Syntactic limits on High-tone spreading

Aida Tali (University of Connecticut); aida.talic@uconn.edu

Linguistic Society of America 91, Austin, Texas - Jan 5th 2017
In this talk I focus on a previously unexplored phenomenon of High-tone spreading from enclitics
to their hosts in a dialect of Bosnian/Croatian/Serbian (BCS) spoken in central Bosnia and
Herzegovina.
The puzzle: In the absence of enclitics:
wh-words like ta what have a falling accent: t
the complementizer da that has no accent: da

Diacritics:
falling accent: [ ` ]
rising accent: [ ]

The presence of enclitics:

in some contexts has no effect on the accent of ta or da:
(1) a. t mu
govori?
(interaction)
what him.DAT says
What is he telling him?
b. Znam da
mi
je donijela knjigu.
(interaction)
know that me.DAT is bring
book
I know she brought me the book.
in other contexts these hosts get a rising accent (High-tone spreading from the enclitic):
(2) a. t mu
je rekao?
what him.DAT is said
What did he say to him?
b. D mi
je da
provedem ljeto
na
that me.DAT is that spend
summer on
I wish I could spend the summer on a mountain.

(interaction)

planini.
mountain

(interaction)

In identical Host+Enclitic sequences, enclitics with a lexical High tone can spread that tone to
their host in some constructions (2), but not in others (1).
This interaction reveals an important syntactic condition on prosodic mapping of BCS clitics, and
supports the idea that the mapping from the syntactic to the prosodic structure is phase bound (see
e.g. Dobashi 2003; Kahnemuyipour 2004, 2009; Kratzer and Selkirk 2007; Sato 2012; Sato and
Dobashi 2016; among others).
In my analysis of the split between (1)&(2), I follow Selkirk (1996) in assuming that clitics can
attach to the prosodic structure in the three ways given in (3):
(3) a. internal clitics: ( )
-inside the (minimal) prosodic word of the host
b. affixal clitics: (() )
-adjoined to the prosodic word of the host
c. free clitics:
(() ) -sister to the prosodic word of the host

LSA 91 (2017)

Tali

I propose:
(4) Simultaneous Spell-Out Condition (SSC):
A clitic CL incorporates into the prosodic word of the host H iff CL and H are in the same
spell-out domain (SOD) and they are immediately adjacent.
(5)

a.

XP
HOST

b.
YP

SOD

XP
HOST

ECL

YP

SOD

ECL

(interaction)
(interaction)
Finally, a clitic that is not in the same SOD as its host can be moved into the SOD of its host
under very narrowly defined phonological conditions which stem from idiosyncratic needs of
one particular clitic (the clitic je be.3sg).
The structure of the talk:

1. BCS accent assignment: Under what phonological conditions can an enclitic interact with
the accent of the host?
2. The main pattern
3. The position of BCS enclitics in the syntax and phases
4. Cyclic incorporation

1. BCS accent assignment and weak elements following the host

BCS has pitch accents falling or rising accent on prominent syllables.
Accent is contrastive, as we can see in the minimal pairs (6a&c) and (6b&d).
(6)

falling
a. k:se:
steep.NOM.PL.FEM.LF/
steep.GEN.SG.FEM.LF
b. kse:
mow.3PL.PRES

rising
c. k:se:
long
steep.GEN.SG.FEM.SF
d. kse:
hair.GEN.SG

Diacritics:
falling accent: [ ` ]
rising accent: [ ]

short

The falling and rising accents partially result from vocabulary items (roots, affixes, and clitics)
in BCS having or lacking a lexical High tone (e.g. Inkelas and Zec 1988), but not all lexical H
tones in a prosodic word are realized due to accent-assignment rules in BCS.
Under what phonological conditions can the H tone of a weak element following a
monosyllabic host be realized?
2

LSA 91 (2017)

Tali

Given the lexical tone marking, there are four possible contexts with a monosyllabic host
followed by a weak element (an enclitic or a suffix):
(7) a. ( + )
No lexical H; default initial H insertion Falling initial accent
The presence of a lexical H bleeds the default initial High tone insertion:
H

b. (
c. (H

+ )
+ H)
H
d. ( + )

