As usual, I am indebted to Pe Mamy for his help in corpus annotation and for checking the grammaticality of modified corpus examples. This manuscript, whose initial version was presented at the SLE-2017 in Zürich, has benefitted from comments from and conversations with Balthasar Bickel, Oleg Belyaev, and Michael Yoshitaka Erlewine (mitcho). I am also thankful to both anonymous reviewers for their excellent comments and suggestions and to Albion M. Butters for proofreading. Finally, I gratefully acknowledge support for the research that this manuscript is based on, both from the Helsinki Collegium for Advanced Studies and from the Kone Foundation.
1It has long been argued that subordination and non-subordination (including parataxis and coordination) do not form a dichotomy but rather a scale (see Haiman & Thomason 1984 for a pioneering statement). Indeed, for example, the English “left-subordinating and” construction, such as (1), has a conditional semantic interpretation and mixed syntactic properties that place it “in between” subordination and coordination (Culicover & Jackendoff 1997; Belyaev 2015).
2(1) You drink another can of beer and I’m leaving. (Belyaev 2015: 268)
3The aim of this paper is to show that relativization strategies in Mano, including the most prominent one, a correlative strategy, are similarly located in between subordination and parataxis. This is expected, given that cross-linguistically, correlative clauses are often adjoined high, and diachronically they can be considered a prime example of an erstwhile paratactic construction acquiring subordinating properties (Givón 2009). Yet, the extent of subordinating properties of correlatives seems to vary from language to language and should be subject to empirical scrutiny, which is the main aim of the present paper. I argue that in Mano, while the overall construction functions as a complex clause and may under certain conditions receive a relativization interpretation, it is syntactically non-subordinating under all syntactic criteria applied.
4The correlative strategy is a subtype of non-reduction relativization strategy where “the head noun appears as a full-fledged noun phrase in the relative clause and is taken up again at least by a pronoun or other pronominal element in the main clause” (Comrie 1998: 62). The example (2) from Hindi illustrates the correlative strategy. Here, the relativized NP (which I will subsequently call Rel-NP) is CD introduced by the relativizer jo in the correlative clause (CorC). The referent is taken up again in the matrix clause (MatC) by a correlate NP us CD-ko (which I will subsequently call Mat-NP), where CD is in the accusative case and accompanied by the demonstrative us. The indexes here and in what follows indicate co-reference.
|
(2)
|
[CorC
|
[jo
|
CD]i
|
sale-par
|
hai]
|
[MatC
|
Maya
|
[us
|
CD-ko]i
|
|
|
|
rel
|
CD
|
sale-on
|
be.prs
|
|
Maya.f
|
dem
|
CD-acc
|
- 1 Here and in what follows, I render long vowels by double vowel characters, while Bhatt renders them (...)
5‘Maya will buy the CD that is on sale. (Lit. ‘[Which CD is on sale], Maya will buy that CD.’) (Bhatt 2003: 486)1
- 2 Note that many definitions of relativization presuppose that the relative clause must be subordinat (...)
6Parataxis is another subtype of non-reduction relativization strategy where “the ‘relative’ clause contains the full-fledged head and is at the same time an un- marked simple (declarative) clause; the relative and main clauses are only loosely joined together” (Comrie & Kuteva 2006: 212). In the example (3), the Rel-NP in the clause that functions like a relative clause is that man, and the Mat-NP is a pronoun he. In the following discussion, I adopt Comrie and Kuteva’s convention by putting “relative” in inverted commas whenever the discussion concerns a paratactic strategy of relativization.2
7(3) That mani just passed by us, hei introduced me to the Chancellor of the University yesterday. (Comrie & Kuteva 2006: 212)
8Following these definitions, then, the difference between the correlative and the paratactic type is in the degree of syntactic embeddedness: paratactic relative clauses are non-embedded and correlative clauses are embedded at least to some extent.
- 3 The motivation for the gloss is the following. The same marker lɛ́ (and its variants) are used in f (...)
9Let us now turn to Mano (ex. 4). Similarly to the Hindi example above, the Mano example features a Rel-NP ŋwɔ́ ‘problem’ accompanied by a relativizer lɛ́ (glossed here and below as att).3 The referent is the subject of the matrix clause and is taken up again by a 3sg pronoun à. Thus, at least on the surface, the relativization strategy in Mano appears to be more like the correlative strategy in Hindi (2) than the paratactic strategy in English (3).
|
(4)
|
[CorC
|
[Ŋwɔ́i
|
lɛ́
|
ā
|
gèē
|
wɛ̄]
|
|
|
|
problem
|
att
|
3sg.pst>3sg
|
say
|
dem
|
|
[MatC
|
ŋ́ŋ̀
|
lō
|
ài
|
kɛ̄-ɛ̀.]
|
|
|
1sg.ipfv
|
go:ipfv
|
3sg
|
do-ger
|
10‘The thingi that he said, I will do iti.’ [MOC]
- 4 The examples straight from the corpus are marked as [MOC]. Some examples have been modified for sim (...)
11Despite the first appearances, using data primarily from an oral corpus of Mano texts,4 I will argue that in Mano, the relation between the matrix clause and the “relative” clause in what appears to be a correlative construction (4) is not of syntactic subordination and, thus, more like the English (3) and (1) than the Hindi (2).
12Cross-linguistically, correlative clauses often—if not always, see Lipták (2009)—behave as syntactic adjuncts either to the argument, in the case of clause-internal correlatives, or to the clause, in the case of utterance-peripheral correlatives. See Nikitina (2012) and Creissels (2009) for Mande, Lipták (2012) for Hungarian, Bhatt (2003), Srivastav (1991) and Dayal (1996) for Hindi, Belyaev and Haug (2014) for Ossetic and Motter (2023) for Hittite. I provide further syntactic and semantic arguments to support that position.
13Moreover, I make a stronger claim that even if one adopts a transformational approach, Mano correlative clauses do not appear in the adjunct-to-matrix-clause position as a result of movement. To prove that, I apply to Mano syntactic arguments developed by Bhatt (2003) for Hindi (Section 3). Unlike Hindi, where these arguments speak in favor of treating (single-headed) correlative clauses as generated clause-internally, in Mano they give an opposite result and speak in favor of considering correlative clauses as generated in the adjunct position to the matrix clause. Essentially, while Bhatt argues that the syntactic representation of (2) is best illustrated as Figure 1a and not as Figure 1b, in Mano the same criteria suggest that the interpretation in Figure 1b is more suitable.
14In Section 4, using arguments by Belyaev and Haug developed for Ossetic (2014), I show that because the Rel-NP and the Mat-NP do not always co-refer, the relationship between the relativized NP and its counterpart in the matrix clause is instead one of pragmatic resolution of anaphora, which further speaks in favor of non-embeddedness. In the discussion in section 5, I further explain that the correlative and matrix clauses are linked through parataxis, rather than adjunction, partly following Motter (2023), who also used Belyaev and Haug’s diagnostics and arrived at a similar conclusion, namely, that Hittite correlatives are paratactic. The paratactic structure of (4) shared with (1) and (3) is illustrated by Figure 2. Whether or not the higher CP node is needed is discussed in Section 5.
15First, however, I introduce some basic facts about Mano grammar, including subordination and relativization strategies.
16Mano is a South Mande language spoken in Guinea and Liberia by about 400,000 speakers. It has a rigid S Aux O V X word order. Aux stands for auxiliary expressing tense, aspect, modality, and polarity (TAMP) categories and X stands for postverbal arguments, typically expressed as postpositional phrases, and adjuncts. The auxiliary also indexes the subject’s person and number and are thus organized into series with the same person and number distinctions as pronouns. Most finite predications are expressed by auxiliary constructions (on constructions with predicators, see Khachaturyan 2023), where the subject NP is optional.
17Ex. (5) provides a single utterance, produced within the same breath group and with a continuation intonation throughout (on continuation intonation, see below), with two coordinated clauses and no explicit coordination marker beyond prosody. The first clause has an overt subject NP, sìī ‘spider,’ which is also indexed on the 3sg past auxiliary, ā; in the second, the subject is only expressed through indexation on the negative auxiliary, lɛ̀ɛ́.
|
(5)
|
Sìī
|
ā
|
ŋwɔ̀
|
tā̰ā̰,
|
lɛ̀ɛ́
|
yī
|
zɛ̄.
|
|
|
spider
|
3sg.pst>bridg
|
thing:cstr
|
exaggerate
|
3sg.neg
|
slumber
|
kill
|
18‘[As the moment was getting nearer,] the Spider made such a big deal out of it that it did not sleep anymore.’ [MOC]
19Portmanteau auxiliaries also incorporate the 3sg pronoun à or the bridging marker à (on the latter, see Section 4.4). In (5) above, the past 3sg auxiliary belongs to the portmanteau type. The corresponding non-portmanteau form is ē ‘3sg.pst.’
