The deep sea · a prediction from 2014 · the fish that came after

A Line Drawn Ahead of the Fish

Deep-sea fish are not crushed, because water barely compresses: at 8,336 metres, where the deepest fish ever seen was filmed, seawater has shrunk by about 3.5 percent. What seems to stop them is chemistry. In 2014 a straight line through how salty fishes are inside, from 500 metres down to a trench at 7,000, predicted that they would stop near 8,200 metres. Below, choose which fish the line gets to see, and watch the limit move.

Why nothing gets crushed

Every ten metres of seawater adds roughly another atmosphere of pressure, so the bottom of the deepest trenches is under more than a thousand. That sounds as if it should flatten anything soft. It does not, because a fish's tissues are mostly water (the jelly-like layer under the skin of many deep-sea fishes measured 96.53 percent water across nine species), and water barely gives. Pressure crushes empty spaces, and it is air that fills them.

At 8,336 m the pressure is about 846 atmospheres. A litre of deep seawater (salinity 35, 2 °C) carried down from the surface ends up about 965 millilitres: it has shrunk by 3.46 percent. A litre of air, by Boyle's law, would shrink 846-fold; real air at that pressure is stiffer than Boyle's law allows and shrinks about 474-fold, to roughly two millilitres. So a gas-filled swim bladder is the part of a fish that pressure really fights. A thesis on the hadal snailfishes gives that as a reason those low-density, jelly-like tissues may help: they “would be adaptive under the high hydrostatic pressures of the deep sea, where the inflation of a swimbladder becomes increasingly difficult”.

That leaves the molecules. Pressure pushes on the shapes proteins fold into, and the fish's answer is a small molecule called trimethylamine N-oxide, TMAO. The 2014 paper this page follows calls it “a potent stabilizer capable of counteracting the destabilization of proteins by pressure”, and it “is known to increase with depth in bony fishes (teleosts) down to 4,900 m.” The deeper a fish lives, the more of it the fish carries in its muscle.

A line with somewhere to stop

Here is the problem that creates. TMAO is dissolved in the fish's body water, and everything dissolved counts toward the body's saltiness, its osmolality. A shallow marine fish is much less salty inside than the sea around it:

“typically 350 mOsmol/kg in shallow species compared with seawater’s 1,100 mOsmol/kg”

Yancey, Gerringer, Drazen, Rowden & Jamieson, PNAS 111 (2014), abstract

Pile on TMAO with depth and the inside climbs toward the outside. If it climbs in a straight line, it reaches seawater at some depth. Below that, a fish would have to be saltier than the sea, which, in the paper's words, “would require reversal of osmotic gradients and, thus, osmoregulatory systems.” So Paul Yancey and his colleagues drew the line. Their figure has twelve points for the muscle osmolality of fishes from 500 to 3,000 m (two of them, drawn with error bars, are averages of several fish), and then they went to the Kermadec Trench and caught five hadal snailfish at 7,000 m. Their abstract:

“We found their muscles to have a TMAO content of 386 ± 18 mmol/kg and osmolality of 991 ± 22 mOsmol/kg. These data fit previous extrapolations and, combined with new osmolalities from bathyal and abyssal fishes, predict isosmotic state at 8,200 m.”

Yancey et al. 2014, abstract

The points below are read off the paper's own figures (how, and how well, is under what this page does not show). Start with the twelve shallow points alone, without the trench. Then add the rest in order, or click any single point to leave it out, and watch where the line reaches seawater and how wide the 95% interval on that depth is.

A prediction that could have failed

Look at what the twelve shallow points alone can say. Their line reaches seawater at 8,386 m, but the 95% interval runs from 7,410 to 9,758 m: over two kilometres wide, because the points stop at 3,000 m and the line has to be projected almost three times further than the data reaches. (The figure's own line, 326 + 0.0916 × depth, reaches 1,100 at 8,450 m; our refit of the points as read off the figure has a slope of 0.0924 and lands at 8,386. A difference in the third decimal place of the slope moves the answer 64 metres. That sensitivity is the lesson in miniature.)

