https://doi.org/10.53656/bel2022-5-3MD

2022/5, стр. 476 - 489

ON BULGARIAN DALI-QUESTIONS AND THE RELATION TO THE SUBJUNCTIVE MOOD

Резюме: The subjunctive mood has been widely discussed from the perspective its selection in complements of volitional, directive and emotivefactive predicates (Picallo 1984, Raposo 1985, Ambar 1988, a.o.). Subjunctive questions and other type of subjunctive main clauses, on the other hand, have not been subject to much systematic research. Some works (Ambar 2016, Giannakidou 2016) suggest that, in main clauses, the selection of the subjunctive mood codifies the expression of the speaker’s evaluations. In line with this view, we discuss the syntactic expression of the Bulgarian dali-questions which do not consist in true requests for information but rather display a flavour of wondering encoded in the occurrence of dali and its morphological make up combining the subjunctive particle da and the interrogative particle li. Considering the properties of the Balkan subjunctive particles, we discuss the syntactic expression of Bulgarian dali-questions building on previous analyses of Bulgarian yes-no questions (Dimitrova 2020).

Ключови думи: Subjunctive mood; dali-questions; Balkan subjunctive particles

1. Introduction

Our goal in this paper is to address the selection of the subjunctive mood in yesno questions and other types of main clauses focusing on data from both Balkan and Romance languages and on the case of the Bulgarian morphologically complex interrogative word dali considering its relation with the subjunctive mood.

The subjunctive mood has traditionally been addressed with respect to its obligatory selection in complements of volitional, directive and emotive-factive predicates. What is more, it has been regarded as the mood of subordination encoding a tense-dependency (Picallo 1984, Raposo 1985, Ambar 1988 a.o). By virtue of this dependency, the subjunctive embedded domain has been considered an anaphoric tense domain, which extends to the matrix clause, giving rise to obligatory obviation effects.

Nevertheless, some recent works (Marques 2010, Giannakidou 2016, Ambar 2016) have convincingly shown that the selection of the subjunctive and the indicative is more complex than it may seem at first glance, especially when it comes to the data from subjunctive main clauses. Importantly, the subjunctive main clauses present a strong argument against the traditional view according to which the subjunctive is dependent or defective. The evidence from subjunctive main clauses in fact shows that the subjunctive is able to occur on its own.

Additional evidence illustrating the complex properties of mood selection comes from some special predicates which can select both the indicative and the subjunctive (Marques 2009, 2010, a.o.). As shown by the data in (1) from Portuguese, even though only the subjunctive is felicitous in complements of volitional predicates, the indicative being ruled out, there exist predicates that are compatible with both the indicative and the subjunctive. One such example comes from Portuguese acreditar “believe”, as illustrated below:

(1) a. Eu acredito que a Ana ganhe as eleições.

I believe that the Ana wins-SUBJ the elections

b. Eu acredito que a Ana ganha as eleições.

I believe that the Ana wins-IND the elections

‘I believe that Ana will win the elections.’ (Marques 2010, 140)

According to Marques (2010), the selection of the subjunctive mood in (1a) conveys the speaker’s low degree of certainty with respect to the truth of the proposition. In contrast, the selection of the indicative mood in (1b) codifies the speaker’s high degree of certainty regarding the truth of the proposition.

Putting aside the general properties of the subjunctive mood to which we come back in Section 2 of the present work, another well-known fact about the subjunctive concerns the crosslinguistic variation with respect to its licensing. Romance languages display special verbal morphology for the subjunctive (Picallo 1984, Raposo 1985, Ambar 1988, Kempchinsky 2009, Quer 1998, Giannakidou 1998, a.o.). Balkan languages (Dobrovie-Sorin 1994, 2001, Krapova 2001, Giannakidou 2009, Roussou 2000 a.o.), on the other hand, display the so-called subjunctive particles, like Greek na and Bulgarian da. The properties of the subjunctive particles and the structural position they occupy have been subject to many discussions, especially when it comes to some intriguing facts like the strict verb-adjacency requirement they obey to and the temporal and aspectual restrictions they impose on the verb they occur with.

