How brain networks work together is key to human intelligence

Modern neuroscience understands the brain as a set of specialized systems. Aspects of brain function such as attention, perception, memory, language, and thinking have been mapped into separate brain networks and each has been examined largely in isolation.

Although this approach has made great strides, it has left unresolved one of the most fundamental facts about human cognition: its overall unity as a coherent system.

“Neuroscience has been very successful in explaining what specific networks do, but much less successful in explaining how a single, coherent mind emerges from their interactions,” said Aaron Barbe, professor of psychology in the University of Notre Dame’s Department of Psychology.

How cognitive relationships constitute ‘general intelligence’

Psychologists have long known that areas as diverse as attention, perception, memory, and language are interrelated, forming what they call “general intelligence.” It accounts for how people function and adapt in a wide range of academic, occupational, social and health contexts. It shapes how efficiently we learn, reason, and perform in response to many everyday problems and tasks.

For more than a century, this framework has suggested that knowledge is unified at a fundamental level. What has been missing is a theory to explain why such unity exists.

“The problem of intelligence is not one of functional localization,” says Barbe, director of the Notre Dame Human Neuroimaging Center and Decision Neuroscience Laboratory.

“Contemporary research often asks where general intelligence originates in the brain—focusing primarily on a specific network of regions within the frontal and parietal cortex. But a more fundamental question is how intelligence emerges from principles that govern global brain function—how distributed networks communicate and collectively process information.”

Barbe and his research team, including Notre Dame graduate student and lead author Ramsey Wilcox, investigated the predictions of a unifying framework known as network neuroscience theory.

Their research appears in journals Nature communication.

Network Neuroscience Theory

General intelligence itself is not a skill or a technique, the researchers argue. It tends to be positively correlated with a pattern-differentiated ability. The research argues that this pattern reflects differences in how efficiently brain networks are organized and work together.

To test this claim, cognitive neuroscientists analyzed brain imaging and cognitive data from one of the largest studies conducted to date, examining 831 adults in the Human Connectome Project, along with an independent sample of 145 adults in the INSIGHT Study, which was funded by the Intelligence Advanced Research Projects Program. Researchers have integrated measurements of both brain structure and function to enable more precise characterization of the human brain.

Rather than identifying intelligence with a specific cognitive function or brain network, network neuroscience theory identifies it as a property of how the brain works as a whole. In this view, intelligence reflects how the brain’s networks are coordinated and dynamically reorganized to meet the various challenges we face in life.

According to Barbe and Wilcox, this study represents an important change.

“We found evidence of system-wide coordination in the brain that is both robust and adaptive,” says Wilcox. “This coordination does not itself constitute cognition, but rather determines the range of cognitive activities that the system can support.”

“In this framework, the brain is modeled as a network whose behavior is constrained by global properties such as efficiency, flexibility and integration,” Wilcox said. “These properties are not tied to individual tasks or brain networks, but properties of the system as a whole, shaping each cognitive function without being reducible to any one of them.”

“Once the question shifts from where intelligence is to how the system is organized,” Wilcox notes, “the empirical goals change.”

Integrated system of networks

The researchers found evidence to support four predictions of network neuroscience theory.

First, the theory predicts that intelligence is not localized to a single brain network but arises from processing distributed across multiple networks. So intelligence depends on how the brain manages the division of labor in different networks and brings them together as needed.

Second, for the brain to manage this distributed processing, it requires integration and effective long-range communication. To synchronize these efforts, Barbe says, “there is a large and complex system of connections that act as ‘shortcuts’ that connect distant brain regions and integrate information across the network.” These pathways connect structurally distant regions of the brain, enabling efficient communication and supporting coordinated processing across systems.

Third, functional integration requires regulatory control that coordinates interactions between networks that shape how information flows throughout the brain. These regions act as regulatory centers, reaching out to other networks to orchestrate the brain’s ongoing activities. They selectively employ the appropriate network for the specific task at hand – integrating subtle cues to understand a problem, learning a new skill, or deciding whether a situation requires caution or a quick, intuitive response.

Finally, Barbe says that general intelligence depends on the brain’s ability to balance local specialization with global integration. In other words, the brain works best when tightly connected local clusters communicate well, but are still able to connect to distant regions of the brain across short communication pathways. According to the authors this makes possible the most effective problem solving.

The research suggests that intelligence is unified not because the brain relies on a single general-purpose processor, but because the same organizational principles shape how all cognitive functions work together.

Across both datasets, individual differences in general intelligence were consistently associated with these system-level characteristics. No single region or canonical “intelligence network” accounts for the effect.

“General intelligence becomes visible when knowledge is integrated,” Barbe notes, “when many processes must work together under system-level constraints.”

Applications for Artificial Intelligence

The implications of this research extend beyond intelligence research, he added. Grounding knowledge in large-scale organization, the study provides a principled account of why minds come together.

This framework helps explain why intelligence develops greatly during childhood, declines with aging, and is particularly susceptible to brain injury. In each case, it is the large-scale combination—not the discrete function—that changes.

The findings also inform ongoing debates about artificial intelligence and how AI models are built. If general intelligence in humans arises from system-level organization rather than a dedicated general-purpose process, then achieving general intelligence in artificial systems may require more than the accumulation or scaling of specialized abilities.

“This research can inform thinking about how the design features of the human brain can be used to inspire the advancement of human-centered, biologically inspired artificial intelligence,” Barbe said.

“Many AI systems can perform certain tasks very well, but they still struggle to apply what they know to different situations.” Barbe said. “Human intelligence is defined by this flexibility—and it reflects the unique organization of the human brain.”

The study was conducted with co-authors Babak Hemmatian and Love Version of Stony Brook University.

Source: University of Notre Dame