The Creole Prototype and the Typology of Sinitic

Stephen Matthews    &   Umberto Ansaldo
University of Hong Kong   University of Stockholm

(draft as presented at the Yuen Ren Centre for Chinese Linguistics 5th Annual Symposium, UC Berkeley 1997;
please check with authors before citing)

This paper takes up the challenge posed to creolistics and typology by the structural similarities between creoles and the Sinitic languages. The solution to this problem has far-reaching significance for Chinese linguistics in terms of explaining some of the characteristic features of Chinese, as well as for the idea of creoles as a distinct typological class. McWhorter (1998) argues that creoles constitute a unique structural class of languages, characterised by little or absent inflectional affixation, semantically regular derivational affixation and little or no contrastive use of tone. The tonal criterion, which has the effect of excluding Sinitic and "sinospheric" languages from this class, is problematic to the extent that some creoles do have lexical tone. Moreover, many of the typical creole structural features including preverbal TMA particles, verb serialization and 'exceed' comparatives are shared or paralleled in Sinitic. To account for these observations we discuss two possible avenues of enquiry:

(i) The creole prototype represents one of a finite number of possible language types, which happens to be represented by Sinitic as well as in creoles (which are forced to be that way by the circumstances of creole genesis) and entails the shared properties;

(ii) The typology of Sinitic may represent the result of early creolization, or at least contact-induced change involving a subset of the factors involved in creolization. The long-standing multilectal situation of China as described in Escure (1997), with a target language spoken by a small, highly-educated part of the population and a great variety of local languages spoken by the masses, presents a socio-historical scenario not unlike that in which creoles are situated. These circumstances could account for the maintenance, if not the origin, of creoloid characteristics in Sinitic. Our account appeals to elements of both scenarios: a certain number of features at issue are the typological consequences of isolating morphology, while others may be motivated by the creole-like scenario in which the varieties of Sinitic appear to have developed.

1. The problem

Creole-Sinitic parallels Parallels have been drawn between Mandarin and creoles before. However, these parallels have not to our knowledge been presented in a systematic fashion, nor have they been extended to other Sinitic varieties. In this paper we try to:

It has long been noted that Chinese presents several features which are also found in creole languages. It is easy enough to find prima facie resemblances where the creole translates word-for-word into Chinese:

(1) Di pikni ron kom hoom. (Guyanese creole: Winford 1993)   (2) Xiao haizi pao hui jia (le). (Mandarin)
      the child run come home                                                          small child run return home (PRT)
     "The child ran back home."                                                    "The child ran back home."

(3) A  ska  fe  wan  xadyi  da  na- namay (Fa d'Ambu: Post 1995: 200)     (4) Keoidei haidou hei gaan uk bei di ga-jan. (Cantonese)
    3gen ASP make a house give ART-family                                                     they PROG raise CL house give CL family
    "They were making a house for the family."                                                  "They were making a house for the family."

With respect to certain topics -- serial verbs for example, as seen in (1-4) -- there has been some cross-fertilization between Chinese linguistics and grammatical studies of creole languages (e.g. Wu 1992, Veenstra 1996). To our knowledge, however, the similarities have not been presented as a whole, nor have serious attempts been made to explain why they should hold. This is partly a consequence of disciplinary boundaries: creole studies grew out of sociolinguistics and now operates to a large extent as an independent field, while Chinese linguistics has mushroomed so that it rarely looks beyond (what we would call) the Sinitic family. Some interaction between the two fields appears overdue: after all, China as a linguistic area offers plenty of examples of language contact, pidginization and creolization.1 In this paper we will:

(i) Review what is meant by the creole prototype;
(ii) Look at the typology of Sinitic in relation to the creole prototype and some specific creole/Sinitic parallels;
(iii) Raise the question of how these similarities may be explained.

The similarities and their explanation have far-reaching significance on both Chinese linguistics and creole studies. For Chinese linguistics, the questions are, why the creoloid characteristics - are they somehow accidental? And why the typological chasm between Sinitic and Tibeto-Burman? This chasm is well illustrated by DeLancey's typological description of the Tibeto-Burman family: "With the exception of Karen, all the Tibeto-Burman languages are postpositional SOV languages with predominantly agglutinative morphology... and this must also have been true of Proto-Sino-Tibetan... Proto-Tibeto-Burman is now reconstructed with a split ergative case marking and verbal agreement system." (DeLancey 1987: 806) Even in Archaic Chinese there is barely a trace of these characteristics: SVO order dominates from the start and most of the other SOV characteristics have also been lost (except relative clauses: the head-final NPs are presumably inherited from Sino-Tibetan). For creole studies, the creole/Sinitic parallels suggest various important conclusions: either creoles are not structurally unique, as some have claimed, or Chinese has itself undergone some degree of creolization and the sociohistorical classification of creoles is insufficient.

