This spring, four years after the nuclear accident at Fukushima, a small group of scientists met in Tokyo to evaluate the deadly aftermath.

No one has been killed or sickened by the radiation — a point confirmed last month by the International Atomic Energy Agency. Even among Fukushima workers, the number of additional cancercases in coming years is expected to be so low as to be undetectable, a blip impossible to discern against the statistical background noise.

But about 1,600 people died from the stress of the evacuation — one that some scientists believe was not justified by the relatively moderate radiation levels at the Japanese nuclear plant.

Epidemiologists speak of “stochastic deaths,” those they predict will happen in the future because of radiation or some other risk. With no names attached to the numbers, they remain an abstraction.

But these other deaths were immediate and unequivocally real.

“The government basically panicked,” said Dr. Mohan Doss, a medical physicist who spoke at the Tokyo meeting, when I called him at his office at Fox Chase Cancer Center in Philadelphia. “When you evacuate a hospital intensive care unit, you cannot take patients to a high school and expect them to survive.”

Among other victims were residents of nursing homes. And there were the suicides. “It was the fear of radiation that ended up killing people,” he said.

Most of the fallout was swept out to sea by easterly winds, and the rest was dispersed and diluted over the land. Had the evacuees stayed home, their cumulative exposure over four years, in the most intensely radioactive locations, would have been about 70 millisieverts — roughly comparable to receiving a high-resolution whole-body diagnostic scan each year. But those hot spots were anomalies.

By Dr. Doss’s calculations, most residents would have received much less, about 4 millisieverts a year. The average annual exposure from the natural background radiation of the earth is 2.4 millisieverts.

How the added effect of the fallout would have compared with that of the evacuation depends on the validity of the “linear no-threshold model,” which assumes that any amount of radiation, no matter how small, causes some harm.

Dr. Doss is among scientists who question that supposition, one built into the world’s radiation standards. Below a certain threshold, they argue, low doses are harmless and possibly even beneficial — a long-debated phenomenon called radiation hormesis.

Recently he and two other researchers, Carol S. Marcus of Harbor-U.C.L.A. Medical Center in Los Angeles and Mark L. Miller of Sandia National Laboratories in Albuquerque, petitioned the Nuclear Regulatory Commission to revise its rulesto avoid overreactions to what may be nonexistent threats.

The period for public comments is still open, and when it is over, there will be a mass of conflicting evidence to puzzle through.

A full sievert of radiation is believed to eventually cause fatal cancers in about 5 percent of the people exposed. Under the linear no-threshold model, a millisievert would impose one-one thousandth of the risk: 0.005 percent, or five deadly cancers in a population of 100,000.

About twice that many people were evacuated from a 20-kilometer area near the Fukushima reactors. By avoiding what would have been an average cumulative exposure of 16 millisieverts, the number of cancer deaths prevented was perhaps 160, or 10 percent of the total who died in the evacuation itself.

But that estimate assumes the validity of the current standards. If low levels of radiation are less harmful, then the fallout might not have caused any increase in the cancer rate.

The idea of hormesis goes further, proposing that weak radiation can actually reduce a person’s risk. Life evolved in a mildly radioactive environment, and some laboratory experiments and animal studies indicate that low exposures unleash protective antioxidants and stimulate the immune system, conceivably protecting against cancers of all kinds.

Epidemiological studies of survivors of Hiroshima and Nagasaki have been interpreted both ways — as demonstrating and refuting hormesis. But because radiation regulations assume there is no safe level, clinical trials testing low-dose therapy have been impossible to conduct.

One experiment, however, occurred inadvertently three decades ago in Taiwan after about 200 buildings housing 10,000 people were constructed from steel contaminated with radioactive cobalt. Over the years, residents were exposed to an average dose of about 10.5 millisieverts a year, more than double the estimated average for Fukushima.

Yet a study in 2006 found fewer cancer cases compared with the general public: 95, when 115 were expected.

Neither the abstract of the paper nor of a second one published two years latermention the overall decrease. (The authors speculated that the apartment dwellers may have been healthier than the population at large.) The focus instead was on weaker results suggesting a few excess leukemia and breast cancer cases — and on a parsing of the data showing an overall increased cancer risk for residents exposed before age 30.

More recently, a study of radon by a Johns Hopkins scientist suggested that people living with higher concentrations of the radioactive gas had correspondingly lower rates of lung cancer. If so, then homeowners investing in radon mitigation to meet federal safety standards may be slightly increasing their cancer risk. These and similar findings have also been disputed.

All research like this is bedeviled by “confounders” — differences between populations that must be accounted for. Some are fairly easy (older people and smokers naturally get more cancer), but there is always some statistical wiggle room. As with so many issues, what should be a scientific argument becomes rhetorical, with opposing interest groups looking at the data with just the rightsquint to resolve it according to their needs.

There is more here at stake than agonizing over irreversible acts, like the evacuation of Fukushima. Fear of radiation, even when diluted to homeopathic portions, compels people to forgo lifesaving diagnostic tests and radiotherapies.

We’re bad at balancing risks, we humans, and we live in a world of continual uncertainty. Trying to avoid the horrors we imagine, we risk creating ones that are real.