In the creole languages of Suriname, epenthetic -m(i) is obligatorily inserted between certain verbs and certain objects. Phonological and syntactic features of both verb and object enter into the definition of the environments in which such epenthesis occurs, with different sets of features required for the different languages. The paper gives most attention to Ndyuka, where epenthesis occurs after monosyllabic verbs ending in a high tone and followed by direct objects or copula complements beginning with a vowel-initial monosyllabic morpheme. The parallel epenthesis in Sranan is also described in some detail, epenthesis here depending not on tone but on nasality of the verb-final vowel. Saramaccan, Matawai, Paramaccan, Aluku, and Kwinti, though treated briefly, provide additional comparative data relevant to conjectures regarding the origins and development of these processes.
The phenomenon with which this paper is concerned is a morphophonological one, both synchronically unusual and diachronically puzzling. It is unusual in that the conditions governing the insertion of epenthetic material are both syntactic and phonological, and both precede and follow the point of insertion. It is puzzling in that, although the languages involved are creole languages for which the historical sources of many of their features are clear, there is no clear source for this insertion--either of the form of what is inserted or of the conditions governing the insertion.
The phenomenon in question, insertion of m and similar forms after certain verbs before certain direct objects, was first described in any detail for Sranan. Voorhoeve (1982, 1985) described the distribution of what he termed the "epenthetic transitive /m/" in Sranan thus:
Monosyllabic verbs ending on a nasal vowel develop an epenthetic /m/ before an object starting with a monosyllabic vowel-initial element (1985:93).
The analogous epenthesis in Ndyuka, the main creole of eastern Suriname, involves mi rather than m, and is conditioned by tone rather than by nasalization. In Saramaccan, the main creole of central Suriname, the epenthesis occurs in a much more restricted set of environments than in Sranan and Ndyuka. In this paper I seek a better understanding of the diachronic processes resulting in the current situation by comparing this epenthesis in these and other creoles, and by drawing on information about their extralinguistic history.
Creole languages are generally characterized by quite 'shallow' morphophonology: to a great extent, underlying form = surface form. Givón (1979:23) suggests that this is particularly the case when the substrate speakers involved do not share a common language-group background, which is the case with Ndyuka. So cases of morphophonological rules in creoles may be of special interest on the one hand; and on the other hand they may be fairly uncomplicated to describe, not having to be integrated with a large number of other such rules.
With regard to the two creoles focused on in this paper, while native speakers of Sranan and Ndyuka can understand a great deal of each other's languages, yet the two languages have had fairly independent histories (Smith 1987) and differ both lexically and structurally. In particular they differ in their phonological systems, the area that turns out to be of prime interest in comparing epenthesis in the two languages. I will begin, then, with a brief summary of the main differences between the phonologies of the two languages.
Both languages have five oral vowels /i e a o u/ and their nasalized counterparts. The latter are symbolized in the practical orthographies commonly in use for both languages by an n after the oral vowel symbol: in, en, etc. Both languages have diphthongs /ai, ei, oi, au, ou/ (also /eu/ in Sranan), but only Ndyuka has sequences of two identical vowels (with the possible exception of one or two forms in Sranan). Ndyuka may, in fact, have three vowels in succession, the first two being identical, as in daái 'turn' and faáu 'faint.' 
Ndyuka long vowels, as we will hereafter refer to sequences of identical vowels, almost all appear to derive from the loss of intervocalic or prevocalic liquids. If there is a cognate in Sranan, the liquid is preserved there:
|1.||koo||----||Ewe klo, etc.||turtle|
|3.||boó||blo||E blow||blow, rest|
|5.||kaí -||---||P cair||fall|
From these examples it can be seen that, in cases where the two languages have cognates, a CVV' sequence in Ndyuka monosyllabic stems corresponds to a CLV sequence in Sranan, while a C V'V one corresponds to a CVLV sequence in Sranan. (An exception is Ndyuka waán, Sranan waran 'warm.') In the former case, we have an additional difference between the two languages: Sranan has CL clusters, Ndyuka does not.
The languages also differ in surface syllable structure in that unstressed vowels are frequently deleted in Sranan, but not in Ndyuka. In particular, word-final vowels in Sranan (often the result of an epenthetic vowel being added to an etymon ending in a consonant) are often dropped before a following vowel, whereas in the corresponding sequences in Ndyuka the two vowels are preserved:
This difference is important to keep in mind in considering the historical relationship between Sranan m and Ndyuka mi, to which I return below.
Ndyuka has lexical tones high and low in contrast, as illustrated by koo 'turtle' and kóo 'cold' and the pairs 2-3, 5-6, and 7-8 above. Sranan does not have such a contrast, with the possible exception of a very few forms. It turns out that the mi-insertion rule in Ndyuka is sensitive to tone.
