Way back in 1997, string theorists were already getting rather touchy about people pointing out string theory’s testability problems. At that time, Gordon Kane published an article in Physics Today with the title String Theory is Testable, Even Supertestable in which he wrote:
A decade ago in PHYSICS TODAY (May 1986, page 7), Paul Ginsparg and Sheldon Glashow raised this question dramatically, and effectively began a widely repeated myth that string theories, candidates for a primary theory, are not testable. Here I want to dispel this myth, and describe some of the many ways in which string theories are testable. If nature is supersymmetric on the electroweak scale, for which there is exciting but not yet compelling evidence, then string theories are even testable in essentially the same ways as traditional ones. All the tests I describe are doable now or in the foreseeable future with existing or proposed facilities or projects.
Kane went on to give a long list of testable things that string theory was going to predict, including as an example a detailed spectrum of superpartner masses, all in the range 50-300 GeV (he assures us that supersymmetry is a prediction of string theory, quoting something Gross and Witten wrote for the Wall Street Journal).
Now that LHC data has finally started to arrive, in amounts large enough to soon start seeing all of the superpartners advertised back in 1997, Physics Today has decided to put out a rather spectacular piece of string theory hype from Kane as their cover story, under the title String Theory and the Real World. In the story, the main theme is the same as 13 years ago: it’s a myth that string theory doesn’t make testable predictions. Now though, the many 1997 predictions are forgotten, and much of the article is devoted to a tendentious discussion of what it means to test a scientific theory. In the 2010 version, there’s no longer a detailed list of things that string theory should be able to predict, instead Kane describes two specific predictions of string theory:
We showed that in no case could the theory generate light but not massless neutrinos. That work represents a clear example of a test of string theory.
So, one specific string theory compactification is known to not look like the real world. That’s a test of string theory????
Whatever one thinks of these latest “tests”, the difference between what string theory tests looked like back in 1997 and what they look like in 2010 is rather remarkable.
“We showed that in no case could the theory generate light but not massless neutrinos. That work represents a clear example of a test of string theory.” means that string theory predicts, correctly, that neutrinos have mass. I think you were confused by the multiple negations in the statement.
Anonymous you are the one being confused, if it said “string theory predicts light but not massless neutrinos.” it would mean string theory predicts neutrinos have small mass (not massless =with mass).
But the actual statement is the negation of the above since it says that in *NO* case could the theory generate light but not massless neutrinos.
I have been reading this blog with increased enthusiasm.
Unfortunately, I do not have a background in theoretical physics
and so it is difficult for me to evaluate the testability of this hypothesis.
My basic question was:
What is the endpoint, if there is one, of string theory? If there are major disagreements regarding the experimental evidence of a theory that has been around for 30 years, what can one conclude about the nature of the propositions that will be posited in the next 30 years?
I read the article yesterday, and found it to be really disappointing. I was very sure you’d comment on it 🙂 The most interesting sentence is this one
“interested readers can find others in talks accessible from the website of the international String Phenomenology 2010 conference held in Paris this summer (http://stringpheno.cpht.polytechnique.fr)”
I’d have expected that his article provides a summary of these efforts, rather than just advertising two of his own papers.
Anonymous,
I still think my interpretation of that “prediction” is correct. After what I quoted, the paragraph goes on:
“Although the particular compactification we studied did not yield the desired neutrino masses, different compactifications may allow for neutrino masses consistent with experiment and offer explanations of observed neutrino properties.”
Maybe though Kane is claiming that string theory “predicts” light but not massless neutrinos because they occur in one particular compactification. If so, “bizarre” would still be an accurate description of the argument.
String theory is testable in a certain range of its fundamendal mass scale. Its tests are high on the LHC priority list and possible with only 2.9 inverse picobarns of data. See
http://128.84.158.119/abs/1010.0203
for latest results.
The statement that string theory is not testable is false.
OK T.T.,
Since the LHC experiments have nearly 50 inverse picobarns collected, at the Winter conferences we should see announcements of the results of testing string theory. If the result is failure, I guess that’s it for string theory, right?
Or maybe string theory gets to have it’s own special kind of test, the kind where it doesn’t matter if the test-taker fails…
Peter,
The main premise of your blog was that “string theory is not testable”. On many occasions you stressed that it does not matter for you whether it was right or wrong, so I will not answer your question. Furthermore, in experimental physics, we do not use the word “failure” to describe results — we rule out or confirm theoretical predictions.
