To find out how much DNA signal bathwater actually holds and how quickly it accumulates, Mitsuyo Machida and Kazuhiko Kibayashi of Tokyo Women's Medical University's Department of Forensic Medicine recruited 11 healthy Japanese volunteers, several of them family members sharing a household. Each volunteer bathed alone in freshly drawn water, following the deep, shoulder-high soak customary in Japanese bathing, with no restrictions on washing routine, tub cleaning, or water temperature. Two liters of bathwater were collected before immersion and again after 1, 2, 5, and 10 minutes (Machida & Kibayashi, 2026).

DNA recovery ran through a glass microfiber filter with a 0.7-micron pore size and a positively charged surface built to pull DNA out of solution. Each 2-liter sample took eight passes through a 250-milliliter funnel. The filters were cut down, digested overnight, and carried through standard extraction and STR chemistry: AmpFlSTR Identifiler Plus, 15 autosomal loci, a 175 RFU analytical threshold. Promega's DNA IQ Spin Baskets separated the digested filter fragments from the extract before cleanup (Machida & Kibayashi, 2026). The study ran under Research Ethics Committee approval at Tokyo Women's Medical University, with written informed consent from every volunteer (Machida & Kibayashi, 2026).

DNA quantity jumped noticeably after the first minute, then climbed more gradually out to ten. The degradation index, a ratio comparing a short 41-base-pair amplicon to a longer 129-base-pair one, declined the longer a volunteer stayed in the water, meaning the DNA recovered later in the soak was more intact, not less. The proportion of loci matching a volunteer's own reference profile increased significantly after two minutes and leveled off by ten. Loci with no detectable match to the bather at all dropped to zero by the ten-minute mark (Machida & Kibayashi, 2026).

The tradeoff showed up in the mixture data. Loci carrying a combination of the bather's alleles and someone else's became more common after five minutes. Allele drop-ins, alleles attributable to neither the bather nor any known contributor, peaked at five minutes and fell off by ten, consistent with the idea that more template DNA settles down the kind of stochastic amplification effects that produce them. Mixed profiles turned up in 6, 7, 9, and 7 of the 11 samples at 1, 2, 5, and 10 minutes respectively. No pull-up artifacts were observed at any interval (Machida & Kibayashi, 2026).

One volunteer, No. 5, produced consistently higher DNA quantities than everyone else at every immersion time tested. Two others, Nos. 3 and 11, started low and stayed low through the first two minutes, and Nos. 2 and 11 were still low at ten. After five minutes, more than 80% of loci matched the reference profile in 7 of the 11 volunteers. By ten minutes, four volunteers, Nos. 1, 3, 5, and 9, had complete profiles (Machida & Kibayashi, 2026).

The authors tie this to prior research distinguishing “good” and “poor” shedders, and they flag a specific risk it creates: residual DNA from a high-shedding household member, like volunteer No. 5, could plausibly contaminate a sample from a low shedder like volunteer No. 6, even after the bathwater is replaced. They also point to an existing case in the literature they cite, where two people immersed together in a bathtub for roughly the same length of time produced a complete profile for one person and a barely detectable one for the other, the same asymmetry showing up outside a research setting (Machida & Kibayashi, 2026).

The department's own casework gave the method its second test. Between February 2008 and April 2023, the department conducted 1,325 forensic autopsies. Of those, 181 were classified as drowning. In 11 of those cases, DNA profiling was requested for two stated reasons: assessing potential co-presence, and identifying individuals who had been present at the scene (Machida & Kibayashi, 2026). Five of the eleven involved victims transported by emergency services before investigators could reach the original scene, meaning whatever the water held had already been sitting untouched by the time anyone arrived to sample it. The other six involved drowning at non-standard sites, a neighbor's pond among them, or victims who weren't immediately identified. A twelfth case, involving a burn injury rather than drowning, was added to test the method further (Machida & Kibayashi, 2026). The autopsy cases went through a separate ethics approval, with a waiver of informed consent from next of kin under national guidelines, and every sample was fully pseudonymized before the investigators saw it (Machida & Kibayashi, 2026).

Complete STR profiles came back from both bathtub-drowning cases in the series, including the case that opens this piece. One case in the series looked different from the rest: the individual had sustained burn injuries and made their own way to the bathtub, where immersion time was estimated at under a few minutes before death, which was ruled a suicide. Bathwater from that case still returned a partial profile, 23% of loci matching, from a DNA quantity the authors describe as sufficient for STR analysis despite the short exposure (Machida & Kibayashi, 2026).

Outside the bathtub, the method came up empty for identification purposes. None of the pond, swimming pool, river, or harbor samples in the case series produced loci that matched a reference profile, even in the four non-bathtub cases, one swimming pool and three river or harbor sites, where some human DNA was detected. The authors point to chlorine and disinfection byproducts in pool water, tannins and humic acid acting as PCR inhibitors in natural water, and UV-driven strand breakage as likely explanations. Bathwater, contained and shielded from those forces, held up. Open water did not (Machida & Kibayashi, 2026).

Machida, M., & Kibayashi, K. (2026). STR analysis of human DNA recovered from bathwater and other water samples for forensic identification. PLOS One, 21(3), e0345878. https://doi.org/10.1371/journal.pone.0345878