Regeneration is rebuilding a lost structure to its original pattern, and it is not one ability. Salamanders rebuild limbs faithfully, lizards regrow a tail that is not the tail they lost, and starfish do something closer to reproduction. Mammals mostly scar instead — and scarring appears to be what prevents it.
The word covers four different things and the popular examples move between them without saying so. A salamander rebuilds an amputated limb complete with bone, muscle, nerve and the correct number of digits: that is regeneration in the strict sense, restoring the original pattern. A lizard regrows a shed tail, but the replacement has a cartilage rod where vertebrae should be and is a functional substitute rather than a rebuild. A starfish arm can regrow an entire animal, which is nearer to asexual reproduction than to repair. And a mammal heals — closes the wound with collagen, quickly and permanently, and rebuilds nothing. Lined up that way, the interesting question stops being "why can axolotls do it" and becomes "what is it that mammals do instead". The evidence points at the scar. Salamander wounds do not lay down the collagen barrier a mammal does, and that absence appears to be a precondition rather than a coincidence: the blastema — the mass of partly de-specialised cells that does the rebuilding — cannot form under a scar. Which suggests our failure is not an absence of machinery but the presence of a faster, more infection-proof response that gets there first.
Developed record · 86% complete · reviewed 2026-08-11
What this page covers
A capability rather than a group. Best characterised in salamanders, planarians, starfish and a small number of invertebrates, with mammals as the informative failure.
Often confused with: Wound healing, which every animal does and which is not the same thing; Autotomy, deliberately shedding a part — a separate ability that often accompanies regeneration; Asexual reproduction, which fragmenting starfish approach and axolotls do not
Quick facts
The best vertebrate case
Salamanders — limbs, tail, jaw, spinal cord, parts of heart and brain
What lizards do
Regrow a tail, with cartilage instead of vertebrae
Almost every confusion here comes from treating these as the same thing.
Before comparing animals it is worth separating what is being compared, because the popular examples sit in different categories and are quoted as though they were one.
What "regeneration" is being used to mean
Kind
What happens
Example
What it is not
Epimorphic regeneration
The lost structure is rebuilt to its original pattern, via a blastema
Salamander limb, complete with bone, nerve and correct digit count
Not available to any mammal beyond fingertips in young children
Substitute regrowth
Something functional grows back, built differently from the original
Lizard tail — a cartilage rod, not vertebrae, and it cannot be shed the same way twice
Not a rebuild; the animal ends up with a different tail
Fragmentation into whole animals
A piece becomes an entire new individual
Some starfish and planarians
Closer to asexual reproduction than to repair
Wound healing
The wound is closed with scar tissue and nothing is rebuilt
Mammals, including us
Not regeneration at all, and the reason regeneration cannot happen
The starfish case is the one most often overstated. "A starfish grows back from any piece" is true of a small number of species and false of most: in the great majority, an arm alone dies, because regenerating a whole animal requires a portion of the central disc. Linckia is the genus that made the reputation, and it is exceptional rather than representative.
Words used here
Epimorphic regeneration
Rebuilding a lost structure via a blastema, restoring the original pattern. The strict sense of the word.
Autotomy
Deliberately shedding a body part to escape, at a designated breakage plane. A separate ability from regrowing it.
Limbs, tail, jaw, spinal cord, parts of heart and brain
Faithful rebuild via a blastema, requiring nerve supply and positional information
Other salamanders
Much the same, less reliably
The axolotl is the model because it breeds in captivity, not because it is the best at it
Planarian flatworms
An entire worm from a small fragment, including the brain
Driven by a reserve of genuine stem cells — a different mechanism from the axolotl
Starfish
Arms; whole animals in a few species
Needs central disc tissue in most species; the "any piece" version is one exceptional genus
Spiders
A lost leg, at the next moult
Only while still growing. An adult spider has stopped moulting and cannot
Octopuses
Arms, fully, including the nerve cord
One of the better vertebrate-scale examples in an invertebrate
Lizards
The tail, as a cartilage rod
A substitute rather than a rebuild, and the trick works poorly a second time
Mammals
Liver mass, fingertips in young children, and little else
Liver regrowth restores mass without restoring shape — not the same as rebuilding a pattern
Reading down the table, the pattern that stands out is not a ladder of complexity. Octopuses regenerate arms well and are not simple animals; planarians and salamanders reach similar outcomes by different routes, one with stem cells and one without. Whatever explains the distribution, "simpler animals regenerate better" is not it.
