Plant Journal

Imagine: the fungal threads beneath the forest are a nervous system. Who or what, then, is an oak tree? A fern? A tuft of grass?

First, the honest truth: that metaphor is too strong. What is established is that mycorrhizae connect roots and transport substances — plants that do not photosynthesize themselves, such as mycoheterotrophic species, even depend entirely on them. What is *not* proven: that such networks exist everywhere in forests, that trees feed and warn each other through them, and that they recognize kin. Of the 28 field studies involving a network blockage, five found a positive effect on seedlings — and just as many found a negative effect. The "nervous system" is therefore more of a shared nutrient and signaling network than a brain.

In that network, plants are not passengers but three things at once:

• the power plant: the network runs on sugars from photosynthesis; without a canopy, there is no power;
• the sensor: each plant locally measures light, water, and damage, feeding that data into the network;
• the endpoint: what the fungus extracts from the soil — phosphorus, nitrogen — goes back into the plant.

The oak is the long-lived backbone in this setup: decades to centuries at the same node, with substantial reserves. The fern is the outsider: an ancient plant lineage that often forms weak mycorrhizae or none at all — a node that barely participates. The grasses are the fine-meshed edge: arbuscular fungi, dense root systems, short lifespans — high bandwidth, rapid turnover.

And the quantum computer? There is one profound similarity: both compute within the material itself. The slime mold converging on the shortest path is an analog computational process; a quantum computer allows amplitudes to interfere so that incorrect paths cancel each other out. In both cases, you let physics find the answer. But the differences are vast: the forest does not employ superposition or entanglement as computational tools — signals between trees are classical correlations, not entanglement. Quantum effects *do* exist in the forest (in photosynthesis), but that is quantum biology, not quantum computing.

The forest, therefore, is not a quantum computer. It is an analog computer without a programmer — and the slime mold is proof that such computers exist.

Source(s): Karst et al. - 'Positive citation bias and overinterpreted results ... mycorrhizal networks' (Nature Ecology & Evolution, 2023, https://doi.org/10.1038/s41559-023-01986-1); Merckx et al. - 'Mycoheterotrophy in the wood-wide web' (Nature Plants, 2024, https://doi.org/10.1038/s41477-024-01677-0); Engel et al. - 'Evidence for wavelike energy transfer through quantum coherence in photosynthetic systems' (Nature, 2007, https://doi.org/10.1038/nature05678); Adamatzky - 'Physarum Machines' (World Scientific, 2010)

Place two oat flakes in a petri dish with agar, put a piece of yellow slime mold between them — and wait. Within twenty-four hours, the organism has spread over everything between the flakes; after a few days, almost everything has retracted again, except for one thick tube: the shortest route between the two flakes. This is the most famous experiment with Physarum polycephalum, a single-celled organism that can grow up to a square meter in size — not a fungus, despite its name, but an amoeba-like protist with thousands of nuclei in a single cell mass.

How it works: the mold explores broadly, and in every tube, the flow determines its fate. Where a high flow of nutrients passes, the tube grows thicker; where little flows, it withers. That single principle — more flow, thicker tube — is enough to ultimately leave only the shortest path. Mathematicians have proven that this mechanism always converges to the shortest route, regardless of the network or the initial distribution.

Even better known than the maze is the map of Tokyo: with oat flakes placed at the locations of the cities, the slime mold built a network strikingly similar to the Japanese railway network — comparably efficient and even a touch more robust. On a 3D map of Los Angeles, it did the same: its network was, in scaled terms, over 65 km shorter than the human road network.

And what does it like to eat most? Sugars — but far from all of them. In the classic chemotaxis experiments (1970–1978), the mold was most strongly attracted to glucose (starting at just 0.25 mM), followed by maltose, galactose, and mannose; lactose and N-acetylglucosamine also lure it. It showed no response to sucrose, fructose, and ribose — even though it can digest fructose. Sensing and eating are therefore two separate systems: 2-deoxy-glucose attracts it even though it has no nutritional value. And there is a limit to the love: around 100 mM, attraction turns into repulsion (osmosis).

