Functional MRI (fMRI) has become one of the most commonly used neuroimaging tools to assess the cortical effects associated with rehabilitation, learning, or disease recovery in subjects with stroke. Despite this, there has been no systematic study of the reliability of the fMR signal in this population. The purpose of this study was to examine the within- and between-session reliability of fMRI in cortical and cerebellar structures in subjects with stroke during a complex, continuous visual motor task performed with the less affected hand. Nine subjects with stroke underwent four testing trials during two sessions separated by three weeks. Subjects performed a drawing task using an MRI compatible joystick while in the MRI. Methods of analysis evaluated included: percent signal intensity change, active voxel count and a voxel by voxel stat value analysis within and between testing sessions. Reliability was determined with Interclass correlation coefficients (ICC) in the following regions of interest: primary motor (M1), primary sensory (S1), premotor cortex (PMC), medial cerebellum (MCB), and lateral cerebellum (LCB). Results indicate that intensity change has superior reliability to the other methods of analysis (Average ICC across brain regions and trials: intensity change: 0.73, voxel count: 0.58, voxel by voxel: 0.67) and that generally with any analysis method, within-session reliability was higher than between-session, as indicated by higher ICC values across brain regions. Overall, when comparing between-session results, moderate to good reliability was obtained with intensity change (ICC: M1: 0.52, S1: 0.80, SMA: 0.78, PMC: 0.94, MCB: 0.86, and LCB: 0.59). These results show good reliability in subjects with stroke when performing a continuous motor task. These findings give confidence for interpreting fMRI test/retest research in subjects with stroke.