37 Mushrooms, One Underground Network: How Fungi Whisper Electrical News Through the Forest Floor

The Experiment: Eavesdropping on 37 Mushrooms at Midnight

On October 23, 2022, at 11:45 PM, Yu Fukasawa's team entered a secondary mixed forest dominated by *Quercus serrata* in Kami, Miyagi Prefecture, Japan, carrying an industrial voltage data logger (GRAPHTEC GL840-Midi)—equipment you would normally find in a factory, not a forest.

Six months earlier, they had applied urea to a 5 m × 5 m plot at an extreme dose of 640 g/m². Urea decomposes into ammonia, which awakens so-called "ammonia fungi"—species that rarely fruit under normal conditions but erupt when nitrogen levels spike. Tonight, 37 mushrooms stood on the plot: 29 *Hebeloma danicum* and 8 *Hebeloma cylindrosporum*, belonging to two species and multiple distinct genetic clones (genets).

The team inserted 37 pairs of stainless-steel subcutaneous needle electrodes—one needle in each cap, one at the base of each stipe—measuring potential differences at one-second intervals within a ±100 mV range. Recording ran for 3.5 days, from 23:45 on October 23 to 8:00 AM on October 27.

During that window, they applied three stimuli at defined moments:

• Local water: 200 ml of tap water poured at the base of one mushroom (#1)
• Local urine: 200 ml of urine from one of the authors (a healthy 44-year-old male)
• Global water: uniform spraying of 10 L/m² of tap water across the whole plot, simulating moderate rain

The results, published March 2, 2026, in *Scientific Reports* as *"Electrical information flows across the sporocarps of two ectomycorrhizal fungi in the field,"* point to a striking conclusion: the mushrooms were communicating—and the pattern of that communication follows the logic of information theory.

Background: How Mushrooms "Talk"

Mushrooms have no nervous system and no centralized information-processing organs. But the vast underground mycelial network behind them does have electrical activity: ion gradients across fungal cell membranes produce potential changes when ion channels open, and these changes propagate along hyphae.

Physically, this resembles neuronal action potentials—both are electrical signals from ion flows. The key difference: neuronal spikes are all-or-nothing digital signals, while fungal signals are graded analog ones, continuously varying from a few millivolts to tens of millivolts.

In 2022, Andrew Adamatzky (University of the West of England) recorded 24 hours of electrical activity from four fungal species—*Omphalotus nidiformis*, *Flammulina velutipes* (enoki), *Schizophyllum commune*, and *Cordyceps militaris*—and analyzed the spikes using algorithms borrowed from human linguistics. He found spike clusters resembling "words," with vocabularies of 14–50 per species and word-interval distributions statistically similar to human language. Published in *Royal Society Open Science*, the study was widely (and somewhat over-)reported as "mushrooms have a 50-word language." Adamatzky himself was cautious—but the work established that fungal electrical signals are structured, not noise. And structure implies information. And information implies someone may be listening.

Findings: A 37×37 Information-Flow Matrix

Fukasawa's team asked a harder question than Adamatzky: is electrical signal between mushrooms actually *transmitting information*? To answer it, mere correlation is not enough—you need causal analysis. They used Granger causality: if the past values of time series A improve prediction of time series B beyond what B's own past provides, then A carries information for B. The result is a 37×37 information-flow matrix, which revealed three surprises:

1. Information flow is not restricted to the same clone. If signals traveled only within a single mycelial network, cross-clone and cross-species pairs should show no flow. The data showed flow across both—arguably via the soil itself: one mushroom's hyphal activity alters local ion concentrations, which neighboring, physically unconnected fungi can sense. A "chemical telegraph" transmitted through the environment rather than wires.

2. Information flow decays with genetic and spatial distance. Cross-species flow is weaker than same-species; cross-clone weaker than same-clone; farther apart means weaker. This "double decay" is itself evidence of genuine transmission—if mushrooms merely responded synchronously to shared environmental factors, the decay pattern would not track both distances coherently. Signals attenuate in transit, plausibly because cross-species "translation" costs information.

3. Local watering increased information flow; global watering decreased it. This is the experiment's most striking contrast, elaborated below.