One word covering genuinely different processes, and the difference is old.
Some animals rebuild by growing, others by rescaling what is left
Established
Specialists would state this without hedging. Multiple independent lines of evidence agree.
Regeneration proceeds by two broadly distinguishable routes. In epimorphosis, exemplified by salamander limb regeneration, new tissue is generated from a proliferative blastema at the wound. In morphallaxis, exemplified by Hydra and prominent in planarians, existing tissue is repatterned and rescaled into the proportions of a smaller complete animal with limited new growth. Planarians combine substantial remodelling with a pluripotent adult stem cell population, the neoblasts, which are the animal's only dividing cells. Hydra maintains continuous cell turnover such that regeneration is an extension of ordinary maintenance rather than a wound-triggered programme.
Who this applies to
the principal experimental models of animal regeneration
“Regeneration means an animal can grow back a lost body part”
That describes one of the two ways it happens and misses the other, which is stranger. A salamander grows the missing limb from a bud at the wound. A planarian cut into pieces does something else: each fragment rescales — existing tissue is remodelled into the proportions of a smaller complete animal, so the result is a whole worm rather than a stump with a new head. Hydra goes further still, replacing its whole body continuously anyway, so cutting it merely interrupts a process already running. One word, three quite different things.
Why we rate it this way, and what the caveats are
EstablishedHigh confidence
The distinction is over a century old, has survived molecular characterisation of all three systems, and remains the organising framework of the field.
How far it can be extended
The distinction is drawn from direct comparison of the best-characterised regenerating systems, though most real cases combine both routes to some degree.
Caveats
Most real cases combine both routes; the categories are analytical rather than exclusive.
Neoblast heterogeneity has been substantially revised by later single-cell work.
Species differ greatly in regenerative capacity even within a single group.
Still unanswered
Are the two routes mechanistically related, or convergent answers to the same problem?
The epimorphic route in the axolotl, and the lineage-restricted rather than pluripotent character of its progenitors.
The classical distinction is over a century old and has survived molecular characterisation of every system it describes. In *epimorphosis*, new tissue is generated from a proliferative zone at the wound — the salamander route, where a blastema forms and a limb grows out of it. In *morphallaxis*, existing tissue is repatterned into the proportions of a smaller complete animal, with relatively little new growth. The animal does not grow the missing part; it rescales into a whole one.
Planarians combine both, and they are the reason the distinction is worth teaching. A fragment regenerates a complete worm, and much of the work is remodelling rather than building — which is why the result is a small complete flatworm rather than a normal-sized one with a new head. Behind it is a population of genuinely pluripotent adult stem cells, the neoblasts, which are the only dividing cells the animal has.
Hydra pushes further and stops being repair at all. Its cells turn over continuously, so the whole animal is effectively replaced on a timescale of weeks; regeneration is an extension of maintenance that never switches off rather than a response to injury. Cutting a Hydra interrupts a process already running. That is also why it turns up on "immortal animal" lists — a body under permanent reconstruction does not accumulate ageing in the ordinary way, which is a fact about tissue turnover rather than about not dying.
Three answers to the same problem
System
Route
What supplies the cells
Salamander limb
Epimorphosis — new growth from a blastema
Local cells that partially dedifferentiate and stay lineage-restricted
Morphallaxis, as an extension of continuous turnover
Continuously dividing cells the animal already maintains
A regenerating limb is not built from blank cells — each one remembers its job
Established
Specialists would state this without hedging. Multiple independent lines of evidence agree.
Lineage tracing using grafted tissue from axolotls expressing green fluorescent protein shows that blastema cells overwhelmingly rebuild the tissue type they originated from: grafted muscle produces muscle, cartilage produces cartilage. The blastema is therefore a heterogeneous population of lineage-restricted progenitors that have partially dedifferentiated, not a pool of cells reverted to a pluripotent state. Dermal and cartilage lineages show limited interchange; muscle does not.
