Sarnami: the Largest Language in Caribbean Hindustani


Out of India: Language contact and change in
Sarnami (Caribbean Hindustani)
Kofi Yakpo (The University of Hong Kong)
1 Introduction
Sarnami, spoken in Suriname, is the only variety of Caribbean Hindustani that
still has a sizeable speaker community. Sarnami is the result of the koineization
of several northern Indian languages during Dutch colonial rule. A comparison
of Sarnami with its closest Indian relatives suggests that contact with Sranan
and Dutch has led to syntactic change, with an inherited head-final order giving
way to head-initial order. SVO is far more frequent in Sarnami than in the Indian
control group. In relative constructions and with certain types of modal and
aspectual auxiliary constructions, the transition has been made to NRel (postposed relative clauses) and AuxV (auxiliary-verb order). However, diachronically
more stable constituent orders like noun vs. adjective, noun vs. adposition and
noun vs. genitive have not been affected by change. Constituent order change in
Sarnami is an example of the kind of convergence that characterizes Suriname
as a linguistic area, where the two dominant languages Sranan and Dutch simultaneously exert pressure toward typological change.1 Contact-induced change
in constituent order has been reported for Sarnami (Damsteegt 1988) and other
Indian diaspora languages like Fiji Hindustani (Siegel 1985, 1987) and South
African Bhojpuri (Mesthrie 1991), but this is the first study to corroborate claims
with statistical evidence from a primary corpus.
In this chapter, I will look at basic word order, relative constructions and
auxiliary constructions and show that Sarnami has undergone substantial
change away from head-final order, typical of Indo-Aryan, towards head-initial
order. The speaker community of Sarnami is characterized by extensive trilingualism in Sarnami, Sranantongo and Dutch, and I will claim that contact with
Dutch and Sranantongo (henceforth Sranan) is responsible for these changes.2
1 I am indebted to Lila Gobardhan-Rambocus, without whose support and advice the research
on which thisstudy relies could not have been carried out. I am equally grateful to Motilal Marhé
for his transcription of the Sarnami, Bhojpuri, Maithili and Magahi corpus. The linguistic annotation of the corpus was done by me with the SIL software Toolbox. I also wish to thank Jeff Siegel
for his valuable comments on a first draft of this chapter.
2 In the 2004 national census about 27% of the total Surinamese population of half a million
self-identifies as “Hindoestaans” (Indian-descended) (SIC 213-2005/02).
130 Kofi Yakpo
Sarnami is actively used by all generations within the Indo-Surinamese community in a pattern of trilingualism including Sranan and Dutch. Sarnami is
the only Indian diaspora language of the Caribbean with roots in the colonial
period that is still vital. The two other languages Guyanese Bhojpuri (Gambhir
1981) and Trinidad Bhojpuri (Mohan 1978; 1984; 1990; Mohan and Zador 1986)
are on the verge of extinction, with only few active speakers left in both countries. The socio-history and many linguistic aspects of the evolution of Sarnami
and the other Caribbean Hindustani varieties also parallel that of three other
Indian diaspora languages outside the Caribbean, namely South African Bhojpuri (Mesthrie 1991), Mauritian Bhojpuri (Baker and Rahnah 1985; 1988) and
Fiji Hindustani (Siegel 1985, 1987). Barz and Siegel (1988) contains the first and
to date only comprehensive and comparative overview of the Indian diaspora
Sarnami, Guyanese Bhojpuri and Trinidad Bhojpuri are collectively referred
to as Caribbean Hindustani in Ethnologue (Lewis et al. 2016). Although Sarnami
owes very little of its linguistic system to Hindustani (i.e. the Hindi-Urdu continuum), the language is indeed referred to by the majority of its speakers by the
Dutch term “Hindoestaans”. Sarnami, literally “Surinamese”, a term coined by
Surinamese intellectuals in the 1960s to reflect the local rootedness of the language, has gained a wider currency in recent decades and is the common designation in academic works and creative literature.
I provide a brief background to the koineization of Sarnami in section 2. In
section 3, I describe the data sources. In section 4, the main part of this chapter,
I look at contact-induced variation and change in Sarnami constituent order, first
turning to basic word order in 4.1, then to relative constructions in 4.2 and to
auxiliary constructions in 4.3. Section 4.4 summarizes the findings and section 5
places them within the broader typological context of contact-induced change in
constituent order.
2 Sarnami as a koiné
Sarnami is the result of the koineization in Suriname of several closely related
languages spoken in the present-day Indian federal states of Uttar Pradesh, Bihar
and Jharkhand (Marhé 1985: 8–13; Damsteegt 2002). In this section, I show how
koineization manifests itself in the verbal system. Damsteegt (1988) provides an
overview of other linguistic features of koineization in Sarnami.
The languages that were to merge and become Sarnami were transplanted to Suriname during the indentured labor trade of the nineteenth century,
Out of India: Language contact and change in Sarnami (Caribbean Hindustani) 131
during which hundreds of thousands of people were shipped from India to the
Caribbean by the European colonial powers in order to substitute for the labor
of enslaved Africans after the official end of slavery. It is unlikely that the koineization process through which Sarnami emerged involved substantial influence from Sranan and Dutch from the very outset. The widespread acquisition of
Sranan by Indian-descended Surinamese probably only began in the course of
socio-economic transformations in the first half of the twentieth century (Yakpo
2015; for the socio-economic background, see Dusseldorp 1963; Hassankhan,
Ligeon and Scheepers 1995; Boer 1998; Hira 1998). The large-scale acquisition
of Dutch began even later, accelerating in particular from the mid-20th century
onwards through the expansion of the educational system and circular migration between Suriname and the Netherlands (cf. de Kleine 2007: 33). I therefore
assume that the changes described in this chapter began to take hold when
Sarnami had already consolidated itself as a fairly stable system. In this section,
I focus on aspects of koineization that Sarnami underwent before the changes
induced by contact with Sranan and Dutch began. The latter are covered in the
subsequent sections.
