A decade ago, as part of a study on diet, psychologist Tom Baranowski was asked to recall everything he had eaten the previous day. A chicken dinner, he said confidently, remembering that he had prepared it for himself and his wife Janice. The thing was, he hadn’t made chicken that night. It was only later that he realized he’d treated himself to a hamburger.
If Baranowski, who studies children’s diets at Baylor College of Medicine in Houston, Texas, was an unlikely candidate for making such a mistake, consider how abysmal the dietary memories of everyone else must be. By observing his study subjects one day and following up the next, Baranowski has found that children routinely forget about 15% of the foods they have eaten, and more than 30% of the foods they do recall turn out to be figments of their imagination. Adults show similar patterns. “The errors of dietary assessment are overwhelming,” says Baranowski.
These mistakes are more than a reminder of the human memory’s fallibility: they threaten to undermine the foundations of modern medical epidemiology. In this field, researchers make associations between past events and experiences, and later ones such as the emergence of cancer or other diseases. But if the initial records are inaccurate, these associations can be weak, misleading or plain wrong. Although the problem is most jarring in studies of diet, it also infects investigations of exercise, stress, pollution or smoking — basically, anything that relies on people reporting their own exposures through interviews or questionnaires. “This is the weak part of epidemiology,” says Paolo Vineis, an environmental epidemiologist at Imperial College London.