A signal leaves an aerial in the Netherlands. It crosses the North Sea at the speed of light, which is to say it takes no time worth naming.
Then, eight seconds later, it comes back.
Eight seconds is an enormous amount of time for a radio wave. In eight seconds a signal can reach out three times as far as the Moon and return. In 1928 there was nothing out there to reach. There was nothing out there at all.
The engineer who kept hearing it twice
Jorgen Hals was a Norwegian engineer with a receiver and the habit of listening late.
In 1927, at his home near Oslo, he started hearing something wrong. Signals from the Dutch shortwave station PCJJ at Eindhoven came in clean. Then they came in again. The second copy arrived roughly three seconds behind the first, and it was too crisp to dismiss as noise.
He knew what an ordinary echo sounded like. A short wave that runs the whole way around the planet returns in about an eighth of a second. That was the familiar one, and radio men heard it all the time.
Hals was hearing gaps more than twenty times longer.
So he wrote a letter. It went to Balthasar van der Pol, the physicist at the Philips laboratory in Eindhoven who ran the transmitter and whose name is still attached to an equation about oscillators. Van der Pol did not put the letter in a drawer.
He passed it to Carl Stormer.
The man who computed the invisible
Stormer was a professor of mathematics in Oslo, and he was not a casual choice.
He had spent decades on the aurora. He built a network of camera stations across Norway, photographed the lights from two places at once, and worked out how high they burned. Thousands of plates. It was the most patient observational program anybody had aimed at the sky above Norway.
And alongside it he did something stranger. He computed, by hand, the paths that electrically charged particles would follow through the Earth's magnetic field. The arithmetic was brutal and he did it anyway.
His numbers said particles could be caught. Held. Circling the planet in rings, going nowhere, waiting.
Nobody could check that in 1928. The instrument capable of checking it did not fly until 1958, when the first American satellite carried a detector built under James Van Allen and found the radiation belts wrapped around us exactly where the mathematics had put them.
That is the man who now sat in Oslo with a stopwatch, listening for something late.
Eleven nights and a column of numbers
In the autumn of 1928 the two of them ran it properly.
Van der Pol transmitted from Eindhoven on a fixed schedule, a short signal repeated at regular intervals so that any echo would be unmistakable and any delay could be timed. Stormer and Hals listened in Oslo. The arrangement ran across October and into November.
The echoes came. The first of them arrived on 11 October 1928.
They were not consistent, and that is the part that has never stopped bothering people. Three seconds. Eight. Eleven. Fifteen. Then eight again, and again, and again. The delays clustered but they would not settle, and no simple pattern held them together.
Stormer wrote it up for Nature before the year was out, under a title as flat as a ledger entry: short wave echoes and the aurora borealis. Van der Pol reported his side in the same journal. Two men with nothing to gain from a mystery, describing one in the most sober language available.
What eight seconds actually means
Radio has one speed. That is the whole difficulty.
A three second round trip puts the reflector about 450,000 kilometers away, just past the Moon. Eight seconds puts it near 1.2 million kilometers out, three times the Moon's distance. Fifteen seconds pushes it to something like 2.2 million kilometers.
Nothing sits there. No moon, no cloud, no known surface. The space between Earth and that emptiness was, as far as anyone could then say, simply empty.
So either the signal went somewhere, or the delay was never about distance in the first place.
The star map in the delays
Forty-five years later the numbers got a second life.
In 1973 the Scottish writer and astronomer Duncan Lunan published an interpretation in Spaceflight, the magazine of the British Interplanetary Society. He took the 1928 delay sequence and plotted it, echo order against seconds, treating the list as a message rather than a fault.
What came out, he argued, looked like a star chart. Specifically the constellation Bootes, with an origin implied at Epsilon Bootis.
He had a framework ready for it. In 1960 the radio astronomer Ronald Bracewell had proposed in Nature that a civilization wanting to make contact would not broadcast across the light years at all. It would send a machine, park it near a promising star, and have it wait. And a waiting probe announcing itself would do the cheapest possible thing. It would take our own signal and hand it back, late.
Lunan's version traveled fast and far, and he did not hold it. He reported errors in his own working and withdrew the identification, though he has returned to the material more than once since. The retraction never quite caught up with the story.
The case against
Nobody can produce one on demand. That is the blunt version and it needs saying first.
Long delayed echoes are still reported, largely by amateur operators, and they remain maddeningly occasional. Later campaigns with better receivers than Stormer ever had did not turn them into a reliable phenomenon.
The standard account does not need distant space at all. It holds that the delay is not travel but crawl. A radio wave moving through plasma near its resonant frequency slows down, and in the right conditions it slows down enormously. Add ducting, in which a wave gets trapped along a channel in the ionosphere or a magnetic field line and bounces along it, and you can in principle stretch a fraction of a second into several. Radio scientists at Stanford pursued exactly this line in the 1970s, and magnetospheric ducting remains the leading candidate.
On the Bootes reading, the objection is statistical. Critics argue that a short list of numbers plotted freely will resemble something, and that constellations are among the easiest shapes to find by accident.
The part that will not go away
Here is the awkward residue.
The leading explanation is a mechanism, not a measurement. Ducting and plasma delay are real physics, well grounded and widely accepted. What has not happened, in nearly a century, is the clean demonstration: an echo produced deliberately, timed, and traced along a known path.
The irregularity is the sticking point. A duct implies a route, and a route implies a repeatable delay. What Stormer logged was three seconds and fifteen seconds and eight seconds over and over, from the same transmitter on the same nights.
And there is the matter of who was holding the watch. Stormer had spent his working life on the physics of the space just above our heads. He had already reasoned his way to trapped particle belts that no instrument could yet see. If anyone in 1928 was equipped to recognize an ionospheric artifact and name it, it was him.
He wrote the numbers down and let them stand.
What is left is a delay
Three to fifteen seconds. Logged by careful people, published in the most conventional journal available, never withdrawn.
Either it is the plasma around this planet doing something slow and strange that we understand in outline and have never pinned down in a single controlled trial.
Or a signal from Eindhoven went somewhere in 1928, and came back.
Ninety-odd years on, nobody has produced the mechanism on command, and nobody has produced the echo on command either.