Considering the distribution and properties of the Balkan subjunctive particles, like Bulgarian da and comparing with their Romance counterparts, our goal in this paper is to address the subjunctive yes-no questions focusing on the syntactic expression of Bulgarian dali-questions.

Interestingly, dali combines the subjunctive particle da and the interrogative particle li. By virtue of its morphological make-up, dali conveys a flavour of wondering and doubt to the structure which distinguishes dali-questions, as in (2), from standard yes-no questions in which the interrogative particle li occurs, as in (3):

(2) Dali Ivan kupi knigata?

Dali Ivan bought.3p.sg. book.def

“Did John buy the book?” (I wonder)

(3) Ivan kupi li knigata?

Ivan bought.3p.sg. Q book.def

“Did Ivan buy the book?

In line with the previous studies dedicated on the properties of subjunctive main clauses and on the relation with the expression of the speaker’s evaluations and type of attitude (Ambar 2016, Giannakidou 2016), we will discuss an analysis for Bulgarian dali-questions considering the functional projection EvaluativeP (Ambar 2000, 2003) responsible for codifying the speaker’s evaluations and further building on the analysis of Bulgarian yes-no questions proposed in Dimitrova (2020).

The paper is organized as follows. In section 2 we discuss the subjunctive main clauses focusing on the existence of a relation with the expression of evaluation (Ambar 2016, Giannakidou 2016). In section 3 we focus on the Balkan subjunctive particles considering their distribution and main properties. In section 4, we discuss Bulgarian dali-questions comparing with standard yes-no questions licensed by the interrogative particle li. In section 5 we put forward a proposal for analysis of daliquestions. Section 6 concludes the paper.

2. Subjunctive main clauses and the expression of evaluation

As mentioned in section 1, the factors triggering the selection of the subjunctive in main clauses have not been studied in depth in the literature. Still, the data from main subjunctive clauses is particularly revealing when it comes to the properties of this mood. Importantly, such structures present a strong evidence that the subjunctive mood can occur on its own. Also, when it comes to discussing the selection of the subjunctive in main clauses, additional questions related to selection arise. Note that in subjunctive main clauses the selection of the subjunctive is clearly not triggered by the semantic properties of a matrix verb but rather appear to be related to given speaker-related factors.

Discussing the distribution of the indicative and the subjunctive mood, Ambar (2016) suggests that each mood is associated with the expression of a specific type of properties. Observe that, the indicative, though not the subjunctive, is felicitous is simple assertions like (4) below:

(4) a Ele vai ao cinema.

he go.IND.PRES.3sg to.the movies

‘He goes to the movies.’

b. *Ele vá ao cinema.

he go.SUBJ.PRES..3sg to.the movies (Ambar 2016, 125)

On the other hand, it is the subjunctive and not the indicative the mood contributing for the expression of evaluations, as in (5) below:

(5) a. Vá ele às aulas! (e /assim terá êxito nos seus estudos)

Go.SUBJ.PReS.3sg he to-the courses (and/so he will succeed in his studies)

‘Let him take the courses!’

b. *Vai ele às aulas!

Go.IND.PRES.3sg he to.the courses

‘He takes the courses!’ (Ambar 2016, 131)

The examples in (4) and (5) above are particularly important when it comes to the properties of the indicative and the subjunctive. In view of the clear-cut opposition between (4) and (5) and considering Giannakidou’s (1998) concept of nonveridicality, Ambar (2016) suggests that the properties of each mood, namely the type of illocutionary force each mood denotes, are captured under the projections AssertiveP and EvaluationP proposed in earlier works (Ambar 2000, Ambar 2003). Roughly, according to this analysis, the indicative-subjunctive interplay and the properties of each mood, in view of the type of predicates selecting them, are accounted for under the functional projections Assertive and Evaluative. Assertive is seen as the domain of the indicative, accounting for what the speaker knows (Ambar 2003). Considering the concept of (non)veridicality (Giannakidou 1998), Assertive can be regarded as the domain of veridicality codifying that the proposition is taken to be true. Evaluative, on the other hand, is the domain of nonveridicality, encoding the speaker’s evaluations and type of attitude with respect to the state of affairs described.

