I noticed today that BBC News has a story headlined ‘Multiverse’ theory suggested by microwave background that assures us that:
The idea that other universes – as well as our own – lie within “bubbles” of space and time has received a boost.
After taking a look at the PRL and PRD papers that are behind this, it’s clear that a more accurate title for the story would have been “‘Multiverse’ theory suggested by microwave background – NOT”. As usual, the source of the problem here is a misleading university press release, one from University College London entitled First observational test of the ‘multiverse’. Somehow the press release neglected to mention something one might think was an important detail, the fact that this “First observational test” had a null result.
It’s well-known that one can find Stephen Hawking’s initials, and just about any other pattern one can think of somewhere in the CMB data. The authors of the PRL and PRD papers first put out preprints last December (see here and here). In these preprints they essentially claimed to have found four specific features in the CMB where the hypothesis that they were due to bubble collisions was statistically preferred. A guest post by Matthew Johnson at Cosmic Variance explained more about the preprints. I didn’t understand their statistical measure, so asked about it in the comment section, where Matthew explained that, by more conventional measure, the statistical significance was “near 3 sigma“.
It turns out that the PRL and PRD papers differ significantly from the preprint versions. In the acknowledgements section of the PRD paper we read that:
A preprint version of this paper presented only evidence ratios confined to patches. We thank an anonymous referee who encouraged us to develop this algorithm into a full-sky formalism.
and the result of the new analysis asked for by the referee is summarized in the conclusion of the paper:
The posterior evaluated using the WMAP 7-year data is maximized at Ns = 0 [Ns is the average number of observable bubble collisions over the full sky], and constrains Ns < 1.6 at 68% confidence. We therefore conclude that this data set does not favor the bubble collision hypothesis for any value of Ns. In light of this null detection, comparing with the simulated bubble collisions... [various bounds ensue]
So, the bottom line is that they see nothing, but a press release has been issued about how wonderful it is that they have looked for evidence of a Multiverse, without mentioning that they found nothing. As one would expect, this kind of behavior leads to BBC stories about how the Multiverse has “received a boost”, exactly the opposite of what the scientific evidence shows.
Update: The FQXI web-site has an article about this. In it, the authors seem far more interested in promoting their PRL paper as “first test of the multiverse” than in acknowledging that a referee made them do a better test of the idea and they got a null result. There’s no mention of the null result in the article.
Update: News stories based on this keep on coming. The latest: Proof of a multiverse discovered?