Who this applies to
axolotl limb regeneration, with the same broad pattern expected across salamanders
Studied in
Ambystoma mexicanum, Caudata
You may have heard
“Axolotls turn cells back into stem cells and grow the limb again from scratch”
That is the intuitive answer and the tracing rules it out. Cells relax their identity just enough to divide and move, and every one of them keeps knowing what it is — muscle rebuilds muscle, cartilage rebuilds cartilage. That makes regeneration a coordination problem rather than a stem-cell problem, which is precisely why it has been so hard to borrow. What we are missing is not a cell type we lack but an instruction system we do not run.
Why we rate it this way, and what the caveats are
EstablishedHigh confidence
Direct lineage tracing with an unambiguous marker, following cells of known origin through to their contribution in the regenerate.
How far it can be extended
Blastema-based regeneration is general to salamanders; the lineage tracing that establishes restriction was performed in the axolotl.
Caveats
Grafting is itself a perturbation and may influence lineage behaviour.
Small proportions of cells crossing lineage boundaries are difficult to exclude at these sample sizes.
One species; the extent of generalisation to other regenerating animals is not established here.
Still unanswered
How is tissue identity maintained through partial dedifferentiation, division and migration?
Grafted fluorescent tissue traced through regeneration, establishing lineage restriction within the blastema.
How we know
Following individual cells through a regrowing limb
When an axolotl regrows a limb, does it turn cells back into stem cells and start over?
A blastema — the bud of cells that forms at the amputation site — looks uniform down a microscope, which is why the standard assumption was that its cells had reverted to something embryonic and pluripotent. Testing that requires knowing where each cell came from, which a microscope cannot tell you. So tissue from axolotls genetically engineered to express green fluorescent protein was grafted into ordinary animals: muscle into muscle, cartilage into cartilage, one tissue at a time. Every descendant of the grafted tissue then glows and can be found anywhere it ends up. The limb was amputated and the regenerate examined to see which tissues the glowing cells had built.
What happened
Cells overwhelmingly rebuilt the tissue they came from. Grafted muscle produced muscle and nothing else; cartilage produced cartilage. Some interchange appeared between dermal and cartilage lineages, and none in muscle. The blastema is a mixed bag of lineage-restricted progenitors rather than a pool of pluripotent cells.
What it shows
That regeneration works by cells relaxing their identity just enough to divide and move while still knowing what they are. That reframes the whole problem: it is not a stem-cell problem but a coordination problem, which is precisely why it has proved so hard to borrow. What we lack is not a cell type but an instruction system.
What it does not show
Grafting is itself a perturbation and could influence how cells behave. Small proportions of cells crossing lineage boundaries are hard to exclude at these sample sizes. And it is the axolotl: whether other regenerating animals work the same way does not follow, and planarians demonstrably do not.
The controls — what makes this evidence rather than a story
One tissue type grafted at a time, so the origin of any glowing cell is unambiguous.
Ungrafted animals regenerating normally, confirming the graft itself does not derail the process.
The whole regenerate examined rather than the tissue expected, so cells crossing lineage boundaries would have been found.
A genetic marker rather than a dye, so the label is inherited by every descendant and cannot leak between cells.
The salamander route holds a surprise that reorganises the whole subject. A blastema looks uniform down a microscope, so the natural assumption was that its cells had reverted to something embryonic and were starting again. Grafting glowing tissue into ordinary animals and following it through regeneration showed otherwise: cells overwhelmingly rebuild the tissue they came from. Muscle makes muscle. Cartilage makes cartilage. They relax their identity just enough to divide and move, and never forget what they are.
That is the most useful fact on this page for anyone asking why humans cannot do it. Regeneration is not a stem-cell problem — it is a coordination problem, and coordination is far harder to borrow. What we lack is not a cell type but an instruction system.
Words used here
Epimorphosis
Rebuilding by growing new tissue from a proliferative zone at the wound. The salamander limb route.
