Twenty obese inpatients who claimed to crave carbohydrate-rich foods were given d-fenfluramine (15 mg p.o., twice daily) or its placebo, double-blind, for two consecutive eight-day periods. Food choices were measured on treatment days 1, 7, and 8 by giving the subjects access to unlimited portions of six isocaloric meal foods (three high in carbohydrate and three high in protein) and of 10 isocaloric snack foods (five high in protein and five high in carbohydrate) available 24 hours a day in a computerized vending machine. d-fenfluramine reduced mealtime calorie intake by only 16% (from 1940 +/- 94 to 1630 +/- 92; p < .001), mealtime carbohydrate by 22%, and had no significant effect on mealtime protein consumption; in contrast, snack calorie intake was reduced by 41% (from 707 +/- 97 to 414 +/- 46; p < .001), and snack carbohydrate intake by the same proportion. The mean number of carbohydrate-rich snacks consumed per day decreased from 5.8 +/- 0.8 to 3.4 +/- 0.4 (p < .01), while that of protein-rich snacks failed to change signficantly (i.e., from 0.7 +/- 0.2 to 0.5 +/- 0.2).