Given that there is an effect of phonological length on (average) vowel quality, how can we best characterize this effect? In most dialects of English, phonological vowel length translates into phonetic peripherality (a better-defined term than the more traditional feature, tenseness -- see Labov, Yaeger & Steiner (1972, ``LYS''). A peripheral vowel is farther from the center of vowel space than a non-peripheral vowel.
In Jamaican, though, the only vowels that are not peripheral -- that is, occurring on the edges of the equilateral triangle formed by long /ii, aa, uu/ -- are short /i, a, u/. All the long vowels are peripheral, and even short /e/ is peripheral, since it too is adjacent to the edges of this triangle.
Comparing the long vowel space with the short vowel space, the clearest effect is that length is not realized as a ``peripheral'' location in phonetic vowel space in the simplest sense, but rather it means, ``in the corners'' of the vowel space. The short vowels occupy the center, while the long vowels occupy the corners of the vowel space.
It is again useful to compare this picture with the definition of the feature, peripherality in LYS (1972:p41ff, p106, etc.). Peripheral vowels are those that are distributed along the outer edges of a vowel space, while non-peripheral vowels are a step inward from the outer edges. In the current data, however, the phonetic feature associated with length is not peripherality in that sense of ``distributed along the edges'', but rather a different kind of peripherality, which may be described as ``distributed in the corners''.
The JC picture is somewhat more compatible with LYS' definition when it is combined with their principles of vowel shifting (p. 106), which state: In chain shifts, peripheral vowels rise, non-peripheral vowels fall, and back vowels move to the front. Long vowels are generally peripheral, so the long vowels should over time end up in the upper corners, while the short vowels should be relatively low. Juba's chart shows just this pattern: the short vowel system is relatively low, the long front-and-back vowels have raised to the corners, and the only vowel that is out of place, /u/, is a back vowel which has fronted. Thus all three of Labov's principles of vowel shifting are supported by this data. We will see below how these principles can be incorporated into a synchronic grammar.
In conclusion, we have seen that the phonological distinction of vowel length in Jamaican Creole is realized by large differences in average duration (long vowels are twice as long as short vowels), and by peripherality differences which have resulted in extremely significant vowel quality differences as shown by distributions in F1-F2 space.
The old story is confirmed: while one may analyse a set of phonological categories as though they are distinguished by a minimal set of structural dimensions, the phonetic realizations of the categories are distinguished by multiple, redundant phonetic differences.