OBJECTIVE This article aims: (i) to re-examine the use and usefulness of categorisation based on 'race', ethnicity and 'population group' membership in public health research; and (ii) to assess the consequences of using these categories for describing, analysing and redressing disparities in health within South Africa. The utility of categorisation based on 'race', ethnicity and 'population group'. Categorising populations and comparing patterns of disease between different groups of people can be a useful technique for identifying potential causes of disease. In this context, ethnicity is a valid social concept that could be used to investigate the consequences of self-ascribed identity on health. Likewise, 'population group', as defined during apartheid in South Africa, represents a valid political concept that could be used for assessing the impact of social discrimination on health. However, both these concepts are often seen, and used, as euphemisms for 'race', even though there are no genetically distinct human subspecies that can be identified and categorised as discrete 'races'. Indeed, 'race' as a biological concept has no validity in human biology. Nevertheless, categories based on 'race', ethnicity and 'population group; continue to be used in health research, and reinforce the perception that differences in disease between different 'racial', ethnic and 'population' groups are the result of heritable biological characteristics. In so doing, they undermine support for health interventions that would otherwise address the social and political origins of 'racial' and ethnic disparities in health. The utility of 'population group' for redressing the consequences of apartheid. Despite these problems, 'population group' classification provides important information for assessing the impact of apartheid on disparities in health within South Africa. Yet, the abolition of apartheid legislation is likely to result in extensive socio-economic and geographical migration that will weaken the sensitivity and specificity of 'population group' as an indicator for identifying inequalities in health. For this reason, targeting corrective action at specific population groups in order to tackle disparities in health, runs the risk of ignoring alternative social causes of inequalities in health, and ignoring disadvantaged individuals from elsewhere in the population. The continued use of 'population group' classification might also perpetuate the root cause of disparities in health, by maintaining the process used to formalise discrimination. CONCLUSIONS If the purpose of health research is to monitor inequalities in health and to help target resources aimed at reducing these inequalities, then it should seek, in its language, concepts and methods to undermine the root cause of disadvantage. Health research aimed at monitoring and redressing the consequences of social disadvantage on health should therefore focus on non-biological determinants of social disparities in health. As a general rule, health researchers should avoid using categories based on 'race', ethnicity and 'population group' when collecting and analysing health data; journal editors should not accept articles that use these categories without justification; and health authorities should not collect data routinely segregated by 'race', ethnicity or 'population group'.