Apparent stress &tgr;a is defined as &tgr;a=&eegr;&tgr;¿, where &tgr;¿ is the average shear stress loading the fault plane to cause slip and &eegr; is the seismic efficiency, defined as Ea/W, where Ea is the energy radiated seismically and W is the total energy released by the earthquake. The results of a recent study in which apparent stresses of mining-induced earthquakes were compared to those measured for laboratory stick-slip friction events led to the hypothesis that &tgr;a/&tgr;¿≤0.06. This hypothesis is tested here against a substantially augmented data set of earthquakes for which &tgr;¿ can be estimated, mostly from in situ stress measurements, for comparison with &tgr;a. The expanded data set, which includes earthquakes artificially triggered at a depth of 9 km in the German Kontinentales Tiefbohrprogramm der Bundesrepublik Deutschland (KTB) borehole and natural tectonic earthquakes, covers a broad range of hypocentral depths, rock types, pore pressures, and tectonic settings. Nonetheless, over ~14 orders of magnitude in seismic moment, apparent stresses exhibit distinct upper bounds defined by a maximum seismic efficiency of ~0.06, consistent with the hypothesis proposed before. This behavior of &tgr;a and &eegr; can be expressed in terms of two parameters measured for stick-slip friction events in the laboratory: the ratio of the static to the dynamic coefficient of friction and the fault slip overshoot. Typical values for these two parameters yield seismic efficiencies of ~0.06. In contrast to efficiencies for laboratory events for which &eegr; is always near 0.06, those for earthquakes tend to be less than this bounding value because Ea for earthquakes is usually underestimated due to factors such as band-limited recording. Thus upper bounds on &tgr;a/&tgr;¿ appear to be controlled by just a few fundamental aspects of frictional stick-slip behavior that are common to shallow earthquakes everywhere. Estimates of &tgr;¿ from measurements of &tgr;a for suites of earthquakes, using &tgr;a/&tgr;¿≤0.06, are found to be comparable in magnitude to estimates of shear stress on the basis of extrapolating in situ stress data to seismogenic depths.