Predicate Clefts in Bulgarian

Full text

Predicate clefts in Bulgarian*
Elena Karagjosova Katja Jasinskaja
Stuttgart University Cologne University
1 Introduction
Predicate cleft constructions (PCCs) have been observed in a number of lan-
guages, including Slavic languages (cf. e.g. Abels 2001, Aboh and Dyakonova
2009 on Russian; Bondaruk 2009 on Polish). According to Abels (2001), whose
terminology we adopt here, a PCC consists of a HEAD and a BODY. The HEAD is a
phrase, usually infinitival. The BODY is a finite clause whose main verb repeats
the main verb of the HEAD. In languages where this construction results from VP
or V movement, one refers to it as predicate fronting with doubling, where the
HEAD contains the fronted copy of the verb, and the BODY the lower copy. An ex-
ample of a Russian PCC is given in (1), where the first clause is the predicate
cleft, and the second is a typical adversative context in which PCCs occur.
(1) [Rabotat’ (to)]HEAD [on rabotaet]BODY, no ničego ne zarabatyvaet.
work.inf prt he work.3sg but nothing not earns
‘As for working, he does work, but he doesn’t earn anything.’
The goal of this paper is to describe predicate clefts in a language that lacks
infinitives – Bulgarian. What does a predicate cleft look like if fronting an infini-
tival phrase is not an option? We provide an account of previously undescribed
Bulgarian PCCs that are unique in their range of options for the morphosyntactic
realization of HEADs. More specifically, we argue that Bulgarian has three differ-
ent ways of structuring the HEAD of a predicate cleft: za-PCCs, če-PCCs, and da-
PCCs. The three constructions exhibit subtle differences in their pragmatic func-
tion and have different restrictions on use. We show how to derive those differ-
ences from the differences in the structure of the HEAD. We argue, moreover, that
at least in the case of če-and da-PCCs, the construction is not derived by move-
ment, unlike Russian and Polish PCCs. Finally, we raise the questions: Which
parts of the HEAD are doubled, and which are not doubled in the BODY, and why?
Previous answers to these questions given for other languages are based on the
nature of movement, and therefore are not applicable to the base-generated
* The research presented here has profited a lot from the second author’s guest stay at the
ACLC, University of Amsterdam. The research was supported by the German Research
Foundation (Grant Nr. BE 4348/2-1, ZAS, and the SFB 732, University of Stuttgart) and
the Emerging Group Dynamic Structuring in Language and Communication funded
through the Institutional Strategy of the University of Cologne (ZUK 81/1).

2 Elena Karagjosova and Katja Jasinskaja
PCCs in Bulgarian. We observe a common structural pattern in the BODY of Bul-
garian PCCs on the one hand and in denials on the other and argue that in both
cases the material that is doubled is the material that cannot be omitted for in-
formation-structural reasons and due to constraints on VP-ellipsis.
We first introduce the 3 types of Bulgarian PCC (section 2). Then we dis-
cuss the pragmatic function and the semantic properties of the HEADs of če-and
da-PCCs (section 3). Finally, in section 4 we address the structure of Bulgarian
če-and da-PCCs and present our explanation of the doubling pattern.
2 Predicate clefts in Bulgarian
The first type is the za-PCC, exemplified in (2), where the first sentence repre-
sents the context for the PCC:
(2) A kak sa dečizata s govoreneto, momičeta? (Internet)
‘And how are the kids doing with talking, girls?’
[Za govorene]HEAD -[ govori]BODY, no mnogo nepravilno.
as for talking talk.3sg but very incorrectly
‘As for talking - (s)he talks, but very incorrectly.’
The HEAD of a za-PCC consists of a PP headed by the preposition za (‘as
for’) and a verbal noun which may be either definite or indefinite. The BODY con-
tains the tensed copy of the verb from which the verbal noun is derived. The
PCC is followed by an adversative clause as is typical for PCCs in other lan-
guages (cf. Abels 2001). The arguments and adjuncts can in general be realized
in the BODY or in the HEAD of the za-PCC.
