Friday, July 5, 2013

More on Falsification

To add fuel to the fire, here's a snippet from a larger piece on falsifiability from Sean Carroll (Cal Tech cosmologist). The whole post is here. I find lots to disagree with in the larger post, but I find the tone about Popper (or naive Popperians) quite congenial. Note his use of 'cudgel,' 'charge,' and the 'don't have to think too hard.' 

"The falsifiability question is a trickier one, to which I will not do justice here. It’s a charge that is frequently leveled against string theory and the multiverse, as you probably have heard. People who like to wield the falsifiability cudgel often cite Karl Popper, who purportedly solved the demarcation problem by stating that scientific theories are ones that could in principle be falsified. (Lenny Susskind calls these folks the “Popperazzi.”) It’s the kind of simple, robust, don’t-have-to-think-too-hard philosophy that even a scientist can get behind. Of course, string theory and the multiverse aren’t at all the kinds of things Popper had in mind when he criticized “unfalsifiable” ideas. His bugaboos were Marx’s theory of history, Freudian psychoanalysis, and Adlerian psychology. The problem with these theories, he (correctly) pointed out, was that they told stories that could be made to fit literally any collection of data. Not just “data we could realistically acquire,” but absolutely anything you could imagine happening in the world. That’s completely different from the examples of string theory or the multiverse, which clearly are saying something concrete about the world (the ultraviolet completion of quantum gravity, or conditions in the universe far outside our observable region), but to which we have no experimental access (or almost none). Of course, there’s also the issue that the demarcation problem is a lot trickier than naive Popperianism makes it out to be, but that’s another discussion. The right strategy, once again, is to look at what actual scientists would do or are doing. When faced with difficult problems concerning quantum gravity or the early universe, they follow precisely the outlined program: they invent hypotheses and try to see which one is the best explanation for the data. The fact that the data are relatively crude (the existence of gravity and gauge theory, the known cosmological parameters) doesn’t prevent it from being science."


  1. Another fun read is Larry Laudan's The Demise of the Demarcation Problem. It doesn't focus specifically on falsifiability but on the broader problem of finding criteria that distinguish science from non-science.

    It does seem a bit silly that syntacticians should have to make a case for their theories being falsifiable given that the majority of philosophers of science are not (have never been?) Popperians. Nonetheless, I think it's worth emphasizing that any working syntactician confronts falsifying data and arguments all the time. Anyone who looks at the syntax literature, or who attends a few conferences, can see that syntacticians are not living in some sort of fluffy informal paradise where everything is too vague to be wrong. There are some overarching theoretical proposals which are not immediately and straightforwardly falsifiable (e.g. antisymmetry, hypothesized abstract universals), but that is just what you'd expect to find in any field. As Carroll's comments suggest, adopting a Popperian stance can serve as a convenient means of dismissing the literature without really engaging it. (And without really engaging Popper's philosophy of science either, which is quite radical and counterintuitive in some respects.)

  2. First I thank Alex D. for reminding us that, for the most part, sytnacticians work on syntax, and do great scientific work that does not need to hide behind what say physicists do. For that reason it is a bit odd that again and again Norbert drags in unobvious analogies between particle physics or string theory and generative linguistics. Just WHAT is it that makes this "branch" of physics relevant to work on say language acquisition? We have no ‘experimental access’ to the beginning of the universe or postulated multiverses but we have easy access to human language. Except for Chomsky in ‘off the cuff remarks’ no one believes that it is impossible to learn ANYTHING about the language faculty with the experimental methods we have access to at the moment. Angela Friederici is not so well known because she speculates about multiverses but because she does innovative brain-research! As do many other psychologists of with very different [language related] theoretic committents. So would it not make much more sense to compare work on syntax to work of psychologists instead of to abstract hypotheses in some distant corner of physics?

    Now given that Norbert refers recursively to Sean Caroll, it might be helpful to have a look at

    Especially relevant for our discussion is:

    “The problem that Sean doesn’t mention is one of circularity. Since you can’t observe anything about it directly, the multiverse must be justified in terms of another theory that can be tested and this is string theory. But if you talk to string theorists these days about how they’re going to test the unified theory that string theory is supposed to provide, their answer is that, alas, there is no way to do this, because of the multiverse. You see, the multiverse implies that all the things you would think that string theory might be able to predict turn out to be unpredictable local environmental accidents.
    So, the multiverse can’t be tested, but we should believe in it since it’s an implication of string theory, but string theory can’t be tested because of the multiverse.
    Until recently, string theorists would sometimes hold out hope that the LHC would see low-energy strings, extra dimensions, or supersymmetry, and that these discoveries would somehow pick out a predictive version of string theory from the landscape of the multiverse. This year’s data from the LHC has pretty much destroyed such hopes.

