Forbidden clitic clusters in Zapotec:Implications for the Person–Case Constraint

Steven Foley, Nick Kalivoda, and Maziar ToosarvandaniUniversity of California, Santa Cruz{srfoley, nkalivod, mtoosarv}@ucsc.edu

CLS 53 – May 25, 2017

1 Introduction

Many languages prohibit certain combinations of clitic arguments based on person: e.g. the Person–Case Con-straint (PCC) and “3–3 Effects” (Perlmutter 1971, Bonet 1991).

(1) Person–Case Constraint (Spanish)a. {1, 2} � 3

PedroPedro

{me,{1SG.DAT,

te}2SG.DAT}

lo3SG.M.ACC

envıa.send.PRES.3SG

‘Pedro sends it to {me, you}.’b. 3� {1, 2}

* PedroPedro

le3SG.DAT

{me,{1SG.ACC,

te}2SG.ACC}

envıa.send.PRES.3SG

Intended: ‘Pedro sends {me, you} to him.’ (Ormazabal and Romero 2007:316–317)

(2) 3–3 Effect (Spanish): *3� 3

* Le3SG.DAT

lo3SG.M.ACC

recomende.recommend.PAST.1SG

Intended: ‘I recommended it to him.’ (Perlmutter 1971:132)

These constraints are often taken to have distinct sources: the PCC is syntactic, while 3–3 effects have a morpho-logical source (Anagnostopoulou 2003, Nevins 2007, 2011, Rezac 2011, a.o.).

To support this division, a certain line of reasoning is sometimes advanced (Nevins 2007, 2011, Rezac 2011):

(3) Division of Labor (Nevins 2011:948)a. Syntactic restrictions are hierarchical, and asymmetric. Based on principles of Multiple Agree.

Repairs involve periphrasis.b. Postsyntactic restrictions may be idiosyncratic and symmetric. Based on principles of syntagmatic

markedness. Repairs involve deletion.

Since the repair for the PCC involves periphrasis, it arises from a syntactic constraint. By contrast, since the 3–3Effect in Spanish is repaired through replacement of a clitic by a “spurious se,” it has a morphological source.

1

(4) 3� 1

Me1SG.ACC

enviaronsend.PAST.3PL

ato

el.3SG.M

‘They sent me to him.’ (Ormazabal and Romero 2007:318)

(5) 3� 3

Se3SG

lo3SG.M.ACC

recomende.recommend.PAST.1SG

‘I recommended it to him.’ (Perlmutter 1971:132)

We examine clitic clusters in several Northern Zapotec varieties (Oto–Manguean: Oaxaca), which have a four-waygender distinction. They forbid certain combinations of third-person clitics in somewhat familiar ways:

◦ Gender–Case Constraint (GCC) (general)Some combinations of subject and object clitics are not permitted depending on their position in a hierarchyof gender.

(6) Guiloxi Zapoteca. 3.EL� 3.AN

Bdel=e’=b.hug.COMP=3.EL=3.AN

‘S/he (an elder) hugged it (an animal).’ (RM and FA, GZYZ012, 32:15)b. 3.AN� 3.EL

* Bdi’inn=b=ne’.bite.COMP=3.AN=3.EL

Intended: ‘It (an animal) bit her/him (an elder).’ (RM, GZYZ014, 32:32)

◦ *X–X Constraint (general)Combinations of identical subject and object clitics are not allowed.

(7) Guiloxi Zapotec* Ba

alreadybetw=ba’=ba’.hit.COMP=3.HU=3.HU

Intended: ‘S/he already hit her/him.’ (RM and FA, GZYZ014, 18:18)

Methodological Point The source of these constraints is not necessarily indicated by their repair: though the*X–X Constraint is repaired through periphrasis, it has a morphological source.

Theoretical Point The GCC — and, by analogy, the PCC — must be able to rule in combinations of identicalclitics as syntactically well-formed; that is, it is grammatically distinct from the *X–X Constraint (Anagnos-topoulou 2003, Nevins 2007, 2011; contra Walkow 2012).