The only H present is realized

The leftmost H is realized
High tone spreading

Falling initial accent

Falling initial accent
Rising initial accent

The H tone of a weak element is realized only if it follows a host without a lexical H tone, in
which case H-tone spreading takes place and gives the host a rising accent.
(8)

Combination
( + H)

Example
n-aH
woman-F.NOM.SG

Accent
Rising

(interaction)

However, if the same suffix (-aH) is outside of the minimal prosodic word of the same toneless
host (en-), then the host gets the default initial H tone and H-tone spreading from the suffix is
blocked, as in (9) with an additional derivational suffix.
(9)

Combination
(() + H)

Example
Accent
H
:n-k-a
Falling
woman-N-F.NOM.SG
female specimen

(interaction)

The H tone of a weak element is realized only if it is in the same minimal prosodic word with
a host without a lexical H tone, in which case H-tone spreading takes place and gives the host a
rising accent.
Turning to enclitics, there is exactly one phonological environment where interaction with the
host (H-tone spreading) is possible:
(10) An enclitic can interact with the accent of its host only if:
(i) the enclitic has a lexical High tone;
(ii) the host does not have a High tone; and
(iii) the enclitic is incorporated into the minimal prosodic word.
interaction = rising tone on the immediately preceding syllable
interaction = falling or no tone on the immediately preceding syllable

( + H)

LSA 91 (2017)

Tali

2. The main pattern

Hosts that allow interaction are all monosyllabic1; and they are found in questions, conditionals
and a certain type of imperatives:
(11) a. the particle da that
b. question words: ko who, ta what, to why, gdje where
Observations for individual enclitics:
(12) The particle da2 or a question word preceding the interrogative particle li always
get a rising accent.
(13) a. D li eli sladoled?
that Q want ice.cream
Would you like some ice cream?
b. t li hoe?
what Q wants
I wonder what he wants.

(interaction)

(14) a. The particle da preceding an auxiliary enclitic has no accent.

b. A question word preceding an auxiliary enclitic (except je)3 always has a rising
accent.
(15) a. Znam [da
su
(oni) doli].
know that are they come
I know that they came.
b. t
su
rekli?
what are say
What did they tell you?

(interaction)

(interaction)

(16) a. The particle da preceding object enclitics sometimes has a rising accent and
sometimes no accent.
b. A question word preceding object enclitics sometimes has a rising accent and
sometimes a falling accent.

It also seems to matter that these hosts are open syllables, since there is no interaction between enclitics and toneless hosts
such as most bridge, rad work, dan day etc, or the question word kad when.
2
The particle da has many different usages, some of which are discussed below. However, it is most commonly used as a
complementizer, hence I will gloss it as that throughout the paper for ease of exposition.
3
The clitic je is has several idiosyncratic properties (see e.g. Browne 1974; Bokovi 2001) that separate it from other
auxiliary clitics. Thus, I put je aside here, discussing it separately below.

LSA 91 (2017)

Tali

(17) a. D mi
je da
provedem ljeto
na planini.
that me.DAT is that spend
summer on mountain
I wish I could spend the summer on a mountain.
b. Znam da
mi
je donijela knjigu.
know that me.DAT is bring
book
I know she brought me the book.
c. t mu
je rekao?
what him.DAT is said
What did he say to him?
d. t mu
govori?
what him.DAT says
What is he telling him?

(interaction)

(interaction)

(interaction)

(interaction)

(18) a. The particle da preceding the enclitic je sometimes gets a falling accent4 and
sometimes no accent.
b. A question word preceding the enclitic je always has a falling accent.
(19) a. D je meni
(da pojedem) sladoled.
that is me.DAT
that eat
ice.cream
I wish I could eat an ice cream.
b. Znam da je meni
donijela knjigu.
know that is me.DAT bring
book
I know she brought the book for me.
c. t je rekao?
what is said
What did he say?

(interaction)

(20) a. The particle da preceding the enclitic se is never accented.

b. A question word preceding the enclitic se always has a falling accent.
(21) a. Znam da se umio.
know that SE washed
I know he washed his face.
b. t se skinulo bijelim vinom?
what SE removed white wine
What removed with white wine?

