Prolonged static strain on the muscles of the neck-shoulder region is believed to be linked to the development of musculoskeletal problems. To quantify the static strain on the basis of EMG, the level as well as the duration of the muscle load should be analysed on temporal basis. In this paper, some methods for the temporal analysis of EMG recordings are proposed with an aim of quantifying the long-term static strain on the muscle. The use of nonlinear median prefilters for decomposing the EMG activity according both to amplitude level and duration of the activity at different levels is proposed. The prefiltering methods were also evaluated using laboratory studies. The main aim of the studies was to compare the estimation errors between EMG and force for different types of prefilters especially when the static load component was analysed. The average estimation error for sequences having a duration longer than 1 s was found to be 8% of MVC in the case of trapezius muscle and 14% of MVC in the case of biceps brachii muscle. Linear relation was found on the basis of linear least squares curve fitting to give the largest correlation coefficients between EMG and force, when the static load component was analysed.