How do the labels left and right take on meaning in new democracies? Existing explanations point to the universality of the left–right scheme or, reversely, emphasize regionally dominant social cleavages. We propose an alternative legacy-focused theory based on two observations: Dictatorships are not ideologically neutral and are negatively evaluated by most citizens and elites after democratization. These premises lead us to expect that when the authoritarian regime is associated with the left (right), the citizens of a new democracy will display an antileft (antiright) bias in their left–right self-identification. We test this hypothesis across Latin American and European new democracies. We find significant bias, which in the case of new democracies following left-wing regimes is concealed due to intercohort heterogeneity. Although older cohorts denote a positive bias, cohorts born after Stalin’s era denote negative bias against the left. Consistent with our expectations, repression exacerbates this bias whereas indoctrination mitigates it. Finally, we look at how these biases apply to party preferences. The findings have important implications for understanding authoritarian legacies and party system development in new democracies.