In this paper we present a conceptual model and a methodology for the review of translated tests in the context of such international comparisons as TIMSS and PISA. We also discuss results from a research on the quality of the Mexican, Spanish translation of TIMSS-1995. We identified translation errors in a considerably high percentage of items and observed relatively high correlations between the severity of translation error and the items’ p values. These findings indicate that our approach for error coding is very sensitive to test translation error. These findings underscore the need for improved translation and translation review procedures in international comparisons. In our opinion, in order to properly implement guidelines for test translation in international comparisons, each participating country needs to have internal procedures that ensure a rigorous review of its own translations. As a conclusion, we offer four recommendations for countries participating in international comparisons. These recommendations address: (a) the characteristics of the individuals in charge of translating instruments, (b) the use of review not simply at the end of the process but during the process of test translation, (c) the minimum time needed for various translation review iterations to take place, and (d) the need for properly documenting the entire process of test translation.

Key words: Educational tests, international tests, test translation, TIMSS, PISA.