Why can salamanders regrow limbs when a person who loses an arm gets a scar and a stump? An axolotl that loses a leg builds a new one, bones, muscles, nerves and skin, in the right order and at the right size. Mammals mostly cannot, although lab mice can regrow the tip of a digit (Aztekin and Storer 2022). The cellular biology of how salamanders do it has been studied for two centuries. The evolutionary question, why the ability exists in some lineages and is absent or weak in others, is still open.
It does not settle anything. It does make clear which numbers someone would need to measure.
The question, stated precisely
Two questions often get blurred together. One is mechanistic: what do salamander cells do after amputation that human cells do not? The other is evolutionary: was limb regeneration present in the common ancestor of salamanders and mammals and later lost in our line, or did salamanders gain or keep something special? And if it was lost, was the loss favored by natural selection, or did it simply drift away because nothing kept it in place?
The second question is the subject here. Review authors are candid about the state of the evidence. One review of appendage regeneration across vertebrates puts it flatly: “The mechanism of evolutionary loss of regeneration in mammals is not understood.” (Daponte et al. 2021). A cellular review of tetrapod limbs likewise describes the reasons for these differences as still largely a mystery (Miller et al. 2019).
Public interest is easy to see. A 2013 study on salamander immune cells was covered by ScienceDaily and by a Royal Society of Biology blog post that framed it around human therapy, and Scientific American ran a podcast episode titled “Warm-Blooded Animals Lost Ability to Heal the Heart.” The appeal is obvious. If regrowth was switched off, perhaps it could be switched back on.

What is established
Several points are well supported by experiments, as opposed to inference.
- Regeneration follows a shared sequence across tetrapods that can do it: the wound closes, a mass of progenitor cells called a blastema forms, and those cells proliferate, differentiate and grow out in a patterned way (Miller et al. 2019).
- The wound epidermis is required for salamander limb regeneration. Nerves, macrophages, systemic hormones and p53 regulation all contribute, with roles that differ between species (Miller et al. 2019).
- Ability varies widely. Frogs and salamanders can regrow appendages; mammals, lizards and adult frogs regenerate limbs poorly or not at all, and capacity declines with age in some species (Aztekin and Storer 2022; Miller et al. 2019).
- Mammals are not missing the raw materials. Aztekin and Storer (2022) write that “Current evidence strongly argues that mammals contain the necessary cells,” which pushes the explanation toward how those cells are regulated.
- When salamander macrophages were removed systemically, limbs failed to regenerate and the stump scarred instead (Godwin et al. 2013). The Royal Society of Biology summary of that work reports that the animals regrew limbs once the immune cells were restored.
- One mammal does better than the rest at skin and ear repair. African spiny mice (Acomys) shed skin under very low tension and regrow ear-hole tissue including hair follicles, sebaceous glands, dermis and cartilage, possibly through a blastema-like structure (Seifert et al. 2012).
That last finding matters for the evolutionary argument. A mammal can evolve, or retain, scar-free regrowth of a complex tissue when its ecology rewards it. The spiny mouse sheds skin to escape predators, so it gets wounded often. Whatever stops most mammals, then, is probably not an impassable developmental wall.
Beyond this, the field is short on direct tests. Most of the evolutionary story is inferred from comparing a handful of model organisms, which is a thin basis for claims about the whole vertebrate tree.
Hypothesis 1: scarring is a defensive trade
The most popular family of explanations treats regeneration and scarring as rival programs. Miller and colleagues summarize it in one line: “The most common hypotheses posit a trade-off between regeneration and a program antagonistic to regeneration.” The usual candidate for the rival is the immune response.
The steelman goes like this. A land animal with a big open wound faces infection and blood loss. Closing the wound fast with dense collagen, then living with the scar, buys survival now at the cost of function later. A strong adaptive immune system may also have pushed macrophages toward defense and away from tissue remodeling; Daponte et al. (2021) describe this as a possibility, and point to frogs, which lose much of their regenerative ability after metamorphosis, when immune responses strengthen. In lizards, repeated tail amputation was associated with more immune cells and more scarring, according to the same review.
Evidence for: the frog transition and the lizard observations. Aztekin and Storer (2022) add a speculative twist, that inflammatory scarring might have been useful in some contexts, while saying plainly that this is not well established.
