On March 17 Paul Steinhardt, a physicist at Princeton University, abandoned a theory he’d been championing for more than a decade. Known as the “ekpyrotic universe” model, it was an alternative to the prevailing theory of inflation, which says the cosmos expanded faster than the speed of light in the first fraction of a fraction of a second of the big bang. If inflation is true, then the process should have released a burst of gravitational waves; in Steinhardt’s model, they shouldn’t exist. On that day in March a team of observers announced at a major press conference at the Harvard–Smithsonian Center for Astrophysics that they had indeed detected the waves, thus providing the first clear look at the universe’s earliest moments. The announcement made a huge splash. “Space Ripples Reveal Big Bang’s Smoking Gun,” trumpeted The New York Times front page. “Discovery Bolsters Big Bang Theory,” proclaimed The Wall Street Journal. Dozens of similar headlines appeared, seemingly everywhere. Steinhardt promptly pronounced his theory dead.

But now he’s not so sure. “The situation,” Steinhardt says, “has changed.” Right from the moment results from the BICEP2 microwave telescope at the South Pole were released, many cosmologists had a sense that the discovery rested on shaky ground. “I think it’s fair to say,” argues William Jones, a physicist also at Princeton, “that the claims struck a lot of people, myself included, as far overreaching what the data can support.” Charles Bennett, a physicist and astronomer at Johns Hopkins University who led research on the Wilkinson Microwave Anisotropy Probe (WMAP) satellite, agrees. “Several of the plots in their paper looked odd to me,” he says.

In the ensuing two months, the doubts have only grown stronger, as physicists have attempted, and failed, to reproduce the BICEP2 team’s calculations—admittedly, without the original data, which the team hasn’t yet provided, and without the “systematics” paper, laying out the possible sources of error, which the team has promised but not yet completed. The paper describing the results themselves has not yet been published by a peer-reviewed journal, although that process is underway.

Growing doubts in the astronomical community, meanwhile, have been raised, first in private and over e-mail, then in a blog post by physicist Adam Falkowski, of the French National Center for Scientific Research, in Paris, and most recently by articles in The Washington Post, New Scientist, Science News and other outlets.

If the detection of primordial gravitational waves weren’t such an enormous deal in the first place, nobody would be making such a fuss. But the BICEP2 results are crucial to verifying inflation, a cornerstone of modern cosmology. The theory was first proposed back in the early 1980s as a solution to a number of cosmological puzzles. One is the fact that the universe appears the same in all directions even though the opposite sides of the visible cosmos could never, under normal circumstances, have been in contact with one another, even at the very beginning. Another is that the universe appears to be flat—two parallel lines won’t touch even if they traverse the entire cosmos. Inflation explained all of these puzzles by positing an episode of superfast expansion long before the cosmos was a billionth of a billionth of a second old.

At first inflation was purely theoretical (although University of California, Santa Cruz, physicist Joel R. Primack said from the beginning that “no theory as beautiful as this has ever been wrong before”). A number of measurements—especially the maps of temperature differences in the cosmic microwave background (CMB) radiation left over from the big bang—have bolstered the theory, but the detection of the gravitational waves triggered by inflation would be an especially powerful validation.

That’s what BICEP2 evidently saw—not the waves themselves, but their imprint on the CMB. Gravitational waves from inflation would subtly twist light the microwaves, creating an effect called “B-mode polarization.” But the signal can also be caused by microwaves bouncing off dust in the Milky Way or by galactic magnetic fields. And according to the critics, the BICEP2 team didn’t convincingly disentangle all these effects. It’s clear they saw something, says Raphael Flauger, a physicist at the Institute for Advanced Study, who did an independent reanalysis of the BICEP2 data. “But it’s hard to assess,” he says, “how much of the signal is due to foregrounds and how much of it—or in fact whether any of it—is from inflation.”

One of the strongest critiques has to do with possible dust contamination. BICEP2 can observe the sky in only one wavelength of microwave radiation, which makes it harder for the researchers to rule out this source of confusion. So the team relied on a map of dust concentrations from the European Space Agency’s Planck satellite, which is also making a map of the CMB. The Planck data haven’t been released, however: the BICEP2 team extracted the information from a pdf file  of a slide flashed at a conference. The slide reflected observations that are likely to be incorporated in a future scientific paper, but that information is very different from raw data that can be plugged into a formal analysis. It’s also different from a final map of dust concentrations that might ultimately be released by the Planck team. “It’s all we had to work with,” says BICEP2 principal investigator John Kovac, from Harvard University, by way of defense. But Princeton’s David  Spergel, who led the analysis of data from the WMAP satellite, calls the strategy “weird. It’s an unusual and risky thing to do, because the slide was not intended for that purpose.”

Spergel also says the BICEP2 team evidently failed to factor in contamination from the cosmic infrared background radiation that comes from distant, dusty galaxies. “When you do that,” Spergel says, “it’s probably enough to account for the entire signal they’re seeing. We pointed that out to them within a week of the announcement. They said they’d look into it.” He says he hasn’t heard anything since. He plans to submit a paper within a week based on Flauger’s analysis. “We’ll argue that they made a mistake.”

But Kovac rejects that claim. “The result we reported came after very careful analysis. We described the uncertainties [inherent in using the Planck slide], and it’s important to remember that this was just one of six models we used to characterize the contribution from dust. We certainly stand by our results.”

It’s important to emphasize that nobody is saying that the telescope or the scientific team are anything less than top-notch, or even that the results are necessarily wrong. But there’s widespread agreement that the results have been overstated. “I just think they got excited and overinterpreted their data,” Spergel says. For his part, Bennett says, “I wouldn’t have held a big press conference about it.”

The controversy won’t last long in any case. By early fall, the Planck team will be publishing their own results about dust polarization, and at least 10 other groups are working on polarization experiments from the ground and with balloon-borne instruments. “If the signal is there,” Spergel says, “we’ll know for sure within two or three years.”

As for the vigorous dispute over how reliable the BICEP2 analysis really is, people on both sides of the argument agree that this is how science is done—albeit usually in a less public fashion. “This,” Spergel says, “is how the sausage is made.”

*Editor's Note (6/6/14): This article was edited after posting. It originally used the term "gravity waves" rather than "gravitational waves."