<p>Comparing mouse and marmoset brain architectures is a problem of current interest. This is challenging since these brains have distinct spatial organizations and expert neuroanatomists disagree. We propose a theoretical framework to relate named atlas compartments across taxa and use it to establish correspondences between marmoset and mice brains. Contrary to conventional wisdom that brain structures may be easier to relate at higher levels of the atlas hierarchy, we find that finer parcellations at the leaf levels, namely cytoarchitectonically distinct non-overlapping regions at the lowest level of the naming hierarchy, offer greater reconcilability despite naming discrepancies. The disagreements among existing atlases are often about how to group brain regions hierarchically into larger structures. We created a list of leaf-level structures for both species and established five types of correspondence between them. One-to-one relations were found between 43% of the structures in mouse and 47% in marmoset. 25% of mouse and 10% of marmoset structures were not relatable and the remainder show more complex mappings. We developed a computational tool for querying and visualizing relationships between the corresponding brains. Our findings provide a basis for computational comparative analyses of the laboratory mouse and the common marmoset brains.</p>