Over the last 15 years, some creolists have debated the question of creole distinctiveness (or “exceptionalism”). Some creolists such as McWhorter (2005) and Parkvall (2008) have argued that creoles are typologically special, while others (such as Chaudenson, DeGraff, Mufwene and Ansaldo) have emphasized the continuity between creoles and other languages.
Bakker et al. 2011 is only the second paper that tried to address this issue from a quantitative perspective (after Parkvall 2008, the first one to my knowledge), and the first one to appear in a journal of creole studies. The authors examine data from 18 creole languages assembled in Holm & Patrick (2007) (Comparative Creole Syntax, CCS) and conclude that creoles are indeed “a special class of languages with specific typological properties”. But what is the empirical question here?
It would be a great advance in our understanding of language change if we could (fully or largely) predict the outcomes of language contact from the kind of social contact situation that has given rise to the contact-induced change. So suppose that we can predict that a situation of large-scale labour migration of a multilingual workforce leads to adoption of the elite’s language as lexifier and to a particular type of restructuring of this language, notably with reduction of grammatical irregularities and elimination of grammatical markers (“inflection”). That would certainly be an important insight. So let’s call the above social contact situation “social creolization”. The prediction would then be that all languages that recently underwent social creolization must lack irregularities and grammatical markers. We could thus look at existing creoles, assume that they all underwent social creolization recently enough (i.e. there wasn’t enough time for new complexity to arise during subsequent changes), and examine their structural properties.
This is what Bakker et al. 2011 attempt to do, but I see a serious problem with the results based on the CCS data (Holm & Patrick 2007): The editors of the CCS volume specifically selected the features examined there on the basis of their interest for creole studies. Thus, the 97 features can in no way be seen as a representative sample of the structures of these languages – on the contrary, they represent a highly biased sample of features. Thus, it is not the least bit surprising that we and up with a SplitsTree network as in their Figure 9:
Moreover, Bakker et al. do not say explicitly how they define “creole”. Maybe their 12 “non-creoles” (or many other languages that they did not look at) are creoles as well – unless we have a definition of “creole”, we cannot tell. Confusingly, Bakker at al. refuse to give a definition:
“Creoles are sometimes characterized on the basis of socio-historical circumstances…, or on the basis of structural features, i.e. a language which, after having undergone restructuring from a pre-existing language, now has e.g. SVO word order, preverbal TMA marking and little or no morphology.”
In practice, the authors accept a language as a creole if it is part of the usual canon of creole studies, but arguably this canon is to a large extent defined by the structural features. This means that the argumentation becomes circular. But even if we admitted only those creoles that we know from historical records underwent social creolization (as defined above), we could still not be sure that we have a representative sample of languages. There may well be languages that underwent social creolization but have little signs of structural simplification. For example, Spanish on Cuba arguably has a similar history as African American English, but there is no restructured Spanish spoken on Cuba (see, e.g, McWhorter 2000 for discussion). But by a social definition, Cuban Spanish would probably count as a creole, and thus be a counterexample to the generalizations that Bakker et al. are trying to make. How many other counterexamples might there be out there?
Traditionally, linguists have classified languages typologically by their structural features – but one might classify them by their social features, or their social history as well. This has hardly been done. There is no “world atlas of sociolinguistic features”, and there is no good typology of social contact types underlying contact-induced change. But these are the kinds of data we know to answer the really interesting questions about the connections between social situations and the outcomes of language contact and change.
Instead, creolists seem to be stuck with the question of (synchronic) creole distinctiveness. But this is not really an interesting question, because there is no theory that could explain it, and anyone would go to diachrony for explanations anyway. Moreover, creole distinctiveness has given rise to acrimonious ideological debates, as well as to misunderstandings. For example, should one go as far as to say that a language that shares the typical structural features of the classical creoles is a creole, even if we know nothing about its history (cf. McWhorter 2005)? But how can one then avoid circularity? And if creoles are non-distinctive, does this automatically imply that they are not separate languages, but daughters (or even dialects) of their lexifiers? Bakker at al., citing Chaudenson and Mufwene, mention this as a relevant issue, but it seems that it is totally separate from the question of typological distinctiveness.
Thus, I would think that we should not be debating synchronic creole distinctiveness, but ways of predicting the outcomes of language change on the basis of social situations (along the lines of Trudgill 2011).
Bakker, Peter, Aymeric Daval-Markussen, Mikael Parkvall & Ingo Plag. 2011. Creoles are typologically distinct from non-creoles. Journal of Pidgin and Creole Languages 26(1). 5–42.