2. Creoles as a typological class

The extent to which creoles, or more generally contact languages, are structurally unique has become a major point of controversy. On the one hand, Bickerton's Language Bioprogram Hypothesis (Bickerton 1981, 1984) strongly implied that they were: "these languages show similarities which go far beyond the possibility of coincidental resemblance and which are not explicable in terms of conventional transmission processes such as diffusion or substratum influence." (1981: 132) On the other hand, Thomason (1996), for example, argues that the class of contact languages does not correspond to any structural typology.2 "This definition [of contact languages] is fundamentally historical; it is based on diversity in the sources of linguistic structures, rather than on (say) typological characteristics of the language. The reason for insisting on a historical definition is that synchronic definitions don't work; there is, for instance, no such thing as a master list of linguistic features that are universally shared by, and exclusive to, contact languages, or even pidgins and creoles as a set." (Thomason 1996: 3) Nonetheless McWhorter (1998) argues that creoles do constitute "a synchronically definable typological class of languages", characterised by the conjunction of three traits:3

(i) little or no inflectional affixation
(ii) little or no use of tone (a) to distinguish monosyllables, or (b) to encode syntactic distinctions
(iii) semantically regular derivational affixation

Crucially, it is the simultaneous occurrence of all three properties which defines the creole prototype. McWhorter argues that they are all the result of imperfect acquisition by "emergency learners"; inflection, irregular derivation and extensive use of tone are opaque to learners in such situations, and appear only as a result of protracted development over time. We believe that McWhorter's characterization of creoles is basically correct, especially with regard to criteria (i) and (iii) which we find convincing and coherent. We have some reservations concerning the tonal criterion as applied to creoles, however, which are detailed below.

2.1 Lexical tone

The tonal criterion (ii), which has the effect of excluding Sinitic and "sinospheric" languages from this class, appears problematic to the extent that some creoles do have lexical tone. In Nigerian English Creole, for example, we find the following minimal pairs:

ba/ba/ = 'father'; baba = 'barber'

fa(da( = Catholic priest'; fa(da( = 'father' (Mafeni 1971)

These are by no means marginal creoles: Saramaccan has been seen as the "purest" and most radical creole. Yet if we remove the tonal criterion, we are faced with a large group of Sinitic and "sinospheric" languages which share McWhorter's other two characteristics. McWhorter appreciates the potential problem posed by the creoloid characteristics of Chinese, such as lack of inflectional morphology and verb serialization, and relies on the tonal criterion to circumvent it. To make the criterion more precise, he argues that creoles do not use tone (a) to distinguish monosyllables, or (b) for syntactic purposes: those creoles which show tone make almost no grammatical use of it. This may represent a striking contrast with West African tone languages, but a less strong contrast with Chinese which (as far as we know) makes little grammatical use of tone.4

We feel that of McWhorter's three features making up the creole prototype, the tonal criterion is the most problematic, for the following reasons:

(a) some creoles show quite extensive use of tone, even if not for grammatical purposes. McWhorter (1998) argues that such cases arise only where substrate and superstrate are closely related, as in Sango, Kituba and Lingala.
(b) It seems to be the case that tone is typically lost in the course of creolization. This, however, is largely true with regard to situations where only one or more substrate languages have tone. Where both substrate and superstrate are tonal, as in the African creoles, tone is frequently retained in the creole (e.g. Kituba, Lingala). Moreover, the studied cases involve African tonal systems; tone systems such as those in southeast Asian languages are very different and could behave differently. Indeed there are cases of southwest China where the number of tones increases under language contact (e.g. the 15 tones of Dong, a Kam-Sui language).
(c) the account assumes that the acquisition of tone is somehow a complicated task for learners - a point which needs to be demonstrated. For example, acquisition of tone does not seem to be particularly problematic in Sinitic languages - even for bidialectal children acquiring two separate tone systems.

