Every year, in a handful of state laboratories, a technician holds a piece of fruit under cold running water for about 15 to 20 seconds. No soap. No produce wash. Just the rinse an ordinary person gives an apple at the sink.
Then the fruit goes into a blender, and the sample runs through a screen looking for 597 pesticides, metabolites and breakdown products.
That is the Pesticide Data Program, run by the Agricultural Marketing Service at the U.S. Department of Agriculture. In December 2025 it published results for 9,872 samples collected during calendar year 2024. The data is free, public and almost never reported in full.
Here is what those numbers say, including the one usually dropped on the way to the headline. You will not find a shopping list here. Our article on pesticide products handles labels and household use, and our hazard and dose article handles why finding a chemical is not the same as being harmed by it.
What the 2024 testing round found
Four numbers carry the story.
42.3 percent of samples had no detectable residue at all. Not a trace, at detection limits far below any legal limit.
More than 99 percent of sampled commodities carried residues below EPA tolerances. A tolerance is the maximum residue the Environmental Protection Agency allows to remain on a food.
76 samples, or 0.77 percent, carried a residue above a tolerance. Twelve domestic, 63 imported, one unknown.
361 samples, or 3.7 percent, carried a residue for which no tolerance has been established. Hold onto that one. It gets its own section below.
Of the total, 9,165 samples were fresh and processed fruits and vegetables, with 531 almond and 176 salmon samples alongside. Nine states collected, seven state laboratories tested, and domestic food made up 60.1 percent of samples against 38.9 percent imported.
Two details about method matter. Samples come from terminal markets, distribution centers and retail sites, close to where people buy food rather than at the farm gate. And the laboratory prepares each one the way a household would. Oranges get peeled. Pineapples get cored. Avocados lose the pit and skin. Apples stay unpeeled, because people do not peel them. So these figures describe food as it is eaten, not food as it was sprayed.
One caution before anyone treats the 76 over-tolerance samples as a seizure list. USDA says the program “is not designed for enforcement of EPA pesticide residue tolerances.” It reports findings to the Food and Drug Administration, which enforces. The program began in 1991, and Congress wrote it into law in 1996.
A tolerance is a legal ceiling, not the line where harm starts
EPA describes a tolerance as the maximum amount of a pesticide allowed to remain in or on a food, which most other countries call a maximum residue limit. It covers imports as well as domestic food.
Getting one is not a formality. A pesticide company must submit toxicity studies on the active ingredient and its breakdown products, field trials run at the highest rate the label allows, data on what survives into processed foods, an accounting of other exposure routes such as home use, and a laboratory method precise enough to check results later.
The standard those submissions clear sits in the Federal Food, Drug, and Cosmetic Act. Congress defined a safe residue as one where “there is a reasonable certainty that no harm will result from aggregate exposure to the pesticide chemical residue.” Aggregate does real work in that sentence. EPA adds up food, drinking water and residential exposure rather than looking at the apple alone.
Three more features come from the Food Quality Protection Act of 1996.
A tenfold factor for children. For threshold effects, the law directs EPA to apply an additional tenfold margin of safety for infants and children, and allows a different margin only when reliable data show it will be safe for them. That sits on top of the uncertainty factors already inside the toxicology.
Cumulative effects, but only within a group. EPA must consider residues together with other substances sharing a common mechanism of toxicity. Pesticides working through different mechanisms are not added together. Remember that two sections from now.
A clean sweep of the old limits. EPA reassessed the 9,721 tolerances in force when the law passed and revoked or modified close to 4,000.
Now the part that puts the 99 percent in context. A tolerance is set at the level residues are not expected to exceed when a product is used as the label directs. It is calibrated to legal agricultural practice and bounded by a safety finding, not set at a biological threshold where effects begin.
So a residue above a tolerance is first of all evidence about how a pesticide was used, not a measurement of harm. The gap between a tolerance and a dose that produced any effect in a laboratory animal is usually enormous, and our hazard and dose article walks through reference doses and the uncertainty factors inside them.
FDA acts on violations. It can keep domestic product off the market, refuse imports entry, and place a firm on an import alert, which lets FDA detain its shipments without examining each one.