    Sean ends with the inspirational admonition:

    The proper scientific approach is to take every reasonable possibility seriously, no matter how heretical it may seem, and to work as hard as we can to match our theoretical speculations to the cold data of our experiments.

    What’s going on in this story though is not a concerted effort to match theoretical speculation to experimental data, but something very different, a concerted effort to build a theoretical framework perfectly insulated from testability, and sell it to the rest of the physics community and the public, hoping no one notices the circularity.”

    Main stream physicists have pretty much given up on string-theory/multiverse because the Large Hadron Collider, which managed to discover the Higgs almost the first time out, has turned up absolutely 0 indications of supersymmetric partner particles. Evidence counts.

    So maybe it is not such a good idea to construct parallels between string theory/multiverse and Minimalist UG….

    1. I think what syntax does have in common with particle physics and gravitational physics is virtually no useful lower level theories to constrain hypotheses. A herpetologist studying a population of lizards in a rock outcropping can draw on a vast collection of scientific results and solid common sense to justify her conclusion that the population will go extinct if a proposed road is built: she knows that they can't fly over it, tunnel through 70 meters of subsoil under it, live off photosynthesis (or, if they are one of the few vertebrates that can get a bit of help this way, there are ways of ascertaining this), or by fusing heavy hydrogen in the morning dew in little reactors at the bases of their tails. They can only live of whatever kinds of bugs or vegetation they eat, and will have to get across the road, if at all, by crawling.

      Syntacticians on the other hand have virtually no useful such constraints on their theories, beyond the ideas that A)languages cannot be transmitted by telepathy, but only by witnessing overt performances (including, of course, classroom reading of archaic texts such as Shakespeare, for some people), B) generation and interpretation should be possible by algorithms in the cases that arise in practice. This isn't nothing (probabably more than the physicists have), but it's pretty thin compared to what real biologists get.

      Note that B) does not mean formal intractability, for example the algorithms used in 90-s era 'corridor shooter' computer games to compute what bits of the map were visible from what other bits (the VIS method of Quake engine games for example) were computational intractable, but feasible by algorithms for the cases that arose in practice (i.e. the maps were typically small, & there was/is a certain amount of craft involved in managing the limitations imposed by the intractability).

    2. oops 'does not mean formal intractibility' -> 'does not require formal tractability'.

  3. I think this thread illustrates how arguments over methodology can so easily become completely tangential and unproductive (I should take some of the blame). Someone kicks off with a bit of Popperian griping and before we know it, we're having an argument over whether string theory qualifies as properly scientific. Popper lite can be useful to the working scientist as a methodological maxim ("try to figure out ways of testing your theory"), but any deeper discussion of what qualifies as scientific and why is probably best left to philosophers of science.

    1. Exactly what would be wrong with giving SCIENTISTS a voice in the discussion of what qualifies as science? Besides my point was not that it would be productive to debate whether theories in fields so remote from linguistics as string theory are or are not scientific. I only suggested that IF someone like Norbert feels an urge to drag such theories again and again into debates about linguistics it would be in his own best interest to select theories that are uncontroversially accepted by the scientific community. So my link was merely intended as an invitation to educate yourselves about what the scientific community at large thinks about theories Norbert seems to favour because.... [I really have no clue WHY he would].

      Finally, given the quality of the *philosophizing* Norbert engages in, you certainly have a point - he ought to leave this up to professional philosophers of science. At least some of us find him not terribly amusing. I know our work looks from a distance as if we're just sitting merrily in comfy armchairs and jot down a smart sentence every once in a while. But believe it or not, we have to WORK as well...

    2. I don't think Norbert was taking any position on string theory itself, he was just quoting Carroll's remarks on falsifiability which happened to be made in that context. As a philosopher of science, you must be aware that what Carroll says on this point is not very controversial. I.e., naive falsificationism is not to be taken too seriously.

    3. The point that "naive falsificationism is not to be taken too seriously" is indeed so uncontroversial, that it does not need to be made over and over again. Far less is it necessary that Norbert invokes the 'authority' of Caroll [whom many physicists would refuse to be mentioned in the same sentence with] - he can just assert it based on his own authority. And if he would follow up such assertion with an example from LINGUISTICS it would probably even convince people who still need convincing. The constant analogies to physics are odd; especially coming from someone who also writes: "Let me leave you with a simple piece of advice that has served me well: when you hear the word ‘analogy’ reach for your wallet." [Norbert. October 16, 1912, ]

  4. The point that "naive falsificationism is not to be taken too seriously" is indeed so uncontroversial, that it does not need to be made over and over again.

    Well, if only it were so, but the point is being made because some of the methodological criticisms of generative syntax that have come up here do appear to have falsificationist presuppositions.

    As for who doesn't want to be mentioned with whom in physics, who cares? There's no point is linguists/philosophers discussing string theory on a linguistics blog. None of us have anything to say about it.