The second type is the če-PCC, cf. (3):
(3) Vsičko e mnogo obârkano. (BulNC1)
‘This is a complete mess.’
[Če e obârkano]HEAD, [obârkano e]BODY Ne tova e važnoto.
that is.3sg confused confused is Not this is important.def
‘A mess it is. This is not what matters most.’
The HEAD of a če-PCC consists of a full CP headed by the complementizer če
(‘that’) and a VP (here consisting of the verb ‘to be’ plus an adjectival predi-
cate). The entire VP is copied in the BODY. Again, the PCC is followed by a con-
trasting utterance. In contrast to za-PCCs, arguments and adjuncts are normally
realized in the HEAD.
The third type is the da-PCC, as in (4):
1 Bulgarian National Corpus, URL: http://dcl.bas.bg/bulnc/home/?lang=en

Predicate clefts in Bulgarian 3
(4) Zašto Toromanova si šie takiva prosti rokli? (BulNC)
Why does Toromanova have sewn such simple dresses?’
[Da e ot skromnost]HEAD [ ne e]BODY. A može bi veče ne
to is of modesty not is and maybe yet not
im stigat parite?
them suffice money.def
‘Out of modesty - no. Maybe they are getting short of money?’
The HEAD of a da-PCC is a full CP2 headed by complementizer da (‘(in or-
der) to’) and the finite verb is copied in the BODY. Again, a contrasting utterance
follows. As with če-PCCs, all arguments and adjuncts are normally realized in
the HEAD, whereas the BODY only contains a copy of the lexical verb and/or the
auxiliary and sometimes the pronominal clitics.
In the rest of the paper we concentrate on če-and da-PCCs.
3 Contrastive topic and the structure of the HEAD
In general, the function of a PCC is to mark the VP as either a focus or a con-
trastive topic. When it marks focus, as in some West-African languages, the
PCC can be roughly paraphrased as an it-cleft, e.g. It is [working]F that he does.
(cf. Lefebre 1992). In Russian and Polish (and many other languages, e.g. Span-
ish, Hebrew, and Korean), the PCC is assumed to mark contrastive topic (hence-
forth CT, cf. Abels 2001): the HEAD contains a CT-exponent, while the BODY con-
tains a focus exponent: [As for working]CT [he works]F. CT on the PCC-HEAD is
interpreted semantically in terms of a set of alternative questions called its topic
semantic value (Büring 1995), e.g. the topic semantic value for (1) is the set of
questions {Does he work?, Does he earn money?, ...?}. The question set indi-
cates at the same time a discourse strategy, i.e. a hierarchy of questions and an-
swers to them along which the discourse is structured (Büring 2003).
The focus in the BODY of a PCC is interpreted in terms of a set of alternative
answers to one of the topic questions (focus semantic value, Rooth 1992). In
general, the focus exponent in the BODY of a PCC tends to be polarity focus (e.g.
in Russian, Abels 2001) or verum focus3 (e.g. Brazilian Portugese, Hungarian,
Mandarin Chinese, Spanish, cf. Vicente 2009), but in some languages constitu-
ent focus is also possible (e.g. Hungarian, Vicente 2007). These types of focus
2 The status of complements containing da-forms is a subject of ongoing debate: the
traditional view of da as complementizer is challenged by the view that it belongs to the
verbal morphology, cf. e.g. Smirnova (2011, 185). For our purposes, however, the
precise status of da is not of crucial importance.
3 Focus on the truth (or falsity) of a proposition, cf. Höhle (1992).

4 Elena Karagjosova and Katja Jasinskaja
differ with respect to the type of alternatives they evoke and are related to dif-
ferent types of topic questions (polar, verum-4, or constituent questions):
(5) {he works, he doesn’t work} polarity focus
he worksF: {it’s true that he works, it isn’t true that he works} verum focus
{he works, he sleeps, he watches TV, ...} constituent focus
The CT-analysis applies in general to all three Bulgarian PCCs. We argue,
however, that the constructions differ in terms of the type of questions and an-
swers they involve. While the nominalized verbs in the HEAD of za-PCCs are
comparable to Russian infinitives, da- and če-clauses exhibit semantic properties
that influence the interpretation of the CT and lead to different restrictions on
the polarity of the HEAD and the BODY, as detailed in the following sections.
