Relatively Speaking (in Circassian)

Ivano Caponigro and Maria Polinsky

University of California, San Diego and Harvard University

1. Taking things for granted?

It is a standard assumption that subordinate clauses like relative, embedded declarative, and

interrogative clauses in natural language vary according to their morphosyntactic and semantic

characteristics. Their morphosyntactic variation is expressed in the presence/absence of a wh-word, a

relative pronoun, or an overt complementizer; syntactic transparency (embedded declaratives) or

opacity (embedded interrogatives and relative clauses); differences in the nature of the

complementizer (Rizzi 1990: 45) and differences in the relationship between the embedded clause and

the matrix clause (complementation vs. adjunction). These embedded clauses are also assigned very

different interpretations. A headed relative like which Adam likes is standardly assumed to denote the

set of inanimate individuals that Adam likes (Quine 1960; Montague 1973); a free/headless relative

like what Adam likes denotes the (plural) inanimate individual that Adam likes (Jacobson 1995,

Caponigro 2004); an embedded declarative like that Adam likes vegetables denotes the proposition

'that Adam likes vegetables'; an embedded polar interrogative like whether Adam likes vegetables

denotes a set containing the proposition 'that Adam likes vegetables' and/or its negation 'that Adam

does not like vegetables'; finally, an embedded constituent interrogative like which food Adam likes

denotes the set of propositions that are appropriate (true) answers to the question 'which food does

Adam like?' (see Hamblin 1973 and Karttunen 1977 for the semantics of both types of interrogatives).

We present and analyze new empirical evidence suggesting that such morphosyntax/semantics

mapping may not be universal. The evidence comes from Adyghe, a Northwest Caucasian language,

in which what looks and behaves like the very same morphosyntactic construction (henceforth

"mystery clause", MC) is used to convey the five different meanings above. This raises at least two

main questions: (i) Is the MC truly one and the same construction and what construction is it? This is

the issue this paper focuses on. (ii) How is the same construction mapped into very different

meanings? For reasons of space, we cannot address this issue here but the interested reader is referred

to Caponigro and Polinsky (to appear).

The paper is structured as follows. Section 2 gives a brief background on Adyghe. Section 3

introduces the main morphological and syntactic properties of the MC. Section 4 develops our

proposal concerning the syntax of the MC. Section 5 concludes.

2. A brief introduction to Adyghe

Adyghe (ady; also known as West Circassian), spoken by about 150,000 people in the south of

Russia, is a morphologically rich language with an absolutive/ergative case system (syncretic for first

and second person). Nouns are inflected for what we will refer to as “specificity” (currently, it is

unclear to us what the precise semantic import of case marking is). Specific DPs have overt marking

both in the ergative (ERG) and absolutive (ABS), as shown in (1); non-specific forms of both cases

have zero marking, as shown in (2).

Other cases include the generalized oblique (-m) and

instrumental -č'e, always overtly marked.

∗

This work was supported in part by the NSF grant BCS-0131993 and a grant from the Davis Center for

Russian and Eurasian Studies at Harvard to the second author. We are grateful to our language consultants

Raxmet Esheva, Raxmet Gisheva, Svetlana Kinokova, Mira Unarokova, and Acherdan Abzhegov. We would like

to thank R. Bhatt, G. Chierchia, D. Gerasimov, A. Giannakidou, S.-S.Kim, Y. Lander, A. Letuchiy, E. Potsdam,

Y. Testelets, and the audiences at WCCFL 27, GLOW 31, and SALT XVIII for helpful comments on this paper.

Standard disclaimers apply.

3rd person absolutive has a null agreement marker, and we will not show it beyond (1) through (3).

Linguistics, ed. Natasha Abner and Jason Bishop, 81-89. Somerville, MA: Cascadilla Proceedings Project.

(1) B’ale-m mE maSine-r Ø-E-qWEta-R

boy-

ERG this car-ABS 3SG.ABS-3SG.ERG-break-PAST.DECL

‘The boy broke this car.’

(2) B’ale-xe-Ø maSine-Ø Ø-a-qWEta

boy-

PL-ERG car-ABS 3PL.ABS-3PL.ERG-break.PRES.DECL

‘Boys break cars.’