(interaction)

Note that the falling tone on the particle da does not indicate that a High tone spreads from je to da, i.e. je does not
interact with the accent of its host here either.

LSA 91 (2017)

(12)-(13)

Tali

(14)-(17)

(18)-(21)

li
Aux
Dat
Acc
Gen
je, se
(interaction) (interaction) (interaction) (interaction) (interaction)
(interaction) (interaction) (interaction) (interaction) (interaction)

Table 1: Interaction of enclitics with the accent of their host

The main puzzle: Why is the interaction sometimes blocked with clitics that have a High tone
(even in identical host+clitic sequences)?

3. Syntactic positions of clitics and phases

Browne (1974): BCS enclitics occur in a fixed order (22a).
I assume that enclitics are in different syntactic positions (see e.g. Bokovi 1995, 2001;
Stjepanovi 1998; Graanin-Yuksek 2016):
(22) a.
li (Q/Foc particle) + Aux +
Dat + Acc/Gen + {se, je}
b. [CP C
[TP
T
[vP IOcl
DOcl [vP
v
[VP V.]]]
Some arguments for placing clitics in different positions: Ellipsis (Stjepanovi 1998; Bokovi
2001) and VP-fronting (Bokovi 2001 (based on p.c. with D. avar)):
(23) a. Mi smo mu
ga
dali, a
i
vi ste (mu
ga
dali) takodje.
we are him.DAT it.ACC given, and also you are (him.DAT it.ACC given) too.
We gave it to him, and you did, too.
b. [VP Dali ga
Mariji] su Ivan i
Stipe.
given it.ACC Marija are Ivan and Stipe
Give it to Marija, Ivan and Stipe did.
Auxiliaries higher than object clitics
(24) a. On mi
ga
je dao, a
i
ona
he mi.DAT it.ACC is gave, and also she
He gave it to me, and she did, too.
b. Dao ga
Mariji je Ivan.
given it.ACC Marija is Ivan
Given it to Mary, Ivan did.
c. *Dao je Mariji ga Ivan.
Even je be.3sg.pres, which follows object clitics in
syntax!5
5

je <e> takoe.
is
too

(Stjepanovi 1998: 532)

PF, is higher than object clitics in the

The clitic se: difficult to pin down to one position in the structure because it is found in many different constructions:
reflexive, reciprocal, middle, passive, unaccusative, object-arbitrarization, impersonal, and frozen se-constructions (see
Marelj 2004). Marelj (2004) argues that se is a Case-absorbing morpheme, which can occupy either a high functional
projection or a low one (see also Progovac 1997).

LSA 91 (2017)

Tali

I adopt a contextual approach to phases: The highest projection in every extended domain is a
phase. (Bobaljik and Wurmbrand 2005; Bokovi 2005, 2013, 2014; Gallego and Uriagereka 2007;
den Dikken 2007; Despi 2013; M.Takahashi 2011; Wurmbrand 2014, a.o).
(25)
(26)

[FP2 F2 [FP1 F1 [LP

[FP1 F1 [LP L]]

L]]]

Following standard assumptions, spell-out domains (SODs) are:

(i) complements of phase heads (Chomsky 2000, 2001)
(ii) the topmost projection in a sentence as the final spell-out

4. Cyclic clitic spell-out and High tone spreading

Spelling out the hosts when there are no enclitics following them:
A wh-host without a clitic following it has a falling accent:
(27) t (on) (esto) govori?
what (he) (often) says
What is he saying? What does he (often) say?
(28)

a. Linearize: ta << [(on)<< (esto)<<govori]

b. wh-word in PF: (ta) default initial H-tone insertion

SOD
CP
TA

TP

SOD

(ON)<<(ESTO)<<GOVORI

da-complementizer without a clitic following it has no accent

(29) Kae [da (esto) govori istinu].
says that (often) says truth
He says that he (often) tells the truth.
(30)

VP
(KAE)

a. Linearize: kae da << [(esto)<<govori<<istinu]

b. da in PF: ((da) (esto) ) free clitic; no H-tone insertion

SOD
CP
C

TP

SOD

DA

(ESTO)<<GOVORI<<ISTINU

When hosts retain this prosody even when followed by an enclitic, this indicates that the host
and the clitic are not in the same minimal prosodic word (cf. (10)).
7

LSA 91 (2017)

Tali

4.1. The host and the clitic in different SODs

Object clitics low in the verbal domain:
(31) a. t mu
govori?
what him.DAT says
What is he telling him?
b. Znam da mu
govori istinu.
know that him.DAT says truth
I know he is telling him the truth.
(32)

a.

falling accent

no accent

b.

