Vincent van Gogh’s The Starry Night is usually described as imagined — a swirling sky out of the painter’s head rather than out of a window. A new study argues the opposite: that the sky in the painting is a real dawn, and that it can be dated.
The method
A team of astronomers at Kyung Hee University in South Korea used astronomy software to reconstruct the dawn sky over southern France for every day in June and July 1889, then compared each reconstruction with the arrangement of celestial objects in the painting.
The constraint they worked from is the one visible detail that is hard to arrange by accident: in the painting the moon sits above Venus. They looked for a dawn on which both bodies fell inside the narrow frame the canvas shows.
Only three dates in those two months matched at all. The closest match was 22 July 1889, which shows a thin moon, the planet Venus, and the Hyades star cluster — all three of which appear in the work.
Until now the painting has been dated to June. The new reading moves it a month later.
What the researchers say
“I stood exactly where Vincent stood 130 years ago,” said astronomer Elijah Jong-hyun Kim of Kyung Hee University.
He was careful about the standard of proof. “To bring this subject under the scope of science we had to measure the similarity between the painting’s star pattern and the real sky using the geometry of shapes and statistics,” he said. The full study is published in the journal Scientific Reports.
That framing is the interesting part. The claim is not that the picture is a photograph. It is that the pattern of objects matches one reconstructed sky more closely than the others, by a measure you can state and someone else can check.
Why the museum is not persuaded
The Van Gogh Museum in Amsterdam was not involved in the research, and does not accept the conclusion.
Its position rests on documentary evidence rather than on the sky. Experts have established that Van Gogh painted The Starry Night in June 1889, on a canvas he obtained at the start of that month. And around 22 July — the very date the astronomers propose — the artist suffered a severe mental health crisis that left him unable to paint for more than a month.
A museum spokesperson added the deeper objection: Van Gogh was not concerned with exact representation. The positions of the celestial objects he depicted cannot therefore serve as a reliable basis for dating the work.
Two kinds of evidence, pointing different ways
This is a clean example of a problem that turns up well outside art history. One side has a computational match with a stated error measure. The other has a purchase record, a documented crisis and a known fact about the subject’s working habits.
The astronomy is not wrong about the sky. 22 July 1889 really is the dawn on which that arrangement appeared. What the calculation cannot establish is the step everyone wants it to establish: that Van Gogh was at a window, on that morning, transcribing it. A painter who reorders what he sees breaks the chain between the sky and the canvas, and no amount of positional precision repairs it.
Which leaves a narrower, more durable finding than the headline suggests, and arguably a more interesting one. The sky in The Starry Night is not invented. Whatever Van Gogh did with the moon, Venus and the Hyades, he had seen them in roughly that relation from that place. The date remains contested. The observation does not.
Related: the metal reservoir around our galaxy’s black hole.
