• Aucun résultat trouvé

Syntactic derivation of Croatian subjunctives

3 Subjunctive complements in Balkan Slavic (BlkS)

3.1 Morpho-syntactic realization of BlkS

3.1.4 Croatian subjunctive complements

3.1.4.3 Syntactic derivation of Croatian subjunctives

The basic syntactic analysis that was proposed earlier on in the context of Bulgarian subjunctive, which I will now assess in light of Croatian data, can be summarized as in (200):

(200) [ CP C[uDeo] [ModP Mod [TP TDA [iDeo] [AspP V [vP…]]]]]

Agree/Check

The most important syntactic claims related to the analysis represented in (200) had to do with the syntax of the Bulgarian mood marker da, as well as the feature-checking mechanism pertaining to the uDeo feature in C. The former was claimed to be inserted under the temporal T-head position, which allowed to account, among other things, for the observed temporal properties related to the Bulgarian subjunctive particle, as well as for its syntactic contiguity with the embedded verb (the latter having been analyzed as situated within the AsP-head, which is contiguous to T). As for the feature-checking mechanism related to Deo, it was claimed to be accomplished through a simple, long-distance Agree relation between uDeo in C and iDeo associated with the particle da in T (the latter having received iDeo through the mechanism of feature transfer between Mod and T, as shown earlier on in 3.1.3.3), which was sufficient to check uDeo because the latter was analyzed as weak in Bulgarian. In the following paragraphs, I will assess whether a similar analysis can be applied to Croatian subjunctives as well.

The first issue I will focus on in this context is the place of insertion of the Croatian mood particle da, and then I will look at the derivation of subjunctive complements in this language from a broader perspective. First of all, the data such as those we observed in (185), where we could see the item da co-occurring with the indicative Comp, clearly show that the subjunctive da in Croatian cannot be seen as inserted in C but must be analyzed as generated somewhere lower down in the structure, which is thus another aspect in which this item patterns with its BlkS counterparts. Given the basic structural description I assume in my analysis of BlkS in general, corresponding to the syntactic representation in (200), the only remaining candidates that can host da are the modality-related Mod-head and the temporal T-head. There are several different types of arguments that favor the latter over the former in this context.

144

The arguments in favor of the T-insertion of Croatian da are roughly the same as those that were advanced earlier on to argue for the same analysis with regards to its Bulgarian counterpart. The first argument is related to the observed temporal properties of these items:

both the Bulgarian and the Croatian particle da were shown to function as tense operators, syntactically binding the dependent t variable contained within the PNP verb and semantically providing a temporal anchor for the interval within which such verbs can be interpreted.

Another, even more important argument that was advanced in favor of the T-insertion analysis of da in Bulgarian was related to the distribution of subjunctive da-complements in this language and, more specifically, the fact that the particle da was shown to appear in some subjunctive clauses, such as the one we observed earlier on in (160) (reproduced below), that are not semantically associated with any modality.

(160) Ivan zapochva da kara kolata.

John begins SUBJ drive3.sg. car-the

‘John begins to drive the car.’

Given the overall syntax-semantics mapping perspective assumed in this thesis, the existence of clauses of the type exemplified in (160) argued against the analysis of the Bulgarian da as inserted under any type of higher modal head, because the latter would preclude it from appearing in such non-modalized semantic environments.

As we can see on the basis of the example in (201) below, the same type of issue is at play when it comes to the distribution of the Croatian particle da as well: just like its Bulgarian counterpart, this item can also appear in non-modalized contexts.73

73 Even though the complements of the type exemplified in (201), which are syntactically associated with obligatory subject control, are more typical of Serbian than they are of Croatian, for the reasons I briefly sketched out at the beginning of 3.1.4 (and which I will explain in more detail later on in 3.2), there is no reason to presume that the particle da we observe there is not the same type of formal item as the particle da we observed in Croatian subjunctive examples so far. If we look at subjunctive distribution in other Balkan languages (as we will do in 3.2 and 3.3 in more detail), where BlkS markers are more overtly marked, we can note that there is no difference in the formal make-up of subjunctive particles regardless of whether they appear in non-control or in control syntactic environments, so there is also no reason to assume that the particle da appearing in non-control contexts (as in Croatian subjunctive complements we focused on so far) and the particle da appearing in a control complement such as the one in (201) are a different formal item either. The only difference pertaining to the particle da in these two types of contexts will be seen as related to the syntactic operations that this item undergoes after it has been inserted in T. More on that shortly.

