THEORETICAL and experimentally based laws for seismic faulting contain a critical slip distance1-5, Dc, which is the slip over which strength breaks down during earthquake nucleation. On an earthquake-generating fault, this distance plays a key role in determining the rupture nucleation dimension6, the amount of premonitory and post-seismic slip7-10, and the maximum seismic ground acceleration1,11. In laboratory friction experiments, Dc has been related to the size of surface contact junctions2,5,12; thus, the discrepancy between laboratory measurements of Dc (??? 10-5 m) and values obtained from modelling earthquakes (??? 10-2 m) has been attributed to differences in roughness between laboratory surfaces and natural faults5. This interpretation predicts a dependence of Dc on the particle size of fault gouge 2 (breccia and wear material) but not on shear strain. Here we present experimental results showing that Dc scales with shear strain in simulated fault gouge. Our data suggest a new physical interpretation for the critical slip distance, in which Dc is controlled by the thickness of the zone of localized shear strain. As gouge zones of mature faults are commonly 102-103 m thick13-17, whereas laboratory gouge layers are 1-10 mm thick, our data offer an alternative interpretation of the discrepancy between laboratory and field-based estimates of Dc.