The Quiet Galaxy

The equation · piece 4 of 7 · fl

The share of planets that could support life on which life actually starts

Given a planet where water is liquid and the chemistry is right, how likely is it that life gets going at all. We know of exactly one case, and on that planet life appeared early, within a few hundred million years of the surface becoming habitable, which optimists read as a sign it is easy. Pessimists answer that we would be here to notice however unlikely it was, and that the step from chemistry to a self-copying molecule may be so improbable that it happens once per galaxy or once per universe. No measurement exists, and this is the factor on which the whole answer most depends. A second, independent origin of life anywhere (Mars, an icy moon, an exoplanet atmosphere) would settle it.

The range this site uses a rate centred on one start per planet, spread over fifty orders of magnitude reviewed
set on 16 October 2026

When the number on the home page is computed, what is picked is not the share itself but the number of times life starts on such a planet over its life, from a bell curve in its powers of ten; the share is then one minus e to the minus that number, so about half the picks land near 1 and the rest spread evenly across the powers of ten below. The chance at the centre of the curve is 0.632, which is where the calculator's slider for this piece starts.

Why that range

The widest range in the equation, and the one that most decides the headline, so the reason matters most here. The shape is the 2018 paper's (Sandberg, Drexler and Ord, section 3.1), which John ruled on 2026-10-16 should be used here: instead of a low and a high for the chance itself, the number of times life starts on a habitable planet over its life is given a bell curve in its powers of ten, centred on one start per planet and fifty powers of ten wide, and the chance is one minus e to the minus that number. Half the weight then sits near 1 (Drake's and Westby and Conselice's assumption, with Lineweaver and Davis's 95% floor of 0.13 and Kipping's three-to-one for 'life starts easily' close behind), and the other half is spread evenly across fifty powers of ten below, which is where the published pessimists (Koonin 2007, Totani 2020) place it without giving a usable number. The centre at one start per planet is the paper's, taken from life's early start on Earth; the fifty-power spread is the paper's reading of how little is known about the chemistry. A plain low-to-high range cannot carry this two-sided argument: the first pass used one (every power of ten from one in ten to the 22 up to 1), and docs/method.md shows that the choice between the two moved the chance we are alone in the galaxy between 88% and 59%.

Rests on: Sandberg and others, 2018; Koonin, 2007; Totani, 2020.

How much this one piece moves the answer

The same six other pieces, the same number of runs; only the way this piece is picked changes. Each row is a position someone could argue, and the three columns are the three numbers on the home page. The row in use is marked.

How the chance life starts is pickedMiddle estimateAlone in the galaxyAlone in the observable universe
every power of ten from 0.001 up to 1 equally likely 9.66 38% under 0.1%
every power of ten from 1 in 100,000 up to 1 equally likely 0.974 50% under 0.1%
every power of ten from 1 in 10 to the 10 up to 1 equally likely 0.0032 72% 1.2%
every power of ten from 1 in 10 to the 22 (one start per observable universe) up to 1 equally likely 1 in 318 million 87% 39%
every power of ten from 1 in 10 to the 30 up to 1 equally likely 1 in 3.2 trillion 91% 55%
every power of ten from 1 in 10 to the 50 up to 1 equally likely about 1 in 10 to the 22 94% 73%
the 2018 paper's shape: a rate centred on one start per planet, spread over fifty orders of magnitudeThe one this site uses. 0.021 58% 39%

The lesson is the one the 2018 paper drew: treat this piece as known and the silence looks like a paradox; carry the real uncertainty and it does not. The middle estimate in use, 0.021, is about one detectable civilisation per 47 galaxies like ours.

The published estimates

5 estimates on the record, oldest first. Each says who made it, when, how, and where to find it. A range is set from these by hand, with the reason above, never computed from them: the reason is the arguable part.

