"I don't know."

"Neither did I," Hildebrand said. "After the dinner last month I went home with a headache and tossed and turned all night. I woke up knowing something was very wrong. You've got a group of men in their late fifties and early sixties, you're going to have some losses. Death is going to start making in-roads.

"But it seemed to me we were way over the probabilities. My mind kept coming up with different explanations, and I decided the first thing to do was find out if my sense of things was accurate. So I called up a fellow I know who's always trying to sell me more life insurance and told him I had an actuarial problem for him. I ran the numbers for him and asked him what percentage of deaths you'd expect over that span of time in a group like that. He said he'd make a couple of calls and get back to me. Take a guess, Matt. How many deaths would you expect in a group of thirty?"

"I don't know. Eight or ten?"

"Four or five. There ought to be twenty-five of us left and instead we're down to fourteen. What does that say to you?"

"I'm not sure," I said, "but it certainly gets my attention. The first thing I'd do is ask your friend another question."

"That's just what I did. Tell me your question."

"I'd ask him to gauge the significance of a sample with three or four times the expected number of deaths."

He nodded. "That was my question, and he had to call somebody to find out. The answer that came back to me was that sixteen deaths out of thirty was remarkable, but it wasn't significant. Do you know what he meant by that?"

"No."

"According to him, the sample's too small for any result to be significant. We could have one hundred percent surviving or one hundred percent dying and it wouldn't really signify anything. Now if we had the same percentage in a substantially larger group, then it would mean something from an actuarial standpoint.