In attitude research, behaviors are often used as proxies for attitudes and attitudinal processes. This practice is problematic because it conflates the behaviors that need to be explained (explanandum) with the mental constructs that are used to explain these behaviors (explanans). In the current article, we propose a meta-theoretical framework that resolves this problem by distinguishing between two levels of analysis. According to the proposed framework, attitude research can be conceptualized as the scientific study of evaluation. Evaluation is defined not in terms of mental constructs but in terms of elements in the environment, more specifically, as the effect of stimuli on evaluative responses. From this perspective, attitude research provides answers to two questions: (1) Which elements in the environment moderate evaluation? (2) What mental processes and representations mediate evaluation? Research on the first question provides explanations of evaluative responses in terms of elements in the environment (functional level of analysis); research on the second question offers explanations of evaluation in terms of mental processes and representations (cognitive level of analysis). These two levels of analysis are mutually supportive, in that better explanations at one level lead to better explanations at the other level. However, their mutually supportive relation requires a clear distinction between the concepts of their explanans and explanandum, which are conflated if behaviors are treated as proxies for mental constructs. The value of this functional-cognitive framework is illustrated by applying it to four central questions of attitude research.