There is a well-documented tendency for cognitive tests to underestimate the abilities of older people in black and minority ethnic groups. This gives rise to a substantially higher risk of mistaken diagnosis of dementia. Reasons include differences in extent or focus of formal education, lack of familiarity with English, lack of literacy in own first language, and culture-specific factors related to individual test items. Attempts to improve the accuracy of screening for these groups have included adaptation of existing tests, including adjustment of cut-points, translation and replacement of culture-specific items. So-called 'culture-free' tests have also been developed, which are less dependent on language, literacy and other skills developed during formal education. Cultural modifications and evidence of cross-cultural performance are summarized here for traditional tests (Mini-Mental State Examination, Short Portable Mental Status Questionnaire, Short Orientation-Memory-Concentration Test, Abbreviated Mental Test Score, Clifton Assessment Procedures for the Elderly), and for culture-free tests (Clock Drawing Test, Mini-Cog, 7-minute screening battery, Time and Change Test). The evidence on unadapted traditional tests shows that short ones perform at least as well as longer ones, and are more consistent across cultural and educational groups. Cut-point adjustments have not been universally found successful in improving accuracy, and do not address issues of acceptability. Translated and/or culturally adapted versions exist for a number of tests: it is important to establish cut-points appropriate to the target populations. There are promising results on culture-free tests, which are seen as less threatening and require little language interpretation, but they require further evaluation.