In its traditional guise, cross-cultural psychology is a science of comparative measurement. Its chief method is the regulated observation or recording of behaviour in different cultural milieux and its quantitative conversion into a common metric. This method allows for the estimation of differences across cultures on the strength or magnitude of the behaviour. We can therefore speak of one cultural group doing more of Y, or being more Y, than some other group, where Y is a conceptually and operationally defined behaviour taken to apply equally well to both groups. The culture-independence of the definition supports examination of how the magnitude of the behaviour may be determined or conditioned by some cultural variable X. According to this picture, culture (the independent variable) differs from place to place and the features by which it differs produce more or less of the universally defined behaviour (the dependent variable) under study. The hypothetical contingency of ‘culture affecting behaviour’ that underlies this approach is thereby maintained. In this paper, I use the problem of translation to cast doubt on this picture of cross-cultural research and the inferential strength of its comparative method. In opposition, I argue that we never measure the same behaviour across cultures if behaviour is understood as socially significant action. Rather, the specification of analogous actions across cultures is a highly uncertain approximation involving the inevitable projection of one’s own conceptual scheme. I argue further that most social phenomena of interest to cross-cultural psychologists cannot be adequately defined in a manner that divorces them from the local linguistic conventions and normative frameworks through which they are realized as part of cultural life. For this reason, cross-cultural psychology cannot effectively model itself on the natural sciences. It is as much hermeneutics as psychometrics.