Survey chapter: Yimas-Arafundi Pidgin

Structure data for these languages can be found in structure dataset 69.

1. Introduction

Yimas-Arafundi Pidgin (autoglossonym: karəm taŋgwapasambi malalək, literally ‘talk for turning and conversing’, Tok Pisin tainim tok) is/was spoken in the villages along the Arafundi River in the East Sepik Province of Papua New Guinea. It is/was spoken in villages speaking two unrelated native languages, Yimas and Arafundi, and the population of these villages currently is around a thousand (Yimas villages, c. 250, Arafundi villages, c. 750), but it was never spoken by more than a handful of men in each of these villages, those who had the clan-based right to carry on trading relationships with men in villages speaking the other language, and there were never any speakers outside of the Yimas and Arafundi villages. Yimas-Arafundi Pidgin is now extremely endangered, at least moribund, and quite probably extinct. I suspect it has been moribund since at least the 1960s. The official language of Papua New Guinea is English, but in the isolated and rural area in which the Pidgin was spoken the practical everyday language of contact is now Tok Pisin. Not only is Tok Pisin the cause of the probable extinction of Yimas-Arafundi Pidgin, it is well on its way to killing off Yimas itself, as it too now has no speakers under 30, and an estimated total of speakers of 50 out of a population of 250. Arafundi, on the other hand, for the moment still seems viable, particularly in the very isolated villages up in the mountains which form the source of the Arafundi River. It needs to be noted that the corpus on which this sketch is based is rather small, 4 texts, a short lexicon and some grammatical notes, as work on the Pidgin was a quite peripheral interest during my fieldwork in the area. The Pidgin unfortunately was also discovered late in my fieldwork period, quite serendipitously in fact: On overhearing a couple of men from the Arafundi-speaking village of Auwim, who had come to Yimas village to trade. Prior to this I had had no inkling of the pervasive role of pidgin languages in the trading networks of Yimas village. Hence there are many gaps in my knowledge and documentation of the language, and it is highly unlikely that any further fieldwork now could ameliorate the situation given its current status.

2. Sociohistorical background

The Sepik region of Papua New Guinea is highly linguistically diverse, and in precontact times facility in more than one’s native language was essential in carrying out the regular commercial exchange relationships among villages that are so typical of the area. Yimas-Arafundi Pidgin was derived from Yimas and Arafundi, two adjoining languages along the Arafundi River in this region. It is actually one of a family of Yimas-based pidgins used by male Yimas villagers in their trading relations with neighbouring groups. There are at least three others (Williams 2000), but Yimas is the dominant lexifier language in all of them (see Foley (1991) for a grammar of Yimas). As in keeping with the general Sepik pattern (Gewertz 1983) the primary trading network is Arafundi sago, the carbohydrate staple, exchanged for Yimas fish. These trade contacts are not sporadic, but are sustained and passed down in clans from generation to generation. Trading partners as well as the proper pidgin language used in interactions with them was part of the clan-based patrimony passed from father to son. A clear status differential marks these exchanges: The Yimas fish suppliers occupy a superior position socially and economically vis à vis the sago suppliers. Other goods besides fish and sago are exchanged between the two groups, but in all cases it is the Yimas who supply the prestige goods. Given these social circumstances, it is expected that the pidgin used will have Yimas as the superstrate language and Arafundi as the substrate, hence Yimas-Arafundi Pidgin. Yimas-Arafundi Pidgin is a relatively stable pidgin; native speakers of both Yimas and Arafundi speak it much the same, the small differences being phonological. Lexically, Yimas-Arafundi Pidgin is largely Yimas: Over 80% of the lexicon is Yimas-derived, the rest coming from Arafundi or from unplaced sources (plus a small mixture of Karawari forms, probably a transfer from the Yimas-Karawari Pidgin). Grammatically, Yimas-Arafundi Pidgin represents a massive simplification of Yimas, a highly morphologically complex polysynthetic language, with some admixture of simplified Arafundi structures. It is demonstrably not Arafundi relexified with Yimas content morphemes. It is, in turn, quite different from the Yimas pidgin used to trading partners speaking Alamblak (Williams 2000), demonstrating that it is indeed a true conventionalized pidgin, not just a foreigner talk register of Yimas. It is particularly noteworthy that still another Yimas-lexified pidgin was used in trading contacts with Karawari speakers (Williams 2000). Yimas and Karawari are closely related languages (Yimas is not related genealogically to either Arafundi or Alamblak, nor they to each other), almost dialects of the same language: Karawari speakers within earshot can often understand a good deal of what is said when overhearing casual conversations in Yimas. The fact that a pidgin was employed in Yimas-Karawari trade encounters, not stable bilingualism or a koine derived from the two, is powerful testimony to the salience of these clan-based conventionalized codes, the pidgins, as indexes of the intervillage trading encounter.

