Computational modeling of the spine has become a viable option for evaluating new implants and procedures. Most models described in the literature, however, represent only a single subject and neglect the normal variation that exists among specimens. A probabilistic simulation comprised of virtual specimens representing a broad population of subjects can address this limitation and be used to evaluate implants or procedures pre-clinically. Challenges exist to applying probabilistic modeling techniques to biologic systems, and perhaps the greatest is parameterization of the anatomy to capture normal variation in shape from specimen to specimen. It’s also critical to implement soft tissues in a robust, automated manner that produces representative biomechanics. The purpose of our research is to overcome these challenges and develop a probabilistic framework to perform population-based studies of lumbar spine biomechanics. This paper describes our results to date for the automated generation of virtual lumbar motion segments.