The evaluative flavour denoted in subjunctive main clauses is not new to the literature. When it comes to the case of subjunctive questions the properties distinguishing these structures from standard yes-no questions have been previously noticed in the literature. For instance, Rouchota (1994) on Greek subjunctive questions points out to the fact that these structures denote a flavour of doubt and wondering. Bîlbîie & Mardale (2015), considering Romanian subjunctive questions, propose that they are associated with the expression of different types of modality. Giannakidou (2016) discussing data from Modern Greek calls the type of subjunctive occurring in questions epistemic subjunctive and suggests that these uses of the subjunctive can be captured under the possibility modal might, as illustrated by the example below:

(6) Na tou milise (arage)?

SUBJV him talked-3SG Q-particle

‘Might she have talked to him? (Giannakidou 2016, 200)

In Giannakidou (2016) the properties of the subjunctive and the standard yes-no questions are discussed from the perspective of the denotation of the variables {p, ⌐p} (Hamblin, 1973). Giannakidou (2016) points out that, differently from standard yes-no questions that denote the opposition {p, ⌐p}, in subjunctive questions “the speaker asks the hearer about the possibility of p rather than p itself.” (Giannakidou 2016, 200).

This analysis can be furthermore extended to the data from Bulgarian subjunctive questions which, similarly to their Modern Greek counterparts, denote a flavour of wondering and doubt. With these observations, in section 5 we extend the discussion to Bulgarian dali-questions proposing that they consist in a case of the so called epistemic subjunctive (Giannakidou 2016). Moreover, we observe that the epistemic meaning denoted by dali-questions appears to be a result of the morphological make-up of the interrogative word dali which incorporates an instantiation of the subjunctive particle da and the interrogative particle li.

3. The Balkan subjunctive particles

In this section we focus on the properties and distribution of the Balkan subjunctive particles capitalising on data from Bulgarian. Balkan languages differ from Romance when it comes to the licensing of the subjunctive, as they rely on the insertion of the so called subjunctive particles like Bulgarian da, Modern Greek (henceforth MG) na and Romanian să.

Many works have been dedicated to the syntactic expression of the Balkan subjunctive clauses and to the categorial status of the subjunctive particles. The properties of the Balkan subjunctive particles are particularly intriguing when it comes to three important aspects: (i) verb adjacency (ii) Tense and aspect and (iii) obviation effects. In this section we discuss the aspects in (i)-(iii) considering the distribution of the Bulgarian subjunctive particle da and comparing, when possible, with data from other languages displaying subjunctive particles.

3.1. Main properties of the Balkan subjunctive particles

A first important aspect of the characterisation of the Balkan subjunctive particles concerns the fact that these elements are obligatorily verb-adjacent. As shown by the examples in (7) from Bulgarian, nothing but pronominal clitics and negation (as in (7b)) can intervene between da and the verb:

(7) a. Iskam Ivan da (*Ivan) kaže istinata.

Want1p.sg Ivan SUBJ Ivan say.PERF.PRES.3p.sg truth.def

“I want Ivan to say the truth.”

b. Iskam Ivan da ne mu kazva istinata.

Want1p.sg Ivan SUBJ Neg cl.dat say.IMPERF.PRES. truth.def

“I want that Ivan doesn’t tell him the truth.”

Obligatory verb-adjacency is particularly revealing when it comes to the structural position occupied by the Balkan subjunctive particles. Discussing similar data from MG, Giannakidou (2009) takes the strict verb-adjacency requirement to argue against the proposal put forward in Roussou (2000) under which Modern Greek subjunctive particle na is a complementiser. In Roussou’s (2000) terms the fact that na and the indicative complementiser oti do not cooccur suggests that they compete for the same structural position. Nevertheless, as shown in Giannakidou (2009), the subjunctive particle na and the indicative complementiser oti sharply diverge when it comes to allowing intervening material, as this is only possible with the indicative complementiser oti, though not with na.