Morphallaxis
Rebuilding by repatterning and rescaling existing tissue, with little new growth. The Hydra route, and much of the planarian one.
Neoblast
A pluripotent adult stem cell in planarians, and the only cell type in the animal that divides.
Dedifferentiation
A specialised cell reverting towards a less specialised state. In salamanders it is partial: the cell can divide again without forgetting its tissue.
The first problem is not growing — it is knowing which end is the head
Established
Specialists would state this without hedging. Multiple independent lines of evidence agree.
In planarians, anteroposterior polarity of the regenerate is specified by Wnt/β-catenin signalling. Knocking down β-catenin causes a fragment to regenerate a head at both wounds and no tail; increasing Wnt signalling produces the opposite, tails where heads should form. Polarity is therefore set by the level of this pathway rather than by the geometry of the cut, and the decision is made before and independently of the growth that follows.
Who this applies to
planarian regeneration; Wnt-based axial patterning is broadly conserved in animals
Studied in
Schmidtea mediterranea
You may have heard
“Cut a planarian in half and each half grows into a new worm”
True, and it skips the interesting part. A piece cut from the middle has two wounds and must build a head at one and a tail at the other — so before anything grows, the fragment has to work out which end of the animal each wound was. That decision turns on the level of a single signalling pathway, and it can be broken in both directions on purpose: suppress it and you get a worm with a head at each end and no tail. The famous trick is not regrowth. It is knowing where you are.
Why we rate it this way, and what the caveats are
EstablishedHigh confidence
A manipulation producing an unambiguous and reversible phenotype in both directions, which is about as clean as developmental evidence gets.
How far it can be extended
Wnt signalling specifies the posterior end across bilaterian animals, though its role in regeneration specifically is demonstrated chiefly in planarians.
Caveats
RNA interference reduces rather than eliminates gene function, and residual activity varies.
Polarity involves more than one pathway; this identifies a necessary component rather than the whole system.
One species, in a laboratory strain maintained asexually.
Still unanswered
What holds the positional information in an intact animal so that any fragment can read its own address?
Reddien and Sánchez Alvarado, 2004 · Annual Review of Cell and Developmental Biology
The classical cutting experiments and the polarity problem they revealed.
How we know
Making a worm build a head at both ends
How does a fragment of a worm know which of its two wounds should become a head and which a tail?
The famous fact about planarians is that a piece regrows the whole animal. The harder question is polarity: a piece cut from the middle has a wound at each end and must build a head at one and a tail at the other, which means it has to work out which end of the original animal each wound was. The suspicion was that this decision runs on Wnt signalling, a pathway that specifies the back end of animals generally. RNA interference was used to knock down β-catenin, the pathway's central component, and the animals were then cut and left to regenerate. Crucially the pathway was pushed the other way as well, by knocking down its inhibitor, since a manipulation that only breaks things is much weaker evidence than one that can produce either outcome on demand.
What happened
Animals with β-catenin knocked down regenerated a head at both wounds and no tail at all. Increasing Wnt signalling produced the opposite: tails where heads should have been. Controls regenerated normally.
What it shows
That the polarity of the regenerate turns on the level of one signalling pathway, and that the decision is made before and independently of the growth that follows. Being able to produce the error in both directions is what makes this causal rather than correlational.
What it does not show
RNA interference reduces rather than abolishes gene function, and residual activity varies between animals. Polarity involves more than this pathway, so this identifies a necessary component and not the whole system. It also does not explain how positional information is stored in the intact animal so that any fragment can read its own address — which is the deeper question and remains open.
The controls — what makes this evidence rather than a story
Animals treated with a control RNA and cut identically, regenerating normally.
The pathway manipulated in both directions, so the result is a dial rather than a break.
Cuts made at defined positions so that the expected outcome for each wound was known in advance.
Knockdown efficiency verified rather than assumed.
Before anything can regrow, the tissue has to answer a question that is easy to overlook: which part of the animal am I, and therefore what is missing? A fragment cut from the middle of a planarian has a wound at each end and must build a head at one and a tail at the other. Get that wrong and you do not get a smaller worm; you get something unviable.
