The Infrastructure Beneath the Forest: Mycorrhizal Networks, the Wood Wide Web, and the Limits of Evidence

The Hook: An Image Too Good to Be Checked

In August 1997, Nature published a short paper that launched one of the most successful metaphors in modern biology. Suzanne Simard and colleagues had labelled two tree species in the field with different carbon isotopes – paper birch and Douglas fir – and shown that labelled carbon travelled in both directions, with a net gain for the Douglas fir. David Read's accompanying commentary was titled "The ties that bind." The phrase that stuck was a different one: wood-wide web, usually attributed to Nature's coverage of that issue.

What followed was not the normal course of a research programme. The metaphor became a worldview. Fungal threads between roots turned into "mother trees" feeding their offspring; isotope transfer turned into "communication"; a symbiosis turned into a forest that thinks, learns, and remembers as a superorganism. Bestsellers, documentaries, Hollywood borrowings, and billions of clicks later, the wood wide web was one of the best-known biological facts in the world – with one peculiarity: very few of the people citing it could have answered the question of how many forests such a network had ever actually been mapped in.

The answer is: in two forest types, in five studies, for two tree species and three mycorrhizal fungal species. There are an estimated 73,300 tree species worldwide.

That figure comes from a paper published in Nature Ecology & Evolution in 2023 that landed hard in the field: Justine Karst, Melanie Jones, and Jason Hoeksema examined three popular claims about common mycorrhizal networks – and, in addition, how the primary literature is cited. Their bibliometric finding was more uncomfortable than their biological one: unsupported claims in the scientific literature had doubled over 25 years.

To conclude from this that mycorrhizal networks are a myth would be the same error running in the opposite direction. Because alongside this debate, something has happened over the past three years that received far less attention and rests on far firmer ground: the mycorrhizal symbiosis has been quantified globally for the first time. The result is an infrastructure whose dimensions are hard to picture – 110 quadrillion kilometres of fungal thread in the Earth's topsoil, and a carbon flux on the order of one-third of fossil fuel emissions.

This article separates two stories that are constantly conflated. The first is mycorrhiza as a planetary metabolic organ: solid, measured, astonishing. The second is the tree-to-tree network as a channel of communication and care: far thinner than the public believes. And at the end stands a methodological lesson that reaches well beyond biology – namely, that a demonstrated connection is not yet a demonstrated function.

Part 1: The Symbiosis as Planetary Infrastructure

What a Mycorrhiza Actually Is

A mycorrhiza is not a plant with a fungus next to it; it is a shared organ. Fungal hyphae grow through the fine root and form an exchange surface across which two material flows run in opposite directions: the fungus delivers phosphate, nitrogen compounds, water, and micronutrients; the plant delivers carbon from photosynthesis – and in the case of arbuscular mycorrhizal fungi, not only sugars but also fatty acids, which these fungi cannot make themselves. They are obligately dependent on their host.

The physical reason for this trade is geometry. Phosphate diffuses extremely slowly in soil; a depletion zone forms around a root within days. A root can only grow its way out of that zone – which is expensive. A hypha with a radius of roughly 2.7 micrometres reaches the same soil volume for a fraction of the construction cost. Viewed soberly, the mycorrhiza is the outsourcing of nutrient acquisition to a provider with a better surface-to-volume ratio.

This arrangement is ancient. Fossils from the Rhynie chert, some 400 million years old, show arbuscular structures in early land plants – the symbiosis is probably older than roots in the modern sense and was presumably a precondition for the colonisation of land. Today roughly 70 percent of all plant species form an arbuscular mycorrhiza (Glomeromycotina: grasses, herbs, most crops, many tropical trees). The trees of temperate and boreal forests – pines, spruces, beeches, oaks, birches – mostly use ectomycorrhiza with basidiomycetes and ascomycetes, which sheathe the root from outside and lay a Hartig net between its cells. To these are added the ericoid mycorrhiza of the heath family and the orchid mycorrhiza.

The distinction is not botanical pedantry: it determines which plants can hang on the same fungal network at all. A Douglas fir and a western red cedar share no ectomycorrhizal fungus. That becomes important in Part 2.

What the Symbiosis Costs

The plant pays. Estimates of the share of net photosynthesis that goes into the mycorrhiza typically range between 4 and 20 percent – an operating expense that competes with foliage.

