We investigated several methods suitable for the decomposition of the superimposed signals. The main attention has been paid to: - short-time cepstral decomposition, - decomposition via higher-order statistics using MISO and MIMO modelling, - decomposition based on the wavelet transform, i.e. done in the time-scale plane, and - decomposition with iterative multimodal parametric search. At the same time, a few modelling approaches were applied to describe the nature of real EMG signals at higher contraction forces or taken by surface electrodes. on the one hand, that were the physiological models based on volume-conductor line source and MIMO modelling, on the other hand, the ARMA models were investigated as well. The modelling experience was gathered in a computer simulator of electromyografic signals called EmgSim. It runs under the Windows environment, builds up a muscle model and simulates measurements of the EMG signal at different parameter settings using single-fibre electrode, concentric electrode, or surface electrode. Along with the screen displays, also the computer files are created in the form appropriate for further verification of the decomposition tools. The afore-mentioned decomposition methods have been extensively tested on the simulated EMGs, as well as with the real needle and surface signals. Evaluation results were presented at the V RAFT Workshop in September 1995 in Garmisch-Partenkirchen, Germany, whereas more detailed and advanced findings are revealed in this final report. The basic conclusions are gathered in the following tables: Table 1 depicts the EMG decomposition capatibilities of the validated methods, while Table 2 surveys several important features of these methods.