At about a quarter to eight on the foggy morning of October 23, 1947, a fifty-two-year-old fisheries biologist named Alexander Bajkov was eating eggs in a restaurant in Marksville, Louisiana, when the waitress came over and told him that fish were falling from the sky outside. Bajkov happened to be one of the country’s leading authorities on freshwater fish, in Louisiana on official business for the state wildlife department. He abandoned his breakfast and went out to look. Fish were lying on the road, in yards, on the roofs of houses, and on the tops of parked cars, across a strip of town roughly three hundred metres long and twenty-five wide. Outside the bank on Main Street they were dense enough to average one per square yard, and traffic was rolling over them. He collected specimens, preserved them in formalin, and later published a short note in Science asking whether fish fall from the sky, having satisfied himself that on at least one occasion they demonstrably had.
Animal rain has one of the longest and best-attested pedigrees of anything filed under the inexplicable. Fish fell on Marksville. Hundreds of spangled perch fell on Lajamanu in the Australian desert, hundreds of kilometres from any sizeable water. Thousands of frogs came down on a Serbian town in 2005. Snakes reportedly fell on Memphis in 1877, rats on Algeria in 1902, sardines and mackerel on a Japanese city, fish on Telangana in 2022. The reports are ancient, global, and consistent, which is unusual in this field, and they share a specific feature that makes them feel genuinely impossible rather than merely odd. It is not that animals came down. It is that they came down sorted: all one species, all a similar size, with no accompanying mud, weed, pond water, or anything else the sky should have brought along. That sorting is the objection everyone raises, and it is the reason this belongs less among the things a culture talks itself into or the places that never existed than among the phenomena where the strangest detail turns out to be the answer.
What Animal Rain Actually Looks Like
Strip the folklore away and the reports converge on a recognisable profile. The animals are small and light: minnows, sunfish, perch a few inches long, juvenile frogs and toads, occasionally worms or spiders. They arrive over a limited area, frequently an elongated strip rather than a circle, which is the shape produced by something moving overhead rather than by a point source. They are usually intact and often alive on landing, or at least fresh, which puts real constraints on how long they can have been aloft and how violently they can have been handled. And the fall is brief, generally accompanying or immediately following a storm.
The age profile is nearly as telling. Falls overwhelmingly involve juveniles and small adults rather than large specimens, which is what a mass filter would produce: an updraft strong enough to lift a two-inch minnow may be nowhere near strong enough to lift a mature bass of several pounds, so the animals that go up are pre-selected by weight before any sorting downstream even begins. The same holds for amphibians, where reports concentrate on newly metamorphosed froglets rather than adults. Animal rain is therefore biased at both ends, once at pickup by what the wind can move and again at delivery by what the air carries furthest, which compounds the impression of a deliberately assorted consignment.
The species composition does the most work. Bajkov’s haul was largemouth black bass, goggle-eye, two kinds of sunfish, minnows, and hickory shad, all ordinary local freshwater fish, the largest a nine-inch bass. That is a local pond emptied onto a town, not an exotic delivery from elsewhere. Where reports describe animals unknown to the surrounding country, that is a genuine anomaly demanding a different explanation, and it happens far less often than the legends suggest. Most animal rain, documented properly, is the local fauna arriving from above, which is exactly what a transport mechanism operating over a short range should produce and exactly not what anything more exotic would produce. That makes these events an unplanned sample of regional ecology, informative in the way any accidental census is, comparable to what can be learned from the study of what animals do and where they live, while remaining permanently attractive to the same appetite for the extraordinary that surrounds reports of things seen overhead. The fish are local. That is the first clue.
The Sorting Problem
Now the objection, which is a good one and deserves stating at full strength. The standard explanation for animal rain is that a waterspout or tornado passed over water, lifted its contents, and dropped them somewhere downwind. But a violent rotating column of air is a blender, not a sieve. If it emptied a pond it should have taken everything in the pond: water, silt, weed, sticks, snails, insect larvae, frogs and fish together, mixed and battered. The fall should arrive as a filthy slurry of assorted pond life.