Sarnami is the result of language contact processes in a situation of diaspora
(Damsteegt 1988; 2002). These are koineization, influence from the colonial language Dutch, and influence from the Surinamese national vernacular Sranan.
The combination of these three processes has yielded a unique new language.
Structural and lexical features seem to indicate that Sarnami is the result of
the mixing of various languages. The three languages considered to have had
the largest input into Sarnami are Bhojpuri, Magahi and Maithili spoken in the
present-day Indian federal states of Uttar Pradesh, Bihar, Jharkhand and West
Bengal. These three languages are classified as languages in their own right of
the Bihari group of Indo-Aryan (Tiwari 1960; Jeffers 1976; Masica 1993: 12–13).
Grierson (1901) however classifies them as varieties of Hindi, the two positions
reflecting the continuum character of north India as a linguistic area (Jha 1994;
Abbi 1997). The grammar and lexicon of Sarnami also reflect the influence of
varieties of Eastern Hindi, in particular Awadhi. Due to their mixed nature, I will
henceforth collectively refer to the languages that emerged during the indentured labor trade as Overseas Indic (rather than overseas “Hindi”, cf. Bartz and
Siegel 1988).
Sarnami showsthe characteristic effects of koineization that have been documented in the literature, namely mixing, leveling,simplification and reallocation
(Kerswill 2002). I will provide an example of the processes of mixing and leveling
by looking at the formation of past tense in Sarnami.
Tab. 1 below features verbal suffixes that serve to express past tense in
Sarnami, Bhojpuri, Awadhi, Maithili and Magahi (Yakpo and Muysken 2014).
132 Kofi Yakpo
I provide two varieties of Bhojpuri in order to exemplify the degree of intra-lectal
variation in one of these languages (and the possibility of mutual borrowing by
the contributing languages): The southern standard variety, described by Tiwari
(1960) and Sadani Bhojpuri, spoken in southern Jharkhand (Jordan-Horstmann
1969). The table only contains non-honorific suffixes. Variants are separated by
a comma, feminine gender forms are provided in parentheses where they exist:
Tab. 1: Past tense suffixes in Sarnami and North Indian languages. (Taken from Yakpo &
Muysken 2014)
1sg -li, -lin
1pl -li, -lin
2sg -le
2pl -le
3sg -l, -is
3pl -l, -is, -lẽ
-lā, (-liu)
-lā, (-liu)
-l, (-li)
-lẽ, (-lini)
-le, -lẽ
-le, -lẽ
-l, -lək
-l, -lək
-l, -lak
-l, -lak
(Sources: Saksena 1971 for Awadhi; Tiwari 1960 and Shukla 1981 for Bhojpuri; JordanHorstmann 1969 for Sadani Bhojpuri; Yadav 1996 for Maithili; Verma 1985 for Magahi)
Table 1 shows that the Sarnami past tense suffixes have multiple sources and are
of mixed origin. Leveling has also taken place in Sarnami: Specific forms have
been picked out while others have not survived the koineization process. The
following observations can be made regarding Tab. 1:
– The 1sg/pl forms are found in all potential contributing languages except
Sadani Bhojpuri and Awadhi (with minor adaptations in Sarnami,such as an
optional final nasal instead of nasalization);
– The 2sg/pl forms are found in Maithili, while the /l/ consonant isfound in all
contributing languages except Awadhi;
– The /-l/ variant of 3sg is found in all contributing languages except Sadani
Bhojpuri and Awadhi, while the /-is/ variant is unmistakably of Awadhi origin;
– The /-lẽ/ variant of 3pl is found in Sarnami and Southern Bhojpuri alone.
Southern Bhojpuri features gendered suffixes in the second and third persons
and separate suffixes for plural number. Some varieties of Awadhi also show
gendered verb suffixation in intransitive clauses (Saksena 1971: 249). However,
the absence of a gender distinction in the verbal morphology of Sarnami may
Out of India: Language contact and change in Sarnami (Caribbean Hindustani) 133
not be the result of simplification. The use of feminine gender has been optional
in Bhojpuri for a long period of time (Gambhir 1981: 249) and the distinction
is not strict in Awadhi either (cf. Saksena 1971: 249, 139). The disappearance of
feminine gender verbal suffixes in Sarnami might therefore simply reflect the
end-point of a development foreshadowed in Bhojpuri and other languages to
the east of the Hindi-Urdu heartlands.
As shown by Damsteegt (1988; 2002) koineization is also characteristic of
other sub-systems of Sarnami. Language contact is therefore intrinsic to the linguistic system of Sarnami.
3 The data
The analyses of constituent order in the following sections are based on a Sarnami
corpus of about 30,000 words of primary data collected in Suriname in 2011–14.
A corpus of about 15,000 words of control data of Bhojpuri, Magahi and Maithili was gathered in India in 2011. The analyses of basic word order frequencies
(see Tab. 2) are based on recordings of the “Frog Story” (Mayer 1969). The comparison of frequencies of constituent order in relative and auxiliary constructions
in section 4.2 (see Tab. 3) and section 4.3 (see Tab. 4) is based on data elicited
via “staged events” (see Gippert, Himmelmann and Mosel 2006), i.e. short video
clips depicting scenes that linguistic informants were invited to describe. Due to
the nature of the narrative task, “Frog Stories” contained sufficient instances of
basic word order, but insufficient instances of relative and auxiliary constructions, hence the different sources.