3.1 če-PCCs
We start with the observation that če-PCCs obey particular restrictions with re-
spect to the polarity of the two parts of the construction: they require the same
polarity in the HEAD and the BODY. Thus whereas the sentences in (6a, b) are well-
formed, the cross-over combinations (6c, d) are unacceptable:5
(6) a. Če čete, čete. b. Če ne čete, ne čete.
that read.3sg read.3sg that not read.3sg not read.3sg
‘As for reading, he does read.’ ‘As for reading, he does not read.’
c. *Če čete, ne čete. d. *Če ne čete, čete.
that read.3sg not read.3sg that not read.3sg read.3sg
This restriction cannot be attributed to the interpretation of CT: in principle,
both positive and negative topic questions are possible; moreover, the polarity of
the answer is independent from the polarity of the question. On the other hand,
the semantics of the če-HEAD offers a natural explanation for this restriction.
It has been observed that the complementizer če expresses factivity, whereas
da expresses non-factivity (cf. e.g. Pitsch 2010, 316-318 and the literature men-
tioned there). Pitsch notes that this property of če is related to a general differ-
ence between factive (transparent) and non-factive (opaque, intensional) com-
plement sentences (observed in e.g. Kiparski and Kiparski 1970; Maienborn
4 I.e. questions asking whether a proposition is true and should be added to the Common
Ground or not, cf. Romero and Han (2004) on yn-interrogatives with preposed negation.
5 For comparison, the polarity of the za- HEAD is always positive, which may be related to
the fact that negating verbal nouns in Bulgarian with the negation prefix ne- does not
always yield acceptable results, cf. nepušene (‘non-smoking’) vs. ??neudrjane (‘non-
hitting’). Note that a negative za- HEAD is only possible with negative polarity of the BODY.

Predicate clefts in Bulgarian 5
2003), where only the former yield a reality reading of the proposition. A more
cautious view advocated in Smirnova (2011) is that the difference between če
and da should be analysed in terms of the indicative-subjunctive opposition. In
any case, there seems to be a clear preference for če to combine with predicates
expressing factivity (factive or semi-factive predicates, cf. Karttunen 1971),
whereas predicates expressing non-factivity tend to combine with da.6 For in-
stance, whereas znaja (‘know’) and viždam (‘see’) combine with če, iskam
(‘want’) and želaja (‘wish’) do not. On the other hand, while some non-factive
predicates like mislja (‘believe’), kazvam (‘say’), or ne e vjarno (‘is not true’)
also combine with če, as in (7a), this is not possible if the če-clause is fronted,
whereas fronting with a factive predicate is perfectly acceptable (7b):
(7) a. Mislja/ Kazvam/Ne e vjarno, če čete.
think.1sg/ say.1sg/ not is true that read.3sg
‘I think/I say/It is not true that he reads. ’
b. Če čete, *milsja/ *kaz vam/*ne e vjarno/znam/ viždam/e vjarno.
that read.3sg think.1sg/say.1sg/ not is true / know.1sg/see.1sg/ is true
‘That he reads, I think/I say/it is not true/I know/I see/it is true.’
This observation suggests that če’s general preference towards factive predi-
cates becomes even stronger when the če-clause is fronted. The association of če
with factive predicates in fronted clauses leads to a factive interpretation of the
če-complement. This is also what seems to happen with the HEAD of the če-PCC,
in spite of the fact that the clause representing the HEAD of a če-PCC is obviously
unembedded. A possible explanation for this is offered by the observation that
če-PCCs seem semantically and pragmatically equivalent to fronted če-clauses
embedded under the predicate it is true: both receive a CT-interpretation (8a); in
addition, the BODY of the PCC and the embedding predicate offer equivalent an-
swers to a question about the truth value of a proposition (8b):
(8) a. [Če čete]CT, četeF/e vjarnoF, no ništo ne razbira.
that reads reads/is true but nothing not understands
‘That he reads, he reads/is true, but he doesn’t understand anything.’
b. A: Vjarno li e, če čete? B: Čete./ Vjarno e.
true prt is that read read/ true is
‘Is it true that he reads?’ ‘He reads./It is true.’