Interestingly, Giannakidou (2009) further observes that MG subjunctive particle na is restricted with respect to the Tense and Aspect of the verb it occurs with. According to this author, MG na is only compatible with verbs in Perfective Present or, as dubbed in Giannakidou (2009), Perfective Non-Past (henceforth, PNP).

Curiously, PNP patterns Romance subjunctive when it comes to its dependent nature. Note that, PNP is unable to occur on its own:

(8) a. {Na/as} to pis.

SUBJ it say.PNP.2p.sg

‘You may say it.’

b. *To pis. (Giannakidou 2009, 1885)

According to Giannakidou (2009), PNP is referentially deficient and, therefore, needs a particle, like MG na, that introduces the variable now (n) into the syntax. Bulgarian patterns MG regarding the tense restrictions of the verb the particle da occurs with: Bulgarian da occurs with verbs in Perfective Present, or PNP as in (9a), or with verbs in Imperfective Present, as in (9b). As illustrated by (9c), Bulgarian da is incompatible with verbs in Past or Future tenses:

(9) a. Iskam Marija da napiše kniga.

Want.1p.sg. Mary SUBJ write.PNP.3p.sg book.

‘I want Mary to write a book.’

b. Iskam Marija da piše knigi.

Want.1p.sg Mary SUBJ write.IMPERF.3p.sg. books ‘I want Mary to write books.’

c. Iskam Marija da pišeše / napisa

Want.1p.sg Mary SUBJ write.PAST.IMPERF.3sg. /write. PAST.PERF.3sg šte napiše knigi .

FUT write.3sg books

In addition, the data from Bulgarian also shows that PNP is unable to occur on its own (10a) or under the scope of the indicative complementiser če (10b), patterning MG in (8) above:

(10) a. *Kaže istinata.

say. PNP truth.def

b. *Znam če kaže istinata

know.PRES.1p.sg that say. PNP truth.def

Another property of the Balkan subjunctive clauses we discuss here concerns obviations effects. As discussed in section 1, obligatory obviation in Romance has been regarded as a result of tense-defectiveness: as subjunctive clauses have [-T] on C, the binding domain of the embedded clause is extended to the matrix clause preventing the co-reference between the embedded and the matrix subject. Nevertheless, differently from Romance, Balkan subjunctive clauses allow the co-reference between the matrix and the embedded subject and therefore void obligatory obviation:

(11) a. Ivan iska Marija da spečeli.

John wants Mary SUBJ win.PNP.3p.sg

‘John wants Mary to win.’

b. Ivan iska da spečeli.

John wants SUBJ win.PNP.3p.sg

‘John wants to win.’

c. Ivan se opita da spečeli.

John REFL tried SUBJ win.PNP.3p.sg.

‘John tried to win.’

The data in (11) illustrates that Bulgarian da-clauses pattern both Romance subjunctive and infinitival clauses Observe that obviation is voided in (11b) and therefore the subjects of the matrix and the embedded clause are co-referent. What is more, da-clauses further occur in obligatory control structure like (11c): a context in which infinitival clauses occur in Romance.

3.2. Analysing the Balkan subjunctive clauses

The structural position of the Balkan subjunctive particles is a complex matter that has been subject to many discussions. On the one hand, they behave as complementisers (Roussou 2000), on the other, they obey to a strict verb-adjacency requirement which may suggest that they are generated in a lower structural position. According to Giannakidou (2009), oti is the true complementiser, while na heads the functional projection MoodP situated above NegP.

Another line of enquiry according to which the Balkan subjunctive particles are generated in a position below the CP is the one proposed in Dobrovie-Sorin (1994). Dobrovie-Sorin (1994) proposes that the Romanian subjunctive particle să is part of the verbal inflection incorporating into the verbal cluster. Dobrovie-Sorin (1994) moreover points out that, even though the analysis under which Romanian să is part of the verbal inflection fully captures the inflection-like behaviour of this element, it fails to account for its complementiser-like behaviour, namely for the fact that să precedes negation and heads embedded clauses. Considering the dual nature of Romanian să which displays both inflection and complementiser properties, Dobrovie-Sorin (1994) proposes that să is generated in an XP position above Tense (or in the CP) but it merges with the verb via incorporation: a proposal that accounts for its dual nature.

