The desiderata for cross-cultural research are reviewed showing how this is dominated by the presumption of transcultural equivalence hidden by the technical problems of translation. This model requires intensive translation work on measures to preserve "equivalence". This work can be prohibitively intensive, might lead to spurious evidence of "equivalence" and has not produced much evidence of equivalence to date. A psychometric axiom--that a measure with low reliability cannot have validity--suggests a pragmatic way to explore these issues and current statistical software and personal computers make this feasible. We recommend that all cross-cultural research should present the internal reliability of the data as this will expose all but one theoretical problem in the detection of "equivalence". We also suggest that non-equivalence may indicate cross-cultural differences in the dimensionality of eating and body image concerns rather than just a problem to be exorcised by more translation work or to be hidden by not reporting reliability. These issues are not specific to work across languages but also apply to comparisons between any social groups within cultures. This is shown using data from questionnaires in their original English with respondents who were fluent in that language.