat) wanted to send cl.(acc) ‘Piero wanted to send it to me’

Similarly, if long NP-Movement interacts with clitic climbing, it is not possible to have a sentence where the clitic has not climbed: (8)

* Queste case si vogliono vendergli a caro prezzo These houses SI want to sell cl.(dat) at high price ‘These houses are wanted to be sold to him at a high price’

In this example both queste case and gli are complements of vendere, it is not possible to have a structure where the NP is preposed while the clitic stays behind. Only restructuring verbs can trigger clitic climbing and long NP-Movement; therefore if a verb that does not belong to the restructuring class is present, the clitic cannot climb: (9)

* Martina lo promette di leggere Martina cl.(acc) promises to read ‘Martina promises to read it’

The same holds for long NP-Movement; the object cannot become the subject of the main verb if this does not belong to the restructuring class: (10)

* Le nuove case popolari si sono promesse di costruire The new houses council SI are promised to build ‘The new council houses are promised to be built’

3 Middle si construction in Italian Italian forms passive sentences in a way parallel to English: (11) Quei libri sono stati dimenticati (dagli studenti) Those books are forgotten (by the students) In this construction quei libri which is the object of dimenticati becomes the subject, while the subject is replaced by the optional by-phrase. The lexical rule for passive proposed for English by Pollard & Sag(1987) can account for Italian passive as well.2 The lexical rule operates on the subcategorization requirements of transitive verbs, making the object become the subject while the subject becomes the optional by-phrase. The rule makes no reference to the notion of grammatical function since this is a notion which is implicit in the order elements have within the subcategorization list.3 In Italian there is another type of construction which has in common with passive the property that the direct object functions as the superficial subject. This is the middle si construction: (12) Quei libri si dimenticano facilmente Those books SI forget easily ‘Those books are easy to forget’ In sentence (12) the clitic si acts as a passivizing element and the sentence receives a middle passive interpretation.4 This construction, which is fully productive, occurs with transitive verbs which select a human subject. In middle si constructions the direct object functions as a subject; in particular it agrees with the verb and it can be replaced by a null subject.5 A difference between a middle si construction and a passive construction is that the former doesn’t admit a “by-phrase”; in this way it is similar to agentless passives. A lexical rule to account for middle si constructions (SI-LR) can be proposed. The rule, which is similar to the lexical rule for passive, creates middle 2

See also Bresnan(1982) for a similar treatment of passive within the LFG framework and for a justification of a lexical analysis of passive. 3 In Pollard & Sag(1987) there is a SUBCAT list which encodes the subcategorization requirements of the sign; the order of elements there corresponds to the traditional grammatical notion of obliqueness of grammatical relations. The analysis proposed below is based on the framework developed in Chapter 9 of Pollard & Sag(1993) which incorporates innovations due to Borsley (Borsley 1987, Borsley 1989). The analysis makes use of valence features (SUBJ, COMPS, SPEC) which encode the subcategorization requirements of the sign. In this framework partial reference is made to grammatical functions. 4 Clitic si can have different interpretations in Italian, there is an impersonal si, a reflexive si, an inherent-reflexive si and an ergative si. See Monachesi(1993a) for an analysis of these other types of si within the HPSG framework. 5 See Manzini(1983) for a more detailed discussion about the subject status of the element in preverbal position. See also Aissen & Perlmutter(1983) for Spanish.

verb forms.6 It operates on the subcategorization requirements of transitive verbs, removing the subject slot and making the object become the subject. It should look roughly as follows:7 (13) Lexical rule for middle verb forms (SI-LR) 2

2

HEAD V

6 6 6 6 i 6 CAT 6 SUBJh 1 6 4 6 COMPSh j 3 i 6 2 [3rd] 6 " # 6 6 4 CONTENT AGENT 1

NP NP

THEME 2

2

3 3 7 7 7 5

7 7 7 7 7 7 7 7 7 5

2

HEAD V 6 SUBJh 6

3 3

NP [nom]

=)

7 6 i 7 6 7 7 2 7 7 6 CAT 6 6 6 7 7 6 4 COMPSh 3 i 5 7 7 6 7 6 CLTS f g 7 6 " # 7 6 7 6 5 4 CONTENT AGENT 1

si

THEME 2

In middle si constructions the object can be preposed and become the superficial subject only if it is in the third person, therefore the restriction imposed on the NP will rule out all those sentences where the superficial subject is not third person, as in the following example: (14)

