This thesis offers an original account of the South African Constitutional Court's reasoning in identifying unfair discrimination perpetrated by the state. I use this account to develop proposals for improving the Court's jurisprudence, in line with its stated aim of addressing patterns of group disadvantage. The Court's Harksen test for unfair discrimination makes dignity the touchstone for identifying this wrong. However, the Court has not explained what is required to prove a violation of dignity or how dignity fits with its concern for group disadvantage. I demonstrate that three necessary conditions must be satisfied for the Court to conclude that dignity has been violated: there must be a) unfavourable treatment on the basis of protected grounds; b) that threatens to create or perpetuate patterns of group disadvantage; and c) that lacks adequate justification. I also investigate important features of the Court's reasoning that have been overlooked in the existing literature, including its concern for messages expressed by discrimination and the fluctuating intensity with which it reviews justifications. Among my proposals for developing this reasoning, I argue that the Court should remove human dignity from the Harksen test and openly acknowledge the considerations doing the work in its decisions. I also provide a detailed critique of five of the Court's most controversial decisions where it found discrimination to be fair despite clear indications that it entrenched patterns of disadvantage. I show that in all five cases the Court applied an indefensibly weak intensity of review, falling below the baseline level of scrutiny which ought to be applied in unfair discrimination cases. I contend that consistent application of this baseline will help to make the Court better at preventing and addressing patterns of group disadvantage. I conclude with a restatement of the Harksen test that consolidates the Court's reasoning and my proposals.