Why rolled oats, then? An oat flake is about 70% carbohydrate — starch that gradually breaks down into the exact glucose and maltose that the mold is most attracted to, plus proteins, fats, and minerals for growth. Moreover, it can smell the flake from a distance: the sugars diffuse through the agar. No wonder rolled oats have long been the standard food in every slime mold lab.

What this teaches us: a network that finds the shortest paths needs no brain — just a single rule and patience. Much like the plants in our installation: the answer lies within the system, not in a central hub.

How to run the test yourself: 2% agar in a petri dish, oat flakes in fixed positions, dark and humid at 25 °C, a photo every few hours, and measure the routes after 48–72 hours (e.g., in ImageJ).

Source(s): Nakagaki, Yamada & Toth - 'Maze-solving by an amoeboid organism' (Nature, 2000, https://doi.org/10.1038/35035159); Tero et al. - 'Rules for Biologically Inspired Adaptive Network Design' (Science, 2010, https://doi.org/10.1126/science.1177894); Bonifaci, Mehlhorn & Varma - 'Physarum can compute shortest paths' (2012, https://arxiv.org/abs/1106.0423); 'Investigating Los Angeles' urban roadway network' (2020, https://pmc.ncbi.nlm.nih.gov/articles/PMC6964692); Knowles & Carlile - 'Chemotactic response ... to sugars' (1978, https://doi.org/10.1099/00221287-108-1-17); Carlile - 'Nutrition and chemotaxis' (1970, https://doi.org/10.1099/00221287-63-2-221); Whiting & de Lacy Costello - 'Sensory fusion in Physarum' (2014, https://arxiv.org/abs/1403.4795)

First, clearing up a misconception: a plant does not "produce" voltage per day. Voltage is an instantaneous state — what is measurable are signals: spikes and slow waves following events.

The orders of magnitude:

• Inside the cell (membrane potential): resting potential of tens of millivolts (negative inside); action potentials from several tens to well over 100 mV. Classic examples: soybean ~60 mV upon light, potato 40 ± 10 mV upon herbivory, Venus flytrap well over 100 mV.
• At the surface (our setup — electrodes on leaf or stem, reference in the soil): much smaller. Measured stress responses: ~7–10 mV (thale cress, cassava), with outliers up to tens of millivolts upon wounding; weak or slow events are in the sub-mV range. And plant signals are slow: virtually everything below 0.2 Hz.

And then the 35-degree day. Heat is a stimulus: moderate warming hyperpolarizes the membrane, but above the optimum (around 30 °C for many plants) it shifts to depolarization — real signals. In thale cress, a heat shock triggers a single action potential. On a hot day, expect more signals (heat plus dehydration), not necessarily higher voltages: it remains a few millivolts at the surface.

Two caveats for the measurement itself on such a day: Ag/AgCl electrodes are temperature-sensitive (thermal voltages and drift), and the plant transpires more while the soil dries out — exactly why laboratories keep the temperature constant and measure inside a Faraday cage. At 35 degrees you therefore measure more plant and more artefact; the distinction lies in the pattern: slow, stimulus-coupled waves versus drift and hum.

Source(s): Volkov - 'Plants as Environmental Biosensors' (2012, https://pmc.ncbi.nlm.nih.gov/articles/PMC2635006); 'Evaluating Bacterial Nanocellulose Interfaces for Recording Surface Biopotentials from Plants' (2024, https://pmc.ncbi.nlm.nih.gov/articles/PMC11014089); Sukhov et al. - 'Electrical Signaling of Plants under Abiotic Stressors' (2021, https://pmc.ncbi.nlm.nih.gov/articles/PMC8509212); Fromm & Lautner - 'Electrical signals and their physiological significance in plants' (2007, https://onlinelibrary.wiley.com/doi/full/10.1111/j.1365-3040.2006.01614.x)

Plants turn light into energy — sugars. Can sound come out of that? There are three routes, from literal to poetic.

