We present an ART-based neural network model (adapted from ) of the development of discrimination-shift learning that models the trial-by-trial learning process in great detail. In agreement with the results of human participants (420 years of age) in  the model revealed two distinct learning modes in the learning process: (1) a discontinuous rational learning process by means of hypothesis testing; and (2) a slow, yet discontinuous learning process. Categorical differences in behavior are the result of uniformly distributed dimensional preferences. In addition, it models the developmental differences between reversal and nonreversal-shift learning. The network implements attention-guided learning by selective sensory processing based on dimensional preferences mediated through reinforcement. The developmental differences consist of separate adjustment of the valuation of negative reinforcement, which is proposed in the empirical neuroscience literature .