The accident at Three Mile Island Unit 2 in March 1979 released little radiation to the public, and no death or illness has been confirmed as caused by that radiation. The emergency response around it went badly in ways that still shape planning. Over several days the utility, the Nuclear Regulatory Commission and the Commonwealth of Pennsylvania gave residents statements that did not agree, and surveys afterward found that far more people left than any advisory had asked to leave. The planning zones, classification levels and parallel notification rules at every U.S. nuclear plant today were written largely in answer to that week.

March 28 to April 9, 1979: the week the response worked from

At about 4 a.m. on Wednesday, March 28, 1979, Unit 2 at the Three Mile Island station lost the flow of feedwater to its steam generators. The station sits on the Susquehanna River a few miles southeast of Harrisburg. The turbine and then the reactor tripped as designed. A pressure relief valve on the reactor’s primary cooling system opened and failed to close, and the control room light associated with it showed only that a signal to close had been sent. Operators believed the system was filling with too much water, so they reduced the emergency cooling flow, and over the following hours the top of the core was uncovered and a large part of it melted. Metropolitan Edison, a subsidiary of General Public Utilities, operated the plant. The NRC’s current public fact sheet puts the damage at about half the core. Nobody could confirm that figure until cameras and sampling reached the reactor vessel years later.

Two final investigations settled that technical sequence. The first was the President’s Commission on the Accident at Three Mile Island, chaired by Dartmouth president John Kemeny, which reported in October 1979. The second was the NRC’s own Special Inquiry Group under Washington attorney Mitchell Rogovin, which reported in January 1980. This article leaves the reactor to them and looks at the response outside the fence, which ran on its own calendar. The utility declared a general emergency early on Wednesday morning. On Friday, March 30, an unplanned release and a garbled recommendation produced the conflicting advisories this piece is mostly about, and that afternoon President Carter sent Harold Denton, the NRC’s director of nuclear reactor regulation, to the site. Concern over a hydrogen bubble in the reactor vessel peaked over the weekend. Carter toured the plant on Sunday, April 1, and Governor Dick Thornburgh withdrew the last of his advisories on April 9.

I was not in Pennsylvania in 1979 and had no part in the response, so everything here comes from the published record. My main sources are the Kemeny report, the NRC’s fact sheet and the health studies named below. Where those sources give a precise time or figure, I use it with the source attached. Where accounts differ or I could not pin a detail down, I say so or leave it out. That discipline matters more than usual with Three Mile Island, because advocates on both sides of the nuclear power argument have retold the accident so often that much of what circulates about it has drifted away from the documents.

Wednesday: a utility, a lieutenant governor and two different stories

The first failure of public information happened on the first day, and it set the pattern for the week. The utility notified the NRC and Pennsylvania officials that morning. Its early public statements described the event in limited terms, with spokesmen saying in substance that the situation was under control and that releases off the site were not significant. The Kemeny Commission’s treatment of public information criticizes the utility’s statements during the accident as inconsistent, both with what was happening inside the plant and with what other officials were saying. I am paraphrasing the commission there, and the report itself is the place to read its exact language.

Lieutenant Governor William Scranton III served as the state’s public spokesman. At a briefing that Wednesday morning he passed on the reassuring picture the utility had given the state. Later the same day the state learned that the plant had been releasing radioactive material to the atmosphere, and Scranton told reporters, in substance, that the situation was more complex than the company had first led state officials to believe. That reversal did lasting damage. Residents and reporters drew the reasonable conclusion that official statements could change within hours and that the utility could not be counted on to report bad news promptly.

Behind the public statements sat an information problem that anyone in public safety communications will recognize. The state’s radiation protection staff was small, and for much of the first day state officials depended heavily on the utility for what was happening inside the plant. NRC headquarters outside Washington got its information through telephone calls to the site and to its regional office, relayed by people who were themselves busy with the accident. No direct data connection ran from the control room to anyone outside it. Each organization built its own picture from secondhand reports and briefed its own audience from that picture, so differences that started as gaps in information came out in public as contradictions.

Friday, March 30: a stack reading and advisories from several directions

Early on Friday morning, during a transfer of radioactive gas inside the plant, an unplanned release went out through the vent stack. A helicopter flying over the stack measured radiation at about 1,200 millirem per hour, the figure given in the Kemeny Commission’s account and repeated in most later histories. That reading was taken in the plume directly above the stack, well away from the houses at ground level. As the number moved by telephone to NRC headquarters, it was treated as though it might represent what people downwind could receive, and NRC staff there passed a recommendation to evacuate to the Pennsylvania Emergency Management Agency. Accounts of the accident describe that recommendation as resting on a misunderstanding of where the measurement came from. The advice that ultimately reached the governor from the NRC chairman that day was far narrower.

