The accident that began at about four in the morning on Wednesday 28 March 1979 at Three Mile Island Unit 2, on an island in the Susquehanna River south of Harrisburg, damaged roughly half a reactor core and killed nobody. What it did do, over five days, was put a utility, a federal regulator and a brand new governor in front of the same reporters with different information, and the resulting collapse of public confidence is the reason American radiological emergency planning looks the way it does today. This is a look at the notification chain and the public information failures, not at the reactor.

The first four hours and who got told

Unit 2 was owned and operated by Metropolitan Edison, a subsidiary of General Public Utilities, and it tripped in the early hours of 28 March after a loss of feedwater, followed by a pressurizer relief valve that opened and failed to reseat while the control room indication suggested it had closed. The plant staff spent the first hours working a problem they had misdiagnosed. Published timelines, including the Nuclear Regulatory Commission’s own backgrounder on the accident and the report of the President’s Commission on the Accident at Three Mile Island, put the declaration of a site emergency at roughly 6:55 a.m. and a general emergency about half an hour later, and if you are going to quote clock times in a class, take them from those documents rather than from a summary like this one.

The outward notifications went in the order the 1979 arrangements specified, which meant the company called state and county officials before anyone with radiological expertise was in a position to interpret what was happening. Pennsylvania Emergency Management Agency and the state’s radiation protection staff were notified during the seven o’clock hour, Dauphin County and the adjacent counties were brought in, and the NRC’s regional office learned of the event shortly afterward rather than simultaneously. Nothing in that chain was built to carry technical uncertainty, because it was built to carry a classification word and a request, and the situation on the island that morning was one that nobody in the control room could yet describe accurately.

By the middle of the morning the story was on the wires, a radio station in Harrisburg had it from a traffic reporter and from listeners, and the state was fielding press calls before it had an agreed account to give. The film “The China Syndrome” had opened in theaters twelve days earlier, on 16 March 1979, which is a coincidence rather than a cause but which shaped the questions reporters arrived with. Within a day the area’s telephone circuits were carrying a volume of press and public traffic that the utility’s small number of lines could not absorb, and the people who most needed to reach each other were competing with everyone else for the same copper.

Three organizations, three versions

The core failure was structural rather than personal. Three organizations were speaking publicly about the same reactor, and each of them had a different and incomplete picture: Metropolitan Edison, which owned the plant and the data and also owned an obvious interest in how the event was characterized; the NRC, which had people at the site, people in the regional office and commissioners in Bethesda who were not receiving the same information at the same time; and the governor’s office, which had no independent radiological measurement capability and was being asked to make protective action decisions anyway. Governor Richard Thornburgh had been in office since January, roughly ten weeks.

The conflicts were not subtle. The company’s public characterization of the release and of the condition of the core was more reassuring than what state officials were learning through other channels, and Lieutenant Governor William Scranton III said publicly on the first day that the situation was more complicated than the company had initially indicated, which is a paraphrase and the exact wording is worth pulling from the contemporary transcripts if you need to quote it. A utility executive’s remark at a press conference about not needing to tell reporters every single thing the company did became one of the most repeated lines of the week, and here again I would go to the contemporary record for the wording rather than trust the version that circulates.

What made it worse is that none of the three was lying in any way a later investigation established. Each was describing the fragment of the event it could see, and because no mechanism existed to reconcile the fragments before they reached the microphone, the public heard the disagreement rather than the reconciliation. A resident in Middletown listening to the radio on Thursday could hear the utility, the state and the federal regulator characterize the same release differently within an hour, and the only rational conclusion available to that resident was that nobody in charge knew what was happening, which happened to be closer to the truth than any of the individual statements.

Friday morning, the helicopter reading, and the advisory

Friday 30 March is the day the communications chain failed in a way that is worth studying line by line. A release of radioactive gas from the auxiliary building prompted a radiation measurement taken from a helicopter directly above the plant’s vent stack, and the reading, in the range of 1,200 millirem per hour, was passed up the line without the qualifier that made it interpretable, which was that it had been taken in the plume immediately above the stack rather than at ground level where people were. Senior NRC staff who received the number reasonably read it as an offsite dose rate, and an evacuation recommendation began moving from Bethesda toward Pennsylvania on that basis.

