Abstract
Cross-national comparison depends on instruments that measure the same construct across languages, an assumption tested less often than it is asserted. We back-translated a widely used well-being scale into Uzbek and Russian and administered all three versions to bilingual respondents. Two items showed systematic divergence exceeding half a scale point, both relating to expressions of contentment that lack close equivalents. Aggregate scores concealed this drift entirely, which cautions against comparing national means without item-level examination.