One might suppose that stressed vowels are more likely to attain the phonetic target of the vowel phoneme than unstressed vowels. Unstressed vowels should be more reduced and more coarticulated with adjacent sounds, which amounts to being more spread out over vowel space than their stressed counterparts. This section tests this supposition, the ``Stressed-Target'' hypothesis, in a simple way, with Rita's data.
If stressed vowels reach a target more than unstressed vowels, their nuclei should be more tightly distributed across vowel space, while unstressed vowels should be more spread out, since they are more likely to be modified by adjacent sounds, etc. In order to compare the scatter in these distributions, then, Table shows the standard deviations of F1 and F2 for stressed and unstressed tokens of each vowel.
If the stressed vowels are less widely scattered than the unstressed vowels, then the stressed standard deviations should be smaller than the unstressed standard deviations. The claim is that a target is more likely to be achieved in a stressed vowel than in an unstressed vowel. These data give only mild support to this theory. 11 of 17 vowels (about 2/3) have a smaller F1 standard deviation when stressed than when unstressed; a different subset of 11 of the 17 have a smaller F2 standard deviation when stressed than when unstressed. So about two-thirds of the vowels are more tightly distributed in each dimension when stressed than when unstressed.
However, since about a third of the vowels fail to observe the predicted pattern, the stressed-target hypothesis is at best a fairly weak generalization.