1\. Harvesting the energy. Bacteria that release electrons live in the soil around roots. With two electrodes, you turn a planter into a fuel cell — a plant-microbial fuel cell. It works: companies like Plant-e have been building installations with it for years. But the yield is small, milliwatts per square meter: enough for a tiny audio signal or a small light, not for a loudspeaker. The plant can therefore power its own voice — but softly.

2\. Making the signals audible. This follows the same path as our light installation: electrodes on the leaf or stem measure voltage fluctuations. Where we translate them into light, you can also translate them into tone — devices like Music of the Plants (since the 1970s, Damanhur) and PlantWave have been doing this for decades. A single codebase can control both light and sound: the same signal, two senses.

3\. The plant is already making sound. The strongest answer: plants emit ultrasonic clicks (20–100 kHz), especially under stress. In drought conditions, researchers measured 35 clicks per hour for tomato and 11 for tobacco; healthy plants stayed below one per hour. Those clicks can be measured meters away and reveal the state of the plant — they likely occur because air bubbles pop in the vascular bundles. Slowed down to the audible range, a plant sounds like a gentle rain of clicks.

For our installation, this means: sound can be included — not as a translation of "what the plant is saying", but as an extra sense applied to the same signals and measurements.

Source(s): Khait et al. - 'Sounds emitted by plants under stress are airborne and informative' (Cell, 2023, https://doi.org/10.1016/j.cell.2023.03.009); Plant-e - elektriciteit uit levende planten ( https://www.plant-e.com); Music of the Plants (Damanhur, https://www.musicoftheplants.com); PlantWave ( https://plantwave.com)

Plants do not see light, but they do measure it. Photoreceptors in the leaf register how much and what kind of light is present, thereby controlling the plant's day and night clock: when the stomata open, when sugars are produced and distributed, when cells elongate.

That rhythm is both the noise and the signal in every electrophysiological experiment. If you turn on the light, the voltage across the cell membrane changes within seconds in numerous places — a natural response that you can use as a stimulus, provided you separate it from the rest. That is why laboratories work with fixed light protocols: periods with light on and off, fixed patterns, and control periods in which nothing changes. This way, you know which change was caused by the light and which the plant generated itself.

Light can also come from the plant itself: bioluminescence, like the faint glow of glow-worms at dusk, is the reverse phenomenon — biological light that communicates something to other organisms. In the installation, such a subtle light occasionally appears among the plants.

For the measurements, this means one rule: the light cycle is a measurement variable, not decor.

Beneath the pot, between the roots, lies a world that measurably flows. Soils are full of microbial life, and part of that life is electrochemically active: microorganisms that exchange electrons with their environment. Researchers use this activity as a measure of soil health — the electrical signals between plant and soil life reveal what the system needs.

In addition, there is the network that literally connects the soil: fungi that link the roots of different plants together via mycorrhiza. They transport nutrients — and according to a growing number of studies, also information: a warning or stress signal can travel from one plant to another through the fungal network. Exactly how significant that role is remains the subject of ongoing research; the findings are promising but not yet conclusive.

For our installation, this is the third layer: the plants in the hall stand in potting soil, and that soil is not a dead substrate, but a living network. What happens in the pot is perhaps the oldest communication system in the room.

That layer is particularly interesting for monitoring: soil signals can be measured without touching the plant itself — precisely what is needed for vulnerable collections.

