Cortical areas are linked through pathways which originate and terminate in specific layers. The factors underlying which layers are involved in specific connections are not well understood. Here we tested whether cortical structure can predict the pattern as well as the relative distribution of projection neurons and axonal terminals in cortical layers, studied with retrograde and anterograde tracers. We used the prefrontal cortices in the rhesus monkey as a model system because their laminar organization varies systematically, ranging from areas that have only three identifiable layers, to those that have six layers. We rated each prefrontal area based on the number and definition of its cortical layers (level 1, lowest; level 5, highest). The structural model accurately predicted the laminar pattern of connections in approximately 80% of the cases. Thus, projection neurons from a higher-level cortex originated mostly in the upper layers and their axons terminated predominantly in the deep layers (4-6) of a lower-level cortex. Conversely, most projection neurons from a lower-level area originated in the deep layers and their axons terminated predominantly in the upper layers (1-3) of a higher-level area. In addition, the structural model accurately predicted that the proportion of projection neurons or axonal terminals in the upper to the deep layers would vary as a function of the number of levels between the connected cortices. The power of this structural model lies in its potential to predict patterns of connections in the human cortex, where invasive procedures are precluded.