3. Sociolinguistic situation

Yimas-Arafundi Pidgin never creolized, and hence never had any native speakers. Its usage was restricted to the male members of the clans who had trading rights to the Arafundi-speaking villages, particularly Auwim village, and never spread beyond that. It was also contextually restricted to situations of inter-language trading and exchange. Today it is on the verge of dying out or more probably already has done so, as its function has been usurped by Tok Pisin. I estimate that currently there are less than five speakers among both the Yimas and Arafundi speaking villages, all over the age of 60 in a country whose male life expectancy is 55 years. There is a very high likelihood that there are no speakers remaining, and certainly there will be none in the very near future.

4. Phonology

Yimas-Arafundi Pidgin has a simple system of 4 vowel phonemes, with no length or nasalization distinctions (see Table 1). The vowel system is identical to that of Yimas itself and quite different from the 7-vowel phoneme system of Arafundi. Again, as in Yimas, the /ə/ is usually (maybe always) epenthetic to break up unlicensed consonant clusters, so underlyingly the language could probably be analyzed as having only three vowel phonemes.

Table 1. Vowels

front

central

back

close

i

u

mid

ə

open

a

There are 15–16 consonants in Yimas-Arafundi Pidgin (see Table 2), but it is not clear that all are phonemic. In particular, the status of the prenasalized stops is problematic. In Yimas the equivalent series are best analyzed as clusters, consisting of a nasal plus an underlying voiceless stop, realized allophonically as voiced. There is some marginal evidence that the former analysis may be correct for the Pidgin as well, as non-homorganic clusters of nasal plus stop are found in the language: manba ‘many’. This could be analyzed as /manpa/ with allophonic voicing as in Yimas, but an alternative /man(m)ba/ with loss of prenasalization following a preceding nasal cannot be entirely ruled out on the basis of the data (much would depend on the linked analysis of epenthetic /ə/).

Table 2. Consonants

bilabial

alveolar

palatal

velar

plosive

voiceless

p

t

k

voiced

(m)b

(n)d

(ɲ)ɟ <j>

(ŋ)g

nasal

m

n

ɲ

ŋ

tap

r

fricative

voiceless

s

lateral

(l)

glide

w

j <y>

The voiceless palatal plosive is missing, because its usual actual phonetic realization is [s], although some speakers do pronounce it as [c] or [tʃ], and for these speakers there are no fricatives at all in the phonemic inventory. Generally, although not always, the palatal lateral is lacking in the Pidgin in contrast to Yimas itself, and typically occurs in the Pidgin through hypercorrection by native speakers of Yimas; the forms in the Pidgin are normally the historically antecedent forms of the synchronic Yimas forms with the palatal lateral: Yimas-Arafundi Pidgin: miria, Yimas mila ‘moon’. All the palatals in Yimas itself have resulted probably relatively recently from a productive rule of palatalization of coronal consonants, while they are an old entrenched series in Arafundi (there are only palatal plosives and the nasal in this language, no palatalized liquids or rhotics). Evidence suggests that the Yimas-Arafundi Pidgin is more like Yimas in this regard, as affixes do exhibit allomorphy between coronal and palatal allomorphs, as with ‑nan ~ ‑ɲan (non-future tense).