The remainder of this talk is structured as follows:

◦ Section 2: Background on clitic pronouns in Zapotec

◦ Section 3: The manifestations of the *X–X Constraint

◦ Section 4: Distinguishing the *X–X Constraint from the GCC

◦ Section 5: Conclusions and further consequences

2

2 Clitic pronouns in Zapotec languages

We focus on several closely related Northern Zapotec varieties:

◦ San Sebastian Guiloxi (original field work, Toosarvandani 2017)

◦ Hidalgo Yalalag (Avelino Becerra 2004, Lopez and Newberg 2005)

◦ San Baltazar Yatzachi el Bajo (Butler 1980)

◦ San Bartolome Zoogocho (Long and Cruz 2000, Sonnenschein 2004)

Figure 1: Zapotec varieties of the Sierra Norte, Oaxaca (Butler 1980)

These varieties all exhibit the same four-way gender distinction:1

ELder human vs. non-elder HUman vs. ANimal vs. INanimate

These gender distinction are realized in the languages’ third-person pronouns, which come in both strong andclitic forms.

(8) Strong and clitic pronouns in Guiloxi Zapotec

STRONG CLITIC

subject object1SG nada’ =a’ –2SG le’ =o’ –3.EL le’ =e’ =ne’3.HU leba’ =ba’ =ba’3.AN leb =b(a) =b(a)3.IN len =(e)n =(e)n

Across the Zapotec languages, clitic pronouns are syntactically and prosodically dependent, while the strongpronouns occur elsewhere (though see Marlett 2010a,b for further details).

1Other Zapotec languages have genders specific to adult males, adult females, children, babies, young unmarried men, deities,celestial bodies, liquids, trees and wooden objects, and disparaged/pejorative referents; see Marlett 2010a for a survey.

3

2.1 The distribution of clitic pronouns

Clitic pronouns can serve as arguments of the verb, possessors of inalienable nouns, and arguments of prepositions.

(9) a. Subject or objectBdel=ba’=ne’.hug.COMP=3.HU=3.EL

‘S/he hugged her/him.’ (Guiloxi: RM, GZYZ012-s, 22)b. Inalienable possessor

bibil=a’sister=1SG

‘my sister’ (Guiloxi: RM, GZYZ003-s, 27)c. Object of preposition

Dzun=ba’do.CONT=3.HU

shbabthought

tse=ba’.of=3.HU

‘S/he is thinking of her/him.’ (Guiloxi: RM, GZYZ018-s, 30)

In subject position, R-expressions and third person strong pronouns are in complementary distribution with clitics.

(10) R-expression

Baalready

bdel(*=ba’)hug.COMP=3.HU

bidao’child

nithis

Pedro.Pedro

‘This child already hugged Pedro.’ (Guiloxi: FA and RM, GZYZ014, 27:09)

(11) Third person strong pronouna. Ba

alreadyshtahs(*=e’)sleep.CONT=3.EL

le’.3.EL

‘S/he (an elder) is sleeping.’ (Guiloxi: FA and RM, GZYZ014, 47:53)b. Ba

alreadybdel(*=ba’)hug.COMP=3.HU

leba’3.HU

beku’.dog

‘S/he (a non-elder) already hugged the dog.’ (Guiloxi: RM and FM, GZYZ013, 11:35)c. Shtahs(*=b)

sleep.CONT=3.AN

leb.3.AN

‘It (an animal) is sleeping.’ (Guiloxi: RM and FA, GZYZ013, 10:48)d. Ba

alreadybzxup(*=en)fall.COMP=3.IN

len.3.IN

‘It fell.’ (Guiloxi: RM and FA, GZYZ014, 49:55)

Local person strong pronouns in subject position, however, must be doubled by a clitic (Toosarvandani 2017), cf.Kalivoda (2015) for Teotitlan del Valle Zapotec.

(12) Local person strong pronouna. Tzxizh*(=a’)

laugh.CONT=1SG

nada’.1SG

‘I am laughing.’ (Guiloxi: FA and RM, GZYZ013, 5:27)b. Tzxizh*(=o’)

laugh.CONT=2SG

le’.2SG

‘You are laughing.’ (Guiloxi: FA and RM, GZYZ013, 6:07)

4

The verb can bear a clitic cross-referencing the direct object only when a subject clitic is present. A direct objectcannot encliticize across an R-expression subject, across a trace of the subject, or onto the subject itself.

