Three weeks pregnant. Coffee makes you gag. Meat smells off in a way it never did before. Vegetables you used to eat — broccoli, brussels sprouts, cabbage — now seem unbearable. The smell of barbecue from the neighbor's yard sends you to the bathroom. Bread is fine. Cereal is fine. Toast is fine. Plain rice is fine. The foods you can still eat are bland. The foods you can't are, with strange specificity, the ones that have until this morning been your favorites.
This is pregnancy sickness, and somewhere between 75 and 89 percent of women experience some version of it during the first trimester.1 About 55 percent will actually vomit. If food aversions are added to the definition, close to 100 percent of pregnant women experience some form of it. The medical name is morning sickness, although it is not confined to mornings. The medical category is illness, although for most of the twentieth century no one knew what was being treated by the misery of the first trimester or why the body was producing it.
Margie Profet's 1992 hypothesis flipped the framing.2 Pregnancy sickness is not an illness in the sense of a malfunction. It is an adaptation — exquisitely calibrated, precisely timed, evolutionarily ancient — designed to keep the developing fetus alive against a specific class of threats. The threat is teratogens: chemicals that cause birth defects when consumed by pregnant women, especially during the first trimester when the fetus's organs are forming. Profet's argument: the foods that pregnant women find aversive correspond to the foods that carry the most teratogens. The timing of the aversion corresponds to the period when the fetus is most vulnerable. The intensity of the aversion correlates inversely with miscarriage risk. The whole picture — aversion content, timing, severity — fits a designed-by-selection mechanism rather than a random pathology.
Buss treats pregnancy sickness as one of the cleanest demonstrations of adaptationist thinking in medicine.3 His framing is direct: "A phenomenon previously regarded as an illness appears to be an exquisitely tailored mechanism designed to combat a hostile force of nature — one that would impair the survival of a child even before it is born." The reframing matters. It changes the clinical default from suppress-the-symptom to respect-the-design.
State the hypothesis precisely.
Profet 1992 proposes that the nausea, food aversions, and vomiting characteristic of first-trimester pregnancy are evolved psychological and physiological adaptations whose function is to prevent maternal consumption and absorption of teratogens — chemicals that disrupt fetal development.2 The mechanism takes specific input (smell and taste cues correlated with toxin presence), runs decision rules (heightened sensitivity, aversion, vomiting if the cues exceed threshold), and produces specific output (avoidance of toxin-bearing foods, expulsion of toxin-bearing foods if consumed).
Three predictions follow.
First, the foods pregnant women find aversive should correspond to those carrying the highest doses of teratogens. The correspondence should not be arbitrary. Foods that don't carry teratogens should not be aversive even if they're disliked for other reasons.
Second, the timing of the aversion should match the period of greatest fetal vulnerability. Fetal organs form during weeks 2-4 after conception, with the sensitive period extending through about week 14. Pregnancy sickness should appear during this window and disappear after it. Aversions before conception or after organogenesis would be evolutionarily irrelevant.
Third, women who experience pregnancy sickness should have better fetal outcomes than women who don't. If the mechanism is functional, its operation should produce measurable benefit in the form of reduced miscarriage rates, reduced birth defects, or improved fetal viability.
The empirical evidence supports all three predictions.4
The food aversion content tracks teratogen content. Tierson, Olsen, and Hook 1985 surveyed 400 pregnant women on what foods they found aversive during the first trimester. The most-aversive foods were coffee (129 women), meat (124), alcohol (79), and vegetables (44). Almost no aversions to bread (3) or cereal (0). The pattern is teratogen-tracking. Coffee contains caffeine and several other compounds. Meat carries fungal and bacterial decomposition products as well as endogenous animal compounds that can be teratogenic. Alcohol is a known teratogen. Vegetables — particularly cruciferous vegetables like broccoli and brussels sprouts — contain allyl isothiocyanate and other secondary plant defense compounds. Black pepper contains safrole, which is both carcinogenic and mutagenic. Rhubarb leaves contain oxalate. The food categories pregnant women avoid are precisely the categories carrying the most chemical defense compounds, while bland starches that don't carry teratogens (bread, cereal) escape the aversion.4
The timing matches organogenesis. Pregnancy sickness typically begins around week 4-5 after conception, peaks at week 8-10, and resolves by week 14-16. The window matches the period of greatest fetal vulnerability to teratogens, which spans weeks 2-14. The mechanism switches on as the fetus becomes vulnerable and switches off as the fetus's organs complete formation. The temporal precision is hard to explain by any non-adaptive theory.