* Io si guardo I SI look ‘I am looked at’

Furthermore the subject in middle si constructions is assigned nominative case since sentences like the following where the subject bears accusative case are not grammatical: (15)

* Le si mangiano cl.(acc) SI eat ‘Them are eaten’

The presence of the clitic si attached to the verb is triggered by the feature CLTS, in a way that will be explained in the discussion further below on clitics. Subject-verb agreement can be handled as in Pollard & Sag(1993), namely the verb assigns the required value to the index associated with its subject. Role assignment proceeds also as in Pollard & Sag(1993). In HPSG roles are assigned within the lexical entry; a role is assigned by means of structure sharing between the index of an element in the SUBJ/COMPS list and the value of some attribute of the verb’s CONTENT value. Role assignment, as well as case assignment, occurs within the lexicon. In this specific case, the superficial subject is coindexed with the THEME role, while the AGENT role should be filled by an external agent which is understood as human and doesn’t surface in Italian. 6

See also Grimshaw(1982) for an analysis in terms of lexical rule of the French se moyen within the LFG framework. 7 Details which are not relevant for the discussion are omitted for reasons of simplicity in this and in all the other lexical rules which will be presented in the paper.

4 An analysis in terms of nonlocal features Restructuring verbs trigger both clitic climbing and long NP-Movement; data presented in section 2 have shown that the two constructions are subject to similar constraints. It is therefore desirable to have an analysis which is able to account for the two phenomena in a uniform way. Given the nonlocal character of these constructions, it seems appropriate to propose an analysis based on the use of nonlocal features and the Nonlocal Feature Principle, which is the mechanism discussed in Pollard & Sag(1993) to deal with Unbounded Dependency Constructions. An analysis of clitic climbing which relies on this mechanism can be formulated.8 The treatment makes use of lexical rules to update the subcategorization requirements of the verbal head and of nonlocal features to encode the information that a clitic can appear at some point in the tree. Within this analysis, a sentence like (2), repeated here: (2) Martina lo vuole leggere Martina cl.(acc) wants to read ‘Martina wants to read it’ will have the following representation: (16) Example of derivation

HHH   HH 

VP [TO-BINDjOC fg, INHjOCfg] V [TO-BINDjOC f 1 g]

VP[INHjOC f 1 g] V[INHjOC f 1 g]

lo vuole

leggere

In this case, leggere undergoes a lexical rule that changes its subcategorization requirements. Since in Italian clitics and full complements are in complementary distribution, the slot related to the relevant complement is removed from the COMPS list of the verb, while a nonlocal feature OC (object clitic) is added: it encodes the agreement information relative to the clitic. The Nonlocal Feature Principle (Pollard & Sag 1993) guarantees the percolation of the feature and ensures that the percolation will stop when a TO-BINDjOC feature (which is assigned to the cliticized verb form) is found with the same value of the INHERjOC one. A problem with this mechanism is that it overgenerates, allowing clitic climbing also with verbs that do not trigger it; therefore a sentence like *Martina 8

See Monachesi(1992) for a more detailed description of the analysis and Miller(1992b) for a similar analysis of related facts in French within a GPSG/HPSG framework.

lo promette di leggere ((9) above) will be accepted. The approach doesn’t account for the general fact that only restructuring verbs can trigger clitic climbing. This is because Unbounded Dependency Constructions are subject to different locality constraints than clitic climbing constructions. In order to rule out sentences like (9), it is necessary to add specific constraints on the path of the OC feature. The percolation of the feature must be stopped if there is a verb that doesn’t trigger clitic climbing and must be allowed if there is a verb that triggers it. It seems therefore that this approach doesn’t naturally capture the fact that clitic climbing is triggered only by a specific class of verbs since constraints need to be imposed to obtain this result. Furthermore, the constraints need to be imposed both on verbs that don’t trigger clitic climbing and on those that trigger it; this is clearly not very economical. Some generalizations are missed if the analysis is extended to account for long NP-Movement as in sentence (4), repeated here: (4) Queste case si vogliono vendere a caro prezzo These houses SI want to sell at high price ‘These houses are wanted to be sold at a high price’ In this construction the direct object of the verb in the embedded clause is preposed to become the subject of the restructuring verb, provided the clitic si is present, acting as a passivizing element. Therefore, sentence (4) should be analysed by means of an interaction between the rule proposed for middle si constructions and the mechanism based on nonlocal features presented above. There are some problems with this as SI-LR operates on the subcategorization requirements of the verb, while in this case the relevant information is encoded in a nonlocal feature. One could imagine a specific rule for middle si constructions that applies on restructuring verbs which operates on information encoded in a nonlocal feature. The rule should change the grammatical function of the element present in the nonlocal feature. This would imply the introduction of a different way to encode information about grammatical functions of elements, since in this case it is not possible to rely on the order of elements within the COMPS list. In this way there would be two lexical rules which handle middle si constructions: one applies to restructuring verbs as in (4) while the other applies to the verbs that do not belong to this class, as in (12). The same phenomenon is analysed in two different ways. It is evident that an analysis in terms of nonlocal features doesn’t provide an elegant account of long NP-Movement, since some relevant generalizations are missed. In the next section it will be shown that an alternative analysis in terms of argument composition can provide an adequate solution to the problems presented above.

