This section will present and discuss estimates of means of distributions in formant-space of this speaker's vowels. For each vowel, the estimated range of possible values for the mean of the vowel's F1-F2 distribution was calculated by using the bootstrap technique. The discussion of the bootstrap technique in the Methods chapter is prerequisite to understanding these distributions. 200 ``bootstrap'' resamplings were done of each vowel class, and the mean for each resampling was plotted. The distribution of resampled means gives a near-optimal estimate of the range within which the true mean lies. Of course these clouds do not represent the distribution of all the tokens of the particular sound-classes, but only of estimates for the mean of that distribution. The sound-classes themselves overlap considerably, presenting a problem of classification which may or may not be completely solved by additional information such as formant-trajectories, duration, stress level, and phonetic context.
Figure shows the distributions of bootstrapped means for 13 non-rhotic Alabama vowels in F1-F2 space, as estimated from 1637 measurements of vowels in the conversational speech of one rural, working class, older white male speaker. Each cloud of points represents the location and inherent accuracy of the estimated mean for the given vowel class, using the particular set of measurements made here. The way to interpret each distribution is that the true mean of the vowel class (assuming a similarly constructed sample whose size increases towards infinity) could be anywhere within the cloud with equal likelihood.
F1 is a measure of the degree of mouth-opening, while F2 reflects the degree of tongue-body frontness (Chapter 2, Acoustics, justifies these statements), which are the primary phonetic parameters distinguishing vowel quality. Thus the chart is an acoustic representation of vowel quality at the nucleus measurement point within the syllable. The origin (0,0) of the axes is to the upper right of the charts; this reverse orientation, created by plotting the negatives of the measurements, makes this graph correspond quite precisely with the usual auditory chart and articulatory vowel-triangles: high-front is located on the upper left, low central is in the middle on the lower edge of the graph. This representation shows in fine detail many differences in phonetic quality among different vowel sounds.
In comparing the locations of the vowels for this Alabama speaker with those that may be expected for Reference American or with those found in other dialects, a number of striking vowel shifts can be observed. The back vowels undergo a number of changes. First, /uw/ shifts very far to the front, so that it is adjacent to /iy/. /U/ also fronts quite far, especially in unstressed syllables. /U/ is farther front than //. These facts confirm the impressionistic results presented above.
The distribution of bootstrapped means for /U/ is much more widely dispersed than the those of other vowels, because there are only 18 tokens of /U/ in the data for this speaker. In fact, the generally more dispersed clouds in this data as compared with, say, the primary Jamaican speaker (Juba B.), is due to the smaller number of measurements (1637 as opposed to 2680).
/ow/, the mid-back long vowel, may undergo some raising, and rather more fronting, judging from the impressionistic transcriptions above. /ow/ is less peripheral than the THOUGHT-class vowel, /w/, which has become a back-raising diphthong, and is quite high; it is certainly not phonetically a low-back vowel, but more like [oU]. With this form, the contrast between /ow/ and /w/ for this speaker is based on a very small phonetic difference, as can be seen by the fact that they are immediately adjacent to each other in Figure . These vowels, which share the phonetic feature of a high-back glide, and whose nuclei are distinguished only by degree of peripherality in the mid-back region, may form a near-merger, an interesting matter for future perceptual research in this dialect. This approximation of the two classes is quite different from the expected phonetic contrast for younger or more upper-class speakers documented by Feagin, Foley, C.-J. Bailey, and others. For example, Feagin (1990:143) gives vowel charts for three other (upper-class or young or both) Anniston speakers, where all three clearly have low nuclei for /w/ and mid or upper-mid nuclei for /ow/. The expected contrast might be written as [U] versus [o]. Further work would be necessary to determine the social and linguistic patterning of these interesting and mysterious differences.
// is shifted up and back from the nucleus of /ay/, which it is identified with in dialects like Reference American. The back-round status of the LOT and THOUGHT classes (/, w/) comes from one of two sources. It may be a historical relic: these vowels are the long and short mid, back, round vowels of earlier forms of English, and in this dialect they may not have followed the lead of other American dialects in the lowering of LOT from [o] to  and of THOUGHT from [o:] to [:]. The other hypothesis is that they had previously lowered and fronted, and now have undergone a retrograde shift; this is somewhat less plausible, since it requires a greater number of changes.
Another long low vowel, /aw/ (MOUTH), undergoes a change that may be related to the diphthongization of /w/ (THOUGHT). When impressionistically transcribed, the nucleus of /aw/ is found fronted and variably raised to the area of [æ, e]. If /w/ moves from the long-vowel subsystem into the back-upgliding subsystem, then it becomes sensible that the nuclei of the two vowels should move apart so as to maximize the phonetic difference between the two. This may explain both the fronting of /aw/ [æo] and the raising of /w/ [oU]. Additionally, the glide of /aw/ is not always an up-glide; sometimes it glides directly back, and even back and down from the nucleus, as in [e] (as noticed in Philadelphia by Labov, and documented widely through the South by G. Bailey, and colleagues).
A smaller number of shifts operate on the front vowels. /æ/ raises nearly to mid (this may be an effect of the general close-mouthed character of James H.'s speech, though, since /æ/ is the lowest of the front vowels), and often has high-front glides. /ey/ shifts in the opposite direction, falling and laxing so that it overlaps with //. This overlap of the nuclei of /ey, / does not result in homophony, since glides distinguish the two.
All of these vowel shifts obey the rules for vowel-shifting in Labov, Yaeger and Steiner (1972), mentioned also in other chapters: In chain shifts, back vowels front, peripheral (long) vowels raise, and non-peripheral (short) vowels lower. Thus back vowels /uw, U/, and, to a slight extent, /ow/ move to the front, short // falls to a low position, and long /w/ raises along with /æ/.
Additionally, nucleus-glide differentiation applies to both /aw/ and /ey/ to front the nucleus of /aw/ away from the back and to centralize and then lower the nucleus of /ey/ away from the high-front-peripheral location of its offglide. /ey/ is lowered and centralized so far that its nucleus is indistinguishable in F1-F2 space from that of //, as shown in Figure .