Measurement of gross alpha and gross beta activity using liquid scintillation counting is a screening method used for e.g. drinking water. By the use of pulse shape analysis an instrument can discriminate pulses caused by alpha particles from pulses caused by beta particles. However, misclassification (also called spillover) occurs, i.e. one of the pulse types are wrongly analysed as the other one and vice versa. Measurements are usually done for instrument settings where the alpha and beta spillover are equal. This is referred to as optimum. However, this instrument setting may restrict the alpha/ beta activity ratios that can be measured with a measurement uncertainty that would be fitfor-purpose. If measurements are to be done at low- or high activity ratios, it would be better to further lower the misclassification from the most abundant pulse type. For example, if a sample contains much more beta than alpha activity, a further decreased beta misclassification would then facilitate that a lower gross alpha/gross beta activity ratio could be measured. In this work, calibrations at different discrimination levels were done for some alpha and beta emitting radionuclides, which then formed the basis for modelling misclassification factors and measurement efficiencies. Furthermore, the method was validated concerning sensitivity for parameters that might be difficult to control in samples. The method was also studied with respect to uncertainty modelling as a function of the gross alpha/gross beta activity ratio, resulting in the lowest and highest activity ratio for which the method would result in measurement uncertainties that could be used for decision-making. Finally, uncertainty propagation was compared with Monte Carlo calculation in order to study deviations in the resulting activity distributions. The work resulted in a method with acceptable measurement uncertainties for samples where the quench levels are not too high (SQPE > 740) and for alpha/beta activity ratios from about 0.001 to about 10. The calculated measurement uncertainty includes correction for systematic deviation caused by sample parameters that might be difficult to control in a measurement.