The case against

Every date in this chapter is an inference, and the methods disagree. Oxygen isotopes in fossil bones and teeth point to raised metabolism in the middle to late Permian. Ridges in fossil noses once suggested a Late Permian start too; they probably anchored turbinates, thin scrolls of bone that cut water loss from breathing. But turbinate size in living mammals does not track metabolic rate or body temperature, and five scanned Triassic skulls of mammal relatives show no bony turbinates. The inner-ear method rests on fluid viscosity measured in only eight species. Critics of the tooth-ring argument note that birds and bats live long despite high metabolic rates; the tooth team replied that fliers should be analyzed separately. A minority argues that warm blood is ancient, shared by the ancestors of mammals, birds and crocodilians. The fairest reading is that warm blood was assembled in stages, and "when" depends on which part you measure.

How often did anyone actually run prey down? Travis Pickering and Henry Bunn argued that the tactic is rare in anthropologists' accounts of foragers and confined to very open, very hot habitats. Over 20 years and well over 100 successful Hadza bow hunts, they wrote, "we have never observed persistence hunting by the Hadza, nor have informant interviews revealed this as a strategy they practice." By their count, only two of Liebenberg's 10 hunts were spontaneous; the rest were prompted for filming, started from a vehicle and allowed water refills, and only three of those succeeded. Leaving out the hunts that would not have happened without the cameras, Liebenberg counted four kills in five attempts. Lieberman and colleagues replied that recent foragers, with bows, spear-throwers and nets, are poor stand-ins for early Homo, a trap they called "the tyranny of ethnography." Tracking, they wrote, "must be as old as hunting." Hundreds of historical accounts show the tactic was widespread, not that it was common. And the Namib hunters mostly walked; even !Nam!kabe told Liebenberg that a hunter could run a kudu down while walking some of the time. If walking suffices, persistence hunting may predate Homo's running anatomy and so explain it less well.

The gut-for-brain swap is the weakest link in this story; the measured rise in energy use stands on firmer ground. Cooking's timing is also unresolved, and meat is one alternative. From chewing experiments, Zink and Lieberman estimated that a diet one-third meat would save nearly two million chewing cycles a year, and slicing and pounding more. Cooking, they noted, "was apparently uncommon until 500,000 years ago". Yet a synthesis that corrected for uneven sampling found no sustained rise in evidence of meat-eating in eastern Africa after Homo erectus appeared. And at two French sites, Neanderthals left hearths in warm periods but almost none in cold ones. How early Homo paid its bill remains open.

The case against farming is strong on average and weak as a universal law, and skeletons are a biased record. Under the so-called osteological paradox, a person who dies quickly of an infection leaves no mark on the bone. A cemetery full of lesions can therefore mean a population strong enough to survive illness rather than a sicker one. Growing populations also fill cemeteries with children, and that skews estimates of how long people lived.

Evolutionary explanations are easy to invent. In 1979 Stephen Jay Gould and Richard Lewontin faulted the adaptationist program, the habit of explaining each trait as an adaptation, for accepting speculative tales on plausibility alone. Konner and Eaton recalled saying from the start that their evidence "could only suggest testable hypotheses." Lieberman's own table of about 50 conditions, from acne to flat feet, lists "suspected" mismatch diseases. Suspects are not convictions.

The chair's case is weaker than its reputation. The Hadza rest as long as industrialized people do, so sitting alone cannot be the whole story. In pooled data on more than a million people, those who sat more than eight hours a day but were highly active had no clear excess mortality. High levels of moderate activity, an hour or a little more a day, "seem to eliminate the increased risk of death associated with high sitting time," the authors concluded. Five or more hours of daily television, though, stayed linked to higher risk even in the most active. Motion-sensor studies point the same way.

The barefoot boom promised fewer running injuries, and the trials have not borne that out. A Cochrane review of 12 trials found that "most evidence demonstrates no reduction in lower-limb running injuries in adults when comparing different types of running shoes", though most of it was of low or very low certainty. About 40% of runners are injured during follow-up, and injury rates do not seem to have shifted despite decades of changing shoe design. Another review found only limited, backward-looking evidence linking a non-heel strike to fewer injuries, and would not recommend changing strike for an uninjured heel-striker.

