Chapter 19 — The Seed Oil Debate: Omega-6, Inflammation, Oxidation, and What Actually Changed

The Hook: Theo's bin

Theo Vasquez threw away four bottles of oil.

Canola, sunflower, a large jug of vegetable oil, and — after some hesitation — the grapeseed oil he'd bought for a recipe once and never used again. About $34 of oil into a bin.

Then he bought a jar of beef tallow for $18**, a bottle of avocado oil for **$16, and — because the video had mentioned it — ghee at $14.

Forty-eight dollars, to replace thirty-four dollars of oil he'd already paid for.

I asked him what he'd expected to change. He said: "Inflammation, I think. It said they're inflammatory."

So I asked the question this chapter is about: inflammation of what, measured how?

He didn't know. And neither, it turns out, does the field — which is a more interesting answer than either of us expected, and it is not the answer the video gave him.

Sugar had a century of research behind it and an argument between qualified people. Seed oils have a smaller literature, a much louder public argument, and — this is the part that makes the chapter difficult — a genuine finding at the centre that almost nobody on either side describes accurately.

⚠️ This chapter will not resolve the debate, because it isn't resolved. What it will do is separate the four things being claimed, show you which one has evidence, and be explicit about which parts I hold with low confidence.

🏃 Fast Track: §19.3 (what actually changed — the strongest fact on the sceptic side), §19.11 (the confounding that swallows everything), §19.13 (what to do). Twenty-five minutes.

🔬 Deep Dive: §19.5 (testing the mechanism), §19.6 (the recovered trials — the best evidence the sceptics have, and I present it in full), §19.9 (smoke points, where the popular argument is backwards).

19.1 What "seed oils" means — and doesn't

The term is not scientific. It's a list, usually eight items, sometimes called "the hateful eight":

Canola (rapeseed) · corn · cottonseed · soybean · sunflower · safflower · grapeseed · rice bran.

What they mostly have in common: high in linoleic acid (LA), an omega-6 polyunsaturated fat; mechanically or solvent-extracted from seeds; refined; industrially produced at scale; and inexpensive.

⚠️ What the category conceals is more interesting than what it captures.

Canola is on the list and is closer to olive oil than to safflower in fatty acid profile. ⚠️ Peanut oil is usually not on the list and has three times canola's linoleic acid. High-oleic sunflower is bred to be low in the compound the whole argument is about, and it's on the list because of where it comes from rather than what's in it.

💡 This is the tell. ⚠️ The category is defined by provenance — seeds, industry, refining — not by composition.

Chapter 18's Case Study 1 named this exact failure: a provenance claim answering a dose question. "Seed oils" tells you where something came from. The argument is about linoleic acid, which varies more than sevenfold within the category and overlaps completely with oils outside it.

That doesn't make the concern wrong. It makes the category the wrong instrument for investigating it — and it's why §19.12's five slots have to be filled before anything here is answerable.

19.2 The claim, stated properly

Four numbered claims, as in Chapter 18 §18.1 — because most of the argument is people answering different ones.

Claim 2 is not in dispute and I want to be clear about that before anything else, because the sceptical side is often treated as though it invented its premises. It didn't. §19.3 is real.

Claim 3 is the argument. Claim 4 is where it's usually sold.

🪞 Learning Check-In: run the check yourself, before I tell you anything

Stop here. You have Chapter 17 §17.14's ten-minute check and Chapter 18's five slots. This is the first live debate you're being asked to assess before the verdicts.

The claim, as you'd meet it:

"Seed oils are inflammatory. They were never part of the human diet until industrial processing made them cheap, and now they're in everything. The omega-6 to omega-3 ratio has gone from about 1:1 to 20:1, and that imbalance drives the chronic inflammation behind heart disease, obesity, diabetes and autoimmune conditions. They're also oxidized and unstable — they turn rancid and produce toxic aldehydes when heated."

Work through it and write your answers down before reading on.

1. Restate it so it could be wrong. ⚠️ Fill the five slots: at what dose · in what form · over what period · in whom · compared to what? Which slots does the claim leave empty?

2. Run the eight tells (§17.1). Which fire? Be honest — including where they don't fire, because this claim is better constructed than most.

3. What rung does the evidence need to sit on for each of the four claims in §19.2? Are they the same rung?

4. What's the comparator? ⚠️ If seed oils are removed, what replaces them? This is the question the claim never states and it changes the answer completely.

