There is a version of this page that would be easier to write. It would pick the most affecting studies, drop the failed replications, and round every number up. It would probably work better in the short term.

I am not writing that one, for the same reason the kitchen publishes the recipes that failed: the moment somebody finds one thing here they can disprove, everything else on the site becomes suspect. So the sections below include the studies that did not replicate, the claims I could not source, and one live scientific dispute that I am not going to pretend is settled.

If you came here already convinced, the useful part is section seven — the things you may be repeating that do not hold up. If you came here sceptical, start at the top; the formal position of the field is more careful and more interesting than either side usually reports.

1 · What scientists actually signed

Two documents get cited constantly and quoted badly. Here is what each one really says.

The Cambridge Declaration on Consciousness — 7 July 2012

Signed at the Francis Crick Memorial Conference at Churchill College, Cambridge, in the presence of Stephen Hawking. The operative sentence:

“The absence of a neocortex does not appear to preclude an organism from experiencing affective states… the weight of evidence indicates that humans are not unique in possessing the neurological substrates that generate consciousness. Non-human animals, including all mammals and birds, and many other creatures, including octopuses, also possess these neurological substrates.”

Cambridge Declaration on Consciousness, 2012

Read it carefully. It names mammals, birds and octopuses. It does not name fish, and it makes a claim about neurological substrates — the hardware — not a claim about what any particular animal experiences. It was a declaration by conference participants, not a peer-reviewed paper and not an institutional position of the University of Cambridge.

The New York Declaration on Animal Consciousness — 19 April 2024

Four sentences, hosted at NYU. The middle two are the ones that matter, and the distinction between them is deliberate:

“First, there is strong scientific support for attributions of conscious experience to other mammals and to birds.

Second, the empirical evidence indicates at least a realistic possibility of conscious experience in all vertebrates (including reptiles, amphibians, and fishes) and many invertebrates (including, at minimum, cephalopod mollusks, decapod crustaceans, and insects).”

New York Declaration on Animal Consciousness, 2024

“Strong scientific support” and “realistic possibility” are two different epistemic claims, and the document is precise about which animals get which. Its organisers wrote separately that it would be “inappropriate to talk about ‘proof,’ ‘certainty,’ or ‘conclusive evidence’” in this area. The declaration makes no policy recommendation of any kind.

2 · What the law recognises

Legal recognition of sentience is real, recent, and much weaker than it sounds.

United Kingdom — Animal Welfare (Sentience) Act 2022

Royal Assent 28 April 2022. Section 5(1) defines “animal” as any vertebrate other than humans, any cephalopod mollusc, and any decapod crustacean — octopus, squid, crab, lobster, crayfish, prawn.

What the Act actually does: it requires the Secretary of State to maintain an Animal Sentience Committee, which may publish reports on whether a government policy had “all due regard” to animal welfare, to which ministers must respond within three months.

Why crabs and octopuses were added

Because of a specific evidence review: Birch and colleagues at the London School of Economics, commissioned by DEFRA, November 2021. They assessed eight criteria — nociceptors; integrative brain regions; neural pathways connecting the two; responses modulated by anaesthetics; motivational trade-offs between threat and reward; flexible self-protective behaviour after injury; associative learning beyond habituation; and behaviour showing the animal values pain relief when injured.

Their gradings were graded, not blanket: very strong evidence for octopods, strong for true crabs, substantial for cuttlefish, squid, lobsters, crayfish and caridean shrimps, and explicitly weak or uncertain for nautiloids and penaeid shrimps. Their recommendation was that all cephalopod molluscs and decapod crustaceans be treated as sentient in UK welfare law. The government took it.

European Union — Article 13

Worth quoting in full rather than stopping where advocacy usually stops:

“…the Union and the Member States shall, since animals are sentient beings, pay full regard to the welfare requirements of animals, while respecting the legislative or administrative provisions and customs of the Member States relating in particular to religious rites, cultural traditions and regional heritage.”

Treaty on the Functioning of the European Union, Article 13

The carve-out is part of the sentence. Quoting only the first half is the kind of thing that makes a reader stop trusting a page.

United States — the gap

There is no federal statute recognising animal sentience. The Animal Welfare Act’s definition of “animal” expressly excludes birds, rats and mice bred for research, and “other farm animals…used or intended for use as food or fiber.” The Humane Methods of Slaughter Act covers “cattle, calves, horses, mules, sheep, swine, and other livestock” — poultry is not named, and USDA has declined petitions to read it in.

The honest version of this: chickens and turkeys sit outside both statutes. That is not the same as “no protection at all” — the Poultry Products Inspection Act and state cruelty codes exist, though most state codes exempt customary farming practices.

3 · Chickens

Roughly nine in every ten land animals killed for food is a chicken. They are also the animal about which people hold the strongest confident opinions and the least evidence.

Newly hatched chicks track quantities

Rugani and colleagues (2009, Proceedings of the Royal Society B) showed chicks a set of identical objects, then hid them behind screens and moved some from one screen to the other while the chick watched. The chicks approached the screen hiding the larger set, with quantities up to five. They were days old and had never been trained.

Hens will wait — but not as long as you have been told

This one is worth getting right, because the popular version is wrong. Abeyesinghe and colleagues (2005, Animal Behaviour) offered hens a choice between a short wait for a small feed and a longer wait for a much larger one.

In the standard condition — 2 seconds’ wait for 3 seconds of food, versus 22 seconds’ wait for 22 seconds of food — the hens were impulsive. They took the small immediate reward. Self-control only appeared in a “jackpot” condition where the longer wait was cut to 6 seconds; there, seven of eight hens chose the delayed larger reward, at a mean rate of 0.93.

The authors read the impulsive result as temporal discounting under uncertainty rather than a cognitive ceiling — the jackpot condition ruled out “an absolute cognitive constraint.” But if you have seen “hens will wait 22 seconds for more food,” that is the study, reported backwards.

Food calls behave like referential signals

Evans & Evans (1999) played recordings of male food calls to isolated hens in a sound-attenuated chamber with no food present. The hens looked down and moved their heads toward the ground as if foraging. Acoustically similar ground alarm calls did not produce it. Contact calls from comparable social contexts did not either.

“Functionally referential” is a technical term: the signal behaves like a referential one. The authors themselves note such signals likely encode several kinds of information at once. It is not a claim that chickens have a word for food, and I am not going to upgrade it into one.

A hen responds to her chicks’ distress

Edgar and colleagues (2011, Proceedings of the Royal Society B, Bristol) directed a puff of air at chicks while the mother watched, against three control conditions. The hens showed increased heart rate, increased maternal vocalisation, increased alertness, reduced preening, and a measurable drop in eye temperature — and did so even though the chicks themselves barely vocalised, which means the hen was responding to the event rather than to a distress call.

The paper’s own conclusion is careful: adult female birds possess “at least one of the essential underpinning attributes of empathy.” That is emotional contagion — one component — not empathy entire. The authors said so; so will I.

4 · Pigs

Mirrors — and the replication that failed

Broom, Sena & Moynihan (2009) found that seven of eight mirror-experienced pigs located hidden food using a reflection, in roughly 23 seconds.

Then Gieling and colleagues (2014) tried to repeat it and published under the title “Lack of mirror use by pigs to locate food.” In replication, 2 of 11 pigs in one trial and 1 of 11 in another used the mirror. Marino & Colvin — whose review is the standard citation for pig cognition — concluded in their own words that the original results “are not as robust as they may seem.”

The same reviewers state plainly that pigs have not demonstrated mirror self-recognition by the mark test, unlike great apes, elephants, dolphins and magpies.

The joystick study, with its caveats attached

Croney & Boysen (2021, Frontiers in Psychology) trained four pigs to move a cursor with a joystick to hit targets on a screen. All four performed above chance on the easiest targets; two Yorkshires reached 42–48% accuracy and one micro pig reached 76%.

The authors’ own limits are as interesting as the result: the apparatus was built for primates, pigs are far-sighted so the monitor sat at 45 cm, and “the pigs’ limited dexterity no doubt constrained their performance.” Four animals is four animals.

One incidental observation from that study has stayed with me. When the food dispenser malfunctioned, the pigs carried on working for verbal encouragement alone.

5 · Cows

The “eureka” study did not replicate

You will find, on a great many vegan sites, the claim that cows experience visible excitement at the moment of solving a problem. It comes from one study: Hagen & Broom (2004), a yoked design in which heifers who controlled the delivery of a reward “got more excited” than heifers who received the same reward without control.

In 2020, Meagher and colleagues at UBC tried to reproduce it in Scientific Reports. Their finding, verbatim: “There was no treatment effect on play in the current study, and unlike Hagen and Broom, we observed play in both learning and control individuals.” They found “no evidence that play is linked with a learning experience,” concluding instead that heifers are motivated to take part in learning tasks but that other aspects of the experience were also rewarding.

Cow and calf

Mother–calf bonds form within about five minutes of birth (Hudson & Mullord, 1977). After separation, mothers walk, urinate, vocalise and butt, and remain restless for over 24 hours. Weary & Chua (2000) found mothers produced contact calls at significantly higher rates when separation happened at four days than at six hours or one day — the longer the pair are together, the stronger the response.

I could not obtain the original abstracts for the Weary and Flower papers directly; those findings are reported at second hand through the Marino & Allen review. So I am not going to publish specific vocalisation counts or durations for them. I also found no source at all for the widely circulated claim that cows cry for their calves for weeks, and I am not going to repeat it.

The health argument for early separation does not hold up either

Beaver and colleagues (2019, Journal of Dairy Science) reviewed 70 peer-reviewed papers on the health effects of separating dairy calves early. For calf scours and mastitis, findings “pointed to beneficial or no effects of suckling.” For cryptosporidiosis, pneumonia, immunity and mortality, results were mixed. Overall: “no consistent evidence in support of early separation.”

