The archive holds three dozen dossiers filed under lights — not ghosts, not creatures, just light: a glow where the record cannot put a source. An apparition can be dismissed as a story; a light keeps showing up in front of parked cars full of people who came specifically to see it. What follows is not a taxonomy of ghosts. It is a tour of the places where the phenomenon is documented and the explanation is not.
The ones with paperwork
The most official light in the file is the one at Marfa, where Texas spent $720,000 of state and federal money on a viewing area — binoculars, bronze plaques, an adobe restroom — for something it could not name. The record there runs back to 1883, when a cowhand named Robert Reed Ellison took the flickers toward the Chinati Mountains for Apache campfires and was told no one who rode out ever found ashes. The counter-record is almost as good: in 2004, a dozen physics students from UT Dallas spent four days with lasers and chase cars and correlated the lights with traffic on US-67. The folklore's answer is in the dates — the sightings predate the highway — and the file keeps both.
Brown Mountain is the only light in the archive the federal government has formally investigated twice. The U.S. Geological Survey sent D. B. Sterrett in 1913; he blamed locomotive headlights, and the verdict held until the flood of 1916 shut the railroads down and the lights kept to schedule. A second geologist, George R. Mansfield, spent two weeks in 1922 with a telescopic alidade and produced a report the file calls a masterpiece of disappointed bookkeeping — trains, automobiles, brush fires. Researchers later found no published account of the lights before about 1910, which means the oldest layer of the legend may be younger than the locomotives blamed for it.
The decapitated brakeman, nationwide
Read enough of these files and one figure recurs: a railroad man who lost his head and now swings his lantern along the grade, looking for it. At Maco, the legend belongs to Joe Baldwin, decapitated in 1867 — but the newspapers document a different man entirely: conductor Charles Baldwin, killed in an 1856 collision at the spot the papers called Rattlesnake Grade, in a wreck that involved a lantern hung where it should not have been. No documentary match for Joe has ever surfaced. The light still fooled engineers badly enough that Maco's signalmen answered it with two lamps, and Grover Cleveland's train got the whole story in 1889.
The same legend is filed at Gurdon, where the documented part is a murder — section foreman Will McClain, beaten to death by a member of his own crew in 1931, the killer executed the next year — and the undocumented part is the light that began appearing along the track soon after. At Saratoga the headless railroad man walks a grade the county bought after the rails came up in 1934 — except witnesses there filed reports while the tracks were still down, which is exactly why the easy explanations keep failing. The archive notes, with some dryness, that the identical story is attached to the Paulding Light in Michigan, where no railroad is even recorded in the valley.
When the explanation arrives on letterhead
The Summerville Light is the rarest thing on this list: a ghost light with a scientific paper behind it. In January 2025, USGS seismologist Susan Hough proposed that the glow over the old railroad grade may be an earthquake light — the town sits inside the epicentral region of the great 1886 Charleston earthquake, and the surviving accounts cluster near small quakes recorded in 1959 and 1960. The folk explanation is a widow's lantern, walking out nightly to meet a train that derailed. Seismology, Hough concedes, cannot say whether ghosts are real. It can say why the reports started when they did.
At Spook Light Road — the Devil's Promenade, where Oklahoma, Missouri and Kansas meet — the witnesses have excellent stories and the debunkers have excellent experiments. Folklorist Vance Randolph recorded seeing the light three times for his Ozark books; skeptics from the Kansas City Star to Popular Mechanics showed that headlights on Route 66, ten miles off, read from the gravel road as a floating ball of fire. And at Cohoke the file is blunt: two legends on record, neither checks out, the light appeared anyway, and the sheriff spent the 1960s ticketing the crowds it drew.
What the file cannot say
The honest reading of these nine dossiers is not that the lights are unexplained — most are explained, at least in part, and the explanations are usually headlights, inversions and geology doing quiet work. What is unexplained is the persistence. The reports at Marfa predate the highway blamed for them; the Brown Mountain lights kept to schedule through a flood that stopped every train; the Maco light retired only when the rails did. The record does not rule out a residuum of something stranger. It does not require one.
What is documented, in every case, is the watching. People have driven to dead ends and dark crossings for a century to look at lights their grandfathers also drove out to see. Whatever draws the cars to the shoulder of US-90 is older than the highway it is blamed on. The file keeps both.