But a line makes a bet at every depth, not only where it ends. At 7,000 m the twelve-point line says one more fish should come in at 972 mOsm/kg, give or take 95. It could easily have come back at 800, and the whole idea would have been in trouble. The Kermadec snailfish came back at 991. With them added the crossing moves to 8,189 m and the interval tightens to 7,746 to 8,701 m. That is the “8,200 m” of the abstract.

Then more fish were published. Yancey's 2025 update replots snailfish osmolalities from data “modified from Linley et al. 2016”, and seven of its snailfish points are not in the 2014 figure: three from the Kermadec Trench (its legend labels them 2014) and four from the Mariana Trench (2016). They sit at 6,453 to 7,898 m, not at the 7,000 m of the 2014 sample, so none of them was in the line above. All seven fall inside the band the twelve shallow points had predicted for a single fish at their depths. All seven also sit above the old line, by 4 to 63 mOsm/kg, and broadly, the deeper they are, the less they overshoot. The line is bending. Watch what that does:

• Fit one straight line through everything so far, twenty points, and it reaches seawater at 7,995 m (7,831 to 8,169). The filmed fish at 8,336 m is outside that interval, deeper than a straight line allows.
• Fit the snailfish alone, eight points from 6,453 to 7,898 m, and the line is much flatter (0.0607 per metre, not 0.0979). It reaches seawater at 8,453 m (8,034 to 9,503), and the fish at 8,336 m sits inside.

Yancey's 2025 update says the estimate was “later refined to 8,300-8,400 m with more data (Linley et al. 2016)”. We could not read how Linley and colleagues fitted it, so this page does not reproduce their number; it shows instead why any single number depends on which fish you let a straight line see, and that near the bottom the relationship is probably not straight.

Where fish have actually been found

In the Izu-Ogasawara Trench the 2022 expedition sampled by trap and video from 4,534 to 9,773 m. Snailfish turned up between 6,824 and 8,336 m, and no deeper. The expedition's chief scientist, Alan Jamieson, said the find was less about the depth than about the prediction:

“The real take-home message for me, is not necessarily that they are living at 8,336m but rather we have enough information on this environment to have predicted that these trenches would be where the deepest fish would be”

University of Western Australia news release, 3 April 2023

(His sentence is about which trenches the deepest fish would be found in; this page tests only the depth.) Every record on the list is shallower than 8,400 m, and the 2014 paper's opening line still stands as a description of what has been found: “Fish appear to be absent from the ocean's greatest depths, the trenches from 8,400–11,000 m.”

The argument is not over

In March 2025 a study in Cell of how fishes adapted to the deep sea (Xu and colleagues) “reported considerably lower TMAO levels in 3 hadal snailfish and 1 cusk eel from the Mariana Trench”, in Yancey's summary of it. Its own abstract says: “Interestingly, our results question the previously assumed linear correlation between trimethylamine oxide (TMAO) content and depth.” Yancey answered in an update deposited on figshare. He argues that fish damaged as they are hauled up leak TMAO and take up sea salt, and that sodium separates them:

“undamaged Mariana Trench snailfish had Na+ contents under 250 mmol/kg, while all damaged fish had random values above 300. Thus, retrieval damage probably causes leakage of TMAO from, and NaCl from seawater into, muscle tissue.”

Yancey 2025, figshare 10.6084/m9.figshare.28692455, version 7

He does not claim to have settled it: “I do acknowledge that Xu et al.'s low-TMAO data, although consisting of only 4 data points, could be accurate, suggesting alterations may be needed to the TMAO piezolyte hypothesis”. He adds that the rise in the fishes' osmolality with depth, which is what this page plots, needs some dissolved substance to account for it whatever that substance turns out to be. He also notes that snailfish from the warmer Mariana Trench carry less TMAO than those from the colder Kermadec at the same depth, a “piezo-thermal” effect that “could alter our hypothesized depth limits for teleosts.” This page has not read the Cell paper's data and does not referee the dispute. What it can show is that the depth limit is a projection, that its width depends on how many deep fish have been measured, and that every central estimate here, from 7,995 to 8,453 m, lies within 350 metres of the deepest fish yet seen.

