Spatial normalization in functional imaging can encompass various processes, including nonlinear warping to correct for intersubject differences, linear transformations to correct for identifiable head movements, and data detrending to remove residual motion correlated artifacts. We describe the use of AIR to create a custom, site-specific, normal averaged brain atlas that can be used to map T2 weighted echo-planar images and coplanar functional images directly into a Talairach-compatible space. We also discuss extraction of characteristic descriptors from sets of linear transformation matrices describing head movements in a functional imaging series. Scores for these descriptors, derived using principal components analysis with singular value decomposition, can be treated as confounds associated with each individual image in the series and systematically removed prior to voxel-by-voxel statistical analysis.