The New Madrid Earthquakes

Summary

Between December 1811 and February 1812 three great earthquakes and thousands of aftershocks struck the Mississippi valley around New Madrid, Missouri, far from any plate boundary. How large they were has been revised more than once, and why they happen there is still not known.

The events themselves are well recorded for a thinly settled frontier: the shaking was felt from New Orleans to New England, and the ground liquefied, slumped, rose and sank. Geologists have since found that the zone produced similar sequences around AD 900 and AD 1450. Estimates of the magnitudes have ranged from about 7.0 to above 8. On the cause, the United States Geological Survey is plain: "No one knows what causes New Madrid earthquakes."

What is documented

The principal shocks. The first struck on 16 December 1811 at about 2:15 in the morning, local time, and a large aftershock followed near dawn the same day (Hough, 2004). The second came on 23 January 1812, at about 9 in the morning according to the USGS and around 8 according to Hough (2004). The third came on 7 February 1812 at about 3:45 in the morning. The USGS records six aftershocks of magnitude 5.5 to 6.3 in the first two days, and hundreds still felt in 1813. Britannica gives 1,874 aftershocks strong enough to be felt in Louisville, Kentucky, about 300 kilometres away.

The effects. According to the USGS, the earthquakes destroyed several settlements along the Mississippi, caused minor structural damage as far away as Cincinnati and St. Louis, and were felt as far away as Hartford, Connecticut, Charleston, South Carolina, and New Orleans. Hough (2004) adds Canada and places along the Atlantic coast. The riverbanks failed, the Chickasaw Bluffs in Kentucky and Tennessee slid, and large tracts of the floodplain were raised or dropped. Buried sediment liquefied over a wide area, splitting the ground and venting water and sediment violently at the surface. Accounts from boats on the Mississippi describe waterfalls forming on the river (Hough, 2004), and Reelfoot Lake in Tennessee formed (Williams and colleagues, 2011). The region was sparsely settled: the same USGS poster notes that St. Louis had only 5,700 inhabitants in 1811.

Earlier sequences. Paleoseismic studies show that the zone produced comparable sequences around AD 900, give or take 100 years, and AD 1450, give or take 150 years (Tuttle and colleagues, 2002). They put the interval between sequences at 200 to 800 years, averaging about 500. The USGS states that the zone has repeatedly produced sequences of major earthquakes over the past 4,500 years.

Since 1812. Later shocks include magnitude 6.0 in 1843, 6.6 in 1895 and 5.0 in 1976 (Williams and colleagues, 2011). Seismometers installed from 1974 onwards have recorded thousands of small to moderate earthquakes, and until 2014 this was the most seismically active area east of the Rocky Mountains (USGS, 2025).

The magnitudes. There were no seismographs, so every estimate is derived from written accounts of the shaking. In 1992 Liu, Zoback and Segall described the three as earthquakes of moment magnitude greater than 8, and in 1996 Johnston and Schweig counted two near magnitude 8. In 2000 Susan Hough and colleagues re-examined the accounts, found that earlier studies had assigned high intensities to many reports that did not describe damage to match, and estimated 7.2 to 7.3, 7.0 and 7.4 to 7.5. Bakun and Hopper (2004) obtained 7.6, 7.5 and 7.8, which they placed between Johnston's figures and Hough's. The USGS now gives about 7.5, 7.3 and 7.5.

The ground motion today. In 1992 Liu, Zoback and Segall reported rapid strain accumulating since the 1950s, from GPS remeasurement of an old triangulation network. Later GPS networks found little or none. Newman and colleagues (1999) concluded that the interval between magnitude 8 earthquakes should well exceed 2,500 years and that the hazard may be overestimated. Calais and Stein (2009) found the zone deforming "too slowly, if at all" to account for its large earthquakes of the past 5,000 years. A USGS fact sheet the same year reported that the experts consulted did not find the GPS data a convincing reason to lower the hazard assessment.

The Browning prediction. Iben Browning proposed a 50 per cent probability that a tidally triggered earthquake of magnitude 6.5 to 7.5 would strike the New Madrid region on 2 to 3 December 1990, plus or minus two days. A working group convened for the National Earthquake Prediction Evaluation Council rejected it in a report released on 18 October 1990. The public response was large all the same. School closures caused up to 40,000 student absences in Missouri on 3 December, and more than 30 television and radio satellite trucks came to New Madrid. No earthquake as large as magnitude 6 occurred anywhere on Earth between 25 November and 10 December 1990 (Spence and colleagues, 1993).

Leading explanations

Old faults in an ancient rift, reactivated by present-day stress. The working hypothesis, not established. The USGS states that the seismicity is spatially associated with the Reelfoot rift and may be produced by movement on old faults in response to compressive stress related to plate motions. That does not explain why this rift, among the many ancient rifts under the central United States, is so active. Pollitz and Mooney (2014), imaging the crust and upper mantle with seismic waves, found the Reelfoot rift unique among them in sitting above a pronounced low-velocity zone in the mantle extending to a depth of at least 200 kilometres, and suggested that stress is concentrated there, leaving it more vulnerable to changes in stress than comparable rifts. Mueller, Hough and Bilham (2004) modelled three of the four largest shocks as ruptures on two connected faults, and proposed that the remaining one may have occurred more than 200 kilometres to the north, in the Wabash Valley.

Too little strain for repeated great earthquakes (Newman, Stein, Calais and colleagues). Contested. On this view the GPS results show the ground deforming too slowly, if at all, to build up the strain for a repeat of 1811 to 1812 every 500 years or so, and the hazard is lower than assessed. The USGS does not accept this. Its current position is that "there is broad agreement in the scientific community that a continuing concern exists for a major destructive earthquake" in the zone.

The size question. Open. How large the earthquakes were depends on how the accounts are read, and on how much the soft sediments along the rivers amplified the shaking at the places the accounts came from. Allowing for that amplification, Hough (2004) obtained 7.0 to 7.5. Other methods give higher values. The USGS describes the three as "between magnitude 7 and 8".

Current status

Partially explained. That the earthquakes happened, when, and what they did to the ground is established, and so is the fact that the zone has produced similar sequences before. The magnitudes are uncertain within the range from about 7 to about 8. The cause is not known: the USGS says so directly. Whether strain is still accumulating is disputed between GPS results that show almost none and a geological record that shows repeated large sequences.

The USGS estimates a 7 to 10 per cent chance of a repeat of the 1811 to 1812 earthquakes in the next 50 years, and a 25 to 40 per cent chance of an earthquake of magnitude 6.0 or larger. Those figures are on its programme page as current in 2025; the registry found no newer estimate of the chance of a repeat, although the USGS released a new National Seismic Hazard Model in January 2024. What would change this record is a physical explanation of why strain concentrates here, or GPS measurements long and precise enough to show whether it is accumulating.

Last reviewed: September 2026. Records are provisional. Where the evidence changes, the entry changes. Found an error? Tell us.