Another attempt to capture the different occurrences of the Balkan particles was made in Dimitrova (2020) discussing the case of the Bulgarian subjunctive da. Considering the behaviour of the Balkan particles which pattern Romance subjunctive and infinitival structures, three types of da-clauses have been distinguished: low, medium and high da-clauses. In broad terms, according to this work, low and medium da-clauses capture those occurrence in which da-clauses pattern Romance infinitival structures and occur with, respectively PNP and the Imperfective Present. High da-clauses, on the other hand, have been considered the true subjunctive da-clauses. According to this analysis, high da-clauses involve higher raising to Evaluative triggered by an unvalued [ueval] feature valued by an Op, responsible for the expression of the speaker’s evaluations (Ambar 2003, 2016, Giannakidou 2016).

In the present work, we will follow the analysis proposed in Dimitrova (2020) in suggesting that subjunctive questions and dali-questions involve raising to Evaluative.

4. On subjunctive dali-questions and standard yes-no questions

In view of the intriguing behaviour of the Balkan subjunctive particles, in this section we will focus on the distribution of the interrogative word dali and its relation to the subjunctive mood comparing with standard yes-no questions, licensed by the interrogative particle li (Rudin et al. 1999, Izvorski 1995, Dimitrova 2020).

Ever since Rudin (1986) it has been noticed that, differently from standard yes-no questions licensed by the particle li, dali-questions denote a strong flavour of wondering and doubt. Nevertheless, dali has traditionally been considered an interrogative complementiser occurring in matrix (12) and embedded (13) questions:

(12) Dali Ivan kupi knigata?

Dali Ivan bought.3p.sg. book.def

“Did John buy the book?” (I wonder)

(13) Popitax dali Ivan kupi knigata.

Asked.1sg dali John bought.3sg book.def

“I asked whether John bought the book.”

Considering embedded questions like (13) above, Krapova (2002) proposes that Bulgarian dali patterns the Italian interrogative complementiser se and is generated in Intº building on the analyses of Rizzi (1997, 1999).

Even though the analysis of Krapova (2002) correctly captures the properties of dali embedded clauses, it fails to account for the fact that dali morphologically encodes an instantiation of the subjunctive particle da. As shown in Smirnova (2011), dali is also plausible in the scope of negated predicates, such as ne săm sigurna ‘I am not certain’ or ne e očevidno ‘it is not obvious’:

(14) a. Ne săm sigurna [da-li ima teč v rezervoara].

NOT be.1SG.PRES certain SUBJ-Q have.3SG.PRES leak in oil.tank

‘I’m not certain whether there is a leak in the oil tank.

b. Ne e očevidno [da-li ima teč v rezervoara].

NOT be.3SG.PRES obvious SUBJ-Q have.3SG.PRES leak in oil.tank

‘It is not obvious whether there is a leak in the oil tank.’

Smirnova (2011, 274)

According to this author, the employment of dali in (14) correlates with the expression of epistemic modality and with the speaker’s low level of commitment to the truth of the proposition. Curiously, in the examples in (14) above, the subjunctive dali can be substituted by the indicative complementiser če:

(15) a. Ne săm sigurna [če ima teč v rezervoara].

NOT be.1SG.PRES certain that have.3SG.PRES leak in oil.tank

‘I’m not certain that there is a leak in the oil tank.’

b. Ne e očevidno [če ima teč v rezervoara].

NOT be.3SG.PRES obvious that have.3SG.PRES leak in oil.tank

‘It is not obvious that there is a leak in the oil tank.’

In the examples in (15) with the indicative complementiser če the speaker considers that the proposition in the embedded domain is true. The opposition between the examples in (14) and (15) thus resembles the contrasts stemming from the selection of the indicative and the subjunctive under predicated like Portuguese acreditar “believe” discussed in section 1 (Marques 2009, 2010). In our view, these parallels are not coincidental.

