Comments

Wednesday, May 6, 2015

My (HOPEFULLY) last ever post on Vyvyan Evans and his endless dodging of the central issues

I was going to write something long and pointed in response to this new piece by Vyvyan Evans (VE). I was going to analyze the article noting its inflated combination of self-pity (those “Chomsky disciples” are saying that my work is junk) and self-aggrandizement (I am the leader of a Kuhnian paradigm shift that will overthrow the Chomskyan orthodoxy that, prior to the heroic efforts of people like me, who are now being viciously pilloried by the intellectually enervated Generative establishment, has had a stranglehold on linguistic pedagogy). I was going to note that VE has yet to address a single criticism leveled against his work, preferring instead to personalize the disagreement in a two fold manner; first by noting his personal travails in fighting the Chomskyan dragon and second by citing the approval of various authorities and luminaries that think him and his work wonderful. However, despite the obvious temptations that VE’s public demeanor make almost irresistible, his latest piece has convinced me that nothing much will be gained by doing this (fun as it may be).[1] The reason is that so far VE has provided every kind of reply to the criticisms leveled against his work except one that addresses the identified problems with his chief “arguments” (and yes, these are scare quotes).  These number three.[2]

1.     VE’s work fails to grasp the difference between a Greenberg and a Chomsky Universal. As such its putative arguments against UG are logically insufficient to get to their desired conclusions (see here and here).[3]
2.     The positions VE’s work identifies and argues against are (regularly and repeatedly) caricatures based on egregious serial misquotation. In particular, the texts that VE’s work cites are very selectively quoted from and when the originals are consulted it is clear that the quoted authors are making the exact opposite points from the ones that VE’s work attributes to them (see here and links therein).
3.     The detailed linguistic arguments VE delivers are inadequate. Not only does the work evidence no understanding of the larger conceptual issues like those in (1) above, it even screws up the technical analytic details (see here for one example examined in detail).

Let’s review these three points again one last time.

First, VE really does not understand what is logically necessary to make his case. This is because VE (still) does not understand how the ‘Universal’ in ‘Universal Grammar’ means. Let me elaborate.

There are two basic conceptions of universal in the linguistics literature. Let’s call them Greenberg Universals (GU) and Chomsky Universals (CU). GUs aim to describe surface patterns common to all languages. GUs thus focus on word order patterns (both prevalent and absent, absolute and statistical) such as the rarity of OVS patterns across the world’s languages and the prevalence of SVO and SOV. GET THE NUMBER OF THESE There are other GUs as well, and some have been the focus of some interesting discussion in the Generative literature in the last several years.[4] However, whatever the theoretical and empirical interest of such GUs, they are different from what Generativists of the Chomsky stripe have meant by ‘universal.’ How so?

Well, CUs do note refer to surface linguistic patterns but to properties of the Faculty of Language (FL) and the formal and substantive restrictions FL imposes on competence grammars (G). Let’s unpack this a bit by taking an example or two of a proposed CU.

A popular CU is the principle that restricts G rules to structure dependent operations (i.e. the G rules are stated in terms of structural features of phrase markers rather than linear properties of strings). Another popular one restricts “movement rules” from applying across islands. A third example is the principle that restricts anaphoric elements from c-commanding their antecedents.  These have been proposed as candidate Universals (i.e. principles that delimit the range of possible grammatical operations and relations, thereby restricting the class of admissible generative procedures available to humans). Note, that CUs refer not to surface outputs like GUs do but to the properties of the Gs that generate a given linguistic object’s meaning-sound outputs. As such, CUs are only indirectly related to surface patterns.

So UG on a CU conception is a description of the restrictions that FL places on the kinds of Gs that human natural languages allow. Thus, on a CU conception, UG describes the general characteristics of human Gs and also thereby limns (some of) the fine structure of the higher-level capacity that humans have which undergirds their capacity to acquire the kinds of Gs that they in fact acquire. Note that empirically investigating CUs logically requires going beyond the inspection of surface distributions (in contrast to GUs). It logically requires investigating the features of Gs. And this requires moving beyond observed surface patterns to the underlying Gs that generates them.

So, say you think that this CU conception of UG and FL is hogwash, how do you argue against it? Well, you need to show (most ambitiously) that Gs can have arbitrary structure (i.e. there is no limit to the kinds of rules, operations and structures that human Gs can have) or (more modestly) you must demonstrate that the universals that GGers have actually proposed are wrong. VE’s work does neither, or does neither competently (see below). Why not? Well here is my diagnosis: VE’s arguments completely fail because VE does not understand the difference between a GU and a CU.  How would this explain VE’s practice? It would explain why VE’s main argument against UG has the structure it does. What is that structure? It amounts to the assertion (i) that natural languages display a very wide variety of surface variations and (ii) that this attested variation in the world’s languages implies (or, more modestly, strongly suggests) that there are no CUs. But this, to repeat, is a non-sequitur, unless, of course one takes CUs to be GUs, which, to repeat, they are not!

More specifically, the confusion between GUs and CUs vitiates VE’s arguments as follows: though surface variation directly bears on GU claims, it has an indirect  relation to whether there are CUs. To demonstrate that surface variation undermines UG in the CU sense requires showing that surface variation implies unbounded G variation. But nobody’s work has ever shown that, and certainly not VE’s. Indeed, VE’s work has not even tried to show this. Rather it has simply asserted (and asserted and asserted and…) that the evident linguistic surface variation found among the world’s languages (which, btw, is not really news to GGers, at least not for the last 40 years) undermines claims for UG.