For a few hours that morning, residents heard several things at once. Word of a possible evacuation spread through official channels and the news media. Thornburgh, trying to establish what the federal government was actually recommending, advised people within ten miles to stay indoors. Around midday he spoke directly with NRC Chairman Joseph Hendrie, who told him in substance that a limited precautionary step was reasonable. Thornburgh then advised that pregnant women and preschool-age children within five miles of the plant leave the area, that others within ten miles remain indoors, and that schools in the five-mile area close. In one morning, people near the plant had heard of a possible evacuation recommended by federal staff, then a ten-mile stay-indoors advisory, then a five-mile advisory that applied only to some of them, while the utility continued to offer its own assessment.

The record I have read does not suggest that any of those officials was trying to mislead anyone that morning. The confusion came from structure, because no plan settled who could recommend protective actions to the public and who could order them, or whose technical assessment a governor should rely on when the regulator’s own staff disagreed. The NRC at the time regulated licensees and had no established role in advising the public. The state held authority over its residents but had very little independent technical capability. The utility had the most information about the plant and had already lost much of its credibility on Wednesday. Each of them filled the gap from its own position, and people in the counties around the plant received the result as a series of instructions that contradicted one another.

The people who left anyway: a small target group and a large departure

Thornburgh’s Friday advisory covered a narrow group: pregnant women and preschool children living within five miles of the plant, a few thousand people in all. Many more people actually left. Surveys done for the NRC in the months after the accident estimated that about 144,000 people living within roughly 15 miles of the plant left at some point during the crisis, and most later planning literature cites that figure. Survey estimates of this kind carry real uncertainty, so I present the number as an estimate from those NRC-sponsored surveys and not as a count. Even so, no source I know of puts the departure anywhere near the size of the group the advisory named.

Most of the people who left stayed with relatives or friends or in hotels, and very few used the mass care facilities opened for them. They made their own decisions based on whatever information they trusted, and after Wednesday’s reversal and Friday’s confusion, the official advisory was not necessarily the information they trusted. Planners later took Three Mile Island as the standard example of what came to be called shadow evacuation, meaning departures by people outside the area an official advisory covers. The term is now part of ordinary evacuation planning for hurricanes and chemical releases as well as for nuclear plants.

The planning consequence is concrete. The NRC requires an evacuation time estimate for each U.S. nuclear plant, and state and county planners use it to judge how long clearing the 10-mile zone would take. These estimates now include an assumption about how many people outside the zone will leave at the same time and load the same roads. NRC guidance and each plant’s study set the specific percentage, and your county emergency management agency can show you the one that applies locally. The lesson carries over directly to agencies far from any reactor. A narrowly drawn evacuation order for a chemical plume or a wildfire will put traffic on the roads from well outside the drawn line, and a traffic control plan built only around the ordered population will be wrong in the first hour.

What went right: a narrow advisory and one federal voice

The response deserves credit in a few specific places, and that credit should be stated as narrowly as the failures. On Friday, Thornburgh limited the advisory to pregnant women and young children within five miles and did not order a general evacuation, and that decision held up against what was later learned about the doses. A general evacuation would have meant moving hospital patients and nursing home residents in a hurry along with everyone else living nearby, and every hurried movement of that kind carries its own risk of injury. Targeting the groups most sensitive to radiation was a defensible public health choice given what the governor could know that afternoon, and the dose estimates discussed below are consistent with his judgment that a wider order was not needed.

The second thing that went right was improvised. When Carter sent Harold Denton to the site on Friday afternoon, Denton became in practice the single federal technical spokesman, briefing the governor and the press near the plant with NRC technical staff behind him. Accounts of the week generally credit his arrival with reducing the number of conflicting federal voices. The hydrogen bubble episode over the weekend showed that the arrangement was incomplete. News reports quoting NRC sources raised the possibility of an explosion in the reactor vessel while technical staff were still working out that the bubble could not burn, because there was no free oxygen in the vessel to support it. Carter’s tour of the plant on Sunday, April 1, is also commonly credited with easing public fear, and the advisory stayed in place another eight days while the reactor was brought under control.

Some of the credit belongs to the plant, and it has to be kept separate from the emergency response. The containment building held, and most of the radioactive material released from the damaged fuel stayed inside the plant. That is why the offsite doses were as small as they were, and nothing the governor, the NRC or the counties did produced that outcome. The fair summary is that officials made a reasonable protective action call on Friday and eventually settled on a single technical voice, both late and under pressure. The low offsite consequences that made those calls look sound came from the engineering of the plant, not from the quality of the emergency planning around it.

