The Last Refuge of Cowards

The talks and panel discussions from the 2018 Breakthrough Prize symposium are available via Facebook video. They ended with the following, from prize winner David Spergel:

Well, alright, I’m going to say something that I probably shouldn’t say in Palo Alto. I don’t think the multiverse is a testable and interesting scientific hypothesis. I think it doesn’t explain anything.

The way the multiverse tends to be used is together with the anthropic principle. The idea is that the universe is the way it is because that’s the way we get to live in it. I find the multiverse solutions to these problems, it’s a lot like if you ask me “why am I wearing a black shirt today”. My answer would be: “you wouldn’t have asked the question if I wasn’t wearing a black shirt”. That’s not a satisfactory answer.

The way we have advanced in science is by falsifiability. By developing hypotheses, testing them (that’s why we do experiments) and ruling things out.

Ideas that are not testable, it’s interesting metaphysics, perhaps interesting for philosophers. What has driven four hundred years of scientific progress is the fact that ideas can be wrong. And, the multiverse, I think is kind of the last refuge of cowards… [nervous laughter from the audience] That’s why it’s great to have tenure.

This entry was posted in Multiverse Mania. Bookmark the permalink.

37 Responses to The Last Refuge of Cowards

  1. Petero says:

    This sentence of the quote looks a bit mangled:
    By developing hypotheses, testing them, that’s what why we do experiments, and ruling things out

  2. Peter Woit says:


    Thanks, fixed.

  3. paddy says:

    David Spergel: Bravo! Especially for avoiding hair-splitting and angels-on-the-head-of-a-pin stuff over falsifiability. We all know it when we see it.

  4. Pierre says:

    It’s not even philosophy or metaphysics; that’s an ignorant slur.

    Philosophers have usually engaged in metaphysics in order to ground moral arguments. That’s a hell of a lot more honest, interesting, and better motivated than the tautological sleight-of-hand upon which the multiverse rests.

  5. Paul says:

    Am I correct in thinking that the Multiverse hypothesis is partly based on geometries of the String Theory?
    Would the proponents of Multiverse base their pro-position on the belief that String Theory is a proven theory?

  6. Cookie says:


    Can one in principle find the correct string theory vacuum if you perform Planck scale experiments? Aren’t there any generic string theory predictions you can in principle test if you can perform those ultra high energy experiments, like extra dimensions or string-like or Brane-like behavior? In principle at least?

  7. Peter Woit says:


    All, enough about string theory. Spergel sensibly thought it not worth even mentioning.

  8. Cookie says:

    Why not? If there’s a reference you can refer me to, that would be much appreciated! I’ve always thought that you could in principle test things with Planck scale experiments.

  9. Casey Leedom says:

    Man, I am so sorry that I didn’t go to see the award ceremony now. I’ve got to get on Stanford’s Physics Seminar email list …

  10. Casey Leedom says:

    By the way, for those of you who want to listen to the comment that Peter quoted, it’s in the “panel discussions” link at 1:25:56 …

  11. Peter Woit says:

    Casey Leedom,
    You can’t get into the actual awards ceremony unless you are a Hollywood starlet or one of our Silicon Valley overlords. The symposium I gather was open to the public.

  12. Alan says:

    Spergel does say about learning from “string theory colleagues” after 1h 15 minutes in relation to duality (linking the small and the large). Yet string theory implies a multiverse? All those vacua? His mention of the anthropic principle and his shirt is just the weak anthropic principle so I didn’t think that was very deep.
    I also think it was David Bohm who said it took 2000 years of theoretical and experimental content before Democritus’ idea of atoms got confirmed. So more time needed for the multiverse?

  13. Peter Woit says:

    I think Spergel is well-informed about the question of testability of multiverse models, not just some naive guy who isn’t happy with anthropics.

  14. BCnrd says:


    The fact that some things took 2000 years doesn’t carry as much weight as it may seem to, since it isn’t as if during most of that time there was a vibrant scientific community working on the task (though there was some scientific work happening during some periods outside Europe). It’s sort of like when Gauss solved a 2000-year-old problem at age 19 by figuring out which regular polygons can be constructed by a straightedge and compass: definitely a landmark accomplishment, but in the intervening 2000 years between Gauss and the ancient Greeks there wasn’t exactly a vibrant community of mathematicians working on that problem, what with the Dark Ages and the absence of an adequate algebraic language and so on.

