The phonology of Jamaican Creole vowels has been studied by several, including Cassidy, LePage, DeCamp, and Wells.6.4 My main source for the following phonological analysis is Wells, who included a short summary of JC phonology in his 3-volumn Accents of English (1982). Let us review the set of phonological contrasts. Jamaican Creole English does not distinguish the following groups of lexical sets, according the description in Wells (1982):6.5
The phonological inventory (of stressed vowels) may be structured using 3 heights, backness, and length, plus 5 underlying diphthongs, as in Table The symbols are Cassidy's.6.6
A distinctive feature of the Jamaican vowel system is the relationship between Middle English //i://, //u://, (corresponding generally to the lexical sets PRICE and MOUTH, Reference American /ay, aw/, and JC /ai, ou/). In some English dialects the nuclei of these vowels have fallen together, but in Jamaican, the nucleus of ME //u:// has not fallen as to low position, and these nuclei remain distinct, as shown by the phonetic measurements on page , and as symbolized in Table by the mid-back form, /ou/.
It should be pointed out that the apparent lack of distinction between the lexical sets PRICE and CHOICE is false, since the CHOICE class in some cases has an onglide /w/, so that by and boy, for example, are distinguished as /bai/ and /bwai/.6.7 However I follows the traditional analysis in which the nuclei and offglides of the two sound classes are considered to be the same, and the distinction between them is no longer in the rhyme of the syllable.
JC /o/, which corresponds to Reference American // (lexical set STRUT), is a mid-back short vowel. This phonetic form is appropriate to the phonological features assigned to this vowel category by the analysis in Phonological Preliminaries. There, // was put into the non-front, mid, short slot; it is phonetically central rather than back in RA. In JC, the corresponding phoneme is a back vowel, which may be taken as support for the earlier analysis.
The structural analysis represented by this chart is simpler and more abstract than the traditional analysis of Cassidy (1967), in which more post-nuclear glides are posited. He transcribes the long-mid vowels in a way close to the phonetic surface (for stressed vowels), namely as /ie, uo/ (rather than /ee, oo/, or /e:, o:/ for example, as per Wells), which are high vowels with downward glides (that is, inglides). The charts below use Cassidy's transcription. In Table , these vowels are located in the gaps in the long-vowel subsystem, and thus may be more abstractly written as /e:, o:/. The more abstract system can generate Cassidy's system by an application of V/_V Raising, which raises the nuclei (but not the glides) of the long mid vowels, as discussed below.
The long vowels, both mid and high, are phonetically raised and shifted to the periphery relative to the corresponding short vowels. (This is shown by the mean phonetic locations of these classes in Figures and , pages and .) /uu/ is more backed than raised relative to /u/, while /ii/ is more raised than fronted relative to /i/, but this basic generalization still applies to the high vowels, as well as to the mid vowels. This would appear to be another instance of Vowel-before-Vowel Raising.
The mid vowel, /e:/ (here labelled ``ie''), is raised as high as /ii/ for Roasta and nearly as high as /ii/ for Juba. This justifies the ``i'' in Cassidy's ``ie'' transcription. /o:/ is labelled ``uo'' for the same reason: its nucleus is raised towards [u], and it has a lowering offglide.6.8 The presence of the offglide and the raised nucleus with the long, phonologically mid vowels provides an opportunity to apply the phonological representations developed in Phonological Preliminaries. If these long mid vowels are underlyingly represented in the same way as the long mid vowels of Reference American, then a plausible derivation from underlying to surface form is in Figure .
The phonological analysis in Figure raises two important technical issues. First, notice that the feature [hi(gh)] after step 3 is not really [high]; rule (3) merely adds a degree of phonetic height, which is not sufficient to reach the full phonetic height of the long high vowels, particularly in the case of /uo/, as shown in Figures , , pages , page. The raising is not really the addition of the feature [high] to a discrete feature specification, since phonologically [high] vowels are actually even more raised. So the issue is, What categories does this raising rule operate over?
