Gender-specific distributions of vowel formants for /i/ appear to diverge from the overall (marginal) distributions (A), whereas for VOT the gender-specific distributions are essentially indistinguishable from the marginal distributions. Intuitively, this makes gender informative for vowel formants, but not for VOT (see also vowels in Perry et al., 2001; vs. VOT in Morris, McCrea, & Herring, 2008). The proposed approach formalises this intuition in a quantitative measure that can be applied to directly compare talker variability across different cues, phonetic contrasts, and socio-indexical grouping variables. Vowel data is drawn from the Nationwide Speech Project, and VOT from the Buckeye corpus (see below for more details).