20Mano distinguishes between alienable possessor constructions, where the possessor is expressed by a dedicated series of possessive pronouns and, in 3rd person, an optional NP, and an inalienable possessor construction, where the pronominal possessor is expressed by a pronoun from the basic series in complementary distribution with a full-fledged NP. The distinction becomes relevant in the discussion of example (38) below.
21Finally, Mano has rich nominal and verbal tonal morphology. In particular, it has a grammatical high tone optionally used with nouns that are followed by a demonstrative and/or the attention marker lɛ́ (6). The high-tone form is also used in correlative clauses, as I show below.
|
(6)
|
(a)
|
mī
|
wɛ̄
|
|
|
|
|
person
|
dem
|
|
|
|
(b)
|
mí
|
wɛ̄
|
|
|
|
|
person:h
|
dem
|
|
|
|
(c)
|
mī
|
lɛ́
|
wɛ̄
|
|
|
|
person
|
att
|
dem
|
|
|
(d)
|
mí
|
lɛ́
|
wɛ̄
|
|
|
|
person:h
|
att
|
dem
|
22‘this person’
23Mano makes use of various morphosyntactic and prosodic means for clause-combining and subordination, including conjunctions, such as yé ‘when’ in (7) or ɓīī ‘because’.
|
(7)
|
Yé
|
à
|
ɲɛ́nɛ́
|
ē
|
kɛ̄
|
gá̰á̰-pɛ̀lɛ̀
|
ā…
|
|
|
when
|
bridg
|
sun
|
3sg.pst
|
do
|
drag-inf
|
dem
|
24‘As the moment was getting nearer [the Spider made such a big deal out of it that it did not sleep anymore.]’ [MOC]
25To mark some dependent clauses, dedicated auxiliary series and verbal forms are used. For example, for forms used in conditional clauses, see Khachaturyan (2020b). The joint auxiliary series is also widely used to express sequences of events or, as in (8), events occurring simultaneously (typically concerning coreferential subjects, but not necessarily). Conjoined clauses are dependent on the preceding context for their temporal interpretation and therefore cannot be considered independent clauses.
|
(8)
|
Ē
|
tó
|
yī
|
áà
|
gèè
|
kélɛ̀…
|
|
|
3sg.pst
|
stay
|
there
|
3sg.jnt>3sg
|
say:jnt
|
that
|
26‘He remained there saying…’ [MOC]
- 5 A reviewer has pointed out that in the function of marking fronted dependent clauses, the demonstra (...)
27The most common means employed in clause-combining are the clause-final demonstratives ā (a variant of yā, which can also assimilate to the preceding vowel resulting in further variants) and ɓɛ̄~wɛ̄~wāā. They are quasi-obligatory in all dependent clauses preceding their main clause. That is, speakers always prefer to have them in elicited utterances, but in natural speech they are not always present and the rules determining the likelihood of omission are not very clear yet. Crucially, in some cases these demonstratives become the only morphosyntactic clause-linking markers. Unlike (7), where the temporal clause includes a conjunction yé ‘when,’ in (9) the clause ‘(when) I drank wine’ does not have any clause-linking marker, only the demonstrative ā.5
|
(9)
|
[Ŋ̄
|
yɔ̄
|
mī
|
ā],
|
kō
|
gèlè
|
gɔ̰̄.
|
|
|
1sg.pst
|
wine
|
drink
|
dem
|
1pl.pst
|
war
|
fight
|
28‘(When) I drank wine, we fought.’ [MOC, modified]
29Finally, Mano distinguishes between prosodic contours marking the end of the utterance vs. those projecting the continuation. In the utterance-final position, the high tone is systematically realized as mid, the low as falling and the mid as mid or mid-falling. Thus, continuation is marked by the absence of final lowering, but also, in the case of some speakers, by a high-rising tone. Since not all instances of clause-combining are marked by dedicated means, prosody allows for distinguishing between several clauses linked to a single utterance (with a paratactic relation, as in 11, or a subordinating one) and clauses produced as separate utterances. See, for example, the contrast between the realization of the final tone of zèí: high in (11), illustrated in Fig. 4 (a sharp rise), and mid in (10), illustrated in Fig. 3 (only a slight rise).
|
(10)
|
Ō
|
nɛ́
|
yà-pɛ̀lɛ̀
|
léé
|
wɛ̄
|
zèí.
|
|
|
|
3pl.exi
|
child
|
sit-inf
|
woman:h
|
dem
|
middle
|
|
30‘They are putting a child on the woman’s back.’ [MOC]
|
(11)
|
[Léé
|
wɛ̄
|
āà
|
nɛ́
|
yà
|
ē
|
zèí]
|
|
|
woman:h
|
dem
|
3sg.prf
|
child
|
sit
|
3sg.refl
|
middle
|
|
[āà
|
sɔ̄
|
yèlè
|
à
|
mɔ̀.]
|
|
3sg.prf
|
cloth
|
attach
|
3sg
|
on
|
31‘This woman has put a child on her back, she is attaching a cloth to it.’ [MOC]
32Mano has three types of relativization strategies: 1) correlative clauses with lɛ́ when the Rel-NP is 1a) utterance-peripheral and 1b) in-situ, 2) detached “relative” clauses with lɛ́ and 3) paratactic clauses. The main focus of this paper is on utterance-peripheral correlative clauses.
331) Correlative clauses are formed with the postnominal relativizer lɛ́, which also has a variant nɛ́ (in nasal contexts) and tɛ́ and can also take the form of the floating high tone. Following the definition provided by Comrie and Kuteva (2006), in correlative relativization the in-situ position of the correlate in the matrix clause (Mat-NP) is filled by a resumptive element. In Mano correlative clauses, Mat-NP is typically expressed by a pronoun (12) or, in the case of relativization of subjects, as agreement on the auxiliary (13). A distinct feature of the correlative clauses is left-dislocation of the Rel-NP and the filling of its in-situ position by a resumptive element. Indeed, in (12) the in-situ position of the Rel-NP is the subject and the resumptive element is indicated by the 3sg.sbjv auxiliary è. In (13), the in-situ position is the argument of a gerund and the resumptive element is the 3sg pronoun à.
|
(12)
|
[CorC
|
Míi
|
è
|
nū
|
zèē
|
bḭ́míà
|
ā]
|
|
|
|
|
person>att
|
3sg.sbjv
|
come:ipfv
|
here
|
night
|
dem
|
|
|
[MatC
|
ŋ́ŋ̀
|
ài
|
zɛ̄.]
|
|
|
1sg.ipfv
|
3sg
|
kill:ipfv
|
34‘Whoever comes here at night, I will kill him/her / I will kill the person who will come here at night.’ [el.]
|
(13)
|
[CorC
|
Míi
|
lɛ́
|
ài
|
gā-à
|
lɛ̄
|
ɛ̄]
|
|
|
|
person:h
|
att
|
3sg
|
die-ger
|
ident
|
dem
|
|
[MatC
|
káà
|
gèē
|
āài
|
wɛ̀lɛ̀
|
gā
|
kɛ̀lɛ̀.]
|
|
|
2pl.ipfv>3sg
|
say:ipfv
|
3sg.prf
|
get.up
|
death
|
hand
|
35‘The person whom you say has died has risen from the dead.’ [MOC, modified]
- 6 Most nouns have the exact same high-tone form that is used with demonstratives and that results fro (...)
36If the head noun is final in the Rel-NP, the head noun can optionally appear as a high-tone form when accompanied by the relativizer (13; see also 6); the high-tone form is obligatory when the relativizer is expressed by a floating high tone (12).6 Crucially, a high-tone form can cooccur with a full-fledged relativizer (as in ex. 13), which means that a high-tone nominal form is not always just a result of association of the floating high tone of the relativizer. Note also that in (12), the Rel-NP can have an indefinite (generic) reading without any special marking. Here, the high tone form of the noun clearly results from the association with the relativizer and is irreducible to other functions, such as definiteness marking.
37The final formal property of correlative clauses is the use of clause-final demonstratives. Note, however, that, as in other dependent clauses, the demonstrative is not obligatory.