Be precise about what that means. It undercuts the industry’s health justification for the practice. It is not itself evidence of emotional harm. Those are different claims and it is worth keeping them apart.

6 · Fish — a real dispute

This is the section where I would most like to give you a clean answer, and cannot. Whether fish consciously feel pain, as opposed to detecting and responding to injury without experiencing it, is an open argument between serious neuroscientists. Here is each side in its own words.

When Key published, the journal ran dozens of open peer commentaries. Some supported him. More opposed him — Broom, Brown, Balcombe, Braithwaite, Striedter, and Anil Seth, whose title is the best single summary of where the field actually sits: “Why fish pain cannot and should not be ruled out.”

That is the honest position. Not “fish definitely feel pain.” Not “fish definitely do not.” It cannot be ruled out, and the precautionary weight falls on assuming they might — which is exactly the New York Declaration’s “realistic possibility” wording, arrived at independently.

7 · What I will not tell you

Every claim in this section is one I have seen on vegan websites, including ones I like. None of them survived a search for a source.

Three general warnings, which cut in both directions

The mirror test is weak evidence either way. De Waal’s objection to the wrasse result is that cleaner wrasse are specifically adapted to spot and remove things stuck to fish, so scraping at a mark may be their day job rather than self-recognition. And in the other direction: dogs reliably fail the visual mirror test, which is routinely reported as evidence they lack self-awareness — but Horowitz’s “olfactory mirror” work suggests the test may simply be in the wrong sense for a scent-dominant animal. A failed test is not evidence of absence, and a passed one is not proof.

The guilty look is not guilt. Horowitz (2009) crossed whether 14 dogs had actually eaten a forbidden treat with whether the owner scolded them, deliberately misinforming owners half the time. Dogs showed no more “guilty” behaviour when they had actually disobeyed. Scolding was the significant predictor — and the effect was strongest when the dog had in fact been obedient. This is the cleanest published demonstration that a confidently-read animal emotion was the observer’s projection, and I keep it in mind on every other page of this site.

Faces are poor evidence. Barrett and colleagues (2019) found that even in humans, facial configurations do not map reliably onto discrete emotions. If reading a human face is unreliable, reading a pig’s is weaker still. The legitimate exception is a validated grimace scale — the sheep pain scale developed against footrot and mastitis, checked against analgesic response. That measures pain, against a known standard. Reading contentment off a photograph does not.

And the objection people raise back

“Plants feel pain too” is usually offered as a gotcha, and it deserves an actual answer. Taiz and colleagues (2019, Trends in Plant Science) reviewed the plant-consciousness literature and concluded the evidence is “highly problematical”: proponents “consistently glossed over the unique and remarkable degree of structural, organizational, and functional complexity that the animal brain had to evolve before consciousness could emerge,” and it is “extremely unlikely” that plants, lacking anything comparable, possess it. On the “root brain” idea they note dryly that if the root tip is a command centre, so is the shoot tip, and you may as well call the whole plant a brain.

Two honest notes. That paper is an opinion piece in a peer-reviewed journal and the plant-signalling community contests it — it is the strongest citation for the mainstream position, not a settled consensus document. And the argument cuts the wrong way for the person making it: most crops grown on earth are eaten by farmed animals, so if plant suffering counted, eating plants directly would still cause less of it. Section one of the earth page has the figures.

8 · The numbers, and their error bars

Almost every animal-count you see quoted is a single number. Almost none of them should be.

The same fact, measured three different ways

Land animals are counted by head. Fish are not — the FAO records catch in tonnes, so every fish number in circulation is a weight divided by an estimated average body weight. The bars are the published ranges; the dot is the midpoint everyone quotes.

Per species, for 2021: chickens 73.79 billion, ducks 4.31 billion, pigs 1.4 billion, sheep 617 million, goats 501 million, cattle 332 million. Faunalytics puts the 2024 land-animal total at just under 88 billion.

9 · What follows from this

Less than you might expect, and that is deliberate.

None of the above proves that a chicken’s experience resembles yours. The research does not reach that, and the researchers are careful not to claim it. What it establishes is narrower and, I think, sufficient: that the animals we farm in the largest numbers have nervous systems built for feeling, respond to injury in ways consistent with experiencing it, and that the formal scientific position on the harder cases is not “no” but “we cannot rule it out.”

The New York Declaration’s third sentence is the only conclusion I would draw out loud: when there is a realistic possibility of conscious experience in an animal, it is irresponsible to ignore that possibility in decisions affecting that animal.

What you do with that is not my business. I am not going to tell you what you owe anyone. I run a kitchen, and the offer at the end of every page on this site is the same one: if you want to eat differently and you do not know how, write to me and I will answer. There is nothing to buy and no follow-up.

10 · Sources

Everything above, in order. Where I could not verify something, the page says so rather than citing it.

1. Cambridge Declaration on Consciousness, Francis Crick Memorial Conference, 7 July 2012 — fcmconference.org
2. The New York Declaration on Animal Consciousness, NYU, 19 April 2024 — declaration and background
3. “Insects and Other Animals Have Consciousness, Experts Declare,” Quanta Magazine, 19 April 2024 — quantamagazine.org
4. Animal Welfare (Sentience) Act 2022 — legislation.gov.uk; House of Commons Library briefing CBP-9423 — commonslibrary.parliament.uk; Jowitt, UK Constitutional Law Association, 2022 — ukconstitutionallaw.org
5. Birch, Burn, Schnell, Browning & Crump, Review of the Evidence of Sentience in Cephalopod Molluscs and Decapod Crustaceans, LSE for DEFRA, November 2021 — lse.ac.uk
6. Treaty on the Functioning of the European Union, Article 13 — eur-lex.europa.eu
7. Animal Welfare Act definition, 7 U.S.C. § 2132 — law.cornell.edu; Humane Methods of Slaughter Act, 7 U.S.C. § 1902 — law.cornell.edu
8. Rugani, Fontanari, Simoni, Regolin & Vallortigara, “Arithmetic in newborn chicks,” Proc. R. Soc. B 276:2451–2460, 2009 — doi.org
9. Abeyesinghe, Nicol, Hartnell & Wathes, “Can domestic fowl show self-control?” Animal Behaviour 70:1–11, 2005 — PDF
10. Evans & Evans, “Chicken food calls are functionally referential,” Animal Behaviour 58:307–319, 1999
11. Edgar, Lowe, Paul & Nicol, “Avian maternal response to chick distress,” Proc. R. Soc. B 278:3129–3134, 2011 — doi.org
12. Marino, “Thinking chickens,” Animal Cognition 20:127–147, 2017 — springer.com
13. Broom, Sena & Moynihan, “Pigs learn what a mirror image represents,” Animal Behaviour 78:1037–1041, 2009; Gieling, Mijdam, van der Staay & Nordquist, “Lack of mirror use by pigs to locate food,” Applied Animal Behaviour Science 154:22–29, 2014 — sciencedirect.com
14. Marino & Colvin, “Thinking Pigs,” Int. J. Comparative Psychology 28, 2015 — escholarship.org
15. Croney & Boysen, “Acquisition of a Joystick-Operated Video Task by Pigs,” Frontiers in Psychology 12:631755, 2021 — frontiersin.org
16. Hagen & Broom, “Emotional reactions to learning in cattle,” Applied Animal Behaviour Science 85:203–213, 2004; Meagher, Strazhnik, von Keyserlingk & Weary, “Assessing the motivation to learn in cattle,” Scientific Reports 10:6847, 2020 — nature.com
17. Marino & Allen, “The Psychology of Cows,” Animal Behavior and Cognition 4:474–498, 2017 — PDF
18. Beaver, Meagher, von Keyserlingk & Weary, “A systematic review of the effects of early separation on dairy cow and calf health,” J. Dairy Science 102:5784–5810, 2019 — journalofdairyscience.org
19. Sneddon, Braithwaite & Gentle, “Do fishes have nociceptors?” Proc. R. Soc. B 270:1115–1121, 2003 — royalsocietypublishing.org
20. Rose, Arlinghaus, Cooke, Diggles, Sawynok, Stevens & Wynne, “Can fish really feel pain?” Fish and Fisheries 15:97–133, 2014 — PDF
21. Key, “Why fish do not feel pain,” Animal Sentience 3(1), 2016, with open peer commentary — issue contents; Sneddon et al., “Fish sentience denial: Muddying the waters,” Animal Sentience 21(1), 2018
22. Kohda et al., “If a fish can pass the mark test…” PLOS Biology 17(2):e3000021, 2019 — plos.org; de Waal, “Fish, mirrors, and a gradualist perspective on self-awareness,” PLOS Biology 17(2):e3000112, 2019
23. Horowitz, “Disambiguating the ‘guilty look’,” Behavioural Processes 81:447–452, 2009 — PDF
24. Barrett, Adolphs, Marsella, Martinez & Pollak, “Emotional Expressions Reconsidered,” Psychological Science in the Public Interest 20(1), 2019 — sagepub.com; McLennan et al., sheep facial expression pain scale, Applied Animal Behaviour Science 176:19–26, 2016
25. Taiz, Alkon, Draguhn, Murphy, Blatt, Hawes, Thiel & Robinson, “Plants Neither Possess nor Require Consciousness,” Trends in Plant Science 24:677–687, 2019 — PDF
26. Our World in Data, animal slaughter counts — ourworldindata.org and animal welfare; Faunalytics global slaughter statistics — faunalytics.org
27. Mood, Lara, Boyland & Brooke, “Estimating global numbers of farmed fishes killed for food annually from 1990 to 2019,” Animal Welfare 32:e12, 2023 — cambridge.org; Mood & Brooke, wild-caught estimates, Animal Welfare 33:e6, 2024 — doi.org

If you only take one thing from this page, take this: the environmental case for eating fewer animal products is strong enough that it does not need help, and every exaggerated version of it hands the other side an easy win.

