• Aucun résultat trouvé

2 Slavic subjunctive as a clausal mood: Indicative vs. subjunctive clause types

2.4 Subjunctive CP as the embedded imperative CP

2.4.1 Common properties of imperative and subjunctive clauses

2.4.1.5 Anti-control

The final common property of imperatives and subjunctives that I will look at here, which is somewhat less obvious on the surface, is related to the area of control or, more precisely, anti-control. The anti-control property that we will go on to observe in this context is manifested in relation to the subjects of matrix imperative and embedded subjunctive clauses, restricting their reference to non-control readings. Recall that we already observed a similar type of anti-control effect in relation to subjunctives earlier on in 2.2.2.3 when we were discussing the phenomenon of subject obviation, which precluded conjoined reference between the matrix and the embedded subject in the context of subjunctive complementation (a couple of relevant Slavic examples are reproduced below).

(109) a. Ivani hochet, chtoby on*i/j ushel. (Russian)

John wants SUBJ he leave

b. Jani chce, zeby (pro*i/j) odszedl. (Polish)

John wants SUBJ he leave

‘John wants *(him) to leave.’

Nevertheless, the subject-obviation data such as those in (109) will be analyzed separately from the common anti-control effect that we will observe in relation to subjunctive and imperative clauses: the latter will be seen as a function of a special feature associated with the imperative/subjunctive CP structure (more on that a bit later on in 2.4.2), whereas the obviation observed in complements such as those in (109) will be analyzed as a result of a broader linguistic phenomenon, which is not directly related to the inherent structural properties of the subjunctive clause type.

The phenomenon in question is the so-called ‘subjunctive-infinitive competition’ (SIC), which has already been proposed as an explanation for the subject-obviation data we observe in subjunctives by a number of different authors (Bouchard, 1984; Farkas, 1992a; Schlenker, 2005 a.o.). As its name implies, SIC suggests that subjunctives and infinitives compete for similar types of embedded syntactic environments (which often involve one and the same

84

selecting predicate), with infinitives restricted to control contexts and subjunctives to non-control contexts.46 This allows to account for the data of the type exemplified below:

(110) a. Ivani hochet, chtoby on*i/j ushel. (Russian)

John wants SUBJ he leave

b. Jani chce, zeby (pro*i/j) odszedl. (Polish)

John wants SUBJ he leave

‘John wants *(him) to leave.’

(111) a. Ivani hochet (PROi/*j) ujti. (Russian)

John wants leaveINF

b. Jani chce (PROi/*j) odejsc. (Polish)

John wants leaveINF

‘John wants (*him) to leave.’

As we can observe in the Slavic examples in (110-111), verbs such as desideratives can introduce the subjunctive in the embedded clause only in the context of non-control readings, while infinitives are restricted to control readings, which can thus be used to explain the subject obviation effect that we noted in relation to complements such as those in (110). The same observation holds for Romance languages as well.

The idea that SIC is responsible for the subject-obviation effect we observe in subjunctives is further reinforced if we look at some languages where, for different types of reasons, the competition between infinitives and subjunctives is somewhat relaxed, or no longer in play.47 This is the case, for instance, in a number of languages situated in the Balkan region, including Slavic ones such as Bulgarian or Serbian, which no longer observe SIC because they have largely replaced their infinitives with subjunctives, for reasons that will be explained in more detail in the following chapter (see Section 3.2 in particular). As a result, such languages do not exhibit subject obviation in subjunctive complements such as those in (109-110), because we no longer have the infinitive variant competing for control contexts:

46 SIC is more straightforwardly applicable to subject-control than to object-control environments. The latter type of syntactic contexts, involving clauses such as those we observed earlier on in (80-82), sometimes allow for the subjunctive to be associated with control readings in some languages that otherwise exhibit SIC in relation to subject control. I will not attempt to account for this contrast here.

47 The relevance of such languages in the context of the SIC approach to subject obviation was first noted by Farkas (1992a).

85

(112) Ivani hoche da (proi/j)ode. (Serbian)

John wants SUBJ he leave3.sg.

‘John wants (him) to leave.’

(113) Ivani iska da (proi/j)dojde. (Bulgarian)

John wants SUBJ he come3.sg.

‘John wants (him) to come.’

As we can see in (112-113), subjunctive complements in these languages are compatible both with control and with non-control readings. A similar, but reverse property is also observed in a language such as English, where the infinitive has largely replaced the subjunctive. As a result, as we can note from the translations below the examples in (112-113), English infinitives introduced under verbs such as desideratives are also compatible both with control and with non-control readings, depending on the type of embedded subject we have (i.e. overt subject vs. PRO).

