Diffusion weighted imaging (DWI) is important for prostate cancer diagnosis but is highly sensitive to artifacts. We developed a method for automatically and quantitatively measuring physically implausible DWI signals (PIDS) that could contribute to diagnostic errors. The level of PIDS is significant and similar in all prostate zones, for scans both with and without an endorectal coil, and is strongly correlated with motion. For scans without the endorectal coil, PIDS is correlated with noise level. In regions with a high percentage of PIDS, PIRADS II criteria may not be optimal, and algorithms that emphasize T2 and DCE-MRI may be preferable.