The Silent Kingdom
Aristotle drew the line between plants and animals at sensation. Plants couldn't feel, couldn't move purposefully, and certainly couldn't communicate by sound. That framework survived more or less intact for two millennia. Even as modern botany discovered that wounded leaves release volatile organic compounds and roots exude chemicals that recruit beneficial soil microbes, the acoustic channel remained off the table. Plants had smells. They had colors. They did not have voices.
The kingdom was silent. This wasn't just philosophical convention. It was physical intuition. Sound requires a mechanism: vocal cords, stridulation, membrane vibration. Plants possess none of these. When researchers in the 1960s attached contact sensors directly to plant stems and detected vibrations during drought—a process called cavitation, in which air bubbles form and collapse inside water-conducting xylem vessels—the finding stayed niche. Those vibrations traveled through tissue, not air. Nobody had shown that a plant could produce a sound recordable from across a room.
In 2023, a team at Tel Aviv University did.
Popcorn in an Acoustic Chamber
Itzhak Khait, working in Lilach Hadany's evolutionary biology lab, placed tomato and tobacco plants inside a soundproofed acoustic chamber with ultrasonic microphones positioned 10 centimeters away. Some plants received normal watering. Others had their water withheld for several days. A third group had their stems cut.
The well-watered plants were almost silent: fewer than one sound per hour. The stressed plants were not. Drought-stressed tomatoes produced an average of 35 clicks per hour. Cut tomatoes emitted about 25. Tobacco plants were quieter but followed the same pattern: 11 clicks per hour under drought stress, 15 after stem cutting, versus near-silence when healthy. The sounds resembled the popping of bubble wrap, rapid and irregular.
They fell in the 20 to 100 kilohertz range, well above the roughly 20 kHz ceiling of human hearing. But at approximately 65 decibels measured at 10 centimeters from the source, they were comparable in loudness to normal human conversation. Insects, bats, mice, and many other mammals hear in this frequency range without difficulty. The researchers estimate such organisms could detect these sounds from up to five meters away.
What generates them? The leading hypothesis is the same cavitation detected by contact sensors decades ago. Transpirational pull creates extreme tension in the xylem sap, and when that tension exceeds a threshold, air is drawn through pit membranes into the water-conducting vessels. The resulting bubbles expand and collapse rapidly, vibrating vessel walls and radiating sound outward into the air. The acoustic pattern fits: sounds increased with dehydration stress, peaked before the plant was visibly wilted, and declined as the plant dried out completely and fewer functional xylem vessels remained to cavitate.
An Algorithm That Learned to Listen
The team then did something no previous plant acoustics study had attempted. They trained a machine learning classifier on the recordings and asked it to distinguish between drought-stressed, cut, and healthy plants based solely on the emitted sounds.
It worked. The algorithm identified the plant's condition with roughly 70% accuracy and could also distinguish whether the source was a tomato or tobacco plant. The researchers then moved to a greenhouse—a far noisier environment with ventilation, irrigation, and equipment hum—and showed the classifier could still detect and classify plant sounds amid the background.
To rule out artifacts, the team recorded empty pots, pots filled with dry soil, and pots filled with moist soil under identical conditions. None produced the acoustic patterns seen in living plants. The sounds were biological, not environmental. Water-stressed plants also began clicking before showing any visible dehydration symptoms. The acoustic signal preceded the droop.
The study wasn't limited to two species. The researchers also recorded corn, wheat, grape, and cactus plants, all of which emitted sounds when stressed. Detailed classification analyses focused on tomato and tobacco, but the phenomenon itself appears to cut across the plant kingdom.
What This Means for a Greenhouse
Here is a calculation that, as far as published literature goes, nobody has run. A typical commercial tomato greenhouse holds 10,000 to 20,000 plants. If drought stress triggers 35 clicks per plant per hour and the entire crop is uniformly water-stressed, that greenhouse is producing 350,000 to 700,000 ultrasonic pops per hour. Roughly 100 to 200 per second. An acoustic firehose of distress signals invisible to every human walking through it.
With an array of ultrasonic microphones and the classification approach from this study, a grower could potentially detect irrigation failures hours or days before any leaf wilts, before any drone flyover spots discoloration, and before any soil moisture sensor flags a dry zone. Commercial ultrasonic microphones cost under $50 each. The signal processing runs on commodity hardware. The barrier was never technology. Nobody knew there was a signal to listen for.
The Strongest Counterargument
The most forceful objection comes from Tom Bennett, a plant biologist at the University of Leeds who was not involved in the study. The sounds do not imply communication. Cavitation is a passive physical process, not a signal that evolved to carry information. "It doesn't mean that they're crying for help," Bennett told Science News. The fact that a machine learning algorithm can decode information from the clicks doesn't mean any organism actually does. Moths, bats, and neighboring plants might be capable of hearing these frequencies, but no study has yet demonstrated that any organism changes its behavior in response to plant-emitted sounds under natural conditions. If cavitation sounds carry no ecological function, they are a curiosity of biophysics, not a revolution in plant biology.
Hadany's counterpoint is that evolution does not require intentions. If stressed plants reliably emit sounds and nearby organisms can hear them, selection pressure exists for those organisms to use the information. A moth choosing not to oviposit on a clicking plant gains a fitness advantage whether or not the plant "meant" to warn her. But this remains a theoretical argument. The ecological link is plausible, not demonstrated.
What We Didn't Prove
This is a single study from one laboratory, not yet independently replicated by another research group. The acoustic chamber experiments focused on two model species, tomato and tobacco, that may not represent the full diversity of plant acoustics. The greenhouse demonstration showed feasibility but did not run the classifier in real time across a full growing season. The 70% classification accuracy is above chance but includes a 30% error rate that would need to shrink substantially before commercial deployment. The proposed cavitation mechanism is inferred rather than directly demonstrated: the study did not simultaneously image bubble formation and record sound, so the causal link rests on correlational evidence and established biophysics. Most critically, no organism has been shown to detect and respond to these signals under field conditions. The ecological implications remain untested.
The Bottom Line
For centuries, biologists classified the plant kingdom as acoustically inert. A team at Tel Aviv University placed microphones next to stressed plants, recorded tens of thousands of ultrasonic clicks, and trained an algorithm to diagnose the plants by ear. The sounds are real, airborne, species-specific, and stress-specific. They begin before any human could tell the plant was struggling. Whether this constitutes communication is an open question. What the data show is that the silent kingdom was never actually silent.
What You Can Do
If you grow plants at home, the practical takeaway is that visible wilting is a late indicator of water stress, not an early one. By the time leaves droop, the plant's vascular system has been under strain for hours or days. Check soil moisture regularly rather than waiting for visible distress. Stick a finger an inch into the soil. If it's dry, water it. Don't wait for the plant to tell you in a language you can see.
For commercial growers and agricultural technologists, this study opens a monitoring channel that did not exist before 2023. Ultrasonic microphone arrays paired with machine learning classifiers could offer an early-warning system for irrigation failures cheaper and faster than thermal imaging or satellite monitoring. The hardware exists and is inexpensive. The algorithms need scaling, but the signal is there, and it has been there all along.