4.1. The particle li and the licensing the yes-no questions in Bulgarian

As discussed above, in our view, the properties of the interrogative word dali and the relation to the subjunctive mood it displays stem from its morphological make up and from the fact that it is a morphologically complex element that combines the subjunctive particle da and the interrogative particle li. With this in mind, in this section we will briefly discuss the properties of Bulgarian yes-no questions focusing on the distribution of the interrogative particle li which is the element responsible for the licensing of Bulgarian yes-no questions.

Two types of li-questions have been distinguished: (i) the neutral V-li questions which li follows the verb, as in (16a), and (ii) the focus XP-li questions in which li follows an XP different from the verb, as in (16b).

(16) a. Ivan kupi li knigata?

Ivan bought.3p.sg. Q book.def

“Did Ivan buy the book?

b. Ivan li kupi knigata?

Ivan Q bought.3p.sg. book.def

“Did IVAN buy the book?”

XP-li questions, such as (16b), have been traditionally associated with focusassignment: the XP that precedes the particle is contrastively focalised (Rudin 1986, Rudin et al. 1999, Izvorski 1995). In fact, the distinction between neutral and focus yes-no questions, their syntactic expression and the structural position occupied by the particle li have been subject to many discussions. Rudin et al (1999) argue that the particle is a complementiser generated in Cº. Izvorski (1995), on the other hand, suggests that li is merged in Focº.

In more recent works, namely Dimitrova (2020), it has been suggested that the particle li introduces the algorithm [x, ⌐x], much as in Hamblin (1973), in which [x] can be the verb or an XP different from the verb. What is more, following Holmberg (2012), Dimitrova (2020) argues that the syntactic expression of Bulgarian yesno questions involves a polarity head PolP, situated above TP, responsible for the denotation of polarity features. Following this line of enquiry, it has been proposed that Bulgarian neutral li-questions display the structure below:

(17) [IntP [Intº [uPol], [uV] [PolP [Polº li [x, ⌐x] [TP [Tº [vP [vº [VP [Vº

In the present work, we will follow the analysis of Bulgarian li-questions proposed in Dimitrova (2020) in assuming that the particle li is generated in PolP denoting a polarity algorithm [x, ⌐x].

4.2. On dali-questions and polarity

Following the analysis proposed in Dimitrova (2020) in assuming that PolP indeed projects in yes-no questions, we can observe that li-questions and daliquestions diverge with respect to the denotation of the alternatives. Observe that, in contrast to standard yes-no questions licensed by li, dali-questions appear to be infelicitous with the coda “or not”:

(18) a. Ivan kupi li knigata ili ne ?

Ivan bought.3p.sg. book.def or not

“Did John buy the book or not?”

b. ?? Dali Ivan kupi knigata ili ne?

Dali Ivan bought.3p.sg. book.def or not

The intriguing data in (18) showing that li-questions are compatible with the coda “or not” is accounted for under Dimitrova’s (2020) proposal according to which the particle li denotes the polarity algorithm [x, ⌐x]. On the other hand, the fact that dali-questions are incompatible with the coda “or not” seems to lay further support to the idea that dali displays a relation to the subjunctive. As mentioned in section 2, Giannakidou (2016) discussing MG subjunctive questions proposes that, differently from standard yes-no questions denoting the set of propositions [p, ⌐p], subjunctive yes-no questions are ‘about the possibility of p, rather than p itself’ (Giannakidou 2016, 200).

Following Giannakidou’s analysis, in the present work we claim that, differently from standard yes-no questions licensed by li, dali-questions rather denote the possibility of p triggered by the occurrence of the subjunctive particle da.

5. Towards an analysis of Bulgarian dali-questions

In this section we put forward a proposal for analysis of Bulgarian subjunctive yes-no questions with dali building on the proposals made in Ambar (2016) and Dimitrova (2020) and assuming with Giannakidou (2016) that, like MG subjunctive questions, Bulgarian dali-questions do not display the set of alternative propositions [p, ⌐p] but rather denote the possibility of p.