That decision turns out to run on the level of a single signalling pathway. Knock down β-catenin, the central component of Wnt signalling, and the fragment builds a head at both wounds and no tail at all. Push the pathway the other way and you get tails where heads should be. Being able to produce the error in both directions on demand is what makes this a mechanism rather than a correlation.
The salamander version of the same problem is positional information held in the connective tissue of the stump, and it is the reason a blastema formed at the wrist builds a hand while one formed at the shoulder builds an entire limb. The tissue knows its own address, retains that knowledge for the life of the animal, and how it is stored remains unknown — which is precisely the part any medical application would need.
Cut the nerve and nothing regrows; move the nerve and a limb can start where it should not
Established
Specialists would state this without hedging. Multiple independent lines of evidence agree.
A denervated salamander limb stump heals its wound but forms no blastema and regenerates nothing. Deviating a nerve to an abnormal position in combination with a wound can induce an ectopic blastema and, under some protocols, a supernumerary limb. Nerve-derived diffusible factors have been identified, and comparable nerve dependence operates in mammalian digit-tip regeneration, indicating that the requirement is general to appendage regeneration rather than a salamander peculiarity.
Who this applies to
salamander limb regeneration, with the same dependence in mammalian digit tips
A direct manipulation with an unambiguous result, reproduced for over a century, and strengthened by the deviation experiments that induce regeneration where it would not otherwise occur.
How far it can be extended
Nerve dependence has been demonstrated by denervation in multiple salamander species and, in a more limited form, in mammalian digit-tip regeneration.
Caveats
Identified nerve-derived factors do not fully account for the requirement.
Mammalian digit-tip regeneration is far more limited, so the comparison is partial.
Denervation has systemic effects beyond removing the signal of interest.
Still unanswered
What is the complete set of nerve-derived signals required, and could they be supplied artificially?
Establishes what the blastema is made of, which is what the nerve requirement gates the formation of.
The other requirement is a nerve, and the experiment that demonstrates it is over a century old and still startling. Denervate a salamander limb and amputate it: the wound heals and nothing regrows, no blastema forms at all. That much says the nerve is permission to proceed. The deviation experiments say something stronger — reroute a nerve to a place it does not belong, add a wound, and a blastema can form there, in some protocols producing a supernumerary limb.
So the nerve is not merely a green light for a plan held elsewhere; it is part of what specifies that a limb is the thing to build. That is also the honest answer to why this has not been lifted into human medicine. The requirement is not a molecule to be supplied but a relationship between nerve, wound and positional information that mammals do not maintain.
Words used here
Positional information
Whatever it is that lets a cell know where in the body it sits. Demonstrated by its consequences; not identified molecularly.
Denervation
Cutting the nerve supply to a limb. A denervated salamander limb heals its wound and regenerates nothing.
Blastema
The mass of proliferating progenitor cells that forms at a salamander amputation and from which the new limb grows.
A mammal cut open does something impressively fast: platelets plug the breach, inflammation clears the debris, fibroblasts flood in and lay down collagen, and the wound is sealed with a permanent patch of tissue that is not what was there before. It is quick, it is strongly resistant to infection, and it rebuilds nothing.
A salamander does not do this. The wound is covered by a thin specialised epidermis, no collagen barrier forms, and the tissue beneath stays accessible — which is what allows cells to dedifferentiate and gather into a blastema. Where a scar forms, no blastema forms. The relationship looks causal rather than incidental, and it turns the question around: mammals may not be missing regeneration so much as running a competing repair that gets there first and locks the site.
The trade-off, if that is what it is, is not obviously a bad one. Fast permanent closure is excellent defence against infection for a large warm-blooded animal, and regeneration is slow — an axolotl limb takes weeks to months, during which the animal is compromised. In cold water with a low metabolic rate that is survivable. For a mammal it may not be.
The exceptions are informative. Young children regrow fingertips amputated beyond the last joint, provided the wound is not surgically closed — and closing it prevents the regrowth. African spiny mice shed skin and regenerate it with hair follicles and cartilage rather than scarring, which shows a mammal can do it.
Scarring is an alternative programme, not a failure to regenerate
Well supported
Good evidence backs this, though some details remain open.
