Posted by Project Phoenix began as an act of stubbornness. In October 1993 Congress cancelled NASA’s High Resolution Microwave Survey, a search for radio transmissions from other civilizations that had been in development for about a decade and had been observing the sky for exactly one year. Nothing had broken. The receivers worked, the software worked, and the line item was roughly $12 million a year, a rounding error in a federal budget. Senator Richard Bryan of Nevada attached the amendment that killed it and told reporters, in so many words, that the taxpayer was done paying to hunt for little green men.
He was wrong about the “done” part. Within a year the SETI Institute in Mountain View had raised private money, rehired most of the engineers, and restarted the targeted half of the dead NASA program under a name nobody needed to have explained to them. Project Phoenix observed from 1995 to 2004, examined roughly 800 nearby stars, and detected no confirmed signal from anybody. That is usually where the story stops. It shouldn’t, because how it found nothing is the part that still shapes the field.
Silicon Valley Bought the Telescope Time
The funding came from people who had made their money on the same technology the search depended on: Paul Allen, William Hewlett, David Packard, Gordon Moore, and Barney Oliver, who had run research at Hewlett-Packard and had been arguing the scientific case for radio SETI since the 1970s. A few million dollars a year, none of it public. Jill Tarter, who had worked on the NASA program at Ames, became project director and stayed through the entire run. If the name is familiar, Jodie Foster spent time with her before playing Ellie Arroway in the 1997 film of Carl Sagan’s Contact.
Private funding changed the science in a way that is easy to miss. NASA’s program had two halves, a targeted search of individual stars and an all-sky survey. Phoenix could only afford one. It took the targeted half, which meant committing in advance to a specific list of stars and defending that list.
A Target List Built Before Anyone Had Found a Planet
Tarter and her colleagues assembled roughly a thousand main-sequence F, G, and K stars within about 200 light-years, sorted by how much they resembled the Sun and how close they were. About 800 of them were eventually observed. Here is the detail that surprises people: when that list was compiled, not one planet was known around any Sun-like star anywhere. Zero. The first, 51 Pegasi b, was announced by Michel Mayor and Didier Queloz in October 1995, months after Phoenix had already started observing, and it turned out to be a Jupiter-mass object whipping around its star in four days, which is exactly nobody’s idea of a nice place to live.
So the target list was built on a proxy. A star like the Sun, near enough that a transmitter of plausible power would still be detectable, was as good a bet as the astronomy of 1994 could offer.
Today you would build it completely differently. Kepler and TESS have handed astronomers thousands of confirmed planets with measured sizes and orbits. TRAPPIST-1 turned out to have seven Earth-sized worlds in 2017. Proxima Centauri, the nearest star to the Sun, got a planet in the habitable zone in 2016. Phoenix had none of that and did the search anyway, which in retrospect looks less like a limitation than like the only honest move available.
Parkes, Green Bank, Arecibo
Phoenix never owned a telescope. It rented time, and the team hauled several tons of custom-built signal-processing hardware to whichever observatory had granted it. The first campaign ran at the 64-metre Parkes dish in New South Wales in early 1995, working through roughly 200 southern stars that no northern telescope could reach. From 1996 the search moved to the 140-foot telescope at Green Bank in West Virginia, which is a nice piece of continuity, because it was at Green Bank in 1960 that Frank Drake pointed the 85-foot Tatel dish at Tau Ceti and Epsilon Eridani and invented the whole enterprise. From 1998 until the end, Phoenix used Arecibo in Puerto Rico in blocks of a few weeks a year.
Arecibo mattered because of raw collecting area. Its 305-metre dish, sitting in a natural sinkhole in the karst hills, gathered more radio energy than anything else on the planet. With it, Phoenix was sensitive enough to have picked up a transmitter comparable to Arecibo’s own planetary radar from a couple of hundred light-years away. The search covered frequencies between roughly 1200 and 3000 megahertz and sliced them into tens of millions of channels at a time, each about one hertz wide.
