This Viewpoint argues that understanding how the brain controls behavior requires an explicitly evolutionary framework. The mammalian brain did not emerge through the replacement of earlier circuits with perfect alternatives but rather through the elaboration of existing circuits along with the addition of new ones, yielding a hierarchical architecture in which many ancient spinal and brainstem circuits remain functionally essential. In this context, cortex does not directly control behavior but exerts its influence via layer 5 projections to evolutionarily older subcortical motor centers. This view challenges the prevailing corticocentric bias in neuroscience, which often treats cortex as a largely self-contained computational system. We propose that key functions such as attention and efference copy are best understood within this layered organization. Attention may reflect competitive filtering of corticofugal outputs at subcortical bottlenecks, while efference copies arise naturally from branching motor pathways distributed across hierarchical levels that reflect evolutionary history. Crucially, these principles expose important limitations in current computational models, which typically omit subcortical circuitry and treat motor output as a terminal stage of processing. An evolutionary perspective instead demands models that integrate cortex with spinal, brainstem, and midbrain systems as interacting components of a unified sensorimotor hierarchy. Incorporating these constraints will be essential for developing biologically grounded theories of brain function.