So this page reports the counter-evidence in the same voice as the evidence. Twice it concedes a point outright. Once it concedes a number that is genuinely worse for the plant-based side than for beef. If that costs the argument something, the argument was not worth much.

1 · Land

The foundational study is Poore & Nemecek, published in Science in 2018. It is not an estimate from a model — it consolidated data from about 38,700 commercially viable farms across 119 countries, covering 40 products that together account for roughly 90% of global protein and calorie consumption, plus 1,600 processors, packaging types and retailers.

The strongest number in the whole debate

What animal farming takes from the earth, set against what it puts on the plate. Nobody contests these three figures — not the FAO, not the livestock sector.

Their scenario for a global shift away from animal products reduces food’s land use by 3.1 billion hectares — a 76% reduction.

The finding almost nobody quotes

The same paper found that impact can vary fifty-fold among producers of the same product. Ninetieth-percentile beef emits 105 kg CO₂eq per 100 g of protein and occupies 370 m²-years; tenth-percentile dairy beef is 12 and 50 times lower respectively. How a thing is farmed matters enormously.

And then the sentence that resolves it, from the same authors: “impacts of the lowest-impact animal products typically exceed those of vegetable substitutes.” Within-category variation is huge. It still does not close the gap between categories.

2 · Emissions

Xu and colleagues (2021, Nature Food) put global food-production emissions at 17.3 gigatonnes CO₂eq a year (±1.7), split:

Where food emissions come from

Total food-production emissions are about 17.3 gigatonnes of CO₂-equivalent a year. This is how they divide.

The single largest commodity is beef at 25% of food emissions. The largest plant is rice at 12%. The paper’s title says animal emissions are “twice” plant emissions — that is 57 against 29, which is accurate, provided you do not quietly drop the 14%.

How big is the food system overall?

Here the honest answer is a range, not a number. Crippa and colleagues (2021) put food systems at 34% of total global greenhouse gas emissions. Poore & Nemecek, via Our World in Data, put it at 26%. The gap is a boundary-definition difference — what counts as inside the food system — not a factual dispute. Quoting “a third” without saying so slightly overstates the case; quoting 26% is defensible.

The finding that cuts against “eat local”

Transport is about 5% of food’s emissions. More than 80% of food-transport CO₂ is trucks; air freight is 0.4% of it. As Hannah Ritchie puts it: “If people switch from imported plant-based foods or chicken to local beef, their carbon footprint will rise.” What you eat matters far more than how far it travelled.

3 · The number that keeps shrinking

If you take one credibility lesson from this page, take this one. The FAO has revised its own livestock emissions figure downward twice, and vegan advocacy is still quoting the 2006 version.

The FAO has revised its own figure down twice

Same organisation, same subject, three reports. The 2006 number is the one you still see quoted everywhere — including by people arguing for what I am arguing for.

An advocacy page citing 18% in 2026 is citing a twenty-year-old figure that its own publisher has superseded twice. I would rather say that first than have someone else say it to me.

And the specific claim that was wrong

Livestock’s Long Shadow stated that livestock produces more greenhouse gases than transport. It is one of the most-repeated lines in the movement, and the methodology behind it was faulty: the FAO applied a full life-cycle assessment to livestock — feed, land-use change, manure, processing, distribution — and compared it against tailpipe-only emissions for transport, excluding vehicle manufacture, road construction, and fuel extraction and refining.

Frank Mitloehner at UC Davis identified this publicly in 2010. FAO co-author Pierre Gerber conceded it to the BBC:

“I must say honestly that he has a point — we factored in everything for meat emissions, and we didn’t do the same thing with transport, we just used the figure from the IPCC…”

Pierre Gerber, FAO, to BBC News, 24 March 2010

On Mitloehner specifically

He is worth engaging with properly rather than dismissing, because on the transport point he was right and everyone now agrees.

4 · Methane is genuinely different

This is the strongest live argument the livestock side has, and treating it as a bad-faith invention would be exactly the error this page exists to avoid. Myles Allen at Oxford, who developed the alternative metric, is a mainstream climate scientist and an IPCC lead author.

The physics, from IPCC AR6: methane’s perturbation lifetime is about 11.8 years. CO₂ has no single lifetime and a substantial fraction persists for centuries. Consequences:

• CO₂ accumulates. Constant emissions mean rising concentration and rising warming.
• Methane is destroyed on a decade timescale. A constant emission rate produces a roughly stable concentration, and therefore roughly stable rather than rising warming.

GWP100 — the standard conversion, which prices biogenic methane at 27× CO₂ over a century — treats a tonne of each as a fixed exchange rate and cannot represent that difference. Allen and colleagues proposed GWP* to fix it: it relates cumulative CO₂ to the rate of change of short-lived pollutants. Lynch and colleagues quantified the result: a stable methane source is worth roughly a quarter of its GWP100 value, and a decline of about 0.3% a year produces negligible additional warming.

The honest summary: the physics of GWP* is not in dispute. What is in dispute is whether a metric that measures change in warming is a legitimate basis for targets, and who benefits when it is applied to one sector or one country rather than globally. Anyone who tells you this is settled — in either direction — is not reading the literature.

5 · Water, and the burger figure

Water footprints come in three kinds, and conflating them is where almost every viral statistic goes wrong.

Mekonnen & Hoekstra’s global averages for beef, per kilogram:

What “15,000 litres of water per kilo of beef” is actually made of

Three physically different things, summed into one alarming number. Green water is rain that fell on that land whatever grew there. Only the blue sliver comes out of a river, a reservoir or an aquifer.

So when you see “660 gallons of water for one burger,” here is what is happening. Run the source data yourself: 15,415 L/kg works out to about 1,847 gallons a pound, so roughly 460 gallons for a quarter-pound patty — the Water Footprint Calculator says 450. The 660 figure does not reconcile with its own source; it is either different beef data or it has quietly added the cheese and the bun.

And of that ~460 gallons, the blue water — the part that comes out of a tap, a river or an aquifer — is around 17 gallons. Still a real number, still favouring the plant-based case. But two orders of magnitude below the headline.

6 · Almonds, avocados, quinoa, soy

Four things get thrown at vegans. Two of them are wrong, one is half right, and one is right — so here they are with the scores marked.

Almond milk Half right

True: almond is the thirstiest common plant milk. Per litre of beverage, almond takes 371 L of water against soy’s 28 and oat’s 48 — and California, where most almonds grow, is water-stressed.

False: that this makes it worse than dairy. Same source data, per litre: dairy takes 628 L of water, 8.9 m² of land and 1.14–2.50 kg CO₂eq. Almond takes 371 L, 0.5 m² and 0.42 kg. It is better on all three; the water gap is narrow, not reversed.

The real answer: if water is your concern, the argument is not almond versus dairy. It is soy or oat versus everything.

Avocado They have a point

This is the one that survives scrutiny, so here it is with the number attached. A 2024 study of two municipalities in Michoacán found avocado blue water at 689–712 litres per kilogram.

Beef’s global-average blue water is 550 L/kg. Irrigated Michoacán avocado uses more withdrawn water per kilogram than beef does. There is also documented deforestation associated with the expansion.

The qualifications are real — that is one irrigated region against a global average, it is per kilogram rather than per calorie or per gram of protein, and beef’s total footprint is still eight to twelve times higher. But the blue-water comparison stands, and conceding it costs almost nothing: avocado is neither a staple nor a protein source. It is a criticism of a fruit, not of plant-based eating.

Quinoa Largely false

The story was that Western demand priced Andean growers out of their own staple. Bellemare and colleagues (2018, World Development) actually measured it: for a 10% rise in the quinoa price, welfare of the average household in producing areas increased by about 0.7%. Producers’ welfare rose faster than the general population’s, and rising prices reduced variability in their income — the crop worked as partial insurance. The Peruvians eating less quinoa were disproportionately wealthier urban households; the producers were poorer, and gained.

There is a separate and more serious literature on soil degradation from monoculture expansion in the Bolivian altiplano, which I have not verified figures for and therefore will not quote.

Soy Backwards

“Soy destroys the Amazon, so your tofu is the problem.” The split:

So most soy-driven deforestation is attributable to meat, not tofu. Two honest additions: in 2015 only 13% of Brazilian soy came from the Amazon and 48% from the Cerrado, which is less protected and more affected; and pasture, not soy, drives most direct Amazon clearance — soy contributes by pushing pasture into forest. The overclaim to avoid is that tofu is therefore deforestation-free. Seven per cent of a crop with a serious land footprint is not zero. Ask where it was grown.

7 · Land that cannot grow crops

The strongest argument on the livestock side, and it deserves accurate numbers.

Mottet and colleagues (2017) found that of roughly 6 billion tonnes of feed consumed by livestock each year, 86% is not edible by humans — crop residues and by-products. Soybean meal is only 4% of global feed intake by mass. On average it takes 3 kg of cereals per kg of meat, and cattle need 0.6 kg of human-edible protein to produce 1 kg of protein in milk and meat.

The counter that actually lands

“This land cannot grow crops” does not mean “this land is best used for livestock.” The alternative is not wheat; it is ecosystem restoration. Hayek and colleagues (2021, Nature Sustainability) put the carbon opportunity cost at 332–547 gigatonnes CO₂ of sequestration available from shifting global food production to plant-based diets by 2050 — equivalent to 99–163% of the emissions budget for a 66% chance of holding warming to 1.5 °C.

And the overclaim from the other direction

Regenerative grazing is often presented as a carbon sink that cancels the problem. The largest review of that literature — Grazed and Confused?, from Oxford’s Food Climate Research Network — found that under generous assumptions soil sequestration could offset 20–60% of emissions from the grass-fed sector, 4–11% of total livestock emissions, and 0.6–1.6% of total human emissions. Its lead author: “Grazing livestock are not a climate solution. Grazing livestock are net contributors to the climate problem.” The sequestration is small, time-limited and reversible.

8 · India is a different problem

Most of what is written about dairy and climate is written about Western dairy, and almost none of it transfers. I cook Indian food, so this section matters to me more than the rest of the page.