Verbal morphology is particularly complex. A verb can agree in person/number with subject,

object, and indirect object (cf. Colarusso 1992: 74, 132-135; O’Herin 2002: 49-69 for agreement in

the closely related Kabardian and Abaza) and has separate positional slots for negation, tense, aspect,

causation, mood, and illocutionary force (Smeets 1984: ch. 5, 6; Rogava and Keraševa 1966: 95-331).

In addition, Adyghe has a rich system of applicative heads (traditionally referred to as preverbs, see

Smeets 1984: 256-67) that incorporate into the verbal complex (see O’Herin 2001 for similar forms in

Abaza). These applicatives take an oblique complement and show agreement with it. The scarcity of

postpositions may be a trade-off of the articulated applicative system; most of the phrases

corresponding to PPs in other languages have to be expressed by a dedicated applicative marker in the

verb and its complement in the oblique case. Applicatives are prefixal and their order mirrors the

syntactic hierarchy: the leftmost applicative is always the highest, as shown in (3).

(3) Ø-t-de-p-fE-a-SHe-R

SG.ABS-1PL-COM-2SG-BEN-3PL.ERG-eat-PAST

[

ApplP-1

1PL-COM-[

ApplP-2

2SG-BEN-[

3PL.ERG-[

eat]]]]-PAST

‘They ate this with us for you.’

Word order in root clauses is extremely free (for instance, in matrix declarative clauses such as

(1), all six word orders are possible), however embedded clauses must be verb-final. The language has

extensive subject and object pro-drop. Matrix polar interrogatives are formed by adding the

interrogative particle (Q) to the predicate, as shown in (4).

(4) B’ale-m mE maSine-r E-qWEta-R-a

boy-

ERG this car-ABS 3SG.ERG-break-PAST-Q

‘Did the boy break this car?’

Matrix constituent interrogatives are clefts. Compare the biclausal structure in (5): the wh-word is

in the predicate position, with the Q marker attached, and either preceeding or following the remainder

of the sentence is a headless relative clause, as shown in (5a) and (5b) respectively.

(5) a. [

[

B’ale-m

E-qWEta-Re]-r ] sEd-a

boy-

ERG 3ABS-3SG.ERG-break-PAST-ABS what-Q

b. sEd-a [

[

B’ale-m

-E-qWEta-Re]-r]

what-Q

boy-ERG 3ABS-3SG.ERG-break-PAST-ABS

‘What did the boy break?’ (Lit. ‘What is it the thing the boy broke?’)

Externally headed relative clauses are prenominal (on internally headed relatives, see Lander

2004), and the relativized argument is cross-referenced by special agreement on the verb in the

relative clause which we will refer to as “wh-agreement” (see O'Herin 2002: ch. 8 for a similar

phenomenon and its analysis in Abaza; see Hewitt 1979 for a similar pattern in Abkhaz).

Wh-agreement is part of the regular verbal agreement paradigm: wh-agreement prefix appears in

the same positional slot as the person agreement marker for the relevant argument (ergative subject,

absolutive subject/object, oblique object, complement of an applicative). However, unlike regular

person agreement, there can be only one wh-agreement marker on a given verb. Unlike person

agreement, wh-agreement indicates the presence of a silent constituent (gap) in a certain syntactic

relation to the verb bound (i.e. coindexed with and c-commanded) by an operator inside the clause.

The gap in the absolutive position is indexed by a null marker, as in (5). The gap in the ergative or

absolutive position, instead, is indexed by the marker z(j)ə- or ze-, which we hypothesize can be

decomposed into the actual wh- marker (z-) and person marker, as shown in (6) and (7).

(6) [

[

mE maSine-r z-E

-qWEta-R] B’ale-r]

this car-

ABS WH-ERG-break-PAST boy-ABS

‘the boy who broke this car’

(7) [

[

B’ale-m

mE maSine-r z-e

-re-jE-qWEtE-R]

boy-

ERG this car-ABS WH-OBL-INSTR-3SG.ERG-break-PAST

enC’E-r]

gun-

ABS

‘the gun that boy broke this car with’

Wh-agreement is marked not only on the verb, which has a null argument, but can also be marked

on the coreferent constituents in the c-command domain of that element. Compare (6) above and (8),

(9); in (8), the boy is the owner of the car, and in (9) the owner is distinct from the hooligan:

(8) [

[

-maSine z-E

-qWEta-R] B’ale-r]

WH-car.ABS WH-ERG-break-PAST boy-ABS

‘the boy who

broke his

car’

(9) [

[

-maSine z-E

-qWEta-R] B’ale-r]

SG.POSS-car.ABS WH-ERG-break-PAST boy-ABS

‘the boy who

broke his

car’

To summarize, Adyghe verbal morphology indexes the main verbal arguments; applicative heads

incorporate into the verb and show agreement with their complements (always non-absolutive); and

finally, the language has special wh-agreement which indexes a gap bound by an operator. Let’s now

turn to the ways this complex morphological machinery is used.

