How Transparent is Creole Morphology?
A study of Early Sranan Word-Formation
MARIA BRAUN & INGO PLAG
University of Siegen
to appear in
Yearbook of Morphology 2002
Version of 21 March, 2002
How Transparent is Creole Morphology?
A study of Early Sranan Word-Formation
MARIA BRAUN & INGO PLAG
The morphology of creole languages has long been a neglected area of study. One
reason for this state of affairs may well have been the wide-spread belief among
linguists that creole languages are characterized (among other things) by little or no
morphology. Evidence for this belief can be found in many publications, two of which
may suffice to illustrate the point. For example, Seuren and Wekker (1986:66) claim
that morphology [is] essentially alien to creole languages”, and in a recently
published textbook on contact languages, we read that “[m]ost pidgins and creoles
either lack morphology entirely or have very limited morphological resources
compared with those of the lexifier and other input languages.” (Thomason 2001:168).
In a similar line of argument, it has been claimed that the derivational
morphemes of the input languages are lost in creolization and are not reconstituted
later (Mühlhäusler 1997, Bickerton 1988, Jones 1995, McWhorter 1998).
It is also a wide-spread belief that, if a creole has morphology at all, it will be
characterized by regular and semantically transparent morphology. This hypothesis is
explicitly argued for by Seuren and Wekker (1986) and, in considerable detail, more
recently by McWhorter (1998, 2000). In Thomason’s words (2001:168), “[m]orphology
also tends to be extremely regular when it does exist in pidgins and creoles, without
the widespread irregularities that are so very common (to the distress of students of
foreign languages) in other languages’ morphological systems.” In what follows, we
will call this ‘the semantic transparency hypothesis’.
There is, however, a growing literature on the morphology of creole languages
in which it is argued that the above claims are wrong or need further qualification. For
example, several authors have shown that affixation, compounding, reduplication and
transposition are major word-formation processes in creoles (DeGraff 2001, Wekker
1996, Dijkhoff 1993, Sebba 1981), and have argued that semantic opacity is not
generally absent from creoles (DeGraff 2001, Plag 2001).
In this paper, we will investigate these issues in more detail to shed new light on
the nature of creole morphology and the role of morphology in creolization, using data
from Early Sranan, an English-based creole language spoken in Suriname in the 18th
and 19th centuries.
We will show that a large proportion of the lexical stock of Early Sranan consists
of complex words, and that affixation, compounding and reduplication are major
word-formation processes in Early Sranan. It will become clear that the derivational
morphemes of the input languages (English, Gbe, Kikongo and Twi) are completely
lost in the creolization of Sranan, which stands in remarkable contrast to other creole
languages like French-based Haitian Creole or Spanish/Portuguese-based Papiamentu,
which have preserved (or reconstituted) bound morphemes of their input languages.
Furthermore, we will show that the semantic transparency hypothesis is
untenable, both on theoretical and on empirical grounds. Semantic opacity in creoles is
inevitable and comes about through the borrowing of complex words from the input
languages. Creoles and non-creoles are therefore synchronically indistinguishable with
regard to their derivational morphology.
The article is structured as follows. In the next section, we explain our data
sources and methodology. Section 3 is devoted to the analysis of the complex words
we find in Early Sranan, before we investigate in section 4 the problem of semantic
transparency. Our results are summarized in section 5, the conclusion.
2. DATA AND PROCEDURE
The roots of Sranan go back to the second half of the 17th century when a group of
English planters and their slaves settled in the colony of Suriname on the Caribbean
coast of South America. The influence of the English in Suriname lasted for only
approximately 30 years, because in 1667 the colony came under the Dutch rule, and by
c. 1680 the English had practically all left the colony. Thus, Sranan stands apart from
many other creole languages because of a relatively short period of contact with its
superstratum English, and a relatively long contact with another European language –
Dutch, whose influence is traceable in a considerable layer of today’s Sranan
vocabulary. Moreover, the massive import of African slaves until the 1850s led to the
fact that the native West African languages of the Surinamese slaves, Gbe and Kikongo
(Arends 1995:248), played a considerable role in the development of Sranan.
The present paper deals with Sranan as it was documented in roughly the first
one hundred years of its existence. We have chosen Early Sranan as an object of
investigation for two reasons. First, as shown in the growing body of diachronic
research, the study of early stages of a creole language can give us new and valuable
insights into the nature of creolization. Second, the data from early stages of a creole
can serve as a good test of the semantic transparency hypothesis: if we prove that
Sranan displayed instances of opaque derivation already in or shortly after its
formative years, this would constitute a strong counterargument to McWhorter’s
(1998:798) assertion that semantic opacity of non-creole languages (if existent at all) is
the result of a long-term semantic drift.
