Anyone who has taken a high dose of DMT and met something on the other side knows the question that follows them home. Was that being real, or was it me?
It sounds like a simple question. But it might not be. A new preprint from two of the most interesting minds working on consciousness right now has just given the question a formal, testable structure, but leading psychedelic researcher Robin Carhart-Harris continues to argue the answer sits much closer to home.
The paper that reopened the case
The standard scientific line on DMT entities has always been straightforward. The brain, freed from its normal top-down constraints, runs its pattern-detection machinery unchecked and conjures autonomous beings out of noise. Elaborate hallucination. Case closed.
A new preprint, Traces of the Other: Are DMT Entities Real?, argues the case is not closed. Written by Andrew Gallimore (Okinawa Institute of Science and Technology), Donald Hoffman (UC Irvine), and mathematician Niffe Hermansson (The Trace Institute), the paper takes the entity question seriously at the level of formal mathematics, derives testable predictions from first principles, and proposes experiments capable of telling the two explanations apart.
To follow the argument, you need Hoffman’s underlying framework: conscious realism.
The Hard Problem (how subjective experience arises from physical matter) remains unsolved despite decades of neuroscience. Hoffman’s response is to flip the question entirely. Rather than asking how matter produces consciousness, he asks what if consciousness is the fundamental thing, and matter is what consciousness looks like from the outside.
The formal engine behind this is what Hoffman calls conscious agent theory. A conscious agent is defined by three components: a space of possible experiences, a space of possible actions, and rules governing how experiences lead to actions and actions lead to new experiences, modelled using Markov chains. These agents combine. Simple ones link together to form more complex ones, scaling in principle from a single perceptual moment up to the layered structure of human consciousness.
Reality, on this view, is a vast network of these interacting agents, and the physical world we perceive is a compressed summary our own agent generates to navigate that network without being overwhelmed by its true complexity.
The evolutionary case for why we’d expect this is the fitness-beats-truth theorem. Evolutionary game theory simulations consistently show that perceptual systems tuned to fitness payoffs outcompete those tuned to objective truth. Natural selection produces minds that see reality usefully, not accurately. The icons on your desktop are designed to let you work, not to show you what’s happening inside the machine.
Why most of reality is invisible
If reality is a network of conscious agents and our perceptual systems are fitness-optimised interfaces, it follows that we only ever perceive a tiny fraction of the agents that actually exist.
The paper formalises this with the concept of a trace. Agent A perceives Agent B when B’s actions propagate through the world and reach A’s perceptual system in a way that actually shapes A’s experience. If B’s actions never land within the subset of world-states A’s system is sensitive to, B is simply imperceptible, not absent, just outside the range A was built to register.
The authors call the range of agents we normally perceive the Consensus Reality Space, a stable region of possible experience shaped by evolutionary pressure to track whatever mattered for our ancestors’ survival. The paper notes that the space of possible conscious agents is effectively infinite. Even for an experience space with just ten states, the mathematical object describing all possible transition structures is ninety-dimensional with ten billion vertices. Scale that up and you have an incomprehensibly vast diversity of possible minds, almost all of which sit permanently outside our perceptual reach.
DMT as interface perturbation
The paper’s central proposal is that DMT perturbs the human perceptual interface enough to push consciousness outside the Consensus Reality Space, into regions where different dynamical rules apply and traces of normally imperceptible agents can be rendered as stable, coherent structure.
The neuroscience lines up with this. Psychedelics increase cortical entropy, flatten the brain’s control energy landscape, and expand the repertoire of states the cortex can explore. Under DMT, the brain measurably visits territory well outside its normal range.
From the mathematics, the authors derive three predictions, and each maps closely onto reported DMT phenomenology.
1. Higher-dimensional phenomenology. Agents requiring richer geometric structure than a three-dimensional interface can support would appear as space itself expanding or warping, more directions than there should be. This is among the most consistently reported features of breakthrough DMT experiences.