(13) a. * Bdel=bhug.COMP=3.AN

Maria.Maria

Intended: ‘Maria hugged it.’ (Guiloxi: FA and RM, GZYZ012, 24:55)b. * No1

whobetwhit.COMP

t1 =eb?=3.AN

Intended: ‘Who hit it?’ (Guiloxi: RM and FA, GZYZ013, 3:34)c. * Bdel

hug.COMP

Maria=b.Maria=3.AN

Intended: ‘Maria hugged it.’ (Guiloxi: RM and FA, GZYZ013, 4:40)

Unlike subject clitics, direct object clitics are always in complementary distribution with an overt argument,regardless of whether it is a strong pronoun or an R-expression.

2.2 A clitic doubling analysis

We take the clitic pronouns to arise through clitic doubling, even when they are in complementary distributionwith an overt argument.

Specifically, we adopt a “big DP” analysis of clitic doubling: the clitic is generated in the specifier of a nominalfunctional projection, e.g. Part(icipant)P or K(ase)P (Torrego 1992, Uriagereka 1995, Nevins 2011, Arregi andNevins 2012, a.o.)

(14)T vP

FP

Cl F′

F DP

Because clitics are deficient, they must Agree with T, triggering movement of the clitic to a position adjacent tothe verb, which surfaces in initial-position through predicate fronting (Lee 2006, Adler et al. 2017).

5

(15) TP

VP1

V

T′

T vP

FP

Cl F′

F DP

t1

The clitic might undergo either phrasal movement plus morphological merger (e.g. Nevins 2011) or head move-ment (e.g. Arregi and Nevins 2012, Preminger 2017).

To account for the contrast between the distribution of local and non-local person clitics, we assume that:

◦ local person strong pronouns are always Merged with a clitic pronoun

◦ third-person clitics can only be Merged with a null pronoun

In the absence of evidence to the contrary, we assume this analysis of clitic pronouns in Guiloxi extends to theother Zapotec varieties.

2.3 Restrictions on clitic clusters

All four Zapotec varieties restrict the clitic clusters that are possible based on their gender (as well as person).

(16) Yalalag: (1 ≹ 2 >) 3.EL > 3.HU > 3.AN > 3.IN

OBJECT

3.EL 3.HU 3.AN 3.IN

SU

BJE

CT 3.EL * � � �

3.HU ∗ * � �3.AN ∗ ∗ * �3.IN ∗ ∗ ∗ *

Though there is variation across the four varieties, these restriction can be broadly divided into two patterns, eachparalleling a constraint found in the domain of person (Perlmutter 1971, Bonet 1991).

◦ *X–X Constraint, cf. 3–3 Effects

◦ Gender–Case Constraint (GCC), cf. Person–Case Constraint (PCC)

6

3 The *X–X Constraint

Generally speaking, across all four Zapotec varieties, combinations of clitics at the same rank on the genderhierarchy are forbidden, e.g. *3.HU� 3.HU, etc.

(17) Yalalaga. * Bchew’=e’=e’.

kick.COMP=3.EL=3.EL

Intended: ‘He kicked him.’ (Avelino Becerra, p.c.)b. * Llne’=be’=be’.

speak.HAB=3.HU=3.HU

Intended: ‘He is speaking to him.’c. * Bdinn=ba’=ba’.

bite.COMP=3.AN=3.AN

Intended: ‘It bit it.’ (Avelino Becerra 2004:35)d. * Bchochj=en=en.

bite.COMP=3.IN=3.IN

Intended: ‘It hit it.’ (following Avelino Becerra 2004:34)

When the subject and object clitics for each gender are phonologically identically, it is not possible to tell whetherthese clusters are forbidden because the clitics share the same form or the same features.