The main source of data used for the present paper is Christian Ludwig
Schumann’s Neger-Englisches Wörterbuch of 1783, which contains 2391 types and 17731
tokens.2 Schumann’s dictionary was chosen mainly because it is the largest and the
most reliable source of Early Sranan (Kramp 1983:3, Arends 1989:19, Bruyn 1995:154-
155). Schumann worked with informants who were native speakers of Sranan and it is
most likely that he was a proficient speaker of the creole himself. His dictionary
provides accurate and abundant information, both linguistic and cultural (cf. e.g.
Bruyn 1995:154-155, Arends 1989:19).
Additionally, six other sources of Early Sranan were consulted in the course of
the analysis: Van den Berg’s (2000) study of Early Sranan in court records of 1667-1767,
Van Dyk’s (c1765) language manual, Herlein’s (1718) Sranan fragment in his
Description of the Colony of Suriname, Nepveu’s Annotations to Herlein’s (1718) Description
of Suriname (1770), Focke’s Neger-Engelsch Woordenboek (1855) and Wullschlägel’s
Deutsch-Negerenglisches Wörterbuch (1856). These sources were used for verification and
falsification of specific analyses, as well as for translations or etymology of certain
words, for which Schumann’s dictionary provided only insufficient information.
We computerized all 18th century sources and extracted word lists from the resulting
files. From Schumann’s word-list we then extracted manually all words that were
putatively complex. In a further step, the entire dictionary was scrutinized manually
for putatively complex words that had not made it into the word-list for orthographical
reasons, spotting complex words that were neither spelled with a hyphon, nor as a
single orthographic word, but as two orthographic words. As shown by the examples
in section 3, all three types of orthographic representation of complex words occur in
Our idea was to apply a rather generous policy of what might count as
‘putatively complex’ in these initial steps of data gathering in order not to miss any
potentially pertinent items. Hence we arrived at a long list of words that was then
subject to a thorough morphological analysis, the first step of which was to exclude all
non-complex words. As is well-known among morphologists, the determination of
what may count as a complex word is not a trivial matter (cf. e.g. Bauer 1988:109ff,
Katamba 1993:19ff), and often the topic of theoretical debates, as for example the
discussion of compounds being either morphological objects, i.e. complex words, or
syntactic objects, i.e. phrases. We have regarded as complex those items that consist of
two and more elements where at least one element was attested elsewhere. Of these
items, only those were considered complex words which, firstly, appear to be items
with a syntactic category specification of the X°-level (i.e. N, V, A, 3 etc.) and, secondly,
are syntactically inseparable. Thus, e.g. the Early Sranan word tinnatu - ‘twelve’ was
regarded as complex because it possesses a syntactic category specification of the X°-
level (it is a numeral) and cannot be syntactically separated without a fundamental
change in meaning. As is quite common in such classification exercises, there are often
borderline cases, where firm evidence for or against a certain decision is lacking. In the
majority of cases, however, matters were rather clear and none of the crucial
arguments presented in section 4 hinges on the potentially controversial status of an
item as a complex word.
In our overview of Early Sranan word-formation patterns given in section 3
below, we have classified the patterns as either affixation, compounding or
reduplication. This is a somewhat simplified picture, since a strict boundary between
affixation and compounding is notoriously hard to draw. Some theories (e.g. Höhle
1982 for German) even deny such a distinction. In a detailed analysis of the Early
Sranan patterns, Braun (2001) breaks up the distinction between affixation and
compounding into four properties:4 boundness (affixes are bound, compound elements
are not bound), selectivity (affixes are highly selective, compound elements are not),
specificity (affixes have a less specific, i.e. more abstract meaning), and serialness
(affixes form larger series of words). In this approach the properties cluster in different
ways with different morphemes, with prototypical affixes at one end of a scale, and
prototypical bases at the other end. For the purposes of the present paper, such a fine-
grained analysis is not necessary and we therefore confine ourselves to the more
traditional classification into affixation and compounding.
3. WORD-FORMATION INVENTORY OF EARLY SRANAN
The most remarkable quantitative finding about Early Sranan word-formation is
perhaps the sheer number of complex words available in Schumann’s dictionary. Of
the 1644 words, 676 (i.e. 41 %) are complex. These words instantiate 32 different word-
formation patterns, of which 11 are productive. These findings demonstrate that earlier
claims about the absence or marginality of creole morphology are incorrect.
In the following sub-sections, we illustrate some patterns of affixation,
compounding and reduplication as found in Schumann (1783) in order to show the
richness of word-formation in Early Sranan (see also Koefoed and Tarenskeen 1996 for
some discussion of the Early and Modern Sranan lexicon). For full documentation and
discussion of individual patterns the reader is referred to Braun (2001).