The factivity reading7 of the če-HEAD explains why the BODY needs to have the
same polarity: if not, the two parts of the PCC convey contradicting information,
6 Cf. also Krapova (2001) where da is claimed to be the subjunctive marker in Bulgarian.
7 We leave here the question of its precise semantic status open.

6 Elena Karagjosova and Katja Jasinskaja
and the utterance becomes inconsistent. It also has an effect on the interpretation
of the CT on the HEAD of the construction: CT on the factive HEAD of the če-PCC
leads to evoking verum-questions, i.e. questions of the type ‘Is the proposition
(really) true?’, rather than of the type ‘Is the proposition true or false?’. Thus in
(3), CT on the če-clause raises the set of verum-questions {Is it true that it is a
mess?, Is it true that this is what matters most?,...?}. A verum-question may be
given either a verum-answer, i.e. a sentence containing verum focus or an epis-
temic adverb like really (cf. Romero and Han 2004)8, or a non-verum answer:9
thus, a question like Is it true that Peter reads the book? seems equally appro-
priately answered by (Yes,) He READS it.; He DOES read it.; or He REALLY reads it.
We argue nevertheless that the focus in the BODY of če-PCCs is verum focus.
Like plain polarity focus, verum focus surfaces usually as an accent on polarity
elements, negation or auxiliary, but may be also on the main verb. This is also
the case in Bulgarian, where verum focus is not always formally distinguishable
from plain polarity focus. A reliable test is therefore the insertion of focused
naistina (‘really’), which in the case of če-PCCs yields a perfectly acceptable
utterance,10 cf. Če govori – NAISTINA govori. (‘As for talking – he REALLY talks.’).
The difference between plain polarity focus and verum focus is functional:11
while both give rise to a set of polarly opposed focus alternatives, plain polarity
focus fixes the truth value of a proposition, whereas verum focus targets an al-
ready established truth value for the proposition in the given discourse. The BODY
of a če-PCC seems therefore a natural environment for verum focus: while the
HEAD suggests that the proposition is true, CT on the HEAD raises the question
whether the proposition is really true. I.e., the truth value of a proposition that is
already proposed in the discourse is being questioned.12 The clause contrastively
attached to a če-PCC may but need not involve verum-interpretation, since
verum-questions allow non-verum answers, as argued above.13
8 Romero and Han (2004, 122) argue that epistemic really, verum focus and yn-
interrogatives with preposed negation introduce the conversational epistemic operator
VERUM whose role is to assert that the speaker is certain that p should be added to the
common ground.
9 And vice versa, a verum-answer may be given to a non-verum polar question, e.g. Does
Peter read the book? - Če čete, čete, no ništo ne razbira. (‘He DOES read it, but.... ’)
10 For comparison, focused naistina is not appropriate in da- or za-PCCs, cf. #Za govorene
- NAISTINA govori, #Da govori -NAISTINA ne govori.
11 Cf. also Romero and Han (2004) who point out that polarilty focus may have different
functions in declaratives, such as contrastive focus or verum focus.
12 That the če-BODY involves verum is also suggested by the observation that it can be
paraphrased as “is true”, cf. (8a). This paraphrase does not apply for the other two PCCs.
13 Cf. Če e obârkano, obârkano e. No sâšto taka e vjarno, če ne tova e važnoto. (‘A mess it
is. But what is also true is that this is not what matters most.’) Such verum-continuations

Predicate clefts in Bulgarian 7
3.2 Da-PCCs
Da-PCCs exhibit a different set of restrictions with respect to the polarity of the
two parts of the construction: the HEAD must have positive polarity, and the BODY
a negative one. Thus only (9a) is well-formed, the rest is unacceptable:
(9) a. Da čete, ne čete. b. *Da čete, čete. c. *Da ne čete, (ne) čete.
to read not read to read read to not read (not) read
‘As for reading, he doesn’t read.’