38In the examples above, the CorC(s) were all situated to the left of the matrix clause. This is the most frequent relativization strategy in my corpus. An alternative position is clause-internal, to the left of the in-situ relativized constituent—as in (14), the argument of a postposition, or in (15), the direct object.
|
(14)
|
[MatC
|
Ā
|
yí-gīnī
|
[CorC
|
gwɛ́ɛ́kòlò
|
vɔ̀i
|
tɛ́
|
|
|
|
3sg.pst>3sg
|
interior-share
|
|
old.man
|
pl
|
att
|
|
ō
|
gbēŋ́
|
yí
|
ā]
|
ōi
|
mɔ̀.]
|
|
3pl.exi
|
block
|
in
|
dem
|
3pl
|
on
|
39‘He shared it [the animal] among all the elders of his neighborhood.’ (Lit.: ‘He shared it, elders who are in the block, among them.’) [MOC]
|
(15)
|
[MatC
|
Zɛ́ɛ́zú
|
lɛ̀ɛ́
|
[CorC
|
kpàléŋwɔ̀i
|
tɛ́
|
ē
|
kɛ̄
|
|
|
|
P.N.
|
3sg.neg
|
|
miracle
|
att
|
3sg.pst
|
do
|
|
à
|
kɛ̄-pɛ̀lɛ̀
|
Cafarnom
|
ɓɛ̄]
|
ài
|
kɛ̄
|
ē
|
dìè
|
pàà]
|
|
3sg
|
do-inf
|
P.N.
|
dem
|
3sg
|
do
|
3sg.refl
|
int
|
at
|
40‘Jesus did not perform in his own country the miracles that he was performing in Capharnaum.’ [MOC, modified]
41Except for the syntactic position, clause-internal correlatives share all other properties with the utterance-peripheral correlatives: left-dislocation of the Rel-NP, the preferential use of the clause-final demonstrative and no restrictions on the syntactic position of the Mat-NP.
- 7 Some Mande languages, like Wan, have a restriction on the syntactic position of the Mat-NP for clau (...)
42Utterance-peripheral CorCs in Mano, like in other Mande languages, should be considered clause-level adjuncts while clause-internal CorCs should be considered adjuncts at the level of the Mat-NP (for arguments, see Nikitina 2012).7
432) Another type of relativization strategy is detached “relative” clauses with a conjunction lɛ́. In contrast to the correlatives, they represent regular finite clauses: there is no left-dislocation of the Rel-NP, and there is no clause-final demonstrative. All syntactic positions are assessable to this type of relativization; see the relativization of the direct object in (16) and of the subject in (17) and (18).
|
(16)
|
[MatC
|
Ŋ̄
|
Gèwúlú
|
gɛ̰̀]
|
[RelC
|
lɛ́
|
ài
|
lèē
|
wà
|
|
|
|
|
1sg.pst
|
P.N.
|
see
|
|
and.so
|
3sg
|
mother
|
3pl.ip
|
|
|
ài
|
dàā
|
ō
|
gā
|
gèlè
|
yí.]
|
|
3sg
|
father
|
3pl.pst
|
die
|
war
|
in
|
44‘I saw Gewulu, whose mother and father died in war.’ [el.]
45Detached “relative” clauses can be situated to the right of the matrix clause, as in (16). They can also appear in a sequence of relative clauses when there are several relative clauses for one and the same Rel-NP (17).
|
(17)
|
[CorC1
|
mīāi
|
lɛ́
|
wāà
|
gā
|
[RelC2
|
lɛ́
|
ō
|
mīā
|
|
|
|
|
person:pl
|
att
|
3pl.prf
|
die
|
|
and.so
|
3pl.exi
|
person:pl
|
|
|
fḭ́
|
dɔ̄
|
pɛ̀lɛ̀]
|
ɛ̄]
|
[MatC
|
í
|
ōi
|
mɔ́ɔ̀wèè
|
tó.]
|
|
rest
|
install
|
place:cstr
|
dem
|
|
2sg.conj
|
3pl
|
pardon
|
leave
|
46‘People who have died and who are in the place of rest, forgive them.’) [MOC, modified]
47If the relativizer is expressed by a floating high tone, a sequence of relative clauses can be formed without the second relativizer at the beginning of the subsequent “relative” clause. Thus, in (18) the marker lɛ́ is optional in the beginning of the second relative clause. More specifically, the actual example from the corpus did not have lɛ́, but the consultant confirmed that it was possible to use it.
|
(18)
|
[CorC1
|
Mí
|
lɛ̀ɛ́
|
nɛ́
|
yē]
|
|
|
|
person>att
|
3sg.neg
|
child
|
give.birth
|
|
[RelC2
|
(lɛ́)
|
lɛ̀ɛ́
|
lēē
|
sí
|
āā]
|
|
|
so.that
|
3sg.neg
|
woman
|
take
|
dem
|
|
[MatC
|
|
mīī
|
kpànāzɛ̀
|
wáá
|
ká.]
|
|
|
|
person
|
important
|
cop.neg>3sg
|
with
|
48‘A person who does not give birth to children, who does not take a wife, he is not important.’ [MOC, modified]
- 8 Another difference between detached “relatives” and correlatives is that the Rel-NP does not appear (...)
49Detached “relative” clauses never appear in the utterance-peripheral position; in that position, only a clause with a left-dislocated Rel-NP is possible.8
- 9 Because in the clause-linking function lɛ́ does not accompany a NP, unlike lɛ́ in the relativizer f (...)
50In detached “relative” clauses, lɛ́ is not used as a relativizer but as a clause-linker with a broad function. Indeed, lɛ́ is amply attested at the beginning of utterances in a narrative and can be roughly translated as a discourse linker ‘and then, and so’ (19). In this function, lɛ́ does not have a nasalized variant and floating high-tone variant, but it has an additional variant yēlɛ́ (which results from adjunction of yē 3sg.emph).9
|
(19)
|
Lɛ́
|
gɔ́
|
à
|
gèē
|
ē
|
nā
|
lɛ̀ɛ̄…
|
|
|
and.so
|
leopard
|
3sg.sbjv>3sg
|
say:ipfv
|
3sg.refl
|
wife
|
to
|
51‘[One day, the leopard’s wife told the leopard: “I want to eat meat, some raw meat.”] And so the leopard says to his wife: …’ [MOC, modified]
52The difference between the utterance-initial lɛ́ and lɛ́ in what appears as a relative clause is subtle: in (18), the clause beginning with lɛ́ is integrated into a single utterance with the preceding clause; the integration is indicated solely by prosody (no pause after or utterance-final intonation on the preceding clause) and the coreference relation between one of the arguments of the first clause and the 3sg pronoun à in the second clause. In other words, detached “relatives” with lɛ́ can be considered a paratactic strategy.
533) The final type of relativization strategy comprises paratactic clauses with no additional marking of relativization, not even discourse linkers. Example (20) illustrates this type. It consists of two clauses linked by continuation intonation and by referential contiguity between them: the direct object of the first clause, mí wɛ̄ ‘this person,’ co-refers to the subject of the second clause. Furthermore, the second clause has a construction with an auxiliary from the conjoint series, which, as example (8) above illustrated, is used for clause-combining and, in particular, expressions of event simultaneity. Such clause sequences can in some cases also be interpreted as a relativization strategy, as in (20), which was translated by a native speaker as a relative clause (‘Vous voyez celui qui marche à pieds?’; ‘Do you[pl] see the person who is walking?’).
|
(20)
|
[Káà
|
mí
|
wɛ̄i
|
gɛ̰̀
|
nɔ́]
|
|
|
2pl.ipfv
|
person:h
|
dem
|
see.ipfv
|
just
|
|
[áài
|
tàà
|
ē
|
gà̰
|
làà?]
|
|
3sg.jnt
|
walk:jnt
|
3sg.refl
|
leg
|
on>q
|
54‘Do you[pl] see the person who is walking?’ (Lit.: ‘Do you[pl] see this guy, he is walking?’) [MOC]
55Indo-Aryan languages are a language group with widespread correlatives, including clause-internal, NP-adjoined and utterance-peripheral IP-adjoined correlative clauses (in other words, adjoined to the entire clause). Example (21) illustrates the phenomenon in Hindi. Unlike Mano, where Mat-NP is usually expressed by a pronoun, in Indo-Aryan languages the Mat-NP is expressed by a demonstrative (woo, according to Davison 2009; ex. 21; vo, according to Bhatt 2003; ex. 23), which can be used pronominally and adnominally (for example, accompanying the noun kitaab in (21)). The relativizer (joo, according to Davison 2009; jo, according to Bhatt 2003) can, in and of itself, fill an argument position within the CorC (22) but usually needs an accompanying element, such as a noun (21).
56(21) Hindi (from Davison 2009: 246–247)
57a) Utterance-peripheral correlatives: [CorC] [MatC]
|
[CorC
|
Joo
|
kitaabi
|
shiilaa-nee
|
likh-ii]
|
|
|
rel
|
book
|
Shila-erg
|
write-prf.f
|
|
[MatC
|
wooi
|
(kitaabi)
|
rahul
|
paṛh
|
rahaa
|
hai.]
|
|
|
that
|
book
|
Rahul
|
read
|
prog
|
is
|
58‘Rahul is reading the book which Shila wrote.’ (Lit.: ‘Which book Shila wrote, that book Rahul is reading.’)
59b) Clause-internal correlatives: [MatC [CorC]]
|
[MatC
|
Rahul
|
[RelC
|
joo
|
kitaabi
|
shiilaa-nee
|
likh-ii]
|
wooi
|
(kitaabi)
|
|
|
Rahul
|
|
rel
|
book
|
Shila-erg
|
write-prf.f
|
that
|
book
|
|
paṛh
|
rahaa
|
hai.]
|
|
|
read
|
prog
|
is
|
|
60‘Rahul is reading the book which Shila wrote.’ (Lit.: ‘Rahul, which book Shila wrote, that book is reading.’)