Note that, despite the fact that dali morphologically encodes an instantiation of the subjunctive particle da, there appear few important differences between dali-clauses and da-clauses, namely when it comes to the properties of Tense and Aspect. As discussed in section 3, the Bulgarian subjunctive particle imposes some strong restrictions on the Tense of the verb it occurs with, as it is only compatible with verbs is Perfective Present, or PNP (Giannakidou 2009) and Imperfective Present. Curiously, in contrast to da-clauses, dali does not impose any restrictions with respect to Tense. As shown in (19), dali is compatible with future, past and present tenses in contrast to da-clauses:

(19) Dali Ivan kazvaše / kaza

Dali Ivan say.IMPERF.PAST.3p.sg. / say.PAST.PERF.3p.sg. šte kaže istinata?

FUT say.PERF.PRES.3p.sg. truth.def.

What is more, another contrast between da-clauses and dali-clauses appears with respect to the defective form PNP. As opposed to da-clauses, the morphologically complex dali appears to be incompatible with the defective PNP similarly to the indicative complementizer če in (10b) above:

(20) *Dali Ivan kaže istinata?

Dali Ivan say.PNP truth.def

Considering that PNP is a deficient form that needs a particle, like the Bulgarian da or MG na (Giannakidou 2009), it seems that, in da-clauses, the particle da is responsible for the valuation of given tense-features of the defective PNP. Nevertheless, the ungrammaticality of the dali-question with PNP in (20) can be felicitously accounted for by virtue of the incorporation of the particle da to the interrogative particle li. As da incorporates with li in dali-questions, the particle does not value the unvalued tense features of PNP and thus gives rise to ungrammatical structures like (20).

Considering all intriguing observations so far and the proposal put forward in Dimitrova (2020) regarding the syntactic structure of Bulgarian yes-no questions, it can be suggested that the morphologically complex dali is formed by external merge of the subjunctive particle da to the interrogative particle li in Polº giving rise to what has been defined as epistemic subjunctive in yes-no questions. Adopting the structure in (17) above and assuming that the element that attaches to li in Polº is the subjunctive particle da, the polarity algorithm [x, ⌐x] li denotes is not absorbed by the verb, as in neutral yes-no questions but rather acts on the subjunctive particle da. By virtue of the way the polarity algorithm applies, such an analyses correctly captures the fact that, like other subjunctive questions, Bulgarian dali-questions convey the possibility of p rather than the opposition [p, ⌐p] (Giannakidou 2016).

In view of the proposal put forth in Ambar (2016) and considering that subjunctive clauses denote the speaker’s evaluations and kind of attitude, it can be further suggested that EvaluativeP projects in Bulgarian dali-questions and that the existence of an evaluative Op in Spec, EvaluativeP triggers higher raising of dali to Evaluativeº. This hypothesis will be further explored in future works.

6. Conclusion remarks

Our goal in this paper was to shed some light on the subjunctive yes-no questions considering the case of the Bulgarian interrogative word dali and its relation to the subjunctive mood. Considering the intriguing properties of the Balkan subjunctive particles and building on previous analyses of the Bulgarian yes-no questions (Dimitrova 2020), we suggested that the complex dali is formed by external merge of the subjunctive particle da to the interrogative particle li in Polº.

Still, many additional questions with respect to the syntactic expression of the Balkan subjunctive clauses and the structural position of the subjunctive particles remain unsettled, especially when it comes to the ambiguity between Romance subjunctive and infinitival structures they denote. We leave the thorough discussion of these topics for future research.

REFERENCES

AMBAR, M. 1988. Para uma Sintaxe da InversŠo Sujeito-Verbo em Português: PhD dissertation: University of Lisbon.

AMBAR, M. 2000. Wh-questions and wh-exclamatives – unifying mirror effects. In C. Beyssade, R. Bok-Bennema, F. Drijkoningen, P. Monachesi (eds): Romance languages and linguistic theory 2000: Selected papers from ‘Going Romance’ 2000. Amsterdam/Philadelphia: John Benjamin, 15 – 40.

