(a) Schematic of recall prediction for one list, in our modeling framework. First, the study period is simulated and the model learns the list. Then retrieval is simulated, and the model calculates the probability of recalling each of the items, as well as the probability of stopping recall. We then record the predicted probability of the observed behavior. If an item was recalled, we update the state of the model conditional on that recall. This process is repeated until the entire observed recall sequence has been simulated. (b) Schematic example of one list with five words. At each step of the recall process, the model makes predictions conditional on the observed behavior up to that point.