Some astronomers are questioning the existence of what might be the most Earth-like planet yet found outside the solar system, based on a reexamination of archival data.

Kepler 452 b was discovered by NASA’s Kepler space telescope and announced in 2015. At the time it seemed like everything astronomers had hoped for in an Earth analogue: slightly larger and more massive than our planet, and in a habitable 385-day orbit around a star remarkably similar to our sun.

But at about 1,000 light-years away, Kepler 452 b is far too faint for easy follow-up studies. Its apparent existence is based solely on data gathered during Kepler’s primary mission, which ran from 2009 to 2013 before being cut short by equipment malfunctions. During this period the spacecraft stared continuously at a single patch of sky, waiting for any of the stars there to almost imperceptibly dim from the shadows of planets passing across their faces. Such “transits” are how Kepler found the vast majority of its planets; but many things besides planets can cause stars to slightly dim, leading to far more false alarms than discoveries of new worlds. For any candidate planet to be confirmed as genuine, it would have to be observed transiting at least three times. Due to its long orbital period, Kepler 452 b barely met that minimal criterion before the telescope’s primary mission ended—but a host of other, more technical tests convinced the Kepler team the planet had a 99 percent chance of being real.

In a new analysis, reported last month in a paper accepted to The Astrophysical Journal, researchers take an extremely statistical approach to considering any given planetary candidate—averaging out errors from the entire span of the Kepler mission, and from every instrument in aggregate. In the process, they say they learned to better distinguish a true planetary signal from astrophysical false alarms or instrumental noise. Armed with a deeper understanding of Kepler’s quirks, the astronomers argue they can more easily flag where and how the spacecraft’s minor defects could compromise data. The authors used this approach to re-vet Kepler’s data from more than 100,000 stars, hoping to find ways to more rapidly confirm strong planetary candidates and boost the odds of validating borderline ones. Their bulk reanalysis showed worlds with orbits of less than 200 days were very easy to confirm, because these planets’ transits repeated enough to display a clear trend outside any background instrumental or astrophysical noise. But the authors found that confirming small, relatively Earth-size planets with longer periods proved tougher due to the premature end of Kepler’s main mission. With this in mind, they set their sights on one of the most marginal targets under these constraints: Kepler 452 b.

“The reason that my co-authors and I focused on Kepler 452 b is because it’s the longest period, [and the] weakest signal that’s been confirmed so far,” says study co-author Jeff Coughlin, a SETI Institute scientist who works on the Kepler mission.

Coughlin and colleagues ran the data for Kepler 452 b through several new simulations under their refined signal-to-noise-ratio threshold, paying particular attention to possible contamination from minor flaws in the spacecraft’s instruments. They found the planet has as high as 92 percent and as low as 16 percent chance of being real. “There’s no evidence that it’s not a planet,” Coughlin says. “But it’s not at 99 percent.” (He adds, however, that he personally believes there’s a “more than 50 percent chance” it is real.)

Some candidates can be checked further using another technique that looks for “wobbles” in the star caused by the gravitational tug of an orbiting body, but Kepler 452 b is too distant and small for that. The great hope in saving it as a candidate is the Hubble Space Telescope, which could be used to watch for the putative planet’s next transit, expected later this month. “We know when the next transit should happen, to within a margin of error of a couple hours,” says Natalie Batalha, Kepler’s former co-investigator and mission scientist and a co-author on the Kepler 452 b discovery paper.

Batalha contends Coughlin and his co-authors are merely highlighting an all-too-common and already well-known problem with many exoplanet claims: Studies often only account for astrophysical phenomena, not problems with instruments that might degrade data quality. But she also claims there are a few potential problems with Coughlin and colleagues’ study—particularly its statistical (rather than case by case) approach to analyzing Kepler data. In averaging out what is and is not a good candidate signal, she says, the study overlooks the factors that made Kepler 452 b so promising in the first place.

Batalha says the method used by Coughlin and colleagues does not take into account Kepler’s actual problem areas, which were discovered and defined through years of painstaking work. For instance, some parts of its detectors functioned better than others. The Kepler instrument has 21 segmented detectors—each mapped to a certain star-filled region of sky—and the segment used to detect Kepler 452 b was known for producing crisper and clearer data than other segments that proved to be problematic, she says. “When you compute the average reliability for the entire catalogue of Kepler’s planet candidates, you’re averaging over all of the detectors,” she says. This does not account for the actual strengths of the Kepler 452 b signal, artificially driving down the confidence of the detection. The initial detection actually has a one-in-3,000 chance of being in error, she adds. “You want to use the best possible processing of the data, and the discovery paper has a slightly cleaner set of data.”

Asked for comment, a NASA spokesperson called doubts about Kepler 452 b “an example of the scientific method at work and the way in which science is an evolving process as new information challenges us to revise our thinking and fine-tune our hypotheses. We welcome debate as it leads to our continued growth in knowledge in exoplanet science and in other fields of astronomy.”

In an e-mail sent to Kepler scientists who work on the project, representatives from NASA’s Exoplanet Science Institute (who manage the institute’s “Exoplanet Archive” catalogue) wrote Kepler 452 b should keep its status as a planet for now: “This new paper has not definitively shown that the Kepler 452 b signal is instrumental in origin. While the external reliability remains high, the internal reliability is lower than what is claimed in the original 2015 paper. For now, the Exoplanet Archive has opted to retain Kepler 452 b within the Confirmed Planets table until a more definitive refutation is published.”

The only instrument functioning right now that could possibly settle this debate—and Coughlin says it will “still be a challenge”—is Hubble. But Kepler 452 b’s next expected transit is on April 18—too soon for astronomers to obtain a prized time slot on the space observatory’s packed schedule. The next transit (if the planet is really there) would not occur until May 8, 2019. Until then the planet’s true status will remain uncertain—and astronomers will likely continue second-guessing a handful of other Kepler finds.