In sum, VE’s work has not demonstrated that surface variation requires different kinds of G rules, nor has it discussed the rather extensive empirical literature that aims to show how very different surface patterns can all be generated using the same basic kinds of Gs. As this has been one of the mainstays of GG research over the last 40 years (and there are literally thousands of pages that have been written on tupological variation by GGers), this means that VE’s work has not discussed the empirical literature relevant to making even the modest claim against UG noted above. Rather, VE’s work reduces to the assertion of a logical non-sequitur (incessantly repeated as if repeating the same thing again and again and again and… boosts its verisimilitude)[5]; because languages display a very large variety of distinct surface patterns therefore the Gs that generate these patterns cannot have common properties.[6]

Let me note, in a charitable vein, that the confusion between GUs and CUs is understandable. Undergrads confuse them all the time. VE’s work does too. Consequently, VE’s arguments are based on a pun (i.e. confusing two uses of the concept ‘universal’) and hence cannot hope to show what they intend. In other words, VE’s arguments are prime examples of the fallacy of equivocation. This is one very important reason why these arguments are worthless (dare I say junk!).[7]

So the first problem with EV’s criticism of Chomskyan GG is that it is based on an equivocation and is thus without intellectual interest (though sociologically, well, that’s another matter).  However, this is not the worst thing about VE’s critique, for one might argue that this is a rather subtle mistake (though, it really isn’t). The two other problems with his work are that it is both dishonest and incompetent.  Let’s take these in turn.

First the dishonesty. It has been amply demonstrated that VE’s work has troubles with quotation. David Pesetsky deserves credit for making this point repeatedly in e-discussions with VE. The same point has also been made by others (see here and links therein for a particularly clear set of detailed examples). This egregious misquotation alone should have made the work unpublishable. It certainly suggests that the “review” process that CUP conducted was far from thorough.[8] I strongly urge those skeptical of my claims to take a careful look at the relevant discussions linked to above. The misquotation is really appalling given that it consistently attributes to people the opposite positions to the ones that they are actually offering and then criticizing them on the basis of the severely distorted misquoted passages.

Last but not least, VE’s work is technically incompetent. VE appears not to know his way around a syntactic argument, at least as regards the work he criticizes. This further vitiates VE’s work. After all, one very good rule of critical thumb is that one know something about what one is criticizing. So if you want to criticize the Chomskyan view of universals then you should know a lot about the arguments for and against them. VE’s work is sadly lacking in this respect. I have gone over one example in detail here. EV’s discussion here gets everything wrong; from the description of the relevant data to be explained to the typological contrast at issue to the sentences relevant to adjudicating the claims. In other words, it is technically wrong in every possible way. An undergrad that submitted this as a homework would not have been happy with the resulting grade. Once again, this suggests that the six who reviewed VE’s work for CUP were either not expert in the relevant fields or, just as likely, did not read the book carefully before pronouncing on its merits. These are intro to syntax errors. CUP clearly needs a better vetting process.

As noted at the outset, I intend this to be my last post on VE’s work. I know that I have sounded harsh, but that’s because I wanted to sound harsh as the work VE has produced is really very bad. It is confused, incompetent and more than a tad dishonest.  IMO, this makes it junk in the sense that it teaches us nothing about the vices or virtues of the Chomskyan Generative program. In fact, it’s worse than that: it cannot teach us anything, as the arguments are based on misquotation, equivocation and technical incompetence. This is why I have repeatedly called it ‘junk.’

However, the fact that this work has been published, and apparently endorsed by many luminaries, is insightful. It tells us a lot about the intellectual atmosphere that surrounds us. IMO, most of this junk stems from a resurgent Empiricism, a particularly crude version in fact. For reasons I actually do not well understand, the idea that humans are built to speak is considered provocative. The further possibility that this native capacity is in some ways domain specific is considered by many to be absurd. Both of these reactions to linguistic nativism are, to my mind, nutty. Here’s why.

That we are biologically built for language is a truism. After all, name another physical object or biological being that does what we do linguistically?  There are none, or at least none that we know of.  In short, the claim that we are biologically “language ready” is about as exciting as noting that the sun rises in the East, unless, of course, you don’t believe that our mental capacities are biologically grounded. And nobody believes that! Hence, that we are “language ready,” biologically built for language, is a truism.

How about the second claim; that what makes us language ready relies on language specific powers and capacities rather than more general cognitive wherewithal? This is not a truism if true. But nor is it absurd. It is what we call in the trade “an empirical issue.” If true, then it argues against simplistic tabula rasa theories of mind. This makes it a favorite target of the Empiricistically inclined. However, one of the glories of GG has been to show how this question can be made empirical rather than philosophical. GG has shown how to argue for and against the language specificity of our linguistic powers. Should linguistic competence be reducible to other more general non-linguistic competences, that would be a fascinating discovery. IMO, this prospect is currently very unlikely (though being a card carrying minimalist I hope (and I know that hoping is one thing and showing another) that many linguistic specific FL/UG details can be off-loaded to more general cognitive operations of other third factors).[9] However, I can see how this might be accomplished in principle. This said, I also know what needs to be done to show this; people who think that linguistic competence does not involve a language specific FL need to show how to explain the mass of data that GGers have discovered and analyzed over the last 60 years without invoking language specific concepts and mechanisms. Nothing short of this will serve. The problem with VE’s work (and that of many others who have preached the same sermon about the demise of GG) is that it completely fails to engage with this work. That’s why it’s intellectual junk, albeit sociologically very revealing.


