Chapter 37 — Exercises

The exercises in this chapter are different from those in most of the book. The lab bench is not the stove; it is the screen, the package, and the table. The skills we are building here are the skills of a careful reader of food claims — claims you will encounter every week of your adult life, in the news, on packages, from friends, in your doctor's office, on social media.

A note before we begin. Some readers come to nutrition with a history of disordered eating, or with bodies that have been the target of weight stigma in medical or social settings. If a particular exercise is going to make your relationship with food worse, skip it. The goal of this chapter is to reduce anxiety about food, not produce more of it. If a "calculate your macros" exercise is bad for you, do not do it. The threshold concept of the chapter — most claims you read in headlines are weaker than the headline implies — does not require any specific lab.

🍳 Lab 1 — How to read a nutrition headline

This is the central data-literacy exercise of the chapter, and Pat Hammond's classroom version, expanded.

Time: 60–90 minutes (one sitting, or three sittings of about 20 minutes if you're going to do all three headlines). Materials: Internet access, pen and paper or a text file, three nutrition headlines from the past two weeks (your choice of source). Try to pick from at least two different outlets — say, one from a major newspaper and one from a popular health website. Allergens / safety: None. ⚠️ The only safety note is mental: do not let yourself spiral on any one headline. Move through the exercise.

Protocol.

1. Pick three headlines. Try for variety: one about a single nutrient ("study says coffee linked to lower stroke risk"), one about a dietary pattern ("Mediterranean diet beats keto for heart health"), and one about a specific food ("avocados linked to weight loss," that kind of thing). If you cannot find three, two will do.

2. For each headline, find the original study. This is the hardest part. The article almost always names the journal or the lead researcher, but rarely links directly. Try the following, in order: - Search the article for the word "published" or "in the journal." Note what's named. - Try Google Scholar with the lead author's surname plus a keyword from the topic. - If you cannot find the study itself, find the press release (often on the institution's website) — it will usually link or quote the abstract. - If even that fails, mark this headline as "study not found" and move on. (Note: this is itself information. Headlines that don't link to their source are weaker than ones that do.)

3. For each study you find, fill in the following table.

4. Compare the honest one-sentence version to the headline. Where did the headline strengthen, simplify, or distort the finding? Is the headline true? Misleadingly true? False?

5. Score the headline 1–5 on faithfulness to the underlying study. Five is "perfectly honest"; one is "wildly misleading." Most will land at 2–3.

Expected results. You will, on average, find that headlines significantly overstate the certainty and the effect size of the studies they describe. You will probably find at least one study where the surrogate marker measured (e.g., a blood lipid level) does not actually correspond to the disease the headline says is reduced. You will find at least one industry-funded study whose conclusion conveniently aligns with the funder's interests.

Troubleshooting. If you cannot find any of the original studies, your data point is the failure itself: you cannot evaluate a headline whose source you cannot trace. Trust those headlines correspondingly less.

🍳 Lab 2 — Reading a nutrition label honestly

Time to translate "X grams of Y per serving" into something a human can use.

Time: 45 minutes. Materials: Three nutrition labels from foods in your pantry. Try for variety: one ultra-processed food (a packaged snack, a frozen meal, a meal-replacement shake), one minimally processed food (rolled oats, dried beans, plain yogurt, peanut butter with a 1- or 2-ingredient label), and one beverage (a sweetened drink, a flavored milk, a kombucha — anything with a label). A calculator or a phone calculator app. Allergens / safety: No food consumed. Just labels.

Protocol.

1. Read the serving size first. This is the trick. Many "single serving" packages contain two or three servings. A 20-oz (590 mL) bottle of soda may have 2.5 servings on the label, with each nutrition number multiplied at consumption. Note the serving size in grams or milliliters and what fraction of the package this represents.

2. For each label, calculate the calories per 100 g (per 100 mL for liquids). This normalizes across products. Most whole foods land between 50 and 400 cal/100 g. Most ultra-processed snacks land between 400 and 600 cal/100 g (very energy-dense). Beverages are different — a sugar-sweetened drink might be 40–50 cal/100 mL while plain milk is around 65 cal/100 mL and "diet" drinks are near zero.

3. Calculate the percent of calories from fat, carbs, and protein. Multiply: fat grams × 9 = fat calories. Carb grams × 4 = carb calories (this includes fiber, but fiber's actual contribution is less; for our rough work, this is fine). Protein grams × 4 = protein calories. The three should add up to roughly the total calories per serving (some difference is normal due to rounding and to alcohol if present). Calculate each as a percent.