What this page does not show

The osmolalities are not the papers' tables. Neither paper's data table could be fetched, so every point was read off a published figure by locating each marker's centre in the image and mapping it through the axis ticks. The check on that is external: a straight line refitted to the twelve digitized 2014 points gives 325.2 + 0.0924 × depth against the printed 326 + 0.0916, the thirteen-point refit gives 320.8 + 0.0952 against the printed 320 + 0.0953, and in the 2025 figure an open square (the legend's style for “Various fish 1996-2007”) sits at 7,004 m and 989, where the 2014 snailfish average of 7,000 and 991 would plot; we take it to be that point, replotted, and use it to check the second figure's axes. Treat each point as good to roughly ten metres and five mOsm/kg. Eight markers between 6,400 and 7,900 m overlap in the 2025 figure and were placed by eye on an enlargement. The years beside the later snailfish are the figure legend's; whether they are the years the fish were caught or the years their data were first published, the figure does not say.

The Kermadec snailfish point is the mean of five fish, entered once, as the 2014 figure draws it. Fitting with a straight line and a 95% interval (Fieller's method, treating each point as independent and equally precise) is our choice. Yancey's 2025 update says the 8,200 m estimate came “with error bars”; we could not read the 2014 paper's body to see them, so every interval on this page is ours, computed from the digitized points. The seawater figure of 1,100 mOsm/kg is the paper's round number. Pressure uses the Saunders (1981) formula at latitude 30°; seawater compression uses the EOS-80 equation of state for water at salinity 35 and 2 °C held at constant temperature, and a real fish is not pure seawater. Nothing here tests whether TMAO is why fish stop: a limit set by anything else that grows harder with depth, food or temperature for instance, would draw the same records table.

Sources

1. P. H. Yancey, M. E. Gerringer, J. C. Drazen, A. A. Rowden, A. Jamieson, “Marine fish may be biochemically constrained from inhabiting the deepest ocean depths”, PNAS 111 (2014) 4461 to 4465, doi:10.1073/pnas.1322003111 (PMC3970477). Figure 3 is the source of the twelve shallow points.
2. P. H. Yancey, “Update on trimethylamine N-oxide (TMAO) as a piezolyte and cryoprotectant: its role in a depth limit for marine fishes and loss from hadal fish during capture”, figshare, version 7 (2025), doi:10.6084/m9.figshare.28692455.v7, CC BY 4.0. Figure 1 (data modified from Linley et al. 2016) is the source of the later snailfish, rattails and eelpouts.
3. T. D. Linley et al., “Fishes of the hadal zone including new species, in situ observations and depth records of Liparidae”, Deep-Sea Research I 114 (2016), doi:10.1016/j.dsr.2016.05.003 (abstract only was read).
4. A. J. Jamieson, P. J. Maroni, T. Bond, Y. Niyazi, J. Kolbusz, P. Arasu, H. Kitazato, “New maximum depth record for bony fish: Teleostei, Scorpaeniformes, Liparidae (8336 m, Izu-Ogasawara Trench)”, Deep-Sea Research I 199 (2023) 104132, doi:10.1016/j.dsr.2023.104132 (abstract); and the UWA news release, 3 April 2023.
5. M. E. Gerringer, On the Success of the Hadal Snailfishes (PhD thesis, University of Hawai‘i at Mānoa, 2017), p. 186 and ch. on gelatinous tissues, ScholarSpace; M. E. Gerringer, “On the success of the hadal snailfishes”, Integrative Organismal Biology 1 (2019), doi:10.1093/iob/obz004, table of hadal snailfish depth ranges.
6. H. Xu et al., “Evolution and genetic adaptation of fishes to the deep sea”, Cell 188 (2025) 1393 to 1408, doi:10.1016/j.cell.2025.01.002 (abstract only was read).
7. Seawater: N. P. Fofonoff and R. C. Millard, Algorithms for computation of fundamental properties of seawater, UNESCO Technical Papers in Marine Science 44 (1983), with coefficients as implemented in the seawater Python package 3.3.5; P. M. Saunders, “Practical conversion of pressure to depth”, J. Phys. Oceanogr. 11 (1981) 573 to 574. Cross-checks: TEOS-10 via gsw 3.6.23, and real-gas air via CoolProp 8.0.0.