On the afternoon of March 18, 1925, people in a town in southern Illinois saw something strange on the horizon. It was not the neat funnel of a twister dropping from the sky. It looked more like an enormous bank of fog rolling along the ground. Some called it “boiling clouds on the ground”; others simply felt a black wall coming toward them. They had only minutes before it reached town. Before sunset that day, 695 people had died and more than 2,000 had been injured across three states: Missouri, Illinois, and Indiana. It was the highest death toll from a single tornado in U.S. history, and a century later that record still stands (Wikipedia).

Photo: unknown photographer · Public domain · Wikimedia Commons
The storm is often called “the worst tornado in the world,” but that is not quite right. The world record belongs to the 1989 Daulatpur–Saturia tornado in Bangladesh, which the World Meteorological Organization estimates killed about 1,300 people (Wikipedia). The Tri-State Tornado is, more precisely, the deadliest tornado in U.S. history and the second deadliest worldwide. So why did this one tornado take so many lives at once? Look one by one at the sky that day and the blind spots of that era, and the disaster turns out to be the result of several conditions locking together.
A Spring Day That Did Not Look Like Tornado Weather
That morning’s sky gave little hint of violent disaster. Snow flurries had fallen the night before, the high in Carbondale, near Murphysboro, reached only 69°F (21°C), and it was cloudy and rainy all morning (Wikipedia). The official forecast the U.S. Weather Bureau issued that day called for nothing more than “rains and strong shifting winds” (NWS Heritage). Nowhere was there a word of danger.
Yet while the ground lay calm, the atmosphere was quietly arranging itself into a dangerous pattern. At about 7 a.m. Central Standard Time on March 18, a surface low of roughly 1,003 hPa sat near northeastern Oklahoma and northwestern Arkansas, and during the day it moved northeast across southeastern Missouri, southern Illinois, and southwestern Indiana (NWS Paducah). Warm, moist air from the Gulf of Mexico surged north along the warm front that stretched east from the low. Temperatures that began in the 50s that morning reached the 60s over most of the eventual track by 1 p.m. (NWS Paducah). Just ahead of the so-called triple point, where the cold front, the warm front, and the dry line met, a huge supercell thunderstorm was born. A layer of warm air aloft, a capping inversion, held down other storms around it, so this one storm could grow alone, undisturbed (Wikipedia, citing the Maddox et al. 2013 reanalysis).

Map: U.S. Weather Bureau (NOAA Library collection) · Public domain · Wikimedia Commons
Three and a Half Hours, Three States, 219 Miles
The tornado touched down around 1 p.m. north-northwest of Ellington in southeastern Missouri and dissipated around 4:30 p.m. southwest of Petersburg, Indiana. For about three and a half hours it left damage without a break. The official path length is 219 miles (352 kilometers), and for 183 miles (295 kilometers) of it the tornado held an almost straight heading toward the northeast (N 69° E) (NWS Paducah).
One major factor in its deadliness was speed. According to the summary by the National Weather Service office in Paducah, Kentucky, the tornado moved at an average of about 62 mph (100 km/h), reaching 73 mph (117 km/h) between Gorham and Murphysboro. A meteorological reanalysis (Maddox et al., 2013) put the overall average at about 59 mph (95 km/h) (Wikipedia). Either way, it was a storm charging forward at around 60 mph. The damage path averaged about 0.75 miles (1.2 kilometers) wide and exceeded 1.5 miles (2.4 kilometers) south of McLeansboro in Hamilton County, Illinois (Wikipedia). There was no official rating system at the time, but judging by the destruction it left, it is widely accepted as the equivalent of an F5, the top of the Fujita scale (Wikipedia). It is worth remembering that the Fujita scale itself does not measure wind; it estimates wind speeds backward from the damage (NWS Norman).

