The causee in the faire-Inf construction of European Portuguese ANABELA GONÇALVES

One of the most interesting aspects of the faire-Inf construction in the Romance languages concerns the properties and the structural position of the causee. In the relevant literature, the most consensual hypothesis is that, in these languages, the causee is the grammatical subject of the embedded domain, although it occurs in a post-verbal position and is introduced by a preposition whenever it is dependent on a transitive verb (see Kayne, 1975; Raposo, 1981; Burzio, 1986; Villalba, 1992; Guasti, 1993, 1997, among others). The aim of this paper is to present some evidence against the idea that the causee, in European Portuguese, is the subject of the embedded domain. I will claim that this domain is the projection of a null affix that incausativizes the embedded verb, suspending the assignment of the external -role of this verb. In consequence, the causee is merged in the positions classically associated to objects.


Introduction
Since Kayne's (1975) pioneering work on causatives, it has been demonstrated that Romance languages differ from English in the sense that only the former allow for the faire-Inf construction, with the general properties listed in (i)-(iii): (i) the causative verb and the embedded verb (Inf) occur in adjacency and the so-called subject of the infinitival domain (the causee) occupies the post-verbal position (see (1a)-(10a)); (ii) the causee surfaces as a DP or a PP, depending on the transitivity of the embedded verb; in the first case, it can be cliticized by an accusative clitic (see (1)-( 5)), in the second case, by a dative clitic (see ( 6)-( 10)); (iii) when cliticized, the causee occurs in adjacency to the causative verb; Clitic Climbing is a consequence of the complex predicate formation (see (1b-10b) vs. (1c-10c)).
Classical approaches to the faire-Inf construction tend to analyze the causee as the grammatical subject of the infinitival domain.The main goal of this paper is to present some arguments against this analysis, at least in what concerns EP.Assuming that the infinitival complement of the causative verb in the relevant construction is defective, in the sense that it lacks some of the functional categories associated to the sentence (see Gonçalves 1999a, 1999band Costa & Gonçalves 1999), I will claim that the above mentioned infinitival domain is the projection of a null affixcall it Incausthat incausativizes the embedded verb, suspending its external -role.In consequence, the causee is always merged in positions classically associated to objects.

Properties of the causee in EP
2.1.Classical approaches: the causee as the grammatical subject of the infinitival domain Kayne (1975), Burzio (1986), Villalba (1992) and Guasti (1997) The main empirical arguments supporting this analysis are listed in A-E bellow: A. The causee is semantically dependent on the embedded verb, that is, in (12), the causee is necessarily human, because the embedded verb ler (Eng: to read) requires an external argument of this nature.the teacher made read the book to-the dog B. In some Romance languages, such as Catalan and Italian, the causee binds anaphoric expressions in object positions, which means that it c-commands these positions, thus creating the appropriate binding configuration.
(14) a. La publicitat li i va fer desitjar de PRO i tenir un cotxe per sortir els caps de setmana. 2 the publicity CL-DAT make want of PRO have-INF a car to go.out the weekends (Villalba 1992: 353; (15a)) b.Ho fatto affermare di PRO i averla visto a Giovanni i .
D. In constructions expressing inalienable possession, the causee corresponds to the possessor and it binds the DP denoting the possessed entity (Kayne 1975;Villalba 1992).
the Maria CL make-FUT put the nose in the business (Villalba 1992: 350; (8a)) E. The causeebut not the higher subjectis the antecedent of reciprocals occurring in the infinitival complement (Kayne 1975).
he make-FUT talk these young girls the one-FEM of the other (id.: 252; ( 175)) The data presented so far apparently confirm the hypothesis that the causee is the subject of the infinitival complement.In fact, the examples (12)--( 16) show that this constituent (i) is thematically dependent on the embedded verb and (ii) binds anaphoric expressions (latu sensu) in embedded object positions.This means that the causee is merged in a position c-commanding the expressions it binds.

