"He didn't do it," Fenton said. "I knew that the instant I started questioning him, but I had to make sure, beyond the faintest possibility of a doubt. All of the things he told me--and at least one-third of the details in his signed confession--are completely wide of the mark. He wasn't there, didn't shoot her, couldn't even tell me how she'd looked in falling fifteen seconds after the shot was fired. How she must have looked, I mean, for her head and shoulders to strike the desk the way they did, upsetting an ashtray, and to be consistent with the position she was in when Miss Prentiss found her. And the body wasn't moved--not a half-inch--until we arrived. "Other things--a dozen, at least. I had the impression he wasn't even sure just where the bullet had entered her head. Oh, a murderer can be mistaken about many things, way off in some respects. But not quite that far off. Not nearly that far off, in fact, when he's studied everything published in the papers about the case, and has been in that office many times. There were details we didn't tell the reporters, details he couldn't possibly have known and he was completely wrong about all of them, with one or two exceptions. And that only strengthens the case against him--the very damaging evidence which proves that he wasn't the killer and couldn't possibly have been." Fenton nodded, his fingers still tapping on the file. "You'd naturally expect him to be right about one or two things, even if they didn't appear in the papers. He's no dumb-bell. He couldn't have written those juvenile delinquency articles and foster-fathered a major movie if he was. He'd naturally be pretty good at guessing games. It's always the exceptions which strengthen, lend real weight, to that kind of rule." Fenton cleared his throat. "It's like in medicine. You take a very rare, unusual kind of disease. Say there are twenty symptoms which are very diagnostic of that particular disease. But one of them only occurs in one case of the disease out of a hundred. And the patient has that one symptom, along with the others. Now ... let's say that in the whole United States, in the course of the year, only about two hundred people die of that rare disease. "Don't you see what I'm driving at Gallison--or do you? It would have to mean that only about two people die from that disease with that particular symptom attached to it every year out of a nation of a hundred and seventy-five million people. According to the law of averages, how likely would the patient be to have the disease? The very symptom which does sometimes occur in connection with the disease--but rarely--which you might think would strengthen the diagnosis, actually helps to weaken it, to make it so unlikely as to practically eliminate the possibility that the patient could have it."