Bron(nen): Washington State University - 'Engineers, plant scientists decode electrochemical soil signals' (2026, https://research.wsu.edu/news/engineers-plant-scientists-decode-electrochemical-soil-signals); arXiv 2509.08434 (preprint) - 'Information and Communication Theoretical Foundations of the Internet of Plants' ( https://arxiv.org/abs/2509.08434)

Where does a signal start when you touch a leaf? Anyone who measures plants electrically knows the limitation: with one or two measurement points, you know that something is happening, but not where it started and where it is heading. Researchers in Singapore and at A\*STAR have changed that: using a soft, flexible array of sensors along the petiole of the Mimosa pudica — the plant that folds its leaves upon touch — they tracked the signal along its entire length.

From the measurements, maps in time and space emerged: you see the signal arise, propagate towards the base of the petiole, and only then does the leaf fold down. Upon touch, cold, and heat, they observed the same pattern: first sensing and signalling, only then moving — the movement itself took place in less than half a second. The same approach also worked in other plants, such as thale cress and pak choi.

That detail is more than just a curiosity. It shows that an event on a single leaf travels electrically through the plant before the visible reaction follows. For anyone wanting to measure plant signals — including in our installation — that is the core: the signal is there before the response.

The sources are listed at the bottom of this piece: the press release from NTU Singapore and the scientific article in Nature Electronics.

Source(s): NTU Singapore - 'Soft electrode array maps electrical signals travelling through plants' (15 sep 2026, https://www.ntu.edu.sg/mse/news-events/news/detail/soft-electrode-array-maps-electrical-signals-travelling-through-plants); wetenschappelijk artikel in Nature Electronics (mei 2026)

A plant that is damaged does not emit sound, but a mixture: volatile organic compounds — VOCs. Small molecules that evaporate from leaves, stems, and flowers, and prime neighboring plants in their defenses. That scents carry information was already known. How a plant perceives them precisely is still largely a mystery — and new overviews on this very topic appeared in 2026.

What makes it remarkable: plants do not recognize some scents directly, but via a detour. Certain receptors appear to first convert the volatile molecule into a more active form. Methyl salicylate, for example, is converted in the leaf into salicylic acid — a classic signaling molecule — and only then does the defense start. And the message does not lie in a single substance: the ratio of the mixture determines which message comes across. Omitting just one component from a mixture can already make the message unreadable.

A new study followed that conversation without touching the plants: with time-series recordings, receivers reacted to the scents of damaged plants, even when the best-known 'green leaf scents' were missing. Damaged plants emitted an average of about 22 compounds; healthy ones about 10.

For our installation, where the electrical side is central, this is the second major language: scent travels through the air, signal travels through the plant. Whoever follows both follows a conversation on two channels.

Source(s): The Plant Journal - 'Molecular insights into volatile organic compound sensing and signaling in plants' (14 mrt 2026, https://pmc.ncbi.nlm.nih.gov/articles/PMC12988820); Frontiers in Plant Science - 'Revisiting volatile organic compounds' role in plant communication' (2026, https://www.frontiersin.org/journals/plant-science/articles/10.3389/fpls.2026.1829872/full)

Put two plants in one pot and send a small voltage signal through one: what carries it to the other — the roots or the soil? The answer is sobering: the soil. Not the sand itself (dry sand conducts nothing, it is glass), but the water in it with its dissolved ions. Wet potting soil is an electrolyte; the entire volume between the two plants participates. The roots are a much thinner, high-impedance parallel path — a negligible share.

That is precisely why "communication" between plants in a shared pot is so suspect. If you measure a signal at plant two, chances are high that you are measuring the voltage distribution in the wet soil, not a message from the plant. Researchers working with electrodes in soil know the problem: when mapping roots, the injection current leaks away through the soil and skews the measurement. And devices like PhytlSigns even measure plant signals against the soil as a reference — so the soil is always part of the circuit.

How to figure it out yourself (four tests, in this order):

1. Without plants: inject the signal between the two plant positions in the same pot. Do you receive it? Then it is the soil. 2. Dry versus wet: repeat with dry and with wet soil — does the signal scale with the moisture? Then it is the water film. 3. Grounding in between: insert a grounded metal plate between the two plants. Does the signal disappear? Then it was traveling through the soil. 4. Two pots: the reference setup. What remains when the soil is separated is what genuinely travels through the plants.

