No. 026 October 13, 2026 10 min read

The Innermost Contour

Every account of where the European eel breeds rests on a 1922 map of nested larval-size contours. When two Danish biologists went back to Schmidt's unpublished station records, the northern edge held and the eastern edge turned out never to have been in the data — and 104 years on, nobody has collected an egg.

The European eel spawns in the Sargasso Sea. I have known that for as long as I have known anything about eels, the way I know Everest is the tallest mountain — a fact with no visible seams. A German research vessel came into Bermuda on 12 April this year after four weeks of towing plankton nets through the place where it happens, the sixth such cruise since 2011. Nobody aboard watched it happen, because nobody ever has.

I went looking for what is actually underneath the sentence. The most load-bearing document is a map from 1922 whose eastern edge was never in the data.

wild eggs ever collected

Atlantic eels, either species

1 of 23

tag pop-ups inside the presumed spawning area

Azores tagging study, 2022

Japanese eel eggs

from a single cruise, May 2009

>90%

fall in recruitment

since the 1970s

The map

Johannes Schmidt, a Danish oceanographer with Carlsberg money behind him, published The Breeding Places of the Eel in 1922 after two decades of hauling nets across the Atlantic. The method was elegant and entirely indirect: catch leptocephali — the transparent willow-leaf larvae nobody had initially recognised as eels at all — measure them, map where the small ones are. Small larvae mean recent hatching; recent hatching means nearby spawning. Draw contours through the sizes, a 10 mm curve inside a 45 mm curve, and you get nested rings closing on a centre in the western Sargasso. That figure, redrawn one way or another, is in every textbook that mentions eels.

In 1985 two Danish biologists, Boëtius and Harding, went back through it with something earlier critics had not had: Schmidt’s own unpublished station records. Their re-examination, in the journal Dana, is neither a debunking nor an endorsement, which is what makes it worth reading.

The north-south boundary holds. Using the combined 1920–21 material they put a 95 per cent northern limit at about 29°30’N — essentially where Schmidt drew the northern edge of his 10 mm curve. The one boundary that can be tested statistically survives the audit at the place he put it.

The east-west boundary does not exist. Their sentence is flat: there is no evidence in Schmidt’s data for an east-west limit.

And the nesting is partly an artefact of the method. A larger-size curve must enclose a smaller one by definition; what the picture adds beyond that is the position of the innermost ring, and that depends on where tiny larvae happened to be caught, which depends on where the ships went. About 37 per cent of all the research effort fell inside the 10 mm curve, and effort outside the central area is described as extremely sparse. The cruises could establish that tiny larvae turn up when you look hard near 25°N, 60°W in April and May. They could not establish absence anywhere else.

What his re-examiners concluded, in the end, is that the spawning grounds lie in the Sargasso area he described, bounded to the north, with an eastern and western extent that cannot be determined from his data. He was right about the thing he is famous for. The map is simply more confident than the material behind it.

Smallest Atlantic eel material collected, mm

Schmidt's smallest, from his station records

5 mm

Modern cruises, 70°W to 50°W

<12 mm, recently hatched

ICES database, ridge box, 1913–1970

23 mm, smallest of 48

'Thor' cruises, Mediterranean, 1908–10

60 mm, smallest of 45

A wild-collected egg

none exists

The inference is strong, and it is an inference

Three independent lines run to the same place, and they are good lines.

Larval sizes, redone with modern gear: recently hatched larvae under 12 mm have been caught across roughly 2,000 km of ocean, from 70°W east to 50°W, in work published between 1982 and 2019. That is a real constraint, and also a 2,000 km one.

Genetics. Als and colleagues genotyped 21 microsatellite markers in 271 European eel larvae taken in the spawning area, 117 American eel larvae and 1,010 glass eels from European coasts, and found panmixia — one randomly mating population across the species’ whole range, once sibling larvae were accounted for. That is as hard a behavioural constraint as genetics can give, and it is silent on location. A panmictic population is consistent with one small site, with three million square kilometres, or with several sites whose offspring mix before anyone samples them.

Tagging. A decade-long effort released 707 adult eels from sites across Europe; 206 tags returned data, more than 80 routes were reconstructed, swim speeds ran 3 to 47 km a day, and every eel moved deeper by day and shallower by night. The conclusion was a mixed migratory strategy: fast eels make the coming spring’s spawning season, slow ones wait for the following year’s. I am working from that paper’s abstract and the coverage of it, the journal’s own page having refused to open for me, and the two do not agree on whether the headline figure is 5,000 or 8,000 km of mapped route.

“First direct evidence”

In 2022 a team tagged 26 female silver eels caught in rivers on the Azorean islands of São Miguel and Flores and released them into the Atlantic. Twenty-three tags reported, detaching between 40 and 366 days later. Bearings ran 208° to 273° — southwest, every one of them. Straight-line distances were 272 to 2,275 km, averaging 1,179. Five pop-up positions fell inside Sargasso Sea boundaries; one fell inside the presumed spawning area bounded by 31°N, 50°W and 24°N, 70°W.

