In the present study, we investigated the accuracy of eyewitness accounts over time from a metacognitive perspective, in which post-retrieval monitoring and control processes play a crucial role in mediating between memory retrieval and ultimate memory performance. In two experiments, participants viewed a narrated slide show depicting ordinary daily events and were questioned about fine-grained event details, either immediately or after a delay (of either 24 or 48 h). High motivation for accurate responding was induced via monetary incentives (Exp. 1) or instructions (Exp. 2). Using Koriat and Goldsmith's (1996) "Quantity-Accuracy Profile" methodology, we were able to isolate the cognitive and metacognitive components posited to underly free-report memory accuracy, and to examine them over time. Our results showed that, even under conditions of free-report and high motivation for accurate responding, the accuracy of memory reports declined substantially over time, largely due to reduced monitoring effectiveness (i.e., monitoring resolution) rather than to changes in control policy (i.e., report criterion). As predicted, the decline over time in confidence was more pronounced for true than for false memories, such that the metacognitive ability to differentiate between correct and incorrect answers deteriorated with time. This poorer monitoring resolution resulted in an increased proportion of errors among the volunteered responses, and consequently, in lower free-report accuracy. Our results shed light on the manner in which memory accuracy over time is affected by changes in the effectiveness of the metacognitive processes that operate during memory reporting.