Percepts are naturally grouped into meaningful categories to process continuous stimulus variations in the environment. Theories of category acquisition have existed for decades, but how they arise in the brain due to learning is not well understood. Here, advanced computational modeling techniques borrowed from educational data mining and cognitive psychology were used to trace the development of auditory categories within a short-term training session. Nonmusicians were rapidly trained for 20 min on musical interval identification (i.e., minor and major 3rd interval dyads) while their brain activity was recorded via EEG. Categorization performance and neural responses were then assessed for the trained (3rds) and novel untrained (major/minor 6ths) continua. Computational modeling was used to predict behavioral identification responses and whether the inclusion of single-trial features of the neural data could predict successful learning performance. Model results revealed meaningful brain-behavior relationships in auditory category learning detectible on the single-trial level; smaller P2 amplitudes were associated with a greater probability of correct interval categorization after learning. These findings highlight the nuanced dynamics of brain-behavior coupling that help explain the temporal emergence of auditory categorical learning in the brain.