In 2023, Heidi-Jayne Hawkins and colleagues attempted to add this expense up globally in Current Biology. Their result: plants channel roughly 13.12 gigatonnes of CO2 equivalent per year – at least temporarily – into the underground mycelium of their mycorrhizal fungi. Broken down: 3.93 Gt to arbuscular, 9.07 Gt to ectomycorrhizal, and 0.12 Gt to ericoid fungi. That corresponds to about 36 percent of fossil fuel CO2 emissions in 2021.

The figure is spectacular, and it is to be read with care – as the authors themselves stress. It rests on extrapolations from single-time-point measurements, "temporarily" is not "stored," and part of this carbon is respired within days. What the number does robustly show is not a sink but a flux: mycorrhiza is not a marginal phenomenon of the carbon cycle but one of its trunk lines.

The First Global Survey of the Network

In June 2026, Science published the most thorough quantification so far. Justin Stewart and a large consortium assembled data from 322 studies with more than 16,000 soil cores across nine biomes, trained models on them to predict hyphal density globally, and calibrated biomass by robotic imaging of more than 300,000 individual hyphae.

The result for arbuscular mycorrhizal fungi alone:

110 quadrillion kilometres is a number without intuition. The authors' accompanying translation helps: a teaspoon of soil can hold up to ten metres of fungal network.

Two of those rows are political. That grassland carries roughly 40 percent of the infrastructure shifts attention away from forests. And that cropland soils have lost nearly half their hyphal density turns a microbiological finding into an agricultural balance sheet.

This is complemented by a paper published in Nature in 2025: the SPUN consortium trained models on 25,000 geolocated soil samples comprising more than 2.8 billion fungal DNA sequences and predicted the richness and rarity of arbuscular and ectomycorrhizal fungi globally at kilometre resolution. The finding: fewer than ten percent of the predicted richness hotspots lie inside protected areas.

Up to this point the story is uncontested, quantitative, and impressive. Now it becomes difficult.

Part 2: From Network of Threads to Network of Trees – and Why the Difference Is Everything

The Definition That Matters

A common mycorrhizal network (CMN) exists when the hyphae of a single fungal individual – a genet – connect the roots of two or more plants. That is a strict condition. Soil full of hyphae is not a CMN. Two trees whose fungal partners meet in the soil are not a CMN. It takes continuous cytoplasm belonging to the same organism.

The proof is correspondingly laborious: one must genotype fungal genets with microsatellites and map their root contacts spatially. Kevin Beiler and colleagues did this in 2010 for Rhizopogon species in Douglas fir stands and drew the architecture of such networks for the first time – an impressive but small evidence base. By Karst and colleagues' accounting, the total remains five studies, two forest types, two tree species, three fungal species.

Four Routes an Isotope Can Travel

Here lies the methodological heart of the whole debate. If labelled carbon is introduced into tree A and later turns up in tree B, at least four pathways are possible:

1. Direct hyphal transport within a shared genet – the pathway the popular story presupposes.
2. Exudation and re-uptake: A releases carbon to the soil, the fungus takes it up and passes it to B. The fungus is then a scavenger, not a pipe.
3. Soil pathway without the fungus: root exudates or dead fine roots of A diffuse, and B takes them up.
4. Microbial detours: soil bacteria process the carbon and B benefits indirectly.

An isotope measurement alone cannot distinguish between these. It shows that something arrived – not how. Every study has to make the pathway plausible by additional construction, and that is precisely where the critique bites.

The Two Key Field Studies

Simard et al. 1997 (Nature 388:579–582) labelled Betula papyrifera and Pseudotsuga menziesii reciprocally with 13C and 14C. They found bidirectional transfer with a net gain for Douglas fir averaging six percent of the isotope taken up through photosynthesis. Their argument for the hyphal pathway rested on a control: Thuja plicata, which forms no ectomycorrhiza, absorbed only small amounts – therefore, the inference went, transfer ran mainly through hyphae directly.

That control is elegant, and it is simultaneously the weak point. "Small amounts" is not "no amounts." The difference between Thuja and Douglas fir shows that the fungal pathway contributes something – it does not quantify how much of that flowed through a shared genet and how much through uptake from soil.

