Comparative studies of genome-wide gene expression must account for variation not only among species, but also within species. Such studies are necessarily large in scale, because they incorporate experiments on multiple individuals of multiple species in multiple developmental stages in multiple environmental conditions. If the experiments are carefully designed and performed, the data they provide are worth the effort. We describe the utility of spotted microarrays for these studies and highlight experimental design criteria that will maximize inferential and statistical power. We conclude with a discussion of experimental protocols that are designed for investigations of differential gene expression and their pitfalls.