The causee as an object: the case of EP
The idea that the causee is the Subject of the infinitival domain, although apparently appropriate to some Romance languages, is not empirically motivated in EP.First, in this language, the causee cannot bind anaphoric expressions in object positions, as illustrated in ( 17 These data strongly suggest that in EP the causee is not the grammatical subject.In consequence, the hypothesis to be evaluated is the following: In EP, the causee is merged in a position lower than subject.That position is classically associated to the DO (if the embedded verb is intransitive) or to the IO (if the embedded verb is transitive).
Besides the data presented in ( 17)-( 18), other empirical arguments support this hypothesis.First, if the embedded verb is transitive, the causee cannot be associated to a floating quantifier: 3 (19) *Aos meninos, o professor mandou ler todos um livro.
To-the children, the teacher made read all a book Notice, however, that typical subjects, in non-causative contexts, can be associated to floating quantifiers: (20) Os professores deram todos um livro aos meninos.
the teachers gave all a book to-the children The contrast between ( 19) and ( 20) confirms the hypothesis that the causee is not a syntactic subject.Interestingly, the causee behaves like typical IOs, in the sense that neither of them can be associated to floating quantifiers.Compare, for instance, the examples in ( 19) and ( 21): (21) *Aos meninos, a Marta deu as bolachas todos.
To-the children, the Marta gave the cookies all The second property that distinguishes the causee from syntactic subjects is related to binding of possessive expressions.As it is well known, when the 3 See Alsina (1996), who presents identical data from Catalan.Notice, however, that, although subjects can be associated to floating quantifiers, it is not true that only subjects exhibit this property.In effect, the adequate generalization seems to be that only DPsnot PPscan be associated to floating quantifiers: (i) Os livros, o Pedro leu(-os) todos.the books, the Peter read(-CL-acc-3plmasc) all The argument is valid to prove that the causee is distinct from typical subjects.
DO integrates a distributive quantifier, it can bind possessive expressions occurring in the causee, as illustrated in ( 22 The contrast between ( 22) and ( 23) can be viewed as a consequence of the fact that the causee and the canonical syntactic subject are merged in different positions.To be more precise, the causee is more embedded than subjects, so, the object asymmetrically c-comands it.As for the IO of non-causative contexts, the example in ( 24) shows clearly that it can be bound by the DO, in the same way that the causee can: (24) O professor deu [cada livro] i ao seu i autor. 4he teacher gave each book to-the its author In sum, taking into consideration the data presented in this section, the hypothesis that the causee is not the subject of the infinitival domain seems to be adequate, at least for EP.Thus, we have to admit that, as far as this language is concerned, the causee is not merged in the embedded [Spec, VP] position.As a consequence of this, the causee does not c-command the DO: on the contrary, it is c-commanded by this constituent, which accounts for the behaviours I have just presented.

The null affix Incaus and the position of the causee
Since the data strongly suggest that in EP the causee is not the grammatical subject of the embedded domain, the main question is to know in what position the causee is merged.Before trying to answer this question, let us concentrate on the nature of the infinitival domain.As it has been defended in the literature, the faire-Inf construction instantiates the phenomenon of complex predicate formation.In fact, the two verbs seem to behave as a syntactic unit: embedded clitics appear on the causative verb (see ( 25)) and the object of the embedded verb can become the subject in passive SE-sentences (see (26)).5 (25) a. Os professors mandaram-no i ler [-] i aos alunos.

Made-3PL-SE build [new houses] to a architect italian
In previous work (Gonçalves 1999a), I have proposed that the phenomenon of complex predicate formation is closely related to the defectivity of the embedded infinitival domain.More precisely, I have argued that the infinitival complement is not sentential, in the sense that neither T nor AgrS project, in accordance to the principle of economy of representation (see Law, 1991;Chomsky, 1995;Grimshaw, 1997;Bošković 1997, a. o.). 6Consequently, there is only one functional domain for checking of verbal features, which explains the complex predicate effects. 7The empirical arguments in favour of sentential defectivity are synthetized on table ( 27 6 Elaborating on Law (1991), Bošković (1997: 25), proposes the following principle of economy (the Minimal Structure Principle: "Provided that lexical requirements of relevant elements are satisfied, if two representations have the same lexical structure and serve the same function, then the representation that has fewer projections is to be chosen as the syntactic representation serving that function.". 7For more details, see Gonçalves (1999a).