As in če-PCCs, this restriction does not follow from the semantics of CT
and should be attributed to the semantics of the da-HEAD. The da-HEAD is best an-
alysed in terms of a positive polar da-question14 expressing a negative speaker-
bias towards the underlying proposition, e.g. Da ste zabeljazali nešto ne-
običajno? (‘Noticed something unusual?’) suggests “I know there’s little chance
that the answer will be ‘yes’ but I thought I'd ask anyway”. If the da- HEAD ex-
presses negative bias towards the underlying proposition, this explains why the
BODY needs to have negative polarity: if not, the BODY contradicts the bias ex-
pressed by the HEAD, and the utterance becomes inconsistent. In addition, the po-
larity of the da-HEAD is fixed by the positive form of the da-question involved in
da-PCCs; negation turns the da-clause into a different type of unembedded da-
question (which also carries a different, positive bias instead). The nature of the
da-HEAD has further the effect that the topic questions evoked by the da-HEAD are
plain polarity questions, rather than verum-questions. Thus, CT in (4) raises the
polar topic questions {Is it out of modesty?, Is it out of poverty?, ...?} and focus
evokes the set {it is out of modesty, it is not out of modesty}.
4 Doubling
This section characterises and explains the doubling pattern in Bulgarian PCCs –
which parts of the VP are doubled and which are not. Existing theories of predi-
cate doubling developed for other languages are mostly based on restrictions on
the realisation of multiple copies in a chain on the assumption that predicate
clefts are derived by movement (Aboh and Dyakonova 2009; Bondaruk 2009).
are inappropriate with the other PCCs (or generally with sentences lacking a verum
element).
14 Based on BulNC-data, we identified different types of unembedded da-clauses –
declarative, exclamative, counterfactual wish, imperative – and several subtypes of
interrogative unembedded da-clauses. Various structural and semantic properties make
the other types of unembedded da-clauses unlikely candidates as HEADs of da-PCCs.

8 Elena Karagjosova and Katja Jasinskaja
Section 4.1 argues that Bulgarian če-and da-PCCs are base-generated. If this is
so, movement-based explanations of doubling cannot apply to Bulgarian. Sec-
tion 4.2 proposes a different kind of analysis: the deletion of material in the BODY
is the result of VP ellipsis, for which the HEAD serves as the licensing context.
4.1 Movement vs. base generation
Abels (2001) argued using a number of tests that Russian PCCs are derived by
movement (in particular, remnant VP movement). Applying the same tests to
Bulgarian suggests that this is not the case at least for če and da-PCCs. First,
unlike Russian PCCs, če and da PCCs resemble hanging topics (cf. Krapova and
Cinque 2008, on Bulgarian hanging topics) in that they are a root phenomenon,
and, second, in that they do not show locality effects. The first point is illustrated
in (10): whereas Russian PCCs can be embedded in a complement clause (10a),
this is not possible for če-and da-PCCs, cf. (10b) and (10c).
(10) a. Petja skazal čto pročitat’ - to on eё pročital. (Ru)
Peter said that read prt he it read
‘Peter said that as for reading he read it.’
b. *Petar kaza, če [če ja e pročel], pročel ja e. (Bg)
Peter said that that it aux read read it aux
‘Peter said that as for reading he read it.’
c. *Petar kaza, če [ da ja e pročel], ne e. (Bg)
Peter said that to it aux read not aux
‘Peter said that as for reading he didn’t read it.’
The second point is shown in (11)–(13). According to the judgements re-
ported by Abels (2001), Russian sentences like (11), where the predicate moves
out of a complement clause, are ungrammatical, i.e. Russian predicate clefts
show locality effects.15 This does not apply to Bulgarian če/da-PCCs:
(11) *Pročitat’- to Petja skazal čto on eё pročital. (Ru)
read prt Peter said that he it read
‘As for reading, Peter said that he read it.’
