For decades, cognitive neuroscience has relied on a powerful but fundamentally limited logic: if a brain region "lights up" during a task, it must be involved in that task. Using functional Magnetic Resonance Imaging (fMRI), researchers like Dirk B. Walther and Diane M. Beck have shown that distributed patterns of activity in regions like the Parahippocampal Place Area (PPA) can predict the categories of scenes a person is viewing Natural Scene Categories Revealed in Distributed Patterns of Activity in the Human Brain.
However, as James J. DiCarlo and colleagues argue in Neural Mechanisms Underlying Visual Object Recognition, identifying a correlation between neural activity and behavior is not the same as proving that activity drives the behavior. A region might be active because it receives "spillover" information from a truly necessary neighbor, or it might be performing a computation that the brain ultimately ignores. To move from "where" to "how," the field is shifting toward methods that establish functional necessity.
The Virtual Lesion: Testing Necessity via Perturbation
The most direct way to prove a brain region is necessary for a function is to see what happens when it is gone. While researchers have long studied patients with focal brain lesions—such as the work by Xiaosi Gu showing that the insula is essential for updating social norms Necessary, Yet Dissociable Contributions of the Insular and Ventromedial Prefrontal Cortices to Norm Adaptation—natural lesions are rare and messy.
Modern neuroscience instead uses "virtual lesions" through Transcranial Magnetic Stimulation (TMS). By applying a targeted magnetic pulse to the scalp, researchers can temporarily disrupt neural firing in a specific area. Daniel D. Dilks and Joshua B. Julian used this to demonstrate that the Occipital Place Area (OPA) is not just active during scene perception, but causally required for it The Occipital Place Area Is Causally and Selectively Involved in Scene Perception. When the OPA was disrupted, participants' ability to recognize scenes dropped, while their ability to recognize faces remained intact.
Dialing the Cortex: Modulation of Social and Cognitive Logic
Establishing necessity isn't just about turning regions off; it’s about "dialing" their activity up or down to observe predictable shifts in complex human behavior. This is often achieved via Transcranial Direct Current Stimulation (tDCS).
For instance, Haoli Zheng and Daqiang Huang have used tDCS to map the causal architecture of social and economic decisions:
- Altruism: Stimulating the ventromedial prefrontal cortex (vmPFC) increases repayment in trust games, suggesting this region causally drives altruistic preference Modulating the Activity of Ventromedial Prefrontal Cortex by Anodal tDCS Enhances the Trustee’s Repayment through Altruism.
- Moral Judgment: Disrupting the Temporoparietal Junction (TPJ) shifts a person’s moral compass toward "consequentialism," where they judge actions by outcomes rather than intentions Modulation of Neural Activity in the Temporoparietal Junction with Transcranial Direct Current Stimulation Changes the Role of Beliefs in Moral Judgment.
- Risk and Revenge: Enhancing the right dorsolateral prefrontal cortex (rDLPFC) can increase revenge behavior Effect of Modulating Activity in the Right DLPFC on Revenge Behavior or alter risk preferences Activation of the DLPFC Reveals an Asymmetric Effect in Risky Decision Making.
These studies provide the "smoking gun" that fMRI cannot: they show that changing the biology of a specific node directly changes the psychology of the individual.
Closing the Loop: Concurrent Stimulation and Imaging
The "gold standard" for establishing causal mechanisms is combining stimulation with real-time imaging. This allows researchers to see not just that a perturbation changes behavior, but how that perturbation ripples through the rest of the brain's networks.
Jade Jackson and colleagues demonstrated this by using TMS while participants were inside an fMRI scanner Concurrent neuroimaging and neurostimulation reveals a causal role for dlPFC in coding of task-relevant information. They found that the dlPFC doesn't just correlate with attention; stimulating it actually "boosts" the neural representation of task-relevant information in other parts of the brain. Similarly, Logan Grosenick and Conor Liston argue that combining optogenetics with fMRI in animal models is essential for moving from "biotypes" of depression to targeted, circuit-level treatments Functional and Optogenetic Approaches to Discovering Stable Subtype-Specific Circuit Mechanisms in Depression.
Silico Causality: Simulating Deficits in Models
As deep learning models become more brain-like, they offer a new, ethical playground for causal testing. In Inducing Dyslexia in Vision Language Models, Melika Honarmand and colleagues "lesioned" specific artificial neurons in Vision-Language Models to replicate human reading impairments. By showing that disabling a specific "word-processing" module in a model creates the same errors seen in human dyslexia, researchers can generate high-fidelity hypotheses about which biological circuits are truly necessary for literacy.
This computational approach, paired with whole-brain decoding frameworks like WAVE Decoding Visual Experience and Mapping Semantics through Whole-Brain Analysis Using fMRI Foundation Models, allows us to test "what-if" scenarios that would be impossible in living humans, providing a bridge between the descriptive "where" of 20th-century neuroscience and the causal "how" of the 21st.