The possible effects of targeted snack food taxes on purchase decisions were assessed via computer-assisted intercept surveys in Canadian supermarkets, utilizing attribute-based stated preference methods (“choice experiments”). Participants were asked to choose between high-fat snacks, some displaying a warning label, and healthier snacks. Latent class models explaining choice were estimated. Results show heterogeneity of consumer response, with notable implications for public health. One class heeds warning labels, another avoids less healthy snacks and becomes more sensitive to price when a warning label is present, and a third class is sensitive to price but not warning labels.