(12) [Če čete], vsički znajat, če čete. (Bg)
that reads everyone know.pl that reads
‘As for reading, everyone knows that he reads.’
(13) [Da e bedna], ti kaza, če ne e bedna. (Bg)
15 There seems to be considerable inter-speaker variation in the acceptance of predicate
clefts with non-local predicate movement in Russian.

Predicate clefts in Bulgarian 9
to is poor you said that not is poor
‘As for being poor, you said that she is not poor.’
Moreover, Bulgarian PCCs do not obey island constraints. In (14), the HEAD
doubles the predicate of the relative clause, but the sentence is grammatical.
(14) Da pečeli, toj e čovek, kojto nikoga ne pečeli. (Bg)
to wins he is person who never not wins
‘As for winning, he is someone who never wins.’
The third difference is in the connectivity effects. Whereas a
tense/aspect/mood mismatch between the fronted and the lower copy of the verb
is not allowed in Russian (Abels 2001; Aboh and Dyakonova 2009), in Bulga-
rian this is possible (in da-PCCs):
(15) a. * Čitat’- to on eё pročital. (Ru)
read.iprf prt he it read.prf
‘As for reading, he has finished it.’
b. [Da čete], ne e pročel nito edna kniga. (Bg)
to read.iprf.prs not aux read.prf neither one book
‘As for reading, he has not finished a single book.’
Finally, in Russian the arguments and adjuncts of the doubled predicate can
be realised in the HEAD or (more often) in the BODY, but not in both, cf. pronoun
eё ‘it’ in (16) adapted from Abels (2001). In Bulgarian they can be realised in
both, as will be illustrated by multiple examples in the next section.
(16) *[Čitat’ (- to) eё (- to)]HEAD [ Ivan eё čitaet]BODY. (Ru)
read prt it prt Ivan it reads
‘As for reading it, Ivan does read it.’
These observations suggest that the syntactic relationship between the HEAD
and the BODY of Bulgarian PCCs is relatively loose. The most straightforward
explanation would be to assume that če-and da-HEADs are base-generated high in
the clause structure. At the same time, there exist clear-cut constraints on what
can be deleted in the BODY and which parts must be realised. In the next section
those constraints are analysed in terms of VP ellipsis.
4.2 Predicate doubling and VP ellipsis
As was shown in section 3, the HEADs of če-and da-PCCs are basically yes/no-
questions. Yes/no-questions require polarity (plain polarity or verum) focus in
the answer. In this section we show that the material obligatorily doubled in the
BODY is the polarity focus exponent. The rest of the BODY (normally) undergoes

10 Elena Karagjosova and Katja Jasinskaja
ellipsis modulo a number of syntactic restrictions. The kind of ellipsis that we
observe in the BODY of predicate clefts is the same as the one we find in denials,
which is another typical context for polarity and verum focus, and for VP ellip-
sis. Predicate clefts and denials behave in a strikingly parallel fashion with re-
spect to their deletion pattern, which supports a uniform treatment.
In languages that mark focus by accentuation, polarity focus is usually real-
ised as accent on the finite verb, whether it is a main verb or an auxiliary (Höhle
1992). Examples (17) and (18) show that the finite lexical verb bears the polarity
focus accent both in denials and in predicate clefts in Bulgarian. This is the only
part that may not be deleted in the BODY. The rest (cf. ja, knigata) can be and
usually is deleted. If it is not, the utterance sounds redundant, but is not un-
grammatical. Presumably, the finite verb moves out of the VP to mark polarity
focus, while the rest of the VP (optionally) undergoes VP ellipsis.
(17) A: Včera toj ošte ne četeše knigata. B: ČETEŠE (ja) (knigata).
yesterday he yet not read.ipf book.def read.ipf sg.f.acc book.def
‘Yesterday he wasn’t reading that book yet.’ ‘He WAS (reading it / that book)!’