The health myth: the dose estimates and the long-term studies

The claim that radiation from Three Mile Island caused a measurable increase in cancer or other disease among people living nearby is a myth, in the specific sense that the major epidemiological studies of the surrounding population have not found such an increase. It should be labeled that way whenever it comes up. The starting point is the dose. The NRC’s fact sheet draws on the dose assessments federal agencies made after the accident. It estimates that the roughly two million people around the plant received an average dose of about 1 millirem above background, and that the maximum dose to a person at the site boundary would have been under 100 millirem. The same fact sheet compares that average to about 6 millirem for a chest X-ray and to a natural background dose in the area of roughly 100 to 125 millirem per year.

The long-term studies point the same way. Researchers at Columbia University examined cancer incidence around the plant and published findings in 1990 that did not show a convincing link between the accident’s emissions and cancer rates. Researchers at the University of Pittsburgh followed a cohort of more than 30,000 people who had lived within five miles, drawn from a registry the Pennsylvania Department of Health set up after the accident. They reported in 2000 and 2003 that they found no consistent evidence of increased cancer mortality attributable to radiation from the accident. Steven Wing and colleagues at the University of North Carolina reanalyzed the Columbia data and published their results in 1997. They reported associations between estimated exposure and some cancers and argued that the doses had been higher than the official estimates. That reanalysis is the main published dissent. Other researchers have disputed it, and it has not displaced the larger body of findings, but anyone presenting the record fairly should mention that it exists.

The accident did harm people, but not through radiation. The Kemeny Commission concluded, in substance, that the most serious health effect of the accident was mental stress among people in the region. Later studies of residents, including mothers of young children, documented elevated distress that in some groups lasted well beyond the week of the crisis. The commission and later researchers tied that distress to the fear and uncertainty of the crisis rather than to radiation exposure. When the subject comes up, I tell people that the record does not show measurable physical harm from radiation and does show psychological harm connected to how that week unfolded.

After 1979: Kemeny, the FEMA assignment and the 10- and 50-mile zones

The Kemeny Commission’s recommendations reached well beyond the plant. Among them was that responsibility for offsite emergency planning and response be consolidated under the Federal Emergency Management Agency, with approved state and local plans required before a plant could operate. In December 1979 Carter directed FEMA to take the lead in offsite planning around nuclear power plants, and the NRC kept responsibility for the licensee and for onsite planning. The two agencies have worked under memoranda of understanding since then. FEMA’s Radiological Emergency Preparedness program, operating under its regulation at 44 CFR Part 350, reviews state and local plans and evaluates their exercises.

The NRC’s 1980 emergency planning rule, codified at 10 CFR 50.47 and Appendix E to Part 50, made emergency planning a condition of operating licenses. Under the rule, the NRC will not issue an operating license unless it finds reasonable assurance that adequate protective measures can and will be taken in a radiological emergency, and for offsite measures the NRC bases that finding on FEMA’s review. Both agencies apply planning criteria they published jointly as NUREG-0654/FEMA-REP-1. It was first issued in 1980 and has been revised since, so anyone working on a plan should get the current revision from the NRC or FEMA rather than a copy found online.

The 10-mile and 50-mile zones are older than the accident, a detail that is routinely lost. A joint NRC and EPA task force proposed them in NUREG-0396, a report published in December 1978. It recommended a plume exposure pathway emergency planning zone of about 10 miles and an ingestion pathway zone of about 50 miles. Three Mile Island turned that guidance into a licensing requirement backed by evaluated exercises. The requirement soon carried real weight. The Shoreham plant on Long Island was built but never entered commercial operation, after New York State and Suffolk County declined to take part in offsite emergency planning and the dispute over evacuating Long Island became central to the plant’s fate.

A related myth holds, wrongly, that the accident created FEMA. The reorganization plan that consolidated federal emergency functions into FEMA went to Congress in 1978. The executive order activating the agency was signed on March 31, 1979, while the accident was under way, but it was the end point of work that had started long before. Three Mile Island gave the new agency one of its first defining assignments when Carter handed it the offsite planning lead nine months later.

