The Trap That Evolution Tried Exactly Once

A Wasp Preserved Mid-Grip

Most fossils in amber are snapshots of an accident. A gnat wanders onto sap, a beetle gets its leg stuck, and 99 million years later somebody in a market in Myanmar sells the resulting yellow lump to a lab. What makes Sirenobethylus charybdis different is not the accident. It is that the accident happened sixteen separate times to sixteen individuals of the same tiny, previously unknown species, and each time it caught a slightly different moment at the tip of the same strange structure.

The wasp itself is unremarkable at a glance: a small, winged, female hymenopteran with an antenna carrying nine flagellomeres and a forewing bearing eight closed cells, traits that place it comfortably in Aculeata, the stinging wasps, and specifically near the base of Chrysidoidea — the superfamily that today includes cuckoo wasps, bethylids and a handful of odder relatives [1]. Where it stops being unremarkable is the last few millimetres of its abdomen, where three overlapping flaps sit fringed with unusually long hairs, in a shape the describing team compares, deliberately and a little uneasily, to a carnivorous plant.

They named the genus for a siren — the mythological kind, and also a nod to Sirenia, the manatees, because in ventral view the tapering abdomen looks like a manatee's tail. The species epithet, charybdis, borrows the sea monster who swallowed and disgorged the ocean three times a day [1]. Naming a wasp after something that swallows and lets go is, it turns out, the whole argument in miniature.

What A Parasitoid Wasp Actually Needs A Hand For

It helps to be specific about what problem a structure like this would even be solving, because "the wasp grabs its host" undersells how hard grabbing is if you are an insect the size of a rice grain and your host does not want to be grabbed.

A parasitoid is not a predator in the usual sense. It does not eat its host outright; it lays an egg on or in a living host, and the larva that hatches eats the host from within or from without, over days or weeks, while the host is often still moving. That creates a mechanical problem predators mostly do not have: the female needs the host to hold still for the seconds or minutes it takes to sting, position an egg and withdraw — not to hold still forever, just for that window. Paralyse it too permanently and, in a lineage where the larva depends on the host staying alive and functioning while it feeds, you have killed your own offspring's food supply.

Biologists split parasitoid strategies along exactly that line. An idiobiont stings a host into permanent paralysis or kills it outright, then the larva develops on a host that will never move again. A koinobiont does something gentler: a temporary sting, sometimes no paralysis at all, and the host goes on living, growing, even moulting, while the parasitoid larva develops inside or on it, in sync [27]. Koinobiosis needs a host that survives the initial encounter. Which means the female needs a way to control a living, resisting animal for just long enough — not a way to end the fight, but a way to pause it.

That is the specific engineering problem an abdominal grasping apparatus would be solving, and it is a problem insects have solved with almost every part of the body except the abdomen. Mantises and mantidflies do it with spined raptorial forelegs. Dragonfly nymphs do it with an extensible, hinged labial "mask" that shoots forward and grabs prey by the head before retracting. Adult dragonflies weave their spiny legs into a basket to snatch insects from the air. Female dryinid wasps — a living chrysidoid family, close cousins of the fossil in question — have one foreleg apex modified into a chelate, pincer-like tarsus that clamps a hopper nymph still during oviposition. Trap-jaw ants spring hair-triggered mandibles. Heelwalkers use raptorial fore- and midlegs. Hangingflies grab prey with enlarged hind tarsal claws [7][8][9][10][11][12]. Every one of those is a leg or a mouthpart doing the grabbing. Sirenobethylus put the trap on the wrong end of the body entirely — the abdomen tip, as far from the head as an insect's anatomy allows.

Reading A Machine You Cannot Watch Move

Here is the actual difficulty facing Wu, Vilhelmsen and colleagues, and it is worth stating plainly because it is the difficulty behind essentially all functional inference in paleontology. You have a fossil. It does not move. You want to know what a hinged structure on it did while the animal was alive, and your only evidence is geometry frozen at the instant of death.

The trick the team used is one paleontology reaches for constantly: turn a single frozen structure into a series by finding it preserved in different states across different individuals, then read the series as if it were stop-motion. No single amber piece shows the trap moving. But amber piece CNU‑HYM‑MA2015124, the holotype, happens to show the lower flap held close against the middle and upper flaps — call it closed. Others in the set of sixteen show that same flap depressed well away from the rest — open [1]. Line those up and you are not looking at sixteen static objects any more. You are looking at sixteen frames of the same mechanism sampled at random points in its cycle, purely because death is not a coordinated event and each wasp got stuck at whatever position it happened to be in.