Evidence against, or at least complicating: macrophages are not simply the enemy of regrowth. Godwin et al. (2013) found that salamanders need them; remove them and the limb scars. So the immune system is a participant in regeneration, not only a competitor. A trade-off may still exist, but it would have to be a matter of how immune cells are tuned, which is harder to measure than presence or absence.
Hypothesis 2: warm blood paid for with lost repair
A second idea links the loss of regeneration to endothermy, the switch to internally generated body heat in mammals and birds. The best direct evidence concerns the heart, not limbs, but it is the clearest test any version of the trade-off hypothesis has had.
Hirose and colleagues surveyed 41 vertebrate species and found that the share of diploid, single-nucleus heart muscle cells, the kind that can still divide, fell as metabolic rate, body temperature and blood thyroxine rose (Hirose et al. 2019). Blocking thyroid hormone signaling in mouse heart cells preserved their ability to divide and regenerate into adulthood, with less scarring after injury. Giving thyroid hormone to adult zebrafish did the opposite. The authors conclude that “loss of heart regenerative capacity in adult mammals is triggered by increasing thyroid hormones” and suggest this may be a trade-off for acquiring endothermy. A later review describes the same inverse relationship between cardiac regenerative potential and endothermy (Xiong 2020).
This is a strong result for one organ. Whether it extends to limbs is untested in the work cited here. A limb is built from many tissue types, and its regrowth relies on a blastema rather than on division of mature heart cells. Xiong (2020) also notes it is unclear whether polyploidy itself causes the loss or merely accompanies it. There is a further puzzle: lizards are ectotherms, yet they cannot regrow limbs either. Endothermy may be one route to losing regeneration, not the only one.
Hypothesis 3: size, time and cancer
A third family of ideas is about scale. Growing a new limb is slow. A large animal would need to rebuild a large structure, and during that period it is impaired, perhaps exposed to infection or predators. A small salamander regrows a small leg. For a 70 kg primate, the rebuilding phase could stretch out long enough that a fast scar is the better bet.
Cancer adds a related cost. Regeneration depends on adult cells re-entering the cell cycle, and Miller et al. (2019) list p53 regulation among the factors involved. Bigger, longer-lived bodies have more cells and more years in which a proliferative program could go wrong. If keeping cells ready to regrow a limb carries any added tumor risk, that risk grows with body size and lifespan. This is an inference from general principles; the sources fetched for this article do not test it directly in regenerating species, so it should be read as a hypothesis.
Evidence for the size argument is mostly indirect: regenerating vertebrates tend to be small, and the mammal that regrows ear tissue well is a small rodent. Evidence against is that some small animals, mice and lizards among them, still fail to regrow limbs. Size cannot be the whole story.

Hypothesis 4: nothing was selected; the ability decayed
A less discussed possibility is that limb regeneration was not traded for anything. If a lineage rarely survives losing a limb, or rarely loses one in the first place, then the genes and regulatory switches that support regrowth face little selection to stay intact. Mutations that damage them would accumulate by drift, the way any unused trait can erode. Readers who want the background on how drift and selection differ can find it in an earlier x9m8 piece on whether evolution is random.
What supports this view? Aztekin and Storer (2022) consider it very probable that incompetence stems from multiple changes rather than one. Many small, scattered losses fit drift better than a single adaptive switch would. What weighs against it: the thyroid result shows at least one case where a specific, adaptive-looking change suppresses repair. And drift does not explain why regeneration would vanish from tissues that are injured often.
These hypotheses are not exclusive. Endothermy could have reduced heart repair in one step, large size could have tilted limbs toward scarring, and drift could have finished the job. The open question is how much each contributed, and in which lineages.
A small cost-benefit model built for this article
To see how the trade-offs interact, I wrote a simple expected-fitness model in Python. It is a thought model, not a fit to data. Every parameter range is an assumption, chosen to be wide, and the output shows only what follows from those assumptions.
