The predominant frequency (a source property) and the resonant frequencies (site effects) have been measured from local seismic events that occurred and were recorded at Campi Flegrei. The predominant frequency is measured as the period of the first half-cycle of the seismogram, which in turn is obtained as six times the standard deviation of the expectation value arrival time. Because the width of the seismic pulse is severely affected by attenuation, this effect has been accounted for through a new method in time domain, based on the broadening of a gaussian pulse propagating in a dispersive medium; the analyzed data suggest that the errors introduced by not correcting by source directivity (very difficult to carry out for local low-magnitude events) are not severe. Data also suggest that the first cycle of the seismogram may contain, accurately enough, information on the source properties, and hence allowing for the characterization of the seismic source in terms of a theoretical spectrum. Under the hypotheses (1) that the first cycles of the seismogram are not severely contaminated by propagation and (2) that resonant effects are mainly generated at later times, spectra of seismic records, corrected by attenuation and source term, provide useful information on the resonant peaks generated by local site effects. The present method of analysis shows that the frequency at which the resonant peaks are located can be accurately measured, whereas the amplitude of the peaks suffer some uncertainties.