A striking feature of sulfate (SO42-) and molybdate (MoO42-) transport proteins, such as SBP and ModA, which specifically bind SO42- and MoO42-, respectively, is their ability to discriminate very similar anions with the same net charge, geometry, and hydrogen-bonding properties. Here, we determine to what extent (1) oxyanion-solvent interactions, (2) oxyanion-amino acid interactions, and (3) the anion-binding pocket sizes of the cognate protein contribute to the anion selectivity process in SO42- and MoO42- transport proteins by computing the free energies for replacing SO42- with MoO42-/WO42- in model SO42--binding sites of varying degrees of solvent exposure using a combined quantum mechanical/continuum dielectric approach. The calculations reveal that MoO42- transport proteins, such as ModA, specifically bind MoO42-/WO42- but not SO42-, mainly because the clesolvation penalty of MoO42-/WO42- is significantly less than that of SO42- and, to a lesser extent, because the large and rigid cavity in these proteins attenuates ligand interactions with SO42-, as compared to MoO42-. On the other hand, SO42- transport proteins prefer SO42- to MoO42-/WO42- because the small anion-binding pocket characteristic of these proteins inhibits binding of the larger MoO42- and WO42- anions. The calculations also help to explain the absence of positively charged Lys/Arg side chains in the anion-binding sites of SBP and ModA. During evolution, these transport proteins may have excluded cationic ligands from their binding sites because, on one hand, Lys/Arg do not contribute to the selectivity of the binding pocket and, on the other, they substantially stabilize the complex between the oxyanion and protein ligands, which in turn would prohibit the rapid release of the bound oxyanion at a certain stage during the transport process.