A thirsty tomato plant can be surprisingly noisy. Not to us: stand beside it in a greenhouse and you will probably hear nothing beyond leaves rustling and the ordinary sounds of the room. But place an ultrasonic microphone nearby and a different world appears. The plant can produce brief, sharp airborne clicks at frequencies far above normal human hearing.
In 2023, researchers at Tel Aviv University reported in the journal Cell that stressed plants emit these ultrasonic sounds in patterns that contain information about their condition. Dehydrated plants became dramatically noisier than unstressed controls. Plants whose stems had been cut also produced many more clicks. Machine-learning algorithms could use the recordings to distinguish different stresses and, in some experiments, different plant species.
The finding inspired irresistible headlines claiming that plants “scream” when they are thirsty. The metaphor captures the drama but not the biology. There is no evidence that a tomato experiences pain and deliberately cries for help. What scientists have demonstrated is subtler: physiological stress inside a plant can generate airborne sound, and other organisms may be capable of using that information.
A sound humans were never meant to hear
The Tel Aviv University team, led by researchers including Itzhak Khait, Lilach Hadany and Yossi Yovel, placed ultrasonic microphones about 10 centimeters from plants. Their main experiments involved tomato and tobacco, although the researchers also recorded other plants including wheat, corn, cactus and henbit.
The microphones were sensitive to frequencies between 20 and 250 kilohertz. Adult human hearing typically reaches only to around 16–20 kilohertz, so the most important plant sounds sat comfortably beyond our auditory range. The recorded clicks were concentrated mainly between about 40 and 80 kilohertz.
To human ears, recordings shifted down into the audible range resemble tiny pops or the crackling of popcorn. Each event is brief. A healthy plant is not continuously broadcasting a hidden ultrasonic song; what matters is how often the clicks occur and how their acoustic characteristics change with the plant's condition.
The difference under stress was striking. Unstressed plants produced fewer than one detected sound per hour on average. Plants subjected to drought or stem cutting produced dozens per hour. In dehydration experiments, the rate did not simply rise forever. As the soil dried, sound production increased toward a peak and later declined as the plant became severely dehydrated.
That changing pattern is one reason the phenomenon is more useful than a simple “thirst alarm.” The sounds appear to track physiological state over time.
What is making the clicks?
The precise physical mechanism remains under investigation, but one leading explanation involves the plant's plumbing.
Water moves upward from roots through microscopic conduits called xylem. The water column inside these vessels can be under substantial tension, especially when evaporation from leaves pulls water upward while the soil is dry. Under sufficiently stressful conditions, the continuous column can break and vapor or gas bubbles can form — a process known as cavitation.
The formation, expansion or movement of these bubbles can create vibrations. Researchers have long detected acoustic emissions and internal vibrations associated with cavitation in plants. One plausible explanation is that some of those events generate vibrations strong enough to propagate through plant tissue and become airborne ultrasonic clicks.
But it is important not to promote a plausible mechanism into a settled fact. The 2023 Cell study established that airborne sounds can be recorded and that their statistics change with plant stress; it did not conclusively identify the source of every click. Cavitation is a strong candidate, and other mechanical processes may contribute.
Either way, no vocal organ is involved. A plant has no lungs, larynx or nervous system equivalent to the machinery humans use to shout. “Screaming” is a metaphor for a physical consequence of stress.
AI could tell when a plant was thirsty
Recording the sounds was only part of the experiment. The researchers also asked whether those clicks contain usable information.
They trained machine-learning models on acoustic features from the recordings. The algorithms could classify whether sounds came from dehydrated or cut plants and distinguish them from controls at rates well above chance. The researchers also demonstrated that plant sounds could be detected and classified in a greenhouse, where fans and other background noises make the acoustic environment far less controlled than a laboratory chamber.
This matters because two ultrasonic clicks may sound virtually identical when converted for human listening, while their frequency structure, duration and other features can contain statistical differences that software detects.
