Figure 1 lb is not correct. The band structure calculations used to create this figure included so-called "ghost states," spurious wave functions which occasionally appear in linearized methods due to the incompleteness of the basis set. A proper choice of energy parameters will eliminate these states. The self-consistent calculations did not contain ghost states. We have eliminated the ghost states in the calculations leading to the new Figure l lb presented here. The conclusions reached in the paper are actually strengthened. We have also changed Figure 1 la by including the Mg 2p semicore states. Finally, the references to Pickett [1985] and Pickett [1986] should be to the journal Comments on Solid State Physics. We would like to thank H. Lu for doing the calculations for the new Figure 11. a V=18.1A :• b V=lO,&, 3 >'8 o• Madelung Potential,• I _ Modelung Pofenfiol •,"""" .. Watson Potential Watson Potential Fig. 11. Differences in eigenvalues at the F, L, X, and K symmetry points for PIB and LAPW charge densities for the B1 phase of MgO. The curves with solid lines are the O 2p states, the dashed curves are the Mg 2p and O 2s states, and the dotted curves are the excited states. The Madelung potentials are indicated by arrows. (a) 18.1 ,•3, 0 GPa. The best fit Watson sphere potential is very close to the Madelung potential. (b) 10 ,•3, 345 GPa. The best Watson sphere potential is somewhat larger than that given by the Madelung potential, but properly chosen spherical ions give an accurate fit.