It is worth noticing that the objection is genuinely scientific in form, which is why it has proved so durable. It takes the proposed mechanism seriously, derives a prediction from it, compares that prediction against the observed record, and reports a mismatch. That is exactly the right way to test an explanation, and anyone raising it is doing better epistemics than the average commentator on animal rain. The trouble is that the prediction was derived from an incomplete model of what storms do to the things they lift, and a good argument built on a wrong premise produces a confident wrong answer, which is a far more dangerous failure than simple credulity.
That is not what gets reported. What gets reported is a clean, graded delivery: fish only, or frogs only, within a narrow size range, without the mud and vegetation, often still alive. People notice this immediately, and it is the single most common reason for rejecting the meteorological explanation, including among serious writers on the subject. Charles Fort made a great deal of it. If you emptied a pond over a town, the town would be covered in pond; instead the town is covered in one species of fish looking as though they came off a market stall. The apparent selectivity feels purposive, in the way any highly ordered outcome does, which is precisely the intuition that makes deliberate-seeming patterns so persuasive throughout the natural world’s arts of misdirection, and the impression is sharpened by the fact that so many of the animals are alive on impact, which raises its own uncomfortable questions about what a fish actually experiences. The delivery looks curated. That is the whole puzzle.
The Man Who Was There
Before resolving it, Marksville deserves its own moment, because it removes the option of denying that falls happen at all. Bajkov was a trained ichthyologist who reached the scene within minutes, examined the material himself, identified every species, measured the affected area, recorded the density of fish per square yard, collected and preserved physical specimens, distributed them to colleagues, and published in a major journal. There is no chain-of-custody problem, no untrained-witness problem, no lost-sample problem, and no century of retelling between the event and the record.
It is worth appreciating how rare that combination is. The overwhelming majority of animal rain reports come from people with no particular reason to record the details a meteorologist would want, arrive filtered through a newspaper, and are written up days or weeks later once the physical material has been swept away, eaten, or rotted. The information that would settle the question, species, sizes, the shape of the affected area, the presence or absence of anything other than the animals, the condition of the specimens, is precisely the information that a startled bystander does not think to preserve. Marksville is famous among researchers not because it was the strangest fall but because it was the only well-instrumented one.
What he described is also diagnostic in ways he could not fully exploit at the time. The strip shape indicates deposition from something moving. The size range, two to nine inches, is narrow but not uniform, which matters more than it sounds. The fish were cold but not frozen, which puts a ceiling on the altitude they reached. They were fresh, which puts a ceiling on elapsed time. They landed on rooftops, which rules out anything walking. And they were all fish, with no frogs, no weed, and no mud reported. Every one of those observations is a measurement, and together they describe not a miracle but a transport event with recoverable parameters, which is the difference between an anecdote and a data point, the same difference separating a real detection from a compelling impression in fields from the training of biological detectors to the study of how nervous systems assemble a report of the world. A specialist stood in the fall with a jar. That is as good as this subject gets.
The Atmosphere Is a Sorting Machine
Here is the resolution, and it inverts the objection completely. A storm updraft is not a blender that delivers its contents in one lump. It is a sorting machine, and it sorts continuously, automatically, and with considerable precision, because everything lofted into moving air is subject to the same physics: terminal velocity. How fast an object falls, and therefore how long it stays aloft and how far the wind carries it before it lands, depends on its mass, its cross-sectional area, and its drag. Two objects entering a storm together will leave it at different times and land in different places if they differ in those properties, and almost everything differs in those properties.
The counterintuitive part is that sorting gets stronger the longer the transport lasts. Over a few metres, differences in fall speed barely matter and everything lands together as an undifferentiated mess, which is what people are unconsciously imagining when they picture a pond being dumped on a town. Over a few kilometres, small differences in terminal velocity compound into large differences in landing position, and the load spreads out along the track. So the very distance that makes animal rain sound implausible is what guarantees the cleanliness of the delivery: the further the cargo travels, the more thoroughly it is separated, and the purer the sample arriving at any single point on the ground.