2. Extreme complexity and otherness. Agents operating in a higher-dimensional experiential space would appear, to a lower-dimensional observer, to be performing godlike feats, much as a routine operation in a higher-dimensional mathematical space can encode interactions that look impossibly complex when computed in ordinary spacetime.
3. Independent agency. An entity driven by an external conscious agent rather than your own cognitive machinery should behave with genuine unpredictability, capable of surprising you in ways you couldn’t have generated yourself. This matches what users consistently describe: entities that teach, demonstrate, and act with complete autonomy from the experiencer’s intentions.
The paper also points to an anomaly the hallucination model struggles with. When the brain’s top-down constraints loosen in dreams or psychotic states, experience fills with familiar humans and animals, exactly what the brain’s pattern-detection machinery evolved to model. DMT does something categorically different. Humans appear in fewer than five percent of entity encounters. The beings reported have no referent in the human ancestral environment at all.
Testing it properly
The paper outlines experiments made feasible by DMTx, a target-controlled intravenous infusion technique that can hold subjects in a stable DMT state for up to two hours, already validated in clinical settings. For the first time, researchers can study extended DMT states under controlled conditions rather than the brief peak of a smoked or vaped dose.
Single-subject paradigms ask whether DMT experiences are constrained by variables outside the subject’s own cognition. In one design, a computer in a separate room randomly switches between two states, unknown to both subject and experimenter, and researchers check whether an entity can track that hidden variable and express it within the subject’s experience at statistically significant rates.
Multi-subject paradigms test for shared structure across independent experiences. Do two isolated subjects encountering the same entity show non-trivial correlations in their reports? Can information be deposited with an entity by one subject and retrieved by another in a separate session?
Push back
Neuroscientist Robin Carhart-Harris is one of leading figures in psychedelic research and is widely recognized for his pioneering work in the field. And he represents the other major position in this debate. Under his framework, psychedelics loosen the predictive models the brain normally uses to organize experience. Deep emotional structures, memories, and archetypal patterns become personified, because personification is how human minds handle complexity they can’t otherwise process.
On this account, the guide you meet is an expression of your own psyche reorganising itself under altered conditions. Nothing external is required. The entity is real in the sense that the experience is real, but it originates entirely within the architecture of your own mind. It’s a framework grounded firmly in what’s currently known about predictive processing, and it doesn’t require any of Hoffman’s more radical metaphysics to work.
Both camps have serious researchers behind them, and for now, neither has settled the argument. The Gallimore, Hermansson and Hoffman paper doesn’t claim victory either. Its authors are explicit that the entities may still turn out to be hallucination, and that their proposed experiments may return null results. That’s science working as intended.
Maybe both are right
None of this settles the metaphysics, and it may not be a question that has to be settled by picking a side. Hoffman’s model is a scientific proposal still being tested. The trace hypothesis is a formal conjecture awaiting the DMTx experiments that could confirm or kill it. Carhart-Harris’s account is grounded in what we currently know about predictive processing in the brain. Anyone who tells you this question is closed, in either direction, is overselling their certainty.
But there’s a version of this where both camps turn out to be describing the same thing from two different vantage points. If Hoffman is right that consciousness is the base layer of reality rather than a byproduct of matter, then the usual boundary between “generated by my mind” and “genuinely other” starts to look less like a fact about the universe and more like a feature of the interface itself, a line drawn for the sake of everyday navigation rather than a line that exists at the deepest level of things.
Which raises the harder question underneath all of this. If every mind, human, animal, or DMT entity, is ultimately a pattern within one and the same underlying consciousness, what does “real” even mean any more? Real as opposed to what? An entity born entirely from your own psyche and an entity that is a genuine trace of another conscious agent could both be, in the end, consciousness meeting itself, simply at different depths and different distances. The dividing line between self and other might be exactly the kind of thing evolution built for us to see, whether or not it holds up once you go looking for it underneath.
Keep up with the research here.