If you want to make true statements, you’ll need to add footnotes or small print.
It looks however that the main purpose of this posting was to make fun of Gordy Kane and other people who are enthusiastic about the prospects of the theory.
Peter wrote:
I don’t have access to physics today and can’t read the article, but I would like to know what this discussion is about: If someone could provide a synopsis I would be thankful.
T.T. wrote:
From the abstract of the paper you refer to:
AFAIK string theory can be adapted to not predict anything below 2.5 TeV (or XY TeV, for that matter), at least that is the state of the art right now. Therefore this is maybe a test of certain versions of string theory, but it is still not possible to falsify the theory. (My apologies to Peter if this remark starts the kind of discussion he has grown weary about).
T.T.,
I think the person who needs to add small print to their claims is you. In the conventional understanding of what it means to “test a theory”, when the experimental result differs from what the theory predicts that carries negative implications for the theory. And yes, people apply the word “failure” when this happens. If the LHC data differs significantly from standard model predictions, this will rightly be described as a failure of the standard model and be taken extremely seriously.
On the other hand, not even the most enthusiastic string theorist seriously expects the LHC to see strings in the first 2.9 inverse picobarns of data. Claiming that your theory is being tested by an experiment when you don’t actually expect the experiment to see anything is just a not very honest abuse of language.
Jupiter,
I think the endpoint of the history of claims about string theory being a unified theory of nature is already just about here: this is an idea that doesn’t work. The lack of any support for such claims in the LHC data over the next few years should put the final nail in that coffin.
On the other hand, there will always be some people and some publications that will never admit what has happened. Extrapolating from the last two examples, by around 2025 we’ll see Physics Today publishing a special issue by Kane on “String theory is too testable, darn it!”. Hard though to even guess what the arguments for this will be by then.
T.T.,
I think Peter really has manners enough not to try to make fun of Gordy Kane. There are bloggers who do make fun of other people (most of us know at least one of those…) but it appears that Peter does something far more intelligent – he points out that some people are making fun of themselves, without intention.
Switching from the meta-topic to the topic, I just wanted to add that there are AFAIK variants of string theory making predictions that are visible with the amount of luminosity you mentioned at the LHC. This is what some people mean when the talk about sensitivity of the LHC to string theory.
It is however not given that we *must* see evidence of this. I believe it takes an average string theory phenomenologist and two hours to set up some model that does explain why we don’t see the signal. This is, why a reasonable man should not say that the LHC is sensitive to string theory, but at most to a tiny tiny fraction of some huge parameter space.
Manuel,
Thanks, although I fear my last comment does show a certain willingness to make fun of this kind of thing. I agree though that mostly what’s going on here is Kane (and Physics Today) choosing to embarrass themselves.
Peter:
There is no problem with the predictions changeing. This is exactly what we should expect of an evolving theory. You accuse the string theorists of claiming that they expect to see nothing, while saying their theory is being tested. Surely, that would not be a wise strategy for them, would it? Rather, the point is the unification of physics is a non-trivial exercise of the humans. It is absolutely incredible to think that a finite machine might reach the energies necesssary to probe this realm, but it is not necessary.
Sam,
The problem isn’t that string theory makes changing predictions, it’s that it makes no legitimate predictions at all. Back in 1997, Kane wasn’t making predictions, he was saying that in the future it would make predictions. Now that the future has arrived, all that has happened is that his past predictions of predictions have collapsed.
G Kane makes a living out of proposing experimental tests for supersymmetric (not necessarily string) theories. He did a lot of work on testing supersymmetric GUTs in the early 80’s before superstrings. I heard him give talks on testing super GUTs in 1983 before the work of Green and Schwartz on anomaly cancellation in type I superstrings.
That’s more than 25 years ago. I wish I could hear Kane admit once that he made a clear prediction that turned out to be wrong but he never does. That’s not right.
Consider on the other hand Ch Thorne who is one of the true fathers of string theory (works on string since the late 60’s) and who very clearly says that string theory as a unified theory of all interactions is not working.
We can do many things with it (come up with new, beautiful mathematics, for example), but at this stage, in this form, unification is not one of them.