4. Look at the ingredient list. How many ingredients? Are there ingredients you cannot pronounce or recognize? Note: "cannot pronounce" is not a reliable indicator of "bad" — pyridoxine is just vitamin B6 — but it is one signal that you are looking at a heavily formulated food. The NOVA classification we discussed in the chapter (page 240's framework) defines "ultra-processed" as foods made primarily from substances extracted from foods (sugars, oils, starches), modified ingredients, and additives. If the ingredient list reads like a chemistry inventory, you are in NOVA group 4.

5. Look at the front of the package. What claims are made? "Heart healthy," "high in fiber," "low in fat," "natural," "real fruit," "no high-fructose corn syrup." Compare each claim to what is actually inside. The mismatch is often instructive.

Expected results.

• The ultra-processed food will likely be calorie-dense (high cal/100 g), with a high carb-to-fat ratio engineered for palatability, low fiber, and a long ingredient list.
• The minimally processed food will likely be less calorie-dense, with simpler ingredients and (often) more fiber.
• The beverage will likely have a serving size that doesn't match the actual container size, and added sugars that look reasonable per serving but become significant when you multiply by the actual servings consumed.

Discussion within the lab. Which of the three did you find most surprising? Which one did the front-of-package marketing distort the most?

🍳 Lab 3 — Evaluating a "superfood" claim

Pick a food currently being marketed as a "superfood." Recent candidates: açaí, goji berries, chia seeds, kale, blueberries, matcha, turmeric, kefir, kombucha, coconut oil, MCT oil, bone broth. Choose one.

Time: 60 minutes. Materials: Internet access and the same skills from Lab 1.

Protocol.

1. Find five health claims about this food. Check the marketing on the food's package, the websites of three brands selling it, and a few wellness articles.

2. For each claim, search for the underlying study. Same protocol as Lab 1. How many of these claims have any peer-reviewed source? How many have a high-quality peer-reviewed source (RCT or meta-analysis)? How many are based on test-tube studies, animal studies, or single small human studies?

3. Compare to a non-superfood control. Look up a comparable, non-marketed food in the same category. Compare blueberries to blackberries (or to a frozen mixed-berry blend). Compare matcha to ordinary green tea. Compare chia seeds to ordinary flaxseed or sesame seeds. Are the claimed benefits unique to the "superfood," or do they exist in much cheaper, less-marketed alternatives?

4. Calculate the cost per gram of the claimed beneficial nutrient. A bag of goji berries at, say, $20 / 200 g compared to a bag of dried cranberries at $4 / 200 g. The marketing premium is rarely matched by a meaningful nutritional premium. (When it is — sometimes it is, for a specific compound — note that.)

Expected results. In nearly every case, the "superfood" claim is either (a) a real but modest property that exists at similar levels in cheaper foods, (b) a property demonstrated only in test-tube or animal studies and not yet in humans, or (c) a property of a single compound that exists in the food at concentrations far below what the studies used.

Troubleshooting. If you find a "superfood" that genuinely does have a unique, well-evidenced benefit, write it up. Those exist (ground flaxseed for cholesterol; oily fish for omega-3s; cocoa flavanols at high enough doses), and recognizing the difference between real and marketed is the whole point of the exercise.

Discussion questions

1. The replication crisis. John Ioannidis showed in 2013 that fifty random ingredients from a cookbook had each been claimed, in published research, to either cause or prevent cancer. What does this tell us about the design of nutrition studies and the incentives in publishing? How should a science teacher communicate this to students without leaving them cynical about all science?

2. Reading a study. A friend sends you a headline: "New study shows 30% lower stroke risk with daily cup of beet juice." Walk through the questions you would ask before believing it. What features of the study would make it credible? What features would make it weak?

3. Cross-cultural balance. The chapter showed that every culture independently paired a grain with a legume to balance amino acids. Pick a cuisine you grew up with — or one a friend or partner grew up with — and identify three or four other examples of "accumulated nutritional wisdom" that exist in that tradition without being formally articulated as nutrition science. (Examples to seed thinking: the use of fermented vegetables alongside heavy meals; the citrus-with-fish pairing; the cooked-tomato-with-fat pairing for lycopene absorption; the iron-rich foods served with vitamin-C-rich foods.)

