Inference in psychological research is often made based on group level data which is aggregated across participants. While this practice is useful to compare general characteristics between groups, it can also draw a misleading picture about processes on the individual level if the properties of the group data do not equal those of the individual data. As has been recognized previously, this is particularly problematic when looking at learning processes (e.g., Estes, 1956; Gallistel, 2004): whereas aggregated learning curves often suggest a gradual increase over time, individual learning curves can be step-like, but show variability in when learning begins. As we point out for the example of repetition learning, this can also lead to misconceptions about the cognitive mechanism underlying the observed data. To overcome this fallacy, we introduce a Bayesian hierarchical modeling approach to model learning curves on the individual level. Our model focuses on three characteristics of the individual curve: 1) Is learning happening at all, 2) When is learning happening, and 3) How fast is learning happening. Whereas this provides a more fine-grained description of individual learning curves, the hierarchical nature of our approach also allows to compare these characteristics about the learning process on the level of groups. As we show in examples from repetition learning experiments, our model allows to compare learning processes on a high resolution between individuals and can provide a better understanding of the mechanisms underlying observed differences. We discuss the applicability of our approach to a broader range of learning tasks.