2.2 Inflectional Morphology

Turning to McWhorter's inflectional criterion, the match between creole and Sinitic typology appears close: both have extremely limited inflection. McWhorter acknowledges that some creoles have at most one or two inflectional affixes. This is also the conclusion that most analyses of Chinese come to: that verbal -le and -zhe, Cantonese -zo2 and -gan2 are suffixes. We might add that any inflection is regular rather than paradigmatic. For example, the Portuguese 3rd person possessive pronoun sua becomes the possessive suffix -sua in Macanese (based on Ferreira 1996: 237):5

Portuguese possessive pronouns     Macanese possessive forms

1 o(s) meu(s) / a(s) minha(s)              iou-sua "my"
2 o(s) teu(s) /a(s) tua(s)                     vo(sso or vo(s-sua
3 o(s) seu(s) / a(s) sua(s)                   e(le-sua / su
1pl o(s) nosso(s) / a(s) nossa(s)        no((sso or no((s-sua
2pl o(s) vosso(s) / a(s) vossa(s)        voso((tro-sua
3pl o(s) seu(s) / a(s) sua(s)               ilo(tro-sua "their"

McWhorter argues further that creole grammatical markers are rarely obligatory. For example, in the extinct Negerhollands Dutch creole the plural morpheme sini appear to be optional. Tagliamonte et al (1997) show that in Nigerian Pidgin English (which, despite the name, is actually considered a creole) plural marking is grammatically optional:

If yu go daun Walkley, at taim na br(d, Ordin(ri br(d, at taims na kek.
"If you go down to Walkley, at times its' ordinary bread, at times it's cake."

Here the word taim appears once without the plural suffix and once with it, with the same meaning. Again this non-obligatory property has often been pointed out with respect to Chinese inflectional morphology such as aspect markers (e.g. Matthews & Yip 1994: 200).

2.3 Derivational Morphology

If creoles and Chinese alike are "impoverished" with respect to inflectional morphology, derivational processes are rather more extensive, including both affixation and reduplication. According to McWhorter's creole prototype, derivational morphology is semantically transparent and regular. Again this applies similarly to Chinese; the examples from three dialects given below show the transparency and regularity of the few derivational affixes found.

Baba Malay offers a striking example of loss of inflectional morphology in creolization: the active meng- and the passive di- prefixes are lost, while the derivational prefix ter- and the suffix -kan are preserved. The prefix ter- in Malay marks a completed action, typically involuntary and often passive in meaning, e.g. (Chia 1994:56): ter-kejut "startled"; ter-peranjat "shocked" Ter- is less than fully productive: not all transitive verbs have a ter- passive, while adjectival ter- "is not inflectional, and has no syntactic significance" (Lewis 1969: 48). It attaches to various word classes, consistent with derivational rather than inflectional status. A similar argument for derivational status may be made for the morpheme kan, which in Malay is a valency-changing derivational suffix covering causative, benefactive and applicative functions. Only a subset of these functions are preserved in BM; typically it forms transitive verbs from intransitive verbs, nouns or adjectives:

(5) Dia pandang kan dia sampay dia tu masok rumah.
      he  stare  KAN  her  until  she  go  house
    `He stares at her until she goes into the house.'

(6) Chek Chek nama kan lu Pai Kiah.
      Chek Chek name KAN you Pai Kiah
    `Chek Chek named you Pai Kiah.'

(7) Kita boleh putus kan harapan anak kita.
     We would broken KAN hope child our
    `We would smash our child's hopes.'

Both ter- and kan, then, function as derivational, rather than inflectional morphemes, even though -kan at least is arguably inflectional in Malay. Holm gives other examples to show that "the status of morphemes transferred from European [i.e. lexifier - SM/UA] to creole lexicons could change from inflectional to derivational, and perhaps from bound to free." (1988: 97).

3. Common properties of creoles and Sinitic languages: the verbal system

Having reviewed the general typological traits the creole prototype we will now focus on some specific properties shared by creoles and Sinitic languages. To make the task manageable we shall focus on the syntax of the verb: tense/mood/aspect, the verbal properties of adjectives and serialization (with particular reference to comparative constructions).