That is comparative anatomy's oldest trick applied to time instead of to species: instead of comparing one structure across many living relatives to infer its ancestral state, you compare one structure across many dead individuals of the same species to infer its range of motion. It works for the same reason both versions of the method work — variation that looks like noise, once you have enough of it, resolves into a shape.

The team backed the comparison with micro-CT scanning of one especially well-preserved paratype, reconstructing the internal geometry in three dimensions rather than relying on the flattened, sometimes distorted view amber gives from the outside [1]. That scan is what let them find the actual pivot — the point where the plate called sternum 6 is overlapped, on its underside, by the sclerite in front of it, sternum 5. A joint like that, found by CT rather than assumed, is what turns "these flaps sit in different positions in different specimens" into "these flaps rotate around this specific axis," which is a much stronger and much more falsifiable claim.

The Hinge, Argued From Two Sclerites That Are Suspiciously Large

Knowing where something pivots tells you it can move. It does not tell you what moved it, and a fossil has no muscles left to inspect — soft tissue essentially never survives amberisation. So the authors reasoned backward from what does survive: the hard exoskeletal plates the muscles would have had to anchor to.

Insect muscles that operate a hinge need two things from the skeleton around them: enough surface area to attach a muscle bulky enough to move the part fast, and a rigid enough sclerite to pull against so the force actually gets transmitted rather than just flexing the body wall. Look at the third tergum and third sternum of Sirenobethylus's abdomen and both happen to be the largest sclerites in the whole abdomen, and both happen to carry a transverse line across their anterior portion — the kind of internal ridge that, in living insects, often marks exactly this sort of muscle attachment site [1]. From there the authors propose a specific, testable arrangement: adductor muscles running from tergum 3 to the lower flap, contracting to pull it up and closed; abductor muscles from sternum 3, contracting to push it back down and open. Closing fast enough to catch a moving host, they note, would require those muscles to be substantial — which is one more reason two unusually oversized sclerites, sitting three whole segments forward of the hinge itself, make anatomical sense as their origin.

None of that is directly observed. All of it is observed structure plus mechanical inference, which is exactly the same category of reasoning a biomechanist uses on a living animal's skeleton, just run in a direction where you cannot go check the answer by watching a video.

Ruling Out The Obvious Alternative

Good functional inference in paleontology is not just building a plausible story for the favoured explanation. It is naming the other plausible stories and explaining, specifically, why they lose. The authors name exactly one serious competitor to "grasps the host," and it is worth walking through why it fails, because the reasoning is more interesting than the conclusion.

Could the apparatus have restrained a mate instead of a host? Female insects controlling males during copulation is not unheard of biologically, but it would be an unusual reversal of the far more common pattern of males restraining females, and — more decisively — no male Sirenobethylus has ever been found. All sixteen known specimens are winged, macropterous females [1]. Kachin amber has produced plenty of male wasps from other Chrysidoidea lineages, but none can confidently be matched to this genus, because its wing venation and its isolated position on the family tree make an association pure guesswork. You cannot test a mating hypothesis against a mate that has never turned up.

That last detail is worth lingering on, because it is a nice example of a structure arguing against its own most dramatic interpretation. If the trap were built to kill, you would expect stiff spines, hard edges, something built to puncture. Instead the setae on the inner face of the lower flap run long and sparse in the middle and short and dense along the border, on a membranous, apparently soft flap surface — a texture the authors describe as more plausibly a cushion than a weapon [1]. A structure this dramatic-looking, on closer inspection, was probably built to be gentle. That is not the twist you expect from something nicknamed a flytrap.

What "Reminiscent Of A Venus Flytrap" Is Not Claiming

It is worth being precise about the comparison the paper itself makes, because the headline is doing more work than the biology behind it. Dionaea muscipula, the actual Venus flytrap, closes two opposing lobes around an insect using rapid changes in cell turgor pressure — a hydraulic, entirely muscle-free mechanism unique to a handful of carnivorous plant lineages [16]. Sirenobethylus's apparatus is a hinged skeletal joint closed by striated muscle, the same basic machinery every insect leg uses to bend. The resemblance the authors point to is not mechanistic. It is the shape: a rounded, fringed structure that closes over something, evoking the plant's toothed lobes by outline rather than by physics. The name is a hook, not a homology claim, and the paper is careful never to claim otherwise.

A Family Of One, Near The Root Of A Bigger Family

Placing a fossil is its own kind of detective work, done with a phylogenetic analysis rather than a magnifying glass. The team scored Sirenobethylus against a matrix of morphological characters shared with living and fossil Aculeata and ran a Bayesian analysis under a model built for exactly this kind of discrete-character data, four independent runs converging after 20 million generations of sampling [1]. The result: Chrysidoidea comes out monophyletic, as it should, and Sirenobethylus sits in its own new family, Sirenobethylidae, positioned as a separate lineage close to the base of that superfamily — an early branch, not a late one.

