The closer we look at the overeating hypothesis, the more counterintuitive its logic becomes. Consider a thought experiment. The subjects are two middle-aged men of similar height and age. One eats three thousand calories a day and is lean. The other eats three thousand calories a day and is obese. (The epidemiologic and metabolic studies of the past century make clear that we could find two such men with little difficulty.) Let’s cut the calorie intake of our obese subject in half and semi-starve him on fifteen hundred calories a day. He will lose weight, although, if Albert Stunkard’s 1959 analysis holds true, there’s only one chance in eight he’ll lose even as much as twenty pounds. Our lean subject will lose weight on this diet as well, as Keys demonstrated with his conscientious objectors in 1944. That’s what the law of energy conservation implies. But they would both be hungry continuously, making it likely they would fall off the diet given time. That’s what common sense, the history of obesity research, and the Carnegie, Minnesota, and Rockefeller experiments tell us. And after some amount of weight loss, their weight will plateau, because their metabolism and energy expenditure will adjust to this new level of calorie intake. “Eventually, calorie balance is re-established at a new (low) plateau of body weight and the calorie deficit is zero,” as Keys explained.

Our intuition is that our obese subject will lose more weight because he has more to lose, but we have little evidence to that effect, one way or the other. And yet, if both our obese and lean subjects fall off the diet and return to eating three thousand calories a day, the obese individual will return to obesity, perhaps even fatter than ever, and thus will satisfy our diagnostic criterion for a character defect; our lean subject will also put back the weight he lost, and perhaps a little more, but will still be lean, and will not have to think of himself as possessed of a perverted appetite or some other character defect.