What happened next is described in the investigation reports and in Thornburgh’s own later accounts. An evacuation recommendation reached state and county officials and was then modified or withdrawn as the origin of the measurement became clear, county civil defense offices spent part of the morning preparing for an evacuation that was not ordered, and published accounts describe civil defense sirens sounding in Harrisburg that morning with no accompanying instruction and no clear account afterward of who had set them off. By late morning the governor, on the advice of NRC Chairman Joseph Hendrie, issued an advisory recommending that pregnant women and preschool age children within five miles of the plant leave the area and that everyone within ten miles stay indoors, and schools in the five mile ring were closed.

No mandatory evacuation was ever ordered. Survey work done afterward for the President’s Commission produced the commonly cited estimate that about 144,000 people within roughly fifteen miles left voluntarily, and that figure is an estimate derived from post event surveys rather than a count, with other published accounts giving somewhat different totals. The advisory itself deserves attention as a piece of drafting, because it covered a population defined by pregnancy and age, drew a five mile line for one group and a ten mile line for another, and was issued into an information environment where a substantial share of the public had already concluded that official statements understated the danger.

The bubble weekend and technical uncertainty in public

Over the weekend of 31 March and 1 April the public story became a hydrogen bubble in the top of the reactor vessel, with the associated question of whether it could burn or explode. The concern was real inside the technical community for a period of roughly two days, and it was resolved when the analysis showed that there was no mechanism to accumulate enough oxygen in the vessel to make the hydrogen flammable, and when the bubble itself was reduced. The NRC’s published account of the accident describes the concern and its resolution, and the Special Inquiry Group report led by Mitchell Rogovin, published in January 1980, goes into how the estimate was formed and corrected.

The communications problem is that the disagreement was aired while it was still a disagreement, without the framing that would have let a listener understand what kind of statement they were hearing. An internal technical dispute about whether a calculation is conservative is normal engineering, and when it reaches the public as competing assertions about whether a reactor might explode, the public has no way to weight them. President Carter’s visit to the site on 1 April, with the First Lady, was in part an act of public reassurance aimed at that specific problem, and Harold Denton, whom Carter had sent to the site as the NRC’s on scene representative and single technical spokesman, became the figure the state relied on for interpretation.

Hendrie’s own description of the decision environment is recorded in the NRC’s internal transcripts from that week and is reproduced in the President’s Commission material. He described himself and the governor as making decisions close to blind, with the governor’s information ambiguous and his own essentially absent. I am paraphrasing deliberately, because the line circulates in several slightly different forms and I have not been able to verify a single authoritative wording; the transcripts are public and that is where to get it if you intend to quote it on a slide.

The health record, and why this is not Chernobyl

There were no deaths and no injuries attributed to radiation from the Three Mile Island accident. The NRC’s published estimates, which come out of the dose assessments done after the event, put the average dose to the roughly two million people in the surrounding area at about one millirem above normal background, with the maximum estimated dose to an individual offsite well under one hundred millirem, against a natural background in that part of Pennsylvania on the order of one hundred to one hundred and twenty five millirem per year. The President’s Commission concluded that the most significant health effect of the accident was psychological stress in the surrounding population.

The epidemiology is not entirely quiet, and an honest treatment says so. Follow up studies of the exposed population, including work from Columbia University published around 1990 and work from the University of Pittsburgh, reported no consistent evidence of increased cancer attributable to the release, while a reanalysis by Steven Wing and colleagues published in the late 1990s argued that the data were compatible with an effect and that the dose estimates themselves carried more uncertainty than the official figures conveyed. Those researchers disagree with each other, the disagreement has not been resolved by any subsequent consensus I can point to, and the appropriate way to state it is that the weight of the published work finds no detectable radiation related mortality while a minority of researchers dispute the dose reconstruction.

