To bridge theory and experiment, we must test brain computational models that operate with neurobiologically plausible subcomponents, explain brain activity measurements, perform complex cognitive tasks, and generalize to new ones. In computational neuroscience, the observation that archetypal neuronal circuits exist in repeated motifs throughout the brain, gave rise to an important class of population density models. These models attempt to explain empirical brain activity measurements with computational microcircuits, hence bridging theory and experimentation. Motivated principally by the organizational principles of neurobiology, these models have evolved to describe a larger dynamic repertoire by incorporating hierarchical characteristics reminiscent of neural networks, yet are typically not used to accomplish tasks or predict behavior. Because information exchange incurs at a cost that scales with distance, we would expect regions that need to interact at higher bandwidth, higher frequency, and shorter latency to be more proximally located in the brain. From this perspective, shorter path lengths are expected to enable higher frequency communication. However, the majority of short distance connections in the brain are within the gray matter, and a large portion of axons in the neuropil are unmyelinated. Here, I will discuss aspects of latency and structural organization and their implications for brain computational models.