Map: John W. Wilson and Stanley A. Changnon Jr., Illinois Tornadoes (Illinois State Water Survey, 1971), as published by the National Weather Service · Public domain · Wikimedia Commons
Still, the statement that “one tornado traveled 219 miles without a break” needs careful handling. Damage surveys in 1925 were far simpler than today’s. There were no aerial surveys and no radar, and surveyors often assumed that a tornado stayed on the ground the entire time (NWS Heritage). So in the early 2000s a team of researchers gathered newspapers, photographs, and survivors’ accounts from the time and redrew the entire path. Their reanalysis, published in 2013, extended the full length of continuous damage evidence to 235 miles (378 kilometers), but it also found 32 gaps of a mile or more within it. The team judged that the beginning and end of the path were probably separate tornadoes. It concluded that the 174 miles (280 kilometers) from central Madison County, Missouri, to Pike County, Indiana, was likely one tornado, and that the 151 miles (243 kilometers) from central Bollinger County, where eyewitness reports are densest, can be considered a continuous path (Johns et al. 2013). Even the most conservative 151 miles would be the second-longest tornado path on U.S. record, after the 2021 Western Kentucky tornado (165 miles) (Wikipedia).
541 Lives in 40 Minutes
The heaviest losses fell on southern Illinois. Around 2:30 p.m. the town of Gorham was demolished, and over the next 40 minutes the tornado swept through Murphysboro, De Soto, Hurst-Bush, and West Frankfort in turn, killing 541 people and seriously injuring 1,423 (NWS Paducah). In Murphysboro alone, a thriving coal and railroad city of 10,000, 234 people died, still the largest toll a single tornado has taken on one U.S. city. Fires broke out across town with people trapped beneath collapsed buildings, and with the water mains broken, there was no way to stop the flames (Wikipedia).

Photo: unknown photographer · Public domain · Wikimedia Commons
In the farming town of De Soto, northeast of Murphysboro, the second floor of the school collapsed and 33 students died. It remains the largest number of children killed at a single school by a tornado in the United States (NWS Paducah). That day, 57 schools across the three states were damaged or destroyed, and about one-third of all the victims were children (Wikipedia). In the mining town of West Frankfort, the death toll is given as anywhere from 102 to the 130s depending on the source, and the village of Parrish was destroyed so completely that it was never rebuilt (Wikipedia; NWS Paducah). After crossing into Indiana, the tornado wiped out the entire town of Griffin and wrecked much of the south side of Princeton before it finally faded (Wikipedia).

Photo: period postcard, scanned by Djohnsonokc · CC BY-SA 4.0 · Wikimedia Commons
The final count was 695 dead (12 in Missouri, 588 in Illinois, 95 in Indiana), 2,027 injured, and about 15,000 homes destroyed (Wikipedia; NWS Paducah). Numbers vary with how they were counted, however, and the true death toll was very likely higher (Wikipedia).

Photo: collection of Edeans · CC BY-SA 3.0 · Wikimedia Commons
An Era When No One Could Say “Tornado”
Today a supercell like this would show up on radar, sirens would sound, and warnings would reach people’s phones. In 1925, however, there was no such thing as a tornado watch or warning at all.
More surprising is that this was not simply a lack of technology. In 1887 the weather service of the U.S. Army Signal Corps barred the word “tornado” from forecasts and directed forecasters to write “severe local storms” instead. The Weather Bureau, created in 1890, kept the policy, arguing that tornado forecasts were unreliable and would only cause panic (Wikipedia). So in 1925, even if the Weather Bureau sensed danger, it was not permitted to call it by name (NWS Heritage).
Change came slowly. Only in 1938 was the word allowed in products sent to emergency management personnel. On March 25, 1948, Air Force weather officers Ernest Fawbush and Robert Miller at Tinker Air Force Base in Oklahoma issued the first official tornado forecast, and it verified that same evening (NWS Norman). The ban on tornado warnings for the general public was formally lifted by a circular letter from the chief of the Weather Bureau on July 12, 1950, and the Bureau’s first experimental public tornado forecast did not come until 1952 (Wikipedia). The year the Tri-State Tornado struck lacked not only the technology to read the sky but even the institutional language to warn of danger. People did not fail to take shelter because they were careless; they had no way of being warned.
The National Weather Service has worked out a hypothetical scenario of what would happen if the same storm came today. It concludes that a High Risk outlook would already have been issued that morning, and that around noon a “particularly dangerous situation” (PDS) tornado watch would have covered parts of Missouri, Illinois, Indiana, and Kentucky. Because the storm moved at 60 to 70 mph, it adds, warnings would have had to cover several counties ahead of the storm to give people time (NWS Paducah).
A Funnel Hidden in Rain
Without warnings, people would at least have had to spot the danger with their own eyes. Another reason this tornado was so deadly is that even this last line of defense failed. Along the Missouri stretch the storm was a fairly classic supercell, and the first witnesses saw a small, distinct funnel in the forested hills (Wikipedia). But as it moved into Illinois, the parent storm changed into a high-precipitation (HP) supercell that poured out heavy rain and hail, and the rotating funnel was often hidden behind curtains of rain, dust, and debris (Johns et al. 2013).