Since T and AgrS do not project, we could conclude that the infinitival complement in the faire-Inf construction is intrinsically verbal, that is, it is a VP, as suggested by Manzini (1983), Alsina (1992Alsina ( , 1996)), Guasti (1993aGuasti ( , 1997)), a. o..However, this hypothesis alone does not account for the DP/PP alternation of the causee.Let us assume, alternatively, the following hypothesis: In the faire-Inf construction of EP, the infinitival domain is the projection of a null affix -Incaus -, which is also empirically motivated in polysynthetic languages (see Baker 1988Baker , 1996)).
The null affix Incaus is the head where the embedded verb is spelled-out when the complex predicate is formed.This accounts for word order phenomena, namely for the position of VP adverbs like bem (Eng: well; see Costa, 1998): (28) a. O editor mandou rever bem o manuscrito ao escritor.
the editor made review well the manuscript to-the writer b. *O editor mandou bem rever o manuscrito ao escritor.
the editor made well review the manuscript to-the writer The central property of this affix is that it suspends the external -role of the embedded verb.This characterization of Incaus derives from the observation of two phenomena: first, as it is well known, the -role of the causee is not the one that the embedded verb assigns to its external argument in non--causative contexts, as it is illustrated in (29).
the Zé made eat the soup to-the João Second, Incaus is incompatible with other lexical morphemes that also trigger the incausativization of the embedded verb, such as the passive morpheme (cf.( 30)) or anticausative SE (cf.(31)): (30) *O professor mandou ser lido o livro pelos meninos.
the teacher made be read the book to-the children (31) *Os soldados mandaram dispersar-se a multidão.
the soldiers made disperse-SE the crowd Notice that the ungrammaticality of ( 30) and ( 31) is explained by Zubizarreta's (1985: 278) Principle of Morphological Nonredundancy, according to which attachment of redundant morphology is prohibited.In fact, the incausativization of the embedded verb is highly inspired on Zubizarreta's work.
However, my analysis is distinct from Zubizarreta (1985) in the following aspects: (i) no additional mechanisms other than checking of verbal features are needed to account for the complex predicate formation; (ii) the trigger of the incausativization is not the causative verb, since there are no empirical arguments to characterize it as a bound morpheme in EP; (iii) the infinitival complement is not sentential; (iv) the causee is not the syntactic subject of the embedded verb.
By virtue of Incaus, the external argument is internalized, i.e., it is merged in positions typically associated to objects.It is worth noting that in my analysis the causee corresponds to an argument of the embedded verb (or, at least, of the complex predicate), so, contrary to Zubizarreta (1985Zubizarreta ( , 1987)), Bordelois (1988), Baker (1996) and Alsina (1997), a. o., the causative verb is not a three-place predicate in the faire-Inf construction of EP.The main empirical arguments that support this approach are the following: The causee is semantically compatible with the embedded verb, as shown by the example in (32), which is ungrammatical because the dog is not able to read a book.the João made read the book to-the Peter (iii) If the causee was internalized as an argument of the causative verb, the asymmetry accusative/dative could not be explained, since this asymmetry is sensitive to the transitivity of the embedded verb only.
(iv) In ditransitive contexts the occurrence of another Goal argument in the embedded domain is disallowed, as illustrated in (34), whose ungrammaticality shows that two arguments are competing for the same -role in the infinitival domain.