5. Noun phrase

Nouns are simple in Yimas-Arafundi Pidgin, with no bound morphology. There is no distinction for gender, which is signalled by distinct words: payum ‘man/male’, aykum ‘woman/female’. Number is often not indicated at all, but if it is, this is done by following the noun with a numeral like kundamwin ‘two’ or the quantifiers asəŋ or manba ‘many’, although the latter two in particular seem restricted to animate nouns in the corpus. There are no determiners per se, but the third person singular pronoun / distal deictic mən very commonly follows nouns or NPs in texts with established referents.

There are two deictics, exhibiting a proximal (nak ‘this’) versus distal (mən ‘that’) contrast, that normally follow the noun, although they can flank it on either side (all examples of prenominal deictics in the corpus involve nak). These are employed in combination with the oblique postposition to indicate location: nak kandək ‘here’ versus mən kandək ‘there’.

The numeral system in Yimas-Arafundi Pidgin is a mixture of base 2 like Arafundi and base 5 and base 10, as in Yimas:

1 mban

2 kundamwin

3 kundamwin mban (i.e. 2+1)

4 man (‘other’) kundamwin (i.e. another 2, 2+2)

5 tam

6 maŋgat (‘other side of something’) mban tam

(i.e. other side 1, and 5)

7 maŋgat kundamwin tam (i.e. other side 2, and 5)

….

10 nuŋgarawri

Numerals always follow the nouns they enumerate (e.g. aykum kundamwin ‘two women’), and if a deictic is present, they follow that too: aykum mən mban [woman that one] ‘one woman’.

Adjectives are invariant in Yimas-Arafundi Pidgin, usually end in -n and precede the noun they modify: kəpan wakən [big snake] ‘a big snake’, man kuraŋ [other coconut] ‘another coconut’. There are no ways to form comparatives or superlatives in Yimas-Arafundi Pidgin; one would simply say for example ‘X is big, Y is small’.

Yimas-Arafundi Pidgin has a single set of independent pronouns that are used in all syntactic functions: subject, object, possessors (see Table 3). They can formally distinguish three persons and three numbers, but in practice the base forms found in the singular are used in non-singular contexts as well, unless it is necessary to emphasize the dual or plural number of the pronouns’ referents. Distinct portmanteau forms for number are only found in the first person: SG ama DU kapa PL paŋgət, but for the non-singular forms number markers kundamwin for dual and asəŋ for plural can optionally be added. For the second and third person, only person is indicated in the base form (2nd mi, 3rd mən), and if number is to be made explicit, the non-singular number markers must be used, e.g. kundamwin for dual and asəŋ or manba for plural. No indefinite pronouns have been identified.

Table 3. Independent personal pronouns

1sg

ama

2sg

mi

3sg

1du

2du

3du

mən

kapa (kundamwin)

mi kundamwin

mən kundamwin

1pl

paŋgət (asəŋ)

2pl

mi asəŋ/manba

3pl

mən asəŋ/manba

Adnominal pronominal possession is indicated in Yimas-Arafundi Pidgin by adding the suffix -nakən to the corresponding independent pronoun. The possessor precedes the possessed noun: mən-(n)akən wi [3SG-POSS name] ‘his name’, ama-nakən tam [1SG-POSS dog] ‘my dog’. Interestingly, when the possessor is non-singular and marked as such by a quantifier or numeral, the bound pronominal possessor is always third person singular: kapa kundamwin mən-(n)akən tam [1DU two 3SG-POSS dog] ‘our (DU) dog’. Nominal possessors can only be expressed through a co-occurring pronominal and precede this, though it is likely this position is not fixed: Paɲanmaŋ mən-(n)akən wi [Paɲanmaŋ 3SG-POSS name] ‘Paɲanmaŋ’s name’, arəŋgawŋ mən-(n)akən tanəm [flying.fox 3SG-POSS bone] ‘a flying fox bone’.

I have no examples of a possessor co-occurring with an adjective in the corpus, and in fact in Yimas itself, such a structure is impossible, nor can either co-occur with a prenominal deictic, so in summary the structure of the NP in Yimas-Arafundi Pidgin is as follows:

possessor/adjective + noun + deictic + quantifier/numeral

where all but the noun are optional, as there is no class of inalienably possessed nouns in the language.

6. Verb phrase

Verbs in Yimas-Arafundi Pidgin are obligatorily inflected with suffixes for tense/status, with a binary distinction of non-future/future (see Table 4).