3.1 TMA and the creole prototype

The tense-mood-aspect systems of creoles were among Bickerton's Bioprogram features. He explained their similarity by arguing that the Bioprogram provides a set of privileged notions, including the non-punctual aspect. Even by those who do not make Bickerton's assumptions, tense/aspect markers show remarkable similarities (Holm 1988, Matthews 1993). According to the Bioprogram these morphemes appear preverbally in the order T-M-A. The Saramaccan forms show this (Bakker et al 1995):

Tense: Anterior                                     Mood:  Irrealis                         Aspect: non-punctual

 Mi bi-nján dí físi                                  Mi o-nján dí físi                           Mi tá-nján dí físi
 I ANT-ate the fish                               I IRR-eat the fish                         I IMP-eat the fish
"I ate/had eaten the fish."                    "I will/would eat the fish."              "I eat/am eating the fish"

Preverbal aspect markers are not a particularly striking feature of Chinese in general, since many Sinitic languages have at least some suffixed forms. However, Chappell (1992) suggests that preverbal marking predominated in older stages of Sinitic, and that this situation is preserved in the isolated Minnan group of dialects which exhibit a number of archaic features. Their TMA systems are also closer to the creole prototype, with preverbal progressive and experiential particles. Chappell (1992: 75) observes that "Preverbal aspect marking can be considered a special characteristic of the Min group". Chaozhou, for instance, has no grammaticalized perfective marker (liau being a shallow recent borrowing from Mandarin).6 Minnan dialects typically have preverbal imperfective markers:

(8) Chaozhou to:     I    to    t'ak   tse
                            s/he PROG read book
                             "He's reading"

(9) Taiwanese te?:     Ts'ia te? kia~
                                vehicle PROG go
                               "The train is going"

The indigenous Minnan experiential marker pak is also preverbal:

(10) I pak khe Pakkia
       s/he EXP go Beijing

One more point on TMA systems: although not part of Bickerton's Bioprogam, many creoles also have completive aspect forms which appear postverbally or at the end of the clause:

(11) A nján dí físi káá (Saramaccan: káá < Portuguese acabar "finish")
       she eat the fish finish
      "He already ate the fish."

(12) Tutu a was-t( fama, a skono (Berbice Dutch: fama < Eastern Ijo)
       when it wash-PFV finish it clean (Kouwenberg 1994)
      "When the washing is done it is clean."

Typically these are derived from verbs meaning "finish". The comparison with Mandarin le < liao is striking in both the grammaticalization pathway and the syntax. Parallels also appear in the expression of modality and futurity: rather than a future tense, Sinitic languages have preverbal modal elements whose meaning includes future and sometimes habitual functions - just like Bickerton's irrealis category (1981: 257-60). Essentially these categories are represented in Sinitic, Cantonese wui5 being an example:

(13) Ngo5dei6 dei6 ji6 si4 wui5 faan1 lai4 ge3 (future)
           we another time will return come PRT
        "Later on we'll come back."

(14) Keoi5 jau5 zan6 si4 wui5 gam2 ge3 (habitual)
         s/he have some time will thus PRT
       "He's like that sometimes."

The relative positions of mood and aspect also match the creole prototype: assuming that tense per se is not present, the order modal - aspect marker - verb matches the T-M-A of the Bioprogram.

3.2 Adjectives as attributive verbs, or the adjectival-verbal continuum

Lesson 1 of Cheung's Practical Chinese Grammar points out that adjectives are verbs. This has been a traditional position and is probably remains the majority view in Chinese linguistics. Similarly for creoles, Bickerton, Sebba and most recently Don Winford (1997) have argued for adjectives as a subclass of verbs. More importantly, the verbal properties of property predicates in creoles are exactly those of Sinitic:

          (15) A liba bradi (Sranan: Winford 1997)
                the river broad
               "The river is wide"

3.3 Serialization and Comparatives

Another characteristic of creoles is the comparative construction using a verb "surpass" which Stassen (1985) terms the "Exceed" comparative:

(22)   a bigi pasa mi (Ndjuka: Holm 1989)
         he big pass me

The defining characteristic of the "Exceed" comparative is a single clause where the standard [of comparison] is constructed as the object of a transitive verb meaning 'to exceed' or to 'surpass'. Particularly clear examples are offered by the Yue comparative with gwo:9

(23) Ngo5dei6 di1 hok6saang1 do1 gwo3 lei5dei6 ge3.
            1-PL CL students more pass you-PL PRT
       "We have more students than you."

This construction type appears to be related both to the categorial status of adjectives and to serialization: in origin, at least, the structure is a serial construction with the "adjective" one of the verbs and the "exceed" verb another. Moreover, some of the verbal properties such as transitivity are often retained in the comparative constructions, as Huttar & Koanting (1993) show for the Ndjuka comparative markers pasa and moo. Similar observations hold for Cantonese comparative marker gwo (Matthews 1997, Ansaldo 1998).