Holm, John, & Peter Patrick (eds.). 2007. Comparative creole syntax. Parallel outlines of 18 creole grammars. (Westminster Creolistics Series 7). London: Battlebridge.
McWhorter, John H. 2000. The missing Spanish creoles : recovering the birth of plantation contact languages. Berkeley: University of California Press.
McWhorter, John H. 2005. Defining creole. Oxford: Oxford University Press.
Parkvall, Mikael. 2008. The simplicity of creoles in a cross-linguistic perspective. In Matti Miestamo, Kaius Sinnemäki, & Fred Karlsson (eds.), Language complexity: typology, contact, change, 265–285. Amsterdam: Benjamins.
Trudgill, Peter. 2011. Sociolinguistic typology: social determinants of linguistic complexity. Oxford: Oxford University Press.
It seems to me that circularity can be avoided if we compare and contrast creoles with their lexifiers and their lexifiers’ language family. Thus, a long time ago, when I was young, silly and thought facts alone could convince people (okay, painfully naive, not just silly) I pointed out on an on-line forum, where certain nameless individuals had stressed that linguistically creoles were just like other languages, that in fact that was not true.
I pointed out that the various Romance creoles lacked grammatical gender (defined the way WALS defines grammatical gender), and that *within the Romance family* this was anomalous. Outside of pidgins and creoles all known Romance languages and dialects, without exception, have a grammatical gender system inherited from Latin. Tellingly, non-creole/non-pidgin Romance languages/dialects in heavy contact situations likewise preserve grammatical gender.
Even more tellingly, Romance pidgins are the only non-creole Romance variety which likewise lack grammatical gender. Indeed they lack a great many things which creoles also lack but which outside of Romance pidgins and creoles are Pan-Romance. I therefore believed (and still do) that we may reasonably conclude that Romance creoles are nativized pidgins. This is because when contrasting pidgins and non-creole Romance varieties typologically Romance creoles quite consistently are pidgin- rather than (non-creole) Romance-like.
Tellingly, in the few rare cases where Romance creoles differ from pidgins and are more non-creole Romance-like, this similarity *always* involves features which, over the short lifespan of the existence of creoles, could easily be accounted for through contact-induced change, AKA decreolization (this phenomenon having been observed in the few instances where we’ve enough written data to follow the history of a creole, as well as being deducible from the comparative structure of isolated versus less isolated forms of an earlier creole, there is no circularity whatsoever here).
The total number of individuals I have managed to convince throught this line of reasoning, on-line or at talks, was zero, much to my frustration and puzzlement at the time. But I believe that the basic method remains valid. If a language (A) has a vocabulary in common with/deriving from B but otherwise differs from B in such a (non-trivial) way that A shares features with pidgins rather than with B, this is in and of itself not enough to claim that A is a creole.
But if B is a language whose history is known, which is a member of a language family and of a language area, both with a well-known history, with clear instances of language contact, and if the pidgin-like features of A are otherwise wholly alien to said language family/area, and if the non-pidgin-like features of A can be accounted for in a non-tautological fashion, then I believe that claiming that language A is a creole, and therefore postulating that it is a nativized pidgin, is justified and heuristically useful.
That in the real world some doubt will remain as whether some suspected creoles in fact are creoles is indubitable. Quite apart from the fact that many language families and language areas are poorly known, it is quite possible that in many parts of the world pidginization cannot be expected to yield a sufficiently distinctive outcome for us to tell apart possible creoles from non-creoles.
But there are clear-cut instances: and from this point of view the pidgin past (i.e. creole status) of the “canonical” Romance and Germanic creoles is to my mind proven quite beyond any reasonable doubt.
I don’t agree that it’s only a question of how hard the work is. Sure, we do know something about how various sociolinguistic factors influence choices and attitudes today, but we have a lot less to go on when it comes to events that took place several centuries ago. And the circumstances surrounding the birth of these languages were rather different from what sociolinguists have studied in modern times (slavery just being one such thing).
After all, had the social circumstances not been extreme, we wouldn’t have got languages which… Oops, circularity agian.
Maybe we’re talking about different things, and maybe you and Martin aren’t referring to a foolproof formula, but perhaps you are admitting speculation and extrapolation about the bad old days based on what we know today. If you admit that, then I think that is pretty much what I am doing, although the stuff referred to on this page doesn’t reflect it.
