We assess the validity of the Thisted-Efron author-ship tests in two stages. First, we construct simulated texts in accordance with the assumptions implicit in the underlying model and use these to validate the basic computations, to determine their range of applicability, and to evaluate their sensitivity to basic lexical parameters. Second, we experiment with actual texts from the Shakespearean canon and the plays of Christopher Marlowe. The results of the tests are mixed, showing good consistency for the Shakespeare plays (with some discrimination among early, middle and late works) but poor consistency between Shakespeare's poems and plays, or among Marlowe's plays.