One of the main goals of memory research is to identify why false memory errors occur. Much work has been done in prior research to identify and define false memory processes using indirect measurements for them, such as using response latencies to hypothesize how one memory process may be faster over another, or inferring how recollecting contextual details of a target is affected by the order in which items were presented during encoding. A weakness in these indirect methods of measuring processes is just that: they are indirect measurements. The advantage of memory models is that they provide direct measurements of memory processes, allowing researchers to test assumptions about how manipulations affect them, and comparing them against each other. The theme of this dissertation is to use new models and modeling techniques to answer questions that previously relied on indirect methods of identifying processes. In Experiment 1, a new model was created to accommodate two source memory designs used to measure overdistribution in episodic memory. The model was able to provide insight that the two designs tapped into different memory processes, and thus were not measuring the same overdistribution metric as previously assumed. In Experiment 2, a new methodology in modeling was implemented so that relative process speeds could be measured alongside process parameter estimates. The new latency extension of a widely used recognition model was fit to previously collected data, providing a first look into how familiarity may not be the fast process responsible for false memory under fast response deadlines as previously believed. Experiment 3 addressed weaknesses from Experiment 2, and supported its findings that context recollection was faster than familiarity. These experiments demonstrate how memory models provide a simple but powerful tool to answer questions that could only be inferred previously.