Creoles exhibiting 'exceed' comparatives                              'Exceed' comparatives in Sinitic

Gullah: I tal pas mi (Turner in Holm)                                                            Cantonese: ngo daai gwo lei [A Va V2 B]
                                                                                                                                     I big surpass you M

Sranan: John bigi moro Peter (Winford 1997)                                              Minnan:   i k'ah kuãi gua [A Adv Va B]
                                                                                                                                  I exceed tall you

Lesser Antillean: U gra~ pase mwe(( (Taylor, cit. Holm)
                           you tall pass me
                        "You are taller than I"

Haitian: Pa-koke makout pi ro pase men ou (Hall, cit. Musyken/Veenstra)
            NEG-hang basket more high pass hand 2sg
            "Don't hang the basket any higher than your hand."

Sao Principe: Rimá mE may( f(rti pasa mi (Valkhoff, cit. Holm)
                    brother my more strong pass me
                 "My brother is stronger than me"

4. Explanations

Turning to explanations, we see basically two options, although they are not mutually exclusive. Chance -- "it just happens to be that way but the reasons cannot be known" --should be last resort. We might call this the "Berkeleyan explanation" (after the philosopher, not the campus). Rather than resorting to this, we shall discuss two possible explanations:

(i) There is a typological class which includes creoles as well as Chinese and other "sinospheric" languages. The creole prototype would then represent one of a finite number of possible language types, which happens to be represented by Sinitic as well as in creoles. For this account to work it is necessary to spell how the type entails the shared properties, preferably in terms of implicational universals.

(ii) Alternatively, there is the possibility that Chinese at some stage underwent some form (or degree) of creolization. This possibility has often been mentioned, but not pursued systematically in the light of what is now known about creolization. Assuming McWhorter's model, the issue becomes a matter of degree to which Chinese - both structurally and sociohistorically - approximates the creole prototype.

4.1 A typological account

Given the similarities we have considered, a typological account might run as follows. The creole prototype represents one of a finite number of possible language types: SVO with isolating morphology. Creoles are forced to be this way by the circumstances of creole genesis, specifically: morphological reduction in the course of pidginization and/or imperfect acquisition of second languages.10 Chinese, on the other hand, happens to belong to the same type, without (necessarily) having gone through any of the same processes. In order to account for the similarities along these lines, it needs to be shown that this basic typology entails the shared properties, whether with absolute implicational universals or probabilistically through statistical universals.

As an example of this line of argumentation, it was suggested in Matthews (1993) that the predominance of imperfective (non-punctual) aspect forms in creoles relates to their isolating morphology. Since perfectives are overwhelmingly inflectional (Dahl 1985) and isolating languages (and creoles) lack inflectional morphology by definition, creoles are forced to express the universal perfective / imperfective distinction through progressive or imperfective forms such as those discussed above. Hence the preponderance of "non-punctual" aspect forms noted by Bickerton. The status of adjectives as attributive verbs, as discussed in 3.2 above, may also be related to the isolating typology shared by Sinitic and creole languages. That is, in inflectional and agglutinating languages there may be affixal morphology which is specific to the adjective class, while in an isolating languages this option is ruled out.

Another extension of this argument is to serialization. Functional arguments lead to the conclusion that creoles are forced by the lack of case and/or paucity of prepositions (see especially Schiller 1993). The complementarity between SVCs and Ps or case morphology on the other has often been noted/suggested (Bickerton & Muysken 1980 in Schiller 1993). A relevant observation here is that at higher levels of the lectal continuum more Ps are adopted from the superstrate language and there is less use of SVCs (Escure 1991). Muysken & Veenstra (1995) note the following typological correlates of serialization: "Verbal derivational morphology is absent... exceptions are reduplication and e.g. obsolete causative formation in Berbice Dutch. Many serializing languages have null case marking and case assignment under strict adjacency." (1995: 292) In the creole literature, the comparative with "pass" is widely treated as an instance of serialization. In Stassen's (1985) typology, serialization is seen as a form of clause-chaining, and the Exceed comparative is implicationally related to serialization. Specifically: If the language has a comparative construction of the 'Exceed' type, the language necessarily has conditional deranking [i.e. SVCs with the same subject]. (Stassen 1995) There is also a strong correlation with SVO word order: If the language has an 'Exceed' comparative then its basic word order is SVO. (ibid)

There thus appear to be some implicational relationships between the shared properties of creoles and Sinitic languages. With respect to creoles, we know why they have these properties: they are the result of the processes of creolization. McWhorter spells out the process: "The rapid adoption of a language as a lingua franca by non-native speakers entails the stripping down of the system to its essentials, for optimal learnability and processability. The natural result is the virtual or complete elimination of affixes and contrastive tonal distinctions, legacies of usage over time which are not essential to a communication system." (McWhorter 1997) The next question to arise is whether the appearance of creoloid properties in Sinitic has any relation to such processes of creolization, or is attributable to other sources.

