We Measured 1,204 Nazca Lines Against The Sky. They Point Downhill.

NAZCA, Peru — Not one of the twenty-seven survived.

That is the result of a test this newspaper wrote down in full on 14 January 2026, before it had a single azimuth: 1,204 straight lines on the Nazca pampa, measured against 27 astronomical targets back-calculated for the window 200 BCE to 650 CE, with the tolerance, the multiple-comparison correction and the stopping rule all published in advance and time-stamped.

Four targets cleared an uncorrected threshold. Twenty-seven targets will do that on their own.

The variable that does predict where a line points is the ground it starts on.

What was fixed before we looked

The pre-registration is published with this story and it was not amended.

It names the 27 targets: the four solstice risings and settings, the equinox rising and setting, the major and minor lunar standstills, and the heliacal risings of nineteen stars, among them the Pleiades, Sirius, Canopus and both Centauri. It fixes the tolerance at ±1.5 degrees. It names the correction. It states the stopping rule, which is the sentence that mattered most: we would run the test once, on the whole survey, and report it.

It also commits, in a paragraph our lawyers read twice, to publishing the result whichever way it came out.

The survey

Dr. Aurelio Chambi-Fenwick flew 21 square kilometres in February under permit, with ground control to four centimetres. The point cloud is published. So are his residuals.

From it, 1,204 lines long enough and straight enough to have a defensible bearing.

Two analysts were then handed the identical data and the identical plan. Dr. Solveig Nakamura-Ruiz and Professor Tobias Aguirre-Lindqvist did not know each other's names until 2 June.

What they got

The same null, by slightly different routes.

Nakamura-Ruiz found four targets clearing p < 0.05 uncorrected. Aguirre-Lindqvist found four as well, but only two of them the same four — which is itself the finding.

After the correction that had been named in January, nothing.

"Four targets cleared the uncorrected threshold," Aguirre-Lindqvist said. "That is what twenty-seven targets do. If I had stopped there I would have had a paper, and the paper would have been about my stopping."

What did predict a bearing

Both analysts, independently, added the same two non-astronomical variables at the stage the plan permitted.

Of the 1,204 lines, 939 — 78 per cent — leave their origin within three degrees of the downhill gradient at that point.

And 734 of them, 61 per cent, end within 400 metres of a dry watercourse.

Neither number is exotic. Both are large, and both are far larger than anything the sky produced.

The lines run where walking is easy, and they run towards where water sometimes is.

Our own correction

That sequence is the reason the plan existed. A newspaper that pre-registers a test and then prints the exciting interim number has pre-registered nothing.

What the desk takes from this

Dr. Rocío Ventura-Halstead's reply runs below and it is the strongest thing on this page. Her argument is that we tested the wrong object: that 1,204 lines made over eight centuries are not one family, that pooling them and correcting across 27 targets is a machine for producing nulls, and — the part this desk has not been able to put down — that gradient and sky were never rivals. You walk downhill to the water. The sky tells you when the water comes.

We cannot answer that with our data, and we are not going to pretend otherwise in the last paragraph of our own story.

What we can say is narrower and we will hold to it. Taken as a whole, this corpus does not point at the sky. It points at the valley floor. Anybody who wants to argue for a smaller family of lines that does point at the sky now has our point cloud, our plan and both working files to do it with, and they will have to say in advance how many targets they intend to try.