On the afternoon of May 31, 1889, the South Fork Dam failed above Johnstown, Pennsylvania. The flood that followed killed 2,209 people, the count the National Park Service publishes. The dam belonged to the South Fork Fishing and Hunting Club. Its members had lowered the crest and fitted screens across the spillway, on a dam that had already lost its discharge pipes before the club bought it. Survivors sued and recovered nothing. The case matters to emergency managers because it shows how far apart a documented warning and legal responsibility can sit, and many private dams sit in that gap today.

A canal reservoir that outlived its canal: 1838 to 1879

The Commonwealth of Pennsylvania built the South Fork Dam to store water for the western division of the Main Line canal, the state’s system of canals and inclined railways between Philadelphia and Pittsburgh. The reservoir sat on the South Fork of the Little Conemaugh River, about fourteen miles upstream of Johnstown and several hundred feet higher. Construction began in the late 1830s and, after long interruptions, was finished in 1853. By then the Pennsylvania Railroad had a line across the Alleghenies. In 1857 the railroad bought the state’s Main Line works, reservoir included, and it had little use for a canal feeder.

As built, the dam was an earth and rock embankment. It is commonly described as about 72 feet high and a little over 900 feet long. Two features of the design matter for what came later. The first was a spillway cut through rock at one end, which carried flood water around the embankment rather than over it. The second was a set of cast iron discharge pipes running through a stone culvert at the base, with valves worked from a tower, so the reservoir could be drawn down. An earthen dam can pass a great deal of water through its spillway safely. Water flowing over the crest of the embankment itself erodes the downstream face, and once that erosion cuts through the crest, the breach widens quickly and releases the reservoir. The pipes let an owner lower the lake before a storm or for repairs, and the spillway kept the crest dry while that happened.

In 1862, with the lake well below full, a section of the dam near the culvert gave way. Little water was stored at the time, so the release did limited damage downstream, and the break was left largely unrepaired for years. In the 1870s the railroad sold the property to John Reilly, a former congressman. During his ownership the discharge pipes were removed and sold for scrap. In 1879 the dam and the surrounding land passed to Benjamin Ruff, who organized the club. The dam therefore reached its last owners with a known past failure and no way to drain it, before anyone from the club had done any work on it.

What the club changed: the crest, the screens and the 1862 breach

The South Fork Fishing and Hunting Club drew its members from Pittsburgh’s industrial and financial leadership. Henry Clay Frick, Andrew Carnegie, Andrew Mellon, Philander Knox and Robert Pitcairn of the Pennsylvania Railroad were among them. The club built a clubhouse and cottages along the shore and stocked the lake with game fish. The membership mattered later, because the most natural defendants had money and standing in Pittsburgh that the families of Johnstown mill workers lacked.

Ruff’s first task was to close the 1862 break and refill the lake. David McCullough’s 1968 history, The Johnstown Flood, remains the account most writers rely on. He describes the fill as rock and earth mixed with brush, hay and horse manure, dumped into the gap without the engineering a dam of that height needed. The club did not replace the discharge pipes. The lake could no longer be lowered on purpose after that, and water could leave only through the spillway or over the crest.

The club also lowered the crest of the dam so the top would be wide enough for a carriage road. Most accounts put the reduction at two to three feet. A bridge carried the road over the spillway. The club fitted iron screens across the spillway to keep the stocked fish from escaping downstream. Screens of that kind catch floating debris in high water and choke the opening they sit in.

Taken together, the changes cut the margin between the largest flood the dam could pass and the flood that would overtop it. A lower crest meant the spillway could carry less water before the embankment itself was overtopped. The bridge and the screens narrowed the opening that water had to pass through. The missing pipes had already taken away the one tool an owner has to lower a lake ahead of a storm. I have found nothing in the published record showing that an engineer reviewed the crest lowering or the screens before they were made. Most state dam safety programs now expect that kind of review for alterations to a dam of that size.

