🦗 A forest in what is now Inner Mongolia was already singing 165 million years ago, and twelve researchers have now worked out roughly what it sounded like. One number in their results lands oddly in Japan: five kilohertz. That is more or less the pitch Japanese people have been catching, caging and holding parties to listen to for the past 1,000 years.

What 20 fossil wings gave up

The paper went online in the Proceedings of the National Academy of Sciences (PNAS) on August 25, 2026. The team worked from 20 specimens of fossil ensiferans, the branch of insects that today includes crickets and katydids, covering nine species: seven in the family Prophalangopsidae, an old group that still has living members (the team measured those living wings), and two in Haglidae, which died out after the Cretaceous.

All 20 came from one place. The Jiulongshan Formation at Daohugou, in Ningcheng County, Nei Mongol (Inner Mongolia), China, is Middle Jurassic rock, about 165 million years old.

Insect song survives in stone in a way that almost nothing else does. A cricket's instrument is a row of hardened teeth on one wing and a scraper on the other, and as the study's authors point out, hardened arthropod cuticle fossilises well while the soft vocal cords of a vertebrate do not. The teeth are still there, and counting them is easy.

Turning a row of teeth into a pitch is the hard part. The researchers measured the resonance, damping and vibrating area of living relatives' wings with micro-scanning laser Doppler vibrometry, and filmed those wings in motion with high-speed video, to build a framework they could test. Fossil wing outlines were then vector-traced and run through finite-element simulations in COMSOL Multiphysics v6.1, with the membrane thickness swept across a plausible range of 5.5 to 18 micrometres because nobody can measure that directly on a flattened fossil. A phylogeny built from 147 living taxa and 1,077 single-copy orthologs placed each fossil on the family tree. An acoustic dataset of 95 species, all of them insects whose song and wing structure are both known, anchored the predictions. Finally a Gaussian process regression model, trained on four things (dominant frequency, syllable duration, total tooth number and the length of the file), filled in the rhythm.

Five kilohertz, and why that number matters in Japan

Five of the nine species turned out to sing low, pure tones between 5 and 7 kilohertz. Others sat above 10 kilohertz. One, Sigmaboilus peregrinus, came out at 20.42 kilohertz. That is ultrasound.

Japan's autumn insects sit in the same territory. The suzumushi bell cricket, the star of the whole tradition, calls at 4.5 kilohertz, and crickets in general run between 4,000 and 5,000 hertz. The kirigirisu katydid sits at 9,500 hertz, the kantan tree cricket down at about 2,000 hertz. The two sets are not the same kind of number, though: the Jurassic values are model predictions, the Japanese ones are measurements of living insects.

More than half of that Jurassic chorus was singing inside the exact band that Japanese poets, shoguns and Edo shopkeepers would later single out as a sound to go out and hear. Had you dropped a Heian courtier into that forest, a good half of what he heard would have been immediately, culturally legible to him. But at 20.42 kilohertz, Sigmaboilus peregrinus was singing at or past the ceiling of adult human hearing. Part of that Jurassic concert was, for a human ear, silence.

Thorin Jonsson of the University of Graz, one of the co-authors, summarised the finding for Smithsonian Magazine: "The world during the Jurassic Period was acoustically far richer and more diverse than previously thought."

Ultrasound 110 million years before the first bat

The standard reason insects go ultrasonic is bats: you shift your signal out of the predator's listening range, or you use ultrasound because your enemy already lives there. But bats appear in the Eocene, about 55 million years ago, and laryngeal echolocation about 50 million years ago. A Jurassic insect calling at 20 kilohertz beat them by roughly 110 million years.

So the paper reaches for a different predator. It argues that eavesdropping by insectivorous mammals with good hearing, including gliding species, was among the main selective pressures behind the diversity of Jurassic insect song, and that the arms race ran in both directions, shaping insect songs and mammal ears together.

The predator is not hypothetical. Volaticotherium antiquum, described in Nature in December 2006, is an insect-eating mammal with a gliding membrane, dated to about 164 million years ago, and it comes out of the same Daohugou beds. Predator and prey, same rock, same air.

The study also suggests the singers were stacked vertically: low-frequency callers signalling close to the ground, higher-frequency ones up in the understory or the branches. A Japanese field in September has the same structure, with different insects in the slots.

Japan has been listening to insects for a thousand years

Japan filed insect noise under music, and kept receipts.

Listening parties show up in The Tale of Genji in the 11th century, with insect-catching outings in Kyoto's Sagano district. By 1687 the fifth shogun, Tokugawa Tsunayoshi, had banned the buying and selling of insects outright, which tells you the trade was worth banning. In the late 18th century an Edo oden vendor named Chuzo started catching bell crickets, a customer named Kiriyama worked out how to breed them, and the whole thing turned into an industry with breeding, cage-making and retail. The Edo Mushi-ko guild capped the number of licensed vendors at 36 and handled around twelve species. By 1858 the guidebook Edo Kacho-reki was listing Ochanomizu and Sugamo as places to go and listen. In 1898 Lafcadio Hearn wrote "Insect Musicians" for Exotics and Retrospectives and carried the whole idea across to an English-reading audience.

Search this topic in English and you will quickly hit the claim, made by Tadanobu Tsunoda, that Japanese speakers uniquely process insect sound in the language-handling left hemisphere. It is repeated constantly in Japan and it is not established science. Peter Dale, writing in 1993, pointed out that Tsunoda brought in his own language categories without expertise in either neurology or linguistics, that those categories are unintelligible to a linguist, and that no corroborating work has followed. The thousand-year habit is real. The brain story attached to it is not.

No Japanese name on this paper, but Japan has a chapter of its own

There is no Japanese involvement in the Jurassic soundscape study. The 12 authors sit at eight institutions in China (Capital Normal University, Sichuan Agricultural University, Hebei University), the UK (Lincoln, Leicester, Bristol), Austria (Graz) and the United States (Arizona State).

Japan does, however, hold one of the field's landmark results. On February 15, 2023, Communications Biology published work by Junki Yoshida of the Hokkaido University Museum and the Fukushima Museum, Yoshitsugu Kobayashi of the Hokkaido University Museum, and Mark Norell of the American Museum of Natural History, describing specimen IGM100/3186 of the ankylosaur Pinacosaurus grangeri. It preserved an ossified larynx, a cricoid and arytenoid pair, the first ever found in a non-avian dinosaur, and large and mobile enough to suggest the animal could vocalise something closer to a bird than to a crocodile.

China and Europe reconstructed the insects; a Japan-led team reconstructed the throat of a dinosaur.

Sound is time, and that is where the ethics start

Reconstructed dinosaur colour, worked out from fossilised pigment structures, gives you a picture. A reconstructed song gives you a duration. A waveform only takes shape as it unfolds in real time, so these teams are not restoring an object; they are restoring an interval.

The heavier version of this, putting back an organism rather than a sound, also runs through Japan. On March 11, 2019, Scientific Reports carried work led from Kindai University in which cell nuclei taken from the muscle of Yuka, a mammoth that had spent 28,000 years in Siberian permafrost, were transferred into mouse oocytes. Some of the nuclei took on structures seen just before cell division. Nothing was born, but the university was clear about the direction of travel: better-preserved nuclei with less DNA damage would be expected to take embryonic development further.

Sound reconstruction creates no organism and carries none of that weight. It also produces something that, once you have heard it, is hard to un-hear.

Is insect sound music where you live, or is it just what summer sounds like? And if someone handed you a recording of a forest from 165 million years ago, would you press play?

References