(18) Strong and clitic pronouns in Yalalag Zapotec (Avelino Becerra 2004)

STRONG CLITIC

subject object1SG nada’ =a’ –2SG lue’ =o’ –3.EL le’e =e’ =e’3.HU lebe’ =be’ =be’3.AN leba’ =ba’ =ba’3.IN len =n =n

In these Zapotec varieties, the form of certain clitics varies by grammatical or thematic role, allowing thesetwo possibilities to be teased apart.

3.1 Case 1: Guiloxi

As we saw in (8), the third-person elder clitic in Guiloxi has two forms: one for subjects and one for objects.

(19) a. Baalready

got=e’.die.COMP=3.EL

‘He already died.’ (RM, GZYZ004-s, 5)b. Ba

alreadybetw=o’=ne’.hit.COMP=2SG=3.EL

‘You already hit her/him.’ (RM, GZYZ011-s, 23)

7

Unlike in Yalalag, two third-person elder clitics can co-occur in Guiloxi, despite other combinations of identicalclitics remaining ungrammatical.

(20) a. Bdel=e’=ne’.hug.COMP=3.EL=3.EL

‘S/he (an elder) hugged her/him (an elder).’ (RM, GZYZ030, 34:15)b. * Ba

alreadybetw=ba’=ba’.hit.COMP=3.HU=3.HU

Intended: ‘S/he (a non-elder) already hit her/him (a non-elder).’ (RM, GZYZ014, 18:18)c. * Ba

alreadybzhige=ba=b.push.COMP=3.AN=3.AN

Intended: ‘It (an animal) already pushed it (an animal).’ (FA and RM, GZYZ014, 19:57)d. * Ba

alreadybdishjw=en=en.break.COMP=3.IN=3.IN

Intended: ‘It (a thing) already broke it (a thing).’ (FA and RM, GZYZ014, 45:05)

3.2 Case 2: Yatzachi

In Yatzachi, the third-person elder clitic also has two forms, conditioned both by thematic and grammatical role.

(21) Strong and clitic pronouns in Yatzachi Zapotec (Butler 1980)

STRONG CLITIC

subject objectagent experiencer

1SG nada’ =a’ =da’ –2SG le’ =o’ =do’ –3.EL l

¯e’ =e’ =ne’ =ne’

3.HU l¯ebo’ =bo’ =bo’ =bo’

3.AN l¯eb =@b’ =b =b

3.IN l¯en =@n =n =n

◦ The form of the third-person elder clitic varies with the thematic role of the external argument: agent vs.“experiencer.”

(22) a. Chol¯sing.CONT

{=a’,=1SG

=o’,=2SG

=e’}.=3.EL

‘{I, You, S/he} am/are/is singing.’ (Butler 1980:56)b. Chyazj-e

need.CONT

{=da’,=1SG

=do’,=2SG

=ne’}.=3.EL

‘{I, You, S/he} need/needs to.’ (Butler 1980:65)

◦ When the clitic cross-references a direct object, it has a form identical to an experiencer clitic.

(23) Goqu@lena’=a=ne’.help.COMP=1SG=3.EL

‘I helped him.’ (Butler 1980:171)

8

Though identical clusters of other third-person clitics are always ruled out (Butler 1980:176–177), elder cliticclusters are only ungrammatical when the subject is an experiencer — when the clitics are identical in form.

(24) a. Chlo’=e’=ne’.teach.CONT=3.EL=3.EL

‘S/he (an elder) teaches her/him (an elder).’b. * Chle’i=ne’=ne’

see.CONT=3.EL=3.EL

Intended: ‘S/he (an elder) sees her/him (an elder).’ (following Butler 1980:176)

3.3 Case 3: Zoogocho

In Zoogocho, the third-person elder clitic has three forms, whose distribution again depends on both thematic andgrammatical role.

(25) Strong and clitic pronouns in Zoogocho Zapotec (Sonnenschein 2004)

STRONG CLITIC

subject objectagent experiencer

1SG neda’ =a’ =da’ –2SG lee (loo) =o’ =do’ –3.EL lhe’ =e’ =de’ =ne’3.HU lhebe’ =be’ =be’ =be’3.AN lheba’ =ba’ =ba’ =ba’3.IN lhen =(e)n =(e)n =(e)n

Any combination of subject and object elder clitics is permitted — because they always differ in their form —though all other identical clitic clusters are impossible (Sonnenschein 2004:54).

