The present study investigated vowel acoustics of conversational speech samples by healthy speakers and speakers with sensorimotor speech disorders. First and second formant frequencies during conversational speech did not exhibit the expected clustering into discrete or quasi-distinct regions of the vowel space. The absence of observable vowel clusters motivated the development of alternative measures to characterize vowel productions during conversational speech tasks. Vowel productions were treated as unlabeled data and variability was quantified in terms of formant changes across productions of frequently produced words. In addition, n-gram analysis quantified the extent to which vowel variability was attributable to a word’s predictability given the preceding three to five words. The results of these analyses are used to elaborate the parameterization of vowel targets during conversational speech and evaluate whether non-signal information, such as word predictability, influences motor planning for speech.