Phylogenetic analyses of gene expression have great potential for addressing a wide range of questions. These analyses will, for example, identify genes that have evolutionary shifts in expression that are correlated with evolutionary changes in morphological, physiological, and developmental characters of interest. This will provide entirely new opportunities to identify genes related to particular phenotypes. There are, however, 3 key challenges that must be addressed for such studies to realize their potential. First, data on gene expression must be measured from multiple species, some of which may be field-collected, and parameterized in such a way that they can be compared across species. Second, it will be necessary to develop comparative phylogenetic methods suitable for large multidimensional datasets. In most phylogenetic comparative studies to date, the number n of independent observations (independent contrasts) has been greater than the number p of variables (characters). The behavior of comparative methods for these classic problems is now well understood under a wide variety of conditions. In studies of gene expression, and in studies based on other high-throughput tools, the number n of samples is dwarfed by the number p of variables. The estimated covariance matrices will be singular, complicating their analysis and interpretation, and prone to spurious results. Third, new approaches are needed to investigate the expression of the many genes whose phylogenies are not congruent with species phylogenies due to gene loss, gene duplication, and incomplete lineage sorting. Here we outline general considerations of project design for phylogenetic analyses of gene expression and suggest solutions to these three categories of challenges. These topics are relevant to high-throughput phenotypic data well beyond gene expression.