The texts represent an even spread of speakers in terms of age, gender,socioeconomic background and geographic origin. The sub-corpus on basic word
order for example (Tab. 2) represents the speech of eight Sarnami speakers in
the age brackets of 15–20 (sar11rev-fe), 21–30 (sar11joe-fe sar11mal2-fe), 31–40
(sar11aar-fe), 41–50 (sar11mal-fe, sar11ram-fe, sar11moe-fe), 51+ (sar11hel-fe). The
sample contains speakers from middle-school (sar11rev-fe), blue collar (sar11malfe, sar11joe-fe), non-university educated white collar (sar11mal2-fe, sar11moe-fe,
sar11hel-fe), and university educated white collar (sar11aar-fe, sar11ram-fe)
occupational backgrounds. The Sarnami sample’s geographical spread ranges
from the Surinamese capital Paramaribo (sar11mal-fe, S2), the district of Wanica
(sar11aar-fe, sar11ram-fe) and Commewijne (sar11moe-fe), to Nickerie (sar11mal2-
fe, sar11rev-fe). The sample consists of three male (sar11mal2-fe, sar11ram-fe,
sar11moe-fe) and five female speakers (sar11mal-fe, sar11mal2-fe, sar11aar-fe,
sar11rev-fe, sar11hel-fe).
134 Kofi Yakpo
The informants in the Indian control group are socially more homogenous,
being university students between 21 and 25 years. There is however, a good
geographical spread encompassing speakers from rural and urban districts of
the Indian federal states of Uttar Pradesh (bho11nee-fe), Bihar (bho11nej-fe,
bho11vai-fe, mat11raj-fe, mat11dep-fe, mat11ata-fe), and Jharkhand (bho11sukfe, mat11arv-fe). Five speakers are male (bho11nej-fe, bho11suk-fe, mat11rajfe, mat11ata-fe, mat11raj-fe) and three female (bho11nee-fe, bho11vai-fe,
mat11dep-fe). The abbreviations used in text names are as follows: sar11mal-fe
= sar [language, i.e. Sarnami]; 11 [recorded in 2011]; mal [name of informant]
-fe [frog story, elicited]. The final two letters ‘-ke’ indicate that the data is
elicited by means of the ‘elicitation kit’ hence the collection of video clips
already referred to.
4 Contact-induced variation and change in
constituent order
The data suggests that contact with Sranan and Dutch has led to a reconfiguration of head-dependent order in some domains. Sarnami shows a far greater
frequency of SVO basic word order than its closest relatives in India (cf. 4.1).
Likewise, Sarnami shows a far greater tendency than the control group towards
NRel (postposed relative clauses) (cf. 4.2) and AuxV (preposed auxiliaries) in
auxiliary constructions (cf. 4.3). In the following sections, I look at the frequency distribution of constituent order in a sub-corpus of Sarnami and its
Indian relatives Bhojpuri, Magahi, and Maithili. I first turn to basic (clausal)
word order.
4.1 From SOV to SVO
Contact with Sranan and Dutch appears to be responsible for an ongoing change
in Sarnami basic word order. The default SOV word order that Sarnami has inherited from its Indo-Aryan contributing languages is ceding to SVO word order in
main clauses. At this point, SOV is still the dominant word order in pragmatically
neutral main clauses. The data however indicates that SVO order has made significant inroads into the language.
Tab. 2 below presents the absolute frequencies (given as a number) and
relative frequencies (expressed as a percentage over the total number of
Out of India: Language contact and change in Sarnami (Caribbean Hindustani) 135
transitive clauses featuring overt objects) of SVO in Sarnami and in the two
Indian control group languages for which “Frog Stories” were recorded
(Bhojpuri and Maithili). Total absolute and relative frequencies are provided in
the final row. Tab. 2 only lists clauses that can be considered prototypically transitive. The table therefore neither includes clauses involving Goal objects of the
verb já ‘go’, nor inherent complements of conjunct verbs, nor object interrogative pronouns. However, nominal and pronominal objects are both included in
the count. I have also excluded clauses that appear to feature SVO order, but in
which the transitive object is separated from the rest of the clause by a pause.
Such occurrences of a surface SVO order constitute instances of afterthought
topicalization of the object.
Tab. 2: Frequencies of SVO compared.
SVO/Total N words
4/9 (44% ) 434
7/16 (44%) 778
5/12 (42%) 544
9/22 (41%) 569
9/23 (39%) 601
10/32 (31%) 820
5/18 (28%) 563
5/18 (28%) 686
54/150 (36%) 4995
Bhojpuri/ Maithili
SVO/Total N words
1/13 (8%) 599
1/37 (3%) 753
1/36 (3%) 983
3/31 (10%) 621
3/26 (12%) 790
0/23 (0%) 503
0/24 (0%) 432
0/12 (0%) 377
9/202 (4%) 5058
Two-tailed p-value (Fisher’s exact test): p <.0001
The percentages in parentheses reveal significant differences in the frequencies
of SVO between Sarnami on the one hand, and Bhojpuri and Maithili on the other.
The lowest percentages of SVO in individual frog stories of the Sarnami corpus(cf.
sar11moe-fe and sar11hel-fe) are still more than twice as high as the highest percentage in the Indian sample (mat11raj-fe), i.e. 28% vs. 12%. The total percentage
of SVO in Sarnami (36%) is nine times higher than that of Bhojpuri and Maithili
(4%). Further, there is no Sarnami text with no occurrence of SVO at all. In contrast, the Indian sub-corpus contains three texts with 0% occurrences of SVO.
