You've probably heard the "wood wide web" version of fungi: a hidden internet under the forest floor, trees talking to each other, mother trees feeding their seedlings through fungal cables in the dark. It's a genuinely beautiful story, and I've told a version of it myself — an earlier issue of mine spent real time on how fungal networks change shape from contact, which is a real and interesting mechanism. But it's also, on its own, an incomplete answer to a more basic question: what is a fungus actually doing down there, before any signal ever moves through it? Set the internet metaphor aside entirely and ask that instead, and you get chemistry and ecology stranger than the network story — and, in places, a lot more honest about what isn't known yet.

"Fungi decompose wood" is at least three different jobs

Say a fungus rots a log, and most people picture one process running slowly. It's actually three distinguishable attack strategies with three different signatures, and they don't leave the same thing behind.

White rot removes lignin, cellulose, and hemicellulose using a real enzymatic toolkit — lignin peroxidase, manganese peroxidase, laccase — and it comes in two flavors that get lumped under one name. Simultaneous white rot takes lignin, cellulose, and hemicellulose apart together, leaving wood that's uniformly lighter and spongier. Selective white rot strips lignin first, faster, leaving behind a bleached, fibrous, cellulose-rich skeleton before cellulose is even touched. Same label, two different residues.

Brown rot is where the actual surprise lives. Brown-rot fungi have lost more than 60% of the genes coding for the lignin-attacking peroxidases that define white rot — which should predict slower, weaker decay. Instead, brown-rot fungi decay wood faster than white rot in single-strain comparisons. The resolution is a mechanism with a great name: chelator-mediated Fenton chemistry. Brown-rot fungi secrete small metabolites — small enough to diffuse ahead of any enzyme into wood's still-intact micropores, where actual enzyme molecules are physically too big to go — that reduce iron and generate hydrogen peroxide. Iron plus peroxide drives the Fenton reaction, throwing off hydroxyl radicals that shred cellulose and hemicellulose non-enzymatically, from a distance, before an enzyme ever shows up. Lignin isn't removed; it's chemically demethylated and left behind, modified and brown — which is both the mechanism's fingerprint and where the name comes from. Brown rot didn't need the genes it lost. It swapped in a smaller, faster molecular tool for the job those genes used to do.

Soft rot is a third case again, mechanistically unrelated to the other two. It's mostly ascomycetes, not the basidiomycetes behind white and brown rot, and it works by boring cavities directly into one specific layer of the wood cell wall, leaving the structure between cells largely intact — unlike white rot's wholesale wall erosion. It matters most in very wet, waterlogged, or otherwise extreme conditions where the other two are outcompeted. And a detail I didn't expect going in: soft rot is documented as a late-stage process on wood that's already been delignified by white rot, not only a stand-alone attack. Decomposition, it turns out, has its own succession — one fungus's residue is another fungus's opening.

Who shows up, and when, isn't one story either

That succession keeps going once you zoom out from a single log to a whole patch of forest floor. Fresh litter gets colonized fast by fast-growing bacteria and ascomycetes. As the easy carbon runs out, more specialized organisms take over — the white-rot fungi above, plus specialist bacteria — not because lignin increases, but because it's relatively enriched: it's what's left once everything easier is gone.

Independent of that timeline, soil pH acts almost like a dial. One field study at Rothamsted Research used a natural, continuous acid-strip gradient — same soil, same site, pH ranging from 8.3 down to 4.0 over about 180 meters — and measured actual growth rates, not just standing biomass. Bacterial growth fell roughly fivefold and fungal growth rose roughly fivefold from pH 8.3 to 4.5, compounding into a roughly 30-fold shift in the fungal-to-bacterial growth ratio. But total decomposition — how much carbon actually got broken down — fell by only about a third across that same huge swing. Who does the decomposing changed enormously; how much decomposing happened barely moved. That's read as functional redundancy: different organisms, same job, mostly covered either way.

Moisture works differently again. Under drought, bacterial growth drops by roughly half in field measurements using isotope tracing, while fungal growth holds essentially steady — fungi even shift their own metabolism toward storage compounds instead of just slowing down. The likely reason is structural, not just physiological: a connected hyphal network can move water to where growth is happening in a way individual bacterial cells can't.

The single sharpest distinction in this part of the story, though, is one a litter-transplant experiment made by accident: it separated two questions that are easy to blur into one. Across a nutrient-rich grassland and a nutrient-poor forest soil, decomposition rate — how fast litter broke down — was set mainly by which soil it landed in. But which organisms actually did the decomposing tracked the chemistry of the litter itself, not the soil site. Fungi behaved as "substrate specialists and site generalists." Bacteria (specifically Actinobacteria) behaved as late-stage generalists regardless of litter type. Asking "do fungi or bacteria dominate here?" turns out to be underspecified — you have to ask separately about speed and about identity, because different variables answer each one.

Mycorrhizal fungi are not one relationship with three flavors

The other half of "fungi matter" is symbiosis — fungi living in or on plant roots, trading resources. The popular framing (mine included, in an earlier piece) treats this as an information-and-incentive problem: what gets traded, how cheating gets policed. Set that aside and ask a blunter question instead: what physically differentiates the three major mycorrhizal relationships, and what does each cost?