On the other hand, the phonological systems of Sranan and Ndyuka are alike in some respects relevant to the occurrence of m(i). As already mentioned, they both have a contrast between oral and nasalized vowels, a feature to which the m-insertion rule in Sranan is sensitive. Underlying nasalized vowels in both languages are alternately realized as phonetic nasalized vowel, nasalized vowel plus nasal consonant, or oral vowel plus nasal consonant (for Ndyuka, see Huttar and Huttar 1994:548). In underlying forms, that is, before the abovementioned vowel deletion applies in Sranan, neither language admits any word-final consonant. A surface syllable-final nasal assimilates to the point of articulation of a following consonant, and is velar before pause. Before a vowel, nasalized vowels are denasalized in both Sranan and Ndyuka:
Finally, both languages admit syllables consisting only of a vowel, as in the pronoun and article a and the pronoun en, to name forms especially relevant to our discussion here. Both also have CV as the most common syllable shape, while only Ndyuka has syllables consisting phonetically of only a nasal, as in mboma 'boa constrictor' (cf. Sranan aboma), except for those arising in Sranan from vowel deletion.
We now turn to the distribution of epenthetic mi in Ndyuka. Consider the following examples with the verb fón 'hit':
|a. a fón mi||'he hit me'|
|b. a fón mi ddá||'he hit my father'|
|c. a fón yu||'he hit you'|
|d. a fón yu ddá||'he hit your father'|
|e. a fón wi||'he hit us/you-pl.'|
|f. a fón wi ddá||'he hit our/your-pl. father'|
|g. a fón den||'he hit them'|
|h. a fón den ddá||'he hit their father'|
|i. a fón den pikín||'he hit the children'|
|j. a fómi a pikín||'he hit the child'|
|k. a fómi en||'he hit him'|
|l. a fómi en pikín||'he hit his child'|
|m. a fón ála||'he hit everything/-one'|
|n. a fón sama||'he hit someone'|
|o. a fón a ósu||'he hit at home'|
|p. a fómi a ósu||'he hit the house'|
Such a paradigm defines where mi occurs, as far as its following environment is concerned: mi occurs immediately before the definite article a (12. j., p.) and the 3rd singular oblique pronoun en, the latter constituting either the entire object NP (12. k.) or the possessive constituent thereof (12. l.).
Consonant-initial objects, as in 12. a.-i., n., and objects beginning with a vowel that does not constitute an entire morpheme, as in 12. m., as well as most non-objects, as in 12. o., are not preceded by mi.
The one case of mi-insertion before something other than an object involves the verb toon 'become' and its complement:
|'You become her child.'|
But the insertion of mi depends also on the phonological shape of the verb. Consider the following:
|a. hó:||a hómi en||'he hoed it'|
|b. sá:||a sámi en||'he sawed it'|
|c. fón:||a fómi en||'he hit it'|
|d. nyán:||a nyámi en||'he ate it'|
|e. sí:||a sí en||'he saw it'|
|f. gí:||a gí en a sani||'he gave him the thing'|
|g. dú 'do':||a dú en||'he troubled him'|
|a. boó:||a boómi en||'he blew it'|
|b. boón:||a boómi en||'he burned it'|
|c. deé:||a deémi en||'he dried it'|
|d. deén:||a deémi en||'he dreamed it'|
|e. fií:||a fiími en||'he freed it'|
|a. bóo:||a bóo en||'he bored it'|
|b. jée:||a jée en||'he heard it'|
|c. fáa:||a fáa en||'he felled it'|
|d. bái:||a bái en||'he bought it'|
|e. fíi:||a fíi en||'he felt it'|
|a. kabá:||a kabá en||'he finished it'|
|b. fusután:||a fusután en,|
a fusutámi en
|'he understood it'|
|c. sipán:||a sipán en||'he tautened it'|
Thus in Ndyuka the class of verbs after which mi is inserted is defined differently than in Sranan. In both languages, the verb must be monosyllabic (i.e., no C's after the first V), with the one exception of Sranan ferstan, Ndyuka fusután 'understand'. But in Sranan the verb must end in a nasalized vowel, a constraint not present in Ndyuka (compare 15. a.-b., c.-d). On the other hand, in Ndyuka the final vowel must bear high tone; cf. 15. a. with 16. a., 15. e. with 16. e. This consideration does not apply in Sranan.