WhenWhoEstimateHow it was reached
1961 Frank Drake and the Green Bank meetingGreen Bank meeting of November 1961, as recorded in Drake and Sobel, Is Anyone Out There? (Delacorte, 1992), pages 55 to 62 (drake-1961) 1 Assumed: the meeting took it that life starts wherever it can, since on Earth it appeared as soon as conditions allowed.
1980 Carl SaganSagan, Cosmos (Random House, 1980), chapter 12, 'Encyclopaedia Galactica' 0.33 A round guess, one third, in a worked example of the equation for a general audience, leaning on the ease of making life's building blocks in the laboratory.
September 2002 Charles Lineweaver and Tamara DavisLineweaver and Davis, 'Does the rapid appearance of life on Earth suggest that life is common in the universe?', Astrobiology 2, 293–304 (2002) 0.13 to 1 Treated life's early appearance on Earth as one draw from an unknown rate and asked what rates are consistent with it: with 95% confidence, more than 13% of planets that stay habitable for a billion years get life, taking the Earth as typical.
June 2018 Anders Sandberg, Eric Drexler and Toby OrdSandberg, Drexler and Ord, 'Dissolving the Fermi Paradox', arXiv:1806.02404 (2018), section 3.1 (sandberg-2018) from 1 in 10 to the 50 up to 1 Not a value but a spread: the number of times life starts per habitable planet over its life was given a bell curve in its powers of ten, centred on once per planet and fifty powers of ten wide, following the rate model of Spiegel and Turner (2012). The resulting chance has a mean of 0.5 and a median of 0.63, with half its weight near 1 and the other half spread evenly across fifty powers of ten below; the low figure here marks one standard deviation below the centre.
June 2020 Tom Westby and Christopher ConseliceWestby and Conselice, 'The Astrobiological Copernican Weak and Strong Limits for Intelligent Life', Astrophysical Journal 896, 58 (2020) (westby-2020) 1 Assumed: their 'Copernican' approach takes it that life, intelligence and communication follow on any Earth-like planet around a star of Sun-like composition that is old enough, so that the count of civilisations depends only on the galaxy's stars and their ages.

What would move this

The discovery that would narrow the range

Evidence of a second, independent origin of life (not sharing Earth life's chemistry) or of its chemical signatures anywhere beyond Earth (Mars samples, Enceladus or Europa plumes, exoplanet atmospheres such as K2-18b), or a laboratory demonstration of how a self-copying molecule arises and how improbable the step is.

When a paper or a survey of that kind is published, it is logged in the story so far with a note saying whether it moved this range. A move is written as a dated revision below, with the old number, the new one and the reason.

The last reading

Once a week the records are read against the news, each against the test above. Open a line to see what that week held. Every reading ever made of this record is on the reading page.

  • 2026-10-16 Logged, no change By hand

    No evidence of life beyond Earth in the window, and three results the log did not yet carry that bear on this piece's test: the Perseverance rover's potential biosignature in a Martian mudstone (Hurowitz and others, Nature, 2025-09-10; the sample stays on Mars), complex organic compounds confirmed in ice grains fresh from Enceladus's ocean (Khawaja and others, Nature Astronomy, 2025-10-01), and a 45-letter RNA molecule that copies itself and its complement in ice, at 94.1% accuracy and 0.2% yield over 72 days (Gianni and others, Science, 2026-02-13), the kind of laboratory demonstration the test names. None settles how improbable the first step is; the K2-18 b dispute stands where the 2025 reanalyses already in the log left it. The range was left alone: the shape is John's ruling, and the ingredients of life are not life.

