For twenty years there has been a dilemma in earthquake physics, because the observed scaling law for large earthquakes did not appear to be consistent with the stress-drop invariance of small earthquake scaling. Surprisingly, slip was seen to continue to increase with rupture length L even for events with lengths much longer than the event widths W (the brittle crust down-dip depth), whereas it might have been expected to saturate for lengths much beyond the width. If this implies that the physics of great earthquakes is somehow different from that of their smaller counterparts, this casts serious doubts on predicting the effects of the rare and damaging great events from observations of the more common smaller events. Here we bring together recently compiled observations of very large aspect ratio earthquakes with results of a 3 dimensional dynamic earthquake model to show that slip-length scaling observations are, in fact, consistent with a scale-invariant physics. Further, we discuss the origin of the large earthquake scaling in the model. ¿ 2001 American Geophysical Union