Hardened arteries are not new. Before the Tsimane study, the Horus team CT-scanned 137 mummies spanning more than 4,000 years, from ancient Egypt and Peru to the Ancestral Puebloans and the Unangan hunter-gatherers of the Aleutian Islands. A third had probable or definite atherosclerosis, among them three of the five Unangan. That finding "raises the possibility of a more basic predisposition to the disease." The prehistoric Tyrolean Iceman, Ötzi, had calcified arteries too, and a genetic predisposition to coronary heart disease (Chapter 4).

Part of the brain-washing story is solid. Cerebrospinal fluid does enter the brain alongside its arteries in animals, and in a small tracer MRI study of patients it reached every region of the human brain. Nearly everything after that is contested: whether fluid sweeps through the tissue or dissolved substances mostly diffuse, where it leaves, and whether sleep speeds clearance or slows it. Human support is thin. The key human tracer study of sleep loss compared seven patients under neurosurgical assessment kept awake with 17 who slept, without randomization or brain-wave recording. A 2026 study supporting sleep clearance relied on a computer model and an experimental in-ear device from its funder, Applied Cognition; seven authors declared stock-option interests.

Nasal breathing's reputation outruns its data in places. The oxygen gain from sinus nitric oxide rests on a handful of people. In a small randomized trial in heart-surgery patients, nasal breathing lowered resistance in the lungs' blood vessels but left arterial oxygen unchanged. No one measured how Paleolithic (Old Stone Age) people breathed, so the popular claim that modern people breathe faster or shallower than their ancestors is a guess.

"Ultra-processed" is a blunt category. French food specialists sorting the same foods into NOVA groups agreed poorly. Critics call the definition too broad to guide choices, and its links with obesity likely confounded: people who eat a lot of it differ in other ways too. A US advisory committee graded the evidence tying ultra-processed diets to body fat as "Limited." Yet the 2025-2030 US dietary guidelines told Americans to "Avoid highly processed" foods that are salty or sweet.

Microbiome science has a hype problem. Of 38 published studies that transplanted human gut microbes into rodents, 36 reported transferring a disease, a rate the reviewers called "implausible." Much of the case that microbes shape mood or weight rests on such animals. Diversity is no simple health score: in healthy women, harder, slower stools went with more species. And consumer tests disagree. Given the same reference stool, the companies' results differed about as much as different donors' stools do. The Hadza are modern people, not ancestors. By one estimate, fewer than 200 still follow the traditional foraging life, and their diversity belongs to a whole way of life, hardships included, that few outsiders would choose (Chapter 38).

Carbohydrates or calories? The carbohydrate-insulin model holds that high-glycemic diets, heavy in foods that raise blood sugar fast, trigger a hormonal response that drives fat storage and then overeating. Its authors include Daniel Lieberman, whose work runs through this book. The rival energy balance model answers that the brain is the main organ that regulates body weight. In a tightly controlled test at the NIH Clinical Center, 20 adults eating freely took in 689 kcal a day less on a low-fat plant-based diet than on a ketogenic (very low-carbohydrate) animal-based one. A feeding trial found higher energy spending on a low-carbohydrate diet, but critics fault its analysis. In a year-long trial, healthy low-fat and healthy low-carbohydrate diets did about equally well.

The dose makes the poison, mostly. In 1538 the physician Paracelsus wrote that "Solely the dose determines that a thing is not a poison." In a modern reappraisal, Grandjean adds that timing and genetic susceptibility matter too, and that missing data are no evidence of safety. He calls PFAS "a clear and unfortunate example." For diethylhexyl phthalate (DEHP), a plastic softener, the worry is exposure in the womb. A systematic review rated the evidence "moderate" in humans, and strong in animals, that such exposure shortens baby boys' anogenital distance, the span from anus to genitals (Chapter 32).

