Recently, Bouchon (1979a) reinterpreted strong motion seismograms obtained during the Parkfield earthquake of 1066 using a new method applicable to a finite propagating dislocation source in a layered medium. His results and other pertinent data, interpreted in terms of the barrier model of Das and Aki (1977), suggest that the rupture may be stopped by a barrier with the specific fracture energy of about 109 erg cm-2. Using the formulas of Ida (1973b), we estimated parameters of the barrier as follows: breaking slip of about 20 cm, cohesive stress of about 100 bars, and length of end zone (nonelastic zone) of a few hundred meters. The barrier parameters for the great 1857 earthquake were also obtained from the description of surface fault breaks by Wallace (1968). The result led to the estimation of maximum acceleration of about 1.5g near the fault, under the assumption that the end zone length is proportional to the diameter of individual crack of the barrier model. Barriers for other earthquakes are discussed, and they are classified into geometrical barriers such as fault bend and corner and inhomogeneous barriers such as the high velocity anomaly straddling the San Andreas fault near San Juan Bautista. The barriers act not only as a stopper of rupture but also as an initiator of rupture, as well as a stress concentrator, causing twin earthquakes and migration or progression of major earthquakes along the plate boundary.