Unpacking Gradient Factors

Open your dive computer's settings and you'll find two numbers separated by a slash. Maybe 40/85, maybe 30/70. Most divers pick them once, copying a buddy or an instructor, then never touch them again. Yet those two numbers, your gradient factors, quietly shape every ascent you make. You set them, and from that choice the model works out how deep you first pause on the way up and how much dissolved gas you still carry when you reach the surface. Divers have argued about how to set them since the late 1990s, and for most of that time nobody could say which values were actually safer. Then a US Navy experiment in 2011 produced the clearest answer yet, and it caught a lot of people off guard.

What follows is the story of where those numbers came from and what the evidence really says about them. Every claim links to its source. None of it is dive training, and the physics stays in plain English.

A ceiling you cannot see

Start with what is happening inside you. Breathe compressed gas at depth and your body slowly absorbs it, mostly the nitrogen, which it cannot use but dissolves into your blood and tissues anyway. The deeper you go and the longer you stay, the more you take on. Coming back up runs the process in reverse, because all that gas now has to leave your body again, a slow unloading divers call "off-gassing". Ascend slowly and the gas you took on during the dive filters out through your lungs, to be breathed out as it should. Ascend too fast and it forms bubbles inside your body but outside your lungs, where you cannot breathe it out. A few small bubbles form on almost every dive and usually cause nothing you can feel, which is part of why they are dangerous: a diver can be quietly bubbling and feel completely fine. Let too many form, or let them lodge in the wrong place, and they can bring on joint pain, numbness, paralysis, even death. That is decompression sickness, or DCS, and heading it off is the whole job of a decompression model.

The first model that actually worked was John Scott Haldane's, in 1908. Running goats through pressure chambers, he found something useful: their bodies tolerate a fair amount of supersaturation. A goat can carry more dissolved gas than the surrounding pressure should strictly allow, up to a point, before bubbles appear. From these goats he drew two conclusions and applied them to humans. The first was a rough rule of thumb: you could halve your absolute pressure in a single step, say from about 10 metres, where the pressure is twice what it is at the surface, straight up to the top, without bubbling. The second was a deliberately crude sketch of the body, five imaginary "tissue compartments" that soak up and release gas at different speeds. Their half-times, the minutes each one needs to fill halfway, ran from a fast 5 to a slow 75.^{[1 ]} The compartments were never meant to be real organs. They were a way to hold on to one fact that still drives everything: your blood loads up in minutes, while your fat takes hours.

Half a century later, the US Navy's Robert Workman sharpened the idea. Instead of one ratio that applied at every depth, he gave each compartment its own ceiling, and named it the M-value: the most dissolved gas a tissue can hold at a given depth before bubbles become likely. That ceiling sits higher when you are deep, where the surrounding pressure helps hold the gas in solution, and drops as you rise toward the surface.^{[2 ]} Albert Bühlmann, working in Zürich, then expanded the scheme to sixteen compartments and published the numbers the whole sport still runs on.^{[3 ]} Each compartment carries its own M-value: fast tissues load and unload quickly and tolerate more supersaturation, while slow ones tolerate less, so the ceiling is really a family of lines rather than one. The set he tuned for dive computers, ZH-L16C, is the engine inside most planners today, including Dive Kit's.

Picture the dissolved gas like the fizz held in a capped bottle. Pressure keeps it in solution; let the pressure off too quickly and it bubbles out. The M-value marks how far you can push before it fizzes, which is why the picture below calls it the fizz line. But there is a catch, and it is the reason the rest of this story exists. An M-value is not a hard wall. It is a best estimate, drawn from limited data, of the point past which DCS becomes likely. It is a real limit, but no single number fits every diver: bodies differ in size, build and composition, and even the same diver varies from one day to the next. Some divers have developed DCS well below their M-values; others have surfaced fine above them. So in 1998 an engineer and cave diver named Erik Baker gave divers a way to choose how much of the model's allowance to use, and how much to hold back as a personal margin. He called it the gradient factor.^{[4 ]}

A gradient factor is the diver's own conservatism dial, written as a percentage of the way from no excess gas at all up to the M-value. At 0% you keep no excess dissolved gas, sitting level with the pressure around you, which is as cautious as it gets. At 100% you go right up to the M-value, the model's limit, with no margin to spare. Most divers choose somewhere in between, and a lower number is the more conservative choice. You set it; the model then works out the stops that keep you under it. Move the two handles below to feel how it works.

