A theory of testing under non-standard conditions is developed. By viewing the likelihood as a function of the unknown parameters, empirical process theory enables us to bound the asymptotic distribution of standardized likelihood ratio statistics, even when conventional regularity conditions...