India is the world’s largest milk producer — about a quarter of global output, 230 Mt in 2022–23, with per-capita availability of 459 g a day against a world average of 322. Livestock emissions were about 222 Mt CO₂e in 2018 — roughly 7.5% of India’s economy-wide emissions, of which about 90% is enteric fermentation.

9 · What I actually think

After all of the above, the case that survives is narrower than the one you usually hear — and, I think, more persuasive for it.

• The land figure is the strongest number in the whole debate. 83% of farmland for 18% of calories is not a rounding error and no one contests it.
• Beef is in a category of its own. A quarter of all food emissions from one product.
• What you eat matters far more than where it came from. Transport is 5%.
• Milk is the swap with the best ratio of impact to effort — the biggest single change most people can make without changing anything they consider a meal.
• Methane is genuinely different from CO₂, and pretending otherwise loses the argument to anyone who has read the physics.
• The FAO figure has come down twice. Say it before someone says it to you.

What I will not do is tell you this is the most important thing you can do about climate, or that going vegan is sufficient, or that the numbers are worse than they are. They are bad enough.

And if you are here because the environment is what moved you, the useful next page is not the philosophy — it is the practical one. Change your milk and nothing else this week. That is the whole first step, it costs nothing, and it removes more dairy from most people’s week than everything else put together, because of how much of it goes into tea.

10 · Sources

1. Poore & Nemecek, “Reducing food’s environmental impacts through producers and consumers,” Science 360:987–992, 2018 — science.org · full text PDF
2. Xu et al., “Global greenhouse gas emissions from animal-based foods are twice those of plant-based foods,” Nature Food 2:724–732, 2021 — nature.com
3. Crippa et al., “Food systems are responsible for a third of global anthropogenic GHG emissions,” Nature Food 2:198–209, 2021 — nature.com
4. FAO, Livestock’s Long Shadow, 2006; Tackling Climate Change Through Livestock, 2013 — fao.org; Pathways towards lower emissions, 2023
5. Pierre Gerber, quoted in “UN body to look at meat and climate link,” BBC News, 24 March 2010 — bbc.co.uk; context and the “stand entirely behind” quotation via Vegetarian Resource Group
6. Mitloehner, CLEAR Center — “Bogus burger blame”; EPA, Agriculture Sector Emissions — epa.gov; funding, Greenpeace Unearthed, 31 October 2022 — unearthed.greenpeace.org
7. Allen et al., “A solution to the misrepresentations of CO₂-equivalent emissions of short-lived climate pollutants,” npj Climate and Atmospheric Science 1:16, 2018 — nature.com; Lynch, Cain, Pierrehumbert & Allen, Environmental Research Letters 15:044023, 2020 — open access
8. Rogelj & Schleussner, “Unintentional unfairness when applying new greenhouse gas emissions metrics at country level,” ERL 14:114039, 2019 — iopscience.iop.org; Carbon Brief Q&A on GWP* — carbonbrief.org; scientific statement on methane metrics, 2026 — methanescience.org
9. IPCC AR6 WG1 Chapter 7 — ipcc.ch
10. Mekonnen & Hoekstra, “A Global Assessment of the Water Footprint of Farm Animal Products,” Ecosystems 15:401–415, 2012 — PDF; Water Footprint Calculator
11. Milk comparison figures via The Conversation and World Resources Institute
12. Guerrero-García-Rojas et al., “Avocado Water Footprint for Two Municipalities in Michoacán, Mexico,” Agriculture 14(7):981, 2024 — mdpi.com
13. Bellemare, Fajardo-Gonzalez & Gitter, “Foods and Fads: The Welfare Impacts of Rising Quinoa Prices in Peru,” World Development 112:163–179, 2018 — author summary
14. Our World in Data, where soy goes and environmental impacts of food; Hannah Ritchie on food miles — substack
15. Mottet, de Haan, Falcucci, Tempio, Opio & Gerber, “Livestock: On our plates or eating at our table?” Global Food Security 14:1–8, 2017 — summaries via Feedipedia and ILRI
16. Hayek, Harwatt, Ripple & Mueller, “The carbon opportunity cost of animal-sourced food production on land,” Nature Sustainability 4:21–24, 2021 — nature.com
17. Garnett et al., Grazed and Confused?, Food Climate Research Network, Oxford, 2017 — tabledebates.org; Nature Communications 2023 on soil carbon offset risk — nature.com
18. Press Information Bureau, Government of India, World Milk Day 2024 — pib.gov.in; USDA ERS, India’s Dairy Sector, LDPM-272-01, 2017 — PDF
19. Mech et al., “Carbon footprint of milk production…in smallholder dairy farms in Karnataka, India,” J. Dairy Science 106(12), 2023 — journalofdairyscience.org; Vasudha Foundation / GHG Platform India, 2022 — PDF

1 · How to read an effect size

Most nutrition writing gives you a direction — “lowers cholesterol” — and never a size. The size is the whole question.

2 · Cholesterol

The strongest general evidence is Koch and colleagues (2023, European Heart Journal): a meta-analysis of 30 randomised trials, about 2,370 people, running from 10 days to 5 years.

Note the last row. Triglycerides did not move, and slightly went the wrong way. Nobody quotes that row. It is in the same paper as the row everybody does quote.

A vegan-only pool, 2025. Koch mixed vegetarian and vegan trials. Tow and colleagues (2025, Trends in Food Science & Technology) pooled only vegan diets — 11 randomised trials — and found LDL −0.40 mmol/L (95% CI −0.49 to −0.31) across 9 trials and 826 people, total cholesterol −0.34, and body weight −5.05 kg (−6.99 to −3.10) across 7 trials. Triglycerides again did not move (−0.27, CI crossing zero, p = 0.15). The LDL number is a little larger than Koch’s; the direction is the same; and the same row stays blank.

And the result that is almost always left out

If cholesterol is what you came for, the plant-based diet is not the strongest dietary intervention on record. The portfolio diet is.

Chiavaroli and colleagues (2018) pooled seven controlled comparisons in 439 people with high cholesterol: LDL −0.73 mmol/L (95% CI −0.89 to −0.56), about a 17% fall, against a control diet that was itself a cholesterol-lowering diet. GRADE certainty: high. That is more than twice the effect of “go plant-based” on its own.

It works because it is specific rather than general: nuts, plant sterols, viscous fibre and soy protein, in stated amounts. Which is the honest headline of this whole page — what you add matters at least as much as what you remove.

3 · Blood pressure

Yokoyama and colleagues (2014, JAMA Internal Medicine), 7 controlled trials, 311 people:

That last framing is the authors’ own: the reduction is “comparable to other lifestyle modifications and approximately half the magnitude of pharmaceutical interventions.” Half a drug is a real thing. It is not a drug.

And the pool that found nothing. Yokoyama’s seven trials were mostly vegetarian, not vegan. When Tow and colleagues (2025) pooled only vegan trials, systolic pressure moved −1.45 mmHg (95% CI −4.93 to +2.02) and diastolic −0.65 (−3.13 to +1.84) — six trials, 489 people, and both intervals cross zero. That does not overturn Yokoyama; it is a smaller pool with wider intervals. It does mean the blood-pressure claim is better supported for vegetarian diets in general than for vegan diets specifically, and a page that quotes the first number should quote the second.

4 · Blood sugar

The current best evidence is Lv and colleagues (2025, Nutrition Reviews): 9 randomised trials, 681 people who already had type 2 diabetes.

One detail worth having: HbA1c only moved in trials longer than twelve weeks. Shorter studies found nothing.

And in people without diabetes, it barely moves at all: the 2025 vegan-only pool (Tow et al., 8 trials, 788 mostly non-diabetic adults) found HbA1c −0.07 points (95% CI −0.18 to +0.04), not significant. The effect is real where blood sugar is already high, and close to nothing where it is not — which is what you would expect, and not what the headlines say.

5 · Weight

The BROAD trial (2017) is the most-quoted number: a randomised trial in New Zealand where people ate a whole-food plant-based diet with no calorie counting and no portion limits — told simply to eat until satisfied.

At six months, the between-group difference was −10.6 kg (95% CI ±2.9) and −3.9 kg/m² of BMI. Cholesterol change was not significant (p = 0.1).

6 · What it does not show

This is the longest section on the page, and deliberately.

“A vegan diet reverses heart disease”

This is the single most overstated claim in plant-based health, and it rests on two studies that do not say it.

Ornish (1990, 1998) was a randomised trial, and it did find less arterial narrowing. But look at what the intervention was — five things at once:

• a low-fat vegetarian diet (about 10% fat)
• aerobic exercise, at least three hours a week
• stress management training, at least one hour a day — stretching, breathing, meditation, imagery
• smoking cessation
• group support meetings, four hours a week

Fewer than fifty people completed. No design on earth can tell you which of those five did the work, and the authors do not claim it could.

Esselstyn (2014) is weaker still, and it is the one shared most. It was not randomised and had no control group. 198 self-selected patients who asked for plant-based counselling; outcomes collected by telephone years later, and from family members for those who had died; imaging for only 22% of them.

The Stanford twin study

Landry and colleagues (2023, JAMA Network Open) is a genuinely clever design — 22 identical twin pairs, one twin to each diet, so genetics are held constant. It found LDL −13.9 mg/dL, fasting insulin −2.9 µIU/mL, weight −1.9 kg.

Now the authors’ own limitations, in their words: “Our study was not designed to be isocaloric; thus, changes to LDL-C cannot be separated from weight loss observed in the study.” Also: 44 people, 8 weeks, no follow-up, generally healthy participants, and meals were delivered for only the first four weeks.

Independent reviewers added that the vegan group ate roughly 200 calories a day less, and reported lower dietary satisfaction. It is a real randomised trial. It is not evidence that a vegan diet lowers cholesterol independently of eating less.