5 An analysis in terms of argument composition The idea that underlies the approach was first proposed by Hinrichs & Nakazawa(1990) for the treatment of the German verb cluster and can be easily extended to the analysis of restructuring verbs. In this way it is possible to account for cases of clitic climbing and long NP-Movement, producing an analysis which captures intuitions similar to the ones underlying Rizzi’s restructuring rule (Rizzi 1982) and the clause reduction analysis of Aissen & Perlmutter(1983). Within the argument composition analysis, the subcategorization requirements of a verbal complement which is governed by a restructuring verb are passed up to the governor. It will be shown that this approach can adequately handle the relevant data concerning clitic climbing and can account for the fact that only certain verbs trigger it. Furthermore, it can account for long NP-Movement in an elegant way since there is a natural interaction between the argument composition mechanism and the lexical rule that handles middle si constructions. The analysis can also account for other cases of clitic climbing both in Italian (Monachesi 1993b) and in French (Miller & Sag 1993).

5.1 Clitic climbing The analysis of clitic climbing is based on the assumption that Italian clitics behave essentially as inflectional affixes. There are arguments similar to those proposed by Miller(1992b) for French showing that Italian clitics exhibit many properties suggesting that they are inflectional affixes, as discussed in Monachesi(1992). Italian clitics exhibit a high degree of selection with regard to their host; namely they can attach only to verbs, they present arbitrary gaps since not all combinations of clitics are possible and they show morphophonological idiosyncrasies; thus they behave like affixes according to the criteria proposed by Zwicky & Pullum(1983). Furthermore, they don’t have wide scope over a conjunction of hosts; this is evidence of their affixal status according to the coordination criteria proposed by Miller(1992a). Italian clitics are also rigidly ordered in a flat clitic cluster, being in this way similar to certain morphological affixes. An account in terms of template morphology as in Simpson & Withgott(1986) could account for the restrictions in combination and the rigid ordering. As previously mentioned, Italian clitics are in complementary distribution with full complements; a lexical rule, the Complement Cliticization Lexical Rule (CLT-LR), can be used to account for this fact. Given a verb that subcategorizes for a certain argument, the argument is removed from the COMPS list and the corresponding item is added as a value of the CLTS feature, as proposed by Miller & Sag(1993) for French. It should look roughly as follows:

(17) Complement Cliticization Lexical Rule (CLT-LR) 2

HEAD V 6 4 COMPSh CLTS W

:::; 1 ; ::: i 75

3

2

3

=)

HEAD V 6 4 COMPSh CLTS W

::: i

[ f 1 g

7 5

The CLTS feature acts as an interface to morphology and triggers the construction of the appropriate cliticized verb form. The interaction of the Complement Cliticization Lexical Rule presented above and the Argument Composition Lexical Rule (COMP-LR) will account for clitic climbing. The rule should look as follows:9 (18) Argument Composition Lexical Rule (COMP-LR) 2 6 6 6 6 4

2 6 6 6 6 6 6 4

HEAD V VCLASS modal _ aspectual _ motion SUBJ h NP i   SUBJ h NP i i COMPSh V P COMPS h i

+ HEAD V VCLASS modal _ aspectual _ motion SUBJ h NP i 3 2 CLTS f g COMPS L  h V 4 SUBJ h NP i 5 i COMPS L

