We estimate corner frequencies and stress drops for the 2011 5.65 Mineral, Virginia, earthquake and aftershocks using a multiwindow coda spectral ratio method. We apply a Bayesian approach using a Markov chain Monte Carlo algorithm to fit the observed spectral ratios of selected event pairs with the Brune omega‐square source spectral model to obtain reliable measurements of corner frequency as well as the associated uncertainties. We show that the use of S‐wave coda provides more stable source spectral ratios than using direct S waves, and select event pairs by examining the common decay characteristics of narrowband coda envelopes. We determined an unusually high stress drop of for the mainshock, assuming a simple Brune‐type source model. The results of our modeling of the mainshock spectrum at the nearest recording station suggest similar values of stress drop for the two main subevents responsible for most of the moment release. The estimated stress drops for 46 aftershocks ( 0.71–4.13) range from to 50 MPa with a median value of 3 MPa. The aftershocks with have high stress drops within a small range of , with a median value of 16 MPa. We observe an apparent decrease of stress drop with decreasing magnitude below . Rather than suggesting a possible breakdown of self‐similar source scaling, we attribute this apparent magnitude dependence to limited signal‐to‐noise ratios for small shocks, limited frequency bandwidth, and particularly, to a group of small magnitude earthquakes that occurred in a loose cluster several kilometers to the northeast of the mainshock. Those events have unusually low stress drops, compared with aftershocks in the main aftershock cluster, and may possibly indicate triggering of seismicity on a set of relatively weak fault or joint planes. We found no temporal pattern or depth dependence in the estimated stress drops.