What relevance do the above observations have to the theory of decreolization? Decreolization is hypothesized form of language shift in which parts of the community that formerly spoke a basilect, gradually acquire the forms of the acrolect. In complete decreolization, the basilect gradually disappears, and the population eventually comes to speak the standard language.
However, if mergers occurred in the process of creole formation, then direct descendants of the resulting merged dialect could never unmerge those sounds. If truly complete decreolization in fact occurred, then mergers would have to be reversed, so that the decreolized dialect is just like the (never-merged) standard. Otherwise there would remain differences between the speech descending from the creole and the standard to which it assimilates. Can we reconcile a view of creole development with the impossibility of unmergers? This discussion is inconclusive, with respect to Jamaican Creole, but it proposes a rule of inference from which, once the linguistic facts are clarified, the historical facts which resulted in them may be inferred.
Until recently, a commonly accepted view of the process of creole formation and development was as follows: At the time of the origin of the creole, a linguistic cataclysm occurred, and the deepest creole was formed. At this time the great majority of people in the community were speakers primarily of this basilect. Then in a lengthy historical process of assimilation, a range of intermediate lects emerged which are gradually more and more like the locally relevant standard language, until the continuum of lects emerged that is found today.
Consider the facts of Jamaican Creole, in this context. Notice that a number of lexical distributions are lost in the basilect. As discussed in Chapter 6, vowel of Reference American /ir, er/ (as in NEAR, SQUARE) are merged into the JC /ir/ class; /ur, or/ (as in CURE, FORCE) merged into JC /ur/; /ay, oy/ (as in PRICE and CHOICE) are merged into the /ai/ class; /æ, / (as in TRAP and LOT) are merged into the /a/ class, /:/ (THOUGHT), /a:/ (BATH, PALM), /ar/ (START), /r/ (NORTH, but not FORCE)/ are all merged into the /a:/ class. These mergers presumably occurred at the earliest stage in the development of the creole; they are characteristic of the basilect.
Mergers are irreversible. Therefore none of these mergers could be unmerged in any dialect descended from the original basilect. If acrolectal Jamaican Creole now has some of these distinctions, with the right words in the right sound-classes, then how is this compatible with the irreversibility of merger? Either decreolization goes, or irreversibility goes. We can't have both, in this absolute form. If we maintain the principle of irreversibility of mergers, we must instead propose a different historical process. I therefore propose the following rule of inference for creole studies:
If historically accurate word classes are kept separate at one level of the creole continuum, but are merged at another level, then the level that which keeps them separate must have existed continuously throughout the history of the community.
If the acrolect has become so approximated to the standard, to the point that vowel classes are unmerged, then the reality must be that the unmerged acrolect has been there since the beginning.
Thus our understanding of the history of the acrolect depends on the facts of lexical distributions of sounds. Do any such apparent unmergers actually exist in situations of alleged decreolization? The essential question is the lexical distribution of sounds in the creole. J.C. Wells (1982:576) says that there are low-back vowels in acrolectal speech which can be pronounced in the formerly back-rounded word classes of LOT, THOUGHT, etc. But these classes are said to be fully merged in the basilect with word classes that don't allow the back-rounded vowel quality. I have difficulty believing that Wells is right, that the acrolectal speakers get their //'s in the right words, and don't put them in the wrong words. Unfortunately this matter will not be settled in this work. If Wells is correct, then the consequence is that the acrolect must have a long history, at least as long as the history of the community which now speaks it. This supports a much more finely graded view of the sociolinguistic situation in the early creole: the acrolect, without the basilect's mergers, must have been present from the beginning.