- 10 By contrast, multiple-headed correlatives are generated in the adjunct position, but I will not go (...)
61These two types of correlative clauses display a surface similarity to Mano correlatives. Yet, I will show that a crucial difference between Hindi and Mano may be discerned. Indeed, Bhatt (2003) entertains two further hypotheses in the transformational approach to syntax: either 1) the (single-headed) correlative clauses are base-generated clause-internally and are optionally fronted to the position of adjunct of the matrix clause, which is the hypothesis he ends up arguing for,10 or 2) all correlative clauses are generated in surface adjunct position, which is the analysis advocated by Srivastav (1991). These possibilities are illustrated with example (2), repeated below as (22a). The movement analysis is illustrated by Figure 1a, repeated below as (22b), and the base-generation analysis is illustrated by Figure 1b, repeated below as (22c).
|
(22a)
|
[CorC
|
[jo
|
CD]i
|
sale-par
|
hai]
|
[MatC
|
Maya
|
|
|
|
rel
|
CD
|
sale-on
|
be.prs
|
|
Maya.f
|
|
[us
|
CD-ko]i
|
khariid-egii.]
|
|
dem
|
CD-acc
|
buy-fut.f
|
62‘Maya will buy the CD that is on sale.’ (Lit. ‘[Which CD is on sale], Maya will buy that CD.’) (Bhatt 2003: 486)
63Bhatt’s consideration of these two options reflects the idea that a surface syntactic construction can be a result of an underlying transformation in which some constituents are generated in a particular syntactic position but “moved” from there to a different syntactic position. When they move, they leave a trace behind, which is represented as t.
64Whether movement has or has not taken place can be discerned, for example, by observing whether all the syntactic slots in a particular predication have been filled. Thus, transitive verbs require direct objects, so the utterance-initial position of what in What do you think John bought? is described as a result of a specific kind of movement, called wh-movement, and there is a trace t in the direct object position of bought: What do you think John bought t?. Another diagnostic is the interpretation of particular elements. For example, where in Where do you think you are going? is interpreted as an argument of the second predication (you are going where) and therefore must have moved from there: Where do you think you are going t? In more complicated cases, where the syntactic and semantic properties of verbs and clauses themselves do not provide any evidence for the syntactic structure, movement is discerned by observing ungrammatical structures whose ungrammaticality is best explained by the observation that the trace appears in a universally precluded syntactic and semantic configuration (an example is the Condition-C effect), either by cross-linguistic tendencies on movement restrictions (such as the Island effects) or by language-internal movement affordances (such as capacity of stacking of moved constituents).
- 11 Note that the logic behind each of these three diagnostics differs in fundamental ways. It is not c (...)
65These are the diagnostics that Bhatt applied to Hindi correlatives and concluded from them that Hindi correlatives are generated clause-internally.11 I review these diagnostics and apply them to Mano in the following order:
66Island effects;
67Possibility to stack two utterance-peripheral correlative clauses; and
68Condition-C effects.
69The CorC and the MatC are not necessarily adjacent: in (13) above, they are separated by a predication introducing reported speech (‘The person who you say has died has risen from the dead.’). This applies to Hindi as well, but with a restriction: in Hindi, the Mat-NP cannot be found within another relative clause (23).
70(23) Island effects in Hindi (Bhatt 2003: 500)
71*[CorC Rel-NPi] [MatC … [RelC … Mat-NPi] …]
|
[CorC
|
Joi
|
vahã:
|
rah-ta
|
hai]
|
[MatC
|
mujh-ko
|
[vo
|
kahaanii]j
|
|
|
rel
|
there
|
stay-hab
|
be.prs
|
|
I-DAT
|
that
|
story.f
|
|
[RelC
|
joj
|
Arundhati-ne
|
usi-ke-baare-mẽ
|
likh-ii]
|
pasand
|
hai.]
|
|
|
|
rel
|
P.N.-erg
|
dem-about
|
write-pfv.f
|
like
|
be.prs
|
|
72*Intended meaning: ‘I like the story that Arundhati wrote about the person who lives here.’ (Lit.: ‘Who lives there, I like the story that Arundhati wrote about that (person).’).
- 12 Some solutions that come to mind include moving the second Mat-NP expressed by the demonstrative ph (...)
- 13 It is problematic to claim pre-empirically that all relative clauses, including those in Hindi, are (...)
73It is not entirely clear whether the ungrammaticality of the Hindi example could be rectified by modifying it.12 But let us follow Bhatt in assuming that the utterance’s ungrammaticality results from the fact that the Mat-NP, expressed by a demonstrative us-, is found within another, clause-internal relative clause. In languages like English, relative clauses are recognized to be “syntactic islands” (Ross 1967) precisely because moving a constituent out of a relative clause is prohibited, which then serves as an explanation for why forming an utterance-peripheral correlative with its Mat-NP within another relative clause results in ungrammaticality.13
74In Mano, the Island effect—or, more specifically, the restriction of coreference between the Rel-NP and a NP within a subsequent (cor)relative clause—does not apply. In (24), there are two utterance-peripheral correlative clauses. The referent of the Rel-NP within the first correlative clause is the (zero) subject in the second corelative clause and is indexed on its auxiliary. The first Rel-NP is not taken up again in the matrix clause. The Rel-NP of the second correlative clause is taken up again as an argument of a postposition in the matrix clause.
75(24) Lack of Island constraint on relativization in Mano
- 14 A reviewer suggests that another possible interpretation could be that one correlative is located i (...)
76[CorC1 Rel-NPi…] [CorC2 = MatC1 Rel-NPj …Mat-NPi…] [MatC2 Mat-NPj…]14
|
[CorC
|
Míái
|
lɛ́
|
ō
|
wálàléwèkpɔ́mìà
|
ká
|
ā]
|
|
|
person.pl:h
|
att
|
3pl.exi
|
prophet
|
with
|
dem
|
|
[CorC
|
zīlèēj
|
lɛ́
|
ōi
|
à
|
zɔ̰̀ɔ̰̀-pɛ̀lɛ̀
|
ɛ̄]
|
|
|
road
|
att
|
3pl.exi
|
3sg
|
show-inf
|
dem
|
|
[MatC
|
íìbɛ̀īàmɔ̀
|
tòò
|
ā
|
í
|
lāɓō
|
sɔ̀lɔ̄-ɓō
|
àj
|
kōlà.]
|
|
|
can.be
|
tomorrow
|
dem
|
2sg.conj
|
salvation
|
obtain
|
3sg
|
behind
|
77‘The people who are prophets, the road that they are showing, it is possible that if you stay on that road, you will obtain salvation.’ [MOC, modified]
78A further argument provided by Bhatt concerns the possibility to stack two (or more) utterance-peripheral correlatives.
79It is perfectly possible to have two clause-internal correlatives in Hindi describing two separate Mat-NPs (25).
80(25) Two clause-internal correlatives in Hindi (Bhatt 2003: 507)
81[MatC … [CorCi …] Mat-NPi … [CorCj …] Mat-NPj …]
|
[MatC
|
Ram-ne
|
[CorCi
|
jo
|
laṛkaai
|
tumhaare
|
piichhe
|
hai]
|
|
|
|
Ram-erg
|
|
rel
|
boy
|
your
|
behind
|
be.prs
|
|
|
[us
|
laṛke-ko]i
|
[CorCi
|
jo
|
kitaabj
|
Shantiniketan-ne
|
chhaapii
|
thii]
|
|
dem
|
boy-DAT
|
|
Rel
|
book
|
Shantiniketan-erg
|
print.pfv.f
|
be.pst.f
|
|
[vo
|
kitaab]j
|
dii.]
|
|
dem
|
book
|
give-pfv.f
|
82‘Ram gave the book that Shantiniketan had published to the boy who is standing behind you.’ (Lit. ‘Ram gave [[which book Shantiniketan had published] that book] to [[which boy is behind you] that boy].’)
83Similarly, in Mano such structures are also perfectly possible (26).
84(26) Two clause-internal correlatives in Mano
85[MatC … [CorCi …] Mat-NPi … [CorCj …] Mat-NPj …]
|
[MatC
|
ē
|
[CorCi
|
wììi
|
lɛ́
|
ē
|
à
|
zɛ̄
|
ɛ̄]
|
ài
|
yí-gīnī
|
|
|
2sg.pst
|
|
meat
|
att
|
3sg.pst
|
3sg
|
kill
|
dem
|
3sg
|
interior-lose
|
|
[CorCj
|
gwɛ́ɛ́kòlò
|
vɔ̀
|
lɛ́
|
ō
|
gbēŋ́
|
yí
|
ā]
|
ōj
|
mɔ̀.]
|
|
|
old.man
|
pl
|
att
|
3pl.exi
|
block
|
in
|
dem
|
3pl
|
on
|
86‘He shared the animal that he killed among the elders that are in his neighborhood.’ (Lit.: ‘He, the animali that he killed, shared iti, the eldersj that are in his neighborhood, among themj.’) [el, based on MOC]
87However, it is impossible in Hindi to have two utterance-peripheral correlative clauses (27).