[1] I will even refrain from commenting on VE’s most impressive riposte, namely his threat to sue critics (see here). This indiscrete polemical feint was quietly withdrawn and expunged from the record by VE, sans apology. Given the alternatives, I prefer VE’s current less bellicose strategy of simply avoiding discussion of the main issues.
[2] At least these are the major problems. There are a host of lesser ones that the posts linked to above also review.
[3] VE’s latest post linked to above seems to have trouble with many other conceptual distinctions. For example, VE has a problem distinguishing a program from a theory. Chomsky has always been careful to distinguish the two notions. For example, GB theory was part of the P&P program. At this moment, there are many minimalist theories animated by the leading ideas of the Minimalist Program. Thus, Chomsky’s distinction between program and theory is in no way an empirical retreat in the face of empirical findings, as VE suggests.  It’s just the normal way of conducting scientific  business.
           
[4] Cinque initiated a lively discussion concerning GU 20. And people like David Adger have investigated GU 20 like effects in an artificial language learning setting (as discussed here).
[5] Sadly, it does boost its truthiness, which is precisely what the agnotologically inclined count on. Repetition can serve to provide the appearance of controversiality (i.e. the idea that there are two sides to a question) even when one of the sides is based on equivocation, as in this case where GUs are confused for CUs.
[6] It is worth adding that VE’s work does not discuss the now substantial literature that analyzes GUs from a GG perspective. It appears that some GUs are well grounded empirically. Further, there are some interesting attempts to explain their properties using standard arguments based on UG (and other) principles.  It is unclear why EV’s work does not discuss these given their obvious relevance to his basic claims.
[7] I leave as an exercise for the reader the application of the GU/CU equivocation as applied to Everett’s Piraha discussion. This is the poster child against CUs that is endlessly trotted out. It too suffers from the confusion endemic to VE’s oeuvre. See here and here for more detailed discussion.
[8] I mention this because VE seems to point to the fact that six “world leading experts” blessed the book for CUP before publication. They may have blessed it, but it seems that they didn’t read it, at least not carefully.
[9] Let me mention one more confusion that VE seems to have fallen into. Chomsky has conjectured that most of what earlier theories postulated as linguistic specific features of UG could and should be reinterpreted in less linguistically parochial terms. He as further conjectured that the only linguistically specific operation (and even this may not be proprietary to FL he has at times surmised) is Merge.  It is important to note that this is a conjecture, and that’s what makes the Minimalist Program a program. Whether Chomsky is right, is currently being aggressively investigated. For what it is worth, IMO, we are still far from grounding this conjecture empirically, though we have made some very interesting analytic connections in trying to realize it.  Importantly, for Chomsky to be right requires deriving earlier GG results using these less linguistically specific notions. This has not yet been done. In other words, no current minimalist theory has yet realized the ambitions of the program. Is this a problem? Not obviously. Why not? Because this is just how science works. Let me explain.
Programs are not true or false but fecund or sterile. They are judged fecund to the degree that they engender theories that can be empirically evaluated as true and false. In other words, good science attempts to realize fecund programs in various explanatory empirically grounded theories. Theories are constantly proposed, elaborated, and rejiggered in face of puzzles. This is normal science. VE does not seem to appreciate this (could it be because he is not actually a practitioner?).
Nor does he seem to understand what Kuhn intended by an anomaly. Anomalies are problems that current theory cannot accommodate given its current conceptual apparatus (e.g. black body radiation given classical mechanics) not things that theory appropriately revamped doesn’t happen to cover. The former require deep conceptual revisions (hence the revolution) the former require revamping the details of the current conceptions (hence their being normal science). At any rate, this confusion between program and theory leads VE to say, not surprisingly, several odd things and to misrepresent the state of play in current GG.
           


47 comments:

  1. It looks like you forgot to fill in "GET THE NUMBER OF THESE". Maybe you just read over it and thought it was another one of your acronyms ... :p

    Also, has anybody said anything to CUP? I haven't been following all of this all that closely since it seems like a waste of time. But from what I've ascertained, it seems like you and David Adger have already done a good job of addressing some of the linguistic/conceptual problems. However, it seems to me that the misquotation problem has largely only been addressed in social media. It could be worth putting together a list of all of the examples and turning it into an open letter to CUP, pointing out that they should be embarrassed to have published something that, at best, doesn't meet minimal academic standards and, at worst, is actually dishonest. I also vaguely remember somebody commenting on this blog mentioning that he also got some of the biological facts wrong, too.

    But it might be worth doing something like that and intentionally leaving out the linguistic/conceptual problems, so that people don't see the open letter as some partisan political maneuver. Of course, some people will inevitably see it as that regardless, but at least it would be clear evidence for the people willing to just read through the examples of misquotation and factual inaccuracies in the book that, regardless of any "ideological" motivation, the book doesn't even meet minimal academic standards.

    ReplyDelete
    Replies
    1. Hopefully this doesn't end up outing the source, as Byron's second tweet indicates that the person who noticed this might wish to remain anonymous, but there's also this, which is just sort of comical at this point.