Conclusion for our setup: in a single shared pot with wet soil, the soil is the best conductor — and therefore also the greatest source of spurious correlations. If you want to know what the plants themselves are doing, the soil must be silenced (isolated, grounded, or actively guarded) — exactly as the feasibility study already prescribes.

And is grounding the solution, then?

Depends on what you mean. A measurement reference — a single well-defined zero point, for instance an electrode in the soil of that one plant — is useful and standard: PhytlSigns measures plant against substrate and explicitly pays attention to the grounding of the entire instrument. But mains ground (the wall socket, the metal frame, the water pipe) is precisely the wrong thing: it creates ground loops, pulls in hum, and connects your channels via an alternative path. The rule is: the shield may be grounded, the signal remains floating — battery power, isolated amplifiers, short shielded wires.

And in a single shared pot, grounding does not resolve crosstalk: the soil connects everything anyway, and an extra grounding electrode is simply an additional node in the same soup — it even provides the injection current with an easy return path. What *does* work, in order: separation (two pots), a grounded plate between the plants, active guarding, and measuring differentially per channel with isolated amplifiers — so that the soil becomes common-mode and the amplifier rejects it.

And a shared water bath?

Put two plants together in water with some sand and send the signal via the leaves: then you get the short circuit you are referring to — in a practical sense. Water with ions provides a resistance of hundreds of ohms between the root zones, while the plant path from leaf to root is in the kilo- to mega-ohm range. The current overwhelmingly chooses the water: both root systems become a single electrical node, and what you measure at the other leaf is the potential field of the bath — not of the plant. There is no danger at microvolts: it is not a hazardous short circuit, just a measurement that no longer says anything about the plants. (Only at larger currents do electrolysis, polarization, and root damage come into play.)

What does work: a dedicated water reservoir per plant (separated), or — if a single bath is required — keeping leaves and electrodes dry above water, measuring differentially per plant against its own reference, galvanically isolating channels, and actively guarding the bath if necessary. And the same test as always: perform the measurement without plants first. Do you measure the bath? Then you are measuring the bath.

How do you recognize the short circuit in the spectrum?

The water itself conducts all frequencies — from DC to megahertz — so the short circuit is not frequency-selective. But that is precisely the giveaway: **the plant *is* selective**. Plant tissue with its membranes acts as a low-pass filter (real plant signals are slow, below 0.2 Hz), whereas through the water all the fast clutter comes through as well: 50 Hz hum and its harmonics, and any test tone you inject.

So: receiving high frequencies (tens to thousands of Hz) = suspicious — water, cables, mains power. Low frequencies, slow (below 1 Hz) = potentially the plant. And because the electrodes themselves have a high impedance below ~1 Hz (electrical double layer), the water path actually works best from hundreds of Hz to several kHz — precisely where the plant lets nothing through anymore.

The test: sweep a sine wave from 0.1 Hz to 1 kHz and observe amplitude and phase. Flat or rising transmission at high frequencies → the water. A low-pass characteristic with a roll-off in the low hertz → the tissue.

Source(s): 'Non-Invasive Phenotyping of Sugar Beet and Maize Roots Using Field-Scale Spectral Electrical Impedance Tomography' (2025, https://pmc.ncbi.nlm.nih.gov/articles/PMC12415421); 'Electrophysiological assessment of plant status outside a Faraday cage' (2019, https://pmc.ncbi.nlm.nih.gov/articles/PMC6864072); ERT-basisprincipes (SOIL, 2025, https://soil.copernicus.org/articles/11/67); 'Plant electrical signals: A multidisciplinary challenge' (2021, https://www.sciencedirect.com/science/article/pii/S0176161721000572)