(34) *O João não lhe mandou dar o livro ao Pedro.the João not CL-DAT-3SG made give the book to-the Pedro So, the structures of ( 35) are (partially) represented in ( 36 the João made eat the soup to-the Ana The incausativization of the embedded verb is adequately captured by these representations: in (36a), the above-mentioned verb is intransitive, and the causee is merged in the DO position; in turn, in (35b), a transitive context, the causee is merged as an IO, as expected. 8n order to legitimate the internalization of the external argument, we have to accept that Merge is only sensitive to the number of arguments of the predicate, not to their -roles (Gonçalves, 1999a).So, contrary to the Uniformity of Theta Assignment Hypothesis (UTAH; Baker, 1988), -roles may be discharged in the course of the derivation, in appropriate configurations.9Notice that, in both cases of (36), the number of arguments of each verb is correctly projected: trabalhar (to work) is a one-place predicate, so it is merged with one argument; comer (to eat) is a two-place predicate, then, two arguments are projected.
Suspending the external -role must not be interpreted as suppression.In fact, if the relevant -role was suppressed, the embedded verb would lose one of its arguments, and the derivation would crash, since Full Interpretation would be violated.On the contrary, the sentences in ( 35) are fully interpretable, which means that the mechanism of -role checking has been successful.The internalized argument is interpreted as the maximally prominent argument in the embedded domain, conforming the Thematic Hierarchy in (37) (see Bresnan & Kanerva, 1989;apud Alsina, 1996: 36;(20)): (37) Agent>Benef>Goal/Exp>Instrument>Pat/Theme>Locative This Hierarchy, associated to the effects of Incaus, predicts that, in transitive contexts, where a Theme argument is already combined with the verb, the causee is the Goal; in intransitive contexts, where no other argument is projected, the causee is the Theme.
Before ending this section, it is worth noting that the internalized argument is distinct from basic internal arguments.There are at least two arguments for this distinction: first, the internal argumentbut not the internalized argument can become the subject of matrix passives: (38) a. Estas casas foram mandadas construir a arquitectos famosos.
these houses were made built to architects famous b. *O atleta foi mandado correr.
the athlete was made run Second, in passive SE sentences, the internal argumentbut not the internalized argumentcan become the matrix subject: (39) a. Mandaram-se construir algumas casas a arquitectos famosos.
made-3PL-SE run the athletes

Residual problems
The analysis proposed here presents some advantages over previous work.First, it accounts for the position of the causee and its categorical nature.In fact, if we assume the internalization of the causee, we naturally derive the relative position of this constituent, dispensing with mechanisms that face theoretical and/or empirical problems. 10 Second, the DP/PP alternation is correctly predicted, and is closely related to the accusative/dative alternation: when the transitive verb is transitive and combines with a DP Object (accusative), the causee surfaces as a PP, otherwise, it could not check its (dative) Case features; when the embedded verb is intransitive, the causee is a DP, and exhibits the accusative Case.
the João {made wash-himself i /made-himself i wash} (to-)the Peter i .b. *O professor mandou ler os seus próprios i livros aos meninos i .
the teacher made read their own i books to-the children i In fact, if the causee was merged in [Spec, VP], as claimed by Villalba (1992) and Guasti (1993aGuasti ( , 1997), a. o., (40) ), a. o., (40) should be grammatical, since the causee would c-command the anaphor, so, the required binding configuration would be met.
Finally, if we assume that the internalized argument is the maximally prominent (internal) argument of the infinitival complement, according to the Thematic Hierarchy, the asymmetry Goal/Theme is correctly accounted for: when incausativization operates, intransitive verbs select for a Theme (the 10 See, for example, Burzio (1986), which requires V' extraction from the infinitival complement; Villalba (1992) and Guasti (1993aGuasti ( , 1997)), according to which [Spec, VP], the position of the causee occurs, on the right, contrary to Kayne (1994), andBordelois (1988), who proposes that the causee is the IO of the causative verb, and controls a PRO (the null subject of the infinitival complement).
causee) but transitive verbs select both for a Theme (the basic internal argument) and a Goal (the causee).Despite of these advantages, my proposal seems to face two problems.The first one concerns inalienable possession.Following Zubizarreta (1985), Villalba (1992) claims that contrasts like the ones in (41) suggest that the causee is the subject.( 41) O João i mandou abrir a boca *i/j ao Pedro j .
the João made open the mouth to-the Pedro Elaborating on Zubizarreta (1985), which claims that "it is a lexical property of a certain class of verbs to allow an 'Art+Body Part' object to be referentially dependent on the verb's subject by virtue of directly binding the determiner to the external argument" (op.cit.: 272), Villalba (1992) concludes that the causee is merged in the subject position.Since it c-commands the Object, the required binding condition is met, and the inalienable possession is adequately expressed.