The sweetener evidence splits by study type. In the review the WHO commissioned, randomized trials linked higher sweetener intake to slightly lower body weight, under a kilogram. Cohort studies (which follow groups of people over time) linked sweetened drinks to a 23% higher risk of type 2 diabetes. Both findings were rated low certainty. Cohorts carry a trap: people who are gaining weight may switch to diet drinks because of it.

Genes matter. In twin studies, estimates of heritability (the share of differences between people explained by genes) run from 35% to 81% for crowding of the lower teeth. Overbite and crossbite (upper teeth biting inside the lower) look mostly environmental. The evidence is rated low certainty. Among the Arara of the Amazon, one inbred village was founded by a single expelled couple. It has nearly double the malocclusion, or faulty bite, of its parent village, despite similar tooth wear. Kahn and colleagues call jaw change "much too fast to be evolutionary." That rules out new mutations, but not old genes behaving differently in a new environment.

Much hydration science is funded by drink companies, and the papers say so. Four authors of the bladder-infection trial in the Handrail worked for Danone Research, and two occasional Danone consultants wrote the editorial on the eight-glass mix-up. A review favoring planned drinking in long, hot exercise appeared in a supplement funded by the Gatorade Sports Science Institute, a division of PepsiCo. Funding does not make results wrong, but ask who gains from "drink more."

The hygiene hypothesis is widely misread as a case against cleanliness. In a UK online poll, nearly a quarter of the public agreed that children need exposure to germs to build a strong immune system. Sally Bloomfield, the hygiene researcher who reported it, calls the idea that broad exposure to germs boosts general immunity to infection "a myth." Infections do not seem to protect against allergy. Promoting handwashing in day care centers and schools in rich countries prevents about a third of bouts of diarrhea. The protective exposures in this chapter are animals, siblings, farm dust, outdoor microbes and early foods, not dirty hands or skipped vaccines.

Vitamin D is this chapter's clearest case of an association outrunning its trials. Low levels travel with obesity, poor health and less education, any of which could explain a link to disease. The Institute of Medicine judged the evidence for cancer, heart disease, diabetes and autoimmunity "inconsistent, inconclusive as to causality." A genetic workaround, Mendelian randomization, treats inborn gene variants that nudge vitamin D as a natural experiment. One such study found deaths climbing steeply below about 20 ng/mL. Its method was later questioned, and the authors' own reanalysis weakened the case. Testing boomed anyway: US Medicare vitamin D tests rose 547% in under a decade. What would change the picture is a large trial in people who are truly deficient.

"Let fevers run" is only half right. The animal case is strong but has a ceiling. In Kluger's rabbits, survival rose with fever only up to a point. In the same study, rabbits given fever reducers died less often, perhaps through some other effect of the drugs. In intensive care, infected patients whose temperature peaked around 39°C had the lowest adjusted death rates, yet lowering fever in trials changed nothing. In one trial in septic shock, where infection has sent blood pressure crashing, cooling cut deaths at two weeks from about a third to about a fifth. A plausible reading: a patient who can mount a fever has reserves, and in the sickest, fever's energy cost can outweigh its benefit.

Pain reprocessing therapy rests on one full trial of mostly white, college-educated volunteers with low-to-moderate pain, treated by experts in the method. Two authors have commercial ties to it, and the team calls for trials in more severe pain. Education alone disappoints. Across a dozen trials, its short-term effects on pain and disability were of "low clinical relevance." For new back pain at high risk of lingering, it left pain no lower than sham education did. Exercise eases pain moderately and function only a little, and no trial can hide from people whether they are exercising. Central sensitization is real but often invoked loosely. Woolf himself has called for diagnostic criteria and objective biological markers.

Some of the rise in cancer reflects living longer and looking harder. Three in five US diagnoses come at 65 or older. The US cancer death rate, adjusted for age, has fallen by about a third in three decades, "largely because of smoking reductions, earlier detection, and improved treatment." Screening can conjure epidemics. In South Korea, thyroid cancer diagnoses rose fifteenfold in under two decades while deaths held steady. Its analysts called this a sure sign of overdiagnosis: finding cancers that would never do harm. Deaths from the fastest-rising early-onset cancers have stayed flat overall, so Patel, Adamson and Welch argue that much of that rise appears to reflect closer checking and overdiagnosis. Young adults' colorectal cancer deaths, though, are climbing.