The miscarriage data are the clinching test.5 Yerushalmy and Milkovich 1965 surveyed 3,853 pregnant women on whether they experienced pregnancy sickness during the first trimester and tracked subsequent miscarriage rates. Women without pregnancy sickness experienced spontaneous abortion at 10.4 percent. Women with pregnancy sickness experienced spontaneous abortion at 3.8 percent. The ratio is 3-to-1. Women whose mechanism activated had three times fewer miscarriages than women whose mechanism didn't activate. The pattern is what the embryo-protection hypothesis predicts. The pattern is not what an "illness" interpretation would predict — illness should correlate with worse fetal outcomes, not better.
Profet's hypothesis represents the kind of move evolutionary psychology can make when the framework is applied carefully. A symptom that medical science had treated as malfunctioning was reanalyzed as functioning, with predictions that turned out to hold. The reanalysis matters clinically — drugs that suppress pregnancy sickness without identifying the underlying teratogen exposure may be worsening fetal outcomes by suppressing a useful defense.
The first wave of evidence is the within-population data on aversion content.4 Tierson, Olsen, and Hook 1985 (N=400) and Dickens and Trethowan 1971 (N=100) both found the same aversion pattern. Coffee, meat, alcohol, and vegetables were the most-aversive categories. Bread and cereal were essentially never aversive. The pattern is robust across study populations and matches the toxin-content prediction.
The second wave is the temporal precision. Studies tracking pregnancy-sickness onset and resolution find the window matches organogenesis. The mechanism is not a global "first-trimester illness" — it is a precise window matching when fetal organs are forming. Buss notes that the resolution typically happens around week 14, "coinciding with the end of the sensitive period for organ development."4
The third wave is the cross-cultural evidence.6 Fessler 2002 reviewed 27 traditional societies for the presence of pregnancy sickness. The pattern appeared in 20 of 27 societies. The 20 societies in which pregnancy sickness was observed were significantly more likely to use meat and other animal products, which typically carry pathogens and teratogens at higher rates than plants. The 7 societies without observable pregnancy sickness tended to have plant-heavy diets with less teratogen exposure. The pattern fits the embryo-protection hypothesis: the mechanism is more strongly expressed where the threat is greater.
A separate cross-cultural data point is anecdotal but robust. The !Kung of Botswana, the Efe Pygmies of Zaire, and Australian Aborigines all have explicit cultural recognition of pregnancy sickness. A !Kung woman, Nisa, reported in Shostak's ethnography that her mother knew she was pregnant when she started "throwing up like this" — the recognition is part of the culture's implicit knowledge about reproduction. Pregnancy sickness is not a Western medical category invented by the medical profession. It's a phenomenon that traditional societies have observed and named for as long as we have ethnographic record.
The fourth wave is the miscarriage data. Yerushalmy and Milkovich's 1965 study (N=3,853) is the clinching demonstration.5 Women without pregnancy sickness had 10.4 percent spontaneous abortion rates. Women with pregnancy sickness had 3.8 percent. The 3-to-1 ratio is enormous. It cannot easily be explained by confounding variables — the data control for age and other obvious risk factors. The straightforward reading is that the mechanism produces measurably better fetal outcomes when activated.
A potential complication: Pike 2000 examined 68 Turkana women in Kenya and did not find the predicted pattern of pregnancy-sickness-protective effect.6 The Turkana eat substantial amounts of meat and animal products, so the mechanism should be active. Pike's negative finding doesn't invalidate the broader pattern but does suggest the mechanism isn't equally activated across all conditions, or that some moderator variables haven't been identified.
The pattern of evidence is strong. Aversion content tracks teratogens. Timing tracks organogenesis. Outcome tracks fetal viability. The mechanism is calibrated to a specific adaptive problem with the specificity that EP requires for an adaptation diagnosis.
The hardest tension in the embryo-protection hypothesis is its therapeutic implication: suppressing pregnancy sickness with anti-nausea medications might increase miscarriage risk by removing the protection. The Yerushalmy and Milkovich data suggest the protection effect is large. The clinical literature on anti-nausea medications has not always treated this concern carefully, partly because of the framing of pregnancy sickness as illness rather than function.
Modern obstetric practice has shifted partly toward respect for pregnancy sickness — most clinicians now avoid prescribing anti-nausea medications for mild-to-moderate cases, reserving pharmacological intervention for severe cases (hyperemesis gravidarum) where dehydration and weight loss become serious health threats. The shift is partly driven by adaptationist thinking. The shift has not been complete. The framing of pregnancy sickness as illness still produces over-treatment in some clinical contexts.
A second tension concerns the Pike 2000 negative finding.6 Among the 68 Turkana women studied, the pregnancy-sickness pattern did not appear. The Turkana eat substantial meat, so the prediction would be that pregnancy sickness should be present. The negative finding doesn't invalidate the broader pattern, but it does flag that something in the Turkana case is different. Possibilities: the specific food items don't carry the same teratogens as Western or other ethnographic-record-society foods; cultural reporting practices differ; sample size was small; some moderating variable hasn't been identified. The negative finding is a reminder that the hypothesis is not yet definitively established across all populations.
