This section considers the effects of phrasal stress on vowel quality as represented by F1-F2 measurements. See Chapter 5 for a discussion of what stress means here (that is, how utterances were impressionistically coded for stress). Somewhat different effects are found when clitic words are included or excluded. Since the pattern is simpler, statistically more significant (4 more vowels have significant effects), more similar to the pattern found in other dialects, and more understandable when clitic items are included, I will primarily discuss the pattern found when all tokens are included, though both charts are displayed, for consistency with other chapters.
Consider Figure , which displays what happens to vowel nucleus formant frequencies under stress reduction.
The interpretation of the graphs is as follows. Each arrow is labelled with the symbol from Table for the class it represents. The tail of each arrow is at the mean of the measurements of the stressed (primary and secondary together) tokens of that vowel class. The head of the arrow is at the mean of the unstressed tokens. Thus each arrow shows the average effect of destressing on the phonetic quality of that vowel. Solid arrows represent statistically significant differences between the stressed and unstressed sets for that class, using the two-tailed t-test described on page . Dashed arrows are insignificant at the 5% level on this test. Vowel classes which only occur with stress are plotted at their mean location and marked with ', while vowels that only occur without stress do not have the apostrophe. The clitics-excluded chart displays the effects of stress on vowel classes from which tokens occurring in clitic words are excluded, while the lower chart displays the effects of stress on vowels including both clitic and non-clitic tokens. The clitic/non-clitic distinction is defined in the Methods chapter.
Note that /ə/ does not include tokens of the reduced vowel /ɨ/ argued by Sledd (1966) to be phonologically distinctive in Southern dialects. Instances of this class are classified as unstressed tokens of /ʌ/, which is arguably the stressed vowel corresponding to /ə/. The mean for this vowel class is thus located at the head of the arrow whose tail is labelled with ``ʌ''.
The pattern of reduction is clearer (more consistent with the patterns found in other dialects) in the clitics-included chart than in the clitics-excluded chart, so I will discuss the clitics-included chart. There we find that the front vowels raise and centralize, with increasing centralization from /æ/ to /ey/ to /, I/ and finally /iy/. The central vowels (fronted /U/ as well as /, , /) move up and to the front. All of these are consistent with the theory of reduction as a shift in formant frequencies roughly in the direction of a target around , where  is understood as at least as high as [i]. This tendency is rather rough, since /I, , ay/ point towards a location somewhat to the back of this target.
A second pattern is observed in this chart, namely the reduction of the nucleus of a diphthong in the direction of its offglide, or nucleus/glide assimilation. This effect explains the two anomalous, though statistically weak, shifts of /w/ and /aw/.
/w/ has a high-back glide, and when destressed it shifts toward the high-back corner. Similarly /aw/ frequently has a low-back glide as [e], and when destressed it shifts back and slightly downward. Nucleus/glide assimilation might also be used to explain the anomalous direction of shift for /I, , ay, ow/. /I, / are inglides in stressed monosyllables, according to Foley (1972); these vowels centralize more than they would if they were simply shifting in the direction of the apparent reduction target . /ay/ is impressionistically transcribed above as having central rather than front glides in some cases, and the stress-reduction effect on /ay/ does not include any fronting. If /ay/ simply shifted toward the reduction target when unstressed, it would shift to the front as well as upwards; if instead /ay/ shifts towards its high-central glides, it would not shift to the front -- this is the actual effect found.
/ow/ also does not shift towards high-front-central position. /ow/ has a high-back glide, and under stress-reduction it moves -- just slightly -- toward the high-back corner.
It is important to note that the reduction target,  seems to be quite distinct from the location of the reduced vowel, //. The reduction target is high and well to the front of central, while // is upper-mid, central, at about the same location as the mean of stressed /U/ and unstressed // (// is presumably distinguished by F3 differences). The reduction target for this dialect appears to be high and relatively front. It is different, then, from the mid-central reduction target of Jamaican as well as the high central reduction target of the Chicago vernacular.