Smoking Gun No Longer Smoking

The BICEP2 paper is now out in Physical Review Letters, with major revisions to its conclusions from the preprint/press conference version of last March. For another sort of associated revision, compare this (from a March 17 Stanford press release):

Linde, now a professor of physics at Stanford, could not hide his excitement about the news. “These results are a smoking gun for inflation, because alternative theories do not predict such a signal,” he said. “This is something I have been hoping to see for 30 years.”

to this (from an interview with Linde in the latest New Scientist):

I don’t like the way gravitational waves are being treated as a smoking gun.

If we found no gravitational waves, it wouldn’t mean inflation is wrong. In many versions of the theory, the amplitude of the gravitational waves is miserably small, so they would not be detectable.

Last month, Resonaances broke the news that there was a problem with the BICEP2 claims, specifically with the bottom line (and punch line) of their preprint abstract:

Subtracting the best available estimate for foreground dust modifies the likelihood slightly so that r=0 is disfavored at 5.9σ.

Back then the BICEP official reaction to the Resonaances claim that they were admitting to a mistake was “We’ve done no such thing.” Post-refereeing, there have been extensive changes in the paper (for example, the “DDM2” dust model based on scraped Planck data is gone), and the bottom line of the abstract has been changed to:

Accounting for the contribution of foreground, dust will shift this value [non-zero r at 7.0 sigma] downward by an amount which will be better constrained with upcoming data sets.

If the BICEP collaboration is still not admitting a mistake in their treatment of Planck data or the bottom line of their preprint, then it seems that referees have told them they can’t publish these in PRL.

Back in March the BICEP2 results made the front page of the New York Times with a Dennis Overbye story Space Ripples Reveal Big Bang’s Smoking Gun, but today the NYT has Astronomers Hedge on Big Bang Detection Claim, which explains well what has been going on.

Update: Nature has a story out about this, which includes the news of a recent presentation at a Moscow cosmology conference by Jean-Loup Puget of the Planck collaboration:

Using for the first time the newest Planck maps available, Puget and his collaborators have directly examined the polarization of dust in these high galactic regions rather than extrapolating from dustier regions in the plane of the Milky Way. Averaging over some 350 high-galactic-latitude patches of sky similar in size to the region observed by BICEP2, Puget reported that polarization from interstellar dust grains plays a significant role and might account for much of the BICEP2 signal that had been attributed to inflation-generated gravitational waves. Puget told Nature that an article detailing these findings would be published in about six weeks.

Update: I’ve been watching Paul Steinhardt’s talk at Strings 2014, where he’s giving a dramatic attack on the way inflationary cosmology is being pursued as in violation of the scientific method. One thing he does is put up exactly the Linde quotes from this posting.

This entry was posted in Uncategorized. Bookmark the permalink.

47 Responses to Smoking Gun No Longer Smoking

  1. It’s a good thing PRL decided to make this available for free, but what happened to their length limit? I recall that the maximum accepted size was four. At the 25 page mark, this paper has a War and Peace size by PRL standards. Does this mean that anyone can now submit papers up to 25 pages to PRL? Somehow I strongly doubt it.

  2. Dave says:

    I’ve heard of PRL trying to lure papers from other journals by relaxing the page limit, but that was for six or seven pages, not 20+. Maybe the preprint was intended for ApJL and PRL managed to coax the authors away.

  3. Bernd says:

    There’s an editorial explaining the unusual format:

  4. Jeff M says:

    Pardon the stupid question, but if they’ve removed the scrubbed Planck map, how exactly do they claim to account for dust?

  5. Peter Woit says:

    Jeff M,

    The preliminary Planck data they got off a slide was just one of several things they were using to estimate the effects of dust. My non-expert understanding is that how they used that has been challenged, other ways they were estimating dust are not completely convincing, and since the time of the preprint Planck released more data about dust which might raise their estimates. So, they’re stuck with no solid statement about dust, just that the effect of dust is “an amount which will be better constrained with upcoming data sets.”

    By the way, in the interview, Linde refers to a rumor that Planck will release relevant results about dust sooner than their full polarization results (which are planned for October), possibly this summer, “any time now”.