Table 4. Tense-aspect-mood markers

tense/aspect

mood

-nan

non-future

-n ~ -nak

imperative/hortative

-n anak

future

ta-

imperfective

məndək-

perfective

aka ~ kakan

negative

The non-future is marked with a suffix -nan (allomorph -ɲan following palatal consonants or the high front vowel) and covers all past and present time events:

(1)
aykum
woman
mariak-nan
talk-nfut
The woman spoke.
(2)
mən
3
manba
many
am-nan
eat-nfut
They are eating.

The future has a complex marking: a suffix -n in combination with the invariable copula anak. It is strictly future, expressing events deictically located in future time and not just unreal events:

(3)
ama
1sg
arəŋ
tomorrow/yesterday
wa-n
go-fut
anak
fut
I'll go tomorrow.

Aspect is expressed through the use of serial verb constructions. The imperfective or progressive aspect is indicated with the root ta-, which has no independent meaning:

(4)
manan
cult.house
kandək
obl
t-mbi
put.inside-dep
ta-nan
ipfv-non-future
(He) was putting (them) inside the cult house.

The verb root məndək- ‘finish’ is used to express perfective aspect:

(5)
mən
3sg
yaki
tobacco
wayk-mbi
buy-dep
məndək-nan
pfv-nfut
He has already bought tobacco.

The imperative/hortative mood has two allomorphs; Transitive verbs require -nak (cf. 6), but intransitive verbs can occur with either -n or -nak (cf. 7), although it is more common with intransitive verbs, especially deictic motion verbs like ‘come’ and ‘go’ to find -n (cf. 8).

(6)
paŋgət
1pl
nak
this
kumbut
village
ndak-nak
leave-imp.tr
Let us leave this village.
(7)
man
other
kumbut
village
kandək
obl
wa-n/nak
go-imp.intr/imp.tr
Let's go to another village.
(8)
mi
2sg
nak
this
kandək
obl
ya-n
come-imp.intr
You come here!

Negative polarity has two markers in Yimas-Arafundi Pidgin, aka and kakan, with the latter vastly the more frequent. The difference seems to be essentially determined by placement and clause type: Aka is preverbal and while rare in the corpus always occurs in nominalized clauses:

(9)
mən
3
manba
many
aka
neg
pait
fight
takari-kŋ
make-nmlz
mambi
then
andi
land
kakan
neg
If they didn't fight, they'd have no land.

Kakan functions like a negative copula and can mean ‘there is no’ or ‘don’t have’, as in example (9), but it is also the general marker of negative polarity in the language for independent clauses:

(10)
a.
am-kəkŋ
eat-nmlz
kapawri
food
awri-ɲan
get-nfut
kakan
neg
They didn't get food to eat.
b.
yəm
water
mən
that
muri-mbi
flee-dep
ta-nan
ipfv-nfut
kakan
neg
The water wasn't draining away yet.

The corpus presents no true uses of modal verbs in the Yimas-Arafundi Pidgin, but there is a specialized use of tay- ‘see’ to mean ‘try’ in serial verb constructions rather like modal auxiliaries in other languages:

(11)
andi
land
mən
that
awt-mbi
get-dep
mayn-(m)bi
fit-dep
tay-ɲan
see-nfut
They got some earth and tried it for its suitability.

Yimas-Arafundi Pidgin has a copula which is extensively used; its form is anak and its corresponding negative is kakan, exemplified in (9). The copula often has a presentative meaning ‘it/there is’: ama anak [1SG COP] ‘it’s me’, kapuk mən anak [noise that COP] ‘there is noise, it’s noisy’, but more commonly it is a simple existential connective; nak mamakən andi anak [this bad land COP] ‘this is bad soil’. In combination with the oblique postposition kandək, it normally means ‘have’:

(12)
yangwara
tree
mən
that
pandaŋ
hole
kandək
obl
anak
cop
The tree has a hole.

7. Simple sentences

Yimas-Arafundi Pidgin is a verb-final language, but not strictly so, as oblique phrases can and very commonly do follow the verb (see also examples 20c and 21):

(13)
kuraŋ
coconut
wambaki-ɲan
throw-nfut
yəm
water
kandək
obl
They threw a coconut into the water.