(26) a. Naand

dalate

Dolor=en’Dolores=DEF

dxe=e=ne’. . .say.CONT=3.EL=3.EL

‘And the late Dolores said to him. . . ’ (Sonnenschein 2004:384)b. Bi

NEG

ble’e=de’=ne’.see.COMP=3.EL=3.EL

‘S/he didn’t see her/him.’ (Long and Cruz 2000:467)

3.4 Repairing the *X–X Constraint

Across these Zapotec varieties, two featurally identical clitics are allowed in the same cluster as long as theydo not have the same realization, regardless of their grammatical function.

(27) ag. ex. obj.Guiloxi =e’ =ne’Yatzachi =e’ =ne’Zoogocho =e’ =de’ =ne’

ag.� obj. ex.� obj.Guiloxi � �Yatzachi � *Zoogocho � �

9

Within Distributed Morphology, the ill-formedness of certain combinations might be derived by positing no Vo-cabulary Items for a specific syntactic terminal, resulting in ineffability (Nevins 2014, Arregi and Nevins 2014).

(28) a. GuiloxiCl

[3.EL]↔ =e’ / V

↔ =ne’ /{

Cl[1SG]

,Cl

[2SG],

Cl[3.EL]

}b. Yatzachi

Cl[3.EL]

↔ =e’ /V

[θ AG]

↔ =ne’ /

V[θ EX]{

Cl[1SG]

,Cl

[2SG]

}V

[θ AG]Cl

[3.EL]

c. Zoogocho

Cl[3.EL]

↔ =e’ /V

[θ AG]

↔ =de’ /V

[θ EX]

↔ =ne’ /{

Cl[1SG]

,Cl

[2SG],

Cl[3.EL]

}

However, this misses the generalization that, in every variety we examine, featurally identical clitics may occur ina cluster if and only if they do not have the same form.

Instead, we propose that there is a general constraint on Vocabulary Insertion in these languages, which penalizescombinations of clitics with the same realization.

(29) *X–X ConstraintFor two clitics D and D′ in the same cluster, D and D′ cannot be exponed by the same Vocabulary Item.

Together with some much simpler Vocabulary Items, this derives the pattern of prohibited clitic clusters acrossthese Zapotec varieties in a maximally general way.2

2Support for this account comes from imperatives in Guiloxi, which do not allow an overt subject, even if this is a clitic pronoun. Athird-person elder object can only be exponed as a clitic pronoun by the same Vocabulary Item as a subject in declarative clauses.

(1) a. Bdel=e’!hug.COMP=3.EL

‘Hug her/him!’ (Guiloxi: FA and RM, GZYZ030, 33:44)b. * Bdel=ne’!

hug.COMP=3.EL

Intended: ‘Hug her/him!’ (Guiloxi: FA and RM, GZYZ030, 33:48)

In other words, the =e’ realization occurs in initial position, while the =ne’ realization occurs in non-initial position in a cluster.

10

(30) a. GuiloxiCl

[3.EL]↔ =ne’ / Cl

↔ =e’

b. YatzachiCl

[3.EL]↔ =ne’ /

{V

[θ EX], Cl}

↔ =e’

c. ZoogochoCl

[3.EL]↔ =ne’ / Cl

↔ =de’ /V

[θ EX]

↔ =e’

Despite the morphological nature of the *X–X Constraint — it is a constraint on Vocabulary Insertion — its repairis periphrasis (see Butler 1980:176–177 for Yatzachi, Avelino Becerra 2004:34 and Lopez and Newberg 2005:9for Yalalag, and Sonnenschein 2004:38 for Zoogocho).

(31) Guiloxia. Bdel=ba’

hug.COMP=3.HU

leba’.3.HU

‘S/he hugged her/him.’ (RM, GZYZ012, 36:00)b. Ba

alreadybzhig=bpush.COMP=3.AN

leb.3.AN

‘It already pushed it.’ (FA and RM, GZYZ014, 19:13)c. Ba

alreadybdishjw=enbreak.COMP=3.IN

len.3.IN

‘It broke it.’ (RM and FA, GZYZ014, 44:35)

The grammatical source of the *X–X Constraint is not reflected in how its violations are repaired: contrathe Division of Labor, it is remedied by periphrasis, despite being postsyntactic in nature.

