This thesis deals with the construction of a medical decision support system, and more specifically with the knowledge sources within the system that facilitate its operation. Simulations of some results from a proportion of these knowledge sources are created, the results correspond to the physical and electrophysiological tests carried out on a patient during neuromuscular diagnosis, and various methods of processing the acquired data for interpretation. Chaos as a method of modelling myoelectric activity is assessed for the purpose of creating an EMG simulation knowledge source and for differentiating between disorder types. The construction of phase portraits, correlation dimension analysis and calculation of Lyapunov exponents are all used to attempt to establish the presence of chaotic behaviour in the myoelectric signal. However, it is proven that the dynamics of the EMG are not chaotic in nature, thus a more suitable model for EMG simulation is chosen. The second knowledge source looked at in detail is that of EMG decomposition. Two methods of clustering MUAPs into their classes are assessed. Firstly the use of a neural network to cluster action potentials represented by correlated features and then non correlated factors. The method proves most effective when non-correlated factors are used. The second method looked at is that of multiple database principal component analysis. This method proves capable of clustering MUAP classes in the presence of noise and MUAP variation. The method is tested on real data and, within the limits of the study, the results are confirmed. A study of time requirements is made for resolution of overlapping action potentials. Two methods are considered - a fast and a more thorough one. It is established that it would be appropriate for these methods to be used in complement with one another, in a method for automatic decomposition that includes both clustering methods discussed along with various other appropriate techniques such as firing time analysis.