  6. srp says:

    I find this type of dispute refreshing. It’s about the quality of observations and interpretation of data–BB relic or interfering signal from relatively nearby dust? More data from different instruments and some careful thinking will sort out which of these is more likely in a few years at the outside.

    One interesting sidelight on this issue is the question of what Harry Collins calls “methodological individualism versus methodological collectivism.” The former stance holds that each individual paper and announcement should be a definitive claim of a finding one way or the other, complete in itself. The latter stance holds that each paper is just one piece of equivocal evidence and overall conclusions should be derived by the community after seeing many such papers over time. The advantage of methodological individualism is that researchers are forced to think hard about what they actually believe and do a lot of work to make their results credible, thereby incentivizing quality work and avoiding publication bias of spurious weak but confirming signals . The advantages of methodological collectivism are that weak individual signals may be collectively strong and that other researchers are less likely to toss out intriguing but less-than-definitive findings if they see others are seeing similarly weak-but-suggestive things.

  7. anonymous says:

    Come on, Peter. The vast majority of the paper is the same as before, about as much as you’d expect following a typical peer review. The evidence against synchrotron and synchrotron-correlated dust has, if anything, strengthened. The caveats about the rest of dust were already in the penumbra of the original paper. For all that you decry sensationalist headlines, you have been the pot calling the kettle black for the last few months. If the rumored Planck paper gets arXiv-dumped soon, you should preliminarily treat it as being as reactionary as the rest of the unreviewed Flauger/Seljak/anyone analysis.

  8. Bob says:

    Those two back-tracking quotes from Linde are quite extraordinary, almost funny if it didn’t reveal such a dreadful way to do science: “If the experimental result is true, the theory is definitely proven. And if the result is false, the theory is still probably correct.” Simply dreadful.

    Inflation: well worth a million dollar prize.

  9. Peter Woit says:

    I don’t think a typical peer review changes the conclusion of the paper from “we’ve discovered something revolutionary at greater than 5 sigma” to “could just be dust”.

    The PRL peer review isn’t that significant in that this change in conclusion had already become widely known, with a lot of public discussion of the problems with the 5 sigma claim and a new consensus that “could be dust” is a significant possibility, not a one in 3.5 million possibility. To see the change in consensus, all one has to do is listen to Linde…

  10. OMF says:

    Bottom line, all I see here is the scientific method in action. Public discussion and scrutiny of the data lead to the most rigorous conclusion.

    The real problem here was the media hyping of preliminary results and the pressures which this placed on the entire investigation and potentially the peer review itself. This kind of modern pressure will end up interfering with future large scale experiments unless methods are found for dealing with it.

  11. Alex says:


    Aren’t these two new quotes never the less correct? My understanding of inflation is that it’s a theory with free functional forms (let alone free parameters) and that even if r came out to be some other value, an inflationary theory could be found to be consistent with that value.

  12. Bob says:

    OMF: it wasn’t the media hyping the initial results, it was the bicep team who did that. That last sentence of the original Bicep preprint will live in physics infamy.

  13. Bob says:

    Alex: It would have been more impressive if Linde had said no gravitational waves is a blow for inflation, but it does not neccessarily kill it. He’s basically saying the result is irrelevant to the theory – unless it supports the theory. That’s poor.

  14. Jeff M says:

    Linde’s comments remind me of why I switched from physics to math way back in the early 80’s. Even back then the parameter space was plenty big enough to accommodate all sorts of experimental results, which drove me nuts. At least from what I remember once you got a result and adjusted the relevant parameters you would then restrict things enough that you got some real predictions which could be checked, and if they didn’t work then the theory would be falsified. As far as I can tell, nowadays you can adjust one parameter to account for an experimental result without limiting the value of any other parameters at all, there will always be an infinite number of theories available to fit any data you have.

  15. Peter Woit says:

    Jeff M,
    Linde’s behavior and the whole inherently untestable multiverse approach to fundamental physics that he represents (together with a willingness to say outrageous things to the press…) are very much outliers in physics in general. Few people behave that way. It’s an interesting question what the physics community can do to keep this kind of thing from discrediting more serious parts of the subject.

  16. anon. says:

    anonymous wrote:

    If the rumored Planck paper gets arXiv-dumped soon, you should preliminarily treat it as being as reactionary as the rest of the unreviewed Flauger/Seljak/anyone analysis.

