When the news of the BICEP2 result broke, the mood was euphoric. There was open speculation about Nobel Prizes, a certain video went viral and cosmologists spoke of a radical transformation in our understanding of the early universe. And all of this may still come to pass. But the wave of doubt blowing through the cosmological community in the last week is growing into a consensus that the BICEP2 team has overstated the case for a discovery. (For recent coverage see Science MagazineScientific American and the Washington Post.) 

BICEP2 squeezes information from the cosmic microwave background, light that has been flying across the universe since just after the Big Bang, forming the backdrop against against which we view the rest of the cosmos. Isolating and removing other sources of microwaves -- so-called "foregrounds" -- is a key challenge for all observers of the microwave sky.

Crucially, BICEP2 measures both the intensity and polarisation of the microwave sky, and the polarisation of foregrounds is poorly understood. After two months of scrutiny, there is a realisation that the BICEP2 team may have underestimated the polarisation contribution from the dust in our galaxy and, in addition, was overly optimistic about the quality of their foreground estimate. Consequently, it seems that the BICEP2 team cannot confidently claim to have detected a "primordial B mode", the signature of gravitational waves and the hallmark of an inflationary era in the early universe.

A further twist is that the BICEP2 foreground analysis relied on unreleased data from the Planck satellite, "scraped" from preliminary maps shown at conferences and preserved in PDFs of the speakers' slides. My initial reaction to this was fairly tolerant, but it really was not (and never will be) a good idea. Specifically, scraping itself adds uncertainty, and this uncertainty was not quantified or accounted for by the BICEP2 team. Moreover, preliminary data is preliminary: the measured value shown at any point in a map of the microwave sky is effectively the difference between two much larger and almost equal numbers, but if either of these quantities changes slightly (as the calibration of the instrument improves, for instance), their difference can change substantially -- another source of uncertainty that the BICEP2 error budget ignored. Finally, the data-scraping detracts from the BICEP2 observations themselves, which represent a dramatic advance in our ability to measure the microwave sky. Astrophysics is a poster-child for the Open Science movement, but this moment is a reminder of the distinction between open science and other people's unfinished work.

So where does this leave us? These developments do not rule out the possibility of a large gravitational wave background, and the data needed to understand the foregrounds is currently being gathered and analysed. If our universe does turn out to contain a significant gravitational wave background, BICEP2 will undeniably have participated in its discovery. Conversely, if the signal seen by BICEP2 is shown to consist entirely of foreground, that heartwarming viral video may come to look like an episode of Candid Camera in which not even the host was in on the joke. And for now, cosmologists are living in interesting times.