However, SIC cannot be seen as the comprehensive explanation behind subject obviation data cross-linguistically, because this anti-control effect obtains in certain types of syntactic environments independently of SIC. This can be most easily observed if we look, once again, at some of the languages where SIC is no longer in play. For instance, even though languages such as Bulgarian or Serbian do not exhibit subject obviation across all intensional subjunctives, which is typically the case in Romance or in most Slavic languages, they nonetheless contain a group of subjunctive complements which do exhibit this anti-control phenomenon, as we can see below in (114-115):

(114) Ivani je naredio da (pro*i/j) ode. (Serbian) John has ordered SUBJ he leave3.sg.

‘John ordered *(him) to leave.’

(115) Ivani zapovjada (pro*i/j) da dojde. (Bulgarian) John orders he SUBJ come3.sg.

‘John orders *(him) to come.’

86

The same property is observed in English as well, which also bans subject-control readings in complements such as those in (114-115), even though English clauses of this type involve an infinitive, not a subjunctive construction.

(116) Hei ordered *(himj)i to come.

The group of complements that exhibits subject obviation independently of SIC in these languages is associated with a specific type of semantic properties and selected only by a restricted and well-defined group of predicates. The latter correspond to the definition of directive verbs (e.g. order, command, insist etc.), which introduce complements that denote reported directive speech acts and can thus be defined as embedded imperatives. This is where the cross-linguistic link between subjunctives and imperatives in terms of anti-control will come into play.

Matrix imperatives are associated with a similar type of anti-control property as the one we just observed in the context of subjunctive complements such as those in (114-115), as was originally noted by Kempchinsky (1998; 2009 a.o.). The only difference is that the anti-control effect in matrix imperatives is not established with respect to some higher subject, given that they involve independent clausal structures, but with respect to the speaker: imperative clauses disallow conjoined readings between the subject and the speaker, which is manifested through the ban on imperatives appearing in first person singular (Kempchinsky, 2009). Nevertheless, if we abstract away from the difference between matrix vs. embedded syntactic environments, we can note that the anti-control effect related to subjunctive complements of the type exemplified in (114-115) and matrix imperatives in general functions in essentially the same way. In both cases, it is established with respect to the relevant external illocutionary point: in subjunctive complements, the illocutionary point in question is the higher matrix subject, hence the ban on conjoined reading between the embedded and the matrix subject and the effect of subject obviation; in matrix imperatives, the relevant illocutionary point is the speaker, hence the ban on conjoined reading between the subject and the speaker and the ban on imperatives appearing in first person singular. As I suggested at the beginning of this section, the anti-control effects we observe in this context will be explained by referring to the underlying feature make-up related to the CP projection used to introduce imperative and subjunctive clauses into the structure, which I will analyze in more detail in the following section.

87

2.4.2 Imperative/subjunctive CP: Formalization

All of the data put forward so far here in 2.4 point towards the conclusion that matrix imperatives and embedded subjunctives should be seen as closely related clauses: they were shown to semantically pattern in terms of their propositional content, direction of fit and the type of modality they denote; they exhibit the same type of dependent temporal properties; they pattern in relation to anti-control; and they also exhibit a number of cross-linguistic overlaps in their distribution. The way in which I will propose to account for this wide range of shared properties between the two types of clauses is by claiming that they should be seen as syntactically associated with the same type of CP projection, with the subjunctive CP corresponding to the embedded instance of the matrix imperative CP. This is not a particularly new or radical proposal, because a similar approach was also put forward by authors such as Han (1998) or Kempchinsky (2009), among others. In fact, I will use parts of the analyses proposed by these authors as a basis for my own study of the imperative/subjunctive CP.

The only type of formal point I made so far with regards to the internal make-up of the subjunctive CP (and, by extension, imperative CP as well) was that, unlike its indicative CP counterpart, it does not contain the extra WP projection which was seen as associated with extensional world-anchoring from a semantic point of view and phase closure from a syntactic point of view. At this point, I will focus more closely on some of the intrinsic properties associated with the subjunctive/imperative CP itself. I will begin by analyzing the properties of this CP in matrix imperative environments, and then I will apply the analysis that will be proposed in this context to embedded subjunctives.

2.4.2.1 Matrix imperative contexts

First of all, I am operating under the assumption that all matrix clauses are syntactically introduced under a CP projection, regardless of whether the latter contains some overt item or whether it is empty. This is because such a projection is necessary in order to denote clause type once the syntactic derivation reaches the interface with semantics. In this sense, every C-head should be seen as containing some type of interpretable clause-typing feature, which is related to the prototypical function pertaining to a given type of clause. As a result, just like Chomsky (1995) claimed that the C-head associated with questions contains an iq feature