Each animal produces one unit of offspring per year while intact and dies at a baseline rate. Appendage injuries arrive at some yearly rate. A scarring animal survives each injury with a small risk of death from the wound and then lives with a permanent performance loss. A regenerating animal takes the same wound risk, then spends a regrowth period impaired and exposed to extra hazard (infection, predation), after which it is restored. Regrowth time scales with body mass to a power between 0.25 and 0.5. Staying regeneration-competent also adds a lifelong cancer hazard proportional to body mass. Both strategies have closed-form lifetime outputs, so the comparison is exact for each parameter draw. I drew 200,000 parameter sets from log-uniform ranges with a fixed random seed.
import numpy as np
rng = np.random.default_rng(22)
def fitness(M, mu, k, d, p, h, tau0, b, c0):
tau = tau0 * (M / 0.05) ** b # regrowth time (yr) grows with mass
W_scar = 1.0 / (mu + k * (p + d)) # each injury: death risk p, lasting loss d
impaired = np.minimum(k * tau, 0.99) # share of life spent regrowing
W_regen = (1 - impaired * d) / (mu + k * (p + h * tau) + c0 * M)
return W_scar, W_regen
def draw(n):
lu = lambda lo, hi: np.exp(rng.uniform(np.log(lo), np.log(hi), n))
return dict(k=lu(0.05, 1.0), # appendage injuries per year
d=lu(0.02, 0.3), # permanent loss per scarred injury
p=lu(0.001, 0.05), # death risk from the wound itself
h=lu(0.05, 2.0), # extra hazard per year while regrowing
tau0=lu(0.05, 0.5), # regrowth time (yr) at 50 g
b=rng.uniform(0.25, 0.5, n),
c0=lu(1e-5, 1e-2)) # cancer hazard per kg per yr if competent
P = draw(200_000)
for M in [0.01, 0.05, 0.5, 5, 70]:
shares = []
for mu in [0.5, 0.2, 0.05, 0.02]:
Ws, Wr = fitness(M, mu, **P)
shares.append(round(float(np.mean(Wr > Ws)), 2))
print(M, shares)
Results for the share of draws in which regeneration beats scarring:
| Body mass (kg) | Mortality 0.5/yr | 0.2/yr | 0.05/yr | 0.02/yr |
|---|---|---|---|---|
| 0.01 | 73% | 74% | 74% | 74% |
| 0.05 | 58% | 59% | 60% | 60% |
| 0.5 | 31% | 35% | 37% | 37% |
| 5 | 10% | 13% | 15% | 15% |
| 70 | 1% | 2% | 2% | 2% |
Three things stand out. First, body size dominates. In the mixed-size run, mass had the strongest rank correlation with the advantage of regrowth (Spearman rho = -0.59), followed by the extra hazard while regrowing (-0.50) and baseline regrowth time (-0.32). The permanent cost of a scar pushed the other way (+0.27). Second, lifespan barely mattered: moving baseline mortality from 0.5 to 0.02 per year changed the share by at most six points in any row. Third, within the ranges chosen, cancer cost was a minor driver (rho = -0.07). Switching it off left 6% of draws favoring regrowth at 70 kg, against 2% with it on.
The cancer result depends on where the size effect comes from. When I made regrowth time independent of size, the share at 70 kg rose to 25%, and size mattered only through cancer. Multiplying the cancer cost by ten in that setting brought the share at 70 kg down to 8%. So the model says: either slow regrowth in big bodies or a large cancer penalty can make scarring the better option, and present data cannot say which.
What the model does not show: it says nothing about mechanism, it treats regeneration as all-or-none, it ignores phylogeny, and it assumes injury rates and regrowth hazards that nobody has measured across species. Injury rate had almost no effect here because, in this setup, it scales both the cost of scars and the cost of regrowth. A model in which regrowth machinery costs something to maintain even without injury would make frequent injury favor regeneration, as the spiny mouse case suggests.

Where the evidence points, and what would settle it
My reading, and it is an opinion: the loss of limb regeneration in mammals probably has more than one cause, and the size-and-time argument is underrated relative to the immune story. The thyroid work makes endothermy the best-supported single mechanism for the heart, but lizards show that limb regrowth can be lost without warm blood. The spiny mouse shows it can return, at least for skin and ear tissue, when injury is frequent. That pattern, loss and partial regain tracking ecology, looks more like a cost-benefit outcome than a fixed developmental limit.
Specific, testable predictions follow:
- Across related species of salamanders or lizards, regenerative completeness should fall with adult body mass after correcting for phylogeny, and regrowth time should rise with mass. A slope near zero would count against the size hypothesis.