Gradient factors come as a pair, and now those two slashed numbers make sense. On a dive that takes on enough gas, you cannot head straight for the surface; you pause at a series of decompression stops on the way up to let the gas come off. The first number, GF Low, governs the deep end of that ascent: it sets how close to the M-value you allow yourself at the first and deepest of those stops. Set it low and you allow yourself less excess gas down deep, so the first stop comes earlier and deeper, where you start shedding gas sooner. The second, GF High, governs the shallow end: it sets how close to the M-value you allow yourself at the moment you surface. Set it low and you surface with a wider margin, which usually means more time at the shallow stops. Between the first stop and the surface, the model simply joins the two numbers with a straight line, the diagonal you just tilted with the two handles.

You just saw the fizz line and the safety margin you keep below it. Here is what choosing that margin does to a real dive. The explorer below takes one descent, 45 metres on air for 20 minutes, deep enough to be firmly in technical-diving territory, and shows the actual Dive Kit plan for it at a range of gradient-factor settings. Pick a setting and watch the dive redraw, both where your first stop falls and how long the whole ascent takes.

The case for deep stops

For more than a decade, one idea ran technical diving. If bubbles are what cause DCS, the reasoning went, then stop them before they have a chance to grow, which means pausing deep and early, while the gas is still dissolved and any bubbles are still tiny.

The idea got its name, and its early momentum, from Richard Pyle, a fish biologist who dived deep reefs to collect specimens. The fish he brought up had gas-filled swim bladders that swelled as the pressure dropped, so on the way to the surface he would stop to vent them with a needle, far deeper than his computer's first required stop. He noticed something: on the dives where he made those deep pauses, he felt far less wrecked afterwards.^{[5 ]} It was a personal, uncontrolled observation, but it arrived just as the theory seemed ready to explain it.

That theory came from a different family of models than Haldane's. Where the classic approach tracks dissolved gas, these "bubble models", the Varying Permeability Model^{[6 ]} and the Reduced Gradient Bubble Model,^{[7 ]} assume that microscopic bubble seeds are always present in the body and have to be kept from growing by stopping deep and early. Divers whose computers ran the ordinary Bühlmann model could not load those bubble models, so they mimicked the same deep, early stops with the dial they did have. They wound GF Low right down, to 10 or 20 out of 100, which forces that first stop very deep.

For years this was close to gospel. The awkward part was that nobody had really tested it. When the leading decompression researchers gathered for a workshop on exactly this question in 2008, the only statement they could all sign up to was that the evidence for and against deep stops was, in a word, conflicting.^{[8 ]} In other words, the field's leading researchers could not yet say whether deep stops helped at all.

There had, in fact, already been a warning sign. A few years earlier, a French Navy team had grafted deep stops onto its standard air tables for dives to 50 and 60 metres and tracked the bubbles in the divers' chests by Doppler. The deep stops did not help: of three versions they tried, two made no difference and one made the bubbling worse, badly enough that they stopped running it.^{[23 ]} It was a small study that counted bubbles rather than actual bends, so on its own it settled nothing. But it was the first real hint that the theory might not survive contact with live divers.

The experiment that changed minds

Three years later, the US Navy Experimental Diving Unit settled it, at least for one kind of dive. The design was beautifully simple. Take a single demanding dive, 52 metres (170 feet) on air for 30 minutes, and bring divers up from it two different ways that spend exactly the same total time stopping. One schedule front-loads the stops deep. The other gets shallow sooner and spends those minutes near the surface instead. The dive and the total decompression stay identical; the only thing that changes is where those minutes go.^{[9 ]}

The deep schedule lost, and not by a little. It produced more than three times as much DCS as the shallow one, enough that the researchers halted the trial early rather than keep giving their divers more of it.

The reason is in those tissue compartments. While you hold deep, your fast tissues, your blood and brain, do off-gas. But the pressure down there is still high enough that your slow tissues, your fat and cartilage, are doing the opposite, still soaking gas up. Your decompression budget is just the fixed stretch of stop time the schedule allows you. Spend it down deep and you clear a little extra from the fast tissues, but you load the slow ones at the same time. Slow tissues let go of gas so reluctantly that the extra they took on is still with you when you surface, so even though the fast tissues come up cleaner, the gas debt you carry overall is larger. The Navy put its conclusion bluntly: holding deep to control bubbles was "unwarranted for air decompression dives."^{[9 ]} Two years later, David Doolette and Simon Mitchell, two of the most cited names in the field, put the same point in print. The ordinary dissolved-gas method, the kind researchers call a gas-content algorithm, had turned out to beat the bubble models rather than lose to them.