4 The Gender–Case Constraint

Clitic clusters in Zapotec are also governed by a family of Gender–Case Constraints (GCCs; Toosarvandani 2017,Foley et al., to appear).

Much like the more familiar Person–Case Constraint (Perlmutter 1971, Bonet 1991), the GCC prohibits certaincombinations of clitics based on their position on a feature hierarchy.

11

(32) Zapotec gender hierarchy3.EL 3.HU 3.AN 3.IN

γ

AN

HU

EL

>

γ

AN

HU

>

[γ

AN

]>

[γ]

In Yalalag, for example, the GCC prohibits all clusters where the object clitic outranks the subject clitic on thegender hierarchy.

(33) Yalalag: (1 ≹ 2 >) 3.EL > 3.HU > 3.AN > 3.IN

OBJECT

3.EL 3.HU 3.AN 3.IN

SU

BJE

CT 3.EL V=e’. . . le’e V=e’=be’ V=e’=ba’ V=e’=n

3.HU V=be’. . . le’e V=be’. . . lebe’ V=be’=ba’ V=be’=n3.AN V=ba’. . . le’e V=ba’. . . lebe’ V=ba’. . . leba’ V=ba’=n3.IN V=en. . . le’e V=en. . . lebe’ V=en. . . leba’ V=en. . . len

As shown above, violations of the GCC are repaired through periphrasis — use of a strong object pronoun —just as violations of the *X–X Constraint are.

The grammaticality of clitic clusters with identical gender features in some Zapotec languages can help usdecide between alternative theories of the GCC — and, by extension, the PCC as well.

A Multiple Agree account

One prominent approach to the PCC makes use of an intervention condition on Multiple Agree to derive ungram-matical clitic clusters (Anagnostopoulou 2003, Nevins 2011, 2014).

Elsewhere (Foley et al., to appear), we argue that such an intervention constraint is well-suited to account formicrovariation in the GCC across Zapotec varieties. There are two crucial ingredients in this account:

◦ Clitic pronouns must be licensed through Agreement with a ϕ-probe; unlicensed clitics crash the derivation.

◦ A probe may license more than one clitic simultaneously via Multiple Agree, so long as the followingintervention condition is satisfied, which is a revised version of Nevins’ (2011, 2014) Contiguous Agree:

(34) Intervention Constraint on Multiple AgreeFor a probe P relativized to features [F]/[F′], where [F′] ⊆ [F], with a goal G that bears [F], therecan be no G′ such that:(i) P c-commands G′ and G′ c-commands G, and(ii) G′ does not bear [F′].

Simplifying somewhat, when a probe is relativized to a single gender feature [F], it can license an object clitic justin case the subject clitic has [F] — that is, it is at least as high as F on the gender hierarchy.

12

(35) a. 3.EL� 3.HU

P[EL]/[EL]

�Clγ

AN

HU

EL

�Cl γ

AN

HU

b. *3.HU� 3.EL

P[EL]/[EL]

Cl γ

AN

HU

LClγ

AN

HU

EL

Importantly, combinations of clitics with identical gender features satisfy the intervention condition, so that X�X combinations are always syntactically well-formed.

(36) 3.EL� 3.EL

P[EL]/[EL]

�Clγ

AN

HU

EL

�Cl

γ

AN

HU

EL

This is a desirable consequence: some X� X combinations are grammatical in some Zapotec varieties (Guiloxi,Yatzachi, and Zoogocho). Others, of course, are filtered postsyntactically by the *X–X Constraint.

An alternative account

Not all theories of the PCC will admit X� X combinations. Walkow (2012), for instance, proposes a theory ofthe PCC in Classical Arabic that can be extended to account for the GCC in Yalalag:

◦ Clitics must be licensed by valuing some feature on a ϕ-probe; unlicensed clitics crash the derivation.