5. What would the narrower true version be? Chapter 17 §17.1's interlude question.

6. And name your prior. Do you already believe this? ⚠️ Write down what would change your mind before you read my verdicts — because after you read them it's much harder to know what you'd have said.

Keep your answers. §19.14 asks you to compare them.

19.3 What actually changed — the strongest fact on the sceptic side

⚠️ I'm putting this before the mechanism, and before any verdict, because it's true and because it is routinely omitted by people arguing the mainstream position.

Three real things:

1. Consumption of linoleic-acid-rich oils rose enormously over the twentieth century. Soybean oil in particular went from a negligible part of the US food supply to one of its largest sources of calories. This is a documented change in the food supply, not an inference.

2. ⚠️ Human adipose tissue linoleic acid content rose substantially over the past half-century. This is measured in fat biopsies rather than estimated from food surveys — which makes it one of the better exposure measurements in all of nutrition, because it doesn't depend on anyone remembering what they ate. The concentration of linoleic acid in stored body fat in the US rose markedly across decades.

3. This is genuinely without precedent. ⚠️ No human population before the twentieth century consumed linoleic acid at these levels, because the technology to extract it at scale didn't exist.

💡 Take this seriously, because the rest of the chapter depends on you having taken it seriously.

A tissue-level measurement showing that the composition of human body fat changed over fifty years is not a wellness talking point. It's a finding.

⚠️ And "unprecedented" is not the same as "harmful" — Chapter 17 §17.1's lost-golden-age tell applies, and refrigeration, vaccination and iodized salt are also unprecedented. But "unprecedented" is an excellent reason to investigate, and the people who noticed this first were right to notice it.

What it doesn't establish: whether the change matters, in which direction, at what dose, or through what mechanism. That's §19.4 onward, and it's where the argument stops going the sceptics' way.

19.4 The mechanism: omega-6, arachidonic acid, and the ratio

The argument is mechanistically coherent, which is why it persuades people who understand biochemistry. Here it is, fairly.

Linoleic acid (18:2 n-6) is an essential fatty acid — you must eat some; deficiency causes real disease. The requirement is small, on the order of 1–2% of energy.

In the body, LA can be elongated and desaturated to arachidonic acid (AA), which sits in cell membranes and is the precursor to a family of signalling molecules — prostaglandins, thromboxanes, leukotrienes — many of which are pro-inflammatory.

Alpha-linolenic acid (ALA, 18:3 n-3) is the plant omega-3, and it's converted to EPA and DHA — precursors to a different, largely less inflammatory or actively resolving set of signalling molecules.

⚠️ And here's the competitive step that makes the ratio argument work: both pathways use the same enzymes, particularly delta-6 desaturase. More LA competing for the enzyme means less ALA converted to EPA and DHA.

So the argument runs: LA intake up → competition for desaturase up → EPA/DHA down and AA up → the balance of eicosanoid signalling shifts toward inflammation → chronic low-grade inflammation → chronic disease.

The ratio figures usually quoted are that ancestral diets ran somewhere near 1:1 to 4:1 omega-6 to omega-3, and modern Western diets run 15:1 to 20:1.

Every step in that chain is a real biochemical relationship. ⚠️ Which is exactly why §19.5 matters so much — because a mechanism this clean generates a prediction, and the prediction has been tested.

The fact the "toxic" framing has to explain away

⚠️ Linoleic acid is essential. Not "beneficial" — essential, in the specific technical sense Chapter 13 established: your body cannot synthesize it, and if you don't eat any, you develop a deficiency disease.

This isn't theoretical. Essential fatty acid deficiency was documented in patients maintained on fat-free intravenous nutrition before lipid emulsions were standard, and in infants on inadequate formulas. The presentation is characteristic:

• Scaly, dry dermatitis — the skin's barrier function depends on specific LA-derived lipids
• Impaired wound healing
• Increased susceptibility to infection
• Hair loss and growth failure in infants

It resolves when linoleic acid is provided, which is what makes it a deficiency rather than a coincidence.

💡 ⚠️ A framework that calls linoleic acid a toxin has to account for the fact that removing it entirely causes a named clinical syndrome.