(18) [Če včera četeše knigata], ČETEŠE (ja) (knigata).
that yesterday read.ipf book.def read.ipf sg.f.acc book.def
‘As for reading this book yesterday, he WAS doing it.’
In analytic forms where the finite auxiliary is not a clitic, polarity focus is
realised on the auxiliary. Examples (19) and (20) show this for Bulgarian past
perfect, which is formed by the past imperfect auxiliary beše (non-clitic) plus
the l-participle of the main verb. Since the auxiliary bears focus accent, it can be
deleted neither in the denial (19), nor in the predicate cleft (20).
(19) A: Toj ne beše čel knigata. B: a. BEŠE ( ja cel)! b. *BEŠE ja!
he not aux.ipf read book.def aux.ipf cl.acc read aux.ipf cl.acc
‘He hadn’t read that book.’ ‘He had!’
(20) [Če beše čel knigata], ... a. BEŠE ( ja čel). b. *BEŠE ja.
that aux.ipf read book.def aux.ipf cl.acc read aux.ipf cl.acc
‘As for having read that book, he had indeed.’
As in (17)/(18) the VP may also not be deleted. However, one cannot delete
the main verb, but keep the pronominal clitic ja as in (19Bb) and (20b). Again, it
is equally ungrammatical both in denials and in the BODY of PCCs.
In positive clauses, a clitic auxiliary cannot realise polarity focus and cannot
stand on its own in denials and predicate clefts.16 Both the finite clitic auxiliary
and the lexical non-finite verb are doubled:
16 For some speakers, (21Bc) is marginally acceptable.

Predicate clefts in Bulgarian 11
(21) A: Toj ne e spal. B: a. SPAL e! b. *SPAL! c. *E!
he not cl.aux slept slept cl.aux slept cl.aux
‘He hasn’t slept.’ ‘He has!’
(22) [Če e spal], ... a. SPAL e. b. *SPAL. c. *E.
that cl.aux slept slept cl.aux slept cl.aux
‘Sleep he did indeed.’
This is different in sentences with negation. It is well known that in Bulgari-
an, negation precedes the clitic cluster and assigns stress to the element that im-
mediately follows it (Franks and King 2000). In a way, the first clitic following
the negation turns into a non-clitic. If it is an auxiliary clitic, it can realise polari-
ty focus and the rest of the clause can be deleted, both in denials and in PCCs:
(23) a. A: Te sa spali. B: Ne SA (spali)! b. [Da sa spali], ne SA (spali).
they cl.aux slept not cl.aux slept to cl.aux slept not cl.aux slept
They have slept.’ ‘They haven’t! ‘As for sleeping, they haven’t slept.’
Pronominal clitics can also bear the polarity focus accent after negation, but
only if ellipsis does not apply (24Ba). Some speakers accept denials like ne GO
e!, where the clitic cluster as a whole is stranded, while the lexical verb and the
rest of the VP are deleted. Finally there is a possibility to delete everything but
the finite clitic auxiliary, in which case it gets stressed (24Bb).17 Once again, the
same acceptability judgements apply to predicate clefts, cf. (25).
(24) A: Kupila go e. B: a. Ne GO *(e kupila)! b. Ne E!
bought cl.acc cl.aux not cl.acc cl.aux bought not cl.aux
‘She has bought it.’ ‘She hasn’t! ’
(25) [Da go e kupila], ... a. ne GO *(e kupila) b. ne E.
to cl.acc cl.aux bought not cl.acc cl.aux bought not cl.aux
‘As for bying it, she didn’t (buy it).’
It is beyond the scope of the present paper to provide a detailed analysis of
VP ellipsis in Bulgarian (see e.g. Lambova 2004 on this issue). What is crucial
is that the accentuation and deletion pattern in the BODY of predicate clefts is the
same as in other contexts with polarity and verum focus, in particular in denials.
The polarity focus exponent bears a focus accent and cannot be deleted. The rest
can, in principle, modulo constraints on the deletion of clitics. As was argued in
17 Alternatively, the negation particle gets the stress, in which case the unaccented clitic
auxilicary may (marginally) be deleted in denials and da-PCCs, but not in če-PCCs,
probably due to the fact that a negative answer to a question like Is it true that he hasn’t
read? can be ambiguous, i.e. No, it is not true that he hasn’t read. vs. No, he hasn’t read.