Given these insertion rules in Sranan and Ndyuka and the differences between the two languages, what can we say about how these rules may have arisen? And did they in fact originate only once, in some proto-creole or pidgin from which all the creole languages of Suriname derive at least some of their features? Voorhoeve (1985:92) suggests
a historical explanation....Many pidgins have a suffix -m or -am on transitive verbs. A creole language seems to lose this phenomenon....The epenthetic /m/ might be a relic of such a transitive suffix in older Sranan.
In his admittedly brief article, however, Voorhoeve does not cite evidence for such a V+m form from older Sranan material, material with which he would have been eminently familiar. I think we may presume that if he had had evidence from Sranan sources for this hypothesis, he would have included it here, rather than present the idea merely as speculative. It remains then to see whether a comparison of the languages themselves, or information from other languages, can help us at all.
We begin with the phonological data given above. Looking only at Sranan m, we still find two considerations which suggest an alternative to the proposed V+m. The first concerns the preceding environment: m is inserted only after nasalized Vs. This restriction is more naturally accounted for if the original epenthetic form started with the nasal consonant m than if it had an oral--or even a nasalized--V intervening between the underlying final nasalized V of the verb and the m. The second consideration concerns the following environment: while it is true that an epenthetic form ending in a C is more likely to be inserted before a V than elsewhere, we still will have to posit a deletion rule to derive contemporary surface m from the putative earlier V+m; whereas an underlying m (whatever its origin) would require no deletion rule--and an underlying mi also would require no additional deletion rule. For not only is the deletion of unstressed V before V a universally simple phonological phenomenon, precisely such a deletion rule is needed for Sranan anyway, as exemplified in 9. and 10. above.
And once we turn to Ndyuka, where the epenthetic form is always mi, our choice goes to an underlying (or at least historically prior) mi for Sranan, rather than to m. But such an argument, of course, crucially depends on the historical relationship between Ndyuka and Sranan, and then on that between these two creoles and other creoles having a similar epenthesis rule. Before turning to such extralinguistic historical considerations, however, let us briefly address one more strictly phonological possiblity.
So far we have assumed, with Voorhoeve, that we are dealing with the insertion of phonological material before certain kinds of direct objects, rather than the deletion of material in other environments. What about the possibility of considering, e.g., nyámi, deémi, as the underlying forms, historically derived from etyma nyám, dream by the same process that converted so many other C-final etyma into CV-final ones? Then there would have developed, e.g., an allomorph nyáN (where N=engma), or nya~, or both, in preconsonantal, prepause, and certain prevocalic environments. This pattern of m in some prevocalic environments would then have spread to other verbs not having a nasal, but ending with high tone, such as boó 'blow' (15. a.).
But in both Ndyuka and Sranan, etyma ending in nasal consonants did not develop paragogic vowels. Rather, the preceding vowel was nasalized, as in E man > Sr man, Nd mán.
Turning now to extralinguistic history, it is generally agreed (see e.g., Voorhoeve 1973) that Sranan took form during the period of British control of Suriname, 1651-1668, given that the basic vocabulary of Sranan is, or was, largely derived from English, not Dutch. Ndyuka, as already mentioned, is very similar to Sranan, from which Hancock (1969:17) concludes that "the Djuka were probably at one time Sranan speakers who fled into the bush, there evolving their own distinctive speech." If Hancock is right, then the existence of mi in Ndyuka is directly relevant to the question of the origin of m in Sranan. Careful comparison of specialized portions of the lexicons of Sranan and Ndyuka, however, suggests that Ndyuka may actually owe less of its origins to Sranan than such a scenario proposes (Huttar 1985).
Voorhoeve, on the other hand, argues that while Sranan and Saramaccan have a common origin, Ndyuka "must have developed out of an English Pidgin in the eighteenth century" (1973:144) in West Africa, not Suriname, a position supported by Price (1976:39). But even so, the future Ndyukas would still have had exposure to Sranan on the plantations before their escape, so that Ndyuka mi would still be relevant to our questions about Sranan m.
Two other creoles of the interior of Suriname that are similar to Ndyuka and to Sranan are Aluku (or Boni) and Paramaccan. Each of these is spoken by a society of escaped slaves formed during the mid to late eighteenth century (Price 1976:31), now living in eastern Suriname in close proximity to and in continual contact with the Ndyukas. For a long time, in fact, the Dutch government of Suriname treated with the Alukus and Paramaccans only through the Ndyukas. From young and old speakers of both Aluku and Paramaccan I have recorded instances of mi- or m-insertion. Those in Paramaccan all occur in environments that conform with the above description for Ndyuka:
|grate||it||(cf. giín 'grate')|
|kámi||en||báka||'combed it again'|
|comb||it||again||(cf. kán 'comb')|
In Aluku, however, mi-insertion occurs not only as in Ndyuka and Paramaccan, as before en in 20:
|Yu||man||sábi||sámi||en?||'Do you know how to saw it?|
|you||be-able||know||saw||it||(cf. sa 'saw')|
but also before other vowel-initial objects, as before ala 'all' in 21 and before the 2nd sing. pronoun object i in 22:
|a||be||kiími||ála||'He had cleaned it all.'|
|he||ant.||clean||all||(cf. kiín 'clean')|
|eat||you||eat||you||(cf. nyán 'eat')|
More similar to Sranan than is Ndyuka, both these varieties have some deletion of unstressed V before V, though not as frequently, apparently, as does Sranan. And Aluku, as seen in 22., has both mi and m as the epenthetic form.