4.2 Creolization hypotheses

Here we need to distinguish two rather separate (but again not mutually exclusive) hypotheses:

(i) The possibility of creolization at an early stage in the development of Sinitic;
(ii) The status of Standard Chinese as a lingua franca with creoloid features.

Ballard's (1984) notion of "mother soup" and DeLancey's 'creolization kitchen' describing a stage of intensive and prolonged language contact in the Yellow River Basin as an ancient phenomenon may account for the origin of at least some creoloid features; the possibility of semi-creolization should be considered here.11 Unfortunately the lack of suitable data makes it difficult to pursue this matter: so far it can only be viewed as an hypothesis. However, it is by raising this kind of issue that attention to and further understanding of it can be achieved. What we do know is that a form of common standard, or lingua franca, existed in China already in very early times. Evolved already during the Han Dynasty, reinforced by the Tang classical period, this high variety of Chinese has for centuries been of exclusive 'property' of a small, well-educated class. The only variety to have established written status, it was imposed on the masses that spoke other varieties of Chinese (and non-Sinitic languages) who learned a certain amount of this superstrate imperfectly in a multilectal acquisition situation, as described in Escure (1997). In this study, the author compares acquisition processes of English in the multilingual situation found in Belize with the acquisition of Putonghua in China. We are all aware of the multilectal situation of China; Escure's approach is particularly revealing in that she highlights the multilingual aspect of the acquisition of Putonghua. To her observations we may add that, especially in those regions of China where non-mandarin varieties are spoken, Putonghua enters the picture of a child's life only later on in school, often in primary school. Depending on the training level received by the teacher, the level of Putonghua varies, though it is reportedly rising. Still, as the language spoken at home is often a different variety, as is the language of the community, Putonghua becomes a standard to which the child aims, but never achieves completely. Escure's notion of acrolect is important here: acrolects are innovatory, never identical to the target language:

"native speakers of basilectal creoles produce their own version of the target language. This does not imply that speakers are "unsuccessful" in their attempt at learning the standard. Their goals may simply be different and it seem in appropriate to assume that acrolectal patterns are approximations of the target language. It may be more relevant to refer to acrolects as innovations." (Escure 1997: 65)

Escure also suggests that this situation is by no means new in China: lingua francas have existed for a long time. Indeed, at times in the past even more than today, a highly specialized, intellectual class has reigned in linguistic matters over a vast population. The way in which the standard has been imposed on the masses is not dissimilar from the way in which superstrate languages have influenced creoles. We can envisage a scenario where the imperfect standard has been negotiated with different Sinitic languages and different dialects (as well as non-Sinitic languages) for a period of some 2000 years, resulting thus in an extremely slow, almost frozen creolization process never really fulfilled. This would account for the creoloid features found today in Sinitic. The fact that these features are found not only in Putonghua but in other varieties of Sinitic is explained by the long history of contact and interaction. This scenario could therefore be complementary to the one exposed in (i), as we could envisage combinations of the two: the early creolization may have taken place under circumstances comparable to that in Han times (ii). Furthermore, the situation described in (ii) above could have contributed to the preservation of the creoloid properties. Escure makes a similar observation in the Caribbean context:

"My hypothesis is that the complex acquisition situation which in creole societies leads to the expansion of the repertoire into acrolectal varieties may well be one leading to reactivation of the putative bioprogram. The acquisition of standard varieties occurs under unstable conditions in the Caribbean... and may indeed be compared to creolization, since there, too, the input is so mixed and so confusing that the speaker/learner has to reinterpret the data in order to produce communicative messages. It makes sense to assume that only the availability of universal principles can help him out of the linguistic dilemma." (Escure 1993: 180)