The 3.7 percent with no tolerance to be below
This is the figure that gets cut for length, and it is the most interesting one in the report. Of the 361 samples carrying a residue with no tolerance set on that commodity, 230 were imported, 118 domestic and 13 of unknown origin.
That is a third category, distinct from both the 76 and the comfortable 99 percent. The reassuring number is a comparison: residue measured, limit consulted, residue found lower. For these 361 there is no limit to consult, so the residue is neither below nor above tolerance.
Legally that is unambiguous. The Federal Food, Drug, and Cosmetic Act deems a pesticide residue in or on food unsafe unless a tolerance or exemption covers it, so a residue with no tolerance makes the food adulterated as a matter of law, whatever the quantity.
Toxicologically it is completely ambiguous, and that is the honest answer. Deemed unsafe is a legal status, not a laboratory result. No tolerance usually means nobody applied for one in the United States for that pesticide on that crop, so nobody ran the assessment. USDA’s report notes the program screens for some pesticides that lack U.S. tolerances but are used in countries exporting food here, which fits the import skew inside the 361. The report does not give a reason for each finding, and this article will not supply reasons it cannot source.
So what can a reader take from 3.7 percent? Not that those residues were dangerous, because an unassessed pairing is not one that was assessed and failed. Not that they were safe either, because nothing was measured against anything.
What you can conclude is narrower. Roughly one sample in 27 fell outside the machinery that produces the reassuring figure. An account that gives you the 99 percent and stops has handed you a real number with an incomplete denominator, and an account that calls those 361 samples poisoned food has gone further than the data allows.
The Dirty Dozen, and what peer review has said about it
The Environmental Working Group, a nonprofit registered under section 501(c)(3), publishes an annual Shopper’s Guide to Pesticides in Produce. Its famous outputs are the twelve highest-scoring crops, the Dirty Dozen, and the fifteen lowest, the Clean Fifteen. The 2026 edition draws on more than 54,000 USDA samples covering 47 crops. This article does not reproduce either list, because telling you which produce to avoid is not its job.
The scoring matters more than the names. EWG ranks each crop on four measures and normalizes them into a combined score out of 400. One is the share of samples with any detection, one the average number of different pesticides per sample, one the average total concentration. The fourth is a toxicity term, added in a 2025 methodology update: EWG sums each pesticide’s concentration with its metabolites, divides by a no-observed-adverse-effect level from animal studies, and adds the ratios. Most of those animal figures come from EPA review documents that the Center for Biological Diversity obtained through a Freedom of Information Act request.
Two peer-reviewed papers questioned the earlier version of the method.
Carl winter and Josh Katz, journal of toxicology, 2011. Winter, a food toxicologist at the University of California, Davis, modeled dietary exposure to the ten most frequently detected pesticides on each of the twelve listed commodities, using Pesticide Data Program results from 2004 to 2008 and federal food consumption surveys. Of the 120 pairings the design produced, every estimate came in below the chronic reference dose, the daily amount EPA considers unlikely to cause appreciable harm across a lifetime. Only one exceeded 1 percent of it, and three quarters fell below 0.01 percent.
Their charge was specific. EWG then used six indicators, and only one looked at how much residue was present. None multiplied concentration by how much of the food people eat, which is what exposure means. Four of the six turned on the presence of multiple different residues, which the authors argued pushed crops carrying many tiny detections up the ranking regardless of dose.
Richard Reiss and colleagues, Food and Chemical Toxicology, 2012. This team put risk and benefit on the same page. If half the U.S. population added one serving of fruit and one of vegetables daily, they estimated roughly 20,000 cancer cases a year might be prevented, against an upper-bound estimate of up to 10 cancer cases a year theoretically attributable to the extra residues. They named their own weak points: possible residual confounding in the nutrition studies, and reliance on rodent bioassays for cancer potency.
Read both with affiliations in view. Winter worked at a public university. Reiss worked at Exponent, a consulting firm. That makes a calculation neither wrong nor disinterested, and a reader is entitled to weigh it.
Now the fair part, because the subject has moved since 2011. The central charge, that the list ranks by detection rather than dose, lands less squarely in 2026, because EWG’s fourth metric does compare a measured concentration against an animal no-effect level.
