In image-guided neurosurgery, deformable registration currently is not a clinical routine. Although using it in practice is a goal for image-guided therapy, this goal is hampered because surgeons are wary of the less predictable deformable registration error. In the preoperative- to-intraoperative registration, when surgeons notice a misaligned image pattern, they want to know whether it is a registration error or an actual deformation caused by tumor resection or retraction. Here, surgeons need a spatial distribution of error to help them make a better-informed decision, i.e., ignore locations with high error. However, such an error estimate is difficult to acquire. Alternatively, probabilistic image registration (PIR) methods give measures of registration uncertainty, which is a potential surrogate for assessing the quality of registration results. It is intuitive and believed by a lot of people that high uncertainty indicates a large error. Yet to the best of our knowledge, no such conclusion has been reported in the PIR literature. In this study, we look at one PIR method and give preliminary results showing that point-wise registration error and uncertainty are monotonically correlated.