◦ A probe Agrees cyclically (Bejar and Rezac 2009), first with the lower clitic and then with the higher one.

For a 3.EL� 3.HU combination, a fully specified gender probe is able to Agree with, and license, both clitics.

(37) 3.EL� 3.HU

Clγ

AN

HU

EL

P

uγ

u AN

u HU

u EL

�Cl γ

AN

HU

⇒

�Clγ

AN

HU

EL

P

uγ

u AN

u HU

u EL

�Cl γ

AN

HU

13

But for a *3.HU � 3.EL combination, the probe is satisfied after Agreeing with just the object clitic, so that thesubject clitic remains unlicensed.

(38) *3.HU� 3.EL

LCl γ

AN

HU

Puγ

u AN

u HU

u EL

�Cl

γ

AN

HU

EL

However, this system also rules out all X� X combinations. No matter the probe’s relativization, a subject willnever be able to value any features on the second cycle of Agree if it is featurally identical to the object.

(39) *3.EL� 3.EL

LCLγ

AN

HU

EL

P

uγ

u AN

u HU

u EL

�CL

γ

AN

HU

EL

Ruling out X � X combinations on syntactic grounds necessarily undergenerates. Clusters that would surfacegiven the right morphological properties are nipped in the bud.

Therefore, any theory of the GCC (or the PCC) must not categorically rule out clitic clusters with identicalfeatures.3

3We may not want to make the stronger claim that a theory of the PCC and GCC must categorically rule in such clusters. For SantiagoLaxopa Zapotec, it may be necessary for some X � X combinations to violate the intervention condition on Multiple Agree. In thisvariety, all 3.EL� 3.EL clusters are ungrammatical, even when the clitics have non-identical realizations.

(i) * Blen=e’=ne’.hug.COMP=3.EL=3.EL

‘S/he hugged her/him.’ (FSR, SLZ1012, 14:37)

This suggests that the GCC can, in some languages, rule out clusters whose subject clitic does not strictly outrank the object clitic.

14

5 Conclusions and further consequences

These Zapotec varieties provide a new perspective on two constraints on clitic combinations:

◦ The GCC has a syntactic source in an intervention condition on Agree, while the *X–X Constraint is mor-phological in nature, constraining Vocabulary Insertion.

◦ Despite their different grammatical sources, the GCC and *X–X Constraint have the same repair — pe-riphrasis — which does not support the Division of Labor.

So far, while we have talked about the postsyntactic character of the *X–X Constraint, we have not addressed theprecise mechanism that leads to its repair.

Two possibilities:

1. The *X–X Constraint acts as a filter on syntactically convergent derivations, cf. adjacency requirements formorphological merger (Bobaljik 1995). The intended meaning is then expressed through a distinct deriva-tion involving periphrasis.

2. If, however, the morphology must always deliver an output, the repair might instead involve the spellout ofa lower copy of the clitic’s movement chain (Bonet 1991).

There are subtle ways of teasing these two possibilities apart (Rezac 2011), which we intend to pursue next.

Acknowledgements

We are grateful to Flavio Antonio, Rosa Mendoza, and Fe Silva Robles for teaching us about their language. Weare also grateful to Heriberto Avelino Becerra for contributing data from Yalalag Zapotec.

ReferencesAdler, Jeff, Steven Foley, Jed Pizarro-Guevara, Kelsey Sasaki and Maziar Toosarvandani. 2017. The derivation

of verb-initiality in Santiago Laxopa Zapotec. Annual Meeting of the Society for the Study of the IndigenousLanguages of the Americas, Austin.

Anagnostopoulou, Elena. 2003. The syntax of ditransitives: Evidence from clitics. Berlin: Mouton de Gruyter.Arregi, Karlos and Andrew Nevins. 2012. Morphotactics: Basque auxiliaries and the structure of spellout. Dor-

drecht: Springer.Arregi, Karlos and Andrew Nevins. 2014. A monoradical approach to some cases of disuppletion. Theoretical

Linguistics 40:311–330.Avelino Becerra, Heriberto. 2004. Topics in Yalalag Zapotec, with particular reference to its phonetic structure.

