The spectral characteristics of regional Pn waves recorded at the High Frequency Seismic Element (HFSE) at the NORESS array in southern Norway are investigated to determine if spectral slope can be used to discriminate between earthquakes and chemical explosions. Displacement spectra are calculated for 134 events in southern Norway that include 36 known earthquakes and 36 known chemical explosions. Events range in distance from 230 to 620 km, and most have signal above the noise at frequencies over 30 Hz. The average frequency at the peak signal-to-noise ratio (SNR) is 15.5 Hz. Spectral slopes are measured on a displacement spectrum calculated with a 2-sec window. The slopes range between ω−1.8 and ω−5.1, and there is no strong evidence that the slope steepens with distance for this data set. According to one theory, earthquakes should have steeper slopes than explosions, but when this data set is separated into earthquakes and explosions we do not see separation between sources with respect to the slope of the spectrum. Both source types have events with the same range of spectral slopes, and the mean slopes and standard deviations are similar. Spectra of events at the same locations are often similar. However, explosions from the Blåsjø Mine in western Norway show variability in the spectral slope. Since the path and distance to HFSE are the same for these events, the variability must be due to differences in the source mechanism. For a set of small earthquakes, however, the observed variability in spectral slopes appears to be more related to distance and path effects. Therefore, we observe as much variation in spectral slope for one source type as there is between source types. The time window over which the spectrum is calculated is varied from 1 to 8 sec to determine if a time window consisting of a pure Pn wave was required to discriminate between source types. Spectral slope usually decreases as the window extends deeper into the coda and the energy content of the signal changes. The slopes also become more stable and more reliable as longer windows are used. However, source discrimination is not possible with any time window.