    One important claim that first appeared publicly (as far as I know) in the Mortensen and Seljak paper was that BICEP’s argument that the frequency-dependence of the signal disfavored dust at 2.2 sigma was wrong, because the correct thing to test was not pure dust but dust + lensing. The published version of the BICEP paper now shows a test of dust + lensing and, indeed, the significance has dropped down to 1.7 sigma. The Seljak paper may not be “reviewed” but this part of its argument is validated.

  17. crime says:
    The term “smoking gun” was originally, and is still primarily, a reference to an object or fact that serves as conclusive evidence of a crime

  18. paul hughes says:

    A smoking gun is conclusive proof. The absence of a smoking gun in not conclusive disproof. Maybe that is what Linde was saying, and I actually think from these quotes alone that it was (the absence of a smoking gun does not disprove the theory). To reiterate, it sounds (from these quotes alone) that in the second quote he was saying that, even if the results turn out not to be a smoking gun, the theory is not disproved; and that he is complaining about people misusing his “smoking gun” quote to imply or infer that such absence is disproof. Seems fairly straightforward to me unless one has an axe to grind.

  19. Peter Woit says:

    paul hughes,
    I don’t think there’s anything at all unclear about what Linde is saying, and the comments by Bob above have it right. It’s perfectly consistent, but it also shows clearly that he’s now arguing for a way of pursuing science immune to falsification.

    As for his complaints about people “misusing” what he said back in March, I just don’t think anyone is “misusing” anything. When you’re part of a press release claiming that the discovery of X proves you are right (and possibly getting part of a $1 million prize because of this…), you’re going to regret that press release when it starts to look like the discovery may really have been of “not X”.

  20. Robert Rehbock says:

    “I’m 95 percent convinced it’s true, but extraordinary statements need extraordinary proof. If these results are correct, they are among the most spectacular results in observational cosmology obtained in the 21st century. We should wait a little before they are analyzed and confirmed by other observers.”
    Linde said on March 21st. Your selective quotes are as unwarranted as declaring the discovery is premature.

  21. Peter Woit says:

    Robert rehbock,

    This is about linde’s claims about the significance of the result if it were true, not about any claims concerning the correctness of the result. Of course like everyone he was aware the results might not hold up and needed confirmation.

  22. paddy says:

    I do like the phrase used in the Nature article PW referenced: “…a popular but outlandish theory….”

  23. srp says:

    Not a multiverse fan, but Linde’s logic is in no way anti-scientific. You have a theory you believe is more likely true than not and then someone comes along with a potential piece of clinching evidence that would raise credibility to 100% if it held up. If it doesn’t hold up, that just drives your credibility back to its original more-likely-than-not level. There’s nothing wrong with that–it’s common sense.

  24. Peter Woit says:

    What’s unscientific about this scenario is that you’re claiming that your relation to experiment is a no-lose one. Heads you win, tails you get a do-over. What makes it science is when you can lose. If a significant primordial gravitational wave signal of r around .2 really is a smoking gun for inflation, corresponding to what you expect in the simplest models, with the expected (GUT) scale, then a very small r is evidence against inflation, because you are forced to move away from simplest versions and original scenario to something more complicated. Put differently, if the viability of the idea of inflation is insensitive to r, then r=.2 was not a smoking gun. Linde would like to have it both ways, his predictive gun smoking when it looked like r=.2, not smoking now that maybe r is much less.

  25. Avattoir says:

    “Puget reported that polarization from interstellar dust grains plays a significant role
    and might account for much of the BICEP2 signal … attributed to inflation-generated gravitational waves”

    The key bits of phrasing being: “a significant role” and “might account for much”, in contrast with, say, ‘may explain entirely’, and ‘might account for all’.

    It’s one thing to assert that the BICEP2 team overstated their findings in relation to primordial b-modes; it’s quite another to assert that the Planck satellite team is asserting their own findings utterly negate the former’s findings. The title to this post purports that Planck is promising to gut BICEP2’s findings. In fact, the words used by the Planck representative makes no such promise, directly, indirectly or impliedly, and indeed is not even reasonably interpretable as promising anything of the sort.