    And with the size of the community of scientists that has developed during the 20th century up to today, as well as the funding mechanisms (such as they are…) one can reasonably expect rates of progress to be way faster than anything during earlier centuries and epochs (though coming up with good new ideas remains as challenging as always).

  15. Philip Gibbs says:

    Why does he think that the universe owes him an answer that would satisfy him? Some people don’t like the uncertainty of quantum mechanics because it does not satisfy them, but the universe stubbornly refuses to change its rules.

    As for falsfiability, when someone has a complete theory of quantum gravity that is falsifiable and not falsified using current technology, then they can level that criticism. Meanwhile those who think multiverses of various sorts could be part of how the universe might work, probably won’t stop thinking about it just because other people tell them that they don’t like it for philosophical reasons.

  16. Peter Woit says:

    Philip Gibbs,
    Where do you see him saying that? Spergel is one of the best cosmologists in the world, expert on the issue of what you can test experimentally about cosmology and how to do it. He’s making the uncontroversial point that if it can’t be tested experimentally it is not conventional science. His description of theorists who promote their own ideas that cannot be tested by confrontation with experiment as “cowards” (because they can never lose) is a pointed one, but it’s a characterization that many if not most of his colleagues would agree with privately.

  17. tulpoeid says:

    It’s not philosophical reasons, it’s epistemological.
    By the way, comparing Democritus who had to try and derive nature’s law through human senses (which means arriving at a right answer merely by chance) with privileged physicists of the 21st century … should I say more?
    Spergel made my day and I’ll drink a glass to his health. Although it’s depressing to realize that stating the obvious makes headlines nowadays.

  18. Philip Gibbs says:

    Last time I checked epistemology was considered a sub-branch of philosophy.

  19. The Observer says:


    “The fact that some things took 2000 years doesn’t carry as much weight as it may seem to, since it isn’t as if during most of that time there was a vibrant scientific community working on the task (though there was some scientific work happening during some periods outside Europe).”

    There is also an “unlikeliness” that is ignored in choosing something like Gauss’s construction, after the fact, in that most of the ideas that were worked on 2000 years ago were wrong. They were so wrong that we have never heard of them. Some of them may have been so bad that they were “not even wrong”. Most ideas are bad ideas. Most music is bad music. Most fiction is bad fiction. So the Gauss example has an “unlikeliness” problem, a good metaphor for the way unlikeliness plagues these wild, escapist, ideas of “physics” today.

    Also, who was working on the construction of a 19 sided polynomial 2000 years ago anyway? That was not a problem that took 2000 years to solve. It was an idea Gauss had of his own and solved, presumably, very quickly. Same goes for Democritus (and Lucretius). No one spent 2,000 years trying to prove it. It was not a project of research, just speculation. Modern atomic theory didn’t arise by guys who read Lucretius and decided to try and prove it. The idea rose de novo with the scientific investigations of the modern era. After that, it was commented that Democritus and Lucretius had come up with the idea in ancient days but they in no way founded or inspired the modern research program. There was no connection at all.

    But I like Steinhardt’s description of the anthropic principle.

    “Yuck!” (see his Fermilab presentation of 15 months ago)

  20. tulpoeid says:

    Philip Gibbs,
    You used uncertainty as an example. Let’s say that I don’t like the notion very much, but it is well proven by now. Then, I don’t like it philosophically but nothing can be done about that.
    On the other hand, demanding continued funding and respect on the basis that we can’t dismiss what can’t be either proven or falsified, speaks about a way of defining how science should work (epistemology). It’s not a matter of taste and it has some very real repercussions, and this is what the post is about.

  21. Alan says:

    The Observer, Steinhardt may say “Yuck” in relation to the Anthropic Principle but really it’s misnamed anyway and to do with the inevitability of life of *some kind* in this universe. No matter we get wiped out in a cataclysm, something else will prevail somewhere in the universe into it’s far BY-old future. The question is, is this significant? Science fact essentially meets a valid philosophical question.