SOD

VP

CP
TA

TP

(interaction)

(ZNAM)

SOD

SOD
CP
C

SOD

TP

DA

MU<<GOVORI

MU<<GOVORI<<ISTINU

The object clitic and the host do not form a minimal prosodic word.
Auxiliary clitics in declaratives:
Even with a slightly higher clitic in T, no interaction with da in C.
(33) a. Znam da su govorili istinu.
no accent
know that are said
truth
I know they were telling the truth.
b. Znam da su mu
govorili istinu.
no accent
know that are him.DAT said
truth
I know they were telling him the truth.
(34)
VP
(ZNAM)

(interaction)

SOD
CP
C

TP

SOD

DA
SU<<GOVORILI<<ISTINU

The auxiliary clitic and the host do not form a minimal prosodic word.
8

LSA 91 (2017)

Tali

4.2. The host and the clitic in the same SOD

While finite verbs in BCS declarative clauses are in positions lower than C, it is standardly
assumed that they can move to C in questions.
(35) Jesu li oni govorili istinu?
are Q they told
truth
Were they telling the truth?
Auxiliary clitics in questions raise to C
(36) a. t su (oni) govorili?
what are they said
What were they saying?
b. t ste mu
rekli?
what are him.DAT told
What did you tell him?
(37)

rising accent

(interaction)

rising accent

SOD
CP
TA

TP

SOD

SU

(ONI)<<GOVORILI

The auxiliary and the host do form a minimal prosodic word.

The clitic li in questions
The particle li is standardly assumed to occupy the C position, realizing an interrogative feature
in yes-no questions (38) (Bokovi 2001; Progovac 1996; Rivero 1993; Schtze 1994;
Stjepanovi 1999) or focus feature in emphatic questions (40), where it is preceded by a wh-word
(Bokovi 2001).6
The clitic li always spreads its H tone to the host, as in (38)/(40), giving it a rising accent.
(38)

D li govori istinu?
that Q says truth
Is he telling the truth?

rising accent

(interaction)

Graanin-Yuksek (2016) argues that li is a focus marker in yes-no questions as well, base generated in the head of a
FocusP below CP. For the purposes of this paper, it is enough to assume that the position of li in the final output of the
syntax is in C, putting aside whether this results from Foc-to-C movement or li being base generated in C.

LSA 91 (2017)

Tali

(39)

SOD
CP
C

SOD

TP

DA LI
GOVORI<<ISTINU

(40) t li
govori?
what FOC says
I wonder what he is saying?
(41)

rising accent

(interaction)

SOD
CP
TA

TP

SOD

LI
GOVORI

An enclitic is in the same minimal prosodic word of the host and interacts with its accent if
they are in the same spell-out domain.
(42) Simultaneous Spell-Out Condition (SSC):
A clitic CL incorporates into the prosodic word of the host H iff CL and H are in the same
spell-out domain and they are immediately adjacent.

4.3. je(be.3sg) and the raising of the object clitics

Syntactically, this clitic occupies the same position as other auxiliary clitics, i.e. it is higher than
object clitics, just like all other auxiliaries.
Recall evidence from Stjepanovi (1998), Bokovi (2001): VP-ellipsis
(43) On mi
ga
je dao, a
i
ona je <e> takoe.
he mi.DAT it.ACC is gave and also she is
too
He gave it to me, and she did, too.
(Stjepanovi 1998: 532)
This should result in the clitic je preceding object clitics, just like other auxiliaries do. However,
this is not what we find.
10

LSA 91 (2017)

Tali

(44) a. ta ste mu
rekli?
what are him.DAT told
What did you tell him?
b. *ta mu
ste rekli?
what him.DAT are told
c. *ta je mu
rekao?
what is him.DAT told
d. ta mu
je rekao?
what him.DAT is told
What did he tell him?