Saramaccan and Matawai, spoken in central Suriname, have a much higher proportion of Portuguese-derived items in their everyday lexicon than do the other languages mentioned so far. Both societies were formed in the late seventeenth and early eighteenth centuries, quite likely by slaves coming primarily from plantations owned by Portuguese-speaking Jews (Goodman 1987), although this interpretation of the data is by no means uncontested (see Price 1976:36-39). Saramaccan also has m-insertion, as in 23 (Rountree and Glock 1982:45):
|Boo||nyam||ën||'Let's eat/slander him.'|
Yet the insertion of m is limited in Saramaccan to two verbs of African origin (Huttar 1985), /nyan/ 'eat' and /fon/ 'beat' (Catherine Rountree, personal communication). Thus the distribution of m in Saramaccan is very different from that of the corresponding forms in Sranan and Ndyuka as far as the preceding environment is concerned. Saramaccan has no articles beginning with a vowel, so comparison with Sranan and Ndyuka on this detail about the following environment is not possible. Apparently the m after nyán and fón occurs only before ën when it functions as personal pronoun, not as possessive pronoun (C. Rountree, personal communication), again differing from Sranan and Ndyuka. (In the absence of relevant data I will assume here that Matawai, which is very similar to Saramaccan, behaves similarly with regard to m-insertion.)
Finally, Kwinti is another creole of central Suriname, resembling Saramaccan and Matawai in some ways and Ndyuka, Paramaccan, and Aluku in some ways (Huttar 1988; but see Smith 1993 for a reappraisal). Its lexicon, for example, shows less Portuguese influence than the former group, but more than the latter group. Various accounts of the origin and history of the Kwintis agree that they have had a good deal of contact with the Matawais, near whom they now live (de Beet and Sterman 1980). Kwinti likewise has mi- or m-insertion. Examples available to me are compatible both with the distribution rules in Sranan and those in Ndyuka, but not those in Saramaccan, as illustrated in 24-25:
|fa||i||e||fómi||a||kúnamí||'...how you beat the kunami'|
|i||e||bló:m||en...||i||e||kíím||en||'You burn it [field] ...you clear it.'|
From these examples it can be seen that Kwinti has both mi and m; Kwinti does, like Sranan, allow some deletion of unstressed vowel before another vowel, as in sábi en > sáb' en 'know it.'
From the data available on these six creoles, then, it appears that m or mi after nyan and fon was present early on in Suriname, however it may have arisen, but presumably from an m in the African etyma. The epenthetic element then developed differently in the different (groups of) creoles, in regard to both form (m vs. mi) and distribution. Thus it remained limited to these two verbs in Saramaccan, but spread to other monosyllabic verbs ending in a nasalized vowel in Sranan, and to monosyllabic verbs ending in a high tone in Ndyuka. What data we have on Aluku suggest spreading to additional objects than those with which it occurs in Ndyuka, Saramaccan, or Sranan.
At any rate, all the creoles of Suriname share this feature. On the other hand, this feature is apparently not attested for any other creole or pidgin, including those of the circum-Caribbean area and the Atlantic coast of Africa (Fred Cassidy, Ian Hancock, John Holm, personal communication; see also Schneider 1966, Günther 1973, Barbag-Stoll 1983). Nor is it found in the pidgin language based partly on Ndyuka and used for trade purposes between Ndyukas and Amerindians (Huttar 1982). New Guinea Pidgin, of course, has the well-known "transitive suffix" /im/, but its function and distribution are both very different from those of the Suriname m(i) (Mihalic 1971). The examples given by Voorhoeve (1985:92) from a seventeenth-century form of Dutch spoken in Cape Town do not make clear that "this suffix" functions like those in Suriname rather than like the suffix in New Guinea.
But if the epenthetic m(i) did not arise from some possibly universal feature of pidgins, where did it come from? Is there any evidence of such a form with such a function in areas of Africa from which slaves were brought to Suriname?
