Abstract
An experiment using a probability sample of 263 U.S. consumers was conducted to document the psychometric effects of translation error on a semantic differential scale. The results demonstrate that data distributions and response styles are significantly altered as a result of mistranslation (inequivalence). Unexpectedly, results show attenuation of equivalent items when inequivalent items are present in a scale. This threatens the validity and reliability of cross-cultural scales that suffer equivalence failure. The authors recommend that simple nonparametric diagnostics be employed to detect translation/back-translation errors and that an iterative process of cross-cultural research be adopted when translation inequivalence is detected. Implications for researchers and practitioners are offered, as well as a discussion of limitations of the study.