In my own view of social behavior, attitude is a central concept, and I regard the study of attitudes as a legitimate pursuit in its own right and as useful component of a multimethod research strategy. For example, studies of the structure and distribution of public attitudes on various social issues or of the determinants of attitude change can greatly contribute to our understanding of societal processes. In particular, attitude assessment combined with other research procedures – such as structural indexes in studies of organizational functioning or behavioral observations in studies of child–rearing practices – can add depth and perspective to the analysis of social phenomena. Too often, however, research has focused on attitudes, not so much because of primary interest in attitudes per se or in their contribution to complex social processes, but because the assessment of attitudes was the easiest and most convenient way of obtaining data. It is obviously more convenient to distribute questionnaires in our classes, asking students how they would be likely to react in a variety of situations, than to observe their behavior in such situations, or to ask a sample of some population about the functioning of their organization, community, or society, than to make the detailed observations and develop the elaborate indexes needed to study a social system more directly. In many of our laboratory experiments too, after investing a great deal of energy and ingenuity in manipulation the independent variables, we have used attitudes as the cheapest and most painless dependent variable for assessing the effects of our manipulations.
Kelman, Herbert C. "Attitudes Are Alive and Well and Gainfully Employed in the Sphere of Action." American Psychologist 29 (1974): 310-24.