Klein, Siegwolf, and Körner 2016 (Science 352:342–344) went an order of magnitude further. From a construction crane over a temperate mixed forest near Basel, they fumigated five roughly 40-metre-tall spruces, about 120 years old, with 13C-depleted CO2 over several years. The result: in the fine roots of neighbouring beeches, larches, and pines, about 40 percent of the carbon came from foreign trees; extrapolated, roughly 280 kg C per hectare per year. The label was also found in the ectomycorrhizal mycelium.

This is to date the strongest field evidence for extensive belowground carbon trade among mature trees of different species. And here, too, the limitation holds cleanly: what was measured was carbon in fine roots, not in the trunk, not in the crown. For the question "does tree A feed tree B?" that is an intermediate result, not a final one. The transfer is established; the benefit to the recipient is not.

The Critique of the Mother Tree Hypothesis

In 2023, Nils Henriksson, John Marshall, Mona Högberg, Peter Högberg, Andrea Polle, Oskar Franklin, and Torgny Näsholm published an examination of the mother tree hypothesis in New Phytologist. Their finding has two parts: the empirical basis is thin – and, more seriously, the theoretical mechanism is not worked out. What is missing is an answer to why a fungus living off a large, carbon-rich tree should pass that tree's carbon to a seedling without compensation.

Their counter-model: what looks like sharing is bilateral trade between one tree and one fungus at a time. The fungus is a market participant, not a courier. And this trade can increase competitive pressure on seedlings rather than relieve it – because the fungus draws carbon from the large tree and invests it where the return is highest, not where need is greatest.

Part 3: Three Claims and Their Evidence

Karst, Jones, and Hoeksema (Nature Ecology & Evolution 7:501–511, 2023) examined exactly three statements that the popular account treats as established.

By their assessment, the best evidence for resource movement within a CMN among woody plants remains the direct observation of 14C travelling more than 20 centimetres from one seedling to the roots of another – a laboratory finding, not a forest finding.

To this comes the bibliometric part, which touched the actual nerve: the authors examined how CMN results are cited and found that unsupported claims had doubled in 25 years, with a bias towards citing positive effects. Their conclusion: knowledge of CMNs is at present too sparse and too unsettled to inform forest management. (The paper received an author correction in March 2023 – a detail one mentions in fairness.)

The Rebuttals

The critique did not go unanswered, and the answers are worth reading because they show what is actually in dispute.

Tamir Klein, Ido Rog, Stav Livne-Luzon, Marcel van der Heijden, and Christian Körner published a reply in Open Research Europe under the programmatic subtitle "Facts, not fantasy": some criticisms are valid, but the evidence for belowground carbon transfer across mycorrhizal networks is solid and accumulating – from Reid and Woods in 1969 through Simard in 1997 to their own field data.

Suzanne Simard, T. L. Ryan, and David Perry responded in 2025 in Frontiers in Forests and Global Change. Their core argument: the critique attacks claims they never made. They never claimed that CMNs are the only transfer pathway, that competition is abolished in forests, that findings are globally transferable, or that trees behave altruistically – altruism, they say, is an assumption imposed by critics. Their findings are place-based and context-dependent.

What Is Actually in Dispute

Untangled, a manageable list of open questions remains – and a set of points that are no longer contested.

No longer contested: mycorrhizal fungi regularly connect the roots of several plants in soil. Carbon and nutrients move between plants. The amounts are measurable in the field.

Contested: what share runs through a shared genet rather than through soil and re-uptake. Whether the amounts arriving at the recipient measurably change its growth, survival, or stress tolerance. Whether there are directional decisions that go beyond physical source–sink gradients. And whether any forestry recommendations follow from all this.

I am of the opinion that this debate is largely settled on the substance and largely unsettled in the rhetoric: the critics are right that preference for offspring and directed signal transmission are not established; the defenders are right that transfer occurs and can be ecologically relevant. What makes the dispute so protracted is not the data but the fact that the two sides are talking about different statements – one about mechanism, the other about phenomenon.

Part 4: Signals – the Weakest Part of the Story

Especially popular is the idea that trees warn one another through the fungal network. The evidence is thinnest here, and two examples show why.