Adult zebrafish regenerate excised ventricular muscle within approximately two months through cardiomyocyte proliferation with minimal permanent scarring, while a mutant line unable to complete the process forms a scar resembling the mammalian outcome — indicating that scarring and regeneration are alternative programmes rather than success and failure. Separately, African spiny mice heal large skin wounds and ear punches by regenerating tissue including hair follicles, glands and cartilage where laboratory mice scar, demonstrating that scar-free regeneration of complex structures is not absent from mammals as a class.
Who this applies to
zebrafish cardiac regeneration and spiny mouse skin regeneration, against the mammalian defaultDo not extend this beyond the taxa listed — the popular version over-reaches.
“Mammals cannot regenerate; only simpler animals can”
The line is tidier than the biology. Scarring is not a failure to do something — it is a fast, effective repair that seals a wound at the cost of permanently losing function, and it appears to be an alternative programme rather than an absence. The clearest evidence is a zebrafish mutant that cannot complete regeneration and scars instead, exactly as we do. And a rodent, the African spiny mouse, regenerates skin complete with hair follicles and cartilage where a laboratory mouse scars. The machinery is not missing from mammals; we default to the other option.
Why we rate it this way, and what the caveats are
Well supportedModerate confidence
Both results are solid within their systems, and the mutant zebrafish line is a particularly direct demonstration that the two outcomes are alternatives. The general framing extends further than either result strictly licenses.
How far it can be extended
Both are specific systems. The general claim — that scarring is a programme rather than an absence — is an inference from them and is not established across tissues.
Caveats
Spiny mouse regeneration covers skin and ear tissue, not limbs or organs.
Zebrafish hearts are far smaller, thinner and less metabolically demanding than mammalian hearts.
Surgical resection differs from the ischaemic damage that causes human heart attacks.
The spiny mouse trait comes packaged with skin that tears off easily, which is not a desirable adaptation in itself.
Still unanswered
What determines whether a wound takes the scarring or the regenerative route, and can that be switched?
The mammalian organ that does restore itself, and the specific sense in which it does not regenerate.
How we know
Cutting away part of a heart to see whether it comes back
Can a vertebrate heart regenerate, or is scarring the only option?
A mammalian heart responds to damage by scarring: the wound is sealed quickly and the muscle is gone for good. Whether that is a limit of vertebrate hearts or a choice mammals make required an animal that might do otherwise. Roughly a fifth of the ventricle was surgically removed from adult zebrafish — a defined, reproducible injury rather than the messy ischaemic damage of a heart attack — and the animals followed over the following weeks with histology and with labelling that marks dividing cells, so that new muscle could be distinguished from surviving muscle.
What happened
The fish regenerated the excised muscle within about two months, producing working cardiomyocytes with minimal permanent scarring. The mutant line, unable to complete the process, formed a scar instead — closely resembling the mammalian outcome.
What it shows
That a vertebrate heart can rebuild lost muscle, and — through the mutant — that scarring and regeneration are alternative programmes rather than success and failure. That is the most useful result here: it turns the human question from "why can we not regenerate" into "why do we take the other route".
What it does not show
Surgical resection is not a heart attack: ischaemic damage is diffuse, involves dying rather than removed tissue, and occurs in an organ already unwell. A zebrafish heart is also far smaller, thinner and less metabolically demanding than a human one, and the cellular source of the new muscle was refined by later work. Nothing here is a therapy.
The controls — what makes this evidence rather than a story
Sham-operated fish undergoing the surgery without resection, separating the effect of the operation from the effect of the loss.
Cell proliferation labelling, so new muscle is identified as new rather than assumed.
Time series rather than a single endpoint, so the process could be followed rather than only its outcome.
A mutant line unable to complete regeneration, which turns the comparison into an internal one.
The cleanest evidence that these are alternatives rather than success and failure comes from a fish heart. Remove a fifth of an adult zebrafish ventricle and it rebuilds the muscle within about two months, with almost no permanent scarring. A mutant line unable to complete that process forms a scar instead — closely resembling what a mammalian heart does. Same injury, same animal, one gene, two outcomes.