4. The Mediterranean question. PREDIMED is one of the most-cited nutrition trials of the 21st century. It tested a Mediterranean dietary pattern in a Spanish population at high cardiovascular risk and showed about a 30% reduction in major cardiovascular events. (a) What are the strengths of this study compared to most observational nutrition research? (b) What are its limits — what populations and questions does it not address? (c) Why would a randomized trial of "the West African pattern" or "the traditional Korean pattern" be much harder to fund and run, even though the underlying patterns may be equally healthful?

5. Industry funding. A study finds that whole milk is associated with lower cardiovascular risk than skim milk. The study is funded by the dairy industry. (a) Does this mean the study is wrong? (b) How should you weight it? (c) If the same finding emerged from a government-funded study, would your weighting change? Why or why not?

6. Body positivity and behavioral framing. The chapter argues for focusing on behaviors within your control — eating patterns, movement, sleep, stress, social connection — rather than on a number on a scale. How is this different from "eat less, move more"? What is the science behind the difference?

7. Talking to a family member. Maya's father has been diagnosed with type 2 diabetes, and his doctor has handed him a glycemic-index handout. He calls Maya for advice. He has been eating a traditional Nigerian diet — pounded yam, jollof rice, egusi soup, plantain, eba — for sixty years. The handout categorizes most of these foods as "high glycemic index" and says to avoid them. Help Maya think through how to respond. What does the science actually say? What are the trade-offs? Where is the doctor right and where is the doctor missing context?

8. The ultra-processed food finding. Kevin Hall's 2019 NIH study found that subjects ate about 500 calories per day more on a calorie-matched ultra-processed diet than on a calorie-matched whole-food diet. (a) Why is this a striking finding? (b) What hypotheses does the chapter list for why this happens? (c) What kinds of follow-up studies would help adjudicate between the hypotheses? (d) How does this finding interact with the "calories in, calories out" framework — does it contradict it, complicate it, or both?

9. Eating disorders and nutrition discourse. The chapter explicitly names eating disorders as a population for whom much nutrition messaging is harmful. (a) Why is this true — what mechanisms are at play? (b) How might a health educator communicate evidence-based nutrition information to a general audience without harming the subset of that audience with disordered eating histories? (c) What is the line between giving information and giving rules?

10. The Aroon perspective. Aroon's grandmother fed three generations on what she could grow and trade for. She did not read studies. What does this say — about the limits of formal nutrition science, about the genuine knowledge of attentive cooks across cultures, and about the relationship between the two? (Bonus: write down one piece of food wisdom you got from someone in your life — a parent, grandparent, friend — and ask: does the formal nutrition science support this wisdom, contradict it, or have nothing to say about it?)

Advanced sidebar expansions

The Bradford Hill criteria, in practice

We listed the nine criteria in the chapter. Here is how to apply them, with a worked example.

Take the claim: eating processed meat increases colorectal cancer risk. Run the criteria.

• Strength. The relative risk reported by the World Health Organization's IARC working group in 2015 is about 1.18 per 50 g/day of processed meat. That is small. Caveat: small effects are more likely to be confounded.
• Consistency. The finding has been seen in multiple cohort studies in different populations (American, European, Asian).
• Specificity. The effect appears specific to colorectal cancer, less so to other cancers.
• Temporality. Cohort studies establish that the meat-eating preceded the cancer — yes.
• Biological gradient. Higher consumption is associated with higher risk in a roughly monotonic way.
• Plausibility. N-nitroso compounds and heme iron mechanisms have been demonstrated in vitro.
• Coherence. The finding is coherent with the broader literature on cured-meat compounds and colorectal cell biology.
• Experiment. No randomized trial has tested processed meat reduction. (You cannot assign humans to eat 50 g/day of bacon for thirty years.)
• Analogy. Other dietary carcinogen studies (aflatoxin, alcohol) follow similar patterns.

Result: most criteria are met, but the absence of a randomized trial and the small effect size means the claim is probable and cautionary, not certain. The IARC classification of processed meat as Group 1 ("carcinogenic to humans") refers to the strength of evidence, not to the strength of risk. Tobacco is also Group 1, and tobacco is enormously more harmful per unit. The classifications collapse a distinction Bradford Hill kept clear.