The 1880 inspection and the warnings Johnstown learned to discount

The Cambria Iron Company was Johnstown’s largest employer, and its works sat in the valley below the dam, so it had good reason to care about the dam’s condition. McCullough reports that Daniel J. Morrell, the company’s head, sent John Fulton, one of the company’s engineers, to inspect the dam in 1880. As McCullough summarizes Fulton’s report, it criticized the quality of the repair and the leakage through it, and it identified the missing discharge pipes as a serious defect.

Ruff answered that the dam was safe and that Fulton’s concerns were unfounded. McCullough also describes an offer from Morrell to contribute toward repairs, and the club never carried out the work Fulton’s findings called for. I have not seen the original letters, so I am paraphrasing rather than quoting either side. The inspection and the club’s answer are still the most important facts in the case for anyone thinking about liability. They show that the hazard was identified in writing nine years before the failure, by an engineer working for the people downstream.

The town’s own relationship with the dam is the other half of the record. Johnstown sat at the junction of the Little Conemaugh and Stonycreek rivers and flooded often. Talk that the dam would break someday circulated for years. McCullough describes spring high water in the years before 1889 bringing alarms about the dam that came to nothing. By 1889, a warning about the dam was something residents half expected and had learned to set aside. People in public safety communications would call this warning fatigue. The messages sent down the valley on May 31 arrived in a town that had heard the same message in earlier springs, when the dam had held.

May 31, 1889: the rain, the work on the crest, and the hour downstream

A storm brought heavy rain over the upper Conemaugh basin through the night of May 30 and into May 31. Published estimates of the total commonly run from six to ten inches in about a day, and the storm is regularly described as the heaviest rainfall then on record for that part of Pennsylvania. Rain gauges in the mountains were few, so those figures are reconstructions, and anyone citing a single number should name its source. By the morning of May 31 the rivers in Johnstown were already out of their banks. Many residents had moved their belongings and families upstairs to wait out what they took to be an ordinary flood.

At the dam, the lake rose through the morning. The club’s young resident engineer, John Parke, worked with laborers and club employees under the club’s president, Elias Unger, who had taken over after Ruff’s death in 1887. They tried to cut an additional spillway through the hillside at the far end of the dam and to raise the low points on the crest with earth. Attempts to clear the spillway screens failed because the screens were jammed with debris. Parke rode to the village of South Fork to raise the alarm, and messages were passed to the railroad telegraph tower and down the line toward Johnstown. Accounts differ on which messages reached whom and how they were received, so I would not present any one version of the telegraph sequence as settled.

Water was flowing over the crest before the embankment gave way, shortly after three in the afternoon. The lake, which the Park Service estimates at about 20 million tons of water, emptied in well under an hour. The flood wave moved down the narrow valley through South Fork, Mineral Point, East Conemaugh and Woodvale, carrying houses, rail cars, trees and wire from a wire works at Woodvale. It reached Johnstown roughly an hour after the failure. In the city, wreckage piled against the Pennsylvania Railroad’s stone bridge below the junction of the rivers. The pile caught fire with people still trapped in it.

Roughly an hour separated the failure from the arrival of the water in the city. That is more warning time than many communities directly below modern dams would get. What that hour could have meant depended on warnings that had already been discounted for years. I will not put a number on the lives a credible warning might have saved, because that is a counterfactual nobody can test.

The toll: why 2,209 holds up and the rounded versions drift

The National Park Service manages the Johnstown Flood National Memorial at the site of the dam, and it publishes the death toll as 2,209. The Johnstown Area Heritage Association, which runs the Johnstown Flood Museum, uses the same figure. The Park Service also reports that 99 entire families died and that 396 of the dead were children. Johnstown’s count has held steady for a long time, and I know of no reputable source that seriously disputes it. That sets Johnstown apart from disasters where the dead were lost without records and the toll can only be published as a range.

The figure includes people who were never identified. In Grandview Cemetery in Johnstown, the Plot of the Unknown holds 777 unidentified victims under rows of identical markers. Remains continued to be recovered downstream long after the flood. Early tallies moved as bodies were found and matched to the missing.