What still separates them:
A no-observed-adverse-effect level is not a reference dose. EPA’s reference dose is that animal figure divided by uncertainty factors, commonly 100 or 1,000, before any children’s factor. Comparing concentrations against undivided animal numbers is different arithmetic from comparing modeled exposures against reference doses.
Three of the four metrics still count detections rather than doses, and none multiplies by consumption, so a pesticide on a crop almost nobody eats and the same pesticide on a staple score alike.
Some underlying data is old. EWG’s own methodology page notes spinach scores rest on 2015 and 2016 samples, because USDA retests each crop on a multi-year cycle.
The mixtures question is genuinely unresolved, and here EWG has a point. Regulators mostly assess pesticides one at a time, and the statute requires cumulative assessment only for substances sharing a common mechanism. Residues of several pesticides that do not share one are not summed. Whether that matters at the concentrations found is insufficiently studied, which differs from both reassurance and alarm.
EWG itself says the guide should not discourage people from buying and eating produce, and states that a diet high in fruits and vegetables, organic or conventional, matters for health.
This article declares no winner. The exposure modeling in those two papers is established-method work with disclosed uncertainty. Low-dose mixtures over decades are an open question rather than a finding either way, and mixing those two evidence tiers is how both sides overclaim.
Five questions for any residue claim
Our hazard and dose article has a general checklist for chemical headlines. These five are specific to food.
- Detected at what level? Found in is a yes or no, not a measurement.
- Compared against which benchmark: a tolerance, a reference dose, or an undivided animal no-effect level? Those three differ by orders of magnitude.
- How much of the food does the comparison assume a person eats? A concentration without a consumption figure is not an exposure.
- Is the residue on the surface or inside the tissue? That decides whether washing changes anything.
- Where was the sample taken, and was it prepared the way you would eat it?
Washing, peeling, and what each one removes
Start with a fact that surprises people. The federal numbers are already post-rinse numbers, because the laboratory washed the produce first, so whatever washing achieves is baked into the 42.3 percent.
Tianxi Yang and colleagues at the University of Massachusetts Amherst, in the Journal of Agricultural and Food Chemistry in 2017, treated apples with two pesticides and left them 24 hours. A baking soda solution cleared surface residues better than tap water or bleach, taking 12 minutes for one compound and 15 for the other. The important finding was what no wash could reach. Within those 24 hours, 20 percent of the applied thiabendazole and 4.4 percent of the applied phosmet had moved into the fruit. Thiabendazole, which plants absorb and distribute internally, penetrated four times deeper into the peel than phosmet, which stays outside. Peeling removed the rest, and takes the peel’s nutrients with it. A 2025 follow-up from the same laboratory, in the journal Foods, found a corn starch soak followed by baking soda took off roughly 94 percent of surface thiabendazole.
Both studies applied pesticide deliberately at known doses, which is not the same as residues on fruit that left a field weeks earlier, so treat those percentages as laboratory behavior rather than a forecast for your groceries.
The honest summary is short. Rinsing under running water, with no soap, reduces what sits on the surface. It does nothing about a systemic pesticide, one the plant took up and moved through its tissues. That is chemistry rather than a loophole.
The number that is larger than everything above
In Morbidity and Mortality Weekly Report in 2022, Seung Hee Lee and colleagues at the Centers for Disease Control and Prevention analyzed 2019 survey data from 49 states and the District of Columbia. Adults are advised to eat 1.5 to 2 cup-equivalents of fruit and 2 to 3 cup-equivalents of vegetables a day. Nationally, 12.3 percent met the fruit recommendation, from 8.4 percent in West Virginia to 16.1 percent in Connecticut. For vegetables, 10.0 percent met it, from 5.6 percent in Kentucky to 16.0 percent in Vermont.
Round that off. Nine adults in ten are not eating enough vegetables.
Now set it beside the residue findings. In the modeling above, dietary exposures to the most commonly detected pesticides typically sat three to five orders of magnitude below reference doses that already carry hundredfold or thousandfold safety margins. CDC’s framing in that report is that a healthy diet helps prevent obesity, type 2 diabetes, cardiovascular disease and some cancers.
The documented harm from eating too little produce is larger, better measured and better established than anything in the residue data. That is not an opinion about pesticides. It is what the two bodies of evidence look like side by side.