Zsofia Babikova and colleagues reported in Ecology Letters in 2013 that bean plants connected through a mycorrhizal network to a neighbour infested with aphids produced aphid-repelling and parasitoid-attracting volatiles – without being infested themselves. The finding remains the most frequently cited support for "underground warning." The discussion turns on the controls: how convincingly do mesh and severing treatments separate the hyphal pathway from soil and root pathways, and how robust are the effect sizes given small samples?

The second case is clearer. In 2022 a paper appeared reporting electrical wound signals travelling from plant to plant through fungal hyphae. Michael Blatt, Andreas Draguhn, Lincoln Taiz, and David Robinson took it apart in Plant Signaling & Behavior in 2023: the measurements used surface electrodes on agar and salt bridges that record extracellular potentials, and such arrangements systematically generate voltage transients through Donnan potentials without any biological source at all. The damning point stood in the original paper itself: its own control – threads soaked in salt solution – transmitted the signals just as well as fungal hyphae. A biological bridge was therefore not required for the effect.

From this one can distil a checklist that applies to any signalling claim, in biology and elsewhere:

1. Is a physical carrier named – molecule, ion, potential – and not merely a channel?
2. Is there a control that removes exactly that carrier and nothing else?
3. Is a response in the receiver measured, not just an arrival?

A study that fails point 2 usually leaves physics without biology.

Part 5: Why "Market" Fits Better Than "Family"

If not care, then what? The answer that has been gaining ground for a decade comes from the evolutionary economics of symbiosis.

Toby Kiers and colleagues showed in Science in 2011 that the mycorrhizal symbiosis is kept stable by reciprocal rewards: plants direct more carbon to the more cooperative fungal partners, and fungi deliver more phosphate to the more generous roots. Both sides can punish bad partners. No altruism is needed, only the capacity to sanction – which is precisely the opposite of a charitable network.

More revealing still is the work of Matthew Whiteside and colleagues, published in Current Biology in 2019: they tagged phosphate with quantum dots so its path through the network could be followed individually, and observed the fungi moving phosphorus from rich into poor soil patches – that is, to where it commands the highest price from the plant partner. That includes withholding it inside the hyphae: warehousing, not conduction.

On this picture a CMN is not a family intranet but an infrastructure with an operator that has interests of its own. The fungus connects many plants and is not neutral while doing so – it is a broker. That the same line benefits one tree and harms another is, in this model, not a contradiction but what one should expect.

It matters what this model does not claim: no intention, no strategy, no calculation. It is a statement about incentive structure and selection, not about inner life. The same discipline is familiar from bacterial threshold systems: a quorum is not a vote, however much the term suggests it.

Part 6: The Overarching Principle

Step back, and this case tells two stories layered on top of each other.

The first is about how a metaphor becomes a mechanistic claim. "Wood-wide web" was an editorial turn of phrase attached to a paper about a six-percent net isotope transfer. But the image brought a whole model with it: nodes, edges, packets, addressing, intent. Think "internet" and you ask about messages; think "marketplace" and you ask about prices; think "depletion zone" and you ask about diffusion gradients. The metaphor pre-decides what gets measured – and what passes as confirmation.

The second is about citation drift as a systemic property. Karst and colleagues' finding describes no fraud but a gradient: every citation drops one qualification, because qualifications cost space and punchlines earn attention. After enough relays, "in one stand, in fine roots, with alternative explanations" becomes a smooth sentence. That unsupported claims doubled in 25 years is not a moral finding but a thermodynamic one about information.

The same mechanic exists in engineering. "The service mesh handles the retries." "Kubernetes makes it resilient." "The systems are decoupled via the bus." These are architecture diagrams in the state of a metaphor: the edge exists, the function is unverified. In almost every postmortem there is, at the end, an arrow everyone took for a pipe and on which nobody had ever measured traffic.

The Central Takeaway

The central lesson is this: a demonstrated connection is not a demonstrated function. Anyone who wants to test a network claim must ask three separate questions – does the path exist, does anything flow along it (and how much, relative to the alternative paths), and does it change the behaviour of the recipient?

The mycorrhizal debate is the textbook case, because in it all three questions have different answers: path partly mapped, flux measured in the field, effect on the recipient open. Merge all three into one statement and you get the mother tree.