3 7 7 7 7 5

3 7 7 7 7 7 7 5

In this way, restructuring verbs subcategorize generally for a saturated VP, but COMP-LR will allow those verbs to alternatively subcategorize for a verbal complement and for the arguments of that verb. This means that the subcategorization requirements of the complement are passed up to the restructuring verb.The result will be a flat structure where the restructuring verb, the verbal complement and its arguments form a constituent.10 If clitic climbing applies, the verbal complement and possible phrasal complements don’t form a constituent as argued in Rizzi(1982); therefore the possibility of having a partial VP should be excluded.11 It should be noted that the raised complements of the lower verb precede the verbal 9

Here L stands for the list of complements subcategorized by the lower verb. See also Abeille’ & Godard(1993) where arguments are given in favour of a flat structure for French tense auxiliaries which also trigger clitic climbing. 11 Given the lexical rule, two structures will be created if the subcategorization slots are filled by phrasal complements instead of clitics: a flat one and one where the lower verb and its complements form a constituent. Skytte, Salvi & Manzini(1991) observe that there are no semantic differences between the two structures, but regional differences in their use. In Central Italy both structures are present, in Southern dialects only the former is present while in Northern dialects only the latter is present. 10

complement; this is because the position in the COMPS list is closely related to grammatical functions. In this way the first complement in the COMPS list will be identified as the direct object. This information will be relevant for the treatment of long NP-Movement. If clitic climbing occurs as in (2), repeated here: (2) Martina lo vuole leggere Martina cl.(acc) wants to read ‘Martina wants to read it’ the Argument Composition Lexical Rule will account for it. Given the basic lexical entry for vuole, which has the following COMPS list: (19) Example of the complement list of vuole

V [COMPS h V P [COMPS h i] i] COMP-LR will apply to it and produce an alternative entry where the restructuring verb will subcategorize for the verbal complement and for the argument of the verbal complement: (20) Example of the complement list of vuole after the application of COMP-LR

V [COMPS h 1 NP [acc]; V [COMPS h

1

i] i]

This output feeds the Complement Cliticization Lexical Rule that will remove the complement and will add it as a value of the CLTS feature, as it is shown in the following example: (21) Example of the complement list of lo vuole after CLT-LR has applied

V [COMPSh V [COMPSh

1

i] i ; CLTS f 1 NP [acc] 3Msg g] [

]

The cliticized verb form will eventually be produced. As for semantic role assignment, it proceeds in a very straightforward way: if a clitic is present, it will be assigned semantic role and case through the link created via structure sharing. Cases where the clitic doesn’t climb as sentence (3), repeated here: (3) Martina vuole leggerlo Martina wants to read cl.(acc) ‘Martina wants to read it’ are analysed by an application of CLT-LR to the lower verb. The slot related to the full complement will be checked off, while the cliticized verb form will be triggered through the presence of the feature CLTS. In this case the restructuring verb subcategorizes for a saturated VP. If there is more than one restructuring verb, one can get sequences where the clitic is attached to the higher verb as is shown in (5), repeated here:

(5) Martina lo vuole poter leggere Martina cl.(acc) wants can read ‘Martina wants to be able to read it’ The application of the Argument Composition Lexical Rule is therefore triggered twice; in this way the subcategorization requirements of the lower verb are passed up to the highest one. The Complement Cliticization Lexical Rule will be responsible for the cliticized verb form. In Italian, if there are two clitics that originate as complements of the same verb, they must cliticize together; they cannot be split as shown in (7), repeated here: (7)

* Piero mi voleva spedirlo Piero cl.(dat) wanted to send cl.(acc) ‘Piero wanted to send it to me’

Since restructuring verbs subcategorize for a verbal complement with an empty CLTS feature, sentences like (7) will be ruled out. The restriction on the verbal complement doesn’t allow argument composition if the verb has already combined with a clitic, namely if CLT-LR has applied. In the specific case of sentence (7), spedirlo will undergo the Complement Cliticization Lexical Rule and it will g. Since this doesn’t match the requirements be marked CLTS fNP[acc] 1 [3msg] imposed on the verb, no argument composition will take place and the sentence will be correctly ruled out. Sentences where more than one clitic is present as: (22) Piero me lo voleva spedire Piero cl.(dat) cl.(acc) wanted to send ‘Piero wanted to send it to me’ are simply analysed by the application of the Argument Composition Lexical Rule; in this way the verb voleva will subcategorize for the two phrasal complements. The application of the Complement Cliticization Lexical Rule will be triggered twice in order to remove the complements from the COMPS list and add them to the set value of the CLTS feature and produce eventually the cliticized verb form me lo voleva. An advantage of this approach is that only restructuring verbs will trigger argument composition, therefore no special locality constraints are necessary to rule out sentences like (9): (9)