That last number is the whole logic of radio SETI. Nature does not produce one-hertz tones. Hydrogen clouds, pulsars, and masers all smear their emission across far wider bands, because the atoms and electrons doing the emitting are moving in every direction at once. A signal squeezed into a single hertz has to come from a transmitter. The trouble is that Earth is covered in transmitters.
The Trick With Jodrell Bank
Interference is not an occasional nuisance in this work. It is the permanent condition. Radar, aircraft, satellite downlinks, GPS, a failing motor in a nearby building, the observatory’s own electronics: a sensitive dish hears all of it, and much of it is narrowband, because human engineers like narrowband for the same reason a hypothetical alien engineer would.
Phoenix solved this by always observing with a second telescope somewhere else. At Parkes the partner was the 22-metre Mopra dish a couple of hundred kilometres away. At Green Bank it was a 30-metre antenna at Woodbury, Georgia. At Arecibo the partner was the 76-metre Lovell Telescope at Jodrell Bank in England, linked over the internet, more than 6,000 kilometres from the primary dish and pointed at the same star.
The reasoning is beautifully simple. Local interference at Arecibo does not exist at Jodrell Bank, and vice versa. Better still, a genuinely distant source drifts in frequency at a slightly different rate as seen from Puerto Rico and from Cheshire, because the two sites sit on a rotating Earth and are moving toward or away from the star at different speeds. That differential drift can be calculated in advance for any target. A candidate had to show up at both sites with the right drift difference or it was thrown out. The check ran automatically, within minutes, while the telescope was still on the star.
Everything It Found Turned Out to Be Us
Candidates were not rare. The system threw up millions of hits per observing run, almost all of them killed instantly by the automated tests. A handful survived long enough to be interesting.
The one everybody remembers happened at Green Bank in June 1997. A narrowband signal held up through the on-source and off-source pointing tests, drifted the way a distant source should, and stayed put for hours while the team worked through the checklist and slowly ran out of ways to explain it. Somebody eventually thought to check the catalogue of spacecraft. It was SOHO, the solar observatory, a million and a half kilometres away in the same patch of sky. Tarter has said that the hours before that identification were the closest she came to believing it had happened.
Nothing else got that far. The final observations ran at Arecibo in March 2004, the target list was declared finished, and the SETI Institute put its effort into building its own instrument instead.
What a Non-Detection Is Actually Worth
It is worth being precise about what Phoenix ruled out, because the popular version (“scientists listened, heard nothing, aliens probably aren’t there”) is not what happened. Phoenix looked at 800 stars out of a few hundred billion in the galaxy, in one frequency band out of an enormous range, for a few minutes each, for signals that were continuous and narrowband and pointed roughly our way while someone was listening. Tarter’s own comparison, from a later analysis of how much of the search space SETI has covered, is a single glass of water dipped out of the ocean. That is not a result about aliens. It is a result about one glass.
What Project Phoenix did settle was methodological, and the field still runs on it. A privately funded team can win time on the world’s best telescopes and use it competently. Real-time verification against a second dish on another continent is practical, not theoretical. And a candidate signal is guilty until proven innocent, which is now simply how this is done. When Breakthrough Listen announced BLC1, a narrowband candidate found in Parkes data pointed at Proxima Centauri, the follow-up analysis that identified it as human interference used exactly the reasoning Phoenix had industrialised twenty years earlier.
The Allen Telescope Array in northern California, funded by the same Paul Allen and observing since 2007, is the direct institutional descendant: an array of small dishes dedicated full-time to the search rather than borrowed by the fortnight. Arecibo itself is gone. The platform came down in December 2020, and no instrument on Earth currently matches what Phoenix had at its disposal between 1998 and 2004.
Which leaves the question exactly where Tarter left it, unanswered and still worth asking. If you want to follow where the search has gone since, the arguments about technosignatures, the candidate signals and the ones that fell apart under scrutiny, come and read along with us at SETIworld. There is a lot of ocean left.
