We examine two closely located earthquakes in Japan that had identical moment magnitudes Mw but significantly different energy magnitudes Me. We use teleseismic data from the Global Seismograph Network and strong-motion data from the National Research Institute for Earth Science and Disaster Prevention’s K-Net to analyze the 19 October 1996 Kyushu earthquake (Mw 6.7, Me 6.6) and the 6 October 2000 Tottori earthquake (Mw 6.7, Me 7.4). To obtain regional estimates of radiated energy ES we apply a spectral technique to regional (<200 km) waveforms that are dominated by S and Lg waves. For the thrust-fault Kyushu earthquake, we estimate an average regional attenuation Q(f)=230f0.65. For the strike-slip Tottori earthquake, the average regional attenuation is Q(f)=180f0.6. These attenuation functions are similar to those derived from studies of both California and Japan earthquakes. The regional estimate of ES for the Kyushu earthquake, 3.8×1014 J, is significantly smaller than that for the Tottori earthquake, ES 1.3×1015 J. These estimates correspond well with the teleseismic estimates of 3.9×1014 J and 1.8×1015 J, respectively. The apparent stress (τa=μES/M0, with μ equal to rigidity) for the Kyushu earthquake is 4 times smaller than the apparent stress for the Tottori earthquake. In terms of the fault maturity model, the significantly greater release of energy by the strike-slip Tottori earthquake can be related to strong deformation in an immature intraplate setting. The relatively lower energy release of the thrust-fault Kyushu earthquake can be related to rupture on mature faults at a subduction environment. The consistence between teleseismic and regional estimates of ES is particularly significant as teleseismic data for computing ES are routinely available for all large earthquakes whereas often there are no near-field data.