      Delete
    2. Given that things seem to have gotten away from the point of the post in the discussion below, I want to reiterate the question/suggestion about writing an open letter to CUP. I'd be curious to hear other folks' thoughts on this. Is this a bad idea? If it's not a bad idea, is it worth the time, and would folks be willing to sign it?

      Also, I hope you'll forgive me for poking fun at you, Norbert, because of your acronym problem (YAP). But I promise that me poking fun at YAP was just a bit of lighthearted fun. :) I hope you're not offended!

      Delete
  2. Greenberg 1963 listed 45 universals.

    ReplyDelete
    Replies
    1. My reading of Norbert's comment to himself was that he wanted the numbers for word order patterns. According to Russell S. Tomlin's book Basic Word Order: Functional Principles, the numbers (frequencies) seem to be: SOV, 44.78%; SVO 41.79%; VSO, 9.20%; VOS, 2.99%; OVS, 1.24%; OSV, 0%. This is based on a sample of 402 languages. I got the numbers from this (paywalled) review of the book.

      But perhaps you were right and he was talking about the number of Greenberg Universals.

      Delete
    2. Thanks to both of you. I cannot recall what I intended but both seem apposite.

      Delete
  3. "Programs are not true or false but fecund or sterile. They are judged fecund to the degree that they engender theories that can be empirically evaluated as true and false"

    Can someone be so kind and provide a complete list of the theories that have been engendered by the minimalist program up to date?

    ReplyDelete
    Replies
    1. This is by no means a complete list, but offhand there are a few that readily come to mind. One is espoused by Norbert in his (2009) book A Theory of Syntax: Minimal Operations and Universal Grammar, which holds that recursive hierarchical structure building is a result of Label, not Merge. There's also Chomsky's theory espoused in his latest series of lectures that Merge is what gives us recursive hierarchical structure and that there are no labels in the syntax, just at the interfaces. (It's also worth reading his Problems of Projection paper.)

      Ones that I'm less familiar with include Hiroki Narita's book, which, to the extent that I'm familiar with it, takes a similar approach to Chomsky but focuses in particular on a theory of phases, purportedly giving us a better understanding of linearization at PF and reduced computational complexity in the narrow syntax. (I'm currently reading Narita's dissertation and only halfway through it, so don't take this as necessarily an accurate characterization of his work.)

      David Adger also has a theory that eschews the existence of functional heads. Haven't read this yet, so that's about all that I can say about it.

      Cedric Boeckx has a theory of syntax that eschews features. Also haven't read this yet, so I can't say much more about it.

      John Bowers has a theory of argument structure that turns traditional views of argument structure on its head, arguing instead that the external argument is merged first with the verb, rather than the internal argument.

      Also, in terms of phonology, Bridget Samuels has a neat theory of phonological derivation by phase, which essentially eliminates any domain-specific phonological primitives.

      Those are the things that immediately come to mind. No doubt somebody more knowledgeable than me could provide a comprehensive list. As for an exhaustive list, I'm not sure anybody could provide that as it would require a near omniscient knowledge of the literature. But perhaps there is such a person with such knowledge.

      Delete
    2. Thank you Adam, I am looking forward to others adding to your list. Just one more question for you: which of these theories is based on empirical data from brain-research? As Norbert stresses [again] Minimalists work on Chomsky-Universals and these "CUs do not refer to surface linguistic patterns but to properties of the Faculty of Language (FL)". FL, we are told is a biological organ, hence its properties should be biological properties and theories about those properties should be supported by biological research. [This is not a trick question, in case you're not aware of biological research supporting any of the theories you mentioned "I don't know of any" is a perfectly fine answer]. Thanks again.

      Delete
    3. @Christina: "we are told is a biological organ, hence its properties should be biological properties...".

      Notice, the claim about language being viewed as a biological organ is perhaps Chomskyan, but the inference "its properties should be biological properties" is yours.

      Could you give us a list of common properties that all biological organs in nature share? This is not a trick question; maybe, I just want to make sure that we are not stuck on terminological issues. So, if you could help list such properties that all biological organs share in nature, maybe we can make progress in the way that you would like.

      Delete
    4. Ah, I suppose it's also worth adding a semantic theory to the mix. There's Paul Pietroski's theory of semantics, which argues that meanings are built up by the operation of Conjunction, rather than Function Application.

      And your insinuation is probably correct that there isn't any empirical brain-research that these theories are based on. At the very least, I'm fine admitting that "I don't know of any". But I also don't follow much in the neurolinguistics literature because I'm not sure that I find any of the claims and inferences that they make convincing (since I think a lot of the research is premature since we don't have a great understanding of how competence relates to performance). Anyway, point being that perhaps I'm wrong about this.

      That being said, if I am right, I don't think there's any reason to think this is a bad thing. As Norbert notes, the fact that the faculty of language is a biological one is a truism, and I don't see how anybody could think otherwise unless they adopt the position of philosophical dualism.

      And as history of science tells us and as Chomsky/Norbert/Gallistel have repeatedly emphasized, it has been the more fundamental theory that has needed to change, not the theory being reduced to the more fundamental one.

      Delete
    5. Also, Karthik makes a good point.

      Delete
    6. Karthik: I am not trying to avoid answering your question [but Fitch [2010) might be a good place to look for answers - his biology is much more recent than mine]. But we are NOT looking for biological properties shared by other biological organs but for those that make language unique: "“There are in fact crucial features of human language that appear to be isolated in the biological world” (Chomsky, 2013: 35) So it is those isolated crucial features I assume minimalists are looking for...

