Language

This is and is not a new book. It is not new since nearly all chapters are reprints of already published articles. It is new insofar as those parts that are not reprints, in particular the general introduction and Ch. 1—already published to be sure, but in 2011, so almost first-hand—are witness to serious inflections in the thought of one of today's leading creolists.

Since creolists, nay scholars generally, who admit to their not having been quite right are something of a rarity, I devote most of the present review to this evolution in McWhorter's view of what kind of languages creoles are and how they come to be. This must indeed be examined at length since it is what gives coherence to a string of essays that would otherwise look more like a miscellany than a collection.

In 1998, M published an article in Language (later reprinted in McWhorter 2005) where he proposed that creole languages could infallibly be identified by searching them for the following three features (the 'creole litmus test'): (i) 'little or no inflectional affixation', (ii) 'little or no use of tone to distinguish monosyllabic lexical items or morphosyntactic categories', and (iii) 'little or no noncompositional combinations of derivational morphemes with roots' (39). Together, the three features make up the 'creole prototype'.

This proposal met with ample discussion and sharp criticism. Part of the criticism was ill-founded as it stemmed from failure to understand the gist of M's hypothesis. It is futile, for instance, to raise the objection that many languages one cannot suspect of creoleness show one or the other of the three features. Indeed, what the creole prototype hypothesis implies is that no language that is not a creole will show all three features TOGETHER. [End Page 657]

Yet, the objection is so often raised that M takes great pains to demonstrate that languages may show features (i) and (ii), but fail the creole test on feature (iii). He illustrates with the Papuan language Abun (7-10). M could not fail to realize, however, that such a drawn out misunderstanding might be due to the fact that his original three criteria were in need of some qualifications, which he provides in Ch. 1.

The main problem with the criteria, however, lies in the 'little or no' quantification. To be sure, it reflects M's basic honesty. Being human devices, phenomenal languages are not perfect systems— whatever they may be 'essentially'. Categorical assertions may therefore be factually true for easily individuated features of particular languages—for example, Modern English has no /y/ in its phonemic inventory—but they are bound to be ridden with exceptions at any more general level. M is well aware of this. He knows that, had he bluntly written that creole languages have 'no inflectional affixation', disclaimers would instantly have rained down upon his head.

The 'little' nevertheless poses a serious problem, since it cannot but raise the question of how little should little be not to become (too) much, and doesn't M's theory run the risk of unfalsifiability if no precise upper bound can be fixed—as none can be in all scientific honesty? The only thing we can reasonably be sure of is that the amount must be small: no known creole language has five or more inflectional affixes in its verbal system, for instance. Hence, M's 'little'.

What this argument actually points to, I think, is that the creole prototype more readily tells us which languages certainly are NOT creoles rather than which ones certainly are. In other words, it tells us that Classical Arabic or Mazatec cannot be creolized languages, while leaving us in doubt—if sticking to purely formal criteria—with Abun or Guinea-Bissau Kriyol.

This perceived difficulty, inherent to all typological endeavors given the untidiness of linguistic evidence, is probably what led M to significantly change his tack. From the start, the creole prototype is related to a notion of grammatical simplicity/complexity, an issue that has figured prominently on the linguistic agenda for quite a few years now (see e.g. Dahl 2004, Hawkins 2004).

Discussions of the issue often conceal an ambiguity: simplicity/complexity for whom? The language user or the language descriptor? We cannot assume both perspectives to be identical, given the ease with which children learn and adults practice languages whose morphosyntactic structures baffle our linguists' minds. By contrast, if a language's morphosyntax is easy to describe— because paradigms, rules, and so on are few and/or transparent—there is no reason to suppose it should be more difficult to acquire as an L1 or an L2. In fact, it should be even easier. Both perspectives may therefore coincide at the bottom of the complexity scale, even though they probably do not at the top.

The issue is then not so much whether creole grammars (i.e. morphosyntaxes) really are 'the world's simplest grammars'. What they certainly always are is SIMPLER than the grammars of their lexifiers. And since the latter (even English!) include a goodish amount of inflectional affixation and much noncompositional derivation, being significantly simpler means showing much less of each, that is, 'little or no'.¹ Moreover, given the asymmetric relation between both views of simplicity, such relative simplification has to mean something in terms of the actual processes of creole emergence.

