The fidelity of an English-to-Japanese translation of a circadian rhythm questionnaire was examined through simultaneous factor analysis in several populations (Jöreskog, 1971a). Results indicate significant differences in item responses between populations, although between-population convergence was obtained on one factor. Back translations revealed both major and minor content discrepancies between the original and translated scales, which preclude clearly separating linguistic or semantic and population differences in item responses. Within-population results based on classic measurement techniques were compared with results based on structural equation techniques (in the American sample only); each technique led to different data-based conclusions. On the basis of the structural equation results, refinements in the source (English) scale items are suggested.