Broadly defined, affirmative action encompasses any measure that allocates goods ? such as admission into universities, jobs, promotions, public contracts, business loans, and rights to buy and sell land ? on the basis of membership in a designated group, for the purpose of increasing the proportion of members of that group in the relevant labor force, entrepreneurial class, or university student population, where they are currently underrepresented as a result of past or present discrimination. The programs involved vary substantially across countries, regarding the identification of their intended beneficiaries (ethnic, national, or racial groups held to be economically or socially disadvantaged, aboriginal peoples, women, people with disabilities, war veterans?), the form of the programs involved (quota/non-quota), the level of the legal norms from which they derive (constitutional, legislative, administrative), and their domain of implementation (whether restricted to the public sector or not). Drawing from previous research on the United States (« Judicial Uses of Subterfuge: Affirmative Action Reconsidered », Political Science Quarterly, 18 (3), 2003, p. 411-436) and France (« Affirmative Action at Sciences Po », French Politics, Culture, and Society, 20 (3), 2002, p. 52-64), the paper will compare affirmative action programs in the United States, India, Malaysia and South Africa, considering each of these four parameters and the relations between them. It will focus both on the intended results and on the different side effects of the policies and of the justifications advanced on their behalf. In particular, I shall argue that in India and Malaysia, reducing the inequality in the distribution of social goods between the different groups involved is not conceived as partaking of the more utopian project of reducing the saliency of group boundaries within the nation-state. Giving communities their proportional share of social benefits is not supposed to lead to the blurring of the lines that keep them separate from each other in private, non-allocative spheres. In the United States and in South Africa, on the other hand, the connection between affirmative action policies equalizing patterns of economic distribution among various ascriptive groups and the ideal of societal integration is a much stronger one. This is reflected in the definition of the social outcome whose attainment would justify the eventual termination of affirmative action programs. In India and Malaysia, the proportionality criterion provides an obvious « focal point » in this respect, generally acknowledged as such. In the United States ? and, to a lesser extent, in South Africa ?, proportional representation is emphatically rejected as a distributive principle, even though it arguably operates covertly at the policymaking level, by providing the benchmark against which « discrepancies » and « deficiencies » will be identified and compensated for. Last but not least, the paper will distinguish among affirmative action regimes according to their degree of internal differentiation, attempting to account for the highly specific configuration found in India, where members of the group generally considered as the most disadvantaged ? the Untouchables ? are being granted some kind of preferential treatment also in relation to the policy?s other beneficiaries. [ABSTRACT FROM AUTHOR]