How many data points do you need? The rarity of an accident is part of statistical comparisons, which is why most people don't worry that much about an asteroid hitting their city.

We have been running hundreds of nuclear power plants around the world for decades. There have been about 27 serious civilian nuclear reactor incidents in the last half century. Of these, you can count on one hand those that had any significant radiation release off the plant site itself.

The anti-nukes keep telling us how many could die. The bar chart shows how many have actually died. That's not a bell curve. That's history, and many years of it. The writer is postulating some kind of accident that is worse than Chernobyl, worse than Japan, and infinitely worse than Three Mile Island.

You want a statistic? The article linked to my original link talks about 1 million deaths per year around the world from coal air pollution. Estimates for the number of deaths from the Hiroshima bombing, including later effects, are as high as 200,000. So one year of coal seems to equal five actual nuclear attacks on a city. How many nuclear reactor accidents per year would be needed to kill 1 million people?

We're watching right now a nuclear recovery in Japan from a worst-case scenario, and then some. It stands a good chance of winding up with no radiation-related deaths at all. Frankly, the loss of the plant's electricity to a disaster-ridden Japan might end up killing more Japanese than the radiation.