Arbuscular mycorrhizae (AM) are the oldest, least picky association — colonizing roughly 72% of vascular plant species, including nearly every major food crop, through structures that penetrate directly into root cells. Ectomycorrhizae (ECM) associate with only about 2% of plant species — mostly temperate and boreal forest trees: pines, oaks, beeches, eucalypts — wrapping roots externally rather than entering cells. Ericoid mycorrhizae (ERM) are narrower still: restricted to one plant family (rhododendrons, blueberries, heaths), confined almost entirely to acidic, nutrient-poor, organic-rich soils like bogs and heathlands.

Host range and carbon cost move in opposite directions. Despite touching a tiny fraction of plant species, ECM fungi collectively draw more aggregate global carbon than AM fungi do — a genuinely counterintuitive number, though one I want to be careful with: that's a global total across very different numbers of host plants, not a proven "cost more per plant" claim, and I'm not going to dress it up as more precise than it is.

The real correction is what ECM fungi are actually made of, biochemically. Since they associate with trees and clearly help move nutrients, it seems reasonable to assume they still moonlight as decomposers — after all, they evolved from free-living decomposer ancestors. That assumption doesn't survive contact with the genetics. A 2019 transcriptomics study comparing two ECM fungi from two independently evolved ECM lineages found both had lost most of the plant-cell-wall-degrading enzymes — the exact ligninolytic peroxidases and cellulases from the decomposer chemistry above — that their free-living ancestors carried. What they kept instead is a narrower, oxidative route built specifically to liberate nitrogen, not carbon, from soil organic matter. And here's the stranger part: both independently-evolved lineages arrived at that same narrow capability through recognizably different genes. Same function, different mechanism, evolved separately at least 70 times. Ericoid fungi sit closer to the opposite end of the spectrum — they carry real decomposer enzymology (multicopper oxidases closely related to known ferroxidases and laccases), genuine leftover decomposer machinery their ECM cousins mostly shed.

Ranking what's actually earned, against what's just a good story

Put the pieces together and "fungi hold the forest together" splits into tiers that don't line up with how often each claim gets repeated.

The best-evidenced tier is also the least glamorous: the decomposer-guild chemistry above, the succession and pH/moisture drivers of who decomposes what and when, and the mycorrhizal enzyme-retention pattern. Every one of these rests on multiple independently authored field or genetic studies, read directly, not inferred from a citation chain.

The middle tier is real but genuinely underspecified: fungal hyphae do measurably help hold soil aggregates together — that's an accepted starting point in a 2025 review, not a contested claim. But the specific popular mechanism — "glomalin," the famous fungal glue protein — turns out to be a case of the story outrunning the measurement. Glomalin has never actually been isolated and characterized as a defined molecule. What gets measured under that name is an operationally defined extraction fraction — heat, citrate buffer, autoclave, then an antibody test — that multiple independent papers report also picks up non-fungal humic material, lipids, and other heat-stable proteins. The correlation between hyphal length and that extraction fraction is real and reproducible. "Glomalin is the glue" is a simplification of much messier, still-disputed extraction chemistry.

The overreach tier is where the popular story goes furthest past its evidence: forest-scale network altruism, and specifically the "mother tree" claim — that mature trees preferentially send resources and warning signals to their own offspring through shared fungal networks. A 2023 Nature Ecology & Evolution paper draws a sharp line here: the existence of belowground carbon transfer between plants sharing fungal connections is real and documented since the 1960s, and that part isn't disputed. What the authors directly dispute is the leap from "transfer exists" to "these networks function as purposeful, cooperative, forest-scale support systems" — they find the field evidence for that leap "insufficiently supported," and state plainly that the specific mother-tree-feeding-offspring version has no peer-reviewed evidence at all. The sharpest forensic detail in that paper: the rate of unsupported network claims in the scientific literature itself has doubled over 25 years — meaning the field's own citation habits have been amplifying the overreach, not just popular science doing it alone. A direct 2023 rebuttal from mycorrhizal-network researchers pushes back hard, arguing carbon transfer is "solid and accumulating" evidence and citing a 2022 DNA-isotope study as the clearest recent confirmation. Both sides are real, sourced, and currently unresolved — which is itself the honest answer, more useful than picking whichever side makes the better campfire story.

What actually changed for me

Before this, "fungi decompose wood" and "fungi hold the forest together" were each one vague, roughly-true-sounding sentence in my head. They're not. One splits into at least three chemically distinct attack strategies that leave different residues and hand off to each other in sequence. The other splits into three tiers of evidence that run almost backwards from how often each part gets repeated out loud: the part with the best colorful story (talking trees, feeding mothers) has the thinnest peer-reviewed support; the part with zero narrative appeal (a fungus that lost most of its lignin-eating genes and got faster anyway by switching to iron chemistry) is the best-documented fact in the whole subject. That inversion is worth sitting with the next time a forest story sounds too good to need a source.