That reframes the human question usefully. It is not "why can we not regenerate" but "why do we default to the other option", which is a question with a mechanism behind it rather than an absence.
The human liver restores its mass, not the piece that was removed
Established
Specialists would state this without hedging. Multiple independent lines of evidence agree.
Following removal of up to approximately two thirds of the liver, the remaining lobes enlarge through proliferation of existing mature hepatocytes until the original total organ mass is restored, typically within weeks, and growth then stops precisely at that point. The excised lobes do not regrow, so the restored organ matches the original in mass and function while differing in shape. The process is compensatory hyperplasia rather than regeneration of a missing structure.
Who this applies to
mammalian liver; mechanistic detail largely from rodents, clinical data from humans
Studied in
Homo sapiens, Rattus norvegicus, Mus musculus
You may have heard
“The human liver can grow back completely”
It restores how much liver there is, not which liver. Remove a lobe and the lobe does not return; the remainder enlarges until the total mass matches what it was, and then stops — with striking precision. Function comes back and anatomy does not, which is a different achievement from an axolotl rebuilding a limb with the right fingers in the right order. The technical name, compensatory hyperplasia, marks exactly that distinction: the liver has solved how much without solving what shape, and shape is the hard problem.
Why we rate it this way, and what the caveats are
EstablishedHigh confidence
Extensively studied experimentally in rodents and directly observed clinically in humans, with a consistent and quantitatively precise endpoint.
How far it can be extended
The response is documented across mammals, with human evidence from partial hepatectomy and living-donor transplantation.
Caveats
Most mechanistic detail comes from rodents; human data are largely clinical.
Regrowth is impaired in diseased or cirrhotic livers, which is where most clinical interest lies.
The signalling account has been substantially extended since this review.
Still unanswered
How does the organ measure its own mass precisely enough to stop growing at exactly the right point?
The cellular basis of the response, the restoration of mass rather than structure, and the precision with which growth terminates.
The liver is the mammalian organ people reach for as a counterexample, and it deserves a precise answer because it is doing something real and not what the question assumes. Remove up to two thirds of a liver and the remainder enlarges — by division of ordinary mature liver cells — until the original total mass is restored, then stops, with striking precision. The removed lobes do not come back. Function returns; anatomy does not.
The technical name is compensatory hyperplasia, and the distinction it draws is the one that matters here. The liver has solved "how much" without solving "what shape", and shape is the hard part — it is the whole of what an axolotl does when it rebuilds a limb with the right number of fingers in the right order.
Why it matters: If mammals scar because regeneration is unavailable, the goal is to supply what is missing. If they scar because fast closure is worth more, the goal is to suppress a working system — which is a different and riskier proposition.
What would settle it: Identifying what in the mammalian response blocks blastema formation, and testing whether suppressing it yields regeneration rather than a tumour.
Why is the distribution so patchy?
Why it matters: Regeneration appears and disappears across the animal tree with no obvious pattern, in both simple and complex animals. Whether it was repeatedly lost or repeatedly invented is unresolved.
How is positional identity retained into adulthood?
Why it matters: Adult tissue behaves differently depending on where in the limb it sits, long after development. The storage mechanism is unknown, and it is the part a medical application would need.
What is the relationship between regeneration and cancer?
Why it matters: Both involve sustained proliferation, and animals that regenerate well appear to get relatively few tumours. Whether the same control explains both is an open and consequential question.
NatureHQ publishes its own gaps. This record is at 86% completeness against what we would call a finished subject.
1 high-priority search intent(s) not yet covered
no research from the last few years is attached — check for newer work
Plant regeneration, which is routine and mechanistically unrelated, is out of scope here.
The regeneration–cancer relationship is raised as an open question and not treated, and it is probably the most consequential open question on the page.
Regeneration in invertebrates beyond planarians and Hydra — starfish, annelids, colonial ascidians — is represented only in the comparison table.
How positional information is physically stored in adult tissue is unknown, so the page can describe its consequences and not its mechanism.
Last reviewed 2026-08-11 · 7 claims · 35 search questions answered on this page