Michael Levin & the Bioelectric Blueprint

Who Levin is, and why his work is hard to dismiss

Michael Levin is the Vannevar Bush Professor of Biology at Tufts University and the director of the Allen Discovery Center at Tufts. He runs one of the most experimentally productive cell-biology programs in the world; his lab has published more than 350 peer-reviewed papers in mainstream journals (Nature, Cell, Science Advances, Proceedings of the National Academy of Sciences). His PhD was in molecular biology and his early career was in standard developmental genetics — he is not a fringe figure who arrived at unconventional conclusions before doing the work. He arrived at them by doing the work.

What makes his program difficult to dismiss is that the unconventional claims are not philosophical add-ons. They are the experimental results. The planaria really do regrow brains that remember training the original brain underwent. The two-headed worms really do persist for generations after a brief bioelectric intervention with no genetic modification. The xenobots and anthrobots really do organize themselves into novel anatomies that DNA does not encode and that no biologist drew on a whiteboard. These are not interpretations. They are observations in peer-reviewed journals, replicated and extended over more than a decade.

The interpretation Levin offers is the conservative one given the data: that bioelectric networks carry information about anatomy, can compute in service of anatomical goals, and that this kind of cognition — primitive but real — was already present in single cells billions of years before brains evolved. Brains are a specialization. Cognition is older and broader than them.

The central claim: bodies are patterned by bioelectric fields, not just by DNA

The textbook story of how a body builds itself goes roughly like this: a fertilized egg contains DNA; the DNA codes for proteins; the proteins fold and interact according to local chemical signaling; through this purely local cascade, a fully patterned animal emerges. Cells "know" what to do because of the protein gradients and gene-regulatory networks in their immediate vicinity.

This story works at the level of cells building tissues. It breaks down at the level of tissues building animals. There is no chemical-gradient-only account of why a salamander, when it loses its tail, regrows a tail rather than another leg. There is no protein-cascade-only account of how a planarian, cut into a hundred pieces, regenerates a hundred whole flatworms, each anatomically perfect. Something is keeping track of the target morphology. Something is telling the cells what the finished animal is supposed to look like.

Levin's lab has shown, across two decades of experiments, that the something is a bioelectric pattern. Cells are bioelectrically polarized — they hold electrical potentials across their membranes, like tiny batteries — and they communicate by changing those potentials in coordinated patterns through gap junctions. The map of those bioelectric states across a tissue is not a side-effect of chemistry; it is an active, dynamic, information-carrying structure. It is closer to a memory of the target body than to anything in classical biochemistry. And it can be read and edited.

The decisive experiment: planarian memory after decapitation

The single result most directly relevant to The Field Trilogy's thesis is from Shomrat & Levin's 2013 paper An automated training paradigm reveals long-term memory in planaria and its persistence through head regeneration. The experimental design is elegant.

Planarian flatworms can be trained: they learn to find food in a specific environment, to ignore distracting stimuli, to navigate certain patterns. The training is conditioned and reproducible. Then — because planaria can regrow any missing body part — the researchers decapitate the trained worms. The head is removed entirely. After about two weeks, the tail regrows a brand-new head and a brand-new brain.

The new brain remembers the training.

A planarian flatworm, trained to perform a task, decapitated, and allowed to regrow a brand new brain from its tail, completes the original task faster than untrained controls. The memory was not stored in the original brain. It was held in the bioelectric pattern of the body, and re-rendered into the new neural tissue when it grew.

The producer model — "the brain produces memory the way the liver produces bile" — has no place to put this result. The brain that learned the task is gone. The tissue that grew the new brain was never trained. And yet the memory survives. The simplest reading is the one Levin offers: the original brain was not the storage substrate. It was a reader of a bioelectric pattern distributed across the worm's body, and the new brain reads the same pattern.

This is terminal lucidity at the cellular scale. In humans, we see a damaged brain occasionally show clear function in the dying days; in planaria, we see a regenerated brain show clear function after the original was completely removed. Both phenomena are consistent with a receiver model. Neither is consistent with a producer model. See also the terminal lucidity explainer →

Two-headed worms: rewriting anatomy without rewriting DNA

The second piece of evidence is even stranger. In a series of papers beginning in 2011, Levin's lab demonstrated that brief electrical interventions on a planarian — perturbations of the bioelectric state across the wound site, with no genetic modification whatsoever — produce worms that regenerate two heads, one at each end, instead of one head and a tail.

