Quantitative models of human memory often rely on assumed latent memory processes, such as patterns of rehearsal of the words on a study list. Consequently, the application of memory models that assume latent rehearsals typically make use of overt rehearsal data. However, these data are not always available in some settings where the application of memory models has proven useful (e.g., clinical assessments of memory performance). In this paper, we show Bayesian statistical methodology can be used to infer the latent pattern of rehearsals needed to successfully apply a temporal model of memory to a clinical data set. We discuss the relevance of this research for those interested in neuropsychological assessment as well as cognitive psychologists interested in basic memory research.