How the Brain Runs an Internal Model of Reality

By Scott Welsh

Introduction

In philosophy, “simulation theory” holds that we are all living inside a computer-generated reality, in the spirit of The Matrix. This paper argues that something very close to this is literally true — except that the “computer” is biological rather than electronic. That computer is the human brain, and the world each of us experiences moment to moment is, in an important sense, a running simulation constructed by it. We step outside that simulation only briefly and partially, whenever we act directly on the physical world.

Like the prisoners in Plato’s allegory of the cave, our access to reality is filtered through a small number of senses operating across a narrow slice of the physical spectrum. What those senses register must then be reconstructed by the brain into something usable — and that reconstruction is, in essence, a simulation of the world rather than the world itself.

From Reflex to Representation: The Evolutionary Origins of Internal Simulation

All life shares three essential imperatives: to eat, to grow, and to reproduce. The first of these — finding food — has historically been the rate-limiting step for survival. Early organisms relied on chance alone, but over evolutionary time they developed strategies for locomotion and simple food-sensing mechanisms such as chemotaxis. As organisms became multicellular, more sophisticated survival strategies became possible, but these required coordinating the movement of many body parts, which in turn drove the emergence of a nervous system.

A further advance was the capacity to adapt rapidly to new threats and opportunities through learning — compressing into a single lifetime what would otherwise take many generations of evolutionary trial and error. With a brain, an organism could store a repertoire of responses to varying circumstances and adjust them based on experience.

As sophisticated as these strategies became, they remained essentially reactive: an animal senses prey and swims toward it, or senses a threat and flees. Even learning of this kind required first stumbling on an opportunity by chance; it did not allow an organism to be proactive. True proactivity requires something more than reflexive response — it requires an internal simulation of the outside world. The expansion of the forebrain made this possible. With it, animals gained the ability to test and learn from hypothetical strategies before acting on them. In effect, the brain could “time travel”: looking forward to ask “what if I did this?” and looking backward to ask “what could I have done differently?” This same capacity underlies delayed gratification, since it allows an organism to weigh outcomes that are separated from the present by a long stretch of time.

This is closely related to what some psychologists call “mental time travel”: the ability to mentally reconstruct past episodes and construct hypothetical future ones, widely regarded as a defining feature of advanced cognition built on top of ordinary episodic memory.

Two Modes of Operation: Real-Time Action and Simulated Deliberation

In effect, the brain can operate in two modes: reasoning in real time, or reasoning within a simulation. Acting in real time has the advantage of speed; consulting the internal simulation before acting allows for a more cautious, deliberate approach. The brain region most associated with rapid, real-time responses is the cerebellum, while the forebrain — and the prefrontal cortex in particular — appears more involved in slower, more contemplative processing.

This division has a computational analogue in current models of motor control, where the cerebellum is thought to maintain “internal models” that predict the consequences of an action quickly enough to guide movement in real time, while regions of prefrontal cortex hold candidate actions in working memory and evaluate them before commitment. More generally, model-based planning — the neural equivalent of internally simulating the outcome of a hypothetical action before taking it — appears to recruit a distinct prefrontal–cerebellar–striatal network, separate from the circuits used for fast, habitual, or memorized responses.

Of course, the internal simulation must be continuously updated and corrected by direct interaction with the world in order to remain accurate. What is learned through real-world action constantly feeds back to refine the simulated model of that world. This dependence becomes strikingly clear when sensory input is removed altogether. In sensory-deprivation experiments, healthy volunteers isolated from sound and light for as little as fifteen minutes reliably report perceptual disturbances, and a substantial minority experience outright hallucinations — seeing faces or objects that are not there, or sensing a presence in the room. The leading explanation is a form of “faulty source monitoring”: deprived of external data to anchor it, the brain’s internal simulation continues running and begins to misattribute its own self-generated content to the outside world.

Neural Evidence: The Default Mode Network

There is direct neurobiological evidence for an internal simulation of this kind, in the form of the default mode network (DMN) — a set of interconnected brain regions, including the medial prefrontal cortex and posterior cingulate cortex, that becomes active during internally directed thought and often quiets down during demanding, externally focused tasks. The DMN is reliably engaged during autobiographical memory, imagining future events, mind-wandering, and inferring the mental states of other people — the very activities one would expect of a network responsible for running and consulting an internal model of the world. Notably, the DMN is not simply idle “default” activity; it also engages strongly during major cognitive transitions, such as returning from rest to a new task, consistent with a role in constructing and updating context and scene representations rather than passively resting.