The coherent version of the sceptical position accepts this and argues about dose — that the requirement is 1–2% of energy, that intakes are now 5–8%, and that the excess is the problem. That's a real argument and it's the one §19.12 asks to see specified.

The incoherent version calls it poison, which requires ignoring the deficiency literature entirely. ⚠️ Chapter 18's threshold, one more time: the dose makes the poison — and an essential nutrient is the clearest possible demonstration that "is it toxic?" is the wrong question.

And it should tell you something that the requirement is easily met. A tablespoon of most seed oils covers it several times over; so do nuts, seeds, poultry and eggs. ⚠️ Nobody in an industrialized food supply is deficient in linoleic acid — which is why the argument is about excess, and why the sceptics who frame it that way are arguing the defensible version.

19.5 Testing the mechanism

This is the section that moved me, and it's the one most consistently absent from popular coverage in both directions.

The mechanism makes two testable predictions. Both have been examined.

Prediction 1: raising LA intake should raise inflammatory markers

It's directly testable. Feed people more or less linoleic acid in controlled conditions, measure inflammatory markers — CRP, IL-6, TNF-α and others.

⚠️ Systematic reviews of controlled human trials have generally not found that increasing dietary linoleic acid raises circulating inflammatory markers. The trials are mostly modest in size and duration, and the markers are proxies rather than disease — both genuine limitations. But the prediction is clear, it has been tested repeatedly, and it has generally not been confirmed.

Prediction 2: raising LA intake should raise tissue arachidonic acid

⚠️ This is the more interesting failure, and the more decisive one.

Increasing dietary linoleic acid does not produce a proportional rise in tissue arachidonic acid. The conversion step is regulated and saturable — the enzyme system does not simply pass more substrate through when more arrives. Adipose LA rises with intake, as §19.3 documented. Membrane AA largely doesn't follow.

💡 Aha moment. ⚠️ The chain in §19.4 has a broken link, and it's the link the whole argument depends on.

The chain runs: more LA → more AA → more inflammatory eicosanoids → disease.

The first arrow is well documented. §19.3 measured it in fat biopsies. ⚠️ The second arrow is where it fails — the body regulates that conversion rather than performing it in proportion to supply.

This is Chapter 6's lesson arriving again in a different system: a pathway existing is not the same as a pathway being the rate-limiting step. A biochemistry diagram shows you what can happen. It does not show you what does happen, or how much, or whether anything regulates it.

⚠️ And this is exactly the mistake Chapter 13's antioxidant supplements made — a real mechanism, a clean diagram, an intervention that didn't work.

What about the ratio itself?

The ratio framing has a further problem, separate from the above: ⚠️ a ratio can be changed at either end, and the two are not equivalent.

Lowering omega-6 and raising omega-3 both improve the ratio. They are different interventions with different effects, and the evidence for raising omega-3 intake is considerably better than the evidence for lowering omega-6. ⚠️ Reporting a ratio hides which one you did — which is a general problem with ratios and not specific to fats.

A worked illustration. Two people both move from 16:1 to 6:1.

⚠️ The same ratio, achieved two ways, with completely different physiological meaning — and the number on the page is identical.

This matters because ALA-to-EPA conversion in humans is poor — a small percentage, lower still for DHA, and lower in men than women. ⚠️ So "improving your ratio" by cutting omega-6 does not reliably put more EPA and DHA in your tissues. Eating fish does.

🔬 Claim → Evidence → Verdict

The claim: "Most people would benefit from more omega-3 in the form of EPA and DHA."

What the evidence shows: Regular oily fish consumption is associated with lower cardiovascular mortality across cohorts, is a component of the dietary patterns with the best outcome evidence (Chapter 10 §10.5), and supplies preformed EPA and DHA that bypass the conversion bottleneck. ⚠️ The supplement trials are considerably more mixed than the food evidence — a familiar pattern by now (Chapter 13, Chapter 16), and one that should make you suspicious of anyone selling capsules.

Verdict: 🟢 Probably true — as food. ⚠️ Two portions of oily fish a week is the cheapest, best-evidenced intervention available in this entire chapter, and it is at the opposite end of the ratio from the argument everyone is having.