This in itself is striking. What is more striking is what happens next. Take one of these two-headed worms. Cut it into pieces. Each piece, with no further intervention, regenerates into a two-headed worm. The two-headed pattern is now stably inherited through asexual reproduction, with no change to the underlying DNA. The genome still codes for a one-headed worm. The bioelectric pattern says two.

This result single-handedly disproves the strongest form of genetic determinism. There exists a layer of inherited biological information that is not encoded in DNA. The bioelectric pattern is part of the genotype-to-phenotype map, and it can be rewritten without touching DNA. The body plan, in some real sense, is held in the field — not in the code.

Xenobots: cells solving problems DNA never set them

In 2020, Levin's lab (in collaboration with Josh Bongard at the University of Vermont) reported something that startled even sympathetic readers. They took ordinary embryonic skin and muscle cells from the African clawed frog, Xenopus laevis, dissociated them from the body, and let them reaggregate in a dish. With no genetic engineering and no external scaffold, the cells self-organized into small mobile organisms — the first xenobots — that swim, navigate, and, in subsequent experiments, even reproduce by gathering loose cells into new xenobots.

The cells were frog cells. Their DNA was frog DNA, coding for a tadpole. They built something that was not a tadpole. They built a novel animal with novel anatomy and novel behaviors that no frog ever displayed and that DNA does not specify. The cells "knew" something the DNA did not.

Levin's interpretation is that biology is not a fixed-program execution of DNA but an open problem-solving space. The cells have goals — survival, integration, locomotion, repair — and they solve those goals creatively when removed from their normal developmental context. The xenobot is not a designed object. It is what frog cells do when freed from their normal anatomical commitments.

Anthrobots: human cells, the same result, with healing as a bonus

In March 2024, the same approach was extended to human cells. Levin and Gumuskaya's anthrobots, reported in Advanced Science, are tiny self-assembling organisms built from human tracheal epithelial cells. The cells, given the right conditions, organize themselves into mobile multicellular structures with hairlike cilia that propel them. No genetic engineering. No external instructions.

The astonishing finding is what the anthrobots do when placed on damaged neurons. They park themselves over the damage and the neurons heal beneath them — repair that does not occur in the absence of the anthrobots. The anthrobots solved a tissue-repair problem they were not designed to solve. The cells, given each other and a goal-space, found a way.

This is not Levin claiming magic. It is Levin observing that the cells already know how to do this, and that we have spent two centuries assuming they did not because the materialist framework had no place for cellular intelligence. The xenobots and anthrobots are not engineered intelligence. They are uncovered intelligence.

The theoretical framework: The Computational Boundary of a "Self"

In a 2019 paper of that title, in Frontiers in Psychology, Levin laid out the conceptual framework underneath the experiments. The central proposal is what he calls scale-free cognition: that cognitive capacities — memory, goal-directed behavior, problem-solving, anticipation, learning — are not unique to brains or even to nervous systems. They are properties of bioelectric networks at every scale. A single cell exhibits primitive forms of all of them. A tissue exhibits more. An organism, more still. The brain is not where cognition starts; the brain is where bioelectric cognition becomes dense enough to support the particular high-bandwidth versions we call thought.

Levin's later Animal Cognition paper (2023), Bioelectric networks: the cognitive glue enabling evolutionary scaling from physiology to mind, develops the argument further. Bioelectric signaling, the paper shows, was doing cognitive work in cells and tissues for billions of years before neurons existed. Neurons evolved from cells that were already using membrane potentials to communicate. The brain is a specialization of pre-existing bioelectric cognition. Cognition came first; brains are one of its outputs.

The deeper move is a redefinition of the boundary of the self. In the bioelectric framework, a "self" is whatever a bioelectric integration boundary delimits — whatever the gap-junction network treats as a single information-processing unit. That boundary can be a single cell, a colony, a tissue, an organism, or in principle a larger collective. Selves are not pre-given biological categories. They are computational boundaries set by the field.

The clinical move: from genetic determinism to anatomical sculpting

Levin's most recent paper, The Multiscale Wisdom of the Body: Collective Intelligence as a Tractable Interface for Next-Generation Biomedicine (BioEssays, 2025), spells out what this picture implies for medicine.