Four concrete steps follow:

1. When reading: look up the cited source, not the citing one. If a claim is interesting enough to pass on, it is interesting enough to check the original paper's "Limitations" section. The doubling of unsupported claims over 25 years is built out of exactly that skipped minute.
2. In the garden and the forest: don't turn the soil, don't buy the infrastructure. The most robust practical finding from the global survey is that cropland soils have lost nearly half their hyphal density – so mechanical disturbance matters. I am of the opinion that restraint towards commercial mycorrhizal inoculants is warranted: the evidence that introduced foreign strains establish durably and deliver a yield benefit is considerably weaker than the advertising, whereas the effect of reduced tillage and continuous ground cover is well documented.
3. In engineering: measure the edge, don't trust the diagram. For every connection treated as a safety net – replication, failover, retry path, backup restore – an operational measurement belongs in place showing that something actually flows there. A replica that is never read from is a fungal thread without demonstrated transfer.
4. In communication: make the metaphor explicit. Anyone explaining a new system chooses an image. It pays to say out loud which one – and what it does not claim. "Like a market, but without intent" is a ten times more useful opening than "like an internet."

A Question to Think About

The mycorrhizal story is so persistent because it serves a longing: that beneath visible competition lies a hidden cooperation. The data say something more sober and at the same time larger – there is a gigantic planetary infrastructure, but it belongs to no one, serves no one, and has interests of its own.

From which an uncomfortable question follows: which of the connections you rely on – technical, organisational, personal – have you ever measured for flow, and which do you know only from the diagram? And for which of them would the result of a measurement be uncomfortable enough that this is why you have not measured it yet?

Cross-References in the Vault

• When Bacteria Take a Vote: Quorum Sensing and the Secret Language of Microbes – the sister question one level down: when is a chemical threshold a "language," and when is that just our metaphor?
• The Enemy That Became a Power Plant: Endosymbiosis and the Bacterial Origin of Complex Life – symbiosis as a permanent state rather than an exception, with the same basic pattern: trade that becomes irreversible.
• The Strongest Bond in Nature: Nitrogenase and How Bacteria Split Nitrogen at Room Temperature – the other great nutrient symbiosis of land plants; phosphorus and nitrogen are the two currencies of this market.
• The Machine That Splits Water: Photosystem II and the Origin of Oxygen – the source of the carbon that, in this article, travels down into the soil.
• The 38 Witnesses Who Never Were: The Bystander Effect Between Laboratory Finding and Video Evidence – the same epistemic case with the opposite sign: there the finding outlives the myth, here it shrinks along with it.
• The Peak of Not Knowing: Why Psychology's Most Famous Effect May Be Just a Statistical Mirage – a textbook finding that breaks under methodological scrutiny, and how to spot it.
• The Invented Memory: Why Recollection Is Not a Recording – and How Elizabeth Loftus Showed You Can Remember Things That Never Happened – citation drift inside a single brain: how a catchy version takes the place of the observation.
• The Whisper of Machines: Gossip Protocols, SWIM, and How a Cluster Learns Who Is Still Alive – a genuine decentralised network with demonstrated message paths, as a contrast to the presumed fungal internet.
• The Long Tail of Latency: Tail Latency and Why Averages Lie in the Cloud – why aggregated averages (here: transfer amounts) hide the interesting distribution.
• Four That Act in Concert: Allostery, Hemoglobin, and the Logic of Cooperative Binding – source–sink gradients and cooperativity without any intent, as the molecular version of the same fallacy.

