Most of our understanding of the human brain comes from non-invasive tools like fMRI, which measures blood flow as a proxy for activity, or scalp EEG, which captures the muffled electrical echoes of billions of neurons through the skull. While powerful, these methods lack the resolution to see the "gears" of thought.
A parallel, high-resolution stream of discovery has emerged from a unique clinical setting: the neurosurgical ward. When patients with drug-resistant epilepsy or Parkinson’s disease undergo surgery, doctors often implant electrodes directly into the brain to locate seizure foci or guide stimulators. For a brief window, researchers can record the firing of single neurons and high-frequency oscillations in humans as they speak, remember, and feel.
Mapping the Stream of Processing
One of the core insights from this field is that human cognition is organized into a clear, hierarchical stream. In a landmark study, Florian Mormann, Itzhak Fried, and colleagues demonstrated that as visual information moves deeper into the Medial Temporal Lobe (MTL), neurons become increasingly selective, firing only for specific, high-level concepts Latency and Selectivity of Single Neurons Indicate Hierarchical Processing in the Human Medial Temporal Lobe.
This isn't just true for vision. Michał T. Kucewicz and his team used direct recordings from 186 patients to show that verbal memory encoding is a continuous "stream" distributed across the cortex Human Verbal Memory Encoding Is Hierarchically Distributed in a Continuous Processing Stream. By tracking activity across the brain, they found that the strongest signals occur at the transition points—from initial visual perception of a word to the high-level linguistic processing required to store it.
The Mechanics of Episodic Memory
How does the brain turn a continuous experience—like watching a movie—into discrete memories? Intracranial recordings have revealed the cellular "editors" responsible for this segmentation.
Jie Zheng and Ueli Rutishauser identified boundary cells that fire at the start or end of an event and event cells that fire in response to the content within that segment Cognitive boundary signals in the human medial temporal lobe shape episodic memory representation. These cells act as bookmarks, helping the brain package experiences into manageable chapters.
This packaging is driven by specific electrical signatures. Xiongbo Wu found that high-frequency bursts called ripples occur both during events and at boundaries, acting as the mechanical gears that encode these "packages" into long-term storage Movie-watching evokes ripple-like activity within events and at event boundaries. Interestingly, these same ripples appear to drive our spontaneous, self-generated thoughts during quiet moments, as shown by Takamitsu Iwata Hippocampal sharp-wave ripples correlate with periods of naturally occurring self-generated thoughts in humans.
The Generalizability Challenge
While these studies provide unparalleled detail, they face a recurring fundamental question: Does a brain with epilepsy work the same way as a healthy brain?
Across nearly all recent literature, researchers like Jules Brochard and Masaki Sonoda emphasize that findings from clinical populations may be influenced by long-term medication, underlying pathology, or compensatory plasticity Neural shape mediation analysis. For instance, while we can map language "highways" in children with epilepsy using dynamic tractography, we must consider if the disease itself has forced the brain to reroute these paths 6D dynamic tractography visualizes the effective connectivity supporting specific linguistic stages in the developing brain.
To mitigate this, the field is moving toward massive, standardized datasets. Projects like the Brain Treebank by Andrei Barbu and colleagues are aggregating tens of hours of intracranial recordings from multiple patients, allowing researchers to find patterns that persist across individuals despite their specific clinical histories Brain Treebank: Large-scale intracranial recordings from naturalistic language stimuli.
Beyond Observation: Direct Modulation
The proximity of these electrodes allows for more than just listening; it allows for "talking" back to the brain. Flavio Frӧhlich demonstrated that the effect of direct cortical stimulation depends entirely on the brain's current "state"—the same electrical pulse can have different effects depending on what the patient is doing Modulation of Cortical Oscillations by Low-Frequency Direct Cortical Stimulation Is State-Dependent.
This state-dependency is being leveraged for therapy. Researchers are testing whether patients can use real-time feedback from their own intracranial signals to "tune" their brain activity, such as downmodulating the insula to increase pain thresholds Neurofeedback modulation of insula activity via MEG-based brain-machine interface or using deep brain stimulation of the zona incerta to selectively block pain signals Stimulation of zona incerta selectively modulates pain in humans.