We analyze a set of 76 mapped surface ruptures for relationships between geometrical discontinuities in fault traces and earthquake rupture extent. The combined set includes 46 strike‐slip, 16 normal, and 14 reverse mechanism events. The survey shows ∼90% of ruptures have at least one end at a mappable discontinuity, either a fault end or a step of 1 km or greater. Dip‐slip ruptures cross larger steps than strike‐slip earthquakes, with maxima of ∼12 versus ∼5  km, respectively. Large steps inside strike‐slip ruptures are rare; only 8% (5 of 62) are ≥4  km. A geometric probability distribution model of steps as “challenges” to rupture propagation predicts that steps of 1 km or greater will be effective in stopping rupture about 46% of the time. The rate is similar for dip‐slip earthquakes, but, within this set, steps are relatively more effective in stopping reverse ruptures and less effective in stopping normal ruptures. By comparing steps at rupture terminations to the set of steps broken in rupture, we can estimate the importance of step size for stopping rupture. We define the passing ratio for a given step size as the fraction of steps broken divided by the corresponding fraction that stop rupture. A linear model for steps from 1 to 6 km in strike‐slip ruptures leads to the passing ratio =1.89–0.31× step width. Steps of ∼3  km are equally likely to be broken or to terminate rupture, and steps ≥6  km should almost always stop rupture. A similar comparison suggests that extensional steps are somewhat more effective than compressional steps in stopping ruptures. We also compiled the incidence of gaps of 1 km and longer in surface ruptures. Gaps occur in ∼43% of ruptures and occur more frequently in dip‐slip than strike‐slip ruptures.

Online Material: Figures of annotated surface rupture maps for 40 earthquakes.

You do not currently have access to this article.