I present a novel approach to modelling spatial mental im-agery within the ACT-R cognitive architecture. The proposedmethod augments ACT-R’s representation of visual objects toenable the processing of spatial extent and incorporates a set oflinear and affine transformation functions to allow the manip-ulation of internal spatial representations. The assumptions ofthe modified architecture are then tested by using it to developmodels of two classic mental imagery phenomena: the mentalscanning study of Kosslyn, Ball, and Reiser (1978) and mentalrotation (Shepard & Metzler, 1971). Both models provide veryclose fits to human response time data.