🔬 Verdict on Claim 3 — the omega-6:omega-3 ratio drives chronic inflammation and disease: 🟠 Probably false as stated. ⚠️ The mechanism is real and the prediction it generates has not survived direct testing. (I hold this less firmly than the ❌s in Chapter 17 — the trials are modest and the markers are proxies. But this is what testing a mechanism looks like, and the result was not what the mechanism predicted.)

19.6 The recovered trials

⚠️ The strongest evidence the sceptical side has, and I'm going to present it properly rather than mention it.

Two randomized controlled trials were run decades ago, testing exactly the intervention at issue: replacing saturated fat with linoleic-acid-rich vegetable oil. Both were incompletely reported at the time. Both had their original data recovered and reanalyzed and published in the BMJ — the Sydney Diet Heart Study reanalysis in 2013, and the Minnesota Coronary Experiment in 2016, both led by Christopher Ramsden and colleagues.

What they found:

• ⚠️ The intervention worked on the intermediate outcome. Replacing saturated fat with linoleic acid lowered serum cholesterol, as expected.
• ⚠️ It did not reduce mortality. In these trials, the cholesterol-lowering did not translate into survival benefit, and in some analyses the intervention groups fared worse.

This is real, it was published in a major journal, and it is uncomfortable for the mainstream position. ⚠️ Anyone presenting the seed oil debate without mentioning these trials is not giving you the argument.

The counter-arguments, which are also real:

1. ⚠️ The oils used were not the oils in question. The Sydney study's intervention used margarine of the era, which was high in industrial trans fat — a substance Chapter 9 established as genuinely harmful and now largely removed from the food supply. A trial of trans-fat-containing margarine is not a trial of canola oil.

2. Design problems. High dropout, institutionalized populations with short exposure periods, incomplete follow-up, and outcome ascertainment that wouldn't meet current standards.

3. They are outliers against a larger body of evidence (§19.7), and a single recovered trial that disagrees with a meta-analysis is a reason to examine both, not to discard the larger one.

4. ⚠️ And the recovery cuts both ways as a lesson. These trials went unreported for decades — which is a real publication-bias scandal, and Chapter 2 §2.8's warning about the file drawer, vindicated. A person who cites this as evidence that the literature is unreliable is making a fair point that applies to all of it, including the parts they like.

🔬 Verdict: 🟡 Unclear / it depends — and this is a genuine 🟡, not a hedge.

The trials are real, the reanalyses are competent, and they did not find what the diet-heart hypothesis predicted. They also used oils that no longer exist in the food supply and have methodological problems that would prevent publication today.

⚠️ The honest reading is that they weaken confidence in the strong version of the replace-saturated-fat-with-vegetable-oil recommendation, and do not establish that linoleic acid is harmful — which is a smaller conclusion than either side wants.

19.7 What the broader evidence says

Against those two trials sits a much larger body of work.

On the intermediate outcome — well established: ⚠️ replacing saturated fat with polyunsaturated fat lowers LDL cholesterol. Chapter 9 §9.5's substitution table. This is not contested by anyone.

On events: meta-analyses of randomized trials replacing saturated fat with polyunsaturated fat have generally found reduced cardiovascular events, with moderate certainty — Cochrane's assessments are worth reading directly for how carefully they qualify this.

⚠️ Note "moderate certainty," and note what it means: further research could change the estimate. This is not a settled question presented as settled; it's a moderately-supported conclusion that the recovered trials legitimately complicate.

And Chapter 9's framing still governs: ⚠️ the question is never "is this fat good or bad." It's "replacing what with what." Replacing saturated fat with refined carbohydrate looks worse than replacing it with polyunsaturated fat, which looks similar to or slightly worse than replacing it with monounsaturated fat. The comparator is the finding.

19.8 Oxidation and heat — the real concern, stated precisely

⚠️ This is where the sceptical side is largely right, and where the concern gets mis-stated into something false.

The chemistry: polyunsaturated fats have multiple double bonds, and double bonds are where oxidation happens. Heat, light, oxygen and time all promote it. Oxidation products include aldehydes — among them 4-hydroxynonenal (4-HNE) and related compounds — which are genuinely reactive and genuinely undesirable.

⚠️ This is real chemistry with real toxicology behind it, and it is not a wellness invention.

But the dose and the form matter enormously, and this is exactly Chapter 18's threshold:

💡 ⚠️ Almost all of the alarming oxidation research is about repeatedly heated oil, which is the condition in commercial frying, not the condition of a bottle in your cupboard or a pan you used once.

