We explore whether observations of average surface rupture properties among strike-slip earthquakes reflect the underlying mechanics. We compare the observed relationship between average slip and rupture length for 27 surface ruptures (18 plate-boundary earthquakes and 9 away from transform plate boundaries) with predictions from two families of uniform-stress-drop models. Purely elastic models with a rupture-limiting locking depth predict a nonlinear relationship while a quasidynamic model with no locking depth predicts a linear relationship. We explore whether observations of fault slip at the Earth's surface distinguish which, if either, of these two families of models may be favored. We find that the data provide insufficient constraints to rule out either a linear or a nonlinear relationship. This might arise from uncertainties in the observations or from reasonable (but unrecoverable) variations in locking depths and uniform stress drops among earthquakes. We advance an alternative interpretation that the complexity among the observations is consistent with dynamic rupture models featuring spatially varying stress drops.