Subjects varied their estimates of the calorie content of a food depending on the assumed negative or positive healthful qualities of the food item they had previously been shown--with weird consequences. Cynthia Graber reports