The claim "seed oils oxidize and produce toxic aldehydes" is true. The implied claim — "therefore the oil you cook with at home is delivering them" — does not follow from the studies making the first claim.

⚠️ And notice where this points: at deep-fried takeaway food, which is also ultra-processed, also energy-dense, and also the thing §19.11 is about. The concern converges on the same target as everything else in Part IV, from a different direction.

What's actually measured, and where the evidence stops

Worth knowing, because it's the difference between a chemistry finding and a health finding.

What's well established: ⚠️ repeatedly heated oil accumulates measurable oxidation products. This is measured directly in the oil — peroxide values, polar compound content, specific aldehydes — and many jurisdictions regulate commercial fryer oil on exactly these measures, requiring it to be discarded above a threshold of total polar compounds. A regulation exists because the chemistry is not in doubt.

What's reasonably established: these compounds are absorbed from fried food, and in animal and cell studies they do the things reactive aldehydes do — form adducts with proteins and DNA, promote oxidative stress.

⚠️ What is not established: the human outcome evidence at realistic dietary exposures. There is observational data associating fried food consumption with cardiometabolic disease — and fried food is confounded with everything §19.11 describes. ⚠️ You cannot tell from a cohort whether the harm came from the aldehydes, the energy density, the salt, the refined starch, or the life that includes eating a lot of fried food.

So the honest chain is: ✅ the chemistry is certain · 🟢 the biological plausibility is strong · 🟡 the human dose-response is unquantified.

⚠️ That is three different confidence levels in one argument, and collapsing them into "fried food is toxic" or "there's no evidence" both misreport it.

🔬 Verdict on Claim 1 — repeatedly heated frying oil produces harmful oxidation products: 🟢 Probably true, and the mechanism and chemistry are solid. ⚠️ This is the part of the seed oil argument that deserved a hearing and mostly hasn't had one, because it got bundled with Claim 4.

19.9 Smoke points, and why the stability argument is half backwards

A specific correction, because this one is widely repeated and easy to check.

The claim: "Refined seed oils are unstable and shouldn't be heated. Use butter, tallow or extra virgin olive oil instead."

⚠️ Refined oils have high smoke points — often higher than butter or extra virgin olive oil.

Refining removes free fatty acids and particulates — the things that smoke. ⚠️ So the refining that the argument objects to is the thing that raises the smoke point.

But smoke point isn't the whole story, and the sceptics have a point buried here:

Oxidative stability depends on degree of unsaturation (more double bonds = less stable), and on antioxidant content (which refining reduces). ⚠️ Extra virgin olive oil has a lower smoke point and better oxidative stability than its smoke point suggests, because it's high in monounsaturated fat and retains polyphenols that refining strips out.

So the accurate version: ⚠️ smoke point and oxidative stability are different properties, and the popular argument confuses them in both directions. Refined high-PUFA oils smoke late and oxidize readily. EVOO smokes early and oxidizes slowly. Highly saturated fats — tallow, ghee, coconut — are the most oxidatively stable, and that is the sceptics' best point on this specific question.

The practical upshot is smaller than either side implies: for home cooking, at ordinary temperatures, with oil used once, this difference is unlikely to matter much. For repeated high-temperature frying, it matters more (§19.8).

19.10 Processing: hexane, deodorizing, and an honest concession

Three separate objections, with three different answers.

1. Hexane extraction. Most refined seed oils are extracted using hexane, a petroleum-derived solvent, which is then removed. ⚠️ Residual hexane in finished oil is in the parts-per-million range, far below established safety limits, and the solvent is largely recovered and reused.

🔬 Verdict: ❌ Not supported as a meaningful health risk in the finished product. (The occupational and environmental case for hexane is a separate and more legitimate discussion, and it is not about your cooking oil.)

2. Bleaching and deodorizing. These are industrial steps using clay filtration and high-temperature steam. They remove colour, odour and free fatty acids — and, unavoidably, most of the antioxidants and phytochemicals in the crude oil.

⚠️ That's a real nutritional cost, and it's the strongest version of the processing objection: refined oil is a purified fat with the accompanying compounds removed. Chapter 7's intact-versus- refined continuum applies to oils exactly as it applies to grains.