One distributional fact belongs here, reported rather than preached. In that CDC analysis, vegetable intake was lowest among adults living below or close to the poverty level, at 6.8 percent. The people furthest from the recommendation are the ones for whom a price premium is hardest to absorb.
When a residue question is a medical question
Dietary residues at these levels are not a poisoning scenario. Real acute exposures look different: swallowing a concentrate, a spill on skin, a product sprayed indoors in a closed room.
For any of those, call Poison Help at 1-800-222-1222, staffed 24 hours a day, and you do not need an emergency to call. Call 911 instead if the person is unconscious, having trouble breathing or having convulsions. Keep the container, because the label tells a responder what the product is. Our article on pesticide products has the full sequence.
If your concern is an occupational exposure, something you met while pregnant, or a child who got into a product, that is a conversation with your own clinician or an occupational medicine service. No general-purpose residue panel exists that will settle it.
What to take from 9,872 samples
Read the source rather than the summary. USDA publishes these results annually, and four numbers describe any year: the share with no detectable residue, the share below tolerance, the count above tolerance, and the count with no tolerance established. A report handing you only the second has handed you half a picture.
Then keep three numbers straight, because almost every confusing residue story mixes up at least two. A tolerance is a regulatory ceiling tied to legal agricultural use. A reference dose is a health benchmark with large safety factors already divided in. A no-observed-adverse-effect level is the raw animal figure those factors get applied to. Comparing a residue against the wrong one of the three produces false alarm or false comfort.
Whatever you decide about any list, decide it without eating less produce. On the evidence above, that trade goes the wrong way by a wide margin.
Keep exploring: USDA organic.
Sources
- USDA Agricultural Marketing Service. “Pesticide Data Program Annual Summary, Calendar Year 2024 (December 2025).” ams.usda.gov
- USDA Agricultural Marketing Service. “Pesticide Data Program.” ams.usda.gov
- U.S. Environmental Protection Agency. “Pesticide Tolerances.” epa.gov
- U.S. Environmental Protection Agency. “Setting Tolerances for Pesticide Residues in Foods.” epa.gov
- U.S. Environmental Protection Agency. “Summary of the Food Quality Protection Act.” epa.gov
- U.S. Food and Drug Administration. “Pesticides (Chemical Contaminants and Pesticides).” fda.gov
- Legal Information Institute, Cornell Law School. “21 U.S. Code 346a, Tolerances and exemptions for pesticide chemical residues.” law.cornell.edu
- Environmental Working Group. “EWG's Shopper's Guide to Pesticides in Produce, 2026 edition summary.” ewg.org
- Environmental Working Group. “Shopper's Guide to Pesticides in Produce: Methodology.” ewg.org
- Winter CK, Katz JM.. “Dietary exposure to pesticide residues from commodities alleged to contain the highest contamination levels. Journal of Toxicology, 2011. PubMed. DOI: 10.1155/2011/589674.” doi.org
- Reiss R, Johnston J, Tucker K, DeSesso JM, Keen CL.. “Estimation of cancer risks and benefits associated with a potential increased consumption of fruits and vegetables. Food and Chemical Toxicology, 2012. PubMed. DOI: 10.1016/j.fct.2012.08.055.” doi.org
- Lee SH, Moore LV, Park S, Harris DM, Blanck HM.. “Adults Meeting Fruit and Vegetable Intake Recommendations, United States, 2019. MMWR Morbidity and Mortality Weekly Report, 2022. PubMed. DOI: 10.15585/mmwr.mm7101a1.” doi.org
- Yang T, Doherty J, Zhao B, Kinchla AJ, Clark JM, He L.. “Effectiveness of Commercial and Homemade Washing Agents in Removing Pesticide Residues on and in Apples. Journal of Agricultural and Food Chemistry, 2017. PubMed. DOI: 10.1021/acs.jafc.7b03118.” doi.org
- Du X, Ho L, Li S, Doherty J, Lee J, Clark JM, He L.. “Efficacy of Household and Commercial Washing Agents in Removing the Pesticide Thiabendazole Residues from Fruits. Foods, 2025. PubMed. DOI: 10.3390/foods14020318.” doi.org