Mortality — where the cohorts actively disagree

One thing that explains a lot of the gap: in EPIC-Oxford the meat eaters ate about 79 g a day. These cohorts compare vegetarians against moderate meat eaters, not against a typical Western diet.

And the finding nobody on my side likes

7 · Where it carries real risk

Not theoretical risk. Measured, in vegans, in published data.

Vitamin B12 The clearest one

Niklewicz and colleagues (2024) pooled 930 vegans, 1,019 vegetarians and 1,166 omnivores:

Read the bold rows. The pooled vegan average for two functional markers sits past the deficiency threshold. Not one unlucky subgroup — the average. Supplemented vegans did significantly better on every measure.

Iodine The one nobody talks about

Eveleigh and colleagues (2023, British Journal of Nutrition), 11 studies, 4,421 adults: pooled vegan urinary iodine 12.2 µg/L, against a WHO severe-deficiency threshold of under 20 and an optimal range of 100–200. Pooled intake 17.3 µg a day against a requirement of 150.

That is not a marginal shortfall. If you have read a hundred vegan articles and none of them mentioned iodine, that is the point of this section.

Bones Real, and fixable

EPIC-Oxford (54,898 people, 17.6 years) found, in vegans against meat eaters: total fractures HR 1.43, hip fractures HR 2.31. In absolute terms, 14.9 more hip fractures per 1,000 over ten years against 2.9 for vegetarians.

But a 2025 meta-analysis found the modifier that matters: vegan women taking no supplements had nearly three times the hip-fracture risk — and combined calcium and vitamin D supplementation eliminated the excess. The 2025 Academy position paper says the same.

8 · The official position, and how it changed

The Academy of Nutrition and Dietetics replaced its position in February 2025. Almost every website still quotes the 2016 one.

Their own caveat is the sentence this whole site is built around: “vegetarian dietary patterns are about more than avoiding animal products, and, like all diets, should be planned appropriately to avoid common nutrient deficiencies.”

The other official position: the US Dietary Guidelines, 2025–2030

The federal guidelines were reissued in January 2026, and they moved in the opposite direction from most nutrition writing of the last decade. The document is short — nine pages — and worth reading in full rather than through anybody’s summary, including this one.

Three things to hold at once. The guidelines still describe a vegan diet as workable — whole foods, legumes, tofu, a supplement plan — which is exactly the plan on the nutrients section. They also now recommend butter and beef tallow on one page and a saturated-fat ceiling on another, and the two do not reconcile; several nutrition bodies said so in writing the week they came out. And the 1.2–1.6 g/kg protein figure is not where most of the trial evidence sits: it is at the upper end of what is studied for older adults and athletes, and well above what the people in the trials on this page were eating when their cholesterol fell. Quote the guidelines by all means. Quote the page number too.

9 · Sources

1. Koch et al., “Vegan diet and cardiometabolic risk factors,” European Heart Journal 44(28):2609, 2023 — academic.oup.com
2. Chiavaroli et al., portfolio diet meta-analysis, Progress in Cardiovascular Diseases, 2018
3. Yokoyama et al., “Vegetarian diets and blood pressure,” JAMA Internal Medicine, 2014 — structured abstract
4. Lv et al., Nutrition Reviews 83(8):1438, 2025 — academic.oup.com; Yokoyama et al., Cardiovascular Diagnosis and Therapy, 2014 — cdt.amegroups.org
5. Barnard et al., Diabetes Care 29(8):1777, 2006
6. Wright et al., the BROAD study, Nutrition & Diabetes, 2017 — nature.com
7. Ornish et al., The Lancet 336:129, 1990 — PDF; five-year follow-up, JAMA 280:2001, 1998 — PDF
8. Esselstyn et al., Journal of Family Practice 63(7), 2014 — PDF
9. Landry et al., the Stanford twin trial, JAMA Network Open, 2023 — jamanetwork.com; independent expert reaction — Science Media Centre
10. Tong et al., EPIC-Oxford IHD and stroke, BMJ 366:l4897, 2019 — bmj.com; Chiu et al., Neurology 94(11):e1112, 2020
11. Orlich et al., Adventist Health Study-2, JAMA Internal Medicine, 2013 — jamanetwork.com; Appleby et al., AJCN 103(1):218, 2016 — Oxford repository
12. Tong et al., EPIC-Oxford fractures, BMC Medicine, 2020 — springer.com; Ballarin et al., Nutrition Reviews, 2025 — academic.oup.com
13. Niklewicz et al., B12 status, Nutrition Bulletin 49:463, 2024 — PDF
14. Eveleigh, Coneyworth & Welham, iodine, British Journal of Nutrition, 2023
15. Reynolds et al., fibre and whole grains, The Lancet 393:434, 2019 — repository PDF
16. Academy of Nutrition and Dietetics position paper, 2025 — PDF; 2016 position — record
17. Tow, Looi, Nithusharini, Lakshmipathy, Palanisamy & Sundralingam, “The green plate effect: systematic review and meta-analyses of vegan diets and metabolic health in adults,” Trends in Food Science & Technology 164:105227, 2025 — sciencedirect.com
18. US Department of Agriculture and Department of Health and Human Services, Dietary Guidelines for Americans, 2025–2030, January 2026 — the nine-page document (vegetarians and vegans, p. 9; protein, p. 3; fats, p. 4)

1 · How to read this

Almost every page like this one is written to alarm you. This one is written to save you time, which means saying “this is not a problem where you live” as often as it says “check this.”

Every item carries one of three verdicts:

Where I could not find a source, the page says so rather than filling the gap with a number. Section thirteen is the full list of things I could not verify, and it is long, because this subject is full of figures everyone repeats and nobody can trace.

2 · India: the green dot is not a vegan mark

If you buy food in India, this is the most useful thing on this page, and it is written into the regulation in black and white.

Every packaged food in India carries a green circle or a brown triangle. Most people read the green one as “no animals were involved.” Here is what the Food Safety and Standards labelling regulations actually say:

“Non-Vegetarian Food” means an article of food which contains whole or part of any animal including birds, insects, fresh water or marine animals or eggs or products of any animal origin, but does not include milk, milk products, honey or bees wax or carnauba wax or shellac.

FSSAI, Compendium of Labelling & Display Regulations, Version VIII, 9 September 2025

What the marks are actually worth:

And note what carries no mark at all: the regulation exempts liquid milk, milk powders and honey entirely.

The one place the green dot is quietly useful is emulsifiers. Because the definition captures “products of any animal origin”, an animal-fat-derived emulsifier like E471 should trigger the brown mark in India, which is more than any European label will tell you. That is my inference from the wording, not a ruling I could find — treat it as a hint, not a guarantee.

3 · Sugar and bone char Depends where you live

The most-repeated claim on the internet, and the one most often overstated. Bone char is cattle bone, calcined and used as a decolourising filter bed that cane sugar passes through. Nothing of it remains in the sugar; the objection is to its use in the process.

Two facts defuse most of the panic:

• Beet sugar never uses bone char. The process is completely different — lime and CO₂, no refining stage. In the US, beet is 55–60% of domestic production.
• Bone char is a cane-refining technology only, and only in some refineries.

4 · Beer and wine The law is against you

This is the section where the honest answer is: you cannot tell from the label, and that is deliberate.

Beer and wine are cleared using finings — and the traditional ones are animal. Isinglass is collagen from fish swim bladders and is, in CAMRA’s words, “the most prevalent type used in brewing.” Wine uses gelatine, isinglass, egg white and casein at doses from 10 to 600 mg per litre.

Finings are process aids, not ingredients, so they do not have to be declared. And the exemption is explicit in EU and UK law:

“Fish and products thereof, except… (b) fish gelatine or Isinglass used as fining agent in beer and wine”

Regulation (EU) 1169/2011, Annex II, item 4 — the allergen labelling list

So isinglass never has to appear on a beer or wine label in the UK or EU. Egg and milk finings in wine must be declared only if they are still detectable in the finished bottle. In the United States nothing is required at all, though a rule proposed in January 2025 would change that — I have not confirmed whether it has been finalised.

The practical answers are: look for a vegan mark, choose unfined beers (New England IPAs and hazy styles are usually unfined by design), or check Barnivore, a crowd-sourced database of over 65,000 drinks built from replies producers gave the community. It is not certification and entries can be years old, but it is the best thing that exists.

5 · Bread, colours and coatings

L-cysteine / E920 Sometimes — and the “human hair” line is out of date

A dough conditioner made either by hydrolysing keratin — human hair, duck feathers, hog bristles — or by fermentation. The famous figure that 90% of the market is animal-derived is from 2011 and I found nothing current.

The UK picture is different: in 2013 the Food Standards Agency said the only L-cysteine its members would use is the synthetic kind, Asda and Allied Bakeries confirmed synthetic across their ranges, and industry said E920 is now very rarely used in UK bread. A vegan L-cysteine exists commercially. And in the UK and EU it must always be labelled — look for “L-cysteine” or “E920”. If it is there, the source is not stated and only the manufacturer can tell you.

Carmine / cochineal / E120 True — and easy to spot

Crushed scale insects, used as a red colour. This is one of the very few hidden animal ingredients that is legally required to be named in both the US and the EU. The FDA banned generic “artificial colour” for it from January 2011; it must say “cochineal extract” or “carmine”. In the EU it is E120, carminic acid, carmines, Natural Red 4 or CI 75470.

Treat this one as reassuring rather than alarming: it is on the label if it is in there.

Shellac / E904 True — but not what you have been told

A resin secreted by the lac insect on trees in India and Southeast Asia, used to glaze sweets, sprinkles, panned chocolate, coated fruit and pills. On labels: shellac, E904, confectioner’s glaze, resinous glaze, food glaze, pharmaceutical glaze, lac resin.

A precision point, because this gets said badly: refined shellac is a secretion, not crushed insects. The raw resin is dissolved, heated and filtered to remove debris and insect parts. Harvesting it does kill lac insects, which is why it fails a vegan standard — but “your sweets are coated in crushed bugs” is carmine, not shellac. Say the true thing.