The two tracks today: classification, notification and who decides

The system that came out of those reforms runs on two parallel tracks, and its main features answer the problems that surfaced on March 30. On the federal track, the licensee reports to the NRC, which oversees the plant’s response and assesses the technical situation. For an incident at a facility it licenses, the NRC serves as the coordinating agency under the Nuclear/Radiological Incident Annex to the National Response Framework. On the state and local track, the licensee notifies the designated offsite response organizations directly, and state and local officials make and announce protective action decisions for the public, with FEMA supporting them and evaluating their readiness ahead of time. The design deliberately puts technical assessment and protective action authority in different hands. It also requires the two tracks to share information continuously, so that the officials who decide have the plant’s technical assessment in hand when they do.

The common language starts with four emergency classification levels standardized in NUREG-0654: Notification of Unusual Event, Alert, Site Area Emergency and General Emergency. Each level has defined triggers in the plant’s emergency action levels and defined actions offsite. Under NRC rules, the licensee must notify state and local authorities within 15 minutes of declaring an emergency classification, and must notify the NRC immediately after that and within one hour. At a General Emergency the licensee must also give offsite officials a protective action recommendation. The recommendation is the licensee’s technical advice. The decision belongs to the state and local officials who hold legal authority over the public, and that is exactly the distinction nobody had written down in 1979.

Public information goes through a joint information center, where the utility, the state, the counties and the federal agencies coordinate releases before they go out. A near-site emergency operations facility brings decision makers and technical staff into the same building. Sirens and the Emergency Alert System alert the public in the 10-mile zone. Where a siren does not reach, or when one fails, fire and police units carry out route alerting, and FEMA guidance sets a design objective for how quickly the public can be alerted. FEMA evaluates offsite plans in exercises held every two years at each plant. As a result, county fire, EMS, law enforcement and communications staff near a plant have assigned roles in those plans and an evaluated record of performing them, which is the practical difference between the planning of 1979 and the planning of today.

What to do at your agency

• Have your emergency manager confirm in writing whether any part of the jurisdiction lies inside the plume exposure or ingestion pathway emergency planning zone of a nuclear plant, including plants in neighboring states, and file the answer with the hazard analysis in the emergency operations plan.
• If any part of the jurisdiction is inside a zone, have each fire, EMS and law enforcement chief get from the county emergency management agency the pages of the radiological emergency plan that assign tasks to their agency, such as traffic control points or route alerting, and confirm that the named point of contact for each task is still in the job.
• Have your radio system administrator check that the talkgroups and notification paths named in the radiological plan, or in the hazardous materials annex for agencies outside any zone, still exist under those names in current radio programming, and correct whichever document is out of date.
• Have your public information officer write one paragraph into the existing crisis communications plan stating which agency issues protective action messages to the public and how a correction to an earlier statement will be coordinated with partner agencies before release.
• Add one item to the agenda of the next local emergency planning committee meeting asking whether the traffic plan for the largest planned evacuation zone accounts for people leaving from outside the zone, and record the answer in the minutes.
• Have your training officer search hazmat and radiological awareness materials for any statement that Three Mile Island caused a measurable increase in cancer, and replace it with the NRC dose figures and the Columbia and Pittsburgh study findings, with the 1997 reanalysis noted.
• Have your communications center manager add a requirement to the hazardous materials dispatch procedure that any air monitoring or radiation reading relayed by radio include the location, time and instrument, and brief each shift on the change before the end of the month.

Takeaways

• The Three Mile Island Unit 2 accident began about 4 a.m. on March 28, 1979, and the NRC’s fact sheet says about half the core melted, although containment kept most of the radioactive material inside the plant.
• Public confidence began to erode on the first day, when the state relayed the utility’s reassurances in the morning and Lieutenant Governor Scranton revised them that afternoon after learning of releases.
• On Friday, March 30, a helicopter reading of about 1,200 millirem per hour taken directly above the vent stack was misread as a possible downwind dose and set off a chain of conflicting advisories.
• Governor Thornburgh advised only pregnant women and preschool children within five miles to leave, yet NRC-sponsored surveys estimated that about 144,000 people within roughly 15 miles left during the crisis.
• The claim that radiation from the accident caused measurable public health harm is a myth according to the major studies, which found no attributable cancer increase, although a 1997 University of North Carolina reanalysis dissented and documented psychological stress was real.
• The 10-mile and 50-mile emergency planning zones were proposed in NUREG-0396 in December 1978, before the accident, and the 1980 NRC rule made offsite planning a condition of operating licenses.
• The claim that Three Mile Island created FEMA is also a myth, because the reorganization plan predated the accident, although Carter gave FEMA the offsite planning lead in December 1979.
• Today the licensee notifies state and local officials within 15 minutes of a classification and the NRC within one hour, and the licensee recommends protective actions while state and local officials decide them.