145

(201) Ivan pochinje da vozi auto.

John begins SUBJ drive3.sg. car

‘John begins to drive the car.’

The best way to account for all of these data is to claim that the Croatian subjunctive particle da, just like its Bulgarian counterpart, is also inserted under the temporal T-head position, which appears below any type of projection related to modality, given the structure in (200).

The only important contrast between Croatian and Bulgarian when it comes to the syntax of the particle da has to do with the positioning of this item with respect to the embedded verb in subjunctive complements. Even though the Croatian data we looked at so far in this context seemed to suggest that the Croatian da exhibits the same type of contiguity with the lower verb as its Bulgarian counterpart, because this was shown to be the case in future-tense constructions such as those in (186), the same degree of contiguity is not observed in subjunctive complements in the two languages. The relevant contrast between Croatian and Bulgarian in this context can be most easily noticed by looking at the syntactic positioning of the embedded subject in such clauses. So far, all of the Croatian subjunctive examples I used have contained an empty pro subject, which did not allow us to observe any type of syntactic difference between the two languages in this context. However, if we introduce a nominal subject in an embedded subjunctive complement in Croatian, we can observe a clear syntactic contrast with respect to Bulgarian, as shown in the examples below:

(202) a. Iskam Ivan da (*Ivan) dojde. (Bulgarian) want1.sg. John SUBJ John come3.sg.

b. Hochu da Ivan dodje. (Croatian) want1.sg. SUBJ John come3.sg.

‘I want John to come.’

As we can see in (202), the canonical position for the embedded subject in Croatian subjunctive complements is the one between da and the verb, and in this sense subjunctives do not syntactically differ from indicatives, whereas the same type of configuration results in ungrammaticality in Bulgarian.

The contrast between the two languages in this context is not just observed when it comes to the embedded subject intervening between the subjunctive particle and the verb, but

146

also when it comes to the possibility of inserting other elements, such as adverbs or focus, between these two items, all of which can be done in the context of Croatian subjunctives, as we can observe through the examples in (203):

(203) a. Hochu da sutra dodjesh.

want1.sg. SUBJ tomorrow come2.sg.

‘I want you to come tomorrow.’

b. Hochu da IVANA pozovete.

want1.sg. SUBJ JOHN invite2.pl.

‘It is JOHN that I want you to invite.’

The data in (202-203) therefore suggest that Croatian particle da does not have identical syntactic properties as those we observed with its Bulgarian counterpart (or with BlkS markers more generally), but that it occupies a higher structural position than the latter, which is further removed from the embedded verb, explaining the lack of syntactic contiguity in (202-203).

Therefore, we now have two contrasting sets of data related to the Croatian subjunctive particle da that need to be accounted for: on the one hand, the latter patterns with its Bulgarian counterpart in that it accomplishes the same temporal function of binding the dependent t variable in the PNP verb, and in that it exhibits the same sort of contiguity with the lower verb in certain types of syntactic environments (namely in future-tense constructions such as those we observed in (186)); on the other hand, the Croatian particle da is further removed from the embedded verb in the context of typical subjunctive complementation than is the case with its Bulgarian counterpart. The syntactic configuration exemplified in (203b) in particular, which shows that the Croatian particle da appears above fronted focus, would suggest that this item occupies a very high position within the embedded left periphery, which is higher than the projection serving as the locus for focalization. The way in which I will propose to account for these different sets of syntactic patterns is simple: I will claim that the Croatian da is inserted under the same T-head as its Bulgarian counterpart, hence their shared properties, but then, unlike the latter, it moves from T to C, hence the contrasts between Bulgarian and Croatian subjunctives in relation to data such as those in (202-203). In order to explain the underlying reasons for this difference in the syntactic behavior of the particle da in two languages, I will now look at the step-by-step derivation of Croatian subjunctives, similarly as was done earlier on in the context of Bulgarian.