Milk, hiding in plain sight True, but findable

Whey, casein, caseinate, lactose and milk solids must be declared in the UK, EU and US, so this is a reading problem rather than a hidden one. The Vegan Society’s list of places people forget to look: soya milk (some manufacturers have added whey powder), breakfast cereal, crisps, and “dairy-free” chocolate that is really just milk-chocolate-adjacent.

6 · Cheese and rennet Sometimes — and Indian paneer (cottage cheese) is the exception

Two European cheeses are legally required to use calf rennet, and I checked both specifications directly. Parmigiano Reggiano: “Milk coagulation, obtained exclusively by the use of calf rennet.” Grana Padano: “Coagulation is achieved using calves’ rennet.”

Grana Padano has a second animal ingredient almost nobody mentions: its specification permits lysozyme — from egg white — up to 2.5 g per hundred kilograms of milk, everywhere except Trentingrana. So standard Grana Padano is not even egg-free, and Trentingrana is the lysozyme-free designation.

The shopper’s line: generic “Parmesan” may well be microbial. “Parmigiano Reggiano” is calf rennet by law. I did not check Gorgonzola, Pecorino Romano, Manchego, Roquefort or Comté, so I am not going to list them.

7 · Vitamins

8 · Two more Indian ones almost nobody covers

Silver leaf — vark Historically true, now unclear

The silver leaf on mithai was traditionally beaten between layers of ox gut or cow hide, because the leaf separates more cleanly from animal tissue than from paper.

In 2016 FSSAI prohibited animal-origin material in silver leaf manufacture — silver leaf “should not be manufactured using any material of animal origin at any stage” — and approved a route using specially treated paper coated with food-grade calcium powder. Traditional manufacturers challenged the regulation and a High Court stayed it in July 2017; FSSAI was still pushing enforcement in 2018.

I could not establish how that litigation ended, so I am not going to tell you vark is now safe. The industry has mostly converted, but vark on mithai from a halwai you do not know cannot be assumed animal-free. Ask, or scrape it off.

Ayurvedic products True, and outside the labelling system entirely

Beauty Without Cruelty India: “It is a common misconception that Ayurvedic products are always vegetarian. Today, they commonly contain about 50 different animal substances.”

Named ones include gorochana (bile of ox or cow), musk from deer, conch, pearl oyster, coral, deer horn, egg shell, lac, beeswax, honey and ghee. Named formulations with goat meat include Ajamamsa Rasayana and Aswagandhadi Lehyam.

The regulatory gap is the real story. Ayurvedic products fall under the Drugs and Cosmetics Act, not FSSAI, so the green and brown marks are not mandatory on them at all. BWC cites a Patanjali badam pak containing pearl oyster while displaying a green vegetarian symbol.

9 · Not food at all

10 · What vegans get wrong

Every claim in this section is one I have seen repeated by people on my own side. Getting these wrong is how a movement loses arguments it should win.

Figs and the wasp — and the correction is also wrong

You have heard that figs are not vegan because a wasp dies inside every one. You may also have heard the rebuttal: that commercial figs are self-pollinating and there is no wasp. Both of those are wrong, and the true version is more interesting than either.

The biology. Common figs — Brown Turkey, Celeste, Black Mission, Kadota, Adriatic — are parthenocarpic and set fruit with no pollination at all. For those, in the words of a University of Florida extension service: “chances of finding a wasp inside a commercial common types fig fruits is zero.”

But Smyrna-type figs are not a rounding error — they dominate the world dried-fig trade. Turkey supplies over half the world’s dried figs, and its main cultivar, Sarılop, is a Smyrna type for which “caprification is compulsory.” In California, Calimyrna — the same variety under another name, and wasp-pollinated — was about 45% of fig acreage in the 1998 crop profile, which is the most recent breakdown I could find.

And even in those, there is no wasp. The female enters, pollinates, and dies — she cannot reach the long female flowers to lay in, so she leaves seeds rather than offspring. The fig then produces ficin, an enzyme that digests her, and absorbs the nutrients.

The rest

11 · A hundred things that surprise people

A hundred things people get wrong in both directions: things that are quietly not vegan, and a few that are vegan when everyone assumes they are not. Short version first; the sections above have the long ones.

Three verdicts here: Not vegan means animal-derived as usually sold. Check means it depends on the brand, the country or the restaurant. Vegan means it surprises people by being fine.

Food and cooking 26

Most of these are not about the obvious ingredient: they are about the stock, the fat, the fish sauce or the glaze that nobody mentions on the menu.

Sweets and snacks 12

Gelatin, milk, honey and two shiny coatings (shellac and beeswax) do most of the damage here.

Drinks 5

The drink itself is plant, but how it is cleared, sweetened or topped can be animal.

Medicine and supplements 16

Never stop or swap a prescribed medicine on your own. The Vegan Society’s own definition says “as far as is possible and practicable”, and that includes your health. Ask your pharmacist whether a vegan version exists.

Clothes and accessories 11

If it is soft, warm or shiny and came from nature, check it. No need to throw out what you own; just buy differently next time.

Beauty and personal care 12

Two labels help: “vegan” for ingredients and the Leaping Bunny for animal testing. They are not the same thing.

Home, hobbies and the rest 6

The less obvious corners, from the kitchen cupboard to the music room.

Surprisingly vegan 12

And the good news: things people assume are not vegan, but are.

12 · E-numbers on the label

E-numbers are the European system, also used in the UK, India (as INS numbers, the same digits without the E), Australia and much of the world. US labels use the names instead, so both are here.

Most E-numbers are plant or synthetic. These are the ones worth knowing: nine that are always animal, and the fat-based emulsifiers that can go either way. For a Check, only a vegan label or the maker can tell you which.

13 · What I could not verify

This subject is full of numbers everyone repeats and nobody can trace. Here is what I looked for and did not find. If you know better than me on any of these, tell me and I will correct it.

• Any industry-wide figure for how much US cane sugar goes through bone char. No such number exists in any source I found. ASR’s own 65% covers one company.
• Whether the Crockett and Arabi refineries still use bone char in 2026. Last confirmed 2021 — and the same goes for the lot-code trick.
• The current share of L-cysteine from keratin versus fermentation. The 90% figure is from 2011.
• Whether human hair is legally barred as an E920 source in the EU. Widely repeated online; I could not retrieve the operative text from three separate routes.
• Which PDO cheeses beyond Parmigiano Reggiano and Grana Padano legally require animal rennet.
• What coagulant Indian processed cheese uses.
• The final legal status of the 2016 vark regulation after the 2017 stay.
• Whether India’s brown mark is actually applied to animal-derived emulsifiers. That was my inference from the wording.
• Anything on tyres, candles or fabric softener beyond blogs. The weakest items here.
• Any silk figure — silkworms per garment, or how common ahimsa silk really is.
• Whether the January 2025 US allergen-labelling rule for alcohol has been finalised.
• Dates for the Skittles gelatin and carmine removals. Secondary sources contradict each other.
• A single specific Indian packaged food with a genuinely hidden animal ingredient. I looked. I did not find one. What I found instead was the opposite — a viral claim that turned out to be false.

14 · Sources

1. FSSAI, Compendium of Labelling & Display Regulations, Version VIII, 9 September 2025 — fssai.gov.in; legal commentary — Obhan & Associates, 2021
2. Bone char and US refineries — Vegetarian Resource Group, 2021; PETA US; beet/cane production share — USDA ERS
3. Organic rules — 7 CFR 205.605; EU comparison — VRG, 2023
4. UK sugar — Tate & Lyle FAQ; PETA UK, January 2026
5. Indian sugar and jaggery processing — SugarProcessTech; Beauty Without Cruelty India
6. Finings — CAMRA, 2024; Australian Wine Research Institute, 2024; Barnivore
7. Allergen labelling exemption — Regulation (EU) 1169/2011, Annex II; wine fining declaration — Food Safety Authority of Ireland; US proposed rule — Federal Register, January 2025
8. Guinness timeline — Barnivore; VRG, 2017
9. L-cysteine — VRG, 2011; UK position — UK Food Guide, 2013
10. Carmine labelling — FDA final rule, 2009; shellac — USDA AMS technical report, 2002; STAR-K, 2016
11. E-number watch list and hidden dairy — The Vegan Society
12. Vitamin D3 — Nutritional Outlook; B12 — VeganHealth.org, 2023; vaccine excipients — Johns Hopkins, October 2024
13. FPC prevalence — VRG, 2012, reconfirmed Genetic Literacy Project, 2024
14. PDO specifications — Parmigiano Reggiano; Grana Padano
15. Paneer (Indian cottage cheese) coagulation — NIFTEM-T, PMFME
16. Vark — FSSAI media compilation, 2018; the stay — Deccan Herald, July 2017
17. Ayurvedic ingredients — Beauty Without Cruelty India; EasyAyurveda
18. Fig biology — UF/IFAS Extension, 2023; Turkish production — IntechOpen, 2019; California acreage — Crop Profile for Figs in California, 1999
19. “Natural flavor” definition — 21 CFR 101.22
20. Cadbury India fact-check — Business Today, 2021; Fact Crescendo, 2023
21. No legal definition of “vegan” — Chartered Trading Standards Institute, 2023; certification standards — Vegan Action
22. Stearic acid — VRG, 2015; bone china — Wedgwood

1 · Why I am the one writing this

Because when a white vegan tells an Indian family their dairy is cruel, the conversation is over before it starts — and they are right that it is over. It sounds like one more person who does not understand the house telling them how to run it.

I cook Indian food for a living. I make paneer (Indian cottage cheese), ghee, khoya, dahi (yogurt), rasmalai and gulab jamun without a cow, and I make them well enough that people ask for the recipe before they ask what is in them. My argument is not that the tradition is wrong. It is that the tradition made a promise about how the animal would be treated, and the modern industry does not keep it.

