• Aucun résultat trouvé

BlkS distribution in control environments

3 Subjunctive complements in Balkan Slavic (BlkS)

3.2 BlkS distribution: Description

3.2.2 BlkS distribution in control environments

Control subjunctive complements that we will look at here are more idiosyncratic to the languages of the Balkan region, because they exhibit various different types of properties that are not typically observed with subjunctives in non-Balkan languages, as will be explained in more detail shortly. In (226-232) below, we can observe some examples of different types of control predicates that introduce embedded subjunctive complements across a wide array of Balkan languages:

(226) Ivan trjabva da dojde. (Bulgarian) John has-to SUBJ come3.sg.

‘John has to come.’

(227) O Kostas bori na odhiji. (Greek) the Kostas can3.sg. SUBJ drive3.sg.

‘Kostas can drive.’

(Roussou, 2009: 1815)

(228) Ivan zna da pliva. (Serbian) John knows(how to) SUBJ swim3.sg.

‘John knows how to swim.’

(229) Nauchio je da vozi auto. (Serbian)

learned has SUBJ drive3.sg. car

‘He learned to drive the car.’

(230) Maria perpiqet te shkruaje. (Albanian)

Mary try3.sg. SUBJ write3.sg.

‘Mary is trying to write.’

(231) Ion a reusit sa vina. (Romanian) John has managed SUBJ come3.sg.

‘John managed to come.’

162

(232) Toi pochna da studira pravo. (Macedonian) he began3.sg. SUBJ study3.sg. law

‘He began to study law.’

BlkS control complements are thus typically selected by predicates such as modal verbs (226-227), verbs of knowing (228-229), implicatives (230-231), or aspectual verbs (232), the latter of which we already observed earlier on thanks to some Balkan Slavic examples.

BlkS clauses in (226-232) exhibit a number of atypical patterns in the context of the cross-linguistic properties of the subjunctive mood. They are unusual from a semantic point of view because they exhibit a degree of semantic diversity that is not typically observed with the subjunctive in non-Balkan languages. More on this will be said a bit later on, once I assess complements of the type exemplified in (226-232) in light of my broader analysis of the subjunctive clausal mood. Furthermore, these complements are also unusual from a syntactic perspective, because they exhibit obligatory subject control, whereas subjunctives across languages were shown to be typically associated with the reverse, anti-control property of subject obviation in this context (see 2.2.2.3 or 2.4.1.5, for instance).

Note, moreover, that subject control in the context of BlkS is not achieved through the usual cross-linguistic means- i.e. via non-finite verb forms, such as infinitives or gerunds.

Rather, predicates such as those in (226-232) introduce complements associated with finite verbal morphology, where the control reading obtains due to the relationship of agreement sharing that is established between the matrix and the embedded predicate.79 The obligatory control associated with such complements therefore entails that the matrix and the embedded predicate cannot exhibit distinct φ-features, as we can observe thanks to the ungrammatical nature of the examples below:

(233) * Ivan trabvja da dojdete. (Bulgarian) John must3.sg. SUBJ come2.pl.

(234) * Pochinje da studirash pravo. (Serbian) begin3.sg. SUBJ study2.sg. law

79 See Landau (2004) and the references therein for a more detailed formal analysis of this BlkS-related control mechanism.

163

(235) * Ion a reusit sa vin. (Romanian) John has3.sg. managed SUBJ come1.sg.

The atypical control patterns that we observe with such BlkS complements are related to some diachronic developments that affected languages situated in the Balkan peninsula, which I will briefly explain before I move on with the analysis.

The fact that finite subjunctive complements appear in control structures in Balkan languages is related to linguistic phenomena subsumed under the term Balkan sprachbund, as I already briefly mentioned earlier on in 2.6. The relevant phenomenon in this context is the one of infinitive loss: many languages situated in the Balkan region have lost their infinitives (to a greater or lesser degree) and replaced them with finite subjunctives at some stage in their evolution, hence the wide distribution of BlkS complements that we just observed.80 The phenomenon of infinitive loss took place across Balkan languages independently of the broader typological differences between them, but it was subject to some regional variation: to put it in relatively simple terms, those languages that are situated more to the South-East of the Balkan region (e.g. Greek, Bulgarian, Albanian) have lost their infinitives in greater measure than languages situated more to the West (e.g. Croatian) or to the North (e.g. Romanian) of the Balkans (Joseph, 1983; Miseska Tomic, 2006, a.o.).