88(27) Ungrammaticality of two utterance-peripheral correlative clauses in Hindi (Bhatt 2003: 508)
89*[CorCi] [CorCj] [MatC … Mat-NPi … Mat-NPj …]
|
[CorCi
|
Jo
|
laṛkaai
|
tumhaare
|
piichhe
|
hai]
|
[CorCj
|
jo
|
kitaabj
|
|
|
rel
|
boy
|
your
|
behind
|
be.prs
|
|
rel
|
book
|
|
Shantiniketan-ne
|
chhaapii
|
thii]
|
[MatC
|
Ram-ne
|
|
Shantiniketan-erg
|
print-pfv.f
|
be.pst.f
|
|
Ram-erg
|
|
[us
|
laṛke-ko]i
|
[vo
|
kitaab]j
|
dii.]
|
|
dem
|
boy-DAT
|
dem
|
book
|
give-pfv.f
|
90‘Ram gave the book that Shantiniketan had published to the boy who is behind you.’ (Lit. ‘[Which book Shantiniketan had published] [which boy is behind you] Ram gave [that book] to [that boy].’)
91If utterance-peripheral correlative clauses were generated in the adjunct-to-matrix-clause position, it would be difficult to explain such a restriction, while it follows from the movement hypothesis. Bhatt explains the restriction by observing that while it is possible to front one, it is impossible to front two adjuncts via wh-movement (28). If clause-internal correlatives are base-generated as adjuncts to the Mat-NPs, it should be impossible to front more than one of them, and this prediction is borne out.
92(28) Impossibility of two utterance-peripheral adjuncts in Hindi (Bhatt 2003: 510)
93wherei whenj [CP… [CP… ti … tj …]]
|
*kahããi
|
kabj
|
[CP
|
Radha
|
soch-ti
|
hai,
|
|
where
|
when
|
|
Radha
|
thing-hab.f
|
be.prs
|
|
[CP
|
ki
|
Ram-ne
|
Sita-ko
|
tohfe
|
ti tj
|
di-ye
|
the?]]
|
|
|
that
|
Ram-erg
|
Sita-DAT
|
presents
|
|
give-pfv.pl
|
be.pst.M.pl
|
94Intended reading: *‘Where and when Radha thinks [that Ram gave presents to Sita]?’
95In Mano, by contrast, it is possible to have two utterance-peripheral correlative clauses (29).
96(29) Two utterance-peripheral correlative clauses in Mano
97[CorCi] [CorCj] [MatC Mat-NPi … Mat-NPj …]
|
[CorCi
|
Wììi
|
lɛ́
|
ē
|
à
|
zɛ̄
|
ɛ̄]
|
|
|
meat
|
att
|
3sg.pst
|
3sg
|
kill
|
dem
|
|
[CorCj
|
gwɛ́ɛ́kòlò
|
vɔ̀j
|
lɛ́
|
ō
|
gbēŋ́
|
yí
|
ā]
|
|
|
old.man
|
pl
|
att
|
3pl.exi
|
block
|
in
|
dem
|
|
[MatC
|
ē
|
ài
|
yí-gīnī
|
ōj
|
mɔ̀.]
|
|
|
2sg.pst
|
3sg
|
interior-lose
|
3pl
|
on
|
98‘He shared the animal that he killed among the elders that are in his neighborhood.’ (Lit.: ‘He, the animal that he killed, shared it, the elders that are in his neighborhood, among them.’) [el, based on MOC]
99The restriction of having two utterance-peripheral adjuncts also does not apply (whether or not it is related to the possibility of having two utterance-peripheral correlatives). In (30), the two adjuncts kálémɔ̀ ‘at home’ and lúó nɔ́fé ‘every day’ are situated in the utterance-initial position and are accompanied by the demonstrative wɛ̄. They can modify the main clause (‘we struggle at home every day’, 30a) as well as the dependent purpose clause (‘so that our kids have food at home every day’, 30b) and any possible combination of these (‘we struggle at home so that our kids have food every day’, 30c, and ‘we struggle at home so that our kids have food every day’, 30d).
100(30) Two utterance-initial adjuncts in Mano
101(a) at homei every dayj [CP… ti tj [CP…]]
102‘We struggle at home every day so that our kids have food.’ [el]
103(b) at homei every dayj [CP… [CP… ti tj ]]
104‘We struggle so that our kids have food at home every day.’ [el]
105(c) at homei every dayj [CP… ti [CP… tj]]
106‘We struggle at home so that our kids have food every day.’ [el]
107(d) at homei every dayj [CP… tj [CP… ti]]
108‘We struggle every day so that our kids have food at home.’ [el]
|
[Kálémɔ̀
|
wɛ̄]
|
[lúó
|
nɔ́fɛ́
|
wɛ̄]
|
[CP
|
kóò
|
gèlè
|
gɔ̰̄
|
|
home
|
dem
|
day
|
every
|
dem
|
|
1sg.ipfv
|
war
|
fight
|
|
[CP
|
kɛ̄
|
kɔ́nɔ́
|
é
|
kɛ̄
|
kò
|
nɔ́ɔ̀
|
nì
|
kɛ̀lɛ̀.]
|
|
|
|
so.that
|
food
|
3sg.conj
|
be
|
1pl.poss
|
child.pl
|
pl
|
hand
|
|
109The rationale behind comparing the movement of adjuncts to IPs (the temporal and place adjuncts in (28)) and adjuncts to NPs in (27) (which are correlative clauses under the movement account) is not sufficiently explained by Bhatt (2003). In other words, it is not clear why the two are expected to be synchronized (i.e., if movement of multiple adjuncts to CPs is impossible, why it should predict the impossibility of movement of multiple adjuncts to NPs). Crucially, assuming that there is a correlation, the fact that movements of multiple adjuncts to CPs is admissible in Mano (30) makes possible the analysis of fronting of multiple correlatives in terms of movement (29). So, this criterion does not provide conclusive evidence about the movement vs. base-generation accounts. The next criterion—Condition-C effects—does.
110In Hindi, if a pronoun c-commands the Mat-NP (the demonstrative phrase associated with a correlative clause), then the pronoun cannot (or at least the speakers prefer that it does not) co-refer to a proper name contained inside that correlative clause. By contrast, if the Mat-NP itself c-commands a pronoun, the pronoun can co-refer to a proper name contained inside the correlative clause. For example, the correlative clause ‘The girl who loves Sita’ in (31) (Lit.: ‘Which girl loves Sita’) has two NPs: ‘the girl’ and ‘Sita.’ The matrix clause ‘she rejected her’ has two pronominal elements with demonstratives, one marked with the ergative (the c-commanding one) and another with the accusative. Now, only the configuration where Sita co-refers with the non-c-commanding NP (‘The girli who loves Sitaj, shei rejected herj’) or with none of the NPs (‘The girli who loves Sitaj, shei rejected herk,’ ‘The girli who loves Sitaj, shek rejected heri’) is permissible, but not the configuration where Sita co-refers to the c-commanding NP (*‘The girli who loves Sitaj, shej rejected heri). The observed restriction on the interpretation of this correlative follows naturally from the more general Condition-C restriction (a referential expression, such as a name, cannot have an antecedent that c-commands it). Indeed, provided that the correlative clause is base-generated in a clause-internal position near the Mat-NP, the proper name within the correlative clause cannot be c-commanded by its antecedent in the matrix clause (*‘The girlj who loves Sitai, shej rejected heri). In example (31) below, one and the same formal structure can have several readings (including an ungrammatical reading in (31d), which is a violation of Condition C).
111(31) Condition C in Hindi (modification of Bhatt 2003: 513)
112(a) [CorCP Rel-NPi… Namej …] […Mat-NPi … Pronj …]
113The girl who loves Sitaj rejected herj
114(b) [CorCP Rel-NPi… Namej …] [ …Mat-NPi … Pronk …]
115The girl who loves Sitaj rejected herk
116(c) [CorCP Rel-NPi… Namej …] [ …Pronk… Mat-NPi]
117Shek rejected the girli who loves Sita
118(d) *[CorCP Rel-NPi… Namej …] [… Mat-NPj … Proni …]
119Shej rejected the girl who loves Sitaj
|
[CorC
|
Jo
|
laṛkii
|
Sita-ko
|
pyaar
|
kar-tii
|
hai]
|
|
|
rel
|
girl
|
Sita-acc
|
love
|
do-hab.f
|
is
|
|
[MatC
|
us-ne
|
us-ko
|
ṭhukraa
|
di-yaa.]
|
|
|
dem-erg
|
dem-acc
|
reject
|
give-pfv
|
120In Mano, however, this restriction does not apply and a proper name within a correlative clause can refer to any NP within the matrix clause. In (32), the proper name within the first correlative clause—Zɛ́ɛ́zú ‘Jesus’—co-refers to the subject of the matrix clause and is therefore indexed on the auxiliary áà. The auxiliary c-commands the Mat-NP, a 3sg pronoun à situated within the PP.