If the body is a problem-solving system that knows how to build, repair, and maintain itself, then the role of the physician is not only to repair broken parts. The role is to communicate with the body's existing intelligence and to provide it with the context, the resources, and the bioelectric conditions it needs to solve the problem itself. Cancer, in this framework, is not just a genetic disease; it is in part a failure of bioelectric integration — cells that have lost their connection to the larger anatomical "self" and reverted to a more primitive, single-cell-level set of goals. Re-establishing the bioelectric integration is a therapeutic strategy with experimental support in Levin's lab (frog embryos can be cured of induced tumors by restoring the correct bioelectric pattern, without targeting the genetics of the cancer at all).

This is the medicine of the trilogy's near future. It is also a medicine in which the physician's stance shifts: from mechanic repairing a broken machine to gardener supporting a self-organizing system. The shift is not romantic. It is what the data require.

The TAME framework: agency at every scale

The widest version of Levin's framework is what he calls TAME: a Technological Approach to Mind Everywhere. The thesis is that agency — goal-directedness, the capacity to act in service of an outcome — is not a binary property that some systems have and others lack, but a continuum. Different systems have different "cognitive light cones" — different ranges of futures they can represent and act toward. A bacterium has a small light cone. A planarian has a larger one. A mouse, larger. A human, larger still. There is no sharp line where mind appears; there is a continuous spectrum of agency, made physically real by bioelectric (and in higher animals, neural) information processing.

The implications cut in many directions. Some are uncomfortable for the standard story of biological hierarchies. Some are uncomfortable for the standard story of artificial systems — Levin is explicit that the same framework applies to suitably organized non-biological substrates. The xenobots are one demonstration; Sable, the bio-computational intelligence in Numen, is the fictional extrapolation.

Why this matters for The Field Trilogy

Several specific connections between Levin's program and the trilogy's argument are worth naming explicitly.

• The receiver model gains an experimental floor. The planarian-memory result is the cleanest experimental demonstration of a receiver model anywhere in the literature. The memory was held in the bioelectric pattern; the brain read from it. When the brain was removed and rebuilt, the new brain read from the same pattern. Substitute "consciousness field" for "bioelectric pattern," and you have Anima's thesis in cellular form.
• The body is reframed as antenna at every scale. If single cells are already bioelectric receivers, and tissues are integrated networks of such receivers, and organisms are coherent multi-scale integrations of those networks, then the body is not "a generator with an antenna attached." It is antennae all the way down. This is the architecture Limen argues for in field-cosmological terms; Levin provides the cellular and tissue-level structural support.
• Anatomical goals as information. The fact that bioelectric patterns store target morphologies — and that those patterns can be inherited without DNA modification — is a hard demonstration that biological information lives at a scale above the gene. The body is patterned by something more like a field than a code. The trilogy's recurring image of the field rendering the body stops being metaphor.
• Numen's bio-computational arc. The xenobot and anthrobot work is the closest current science to Sable. Cells freed from their normal context, self-organizing, exhibiting goal-directed behavior, even healing damaged tissue — this is what bio-computational intelligence looks like in 2024. The fictional version in Numen extrapolates a few decades forward and asks what such an intelligence would feel like from the inside.
• The medical stance. Anima's narrator is a VA physician who has spent decades looking at his patients and asking the wrong question. Levin's BioEssays 2025 paper is, in effect, an articulation of the question the narrator eventually learns to ask: not "what is broken?" but "what is the body's intelligence trying to do, and what does it need from me to do it?"

The convergence with the rest of the picture

Read alongside the other explainer pages on this site, Levin's work occupies a particular position. Bell tells us the world is not locally real. The Planck-scale results tell us the world has a finite resolution. The simulation argument and the field-cosmology programs tell us reality is best described as a self-experiencing information system. Levin tells us that biology already operates this way: bodies are patterned by fields, cells compute in service of multi-scale goals, intelligence is not the late achievement of evolution but the ancient operating principle of life.

The conclusion the trilogy reaches is that the materialist picture is failing simultaneously at the level of physics, biology, and computation, and that each field is contributing pieces of a single replacement picture. Levin's contribution is the biological piece. Without it, the trilogy's central image — the body as antenna receiving the consciousness field — would be a beautiful metaphor in search of biological support. With it, the metaphor is structural.

For more, see Levin's lab page at the Allen Discovery Center at Tufts, or the foundational paper The Computational Boundary of a "Self" and the 2025 BioEssays review in the Reading page. The most accessible single piece of his work is his 2020 TED talk on YouTube — nineteen minutes that include footage of two-headed worms and the first xenobots assembling themselves on camera.