"If Everyone Already Knows, There's Nothing to Say": Information Theory Underground

Fukasawa's own explanation is arguably the most elegant line of the paper:

> "Applying water to all the mushrooms may mean that there's no need to share information since the whole network already knows what's going on, which could be why the flow of information decreased in this situation."

This maps directly onto a core principle of information theory: information equals a reduction in uncertainty. "The sun will rise tomorrow" carries zero information because you already knew it. When one mushroom gets watered, it "knows" local soil moisture rose—but the others do not, so the information has propagation value, and flow increases. When the entire plot is watered, every node simultaneously holds the same information; redundancy reaches 100%, and communication drops to a minimum.

This is not mushrooms being "lazy"—it is redundancy elimination emerging naturally in a biological network.

Key insight: the network is not propagating the fact "water arrived," but the fact "I know something you don't." The former is mere physical signal conduction; the latter has the structural signature of communication—senders transmit what receivers lack.

Why the Urine Trial Was the Clever Control

The urine experiment looks bizarre but was a carefully designed control. All 37 mushrooms were ammonia fungi—already activated months earlier by urea. Urine's main component is urea: a *food signal* these fungi had already received. Water was a *neutral, unexpected* signal. The logic:

• Local water: neutral signal, local perturbation → information flow increased
• Local urine: food signal, local perturbation → smallest change in information flow
• Global water: neutral signal, global perturbation → sharp decrease in information flow

Urine's minimal effect seems counterintuitive—shouldn't food matter more? But it confirms the information-theoretic reading: to a nitrogen-activated network, urine is "old news" with maximal redundancy. Water is "news." News gets transmitted; old news doesn't. If the mushrooms merely reacted to any perturbation, water and urine would have elicited similar responses. They didn't. Distinguishing "news" from "old news" is a hallmark of a communication system, not passive physical conduction.

A Brainless, Centerless Information Network

The "Wood Wide Web" concept—mycorrhizal networks exchanging resources between trees and fungi—has long faced a skeptical question: is this *communication*, or just passive physical diffusion? Sugar flows down concentration gradients without intent; so does water. If mycelial networks are passive plumbing, "Wood Wide Web" is a romantic metaphor, not science.

Fukasawa's electrical data offers a forceful rebuttal: electrical signals are actively generated, structured, and modulated by informational redundancy—following the laws of information theory rather than biochemistry alone. Mycorrhizal networks are not just resource pipelines; they are information networks, exhibiting redundancy elimination, distance decay, and cross-node propagation—all without a brain, a center, or a nervous system.

We in AI are re-inventing, over decades, what these networks have optimized over roughly 400 million years: distributed learning, decentralized communication, redundancy-aware messaging.

Back to the Silent Morning

At 5:55 AM on October 25, the team watered mushroom #1. In the hours that followed, Granger analysis showed information flow radiating outward. At another moment, they sprayed the entire plot: the voltage traces still fluctuated, but causal links among the 37 mushrooms collapsed. All moving—but no longer "talking." Because there was no need.

When every node knows what every other node knows, redundancy hits 100% and the communication channel goes quiet. That silence is not failure—it is the network reporting that communication costs can fall to their minimum. It is efficiency at its limit.

Our human networks, by contrast, forward news everyone has already read and repeat statements everyone has already heard—high redundancy without the bandwidth reduction, because we lack a mechanism to detect "does the other side already know this?" The mushroom network has one. We don't—yet. Ours is 50 years old; theirs has had 400 million years to learn when to stay silent.

---

References:

1. Fukasawa, Y., Akai, D., Takehi, T., Takahashi, D., & Osada, Y. (2026). Electrical information flows across the sporocarps of two ectomycorrhizal fungi in the field. *Scientific Reports*. DOI: 10.1038/s41598-026-42673-y 2. Adamatzky, A. (2022). Language of fungi derived from their electrical spiking activity. *Royal Society Open Science*, 9(4), 211926. DOI: 10.1098/rsos.211926 3. Fukasawa, Y., et al. (2023). Mushroom's electrical conversation after the rain. *SSRN Electronic Journal*.