* Martina lo promette di leggere Martina cl.(acc) promises to read ‘Martina promises to read it’

Since the main verb is not lexically marked as a clitic climbing trigger, it will not undergo COMP-LR which allows argument composition and the sentence will be correctly ruled out. The fact that clitic climbing occurs only with certain verbs can be naturally captured by the mechanism; the argument composition approach therefore allows a more elegant analysis of clitic climbing than an approach based on the use of nonlocal features.

5.2 Long NP-Movement A further argument to show that a treatment in terms of an argument composition based approach is superior to a nonlocal one can be found in the analysis of long NP-Movement. There is a natural interaction between the Argument Composition Lexical Rule and the Lexical Rule for middle si constructions and this interaction accounts in an elegant way for cases of long NP-Movement. It has been shown that COMP-LR makes dependents of the complement verb become dependents of the restructuring verb, while SI-LR makes the direct object of a verb become its subject. The interaction of the two lexical rules makes some predictions. COMPLR makes the object of a complement verb become the object of the restructuring verb; this creates the appropriate input for the application of SI-LR which acts on subcategorization requirements. The rule operates on the COMPS list and will make the object of the restructuring verb become the subject. This is the desired result. In a sentence like (4), repeated here: (4) Queste case si vogliono vendere a caro prezzo These houses SI want to sell at high price ‘These houses are wanted to be sold at a high price’ The Argument Composition Lexical Rule applies to vogliono so that queste case, which is the object of vendere, becomes the object of vogliono: (23) Example of the complement list of vogliono after the application of COMP-LR

V [COMPS h 1 NP; V [COMPS h

1

i] i]

The output of COMP-LR feeds SI-LR which makes queste case function as the subject of vogliono as it is shown in the following example: (24) Example of the complement list of si vogliono after the application of SI-LR

V [SUBJ h 1 NP i; COMPS h V [COMPS h

1

i] i; CLTSfsig]

In this case, queste case is a full phrase which fills the SUBJ slot of the verb; therefore the Valence Principle can normally apply and check off the SUBJ slot.12 Sentences where more than a restructuring verb is present such as (6): (6) Queste case si vogliono poter vendere a caro prezzo These houses SI want can sell at high price ‘These houses are wanted to be able to be sold at a high price’ are accounted for by a double application of the Argument Composition Lexical Rule which makes queste case the object of vogliono and then by an application of SI-LR.13 Restructuring verbs subcategorize for a verbal complement with an empty CLTS feature; in this way sentences like (7) *Piero mi voleva spedirlo are correctly ruled out, as has been shown. Sentences like (8), where there is an interaction among clitic climbing, long NP-Movement and middle si: (8)

* Queste case si vogliono vendergli a caro prezzo These houses SI want to sell cl.(dat) at high price ‘These houses are wanted to be sold to him at a high price’

are also correctly ruled out by the same restriction imposed on the verbal complement. The cliticized verb form vendergli doesn’t match the requirements on the verbal complement since the CLTS feature will not be empty and therefore no object preposing will be permitted. On the other hand, sentences like: (25) Queste case gli si vogliono vendere a caro prezzo These houses cl.(dat) SI want to sell at high price ‘These houses are wanted to be sold to him at a high price’ are correctly allowed. Long NP-Movement is triggered only by restructuring verbs; therefore sentences like (10), repeated here, are correctly ruled out: (10)

* Le nuove case popolari si sono promesse di costruire The new houses council SI are promised to build ‘The new council houses are promised to be built’

12 The Valence Principle is the principle of the grammar responsible for checking off subcategorization requirements that have been satisfied. 13 A sentence like:

* Si vogliono queste case poter vendere a caro prezzo SI want these houses can sell at high price ‘These houses are wanted to be able to be sold at a high price’ is not grammatical and it should be ruled out by a Linear Precedence rule that orders a verbal complement before other phrasal complements.