The problem in popular retellings is rounding. Phrases such as “more than 2,000” or “over 2,200” are honest approximations of 2,209. Some retellings go further and repeat early press estimates from the first days, when reporters arriving in a destroyed city published figures well above the final count. A retelling that gives a number much higher than 2,209 without a source is usually repeating one of those early estimates or inflating a rounded figure. There is no reason to prefer it over the Park Service count.

The suits against the club and the defense that held

After the flood, survivors and the families of the dead sued the South Fork Fishing and Hunting Club. McCullough’s account, which most writers follow, reports that none of the suits produced any recovery. I have found no reputable source reporting a judgment or settlement paid by the club, or by any member, as a result of litigation. I am not naming individual cases or courts here, because the secondary sources do not give a consistent record at that level of detail, and a wrong case name would be worse than none. Anyone who needs the litigation record should go to the county court archives in western Pennsylvania and to the collections of the Johnstown Area Heritage Association.

The club’s position, as it comes through in the record, was that the flood was an act of God. On this argument the rainfall went beyond anything that could have been anticipated, and the dam would have failed under it whoever owned it and however it had been kept. The members said little in public. Several gave money to the relief effort, and Andrew Carnegie paid for a new library for the city, the building that now houses the Johnstown Flood Museum. Giving to relief is not an admission of fault in law, and nobody treated it as one.

The club’s legal structure also worked against the plaintiffs. The club was a chartered corporation, so a suit against it could reach only the corporation’s assets. After the flood those assets amounted to land around an empty lakebed, a clubhouse and some cottages. Reaching the fortunes of individual members would have required proving personal fault against each of them under the negligence rules of the day. The Pennsylvania courts of the period were not a promising forum for that kind of claim. In 1886 the state’s Supreme Court had ruled for a coal company in Pennsylvania Coal Co. v. Sanderson, a well-known case over mine water fouling a stream, and the decision is widely read as protective of industry. Philander Knox, one of the club’s members, later served as United States Attorney General and Secretary of State, which gives some sense of who was on the membership list.

The 1891 engineers’ report and a modern reanalysis that disagrees

The American Society of Civil Engineers appointed a committee to investigate the failure soon after the flood, and the committee visited the site. Its report was not presented to the society until 1891, two years after the flood. In substance, the committee concluded that the flood exceeded what the dam could have passed even in its original condition, so the club’s alterations did not decide whether it failed. That conclusion supported the club’s act of God position.

In 2016, a group of geologists at the University of Pittsburgh at Johnstown published a reanalysis that challenged the 1891 findings. They modeled the hydraulics of the original and altered dam. They argued that the lowered crest, together with a sag in the middle of the embankment, substantially reduced the dam’s capacity to pass flood water, and that the 1891 committee had understated what the alterations did. Readers who want the method and the numbers should go to the published paper rather than to news summaries of it.

Neither document is a court finding, and they disagree on whether the alterations caused the failure. The 1891 committee worked from the site itself and from witnesses who had seen the dam, using the hydraulic methods of its day. The 2016 group applied modern modeling to historical measurements of a structure that has not existed in its 1889 form for well over a century. I would put both in front of a class and say plainly that they disagree. Both support a narrower point that does not depend on settling the counterfactual. A dam that cannot be drawn down, and whose overflow capacity has been reduced, has less margin in a large storm than the dam its designers built.

Rylands v. Fletcher, the American courts, and foreseeable negligence

Under the negligence rule that governed most American injury law in 1889, a plaintiff had to prove two things. The first was that the defendant failed to take the care a reasonable person would have taken against a foreseeable harm. The second was that this failure caused the injury. The Fulton report went to foreseeability, because it showed the club had been told about the defects in writing. The act of God defense attacked causation, arguing that the rain, rather than the condition of the dam, killed the people of Johnstown.

English law already had a different rule, and it came out of a reservoir case. In Rylands v. Fletcher, water from a mill owner’s newly built reservoir broke into old mine workings and flooded a neighboring coal mine. In the Exchequer Chamber in 1866, Justice Blackburn held that “the person who for his own purposes brings on his lands and collects and keeps there anything likely to do mischief if it escapes, must keep it in at his peril.” The House of Lords upheld the result in 1868, and under that rule the question becomes whether the water escaped and did damage, without the plaintiff having to prove that the owner was at fault.