3. Mystery clauses

3.1. MCs as relative clauses

The first subset of mystery clauses is quite prosaic: headed and headless relative clauses. We just

introduced relative clauses in our discussion of wh-agreement. We will now briefly discuss their

analysis. Adyghe relative clauses do not show reconstruction effects and do not preserve idiomatic

meaning under relativization:

(10) a. E-pe hWEzE-r qErexE

SG.POSS-nose smoke-ABS blow.PRES

‘S/He is furious.’ (lit. “Smoke is coming out of his/her nose.”)

b. [E-pe

qErexE] hWEzE-r

SG.POSS-nose blow.PRES smoke-ABS

‘the smoke that is coming out of his/her nose' (literal meaning)

# ‘his/her anger' (idiomatic)

These data argue against the head-raising analysis (Kayne 1994, Bianchi 2002, a.o.) and point to

the operator-variable analysis of relative clauses. On the operator analysis, there are two further

possibilities: the operator moves to the periphery of the relative clause (11a); or it is base-generated

there and binds an empty category in the clause (11b).

(11) a. [

[

… ]] MOVEMENT ANALYSIS

b. [

[

pro

… ]] BINDING ANALYSIS

At this juncture, we do not have sufficient evidence to distinguish between these two possibilities;

Adyghe also has homophonous prefix ze-, reflexive/reciprocal; this prefix can co-occur with the ze-

discussed here (Smeets 1984: 256).

while the movement analysis is more traditional, we should not lose sight of the fact that Adyghe has

very rich pro-drop and that unselective binding of a null pronominal indexed on the verb by agreement

cannot be ruled out. Both analyses predict island effects, and as (12) shows, this prediction is borne

out: while scrambling in the root clause is free (12b), scrambling out of the relative clause is

ungrammatical (12c):

(12) a. [C’ale-m tERWase

qEsfjE{WetaRe] qebar-er

njEpe zexesxRere

boy-

ERG yesterday spread.PAST rumor-ABS today reaches.me

‘The rumor that the boy spread yesterday has reached me today.’

b. njEpe zexesxRere [C’alem tERWase

qEsfjE{WetaRe] qebarer

today reaches.me boy yesterday spread rumor

‘The rumor that the boy spread yesterday has reached me today.’

c. * C’alem

njEpe zexesxRere [ t

tERWase

qEsfjE{WetaRe] qebarer

boy-

ERG today reaches.me yesterday spread rumor

Free relatives are very similar to headed relatives; they just do not have an overt head and the case

marker attaches directly to the right edge of the relative clause. They appear in wh-questions, which

are clefts (see ex. (5) above). Just as headed relatives, free relatives are syntactic islands (13).

(13) mjErE [

[

mE maSine-r z-E-qWEta-Re]-r] jEwaR

Mira this car-

ABS WH-ERG-break-PAST-ABS beat

‘Mira beat the one/those who broke this car.’

In sum, Adyghe allows free relatives and prenominal headed relatives, which are fully finite (cf.

the tense marking in all the relatives above), a rather unsurprising situation paralleled by many head-

final languages.

3.2. MCs as “embedded matrix declarative clauses”

The mystery starts to develop when we look at MCs in other positions. Adyghe has a series of

non-finite embeddings with the adverbial ending -ew (Smeets 1984; Lander 2004; Polinsky and

Potsdam 2006) and direct quotations, but embedding of finite clauses like declaratives is impossible.

