Carbohydrates were the hardest. There was no test then, nor is there now, for identifying the concentration of carbohydrates in general. But Atwater knew the main organic matter in foods were the three big items, protein, fat, and carbohydrate. He also knew how to calculate the total amount of organic matter. He simply burned the food completely, leaving only the mineral ash that did not burn and was therefore the inorganic part. Knowing how much organic matter the food contained and how much fat and protein it held, he obtained the amount of carbohydrate by subtraction: the weight of carbohydrate was what was left when the weights of the fat, protein, and mineral ash had been subtracted from the total weight of the original food item.
Atwater was thus able to estimate the amount of protein, lipid, and carbohydrate in his food. The second piece of information he needed was how much of the food a person eats is digested, as opposed to being passed through the body unused. This required him to analyze the feces of people who were eating precisely measured diets, which he duly did. He was then able to estimate, for each of the three nutrients, how much of what was eaten was also digested. Once again he found that there was little variation within the categories of protein, fat, and carbohydrate, so he assumed the variation could be ignored.
The chemist now had what he wanted. He knew how much energy each of the three big types of macronutrient contained, how much of each macronutrient was present in the food, and how much of it was used in the body. Ignoring variation within each type of macronutrient, he proposed the convention that still dominates the food industry and government standards. By taking into account the proportion of the food that he found was not digested, which was rarely more than 10 percent, he claimed that on average proteins and carbohydrates each yield four kcal/gram, while lipids yield nine kcal/gram. These are known as Atwater’s general factors.
This simple and convenient system forms the basis of the Atwater convention and is essentially what the U.S. Department of Agriculture’s National Nutrient Database and McCance and Widdowson’s
The Composition of Foods
use to produce their tables of nutrient composition. But nutritionists have long recognized important limitations in the Atwater convention, so it has been modified in various ways. One way was to make the general factors more specific. In 1955 the Atwater specific-factor system was introduced to take advantage of a half century of nutritional biochemistry research. For example, the energy value of different types of protein is known to vary: egg protein produces 4.36 kcal/gram, whereas brown rice protein produces 3.41 kcal/gram, and so on. An exhaustive list of such variants has been compiled.
Modifications have also been made to the systems for analyzing nutrient composition. Atwater assumed that all the nitrogen in a food was part of a protein and that all proteins contained 16 percent by weight of nitrogen. However, nitrogen can be found in other molecules that may or may not be digestible, such as nonprotein amino acids and nucleic acids, and some proteins have more or less than 16 percent of nitrogen. So for several decades Atwater’s general average of 16 percent nitrogen per protein has been replaced by specific figures, such as 17.54 percent for macaroni protein and 15.67 percent for milk protein.
I mention these modifications to the Atwater system to show that nutritionists have been actively engaged in trying to improve it, and to show that the changes they have proposed have on the whole been rather minor. For example, although egg protein produces more kilocalories per gram (4.36) than brown rice protein (3.41), neither figure is very far from Atwater’s estimate of 4 kcal/gram. In fact, although the specific-factor system lends greater precision, the overall effects of the changes are so small that some nutritionists (particularly those in Britain) still prefer to use general factors, albeit modified since Atwater’s time.
The general factors have never been static; more factors have been added over time. Even Atwater modified his own system by separating alcohol into its own category (he gave it a rounded value of seven kcal/gram). Much later, in 1970, a general factor was added for the class of carbohydrates called monosaccharides, or simple sugars. New general factors have also been proposed for dietary fiber (or nonstarch polysaccharides), which are so much less well digested than other carbohydrates that they clearly deserve a lower energy value than four kcal/gram; a figure of two kcal/gram has been proposed. The system has also been modified to allow for energy lost in urine and gas production. These and similar modifications continue to tweak the original Atwater system while retaining its essential philosophy.
The Atwater system is thus a flexible convention that is continuously modified but still provides the fundamental basis for assessment of energy value in today’s foods. It allows people eating ordinary cooked foods to track their caloric intake sufficiently to get a good idea of when they are overeating or undereating. But it has two critical problems that undermine its ability to assess the food value of items of low digestibility, such as raw foods or foods like whole-grain flour with large particles.
The first problem is that the Atwater convention does not recognize that digestion is a costly process. When we eat, our metabolic rate rises, the maximum increase averaging 25 percent. The corresponding figures for fish (136 percent) and for snakes (687 percent) are vastly higher, showing that humans pay less for digestion than other species, presumably due partly to our food being cooked. But the cost of digestion is still significant for humans and can be reduced or raised depending on the food type.
