Global and regional surveys of earthquakes show that empirically determined b-values of the Gutenberg-Richter distribution are remarkably stable, generally limited to values of -1 +/- 0.2. Here I interpret observations from California, New Zealand, and Japan to suggest that the stability of the b-value is a manifestation of a physical process; specifically, the tendency of crustal strains to organize along relatively discrete zones. Given a portion of the earth's crust subject to a displacement field, displacement is accommodated by a system of fault lengths that obey a power law distribution. With continued displacement, longer faults develop at the expense of shorter faults and take up an increasingly greater portion of the displacement budget. Shifts in regional displacement directions lead to reversals of these trends. The changes in the ratio of long faults to short faults in a region is thus accompanied by changes in the relationship between fault length and fault slip rate. Because the recurrence of earthquakes on faults is a function of both fault length and fault slip rate, the deformation process may result in a system of feedback that inhibits changes in the b-value as fault populations change. A corollary to the idea is that the magnitude-frequency statistics of seismicity may be attributed to the same physical processes that are responsible for the development of plate tectonic boundaries and the gross spatial distribution of seismicity around the globe.