In fact, descriptions of Northwest Caucasian emphasize that these languages lack embedding

complementizers (Jakovlev and Ašhamaf 1940, Rogava and Keraševa 1966, Gerasimov and Lander

2008) and what appears as an embedded finite clausal complement in other languages is introduced

either as direct speech or as a specially marked verb form. For instance, the matrix verb ‘tell’ can take

a DP object (14a), but not an embedded clause, as shown in (14b) where we try to embed our initial

matrix declarative in (1):

(14) a. qebar-er Q-qE-s-jE{Wa-R

news/rumor-

ABS 3SG.ASB-DIR-1SG.IO-3SG.ERG-tell-PAST

‘S/he told me the news/rumor.'

b.* [B’ale-m mE maSine-r E-qWEta-R] qE-s-jE{WaR

boy-

ERG this car-ABS 3SG.ERG-break-PAST DIR-1SG.IO-3SG.ERG-told

(‘S/he told me that the boy broke that car.’)

In order to embed the proposition expressed by a declarative sentence like (1), the embedded

predicate has to carry the following two extra markers, as shown in (15): (i) the additional prefix zere-,

which Smeets characterizes as ‘that/how’ (1984: 254), and (ii) the case suffix assigned by the

embedding predicate (absolutive in (15), just as the DP complement in (14a)).

(15) [

[B’ale-m mE maSine-r *(zere)-qWEta-Re]-*(r)] qEsjE{WaR

boy-

ERG this car-ABS zere-break-PAST-ABS told.me

‘S/he told me that the boy had broken this car.’

Thus, a special MC, which occurs in a DP position and looks identical to a headless relative (cf. (15)

with (13) above), is needed to express an embedded proposition. Further evidence that this MC occurs

in the DP position, not CP position, comes from its co-occurrence with postposition (16a) or as

possessor DP (16b):

(16) a. [

[ EqWe qe-zere-KWeZ’e-Re-]m paje] jane mEgWEIWe

son.

ABS DIR-zere-come_back-PAST-OBL for mother glad.PRES

‘The mother is glad that her son is back (lit.: glad for).’ (Sumbatova 2005)

b. [

[

B’aler qE-zere-KWeZ’e-S’t-E]m] jE-IepqE] -r]

boy.ABS DIR-zere-come_back-FUT-OBL POSS-verity-ABS

‘the validity/news/fact of the boy arriving’

Even more surprisingly, the very same form is observed in embedded polar interrogatives, as

discussed in the next section.

3.3. MCs as “embedded polar interrogatives”

Like declaratives, polar interrogatives cannot be embedded directly, with or without the

interrogative:

(17) * [ B’ale-m mE maSine-r EqWEtaR(-a)] qEKeWEpBaR

boy-

ERG this car-ABS broke-Q asked

('S/he asked if the boy had broken this car.’)

Instead, they can either be introduced as direct quotations or appear in the nominal form. In what

follows, the verb ‘ask/inquire’ takes an oblique object (18a), and a complex noun phrase with the same

mystery verb form as above can appear in that position (18b):

(18) a. [

sE-Ve–m] qEKeWEpBaR

SG.POSS-name-OBL asked

‘S/he asked my name.’

b. [

[ B’ale-m mE maSine-r *(zere)-qWEta-Re]-*(m)] qEKeWpBaR

boy-

ERG this car-ABS zere-break-PAST-OBL asked

‘S/he asked if the boy had broken this car.’

The form of the embedding in (18b) is exactly the same as that in (15) above, the only difference has

to do with the case assigned by the matrix verb (ABS -r by ‘tell’, OBL -m by ‘ask’).

3.4. MCs as “embedded constituent interrogatives”

We have already seen that polar interrogatives cannot embed directly, so it comes as no surprise

that constituent interrogatives do not embed either, regardless of the order or presence of the

interrogative marker:

(19) * [[

[ mE maSine-r z-E-qWEta-Re] Het-(a)] qEKeWpBaR

this car-

ABS WH-ERG-break-PAST who-Q asked

(‘S/he asked who broke this car.’)

The meaning of an embedded constituent interrogative can be conveyed as follows:

(20) mjErE [

[mE maSine-r z-E-qWEta-Re]-m] qEKeWpBaR

Mira this car-

ABS WH-ERG-break-PAST-OBL asked

‘Mira asked who broke this car.’

This MC is identical to headless relatives we have already seen in (13) above: it has the overt

wh-marking indicating that the gap is in the ergative position, and it is in the oblique case, as

determined by the matrix verb. If Mira’s question is about the broken car, the MC looks as follows:

(21) mjErE [

[C’ale-m Ø-E-qWEta-Re]-m] qEKeWpBaR

Mira boy-

ERG WH.ABS-3SG.ERG-break-PAST-OBL asked

‘Mira asked what the boy broke.’