    But that doesn’t even get to the half of it, because the phrase “might account for much” means, more precisely, by necessary implication, ‘might account for some of what BICEP2 detected, but not all’. And thus, the question becomes, Even accounting for the full potential of possible signal noise due to dust within this galaxy, what, OTHER THAN PRIMORDIAL B-MODES, would account for the rest?

    Resume dancing on the BICEP2’s coffin while you still can without appearing ridiculous.

  26. Dan C says:

    What’s the difference between the situation here, i.e. parameterizing some cosmological parameters (say, the inflationary Hubble parameter via the tensor-to-scalar ratio), and something like dark matter searches, where one is looking for, say, the mass of some dark particle?

    In both cases one cannot seriously argue that a non-detection, i.e. a bound on the parameters, is not scientific progress because it is not a “win-or-lose” situation.

    They are both win-or-lose situations for a variety of concrete models with some explicit parameters, dynamics, etc. But they are hardly win-or-lose for the general paradigms.

    On the other hand, an actual detection of either of the parameters would be a huge triumph for the theory.

    I think this is all Linde is pointing out, although he can indeed say it somewhat dramatically. BTW, I don’t agree that the GUT scale is “expected” in inflation unless one has a gambling addiction. An enormous range of energy scales can be easily made consistent with observation.

  27. Peter Woit says:

    The title to the post is obviously mainly referring to Linde’s change of story, secondarily to the related change in the BICEP2 paper conclusions. The Planck news was added later as an update. Of course I have no idea what they will show about dust and how much if any of the BICEP2 result will survive.

  28. Stuart says:

    Someone once said “Absence of evidence is not evidence of absence” Perhaps this is what Linde is alluding to.

  29. Tom Cohoe says:

    If it goes through here, it’s a goal. If it doesn’t, it’s not a miss.

    Nice game.

  30. srp says:

    If somebody finds a murder weapon in a wastebasket, that might clinch a theory that X is guilty, but not finding it in the wastebasket might not drive the probability much below the original belief in X’s guilt. Especially if there are thousands of places the weapon could have ended up.

    There are many asymmetric evidentiary situations where the impact of finding something specific is much more consequential than not finding it. Perhaps you could make the reverse argument, though, that even a finding of the polarization might not support inflation, because other things could have caused it. In fact, I heard Roger Penrose on NPR make exactly that point a few weeks ago.

  31. Curious Mayhem says:

    I agree with srp. No evidence for or against inflation right now means that inflation remains a viable, theoretically compelling, but unproven theory, just as it was on March 16.

    Meanwhile, I strongly suspect that the BICEP2 result will be watered down to lower confidence, but not enough to scotch the whole conclusion.

    We do need a way to deal with prepublication / preprint / media hype mania. The media makes para-scientific pronouncements that are continuously pseudo-authoritative. Results like that in science are very important but uncommon.

  32. Peter Woit says:

    Curious Mayhem,
    I don’t see the problem here being the media. The media coverage of BICEP2 has quite accurately reflected what prominent scientists are saying. Back in March the experimenters were holding press conferences to announce a 5 sigma discovery, most experts were saying this looked good (none of them were bringing up the dust problem), and Linde and other prominent theorists were issuing press releases about this being a “smoking gun” for inflation, providing evidence for the multiverse, etc. This got accurately reported.
    More recently, the media coverage has also done a good job of reporting the news that problems were found with the BICEP2 claims. Linde is unhappy that his own “smoking gun” press release claims will make him look bad if the BICEP2 result collapses, but I don’t think he has a legitimate complaint.

  33. John Gribbin says:

    A small but important point. None of this is about the Big Bang. Inflation preceded the BB. That’s the whole point of the idea.