      That the faculty of language is a biological faculty is only a truism, Adam, if such faculty actually exists. Olivier was so kind to remind us that for hundreds of years astronomers believed in epicycles - it did not follow such epicycles existed outside their best theories. What is at issue in the debate with Vyv is whether or not there is good empirical evidence for a biological LF. Directing the readers of this blog to such [biological] evidence would be a much more productive way for Norbert and David Pesetsky to spend their time vs. worrying whether or not Vyv got some Chomskyan text from 1980 right. [otherwise some Hornstein-Bot or Pesetsky-Bot might show up on this blog, mocking them every time they use the word 'Chomsky' - apparently that is a very bad thing to do]

      This next question is specifically for Norbert who is given permission to break his otherwise much appreciated promise not to talk to me. On p. 29 of SoL NC talks about "Norbert Hornstein's theory of control". There is no entry in the reference section for it. Am I correct to assume that he refers to your "Movement Theory of Control" or does he refer to something else?

      Delete
    7. This comment has been removed by the author.

      Delete
    8. @Christina: A second time in a row, you have avoided answering my questions and instead replied tangentially. Since you are the one who made the inference I pointed out above, and have repeatedly requested lists of such things in the past, I think it is incumbent on you to explain clearly what you mean by such a phrase or such an inference. I don't think giving me a reference to what someone else has said is either useful or constructive.

      Delete
    9. Karthik: I am NOT the one who claims there is a biological LF. So why would it be incumbent on me to provide a list of what it might contain? If the inference I draw from the literature [that a biological organ has biological properties] is incorrect then please tell me why. Here is a relevant quote from John Collins:

      "From the I-language perspective, linguistics is a straightforward biological inquiry, much as if our concern was for the mammalian visual system or the auditory capacity of bats."

      So instead of me making any inference, maybe you can explain: what does John Collins mean by straightforward biological inquiry? Is Norbert's 'theory of control' a part of this straightforward biological inquiry? Norbert claims not to know anything about biology but he certainly DOES know a lot about linguistics. If someone who knows no biology but a lot of linguistics can propose theories, what makes this activity part of straightforward biological inquiry?

      Delete
    10. @Christina: I appreciate that you think I am John Collins' spokesperson, but I would recommend that you ask him directly if you want an answer. The same goes for the question you asked Norbert.

      As far as the biological organ discussion goes: minimally, it is a useful metaphor to me. Chomsky at least has argued for this metaphor, and perhaps the realist extension, in many places. I am not going to do your homework for you if you are not willing to participate in a constructive debate.

      Now, if you don't like the realist extensions, or you don't like the metaphor, I am not one to stop you. It is fine with me. lol! But, if you would like to contribute positively to the field, then I think it would be useful if you could articulate precisely why. And I don't mean more quotes from others or more of what Chomsky said. You have expressed your disagreement simply by saying nothing, and asking question after question. Which is why I ask you (now a third time) what exactly do you mean by "hence its properties should be biological properties..."?

      Delete
    11. Karthik: my apologies: I was not aware that for you talking about a biological [language] organ is merely a useful metaphor and the properties of that metaphorical organ are also metaphorical properties. In that case we talk about very different things and can just agree to disagree. And, if the field you accuse me of not contributing to is also a metaphorical field, then, sorry, I have no interest in contributing to that. I am, however interested in learning about real biological properties of LF that have been discovered by those engaged in the straightforward biological inquiry. If you do not know any maybe you can allow those who do to get a word in. Thank you.

      Delete
    12. So, given your non-answer, I will take it you want answers to questions that contain words that you yourself can't clarify. Interesting...

      And, I accept you apology, though, of course you have blatantly misrepresented my statements above. I simply don't understand your position...it seems so pointless. Perhaps, as pointless as my attempt to engage you in good faith by asking you to participate in a constructive dialogue, though we seem to have very few shared assumptions. Perhaps, the fault is mine.

      Delete
    13. We have already worked through this particular issue before; over a year ago we had a discussion about biological VS abstract properties. It is pretty much a non-issue for anyone who understands what Marr's levels of description are about. Note that this is not some fancy trick linguists whipped up, it is a widely accepted methodology that has proven extremely useful as can bee seen in computer science, for instance, which is only tangentially concerned with the physical devices doing the computing.

      Delete
    14. What I find remarkable in that exchange is not so much the misrepresentation of the position of the interlocutor, nor the requests other people carry on ill-defined research projects or the constant goal-shifting (this is, after all, quite usual). No, I was surprised by the flaunting of general ignorance: a basic knowledge of biology, linguistics and history of science are all at one point or another deemed unnecessary for the conversation to go on (or perhaps even characterized as impediment).

      Anyways...

      Delete
    15. Olivier: the Behme Bot thinks you are talking about others, and not about it.
      (the Behme Bot is also tempted to call you "Oliver")

      Delete
    16. Thank you for the link, Thomas. It illustrates the same problem; there I ask a question that could have been answered with a sentence - instead we spiral into an endless debate about issues not relevant to my question. Here I ask for a list of theories [and what theory of Norbert was referred to in SoL]. Adam was kind enough to provide an answer but then Karthik took it upon himself to derail any further attempts at clarification of the position Norbert states Vyv misunderstands. I have to conclude Karthik wishes this position to remain obscure.

