Martin Haspelmath posted a blog text a few days ago in which he mentioned the interesting question whether one can define creoles on structural grounds, independently of what we know about social factors. And he mentioned the important question of circularity.
The idea that one has to start with social factors and then look at the linguistic outcomes is very important – I have in fact always emphasized that for pidgins, for instance in my 2008 article on pidgins. This should also be the case for creole studies.
In contrast to other linguists, who often try to generalize over both pidgins and creoles, I am personally interested in where they differ. When you want to find out what the typological properties of pidgins are, you need to define pidgins first on the basis of sociohistorical facts, along the lines of that a pidgin is a language with norms that came into being in a situation where people who had no language in common, who needed and wanted to communicate and created a simplified version of one (or more) of the languages in contact. Simplification is a necessary part of the definition: it is from a communicative viewpoint quite smart to strip a language of its unnecessary components in a pidgin-generating situation. This special situation leading to significant simplification must be part of a definition of a pidgin.
What about creoles? Is it possible to isolate a set of sociohistorical situations that can be considered necessary for creolization? Are creoles solely defined by their socio-history, or do specific socio-historical events lead to languages with certain structural/typological properties? Here some circularity is unavoidable, unfortunately, even more so than with pidgins.
Yes, there is a risk of circular reasoning when one says that language X is a creole just because creolists call them creoles. The fact that Cubans speak something close to what is spoken in parts of the Iberian peninsula, with no symptoms of reduction and expansion, is interesting, but apparently some social factors were different there. On the other hand, we have good reasons NOT call Cuban Spanish a creole from a linguistic point of view.
Still, there is a group of languages that linguists, including creolists, call creoles – in many cases because the speakers call the language something like kriol, criollo, creole, cariolisch, krio, krioro, crioulo – but that is not a watertight criterion, as not all creole languages are so called by their speakers, and perhaps there are languages called “creole” which do not qualify as such (Copper Island Aleut?).
In our 2011 article (Bakker et al. 2011) we did not ourselves select the creoles for the occasion: one selection was taken from Holm & Patrick (2007), which contains at least two languages whose creole status has been put under discussion at some stage, Berbice Dutch and Nagamese. Still, these clearly group with the other creoles and against the non-creoles in the test you refer to (I admit here that that was against my own expectations).
But what Martin does not mention is that we also used the features set up by respected typologists such as Haspelmath, Gil, Dryer and Comrie: also using the WALS features, the creoles are clearly set apart from the non-creoles, with no creole language outside the cluster and only one of more than 150 languages of the world appeared inside the creole cluster of the network. Non-creoles that are sometimes compared with creoles (West African languages, Sinitic languages) are found almost evenly spread among the non-creoles: in the table on p. 44 they are marked with stars, and include Igbo and Yoruba.
(in this graph, all pidgins are located on the top between the dotted line; languages spoken by the creators of the creoles are indicated with stars (substrates) or ellipses (lexifiers))
All our data are online, open for control and scrutiny.
The conclusion is robust and unavoidable: there is a group of languages called creoles by linguists and these share a certain set of structural properties that set them apart from all other languages. Which typological properties creoles share, is surprisingly difficult to define. Aymeric Daval-Markussen and I have a paper in press in which we provide an overview of almost all the properties that have been proposed in the literature in the past half century, as being somehow “typical” for creoles. Strikingly, there was hardly any overlap between the different authors. In addition, many proposed features are only found in a minority of creoles. Creoles are clearly quite diverse from a typological perspective. Still, whatever test or sample you use, the creoles stand out from the non-creoles. As Aymeric Daval-Markussen has shown, as few as two properties can be sufficient to distinguish creoles from (almost all) non-creoles: tense is not expressed with inflectional means and the indefinite article is derived from the numeral “one”.