For any who are looking for the original article by Gordon Kane but have not any clue where to access it, I have uploaded a copy on file-hosting websites
[Sorry, but using a Columbia University web-site to provide a link to less-than-legal copies of documents is asking for trouble I don’t want. If anyone has a better, legally non-problematic, link to this content, I can post that]
Write a letter to Physics Today about Kane’s article. Quote from his 1997 article, for example the detailed spectrum of superpartner masses, all in the range 50-300 GeV. Point out that many 1997 predictions are forgotten in the 2010 article. Ask Kane how many of the 1997 predictions have been verified, or point out that none have been successful.
PT may not accept the letter, or may edit it heavily. KISS ~ I think you know what that means.
On the matter of scientific prediction, I just wanted to mention something I explained in my post What is a scientific prediction?. It is in practice very rarely the case that a theory is falsified. More often, it is what one could call “implausified”: You constrain parameters so much the theory becomes implausible (most often because the parameters are unnatural). The problem with the LHC “predictions” is that if nothing is found, you can just go and say “naturally” we wouldn’t have expected to see anything, because that particular model wasn’t the right one or because this compactification scale, mass spectrum, etc wasn’t plausible to begin with or swhatever. That’s very likely what’s going to happen within the next couple of years. (But you know, we would totally expect to see something at the ILC!!) I think you’re actually better off sticking to cosmology, though even there you have the problem that the model doesn’t uniquely follow from the theory, so there’s always an escape route. In essence, one has to be careful what one calls a prediction of the theory itself and what a prediction of a particular model that is based on the theory. For all I know, everything you’ll hear at the string pheno is of the latter, not of the former type. Consequently, if you try to reverse the logic, falsifying the model doesn’t falsify the theory. (Besides that, it also isn’t much of a “pre”diction to claim neutrinos have small but nonzero masses.) Not that this problem is unique to string theory in particular. In any case, it’s unfortunate Kane is making his points so badly because I welcome the string pheno efforts.
So, I am looking at T.T.’s link and trying to understand exactly what the “string resonances” that have been excluded at the 95% level under 2.5 TeV are.
Poking at google I am seeing explanations that the thing being tested by the search described in T.T.’s link is not string theory in general, but rather the “low-scale string” scenario, where the “fundamental string mass scale” is in the TeV range. Is this correct?
It furthermore appears that even in a universe where string theory is correct, it would be more surprising than not if the “low-scale string” scenario were true. So in contrast to, say, Supersymmetry or the Higgs Boson (which, if they are not found at the LHC scale, this will be considered alarming and potentially threatening to the acceptance of those theories, because although you can still rescue the theory you have to do contrived things to the models to do so), if the fundamental string mass scale is not found in LHC ranges then string theorists would legitimately consider this basically to be the expected result, because the low-scale scenario is relatively exotic among string theories. Is this also correct?
Also, I am seeing suggestions that low-scale string theories require “large extra dimensions”. Are the low-scale string theories, then, the same ones where the LHC is predicted to eventually generate the vaunted “micro black holes”? But it looks like the search is not looking for signs of micro black holes being created, it is looking for “string Regge excitations” ( http://arxiv.org/abs/hep-ph/0001166 , reference [3] in T.T.’s link ) of quarks and gluons. Is there a simple explanation of why these “excitations” occur in the scenario being tested?
Taking these statements at face value, it is fair to say that string theory is indeed testable, or at least one specific compactification of it. Kane claims that this theory is inconsistent with light, non-massless neutrinos, and exactly that form of neutrino has in fact been observed, so this specification of string theory has been tested and refuted. Is this not right?
neo,
The criticism of string theory is that it predicts nothing since you can find a “solution” that has any feature you want. Showing that there is a solution with the wrong neutrino masses doesn’t exactly refute this.
Strings vibrate. The quantized modes are Regge excitations. This is a universal property and is valid for all models, not a “feature” of any particular “solution”. If Nature chose just one parameter, the fundamental scale of order TeV, LHC will discover string resonances in dijets and many other places. This is very simple, so if you have a scientific (as opposed to social group-thinking that you despise) argument why Nature made a different choice, I’ll be glad to hear it.