147

The derivation of Croatian and Bulgarian subjunctives will not differ when it comes to the Merge operations that underlie the structural build-up of subjunctive complements, but only with regards to the syntactic operations that take place within the overall structural framework associated with the subjunctive clause type. The first relevant Merge operation when it comes to the derivation of Croatian subjunctive complements is thus the same as the one we observed earlier on in the context of Bulgarian, i.e. the operation whereby the head T, and its projection TP, are merged on top of vP (once again, I ignore vP-internal Merge operations):

(204) TP

T vP

da

As explained earlier on, this is also the derivational step where the subjunctive particle da is externally merged in the structure under T. Given that the exact same situation was argued to obtain at this derivational stage in the context of Bulgarian as well, this allows me to account for all the common patterns that we observed between the Bulgarian and Croatian particle da.

The derivational step in (204) can explain, first of all, the temporal properties associated with both the Croatian and the Bulgarian da: given that they are both analyzed as inserted in T, they can both be argued to function as temporal operators, binding the dependent tense variable associated with the verb situated lower down within vP. Moreover, this analysis can also explain the fact that the Croatian particle da is structurally contiguous with the lower verb in the future-tense constructions such as those in (186). In this context, we can apply a similar type of explanation as the one proposed earlier on for Bulgarian subjunctive complementation more generally: da in such cases remains under T, in a syntactically contiguous position with respect to the lower verb (the latter will be argued to be situated in Asp, which is immediately contiguous to T, in Croatian as well). The reason why Croatian da does not move up to C in such cases is because this movement will be seen as crucially related to the feature-checking requirements associated with the subjunctive CP, which is not present in structures such as those related to clauses in (186), given that they do not involve subjunctive complementation.

Finally, the analysis in (204) will, once again, be used to explain why the Croatian particle da, like its Bulgarian counterpart, can appear in non-modalized semantic environments:

complements such as those in (160) or (201), which do not denote any modality, will be shown

148

to constitute non-phasal syntactic domains, whose structural build-up stops at the derivational step in (204), before any modality-related projection is merged within the structure.

The next derivational step in the build-up of Croatian subjunctive complements is, once again, identical to the one we observed at the same stage in the context of Bulgarian subjunctives earlier on: it is the step whereby the modal head Mod, which contains the interpretable deontic feature iDeo, is merged on top of TP:

(205) ModP Mod[iDeo] TP

T vP da

Just like it was argued earlier on in the context of Bulgarian, this is also the derivational stage that endows the particle da with its modal properties. Once again, the particle da in T establishes an Agree relationship with Mod, which allows it to acquire the modal feature iDeo through the mechanism of feature transfer, as illustrated below in (206):

(206) ModP Mod[iDeo] TP

T vP da[iDeo]

Agree/F-tranfer

At this point, therefore, the particle da, in addition to being endowed with semantic modal properties, can also syntactically function as a potential feature-checker for uDeo in C.

The feature-checking property associated with the Croatian particle da becomes relevant in the context of the final derivational step in the structural build-up of Croatian subjunctive complements, i.e. the step whereby the subjunctive C, containing uDeo and projecting CP, is merged on top of Mod, as illustrated below:

149

(207) CP

C[uDeo] ModP

Mod TP T vP da[iDeo]