- Species that are injured often (by autotomy, skin tearing or fighting) should regenerate better than close relatives that are injured rarely. Field injury rates are the missing number here.
- If endothermy drives limb loss as it appears to for the heart, then mammals with low body temperature and low thyroxine should show more limb or digit repair than their relatives. Hirose et al. (2019) found that platypus and some bats keep far more diploid heart muscle cells than mice; their digit and ear repair would be a direct test.
- If the cancer cost matters, regenerating species should carry extra tumor-suppression features, or lower tumor rates per cell, compared with non-regenerating relatives of similar size.
- If drift explains the loss, the genes and enhancers that drive regeneration should show relaxed selection, more damaging changes and pseudogenes, in lineages that lost the ability, while lineages that kept it should show conservation.
Any of these is within reach of current comparative genomics and field biology. For context on how fast human traits change under selection, see our piece on recent human evolution; for why age-related decline in repair is hard to measure, the discussion of epigenetic clocks is a useful companion.
Common questions
Did human ancestors once regrow limbs?
Nobody knows. Salamanders and frogs regenerate limbs, which many researchers take as a hint that the ability is old, but that is a reading of the family tree. Whether the specific ancestors of mammals could regrow whole limbs is an inference, not an observation.
Can humans regenerate anything?
Some tissues, yes, but not whole limbs. Mice regrow digit tips, which shows that mammalian limbs keep some capacity at their extremes (Aztekin and Storer 2022). The cells needed appear to be present; the regulation is different.
Is the immune system the reason we scar?
It is part of the story. Salamanders without macrophages scar instead of regrowing, so immune cells help regeneration rather than simply blocking it (Godwin et al. 2013). How immune responses are tuned, not whether they exist, is the likelier difference.
Does warm-bloodedness explain it?
For the heart, the evidence is good: thyroid hormone levels tied to endothermy suppress heart cell division, and blocking that signal in mice preserved repair (Hirose et al. 2019). Limbs have not been tested the same way, and cold-blooded lizards also fail to regrow limbs.
What does the model in this article prove?
Nothing about real animals. It shows that, under wide assumptions, large body size tips the balance toward scarring through slow regrowth or cancer cost, and it identifies which measurements would make the argument testable.
References
- Aztekin C, Storer MA. 2022. To regenerate or not to regenerate: vertebrate model organisms of regeneration-competency and -incompetency. Wound Repair and Regeneration. DOI: 10.1111/wrr.13000. PMC7613846
- Daponte V, Tylzanowski P, Forlino A. 2021. Appendage regeneration in vertebrates: what makes this possible? Cells 10(2):242. DOI: 10.3390/cells10020242. mdpi.com/2073-4409/10/2/242
- Godwin JW, Pinto AR, Rosenthal NA. 2013. Macrophages are required for adult salamander limb regeneration. Proceedings of the National Academy of Sciences. DOI: 10.1073/pnas.1300290110. PMC3677454
- Hirose K, Payumo AY, Cutie S, Hoang A, Zhang H, et al., Huang GN. 2019. Evidence for hormonal control of heart regenerative capacity during endothermy acquisition. Science 364(6436):184-188. DOI: 10.1126/science.aar2038. Author PDF
- Miller BM, Johnson K, Whited JL. 2019. Common themes in tetrapod appendage regeneration: a cellular perspective. EvoDevo. DOI: 10.1186/s13227-019-0124-7. PMC6572735
- Seifert AW, Kiama SG, Seifert MG, Goheen JR, Palmer TM, Maden M. 2012. Skin shedding and tissue regeneration in African spiny mice (Acomys). Nature 489(7417):561-565. DOI: 10.1038/nature11499. PMC3480082
- Xiong JW. 2020. Evolutionary insights into heart regeneration. Cell Regeneration 9:23. DOI: 10.1186/s13619-020-00069-x. Open access
- ScienceDaily. 2013 (May 20). Press release on Godwin et al. 2013, Monash University. sciencedaily.com
- Devonport J. 2013 (date inferred). Salamanders, Doctor Who and the secret of regeneration. Royal Society of Biology blog. blog.rsb.org.uk
- Scientific American. Warm-blooded animals lost ability to heal the heart (podcast episode). scientificamerican.com