AMBAR, M. 2003. Wh-asymmetries. In Anna Maria Di Sciullo (ed.): Asymmetry in Grammar. Amsterdam, USA: John Benjamins, 208 – 249.

AMBAR, M. 2016. On Finiteness and the Left Periphery: Focusing on Subjunctive. In Blaszack, J. Klimek-Jankowska, D. Mygdalski, K. Giannakidou, A. (eds.). Mood, Aspect, Modality Revisited: New Answers to Old Questions. Chicago: University of Chicago Press. 125 – 176.

BÎLBÎIE, G. and MARDALE, A. 2018. The Romanian Subjunctive from a Balkan perspective. In Krapova, I. and Joseph, B. (eds): Balkan Syntax and (Universal) Principles of Grammar. Berlin/Boston: Mouton de Gruyter, 278 – 314.

DIMITROVA, M. 2020. On the Syntax of Yes-no questions in Bulgarian and Portuguese. PhD dissertation: University of Lisbon.

DOBROVIE-SORIN, C. 1994. The Syntax of Romanian. Berlin: Mouton de Gruyter.

DOBROVIE-SORIN, C. 2001. Head-to-Head merge in Balkan subjunctives and locality. In Rivero M. L. and Ralli, A. (eds): Comparative Syntax of Balkan languages. New York: Oxford University Press, 44 – 73.

GIANNAKIDOU, A. 1998. Polarity Sensitivity as (Non)veridical Dependency. Amsterdam-Philadelphia: John Benjamins.

GIANNAKIDOU, A. 2009. The dependency of the subjunctive revisited: Temporal semantics and polarity. Lingua 120, 1883 – 1908.

GIANNAKIDOU, A. 2016. Evaluative subjunctive and nonveridicality. In Blaszack, J. Klimek-Jankowska, D. Mygdalski, K. and Giannakidou, A. (eds): Mood, Aspect, Modality revisited: New Answers to Old Questions. Chicago: University of Chicago Press, 177 – 217.

HAMBLIN, Ch. 1973. Questions in Montague English. Foundations of Language 10, 41 – 53.

HOLMBERG, A. 2012. On the Syntax of Yes and No in English. Newcastle Working Papers in Linguistics 18, 52 – 72.

HOLMBERG, A. 2016. The syntax of Yes and No. Oxford: Oxford University Press.

IZVORSKI, R. 1995. Wh-movement and focus movement in Bulgarian. In R. Eckardt V. van Geenhoven (eds.). ConSole II Proceedings. The Hague: Holland Academic Graphics, 54 – 67.

KEMPCHINSKY, P. 2009. What can the subjunctive disjoint reference effect tell us about the subjunctive? Lingua 119, 1788 – 1810.

KRAPOVA, I. 2001. Subjunctives in Bulgarian and Modern Greek. In Rivero M. L. and Ralli, A. (eds): Comparative Syntax of Balkan Languages. New York: Oxford University Press, 105 – 126.

KRAPOVA, I. 2002. On the Left Periphery of the Bulgarian sentence. University of Venice Working Papers in Linguistics 12, 107 – 128.

MARQUES, R. 2010. On the selection of mood in complement clauses. In Hogeweg et al. (eds.): Crosslinguistic Semantics of Tense, Aspect and Modality. Amsterdam and Philadelphia: John Benjamins, 179 – 204.

PICALLO, C. 1984. The infl node and the null subject parameter. Linguistic Inquiry 15, 75 – 102

QUER, Jo. 1998. Mood at the Interface. PhD dissertation: University of Utrecht

RAPOSO, E. 1985. Some asymmetries in the binding theory in Romance. Linguistic Review 5, 75 – 109.

ROUCHOTA, V. 1994. The semantics and pragmatics of subjunctive in Modern Greek. PhD dissertation: University College London.

RUDIN, C. 1986. Aspects of Bulgarian Syntax: Complementizers & Wh Constructions. Slavica Publishers.

RUDIN, C, KRAMER, C., BILLINGS L. and BAERMAN, M. 1999. Macedonian and Bulgarian li Questions: Beyond Syntax. Natural language and Linguistic Theory 17, 541 – 586.