Questioning about other constituents is also expressed by MCs that look like headless relatives:

(22) mjErE [[C’alem mE maSiner zE-d-jE-qWEta-Re]-m] qEKeWpBaR

Mira boy.

ERG this car.ABS WH-COM-3SG.ERG-break-PST-OBL asked

‘Mira asked with whom the boy broke this car.’

Unlike matrix interrogatives, no overt wh-word can occur in embedded interrogatives. Both in root

and embedded structures, multiple wh-questions are impossible, and have to be expressed by

coordination of single wh-questions:

(23) mjErE [[z-E-qWEta-Re]-m]-jE [[Ø-E-qWEta-Re]-m]-jE qEKeWpBaR

Mira

WH-ERG-break-PST-OBL-and WH.ABS-3SG.ERG-break-PAST-OBL-and asked

‘Mira asked who broke and what.’ (no pair list reading)

The surface form of embedded interrogatives offers the initial key to our mystery clauses. Embedded

wh-questions seem to be expressed by headless relative clauses, essentially identical to the relative

clause used in the root cleft question (see (5) above), and the case of the headless relative clause is

determined by the matrix verb. This leads us to the proposal concerning the structure of Adyghe MCs.

4. The syntax of Adyghe mystery clauses

4.1. All MCs are relative clauses

Our proposal is that all the MCs in Adyghe are complex DPs containing a relative clause with or

without a nominal head (N) as the complement of the determiner head (D):

(24) [

[

Relative Clause] (N)] D]

The evidence in support of the DP nature of MCs is threefold. First, the –r/-m marker, which

occurs as a suffix on standard DPs, has to appear at the right edge of the MC as well (unless there is an

overt head noun). Second, MCs functioning as embedded declaratives or interrogatives have the same

distribution as DPs, namely, they can occur as complements of postpositions (17), which would be

unexpected for CP complements. Finally, these complements conjoin the same way as DPs, with the

particle –(r)j

(23). Unlike DPs, CPs do not coordinate—for instance, two relative clauses can neither

stack nor conjoin.

As far as the nominal head goes, we are assuming it is optional since, although often missing,

sometimes it occurs: for instance, the sentence in (21) above can be also paraphrased as having an

overt head noun ‘thing,’ and (22) can have an overt head noun ‘person,’ as shown in (25). “Embedded

declarative” and “embedded polar interrogative” MCs can have a nominal head too. It is represented

by abstract nouns like ‘news,’ ‘verity,’ ‘statement,’ as shown in (26).

(25) mjErE [[C’alem mE maSiner zEdjEqWEtaRe] cEfE-m ] qEKeWpBaR

Mira boy.

ERG this car.ABS WH-COM.broke person-OBL asked

‘Mira asked with whom (lit.: what person) the boy broke this car.’

(26) [[C’ale-r njEpe qE-zere-KweZjE-S’tE] (IEpqE)-r aS’ qEsjE{WaR

boy-

ABS today DIR-zere-come-FUT verity-ABS 3SG.ERG said

‘S/he said that the boy was coming back today.’

MCs like the one in (26) could hold a very different relation to this overt head than the head of a

The reasons why relative clauses do not stack are not entirely clear, but it is worth noting that the stacking

of relative clauses, generally available for postnominal relatives (Hudson 1990; Sag 1997), is difficult or even

impossible in head final languages (Suzuki 2005, Larson and Takahashi 2005).

relative clause: the MC could be a complement of the abstract noun, just as it is the case with familiar

fact-clauses across languages, which superficially resemble relative clauses (Kiparsky and Kiparsky

1970, a.o.). However, two observations suggest that this is not the right proposal and that MCs with

nominal heads are actually relative clauses. First, in languages where complement clauses can appear

with and without a head noun, the two constructions have different distribution and interpretation (e.g.

