Learning to learn is a separate function from classical learning, which makes our learning more efficient and flexible. It is particularly associated with primates (Harlow, 1953) and linked to the lateral prefrontal cortex (LPFC) by lesions studies (Browning et al., 2007). LPFC forms an interacting circuit with the midcingulate cortex (MCC) which plays an important role in outcome and performance evaluation (Procyk et al., 2016). Most electrophysiological recordings are made after extensive training, so they ignore neurophysiological changes that permit learning to learn. This project identifies these crucial changes using longitudinal recordings, starting with task naïve animals. We recorded two monkeys with 256 chronic intracortical electrodes (FMA, Microprobes) in the LPFC and MCC of both hemispheres. They performed a series of touchscreen tasks, progressively advancing from simple touch tasks, through classical discrimination and reversal learning to more complex problem solving tasks. The tasks provided scope for both simple problem learning and also learning to learn, through which acquired strategies improved long-term performance. We performed single-unit, local-field, and evoked potential analyses to investigate the longitudinal changes in the neural correlates of choice and outcome processing before, during, and after the animal learned to learn. We were able to track the emergence of previously described outcome derived responses in monkeys making cognitive task choices for the very first time. Signals differentiating between correct and incorrect feedback showed several distinct progressions throughout learning, describing specific markers of learning generalization. These results provide important context to previous results obtained only from over-trained subjects.