Functional magnetic resonance imaging (fMRI) is a rapidly growing technique for studying the brain in action. Since its creation [1], [2], cognitive scientists have been using fMRI to understand how we remember, manipulate, and act on information in our environment. Working with magnetic resonance physicists, statisticians, and engineers, these scientists are pushing the frontiers of knowledge of how the human brain works. The design and analysis of single-subject fMRI studies has been well described. For example, [3], chapters 10 and 11 of [4], and chapters 11 and 14 of [5] all give accessible overviews of fMRI methods for one subject. In contrast, while the appropriate manner to analyze a group of subjects has been the topic of several recent papers, we do not feel it has been covered well in introductory texts and review papers. Therefore, in this article, we bring together old and new work on so-called group modeling of fMRI data using a consistent notation to make the methods more accessible and comparable.