She remembered/*thought the (fact/claim/rumor) that it was sunny); in Adyghe, there is no

interpretive difference between the variants of (26) with and without

IEpqEr

. Second, complement

clauses without the head noun are syntactically transparent, as shown by the English translation of

(27) and (28). On the other hand, all Adyghe MCs, even “embedded declaratives”, are strong syntactic

islands, as shown by the only available reading in (27) and the unacceptability of (28):

(27) njEpe aS’ [[ C’ale-r qE-zere-KweZjE-S’tE] qEsjE{WaR

today 3

SG.ERG boy-ABS DIR-zere-come-FUT said

‘Today, s/he said that the boy was coming back.’ (said today/*return today)

(28) * aS’ [[ C’ale-r sEdjERWa qE-zere-KweZjE-S’tE] qEsjE{WaR

SG.ERG boy-ABS when DIR-zere-come-FUT said

(‘When did she say the boy was coming back?’)

One could counter this with the proposal that all MC, regardless of the overt head noun, resist

extraction as complex DPs. However, we find that DPs and NPs in Adyghe are actually transparent; as

(29) shows, subconstituents of a DP can scramble. Therefore, the DP status fails to account for the

strong island effects in MCs.

(29) a. we-I&-a sSa-m jE-prezjEdent-Er ?

SG-know-INTERR USA-OBL POSS-president-ABS

b. jE-prezjEdent-Er

we-I&-a [ t

sSa-m] ?

POSS-president-ABS 2SG-know-INTERR USA-OBL

c. sSa-m

we-I&-a [ t

jE-prezjEdent-Er ]?

USA-

OBL 2SG-know-INTERR POSS-president-ABS

‘Do you know the president of the USA?’

To sum up, MCs that behave like relative clauses have wh-agreement on the verb. MCs that

behave like embedded wh-interrogatives are identical to relative clauses, in particular they exhbit the

same wh-marking. MCs that behave like embedded declaratives and polar interrogatives carry the

invariable marker ze-re, which appears at the left periphery of the embedded verb. We suggest that ze-

is the usual marker of wh-agreement (fused with the oblique marker), and re- is a high applicative

whose function we discuss below. If this is on the right track, then the zV- marking on predicates of all

MCs is that of wh-agreement. On the basis of their distributional properties, we concluded that MCs

are all DPs. Next, based on the syntactic opacity (which is not found in complex DPs), uniform wh-

agreement, and constraint against stacking, we conclude that these complex DPs all contain a relative

clause with an operator-variable configuration.

4.2. What relativizes in the ze-re- clause?

In order for this proposal to go through, we need to consider what can relativize in ze-re- clauses,

the ones we encounter as embedded declaratives and embedded polar interrogatives. We just

concluded that ze- in ze-re- is the usual wh-agreement that signals/licenses an operator-variable

configuration. What is -re- though and what does it contribute? Recall that the morphological order of

applicatives mirrors the syntactic hierarchy, with the leftmost applicative corresponding to the highest

position. Whatever variable is indexed by ze-re-, it is quite high, since it is the highest applicative. We

suggest that -re- signals and licenses a specific kind of operator-variable configuration: a variable over

polarity operators, as shown in (30).

For reasons of space, we illustrate just with one example, but all the MCs are equally strong islands.

Gerasimov and Lander 2008 propose that ze- indexes the (silent) noun ‘fact’ and -re- instantiates one of the

functions of the instrumental operator. The abstract head noun ‘fact’, however, is incompatible with polar

(30) [

[

i-polarity

[

… [

ApplP

[Appl° re] [

… ]]]]]

A polarity operator is a function that takes a proposition p and returns either the very same

proposition p (positive polarity operator) or its negation ~p (negative polarity operator). The whole

mystery clause would end up denoting a set containing the two polarity operators, after standard

lambda abstraction over the variable has applied (see Caponigro and Polinsky to appear for the details

of the semantic analysis and independent evidence in favor of polarity operators and variables over

them). The intuition that we want to capture by means of polar operators is that embedded declaratives

and embedded polar interrogatives share a basic feature at the level of their semantic contribution:

their denotations are built on the same proposition. For instance, the denotations of the embedded

declarative (that) Mary left and of the embedded polar interrogative if/whether Mary left both depends

on the proposition 'that Mary left'. This proposition is either the actual denotation of the clause, as in

the case of the embedded declarative, or the proposition that together with its negation occurs as a

member of the set-denotation, as in the case of the embedded polar interrogative.

A possible additional argument for the relative clause analysis of ze-re- clauses comes from the

observation that these clauses never stack with other MCs. In more familiar languages, a fact-clause

and a relative clause can stack, as shown for English in (31). In Adyghe, no such stacking is possible

in either order (32), which suggests that these clauses are all of the same nature (recall that relative

clauses proper do not stack).

