The Quiet Galaxy

Explanations · They are there and we cannot see them

Expansion stalls: civilisations do spread between stars, but settlement spreads like a stain, not a flood, and leaves large regions, perhaps ours, never reached

Proposed by Geoffrey A. Landis, 1998 fc left alone

The argument that someone should already be here assumes that once a civilisation starts settling other stars it keeps going until it has settled them all. Geoffrey Landis pointed out that this is not how spreading works. If each new settlement has some chance of becoming the kind that sends out more ships and some chance of becoming the kind that does not, then the mathematics of how things spread through a network, called percolation, says the result depends on that chance. Below a threshold, settlement fizzles out in a small cluster. Just above it, settlement spreads but leaves large unsettled holes. Earth could sit in one of the holes, with settled stars a few hundred light-years away in several directions and none nearby.

Geoffrey Landis, a NASA scientist who also writes science fiction, published the idea in 1998 as a direct answer to Hart's argument that any interstellar civilisation would have settled the whole galaxy. He made three assumptions. First, interstellar travel is possible but difficult, so a civilisation settles only nearby stars, within a few tens of light-years. Second, a new settlement is in effect a new civilisation, with its own culture, because the distances prevent any central control. Third, each new settlement has some probability of being expansionist, sending out ships of its own, and otherwise is content to stay put. Under these assumptions, settlement spreads through the network of stars the way water spreads through coffee grounds or a disease spreads through a population, and the theory of that process was already well developed.

The result has a sharp threshold. If the probability of a settlement being expansionist is below a critical value, every expansion dies out after a few generations and the civilisation stays in a small cluster around its home star. Above the value, expansion goes on without limit, but it does not fill space evenly: it leaves voids, regions surrounded by settlements but never entered, because the settlements at the edge of the void all happened to be the non-expanding kind. Near the threshold the voids are large, and a planet in one could be hundreds of light-years from the nearest settled star while the galaxy as a whole was well occupied. Landis argued that Earth may be in such a void, and that the silence is a local fact about our neighbourhood rather than a global fact about the galaxy.

William Newman and Carl Sagan had argued along related lines in 1981, modelling settlement as a slow diffusion rather than a rapid wave and concluding that a galaxy could take far longer to fill than Hart assumed, especially if populations grew slowly. Landis's contribution was the threshold and the voids.

The bearing on the Drake Equation is nil, in the sense that the explanation says nothing about how many civilisations exist or how long they last; it answers Hart's question about settlement rather than Drake's about detectability. For the catalogue the factors are left unchanged and the silence is put down to geography. If the galaxy is settled with voids, there may be many civilisations within radio range beyond the void's edge, and the explanation predicts that they will be found by looking farther rather than by looking longer.

The objections: the model assumes that an expansionist settlement, once established, never later changes its mind and that a content one never later starts expanding, which over millions of years is a strong assumption; later simulations with more realistic assumptions, including the motion of stars over time (Carroll-Nellenback, Frank, Wright and Scharf, 2019), found that stellar drift keeps a settlement front from stalling and fills in the systems behind it in less than the age of the galaxy, so the permanent voids of a fixed map do not last; they also found that unsettled regions can still exist at any one time if civilisations are short-lived or suitable worlds are clustered. The model also says nothing about why no settled region, anywhere, has produced a signal or a visible mark, which is a question about detectability rather than settlement. Landis's reply was that the model was meant to show that Hart's conclusion did not follow from Hart's premises, and on that narrow point it is generally agreed to succeed.

What it says about the equation

Every explanation is a claim about one or more of the seven pieces that are multiplied to give the number of detectable civilisations. This one bears on 1 piece:

The test: what would strengthen or weaken it

Would strengthen it: Detection of settled or technological systems at a distance of a few hundred light-years with none nearer, which is the pattern the model predicts; evidence from Earth's own future settlement of the solar system that daughter settlements often stop expanding.

Would weaken it: A galaxy-wide survey finding no settlement anywhere, which the model does not predict; simulations confirming that stellar motion erases voids on timescales short compared with the galaxy's age; a detected nearby civilisation, which would put us outside any void.

These two lines are what the record is read against when a new paper or discovery is logged. A result of the kind named here, published with its evidence, is what would settle it; a press claim on its own settles nothing.

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

    Two models with settlement that leaves gaps: Roudenko and Pierre-Boyer (2024-10-31) simulate competing civilisations and find even dominant ones leave large gaps between colonies; Ivliev (2026-06-11, unreviewed) argues expansion past a threshold is quiet and saturates a neighbourhood within ten million years, so none has reached us. No simulation on whether stellar motion erases voids. Logged both.

    Roudenko, 2024Ivliev, 2026 source 1source 2

How seriously it is taken

Widely cited and accepted as a valid answer to the strong form of Hart's argument; later work suggests voids are not permanent when stars move, so most researchers see it as weakening Hart's case rather than resolving the question.

The papers that bear on it

Books, films and other work that take it up

Read next

Notes

Checked 2026-10-16. Borne out: Landis, 'The Fermi Paradox: An Approach Based on Percolation Theory', Journal of the British Interplanetary Society 51, 163-166 (1998), per the society's own listing (https://bis-space.com/shop/?p=13987); Newman and Sagan, 'Galactic civilizations: Population dynamics and interstellar diffusion', Icarus 46, 293-327 (1981), https://doi.org/10.1016/0019-1035(81)90135-4, whose analysis the 2019 paper summarises as finding that the settlement wave could be slow, with the time to fill the galaxy approaching or exceeding its age. Carroll-Nellenback, Frank, Wright and Scharf, 'The Fermi Paradox and the Aurora Effect', Astronomical Journal 158, 117 (2019), https://arxiv.org/abs/1902.04450, say stellar motion stops settlement fronts from stalling, that systems behind a front fill in within less than the galaxy's age, and that earlier fixed-map models such as Landis's cannot support the persistence of large uninhabited regions over long times; but they also find that with finite civilisation lifetimes the galaxy may sit in a steady state with unsettled regions. Corrected: the body said only that voids fill in; it now states both findings. The standing and would_weaken lines were left as they were and still read correctly.

Think this page gets the explanation wrong, or misses an objection? Say so, or read how disagreement gets folded in. The record itself is data/explanations/expansion-stalls.toml in the public repository. This page's address will never change; if the explanation is ever retired, the address will say what replaced it.