However, I think that it is not true that in inalienable possession constructions the object is referentially dependent on the subject only.This fact is illustrated by examples like (42), inspired on Miguel (1996): (42) A Maria i lavou as mãos *i/j à Ana j .
the Maria washed the hands to-the Ana In this case, the subject a Maria does not denote the possessor (of the hands); on the contrary, the possession relation is established between the a-DP (à Ana) and the DP as mãos.Notice, however, that, as Miguel (1996) remarks, the a-DP in (42), à Ana is not a true IO, although it is replaceable by the dative clitic lhe: (43) A Maria lavou-lhe as mãos.
the Maria washed-CL(DAT, 3SG) the hands In turn, this constituent is distributionally equivalent to a genitive PP, introduced by de (Eng: of): (44) A Maria lavou as mãos da Ana.
the Maria washed the hands of-the Ana Based on the behaviour of the a-DP, Miguel (1996) suggests that, in (42), as mãos à Ana is a complex DO, which has the structure of a small clause: the The causee in the faire-Inf construction 211 a-DP is the subject and the DP as mãos (Eng: the hands) is the predicate.11Thus, following Miguel (1996), we can formulate the following hypothesis: the expression of inalienable possession is dependent on the existence of a predication domain.This hypothesis is adequate not only to Miguel's data, but also to the faire--Inf construction illustrated in (41) and repeated in (45): (45) O João i mandou abrir a boca *i/j ao Pedro j .
the João made open the mouth to-the Pedro In fact, I have suggested above that the PP ao Pedro (Eng: to-the Peter) corresponds to an internalized argument, which results from the incausativization of the embedded verb.This argument is interpreted as the maximally prominent argument in the embedded domain, conforming the Thematic Hierarchy in (37) -Theme, in intransitive contexts, Goal, in transitive contexts.In a certain sense, we can say that the causee is still the logical subject of the embedded domain, so, the required predication relation is obtained, and inalienable possession in ( 45) is correctly expressed.I leave the development of this question for future work.
The second (potential) problem concerns reciprocals.Kayne (1975) considers that the behaviour of French reciprocals is an additional argument for the analysis of the causee as the syntactic subject of a sentential domain.This way, a sequence like ( 46) is ungrammatical because the antecedent and the reciprocal occur in different domains: (46) *Ils feront parler cette jeune fille l'un de l'autre.
they make-FUT talk this young girl the one-MASC of the other (Kayne, 1975(Kayne, /1977: 250;: 250;( 171)) If we assume, as I propose in the present work, that (i) the infinitival domain is not sentential and (ii) the causee is not a subject, the matrix subject can qualify as the antecedent of a reciprocal occurring in the embedded domain.However, this is not always true, as the contrast between (47) and (48) shows: (47) As professoras i mandaram telefonar as meninas j umas às outras *i/j .the teachers made call the girls one-PL-FEM to the others-PL-FEM (48) Os professores i mandaram corrigir os trabalhos j uns aos outros i/*j .
