The frequency dependence of the seismic quality factor, Q, is important in interpreting seismic wave spectra for nuclear test descrimination, earthquake damage studies, and in exploration. In this study, P and S wave spectra from 18 events recorded during three 1977 microearthquake swarms are used to compute Q(f) (frequency range =3--40 Hz) for a seismically active geothermal region of disturbed upper crust in the central Rio Grande rift near Socorro, New Mexico. Numerically differentiated spectra are fit with a model of the form Q(f)=Q0f&eegr; and standard linear regression techniques used to compute Q0 and &eegr;, along with their uncertainties. Only two out of 18 events in these swarms exhibit significant Qp or Qs frequency dependence at the 95% confidence level. This is primarily due to large uncertainties in &eegr; generated by upper-crustal reverberations. Averaging Qp and Qs over events in each swarm shows that for two swarms, a significant frequency dependent Qp exists with &eegr; ranging from 0.29 to 0.67 and Qop ranging from 5 to 17 (95% confidence intevals). A strong decrease in Qp or Qs with frequency (&eegr;<-0.2) appears to be ruled out. Likewise, the frequency dependence of Qp and Qs does not appear to change with event distance over the range 15.4--16.9 km. Smoothed spectra produce less frequency dependence and higher Qo values than smoothed spectra (the only significant Q(f) is Qos=123 and &eegr;=0.33 for on swarm). This suggests that much of the apparent freqency may be due to ripple in the spectra (caused by scattered phases in the time window, or the process of windowing itself) and that this technique should only be used with ''smooth'' spectra. ¿ American Geophysical Union 1987