Fisher’s exact test renders a p-value well below a significance level of 0.05, hence
showing a significant difference in the frequency of SVO between Sarnami and
Bhojpuri/Maithili. In what follows, I interpret the findings in Tab. 2 in the light of
the word order typology of Indo-Aryan, Dutch and Sranan.
136 Kofi Yakpo
Pragmatically unmarked word order of Sarnami basic clauses is generally
Subject – Object – Verb (SOV) (cf. also Marhé 1985: 26), as in (1) (O is set in bold in
the examples in this section).3
(1) ego manai ego dosu
a person a box
lá-il hai
bring-pst.3 be.prs
‘A person has brought a box.’ (Sarnami)
Sarnami shares SOV basic word order with other Bihari languages and other
branches of Indo-Aryan. Sarnami’s close relatives in India, Maithili, Magahi and
Bhojpuri feature SOV word order, cf. (2), and so does Hindi, cf. (3).
(2) kuttaa bhii beng
dog also frog
‘The dog also began to look for the frog.’ (Bhojpuri)
(3) pita jī
father hon
pəɽʰ rəhe hɛ̃.
‘Father is reading the newspaper.’ (Hindi)
(Kachru 2006: 251)
Word order in many Indo-Aryan languages nevertheless variesin accordance with
syntactic and pragmatic factors. In Hindi, nominal constituents that immediately precede the finite verb are under new information focus (Kachru 2006: 251).
Conversely, word orders that diverge from the SOV basic pattern are exploited for
the expression of contrastive focus.
(4) Mohən
ne de
erg ‘give’
book.f pl
Shyam acc/dat
‘Mohan has given his books to Shyam.’ (Hindi)
(Kachru 2006: 126)
In the Hindi example in (4), SVO word order is therefore employed for contrastive
focus of the object (Kachru 2006: 159). In Sarnami, SVO word order, asin (5), enjoys a
much higherfrequency than in the Indian control group,i.e. 36%, versus 4% in India.
3 Sarnami has a standard orthography with the following conventions: Retroflex consonantsfeature an underscore, hence ṟ[ ɽ ], ḏ[ ɖ ]. A nasalized vowel is rendered by a following ṉas in meṉ
[mẽ] ‘in’. Long vowels bear an acute accent, hence á[aː]. With Indian languages, I follow the established convention of capitalizing retroflex consonants (e.g. R, D), representing phonemic nasalization of vowels with a following capitalized N, and long vowels by vowel doubling (e.g. aa [aː]).
Out of India: Language contact and change in Sarnami (Caribbean Hindustani) 137
(5) tab u
then 3sg
ego hol jamin meṉ.
one hole ground in
‘Then he saw a hole in the ground.’ (Sarnami)
The high frequency of SVO in Sarnami is subject to little variation between speakers (see Tab. 2), and it is also higher than could be attributed to the pragmatic
function of focus alone. SVO therefore seems to be competing with SOV as an
unmarked basic word order. I conclude that the high incidence of SVO in Sarnami
compared to Bhojpuri/Maithili is due to contact with Sranan and Dutch. In
Sranan, SVO is the only possible word order in basic (and in complex) clauses,
compare (5) above and (6) below.
(6) dan a boi si wan olo.
then the boy see a hole
‘Then the boy saw a hole.’ (Sranan)
Dutch (as spoken in Suriname and in The Netherlands), features SVO word order
in basic clauses, cf. (7).
(7) het jongetje
the boy
ziet een gat
sees a hole
in een boom.
in a tree
‘The boy sees a hole in a tree.’ (Dutch)
However we also find head-final order in basic clauses featuring periphrastic
tenses and moodsin Dutch. Compare example (8), with its S-AUX-O-V word order,
where the transitive object follows the inflected auxiliary verb (heeft ‘has’) but
precedes the lexical verb (gezien ‘seen’, a participial form):
(8) het jongetje heeft een gat
the boy has a hole
in een boom
in a tree
‘The boy has seen a hole in a tree.’ (Dutch)
Constructions with an S-AUX-O-V order as in (8) can be seen as somewhat intermediate between SOV and SVO (cf. Gensler 1994). They are therefore not typical
exponents of SOV. Nevertheless, many types of Dutch subordinate clauses are
clearly characterized by SOV order, as exemplified by (9):
(9) ik hou deze vast
I hold this tight
tot ik die ball
until I that ball
‘I’ll hold on to this until I get back that ball.’ (Dutch)
The relatively few complex clauses in the corpus of Sarnami frog stories feature
SOV order as well. It is possible that the existence of SOV order in Dutch subordinate clauses has contributed to the retention of SOV in Sarnami subordinate
clauses. This assumption would have to be verified by further analyses.
138 Kofi Yakpo
The hypothesisthat language contact isresponsible forthe change in Sarnami
basic word order is supported by evidence from other Overseas Indic varieties.
Mesthrie (1991: 183–184) reports that South African Bhojpuri is characterized by
a higher-than-usual frequency of SVO clauses due to extensive contact with, and
language shift to English, and Siegel says the same for Fiji Hindustani (Siegel
1987: 144), although these two sources do not provide statistical evidence. Main
clauses are not the only type of structure affected by change in constituent order.
In the following section, I show that Sarnami relative constructionsshow an even
greater tendency to diverge from inherited patterns.