There appears to be a gradient dimension of phonetic height, since there are a fairly large number of different heights distinguishing raised and non-raised vowels. The back vowels may be arranged in a scale from high to low according to the mean F2 for stressed tokens of the vowel, thus: /uu, uo, ou, o, aa/. The front vowels may be arranged in a similar scale, /ii, ie, i, e, a, aa/. The five or so heights phonetically necessary here may be distinguished by using a small-n numerical system of height specification such as Clements', Schane's, or Ladefoged's (see Phonological Preliminaries), or may be represented by a more continuous gradient, as assumed in Labov, Yaeger and Steiner (1972). Since the phonetic height represented in the chart of acoustic measurements must be derived at some point, we may economize on intermediate steps in the derivation by considering this a phonetic raising rule that operates on a continuous gradient. The continuous gradient must be brought in at some point anyway, since it is necessary to account for the acoustic observations and the articulatory differences that they reflect. If we account for the phonetic raising of long non-low vowels by applying the rule used in step (3) above at this low level in the derivation of phonetic forms, we can avoid postulating an additional, unnecessary level of phonetic derivation.
In fact this general line of argument suggests that small-n numerical height specifications are to be avoided if possible, in favor of phonetic rules that interpret phonological categories in terms of continuous, gradient phonetic dimensions. The precise form of such rules is yet to be ascertained, but in this case, their effect is clear: the nuclei of the long, mid vowels are raised to some degree; the nuclei of short vowels are lowered by some scalar value.
A second technical question is, How is the natural class of mid vowels to be specified in the specification of rule (3)? They are neither high nor low, so we may say positively that the height tier6.9 lacks specification, and thereby pick out the mid vowels. Since this raising rule applies to both front and back mid vowels, I don't specify the [front] feature. But the lack of specification may have two quite different meanings, as discussed in Chapter 3, page . The lack of specification of the [front] feature is used to avoid all values on a particular dimension, namely both [front] and non-[front]. The lack of specification of [high] or [low], on the other hand, is used to pick out a particular point on the height dimension, namely the height which is neither [high] nor [low]. There is no formal difference between not specifying a feature in order to generalize over the values of the feature, and not specifying a feature in order to pick out the unmarked value on that dimension. This is a general problem with privative features in phonology.
In the current case, we're lucky, and there is no conflict between these two significances, but such occasions could conceivably arise, posing severe difficulties to theories relying on underspecification, without some solution like that given in Chapter 3.
Despite the improvements in structural symmetry and simplicity, and the elimination of gaps gained by analysing Cassidy's /ie, uo/ vowels as long mid vowels, the base-6 system represented above has two major faults: it makes a fundamental distinction between low-front and low-back vowels that may be unnecessary, and it still contains a great many gaps among the gliding vowel subsystems.
As in the phonology of Reference American vowels, some of the central problems in characterizing Jamaican Creole vowel structure are the low vowels and the glides. The various low vowels classes of other dialects are collapsed in JC into just two: long and short a, variously written as /aa, a/ or /a:, a/. The lexical sets which make up these phonemes are listed above in Figure . However, this picture holds two mysteries: the low-back vowels, and /ar/.
The low-back vowels, marked with 35#35 in Table and Figure above, have an uncertain phonological and historical status. The phonemes /a, a:/, when they occur in words of the historically back-round lexical sets, LOT, THOUGHT, CLOTH, and NORTH, are said to be optionally backed and rounded in acrolectal Jamaican Creole English. However, this backing is said to be impossible for /a, a:/ in words which historically were not back and round, namely PALM, BATH, and TRAP words. This poses a paradox, since this situation is not compatible with the following widely accepted assumptions:
1: Low-back and low-front vowels are merged in the basilectal creole.
2: The more acrolectal (RP-like) levels of the creole are historically descended from the earlier basilect through a process of decreolization.
3: Merger is irreversible.