We have shown that there are extensive structural parallels between Sinitic and creole languages. While this fact has been noted before, a systematic correlation between their typology has not been presented, nor have these similarities been explained satisfactorily. To some extent these are the result of a shared isolating typology which entails the broad properties. However we also believe that part of the explanation for the extent of the parallels lies in the history of language contact in China which in several respects is comparable to the circumstances of creolization. We see this as largely consistent with McWhorter's view of the creole prototype: although a strict application of his criteria would separate Sinitic from creoles on the grounds of tone, we have raised some reservations about this criterion as well as the assumption that tone is difficult in acquisition. With respect to the other two criteria, the parallels which we have spelt out show that Sinitic varieties meet these criteria closely and thus approximate closely to the creole prototype. While McWhorter underlines the importance of all three criteria being met, we also note that inherent in the concept of prototype, is a notion of graduality. The view of creole as a prototype notion allows us to distinguish languages which conform more or less closely to the prototype; and indeed languages such as Réunionnais and Afrikaans have been analysed a semi-creoles. Applying this notion of graduality to Sinitic varieties, it is not unreasonable can suggest a partial creolization within Sinitic. The mechanisms have been described in terms of (a) the possibility of early creolization and (b) the clearly observable lingua franca situation that may have been ongoing for a long period. In this respect we would note that McWhorter (1997) has argued that certain creoles have developed more gradually than had hitherto been thought: an ancestor of Sranan was already present in West Africa in 1630. Thus the difference between the time span of the formation of prototypical creoles and that of Sinitic may ultimately be one of degree.


1 This paper began as a contribution to the symposium on "Interdisciplinary approaches to Chinese linguistics" at the Yuen Ren Chao Centre for Chinese Linguistics at Berkeley in March 1998. We are grateful to Samuel Cheung for organizing this event and to the participants for their feedback.
2 The discussion implies that there are languages without contact, which seems untenable on historical grounds.
3 It is clear that McWhorter intends this characterization to be complementary to the sociohistorical one: the question at issue is whether "creole" is merely a sociohistorical term.
4 The clearest and perhaps best-known case is the perfective derived by tone change in some Yue dialects: sik6 faan6 -> sik2 faan6 eat rice eat-PFV rice "eats" "has eaten"
5 This function of sua appears to involve convergence of an old Portuguese possessive structure [NP sua N] with the substrate Cantonese construction [NP ge3 N].
6 In lieu of a perfective marker, Chaozhou uses a variety of verbal complements such as ho which retain lexical meanings as well as the implication of completion.
7 When attributive verbs are reduplicated they no longer allow aspectual modification, as in Saramaccan: *dí miíi tá-bígi-bígi (Bakker et al 1995: 172) the child PROG-big-big This applies equally to Cantonese: keoi5 fei4-zo2 keoi5 * fei4-fei4-(dei2)-zo2
8 Although the bi comparative is treated as an example of the Exceed type by Stassen (1985) it differs from the Cantonese and other southern varieties and it is not clear that it should be treated as an "exceed" comparative. In Cantonese, on the other hand, comparatives are one of the few criteria that do seem to separate adjectival predicates from stative verbs: ni gin si lei zung cingco gwo ngo (cingco = adjectival V) this matter you even clear than me vs. * ni gin si lei zung sik gwo ngo (sik = stative V) this matter you even know than me (ni gin si lei zung sik dak DO gwo ngo this matter you even know ADV more than me )
9 The Minnan comparatives with k'ah may also be considered as 'exceed' comparatives: although k'ah is not a verb, it does have the "exceed" semantics (meaning specifically "too" in Chaozhou) and the overall construction is a transitive one.
10 Mufwene (1991) also noted that the selection of structural options in creolization is typically in the direction of isolating typology and periphrasis, while not accepting "that pidgins and creoles constitute a distinct formal linguistic type" (124; elsewhere he explicitly denies this).
11 To make the idea less far-fetched, one should consider that there was an extensive slave trade in Han and post-Han times, entailing major shifts in population, as well as many cases of armies being moved around the country. We could almost envisage a series of "fort" situations, such as those recognized in the creole literature (Bickerton 1988) all over China.
12 McWhorter's argument comes dangerously close to the example of ad hoc argumentation given by Comrie (1989): RelN -> SOV is only a tendency since Chinese is a counterexample; Rel N and non-tonal -> SOV would be an absolute universal, but "the proposed replacement universals is incoherent since there is no imaginable connection between a language being tonal or not and having verb-final word order or not (in general, tonal and non-tonal languages are distributed randomly with respect to word order types)." (Comrie 1989: 20) 19


Ansaldo, Umberto. 1998. 'A typology of comparatives in Sinitic'. Paper to be presented at the 7th Conference of the International Association for Chinese Linguistics / 10th North American Conference on Chinese Linguistics, Stanford University.

