Sign statistics are used for change-point detection. We consider the so-called “epidemic” alternative, which is a generalization of the at-most-one-change alternative. Asymptotic and exact finite sample distributions under the null hypothesis of no change have been derived in Gombay (1994). We now provide the tables. Furthermore, asymptotics for the test statistics and the properties of the change-point estimators are investigated under the alternative hypotheses. The properly normalized versions of the test statistics are asymptotically normally distributed or functions of normal random variables.