Drawing: unknown artist (1925) · Public domain · Wikimedia Commons
That is why many witnesses never saw a pointed twister. What they remembered was an “amorphous rolling fog” and “boiling clouds on the ground,” and even weather-wise farmers did not sense the danger until the storm was upon them (Wikipedia). Eugene Porter, who lived in Murphysboro, later put it this way: “Usually you think about a tornado, it has a funnel, and it may be a block or two or three blocks wide. But something about a mile wide…” (NWS Paducah) The 2013 reanalysis also recorded that witnesses saw a broad wedge tornado along most of the path (Johns et al. 2013). The common claim that “this tornado had no funnel at all” is not accurate. It is closer to the truth to say the funnel was there, but it was so wide and so wrapped in rain and dust that it did not look like a tornado. Add a forward speed of around 60 mph, and by the time people realized what was coming, there was almost no time left to escape.
How Is a Tornado Born?
A tornado is “a narrow, violently rotating column of air that extends from a thunderstorm to the ground” (NSSL). The most destructive tornadoes come from supercell thunderstorms, which contain a persistently rotating updraft called a mesocyclone. Their starting point is not the ground but the structure of the atmosphere itself.
The key is how the wind is layered. When the wind grows stronger or changes direction with height (wind shear), the air in between forms a rotating tube lying on its side, like a pencil rolled between two palms. When a growing thunderstorm’s strong updraft lifts this tube upright, a vertically spinning rotation appears inside the storm, and winds that turn with height strengthen one side of that rotation into a mesocyclone (NOAA JetStream). Rising air stretches and narrows, and beneath it a much thinner tornado reaches down toward the ground.

Illustration · AI-generated (Codex/ChatGPT subscription). Made to show the sequence wind shear → horizontal rotation → tilting by the updraft → mesocyclone → tornado, based on explanations from NOAA JetStream and NSSL.
Interestingly, science still cannot fully explain this last step. The National Severe Storms Laboratory (NSSL) describes a theory that temperature differences along the edge of downdraft air wrapping around the mesocyclone are involved, but it also notes modeling studies showing tornadoes can form without such temperature patterns, and says plainly, “The truth is that we don’t fully understand” (NSSL). Most of the time, this circulation of the atmosphere is part of the intricate order God made, carrying rain across the plains to water the wheat fields and bring spring. Yet the very same ingredients of that order, layered winds, updrafts, and temperature differences, condense into destructive rotation in a very rare arrangement. The same laws of the atmosphere that feed us can, at certain moments, become a force we cannot fully comprehend.
“It Was Not God Who Hurt Murphysboro’s Babies”
After the disaster, people asked how they should understand what had happened. Until then, it had been common in America to describe tornadoes as the wrath of God (Wikipedia). But faced with a disaster in which so many of the victims were children, local ministers urged people plainly not to see it as a judgment. At an ecumenical memorial service, one Illinois minister preached, “It was not God who hurt Murphysboro’s babies.” Historian Peter J. Thuesen, in his book Tornado God, sees the event as a turning point in how Americans thought about God, nature, and disaster (PBS American Experience).
The disaster of March 18, 1925, was a rare event in which many conditions came together in one place: a single giant supercell that grew undisturbed, a forward speed of around 60 mph, a funnel so wrapped in rain and dust that it did not look like a tornado, and an institution that would not even use the word that could have warned people. The tornado watches and warnings, sirens, and Doppler radar we now take for granted are the fruit of a long effort to keep disasters like this from happening again. The National Weather Service judges it highly unlikely that a storm like this would take nearly 700 lives today, but adds that even the best warning system does little good if people do not heed the call to take shelter (NWS Paducah). The number 695 quietly testifies to why today’s warning system exists, and why we should not hesitate when we hear a warning.
References
- 1925 Tri-State tornado — Wikipedia
- A Look Back at the Tri-State Tornado (1925) — NWS Paducah
- The Tri-State Tornado of 1925 — NWS/NOAA Heritage
- Johns et al. (2013), The 1925 Tri-State Tornado Damage Path and Associated Storm System — E-Journal of Severe Storms Meteorology
- Tornado warning — Wikipedia
- The First Official Tornado Forecast (25 March 1948) — NWS Norman
- The Fujita and Enhanced Fujita Scales — NWS Norman
- Severe Weather 101: Tornadoes — NOAA NSSL
- Thunderstorm Hazards: Tornadoes — NOAA JetStream
- The Thrill of the Chase — PBS American Experience
- Daulatpur–Saturia tornado — Wikipedia