The evidence since NEDU

NEDU did not end the conversation, but everything since has pointed the same way. Take a 2017 study from Italy. Researchers put 51 divers through the same deep dive, 50 metres for 25 minutes on trimix, and split them between two ways up: a Bühlmann computer set to GF 30/85, and "ratio deco", an older rule-of-thumb method that stops even deeper and takes longer. Both groups surfaced with about the same number of bubbles. But the ratio-deco divers, the ones who had stopped deeper, showed noticeably more inflammation in their blood vessels.^{[11 ]} It is a small study and nobody developed DCS, so it is a hint rather than a verdict, yet once again going deeper did not turn out to be gentler.

A 2023 study for the Belgian Navy came at the question from the opposite end. Instead of diving, the researchers ran the Bühlmann model through a huge range of GF settings on a computer and asked which ones best matched two decompression tables that decades of use had already proven safe. The title gives away the answer: more conservative settings are not necessarily safer. The closest matches kept GF Low high and added caution only by lowering GF High. It is worth pausing on that asymmetry. A low GF High is the cautious choice, since it brings you up further from the edge, but a low GF Low is not, because it forces the risky deep stops; the caution lives at opposite ends of the two dials. The popular "conservative" default of 30/70, with its deep first stops, matched the safe tables worst of all, because those deep stops loaded the slow tissues more.^{[12 ]}

A third line of evidence came from auditing the dive computers themselves. In 2018, Doug Fraedrich tested whether the common algorithms could reproduce the US Navy's known-safe profiles, and found that Bühlmann matched them only when GF Low was kept at 55 or higher and GF High at 70 or lower. Wind GF Low below 55 and the model pushed the first stop deeper than the validated range.^{[24 ]} Once again the deep end of the dial is where the risk creeps in, which is why the recent work keeps nudging GF Low up rather than down.

So the expert view has shifted. The advice now is to stop chasing deep stops and put the time shallower instead, by keeping GF Low higher and leaning on GF High for conservatism. It is telling how carefully the experts phrase their own choices. When David Doolette mentioned the setting he personally uses, GF 70/85, he went out of his way to call it his own judgement, not a number anyone should copy.^{[13 ]}

The variables a computer cannot see

The most humbling result came in 2023, when Doolette and Murphy had the same divers repeat one dive on different days and counted their bubbles each time. Common sense says one person on one profile should bubble about the same every time. They did not. Roughly two-thirds of all the variation in bubble scores showed up within individual divers, day to day, rather than between one diver and the next: the same body on the same dive, bubbling differently from one occasion to the next.^{[14 ]} Simon Mitchell singled the finding out in his comprehensive 2024 review.^{[15 ]}

If the algorithm cannot promise safety, the next question is what actually moves the odds, and the biggest real-world dataset we have offers some answers. An analysis of the DAN diving-safety database looked at 127,957 dives from 5,907 divers.^{[16 ]} Of everything a diver can change, the one thing most tightly linked to developing DCS was how full of dissolved gas they still were the moment they surfaced. The study writes that as a percentage of the model's limit. They call it the surfacing gradient, a different use of the word from gradient factor, and 100% means right at the edge. That surfacing percentage is essentially what GF High sets: choose a lower GF High and you surface at a lower one. Divers who developed DCS had typically surfaced at 87% of the limit; divers who stayed clear, at 74%. Read plainly, the ones who came up closer to the edge developed DCS more often, which is about as direct a piece of real-world support for a lower GF High as you could ask for.

That chart says two things at once. Risk climbs sharply as the surfacing percentage rises, yet about nine in ten of the recorded DCS cases still happened between 70% and 90%, exactly the range most divers would think of as ordinary.^{[16 ]} That is not the contradiction it looks like. The chance of any single dive going wrong is highest right at the edge, but the great majority of dives are made in that ordinary 70% to 90% band, so that is where the bulk of the cases pile up: many slightly risky dives outnumber a few very risky ones. There is no point on the curve where the risk falls to zero. There is only a slope, and you get to choose where on it to stand. Lowering GF High walks you down that slope, and it spends the extra time where off-gassing works best, up shallow.