The conflation with Chernobyl is the error I see most often in training material, and it is a large one. The April 1986 accident at Chernobyl Unit 4 in the Ukrainian SSR involved a destroyed reactor, an open core, a large uncontrolled release and about thirty deaths within months among plant staff and emergency workers from acute radiation syndrome and trauma, according to the assessments published by UNSCEAR and the IAEA, along with a documented later increase in thyroid cancer among people exposed as children. The two events sit in different categories of consequence, and an instructor who blurs them loses the specific and genuinely useful lesson that Three Mile Island teaches, which is about information rather than about radiation.

What the investigations found about information flow

The President’s Commission on the Accident at Three Mile Island, chaired by Dartmouth College president John Kemeny, reported in late October 1979 after taking testimony from the utility, the NRC, state officials and plant staff. Its findings on public information were blunt: the utility’s statements were not consistently reliable, the NRC had no adequate system for informing the public, the various sources of information were uncoordinated, and the absence of a single authoritative technical source until Denton arrived left journalists and officials assembling their own pictures from fragments. The commission also found that the emergency planning arrangements between the plant, the state and the counties were inadequate to the event.

The Rogovin Special Inquiry Group, reporting to the commissioners in January 1980, covered similar ground from inside the agency and examined how information moved between the site, the regional office and headquarters. Both reports identify the same underlying defect, which is that the notification chain had been designed to transmit a classification and trigger a response, and nobody had designed a parallel channel for transmitting what was actually known, what was estimated and what nobody could yet determine, so a chain that could only carry conclusions ended up manufacturing them when it was fed uncertainty.

The organizational consequences followed quickly. The NRC was restructured, with the chairman given clearer authority during emergencies, and the agency built out a permanent operations center and a dedicated emergency notification telephone line to licensee control rooms, so that the regulator was no longer dependent on ordinary commercial circuits and ad hoc calls during an event. If you want the specifics of the recommendations and which ones were implemented, the two reports are public and the NRC maintains a summary of its post accident actions, which is a better source for detail than any retelling.

What got rebuilt: zones, sirens, and the joint information center

The most durable product of 1979 is the two zone structure that still governs planning around every commercial power reactor in the United States. The NRC and the newly created Federal Emergency Management Agency established a plume exposure pathway emergency planning zone of about ten miles radius, within which protective actions such as sheltering and evacuation are planned in detail, and an ingestion pathway zone of about fifty miles, aimed at food and water contamination. The planning criteria were published jointly by the two agencies as NUREG-0654 and FEMA-REP-1, and FEMA took responsibility for reviewing offsite plans prepared by state and local governments, a division of labor that is still in place.

Inside the ten mile zone, the criteria required an alert and notification system capable of reaching essentially the whole population quickly, which in practice produced the siren networks, tone alert radios and broadcast arrangements that utilities fund and counties operate around plants today, with a notification design objective usually described in terms of about fifteen minutes. Sirens alert but do not inform, which was understood at the time, so the design always assumed a broadcast channel carrying the actual instruction, formerly the Emergency Broadcast System and now the Emergency Alert System and Wireless Emergency Alerts through IPAWS. Verify current requirements and the current text of the criteria with the NRC and FEMA rather than relying on this description, because the documents have been revised several times.

The joint information center is the other inheritance, and it is the one that reaches departments with no reactor anywhere near them. The idea that the licensee, the regulator, the state and the affected counties put their public information staff in one room, agree on the facts before release and issue from a single point is a direct response to the contradictions of late March 1979. It was carried into general emergency management through the public information function in the Incident Command System and the Joint Information System described in the National Incident Management System, which is why the same structure now appears in hurricane, wildfire and pipeline responses that have nothing to do with radiation.

Three Mile Island in 2026 and what I could not verify

Unit 2 has not operated since the accident. Defueling was completed around 1990 and the unit sat in a monitored storage condition for decades, with ownership of the damaged unit transferred in 2019 and 2020 to a decommissioning company for eventual dismantlement. Unit 1, the undamaged reactor on the same island, ran until September 2019, when its owner shut it down for economic reasons.

