We have extended our calculations for defect levels at semiconductor free surfaces to ideal semiconductor heterojunctions. We find that the Fermi-energy pinning observed for deposition of Ge and Si (as well as metals and oxygen) on GaAs(110) surfaces is explained very satisfactorily by free-surface antisite defect levels, but cannot be explained, even qualitatively, by antisite defect levels for bulk GaAs or for ideal GaAs/Ge or GaAs/Si interfaces. We conclude that when pinning occurs at heterojunctions and at other semiconductor/overlayer interfaces, the pinning defects are somehow ‘‘sheltered’’ by irregularities in the atomic configurations at the interface.