This chapter works out the ontological commitments of economic theory in the sense of trying to analyse which concept of preference is presupposed by economic theory. Faruk Gul and Pesendorfer specifically argue against the use of some kinds of data: subjective states and hedonic utility are not useful for calibrating and testing standard economic theory. The chapter presents two arguments that show that it is hard to formulate economic explanations while using the behaviouristic preference concept. The first methodological argument emphasises that understanding choice behaviour requires one to understand how people represent the outcomes of choice problems. The second argument shows that to model strategic interactions one needs to assume that agents have preferences over outcomes that they cannot choose. The mentalistic preference concept can integrate the results of behavioural decision research into economic theory. Glymour's bootstrap account of confirmation to show how hypotheses that contain variables whose values are not immediately determinable by observation can still be tested.