Ph.D. Dissertation, University of California, Los Angeles.Bejar, Susana and Milan Rezac. 2009. Cyclic Agree. Linguistic Inquiry 40:35–73.Bobaljik, Jonathan. 1995. Morphosyntax: The syntax of verbal inflection. Ph.D. Dissertation, Massachusetts

Institute of Technology.Bonet, Eulalia. 1991. Morphology after syntax: Pronominal clitics in Romance. Ph.D. Dissertation, Massachusetts

Institute of Technology.

15

Butler, Inez M. 1980. Gramatica zapoteca: Zapoteco de Yatzachi el Bajo. Mexico City: Instituto Linguıstico deVerano.

Foley, Steven, Nick Kalivoda and Maziar Toosarvandani. To appear. Gender–Case Constraints in Zapotec. Work-shop on Structure and Constituency in Languages of the Americas (WSCLA) 22.

Kalivoda, Nick. 2015. Distinguishing doubled clitics and pronominal arguments in Teotitlan del Valle Zapotec.Ms., University of California, Santa Cruz.

Lee, Felicia. 2006. Remnant raising and VSO clausal architecture: A case study of San Lucas Quiavinı Zapotec.Amsterdam: Springer.

Long, Rebecca and Sofronio Cruz. 2000. Diccionaria Zapoteco de San Bartolome Zoogocho Oaxaca. MexicoCity: Instituto Linguıstico de Verano.

Lopez, Filemon and Ronaldo Newberg. 2005. La conjugacion del verbo zapoteco, zapotevco de Yalalag. MexicoCity: Instituto Linguıstico de Verano.

Marlett, Stephen A. 2010a. Personal pronouns: Inventory. In The Zapotec grammar files, eds. Cheryl A. Black,H. Andrew Black, and Stephen A. Marlett. SIL International.

Marlett, Stephen A. 2010b. The la’a ‘base’ plus enclitic pronoun phrase. In The Zapotec grammar files, eds.Cheryl A. Black, H. Andrew Black, and Stephen A. Marlett. SIL International.

Nevins, Andrew. 2007. The representation of third person and its consequences for person-case effects. NaturalLanguage and Linguistic Theory 25:273–313.

Nevins, Andrew. 2011. Multiple agree with clitics: Person complementarity vs. omnivorous number. NaturalLanguage and Linguistic Theory 29:939–971.

Nevins, Andrew. 2014. Book review: “Defective paradigms: Missing forms and what they tell us”. RevistaLinguıstica 10:29–36.

Ormazabal, Javier and Juan Romero. 2007. The object agreement constraint. Natural Language and LinguisticTheory 25:315–347.

Perlmutter, David. 1971. Deep and surface structure constraints in syntax. New York: Holt, Reinhart, andWinston.

Preminger, Omer. 2017. What the PCC tells us about “abstract” agreement, head movement, and locality. Ms.,University of Maryland, URL https://ling.auf.net/lingbuzz/003221.

Rezac, Milan. 2011. Phi-features and the modular architecture of language. Dordrecht: Springer.Sonnenschein, Aaron. 2004. A descriptive grammar of San Bartolome Zoogocho Zapotec. Ph.D. Dissertation,

University of Southern California.Toosarvandani, Maziar. 2017. On reaching agreement early (and late). In Asking the right questions: Essays in

honor of Sandra Chung, eds. Jason Ostrove, Ruth Kramer, and Joseph Sabbagh, 124–138. Santa Cruz, CA:Department of Linguistics, University of California, Santa Cruz.

Torrego, Esther. 1992. Case and agreement structure. Ms., University of Massachusetts, Boston.Uriagereka, Juan. 1995. Aspects of the syntax of clitic placement in Western Romance. Linguistic Inquiry 26:79–

123.Walkow, Martin. 2012. Cyclic agree derives restrictions on cliticization in Classical Arabic. In Papers from Arabic

Linguistics XXVI: Papers from the annual symposium on Arabic linguistics, 2012, eds. Reem Khamis-Dakwarand Karen Froud, 135–160. Amsterdam: John Benjamins.

16