When Atwater burned foods in a bomb calorimeter, he ignored this complexity. He assumed that humans could use all the energy present in a food and digested in the body. If food burns in the bomb calorimeter, Atwater seemed to conclude, it produces the same amount of energy value in our bodies. But the human body is not a bomb calorimeter. We do not ignite food inside our bodies. We digest it, and we use calories to pay for this complex series of operations. The cost varies by nutrient. Protein costs more to digest than carbohydrates, while fat has the lowest digestive cost of all macronutrients. In a 1987 study, people eating a high-fat diet achieved the same weight gain as others eating almost five times the number of calories in the form of carbohydrate. The higher the proportion of protein in the food, the higher the cost of digestion. Based on animal studies, we can expect that the costs of digestion are higher for tougher or harder foods than softer foods; for foods with larger rather than smaller particles; for food eaten in single large meals rather than in several small meals; and for food eaten cold rather than hot. Individuals vary too. Lean people tend to have higher costs of digestion than obese people. Whether obesity leads to a low cost of digestion or results from it is unknown. Either way, the variation is important for someone watching his or her weight. For the same number of measured calories, an obese person, having a lower digestive cost, will put on more pounds than a lean person. Life can be unfair.
Compounding the problem, a second big failure of the Atwater system is closely related and equally important. The Atwater system assumes that the proportion of food digested is always the same, regardless of whether the food is in liquid or solid form, part of a high-fiber or low-fiber diet, or raw or cooked. Recall that one of Atwater’s general factors was the proportion of food that is passed into the feces undigested. He found that this was low—10 percent or less—and he assumed that this proportion was constant. This assumption has long been known to be wrong. When A. L. Merrill and B. K. Watt introduced the Atwater specific-factor system in 1955, they noted specifically that the digestibility of a grain is affected by how finely it is milled. More extensively milled flour might be completely digested, whereas less milling could lead to 30 percent of the flour being excreted unused. So they called for specific data to be applied to the digestibility of every food. Such data, however, are often unavailable. Identifying the digestibility of each food according to its physical state is difficult, because large numbers of experiments are required. To complicate matters further, the digestibility of the same item varies according to the food context in which it is consumed. For example, protein digestibility tends to be lower when the protein is part of a high-fiber food than when part of a low-fiber food. For raw foods, we have only scattered information on how various durations of cooking, down to none at all, influence the proportion of a food that is digested. Very few studies use the only appropriate measure, ileal digestibility, which takes the sample of unused food at the end of the small intestine, rather than at elimination from the body.
All these factors play such an important role in determining the net value of a food item that many nutritionists have called for a major revision of the Atwater convention. But the information needed to account for the effects of variation in the cost of digestion and digestibility is hard to obtain and difficult to incorporate into a food-labeling system. A widespread preference therefore persists among professionals to keep the Atwater general-factor system. Essentially, nutrition science is faced with choosing between the immense effort of accumulating nutritional-value data that are difficult to quantify but accurate, on the one hand, or using easily quantified but physiologically unrealistic measures, yielding only a rough approximation of food value. Given the difficulty of acquiring the actual, contextually adjusted nutritional values of individual foods (and combination of foods), the general public is provided with estimates of food values that do not reflect the realities of the digestive process. The scientists who compiled the National Nutrient Database and
The Composition of Foods
must have known that raw foods would produce less net energy than cooked foods and that a higher proportion of raw food was likely to pass through the body unused. But they were locked into an old, approximate-measurement technique, and the result is a falsehood. The data in standard nutritional tables assume that particle size does not matter and that cooking does nothing to increase the energy value of foods, when abundant evidence shows the opposite to be true.
The physics of food matters because our foods and food processing techniques are changing in ways we can expect to contribute to the obesity crisis—thanks to our inability to assess the real caloric value of our diet. In our grocery stores, we find flour that has been ground ever finer, foods made ever softer, calories in ever greater concentration. Rough breads have given way to Twinkies, apples to apple juice. Consumers are misled by the current food-labeling system into thinking they will get the same number of calories from a given weight of macronutrients regardless of how it has been prepared. People are unlikely to experience consequences of our dietary choices any differently than the snakes that got more food value from eating meat that had been ground up, or the rats that got fat when their food pellets were soft. Only one study has been conducted to test the effect of food hardness on health. It found that Japanese women whose diets were softer had larger waists, which are associated with higher rates of mortality. That was a preliminary study. It will take time to show how consistent such effects are, but the indications are clear. We become fat from eating food that is easy to digest. Calories alone do not tell us what we need to know.