4.2 Relative constructions
In this section, I show that Sarnami has a preference for postposed relative
clauses (NRel) rather than preposed correlative ones (RelN). Preposed correlative
constructions are preferred by Sarnami’sIndian relatives. I take this preference of
Sarnami to be a manifestation of contact-induced change. Tab. 3 below compares
the frequency of postposed relative clauses in Sarnami and its Indian relatives
Bhojpuri, Maithili, and Magahi. The Sarnami corpusshows an overwhelming preference of 72% for postposed relative clauses. At the extreme ends we find texts
with a low incidence of relative clauses (e.g. 100% for sar10sha-ke, with a single
relative clause, and 50% for sar11paramuru, with a total of two relative clauses).
However, texts with a higher incidence of relative clauses all show a dominance
of postposed relative clauses (e.g. 67% for sar10mot-ke and 65% sar11sit_sha-ye).
One text with a relatively high incidence of relative clauses (sar11aar-ke) does not
Tab. 3: Frequencies of NRel (postposed relative clauses) compared.
1/1 (100% )
4/5 (80%)
1/1 (100%)
6/6 (100%)
2/3 (67%)
6/9 (67%)
1/2 (50%)
5/18 (65%)
15/23 (72%)
N words Bhojpuri/
1756 bho11nee-ke
2358 bho11nej-ke
1667 bho11suk-ke
1848 bho11vai-ke
643 mat11raj-ke
2853 mat11dep-ke
1786 mat11arv-ke
1309 mag11gag-ke
14220 Total
NRel/ N words
5/12 (42%) 1035
2/9 (22%) 753
1/1 (100%) 983
0/2 (0%) 621
2/5 (40%) 1324
3/9 (33%) 790
6/16 (38%) 4897
2/7 (29%) 2324
21/61 (34%) 12727
Two-tailed p-value (Fisher’s exact test): p = .0001
Out of India: Language contact and change in Sarnami (Caribbean Hindustani) 139
contain a single instance of a preposed relative clause. On the Indian side, we find
a clear preference for structures involving preposed (cor-)relatives. Leaving aside
the extreme cases (100% of postposed relative clauses for bho11suk-ke and 0% for
bho11vai-ke), we see that texts with a reasonably high incidence of relative clauses
all contain somewhere between a quarter and a third of post-posed relative clauses.
The findings in Tab. 3 show a significant shift of Sarnami away from its IndoAryan relatives. The typological and contact-related context of this development
is the following: in Sarnami’s next-of-kin in India correlative constructions are
the default structures. In these constructions, the relative clause is introduced by
a relativizer, while the main clause is introduced by a correlativizer. The correlativizer is either a specialized function word, or a deictic/referential element,such
as a demonstrative adjective or pronoun or a resumptive pronoun co-referential
with the head noun. Typically, correlative structures feature preposed relative
clauses, i.e. the relative clause precedes the main clause. The following example
from Maithili shows such a correlative construction. It features a left-adjoined
relative clause introduced by the morphologically invariant relativizer je, and a
following main clause introduced by the equally invariant correlativizer se.
(10) [je
r e l
kailh rait
yesterday night
utəl əich.
asleep be.prs
se u nəṯua
corel dist dancer
‘The dancer who danced yesterday night is now asleep.’
(Maithili; Yadav 1996: 356)
The simultaneous use of a relativizer and a correlativizer and the resulting explicit marking of the subordinate and main clauses provides flexibility in the ordering of the two clauses in the Indo-Aryan languages (cf. Kachru 2006: 4). Apart
from preposed relative clauses, we therefore also encounter postposed relative
clauses, among other possibilities. Compare (11) from Maithili.
(11) u nəṯua
dist dancer
ekhən utəl
now asleep
[je rait
r e l night
nac-əl] se
dance-pfv corel
‘The dancer who danced last night is now asleep.’
(Maithili; Yadav 1996: 352)
The Sarnami corpus also features pre- and postposed relative clauses. The
Sarnami sentence in (12) below involves a correlative structure, in which the
object is relativized. The relative clause is preposed and introduced by the same
form as in Maithili above, namely the morphologically invariant relativizer je.
140 Kofi Yakpo
The following main clause contains the resumptive pronoun oke, which functions as a correlativizer, and is made up of the distal/3sg pronoun o and the
acc/dat marker ke:
(12) [je zondig
r e l sinful
vergeven kare
forgive ‘do’
kar-is hai]
‘do’-pst.3 be.prs
‘God forgives him/her who has sinned.’ (Sarnami)
Lit. ‘Who has sinned, God forgives him/her.’
The following example shows another way of forming a correlative construction
in Sarnami. Once again the relative clause is preposed. But this time the relative
clause is introduced by the (morphologically invariant) relativizer jaun:
(13) aur
[jaun balwá
r e l ball
nicwán rahá]
below be.pst
em .prox
kisi-yá meṉ bíg
box-def in throw
‘And he took the ball that was underneath and threw it in the box.’
Lit. ‘‘And the ball that was underneath, that very one he took and
this very one he threw into the box.’ (Sarnami)
In example (13) above, the head noun is contained in the relative clause and
picked up by correlative/resumptive elements in the main clause – the emphatic
distal and proximate pronouns ohu and ohi respectively. This type of relative construction, in which the head noun is overtly expressed in the subordinate relative
clause, and reiterated through a resumptive element in the main clause is also
common in other Indo-Aryan languages (cf. Hindi, Kachru 2006: 220–221). The
alternative way of relative clause formation is shown in (10) above, where the
head noun remains unexpressed in the relative clause, and is instead realized
in the main clause. For the sake of brevity, I will not spell out the often intricate arguments presented with respect to headedness of structures like (10)–(13)
(see e.g. the discussion and references in Cinque 2009). Therefore, I will continue using the term “preposed” when referring to structures in which the relative
clause precedesthe main clause and “postposed” when the main clause precedes
the relative clause. Preposed relative clauses are, of course, a typological correlate of SOV basic word order, while postposed ones are a typological characteristic of SVO languages (see e.g. Dryer 2013).
