Best practice for international research projects
Multi-country studies fail on translation and comparability far more often than on fieldwork. What the cross-cultural survey literature recommends instead of back-translation.
A pan-European or global study is not a domestic study repeated several times. The output is a comparison, and a comparison is only as good as the equivalence of the instruments that produced it. Most of the risk sits in one place: the questionnaire that has to mean the same thing in every country it lands in.
Back-translation is not enough
The traditional safeguard is translate-and-back-translate: one translator renders the questionnaire into the target language, another renders it back, and the two source versions are compared. It is cheap and it catches gross errors. Its weakness is well documented — it tests whether words round-trip, not whether the question works. A back-translation can come back looking perfect while the forward version asks something subtly different, or something culturally unaskable.
The team approach
Current practice in comparative survey research uses a team method, usually summarised as TRAPD: Translation, Review, Adjudication, Pretesting, Documentation.
Several translators work independently first, then meet. The translators are deliberately mixed — survey methodology, translation practice, subject knowledge — because the errors each discipline catches are different. The review session, where versions are read aloud and argued over, is where problems of register and vocabulary surface, and those are exactly the problems a silent desk review misses.
Adjudication settles the final version with someone accountable for the decision. Pretesting puts it in front of real respondents. Documentation records what changed and why, which is what makes the next wave possible.
What has to stay comparable
Translation quality is not only about meaning. Response scales, style and register, layout and presentation all shape the answer, and all of them can drift country by country while every individual translation looks defensible. A five-point scale whose midpoint is labelled in one country and unlabelled in another has stopped being the same instrument.
Plan the timetable around it
The practical consequence is that the questionnaire has to be finished earlier on an international project than on a domestic one — not later, which is the instinct when several agencies have to be coordinated. Translation, review and pretesting across six countries is not an administrative step to be squeezed between sign-off and fieldwork. It is a phase. Projects that treat it as a formality discover the problem at analysis, when the country differences in the data turn out to be differences in the questionnaire.