3. And the honest concession: ⚠️ the high-temperature deodorizing step can generate small amounts of trans fatty acids. Typically well under a few percent, and far below the levels in the partially hydrogenated fats Chapter 9 covered — but not zero, and the sceptics who raise it are raising something real.

🔬 Verdict on processing: 🟡 Unclear / it depends. ❌ Hexane residue — not a meaningful risk. 🟡 Loss of antioxidants and small trans generation — real, of uncertain practical significance. ⚠️ This is a case where the direction of the concern is right and its magnitude is unestablished — which is a legitimate place to stand and is not the same as "seed oils are poison."

19.11 ⚠️ The confounding that swallows everything

The most important section in the chapter, and the reason the epidemiology can't settle this.

Where are seed oils found?

Fried takeaway. Crisps and chips. Packaged baked goods. Salad dressings. Mayonnaise. Fast food. Restaurant food. Essentially every ultra-processed food in the shop.

⚠️ So a person with high seed oil intake is, with overwhelming probability, a person with high ultra-processed food intake, high energy intake, low fiber, more takeaway, and less home cooking.

Which means:

Any observational association between seed oil intake and disease is almost perfectly confounded by ultra-processed food intake, and no statistical adjustment can fully separate them, because they are not separate exposures in the real world.

This cuts in both directions, and I want to be even-handed about it:

Against the sceptics: ⚠️ the population-level correlation between rising seed oil consumption and rising chronic disease is exactly what you'd expect if ultra-processed food were the cause and seed oils merely the marker. Chapter 2's confounding, at maximum strength.

Against the mainstream: ⚠️ it also means the reassuring observational data can't fully exonerate them either. Cohorts that adjust for "diet quality" are adjusting using instruments that partly measure seed oil intake. The confounding is symmetric.

And it explains something that puzzles people about the debate:

💡 ⚠️ People who cut out seed oils frequently do feel better and often lose weight — and they are not imagining it.

Because you cannot remove seed oils without removing nearly all fried food, nearly all packaged snacks, most takeaway, most restaurant meals, and most ultra-processed food.

Chapter 17 §17.1's eighth shape: count the changes. ⚠️ Removing seed oils changes six things at once, and one of them gets the credit.

This is not a debunk. It's a claim that the intervention works and the explanation is wrong — which is the single most common pattern in this entire book, and Chapter 22 is where the actual explanation lives.

Why this argument took off when it did

Worth a moment, because it explains the shape of the debate better than any of the biochemistry.

Seed oils arrived as the internet's villain at roughly the point sugar stopped being one. That's not a coincidence, and there are structural reasons.

1. ⚠️ It has a genuine fact at the centre (§19.3). Unlike the claims in Chapter 17, this one starts from a real, measured, unprecedented change. Movements built on a true premise are much harder to dislodge, and the people who noticed the adipose data first were doing legitimate work.

2. It offers a single, removable villain. Chapter 22's actual answer — ultra-processed food — is a category with fuzzy edges that requires changing how you shop, cook and eat. "Avoid these eight oils" is a rule you can follow on Tuesday. ⚠️ Chapter 17 §17.11's first criterion: it resolves uncertainty.

3. It works (§19.11). Which means every advocate has a genuine personal result.

4. It has a suppressed-truth story available, and industrial oil is an easy antagonist — ⚠️ and Chapter 18 §18.12 established that this shape is sometimes true, so it can't simply be dismissed. Applying that test here: is there a documented episode with a year, a journal, a sponsor and an archive? For seed oils specifically, not of the kind Kearns, Glantz and Schmidt found for sugar. ⚠️ The absence of a documented episode isn't proof of nothing — but it's the difference between a finding and a suspicion, and the distinction is worth insisting on in both directions.

5. And it has a moral vocabulary already built — industrial, refined, processed, unnatural — against tallow, butter, ancestral, traditional. ⚠️ Chapter 17's purity-and-contamination shape, fully loaded.

💡 ⚠️ The uncomfortable observation: shapes 1, 4 and 5 above are Chapter 17 tells firing on a claim whose factual premise is true.

This is the thing Part IV keeps teaching and it's worth saying as plainly as possible: the tells identify how a claim is being sold, not whether it's right. A true finding can be marketed with every rhetorical shape in the book — and Chapter 17 §17.9's warning was that a sceptic who reads the packaging and skips the content will get this exactly wrong.

