While it is a relatively well understood problem to determine hypocentral depths from teleseismic data or from dense local networks, many eastern Canadian earthquakes are too small to be recorded teleseismically and occur in regions where the network density is adequate for reliable epicentral determination but insufficient for hypocentral determination. Using stacked, filtered (2–6 Hz bandpass) regional data (recorded at 250 to 500 km), we have developed a method for rapid depth estimation of small and moderate earthquakes. We pick the largest amplitude arrival in the Pn wavetrain, which we call Pmax, and measure its arrival time relative to the onset of Pn. Because Pmax appears to be related to the same phase, sPn, we can use it reliably to determine depths. The method was calibrated using earthquakes in the Charlevoix, Québec, seismic zone, where the network density is sufficient for depth determination as part of the location procedure. The standard error is 2.3 km, assuming that the Charlevoix network depths are correct. In only one case (out of twelve) did the network and Pn-Pmax depths differ by more than 5 km, implying that the assumption that Pmax is usually related to the same phase is valid. The Pn-Pmax method was then applied to earthquakes in the western Québec and Lower St. Lawrence seismic zones. Results show that earthquakes in western Québec can occur over a wide range of crustal depths but, on average, are shallower than the 18-km depth currently used as a default. Earthquakes in the lower St. Lawrence appear to be deeper than the default, but this conclusion is based on only two earthquakes.