Aux<<Dat

Dat<<Aux
Aux(je)<<Dat
Dat<<Aux(je)

(45) Constraint on je-linearization: *je<<enclitic

Crucially, this is a PF constraint because in the syntax, the clitic je behaves like other auxiliaries
in being higher than object clitics (43).
Recall that object clitics in the absence of any other clitic do not interact with the accent of the
host (see (31)) because they are not in the same SOD.
Interestingly, when je follows an object clitic, the clitic interacts with the accent of its host.
(46)

t mu
je rekao?
what him.DAT is said
What did he tell him?

rising accent

(interaction)

The rising accent on the host in (46) indicates that the object clitic is in the same minimal
prosodic word of the host, and in the same spell-out domain as the host according to (42).
How does it get there (if not in the syntax)?
PF mechanism: The object clitic and je reorder to satisfy the constraint on linearization (45).
(47) a.

(i) At the lower spell-out domain (SOD1), which

contains only the object clitic and the verb in (47a), the
clitic gets linearized to precede the verb in PF.

SOD2
CP
TA

JE

TP

SOD1

MU<<REKAO

b.

TA<<JE<<(MU<<REKAO)

(ii) At the next spell-out domain, ta and je need to be

linearized with respect to each other and with respect
to the elements from SOD1.
(iii) The c-command relations in (47a) yield the result
in (47b) (see e.g. Kayne 1994). However, this violates
the constraint in (45).
11

LSA 91 (2017)

Tali

(iv) PF-reordering to satisfy the idiosyncratic requirement of je in (45):

TA<<MU<<JE<<REKAO

Prediction 1: If both je and object clitics are in the spell-out domain lower than the host, then
PF-reordering of the object clitic in front of je should not have any effect on the prosody of the
host.
Recall: da+mu = no interaction in constructions like (48):
(48) a. Znam da
mu
govori istinu.
know that him.DAT says truth
I know he is telling him the truth.
b. Da mi
govori istinu, vjerovao bih
mu.
that me.DAT says truth
believed would him.DAT
If he were telling me the truth, I would believe him.

(interaction)

(interaction)

In such constructions, even if je follows object clitics, there is no interaction with da:
(49) a. Znam da mu
je govorio istinu.
(interaction)
know that him.DAT is said
truth
I know that he was telling him the truth.
b. Da mi
je govorio istinu, vjerovao bih mu.
that me.DAT is said truth believed would him
If he had been telling me the truth, I would have believed him.
Desiderative construction
For the sake of completeness object clitics interact with the accent of the complementizer in
desiderative constructions:
(50) a. D mi
je sladoled.
rising accent
(interaction)
that me.DAT is ice.cream
I wish I had an ice cream.
b. D mi
je {da
pojedem}/{pojesti} sladoled.
that me.DAT is that eat
eat.INF ice.cream
I wish I could eat an ice cream.
(i) T-to-C movement of je
(ii) PF reordering of object clitic across je
Prediction 2: if the only way for clitics placed in a low projection in the syntax to reach a higher
spell-out domain is by a PF reordering, which takes place due to the presence of je, then when je
is absent, these clitics should never be able to reach the higher spell-out domain and interact with
the accent of elements in that domain.
We have seen this before in constructions where there is no auxiliary (31).
12

LSA 91 (2017)

Tali

Additional contexts, with se

Browne (1974) and Bokovi (2001, 2004): in the presence of the clitic se self, the clitic je is
preferably omitted.
(51) a. On je vratio
knjigu.
he is returned book
He returned the book.
b. On se vratio.
(je dropped)
he SE returned
He returned.

BCS past tense:

Aux + participle

Recall that an object clitic interacts with the accent of the host when je follows it:
(52) t mu
je rekao?
rising accent
(interaction)
what him.DAT is said
What did he tell him?
Crucially, there is no interaction when je is dropped in the presence of se:
(53) t mu
se desilo?
falling accent
(interaction)
what him.DAT SE happened
What happened to him?
Crucially, the presence of se has no effect on the interaction between higher clitics and the accent
of their host.
(54) a. t li
su se dogovorili?
what FOC are SE agreed
I wonder what they agreed.
b. D li su se vidjeli?
that Q are SE seen
Did they see each other?
c. t su se dogovorili?
what are SE agreed
What did they agree?