And should we not admit speculations on olden days based on the modern world, then we’d have to resctrict ourselves to creole-looking languages which are forming today, or which have emerged in the past couple of decades, and thereby, our sample would shrink to (to the best of my knowledge) pretty much zero. (Nicaraguan Sign being the case that I can think of where birth from a pidgin can be or has been invoked).
Understandable points, but a clarification is necessary. I have not stated “All creoles have these features, and therefore if we find a language with these features it’s a creole too.” Obviously one objects that we haven’t defined what a “creole” is — circular indeed. However, if the specification has been that certain languages which we KNOW were born amidst certain types of conditions all seem to combine certain features, then surely it is a reasonable proposal that those conditions have led to a type of language — exactly the approach suggested in this post, and that Trudgill (2011) nicely demonstrates for what I clumsily title “NCSL” languages. It’s not clear to me why that approach can’t be extended successfully to creoles. More specifically, in 2005 I suggested that IF we agree that certain conditions led to a certain kind of combination of features, then this could be useful in suggesting those kinds of social conditions as the source of that combination of features in languages we don’t know the history of. That is, I was suggesting that this paradigm could be useful, not just for rather sterile goals of museum-style taxonomy, but to chart the history of unwritten languages. I find this approach useful in, say, Indonesia.
Mikael, I think it is indeed possible to “quantify and operationalise this”; it’s just that it’s jolly hard work that we are only beginning to do (eg. in the APiCS database). It’s not just a matter of “people’s feelings and wishes”; there are also lots of objective and readily quantifiable dimensions that can be measured, like who speaks what when, how language choice relates to various sociological parameters such as income, political power, etc. etc., and then trying to plot these dimensions over time, in order to get at the history of the language. Such a research program is feasible, non-circular, and would provide answers to questions concerning the relationship between socio-history and grammatical structure. But people such as Peter Trudgill are beginning to address these questions, and the APiCS database provides a foundation for further work in this direction.
Circularity is indeed a problem to anybody taking part in the debate. But I’m not sure your “social creolization” label helps much, since its definition includes “a particular type of restructuring”, which basically points back to the “creole canon” in any case…
I partly disagree with the claim that “no theory could explain it” since “if a language looks that way, it has once (recently) been a pidgin” is such an explanation. I suspect, though, that you mean that no theory predicts when a pidgin will develop and when it won’t (as in your Cuban example). Or if a pidgin does arise, we can’t predict wether it will nativise or not. So by invoking pidginisation, I could be said to just shift the problem to the next level, without actually solving it. That’s a fair criticism, though I would say that recognising the pidgin past is a necessary first step towards understanding.
As for myself, I believe that the two main factor behind piginisation is suddenness (you don’t have the time to learn the other party’s language, because the communication problem is too urgent) and social barriers (you are not interested in becoming part of the other group in the first place, or you are not welcome among them). In both cases, the costs of learning another language outweights the perceived benefits, and a pidgin gives you plenty of bang for the bucks by being easily acquired.
We would of course like to quantify and operationalise this, but I am aware of no method of doing so, since people’s feelings and wishes don’t easily lend themselves to that (as the consistently recurrent failure of economic theories show). Human behaviour is simply too tricky. Even with a gazillion times more data than we currently have, we’d still be stuck with tendencies rather than infallible formulas.
I’m not quite sure that I get the “anyone would go to diachrony for explanations anyway” part, though — since when is the concept of “diachrony” incompatible with that of “theory”? And if it is, how many other characteristics of languages as found in e.g. WALS can be explained by a “theory” that makes no reference to diachrony? (Yes, I am aware of a few, but that is very much the exception rather than the rule). To me, statistical statements of the type “SOV languages tend to have postpositions” are interesting, but don’t constitute a “theory”. Maybe you’re saying something immensely profound here that I just don’t manage to wrap my brains around.
But at the end of the day, I agree about the circularity problem.
Good points. But I read your post as being less of a critique of the creole distinctiveness hypothesis and more a critique of the notion of creole itself, and the way in which the “creole” label is (mis)attributed to various languages. And I think you and your APiCS coeditors present the correct route forward when you characterize the languages included in the volume/database not in terms of labels such as “creole” and “pidgin” but rather in terms of specific objective historical/sociolinguistic features. Still, once languages are characterized in terms of such “extrinsic” features, it does make sense to attempt to correlate such features with other “intrinsic” features pertaining to grammatical structure, as you are indeed suggesting.