(31) a. The fact that gas consumption is diminishing that/which CNN ignored

b. % The fact that/which CNN ignored that gas consumption is diminishing

(32) a. *

[ daKWeme zere-I&WejERWe mWerat qEsfjE{Weta-Re] qebarer

marrying zere-desire_have.

PRES M deliver.me-PAST news.ABS

* [ mWErat qEsfjE{Weta-Re daKWeme zere-I&WejERWe] qebarer

M deliver.me-

PAST marrying zere-desire_have.PRES news.ABS

(‘the

news/rumor that she wants to get married which Murat told me’)

4.3. zere- as a marker of subordination?

All five clauses considered here are syntactically the same: definite DPs that include a CP with an

operator in it, binding an empty category in the TP. However, one could consider them all definite

DPs, with the marker zere- simply indicating subordination or complementation. Such an analysis

would capture the syntactic distribution and island effects and would also capitalize on the well-

attested parallels between relative and non-relative complementizers (English that, Romance que,

Russian čto, etc.). Furthermore, one could draw a parallel between zere- and the English how in a

sentence like I remember how my Great Uncle Jerry would sit on the porch all day long.

However, such an analysis would create significant difficulties in accounting for the polar

interrogative use of MCs. As far as we know, no languages that have homophony between relative and

complement clause heads, also include polar interrogatives in that list. Adyghe MCs do not

presuppose their content, while English how-complements do. Next, Adyghe MCs stand out in that the

operator-variable relationship in them is marked morphologically (by wh-agreement and, if other

conditions are met, an applicative marker).

5. Conclusions

We showed that all Adyghe mystery clauses have the same basic syntactic structure: they are all

complex DPs containing relative clauses, thus instantiations of a syntactic configuration in which an

operator binds an empty category. Given the empirical facts of Adyghe, we do not see any compelling

reasons to posit structural ambiguity for the mystery clauses considered here. Our conclusions are

similar to the ones reached by Gerasimov and Lander (2008) who analyze the embedded declarative

interrogatives (see Caponigro and Polinsky to appear for details); as for the nature of the high -re-, it can co-occur

with the instrumental -re- (which is lower), and that casts doubt on their synchronic identity.

MC as a relative clause (they do not consider other MC types however).

If our proposal is on the right track, Adyghe is simpler than the more familiar languages, where

both relativization and complementation are required. Thus “sentential complements” are not an

indispensable part of grammar; they can be fully represented by relative clauses within complex DPs.

What follows from this simplicity is that the more familiar complementation strategies, ones that we

normally take for granted, may not constitute a structural necessity in natural language.

The Adyghe system offers further empirical support to the idea that relative and non-relative

complementation are qualitatively different (Rizzi 1990; Lasnik and Saito 1992), although the

conceptual motivation for this difference is not yet understood.

There is another challenge left. As we saw, mystery clauses receive five different interpretations,

according to the verbal morphology and/or environment they occur in—so how can the same syntactic

structure can be mapped into five different meanings? We address this puzzle in Caponigro and

Polinsky (to appear) and show that this mapping can be achieved by means of semantic tools that are

independently available.

References

Bianchi, V. 2002. Headed relative clauses in generative syntax. Glot International 6.7, 6.8.

Caponigro, I. 2004. The semantic contribution of wh-words and type shifts: Evidence from free relatives cross-

linguistically. In R. Young (ed.) Proceedings of SALT XIV, 38-55, Ithaca, NY: CLC Publications, Cornell

University.

Caponigro, I., and M. Polinsky. To appear. Almost everything is relative in the Caucasus. Proceedings of SALT

XVIII.

Colarusso, J. 1992. A grammar of the Kabardian language. Calgary: University of Calgary Press.

Gerasimov, D., and Y. Lander. 2008. Reljativizacija pod maskoj nominalizacii i faktivnyj argument v adygejskom

jazyke. In V. Plungyan and S. Tatevosov (eds) Issledovanija po teorii grammatiki, 4: Issledovanija po

glagol’noj derivacii, 290-313. Moscow: Jazyki russkoj kul’tury.

Hamblin, C.L. 1973. Questions in Montague English. Foundations of Language 10: 41-53.

Hewitt, B. G. 1979. The relative clause in Abkhaz (AbZui dialect). Lingua 47: 151-187.