This is the point where we observe the most important difference between the syntactic derivation of Croatian subjunctives and the one associated with their Bulgarian counterparts: in the latter case, the particle da stays under T, checking uDeo in C through a long-distance Agree configuration, but in the former case, the item da must be analyzed as moving to C, given the data of the type exemplified in (202-203). The syntactic contrast between Croatian and Bulgarian in this context can be explained by postulating a single formal distinction with regards to the feature make-up of their subjunctive CP: unlike Bulgarian subjunctives, which contain a weak uDeo in C, their Croatian counterparts contain a strong uDeo, which needs to be checked locally. Given that this type of feature checking can only be achieved by the particle da, the latter is then attracted and head-adjoined to C, as illustrated below in (208):74

(208) CP

C[uDeo] ModP

da[iDeo]

Mod TP

T vP da[iDeo]

Agree/Move/Check

The analysis in (208) can account for all the relevant properties associated with the Croatian subjunctive particle that we observed so far: it explains both the shared patterns that this item was shown to exhibit with respect to its Bulgarian counterpart (which are accounted for under the assumption that they are both inserted under T), as well as the contrasts we observed

74 If we want to maintain the Head Movement Constraint proposed by Travis (1984) in the context of the head movement of da from T to C, then we can simply add that da passes through the modal head on its way to C. Even though this additional stipulation would have no meaningful impact on my analysis of Croatian subjunctive derivation, one would need to provide some broader theoretical justifications for it (especially in light of the Attract (vs. Greed) approach to movement (including head movement) that I am assuming in this thesis). I will not attempt to do so here.

150

between the two (explained by the fact that the Croatian da, unlike its Bulgarian counterpart, subsequently moves to C).

Before I end my analysis of Croatian subjunctive derivation, I will add a couple of words concerning the properties of the verb that appears in this type of clauses. Like its Bulgarian counterpart, the latter will be analyzed as situated under Asp, i.e. the highest head within the verbal vP layer, as illustrated below:

(209) [CP [ModP [TP [AspP Asp V [vp ]]]]]

This analysis allows to explain the shared properties that we observed between Croatian and Bulgarian subjunctive-related verb forms. First of all, given that AspP is immediately contiguous to the T-head where da is inserted, the structural representation in (209) allows to explain the contiguity that is observed between the verb and the particle da in those syntactic environments where the latter does not move up to C (e.g. in future-tense constructions such as those in (186)). The same analysis also accounts for the semantic properties associated with verbs that appear in subjunctives, specifically the fact that, while they are temporally deficient, they are fully specified for aspect on the semantic level. As we can observe in (210) below, these types of verbs in Croatian, just like their Bulgarian counterparts, can denote both perfective and imperfective-type meanings:

(210) a. Hochu da dodjesh sutra u pet.

want1.sg. SUBJ come2.sg.PNP tomorrow at five

‘I want you to come tomorrow at five.’

b. Hochu da dolazish svaki dan.75 want1.sg. SUBJ come2.sg.IMPERF every day

‘I want you to come every day.’

This is, once again, straightforwardly accounted for given the analysis in (209).

At this point, therefore, we have a reasonably complete formal account of the syntactic properties associated with the typical subjunctive complements in the Balkan Slavic languages

75 Once again, even though PNP is the preferred verb form in Croatian subjunctives, the imperfective is not banned outright, and can be used when the subjunctive complement is clearly associated with imperfective aspectual readings, which is the case in (210b).

151

that were the main focus of my analysis so far. Once again, the basic structural description in (209) will be further refined and articulated later on in 3.3 once I focus on some finer syntactic and semantic differences that can be noticed between various subjunctive complements in the context of BlkS distribution more broadly. Before I move on to that part of the analysis, I will first briefly compare my syntactic account of the subjunctive particle da in Croatian with the one proposed in Todorovic (2012) for Serbian, which is the most detailed study devoted to this issue in recent literature. The reason why Todorovic’s study will be interesting in this context is because the author primarily focused on subjunctive complements that exhibit the property of obligatory control, which I haven’t looked at in detail until now, but which will receive a major amount of attention in the remaining parts of this chapter.

3.1.4.4 Socanac (2011) vs. Todorovic (2012): Different theoretical perspectives on the syntax