The title says first direct evidence of adult European eels migrating to their breeding place, and the press ran with a century-old mystery solved. The sentence I keep coming back to is in the paper and in almost none of the coverage: at the speeds measured — no eel exceeded about 12 km a day — none of these animals could have reached the spawning area before the recognised spawning period ended. Extrapolating the 2019 mean bearing and speed to the 14 February peak puts arrival 466 days after release.

That is not a refutation of anything. It is positive support for the slow-migration, spawn-the-year-after picture, and it fits the mixed-strategy result. But “tracked to their breeding grounds” and “one tag of 23 surfaced inside the box, on an extrapolation landing a year and a half later” are the same finding stated at two very different confidence levels, and only one of them travelled.

The paper is scruffy in small ways — the Results give a mean speed of 6.8 km a day and the Discussion 6.5, and a published correction fixes metres per day to kilometres per day — but candid in the ways that matter. Speeds assume a constant rate and bearing along a rhumb line. Some surfacing positions were back-calculated from drift. The eels came from two islands in two years, sex was inferred from body length, and Azorean eels were assumed to take the mainland route. Nothing in the deflation above came from outside the paper.

Why the Pacific version got answered

The Japanese eel was the same problem and is no longer. In May 2009 a cruise collected 31 eggs, in five tows at three stations on 22 and 23 May, on the southern West Mariana Ridge just south of a salinity front. They were spotted by morphology, confirmed by real-time PCR aboard within hours, sequenced afterwards. Back-calculation put spawning on the night of 20–21 May, three and a half days before the new moon — consistent with a 2005–2009 pattern of spawning two to four days ahead of each new moon.

What made that findable was geometry and a clock. The site sits where a seamount chain crosses a salinity front: a line you can steam along, which moves predictably — in May 2009 it crossed the chain near 13°N. The moon reduces when to three days a month. Even with both, it took a 90-station cruise in April 2009 that found nothing, then 86 stations in May with the first eggs on the sixth day, after hundreds of stations in earlier years.

The Sargasso offers neither coordinate. On the 2020 paper’s figures the presumed area is some three million square kilometres of open water whose surface currents average under 5 cm/s, against 10–20 cm/s where the Japanese eel spawns: no front to follow, no ridge crossing at a known latitude, no established lunar window. The Atlantic question is harder not because the animal is more secretive but because that ocean has fewer edges.

The argument that is still live

In 2020 Chang, Miller, Tsukamoto and colleagues proposed that Atlantic eels may also spawn along the Mid-Atlantic Ridge. Their case: tagged eels from many release points converge toward the Azores, which is not on the shortest line to the Sargasso; the Japanese precedent is a seamount chain; males and females leave Europe months apart, so something must serve as a meeting place; and otolith manganese, a hydrothermal signature, looked enriched in European glass eels. Passive particle tracking from 15–29°N, 43–48°W delivered 0.71 per cent of northern-release larvae to 25°W, against 0.14 per cent from Sargasso releases.

A Matters Arising in 2022 took it apart point by point: the ridge’s peaks sit well below the 800–1,000 m the eels actually swim; ridge and Sargasso releases drift similarly because both sit in the same slow westward flow; and the manganese argument reverses its own source, which reported higher Mn/Sr in Sargasso leptocephali than in glass eels, not lower. They also point at the ICES Eggs and Larvae database, where the ridge box holds 48 A. anguilla leptocephali from 10 stations between 1913 and 1970, mean length 41 mm, range 23 to 45 — last year’s cohort drifting through, nothing newly hatched. Though they concede the database records no null stations, so the effort behind those 48 is unknown.

The reply, published the same month, gives up the speculative parts and holds that the gaps are sampling gaps. The detail I find most telling: the 2020 authors say in their own paper that it is a fact that eel spawning occurs in the Sargasso Sea, and claim only that it may not be the sole location. Both sides of the live dispute agree exactly on the inventory of what is missing — no eggs, no spawning adults, no observation.

What I take from it

The strength of an inference and the number of times a thing has been seen are independent quantities, and popular accounts collapse them because the inference is usually the part you wanted anyway. I do think the eels spawn in the Sargasso. The northern limit survived a hostile re-audit at the latitude Schmidt drew it, the genetics permit nothing but a single breeding population, and every tag points southwest.

But the sentence does work nobody has checked at the resolution it implies. The eastern edge of the famous map was never in the material. The recently hatched larvae are spread across 2,000 km of longitude. The nested rings partly describe where the Dana’s nets went. And the one species where this question closed, closed because that ocean happened to supply a ridge, a front and a moon.

The cruise that came home in April is the sixth on a three-year cycle, which puts the next window in 2029 — a hundred and seven years after the map.