This paper demonstrates that the recovery rates of water level changes due to fault creep can be used to infer the position of the slipping zone relative to the well. This technique is applied to five slip-related water level changes observed in a well 460 m from the San Andreas fault near Parkfield, California during January 1989 to July 1990. The water level changes are all characterized by a rapid drop (in less than 8 hours) and a slow recovery of 15 to 30 days. The water level drop and the position of the well constrain right-lateral slip to be predominantly to the southeast of the well. The recoveries are modeled by pore fluid diffusion due to a plane strain dislocation in a porous, fluid-saturated elastic solid. Satisfactory agreement with the observations is obtained by assuming slip occurs instantaneously and extends indefinitely to the southeast of the well. The extent of the slip northwest of the well, which ranges from zero to 322 m, is determined by fitting the variations in recovery times among the five events. Calculations are carried out for the limiting cases of both an impermeable and a permeable fault. The permeability of the fault has little effect on inferences about the slip, but the inferred (horizontal) diffusivity differs for the two cases: 0.15 m2/s for the impermeable and 0.06 m2/s for the permeable. For two of the events the magnitude and distribution of slip inferred from the water level changes agree well with that measured on two nearby creepmeters. For a third event the agreement is moderately good. Discrepancies for the remaining two events suggest that the surface slip measured by the creepmeters may differ significantly from that near the 240 m depth of the well. ¿ American Geophysical Union 1993