People are incredibly good at adapting their movements to altered circumstances, likely engaging multiple learning processes. How multi-process motor learning dynamics depend on the task or change with age is still unclear. In several experiments we investigate how well a two-rate model with a fast and slow process (Smith et al., 2006) can account for changes in task demands, and how age affects two-rate adaptation. Each participant separately learns two rotations in counterbalanced order and separate parts of the workspace. This way we can compare model fits on data from two tasks within each participant. We tested if two-rate model fits are affected by providing continuous or terminal feedback. Terminal feedback disambiguates error size, which would benefit error-based models of motor learning. However, adaptation with terminal feedback could be explained with a single-rate process, so that continuous feedback seems better suited for studying multi-rate motor learning. Second, we tested how older and younger adults learn either an abruptly or gradually introduced rotation. We found no effect of age, but this may be due to our choice of paradigm, in particular the final part of the task that's meant to separate the two model processes. We test this in a third experiment. Each participant does two of four different paradigms, with identical initial learning. The final parts of the task differ, affecting the model fits' fast and slow contributions to initial learning. Our results guide experimental design for two-rate model studies, and show that within-participant paradigms are both possible and useful.Acknowledgments: NSERC; DFG EG 145/1-1