      Of course there is also a disanalogy; back then we talked about Platonism vs. biolinguistics. Here we talk about one position on human biology [Vyv's who provided ample of examples not once touched upon by his critics] and another [the biolinguistic]. The arrival of disembodied Bot clearly signals that it is pointless to engage in any further clarification attempts. I thank Adam for the only genuine answers to my questions.

      Delete
    17. Much as I am sure that Karthick would make an excellent spokesperson, I can't afford his services. Anyway, the quotation from me is perfectly innocent. Inquiry into language is inquiry into a biological system, if we exclude magic or Platonism. It doesn't follow from that that current inquiry needs to be translated into a theory of mitosis, say, to count as biological. The point is to fix the domain of inquiry, not to dictate the means of such inquiry. As an analogy, David Marr produced detailed theories of vision, on the understanding that he was inquiring into a fragment of biology, but all the action in the theory is computational, which in principle could be run on a machine. As a philosopher, I perhaps think in overly abstract terms, but the present point is kinda trivial.

      Delete
    18. Maybe what Christina is looking for is something like Alistair Knott's work. http://mitpress.mit.edu/books/sensorimotor-cognition-and-natural-language-syntax-0

      Delete
    19. Thank you David. i have not read the book but judging by the abstract it is indeed work of the kind I had in mind. Or something like: Di Sciullo, A.M. (in press). On the asymmetry of language and genes, L-words and DNA words. Czech and Slovak Linguistic Review.

      I am merely asking what theories are endorsed by [the majority of] minimalists at this time and which of these theories are supported by/linked to biological research. I have no idea why such inquiry causes so much hostility nor why John sees a need to point out the quotation was 'perfectly innocent' - I had not accused it of anything but used it as an example to illustrate what I am asking for. I did object to the talk of metaphorical biological organs because I think that misrepresents the Chomskyan commitment.

      Delete
    20. I meant that the content of my sentence was innocent, free of any unclarity or conceptual subterfuge; after all, you had asked 'what does John mean?', as if I had said something opaque. Still, I did not take you to be accusing me of anything in particular and did not intend for my reply to be hostile. On the substantial point, I think Grodzinsky's long-standing work might also be cited, although, again, I think the abstract philosophical point stands independently of particular examples.

      Delete
    21. Yes, clearly it is the arrival of the Behme Bot that signals that any meaningful conversation has come to an end.

      Delete
    22. @John: we have been accused of misunderstanding even the most basic points - so i wanted to make sure i do not interpret your sentence in a way you did not intend. I am glad to learn i understood your sentence correctly. However, I misunderstood your use of 'innocent' [as opposite of guilty vs. as meaning clear/unambiguous] - apologies for that. Finally, I did not consider your response in this thread as hostile but only that of Karthik, Bot, Olivier, and Thomas G.

      You all will be pleased to learn that i have decided to stop commenting on this blog. i will post one final comment at the 'important Ewan's message' post because this is a better place to deal with a rather serious issue you [as a group] may want to address. Good bye.

      Delete
    23. Your misinformed vitriol will be missed. (If only for its entertainment value.)

      Delete
  4. Evans is quite the intellectual historian in this post. In a similar spirit, I propose the following history of astronomy, closely modeling the original.

    One of the most basic questions astronomers asked was this: how to account for the movement of celestial bodies? Back in the antiquity there were no good answers to this question, and little to go on. Hence, the idea that bodies were turning on epicycles centered on the Earth was, on the face of it, highly appealing. But the search for “epicycles” keeps turning up nothing. In a previous post in this blog, I observed that the quest for epicycles in the astronomical enterprise has, over the years, been steadily downsized. First, we were told that in fact circles were enough under the hypothesis that it is the Earth that moves around the Sun and not vice-versa (if, like me, you have seen the Sun rise this morning, don’t worry). Then, circles were deemed not enough anymore and everything was supposed to be ellipses. Today, all that remains is the claim for a highly abstract “Law of Universal Gravitation” according to which bodies attract each other. One problem is that various experts disagree as how to define the terms.

    Lucretius, a philosopher and poet with over 30 years experience conducting research, first argued, back in 58 BC, that at least one material, lodestone from Magnesia, fails to exhibit universal attraction: indeed under some circumstances blocks of lodestone repulse iron. For the astronomical enterprise, the real story—and a damning indictment on the failure of the enterprise— is that after over 2500 years of looking, no true epicycles have been found. The last vestige of a once promising project—the claim that what is universal is a highly abstract law of attraction—is all that’s left. And if Lucretius is correct, even this claim appears to be on shaky ground too.

    ReplyDelete
  5. This comment has been removed by the author.

    ReplyDelete
  6. /Behme Bot awakens from its sleep

    /notices the name "Vyvyan Evans" on Norbert's blog

    /thinks we are complaining about her saying "Chomsky"

    /misses point entirely. as usual.

    ReplyDelete
    Replies
    1. Whoever is doing the Behme Bot, it's really not funny, is antagonistic to CB, and is altogether unhelpful.

      Delete
    2. I agree with William on this. I think it's very unpleasant.

      Delete
  7. This comment has been removed by the author.

    ReplyDelete
  8. Can we talk a bit about falsifiability and Chomskyan Universals (CU)?