Also, but rarely, an object noun phrase, but never a subject, can follow the verb:

(14)
wut
night
kandək
obl
mənda
and.then
kra-nan
cut-nfut
taŋgut
rope
mən
that
Then at night they cut the rope.

But this only seems to occur in the corpus when the object noun phrase has established reference and is topical.

When two unmarked preverbal core noun phrases occur in the clause, a rare situation given the pervasive zero anaphora for noun phrases with established referents, no fixed ordering is mandated, with both SOV and OSV orders possible and with the latter, perhaps surprisingly, marginally more common in the corpus:

(15)
a.
mən
3
kundamwin
two
ama
1sg
tay-ɲan
see-nfut
kakan
neg
I didn't see those two.
b.
paŋgət
1pl
nak
this
kumbut
village
ndak-nak
leave-imp.tr
Let us leave this village.

Dative noun phrases or animate arguments with the semantics of recipient or benefactive are typically marked with the postpostion namban, although this is not obligatory:

(16)
a.
mambi
then
aykum
woman
namban
dat
mariak-nan
say-nfut
Then he told the woman.
b.
yaki
tobacco
pwisəŋ
bundle
mən
that
Yimas
Yimas
namban
dat
aca-nan
give-nfut
They gave the Yimas a bundle of tobacco.

The postposition namban can also be used to mark animate objects of transitive verbs, but again as examples like (15a) illustrate, its usage here is not obligatory, and, in fact, is rather uncommon in spontaneous text:

(17)
mən
3sg
ama
1sg
namban
dat
tay-ɲan
see-nfut
/
 
kratək-nan
hit-nfut
He saw/hit me.

There is a single oblique postposition in the language kandək, and it is used to mark all noun phrases functioning in typically oblique semantic roles, locative, source, goal, instrument and accompaniment (cf. also 13, 14 above):

(18)
a.
tam
dog
mən
that
yangwara
tree
kandək
obl
kratək-nan
hit-nfut
He hit the dog with a stick.
b.
aykum
woman
kundamwin
two
irəpm
coconut.palm
kandək
obl
wapat-ɲan
climb-nfut
Two women climbed up a coconut palm.
c.
kumbut
village
kandək
obl
awt-mbi
get-dep
pambaysa-mbi
carry-dep
ya-mbi
come-dep
They got it and brought it to the village.
d.
mən
3sg
naŋga
where
kandək
obl
skul
school
kandək
obl
takat-n
make-fut
anak
fut
Where will he teach?
e.
kra-mbi,
cut-dep
Kayan
Kayan
kandək
obl
Paput
Paput
kandək
obl
api-ɲan
put.inside-nfut
Having cut it up, they put it inside with Kayan and Paput.

The postposition kandək can also be used with temporal words, but here it is not obligatory and this could be a calquing from Tok Pisin:

(19)
a.
mambi
then
man
other
tim
sun
kandək
obl
wa-nan
go-nfut
Then he went on another day.
b.
praŋg-mbi,
sleep-dep
man
other
tim
sun
tumbundut
morning
mambi
then
mariak-nan
say-nfut
Having slept, the next morning he said.

The one exception to this general oblique usage of kandək concerns purpose. Purpose clauses are marked with the aforementioned dative postposition namban, and this is extended to noun phrases whose semantics is specifically purpose rather than just locative goal; compare (18c) with (20c):

(20)
a.
tupwi
sago
pan
pound.sago
namban
dat
wa-nan
go-nfut
They went in order to pound sago.
b.
ama
1sg
wa
go
namban
dat
mariak
say-dep
ta-nan
impfv-nfut
I was talking to them so that they'd go. = I was telling them to go.
c.
awt-mbi
get-dep
pambaysa-mbi
carry-dep
ya-mbi
come-dep
krasaki-mbi
dry.in.sun-dep
pati
party
mən
that
namban
dat
Having gotten it and brought it, they dried it in the sun for the party.