The problem with the original creole prototype was that no real explanation was given of why creole languages should present precisely those three negative traits and in what sense this makes them simpler. Notice that they are all MORPHOLOGICAL traits—as if syntax was not relevant to the simplicity/complexity metrics, which may well be true, but needs to be given some thought.

M's initial attempt at an account was to point out that creoles are young languages, which they undoubtedly are. This amounts to assuming that morphological complexity is like wrinkles: it comes with age, thereby implying that morphological simplicity represents the primeval state of language—for which there is absolutely no evidence. Youth is certainly a factor, but it does not give us a general explanatory principle. [End Page 658]

M's decisive move was to connect his creole prototype to the growing line of research that views adult unguided second language acquisition (SLA) and the formation of interlanguages and 'basic varieties' as the crucial factor in creole emergence (e.g. Klein & Perdue 1997, Plag 2008, Siegel 2008). This led him to a significant reassessment of the whole process. Contrary to what his previous views seemed to imply, he now holds that 'The normal state of language is highly complex' (1) and that 'Languages significantly low in this kind of complexity...owe this state to second-language acquisition in the past' (2).

Two types of languages may thus be distinguished. On the one hand, there are languages that have continuously been transmitted as L1 as far back as recorded time goes. Icelandic seems a good example. On the other hand, there are languages whose history includes one or several episodes when they were acquired as L2 by enough incoming people for the SLA process to affect the target. The obvious example here is English as compared with the other Germanic languages.

English is not a creole, but it is what M calls a 'nonhybrid conventionalized second language' (NCSL). Although he never overtly justifies his choice of terms, we may surmise that 'nonhybrid' means the language continues to belong to the same stock (English is still Germanic), whereas 'conventionalized' probably implies that the changed result of the L2 episode becomes the vernacular of the (enlarged) linguistic community. The term is not really transparent and I am not sure what a hybrid language would be: if family membership is decided by looking at the lexicon— following the time-honored and, I believe, only secure practice—then Portuguese-based creoles, for instance, are obviously Romance languages, but that does not tell us anything interesting about them. We (including, I believe, the author himself) would therefore like to find another name for NCSLs that would be both more adequate and, hopefully, less of a mouthful. I could not come up with anything, however, so I will stick to M's term that has at least the merit of referring to a well-defined entity.

Creoles can now be viewed as extreme instances of NCSLs because they were 'born in a situation in which adult acquisition was universal or nearly so' (3). This should not be seen as a revival of the 'life-cycle' hypothesis. It is a fact that not all creoles can be traced back to a STABLE pidgin. Yet, all creoles are grounded in adult unguided SLA, a crucial episode, however short its duration, because they simply would not exist without it, whereas there would still be English without the Danes and the Normans, only a different English.

Only from this perspective do the features of the creole prototype constitute a rational and necessary set rather than a haphazard collection. The basic fact is that adults struggling with an L2 'on the job' learn isolated words that they cannot organize into paradigms until they achieve a sufficient level of proficiency—which they never do in creole-forming situations. The almost complete loss of affixational inflection results precisely from this failure to acquire synthetic paradigms. Absence of noncompositional derivation also results from acquiring only isolated, unanalyzable forms (98-101). Of course, the ensuing creoles then develop new paradigms, of the analytic type, as well as a measure of noncompositionality. Also notice that such a notion of the process seems to imply an 'abstractivist' model of morphological phenomena (Blevins 2006).

I unfortunately have no space left to comment on M's careful analyses of particular creole facts supporting his assumptions, or on what he has to say about Riau Indonesian, Homo floriensis, and the Celtic hypothesis for the history of English. Convinced or not, the reader will enjoy these well-argued, always interesting, often intriguing pieces of scholarship.

Hugo Schuchardt's dictum that 'all languages are mixed languages' remains in full force. Mixing means contact, which implies SLA. Yet, there is an empirical difference between languages where the scars left by unrecorded 'accidents' (actually ordinary events) have been smoothed away by the passage of time, which show the 'normal' level of complexity, and languages where the 'accident' (contact-cum-SLA) is still recent and recoverable, which show an inordinate level of simplicity, with creoles the extreme case. M's notion of NCSL, although not fully new in its content, greatly helps us to formulate and exploit this difference. This is, I think, the main contribution of his book. [End Page 659]