Did electric light steal our sleep? Less than the story says. Across half a century of studies, measured adult sleep showed no association with study year, and the US short-sleep rate has held steady recently. Children, whose sleep has shortened slightly year by year, are the exception. Pre-industrial sleep was not long or pristine either. What electricity reliably does is make sleep later.

The main objection is genetic: myopia runs in families, and twin studies put its heritability at 84 to 90%. True, and largely beside the point: heritability describes the spread within one population under one set of conditions, not why a whole population changed. "Gene pools simply do not change that fast," as the myopia researcher Ian Morgan puts it. He adds that most children who became myopic in the epidemic did not have myopic parents. Geneticists on the other side hold that heritability of 60 to 90% still makes genes the primary driver. A fair reading: genes set who is susceptible, and environment how many.

The sauna story rests almost entirely on one Finnish cohort, analyzed again and again by one group. Its comparison group is men who took a sauna once a week, not never, and only about 200 men sat in the top group. Letters in reply ran under the title "The link between sauna bathing and mortality may be noncausal." The Japanese study is independent support, and also observational.

The link between hearing loss and dementia is strong as a statistic and unresolved as a cause. Almost all the evidence is observational, and two other explanations are hard to rule out. One is reverse causation: early, undiagnosed dementia can itself reduce how well a person hears, or reports hearing. The other is shared damage: harm to blood vessels could separately injure the cochlea, the inner ear's hearing organ, and the brain's memory regions in the medial temporal lobe, where dementia begins. Reviewers increasingly call hearing loss a potentially modifiable risk marker rather than a confirmed causal factor, and ACHIEVE's null overall result fits that reading. Hearing aids remain worth having for hearing. Nobody should buy them expecting a dementia vaccine.

The central objection has never gone away: healthier, wealthier people live in greener places and visit parks more. The designs built to beat it only half succeed. Identical twins kept the depression association but lost stress and anxiety. Danish adjustment shrank the extra ADHD risk by nearly two-thirds. Movers gained, yet a panel study of more than 10,000 people called individual effects small. Believe the direction; stay modest about the size.

"Cortisol is bad" is a cartoon. Part of the hormone's job is keeping the alarm from overshooting. In the cold studies, cortisol levels did not predict who got sick; receptor resistance did. In Whitehall, morning cortisol and the waking rise predicted nothing about death.

This chapter's claims are not equally strong.

The biggest open question is whether phones and social media lie behind the rise in teenage mental illness, and the evidence does not yet settle it. In The Anxious Generation (2024) Jonathan Haidt argues that between roughly 2010 and 2015 a play-based childhood gave way to a phone-based one. He calls that shift a major cause of a sharp rise in teenage depression, anxiety and self-harm, hitting girls hardest. His remedies: delay access to smartphones and social media, make schools phone-free, and allow more unsupervised free play. The rise itself is well documented in national surveys. The causal claim is the contested part.

Evolutionary psychiatry has a testability problem, and this chapter has been living inside it. The three evolutionary accounts of depression (social defeat, problem-solving rumination, infection defense) were each fitted to existing observations after the fact, and none has been set against the others in an experiment that could rule any of them out. Adaptive stories are cheap to build and expensive to test. Treat them as frames for asking questions, not as findings.

Concealed ovulation is the weakest link in this chapter. Alexander and Noonan's pair-bonding model and its two rivals, paternity confusion and alliance building, all hang together. No comparative test in humans has decided among them. Treat the whole topic as an open question dressed in confident prose, including the prose above.

The temptation in a chapter like this is to file cesareans, formula and hospitals under unnatural and stop thinking. Resist it. The differences are real and specific, but none of them shows that children born by cesarean or fed formula end up worse off. The founding study of birth and microbes followed none of its 10 babies past the first day, and measured no health outcome at all.