Everything below is sourced. Section five is the case against me, researched as hard as the rest, and it contains two concessions I would rather not make.

2 · What the texts actually say

The Hindu texts are in tension with themselves, and pretending otherwise is dishonest

Ahimsa is the first of the five yamas in the Yoga Sutras (2.30). The Mahabharata says it plainly — ahimsa paramo dharma, non-injury is the highest dharma (Anushasana Parva 13.116). The Bhagavad Gita lists it among the divine qualities (16.2) and as a constituent of knowledge (13.7–10).

The Manusmriti contains the sharpest statement of complicity I know of in any religious text:

“He who permits (the slaughter of an animal), he who cuts it up, he who kills it, he who buys or sells (meat), he who cooks it, he who serves it up, and he who eats it, (must all be considered as) the slayers (of the animal).”

Manusmriti 5.51, Bühler translation

That is the entire supply-chain argument, written down two thousand years before anyone said “supply chain.” The buyer is a slayer.

And the older layer is more complicated still

The historian D. N. Jha — Professor of History at Delhi University, a member of the Indian Council of Historical Research, an NCERT textbook author — argued in The Myth of the Holy Cow (2001) that beef was part of the early Indian diet, that cattle were routinely offered in Vedic sacrifice, that the Charaka Samhita prescribes beef medicinally, and that the Brahmanical prohibition on cow-killing developed around the middle of the first millennium CE rather than being present from the start.

Two things must be said about that book, and both matter.

It is an argument, not a survey. Jha set out to demonstrate a thesis and the book reads that way. Reviewers found it dense and hard going for anyone not already inside the Indological literature.

And publishing it was dangerous. Jha received death threats. Hindu nationalist organisations called the book blasphemous. A civil court in Hyderabad restricted its circulation. The Indian publisher came under enough pressure that Verso brought out the international edition from London. I am not going to pretend this is a neutral academic question in India, because it is not. But a claim being politically explosive is not the same as a claim being false, and the historical position Jha describes is the mainstream one among historians of the period.

The Jain texts are less ambiguous

Ahimsa is the first and governing vow in Jainism. Two passages are usually cited on dairy specifically — the Uttarādhyayanasūtra 17.15, which calls one who eats milk and curds a bad śrama&nfrown;a, and the Puruśārthasiddhyupāya verse 71, which lists butter among the great degenerative substances not to be eaten.

The more interesting argument is structural. The Tattvārthasūtra lists five aticāras — transgressions of the ahimsa vow: tethering, beating, piercing the skin, overloading, and withholding food.

The scholar Christopher Jain Miller makes the argument I find hardest to answer: when the scriptures permitted dairy, they were permitting something that no longer exists. Milk taken as surplus from an animal who was not bred to order, not inseminated, not separated from her calf, and not sold on when she stopped producing. The permission assumed a relationship. The relationship is what changed.

3 · Where gau mata comes from

Reverence for the cow is genuinely ancient. The Vedic term aghnya — “not to be killed” — is real and old.

But the political mother-cow, the one that appears on posters beside Bharat Mata, is much more recent than most people assume, and the dates are not in dispute:

Historians of late-colonial North India read the mother-cow icon as constructed in that period, alongside Bharat Mata and Matri Bhasha, as part of a political project. The reverence is old. The movement is about a hundred and sixty years old. Both of those are true and neither cancels the other.

4 · What happens now

These are government figures wherever government figures exist, and I say so where they do not.

Artificial insemination — and the claim I will not make

You will read that Indian dairy cows are artificially inseminated. Most of them are not. The Department of Animal Husbandry and Dairying says so in its own guidelines:

“The average A.I coverage of the country is only 30%… Rest of the 70% are either left unbred or bred with scrub bulls of unknown genetic merit.”

DAHD, Guidelines for the Nationwide Artificial Insemination Programme Phase III, 2021

The national target is 70%. As of March 2025 the programme reported 83.9 million animals covered and 123.4 million inseminations performed. So the accurate statement is not “Indian cows are inseminated” — it is that artificial insemination is official state policy, expanding fast, and deliberately engineered toward female calves: 5.87 million doses of sex-sorted semen produced at government facilities, 90% accurate for females, with the price being cut from ₹800 to ₹250 a dose.

That last fact is the one worth sitting with. The state is spending public money to make sure fewer male calves are born, because of what happens to male calves.

Male calves

The USDA's Economic Research Service, in a report on India's beef exports, states that about 70% of male calves die or are slaughtered within a short period of birth.

What is not an estimate is the language used about male calves in Indian animal-science journals, written by meat scientists arguing they should be better utilised rather than by animal advocates. They describe male calves as “considered as burden and left as such as stray animals,” neglected “with a view to save on mother's milk and feed resources,” and “kept with dispassion, not nourished properly.”

Where the cows go

Nine states and territories ban the slaughter of cows and their progeny outright — Delhi, Gujarat, Haryana, Himachal Pradesh, Jammu & Kashmir, Punjab, Rajasthan, Uttar Pradesh, Uttarakhand. Penalties are severe: Gujarat's amended Act carries a minimum of ten years and a maximum of life.

Assam permits slaughter above ten years of age; West Bengal above fourteen. Kerala, Arunachal Pradesh, Meghalaya, Mizoram and Nagaland have no cow-slaughter legislation at all.

An animal that cannot legally be killed in Rajasthan and can be in Kerala does not stop existing when she stops producing milk. The internal movement is widely reported and I could not find a reliable quantified figure for it, so I am not going to invent one. Cross-border movement into Bangladesh is documented in Border Security Force seizure data, which shows a sharp decline — from 10,600 head seized on the Meghalaya frontier in 2020 to 3,644 in 2023. Seizure figures are not trade figures, and they are the government reporting on its own success. Treat them as directional.

Oxytocin

Injecting oxytocin to force milk let-down is specifically illegal in India. Section 12 of the Prevention of Cruelty to Animals Act 1960 prohibits phooka and, since the 1982 amendment, “any other operation (including injection of any substance) to improve lactation which is injurious to the animal's health.” The penalty is up to two years, a fine, and forfeiture of the animal.

For comparison, the general cruelty provision in Section 11 still carries a fine of ₹10 to ₹50 for a first offence — an amount unchanged since 1960.

How common is it? The Government of India has stated that scientific data on the extent of the practice is not available. The only figure I found is from FIAPO, an Indian animal-advocacy federation, which investigated 451 dairies across ten states in 2017 and reported oxytocin use at 47% of them. That is one NGO's investigation, not a national statistic, and I am labelling it as such.

Gaushalas — and a real problem with the numbers

Uttar Pradesh goes from 582 to 8,112 in twenty months. That is a change in what is being counted, not a building programme — but I could not confirm the definitional change, so both figures are here with their dates.

The government also states plainly: “The data regarding number of cows in those gaushalas are not maintained.” Nobody knows how many animals are in them.

The one peer-reviewed study of gaushala management — 54 shelters, median 232 cattle each — found median annual mortality of 13.8%, against 1–5% on Western dairy farms. 41% provided pasture access. 18% disposed of carcasses in open spaces or creeks. 91% allowed unrestricted breeding.

And a Jain-studies academic makes the observation that stayed with me longest: religiously run gaushalas commonly sustain themselves by selling milk — so a shelter for retired cows funds itself by milking the younger ones.

5 · The case against me

Researched as hard as everything above. Two of these I cannot answer.

Indian dairy really is structurally different, and it is not close

What an Indian dairy animal actually eats

This single chart is the reason Western criticism of Indian dairy so often misfires. Almost two-thirds of the diet is crop residue — the stalks and husks left after a harvest, which no human was going to eat.

And why it is still more carbon-intensive per litre

The same animal that eats waste also produces about a quarter as much milk. She is maintained year-round either way, so her emissions are divided across far fewer litres.

This is not a confined feeding operation. It is an animal living with a family, eating what would otherwise be waste, in a system that converts feed to protein eight times better than the American one. Anyone who argues about Indian dairy without knowing this will be dismissed by Indian readers in the first paragraph, and should be.

The honest counterweight from the same body of research: that Kerala study found 80% of cows in normal body condition and 10.4% lame — alongside the universal nose-piercing and tethering. Both pictures are true at once.

Eighty million livelihoods, and most of them women

Dairy employs more than 80 million farmers in India. Nearly 70% of the dairy workforce is women. There are more than 48,000 women-led village dairy cooperative societies. The cooperative network alone has around 17 million farmer members across 230,000 villages, and dairy contributes roughly 5% of the national economy.

I do not have an answer to this. A transition that impoverished seventy million rural women would not be ahimsa by any reading of the word, and any vegan who waves this away has not thought about it.

Children's growth

An Indian study across two national surveys — 24,670 children in NFHS-3 and 132,767 in NFHS-4 — found milk intake associated with higher height-for-age. A cross-country study of 673,177 observations across seventy low- and middle-income countries found milk consumption associated with a reduction in stunting of about 1.9 percentage points.

The authors' own caveat matters: the effect was stronger in wealthier quintiles, and milk may be standing in for better overall nutrition rather than causing the difference. But in a country with India's rates of childhood undernutrition, this is not a consideration to skip past.

And the sharpest criticism comes from a sympathiser

Yamini Narayanan's Mother Cow, Mother India (Stanford, 2023) is the serious scholarly version of this whole page — that the sacred “mother cow” and the commodity “dairy cow” are two constructs that between them hide what actually happens.

Reviewing it in Contributions to Indian Sociology, Vasundhara Bhojvaid accepts the diagnosis and rejects the conclusion. Her argument is about caste: Dalits and religious minorities perform the “bad touch” of slaughter and carcass disposal while upper castes perform the “good touch” of milk. On that reading the problem is a caste structure, and veganism is a market solution to something markets did not create and cannot fix.

I do not think that is right, but I think it is the strongest objection in the literature, and it comes from inside the same tradition of criticism rather than from the dairy industry.

