The Phoenix Lights, 13 March 1997

On the evening of 13 March 1997, thousands of residents across Arizona, Nevada, the Mexican state of Sonora and the southern California border reported seeing a large V-shaped or triangular formation of lights silently drifting overhead, and, later, a separate series of stationary lights descending along the Sierra Estrella range. The event, known as the Phoenix Lights, remains one of the most widely witnessed mass sightings on record, and usefully divides into two distinct parts: the early-evening V formation and the later lights over the mountains.

The V formation

Beginning around 8:30 local time, witnesses along a path of roughly 300 miles described a solid dark triangular object with lights at each vertex drifting without sound. Observers estimated its size as several times the length of a football field. The consistency of the shape, the silence, and the description across widely separated witnesses distinguish this from the later lights, and it has never been officially explained.

The flares

Roughly 90 minutes later, a series of lights appeared in a formation over the Sierra Estrella range and dropped slowly behind the mountains, winking out one at a time. In March 2007 — ten years later — the Maryland Air National Guard's 104th Fighter Squadron confirmed that its A-10 aircraft had dropped long-burning LUU-2/B illumination flares over the Barry M. Goldwater Range at exactly that time. The flares account, convincingly, for the second event: stationary lights that vanish behind a mountain silhouette. They do not account for the earlier V formation.

The governor

Governor Fife Symington initially defused public pressure with a 1997 press conference in which an aide appeared in a green alien costume. A decade later, in 2007, Symington publicly stated that he himself had witnessed the V formation that night, that it struck him as "otherworldly," and that he had downplayed it at the time to avoid panic.

Why the case is instructive

The Phoenix Lights are valuable as a reference case precisely because the same event contains both a documented conventional explanation and an unexplained component. It illustrates the discipline this archive tries to apply: separate the flares from the V, attribute what can be attributed, and record what cannot. Conflating the two — using the flares to explain the V, or using the V to dismiss the flares — is exactly the failure of analysis the archive is designed to avoid.