In September 2024, Constellation Energy announced a plan to restart Unit 1 under a long term power purchase agreement with Microsoft, renaming the site the Crane Clean Energy Center after the company’s late chief executive. Restart dates have been publicly revised more than once since that announcement, and the process requires a series of regulatory approvals from the NRC along with grid interconnection work. I have not been able to verify the current status of that restart or the current schedule as of today, 29 September 2026, and I am not going to state one. Check the NRC’s public documents on the license amendments and inspections for that unit, and the company’s own filings, before you repeat any date.

The reason this matters for a communications officer rather than an energy analyst is that restarting a reactor restarts an emergency planning zone, with the county sirens, the tone alert radios, the protective action decision chain, the annual exercise cycle and the joint information center that go with it. If your jurisdiction sits inside ten miles of any operating or restarting plant, the practical question is whether the offsite plan and the public notification equipment have been exercised recently with your people in the seats, and that is a question for your county emergency management director with a documented answer, not an assumption.

What to do at your agency

• Ask your emergency management director for the current joint information center procedure that applies to your jurisdiction and confirm, in writing, whether it contains a step requiring technical and elected officials to reconcile their facts before a release goes out, and if that step is missing, draft the paragraph and put it on the next EMA committee agenda.
• Have your communications supervisor write a short standing rule into the dispatch guideline that any radiological, air monitoring or hazmat reading relayed through the center must include instrument, units, location and time, and have supervisors check compliance on the next hazmat incident.
• Have your PIO or a designated officer identify by name and phone number the single technical subject matter expert your agency will call for a radiological or industrial release, whether that is a utility health physicist, a state radiation protection office or a regional hazmat team, and put that contact on the same card as your after hours numbers.
• If any part of your jurisdiction lies within ten miles of an operating or restarting nuclear plant, ask your county emergency management director for the date of the last graded exercise, the siren coverage map and the list of alert and notification equipment your agency is responsible for, and read the corrective actions from the last exercise report.
• Test your end of the public alerting path this quarter by sending a scheduled test message through your EAS or WEA origination method with your alerting authority’s approval, and document who at your agency is authorized to originate, who can approve after hours, and how long the whole path actually took.
• Correct any orientation slide in your department that presents Three Mile Island alongside Chernobyl without distinguishing the consequences, because a student who later learns that nobody died at Three Mile Island will discount the rest of what you taught them.

Takeaways

• The accident at Three Mile Island Unit 2 began at about four in the morning on 28 March 1979, and no deaths or injuries have been attributed to radiation released from it.
• The NRC’s published dose estimates put the average exposure to the roughly two million people in the surrounding area at about one millirem above background, with the maximum estimated individual offsite dose well under one hundred millirem, while a minority of researchers, notably Steven Wing and colleagues in the late 1990s, have disputed the dose reconstruction.
• The President’s Commission chaired by John Kemeny, reporting in late October 1979, concluded that the most significant health effect was psychological stress and found the public information arrangements uncoordinated and at times unreliable.
• The central communications failure was structural, because the utility, the NRC and the governor’s office each held a different fragment of the picture and no mechanism existed to reconcile those fragments before they reached the public.
• On the morning of 30 March a radiation reading taken by helicopter above the plant’s vent stack traveled up the chain without its location, was read as an offsite ground level dose rate, and contributed to an evacuation recommendation that was subsequently modified.
• Governor Thornburgh issued an advisory that day recommending that pregnant women and preschool age children within five miles leave and that everyone within ten miles stay indoors, and no mandatory evacuation was ever ordered, though survey work for the President’s Commission produced the commonly cited estimate that about 144,000 people left voluntarily.
• The ten mile plume exposure and fifty mile ingestion pathway planning zones, the NUREG-0654 and FEMA-REP-1 planning criteria, the siren and broadcast alerting requirements and the joint information center all trace directly to this accident.
• Constellation announced in September 2024 a plan to restart Unit 1 as the Crane Clean Energy Center, the publicly reported schedule has been revised since, and I have not verified its current status, so check NRC documents and company filings before repeating a date.