Detailed mechanisms for substrate water exchange in the oxygen evolving complex in photosystem II have been determined with DFT methods for large models. Existing interpretations of the experimental water exchange results have been quite different. By many groups, these results have been the main argument against the water oxidation mechanism suggested by DFT, in which the oxygen molecule is formed between a bridging oxo and an oxyl radical ligand in the center of the OEC. That mechanism is otherwise in line with most experiments. The problem has been that the mechanism requires a rather fast exchange of a bridging oxo ligand, which is not a common finding for smaller Mn-containing model systems. However, other groups have actually favored a substrate derived oxo ligand partly based on the same experiments. In the present study, three S-states have been studied, and the rates have been well reproduced by the calculations. The surprising experimental finding that water exchange in S1 is slower than the one in S2 is reproduced and explained. The key to this rate difference is the ease by which one of the manganese centers (Mn3) is reduced. This reduction has to occur to release the substrate water from Mn3. The similar rate of the slow exchange in S2 and S3 has been rationalized on the basis of earlier experiments combined with the present calculations. The results strongly support the previous DFT-suggested water oxidation mechanism.