Censored and truncated regression models with unknown distribution are impo
rtant in econometrics. This paper characterizes the class of all conditiona
l moment restrictions that lead to rootn-consistent estimators for these mo
dels, The semiparametric efficiency bound for each conditional moment restr
iction is derived. In the case of a nonzero bound it is shown how an estima
tor can be constructed and that an appropriately weighted version can attai
n the efficiency bound. These estimators also work when the disturbance is
independent of the regressors. The paper discusses combining conditional mo
ment restrictions for more efficient estimation in this case.