They are ruled out for the same reasons as sentences like *Martina lo promette di leggere ((9) above), namely because verbs which do not belong to the restructuring class will not trigger the application of the Argument Composition Lexical Rule. Since argument composition is not allowed, the application of the Lexical Rule for middle si construction will not be triggered and so there is no way that the initial object of costruire can become the subject of promettere. Sentences where a verb that doesn’t belong to the restructuring class intervenes between sequences of restructuring verbs are ruled out for the same reasons.

6 Conclusion An analysis of restructuring verbs within the HPSG framework has been proposed. It has been shown that an analysis in terms of nonlocal features and Nonlocal Feature Principle doesn’t provide an adequate solution for the treatment of clitic climbing and long NP-Movement. An alternative analysis, in terms of argument composition, has been proposed. It has been shown that the analysis can account for the basic facts related to clitic climbing. It can also be extended to account for other situations where clitic climbing occurs, both in Italian (Monachesi 1993b) and in French (Miller & Sag 1993). Furthermore, the analysis interacts in a natural way with the treatment proposed for middle si constructions, producing an elegant account of long NP-Movement.

References Abeille’, A. & D. Godard (1993): “The Syntactic Structure of French Auxiliaries." Paris. Ms. Aissen, J. & D. Perlmutter (1983): “Clause Reduction in Spanish." In: D. Perlmutter (ed.) Studies in Relational Grammar. Chicago University Press, Chicago Borsley, R. (1987): “Subjects and Complements in HPSG." CSLI, Stanford. CSLI Report 107 (1989): “An HPSG approach to Welsh." Journal of Linguistics 25:333–354 Bresnan, J. (1982): “The Passive in Lexical Theory." In: J. Bresnan (ed.) The Mental Representation of Grammatical Relations. MIT Press, Cambridge, MA Grimshaw, J. (1982): “On the Lexical representation of Romance Reflexive Clitics." In: J. Bresnan (ed.) The Mental Representation of Grammatical Relations. MIT Press, Cambridge, MA Hinrichs, E. & T. Nakazawa (1990): “Subcategorization And VP Structure in German." In: Hughes, Shaun & Salmons (eds.) Proceedings of the Third Symposium on Germanic Linguistics. Benjamins, Amsterdam

Manzini, M. R. (1983): “Restructuring and Reanalysis." MIT, PhD Thesis Miller, P. (1992a): “Postlexical Cliticization vs. Affixation: Coordination Criteria." In: C. Canakis, G. Chan & J. Denton (eds.) CLS 28: Papers from the 28th Regional Meeting. (1992b): Clitics and Constituents in Phrase Structure Grammar. Garland, New York Miller, P. & I. Sag (1993): “French Clitic Movement without Clitics or Movement." Paper presented at the LSA annual meeting., Los Angeles. (Ms. Stanford and Lille) Monachesi, P. (1992): “The use of nonlocal features in the analysis of Italian object clitics and clitic climbing." In: Proceedings of the third CLIN meeting, Tilburg. (To appear) (1993a): “On Si Constructions in Italian HPSG Grammar." In: Proceedings of the Tenth Meeting of the Eastern States Conference on Linguistics (ESCOL), Ohio State University, Columbus. (To appear ) (1993b): “Object clitics and clitic climbing in Italian HPSG grammar." In: Proceedings of the Sixth Conference of the European Chapter of the Association for Computational Linguistics. Utrecht Pollard, C. & I. Sag (1987): Information-Based Syntax and Semantics. Fundamentals. Vol.1. CSLI, Stanford (1993): Head-Driven Phrase Structure Grammar. CSLI and University of Chicago Press, Stanford. (To appear) Rizzi, L. (1982): Issues in Italian Syntax. Foris, Dordrecht Simpson, J. & M. Withgott (1986): “Pronominal Clitic Clusters and Templates." In: H. Borer (ed.) The Syntax of Pronominal Clitics. Syntax and Semantics. Vol.19. Academic Press, Orlando Skytte, G., G. Salvi & M. R. Manzini (1991): “Frasi subordinate all’ infinito." In: L. Renzi & G. Salvi (eds.) Grande Grammatica Italiana di Consultazione. Vol.2. Il Mulino, Bologna Zwicky, A. & G. Pullum (1983): “Cliticization vs. Inflection: English n’t." Language 59.3:502–513