A third tension is about variation in pregnancy-sickness severity.1 Some women experience mild aversions; others vomit constantly for weeks. The variation is large. The embryo-protection hypothesis would predict that severity should track exposure to teratogens — women with higher dietary teratogen load should have stronger mechanism activation. The empirical work testing this is less developed than work on the basic effect.
A fourth tension is about the source-specific aversion. Why are coffee and meat the strongest aversions? Buss reports the data without exploring what specific teratogens in these foods drive the response. The mechanism appears to be tracking real toxin content, but the specific molecular triggers haven't been fully identified. This is normal for evolutionary psychology — the framework explains the function without requiring the underlying biochemistry to be fully mapped — but it leaves room for skeptical readers to ask whether the aversion is really tracking teratogens or some other dimension correlated with them.
A fifth tension concerns the maladaptive cases. Some women experience pregnancy sickness so severe that they are hospitalized with hyperemesis gravidarum, lose substantial weight, and risk dehydration. The mechanism in these cases is producing real costs that may exceed the embryo-protection benefits. The framework explains why most cases are protective; it doesn't fully explain why some cases overshoot. Possible answers: individual variation in mechanism calibration; modern dietary toxin loads exceeding EEA levels; gene-environment interactions creating mismatches. Each is plausible. None is fully developed.
Profet's 1992 hypothesis is the foundation, and the textbook treats it as the standard account.2 The convergence between Profet and Buss is essentially complete; Buss's treatment is a faithful summary of Profet's argument with updated empirical support. Where Buss adds emphasis is on framing pregnancy sickness as the cleanest example of adaptationist reframing of medical phenomena. The line at 1377 — "A phenomenon previously regarded as an illness appears to be an exquisitely tailored mechanism designed to combat a hostile force of nature" — is Buss's framing rather than Profet's, but the substance is shared.
Yerushalmy and Milkovich 1965 contributed the clinching empirical data.5 Their N=3,853 study predates the embryo-protection hypothesis by 27 years; the data were collected for other reasons and only later recognized as supporting Profet. The 3-to-1 ratio of miscarriage rates between women without and with pregnancy sickness is the cleanest single piece of evidence.
Fessler 2002 contributed the cross-cultural ethnographic review.6 Twenty of 27 societies show pregnancy sickness; the seven without show plant-heavy diets that correlate with lower teratogen exposure. The cross-cultural pattern fits Profet's prediction. Fessler's work is also where the empirical complications appear — Pike 2000's Turkana data don't fit, and Fessler is the source for the careful framing of cross-cultural variation.
Flaxman and Sherman 2000 contributed independent confirmation across additional societies and a refined analysis of which specific teratogens drive aversion.6 Their work converges with Profet on the basic hypothesis while extending it to specific food categories and toxin types.
Where Buss is silent and the vault has work to do: the broader implications for psychiatric symptom interpretation. If pregnancy sickness can be reanalyzed from illness to function, what other psychiatric symptoms might admit similar reanalysis? Nesse and Williams's 1994 Why We Get Sick extends the adaptationist frame across medicine. The vault could connect pregnancy sickness to broader Darwinian-medicine pages on fever-as-defense, anxiety-as-monitoring, depression-as-disengagement. Each is a candidate for the same reframing — symptom recategorized as function — though each requires its own empirical case.
A separate tension runs to the broader framework of evolutionary medicine. Nesse, Williams, and others have built a research program around adaptationist reframing of medical phenomena. The framework predicts that some "diseases" are actually defenses (fever, cough, vomiting after toxin ingestion, pregnancy sickness, immune response inflammation), some are defects (myopia, lower-back pain), and some are mismatches (obesity, type 2 diabetes, allergies). Profet's pregnancy-sickness hypothesis is the textbook case of the first category. Buss's textbook foregrounds it. The vault's medical and psychological pages should be cross-referenced to build out the broader framework.
The reframing move that Profet makes — symptom reanalyzed as function — is a transferable diagnostic principle. The principle: when a phenomenon looks like malfunction, ask whether it might instead be functioning correctly under conditions different from the ones it was tuned for. Apply the question systematically. Some apparent malfunctions are real malfunctions. Some are functions that have become invisible because the conditions they were tuned for have changed.
The connection to medicine is direct. Fever was treated as illness for most of medical history; modern Darwinian-medicine work treats fever as functional immune response. Cough was treated as illness; the cough is the lungs clearing pathogens. Vomiting after toxin ingestion was treated as illness; the vomiting is removing the toxin. In each case, the symptom is the body doing protective work. Suppressing the symptom suppresses the protection. Treatment that aimed at the symptom rather than at the underlying cause caused harm in some cases. The pregnancy-sickness reanalysis is part of a broader Darwinian-medicine project that the vault should track.