    1. There is a difference between naive falsificationism which is simplistic I think we all agree, and non-naive falsificationism (for example, Lakatos' "sophisticated methodological falsificationism") which are tenable positions for all their faults.

    2. Yes of course. However, it's rather difficult to show that a theory fails to meet the demands imposed by Lakatos (especially since theories aren't even the unit of falsification for him, IIRC), and even if one can show this, it's not clear how significant this is. A proponent of the theory can simply respond: "Ok, but I don't agree with Lakatos' philosophy of science".

    3. "A proponent of the theory can simply respond: "Ok, but I don't agree with Lakatos' philosophy of science".

      S/he could indeed. But why always expect the worst from 'the other gal/guy''? Is that what you would do if someone shows you a serious flaw in your theory? Or would you at least consider the possibility that you are wrong? I suspect that when you're not 'hanging out' here, telling philosophers they're too stupid to do philosophy, you're actually a hard working linguist; trying to make a worthwhile contribution. Why not expect [by default] that other linguists have the same goal. Alex Clark strikes me as an way above average reasonable guy - do you think he would pull this reply if his 'pet theory' is under scrutiny?

      I think linguistics is way to complex and we know way too little so far for one 'camp' to have all the right answers. So any serious progress could just need collaboration with 'the other gals/guys'. For that reason it would be great if you [pl] could finally overcome the kind of war-mongering expressed here: "Like Robing Hood and the Sheriff of Nottingham, Chomsky and Quine have staked out the poles...Within professional philosophy. Chomsky has appeared the brigand, leading raiding parties deep into enemy territory..." [Hornstein, 1991, 104] - jeez such attitude is so last century - just try to TRUST that those who hold different views on say UG are not THE ENEMY - you might just be in for a very pleasant surprise...

    4. Is that what you would do if someone shows you a serious flaw in your theory?

      No, it's what I might do if someone criticized it on the grounds that it didn't meet some methodological criterion proposed by Lakatos. (Although it would depend on the details — I'm not dismissing in advance all conceivable criticisms meeting that description.) I assume that Alex didn't have a criticism of that sort in mind anyway.As I recall, his general point has been that there is a close connection between formalization and falsifiability. In my experience, informally-specified syntactic theories make lots of predictions which can be tested, and I've quite often seen my own or other people's theories refuted quite decisively by a new data point. Part of the disconnect here probably relates to the fact that Alex is less interested in syntactic analyses of specific phenomena and more interested in hypothesized universals and acquisition mechanisms. I do think that theories in this domain are a bit harder to falsify, on the whole, than (say) a particular analysis of short scrambling in Japanese. However, I don't think that formalization is likely to help here except in a rather trivial way. Those things left implicit in informal theories are usually precisely those things which could be revised without modifying the core of the theory. Of course, that's not always true and formalization does occasionally lead to important new insights.

    5. If I understand your correctly [and please correct me if i don't] you suggest those who push for formalization do so because there is [or they perceive there to be] "a close connection between formalization and falsifiability". I think this is a bit mixing cause and effect. From proposals i actually have read I gather that people push for formalization to get deeper insights and to provide better tools for data analysis. Consider:

      "The goal of this paper is to give a precise, formal account of certain fundamental notions in minimalist syntax, including Merge, Select, Transfer, occurrences, workspace, labels, and convergence. We would like this formalization to be useful to minimalist syntacticians in formulating new proposals and evaluating their own proposals, both conceptually and empirically." (Collins & Stabler, 2011)

      At least to me this seems a worthy pursuit. Now it might be true that it is easier to falsify a theory that is formalized. But surely that is not the primari objective for formalization, rather it is a byproduct.

      Further, as you say yourself, it is possible to falsify proposals that have not been formalized. What is important then is that one is clear about what 'the core' of one's theory is. If it is not possible to give a formal account it has to be at minimum a non-circular account. Only then can one tell whether a particular auxiliary hypothesis can be replaced or not.

      Recall Popper did not call Marxism pseudosceince because it had not been formalized but because during the attempts to establish Marxism in the real world our beloved leaders just kept changing auxiliary hypotheses to insulate the core from falsification. And that seems a problem very different from throwing out a theory because of one problematic data point....

    6. I completely agree with you that there may be many different motivations for formalizing a theory, Christina. Alex has mentioned in a few comments that formalized theories tend to be easier to test.

  5. " There's no point is linguists/philosophers discussing string theory on a linguistics blog. None of us have anything to say about it.' - I could not have said this better myself and hope Norbert is reading your emphatic plea - he was the one introducing claims about string theory in the discussion - not me...

    As for methodological criticisms of generative syntax: some may be based on naive falsificationism. But others [e.g., Paul Postal's criticisms expressed here: ] are not and remain unanswered. Maybe you would like to be the first to respond to Postal?