Given these conditions, it is impossible that the merged front and back low vowels should somehow be unmerged in the acro- and mesolects of JC. These assumptions are discussed extensively in Appendix 2. I will assume, for now, that the accepted assumptions 1, 2, and 3 are indeed true, and instead that the claim that /, :/ are phonologically distinct from /a, a:/ is false.
The status of the phoneme /ar/ (with the START and NORTH lexical sets) is also something of a mystery. This vowel is long and has the same nucleus as /a:/. Commonly there is no rhotic glide pronounced with these words, so that it seems that they should be classified with the /a:/ phoneme. However, words in the START and NORTH classes are occasionally pronounced with a rhotic glide, and the other classes are almost never pronounced with a rhotic glide (Patrick 1991). Thus it would appear that /ar/ is a separate phonological category, which is frequently neutralized with /a:/.
A well-designed phonological system, both from the perspective of the linguistic analyst, and from the perspective of the living, changing language itself, is symmetrical and gap free. Gaps in phonological structures count as flaws in two senses. First, they may be real and then often result in historical splits, mergers, or sound shifts, by which the gaps are eliminated. Second, gaps may be due to faulty linguistic analysis. If the system is historically stable, as Jamaican seems to be, then attention may be turned to the analysis.
Notice that there are three kind of glides (/i, u, r/), but only 5 gliding vowel phonemes (excluding /ie, uo/). The situation is somewhat similar to the table from Trager & Bloch (1942), discussed in Phonological Preliminaries, where gaps and fillers correspond with each other across subsystems. We might reduce the Nucleus-Glide sequences from two or three paradigms with as many as 13 gaps to just one, with no gaps, thus:
This rearrangement considerably simplifies the phonological structure at the cost of additional rules. The following four rules will do the necessary rearrangements. They use binary features [15#15high], [15#15low], [15#15rhotic], and [15#15front], labelled with ``N'' for features that occur in the Nucleus slot in the syllable, and with ``G'' for features in the Glide position. Here I assume Glide segments in the N+Glide subsystem are specified as [-high] at the point in the derivation where these rules apply.
1. G[-high] 14#14 G[+high] /N[-high,-low].
2. G[-high] 14#14 G[+rhotic].
3. G[+high] 14#14 [17#17 front] /N[17#17 front].
4. N[+front] 14#14 [+low] /G[+high].
Informally, these rules arrange it so that the non-mid vowels have rhotic glides, while the mid vowels have high glides with a frontness feature inherited from the nucleus. The Glide slot for this set of vowels underlyingly contains the feature [-high], which is the underspecified form of the [rhotic] glide, as argued in Phonological Preliminaries. [-high] is rewritten as [+high] before mid vowels in (1), which is thus a rule of nucleus-glide differentiation. Then the remaining [-high] glides are rhoticized by rule (2), which must follow (1) as stated here. (3) assimilates the non-rhotic, high glides to the frontness of the adjacent nuclei, and must also follow (1).
The last rule, (4), changes the nucleus of the mid, front, upgliding vowel to low. This rule recapitulates a late part of the Great Vowel Shift,6.10 lowering the mid nucleus of a rising diphthong. It is restricted to apply only to the front mid vowel and not the back.6.11
We have hereby eliminated two underlying vowel-glide paradigms or subsystems, as well as the low-back vowels, from the base-6 system represented above in Table . The result of these eliminations is a rather abstract, but extremely simple and symmetric base-5 system of short, long, and gliding vowels, represented in Table .6.12
It is remarkable that the analysis of Reference American given above in Phonological Preliminaries resulted in a structure quite similar in fundamental respects to that argued for here: 3 heights vs. backness vs. length, plus restricted combinations of nuclei with 3 glides. Despite the important phonetic differences and the major differences of inventory, the structure of the vowel system of Jamaican Creole described here has the same basic dimensions as the surface phonological structure of a radically different dialect of English. The differences are that this representation of the Jamaican vowel system has a base-5 rather than a base-6 structure (that is, there is only one low vowel per subsystem), and that the gliding vowels can be collapsed into a single subsystem.