Copper-zinc superoxide dismutase (CuZnSOD) acquires its catalytic copper ion through interaction with another polypeptide termed the copper chaperone for SOD. Here, we combine X-ray crystallographic and analytical ultracentrifugation methods to characterize rigorously both truncated and full-length forms of apo-LYS7, the yeast copper chaperone for SOD. The 1.55 Å crystal structure of LYS7 domain 2 alone (L7D2) was determined by multiple-isomorphous replacement (MIR) methods. The monomeric structure reveals an eight-stranded Greek key β-barrel similar to that found in yeast CuZnSOD, but it is substantially elongated at one end where the loop regions of the β-barrel come together to bind a calcium ion. In agreement with the crystal structure, sedimentation velocity experiments indicate that L7D2 is monomeric in solution under all conditions and concentrations that were tested. In contrast, sedimentation velocity and sedimentation equilibrium experiments show that full-length apo-LYS7 exists in a monomer-dimer equilibrium under nonreducing conditions. This equilibrium is shifted toward the dimer by approximately 1 order of magnitude in the presence of phosphate anion. Although the basis for the specificity of the LYS7-SOD interaction as well as the exact mechanism of copper insertion into SOD is unknown, it has been suggested that a monomer of LYS7 and a monomer of SOD may associate to form a heterodimer via L7D2. The data presented here, however, taken together with previously published crystallographic and analytical gel filtration data on full-length LYS7, suggest an alternative model wherein a dimer of LYS7 interacts with a dimer of yeast CuZnSOD. The advantages of the dimer-dimer model over the heterodimer model are enumerated.