A few months ago the world was stunned by an announcement from Harvard University of what seemed to be the biggest science news of the century: the BICEP2 experiment proved that right after the big bang our universe underwent a rapid phase of cosmic inflation. This finding became front-page news all around the world. Now it seems clear that this statement cannot be upheld, as the European Planck satellite has published a paper indicating that the authors significantly underestimated contributions from dust in our Milky Way. A new science scandal? Not really, but some lessons can be learned.
So, what has happened? BICEP2 measured very accurately the properties of radio signals from the primordial fireball of our universe generated 380,000 years after the big bang. In particular, they measured the polarization – the direction of oscillation of the radio waves – of this cosmic background radiation with unprecedented precision. The observed pattern of polarization on the sky, so called B-modes, seemed to fit predictions made for the inflationary model. In this model the rapid expansion of our baby universe would produce ripples in spacetime that propagate as gravitational waves through the cosmos and leave their imprint in the fireball radiation.
The detection of the cosmic background radiation had already generated two Nobel prizes (1,2), so the third one was now just a question of time. The press officers of the participating institution did their job and made sure this message would not go unnoticed. Rumors spread conveniently before the press conference via Facebook and Twitter and a heart-wrenching youtube video was posted showing how the good news was delivered to Stanford’s Andrei Linde, one of the fathers of the inflationary model.
The news was almost instantly accepted as fact. This worried me a little bit, given that there are so many difficult steps from the actual experiment to the final conclusion. In a previous blog I discussed these worries and the role Facebook and Twitter play nowadays in science communication. Moreover, everyone – including myself – was expecting that inflation was true anyway, which made it easy to embrace the results right away. However, as the scientific misconduct cases of Schön and Stapel have demonstrated, wishful thinking is where things can go wrong most easily in science.
Moreover, for some the finding even seemed to be some kind of an epiphany, delivering the long-sought support of the multiverse theory. After all, if the universe can spontaneously self-inflate, this might have happened countless times already with other universes elsewhere.
Indeed, Andrei Linde himself likes to motivate his scientific talks with the ultimate divine question: Our present universe shows an incredible level of fine-tuning – how can that be? Either, he says, the universe was created by God, or, there is a countless number of different universes spawned by his inflationary model – and God is not necessary. Unfortunately, it is not entirely clear whether that causal connection is correct and whether God agrees with it but, nonetheless, the bar is set very high.
So, at the peak of the hysteria, this was the point where it was time to grab a cup of tea, wait, and let the scientific method take its course. Indeed, doubts surfaced whether the authors of the study, had properly quantified all other – non-cosmological – effects that could have contributed to their measured signal. The community noticed that the paper in question had not yet been peer-reviewed by a major journal. In addition, the BICEP2 team was anticipating results of a European satellite project, Planck, that could scoop them. In fact, the Planck collaboration was in possession of much better data on emission from dust in our Milky Way, that could potentially confuse the cosmic background signal. Hence, the BICEP2 collaboration, not yet having access to the data, scanned a preliminary dust map from a conference presentation of Planck and used this to estimate the magnitude of the effect – not a good idea.
Quickly, Flauger, Hill & Spergel showed that the dust contribution was so uncertain, that the BICEP2 results could be insignificant. And indeed, in the final published version of the BICEP2 paper the claim for detection of inflation was softened. Now it was still up to Planck to determine how strong the dust contribution really was. As it seems now, the dust contribution is in fact much higher than estimated by BICEP2. Hence, at the end of the day there is very little left to support the strong the claim that inflation has been discovered.
So, what does that mean? Well, the experimental finding of BICEP2 is not questioned, however, rather than looking at the face of God – as Nobelprize winner G. Smoot called it – BICEP2 might have just looked at the dusty outskirts of our own Milky Way. Cosmic inflation might still have happened, but currently there is no experimental evidence for it.
Planck and BICEP2 now work on a joint paper, trying to combine their data, so maybe, with a more careful analysis a cosmological signal can still be found – or not.
What are the consequences now? Does one need to retract the BICEP2 paper? Well, the published version was a bit more careful and acknowledged in the end, that dust could spoil their results. So, factually there is probably no ground to retract the paper itself. Will the authors be in trouble? Not necessarily, but they’ll probably have some interesting discussion ahead of them. At least the paper is already quoted close to 700 times within half a year. Hence, formally this will contribute positively to their citation records. On the other hand, most expert committees will know how to put this in perspective. Is the scientific method in danger? Actually also not – peer review via journals, conferences, and twitter has worked remarkably well in this case. Has the BICEP2 team jumped the gun? Absolutely! The “only thing” that was blatantly wrong in this story was the press release (and a few other bits of the media campaign that were a bit over the top).
For scientists it remains a difficult question when to publish results. Science is after all a competitive field. Most scientists seek truth, but if you are too late, someone else may get the fame and the precious grants. It would be naive to ignore these factors.
So, what can we learn from this? Are press releases bad? Not necessarily – they are good entertainment, but they are not always good science.
There is no reason to reject the entire body of scientific knowledge, but the next time you see a press release about a groundbreaking discovery, sit back, take a cup of tea, and exercise some random acts of skepticism – it will sort itself out.