That opens an obvious agricultural possibility. Instead of deciding when to irrigate an entire greenhouse or field according to a fixed schedule, future monitoring systems could potentially listen for acoustic evidence of water stress. Microphones paired with algorithms might identify plants approaching drought stress before obvious wilting becomes severe.
Such technology is not yet equivalent to a universal plant translator. Species, growth conditions, background noise, microphone placement and environmental stresses all complicate the problem. But the 2023 work demonstrated that the acoustic signal contains enough information to make remote stress monitoring scientifically plausible.
Are plants communicating with insects?
The most provocative question is not whether humans can exploit the clicks but whether evolution has already produced animals that do.
Many insects and mammals can hear frequencies humans cannot. The original researchers therefore proposed that moths, bats, rodents and other organisms might be capable of detecting plant-generated ultrasound. At the time, that was an ecological hypothesis rather than proof that animals actually listened to plants.
A later Tel Aviv University study provided a significant next step. Researchers tested female moths making decisions about where to lay their eggs. Moths need suitable plants because their larvae will feed after hatching, so the condition of a potential host matters.
In controlled experiments, females responded to recordings of ultrasonic plant sounds. When the researchers impaired the moths' hearing organs, the behavioral preference disappeared, supporting the conclusion that sound itself was influencing the decision. In another setup involving healthy tomato plants, females tended to avoid the plant associated with playback of dehydration sounds.
This is important because it moves the field beyond the statement that an animal could hear a plant. At least in this experimental system, an insect actually used plant-associated acoustic information in behavior.
Still, “communication” requires careful wording. A signal in evolutionary biology often implies that it evolved because transmitting information benefits the sender, while a cue may simply be an incidental feature that another organism learns to exploit. If cavitation produces clicks as an unavoidable physical consequence of drought, a moth listening to them could be eavesdropping rather than receiving a message deliberately evolved by the plant.
Does this mean plants feel pain?
No such conclusion follows from the acoustic evidence.
Pain, as studied in animals, involves sensory and nervous systems and a subjective experience that cannot be inferred simply because an organism responds to injury. Plants possess extraordinarily sophisticated signaling networks. They detect light, gravity, touch, temperature, pathogens, chemicals and damage; they alter gene expression, hormones, growth and defensive chemistry in response. None of that requires assuming human-like awareness.
The ultrasonic clicks show that plant physiology has an acoustic dimension. They do not demonstrate a mind behind the noise.
This distinction is worth preserving because the real discovery is already remarkable. For centuries humans have lived among plants while perceiving only a narrow slice of their interactions. We see leaves wilt and smell volatile compounds released by flowers, but our ears miss an entire band of frequencies in which stressed stems can become acoustically active.
What sounded like silence was partly a limitation of our senses.
A field may be noisier than it looks
The study also changes how scientists can think about an ecosystem. A dry field is not populated only by plants responding independently to soil and sunlight. It contains insects searching for hosts, predators hunting those insects, mammals moving through vegetation and neighboring plants exposed to vibrations and airborne chemicals. Some of those organisms possess sensory ranges dramatically different from ours.
If plant ultrasound reliably reflects water status or injury, natural selection gives nearby organisms an opportunity to exploit it. A herbivore might avoid a poor-quality host. A predator could conceivably use plant noises as indirect evidence of activity nearby. Whether plants themselves detect and respond to these airborne clicks remains an active research question.
The discovery therefore does not reveal a secret botanical language in the human sense. It reveals something more scientifically useful: information can exist in an environment without being intentionally spoken.
A thirsty tomato does not need to know that it is making noise. Cavitation, tension and tissue mechanics may be enough to create the clicks. A moth does not need the plant's permission to listen. And a farmer equipped with ultrasonic sensors may eventually become another eavesdropper.
Plants are not screaming for water. But when they begin to run dry, they really can become louder — in a part of the acoustic world that humans only recently learned how to hear.
