Slip distribution along a laboratory fault, which consists of eight spring‐connected blocks that are elastically driven to slide on a frictional surface, has been examined for a “long” sequence of slip events to test the applicability of some conceptual models proposed recently in the literature. The distributions of large slip events are found to be quite variable and do not fit the uniform slip or characteristic earthquake models. The rupture initiation points are usually not near the corresponding maximum slip points, in contrast to observations by Thatcher (1990) and by Fukao and Kikuchi (1987) that earthquake hypocenters are commonly near corresponding regions of maximum slip in the fault planes. This contrast may suggest that either the present observations or theirs are not representative or the teleseismically determined hypocenters may not always be true rupture initiation points as usually assumed. Large slip events are also found to be a stress‐roughening process. They are triggered by some small events after the stresses have been adjusted by some earlier small‐to‐moderate events to be near the critical levels at most locations along the fault. This suggests that earthquake prediction monitoring efforts should not be limited to a small region near an asperity but should be spread out to cover the entire fault segment in a seismic gap in order to detect the condition of simultaneous strain buildup.