Run the tells to decide how much scrutiny something deserves. Then do the scrutiny.

19.12 Filling the five slots

Chapter 18's threshold, applied. Watch how much of the argument survives.

⚠️ Four of five slots are empty in the claim as usually stated. That's not a hostile reading — it's why the argument can run for years without resolving.

19.13 So what do you actually do?

The practical answer is unsatisfying, cheap, and barely depends on who's right.

1. ⚠️ The form matters more than the oil. Deep-fried food eaten out is where the studied oxidation condition actually occurs. A bottle in your cupboard is not that.

2. Use olive oil as your default, if you can afford it. ⚠️ Not because seed oils are poison — because extra virgin olive oil has good oxidative stability, retains polyphenols, tastes better, and is the fat in the dietary pattern with the best outcome evidence (Chapter 10 §10.5). That's a positive case, not a defensive one.

3. Keep a high-smoke-point oil for high-heat cooking. Refined avocado, refined olive, or — yes — refined canola. ⚠️ Don't use extra virgin olive oil for searing at 250°C; it smokes at ~190–200°C.

4. Don't reuse frying oil, and don't buy from places that obviously do. §19.8 is the real finding and this is the action it implies.

5. Store oil cool, dark, closed, and buy sizes you'll finish. ⚠️ Free — and it's the one oxidation variable you actually control. A bottle open next to the hob for eighteen months is the household version of §19.8's studied condition. Smell it: rancid oil smells like crayons or old putty, and once you've noticed it once you'll never miss it again.

5b. Eat oily fish twice a week. ⚠️ The best-evidenced positive action in this chapter, and it's at the opposite end of the ratio from the argument. Tinned sardines, mackerel or salmon — under $2 a portion, no conversion bottleneck, and 🟢 rather than 🟠.

6. ⚠️ Don't throw away oil you've already bought. Theo binned $34 of oil and spent $48 replacing it, on the strength of a claim whose central mechanism has failed direct testing. If you want to change, change when the bottle is empty.

7. And check the denominator (Chapter 7 §7.7). If eliminating seed oils means eliminating fried takeaway, that's a real change and a good one — ⚠️ and you should know that's what you did, so that you don't attribute it to the oil and then reintroduce the takeaway once it's fried in tallow.

What we don't know

⚠️ Whether the adipose linoleic acid rise in §19.3 has consequences. It's real, it's unprecedented, and its significance is genuinely unknown. That is an honest answer and it should not be comfortable for anyone.

Whether there's a threshold above which LA intake matters, and where it would be.

Whether the recovered trials are outliers or signal (§19.6).

And whether any of this can ever be separated from ultra-processed food (§19.11) short of a long-term feeding trial that nobody is going to fund.

⚠️ I want to be explicit about my own confidence here, because this chapter has more low-confidence verdicts than any other so far.

Chapter 17's Project Checkpoint asked you to set a three-year reminder because some of this book will be wrong. ⚠️ If you want to know where to look first, look here.

19.14 The two honest positions, side by side

Now go back to your 🪞 Learning Check-In answers.

Three questions:

1. Which slots did you find empty? §19.12 says four of five. Did you get the form slot — the one that separates a cupboard from a commercial fryer?
2. Did you find the true version? §19.3 and §19.8 are both real. ⚠️ A check that returned "it's all nonsense" was run too fast — and Chapter 17's further-reading warned about exactly that.
3. What did you write down as changing your mind — and does the evidence in §19.5 meet it? ⚠️ If you wrote something and then didn't update when you read it, notice that. It's the most useful thing this chapter can show you about yourself, and it's why the check-in came first.

🧾 What it costs

⚠️ Olive oil is defensible on positive evidence (Chapter 10 §10.5) and tastes better. Tallow at $350 a year rests on a mechanism that failed testing. And "don't reuse frying oil" — the best-supported action in this chapter — costs nothing.

Seventh consecutive chapter where the best-evidenced change is the cheapest one on the page.

Spaced Review

Answer before reading on.

1. (Chapter 13) §19.5 says the mechanism failed testing. Which chapter of this book contains the closest parallel, and what's the shared lesson?