  34. Garbage says:

    Say soon after the BEH mechanism was discovered they’d have gone around saying: ‘Here’s the smoking gun: A fundamental scalar around 10 GeV’ In fact, ‘the Higgs’ itself was a response to the referee asking for such thing. Then after many attempts LEP comes and the Higgs is not there at 115 GeV. ‘No longer smoking?’ No so fast, we have a lot of EWPD that fits the SM very well. We also know there’s a threshold. The LHC was built and ultimately the Higgs found around 125 GeV. Inflation fits all data we have from early universe cosmology, but like the SM with the Higgs, it’s got a few free parameters. –We do not have any alternative to inflation that works– There’s a threshold somewhat similar to the TeV scale for the EW theory. In fact, we know there’s an irreducible level of non-Gaussianity (NG) which is correlated to the size of r and ns, the tensor-to-scalar ratio and scalar tilt. Unfortunately the size of the NG may be too small to be detected in the near future, but nonetheless it is a well defined ultimate prediction of inflation. (There is also the celebrated n_t + r/8 ~ 0 for simple models.) For the case of GWs, simple models do predict a size comparable to 1-ns (there’s this factor of 8 in between). But we can easily have a world where those two are unrelated. In this case the level of GWs may be undetectably small in the near future, this is also related to the relation between r and super-Planckian field range. I’d claim that a very small r would be unnatural, but we are still far from ruling out even the simplest models of inflation… as the case of the Higgs, don’t kill a theory that works, just go ahead and keep on pushing experiments!

  35. philh says:

    Here is an additional attack on the inflation hype but also a defence:

    Attack: even if the BICep 2 signal is confirmed there seems to me a further piece of evidence that is needed. That is the tilt in the primordial gravity wave sepctrum . As I understand there are alternatives to inflation that predict blue tilt for this spectrum , with inflation predicting red tilt. If the spectrum was shown to be blue tilted that might disfavour inflation. Hence measuring the existence of the signal is not enough. I saw very little mention of this in any press coverage.

    Defence: As I understand inflation predicts gravitational waves but does not say exactly what energy scale these will be driven from and what value of r it should be detected at.
    Similarly the hypothesis that life exists on other planets does not tell you which planet it will exist on. If you found what you thought was life on Mars , that would be a smoking gun for the Et hypothesis. If you then found the evidence was shoddy that doesn’t mean your hypothesis was wrong.
    Whilst it was overhype to say BICEP 2 confirmed inflation , ( see my attack above) what I think is true is that the discovery of B mode would significantly improve the evidential picture for inflation over certain other rival models as ekpyrotic , CCC etc. But not finding doesn’t disprove inflation, just as not finding life on Mars doesn’t disprove the ET hypothesis.

  36. Neophyte says:

    Garbage says: “don’t kill a theory that works”

    But not everyone thinks it works, e.g. Penrose. It is a contrived model that unsuccessfully tries to explain a small set of facts about the universe. It creates more problems than it tries to solve. Why bother with it?

    I agree “keep on pushing experiments”.

  37. Peter Woit says:

    If back in 1964 after the Anderson-Higgs mechanism was discovered, say Brout and Englert had issued a press release claiming that some recent possible bump at 10 GeV was a “smoking gun” for the Higgs particle, people would have thought they were probably nuts. If, several months later after the bump started to look dubious they had gone to the press to complain that they had been taken seriously earlier, all doubt about their craziness would have been removed.

  38. Curious Mayhem says:

    Inflation is emphatically not a contrived theory. It was designed originally to solve one problem (topological relics) and ended up solving three, and quite elegantly. Penrose is simply wrong about this.

    Penrose is correct that inflation leaves unanswered the cosmic entropy questions that he and others rightly think are important. But before inflation, no one had an answer for those either. Actually, the questions had never been posed that way before the 1980s, although there were Dirac and others with “large cosmic numbers,” an earlier version of this line of thought.

    Peter, you’re right that, on BICEP2, the media has been fairly responsible in its own right, and its hype in this instance is mainly a reflection of hype from prominent scientists. But that doesn’t mean the media hasn’t engaged in its own hype in other cases.

    The pre-publication mania that I referred to has both components, from prominent scientists and from the media itself, and it’s an issue that scientists, journals, and the media all need to recognize and contain.

