AbstractAccess to large portions of our knowledge is unstable, so that recall fluctuates depending upon momentary context. Laboratory interventions can stabilize access for a time, either by preventing the loss of precarious access (preventive maintenance) or by re‐establishing lost access (corrective maintenance). This paper describes two methods of measuring the magnitude and duration of such intervention effects. Preventive maintenance effects are measured by the degree and the persistence of asymmetry of up‐ versus down‐fluctuations on successive attempts to recall. The preventive maintenance effect of a single successful recall trial has a half‐life in excess of 1 month. The methods are suitable for comparing the effectiveness of various schedules and types of maintenance interventions for a variety of semantic memory content and for diverse subject popu