Stereolithography has attracted more attention due to better part build accuracy than other rapid prototyping technologies. However, this build method still limits wider applications due to the unsatisfactory level of dimensional accuracy that remains with the current technology. To improve accuracy and reduce part distortion, understanding the physics involved in the relationship between the operating input parameters and the part dimensional accuracy is prerequisite. In this paper, this causality is identified through a process model obtained via an artificial neural network based upon 140 actual build parts. The network is so constructed that it relates the process input parameters to part dimensional accuracy. The neural network model is found to predict the effects of the input parameters on the accuracy with reasonable accuracy. The prediction performance is discussed in detail for various process parameter ranges.