  39. Curious Mayhem says:

    And if the gun is smoking, what is the gun smoking anyway? 🙂

  40. Garbage says:

    If the bump was consistent with a fundamental scalar particle, nobody would have thought they were nuts. I want to believe people would have been really excited! If after the bump had disappeared in the ‘dust’ they would have said “well there’s still a lot of room for the Higgs to show up,” they’d be absolutely correct, and time would prove them right. You’re purposely ignoring the point. The Higgs particle is a smoking gun of the mechanism, the same way GWs are. But there isn’t a unique possibility for the scale at which it decides to show up. (Some may even argue that the fact that m_h ~ m_Z is a score for SUSY.) Given a set of initial conditions, Inflation *predicts* a scale invariant power spectrum with a small tilt. The simplest models produce negligible non-Gaussianity, as observed by Planck. These models would give r~0.1 . To that extent it is an extremely interesting time for cosmology. However, there are many simple extensions which lower r and yet give basically the same fit with data. In fact, String-inspired models often “predict” lower values. Until we don’t cross the non-Gaussianity threshold (the irreducible level predicted by inflation) we will never be able to tell for sure.

    Speaking about quotes. Steinhardt has changed his discourse so many times after a bound on fnl^loc comes about that Linde’s comments are innocuous. Like he said: “[Inflation] is the worst form of [theory] except all the others that have been tried.”
    There are so many problems with Steinhardt’s proposal there’s no point in arguing. Like a famous mathematician once told me, “as long as you keep using ‘~’ signs you’re good”

  41. Simeon Every says:

    The problem isn’t that stringy/multiversey theories are inconsistent or incompatible with other scientific theories. The problem is that there seems to be no limit to how many experiments must be done to establish which particular stringy/multiversey theory accounts for the existence of the observed universe (if indeed it is even possible to select one particular theory based on observing only the observed universe) and no experiment that can be done that indicates that no stringy/multiversey theories correctly account for the existence of the observed universe. If the latter is the case, how will you ever find out?

  42. Peter Woit says:

    Back in 1964 postulating a Higgs particle explained nothing about anything (the use of the Higgs mechanism to provide an electroweak theory was years in the future), which is why anyone invoking “smoking gun” evidence for it would not have been taken seriously.

    I strongly recommend looking at Steinhardt’s Strings 2014 talk, where he discusses the Linde quotes and explains in detail the obvious problem with them.

  43. Garbage says:

    ‘Model of Leptons’ – 1967. Although it is true the original motivation for BEH is lost in time (same as the YM theory) I think is pretty clear what we’re talking about unless you want to divert the point. The original motivation for Inflation (horizon problem, monopole, etc.) is also lost in time if you want; but the theory survives, like the Weinberg model does, with quantitative predictions. (Here I think Mukhanov deserves a lot of credit by the way.) What people fails to realize is that there’s a very natural connection between an accelerated expansion and quasi-scale invariance, the fact that many models fit data is because of the robustness of this simple (yet powerful) idea.

    I am pretty sure if a Higgs-looking resonance had shown up in the 70’s (or even 80’s) around 10GeV a lot of people would have gotten really excited, and that’s 40 (30) years before they found the Higgs.

    I’ve seen Steinhardt in action many times, making the same points over and over. I also suggest you go back and see his own claims about his model and how he changes his mind all the time. I also wish we would understand how to make sense of the big picture with inflation, but given what we observe, there’s nothing like it, it is just too simple and idea that works extremely well… I think you’re confusing this with the multiverse, the measure, etc. I’m not even touching upon things I cannot measure/prove. But please, learn a little bit cosmology and you’ll see how you’ll be yourself driven towards inflation like the most compelling explanation.

  44. physics PhD (experimental optics) says:

    In fairness to Linde, it seems that his press release was driven by the Stanford media relations office. As is typical in press releases, “quotes” are often written / scripted / massaged by media flack, and quickly run by the “speaker.”

    Had Linde been fortunate enough to work at a publicly traded company, his words would at least be reviewed by a lawyer, to protect the company from any liability or blow-back.

    Yes, he needs to stand by his words, and eat them if necessary. But overactive University PR organizations are also part of the problem.

  45. Jesper says:

    Peter – do you know where one can watch the Paul Steinhardt’s talk you mention? I’ve looked for it but only found his slides.

  46. Peter Woit says:

    I watched the talk in real-time via the streaming site. Presumably video of the talks is being processed and will show up later.

  47. S. Molnar says:

    It’s old news by now, but there is an interview with David Spergel about this in the Simons Foundation magazine:

Comments are closed.