Private gardens, arable fields, permanent grassland and lawns each took delivery of the same object in 2021: standardised white organic cotton briefs, to be buried eight centimetres down with the waistband left showing above the surface. Two further categories went in alongside them, permanent crops and a catch-all of other land uses, six in all.
Two months later the briefs from private gardens had lost, on average, around 58 percent of their fabric. The ones from lawns had lost around 40 percent. Arable fields sat at around 51 percent and permanent grassland at around 47 percent. In the team’s regression model the single strongest predictor of how much cloth disappeared was not any of the chemistry the team measured in the laboratory afterwards. It was what the land was being used for.
The work was published on 26 August 2026 in Plants, People, Planet by a team based at Agroscope, the Swiss federal agricultural research centre, and the University of Zurich, with the participants themselves credited in the author list. The regression model that ranked land use first accounts for 16.5 percent of the variation in the two-month figure, and the authors’ own reading is that their results suggest the test reflects soil fertility more strongly than soil health in the broader ecological sense.
Eleven hundred packs went out; 780 sites came back usable
The call for volunteers drew around 1,400 registrations within a few days. Participation was capped at roughly 1,000, and rather than taking the first thousand names the team selected for a geographic spread across the country, with at least half of those chosen being farmers.
Each participant received a pack containing two pairs of organic cotton briefs, twelve teabags, a bag for a soil sample, illustrated burial instructions in German, French and Italian, and a prepaid box for the return journey. The briefs were a single supermarket product, Coop Naturaline men’s white briefs in size S, ordered direct from the producer so that every site was burying the same thing. They are 100 percent organic cotton apart from the waistband and seams, which are synthetic.
The burial was prescribed down to the hole. Participants dug out roughly 60 by 30 by 30 centimetres, took a soil sample, stood two pairs of briefs upright side by side, then refilled until the leg openings were covered at eight centimetres, with three teabags of each type in front of each pair. The soil went back, was trampled down, and the spot was marked with a yellow wooden peg. The first pair came up after 30 days and the second after 60, air dried and posted to Agroscope.
The numbers that survived that round trip are smaller than the ones the project is usually described by. The paper’s headline totals are 1,000 citizen scientists, 2,000 pairs of underpants and 12,000 teabags. Its methods section reports 1,100 packs posted, and a laboratory that received 883 soil samples, 1,752 pairs of underpants, 7,088 intact teabags and 800 returned questionnaires. After quality filtering, 780 sites remained, and that is the figure the results rest on.
Decomposition was measured both by weight and by area, and the area figure is the one the analysis rests on. Each pair was cut open, spread flat on a blue background and photographed from a fixed height; the blue was filtered out in software and what remained was counted in pixels against a new pair. Weight loss tracked area loss closely, with an R squared of 0.73, but soil particles cling to buried fabric, so the photograph was treated as the more reliable measure.
The order the main land uses came out in
Averaged across every site, the briefs had lost 10.1 percent of their area after one month and 50.45 percent after two. The gap between land-use types opened up inside that average, and the effect was statistically strong.
After two months the ranking ran:
Private gardens, around 58 percent decomposed
Arable fields, around 51 percent
Permanent grassland, around 47 percent
Lawns, around 40 percent
Private gardens and lawns each differed significantly from arable fields and from permanent grassland. Lawns also came in significantly below the catch-all category of other land uses. The remaining comparisons between land-use types, permanent crops among them, were not statistically distinguishable from one another.
Why the tidiest ground rotted the least
Lawns are the surprise in that list, because a lawn is not obviously abused ground. It keeps continuous plant cover and is usually left physically undisturbed, which the paper says provides relatively stable conditions for soil organisms. Set against that, and depending on the type of lawn, frequent mowing, low plant diversity and regular herbicide use can all work against soil biological communities. The paper does not say which of those pressures wins. It reports that the lawns rotted the least.