One of the examples of the regional variation in the spread of the infinitive-loss phenomenon that is particularly relevant when it comes to my study can be observed in the case of Balkan Slavic, specifically Serbian and Croatian, which exhibit one of the most conspicuous grammatical differences in this context. On the whole, Serbian was affected by the infinitive loss in greater degree than Croatian, due to their different geographical position: Serbian is spoken more to the East of the Balkans and Croatian more to the West, which is why Serbian speakers tend to use subjunctives in control contexts much more often than Croatian speakers,81 who prefer to employ the infinitive in control structures such as those in (226-232), given that Croatian was less affected by infinitive loss.

(236) a. Ivan mora dochi. (Croatian)

John must comeINF

80 See Joseph (1983) for a more comprehensive diachronic account of this phenomenon, which I will not develop here.

81 Standard Serbian actually productively employs both control subjunctives and infinitives, with the latter being progressively lost only in some southern dialects. Here I will only focus on Serbian control subjunctives.

164

b. Marija zna plivati.

Mary knows(how to) swimINF c. Pochinjem voziti auto.

begin1.sg. driveINF car

As a result, when I turn to the analysis of BlkS distribution, which is what I will be doing throughout the remainder of this chapter, I will refer to this language as Serbian, given that the latter exhibits more Balkanized properties in the context of subjunctive distribution than Croatian does.82

Now that I have briefly explained some of the diachronic reasons behind the idiosyncratic subjunctive complementation patterns that we observe in Balkan languages, as well as some of their different synchronic manifestations, I will turn to a closer theoretical study of these BlkS complements. First of all, given the unusual syntactic and semantic properties that we observed in relation to BlkS control subjunctives such as those in (226-232), whereby they differ from all Subj1 complements that we looked at so far in this dissertation, I need to begin by assessing whether these complements can nonetheless be analyzed as part of the Subj1 clausal mood, or whether they involve a different type of syntactic complementation, similarly as was proposed in the context of epistemic-type Subj2 BlkS complements such as those exemplified earlier on in (218-219).

If we only remain focused on the more easily observable formal and semantic properties associated with BlkS complements of the type exemplified in (226-232), then the latter type of analysis might appear more plausible. Not only were such complements shown to exhibit obligatory-subject control, which is the opposite formal property from the subject obviation that we typically observe with Subj1 clauses in this context, but a number of these complements are also associated with the type of semantic properties that have little or nothing to do with the typical meaning that we observe with intensional subjunctive or imperative clauses. This is particularly obvious when it comes to BlkS complements to implicative or aspectual predicates, such as those exemplified earlier on in (231-232), which are associated with entirely realis, non-modalized semantic interpretations, completely unrelated to the typical deontic modal interpretations that we observed with Subj1 or imperative clauses so far. Thus, under the view

82 In order to simplify the presentation, I will be referring to the language as Serbian throughout Sections 3.2 and 3.3, even when discussing those cases of subjunctive complementation where the latter exhibits the same properties as Croatian (i.e. non-control subjunctives), which will allow me to avoid unnecessary terminological complications.

165

of Subj1 clausal mood as a cluster of related morpho-syntactic and semantic properties that are shared between complements subsumed under this clause type (see 2.2), it is difficult to view complements such as those in (226-232) as part of the same mood category as the more typical intensional subjunctives.

Nevertheless, if we look at some other properties associated with BlkS control complements, then the analysis that would exclude them from the Subj1 clausal mood no longer appears as obvious. First of all, even though some control subjunctives, such as those in (231-232), do not exhibit the typical modal meanings that we observed with Subj1, others do. This is most obviously the case with control complements to deontic modal verbs, such as the one in (226) for instance, which exhibit the same type of deontic modality and world-to-word fit direction as the one we observed with intensional subjunctives and matrix imperatives.

