We estimate the source parameters for a selection of microearthquakes that occurred at Mount St Helens in the period 1995–1998. Excluding the activity of 2004 September, this time period includes the most intense episode of earthquake activity since the last dome-building eruption in 1986 October. 200 seismograms were processed to obtain seismic moments, source radii, stress drops and average fault slip. The source parameters were determined from the spectral analysis of P waves, after correction for attenuation and site effects. In particular, P-wave quality (Q_p) and site (S) factors have been previously calculated in the frequency ranges 2–7 Hz and 18–30 Hz. Because it was impossible to perform corrections for Q_p and S over the whole spectrum we applied a new approach, based on the notion of ‘holed spectrum’, to estimate spectral parameters. The term ‘holed spectrum’ indicates a spectrum lacking corrected spectral amplitude values at certain frequencies. We carried out a statistical study to verify that dealing with the ‘holed spectrum’ does not lead to significant differences in the estimates of spectral parameters. We also investigated the dependence of spectral parameters (low-frequency level, corner frequency and high-frequency decay) on the bandwidth of spectral hole, and defined the threshold values for three different spectral models. Displacement ‘holed spectra’, corrected by attenuation and site response, are then used to determine spectral parameters in order to calculate seismic source parameters. Seismic moments range from 10¹⁷ to 10¹⁹ dyne-cm, source dimensions from 100 to 350 m, and average fault slip from 0.003 to 0.1 cm. Self-similarity seems to break down in that stress drops are very low (0.1–1 bars). We postulate that seismicity is associated with a brittle shear failure mechanism occurring in a highly heterogeneous material under a relatively low stress regime.