Every generation reports that children no longer play the way they used to. This alone should slow you down. Gray's case is also lopsided: the mental-health side of his trend is measured with standard instruments and repeated samples, while the play side rests on softer survey and observational material with no single clean measure. The widely quoted 50% fall in outdoor play has no traceable source and should be treated as folklore until one appears.

Several companies now sell "biological age" tests to consumers, built on clocks like Horvath's or DunedinPACE. Those clocks were developed and validated as research tools for comparing large groups. No evidence reviewed for this chapter shows that moving your personal score by changing your habits predicts anything about your health. Working for groups does not make a test precise for one person. Chapter 43 has more on tracking a thing in place of changing it.

Fries's compression hypothesis is still a hypothesis. Has illness across whole populations been compressed, expanded, or settled into what demographers call a dynamic equilibrium, worse health but slower dying? That remains an open argument in the field, not a settled result in Fries's favor. The Garmany and Terzic data point toward expansion. But they are national, ecological figures, averages for whole countries rather than individuals followed through life. And health-adjusted life expectancy is itself a modeled quantity, built on disability weights (scores for how far each illness falls short of full health) that reasonable people dispute. A widening gap measured this way is a real signal. It is not a verdict.

Every finding in this chapter describes groups and odds; none of them predicts an individual. Someone with plenty of ancestry from a lactase-persistent population can still be lactose intolerant. Someone with no Tibetan variants can acclimatize to altitude over days to weeks by ordinary routes. The book's thesis, that the environment is the intervention, works inside ranges shaped by genes, not in spite of them.

The main risk in a chapter like this is the one it warns about. "One cause of everything" is a seductive shape for an argument, and chronic inflammation is the current favorite candidate. The first edition of this book ended this chapter by declaring that the mismatch was one problem expressed through thirty systems. That overstates what anyone has shown. Furman's group itself places inflammation among interacting causes, not beneath them. The defensible claim is narrower: these systems are linked, and they can be moved together.

Almost everything in this chapter is observational: it compares people who already live differently rather than assigning them at random. People who walk 8,000 steps, take stairs at speed and lift twice a week differ from people who do none of those things in income, illness, optimism and a hundred unmeasured ways. Randomized trials of exercise exist in quantity, but they run for months and measure blood pressure or fitness, not decades and death. The direction is consistent enough to act on. The precision is not what the round numbers imply.

Four soft spots, in order of how much weight people put on them.

The two largest numbers in this chapter come from observational research and deserve the discount that implies. Okun and colleagues flag the healthy-volunteer problem themselves, and adjustment shrinks it without proving the remainder causal. The purpose finding has the mirror-image weakness: undiagnosed illness can lower reported purpose before it kills, producing the association with no causal path at all. Say "associated with" here, and mean it.

This chapter is no verdict against self-tracking, and neither is Baron's paper. Baron and colleagues frame orthosomnia as a clinical judgment call, weighing doubts about the device against a patient's fair interest in their own data. For a diagnosed sleep disorder, supervised weight management or a chronic disease, tracking is a targeted tool and often a valuable one. The trouble starts when measurement has no question attached, or when it replaces how you feel instead of informing it.

Every information technology has set off this worry, and the first recorded version is the best. In Plato's Phaedrus, Socrates argues that writing will produce forgetfulness, because those who learn it will trust outside marks instead of their own memory. He was right about the mechanism and wrong about the verdict: the Sparrow result confirms his objection in a laboratory, about a technology nobody proposes to give up. Slide rules and pocket calculators drew the same complaint from arithmetic teachers, who expected mental arithmetic to decline, and the sky held.

The urban evidence in this chapter is the weakest kind in the book, and its force as a story far outruns its methods. Barcelona's 667 lives a year come from a health-impact model of a plan only partly built, not from counting deaths in a finished system. Pontevedra's percentages are the city's own, from a city government with a case to make. Neither city was compared with a randomized or even a matched control city, a test close to impossible in urban policy. A mid-sized Spanish city with a walkable medieval core and a mild climate is a friendly test case. Nothing here shows that the policy transfers to a larger, poorer or more car-dependent place.