The connection to clinical psychology is parallel.7 Anxiety disorders, depression, and addiction admit at least partial reanalysis along Profet's lines. Anxiety as evolved monitoring system firing on modern inputs that don't match the EEA. Depression as evolved disengagement signal pulling someone away from low-payoff effort. Addiction as reward-prediction system firing on engineered stimuli. Each requires careful empirical case-making. The diagnostic principle — function-not-malfunction — is the connecting thread.
The connection to existing vault pages: somatic trauma theory hub and cortisol-HPA axis stress cascade both work on phenomena that look like malfunction (chronic stress response, persistent trauma symptoms) but admit partial reframing as functions firing on inputs they weren't tuned for. The reframing doesn't dissolve the suffering. It changes the treatment frame. Treatment that respects the underlying function and seeks to address the inputs (rather than suppressing the response) tends to work better than treatment that targets the symptom.
The insight neither domain generates alone: the standard medical and psychiatric framing of "symptom" obscures the question of whether the phenomenon is functioning. Pregnancy sickness was illness for centuries, then reanalyzed and now treated as function. The same reanalysis is partially under way for fever, anxiety, depression. The vault's medical and psychiatric pages get sharper when each phenomenon is annotated with the question "is this a malfunction, a function under mismatched conditions, or a real defect?" Different answers imply different interventions.
A second handshake runs to behavioral-mechanics and the question of when to interfere with someone else's evolved mechanism. A propagandist trying to suppress someone's gut-level objection to a course of action is suppressing what may be that person's evolved mechanism reading the action correctly. A salesperson trying to overcome buyer's remorse is suppressing what may be the buyer's evolved mechanism reading the purchase correctly. The pregnancy-sickness reanalysis suggests caution about manipulation that targets a specific felt response. The response might be a malfunction worth overriding. It might be a function the manipulator is unhelpfully suppressing.
The connection to the manipulation and influence hub: defensive analysis of manipulation should ask whether the felt resistance the manipulator is trying to overcome is itself an evolved protective response. Sometimes it is. The vault's BM defense pages get sharper when each manipulation technique is annotated with what protective response it tries to suppress.
The insight neither domain generates alone: the felt experience of resistance to manipulation is often an evolved mechanism doing exactly what it should. Suppressing it is the manipulator's goal. Recognizing this changes both how the target experiences the resistance and how the target chooses to respond to the manipulation.
The Sharpest Implication.
Some of what your culture treats as illness or pathology in your body and mind may be your body or mind doing exactly what it should. The default medical framing — symptoms are problems to be eliminated — is often correct, but not always. Sometimes the symptom is a defense that selection installed. Sometimes the suffering is the price of a protection working correctly.
This does not mean you should refuse medical treatment. It does mean that before suppressing a symptom, you can ask what the symptom might be doing. Pregnancy sickness, fever, cough, vomiting after a bad meal — each of these has been reanalyzed at least partially as function. The default of "make this go away" is not always the right default. Sometimes "let this run its course while addressing the underlying cause" is wiser.
The harder version of the insight applies to psychological symptoms. Anxiety, depression, intense aversion to particular foods or situations, strong moral disgust at certain practices — each of these is at least partly evolved machinery doing work. Some of the work is calibrated to current conditions and is doing real protective work. Some of the work is mismatched — calibrated to EEA conditions that no longer apply. Distinguishing which is which is one of the harder pieces of self-knowledge, and the medical and psychiatric professions are still building the diagnostic frameworks to do it well.
The corollary: respect for evolved mechanisms is not the same as fatalism about them. A pregnant woman whose pregnancy sickness becomes hyperemesis gravidarum needs medical intervention; the mechanism is overshooting in ways that cost more than they protect. A person with severe depression that prevents normal functioning may need treatment even if some depression is functionally adaptive disengagement. The framework helps distinguish when to intervene and when to respect the mechanism. It doesn't say "always respect" or "always intervene." It says "ask which case this is."
Generative Questions.
Pregnancy sickness is the cleanest case of the function-not-malfunction reanalysis. What other clinical symptoms admit similar reanalysis with comparable evidence quality? Fever has good empirical support. Anxiety is more contested. Depression is even more contested. Where does the evidence currently stand on each?
The Profet reanalysis worked because the predictions (food content tracking teratogens, timing tracking organogenesis, outcome tracking miscarriage rate) were specific enough to be testable. What does the equivalent test set look like for psychological symptoms? If anxiety is functional, what specific predictions follow that wouldn't otherwise be obvious? The empirical work testing such predictions for psychiatric symptoms is less developed than the work on pregnancy sickness.
If clinical practice should distinguish "function under mismatched conditions" from "real malfunction," what diagnostic protocol does the distinction? Most practicing clinicians don't have a clear method for asking this question. The Darwinian-medicine framework supplies the question; the methods to answer it are still being developed.