Living Chrysidoidea is not shy about parasitoid diversity. Scolebythidae attack wood-boring beetle larvae. Bethylidae parasitise beetle or moth larvae. Within Chrysididae alone, the subfamily Cleptinae goes after sawfly cocoons, Amiseginae and Loboscelidinae lay eggs in the eggs of stick insects, and Chrysidinae are kleptoparasites that sneak eggs into the nests of solitary bees and wasps. Sclerogibbidae parasitise the nymphs of webspinners. Dryinidae and Embolemidae parasitise the nymphs of true hoppers [27]. Seven documented host relationships, seven different tactics, all within one superfamily — and not one of the living branches carries anything resembling an abdominal grasping apparatus.

A Precedent For Structures With No Descendants

The paper does not claim this is the first time Kachin amber has produced a capture mechanism with no living echo. It explicitly places Sirenobethylus in a small company: the haidomyrmecine ants, nicknamed "hell ants," whose scythe-like mandibles worked upward and forward against a horn on the front of the head — a mandible motion found in no ant alive today, in a subfamily that appears to have left no living descendants carrying the trait [1][15]. The authors call their new find "an even more bizarre" addition to that short list [1].

Put those two together and a pattern starts to look less like coincidence and more like something Kachin amber is unusually good at showing us: the mid-Cretaceous had insect lineages actively experimenting with body plans that simply are not represented in the modern fauna, not because they were bad ideas caught mid-failure, but because the fossil record is the only place they survive at all.

What "Abandoned By Evolution" Does And Does Not Mean

It is tempting to hear "no living descendants use this structure" and picture natural selection trying it, weighing it, and voting it off. That is almost certainly not what happened, and the distinction matters.

A trait vanishing from the fossil record is compatible with several very different stories, and the evidence in this paper cannot distinguish between them. It could have been outcompeted by a better solution to the same problem — except no better abdominal solution exists among Chrysidoidea today; the whole lineage of abdominal grasping simply is not represented, better or worse. It could have been tied to a specific host lineage that itself went extinct, taking a hyper-specialised predator down with it, the way many parasites vanish when their one host does. It could have been lost to nothing more dramatic than a small founding population, a genetic bottleneck, a run of bad luck completely unconnected to whether the trap worked well. Evolution has no foresight and no committee; a perfectly functional structure disappears constantly, for reasons that have nothing to do with its function, simply because the population carrying it did not make it through some unrelated narrow passage.

What the fossil record can tell you, honestly, is only the outcome: this structure is documented in exactly one deposit, in exactly one small clade, and it is not documented anywhere since. Explaining why is a separate, much harder question, and the paper does not attempt it. Neither should we pretend to answer it here. What we can do is size the gap.

Sizing The Gap, In The Club's Own Arithmetic

None of the numbers below are in the paper. They are ours, built from two real, independently sourced dates: the amber's own radiometric age and the well-established age of the Cretaceous–Paleogene boundary, the asteroid impact that ended the age of non-avian dinosaurs.

Framed carefully: absence of evidence across roughly 99 million years of subsequent fossil deposits and the entire modern insect fauna is strong circumstantial evidence for "this trait line ended," not proof of it. A structure this specialised, in a lineage this narrow, is exactly the kind of thing the fossil record is worst at capturing more than once. We are confident in the dates. We are not claiming certainty about the extinction itself — only that the documented gap is enormous, and that "no known descendants" is the most honest way the original authors, and we, can currently put it.

The Limits Of What One Amber Deposit Can Tell You

It is worth listing plainly what this study does not and cannot establish, because a fossil this striking invites more certainty than the evidence supports, and the authors themselves are careful about the line.

No host has ever been found attached to or near a Sirenobethylus specimen. The "homopteran hoppers, or small winged insects like flies" suggestion is an informed guess built from what related living chrysidoids attack and from the apparatus's evident size and reach, not a direct observation [1]. No male has been found, so nothing about sexual dimorphism, mating behaviour, or population structure can be said with any confidence. And a functional inference built from geometry and a proposed muscle arrangement is a hypothesis with a clear, stated confidence level — "probably," "we suggest," "the most likely function" — not a settled fact, however well the geometry lines up [1].

What would raise that confidence is exactly the kind of evidence paleontology usually cannot get and occasionally, spectacularly, does: a specimen caught with a host actually gripped in the apparatus. It has not happened for this species. It has happened for other predatory relationships preserved in amber, which is part of why paleontologists keep looking, one purchased lump at a time.

