The argument for raw milk says it is alive and that pasteurized milk has been cooked into something inert. The argument against it usually says nobody tests raw milk at all. The second claim is the weaker of the two, because where retail raw milk is legal it is tested, often on a tighter schedule than pasteurized milk.
The useful question sits one level down. The tests exist. They measure something else.
What the required tests actually measure
State standards for retail raw milk are built on indicator organisms. The numbers move from state to state, which is itself informative, but the shape is consistent:
- Standard plate count, commonly capped at 20,000 per mL for packaged retail raw milk, 30,000 per mL in South Dakota. This counts how many bacteria are present in total.
- Coliform count, commonly capped at 10 per mL. This is a proxy for fecal contamination and general hygiene.
- Somatic cell count, capped at 400,000 per mL for cows and 1,500,000 per mL for goats in some states, 750,000 and 1,000,000 in Texas. This reports udder health, mainly mastitis.
Each of those is a real measurement, honestly performed. None of them identifies Salmonella, Campylobacter, Shiga toxin-producing E. coli or Listeria monocytogenes. A plate count tells you how many organisms are in the sample. It does not tell you which.
The structure repeats a pattern worth recognizing. In honey fraud, the glass of water and the match measure moisture rather than syrup, and the test is sound while the conclusion drawn from it is not. Here a clean indicator panel gets read as a clean bill of health.
Where pathogens are tested, the schedule is the problem
Some states go further. Grade A raw milk carries a pathogen limit of zero, with samples tested for pathogenic bacteria at least twice every six months. That sounds decisive until you ask what a passing result certifies.
Shedding is intermittent. A cow carrying Salmonella or STEC does not excrete it steadily, and contamination can also enter during milking, from hands, equipment, bedding or water, independently of herd health. A pathogen test performed twice in six months certifies a moment. It cannot certify the bottle in your refrigerator, which was filled on some other morning.
That is the real gap, and it is structural rather than a matter of anyone cutting corners. Routine testing is frequent and measures indicators. Pathogen testing is specific and infrequent. Nothing in that arrangement follows a single carton from udder to kitchen.
The number that quantifies the gap
Costard and colleagues modeled outbreak-related disease burden from unpasteurized cow’s milk and cheese in the United States for 2009 to 2014, published in Emerging Infectious Diseases. Dairy outbreaks overall caused an average of 760 illnesses and 22 hospitalizations per year. The split is where it gets interesting.
Unpasteurized dairy products caused 840 times more illnesses (95% interval 611 to 1,158) and 45 times more hospitalizations (interval 34 to 59) than pasteurized products, per unit consumed. Raw milk was consumed by 3.2% of the population and raw cheese by 1.6%, and between them they accounted for 96% of all illnesses linked to contaminated dairy. The same model projects that doubling consumption of raw milk or cheese would raise outbreak-related illnesses by 96%.
Those figures are ratios per unit consumed, which is what makes them portable. The consumption shares are specific to the United States in that period and should not be lifted into another country or another decade.
Somatic cells cannot see Salmonella
The most common misreading goes like this: the somatic cell count is low, therefore the milk is clean. A somatic cell count reports inflammation in the udder. A high number means mastitis. A low number means no mastitis. It says nothing in either direction about whether a pathogen is present.
Plate counts have the same limitation from the opposite side. Milk can sit comfortably under 20,000 per mL and still carry enough E. coli O157:H7 to cause serious illness, because the infectious dose for STEC is very low. Conversely milk with a high plate count may be merely stale.
The cold chain nobody logs
Time and temperature after milking do the rest of the work. Listeria monocytogenes grows from 31.3°F (minus 0.4°C) upward, so refrigeration slows it rather than stopping it. Meanwhile the route from farm to kitchen is rarely instrumented: a cooler in a truck, a farmers market table, a car, a tote bag, and finally a home refrigerator where the measured average is 42.3°F and the top shelf runs 45.9°F.
An industrial chain chills milk immediately after milking and holds it cold until pasteurization. It is the least glamorous part of the process and the part that disappears first when milk moves through informal channels.
What pasteurization does, and what it costs
The standard treatment is 162°F (72°C) for 15 seconds. That is enough to kill vegetative cells of all four pathogens above. It does not kill spores and it does not remove a toxin that has already formed, so it closes one specific class of risk rather than all risk.
Now the claim about destroyed nutrition, which is testable. Macronutrients come through unchanged: protein, fat and carbohydrate. Minerals come through unchanged, calcium and phosphorus included. The fat-soluble vitamins A, D and E and beta-carotene are heat stable and survive. Riboflavin is unaffected. The documented losses are roughly half the vitamin C, about 10% of thiamine, and under 10% of vitamin B12.
Half the vitamin C sounds significant until you remember that milk was never a meaningful source of vitamin C. A glass contains so little that losing half of it changes nothing in a diet. The nutritional case against pasteurization rests on the one nutrient milk barely has.
The processor has a motive, the cooler at the market does not
One more link in the chain is visible only up close. Dairy plants screen incoming milk for antibiotic residues, and not because a rule says so. Residues inhibit starter cultures, so contaminated milk will not become yogurt or cheese and ruins the batch it enters. How that screening works is covered separately in the piece on antibiotics in meat and milk.
A direct seller has no equivalent exposure. The milk goes to a customer rather than into a process, so nothing downstream fails loudly when something is wrong. That is a description of incentives rather than an accusation about any individual farmer.
Goat milk and Toxoplasma
One side route deserves accurate weighting. Transmission of Toxoplasma gondii through unpasteurized goat milk is documented but rare. There is a British goat herd report in which four of six goats were seropositive, the parasite was isolated from one animal’s milk, and the owner’s son developed infection with serological confirmation. A separate case report appeared in 2025. Molecular surveys have found parasite DNA in one of 27 raw goat milk samples.
It is a genuine route and a minor one. Undercooked meat and contaminated water remain the dominant exposures for Toxoplasma, and presenting goat milk as the leading source would be its own kind of exaggeration.
What to do
- Read a passing test as what it is. Plate count, coliform and somatic cell results describe total load, hygiene and udder health. There is no line on that report about pathogens.
- Heat it if it came through an informal channel. Holding milk at 162°F for 15 seconds, or simply bringing it to a boil, removes exactly the class of risk the routine panel does not cover.
- Treat raw-milk cheese as its own decision. In the model, raw cheese produced nearly as many illnesses as raw milk while being eaten by half as many people.
- Shorten the time to cold. Raw milk belongs in the refrigerator before the rest of the shopping, on a lower shelf rather than the door.
- Children, pregnant people and adults over 65 should skip it. In these groups Listeria and Salmonella produce severe disease rather than a bad week.
The short version
Retail raw milk is tested, and the testing is competent. It measures plate count, coliforms and somatic cells, which is to say total bacterial load, hygiene and udder health. Where pathogen testing is required, it happens twice in six months, so it certifies a moment rather than a container. Pasteurization is what substitutes for the test that does not exist.
The size of the gap has been estimated from outbreak data: 840 times more illnesses and 45 times more hospitalizations per unit consumed, with 3.2% and 1.6% of consumers accounting for 96% of dairy-linked illness. Shedding is intermittent and contamination enters during milking, which is why herd health alone cannot close it.
The price of acting on this is small. Heat treatment takes half the vitamin C, which milk hardly contains, and leaves protein, calcium and the fat-soluble vitamins alone. Boiling at home does the same job. What is lost is the word raw.
The same habit of reading a number as something it is not shows up with lead: the CDC reference value of 3.5 ug/dL is a percentile, not a safety threshold. See heavy metals in spices.