Sources

• Simard, S. W., Perry, D. A., Jones, M. D., Myrold, D. D., Durall, D. M., & Molina, R. (1997): Net transfer of carbon between ectomycorrhizal tree species in the field. Nature, 388(6642), 579–582. DOI: 10.1038/41557
• Read, D. (1997): The ties that bind. Nature, 388(6642), 517–518. DOI: 10.1038/41426
• Beiler, K. J., Durall, D. M., Simard, S. W., Maxwell, S. A., & Kretzer, A. M. (2010): Architecture of the wood-wide web: Rhizopogon spp. genets link multiple Douglas-fir cohorts. New Phytologist, 185(2), 543–553. DOI: 10.1111/j.1469-8137.2009.03069.x
• Kiers, E. T., Duhamel, M., Beesetty, Y., Mensah, J. A., Franken, O., Verbruggen, E., Fellbaum, C. R., Kowalchuk, G. A., Hart, M. M., Bago, A., Palmer, T. M., West, S. A., Vandenkoornhuyse, P., Jansa, J., & Bücking, H. (2011): Reciprocal rewards stabilize cooperation in the mycorrhizal symbiosis. Science, 333(6044), 880–882. DOI: 10.1126/science.1208473
• Babikova, Z., Gilbert, L., Bruce, T. J. A., Birkett, M., Caulfield, J. C., Woodcock, C., Pickett, J. A., & Johnson, D. (2013): Underground signals carried through common mycelial networks warn neighbouring plants of aphid attack. Ecology Letters, 16(7), 835–843. DOI: 10.1111/ele.12115
• Klein, T., Siegwolf, R. T. W., & Körner, C. (2016): Belowground carbon trade among tall trees in a temperate forest. Science, 352(6283), 342–344. DOI: 10.1126/science.aad6188
• Whiteside, M. D., Werner, G. D. A., Caldas, V. E. A., van't Padje, A., Dupin, S. E., Elbers, B., Bakker, M., Wyatt, G. A. K., Klein, M., Hink, M. A., Postma, M., Vaitla, B., Noë, R., Shimizu, T. S., West, S. A., & Kiers, E. T. (2019): Mycorrhizal fungi respond to resource inequality by moving phosphorus from rich to poor patches across networks. Current Biology, 29(12), 2043–2050. DOI: 10.1016/j.cub.2019.04.061
• Karst, J., Jones, M. D., & Hoeksema, J. D. (2023): Positive citation bias and overinterpreted results lead to misinformation on common mycorrhizal networks in forests. Nature Ecology & Evolution, 7(4), 501–511. DOI: 10.1038/s41559-023-01986-1 (author correction: DOI 10.1038/s41559-023-02035-7)
• Henriksson, N., Marshall, J., Högberg, M. N., Högberg, P., Polle, A., Franklin, O., & Näsholm, T. (2023): Re-examining the evidence for the mother tree hypothesis – resource sharing among trees via ectomycorrhizal networks. New Phytologist (2023). DOI: 10.1111/nph.18935
• Blatt, M. R., Draguhn, A., Taiz, L., & Robinson, D. G. (2023): A challenge to claims for mycorrhizal-transmitted wound signaling. Plant Signaling & Behavior, 18(1), 2222957. DOI: 10.1080/15592324.2023.2222957
• Hawkins, H.-J., Cargill, R. I. M., Van Nuland, M. E., Hagen, S. C., Field, K. J., Sheldrake, M., Soudzilovskaia, N. A., & Kiers, E. T. (2023): Mycorrhizal mycelium as a global carbon pool. Current Biology, 33(11), R560–R573. DOI: 10.1016/j.cub.2023.02.027
• Klein, T., Rog, I., Livne-Luzon, S., van der Heijden, M. G. A., & Körner, C. (2023): Belowground carbon transfer across mycorrhizal networks among trees: Facts, not fantasy. Open Research Europe, 3, 168. DOI: 10.12688/openreseurope.16594.1
• SPUN Mapping Consortium, Van Nuland, M. E., Averill, C., Stewart, J. D., Prylutskyi, O., Corrales, A., van Galen, L. G., Manley, B. F., Qin, C., Lauber, T., et al. (2025): Global hotspots of mycorrhizal fungal richness are poorly protected. Nature, 645(8080), 414–422. DOI: 10.1038/s41586-025-09277-4
• Simard, S. W., Ryan, T. L., & Perry, D. A. (2025): Opinion: Response to questions about common mycorrhizal networks. Frontiers in Forests and Global Change, 7, 1512518. DOI: 10.3389/ffgc.2024.1512518
• Stewart, J. D., Bisot, C., Cargill, R. I. M., Van Nuland, M. E., Hawkins, H.-J., Oyarte Galvez, L., Klein, M., van Son, M., et al. (2026): Global density and biomass of arbuscular mycorrhizal fungal networks. Science, 392(6803), 1171–1176. DOI: 10.1126/science.adu4373

← All articles