As mentioned above, zero anaphora for noun phrases with established referents is absolutely rife in Yimas-Arafundi Pidgin, so a clause can occur with no overt core arguments, just a verb or sequence of verbs in a serial verb construction, and if noun phrase arguments are present, the vast majority of clauses contain no more than one, which is typically new information or focal. This is in keeping with the general information structure strategies among Papuan languages, and voice alternations like passive or antipassive are lacking. Examples (18c) and (20c) above are typical, as is the following longer textual example, which illustrates the context in which such extensive zero anaphora operates:

(21)
a.
kəpan
big
wakən
snake
mən
that
nuŋgum
people
mwinja
people
kumbut
village
asəŋ
pl
tu-mbi,
kill-dep
mambi
then
kra-mbi
cut-dep
am-nan
eat-nfut
yaŋi
clay.pot
kandək.
cut-obl
The people of the village killed a big snake, then cut it up and ate it from a clay pot.
b.
yaŋi
clay.pot
kandək
obl
kra-mbi
cut-dep
am-(m)bi
eat-dep
tandaw-nan
sit.down-nfut
They cut it up and sat down eating it from a clay pot.

8. Complex sentences

Unlike Yimas itself, which uncharacteristically for a Papuan language leans toward subordination as its main clause linkage device, Yimas-Arafundi Pidgin is more typical of the region and like Arafundi mainly uses the prototypical Papuan interclausal connective strategy of clause chaining, although subordination through a process of nominalization is still used as well. Nominalizations are always non-finite and are marked with the suffix -kŋ ~ -kən ~ -kəkŋ; the conditions which determine the allomorphy of this suffix are not determinable from the restricted corpus available. Nominalizations have two uses. First, they are employed for non-finite subordinate adverbial clauses. These have two functions in Yimas-Arafundi Pidgin: either to express temporal clauses whose described event is simultaneous with that of the main clause:

(22)
a.
pəpənəŋ
lake
kandək
obl
kuriaŋ-kəkŋ,
walk-nmlz
kapawri,
food
yangwara,
tree
kamban,
fish
mən
that
kamda
only
awkura-mbi
gather-dep
While he was walking around the lake, he gathered food, wood, fish.
b.
mambi
then
kra-kŋ
cut-nmlz
mən,
3sg
mambi
then
kapuk
noise
kwarisa-nan
get.up-nfut
While he was cutting it, a noise came up.
c.
mambi
then
paymban
eagle
mən
that
kapa
1du
awkura-mbi
gather-dep
mən-(n)akən
3sg-poss
manan
cult.house
kandək
obl
api-kŋ
put.inside-nmlz
mambi
then
kapa
1du
apanda-nak
shoot-imp.tr
When the eagle, having collected us, puts us in his cult house, let us shoot him.

Or for conditional or counterfactual events:

(23)
mambi
then
nuŋgum
people
mwinja
people
kumbut
village
asəŋ
pl
aka
neg
pait
fight
takari-kŋ,
make-nmlz
mambi
then
andi
land
kakan
neg
If the villagers hadn't fought, then they'd have no land.

The second use of nominalizations is with embedded relative clauses. These are extremely rare in the corpus, but the very few examples illustrate relativization on object noun phrases (24a), subject noun phrases (24b) or with unspecified indefinite subjects (24c, d):

(24)
a.
mambi
then
yangwara
tree
kra-kŋ
cut-nmlz
that
tay-ɲan
see-nfut
Then he looked at the tree he had cut.
b.
man
other
payum
man
aka
neg
aykum
woman
tari-kən
hold-nmlz
mambi
then
ya-nan
come-nfut
Then another man who didn't have a wife came.
c.
am-kəkŋ
eat-nmlz
kapawri
food
Food to eat.
d.
mən
3sg
arəŋ
tomorrow
apanda-n
shoot-fut
anak,
fut
mari-kŋ
die-nmlz
wari
manner
takari-n
make-fut
anak
fut
Tomorrow we will shoot him, we will cause his death.