Ansaldo, Umberto & Stephen Matthews. 1998. 'The Min substrate and creolization in Baba Malay.' To appear in Journal of Chinese Linguistics.

Ballard, W.L. 1984. 'The mother soup: a south China recipe for tonometamorphogenesis.' CAAAL 22.

Bickerton, Derek. 1981. Roots of Language. Ann Arbor, Michigan: Karoma.

____ 1988. 'Creole languages and the bioprogram'. In Newmeyer, F.J. (ed). Linguistics: the Cambridge Survey, Vol. 2, 268-84. Cambridge University Press.

Chappell, Hilary. 1992. 'Towards a typology of aspect in Sinitic languages'. Chinese Language and Linguistics 1. Taipei, Academica Sinica.

Dahl, Oesten. 1985. Tense and Aspect systems. Oxford: Blackwell.

DeLancey, Scott. 1987. 'Sino-Tibetan languages'. In Comrie (ed), The World's Major Languages, 799-854. London: Croom Helm.

Escure, Genevieve. 1991. Serialization in creole oral discourse. In In F.Byrne & T.Huebner (eds): Development and Structures of Creole Languages. Amsterdam: John Benjamins. 179-192.

____ 1997. Creole and Dialect Continua: Standard Acquisition Processes in Belize and China. John Benjamins: Amsterdam/ Philadelphia.

Ferreira, Jose dos Santos. 1996. Papiacám di Macau. Fundacao Macau.

Holm, John. 1989. Pidgins and Creoles: A Reference Survey. Cambridge University Press.

Huttar, George & Evert Koanting. 1993. 'Are Ndjuka comparative markers verbs?' In Atlantic meets Pacific: a Global View of Pidginization and Creolization, ed. by Byrne, F. and J. Holm. John Benjamins: Amsterdam/ Philadelphia. 165-74.

Mafeni, B. 1971. 'Nigerian Pidgin'. In Spencer (ed.) The English Language in West Africa. London: Longman.

Matthews, Stephen. 1993. 'Creole Aspect and Morphological Typology' in Atlantic Meets Pacific: a Global View of Pidginization and Creolization, ed. by Byrne, F. and J. Holm. John Benjamins: Amsterdam/ Philadelphia.

____ 1997. 'On describing Cantonese grammar: methodological and typological perspectives.' Paper presented at the 6th Conference of the International Association for Chinese Linguistics, Leiden University, The Netherlands.

McWhorter, John. 1997. 'It happened at Cormantin: locating the origin of the Atlantic English- based creoles.' Journal of Pidgin and Creole Linguistics 12:1, 59-102.

____ 1998. 'Identifying the Creole Prototype: Vindicating a Typological Class'. MS., University of California at Berkeley.

McCawley, James. 1992. 'Justifying part of speech assignments in Mandarin Chinese.' Journal of Chinese Linguistics 22.

Mufwene, Salikoko. 1991. 'Pidgins, creoles, typology and markedness'. In F.Byrne & T.Huebner (eds): Development and Structures of Creole Languages. Amsterdam: John Benjamins.

Muysken, Pieter & Tonjes Veenstra. 1995. 'Serial verbs'. In Arends et al: Pidgins and Creoles: An Introduction. 289-301. Amsterdam: John Benjamins.

Post, Marike. 1995. 'Fa d'ambu'. In Arends et al: Pidgins and Creoles: An Introduction. 191-204. Amsterdam: John Benjamins.

Sackmann, Robin. 1996. 'The problem of 'adjectives' in Mandarin Chinese', in Sackmann R. & M. Budde (eds.) Theoretical Linguistics and Grammatical Description. Amsterdam: John Benjamins

Seuren, Pieter. 1986. 'Adjectives as adjectives in Sranan: a reply to Sebba'. Journal of Pidgin and Creole Linguistics 1, 123-34.

Tagliamonte, Sali et al. 1997. 'Plural marking patterns in Nigerian Pidgin English'. Journal of Pidgin and Creole Languages 12:1. 103-129.

Winford, Donald. 1993. 'Directional serial verbs in Caribbean English creoles'. In Atlantic Meets Pacific: a Global View of Pidginization and Creolization, ed. by Byrne, F. and J. Holm. John Benjamins: Amsterdam/ Philadelphia.

____ 1997. 'Property Items and Predication in Sranan'. Journal of Pidgin and Creole Linguistics 12: 2, 237-302.