Individual rationality plays a central role in modern economic theory, while how to measure rationality has been quite challenging. In particular, whether the rationality measured in different domains is stable or not is an open question. Combining consumers’ purchase data from a large supermarket chain and their choice in a survey experiment, we directly examine three important questions for rationality measurements: (1) external validity: can rationality measured in the survey experiment successfully predict consumers' actual purchase behaviour; (2) cross validity: is rationality exhibited in risk preference consistent with that in social preference? (3) what leads to these asymmetric patterns of consistency.