We invert peak ground velocity and acceleration (PGV and PGA) to estimate rupture direction and rupture velocity for 47 moderate earthquakes (3.5≥M≥5.4) in northern California. We correct sets of PGAs and PGVs recorded at stations less than 55–125 km, depending on source depth, for site amplification and source–receiver distance, then fit the residual peak motions to the unilateral directivity function of Ben-Menahem (1961). We independently invert PGA and PGV. The rupture direction can be determined using as few as seven peak motions if the station distribution is sufficient. The rupture velocity is unstable, however, if there are no takeoff angles within 30° of the rupture direction. Rupture velocities are generally subsonic (0.5β–0.9β); for stability, we limit the rupture velocity at v=0.92β, the Rayleigh wave speed. For 73 of 94 inversions, the rupture direction clearly identifies one of the nodal planes as the fault plane. The 35 strike-slip earthquakes have rupture directions that range from nearly horizontal (6 events) to directly updip (5 events); the other 24 rupture partly along strike and partly updip. Two strike-slip earthquakes rupture updip in one inversion and downdip in the other. All but 1 of the 11 thrust earthquakes rupture predominantly updip. We compare the rupture directions for 10 M≥4.0 earthquakes to the relative location of the mainshock and the first two weeks of aftershocks. Spatial distributions of 8 of 10 aftershock sequences agree well with the rupture directivity calculated for the mainshock.