This same confounding of calories and carbohydrates might also explain the success stories attributed to low-calorie diets—Albert Stunkard’s one patient in a hundred, as he reported in 1959, who lost as much as forty pounds and managed to keep it off. It’s effectively impossible to restrict calories significantly without also reducing the carbohydrates. Any calorie-restricted diet that restricts all calories equally, restricts carbohydrates, too. Even diets that preferentially reduce fat will have to reduce carbohydrates to achieve a significant reduction in calories (unless the dieters are willing to sacrifice protein in fish and meat, for instance, in order to avoid the fat that accompanies it). If dieters avoid sweets and snacks, and if they drink sugar-free soda but not regular soda, they’re reducing their carbohydrate consumption significantly, and they’re changing the type of carbohydrates they consume. Any benefit may be due to the calories reduced, or the carbohydrates, or even just the relative absence of sugar.

Another issue that complicates this issue of calorie versus carbohydrate restriction is that the effect of weight-loss diets changes over time. The modest benefits of semi-starvation slowly diminish with time, as the calorie restriction induces a compensatory inhibition of energy expenditure. Moreover, much of the initial weight loss comes from losing water, not fat (see Chapter 8). Because of this “tendency to retain water on a carbohydrate diet and to give it out on a rich fat diet,” as Dunlop described it, restricting carbohydrate calories specifically will induce a more dramatic and immediate loss of water. Testing diets for only a few weeks will demonstrate that carbohydrate-restricted diets induce weight loss at a greater rate than calorie-or fat-restricted diets, but whether they induce fat loss at a greater rate is a different question. “Changes in body-weight are to be taken, therefore, as of significance only when the experiment continues for a period of several weeks,” as Francis Benedict cautioned in 1910. “Certainly, for short experiments, body-weight is for the most part wholly without significance.”