12 Elena Karagjosova and Katja Jasinskaja
section 3, the HEAD has the pragmatic effect of a polar question and therefore
provides a suitable context for polarity focus and VP ellipsis, and the BODY con-
tains just those elements that are not deleted by ellipsis. Since ellipsis is in gen-
eral syntactically optional, both denials and predicate clefts are also acceptable
without ellipsis, although the result sounds redundant and not very natural.
5 Concluding remarks
To conclude, we have described in some detail two of the three kinds of predi-
cate clefts in Bulgarian, which are interesting from the typological point of view
because their HEADs are not infinitival. We have derived the semantic and prag-
matic differences between these kinds from the semantic and pragmatic proper-
ties of the complementizers če and da and the notion of contrastive topic. We
have argued that če- and da-PCCs are not derived by movement and have pro-
vided a new kind of explanation for the doubling pattern, which is not based on
movement, but on the mechanisms of polarity focus and ellipsis.
Bibliography
Abels, K. (2001). The predicate cleft construction in Russian. In Franks et al. (eds.) Formal Approaches to Slav-
ic Linguistics, 1–18, Ann Arbor.
Aboh, E. O. and Dyakonova, M. (2009). Predicate doubling and parallel chains. Lingua, 119(7):1035–1065.
Bondaruk, A. (2009). Constraints on predicate clefting in Polish. In Zybatow et al. (eds.) Proceedings of FDSL
7, Leipzig 2007, 65–78. Peter Lang.
Büring, D. (1995). The 59th-Street Bridge Accent. PhD thesis, University of Tübingen.
Büring, D. (2003). On D-trees, beans, and B-accents. Linguistics and Philosophy, 26(5):511–545.
Franks, S. and King, T. H. (2000). A Handbook of Slavic Clitics. Oxford University Press.
Höhle, T. (1992). Über Verum-Fokus im Deutschen. In Jacobs, J. (ed.) Informationsstruktur und Grammatik,
112–142. Westdeutscher Verlag.
Karttunen, L. (1971). Some observations on factivity. Papers in Linguistics, (4):55–69.
Kiparski, P. and Kiparski, C. (1970). Fact. In Bierwisch, M. and Heidolph, K. E. (eds.) Progress in Linguistics.
A Collection of Papers, 143–173. The Hague, Paris.
Krapova, I. (2001). Subjunctives in Bulgarian and modern Greek. In Rivero, M. L. and Ralli, A. (eds.) Compara-
tive Syntax of Balkan Languages, 105–126. Oxford.
Krapova, I. and Cinque, G. (2008). Clitic reduplication constructions in Bulgarian. In Kallulli, D. (ed.) Clitic
Doubling in the Balkan Languages, 257–286. John Benjamins.
Lambova, M. D. (2004). On Information Structure and Clausal Architecture: Evidence from Bulgarian. PhD
thesis, University of Connecticut.
Lefebre, C. (1992). Towards a typology of predicate cleft languages. Journal of West African Languages,
22(1):53–61.
Maienborn, C. (2003). Die logische Form von Kopula-Sätzen. Number 56 in Studia Grammatica. Berlin.
Pitsch, H. (2010). Complex verbs between syntax and morphology in Bulgarian. Russian Linguistics, (34):307–
329.
Romero, M. and Han, C. (2004). On negative yes/no questions. Linguistics and Philosophy, (27):609–658.
Rooth, M. (1992). A theory of focus interpretation. Natural Language Semantics, 1(1):75–116.
Smirnova, A. (2011). Evidentiality and Mood. PhD thesis, Ohio State University.
Vicente, L. (2007). The Syntax of Heads and Phrases. PhD thesis, Leiden University.
Vicente, L. (2009). An alternative to remnant movement for partial predicate fronting. Syntax, 12(2):158–191.