121(32) Violation of Condition-C constraint in Mano
122[CorCP Rel-NPi … Namej …]i [MatC Auxj . . . Mat-NPi . . . ] ~ (31d)
|
[CorC
|
[Ŋwɔ́
|
yīè
|
gbɔ́kò]i
|
lɛ́
|
Zɛ́ɛ́zúj
|
ā
|
kɛ̄
|
kō
|
lɛ̀ɛ̄
|
ɛ̄]
|
|
|
problem
|
good
|
big
|
att
|
P.N.
|
3sg.pst>3sg
|
do
|
1pl
|
to
|
dem
|
|
[MatC
|
áàj
|
ē
|
dìè
|
dɔ̀kɛ̀
|
ài
|
ká
|
kō
|
kɛ̀lɛ̀.]
|
|
|
|
3sg.jnt
|
3sg.refl
|
int
|
give:jnt
|
3sg
|
with
|
1pl
|
hand
|
|
123‘Jesus gave himself to us with the good big thing that he did to us.’ (Lit.: ‘The good big thingi that Jesusj did to us, hej gave himself with iti to us.’) [MOC, modified]
124If the correlative clause is assumed to be generated directly as an adjunct to the matrix clause, there is no violation of the c-commanding restriction between any pronoun in the matrix clause and a proper name in the correlative clause.
125In Hindi, all the three diagnostics listed above—the Island effects, stacking and Condition-C effects—suggest that utterance-peripheral correlatives move from a clause-internal position. Without going into detail about the controversy between surface-based and transformational approaches to syntax, it suffices to observe that all three criteria applied to Mano yielded the opposite results (although the data on stacking is consistent with both movement and base-generation interpretations). This implies that utterance-peripheral correlatives in Mano are generated at least in the adjunct position to the matrix clause, or, as I show in further sections, even higher.
126In all the Mano and Hindi examples reported above, Rel-NP and Mat-NP are coreferential. This also seems to be a requirement of correlative and any other kinds of relative clauses, if we follow the definition of Andrews (2007: 206):
127A relative clause is a subordinate clause which delimits the reference of an NP by specifying the role of the referent of that NP in the situation described by the relative clause.
128This does not hold for all relative clauses across languages, however. In their analysis of Ossetic (Indo-Aryan) correlatives, which can also be clause-internal or utterance-peripheral, Belyaev and Haug (2014) remain agnostic regarding the two hypotheses entertained by Bhatt (whether the utterance-peripheral correlatives are a result of movement or not), pointing out that a choice between transformational and non-transformational approach to syntax is ultimately theory-dependent (the syntactic theory they adopt, LFG, is more surface-oriented). Crucially, they advance additional semantic arguments about a lack of strict coreference between the Rel-NP and Mat-NP, which serve as an additional reason to treat Ossetic correlatives as non-embedded. These arguments are:
129Possibility of a part-whole semantic relationship between Rel-NP and Mat-NP and split antecedence; and
130Possibility of bridging relationship.
131Both these possibilities also hold for Mano. Moreover, as I will show in Section 5, there may not be any Mat-NP strictly or loosely related to the Rel-NP, which points to further semantic independence of the two clauses.
- 15 West African languages, including Mande, are known to have flexible personal indexicality in report (...)
132To begin with, there are cases where the Mat-NP is coreferential to Rel-NP but does not have the same grammatical person. Example (33) begins with a correlative clause in a narrative frame of reference where all the referents are not current speech act participants and therefore are framed in 3rd person, and the past tense is used. The matrix clause is framed as reported speech of one of the participants. The addressee of the reported discourse, which co-refers to the Rel-NP míá ‘the people,’ is encoded with 2pl pronominal markers. This does not reflect the expected indirect strategy of the narrative frame of reference but rather a direct strategy. Because of this break in deictic anchoring15 and despite the coreference, the correlative clause and its Rel-NP (míá ‘people’), on the one hand, and the matrix clause with its Mat-NP (indexed on the 2pl auxiliaries; for another token of Mat-NP expressed by an auxiliary, see example (13)), on the other, are deictically disintegrated.
133(33) Coreference between Rel-NP and Mat-NP in Mano with mismatch in person specification
|
[CorC
|
Míá
|
lɛ́
|
ō
|
kɛ̄
|
yī
|
wɛ̄]
|
|
|
person.pl:h
|
att
|
3pl.pst
|
be
|
there
|
dem
|
|
[MatC
|
kà
|
yíí
|
dò
|
sí
|
ká
|
nɔ̄
|
yèlèmì
|
lɛ̀ɛ̄.]
|
|
|
2pl.sbjv
|
water
|
indef
|
take
|
2pl.conj>3sg
|
give
|
master
|
to
|
134‘The people who were there, you [pl] take some water and give it to the master of the ceremony.’ [MOC, modified]
135Furthermore, in Mano, a pronominal Mat-NP can correspond to a conjunction of several Rel-NPs, a phenomenon called split antecedence. In the example (34), which is about Jesus’ forgiving of the living and the dead, there are two groups of referents: we who follow you1 (the living) and people who died2 (the dead). This example also illustrates a person mismatch, since a 3pl Mat-NP refers to the conjunction of a 1pl referent and a 3pl referent.
136(34) Split antecedence in Mano
137[MatC … [CorCi] [CorCj] Mat-NPi+j …]
|
[CorCj
|
mīāi
|
lɛ́
|
wāà
|
gā
|
ā]
|
ɛ̰́ɛ̰́
|
[CorCi
|
|
|
person:pl
|
att
|
3pl.prf
|
die
|
dem
|
and
|
|
|
[kō
|
dìètíní]j
|
lɛ́
|
kò
|
dɔ̄
|
ī
|
zò
|
|
1pl
|
int
|
att
|
1pl.sbjv
|
install:ipfv
|
1sg
|
heart
|
|
kōlà
|
ā]
|
[MatC
|
í
|
ōi+j
|
mɔ́ɔ̀wèè
|
tó.]
|
|
behind
|
dem
|
|
2sg.conj
|
3pl
|
pardon
|
leave
|
138‘(We are crying unto you) so that you forgive the people who died and us, who follow you.’ (Lit.: ‘So that, we ourselves who stay behind your hearti and people who have diedj, you forgive themi+j.’) [MOC, modified]
139Note that in Hindi, at least according to Bhatt (2003), each relativized Rel-NP must have its own corresponding Mat-NP. Mat-NPs, in turn, are coordinated to express similar constructions (35).
140(35) Conjunction of Mat-NP instead of split antecedence in Hindi (Bhatt 2003: 504)
141[MatC … [CorCPi] Mat-NPi and [CorCPj] Mat-NPj …]
|
[MatC
|
Rahul
|
aajkal
|
[CorCi
|
jo
|
kitaabi
|
Saira-ne
|
likh-ii]
|
voi
|
aur
|
|
|
P.N.
|
nowadays
|
|
rel
|
book.f
|
Saira-erg
|
write-pfv.f
|
dem
|
and
|
|
[CorCj
|
jo
|
cartoonj
|
Shyam-ne
|
banaa-yaa]
|
voj
|
paṛh
|
rahaa
|
hai.]
|
|
|
rel
|
cartoon
|
Shyam-erg
|
make-pfv
|
dem
|
read
|
prog
|
be.prs
|
142‘Nowadays, Rahul is reading the book that Saira wrote and the cartoon that Shyam made.’ (Lit. ‘Nowadays, Rahul is reading [[which book that Saira wrote] that (book)]i and [[which cartoon that Shyam made] that (cartoon)]j.’)
143In Ossetic, however, split antecedence is possible (36).
144(36) Split antecedence in Ossetic (Belyaev & Haug 2014: 11)
145[CorCPi] [CorCPj] [MatC … Mat-NPi+j …]
|
[CorCi
|
Sard-ɜj
|
sə
|
konflikt-tɜi
|
išt-a]
|
[CorCj
|
sə
|
χarakter-tɜj
|
|
|
life-abl
|
what
|
conflict-pl
|
take-pst.3sg
|
|
what
|
character-pl
|
|
ɜvdəšt-a]
|
[MatC
|
wədoni+j
|
wəd-əštə
|
kadɜǯ-ə
|
ɜrmɜg.]
|
|
demonstrate-pst.3sg
|
|
those
|
be-pst.3pl
|
legend-gen
|
material
|
146‘The characters that he took from life and the characters that he demonstrated were legendary material.’ (Lit.: ‘What conflictsi he took from life, what charactersj he demonstrated, theyi+j were legendary material.’)