For me, a creole is a language lexically related to another language, which was first simplified for communicative reasons (one can call it a pidgin stage), and then made into complete and full languages by developing necessary grammatical distinctions not found in the simplified intermediary language. Even if the intermediate pidgin is not documented, there is ample indirect evidence, such as the absence of gender and inflection, present in all lexifiers and their overseas varieties except for those called creoles. Baker (2002) has identified pidgin features in creoles.
I have challenged the audience several times at conferences, and now it is time to repeat it in public to the whole world. The first time the price was 50 euros. Thus far I have increased the stakes with 50 euros each time, so I set it at 250 euros now. It is not enough for a research grant, but I pay it out of my own pocket. How can you win?
If someone can provide a result in which a sample of 12 creoles and 12 non-creoles are selected, I bet that it is IMPOSSIBLE to get a result in which the creoles and non-creoles are blended when 12 typological features are chosen from a broader domain of typological features (e.g. not 12 features relating to numerals). The first one who is able to do so will win my 250 euros. To do this, you have to know which languages are creoles.
Which languages are creoles? Not all languages in APICS are creoles and pidgins.
The following five languages in APICS are intertwined and mixed languages, and not creoles or pidgins: Michif, Ma’a/Mbugu, Media Lengua, Gurindji Kriol and, in a different way, Sri Lanka Malay. None of these underwent a stage of simplification and then expansion. The following five languages have undergone some simplification but not much expansion to speak of: Singlish (English, Chinese), Lingala (Bantu), Kituba (Bantu), Afrikaans (Dutch), Ambon Malay (Malay). Do not take any of those languages in your sample. And some nine of the APICS languages are pidgins, not creoles, and you cannot take those either, as we deal with creoles: Fanakalo, Chinese Pidgin English, Pidgin Hindustani, Chinese Pidgin Russian, Eskimo Pidgin, Chinuk Wawa (except from Grand Ronde), Yimas-Arafundi Pidgin, Bazaar Malay, Pidgin Hawaiian. That leaves at least 52 languages of the 76 APICS languages to choose freely from. From the 7000 other languages of the world, choose 12 from 12 different families, not 12 Kwa or Turkic or Sinitic languages. Thus, I still do not define creoles, of course, but at least I list them here.
The winner will be announced here. If there ever will be one.
Baker, Philip. 2002. No Creolisation without prior pidginisation. Te Reo 44: 31–50.
Bakker, Peter. 2008. Pidgins versus Creoles and Pidgincreoles. In: Handbook of Pidgin and Creole Studies, ed. by Silvia Kouwenberg & John Singler. Oxford: Wiley-Blackwell. 130-157.
Daval-Markussen, Aymeric. 2011. Of networks and trees. New light on the typology of creoles. Thesis, Aarhus University.
Daval-Markussen, Aymeric and Bakker, Peter. 2012. Explorations in creole research with phylogenetic tools. In Proceedings of the EACL 2012 Joint Workshop of LINGVIS & UNCLH, April 2012, Avignon, France. Association for Computational Linguistics, 89–97.
The crux is that the advocates of the position that Peter Bakker represents indulge in circular reasoning, even if this is denied. Bakker writes: “Yes, there is a risk of circular reasoning when one says that language X is a creole just because creolists call them creoles. The fact that Cubans speak something close to what is spoken in parts of the Iberian peninsula, with no symptoms of reduction and expansion, is interesting, but apparently some social factors were different there. On the other hand, we have good reasons NOT call Cuban Spanish a creole from a linguistic point of view.” For many languages labeled creoles as well as for many languages labeled non-creoles we know very little if anything about their genesis. If Cuban Spanish had had the requisite (Papiamentu-like) linguistic creole features, its social history would have been argued in hindsight to be of the type apparently needed to foster a creole language. The only way to be admitted to the “creole club” in linguistics is by possessing certain structural features, while being lexically related to a language not possessing those same features. Being a “creole language” is a relational notion, in this sense.
I received one submission today. We have not had time to look at it in detail, but it could be that there is a winner.
I should have given an email address: linpb at hum.au.dk