Why A One-Off Experiment Is Worth This Much Attention

It would be easy to file this as a curiosity: strange amber wasp, weird trap, moving on. That undersells what a genuinely isolated trait is good for scientifically.

A structure that convergent evolution reinvents constantly — eyes, wings, or the raptorial grasping leg, which by our own count of the examples the paper cites turns up independently at least five separate times across mantises, dragonflies, dryinid wasps, heelwalkers and hangingflies — tells you that structure solves a common, recurring problem well enough that selection keeps finding it. A structure built exactly once, in one small lineage, that then vanishes without leaving descendants who kept using it, tells you something different and, in its way, just as informative: that a viable solution to a real problem existed, was not obviously worse than the leg-based solutions every other lineage kept, and still did not spread or persist. Evolution is not a search for the single best answer. It is millions of separate, contingent experiments, most of which end for reasons that have nothing to do with the quality of the idea. Sirenobethylus is a rare direct look at one of those experiments, preserved at the exact moment before whatever ended it, with nothing to say about which ending it was.

That is the part of "abandoned by evolution" that is worth sitting with, more than the flytrap name or the manatee-tail etymology or any of the amber deposit's other charms. Somewhere in the mid-Cretaceous, an entire, functional, three-part hinged grasping mechanism existed, at the wrong end of a wasp, for reasons its owner did not choose and its descendants did not get to keep testing. It worked well enough to be built by muscle and bone across sixteen individuals we happen to have found. And then, as far as 99 million years of subsequent record can tell us, it simply stopped.

References

1. 1.

   Wu, Q., Vilhelmsen, L., Li, X., Zhuo, D., Ren, D. & Gao, T. (2025). A Cretaceous fly trap? Remarkable abdominal modification in a fossil wasp. BMC Biology 23, 81. doi:10.1186/s12915-025-02190-2

2. 2.

   van de Kamp, T., Mikó, I., Staniczek, A. H., Eggs, B., Bajerlein, D., Faragó, T., et al. (2022). Evolution of flexible biting in hyperdiverse parasitoid wasps. Proceedings of the Royal Society B 289, 20212086.

3. 3.

   Corbet, P. S. (1999). Dragonflies: Behaviour and Ecology of Odonata. Cornell University Press.

4. 4.

   Virla, E. G., Moya-Raygoza, G. & Guglielmino, A. (2023). A review of the biology of the pincer wasps (Hymenoptera: Dryinidae). Austral Entomology 62, 274–299. doi:10.1111/aen.12654

5. 5.

   Gronenberg, W., Tautz, J. & Hölldobler, B. (1993). Fast trap jaws and giant neurons in the ant Odontomachus. Science 262, 561–563. doi:10.1126/science.262.5133.561

6. 6.

   Zompro, O., Adis, J. & Weitschat, W. (2002). A review of the order Mantophasmatodea (Insecta). Zoologischer Anzeiger 241, 269–279.

7. 7.

   Setty, L. R. (1940). Biology and morphology of some North American Bittacidae (Order Mecoptera). American Midland Naturalist 23, 257–353.

8. 8.

   Barden, P., Perrichot, V. & Wang, B. (2020). Specialized predation drives aberrant morphological integration and diversity in the earliest ants. Current Biology 30, 3818–3824. doi:10.1016/j.cub.2020.07.055

9. 9.

   Durak, G. M., Speck, T. & Poppinga, S. (2022). Shapeshifting in the Venus flytrap (Dionaea muscipula): morphological and biomechanical adaptations and the potential costs of a failed hunting cycle. Frontiers in Plant Science 13, 970320. doi:10.3389/fpls.2022.970320

10. 10.

    Cruickshank, R. D. & Ko, K. (2003). Geology of an amber locality in the Hukawng Valley, northern Myanmar. Journal of Asian Earth Sciences 21, 441–455. doi:10.1016/S1367-9120(02)00044-5

11. 11.

    Shi, G. H., Grimaldi, D. A., Harlow, G. E., Wang, J., Wang, J., Yang, M. C., et al. (2012). Age constraint on Burmese amber based on U–Pb dating of zircons. Cretaceous Research 37, 155–163. doi:10.1016/j.cretres.2012.03.014

12. 12.

    Renne, P. R., Deino, A. L., Hilgen, F. J., Kuiper, K. F., Mark, D. F., Mitchell, W. S., et al. (2013). Time scales of critical events around the Cretaceous–Paleogene boundary. Science 339, 684–687. doi:10.1126/science.1230492

13. 13.

    Finnamore, A. T. & Brothers, D. J. (1993). Superfamily Chrysidoidea. In: Goulet, H. & Huber, J. T. (eds.), Hymenoptera of the World: an Identification Guide to Families, pp. 130–160. Agriculture Canada, Ottawa.