    Peter, I don’t really get that Spergel talks of learning from “string theory colleagues” and yet doesn’t like the multiverse – which I thought was kind of inevitable from ST. I mean, if the multiverse is kind of forced theoretically (is it?) then is it the string theorists fault?

    BCnrd, But I thought the point was well made. I’d guess many good thinkers over those 2000 years were pondering Democritus’ idea on atoms and were wondering how to test it.

  22. Peter Woit says:


    “String theory” now refers to a huge variety of different sorts of activities, and I think Spergel was referring to ideas about dualities, which have nothing to do with the multiverse.

    About “string theory implies the multiverse” claims: string theory unification is a failed idea that predicts nothing. It is compatible with pretty much anything, including a multiverse or no multiverse.

  23. Casey Leedom says:


    Thanks for letting me know about the difficulty getting into the awards ceremony. Too bad, but personally, I think I would enjoy the symposium more.

    I’ve recently started going to a bunch of the Stanford Bio-X talks and they’ve been great. I have no idea why I’ve lived this long in Palo Alto and not taken advantage of the many and various free talks at Stanford. Oh well, better late than never I suppose.


  24. Lowell Brown says:

    I do not think that 2,000 years in the past has anything to do with 2,000 years in the future. On must relate past to future years by many scale factors. For example, one should divide the year by the number of scientists living in that year. And divide by the large and multiply by the small energies that can be measured. Same for distances and times. Divide by the number of communication links. Divide by the knowledge acquired in some measure. So, allowing for my crazy big factors, maybe 2,000 years in the past is 2 or 20 years in the future.

  25. Philip Gibbs says:

    “Multiverse” is a word that covers may different ideas in physics. Since Spergel is a cosmologist he may be thinking mostly of the multiverse version of eternal inflation. This is a very incomplete theory built on several layers of speculation. I don’t favour it myself and agree with those who say that it has been overhyped. I don’t think the logic in support of it is particularly good, but I accept that some version of it could be correct and don’t see why others should stop thinking about it.

    The multiverse idea can however also be used to refer to anything from ordinary Hilbert space in quantum mechanics up to the Mathematical Universe of Tegmark. Even the anthropic principle means different things to different people and is obviously a broad philosophical principle rather than a specific physical theory. It is easy for someone to conflate the issues by talking about the multiverse without being clear about which of these things they actually mean. A more concrete question is whether the vacuum state with the spectrum of particles we know is a unique consequence of some fundamental theory, or one of several possible outcomes whose selection involved an element of historical chance. I think this is a perfectly scientific question. It could only be answered in the context of some theory that we do not have yet. It is difficult to see how it could ever be tested or falsified directly, but a satisfactory answer may be given by a theory that is perfectly testable in other ways.

    At this time phenomenology beyond the standard models is very scant and quantum gravity phenomenology is non-existent. I don’t think epistemology or any other branch of philosophy can be applied to answer questions like that. The scientific method determines the soundness of the way scientists operate but it does not offer definitive answers to scientific questions. It is only right that physicists are allowed to explore all options mathematically until the observational prospect improves. Opposing philosophical principles have always played a part in the directions theorists go. Any theory of quantum gravity you could imagine would not be falsifiable while quantum gravity phenomena are inaccessible to experiment, so that kind of criticism is void.

    It is sometimes argued that lack of funding and the influence of a small section of elite theorists is stopping people from exploring all the alternatives, but I suspect that if good new ideas are waiting they will be found sooner or later despite what anyone says. Everyone is free to express their opinions and explore the directions they favour, but it is not helpful to deliver polemic speeches that criticise other people for following the path they find most promising. Surely Spergle is sufficiently competent and well placed to coach his own students to help him research any direction he wants to go in.

  26. Jernej Satler says:

    David. Braaaavo!