Chapter 13's antioxidant supplements. A real, elegant mechanism — oxidative damage causes disease, antioxidants neutralize it — that generated a clear prediction, was tested in large trials, and failed (ATBC and CARET, Chapter 2 §2.2). ⚠️ The shared lesson: a pathway existing is not the same as a pathway being rate-limiting. A diagram shows what can happen, not what does, in what quantity, or what regulates it.

2. (Chapter 9) Why can't you answer "are seed oils bad?" without naming a comparator?

Because Chapter 9 §9.5 established that fats are only evaluable as substitutions. ⚠️ Replacing saturated fat with PUFA, with refined carbohydrate, or with monounsaturated fat gives three different answers — and removing seed oils by not eating fried food is a fourth intervention entirely. "Compared to what" is the fifth slot and it decides the question.

3. (Chapter 17) People who eliminate seed oils often feel better. Which myth shape explains it, and why isn't that a debunk?

Shape 8, the bundled intervention — count the changes. ⚠️ Removing seed oils removes fried takeaway, packaged snacks, most restaurant food and most UPF: six changes, one credited. It isn't a debunk because the intervention genuinely works — the claim being corrected is the explanation, not the result. Chapter 22 has the actual mechanism.

Project Checkpoint: Your Fats Audit

Component nineteen. Thirty minutes, mostly standing in your kitchen.

Step 1 — Inventory every fat you own.

⚠️ Include butter, cooking spray, mayonnaise and salad dressing. Most people's largest seed oil source is a bottle of dressing or a jar of mayonnaise, not the oil they cook with.

Step 2 — The form question, which matters more than the inventory.

How often do I eat deep-fried food prepared outside my home? ____ times/week

Do I ever reuse frying oil at home? ____

⚠️ §19.8 says this is the studied condition. If your answer to the first is 4+, that's your finding — and it isn't really about oil.

Step 3 — Check your storage. ⚠️ Cool, dark, sealed, and finished within a few months. Is anything sitting open next to the hob? How old is the oldest bottle? This is free and it's the only oxidation variable you control.

Step 4 — Two decisions, and only two.

⚠️ Do not throw anything away. Change when a bottle is empty. Theo's $82 afternoon is the cautionary tale and it is entirely avoidable.

Step 5 — And write one sentence:

"If I change nothing about oil but eat fried takeaway _ fewer times a week, I have changed _____."

Fill in the blank honestly. ⚠️ For most people the second blank is "more than any oil swap would have."

Next checkpoint (Chapter 20): your organic and residue audit — what you buy organic, what it costs, and what you'd have to believe for that to be worth it.

Chapter Summary

"Seed oils" is a provenance category, not a compositional one. ⚠️ Canola (~19% LA) sits closer to olive oil (~10%) than to safflower (~70%), and peanut oil (~32%) usually isn't on the list at all.

Four claims:

Other verdicts:

⚠️ §19.5 is the chapter's decisive finding. The mechanism predicted that more LA → more tissue arachidonic acid → more inflammation. Tested directly: LA intake does not raise inflammatory markers, and does not raise tissue AA proportionally — the conversion is regulated and saturable. The chain has a broken link, and it's the load-bearing one. (Chapter 13's antioxidants, again.)

⚠️ §19.11 is why nothing here can be settled observationally. Seed oil intake and ultra-processed food intake are not separable exposures. The confounding is symmetric — it prevents indictment and exoneration equally. And it explains why elimination works: you cannot remove seed oils without removing six other things.

§19.12: four of five slots are empty in the claim as usually stated — dose, form, population, and comparator.

The one thing to remember: Theo binned $34 of oil and spent $48 replacing it, for a claim whose central mechanism failed direct testing. ⚠️ Meanwhile the best-supported action in the chapter — don't eat food fried in repeatedly reused oil — costs nothing and is about a restaurant, not a cupboard.

What's Next

Chapter 20 takes the same threshold into a debate where the dose question is argued almost entirely as a presence question — is it in there? rather than how much, and does that amount do anything?

Pesticide residues and what the measured levels actually are. Organic farming's own pesticides, which exist and are rarely mentioned. Whether organic produce is more nutritious — a genuine literature with a genuine answer. GMOs, the scientific consensus, and the fact that the strongest objections aren't about safety at all. The Dirty Dozen list, who publishes it, and what happens when you check the numbers behind it.

⚠️ And "clean eating" — a phrase with no definition, a large following, and a documented relationship with disordered eating that Chapter 34 will pick up.