Tables , show the effects of stress classifications using raw formant frequency measurements. Vr vowels are not included. 0 represents unstressed tokens, while 1 represents primary stressed tokens and 2 represents secondary stressed tokens. In the chart headings, /i:, u:, e:, o:, æ, , , / are written as /iy, uw, ey, ow, ae, ^ , a, E/, , respectively.
The effect of stress is to lower the vowel's realization for phonemes /æ/, //, /e:/, /w/, and possibly //, //, and /y/; backing applies to /o:/, /U/, and perhaps /u:/, and /oy/(and /o/ in Spanish words); fronting may apply to /I/.
The only vowel for which no effect is evident in these charts is /i:/, for which stressed tokens appear everywhere that unstressed tokens do, and in apparently similar proportions.
Syllabic /r/ (that is, //) is more spread out in all directions when unstressed than when stressed; that is, there appear to be target F1-F2 values for stressed //, whereas when unstressed that target is much more variably realized. With relatively few tokens, // shows the opposite effect: stressed tokens are distributed more widely than unstressed tokens, which occupy only a small part of the range of the stressed tokens.
The general picture fits nicely with the standard qualitative description of the effect of stress on vowels in English: stress generally makes vowels more peripheral and helps vowels to attain their targets. Thus front vowels should be fronter, back vowels backer, and low vowels lower. Further, if stressed vowels are more likely to hit a phonetic target than unstressed vowels, then vowels should be less widely distributed when stressed than when unstressed. This is generally the picture provided by the above charts.
An exception should be noted. Stressed and unstressed /e:/ are equally widely scattered along the F1 dimension, though stressed tokens are lower than unstressed tokens. So for this phoneme, phrasal stress does not make the vowel more likely to attain a particular target, since there is no evident target in either stressed or unstressed sets.