The neural circuits supporting short- and long-term memory are often considered distinct and independent from each other. However, during the pursuit of immediate goals, different timescales of learning and memory systems are in constant interaction. By using unique, longitudinal experimental designs for human neuroimaging we can detail how working memory processes and representations change at different stages of learning. These approaches provide critical bridges between long-term and working memory systems. 

To detail the neural and cognitive machinery for working memory in the prefrontal cortex, we compare functional and anatomical neuroimaging measurements with analyses of large-scale neurophysiological data. This framework uniquely considers the neural architectures across scales from millimeters (voxels) to microns (single-cells). 

Then, we construct multi-area neural network models with neurobiologically inspired parameters (see: psychrnn.readthedocs.io/) trained on cognitive tasks. We can then interrogate what specific computations, algorithms, and training regimes enable successful working memory and flexible cognition, and how these processes may be implemented in the human brain.