• Aucun résultat trouvé

Syntactic derivation of Bulgarian subjunctives

3 Subjunctive complements in Balkan Slavic (BlkS)

3.1 Morpho-syntactic realization of BlkS

3.1.3 Bulgarian subjunctive complements

3.1.3.3 Syntactic derivation of Bulgarian subjunctives

This is the point where the more general syntactic analysis pertaining to the subjunctive CP clause type that I proposed earlier on (Section 2.4 in particular) in the context of a broader study of Slavic subjunctive complementation will begin to receive a more concrete syntactic application. First of all, recall that subjunctive CP was analyzed as the embedded instance of the matrix imperative CP, the latter being a projection of an imperative operator in C, which

124

contains a hierarchical feature cluster consisting of the higher, clause-typing Dir(ective) feature, and the lower modal Deo(ntic) feature, as illustrated below:

(172) [CP C iDir.>uDeo]

For the moment, I will abstract away from Dir and only focus on the Deo feature, given that the latter was argued to have a greater impact on the syntactic operations taking place within a given subjunctive/imperative clause, since it was analyzed as an uninterpretable feature which can function as an attractor Probe and cause movements. I will reintroduce Dir within my analysis later on in 3.3 once I focus on BlkS distribution and the formal and semantic differences that can be observed between various BlkS complements.

Recall that the introduction of an uninterpretable uDeo feature within the imperative/subjunctive CP structure was originally motivated by the fact that various different languages exhibit left-periphery movements in matrix imperative clauses (see (120-123) in 2.4.2.1). Such movements were seen as motivated by the feature-checking requirements of the strong uDeo in C, which needs to be locally checked by the interpretable iDeo instance of the same feature. The latter will be analyzed as initially inserted under the Mod-head that is the locus of deontic modality, and which is situated between the higher C containing uDeo and the lower T-head where the subjunctive particle is externally merged, as illustrated below:

(173) [CP C iDir.>uDeo [ModP Mod iDeo [TP TDA [vP ]]]]

Given the broader syntactic analysis of the subjunctive clause type as an embedded instance of the matrix imperative CP, I will argue that the same type of Agree/feature checking mechanism that is responsible for movements in imperatives obtains in subjunctive complements as well.

The prediction, therefore, is that we should also observe left-periphery movements in embedded subjunctive contexts, specifically in those languages where uDeo is strong (given the overall feature-strength approach to movement based on Chomsky (1995)).

Shifting back to Bulgarian, we already noted earlier on that some authors argued that one can indeed observe such movements within subjunctive complements in this language. For instance, Krapova (1998) claimed that the subjunctive C-head in Bulgarian contains an uninterpretable Mood feature which is checked via verb movement to C. If we recall the discussion from earlier on, this allowed Krapova to reconcile her analysis of the subjunctive da

125

as a C-inserted item with the observed syntactic contiguity between da and the verb.

Nevertheless, even though the presence of V-C movement in Bulgarian subjunctives would be desirable in light of my overall approach to subjunctive syntax, we have already observed various types of evidence that argue against this analysis. As a result, the feature-checking mechanism pertaining to Deo in (173) will need to be accounted for in a different fashion when it comes to Bulgarian subjunctives.

Before I analyze how the feature-checking of uDeo obtains in Bulgarian, I first need to address a separate issue with regards to the structural analysis in (173) that is left hanging, related specifically to the syntax of the subjunctive da in light of the claim that the latter is inserted in T. Recall that one of the original motivations for this claim was the fact that da can appear in non-modalized semantic environments, such as the one associated with the clause we observed earlier on in (160), which would be difficult to account for if da was viewed as externally merged under some modal head or under subjunctive C. The potential problem with this analysis, however, is that the structure proposed in (173) does not allow to straightforwardly account for the modal properties that this element does typically exhibit. Regardless of the existence of clauses such as the one in (160), where da is introduced under realis-type predicates such as aspectuals which do not appear to be associated with any modality, this item nonetheless exhibits modal properties in most contexts of its use, not just in embedded intensional subjunctives, but also in matrix clauses such as those we observed earlier on in (156-157), where the use of da crucially contributes to the deontic modal interpretation of the sentence.

The desired analysis of the syntax of da would therefore be the one that accounts both for the fact that this item is typically associated with modal interpretations, as well as the fact that it can sometimes appear in non-modalized semantic environments.