147Bridging is a relationship between noun phrases that are not coreferential but tightly connected contextually. In example (37), there is no coreference between NPs in the two utterances. However, the windows are understood to be contextually related to the room as the windows of the room.
148(37) Bridging in English (Clark 1975: 171)
149I walked into the room. The windows looked out to the bay.
150In Mano correlative clauses, the referents of the Rel-NP and an NP in the matrix clause which is understood to be its correlate can similarly be in a bridging relationship (38). The marker à in Mano is dedicated to marking that a referent of a given NP is in a bridging relationship to the referents introduced earlier (Khachaturyan 2020c). It is homonymous with the 3sg pronoun, which can be used to express an inalienable possessor. In this context, however, the only interpretation is that of bridging, because ‘stage’ is a relational noun, which, in the function of the possessee, combines with the possessive pronoun: e.g., là stage 3sg.poss internship, *à stage 3sg internship ‘her internship.’
151(38) Bridging in Mano correlatives
|
[CorC
|
Làkólè
|
lɛ́
|
ɓà
|
ɲɛ̄
|
ɛ̄]
|
|
|
school
|
att
|
2sg.sbjv>3sg
|
finish:ipfv
|
dem
|
|
[MatC
|
íì
|
à
|
stage
|
mɛ́
|
dā̰.]
|
|
|
2sg.ipfv
|
bridg
|
[fr]internship
|
prev
|
learn:ipfv
|
152‘The school that you finish, (afterwards) you learn at an internship (related to the school).’ [MOC, modified]
153The school and the internship are understood to be contextually related: similarly as in the French system of higher education, it is typical of Guinean students to undergo some kind of professional training as part of their studies.
154The next example (39) is similar to the previous one. Here, the marker of bridging is part of the portmanteau auxiliary wāà but it is unambiguously present because it triggers a low tone on the head noun (on low-tone head marking in Mano bridging constructions, see Khachaturyan 2020c).
155(39) Bridging in Mano correlatives
|
[CorC
|
Nú
|
kɔ̄āà
|
ɓō
|
ɓɛ̄]
|
|
|
come>att
|
1pl.prf>3sg
|
implement
|
dem
|
|
[MatC
|
míá
|
yā
|
wāà
|
ŋwɔ̀
|
yí
|
dɔ̄.]
|
|
|
person.pl:h
|
dem
|
3pl.prf>bridg
|
problem:cstr
|
interior
|
know
|
156‘That we have come, they had understood it.’ (Lit.: ‘Coming that we have implemented, they have understood the problem.’) [MOC]
157The Rel-NP and a correlate NP in the matrix clause can also belong to the same set (40). Here, there are two possible interpretations of the marker à: either as a 3sg pronoun or as a bridging marker.
|
(40)
|
[CorC
|
Kpàléŋwɔ̀
|
tɛ́
|
ī
|
kɛ̄
|
à
|
kɛ̄-pɛ̀lɛ̀
|
cafarnom
|
ɓɛ̄]
|
|
|
|
|
miracle
|
att
|
2sg.pst
|
do
|
3sg
|
do-inf
|
N.P.
|
dem
|
|
|
[MatC
|
à
|
dò
|
kɛ̄
|
ē
|
dìè
|
pàà!]
|
|
|
3sg/ bridg
|
indef
|
do
|
3sg.refl
|
int
|
at
|
158‘The miracles that you were performing in Capharnaum, perform some of them in your own country!’ [MOC]
159I am not aware of the availability of bridging in Hindi, but Ossetic does allow bridging (41).
160(41) Bridging in Ossetic correlatives (Belyaev & Haug 2014: 12)
|
[MatC
|
ɜž
|
χorž
|
žət:-on
|
[CorC
|
de=
|
št’ol-əl
|
sə
|
k’am
|
iš],
|
|
|
I
|
well
|
know-pst.1sg
|
|
your
|
table-super
|
what
|
photo
|
is
|
|
wə-sə
|
lɜpːu-jə.]
|
|
that-attr
|
boy-gen
|
161‘I knew well the boy whose photo is on your table.’ (Lit.: ‘I knew well, what photo is on your table, that boy.’)
162Given that the relationship between Rel-NP and its correlate in the matrix clause in Ossetic is not restricted to coreference and it allows for split antecedence and bridging, Belyaev and Haug (2014) suggest analyzing it as one of pragmatic anaphora resolution. Since these arguments yield the same results in Mano, it should be concluded that Rel-NP and Mat-NP have an anaphoric link in Mano as well. In addition, as Motter (2023) demonstrates, the semantic mismatches serve as an additional argument against the movement account.
163Section 3 has demonstrated that in Hindi there are strong syntactic arguments in favor of the hypothesis that clause-peripheral correlatives are generated clause-internally and then moved to the position adjoined to the matrix clause. In Mano, by contrast, these arguments do not replicate (except for the stacking argument, which gives inconclusive results). Therefore, correlative clauses should be interpreted as adjoined to the matrix clause and base-generated in that position.
164Section 4 applied the arguments suggested by Belyaev and Haug (2014) concerning semantic mismatches between the Rel-NP and its correlate in the matrix clause. This time, they gave the same result as the Ossetic analyzed by these two authors. In Mano, similar examples can also be found, but in addition, I have demonstrated a case of different person specifications because of a shift of deictic frame of reference within the same utterance with reported speech. These examples further demonstrate that the relative(-like) construction and the matrix clause are loosely connected not only syntactically but also semantically.
165Three more arguments for full syntactic and semantic independence of Mano correlatives should be mentioned.
166First, the relativization marker lɛ́ is not restricted to that function and can be used in independent clauses as a predicative center of identifying constructions (42), a focus marker (48; Khachaturyan 2023) and a marker accompanying adnominal demonstratives (6; Khachaturyan 2020a). Surely, it is still left to be demonstrated how these functions are connected, which is beyond the scope of the present paper. But assuming that they are — and the underlying invariant function seems to be attention-drawing, whence the gloss — it should be clear for our current purposes at least that lɛ́ does not function as a marker of syntactic subordination.
167(42) identifying construction with lɛ́ in Mano
|
Gààzù
|
dōó
|
wɛ̄
|
lɛ́
|
nɔ́
|
wɛ̄.
|
|
mirror
|
one
|
dem
|
att
|
just
|
dem
|
168‘It is just the same mirror.’ (Lit.: ‘It is just one mirror.’) [MOC]
169Second, the demonstrative occurring at the end of the correlative clause is otherwise used in subordinate clauses situated to the left of the matrix clause (7, 9). More broadly still, the demonstratives are also used to accompany left-dislocated constituents. The combination of left-dislocation with demonstratives, in turn, is a common means of introducing topics (43; see also Khachaturyan under review). Therefore, correlatives are best described as minimally integrated topic-like constituents (see Motter 2023 for a similar analysis).
|
(43)
|
Léé
|
wɛ̄
|
ō
|
là
|
nɛ́
|
yà-pìà
|
dīà̰.
|
|
|
woman:h
|
dem
|
3pl.exi
|
3sg.poss
|
child
|
sit-inf>3sg
|
back
|
170‘That woman, they are putting a child on her back.’ [MOC]
171Finally, the same structure that is used for utterance-peripheral correlatives (a left-dislocated NP, followed by lɛ́, followed by a clause where the NP’s referent is taken up by a resumptive element and followed by a demonstrative) can be used in cases where there is no pair of NPs in the correlative and the matrix clause, that are coreferential or connected via bridging and split antecedence.
172(44) Relative-like construction in Mano without pragmatic relationship between the Rel-NP and any of the NPs in the matrix clause.
|
[Ŋwɔ́
|
wā
|
gèē
|
yālá
|
ā]
|
[í
|
ī
|
lìà
|
gā
|
nɔ́!]
|
|
problem
|
3pl.pst>3sg
|
say
|
yesterday
|
dem
|
2sg.conj
|
2sg
|
silence
|
die
|
just
|
173‘The thing that they said yesterday, just shut up!’ [MOC]
174Note that (44) is similar in many ways to example (4) introduced above and repeated below as (45).
|
(45)
|
[CorC
|
[Ŋwɔ́i
|
lɛ́
|
ā
|
gèē
|
wɛ̄]
|
|
|
problem
|
att
|
3sg.pst>3sg
|
say
|
dem
|
|
[MatC
|
ŋ́ŋ̀
|
lō
|
ài
|
kɛ̄-ɛ̀.]
|
|
|
1sg.ipfv
|
go:ipfv
|
3sg
|
do-ger
|
175‘The thingi that he said, I will do iti.’ [MOC]
176The only difference is that in (45), there is a Mat-NP coreferential to the Rel-NP. Therefore, (45) can be interpreted as a relativization strategy, while (44) does not have a corresponding Mat-NP and does not function as a relative clause. Instead, the correlative-like part in (45) functions as a hanging topic.