  27. R LeVitt says:

    Well, alright, I’m going to say something that I probably shouldn’t say here. I don’t think the monoverse is a testable and interesting scientific hypothesis. ;^)

    My point being, we have exactly zero evidence that our Universe is a singleton. The hypothesis that our cosmos is unique seems an unprovable idea with no data supporting it. As the old saying goes, “Absence of evidence is not evidence of absence.” Which in my view makes the monoverse and multiverse more or less equivalent from an empirical point of view.

    If I understand the problem correctly, the problem of one or many cannot be resolved by experiment. If so, theoretical ideas may give us the only hints, however tenuous.

  28. a1 says:

    @RLeVitt: multiverse vs. monoverse, who wears the burden of proof?
    Universal statements are falsifiable, existential statements are verifiable.

  29. I would just like to stress that there is a huge leap between saying that
    1) String theory multiverse is not justified because string theory isn’t,


    2) no scientific theory whatsoever may contain reference to a multiverse.

    While the first statement seems to me to be an accurate description of the current situation, I see the second one as a dogmatic position which can potentially hinders scientific progress.

  30. William Astley says:

    An infinite number of patches with infinite diversity possibilities is also a problem for ‘inflation’.

    If it is theoretically possible for there to be good inflation (Inflation is the name for the 100,000 times faster than the speed of light expansion of space by the hypothesized inflaton field immediately after the BB event) then there would also be bad inflation.

    “What do you mean? Inflation has two major problems: First of all, we have learned that inflation is highly sensitive to initial conditions. This is the opposite of what everyone thought originally. For example, in the 1990s, by considering different initial conditions and parameters, Linde (and others) championed models of inflation that would lead to an open universe rather than a flat universe, because, at the time, observations seemed to point that way.”

    “Second, we have also learned that inflation generically produces a multiverse (“multimess”) of outcomes – literally an infinite number of patches with an infinite diversity of possibilities – and there is currently no criterion to prefer one possibility over another. As Guth has put it, “In an eternally inflating universe, anything that can happen will happen; in fact, it will happen an infinite number of times. Thus, the question of what is possible becomes trivial—anything is possible […] The fraction of universes with any particular property is therefore equal to infinity divided by infinity—a meaningless ratio.” See, highlighted text in the Conclusion section of Guth’s paper published in J.Phys. A40, 2007 (LINK). In other words, there is nothing that says that what we observe in our patch is typical or could be predicted a priori on the basis of the theory.”

  31. Peter Woit says:

    Spergel clearly was referring to the cosmological multiverse, and I’m sure he would say that claims there is only one universe are just as unscientific as multiverse claims. And I doubt he intends to criticize theorists for working on whatever they want, the problem is when they decide to start selling untestable ideas to the public to further their agenda.

  32. R LeVitt says:

    @ai: My point was that neither the multiverse nor the monoverse ideas are falsifiable at present, if ever. The only existential, verifiable statement we can make is that there is at minimum one Universe.

    As for burdens of proof, nature doesn’t require them–and is under no obligation to make all her secrets known to us.

  33. Peter Woit says:

    Unless you have something substantive to say about the Spergel story, please don’t add to this comment section. Most attempts to argue about the multiverse do nothing but reinforce Spergel’s point that nothing is being explained and we’re best off without such arguments.

  34. Pingback: Various and Sundry | Not Even Wrong

  35. JimV says:

    It is probably too late, since I just saw this post today (which is my fault), so if this comment isn’t accepted so be it.

    My thought on the multiverse/Anthropic answer to fine-tuning is that yes, it could promote some sort of understanding because it supplies a mechanism, and once one has a mechanism it is easier to explain a concept. Sort of like when Feynman was asked what he would like to pass on as a starting point to another civilization (or something like that) and he said the concept that matter is composed of tiny bits that are sticky when close together (or something like that). That example is not completely on-point because it’s true and can be tested, but just the availability of the multiverse idea is, to me, sufficient to counter fine-tuning arguments. It demonstrates that there is another logically-possible alternative, whether that alternative is testable and/or true or not.

    And I like to have a mechanism. It makes things so much clearer if I can see how something could occur even if I don’t know whether it occurs. So I’m glad the multiverse concept exists. Granted, we always should preface such speculations with “I don’t know if this is true”, but isn’t it fun to speculate?

Comments are closed.