How Does the Brain Know Itself?

“If the human brain were so simple that we could understand it, we would be so simple that we couldn’t.” — Emerson M. Pugh

In a previous piece, I argued that introspection may be our most direct empirical contact with physical reality. Here, we take a deeper dive into how that might be.

Being an interventional psychiatrist—watching people respond rapidly to treatments like transcranial magnetic stimulation, with measurable, specific mechanistic effects that translate directly into subjective experience—has only consolidated this stance. Interventions need to work, and we are best served when we understand exactly how, yet with the best of intentions, we have operated mainly without a coherent theory of how the objective and subjective are related. We are beginning to have the tools to build a better map—and to change the map and therefore change the terrain itself.

What the Brain Does to Know Itself

Start with the body. Before any high-level narrative of selfhood can form, the brain must continuously register what is happening inside the organism—the heartbeat, the gut, the slow thrum of visceral state. This is interoception, mediated primarily through the insula. Neuroanatomist A.D. Craig (2009) proposed a hierarchical model in which the posterior insula maps raw sensations— pain, temperature, organ state—and then, through successive integrations, these become higher-order emotional representations in the anterior insula. What begins as the body’s raw telemetry ends as something closer to feeling.

The visceral substrate of selfhood is not an add-on to cognitive self-awareness; in many frameworks, it is its foundation. Neuroimaging shows the anterior insula co-activating with the medial prefrontal cortex during self-reflection: The body and the narrative are not cleanly separable.

At a higher level, the Default Mode Network—medial prefrontal cortex, posterior cingulate, precuneus—continuously updates the autobiographical self (the “wallpaper of the mind”), integrating memory with present state, possibly foundational for core sense of self (Brenner, 2025). Depression shifts these same circuits, the DMN dysregulated as self-referential processing becomes a trap.

Higher up the hierarchy, three more regions complete the picture. The anterior prefrontal cortex supports metacognition—thinking about thinking —and structural differences here predict introspective accuracy with enough precision to have an anatomical correlate; its failure produces anosognosia, the patient who has lost the capacity to catch themselves.

Alongside it, the anterior cingulate functions as a continuous internal auditor, generating what EEG researchers call the error-related negativity—a neural spike occurring within milliseconds of a mistake, before conscious awareness has registered it, and without which real-time self-correction breaks down.

Finally, the temporoparietal junction handles the relational dimension: perceiving the self from a third-person vantage, modeling oneself as an object in a social world and not only as the subject of experience—which is why its damage produces out-of-body experiences and failures of body-ownership. Self-awareness, it turns out, requires all three: the capacity to assess one’s own mind, to catch one’s own errors, and to see oneself as others do.

How Do We Feel the Inner Landscape of Present-Moment Experience?

We can map these regions. Explaining why any of this activity produces felt experience is another matter—this is the hard problem of consciousness that Chalmers named decades ago and remains unsolved. We have good accounts of the correlates of consciousness. We do not have a satisfying account of why those correlates are accompanied by experience at all.

Leading theories each capture something real without closing the gap. Global Workspace Theory proposes that consciousness arises when information is broadcast widely enough to become available to multiple processes simultaneously; think of it as the moment a signal stops being local and becomes the whole brain’s business.

Predictive coding frames the brain as a continuous prediction machine, constructing perception from the interplay of expectation and sensory surprise. On this account, the self is the brain’s best current model of the causes of its own states—not a fixed entity but a hypothesis under constant revision. When that model destabilizes—in psychosis, dissociation, severe depression—the self fractures or dissolves.

Consciousness Essential Reads

Beneath the neurons, standard accounts skip layers that may matter. Astrocytes, long dismissed as structural scaffolding, communicate via calcium waves and modulate neurotransmission in ways that likely shape the synchronized activity most theories of consciousness require.

Neuromodulators—dopamine, serotonin, norepinephrine—don’t carry content; they carry information about the brain’s overall state, chemical memos organizing what neurons are doing at a meta level. This is precisely where most pharmacological interventions land. They work. But without a granular theory of what these modulatory shifts actually do to subjective experience, we are turning dials without fully knowing what we are tuning.

Why It Matters Now

Psychotherapy as currently practiced was built largely through intuition and clinical iteration. It works—often substantially. But we do not know, with any precision, which aspects of the therapeutic process drive which neural changes, for which patients, in which timing windows.

Neuromodulation sharpens the stakes: Transcranial magnetic stimulation (TMS) and related emerging methodologies (Brenner, 2025; 2026) can now indirectly and directly modulate the insula, the anterior cingulate cortex, the prefrontal metacognitive systems—the exact circuits above—with increasing specificity. We are intervening in the mechanisms of subjective experience based on a burgeoning theory of how intervening in them changes what a person experiences from the inside.

AI-enabled analysis of neural data is beginning to change this, aided by new computational models and, especially, the capacity for sophisticated AI stacks to derive information in new ways (Brenner, 2026). The bottleneck has not been the richness of the brain’s signals; it has been our capacity to decode them.

That gap is closing. Biomarkers of subjective states—depression, anxiety, the quality of therapeutic engagement—may soon be readable in ways that allow treatment with a precision we have never had. There is a substantial body of clinical research already established, and there are tools we can use today in outpatient clinics like my practice, such fMRI-guided neuronavigation for TMS.

The twentieth century was the century of the gene. The twenty-first may be the century of the brain, and more than that, of diverse intelligence (Michael Levin, 2025). Subjective experience made tractable, of the inner world made available not just to philosophy but to clinical practice and day-to-day life—brightening rather than dulling the shine of conscious experience.

Given what is at stake—for suffering people, for an aging population, for minds increasingly shaped by technologies we barely understand, for our society and future—it is mission critical