Follow a lifted pond through the system and the separation is inevitable. The water goes first and never comes down as water at all; dispersed into droplets inside a cloud, it evaporates, mixes, and rejoins ordinary precipitation, which is why nobody reports pond water falling. Silt and sand are dense and small and drop out almost immediately, close to the source. Weed and vegetation have high drag and low mass and drift off on a completely different trajectory. Snails and stones fall early. Fish of a given size and shape share a characteristic terminal velocity and therefore fly a characteristic distance, and frogs, differently shaped and differently dense, fly a different one. By the time the load is deposited, kilometres downwind, it has been separated into bands by aerodynamic class, and anyone standing under one band experiences a delivery of one species at one approximate size and nothing else. The apparently impossible selectivity is not selection; it is what moving fluids always do to mixed particles, the same sorting that produces graded sediment beds and layered deposits, and that underlies the practical hydrology behind managing water as a system and the engineering of anything that moves material at scale, as catalogued in the history of large transport infrastructure. The sky did not pick the fish. It graded them.
What Tornadoes Do With Debris
This is not a hand-waving argument, because the sorting has been measured directly, in the field, at large scale. After the tornado outbreak of 27 April 2011 devastated Alabama and neighbouring states, residents began finding photographs, documents, and personal effects carried enormous distances, and posted them online trying to return them to their owners. Researchers at the University of Georgia turned that crowd-sourced material into a formal dataset, assembling records of 934 objects lofted by at least fifteen separate tornadoes and reconstructing takeoff and landing points using geographic information systems and trajectory modelling.
The dataset also happens to illustrate why this question was so hard to settle before. Tornado debris trajectories cannot be studied experimentally, since nobody can put a labelled pond into a storm and wait to see where it lands, and until recently they could not be studied observationally either, because the objects were scattered anonymously across several states with no way to match landing points to origins. What changed was not the physics but the ability to identify individual objects and their owners at scale, which turned an intractable problem into a straightforward one almost overnight and produced a sample large enough to answer questions that had been argued over for a century.
The results are the best empirical demonstration available of what a violent updraft does to mixed cargo. As the resulting analysis in the Bulletin of the American Meteorological Society reports, objects travelled as far as 353 kilometres, exceeding the previous record for a documented tornado debris trajectory, with one photograph carried roughly 220 miles across two states. Meanwhile a five-foot metal sign from the same outbreak turned up about eighty kilometres from where it started. A photograph and a metal sign entered the same storm system and were deposited hundreds of kilometres apart, purely because of mass and drag. That is aerodynamic sorting, quantified, with a sample approaching a thousand objects. It also explains why debris fields are banded rather than mixed, an insight with real consequences for emergency planning and for predicting where hazardous material will land, which is why this sort of dispersion modelling matters well beyond meteorology, in domains from the assessment of battlefield and industrial hazards to the propagation calculations behind directed-energy and radar systems. Same storm, same moment: eighty kilometres versus three hundred and fifty.
Waterspouts
The lifting half of the mechanism needs its own correction, because the popular image is wrong in a specific and revealing way. A waterspout is generally pictured as a straw, sucking a column of water up into the cloud. It is not. The visible funnel is condensation, water vapour made visible by the pressure drop inside the rotating column, and not a solid column of lifted lake. Waterspouts do not drink ponds dry.
What they do is generate violent rotating winds at and just above the surface, and those winds can move anything light enough, which certainly includes small fish near the surface and animals at the water’s edge. Meteorologists distinguish tornadic waterspouts, true tornadoes that form in severe storms and travel over water, from the weaker fair-weather variety that develops upward from the surface beneath building cumulus. The tornadic kind is far more capable of lofting animals, and it is worth noting that many documented falls happened well inland, associated with ordinary strong thunderstorm updrafts over ponds, marshes, and flooded fields rather than with anything dramatic over an ocean. An updraft does not need to be a tornado to lift a two-inch minnow a long way, and the mechanism has been proposed since at least the nineteenth century, including for a fall of frogs on Kansas City in 1873 and for frogs found frozen inside hailstones at Dubuque, Iowa, in 1882, a startling detail that also fixes a minimum altitude for the animals involved. The sensory sophistication of the animals is entirely irrelevant here, unlike in the magnetically guided migrations of birds; a fish in an updraft is cargo, subject to the same physics as any other object in a sky now also full of drones and machines. The funnel is not a straw. The wind does the work.
Lajamanu
The Australian case is the one that most tests the explanation, and it is worth taking seriously precisely because it looks hardest. Lajamanu is a small community in the Northern Territory, deep inland, several hundred kilometres from the coast and a long way from any substantial permanent water. On 25 and 26 February 2010, hundreds of small fish fell on the town, identified as spangled perch, and this was not the first time: comparable falls were reported there in earlier decades, making it a repeat location rather than a one-off.