    Norbert, I would expect you to "bite the bullet" on this and say that they are not straightforwardly falsifiable, particularly given your rejection of naive Popperian falsificationism.
    So take "Another popular one restricts “movement rules” from applying across islands"
    This doesn't seem to be falsifiable in any easy way since the islands are defined to be restrictions on movement rules.
    And the same thing applies to each of the other CUs considered in isolation.
    They are generally informally stated using theoretical terms that are not themselves formally stated, and so it's hard to see how one could refute most of them.


    And I don't understand how one could "show (most ambitiously) that Gs can have arbitrary structure" on the basis of the finite sample of languages that we have available to us.

    (edited to correct a typo)

    ReplyDelete
    Replies
    1. @Alex C:

      >>And the same thing applies to each of the other CUs considered in isolation. They are generally informally stated using theoretical terms that are not themselves formally stated, and so it's hard to see how one could refute most of them.

      Did you mean to say that such claims are tough to refute within the framework of discussion, or do you mean, outside the framework?

      With reference to the former, it seems to me that practicing syntacticians have the description sufficiently formalized (even if it is not completely formalized) to allow for a meaningful debate about such claims/hypotheses. Clearly, they are able to apply a claim from a particular paper/analysis to another case or language.

      With reference to the latter, I am not sure any claim/hypothesis has such potential no matter how formalized it is. Isn’t this the Duhem-Quine thesis?

      Delete
    2. I should say that I find Alex C's question important and legitimate. As an aside, as an outsider from the harder corner of science, I find minimalist syntax appealing precisely because of the the wide range of falsifiable predictions it makes.

      Delete
    3. So there is a sense in which you're right that islands are the movement restrictions, so there is a reasonable fear of circularity. Where the person interested in testing this hypothesis gets a foothold is that while the theory of islands is under revision, the GGer would have to admit defeat in the endeavor if it became clear there was no set of structure dependent rules that could constrain movement out of islands, or that the rule set is idiosyncratic based on the language, not just the what the grammar implements them in that specific language. This gets back the the point made in the OP:

      "...you need to show (most ambitiously) that Gs can have arbitrary structure (i.e. there is no limit to the kinds of rules, operations and structures that human Gs can have)..."

      If the rules are arbitrary, the GGer must be wrong. Admittedly, there is a long, incremental walk between where we are now and arbitrary.

      Delete
    4. There is no danger of circularity. When Ross used “island” as a theoretical term, he gave a precise non-circular definition of it. The term is often used now in a broad descriptive sense, but since no-one is proposing a theory in which “island” is defined as “anything that can’t be moved out of”, there is no invidious circularity. The various theories of island effects that have been proposed all identify island configurations in terms of independent properties.

      Delete
    5. @Alex C:

      I am a delighted bullet biter with one caveat: I think that almost nothing in the non mathematical sciences is "falsifiable" in any strict sense. I think that empirical questions can be tested and shown wanting, but what many mean by falsification, I consider a myth. With this in mind, the relevant concerns are whether the claims are narrowly circular (no, as Alex D noted). So the claims re islands are standard issue empirical claims that can be shown to be wanting though can always be saved with enough ancillary apparatus. That's more or less always true for empirical claims. So, I do bite the bullet but it tastes similar to those one bites in any domain of empirical inquiry.

      Delete
  9. "Can we talk a bit about falsifiability and Chomskyan Universals (CU)?"

    Wouldn't a straightforward falsification of the central thesis of minimalism be a language whose syntactic rules were independent from the binary hierarchical structures minimalism posits? Say a language for which cataphoric pronouns systematically differ from anaphoric pronouns, or for which negation is licensed by a fixed position in the linear order except the last (and possibly fist, people more knowledgeable than me can tell me if this is compatible), or which has reflexives whose licensing is entirely pragmatic, or with wh-movement to a position in the middle of the sentence, or with wh-in situ with no intervention effect, or a V3 language in the same sense that German is V2, or which has well-separated classes of verbs and nouns yet the licensing of noun complements of those would be identical etc. etc.

    "And the same thing applies to each of the other CUs considered in isolation."

    I should say that I also reject naïve Popperian falsificationism, but only in the sense that I think people should actually read what Popper wrote. For instance, the actual Popper understood very well the distinction between falsifiable predictions, accounting for problematic data and immunization. Applied to linguistics, it seems to me that the straightforward popperian formulation is under the guise of correlation: so it is not so much that you expect to describe complete precise syntactic rules for English, or abstract general rules that hold for every language; you expect to find correlations (so for instance you certainly don't expect successive cyclic movement to be visible in every language, nor do you, at present, expect to be able to conclusively characterize the lower phase of all languages, or even of a given one, but you do expect that the tests identifying the lower phase will also predict where successive cyclic movement is visible when it is as well as where quantifiers or clitics might strand when they do etc.). So yes, taken in isolation and considering the primitive status of linguistics, you should expect to find a lot of easy falsification. Falsifying the predicted correlations is in principle possible and would be serious (and very interesting) business.

    ReplyDelete
  10. I am not sure I know what you mean by syntactic rules here, Olivier.
    But I think the general problem with that type of example is that one of the diagnostics for syntactic structure is the operation of say movement rules.
    In general we don't have a reliable cross linguistic definition of what the syntactic structure is. Rather the syntactic structure is a theoretical posit that is used to explain among other things the restrictions on certain types of movement (and lots of other things as well!).