Affixation is a common word-formation device in Early Sranan: 177 out of 676 complex
words in Schumann’s dictionary are formed by means of affixation. The data from
Schumann’s dictionary demonstrate that Early Sranan developed a number of affixes
already during the initial stages of its existence.
Early Sranan makes use of two deictic markers -weh (< E. away/?way) and -
dom(m)/-dum(m)/don (< E. down). The affix -weh can be attached to verbal or adjectival
bases, as is shown in (1a, b and c). When attached to verbal bases, it serves to indicate
the direction of the action away from the point of reference. With adjectives, it may
either mark spatial deixis, as in (1b), or temporal deixis (temporal distance away from
the point of reference), as in (1c).
[VERB/ADJECTIVE + weh]VERB/ ADJECTIVE
meaning of base
‘long (spat. & temp.)’
The affix –dom(m)/-dum(m)/don can occur with verbs, as in (2a) or with bound roots, as
in (2b) and, similarly to the affix –weh, denotes spatial deixis (it indicates the direction
down from the point of reference):
[VERB/BOUND ROOT + dom]VERB
‘to bend (down)’
‘to fall (down)’
‘to lie/to lay (down)’ li- (bound root)
‘to sit (down)’
si- (bound root) ‘?sit’
The patterns – in (1) and in (2) seem to be unproductive in Early Sranan.
Another affix is -tentîn (< E. time + D. tien ‘ten’) that is attached to cardinal
numerals from two to nine to form tens from twenty to ninety, as it is shown in (3):
[NUMERAL + tentîn]NUMERAL
Early Sranan also makes use of the person-forming affix -man (< E. man) which can be
attached to nominal bases, as in (4a), to adjectival bases, as in (4b), and to verbal bases,
as in (4c). The meaning of the output nouns is always ‘person having to do with X’,
where X may be N, A or V.
[NOUN/ADJECTIVE/VERB + man]NOUN
‘a deaf person’
‘to do magic’
The pattern introduced in (4) is the most productive affixation pattern attested in
Schumann’s dictionary – 67 words out of the total 177 words produced by affixation
belong to the pattern in (4). Out of the three subpatterns, V+man is the most
Another two affixes attested are the gender markers man(n)- (< E. man) and
uman- (< E. woman) which can be preposed to nouns denoting animals, human beings
or a person’s occupation with the purpose of indicating natural gender:
[man(n)/uman + NOUN]NOUN
futuboi (N) ‘servant’
futuboi (N) ‘servant’
There are two abstract-noun forming affixes in Early Sranan, -sanni (< E. something) and
-fasi (< E. fashion), which attach to adjectives and verbs:
a. [VERB + sanni]NOUN
krukuttu (V)6 ‘be wrong’
‘to play fool’
b. [ADJECTIVE/VERB + fasi]NOUN
kondre (A) ‘worldly’
‘folly/stupidity’ lau (V)
‘to be stupid’
There was no evidence in the early sources (nor in later ones) that any of the
superstratum affixes has survived in Early Sranan. The creole has developed its own
inventory of affixes in the course of creolization, and all English affixation is lost.
Compounding is the most common word-formation process in Early Sranan: the
majority of complex words from Schumann’s dictionary (378 out of a total of 676) are
compounds. This fact confirms claims (as e.g. by Holm 2000:130) that creole languages
favor new combinations of free morphemes rather than new combinations of bound
morphemes. Early Sranan makes use of quite a number of different compounding
patterns, which we will illustrate in the following paragraphs.
The most productive pattern is the combination of two nouns. Different
structural subpatterns can be singled out within this group, depending on whether the
constituents are simplex, complex, or reduplications. As can be inferred from (7), N+N
compounds in Early Sranan can consist of two simplex nouns, as in (7a), or of a
complex noun and a simplex noun, as in (7b) and (7c), or of a simplex noun and a
reduplicated noun, as in (7d and e), or of two complex nouns, one of which is
reduplicated, as in (7f).
[NOUN + NOUN]NOUN
‘honey’ kakka (N) ‘droppings’
‘mouth’ neti (N)
watra (N) ‘water’
jakketi (N) ‘coat’
c. muffe sabbatem
‘stomach’ jamjam (N)
sakka (N) ‘sack/bag’
sakkasakka (N) ‘rattle’
snekki (N) ‘snake’
f. Bakkrakondre-jamjam ‘European Bakkra-kondre ‘Europe’ jamjam
fruits and (N)
Early Sranan N+N compounds can be both endocentric (e.g. (7e)) and exocentric (e.g.
(7b)). However, endocentric compounds with heads in the rightmost position prevail.