Typical microarray or GeneChip experiments now provide genome-wide measurements on gene expression across many conditions. Analysis often focuses on only a few of the genes, looking for those that are "differentially expressed" between conditions or groups of conditions. However, the large number of measurements both present statistical problems to such single gene approaches and offers a tremendous amount of information for methods focused on biological processes rather than individual genes. Here we provide a method to utilize biological annotations in the form of gene ontologies to interpret the results of individual or multiple pattern recognition analyses of a microarray experiment.