Functional magnetic resonance imaging (FMRI) is a major tool for the evaluation of brain function and architecture. It is widely used by physicians, neuroscientists, psychologists and others. In order to process the data collected using FMRI, it is necessary to use post-acquisition processing software that employs motion correction and statistical modeling capabilities. These types of programs permit the user to extract the information about areas of brain activations that have occurred during a study. How well a particular motion-correction technique works and what effect it has on statistical processing are difficult to evaluate, since the level of activation present is not known a priori. This paper provides a description of the construction of a software phantom for use with FMRI post-acquisition processing tools with the properties that it is based on real subject data, has known locations and levels of activation, has known amounts of rigid body motion and noise added, and can be used to evaluate a processing system as if it were a real data set. Versions of the software phantom are available for downloading at the website: .