Going back to the Sarnami corpus, the most common type of relative construction encountered involves postposed relative clauses. Contrary to the correlative
Out of India: Language contact and change in Sarnami (Caribbean Hindustani) 141
structures covered above, such structures are also clearly head-initial: The head
noun is contained in the main clause, and both the head noun and the main clause
precede the relative clause. The relative clause is introduced by the invariant relativizer jaun and there is no correlativizer. Compare (14), involving a relativized prepositional phrase.
(14) u
ego dosu
a box
hoi sake]
become can
hai [jaun pe
be.prs r e l on
‘He has brought a box that he can stand on.’ (Sarnami)
Postposed relative clauses are much less frequent in Sarnami’s Indian relatives
than preposed ones. In addition, postposed relative clauses are generally seen
to be extraposed, i.e. under focus in many Indo-Aryan languages (e.g. Hindi, cf.
Kachru 2006: 221). The Sarnami data does not, however,suggest that these postposed relative clauses are pragmatically marked in any particular way. Relative
constructions involving a single relativizer or relative pronoun, a head noun in
the main clause and a postposed relative clause also represent the most neutral
type of structure in Sranan and Dutch; compare (15) and (16) respectively.
(15) kande
na a
tyari kon].
bring come
lespeki [san
respect r e l
a barba beard
‘Perhaps it’s the respect that the beard brings along.’ (Sranan)
(16) een
vrouw gooit een
woman throws an
bijl op de
axe on the
bord [die dan
plate r e l then
in stukken
in pieces
‘A woman throws an axe on the plate, which then breaks
into pieces.’ (Dutch)
I conclude that head-initial order in Dutch and Sranan relative constructions
has influenced constituent order in Sarnami relative constructions. The Indian
relatives of Sarnami nonetheless still show a far higher variability in the position
of relative clauses than in the position of verbal objects, indicating that there
is already much flexibility in the positioning of relative clauses in the Indian
languages. In the domain of auxiliary constructions, which I now turn to, the
divergence between Sarnami and its Indian kin is yet greater.
142 Kofi Yakpo
4.3 Auxiliary constructions
Sarnami shows an overwhelming tendency towards head-initial order (AuxV) in
auxiliary constructions. Again, Sarnami stands out from among its Indian sister
languages, which show a clear preference for dependent-head order in like constructions. In the following, I employ the term “auxiliary” in a broad sense to
include non-bound elements with differing degrees of grammaticalization that
fulfill modal, aspectual and tense-related functions(cf. Heine 1993). Tab. 4 shows
the frequencies of head-initial (AuxV) order in auxiliary constructions in Sarnami
on the one hand, and in Bhojpuri, Maithili, and Magahi on the other.
The results of the frequency count in Tab. 4 in the two (groups of) languages
provides strong evidence for a difference in constituent order in Sarnami vis-àvis its Indian relatives. Nearly 80% of Sarnami auxiliary constructions involve
head-initial (AuxV) order. In contrast, 95% of auxiliary constructionsin Bhojpuri,
Maithili, and Magahi show head-final (VAux) order. Even in the Sarnami text with
the lowest incidence (sar11poo_kam-ke) only two-thirds of auxiliary constructions show VAux order. On the Indian side, the text with the highest number of
auxiliary constructions (mat11arv-ke) only shows a comparatively low percentage
of AuxV occurrences (13%). The text with the second highest number of auxiliary constructions (mag11gag-ke) features no construction with AuxV order at
all. All other texts, save one (bho11nej-ke), feature no construction with AuxV
order either. The divergence between the Sarnami and the Indian samples is
therefore particularly strong in this domain, both with respect to the individual
texts, as well as to the total. The statistics confirm Damsteegt’s (1988: 166) claim
Tab. 4: Frequencies of AuxV order compared.
AuxV/ N words
9/9 (100%) 1756
11/15 (73%) 2358
10/10 (100%) 1667
17/19 (89%) 1848
3/3 (100%) 643
15/25 (60%) 2853
AuxV/ N
Total words
0/3 (0%) 1035
1/5 (20%) 753
0/4 (0%) 983
0/5 (0%) 621
0/5 (0%) 1324
0/6 (0%) 790
11/12 (92%)
12/19 (63%)
1786 mat11arv-ke
1698 mag11gag-ke
4/32 (13%) 4897
0/25 (0%) 2324
Total 88/112 (79%) 15326 Total 5/102 (5%) 12727
Two-tailed p-value (Fisher’s exact test): p <.0001
Out of India: Language contact and change in Sarnami (Caribbean Hindustani) 143
(not substantiated by figures) of a change in constituent order in auxiliary constructions, and Damsteegt also gives references to it in Fiji Hindustani, Guyanese
Bhojpuri and Trinidad Bhojpuri.