    So if we had a language that really did have some very strange movement rules,
    what I think would happen would be that lots of syntacticians would write lots of papers proposing strange syntactic structures for this language so that the rules
    respected the sorts of constraints that linguists have hypothesized to be universal.
    Indeed some of the most interesting languages for syntacticians seem to be precisely these ones, which
    superficially appear to violate some widely held generalizations. The intellectual game then is to show that these purported counterexamples are not in fact counterexamples. I don't quite know what the best example would be but maybe Icelandic.

    I don't think that this is necessarily a problem for the regular Quine-Duhem reasons.
    But it is useful to know whether or not a particular CU can be falsified independently of a whole bunch of other assumptions.
    My take is generally that CUs (of the relative clauses are islands type variety) are actually harder to falsify than they appear at first glance. Naively you can just say, well, if you see a language where relative clauses aren't islands then that is a counterexample. But maybe what you think is a relative clause isn't or maybe there is something which isn't really movement etc. So of course these have to be, as Thomas G reminded me once, independently motivated, but the methodological point still stands.
    Again this is fine, but but but if you have a whole bunch of different CUs, and each one has to be waggled a bit to accomodate some recalcitrant data, you may end up in a situation where you have a theory that looks restrictive but in fact when formalized ends up being equivalent to some huge class of grammars -- MCFGs or RCGs or Turing machines or whatever. That should be worrying.



    ReplyDelete
    Replies
    1. Alex: I think a distinction needs to be made, in the spirit of Popper, between a hypothesis being falsifiable and what to do when confronted with apparent falsifying phenomena. All of the relevant claims are falsifiable in the first sense simply because they have empirical content, they allow for predictions. On the other hand, what the appropriate response is to apparently falsifying phenomena strikes me as undecidable a priori. I suppose one has to make a judgement on a case by case basis about whether the phenomena are genuinely falsifying or are to be accommodated by a tweak, which ought not to be ad hoc, etc. Still, I don't quite see the worry you entertain. Standard MP approaches are highly restrictive relative to what language could be like (binary branching, restrictions on extraction, no vacuous quantification, structure dependence, etc.). If, within such parameters, one found a class of formally (extensionally) equivalent grammars, that wouldn't refute the approach but only enjoin the theorist to tease out what empirical differences there might be within the class. Only something approaching Turing equivalence would show a lack of serious restrictions.

      Delete
    2. "The intellectual game then is to show that these purported counterexamples are not in fact counterexamples. I don't quite know what the best example would be but maybe Icelandic. "

      But this is exactly where one should distinguish pop-Popper from real Popper (not to mention Quine-Duhem). If a counterexample is found, it is perfectly allowed in standard popperian epistemology to posit a strange syntactic structure to account for it provided, and that's the crucial point, the existence of this strange syntactic structure itself entails new falsifiable predictions (this is why the real Popper, not his caricature, thought falsification was helpful in the demarcation problem, but was not helpful to discriminate between scientific theories: all non-tautological scientific theories are always badly falsified at any given point; however some have stronger epistemological content than others).

      To come back to linguistics, a paradigmatic case study would be binding theory (I'm going to restrict to simple transitive sentences). As far as I know, Tanya Reinhart proposed the hypothesis that a 3rd person pronoun could not refer to a c-commanding antecedent. So far so good, but Frisian is a straightforward falsification. Then Eric Reuland proposed that the real rule is that a 3rd person pronoun at a bottom of chain could not refer to the top of the chain, where links of the chain are (in particular) given by structural case assignment and hypothesized that some transitive verbs in Frisian do not assign structural case. OK, strange structure posited, phenomenon salvaged (but epistemological content of the theory unchanged for the moment, or even arguably diminished: one new phenomenon has been explained but at least one and arguable several new hypotheses have been made). However, our new hypotheses have empirical entailments of their own: for instance, they straightforwardly entail that Frisian verbs that allow coreference behave differently in abstract morphology, and because the only access we have to abstract morphology is real morphology, we expect that they differ in real morphology. Well, do they? Also, the reformulation now allows for the possibility that a reflexive would be licensed without a c-commanding antecedent provided that the head assigning structural case c-commands both the antecedent and the reflexive. Well, is this attested? If the answer to one or both of these questions is no, then Redland's suggestion has to be discarded, not because it has been falsified (everything is always falsified) but because the epistemological content is now clearly weaker than Reinhart's proposal. But if the answers are both yes, then the epistemological content of Reuland's proposal is clearly greater and so progress has been made (though we should definitely expect his proposal to be badly falsified by other problematic data).

      One last thing: by his own account, Reuland came to construct his proposal because he noticed that if narrow syntactic objects were constructed by Merge and probed by functional heads (two standard core assumptions of minimalism), then c-command should play a crucial role (as it is the relevant structural relation in binary trees) but should not be the whole story. So it is not that core assumptions about CU could have really been straightforwardly falsified by such investigations, but they nevertheless played a crucial role at every step.

      All in all, a routine story about ordinary science, it seems to me.

      Delete
    3. For what it is worth; John C and Olivier have expressed what I would have said had I had the eloquence that they have. The general point is that it is very hard in the natural sciences to definitively prove anything refuted. This is no less true in linguistics than it is in physics or biology. From my recollections, it may even be true in math (see Lakatos's Proofs and Refutations). At any rate, so long as the reasoning does not get viciously circular, the process is as innocent in linguistics as it is anywhere else. Of course, this has to be done well, but again this is true for all domains of inquiry. So, I do not see the methodological problem, or at least not one specific to syntax and the practice of syntacticians.

      Delete