Several American courts declined to adopt Rylands in the 1870s, among them the New Hampshire court in Brown v. Collins and the New York court in Losee v. Buchanan, both decided in 1873, and both reasoned that strict liability would burden industrial development. Jed Handelsman Shugerman, in a 2000 Yale Law Journal article titled “The Floodgates of Strict Liability,” argued that the Johnstown flood and other reservoir failures of the period changed that picture. In his account, American courts adopted Rylands far more widely in the decades after 1889 as public opinion turned against owners who stored dangerous volumes of water above other people’s homes. His article is an interpretation of why courts moved, which is a different thing from a court citing Johnstown as its reason. It remains the most direct scholarly link between the flood and the law of liability.

From there the line runs through the twentieth-century Restatements of Torts, which recognize strict liability for ultrahazardous and later abnormally dangerous activities. It continues into state dam safety statutes, which place inspection and maintenance duties on owners directly. In this country, dam regulation has tended to follow failures. Georgia passed its Safe Dams Act in 1978, after the Kelly Barnes Dam above Toccoa Falls College failed in November 1977 and killed 39 people. How a court in your state would treat a dam failure today depends on that state’s statute and case law. That question belongs with your county attorney, not with a history article.

What to do at your agency

• Have your emergency manager pull the state dam safety program’s list of high and significant hazard dams upstream of your jurisdiction, record the owner of record for each one in the hazard annex, and flag every dam owned by a homeowners association, club or other private group.
• Have your emergency manager ask the state dam safety engineer responsible for your area for the date and findings of the most recent inspection of each high hazard dam on that list, and file a copy of each report with the emergency operations plan so the county holds its own record of what was found.
• Ask your public works director whether the county itself owns any dam, including park lakes and water supply reservoirs, and if it does, obtain the latest inspection report and confirm that every listed deficiency has a work order with a recorded status.
• Add one item to the agenda of your existing hazard mitigation planning committee asking the owner representative of each upstream high hazard dam to state in writing whether the crest, spillway or outlet works have been altered since the last inspection, including roads, bridges, fish screens and debris barriers.
• Have your communications center manager place a test call to the owner contact number for each high hazard dam and confirm that someone answers who can describe the dam’s condition. Then add one sentence to the dispatch procedure requiring that every dam owner warning be relayed to the on-duty supervisor and the emergency manager, however many earlier warnings proved unfounded.
• Ask your county attorney for a one-paragraph summary of your state’s dam safety statute covering owner duties and any county responsibilities, and place it in the dam section of the emergency operations plan.

Takeaways

• The South Fork Dam failed on May 31, 1889, and the National Park Service publishes the toll as 2,209 dead, including 99 entire families, 396 children and 777 unidentified victims buried in the Plot of the Unknown.
• Pennsylvania built the dam for its canal system and finished it in 1853, and the dam had partly failed in 1862 and lost its discharge pipes before the club bought it in 1879.
• The club lowered the crest for a carriage road, fitted fish screens across the spillway and never replaced the discharge pipes, which together left the lake impossible to draw down and reduced the dam’s margin in a large storm.
• McCullough reports that a Cambria Iron Company engineer, John Fulton, criticized the dam’s condition in an 1880 inspection and that Benjamin Ruff dismissed the concerns, which shows the hazard was documented nine years before the failure.
• Suits against the club produced no recovery, because the club argued the flood was an act of God and a corporation with few assets stood between the plaintiffs and the members’ fortunes.
• The 1891 engineers’ committee concluded the dam would have failed even unaltered, a 2016 reanalysis from the University of Pittsburgh at Johnstown argued the alterations mattered substantially, and the two disagree.
• Rylands v. Fletcher made the owner of escaping stored water liable without proof of fault, several American courts rejected that rule in 1873, and Jed Handelsman Shugerman has argued that Johnstown helped push American courts toward adopting it.
• Any death toll well above 2,209 that appears without a source is usually an early press estimate or an inflated rounding and should not be used.