The way in which I suggest to account for all this is by proposing a syntactic mechanism whereby the item da situated in T establishes an Agree relationship with the higher deontic Mod head. The precise formal nature of this relationship is illustrated below:

(174) [ModP Mod[iDeo] [TP TDA [iDeo] ]]

Agree/F-transfer

As we can see in (174), the Agreement that is established in this context results in feature transfer, whereby the iDeo feature is transferred from Mod to the particle da under T, thus

126

endowing the latter with the modal properties that this item is typically associated with.60 The Agree mechanism in (174) obtains in certain types of syntactic environments but not in all of them, explaining why da is not always associated with modalized interpretations. In order to illustrate in more detail how this analysis works in the context of subjunctive complementation, let us look at the step-by-step derivation and structural build-up of Bulgarian subjunctives.

As I already implied earlier on in the introductory chapter, I will assume a derivational, bottom-up approach to structure build-up in my syntactic analysis of subjunctive complementation, which is in accordance with the overall minimalist view of syntax (see 1.3).

The first relevant Merge operation when it comes to my analysis of the Bulgarian subjunctive is the one whereby the T-head, projecting TP, is merged on top of vP (I ignore the vP-internal Merge operations, since they are less relevant for the present discussion). This is the operation whereby the subjunctive da is first introduced in the structure, as we can observe below.

(175) TP

T vP

da

At this point, da only functions as a tense operator, in accordance with the analysis proposed earlier on 3.1.3.2, i.e. it binds the dependent t-variable contained in the embedded verb, which has already been merged vP-internally (a bit more on the latter Merge will be said later on).

Crucially, though, the derivational step in (175) does not yet endow the particle da with any type of modal feature, because the relevant Mod projection has not yet been merged in the structure. This is what will ultimately allow me to account for da appearing in non-modalized semantic environments: (175) will be argued to constitute the final derivational step in the structural build-up associated with the atypical, non-modalized Bulgarian subjunctives such as the one in (160) (as well as its BlkS equivalents in general). Such complements will be shown to be associated with syntactically anaphoric, non-phasal properties, which will be explained by claiming that their structure is truncated all the way down to TP. This still allows for the particle da to be inserted in the structure of such clauses, but it does not allow it to acquire any type of modal property, given that the Agree relationship in (174) hasn’t yet obtained, which

60 This proposal is based on some of the more general syntactic analyses of Agreement which view the latter as involving feature transfer or feature sharing (see Frampton, 2000; Legate, 2005; or Bobaljik, 2008 a.o.).

127

therefore explains why da can be found in non-modal subjunctive complements of the type we observed in (160).

The next relevant derivational step in the build-up of Bulgarian subjunctives is the one whereby the Mod-head, which contains iDeo and projects ModP, is merged on top of TP:61

(176) ModP Mod[iDeo] TP

T vP da[iDeo]

Agree/F-transfer

Once the Merge operation in (176) has taken place, the subjunctive marker da in T enters into Agreement with Mod, which results in feature transfer of Deo from Mod to da, endowing the latter with the modal properties that are typically observed with this item. At this point, therefore, da can function both as a tense operator and as a modal particle, which are the usual dual properties we observe with this element.

The next, and final step in the syntactic derivation of Bulgarian subjunctives, is the one that merges the C-head, containing uDeo and projecting subjunctive/imperative CP, on top of ModP, as shown below in (177):

(177) CP

C[uDeo] ModP

Mod TP T vP da[iDeo]

Given that all uF features must be checked and deleted before the structure reaches the interface with semantics, the final question that needs to be answered is how this feature-checking is achieved when it comes to uDeo in Bulgarian subjunctives. Under the standard minimalist

61 For the purposes of the present discussion, I ignore the internal structure related to clausal modality, which will be analyzed later on in terms of a slightly more articulated syntactic layer as opposed to a single projection (see 3.2-3.3). Here I only focus on the projection encoding deontic modality, which contains the iDeo-feature, because this projection is most relevant when it comes to accounting for the properties of the subjunctive particle da, as well as for the derivation of the typical intensional subjunctives in Bulgarian, which are the focus of the present analysis.

128

assumptions, this can be done in one of two ways: either through simple Agree, whereby uF is checked off against the relevant iF in a long-distance configuration; or through Agree combined with Move, whereby the relevant iF is carried by some overt item under the projection that contains uF in order to check the latter locally. There are several reasons to assume that Bulgarian exhibits the former strategy.

