Scientists challenge 70-year-old ‘lizard brain’ theory of human intelligence

For decades, the human brain has been portrayed as an evolutionary skyscraper: ancient instincts at the bottom, emotions in the middle and sophisticated reasoning at the top. New research from Georgia Tech argues that this familiar “lizard brain” story is far too simple.

The study, published in Science Advances, compares biological brains across species with experiments on artificial neural networks. It proposes that brain evolution may have involved competition between fundamentally different ways of wiring neural circuits. Limited space inside the brain could have forced those systems to compete, shaping the evolution of intelligence.

The older model, developed in the 1950s, divided the brain into evolutionary layers. Basic bodily functions occupied the bottom, an emotion-driven reptilian brain sat above them, and the neocortex represented complex human reasoning. Researchers now emphasize that the neocortex also handles vision and perception, while the limbic system, often loosely labeled the reptilian brain, contributes to memory, smell, navigation and emotional regulation.

Rather than treating these regions as separate structures, the Georgia Tech team examined how they change together across species. When one part of the limbic system was relatively large, other parts were generally larger too, while the neocortex tended to occupy a smaller share of the brain. That coordinated pattern suggested that the regions operate as interconnected systems whose sizes shift together during evolution.

The researchers then considered whether differences in wiring could explain the trade-off. Neocortical circuits often use localized spatial maps, linking neighboring body parts or organizing visual and auditory information by location. Limbic connections are more distributed, with patterns of neural activity representing complex information such as smells and memories. In artificial networks, localized connections performed well on vision, sound and touch tasks, while distributed wiring proved better suited to complex information.