Repeat locations are in fact a strong argument for the mundane explanation rather than against it. A genuinely inexplicable phenomenon has no reason to prefer one small town in the Northern Territory over the thousands of comparable settlements around it. A mechanism grounded in local hydrology, local species, local terrain, and prevailing storm tracks has every reason to keep producing the same result in the same place, because all of those inputs are fixed features of the landscape. Recurrence at a fixed site is the fingerprint of a stable natural process, and it is precisely what you would predict if the cause is geography rather than caprice.
The repetition is informative rather than mysterious. Spangled perch are a hardy native Australian freshwater species notorious for exploiting ephemeral water, colonising temporary pools and floodwaters across the arid interior with remarkable speed, which means the region is dotted with short-lived waterbodies holding exactly this fish at exactly the size that gets lofted. Add a monsoonal storm season, towering convective updrafts over a flat landscape, and a town that happens to sit downwind of that geography, and the recurrence stops being a puzzle and becomes a consequence of local ecology plus local meteorology. A specialist species and a particular landscape producing a repeatable outcome is the ordinary business of ecology, and it is exactly the sort of interaction missed by anyone reasoning about a place in isolation, a failure that has undone confident interventions in unfamiliar environments from the industrial plantation swallowed by the Amazon onward, while the fish themselves, surviving both flight and landing, join the ranks of animals whose toughness verges on the absurd, alongside the cases collected in accounts of remarkable animal endurance. It rains fish in Lajamanu because of what lives near Lajamanu.
The Fish That Came Up
Yoro, in north-central Honduras, is the most famous case in the world and is probably not animal rain at all, which makes it the most interesting entry in the file. The town has reported a rain of fish, the Lluvia de Peces, at least once a year since the 1860s, typically between May and July after torrential storms, and it comes with an origin story: a Catholic missionary, Father José Manuel Subirana, is said to have prayed for relief during a famine, whereupon fish began falling. There is an annual festival with parades and religious ceremonies, and the event is a genuine point of civic pride and a tourist draw.
The evidence points somewhere other than the sky. Nobody has ever photographed or filmed the fish falling, and the reason locals give is entirely reasonable, which is that no one goes outside during a storm of that severity; what exists in abundance is documentation of the aftermath. When a National Geographic team was in the area in the 1970s and encountered the phenomenon, they did not witness a fall either, but they did examine the fish, and found the animals were blind. That is close to decisive, because blindness of that kind indicates a population living permanently without light, which points to subterranean streams and flooded cave systems rather than any surface pond, and it fits the further finding that the species recovered do not always match those in local rivers. The likeliest account is therefore not precipitation but hydrology: torrential rain floods a karst system, subterranean fish are flushed to the surface, and they are stranded on the ground as the water retreats, to be found the next morning exactly where a fall would have left them. It is a genuinely remarkable natural event that acquired the wrong label, in a region whose history is thoroughly entangled with outside interpretation, as the record of foreign enterprise in Central America attests, and whose civic identity is now built around a founding miracle in a way familiar from communities organised around a shared providential story. The fish did not come down. They came up.
The Frogs Were Already There
Yoro points to a wider and much underrated category: the fall that never happened. In an enormous proportion of reported animal rain, particularly involving frogs and toads, nobody observed anything descending. What was observed was a great many animals present on the ground after heavy rain, where there had been none the previous evening, and the fall was inferred from their sudden presence.
This matters for how the historical record should be read, because it means the corpus of animal rain reports is almost certainly a mixture of at least three unrelated phenomena wearing one label. There are genuine aerodynamic falls, of the Marksville kind, where animals demonstrably arrived from above. There are groundwater events, of the Yoro kind, where animals arrived from below. And there is a large and probably dominant category of emergence events, where the animals never travelled at all and simply became visible. Lumping these together and then complaining that no single explanation covers every case is a guaranteed route to permanent mystery, since the cases do not share a cause.
