Plants make noise when stressed!

April 4, 2023 • 9:45 am

No, they don’t scream, but two species tested in today’s paper make ultrasonic noises when they’re either cut or stressed from drought.  Or so says a group of scientists from Tel-Aviv University in a new report in Cell.

We’ve known for some time that some species of plants emit volatile compounds when stressed, but this is the first finding that the plants make noises that travel through the air. Granted, you wouldn’t be able to hear these noises (they’re ultrasonic), but some animals might, and it’s possible (though unknown) whether other plants might be able to detect the sounds, too.

Click on the screenshot below to read the report, or find the pdf here (the reference is at the bottom). 

I’ll be brief, or try to.

The authors put potted plants, tomato or tobacco, in an acoustically-insulated chamber 1 X 1 X 0.5 m along with special microphones that could record ultrasonic sounds. They then treated the plants in three ways:

a. control 1: no treatment and no plant: just a pot
b. control 2: potted plant no treatment (they also recorded sounds of neighbor plants next to untreated plants)
c. Plants unwatered, so began to wither from drought.
d. Plants cut off at the base with scissors (cut-off piece was recorded).

Three plants at a time were put in a box, each having two microphones pointed at it, and recordings were made under the treatments above.  Here’s a diagram of the experiment with the caption from the paper. As the bottom of the figure, the sounds were then subject to computer analysis for frequency, duration, number of emissions, and THEN an AI learning program was used to see if the computer could distinguish the treatments from each other and from the controls.

Top: Acoustic box setup. In each recording, three plants are placed inside a 50 × 100 × 150 acoustic box with two directional microphones oriented at each plant. Using two microphones helps eliminating false detections resulting from electrical noise of the recording system and cross-plant interference. Bottom:  Illustration of the procedure used to train a classifier that distinguishes between tomato sounds and greenhouse noises. A background noises library was first generated, by recording inside an empty greenhouse for several days. Using this library and the library of tomato sounds recorded in an acoustic box, we trained a convolution neural network (CNN) classifier to distinguish between tomato sounds and background noises.

If you want to hear what a screaming plant sounds like, manipulated so that human ears can here it, there are audios of tomato sounds here x

The upshot is that the experiment was a success: for both tomato and tobacco, plants emitted many more sounds when stressed than when unstressed. Moreover, using machine learning, in both species cutting could be distinguished from drought, and both could be distinguished from the control—all with a pretty high level of certainty.

The figure below shows the number of sounds emitted per hour by the plants. As you can see: the control (just a pot) and the plant controls (one with the microphone focused on the untreated control plant [“self control”] and the other focused on its neighbor [“neighbor control”], emitted many fewer sounds than the stressed plants. As the authors note,

The mean number of sounds emitted by dry plants was 35.4 ± 6.1 and 11.0 ± 1.4 per hour for tomato and tobacco, respectively, and cut tomato and tobacco plants emitted 25.2 ± 3.2 and 15.2 ± 2.6 sounds per hour, respectively (Figure 1B). In contrast, the mean number of sounds emitted by plants from all the control groups was lower than 1 per hour. Our system did not record any sound in the Pot control (Figure 1B) over >500 h of recordings.

Well, it’s good to know that pots don’t make noise! Here’s the figure with its legend. For both tobacco and tomato, either treatment (drought stress or cut stress) was significantly different from the controls, and the two stresses were significantly different from each other.

I won’t go into the messy details of how the various aspects of the songs were analyzed, but the neural-network analysis was able to distinguish the two stress treatments from each other with at least 70% accuracy, which is pretty good. Here are a pair of plots showing the difference in the amplitude and power (normalized amplitude of the sound spectrum) for both treatments and for both plants. They don’t look that different, but, when combined with other aspects of sound, like frequency, the analysis was able, from the sound spectra, to tell drought from cutting with a high level of accuracy.


Examples of time signals of sounds emitted by: a drought-stressed tomato, a drought-stressed tobacco, a cut tomato, and a cut tobacco, normalized. Peak dBSPL values at 10 cms, relative to 20 μPa, are noted by the arrows (see STAR Methods).

Power (analyzed from [C]):

(D) The normalized spectra of the sounds from (C).

Finally, the accuracy of telling apart pairs of treated plants. The red line is the best one, and reflects the “scattering network” method of sound analysis. As you see, each pair could be distinguished with at least 70% accuracy, and all differences were highly significant; only the first four are relevant to us because the two treatments on the right reflect comparison of sound to the ambient “electrical noise”.

The accuracy of sound classification achieved by different feature extraction methods, with an SVM classifier. The best results were obtained using the scattering network method for feature extraction (red line, p < e−12 for each pair). Using MFCC for feature extraction the results were also highly significant (black dashed line, p < e−4 for each pair) and even basic methods for feature extraction allowed for better-than-random classification (gray line, p < e−6 for each pair apart from one case: Tobacco dry vs. Tobacco cut, which was not significant with the basic method). The comparisons Tomato vs Elect. Noise and Tobacco vs Elect. Noise refer to electrical noise of the recording system. Training set size of the two groups in each pair was equal ( sounds for each pair, see Table S1), and significance levels for each pair were calculated using Wilcoxon rank-sum test with Holm–Bonferroni correction for multiple comparisons. Error bars represent standard deviations. The classification results were reproduced using CNN models (see STAR Methods and Figure S1E).

The authors note in passing that they also recorded sounds from wheat, corn, cabernet sauvignon grapevines, henbit, and pincushion cactus (it’s not clear whether these plants were drought-stressed or cut), but didn’t get sounds from the woody parts of almond or grapevine. They note from this that:

We thus expect that many plants emit sounds, but the diversity of characteristics of these sounds are yet to be researched.

