This paper describes the difficulties encountered in evaluating frozen sperm in both animal and human experiments, and discusses the principles of experimental design used to overcome them. Issues influencing the validity of results include randomization strategies and blindness. Issues which bear on the efficiency of design include stratification, pairing and blocking, multifactor designs, and use of covariates. Fundamental questions concerning the impact of study size and negative results are discussed. One approach to combining the results of many small studies, meta-analysis, is presented.