    Hurowitz, 2025Khawaja, 2025Gianni, 2026 source 1source 2source 3

How the range has moved

DateWhatFromToWhy
16 October 2026 range.shape every power of ten from 1 in 10 to the 22 up to 1 equally likely a rate centred on one start per planet, spread over fifty orders of magnitude John ruled on 2026-10-16: 'Use the 2018 paper's log-normal rate shape for the chance life starts, adding the third shape to the record layout.' The first pass had used a plain low-to-high range as a stand-in because the layout allowed nothing else; that stand-in put most of the weight on 'life is very rare', where the paper splits the weight evenly between 'life is easy' and 'life is hard'.Because of Dissolving the Fermi Paradox.
17 October 2026 what_would_move_this Evidence of life, or of its chemical signatures, anywhere beyond Earth Evidence of a second, independent origin of life (not sharing Earth life's chemistry) or of its chemical signatures anywhere beyond Earth Ruled by John 2026-10-17 on the outside review (item 11): microbes on Mars that share Earth life's chemistry would show only that rocks travel between planets. The factor's own description already said 'a second, independent origin'; the test line now matches it.
17 October 2026 range.sources sandberg-2018 sandberg-2018, koonin-2007, totani-2020 Ruled by John 2026-10-17 on the outside review (item 10): the two pessimistic papers the range's reason already names now have records in the publications log. The range itself is unchanged.Because of The cosmological model of eternal inflation and the transition from chance to biological evolution in the history of life.
17 October 2026 range.status first-pass reviewed The outside review of the seven ranges ran on 2026-10-17 and John ruled on every item the same day (docs/reviews/2026-10-17-rulings.md).

The explanations that lean on this piece

An explanation for the silence is, in the end, a claim about one or more of the seven pieces. 4 explanations say this piece is smaller than it looks.

The papers that bear on it

Books and films that take it up

Notes

Two lines of argument carry no number and so are not estimates but matter for the range. On the pessimistic side, Koonin (2007) and Totani (2020) argue from the chemistry that a self-copying system is so unlikely to form by chance that it would not start even once in the observable universe (Koonin puts the chance below one in ten to the 1,018 for a coupled copying-and-translating system in one observable region, a deliberately generous upper bound; Totani finds random chemistry in the observable universe's ten to the 22 stars makes RNA chains of about 32 units, against the 40 to 100 needed); both then conclude that life can still arise somewhere in a far larger, unobservable universe (Koonin's infinite multiverse; Totani's inflated universe of more than ten to the 100 stars). Their figures are so small that no computation can use them, and the range treats 'once per observable universe' as the floor they imply. On the optimistic side, Kipping's 2020 Bayesian analysis of when life and intelligence appeared on Earth (kipping-2020) finds that life starting quickly is at least 2.8 times more likely than life starting slowly, 8.7 times if disputed 4.1-billion-year-old carbon traces are accepted, while cautioning that the analysis concerns Earth and says nothing about planets unlike it. Spiegel and Turner (2012) had argued the opposite from the same evidence: that with one example and the fact that we must exist to ask, the early start tells us little. The 'hard steps' argument (Carter 1983; Hanson 1998 on the Great Filter, hanson-1998; Hanson and colleagues 2021, hanson-2021-grabby, who put the number of hard steps at about six) says that whichever steps between chemistry and a spacefaring species are the hard ones, Earth's history fits a world that got lucky several times. Checked 2026-10-16 against the papers: Sagan's one third, Westby and Conselice's assumption that this factor is 1 (their section 2.2), and Spiegel and Turner's conclusion (PNAS 109, 395; the choice of prior dominates, and the evidence is consistent with an arbitrarily low chance) are borne out; the Koonin and Totani figures are as described above, which differs from the first draft. The range's reason below, written earlier, says Kipping's 'three-to-one' (the paper gives 2.8 to 1) and that Koonin and Totani place the chance 'without giving a usable number'; neither changes the range, but both should be revisited at the outside review. Written 2026-10-16. Outside review, 2026-10-17, ruled by John the same day: widening the spread from 50 to at least 1,000 powers of ten (item 9) was declined, because Koonin's and Totani's figures bound one mechanism and would, as a spread, put nearly half the weight below one start per observable universe; the engine also cannot draw a spread above about 80 (the draw overflows), which is a limit for the method's own review. Koonin (2007) and Totani (2020) now have records in the publications log (koonin-2007, totani-2020) and are linked from this factor, but stay in these notes and not in the estimate list (item 10), since an estimate here is a number the width can use and a chance of one in ten to the 1,018 is not one. The test line now asks for a second, independent origin (item 11). The reviewer's proposal to add a survey of exoplanets showing life but no technology to this line (item 14) was declined: that bounds the product of this factor, the next and the lifetime, not this one alone.

Think the range should move? Argue it, with a source. Spotted an error on this page? Say so. Or read how disagreement gets folded in. The record itself is data/factors/life.toml in the public repository.