I continue with some more details on constituent order in auxiliary constructions in Sarnami and its Indian sister languages. The constituent order inherited
from Indo-Aryan is exemplified in the two examples below, from Sarnami and
Bhojpuri. Typical for head-final order (VAux), the auxiliary verb kos(h)is(h) kare
‘try’ in the following Sarnami (19) and Bhojpuri (18) sentences, followsthe lexical
verb (pakare ‘get hold of’ and nikale ‘come out’ respectively). “Compound verbs”
like kos(h)is(h) kare, lit. ‘effort do’, have been covered in great detail in the literature on Indo-Aryan, including Sarnami (e.g. Kishna 1979; Appel and Muysken
1987: 126–127; Muysken 2000: 197–202, 208–211):
(17) tab
ke kosis
acc/dat effort
‘Then he tried to get hold of the piece of clothing.’ (Sarnami)
(18) aur donoN
and both
aage paani meN se
front water in abl
koshish kare
effort do
‘And they both begin to try to come out of the water.’ (Bhojpuri)
As shown above in Tab. 4, both Sarnami and the Indian control group feature
head-initial order as well, only with inverse frequencies. Structures like (19)
(nearly identical to (17) above) are the norm in Sarnami (79% of occurrences), but
exceedingly rare (5%) in the Indian languages:
(19) ab kosis
now effort
kare haigá
do be.prs
‘Now he’s trying to get hold of the piece of clothing.’ (Sarnami)
The following two examples from Magahi (20) and Bhojpuri (21) respectively
show AuxV in the Indian control group. Both examples from different speakers
describe the same scene:
(20) ge-l-aii
‘He went to get something.’ (Magahi)
(21) to laik-waa phir
emp boy-def again
baa kuch
be.prs something
lei ke.
get acc/dat
‘That boy went again to get something.’ (Bhojpuri)
144 Kofi Yakpo
In the Indian languages, head-initial (AuxV) constructions like in the two sentences above are often limited to heavier structures involving an object NP, which
would otherwise be preposed to the auxiliary together with the main verb as in
(22) (object in bold). Such departuresfrom canonical constituent order can also be
stylistically and pragmatically marked as is the case with clausal word order (see
e.g. Masica 1993: 332–336):
(22) aur
baad kapaR-waa
after clothing-def
‘Then after that he is trying to make the (piece of) clothing
fall down.’ (Bhojpuri)
The observationsin thissection are limited to constructionsinvolving auxiliary verbs
that (1) retain uses aslexical verbs next to their grammatical functions as modal and
aspectual auxiliaries, and that (2) are not in an advanced stage of grammaticalization, i.e. whose lexical meanings and grammatical functions are relatively close to
each other. The observations are further limited to the four most frequent concepts
found in the corpus, contained in Tab. 5 (in descending order of textual frequency,
with the top-most verb já go (do sth.)’ being the most frequent one).
Tab. 5: Directional and modal auxiliaries.
jámáng-, cáhsakpruberi
kausis karBhojpuri
koshish karMaithili
jacahsək-, pakosis/
koshish kərMagahi
jacahsak-, paa(w)-
koshish karGloss
(do sth)’
Sarnami and the Indian languages nevertheless show canonical head-final
(VAux) order in constructions involving so-called vector verbs as well as copular
auxiliary verbs(e.g. hai in (1)) Vector verbs express diverse and often subtle aspectual nuances, and may themselves be marked for tense, mood, and aspect by
suffixation (e.g. past {-il(e)} in (23) and (24) below). However, the verb root carries
little or no lexical content, even when there is an etymological relation with a
lexical verb, compare the roots and meanings of gá- in (23) below and (21) above.
Out of India: Language contact and change in Sarnami (Caribbean Hindustani) 145
The vector verb gáil(e) (23–24) has verb focusing and passivizing functions while
paRle (24) expresses a suddenness of action:
(23) manai-yá
meṉ gir
in fall
‘The person fell in(to) the middle.’ (Sarnami)
(24) tastarii
gir paR-le
fall sudden-pst.3
au phuT
and break
‘A plate suddenly fell and got broken.’ (Maithili)
Vector verbs are highly grammaticalized and therefore functionally similar to
aspectual suffixes. They are characterized by a high textual frequency, fulfill core
roles in the TMA system of Sarnami and express abstract grammatical functions
in Sarnami and its Indian siblings. Many of these “verbs” are also subjected to a
larger-than-usual degree of phonological erosion and fusion (e.g. the verb
focusing/passivizing vector verb ga-il ‘go -pst.3’ is routinely pronounced [gɛ]).
On a side note, the collocation phuT gaile in (24) is strikingly similar in form and
meaning to expressions in the distantly related West Germanic languages German
and Dutch, i.e. kaputt gegangen ‘got broken’ (German) and kapotgegaan (Dutch).
I hypothesize that language contact is responsible once more for the present
dominance of head-initial AuxV order in the Sarnami auxiliary constructions
covered in this section. We are dealing with a case of structural borrowing from
Sranan and Dutch. Auxiliary constructions in Sranan exclusively involve headinitial (AuxV) order, as shown in the following two examples, featuring the verbs
go ‘go (do sth.)’ and wani ‘want’:
(25) a go teki
3sg.sbj go take
wan moro
one more
langa udu.
long wood
‘He went to take a longer (piece of) wood.’ (Sranan)
(26) den man man
a bon.
def sg tree
wani puru a
want remove
trui ini
pullover in
‘The men want to remove the pullover from the tree.’ (Sranan)
Dutch also has AuxV order in main clause auxiliary constructions, compare the
following example:
(27) ik kan
I can
komen, als
come if
je wilt.
you want
‘I can come tomorrow, if you like.’ (Dutch)
146 Kofi Yakpo
Head-final order (VAux) is also possible in Dutch auxiliary constructions.
However, it is restricted to subordinate clauses and unusual in spoken and
written Dutch (cf. Donaldson 2008: 219). Hence, alternative (a) (AuxV) below is
more common than (VAux):
(28) a. ik
weet dat je
know that you
kan komen.
can come
‘I know that you can come.’ (Dutch)
b. ik weet dat je
I know that you
komen kan.
come can
‘I know that you can come.’ (Dutch)
Once again, Sranan and Dutch pull into the same direction and it is very likely
that constituent order in auxiliary constructions in both languages provides the
pattern onto which Sarnami constructions like (19) above are grafted.