The phlogiston theory was never disproved, per se. The luminiferous ether theory was also never disproved. They were merely forgotten or abandoned, because better theories came along which could explain the facts without increasingly artificial contrivances. So it will be with string theory. One can complain about the state of physics TTWP etc, and commit suicide as Boltzmann did, but that is neither here nor there.
“The criticism of string theory is that it predicts nothing since you can find a “solution” that has any feature you want.”
Wouldn’t it be rather big news if a solution of string theory which looked exactly like the real world was found?
T.T.: Because there’s no deep reason the relevant scale should be at a TeV.
At Mitchell Porter,
No it would not be big news if string theorists found such solution because it wouldn’t explain much: the strength of such explanation would be similar to the Greek ‘explanation’ of positions of planets by epicycles. I am sure that the Greek thought in that time that this was the best exlanation one could get, just like string theorists now believe that science has to ‘progress’ through antropic reasoning… History has shown this type of ‘ideas’ to be wrong again and again and one should not take it seriously.
Mitchell Porter,
If a string theory solution is found in which all the parameters of the standard model can be reliably calculated and they come out right, that will definitely make the newspapers.
So would the observation of a flock of pigs circling the CERN site at an altitude of 100 meters.
Theorem [Larsson 2007, 2009]
Supersymmetry will not be discovered at the LHC.
Proof: String theory predicts supersymmetry (Witten 1984-2002). String theory predictions are always wrong. Hence supersymmetry does not exist, and will in particular not be found at the LHC. QED.
Corollary
Lubos Motl will lose his experimental-susy-by-2006 bet.
Peter – there is an obvious contradiction between “string theory can be made to look like anything” and “string theory can’t be made to look like reality”. So what are you saying – that string theory *can* resemble reality, but that the calculations would be hard?
???? Mitchell Porter, I don’t understand what is so difficult to comprehend from the last comments… There are really a few issues here
(a) String theory hasn’t managed yet to resemble reality
(b) string theory contains 10^{500} solutions of which presumably the large majority do not look like reality
(c) even if string theory would solve (a), it probably never can *explain* (a) in the context of (b).
So like I said, this is the situation the Greeks had with epicycles, they could solve (a) but never got an explanation for it … until galileio and newton came by. Peter appears to think that solving (c) is never going to happen and I tend to agree with him.
Usually, a common feature of all scientific revolutions is that calculating out the old theories was *easy*. It was a piece of cake for Newton to verify the correspondence between the 1/r^2 force of gravity and the equal area swept out in equal times by planetary motions. It was simple for Einstein to deduce a Newtonian limit for the gravitational laws and the mere fact that the classical quantum correspondence is *not* so straightforward – in my mind – indicates that we are missing something. Consistency checks have always been easy … it was calculating the new phenomena which was usually a bit more difficult (see for example the three body problem in Newtonian gravity).
Why is this difficult to understand? It is not because people have a lack of good ideas that we should desperately try to defend the only things we can come up with so far…
Mitchell,
The problem with “string phenomenology” is two-fold:
1. You can’t actually calculate much at all reliably. Even in the limiting cases where perturbation theory is valid, an example of this problem is that you don’t have explicit metrics for the Calabi-Yaus compactification spaces you would like to use. After 25 years of work on this, the state of the art is not much advanced. This is why even Kane has given up on his 1997 claims, and why it’s highly implausible that all of a sudden you will be able to start exactly matching up string theory models and the Standard Model parameters.
2. As far as anyone can tell, when you are able to calculate something, you can get almost any values you want, by choosing a complicated enough string theory model. Even if you solved 1, there’s no reason to believe that this problem would go away, and that all of a sudden a relatively simple model would appear that matched the standard model. One can’t logically rule such a thing out, but believing in its existence at this point is pure wishful thinking.
Mitchell,
It’s not really true that you can’t calculate much reliably in a given string compactification. For example, in Type II string theory with D-branes on toroidal orientifolds, all parameters such as gauge and Yukawa couplings can be calculated reliably. Additionally, there are string models which strongly resemble the Standard Model. The problem is really of uniqueness. In order to fix the values of parameters such as couplings, the problem of moduli-stabilization must be addressed. This is on top of the problem of the huge number of possible vacua. Peter is right that the so-called top-down approach by which the Standard Model may be derived uniquely from string theory has thus far failed. However, the ‘bottom-up’ approach is alive and well.