Abstract The determinants of performance in word translation by unbalanced bilinguals, fairly fluent in their second language, were studied. Translation was both from the subjects′ native (L1) to their second (L2) language and in the reverse direction ("forward" and backward" translation, respectively). The predictor variables were imageability, context availability, definition accuracy, familiarity, word frequency, length (each of these six was determined for the L1 and L2 words separately), and the cognate status of the translation equivalents. Both forward and backward word translation were influenced by meaning variables, familiarity variables, and cognate status. However, meaning played a somewhat more important role in forward than in backward translation, whereas familiarity appeared to have a larger influence in backward translation. A few other differences between forward and backward translation were detected, but, when considering the complete stimulus set, the differences between translation directions were generally small. In some of the subsets of the stimulus materials (particularly noncognates) larger directional differences occurred. Particularly relevant is the finding that meaning affects backward translation, because it suggests a qualification of the "asymmetry model" of word translation as proposed by Kroll and Stewart (1994).