177Thus, even though correlative clauses are associated with a number of morphosyntactic properties, including the marker lɛ́, the final demonstrative and the left-dislocation of the Rel-NP, none of these are specifically tied to correlative constructions, or to subordination. Instead, Mano correlative constructions can be considered a special case of a paratactic topic-comment construction. Furthermore, this construction receives a relativization-like reading under several conditions, including (partial) coreference.
- 16 Moreover, the cooptation framework allows for a fully synchronic reading of a given construction an (...)
178The existence of a relatively conventionalized (but at the same time optional) pragmatic interpretation of a paratactic construction can be considered a case of cooptation (Mauri & Sansò 2011). The term “cooptation” was suggested by Mauri and Sansò to cover cases where the source construction developing new pragmatic and grammatical functions is compatible with its source function and where there is no structural change involved, which is precisely the case of Mano correlatives. Not surprisingly, by the same logic, the two other types of “relative” clauses—detached with lɛ́ and paratactic—can also be considered a result of cooptation of paratactic clause-combining, although perhaps with a much lesser degree of conventionalization.16
179An analysis in terms of parataxis has been recently suggested for Hittite correlatives (Motter 2023). Motter observed the same semantic flexibility in the relationship between the Mat-NP and the Rel-NP—up to the lack of any candidate Mat-NP in the subsequent clause. To the extent that the available Hittite corpus allows, he also observed a lack of movement effects (specifically, a violation of Condition C). He concluded that Hittite correlative clauses are best understood as hanging topics and that the overall correlative structure (correlative clause and the subsequent clause) is a paratactic structure involving syntactically independent clauses conjoined together.
180There are two possible analyses of the conjunction: either the two clauses are conjoined in syntax, which is also put forward by Davison (2009) for Sanskrit, or more loosely in discourse, as Motter (2023) proposes. I illustrate the two analyses for the Mano examples (45) and (44) by (46a) and (46b), respectively; the syntactic representation in (46a) is essentially a modification of the one in Figure 2 above. The difference between the two analyses boils down to whether there is a higher CP projection uniting the two clauses.
181As Motter suggests, the distinction between the two analyses basically revolves around the metatheoretical question of where one wishes to put boundaries on syntactic phenomena. If we admit that clause-combining in Mano is outside the domain of syntax, or at least at the interface with discourse, the usage-based approach to syntactic phenomena offers avenues for further exploration. Such an approach treats an utterance not as an isolated artifact, which can be analyzed post hoc, but as a product of emergent elaboration that, crucially, is embedded in a larger discourse context (Fox & Thompson 2007). Under such an account, correlative constructions become a member of a family of constructions (Hopper 2001) and their relativization-like interpretation can be viewed as epiphenomenal to their more general interactional functions. Although this topic is beyond the scope of the present paper, I would like to illustrate the idea by one last group of examples. As example (39) already showed, the Rel-NP can be expressed by a nominalized verb. Such relativizations of verbal nominalizations appear to be especially common in narrative discourse—in particular, as part of the head-tail linking strategy, as in (47), but also without repeating any previous verbal content—and can be interpreted as a temporal dependent clause.
|
(47)
|
[Ló
|
wā
|
ɓō
|
ā]
|
|
|
go>att
|
3pl.pst>3sg
|
implement
|
dem
|
|
áà
|
gèè
|
ē
|
dàā
|
lɛ̀ɛ̄…
|
|
3sg.jnt
|
say:jnt
|
3sg.refl
|
father
|
to
|
182‘[They acknowledged their guilt, they got up, he and his father, and went to the bush.] While they were going (lit.: going that they did) he said to his father…’
183Crucially, such relativized nominalizations may also receive a causal interpretation, as in (48):
|
(48)
|
[Nū
|
lɛ́
|
Sɔ̄gīpāà
|
ā
|
ɓō
|
ā]
|
kō
|
pɛ̄
|
sɔ̀lɔ̄=ɓō.
|
|
|
come
|
att
|
P.N.
|
3sg.pst>3sg
|
implement
|
dem
|
2pl.pst
|
thing
|
obtain
|
184‘Because Soguipa came we obtained many things.’ (Lit.: ‘Coming that Soquipa did, we obtained thing.’)
185Thus, the interpretation of the correlative clause-like constituent, especially in the absence of an anaphoric relationship with the matrix clause, is essentially open-ended. But the presence of anaphoric relationship does not automatically yield a relative interpretation either. In the next example (49), there is a perfectly coreferential Mat-NP, and thus the construction is indistinguishable from a correlative relativization, and still the primary interpretation is causal. Indeed, the male speaker is asking his wife for some salt, which is unusual given that men are only occasionally involved in cooking and women are responsible both for cooking and seasoning and also for regulating meal hours. Asking for salt outside a regular meal is very unexpected. To explain why he needs it, the man says that he has roasted some cassava on the fire, presumably as a snack, which is an understandable thing to do.
|
(49)
|
[Béì
|
pɛ̰́ɛ̰́
|
mā
|
yà
|
tíè
|
dḭ̀ḭ̄
|
ā̰,]
|
|
|
cassava
|
slice>att
|
1sg.pst>3sg
|
sit
|
on.fire
|
there
|
dem
|
|
ŋ́ŋ̀
|
lō
|
à
|
dà-à
|
à
|
mɔ̀.
|
|
1sg.ipfv
|
go:ipfv
|
3sg
|
fall-ger
|
3sg
|
on
|
186‘[Give me some salt,] because I put cassava slices on the fire, I will put it (salt) on them (the slices).’ (Lit.: ‘The cassava slices that I put on the fire OR It is cassava slices that I put on the fire, I will put it on them.’)
187The reason for a causal interpretation is because the correlative-like constituent is essentially a thetic clause: all its content (except the subject that is topical) is new to the hearer. Indeed, the marker lɛ́ and its variants can also be used in cleft-like (including thetic) constructions (Khachaturyan 2023), which is illustrated by the second variant of the literal translation. Thetic clauses, in turn, are often used in explanations (Vydrina 2020).
188Examples (47–49) suggest that the correlative-like clause should be viewed as an independent discourse contribution—a (re)introduction of a discourse referent and some (old or new) information about it—which is then elaborated upon in the subsequent clause. The resulting interpretation (relative or causal or head-tail linkage or others, not mentioned in this paper) is essentially epiphenomenal to a more general process of emergent syntactic elaboration. A further grounding of this idea should be done in future research.
189Given that, cross-linguistically, correlative clauses are often adjoined highly, they can be considered a prime example of an erstwhile paratactic construction acquiring subordinating properties (Givón 2009). Evidence from Mano based on criteria developed by Bhatt (2003) and Belyaev and Haug (2014) supports this hypothesis and illustrates such development at an early stage. However, the extent of these subordinating properties should be scrutinized, and simply postulating that a correlative clause is a clause-level adjunct misses many important details in which correlative clauses in different languages are similar and different. My close comparison of Mano with Hindi and Ossetic has demonstrated that Mano correlatives are fully syntactically and semantically independent from the matrix clause and the relative interpretation of the sequence “correlative clause–matrix clause” is a result of a pragmatic cooptation of a paratactic topic-comment-like construction.
190Finally, I have paved the way for further usage-based explorations of the discourse functions of correlative-like clause-combining. This has been made possible not by an a priori choice of framework (usage-based vs. generative syntax), but it emerged from rigorous syntactic testing that showed that Mano correlatives are essentially a discourse phenomenon or, at minimum, at the interface between syntax and discourse. Although such explorations are beyond the scope of the present paper, I hope to have laid a foundation for a productive combination of usage-based and generative accounts of the same phenomenon, contributing to establishing a dialogue between these theoretical frameworks.
1911 – first person
1922 – second person
1933 – third person
194abl – ablative
195acc – accusative
196att – attention drawing marker
197attr – attributive
198bridg – bridging
199C – clause
200conj – conjunctive
201Cor – correlative
202CP – complementizer phrase
203cstr – construct form
204dem – demonstrative
205el. – elicited
206emph – emphatic
207erg – ergative
208exi – existential
209f – feminine
210fr – French
211fut – future
212gen – genitive
213ger – gerund
214h – high tone
215hab – habitual
216ident – identifier
217indef – indefinite
218inf – infinitive
219int – intensifier
220ip – inclusory pronoun
221inflection phrase (corresponding to a finite clause)
222ipfv – imperfective
223jnt – conjoint
224Mat – matrix
225MOC – Mano Oral Corpus
226neg – negation
227NP – noun phrase
228P.N. – proper noun
229pfv – perfective
230pl – plural
231poss – possessive
232prev – preverb
233prf – perfect
234prog – progressive
235prs – present
236pst – past tense
237q – question
238refl – reflexive
239rel – relativizer
240sbjv – subjunctive
241sg – singular
242super – superlative