Weighted network of foods, grouped by co-occurrence
Some of our most human moments can be expressed in food: an 11 a.m. whiskey soda, a late night chocolate bar, a skipped lunch. These occasions are often private, hidden in the mundane details of day-to-day existence. If you want to understand how people actually eat, you have to uncover these myriad choices.
Anyone who has tried to keep a food journal knows that this isn’t easy. Try it for yourself right now: try to catalogue what you ate yesterday. Breakfast, lunch, and dinner might spring to mind. What about dessert? Anything to drink? Was there milk or sugar in your coffee? Midnight snacks? Quick trips to the refrigerator just to see what’s in there?
This memory-based method of study forms the basis for much of the past fifty years of nutrition research, but whether or not we can view this kind of dietary data as scientifically valid is hotly debated. Even so, it’s being used to inform dietary guidelines, which in turn alter the way people eat—a cycle of measurement, recommendation, and re-measurement.
Here are some statistics about what Americans say they eat. On any given day, one in eight Americans will eat pizza, and nearly half will have a sandwich of some kind. Half of Americans consume milk at breakfast. A quarter of our food intake will be in the form of snacks.
These results are from the US National Health and Nutrition Examination Survey (NHANES). Since the 1960s, the United States Center for Disease Control has been running NHANES to track the overall health of Americans. Its goal is straightforward, yet grand: “assess the health and nutritional status of adults and children in the United States,” providing the foundation for US government “nutrition surveillance.”
In what is now a continuous study, researchers at the National Center for Health Statistics (part of the CDC) interview and measure new cohorts of 5,000 demographically representative Americans every year. That data is used to provide recommendations to groups like the US Department of Agriculture, who in turn decide things like what kinds of health claims are allowable on dietary supplements and what goes into the (now defunct) food pyramid.
The pages that follow show some of the food trends that a cursory glance at NHANES provide. If we were to believe that the data were totally accurate, it would confirm the stuff of stereotypes: men drink beer, women eat yogurt, older people eat oatmeal, healthy people eat dried fruits and berries. Recently, however, NHANES has come under fire for the what some are calling “inadmissible” evidence in rigorous science: memory-based dietary measurements, the kind that you tried a few pages ago.
These methods introduce difficult-to-control biases. People forget what they’ve eaten, or perhaps they want to please an interviewer with additional information that might not have actually happened. It's well known that people routinely make up memories entirely. When you tell the interviewer that you had a coffee in the morning, they are instructed to respond with "Did you add anything to your coffee?" Suddenly you might wonder—did I?
To draw a parallel with criminal justice, over time we’ve replaced much of witness testimony with the comparatively more reliable DNA forensic evidence. That kind of transition isn’t likely to happen in nutritional epidemiology. Just knowing they’re part of a study causes people to eat differently (often better, since someone’s watching), making it difficult to gather unbiased evidence.
A more subtle effect might also be at play: the role that food plays in our self-image. We all have beliefs (whether conscious or not) about what we eat. We want to believe that we’re happy with the food choices we make, which for many of us equates to treating our bodies well. We think of ourselves as members of one community or another, each of which has different food preferences that become part of our identities.
It would be going too far to say that this effect biases the NHANES data beyond use. It’s an important, imperfect view into what Americans want to think that they ate. The hazard is in taking it at face value. So long as those at the USDA and others that use this data recognize the flaws in it, we might avoid the cycle of nutrition misinformation that, just to name a few examples, demonized cholesterol in eggs and just now is awakening to the role of excessive sugar in the obesity crisis.
Whatever the future of NHANES holds, there’s a beauty in what it’s recorded so far that goes beyond the raw science of nutrition. It catalogues part of our sense of self—both in the choices we make, and in how we choose to explain them to others. The pages that follow are a collection of some of the stories that emerge when you step away from averages and begin to look at the days of individual Americans.