(interaction)

Conclusions:
I have argued that the mapping of clitics from their syntactic positions to the prosodic structure
is constrained by phases.
This was supported by evidence involving High-tone spreading from enclitics to their hosts in
a dialect of BCS, showing that a clitic can incorporate into the minimal prosodic word of the
13

LSA 91 (2017)

Tali

host only if it is in the same spell-out domain, i.e. linear adjacency between the clitic and the
host is not enough for incorporation.
I have argued that one apparent exception to the Simultaneous Spell-Out Condition comes from
idiosyncratic phonological requirements of one particular item that behaves differently from
other BCS enclitics in several respects.
The observation that enclitics in this dialect have a lexical High tone is new (BCS clitics are
usually considered to lack lexical High tones, i.e. to be non-accentogenic). The High tone on
the enclitics indicated by the fact that there are contexts where the host preceding such enclitics
gets a rising accent, which results from the rule of High tone spreading in BCS.
Selected References (PDF with a full list to be uploaded at: www.aidatalic.jimdo.com):
Bobaljik, Jonathan and Susi Wurmbrand. 2005. The domain of agreement. Natural Language and Linguistic
theory 23. 809-865.
Bokovi, eljko. 2001. On the nature of the syntax-phonology interface: Cliticization and related phenomena.
Amsterdam, Elsevier.
Bokovi, eljko. 2013. Phases beyond clauses. In The nominal constructions in Slavic and beyond, ed. Lilla Schrcks,
Anastasia Giannakidou, and Urtzi Exteberria, 75-128. Berlin: Mouton de Gruyter.
Bokovi, eljko. 2014. Now Im a phase, now Im not a phase: On the variability of phases with extraction and ellipsis.
Linguistic Inquiry. 45. 27-89.
Browne, Wayles. 1974. On the problem of clitic placement in Serbo-Croatian. In Richard Brecht and Catherine
V. Chivany, Slavic transformational syntax, 36-52. University of Michigan.
Despi, Miloje. 2013. Binding and the structure of NP in Serbo-Croatian. Linguistic Inquiry 44: 239-270.
Dobashi, Yoshihito. 2003. Phonological phrasing and syntactic derivation. Ithaca, NY: Cornell University
dissertation.
Graanin-Yuksek, Martina. 2016. Size matters: The syntax of disjunctive questions. Linguistic Inquiry 47(2).283305.
Inkelas, Sharon and Draga Zec. 1988. Serbo-Croatian Pitch Accent: The interaction of tone, stress, and intonation.
Language 64 (2). 227-248.
Kahnemuyipour, Arsalan. 2004. The syntax of sentential stress. Toronto, Canada: University of Toronto
dissertation.
Kahnemuyipour, Arsalan. 2009. The syntax of sentential stress. Oxford: Oxford University Press.
Kratzer, Angelika and Elisabeth Selkirk. 2007. Phase theory and prosodic spellout: The case of verbs. The
Linguistic Revew 24. 93-135.
Marelj, Marijana. 2004. Middles and argument structure across languages. Utrecht: University of Utrecht
dissertation.
Progovac, Ljiljana. 1997. Events in Serbian. In M. Dimitrova-Vulchanova, et al (eds.), University of Trondheim
Working Papers in Linguistics 31.79-116.
Sato, Yosuke and Yoshihito Dobashi. 2016. Prosodic phrasing and the that-trace effect. Linguistic Inquiry 47(2).
333-349.
Sato, Yosuke. 2012. Phonological interpretation by phase: Nuclear stress, domain encapsulation and edge
sensitivity. In ngel Gallego (ed.), Phases: Developing the framework, 283-308. Berlin: Mouton de Gruyter.
Selkirk, Elisabeth. 1996. The prosodic structure of function words. In K. Demuth and J. Morgan (eds.), Signal to
Syntax: Bootstrapping from speech to grammar in early acquisition, 187-213. Lawrence Erlbaum.
Stjepanovi, Sandra. 1998. On the placement of Serbo-Croatian clitics: Evidence from VP ellipsis. Linguistic
Inquiry 29. 527-537.

14