Both the Rule Space Method (RSM) and the Neural Network Model (NNM) are techniques of statistical pattern recognition and classification approaches developed for applications from different fields. RSM was developed in the domain of educational statistics. It started from the use of an incidence matrix Q that characterizes the underlying cognitive processes and knowledge (attribute) involved in each item. The examinee's mastered/nonmastered states (knowledge state) for each attribute is determined from item response patterns. RSM uses the multivariate decision theory to classify individuals, and NNM, considered a nonlinear regression method, uses the middle layer of the network structure as classification results. Similarities and differences between the methods, and their supplemental characteristics when both are applied are discussed. This paper compares these approaches by focusing on the structures of NNM and knowledge states in RSM. An application of RSM is shown for a reasoning test in Japan. (Author/SLD)