The Women's Health Initiative re-analysis

The 2006 publication of the WHI Dietary Modification Trial — finding no significant cardiovascular or cancer benefit from a low-fat dietary intervention in 49,000 postmenopausal women — was one of the most consequential nutrition studies of the early 2000s. Critics of the trial pointed out (correctly) that the intervention group never achieved the targeted dietary changes; by year three the dietary differences between groups were small. Defenders of the trial pointed out (also correctly) that this is exactly the real-world condition under which long-term dietary advice is given, and the trial therefore speaks to the effectiveness (not just the efficacy) of low-fat dietary advice. Both are right. The lesson is that long-term dietary intervention trials are hard to do and hard to interpret, and any single result needs to be set against the broader literature.

Sodium: the unresolved debate

We sent you back to chapter 3 for the sodium discussion, but a quick update on the state of the evidence is worth having here. The major position of WHO, AHA, and most national health agencies is that population-level sodium reduction would reduce cardiovascular events. The evidence base is mixed. Strong support: people with hypertension generally see blood-pressure reductions on lower-sodium diets. Mixed support: the long-term cardiovascular outcomes of population-level sodium reduction. Some recent meta-analyses (notably the PURE study, 2014) have found a J-shaped curve, with both very low and very high sodium intakes associated with worse outcomes. The PURE study has critics. The debate is unresolved. The reasonable position for a non-hypertensive person eating a varied diet is: do not stress about it. The reasonable position for a hypertensive person is: follow your doctor's guidance. The sodium-reduction policy debate is not the same as the individual-eating-decision debate.

🥖 Mastery Food Checkpoint

Where this chapter lands for each of the five tracks.

• Bread track. Whole-grain bread has reasonably good evidence for cardiovascular and metabolic benefits relative to refined-grain bread. Sourdough has some evidence for lower glycemic response than commercial yeasted bread of the same flour, due to lactic acid lowering pH and slowing starch digestion. Bread is not medicine; bread is also not poison. If you have followed the bread track, you can now make whole-grain bread that beats commercial whole-grain bread on every dimension that matters — and that is, plausibly, the version most worth eating.
• Cheese track. Cheese is fermented dairy concentrated in fat, protein, salt, and sometimes a great deal of sodium. The evidence on dairy and cardiovascular health has shifted — the older blanket recommendation against full-fat dairy has weakened, and some studies suggest fermented dairy (yogurt, kefir, aged cheeses) may have neutral or even modestly beneficial associations with metabolic health. Eat cheese for the joy of cheese; pay attention to portion as you would for any calorie-dense food; do not panic.
• Chocolate track. Cocoa flavanols at high doses have shown small cardiovascular and cognitive benefits in trials. Eating commercial chocolate to capture those benefits is a poor strategy — the dose required is much higher than commercial chocolate provides, and the sugar load works against the benefit. Dark chocolate (70%+) once a day is reasonable as a small pleasure with a hint of measurable benefit. The "chocolate is health food" headlines are overclaim. Eat chocolate because it is excellent.
• Fermented vegetables track. The Stanford 2021 trial gave fermented foods their best evidence yet — a 10-week intervention increased microbiome diversity and reduced inflammation markers compared to a high-fiber control. The mechanism (live cultures plus fermentation byproducts plus fiber-microbe interactions) is plausible. Eat fermented vegetables for the flavor and the variety and the modest health correlation. Do not buy "probiotic" supplements with billion-CFU labels; the evidence does not support them outside specific medical indications.
• Coffee track. Coffee is one of the most-studied beverages in the human diet, and the verdict is more positive than the older "coffee causes cancer" hypothesis would have predicted. Up to about 400 mg caffeine per day (roughly 4 cups of brewed) is associated, in multiple cohort studies, with small reductions in cardiovascular and neurodegenerative disease risk. Caffeine sensitivity varies; pregnancy guidelines recommend lower intakes. The coffee track is the track most reassured by the evidence: enjoy your coffee, in moderation, without guilt.

What we are NOT learning in this chapter (and why)

We are not learning rules. We are not building a meal plan. We are not optimizing macros. We are not picking a diet.

We are learning a posture: skeptical, generous, careful. The posture is portable. It will serve you in 2030 when the saturated-fat verdict has shifted again. It will serve you in 2040 when whatever now holds the status of "Mediterranean is best-studied" is being challenged by data from cuisines we haven't yet measured. It will serve you next Thursday when a friend forwards you a "groundbreaking study" headline and you can read the actual study and understand it.

The next chapter looks at where food is going — cultured meat, precision fermentation, the kitchen of 2050. The chapter you just finished is the one that will help you read the future as well as the present.