Hudson, R. 1990. English word grammar. Oxford: Blackwell.

Jacobson, P. 1995. On the quantificational force of English free relatives. In. E. Bach, E. Jelinek, A. Kratzer, B.

Partee (eds), Quantification in natural language, 451-486. Dorderecht: Kluwer.

Jakovlev, N.F., and D. Ašxamaf. 1941. Grammatika adygejskogo literaturnogo jazyka. Moscow: Izd. AN SSSR.

Karttunen, L. 1977. Syntax and semantics of questions. Linguistics and Philosophy 1: 3-44.

Kayne, R.1994. The antisymmetry of syntax. Cambridge, MA: MIT Press.

Keraševa, Z. 1984. Predloženija s finitnymi i infinitnymi glagolami v adygskix jazykax. Tbilisi: Mecniereba.

Kiparsky, P., and C. Kiparsky 1970. Fact. In M. Bierwisch and K.E. Heidolph (eds), Progress in linguistics, 143-

73, The Hague: Mouton.

Lander, Y. 2004. A relative intrigue in Adyghe. Paper presented at the First Conference on the Syntax of World’s

Languages, Leipzig, August 2004.

Larson, R., and N. Takahashi. 2005. Order and interpretation in prenominal relative clauses. Paper presented at

the Second Workshop on Altaic Formal Linguistics, WAFL-2.

Lasnik, H., and M. Saito. 1992. Move α. Cambridge, MA: MIT Press.

Montague, R. 1973. The proper treatment of quantification in English. In K.J.J Hintikka, J.M.E. Moravcsik, and

P. Suppes (eds), Approaches to natural language, 221-242. Dordrecht: Reidel.

O’Herin, B. 2001. Abaza applicatives. Language 77: 477-493.

O’Herin, B. 2002. Case and agreement in Abaza. Dallas, TX: SIL International.

Polinsky, M., and E. Potsdam. 2006. Expanding the scope of control and raising. Syntax 9: 171-192.

Quine, W. V.O. 1960. Word and object. Cambridge, MA: MIT Press.

Rizzi, L. 1990. Relativized minimality. Cambridge, MA.: MIT Press.

Rogava, G., and Z. Keraševa. 1966. Grammatika adygejskogo jazyka. Majkop: Krasnodarskoe knižnoe

izdatel’stvo.

Sag, I. 1997. English relative clause constructions. Journal of Linguistics

33: 431-484.

Smeets, R. 1984. Studies in West Circassian phonology and morphology. Leiden: Leiden University Press.

Sumbatova, N. 2005. Information structure and clause combining in Adyghe. Paper presented at INALCO,

December 6.

Suzuki, M. 2005. Clause stacking and antisymmetry in Japanese relative clauses. Ph. D. Diss., CUNY Graduate

Center.

Proceedings of the 27th West Coast

Conference on Formal Linguistics

edited by Natasha Abner

and Jason Bishop

Cascadilla Proceedings Project Somerville, MA 2008

Proceedings of the 27th West Coast Conference on Formal Linguistics

ISBN 978-1-57473-428-7 library binding

A copyright notice for each paper is located at the bottom of the first page of the paper.

Reprints for course packs can be authorized by Cascadilla Proceedings Project.

Ordering information

Orders for the library binding edition are handled by Cascadilla Press.

To place an order, go to www.lingref.com or contact:

Cascadilla Press, P.O. Box 440355, Somerville, MA 02144, USA

phone: 1-617-776-2370, fax: 1-617-776-2271, e-mail: [email protected]

Web access and citation information

This entire proceedings can also be viewed on the web at www.lingref.com. Each paper has a unique document #

which can be added to citations to facilitate access. The document # should not replace the full citation.

This paper can be cited as:

Caponigro, Ivano and Maria Polinsky. 2008. Relatively Speaking (in Circassian). In Proceedings of the 27th

West Coast Conference on Formal Linguistics, ed. Natasha Abner and Jason Bishop, 81-89. Somerville, MA:

Cascadilla Proceedings Project.

or:

Caponigro, Ivano and Maria Polinsky. 2008. Relatively Speaking (in Circassian). In Proceedings of the 27th

West Coast Conference on Formal Linguistics, ed. Natasha Abner and Jason Bishop, 81-89. Somerville, MA:

Cascadilla Proceedings Project. www.lingref.com, document #1819.