Multiple hypothesis testing calls for methods that control Type I error without sacrificing the ability to test individual hypotheses. Major approaches involve ordered p-values or comparison of normally distributed means. Methods using ordered p-values include those based on the first-order Bonferroni inequality, those based on the Simes equality, modified methods for testing hypotheses that are logically related, and methods that control the false discovery rate. Other issues discussed include confidence intervals, robustness and directional versus nondirectional inference.