It is time to modify Atwater’s convention to include the effects of the physical structure of foods in estimates of a food’s nutritional value. And we must educate ourselves. As food writer Michael Pollan has argued, we should choose “real food,” not “nutrients.” For Pollan, real food is natural or only lightly processed, recognizable and familiar. By contrast, nutrients are invisible chemicals, such as essential oils and amino-acids and vitamins, objects of scientific expertise whose significance we must take on faith. The less processed our food, the less intense we can expect the obesity crisis to be.
We once thought of our species as infinitely adaptable, particularly in our diet. Different peoples survive on diets that range from 100 percent plants to 100 percent animals. Such flexibility buttresses a notion that human evolutionary success depended merely on inventiveness. Taken to extremes, our species seems to be free to create our own evolutionary ecology.
The cooking commitment says otherwise. The human ancestral environment was full of uniform problems: how to get fuel, how to regulate feeding competition, how to organize society around fires. The big problem of diet was once how to get enough cooked food, just as it is still for millions of people around the world. But for those of us lucky enough to live with plenty, the challenge has changed. We must find ways to make our ancient dependence on cooked food healthier.
ACKNOWLEDGMENTS
I am indebted to many sources, friends, and colleagues who have guided my attempts to understand the significance of cooking. I owe special thanks to Rachel Carmody, NancyLou Conklin-Brittain, Jamie Jones, Greg Laden, and David Pilbeam for collaboration in research. I am particularly appreciative of those who gave editorial and scholarly advice on earlier versions. Dale Peterson, the late Harry Foster, Martin Muller, Elizabeth Ross, and Bill Frucht commented in generous detail. Rachel Carmody, Felipe Fernandez-Armesto, Elizabeth Marshall Thomas, Victoria Ling, Anne McGuire, David Pilbeam, and Bill Zimmerman also kindly read entire drafts. For comments on individual chapters, I thank Robert Hinde, Kevin Hunt, Geoffrey Livesey, Bill McGrew, Shannon Novak, Lars Rodseth, Kate Ross, Stephen Secor, Melissa Emery Thompson, and Brian Wood. For other kinds of support, ideas, and advice, I am grateful to Leslie Aiello, Ofer Bar-Yosef, Dusha Bateson, Pat Bateson, Joyce Benenson, Jennifer Brand-Miller, Alan Briggs, Michelle Brown, Terry Burnham, Eudald Carbonell, John Coleman, Matthew Collins, Randy Collura, Debby Cox, Meg Crofoot, Roman Devivo, Irven DeVore, Nancy DeVore, Nate Dominy, Katie Duncan, Peter Ellison, Rob Foley, Scott Fruhan, Dan Gilbert, Luke Glowacki, Naama Goren-Inbar, John Gowlett, Peter Gray, Barbara Haber, Karen Hardy, Brian Hare, Jack Harris, Marc Hauser, Kristen Hawkes, Sarah Hlubik, Carole Hooven, Sarah Hrdy, Stephen Hugh-Jones, Kevin Hunt, Dom Johnson, Doug Jones, Sonya Kahlenberg, Ted Kawecki, Meike Köhler, Kat Koops, Marta Lahr, Mark Leighton, Dan Lieberman, Susan Lipson, Julia Lloyd, Peter Lucas, Meg Lynch, Zarin Machanda, Bob Martin, Chase Masters, the late Ernst Mayr, Rob McCarthy, Rose McDermott, Eric Miller, Christina Mulligan, Osbjorn Pearson, Alexander Pullen, Steven Pyne, Eric Rayman, Philip Rightmire, Neil Roach, Diane Rosenberg, Lorna Rosen, Norm Rosen, Kate Ross, Stephen Secor, Diana Sherry, Riley Sinder, Catherine Smith, Barb Smuts, Antje Spors, Michael Steiper, Nina Strohminger, Michael Symons, Mike Wilson, Tory Wobber, Brian Wood, and Kate Wrangham-Briggs. For exceptional collegial support, I acknowledge the late Jeremy Knowles, Doug Melton, and David Pilbeam. For calm writing opportunities, I thank the staff of the Weston Public Library (Massachusetts), Alison and Kenneth Ross (Badachro, Scotland), Robert Foley and Marta Lahr (Leverhulme Center for Human Evolutionary Studies, Cambridge, UK), the Medical Library of the University of Cambridge (UK), and the authorities of Kibale National Park, Uganda, where I wrote the proposal for this book during three weeks under a fig tree in April 2001.