The inference is usually wrong, because amphibians have a well-documented habit of doing exactly this without leaving the ground. Many frogs and toads spend dry periods buried and dormant, sometimes for months, and emerge in enormous synchronised numbers within hours of the first substantial rain, which is a survival strategy rather than a curiosity. Juvenile toads dispersing from a breeding pond can carpet a landscape overnight. Earthworms surface during heavy rain. So the sequence of dry ground, storm, and ground suddenly covered in small animals has a wholly terrestrial explanation requiring nothing to have flown, and telling the two apart demands either seeing the fall or finding animals somewhere nothing could have walked to, such as a rooftop, which is precisely the detail Bajkov recorded at Marksville and precisely the detail most reports lack. Distinguishing an event from an inference about an event is the entire discipline here, and it is the same gap between raw sensation and constructed interpretation that runs through the science of how vision is assembled. Rooftops are evidence. A wet lawn full of toads is not.
Animal Rain in 2026
The modern position is that animal rain is real, uncontroversial among meteorologists, and better documented every year, largely because everyone now carries a camera. Recent falls have been filmed rather than merely described, and the footage consistently shows what the physics predicts: small animals, one type, brief duration, in and around storms. The reference literature has become correspondingly matter-of-fact, and anyone wondering whether fish can fall from the sky can consult the Library of Congress account of the phenomenon, which reproduces the Marksville report and treats the question as settled.
There is also a quiet shift in how such reports are received. A fall of fish in 1876 or 1947 was a newspaper sensation and a theological talking point; a fall of fish today is a short clip, a local news item, and a meteorologist explaining updrafts, and the explanation now generally travels alongside the footage rather than arriving years later. That is a meaningful change in the life cycle of an anomaly, because the window in which a phenomenon can accumulate folklore has narrowed from decades to roughly a news cycle. Whether that makes people less credulous or merely faster is an open question, but animal rain is one of the few cases where the correction now keeps pace with the claim.
The instrumentation has moved further than the folklore. Dual-polarisation weather radar can now identify debris lofted by a tornado in real time, distinguishing the chaotic returns of tumbling non-meteorological objects from raindrops and hail, which means the transport of solid material by storms is routinely observed rather than inferred from what turns up afterwards. Crowd-sourced recovery of lofted objects, of the kind that produced the 2011 debris dataset, has become a standard research technique rather than an improvisation. And the general problem of tracking where storm-borne material goes has obvious application well past curiosity, at a moment when instrumented atmospheric sensing sits among the quietly useful entries in the catalogue of technical undertakings. The question is no longer whether fish fall. It is how far, from where, and sorted how.
Sorted
Strip animal rain down to its mechanism and the feature that made it seem impossible turns out to be the signature that proves the explanation. A fall of one species at one size, clean, without mud or weed or pond water, is not evidence against aerodynamic transport; it is precisely what aerodynamic transport must produce, because a storm that lifts a mixed load begins sorting it by terminal velocity immediately and delivers it not as a lump but as a series of bands strung along its track, each holding objects of a single aerodynamic class. Anyone under one band gets fish. Anyone under another gets the weed, or the silt, or nothing at all because the water evaporated on the way. Measure the debris field of a real tornado and you find a photograph three hundred and fifty kilometres out and a metal sign at eighty, which is the same process operating on office supplies instead of sunfish.
The test this yields is short enough to apply on the spot. Ask whether anyone actually saw the animals descend, or whether a fall was inferred from their presence. Ask whether any landed somewhere nothing could have walked or crawled to. Ask whether the material was one species at one size, which indicates transport, or a mixture, which indicates something else. And ask what lives in the water upwind. Four questions separate the genuine falls from the emergences and the groundwater events, and most reports fail at the first.
The rest of the category dissolves along adjacent lines. Yoro’s celebrated rain of fish is very probably a flooded cave system venting blind fish onto the surface, an event no less extraordinary for arriving from below. A great many frog falls are mass emergences that nobody saw fall because nothing fell. And the residue, the genuine, observed, specimen-collected falls like the one a fisheries biologist walked into outside a Louisiana diner in 1947, are ordinary storms doing an ordinary thing to a nearby pond, which is why these cases sit among the resolved entries in the catalogue of Fortean phenomena rather than the open ones. The strangest thing about a rain of fish was never that fish came down. It was that they arrived in order, and order is what moving air does to everything it carries.