Implications for Self, Consciousness, and Social Cognition

The idea that much of human life is lived within an internal simulation of the world carries several implications for psychology and cognition. It seems reasonable to suppose that this internal simulation includes a simulated model of ourselves as rational actors — and it is conceivable that this self-model is what we experience subjectively as consciousness.

It also seems likely that we use this simulation of ourselves as the template for simulating other people. This would help explain the well-documented psychological pattern known as projection, in which a person unconsciously attributes their own motivations or feelings to others. Consistent with this, the same self-referential, internally directed brain network implicated in modeling one’s own thoughts — the DMN — is also consistently activated when people try to infer the beliefs, intentions, or feelings of others, a process known as theory of mind or mentalizing. Athletes, notably, are often coached to “stay out of their own heads” and avoid “psyching themselves out,” consistent with the idea that fast, real-time athletic performance is better served by direct action than by slow, simulation-based contemplation.

When the Simulation Breaks Down: Injury and Illness

Because the internal simulation depends on constant synchronization with real-world experience, disruptions to that synchronization — or to the neural machinery running the simulation itself — can make the world feel unpredictable, unfamiliar, or frightening. Several clinical conditions illustrate this dynamic.

The historical practice of frontal lobotomy, which severed connections between the prefrontal cortex and the rest of the brain, is thought to have effectively disabled the internal simulation, pushing patients out of deliberative, simulation-based processing and into more direct, reflexive, cerebellar-mediated behavior. This is consistent with the well-established clinical picture of frontal lobe damage more broadly, which produces marked behavioral disinhibition: impulsivity, poor judgment, and a diminished capacity to weigh future consequences before acting.

In schizophrenia, functional imaging studies consistently show abnormal activity and connectivity in the default mode network, and this disruption has been linked to disorganized thought, an altered sense of self, and, in some cases, hallucinations. One plausible interpretation, consistent with the simulation framework, is that when the brain’s internal model becomes decoupled from reliable real-world updating, the resulting simulated world can drift sharply from consensus reality — and an unpredictable inner world may in turn contribute to the fear and paranoia often associated with the illness.

Autism spectrum disorder shows a related but distinct pattern of default mode network disruption, also associated with difficulties in social cognition. One speculative extension of the simulation framework is that autism may involve difficulty building a stable, well-calibrated simulation of the surrounding world, which could likewise make that world feel unpredictable and threatening. The repetitive behaviors often observed in autism may then function, at least in part, to impose a felt sense of predictability on an environment that is otherwise difficult to model internally.

Living Between Two Worlds

Over evolutionary time, the brain’s original role was to coordinate the cells of a multicellular organism toward common goals. Gradually, this expanded into identifying and choosing those goals — not merely how to move the organism, but where and when. To do this, the brain built an internal simulation of the outer world in which it could move forward and backward in time to weigh different courses of action. We all alternate between living inside this simulation, where we plan, and acting in the real world, where we execute.

Trouble arises when this balance tips too far in either direction. Spending too much time in the inner, simulated world produces endless planning without doing — procrastination, along with rumination and worry about a past that cannot be changed or a future that has not yet arrived. Spending too little time there produces the opposite problem: a person fixated on the immediate present who fails to learn from the past, plan for the future, or anticipate the consequences of their own actions. The healthiest position lies somewhere between these two extremes.

Conclusion

Viewed through this lens, human cognition can be understood as the product of two complementary systems: a fast, direct-acting system tuned to the immediate present, and a slower, simulation-based system capable of modeling hypothetical pasts and futures. The latter appears to depend heavily on the brain’s default mode network, is disrupted in characteristic ways by neurological and psychiatric illness, and may be foundational to our sense of self and our capacity to understand other minds. Much remains speculative in this framework, but it offers a unifying perspective on phenomena as varied as delayed gratification, hallucination, projection, and the disorganized thought of schizophrenia.