The problem of estimating reliability of a system undergoing development testing is examined. It is assumed that the test program is conducted inKstages and that similar items are tested within each stage. In addition, it is assumed that the probability of an inherent failure,q0, remains constant throughout the test program while the probability of an assignable cause failure in thei-th stage,qi, does not increase withi. The number of inherent failures, of assignable cause failures, and of successes is recorded in each stage. Maximum likelihood estimates ofq0,qi(i= 1, 2, …,K) and a conservative confidence bound for the reliability in theK-th stage are obtained. Numerical examples to illustrate the methods are given.