In the present research, the authors examined contextual variations in automatic attitudes. Using 2 measures of automatic attitudes, the authors demonstrated that evaluative responses differ qualitatively as perceivers focus on different aspects of a target's social group membership (e.g., race or gender). Contextual variations in automatic attitudes were obtained when the manipulation involved overt categorization (Experiments 1-3) as well as more subtle contextual cues, such as category distinctiveness (Experiments 4-5). Furthermore, participants were shown to be unable to predict such contextual influences on automatic attitudes (Experiment 3). Taken together, these experiments support the idea of automatic attitudes being continuous, online constructions that are inherently flexible and contextually appropriate, despite being outside conscious control.