They also note that tomato plants experiencing a different stress—infection with a virus—also made noises, but didn’t analyze them to see if the noise spectra could be distinguished from those occurring under drought or cutting.

A few questions:

How are these sounds produced? The authors note: “One potential mechanism that may be responsible for the emission of at least part of the sounds we record is cavitation in the stem.” Cavitation, as noted here, is simply the formation of bubbles in the stem when, in the xylem tissue, changes in water pressure cause the formation of bubbles. It’s presumably these bubbles that make the sounds, and the authors adduce more evidence for this, but you can read the paper to see it.

Can anything hear these sounds? Using models of animal hearing, the authors conclude that these sounds could be heard “by many mammals and insects” from 3-5 meters away. I’m not sure if that’s relevant to the animals unless insects, for example, would prefer to eat or lay eggs on damaged plants. Perhaps botanically oriented readers could weigh in here

Is there any practical applications of these findings? The authors suggest that they could use the machine-generated recognition patterns to monitor plants for drought or disease in the field or greenhouse, allowing more efficient watering. But I think there are already ways to determine soil moisture and the drought status of plants.

What are potential problems with the authors’ conclusions? The authors have a section called “limitations of the study” which mentions a couple of potential problems. The most serious to me is that these noises were recorded in acoustically modified chambers that allowed maximum sensitivity of the microphones to sound. In the open field it’s much less likely that the sounds could be detected even a meter or two away.

They also mention that they used only two plant species, but note that sounds were produced in several others that were tested.  Further, we don’t know whether other stressors would cause plants to make noises, and whether those noises could be distinguished from those produced by cutting or drought. (They did say, as I noted above, that viral infection of tomatoes also led to sound production.)  Yes, we need to see this phenomena is general, and, even if not, what caused it in these two species. I suspect the phenomenon is more general since they didn’t choose for analysis just two species they knew made distinctive sounds. How general, though, remains to be determined.

And of course the biological significance of these sounds also remains to be determined. If they’re due to cavitation, they may have no significance other than being simple byproducts of plant stress, and might not affect other plants or animals in the area.

Still, in the end I think it’s pretty cool that plants make detectable sounds when they’re stressed. Think about that the next time you pick a tomato!


Khait, I. et al. 2013. Sounds emitted by plants under stress are airborne and informative. Cell: DOI:

17 thoughts on “Plants make noise when stressed!

  1. Secret life of plants redux? Hope not. The authors may not intend, but the material lends itself to support those who look to Plants for the Ethical treatment of People.

  2. I never read “The Secret Life of Plants”, but I dimly recall a sardonic John Collier story about someone who could hear the screams emitted by plants when cut. As for me, before I retired I sometimes thought I heard ghostly chuckling from the bacterial colonies on my Petri plates.

  3. Kinda makes me want to break out my 1979 vinyl of Stevie Wonder’s Journey Through the Secret Life of Plants and pop it on the turntable. (Here’s the title track, for those of you who don’t remember.)

    Stevie had a run of great albums in the 70s, but this wasn’t one of ’em. But after all the pleasure he’d given me with sides like Innervisions and Songs in the Key of Life and Hotter than July and the others, I figured I owed the cat the benefit of the doubt and picked up a copy on faith alone.

  4. It makes sense that plants can talk. After all, trees can hear the return of birdsong in the spring telling them to let their leaves burst forth in response.

    1. I’m not sure if this is sarcastic or serious…But if you are serious, this seems very surprising to me. For one thing, I think leaf sprouting often pre-dates bird arrival. The timing of plant season behaviors is usually usually determined by day length and temperature and water. Do you have a reference?

      1. Sorry, it’s a Canadian cultural reference. I was repeating what some ways of knowing claim to believe as an alternative to settler science. I should have been clearer that I don’t believe it myself.

        1. Emojis are your friend. An 😉 emoji would have cleared any confusion. Oh, the subtleties of emojis. Wished they didn’t exist…

  5. The paper omits mention of individuals who were assigned to the human species at birth, but believe that they were born in the wrong body and are really plants. Had the study been conducted in the US, individuals of this sort, such as for example a trans-Blue Peter Rhododendron, would have been consulted to translate the sounds.

  6. Well that is pretty cool, but it could just be a happenstance thing from air being burped up into vascular tissue. One can think of More Things To Do, like cutting the stem of a plant while it’s below water. This would minimize entry of air, and this might make the plants scream much more silently.

  7. Don’t let the vegans know!

    Joking aside, this is remarkable and very intriguing. It opens up a whole host of other questions…do trees make sounds when stressed? Perhaps just, “I am Groot.” 😉

  8. This is kinda’ interesting, but maybe not all that interesting. Cutting, drying, heating, manipulating, and chewing off the leaves of plants will have some effect on the plant due to cavitation (possibly), breaking of cell walls, cracking of xylem and phloem, cell walls shrinking, gas exchange disruptions, etc. A sufficiently sensitive transducer will be able to detect these changes, and our modern transducers are extraordinarily sensitive. So, I’d like to know if the observed sounds are anything more than just the sounds of a plant being manipulated, with each form of manipulation having its own tiny acoustic signature. The word “stress” has a certain connotation. I wonder if more is being read into this than is actually there.

    1. Yes, I agree. I’m not sure if this result is of any biological significance rather than being, as I said, an epiphenomenon. But many people wanted me to write about it and the press is covering it like crazy, so I wrote about it. It shouldn’t be completely ignored, though, as nature sometimes can fool us.

  9. That’s it ! I’m joining the Vegetable Liberation Front. We’ll break into
    supermarkets after dark, rescue the veggies and replant them !
    Save the vegetables !!

Leave a Reply