The Cornell model has been extensively used as a mouse model for studying the latent stage of Mycobacterium tuberculosis infection. In this model mice are infected and then given a course of chemotherapy prior to reactivation of infection. We discuss here the importance of using adequate mouse numbers in a Cornell model for the study of relapse rates in order to obtain sufficient statistical power to confirm a hypothesis. Experiments with small sample sizes are useful for ‘screening’ experiments, but will have very little value in ‘confirming’ the objective. When the objective of the experiment is confirmation of an effect through establishment of statistical significance, power calculations are critical in order to assure that the sample size will be sufficient to meet that objective.