Therefore, we cannot systematically exclude all BlkS complements associated with obligatory-subject control from the Subj1 clausal mood analysis just on the basis of their semantic properties. It is also implausible to suggest that only some of these complements, such as the one in (226), should be subsumed under the Subj1 label, while others, such as those in (231-232), should not, because the formal properties associated with all of these complements, when it comes to their subjunctive morphological marking or subject control, for instance (as well as a number of other ones that we will observe later on), are essentially the same. Therefore, a principled analysis of Subj1 clausal mood in this context would need to either include all of the BlkS complements in (226-232) under this label or exclude them. Some of the data that will be presented in the following paragraphs will argue in favor of the former type of approach to these control subjunctives.

Recall that the basic distinction between Subj1-type complementation and other types of clauses where subjunctive morphology can appear, such as Subj2, was related to the notion of selection: Subj1 complements, unlike Subj2, were seen as lexically selected by the matrix predicate under a separate subjunctive clause type. As a result, the use of subjunctive morphology in Subj1 complements was shown to be obligatory, whereas Subj2 clauses exhibited more variability with the indicative mood, as we also observed in the context of BlkS Subj2 earlier on in (218-219). If we now approach BlkS control complements through the same prism, we can note that they pattern with Subj1, not with Subj2-type clauses in this context:

(237) Ivan trjabva da / *che dojde. (Bulgarian) John must SUBJ / IND come

166

(238) Maria perpiqet te / * qe shkruaje. (Albanian)

Mary tries SUBJ / IND write3.sg.

(239) Ion a reusit sa / * ca vina. (Romanian) John has managed SUBJ / IND come3.sg.

The ungrammaticality stemming from the introduction of indicative markers in control BlkS complements exemplified above shows us that the use of subjunctive marking in such cases is obligatory, which suggests that such complements should also be seen as selected under the subjunctive clause type, rather than any other syntactic CP type. 83

83 Some of the complements we noted earlier on in (226-232) may appear to constitute an exception in this context.

This is the case, for instance, with complements introduced under the predicate know, exemplified earlier on in (228). As we can observe on the basis of Bulgarian examples below, this predicate can introduce both the indicative marker (che) and the subjunctive marker (da) in the embedded clause, differing in this sense from predicates such as those in (237-239) above:

The mood variation in (i), however, should not be seen as the same case of syntactic optionality as the one we observed with Subj2-type clauses (e.g. 218-221). This is because the two embedded variants in (i) constitute two distinct clause types: the indicative variant in (ia) exhibits all the clausal properties associated with the indicative clause type, whereas the subjunctive variant in (ib) exhibits clausal properties that are more typical of the Subj1 clause type. As a result, for instance, the latter cannot be associated with the indicative-related independent tense, as shown by the ungrammaticality stemming from the introduction of past-tense markers in the embedded clause:

(ii) * Znam da e pluval vchera. (Bulgarian)

know1.sg SUBJ has swum yesterday

In this sense, the subjunctive complement in (ii) patterns with Subj1 clauses, as opposed to indicatives.

Subj2-type complements, on the other hand, always exhibit the basic clausal properties related to the indicative clause type, regardless of the morphological mood marking (i.e. indicative vs. subjunctive) that appears in the embedded clause. Thus, for instance, as we observed several times already, Subj2 complements, such as the Greek and Bulgarian epistemic-type subjunctives exemplified in (224), can denote all types of temporal relations with respect to the matrix tense, including anteriority, pattering in this sense with indicatives. Later on in 6 we will see that this is only one of many clausal properties that such complements share with indicatives. As a result, the cases of mood-choice optionality involving Subj2, such as those we observed in the context of BlkS earlier on in (218-221), should be distinguished from those exemplified above in (i): in the former case, we have a single, indicative-CP selecting verb, which can optionally introduce subjunctive morphology in the embedded clause, producing only a slight shift in the modal interpretation (more on that in 6) but keeping all the basic indicative-related clausal properties; in the latter case, however, the optionality is only apparent, because we actually have two separate lexical entries (the factive know in (ia) and the dynamic modal know in (ib) in this particular case), one of which selects the indicative clause type, while the other selects the subjunctive (i.e. Subj1) clause type in its complement. Therefore, the use of subjunctive marking with the latter is actually obligatory.