the teachers made correct the works one-PL-MASC to the others-PL-MASC Notice that in (47) the reciprocal is bound by the causee, while in (48) the reciprocal is the causee itself.Then, we can generalize that the causee cannot intervene between the potential antecedent and the reciprocal.Based on this idea, we can formulate the following hypothesis: in the faire-Inf construction, (i) the antecedent c-commands the reciprocal and they share the same set of -features; (ii) the antecedent must be the most prominent argument that is closer to the reciprocal.This hypothesis seems to capture the contrast in (47) vs. (48).In fact, in (47), the antecedent of the reciprocal is the causee, not the matrix subject.Recall that the causee became an internalized argument, but it is still the more prominent argument of the embedded domain; thus it prevents the matrix subject from being the antecedent of the reciprocal.In turn, in (48), the reciprocal is the causee itself, so, in the embedded domain there is no legitimate antecedent, since the occurring Theme is less prominent than the causee.In this case, the matrix subject is able to bind the reciprocal, since there is no sentential barrier between the two elementsrecall that I have proposed that the infinitival domain is not sentential.

Concluding remarks
In this paper I have shown that in the faire-Inf construction of European Portuguese the causee does not exhibit the properties of a canonical subject.This fact is naturally derived from the hypothesis that in the above-mentioned construction the infinitival domain is defective in the sense that it lacks some of the functional heads classically related to the sentence, in particular AgrS and T. I have claimed that the infinitival domain is the projection of a null affix -Incaus -, which suspends the external -role of the embedded verb.As a consequence, this argument is internalized; i.e., it is merged in positions typically associated to objects (DO or IO, depending on the transitivity of the embedded verb).
In order to account for the behavior of the causee in what concerns its position and its -role, I have proposed that Merge is only sensitive to the number of arguments of the predicate, which entails that -roles may be discharged in the course of the derivation, differently from UTAH's claim.
Finally, I have suggested that the causee corresponds to the maximally prominent internal(ized) argument (Goal or Theme) of the embedded predicate, according to the Thematic Hierarchy.The thematic prominence of the causee is able to explain the specific behavior of reciprocals and the way inalienable possession is expressed.
, a. o., analyze the DP/PP italicized in constructions like the ones presented in (11) as the subject of the infinitival domain: (11) a. O João mandou trabalhar os meninos.the João made work the children 'João made the children work.' b.O João mandou comer a sopa à Ana. the João made eat the soup to-the Ana João made Ana eat the soup.'

(
12) a. O professor mandou ler o livro ao Pedro.the teacher made read the book to-the Peter b. *O professor mandou ler o livro ao cão.
): (22) O Presidente mandou defender [cada proposta] i ao seu i autor.the President made defend [each proposal] to its author As opposed to what is shown in (22), in non-causative contexts, the DO cannot bind the possessive expression in the subject constituent: (23) *O seu i autor defendeu [cada proposta] i .its author defended [each proposal Os pais i mandaram comprar eles i o livro às crianças.the parents made buy they the book to-the children b) *O Zé mandou não sair a Maria.the Zé made not leave the Maria c) *O Zé mandou ter saído os meninos.the Zé made have left the children AgrS a) Subject-verb agreement is not triggered b) Nominative SE cannot occur a) *Os professores mandaram lerem o livro aos alunos.the teachers made read-INF-3pl the book to-the children b) *Os professores mandaram acabar-se o trabalho rapidamente.the teachers made finish-SE the work quickly

(
32) *O professor mandou ler o livro ao cão. the teacher made read the book to-the dog (ii) The causee (Goal) always occurs after the infinitival complement (Theme), in opposition to what usually happens when the Theme is sentential (heavy complement effects: see (33a, b) vs. (33c, d)).(33) a. O João disse à Maria ter ido ao cinema.the João told to-the Mary have gone to-the cinema b. *O João disse ter ido ao cinema à Maria.the João told have gone to-the cinema to-the Maria c. *O João mandou ao Pedro ler o livro.the João made to-the Peter read the book d.O João mandou ler o livro ao Pedro.
Second, the causee cannot be the controller of PRO in Control constructions embedded in causatives, contrarily to what happens in Catalan (cf.(14a)) and in Italian (cf.(14b)): (18) *O Miguel mandou dizer PRO i ter visto a Ana ao Zé i .the Miguel made tell PRO i have-INF seen the Ana to-the Zé i