6 · The Jains who already moved

This is not a Western idea arriving in India. It is already happening, and it is happening in the tradition with the strictest ahimsa doctrine of all.

Individual ascetics have gone further. Muni Shri Vihar Sagarji Maharaj called on Jains to discontinue milk in 2014. Aryika Shree Chaitanyamati Mataji, in 2021: “Milk is not vegetarian but it is non-vegetarian and this is why we have renounced it as well.”

7 · What there is to eat

India's plant-based dairy market was about ₹250 crore in 2021 and is projected at ₹891 crore for 2025, with more than 45 brands and 223 products across milk, yoghurt, butter and cheese, according to the Good Food Institute India. Most Indian consumers who buy it use both, and most say they will pay roughly a 20% premium.

But the market is the least interesting part of the answer, because the cuisine already contains most of what you need. Idli, dosa and the whole fermented South Indian repertoire are dairy-free by construction. Every dal. Chana, rajma, chhole. Most sabzi. Roti, paratha made on oil, poori. Coconut milk through the entire South. The dairy in Indian food is concentrated in a much smaller set of places than people assume — chai, paneer (Indian cottage cheese), ghee, dahi (yogurt), and the mithai — and every one of those five now has a version I can make and would serve to anyone.

I am deliberately not citing a source for that paragraph. It is what I know from cooking, not something I read, and I would rather say so than dress it up.

8 · What I am not saying

• I am not saying Hinduism or Jainism is wrong. My argument is that both traditions set a standard for how an animal must be treated, and the industry does not meet it. That is an argument for the tradition, not against it.
• I am not saying Indian dairy is the same as American dairy. It plainly is not, and section five says so with numbers.
• I am not saying every Indian dairy animal is mistreated. Many are loved. That was never the question — the question is what happens to her calf, and to her when she stops producing.
• I am not saying eighty million people should lose their livelihood. I do not know what the answer to that is, and I said so.
• And I am not going to tell you what you owe anyone. Nobody was ever argued into this, including me. I was fed.

9 · Sources

1. Manusmriti 5.48, 5.51, 5.56 — Bühler, The Laws of Manu, Sacred Books of the East vol. 25, 1886 — sacred-texts.com
2. Yoga Sutras 2.30, 2.35; Mahabharata, Anushasana Parva 13.116; Bhagavad Gita 16.2, 13.7–10, 17.14 — verse references via Hindupedia and Centre for Yoga Studies
3. Christopher Jain Miller, “Are Jains Vegan? What Do Jain Scriptures Say About Dairy?” Arihanta Institute — arihantainstitute.org
4. D. N. Jha, The Myth of the Holy Cow, Matrix Books 2001 / Verso 2002; reception and the circulation order — Wikipedia; reviewed in Journal of the American Academy of Religion 72(1), 2004
5. Cow protection movement, 1860s–1893 — Wikipedia; Charu Gupta, “The Icon of Mother in Late Colonial North India” — researchgate.net
6. DAHD, Guidelines for Nationwide Artificial Insemination Programme Phase III, 2021 — dahd.gov.in; Lok Sabha Q.1946, 11 March 2025 — dahd.gov.in; Rashtriya Gokul Mission — pib.gov.in
7. Landes, Melton & Edwards, “From Where the Buffalo Roam: India's Beef Exports,” USDA ERS LDPM-264-01, 2016 — ers.usda.gov
8. Jairath et al., “Buffalo calf: an emerging meat source in India,” MOJ Food Process Technol 2017 — medcraveonline.com; Kumar & Yadav, Journal of Animal Research 2021 — PDF
9. Cow-slaughter legislation by state — Factly; USDA GAIN, “India — Cattle Slaughter Legislations,” 2015
10. Prevention of Cruelty to Animals Act 1960, ss. 11–12 — Animal Legal & Historical Center, Michigan State; oxytocin regulation and litigation — ThePrint
11. FIAPO, Cattle-ogue, 451 dairies across 10 states, 2017 — summary
12. 20th Livestock Census, stray cattle — PIB; DAHD Compendium 2022-23 — PDF
13. Gaushala counts — Lok Sabha Q.3423, 21 March 2023 — dahd.gov.in; Basic Animal Husbandry Statistics 2024 via Lok Sabha reply, 19 December 2024
14. Sharma, Schuetze & Phillips, “The Management of Cow Shelters (Gaushalas) in India,” Animals 2020;10(2):211 — PDF; Jonathan Dickstein on gaushala economics — Arihanta Institute
15. Mullan et al., “Welfare Challenges of Dairy Cows in India Identified Through On-Farm Observations,” Animals 2020;10(4):586 — mdpi.com
16. Herd structure and smallholder economics — Meeta Punjabi for FAO — fao.org; feed efficiency and manure valuation via Scroll.in / PASTRES
17. Dairy livelihoods and women's participation — PIB, India's Dairy Sector; PIB, World Milk Day
18. Thomas et al., “The role of dairy consumption in the relationship between wealth and early life physical growth in India,” BMC Public Health 2024;24:96 — bmcpublichealth.biomedcentral.com; Herber et al., Scientific Reports 2020;10:6730 — nature.com
19. Yamini Narayanan, Mother Cow, Mother India, Stanford University Press 2023 — sup.org; reviewed by Vasundhara Bhojvaid, Contributions to Indian Sociology 2024 — sagepub.com
20. Jain vegan organisational pledges — Jain Vegans UK; ascetic statements — 2014 and 2021
21. Good Food Institute India, plant-based dairy category insights, 2024 — gfi-india.org

1 · The scale of it

ICMR-INDIAB, India’s national survey, covered 113,043 people across 31 states. What it found:

The survey’s lead author on what happens next: “conversion to diabetes is very, very fast in our population; more than 60% of people with pre-diabetes end up converting to diabetes in the next five years.”

2 · What the national survey found

In 2025 the same group published the dietary analysis — 18,090 adults, nationally representative, in Nature Medicine. This is the most important source on this page, and it is Indian data about Indians.

And the associations, comparing the highest carbohydrate consumers to the lowest:

Highest against lowest refined-cereal consumers: 18–25% higher likelihood of type 2 diabetes. For white rice specifically: 19–26%.

What did associate with lower risk was replacing carbohydrate with protein — and notably, plant and pulse protein performed as well as anything else. Dairy protein: diabetes OR 0.89, prediabetes 0.82. Pulse protein: similar. Fish and egg protein: 0.91.

The authors’ own recommendation is “reduced consumption of refined carbohydrates and saturated fat” with “increased consumption of protein, primarily from plant and dairy sources.”

So India’s national data does not say the problem is the absence of meat. It says the problem is carbohydrate volume, refined cereal, and not enough protein — every one of which a well-built plant-based plate fixes better than an Indian vegetarian plate currently does.

3 · Why Indian vegetarianism buys so little

The Indian Migration Study measured it directly — 6,555 people, including sibling pairs, across four Indian cities. Vegetarians against non-vegetarians:

Being vegetarian in India buys you about 0.7 mmHg of diastolic blood pressure. That is a real number and it is almost nothing.

The authors explain why, and it is the key to the whole page: Indian non-vegetarians eat a median of 20.3 g of meat a day, against 49–65 g in the Western studies. You cannot get a large benefit from removing something the comparison group barely eats.

4 · Thin is not safe

A cohort of about 1.5 million people in the UK worked out the BMI at which each ethnic group carries the same type 2 diabetes risk as a white person at BMI 30.

If you are South Asian, a BMI of 24 is not a clean bill of health, and a normal-looking BMI is one of the commonest reasons the diagnosis arrives late.

5 · B12 is not a vegan problem in India

A meta-analysis of 270 Indian studies found deficiency prevalence in the general population — vegetarian and non-vegetarian alike:

Roughly half of all Indians are B12 deficient. Six in ten are vitamin D deficient. This is a population-level problem that predates anybody’s decision to go vegan.

Two things follow, and they point in opposite directions. A vegan diet makes it worse — and it is not the origin of it. If you are Indian and you have never had B12 tested, the odds you are already low are close to a coin flip, whatever you eat.

6 · What follows

Taking India’s own data at face value, rather than importing a Western argument.

• Less rice and roti, not different rice and roti. The national survey is explicit that swapping the grain without cutting the quantity did nothing.
• More protein, and pulses count. Protein is 12% of Indian energy intake against a 15% recommendation, and pulse protein performed as well as animal protein in the substitution analysis. ICMR-NIN recommends 85 g of uncooked pulses a day.
• Watch the oil and the ghee, not the absence of meat. Saturated fat is 8.9% of energy nationally and exceeds recommendations in most states.
• Get B12 tested, then supplement. Not because you are vegan — because you are Indian.
• Use a lower BMI threshold for yourself than the chart on the clinic wall.

7 · Sources

1. Anjana et al., ICMR-INDIAB-17, The Lancet Diabetes & Endocrinology 11(7):474, 2023 — 113,043 people across 31 states
2. Anjana, Sudha, Bhupathiraju, Mohan et al., “Dietary profiles and associated metabolic risk factors in India,” Nature Medicine, 2025 — nature.com
3. Shridhar et al., the Indian Migration Study, PLOS ONE, 2014 — plos.org
4. Misra, Balagopal, Raj & Patel, “Vegetarian Diet and Cardiometabolic Risk among Asian Indians in the United States,” Journal of Diabetes Research, 2018
5. Caleyachetty et al., ethnicity-specific BMI cut-offs, The Lancet Diabetes & Endocrinology, 2021 — summary
6. Venkatesh, Sharma, Ananthan et al., “Micronutrient deficiencies in India,” Journal of Nutritional Science 10, 2021 — cambridge.org
7. ICMR-NIN, Dietary Guidelines for Indians, 2024, and Nutrient Requirements for Indians, 2020 — nin.res.in

Nothing matches that. Try a shorter word — milk, egg, ghee, gelatin, iron.