this paper aims to explore the types of rationality that underlie choice for political regime that intervene in the process of preference formation. the analysis is focused on brazil, a society that recently completed a successful process of transition towards democracy. our strategy is to test the rationality of support for democracy in both aggregate-level time series analysis, and individual level cross-section analysis. at aggregate level, our attention is centered on if, a long term, the performance of democratic governments predict the percentages of support for democracy. at individual level, our interest is focus on the balance between the impact of survey respondents' evaluation of different dimension of democratic performance (that we labeled here "utilitarian rationality"), and the effects of citizens' normative preferences (that we named here, "normative rationality"), as well as their joint impact on molding of individuals' preferences for a particular type of government in brazil. the results show that, at aggregate level, objective measures of performance do not explain support for democracy. in a different way, results at individual level endorse the hypothesis that utilitarian rationality prevails along some signs of normative rationality. finally, we suggest that such difference between two levels could be interpreted in favor of the idea of include other rationalities to explain political attitudes. if we only consider the utilitarian rationality in the preferences formation, we could only have a piece of the portrait.