As mentioned above, clause chaining is the main clause combining mechanism illustrated in the corpus of texts collected. Clause chaining is a phenomenon whereby languages distinguish between two types of clauses, independent and dependent, in which the former are characterized by fully inflected verbs, in particular for tense-aspect-mood, while the latter contain morphologically simpler, stripped down verbs. Dependent clauses coordinate with independent clauses and normally precede them, and many of their semantic features are determined by the following independent clause. Many of the previous examples illustrated clause chaining, and in Yimas-Arafundi Pidgin, the verbs of dependent clauses are marked with ‑mbi [DEP], while those of independent clauses take the tense or mood suffixes discussed in §6. Clauses in clause chaining constructions express sequential events and the ordering is iconic: Events located earlier in time are expressed in dependent clauses linearly preceding those expressing subsequent events. Conjunctions are not necessary, but the language does make heavy use of mambi ‘then’ and, to a lesser extent, mənda ‘and then’. The following examples of clause chaining in Yimas-Arafundi Pidgin are quite typical of narrative texts (for more examples look at the glossed text presented below):

(25)
a.
mambi
then
api-mbi,
put.inside-dep
mambi
then
wambaki-ɲan
throw-nfut
yəm
water
kandək
obl
They put them inside and then threw them into the water.
b.
wark-mbi
seal-dep
mənda
and.then
wambaki-ɲan
throw-nfut
They sealed it and then threw it away.
c.
pəpənəŋ
lake
kandək
obl
mambi
then
kuria-mbi,
walk-dep
way-mbi
turn-dep
wa-nan
go-nfut
kumbut
village
kandək
obl
Having walked to the lake, they went back to the village.

Clause chaining constructions seem to be able to be formed with few syntactic restrictions; in particular, it is not necessary that the clauses share subjects:

(26)
a.
mambi
then
awt-mbi
get-dep
yəm
water
kandək
obl
wambaki-mbi
throw-dep
mambi
then
yəm
water
mənda
and.then
tandaw-mbi
stay-dep
ta-nan
impfv-nfut
Then he got them and threw them into the water, and the water was still there.
b.
awt-mbi
get-dep
asa-mbi,
give-dep
yəm
water
mən
that
am-(m)bi,
drink-dep
mənda
and.then
wa-nan
go-nfut
They got and gave it to him, and he drank the water and then he left.

Serial verb constructions at first look indistinguishable from clause chaining constructions because non-final verbs in serial verb constructions also occur with the suffix -mbi, and the last one, if sentence final, takes the tense-mood suffixes. However, there are three properties that enable us to sort true serial verb constructions from clause chaining constructions. First the order of verbs in serial verb constructions need not be and commonly is not iconic, in clear contrast to clause chaining constructions:

(27)
a.
yaŋi
clay.pot
kandə
obl
am-(m)bi
eat-dep
tandaw-nan
sit-nfut
They were sitting down eating it from a clay pot.
b.
pəpənəŋ
lake
kandək
obl
kuria-mbi
walk-dep
tay-ɲan
see-nfut
He walked around the lake searching. (the walking and looking are simultaneous)

Second, serial verb constructions select their arguments and adjuncts as a unit, so that any sequence of verbs in which an argument or adjunct can only be selected by non-initial verb in the series must be a serial verb construction:

(28)
kumbut
village
kandək
obl
awt-mbi
get-dep
pambaysa-mbi
carry-dep
ya-mbi
come-dep
They got it and brought it to the village.

In this example the goal oblique noun phrase is semantically selected by the final verb only, ya- ‘come’, indicating that the whole sequence is a unit which determines its arguments and adjuncts. In particular, the verb closest to the oblique noun phrase awt- ‘get’ cannot select for this adjunct semantically, as the getting is done in a place which is not the village. A structure like (28) is impossible in a clause chaining construction, for in these any arguments or adjuncts must be in the same clause as their governing verb, i.e.:

(29) awt-mbi, kumbut kandək pambaysa-mbi ya-nan

get-DEP village OBL carry-DEP come-NONFUT

‘They got it, and brought it to the village.’

Finally, prosodic phenomena distinguish serial verb constructions from clause chaining constructions. Verbs in dependent clauses in clause chaining constructions are marked by a distinctive falling pitch (orthographically represented by a comma in the examples above), not as low and not over such a large pitch range as final verbs in independent clauses, but quite distinct nonetheless. Verbs within serial verb constructions have no such fall in pitch: All are spoken with roughly the same pitch except the last, which takes on the distinctive falling pitch of the end of a sentence.