Private gardens carried the highest soil organic carbon in the study at around 3.4 percent, level with the catch-all other category, along with the highest nutrient concentrations, and the authors note they may also be worked and dug regularly. Arable fields are the useful comparison: their organic carbon was as low as the lawns, at around 2.1 percent, and they still rotted more cloth. The authors suggest that difference may be related to greater physical disturbance, which can stimulate microbial activity and decomposition, a process that also costs the soil carbon over time.
Decomposition also fell as sites climbed in altitude, which the authors think is likely due partly to lower temperatures and partly to intensively managed Swiss farmland sitting mostly at lower elevations. Among fertiliser types, sites that had received compost showed the highest decomposition.
The gardens at the top of the table carry an awkward second finding. They averaged 13.74 milligrams of available phosphorus per kilogram of soil, against Swiss fertilisation guidelines that put the optimum for extractable phosphorus, measured in carbon-dioxide-saturated water, between 1.6 and 4.3 milligrams per kilogram depending on soil texture. Some garden samples came in above 60. The authors read that as many gardeners fertilising well past anything their plants can use, with the excess liable to be lost to erosion and leaching into groundwater and surface water, and they conclude that in most cases no additional phosphorus fertiliser is required to maintain plant growth. The paper gives that garden average twice, as 13.74 in its results and 13.4 in its discussion; this article uses the results figure, which sits beside its own test statistic.
Soil organic carbon was not a significant predictor
Soil organic carbon is often treated as a key indicator of a healthy, biologically active soil, especially in managed mineral soils. It did not come out as a significant predictor in the multiple regression model here, even though decomposition rose with soil organic carbon when the two were compared directly. The paper’s explanation is that carbon moves with most of the nutrient measures, and that the model suggests nutrient availability may be more important than the carbon itself.
That is the reasoning behind the sentence the authors put in their own conclusion: their results suggest the underpants index reflects soil fertility, and nutrient availability in particular, more strongly than soil health in the broader ecological sense. They still propose it as a tool for a first impression of soil health, and the paper’s own title calls it a soil health assessment, so this is a lean rather than a retraction.
It matters because it changes what the league table means depending on where the pants are buried. Higher decomposition, and so potentially greater fertility, may be considered beneficial in a garden or a field. In a meadow or a forest, high nutrient availability is often an unwelcome sign, associated with reduced plant diversity and the disappearance of rarer species. To identify what is actually driving the cloth loss, the authors call for controlled experiments that manipulate soil fertility, compost application and organic carbon directly.
The dataset carries two further limits. It is one country and one season, with climate data drawn from May to July 2021, and the volunteers put themselves forward before being picked for geographic spread, which is not a random sample of Swiss ground. The data are openly deposited.
The tea bags said roughly the same thing
The twelve teabags in each pack were there as a check against an established method. Rooibos and green tea in synthetic mesh are the standardised litter used by the Tea Bag Index, and one of the project’s two stated hypotheses was that the tea and the cloth would agree.
The two agreed. The paper calls the correspondence strong; the correlation behind that word is a Spearman’s rho of 0.32, significant but loose. Land use came out as a significant driver of the tea figures as well, with private gardens again at the top, and that model accounted for 21.5 percent of the variation.
The clearest split between the two measures was on nutrients. In the regression models, fertiliser type and available phosphorus predicted the underpants result but not the tea result, and the authors suggest that may be because tea leaves arrive with their own potassium, sodium and micronutrients while cotton is largely low-nutrient cellulose, so the cloth depends more on what the surrounding soil can supply. That is their argument that the underpants may be the more sensitive reading of fertility, though they also report that the two indices showed very similar relationships with soil properties overall and describe both as equally meaningful about soil conditions. The tea bag protocol as published calls for 90 days in the ground, and this project ran to 60.
The team also went back over its own photographs and re-scored them by eye, on a crude scale in steps of ten percent, and found it similarly accurate to the pixel count. A citizen science project in Hungary also found a trend towards higher decomposition in private gardens.
A pair of white briefs on a blue sheet, half the leg gone, is the instrument the project is actually proposing. It needs no balance, no laboratory and no training to read.