167

Another indication that the control subjunctives we observed in (226-232) should be subsumed under the Subj1 label is the fact that they also exhibit other clausal properties that are characteristic of the subjunctive clause type. For instance, they pattern with the more typical intensional Subj1 complements when it comes to tense as well. As we can observe in the examples below, BlkS control subjunctives are subject to similar types of temporal restrictions as the ones we observe with intensional subjunctives, because both of these types of complements ban anterior tense readings.84

(240) a. * Hochu da je doshao juche. (Serbian) want1.sg. SUBJ has come yesterday

b. * Ma astept sa a venit ieri. (Romanian) me expect1.sg. SUBJ has come yesterday

(241) a. * Ivan trjabva da dojde vchera. (Bulgarian) John must SUBJ came3.sg. yesterday

b. * Ivan pokushava da je stigao juche. (Serbian) John tries SUBJ has arrived yesterday

c. * Toi zapochva da e karal kolata vchera. (Bulgarian) he begins SUBJ has driven car-the yesterday

In this sense, control subjunctives such as those in (241) differ not only from indicatives but also from BlkS Subj2 complements that we observed earlier on in 3.2.1, which were shown to be associated with indicative-type independent tense, as evidenced by the fact that they are compatible with anterior temporal readings:

(242) a. Pistevo na elise to provlima echtes. (Greek) believe1.sg. SUBJ solved3.sg. the problem yesterday

‘I think he solved the problem yesterday.’

84 In fact, the control subjunctive complements such as those in (241) will be shown to exhibit even more deficient temporal properties than the ones we observe with the more typical intensional subjunctives such as those in (240).

In this context, the latter will be analyzed as associated with dependent tense, whereas the former will be seen as associated with completely anaphoric tense. Nevertheless, for the moment, the relevant observation is that they all exhibit more deficient temporal properties than indicatives or Subj2.

168

b. Ne vjarvjam da dojde vchera. (Bulgarian)

not believe1.sg. SUBJ came3.sg. yesterday

‘I don’t believe he came yesterday.’

All of this, therefore, strongly suggest that BlkS control complements should also be included under the Subj1 clausal-mood analysis, despite some of the atypical properties that they exhibit in this context. This is because all of them display at least some of the related formal and semantic patterns that characterize Subj1 clausal mood as such. For instance, control subjunctives were shown to exhibit the same relation between the obligatory morpho-syntactic marking for the subjunctive and the dependent temporal properties of the subjunctive that we also observed earlier on in the context of the more typical intensional Subj1 complements.

Moreover, some control subjunctives, such as those selected by deontic modals, also exhibit a wider set of properties that are typical of the Subj1 clause type, because they denote the same type of modality that we observe in the more typical cases of Subj1 complementation as well.

All of these observations point towards the conclusion that control subjunctives should be analyzed as Subj1-type clauses as well. This analysis will be further reinforced later on in 3.3 once we focus more closely on the formal relation that can be established between control subjunctives and intensional subjunctive complements in the context of the Subj1 clause type.

The Subj1-aproach to BlkS control complements, however, brings forth a whole host of additional theoretical problems that will need to be addressed. Most importantly, it implies that BlkS Subj1 mood as such is associated with a much greater degree of formal and semantic diversity than the one we typically observe with the Subj1 clausal mood across languages, which makes it difficult to subsume BlkS under a broader cross-linguistic approach to Subj1.

This is the primary reason why I chose to place the central theoretical focus of my study of Slavic subjunctive on those Slavic languages that are situated in the Balkan region.

The diverse properties associated with BlkS Subj1 will ultimately be explained in light of the broader analysis of the subjunctive that I proposed in the first part of the dissertation, specifically in light of the claim that Subj1 clause type should be seen as the default syntactic option in embedded contexts, as well as the claim that clauses of this type can be subject to varying degrees of structural truncation, which can lead to both formal syntactic contrasts as well as interpretative contrasts once the structure associated with different types of Subj1 complements reaches the interface with semantics. Before I move on to analyze BlkS distribution through this broader theoretical prism, I will first use the next section to precisely

169

list and catalogue all the Subj1-selecting predicates that I will be dealing with throughout the remainder of this chapter.