4.4 Summary
The preceding three sections have shown that Sarnami and its immediate relativesin India go separate waysin constituent order. I have suggested that the divergence of Sarnami from the Indian control group languages is the consequence of
contact with Dutch and Sranan. Tab. 6 below summarizes the findings.
The Indian languages Bhojpuri, Maithili, and Magahi feature the dominant
head-final order characteristic for Indo-Aryan. This manifests itself as object-verb
basic word order (SOV), preposed correlative clauses (RelN), and verb-auxiliary
(VAux) order. By contrast, Sarnami departs markedly from its genetic heritage.
More than a third of Sarnami main clauses show an SVO order, against only
4% in the Bihari control group. Sarnami has nine times as many clauses with
SVO order as the Indian languages. With auxiliary constructions, the divergence
from inherited patterns is even more pronounced. Head initial (AuxV) auxiliary
Tab. 6: Frequencies of head-initial order compared.
Head-initial order
Bhojpuri, Maithili,
Magahi (India)
Sarnami: India ratio
Out of India: Language contact and change in Sarnami (Caribbean Hindustani) 147
constructions are sixteen times more frequent in Sarnami than in the Indian
languages. In this domain the changes have therefore been the most profound.
Nevertheless, I have also shown that head-initial order is not uncommon in
the Indian languages. SVO word orderisrecruited to focusthe object, for example,
and postposed relative clauses are pragmatically marked in a similar way. In
Sarnami, SVO and postposed relative clause (NRel) orders are however competing
with SOV and preposed correlative clauses(RelN) for unmarked status. Even with
respect to relative constructions, where the variability of constituent order is the
greatest in the Indian languages, postposed relative clauses still occur twice as
much in Sarnami as in the Indian languages.
On a whole, the statistics show that head-initial order has made inroads into
the syntax of Sarnami and is competing with or displacing inherited head-final
patterns. I attribute this development to contact with Sranan and Dutch.
5 Conclusion
The changesin constituent order described in the preceding sections are substantial
enough to set Sarnami apart from itsIndian next-of-kin Bhojpuri, Magahi, and Maithili. I have suggested that the cause of these morphosyntactic changes is contact
with Sranan and Dutch.I have shown that there is a statistically significant tendency
away from head-final to head-initial structures in basic word order, relative constructions and auxiliary constructions. In all three domains, Sarnami, like its Indian
relatives, already employs head-initial structures. But the frequency of head-initial
structuresis considerably higher in Sarnami than in the Indian control group.
There is also a considerably higher frequency in the Indian languages of
head-final relative constructions compared to head-final clausal word order and
auxiliary-verb order. The position of relative clauses therefore appears to be very
variable in the Indo-Aryan relatives of Sarnami from the outset. It istherefore possible that the order of relative clauses is generally more prone to change during
language contact than other constituents (see e.g. Collins 2012: 95), because it is
a relatively unstable order in the first place. The order of auxiliaries vs. lexical
verbs in Sarnami diverges most from its Indian relatives in statistical terms, even
if my observations only hold for modal and aspectual auxiliaries that are close
in their meanings to their lexical verb counterparts. More grammaticalized auxiliaries (copular verbs and vector verbs) whose meanings are more opaquely connected to their lexical counterparts, and which fulfill more abstract grammatical
functions have retained their post-verbal position. Many of these vector verbs also
have a much higher textual frequency than the auxiliaries covered in section 4.3.
148 Kofi Yakpo
To sum up, Sarnami still features a dominant head-final order (SOV) in basic
clauses. In relative constructions and with certain types of modal and aspectual
auxiliary constructions head-initial order is predominant. A crucial facilitating
factor for these changes in Sarnami is that head-initial orders already exist in
Sarnami’s Indian relatives, even if pragmatically specialized. Further research
is needed to assess whether the tendency towards head-initial structures can
also be observed with other syntactic units, e.g. the preferred order of verbs vs.
adjuncts, and verbs vs. indirect objects. Constituent orders that have been shown
to be diachronically more stable seem not to have been affected by change in
Sarnami, e.g. noun vs. adjective, noun vs. adposition and noun vs. genitive order
(see e.g. Greenberg 1969; Greenberg 1980; Collins 2012). Sranan and Dutch are
not always consistent either in the way their basic word order (SVO) correlates
with other constituent orders. However, Sranan, which once made extensive
use of postpositions, has all but completed a typological shift to prepositional
structures (Yakpo, this volume), which correlates with its dominant SVO order
(see Greenberg 1969; Dryer 1989; Dryer 1992). It therefore remains to be seen how
much typological pressure unfolds on Sarnami to reconfigure its head-final constituent order in adpositional phrases as well, and borrow or develop prepositions
next to its existing set of postpositions.
The development in Sarnami is a good example of the kind of convergence
that typifies Suriname as a linguistic area, when the two dominant languages
Sranan and Dutch exert pressure for change in the same direction because they
share typological features.
ablative marker
accusative-dative marker
causativizing infix
definite article
distal demonstrative
emphatic form
ergative case
feminine gender
high honorific pronoun
honorific pronoun
Out of India: Language contact and change in Sarnami (Caribbean Hindustani) 149
indf indefinite article
inf infinitive
ipfv imperfective aspect
loc locative preposition
m masculine gender
mhon mid-honorific pronoun
neg negator
nhon non-honorific pronoun
obj object
pa rt participle
pfv perfective aspect
pl plural
poss possessive
p r f perfect tense/aspect
prog progressive aspect
prox proximate demonstrative
prs present tense
pst past tense
quot quotative marker
r e f l reflexive pronoun
r e l relative pronoun
sbj subject
sg singular