First of all, the fact that we observed earlier on that the embedded subject in Bulgarian subjunctives can precede the particle da and appear at the beginning of the clause (see (156)) would suggest that the item da does not end up under the C-head selected by the predicate, the latter remaining empty in this context, because otherwise the configuration of the type exemplified in (156) would imply a violation of the locality constraint on CP-selection. Another argument against the claim that da moves to C has to do with the syntactic contiguity that we observed between the particle da and the embedded verb in Bulgarian subjunctives. The only way to reconcile this contiguity with the idea that da moves to C, given my overall syntactic analysis of Bulgarian subjunctives (specifically the T-insertion analysis of da), would be to claim that somehow both the particle and the verb move up to C in order to check uDeo, which would be difficult to justify in light of the Economy principle given that at least one of these movements seems superfluous. As a result, I will argue that there is no overt C-related movement in Bulgarian subjunctives: the uDeo in C is weak and can hence be checked through a long-distance Agree configuration with iDeo that has been transferred from Mod to da, as illustrated below:62

(178) [CP C[uDeo] [ModP Mod [TP T DA [iDeo] [vP]]]]

Agree/Check

Nevertheless, once we focus on the syntactic derivation of subjunctives in some other Slavic languages (including Croatian in the following section), we will observe that this type of clauses

62 While the feature transfer of iDeo from Mod to da in Bulgarian subjunctives may appear as trivial from a syntactic, feature-checking standpoint, because uDeo in C could have just as well been checked off via Agreement with iDeo in Mod, which would not have required any preceding syntactic operation to take place, this proposal is crucial when it comes to accounting for the semantic properties associated with the Bulgarian subjunctive particle (most importantly, the fact that it can but needn’t be associated with modal interpretations). Moreover, the claim that Deo is transferred from Mod to the particle da in T will also become relevant from a syntactic point of view once I turn to the analysis of subjunctive derivation in Croatian, because it will allow me to account for the movement that the subjunctive particle will be shown to undergo from T to C in this language. See 3.1.4.3 for more detail.

129

can also exhibit overt movements to C, as would be expected under my overall analysis of the Subj1 clause type.

Before I can claim to have the full structural description of the syntactic derivation of Bulgarian subjunctives, I still need to say a word or two about the positioning of the verb within the vP structure, even though the latter is not of central concern for my analysis. Under the assumption that the vP constitutes a relatively articulated syntactic layer, and given the observed syntactic contiguity between da and the verb, it makes sense to claim that the latter is situated in some high vP-related projection that is close to T. In this context, I will claim that the verb appears under the head position of the Asp(ectual)P projection, which has often been analyzed as the highest projection within the verbal layer, situated in a close syntactic configuration with TP (Borer, 1994; Thompson, 2006; Travis, 1992 a.o.).63

(179) [CP [ModP[TP…T da [AspP…Asp V [VP]]]]]

In addition to accounting for the syntactic contiguity observed between the verb and the subjunctive particle in T, this analysis can also account for the more general observation that verbal aspect plays a greater role in the context of mood distinctions in Balkan languages, such as Greek or Bulgarian (as well as Croatian, as we will see in the following section), than is typically the case from a cross-linguistic perspective. Moreover, the representation in (179) can also explain some of the semantic properties associated with verbs that appear in Bulgarian subjunctive complements, specifically the fact that, while being temporally deficient, such verbs are nonetheless fully specified for aspect on a semantic level (regardless of whether or not a given aspectual interpretation is associated with its usual morphological marking). As we can see below, subjunctive-related verbs in Bulgarian can denote both perfective and imperfective-type meaning.

(180) Iskam da dojdesh dnes v sedem chasa / vseki den.

want1.sg. SUBJ come2.sg. today at seven hour / every day

‘I want you to come today at seven o’clock / every day.’

63 The vP-internal movements that allowed the verb to end up in this position are of less concern for my current analysis, so I will ignore this issue here.

130

The representation in (179), which claims that the subjunctive verb is directly associated with the aspectual Asp head, allows to straightforwardly account for all these observations.

At this point, I claim to have reached a comprehensive syntactic analysis related to the derivation of Bulgarian subjunctive complements, which is sufficient to account for the basic properties associated with the typical intensional subjunctives that were the primary focus of my study of BlkS so far. The syntactic account I proposed in this context will be further refined and articulated later on in 3.3, once I focus on the issue of BlkS distribution and look at some finer formal and semantic differences that can be observed between various types of BlkS complements. Before I do that, however, I will first assess whether a similar analysis as the one I just proposed in the context of Bulgarian can be used to account for the properties of intensional subjunctives in the second Slavic language that I will focus on in my study of BlkS, namely Croatian.64