This dissertation investigates an often assumed hypothesis in phonetics and phonology: that there should be relatively less within-category phonetic variation in production in languages which have relatively more phonological contrasts (Lindblom, 1986, on vowels). Although this hypothesis is intuitive, there is little existing evidence to support the claim and it is difficult to generalize outside of vowels. In this dissertation, I argue that this hypothesis is not trivially true and needs additional specification. I propose an extension of this hypothesis, Contrast-Dependent Variation, which predicts relative differences in extent of within-category variation between languages and individual speakers. Contrast-Dependent Variation can make predictions across multiple phonetic spaces as it considers individual phonetic dimensions to be the relevant units of comparison, rather than phonological inventory subsets (stops, vowels, etc.). Relative differences in extent of within-category variation are predicted by differences in cue weight, rather than differences in number of phonemes. The dissertation tests this hypothesis by examining two between-language case studies: stops in Hindi and English and sibilants in French and Polish. I also consider a within-language case study: individual differences in extent of within-category variation in Mandarin sibilants. The results here show that differences in extent of phonetic variability between languages and speakers are systematic; they are structured according to the system of phonological contrasts.