Assessing speech discrimination skills in individual infants from clinical populations (e.g., infants with hearing impairment) has important diagnostic value. However, most infant speech discrimination paradigms have been designed to test group effects rather than individual differences. Other procedures suffer from high attrition rates. In this study, we developed 4 variants of the Visual Habituation Procedure (VHP) and assessed their robustness in detecting individual 9-month-old infants' ability to discriminate highly contrastive nonwords. In each variant, infants were first habituated to audiovisual repetitions of a nonword ("seepug") before entering the test phase. The test phase in Experiment 1 (extended variant) consisted of 7 old trials ("seepug") and 7 novel trials ("boodup") in alternating order. In Experiment 2, we tested 3 novel variants that incorporated methodological features of other behavioral paradigms. For the oddity variant, only 4 novel trials and 10 old trials were used. The stimulus alternation variant was identical to the extended variant except that novel trials were replaced with "alternating" trials--trials that contained repetitions of both the old and novel nonwords. The hybrid variant incorporated elements from both the oddity and the stimulus alternation variants. The hybrid variant proved to be the most successful in detecting statistically significant discrimination in individual infants (8 out of 10), suggesting that both the oddity and the stimulus alternation features contribute to providing a robust methodology for assessing discrimination in individual infants. In Experiment 3, we found that the hybrid variant had good test-retest reliability. Implications of these results for future infant speech perception work with clinical populations are discussed. (Contains 4 tables, 2 figures, and 1 footnote.)