Considering Bantu briefly, we do have Guthrie's (1971, Vol. 4:234) C. S. 2233b *mi, with reflex mi as a prefix in Bali (B.75) and C. Kongo (H.16b), both of which are in a plausible area to have had input into Ndyuka. But the function of this mi in Bantu is so different from that of the forms in the Suriname creoles that the former is very unlikely to be a source of the latter. More likely would be postverbal object forms as attested in specific Bantu languages (see, e.g., Hyman 1981:15, 78 on 1st sg. me in Noni, a Grasslands Bantu language of Cameroon), if such are found in languages that could have reasonably contributed to the Suriname creoles.
Yet there is so far more evidence for Kwa influence on the syntax of West Atlantic anglophone creoles, including the creoles of Suriname, than of Bantu influence (see, e.g., Huttar 1981), however extensive the Bantu lexical influence may have been. Looking then at Kwa languages, which have also made a significant contribution to the lexicons of many Caribbean creoles, we note the rather widespread occurrence of forms like mi, me, or m as objects immediately following verbs. The following are illustrative:
That these forms are all postverbal, designate an object, and are formally similar to the Suriname m(i) means that we cannot dismiss outright the possibility of some Kwa influence in the rise of the Suriname forms. On the other hand, positing such a source raises a number of problems:
The specific course of both functional and phonological reinterpretation of some Kwa object forms that could have led to the Suriname forms, taking into account the four explicanda just mentioned, is something I will not attempt here. It may well be more reasonable to posit an independent development in Suriname, beginning from the final m found on two common verbs. In addition to the Kwa possibility, we should not ignore other language families of West Africa that also attest postverbal object forms with m, such as West Atlantic (e.g., Temne mi '1st sg. obj.', Wilson 1961:47; cf. various pronominal object forms in Fulfulde, McIntosh 1984:197-199).
* An earlier version of this paper was presented at the 8th biennial conference of the Society for Caribbean Linguistics, 1986. Material for this paper was gathered by Mary L. Huttar and me during residence in Suriname in 1968-1973 and 1988-1990, and for nine weeks in 1981 and two weeks in 1982, under the auspices of the Summer Institute of Linguistics. The 1981 research received generous support from the Wenner-Gren Foundation for Anthropological Research, Inc., and from the National Endowment for the Humanities, the former of which also supported the 1982 research. The research in 1988-1990 was supported by the National Science Foundation under Grant #BNS-8811211. Our research was greatly aided by the use of a 47,000-word concordance of Ndyuka texts, produced at the Oklahoma University Research Institute under National Science Foundation Grant #GS-1605. Of those mentioned in the text or notes, I am especially grateful to Hazel Carter for her many helpful suggestions and detailed information about various languages and language families of Africa.
1 Givón is not precise on how diverse the linguistic background of various substrate speakers can be and still count as having a common language-group background; so one could argue that the creators of Ndyuka did have such a common background, if one considers Kwa, Bantu, Voltaic, etc., as all similar enough.
2 Vowel symbols with ' after them indicate vowels with high tone; unmarked vowels bear low tone. With the generalized vowel symbol V, high to is indicated thus: V'
3 Possible exceptions are nyuundú 'otter sp.' (cf. kiKongo n`-yuundú 'otter'), paaká (cf. kiKongo m`-pàka 'a wood grouse') and other words of African origin (orthography modifed due to HTML limitations). See Huttar 1985.
4 Sranan forms are from the 1980 Woordenlijst. See Smith 1977, Sebba 1982, and Smith and Huttar 1984 on the liquids in the Suriname creoles.
5 For more on correspondences between Ndyuka forms with long vowels and Sranan forms, see Huttar 1987 or the references mentioned in note 4.
6 See Huttar and Huttar (1994:470) on the lack of possessive pronouns as a distinct syntactic category in Ndyuka.
7 Data from the mid-1980's from a male speaker then of 25-30 years of age suggest that the insertion of mi is now optional in many environments (Louis Shanks, personal communication).
8 Sranan may well have continued as a pidgin, however, till well into the eighteenth century, as Plag (1993) argues.
9 Some other verbs are modified in other ways before certain objects. E.g., verb-final a changes to ë before the pronoun ën (Rountree and Glock 1982:69). The verbs nyán and fón have yet other shapes in other environments.
10 My thanks to Hazel Carter for bringing this phenomenon to my attention.
11 Mukarovsky 1976, vol. II:278-279 lists a large number of Gur (Voltaic) and Togo Remnant languages, as well as Kwa languages, with 1st sg. forms consisting of m + front vowel. Such a list, of course, does not indicate the position of such forms relative to the verb.