The dial is only part of the picture, though. The same DAN analysis turned up large differences in risk that no GF setting touches. Some of it comes down to who you are: in this data, women developed DCS far more often than men, for reasons still being worked out. Some of it comes from the dive itself. Working hard at depth raised the odds, and so did carrying more than one breathing gas, though that usually just marks a longer or more demanding dive rather than doing harm on its own. Temperature mattered as much as the settings did, and in a way that catches people out: being warm on the bottom helps you soak up gas, while being cold on the way up slows you giving it back, so the dangerous combination is a warm, hard-working bottom followed by a cold, shivery deco.^{[17 ]} And a little of the risk is anatomy. About one diver in four has a small flap between the chambers of the heart, a patent foramen ovale, that can let bubbles slip from the veins across into the arteries. Bubbles on the venous side are usually filtered out harmlessly by the lungs; the ones that cross into the arteries instead ride straight to the brain and other organs, where they do far more harm.^{[22 ]}

One last bit of folklore is worth puncturing: the so-called helium penalty. Because Bühlmann's model treats helium as moving about 2.65 times faster than nitrogen, it asks for longer decompression on trimix, not shorter.^{[10 ]} That feels backwards until you follow it through. Gas that moves out fast also moves in fast, so your tissues take on a lot of helium during the dive, and the model then holds that helium to a tighter limit on the way up. Some divers resent paying the bill. Yet when researchers measured the two gases in sheep, they behaved almost identically in the brain and muscle, the tissues that drive short "bounce" dives, the quick down-and-up kind where only the fast tissues have time to fill. The real difference only showed up in the slowest tissues.^{[20 ]} So the penalty may rest partly on an oversimplification. Even so, Simon Mitchell's warning holds: do not try to dodge it by telling your computer you are on plain nitrox when you are really on trimix.^{[21 ]} Bühlmann may have landed on roughly the right amount of decompression for the wrong reason, and trimming that time just throws away a buffer that works.

Where the agencies stand

Researchers are one thing; what your training tells you is another. Some agencies have begun writing the shift into their published guidance, though they are not all in the same place yet.

The most explicit is CMAS, the world underwater federation. Its 2025 fact sheet on gradient factors tells recreational air and nitrox divers to set GF Low equal to GF High, somewhere in the 80 to 90 range (90/90, 85/85, 80/80), and keeps the lower GF Low values for helium mixes. It says in as many words that on air or nitrox a GF Low below GF High forces deep stops that raise the risk of DCS, and it points to the same French, US Navy and Belgian work covered above.^{[25 ]}

PADI's technical arm moved first among the large agencies. In 2019 it dropped the requirement to teach deep-stop strategies, in an article reviewed by Simon Mitchell that pointed straight at the NEDU result, and in 2023 it rebuilt its Tec 40, 45 and 50 courses around current thinking on gradient factors, gas density and deep stops.^{[26 ]} Others have moved more slowly. GUE still lists 20/85 as the default in its DecoPlanner software and technical courses, even as many of its own instructors have drifted toward higher GF Low values,^{[27 ]} and agencies such as TDI publish no fixed gradient-factor figure at all, leaving the choice to the instructor and the diver. If you were taught to dive 30/85 and never told why, that is the reason: the answer you got depended a great deal on who taught you, and when.

Choosing your numbers

Notice what all of this does not give you: a number. That is not a cop-out. It is the honest state of the science. What the last fifteen years have produced is not a setting but a handful of principles you can actually use.

• More decompression beats less. Time in the water is the most dependable safety lever you have, and cutting a schedule short is almost always the riskier move.
• Deep stops are not free. Every minute you hold deep, your slow tissues may still be filling. The evidence no longer supports forcing very deep stops for their own sake.
• For more margin, reach for GF High, not GF Low. Lowering GF High puts the extra time up shallow, where off-gassing actually works, and that is exactly where the DAN data say the surfacing gradient matters most.
• The setting is not the whole dive. Your temperature, your workload, your hydration, your own body, even the day, can matter more than the gap between two neighbouring GF values.
• Be honest with your computer. The model's safety buffers are calibrated; gaming them to save time mostly just removes the margin you were counting on.

In the end, gradient factors are a dial for balancing your own appetite for risk against your own body, and both of those are personal. A fit, warm, well-rested diver on an easy profile is in a different place from a cold, tired one at the end of a hard week, even on the identical dive. The aim was never to find the "right" numbers. It is to understand what the dial actually does, so that whatever you set is something you chose on purpose, rather than inherited and forgot.

That is the spirit Dive Kit's planner is built in. It runs the same Bühlmann ZH-L16C the rest of this article describes, and it hands the gradient factors to you rather than hiding them behind a single conservative-or-aggressive switch, showing the trade-off each setting makes instead of burying it. But the real tool here is not an app. It is the habit of asking what the evidence says and staying honest about what it does not. Dive within your training, and give your body the time it needs to let the gas go.