In order to diagnose properly quality problems that occur in manufacturing the diagnostician, be it human or computer, must be privy to various sources of information about the process and its behaviour. This paper describes how neural networks and Bayesian discriminant function techniques can be used to provide knowledge of how a product characteristic changed, i.e. shift in mean or variability, when so noted by the control chart application. Such information is useful because there usually exists some underyling knowledge about the physical phenomena in question that relates the behaviour of the observed characteristic to its processing variables. When a change in the process is detected by the appropriate statistical method, a feature vector of process-related statistics is used to identify the change structure as a shift in mean or variance. This paper addresses various issues concerned with this problem, namely: process change detection, feature vector selection, training patterns, and error rates. Simulation experiments are used to test various hypotheses and also compare the effectiveness of the two proposed approaches against two simpler heuristics. Results show the neural network and quadratic discriminant function approaches to be fairly similar, with a success rate of 94% and both to be superior to the simpler heuristic approaches.