Monday, March 4, 2013

The Biolinguistics Program


Programs can be evaluated along many dimensions. They can be fecund or sterile, cramped or expansive. All look forward not back.  They are more like commencement addresses and promissory notes than eulogies or audits. I personally am a fan of the biolinguistic program for it strikes me as intellectually expansive and potentially fecund.  I am not a big fan of Platonist linguistics for it strikes me as conceptually cramped and conservative.  Let me explain.

When I first got into linguistics, the slogan was that linguistics was a sub-branch of cognitive psychology (CP). Chomsky’s review of Skinner, chapter 1 of Aspects and the elaborate philosophical discussion in Reflections on Language set the tone for the enterprise.  Linguists were studying a mental faculty in much the way that perception psychologists were studying vision and audition.  The analogies were evident: our data consisted in asking native speaker’s acceptability judgments. Our problem was to figure out how it was that humans when exposed to a pretty meager subset of the data were able to project a competence with apparently unbounded range.  This latter, the “Projection Problem,” was idealized in the famous picture in (1):

(1)  PLDL à UG à GL

Given linguistic input (aka: primary linguistic data) L UG names the to be identified function that takes a Language Acquisition Device (LAD, aka, kid) to a grammar of L. UG is thus the general recipe for taking kids from PLD to G.  Given the nature of the projection problem, G was understood to be a set of recursive rules and so the problem was to identify what took you from outputs of G, actual uttered linguistic productions like John loves Bill, to the rules that generate the patterns in those productions. The project was complex for to study UG required knowing something about G and to study G required studying judgment patterns of native speakers. Complex, yet methodologically rather straightforward, or so it seemed to me then and now.

It is easy enough to see why this conception of the enterprise caught on so quickly. First, it responded to a deeply held philosophical intuition, viz. that language was a window on the mind. By studying the fine structure of linguistic competence it was possible to identify essential powers of human minds. Second, despite the evident idealizations (recall, Aspects bid us study the ideal speaker-hearer!), given the actual practice it sure looked like we were studying a human capacity. After all, we were plumbing speaker intuitions, making predictions about unqueried judgments, and making models, in general, of what native speakers would find agreeable or not. Third, situating linguistics in a CP context served to stretch the kinds of questions linguists should be interested in.

For example, the CP context encouraged looking for ways to study linguistic behavior using other methods. Thus, there was a burgeoning effort to apply psycho techniques to the study of linguistic behavior. Aside from the early work by Chomsky and Miller, early psycho-ling techniques were also developed, J&J Fodor, Bever, Garrett, Forster, Gleitman, Newport, Crain etc. This work took linguistics seriously enough to investigate it using standard psycho techniques and found some novel effects which have proved to be of lasting interest (think Garden Path or Filled Gap Effects). Some of this proved somewhat distracting (I am thinking of the psychological reality debates) but much of it served to develop techniques that are still in use today to investigate how Gs are used to parse and produce language. Indeed, the early debates around the derivational theory of complexity are the first attempts to see how transparent parsers and grammars are, a topic that Marcus, and Berwick and Weinberg and Phillips have investigated with surprising results.

Second, linguistics was the model science for the whole CP revolution. Linguistics made legitimate the search for native bases for many other cognitive competences: how we object permanence, causality, face recognition etc. 

Thus, locating linguistics in the broader CP world was good both for linguistics (by making it responsive to the concerns of neighboring domains) and good for CP.  Speaking somewhat personally, the ling dept at UMD was constructed on the premise that this kind of CP orientation was both interesting and productive.  From the get-go the department was built to implement the CP vision. The idea was that the distinction between the theoretical areas (e.g. syntax, phonology, semantics, morphology) and the conjunctive ones (e.g. acquisition, processing, computational, neuro) was less one of centrality and more on of alternative platforms.  Different methods can be used to study related problems. The common object of investigation was the structure and properties of FL. Given the diversity of the behavioral effects that linguistic competence gives rise to, it is to be expected that different techniques will be used to investigate it.  At any rate, the unifying theme is that we were all investigating the same mental organ using different methods and tools. In my view, this has proven to be a very fecund approach.  In particular, the CP perspective treats language as a natural phenomenon.  It aims to limn the underlying powers undergirding complex behavior.  Locating linguistics within CP means displacing it (at least somewhat) from a more general cultural setting, or a more philological context. On the CP view, the aim is to find the cognitive mechanism behind the overt behavior. One studies linguistic properties as way stations to these cognitive powers. This is to take a resolutely expansive view of the enterprise, a very good thing in my view.

However, not everyone likes this expansive view for it threatens to challenge and/or displace traditional techniques.  In my view, Platonism can serve as a safe haven from the hurly burly of this CP conception. Locating linguistics within mathematics insulates it from the empirical concerns of CP.  Here’s the comforting soliloquy: “What they find cannot in principle bear on what we find. After all, just as physics is incapable of upsetting mathematical truth so too work in brains, parsing, acquisition cannot bear on linguistic truth.”  This is a cramped view.  Why would anyone choose to endorse it?

So, in my view, the CP setting for linguistics has been extremely fecund and intellectually expansive. What of the biolinguistics vision?  I can see no problem with that either.  The setting is a natural extension of the CP conception on the common understanding (more a modern fashionable prejudice actually, but count me in) that minds are secreted by brains.  In other words, unless one is a dualist, not a widely endorsed position nowadays (I heard someone identify it with believing in ghosts), there is a common assumption that mental capacity supervenes on brain structure.  Nobody really knows how this works, but it is assumed to be true.  If linguistics is a part of CP then linguistic competence too lives in brains. How? Don’t know. But there is a lot of work out there, some of it quite interesting, trying to find out. However, this work only makes sense if one believes that linguistic competence ultimately has a biological basis. Again, Linguistic Platonism questions the rational basis of this sort of inquiry.  If grammars are not concretized in brains then looking for the brain bases of grammatical competence is a fool’s errand.  It’s like looking for the number three in a haystack!

There are other examples that make analogous points: comparing bird song and language structure, studying FOXP2 and its effects on other species, trying to figure out why various brain insults have the effects they do, the whole brain mapping industry, all of these efforts make no sense unless language is viewed in a biolinguistic setting.  Have the results been terrific to date? Not in my view. Is the perspective therefore useless? Not at all.  More so than minimalism, biolinguistics is a program, a vision of language study and its fit in the wider scientific enterprise.  It is possible to study language and its structure from a narrow perspective.  In fact, for many questions and issues this perspective will be perfectly adequate (one reason that linguists tend not to care about the issue).  However, it is hard to see what one gains by so restricting oneself to this perspective.  It is possible to insulate one’s investigations from these wider concerns, but why would you want to?

So, for me the Chomsky revolution in generative grammar started by locating linguistics among the cognitive sciences. Embedding CP within biology more generally is the natural next step.  It is hoped that over time, just as taking linguistics as part of CP proved invigorating for linguistic investigation, so too will looking for the biological roots of linguistic competence. At any rate, the apparent push back against the biolinguistic vision is very hard for me to understand or sympathize with. What are the “critics” afraid of? Too much excitement?

9 comments:

  1. This is a potentially very interesting and relevant topic: why should we prefer biolinguistics over linguistic Platonism. So, naturally, I was a bit disappointed to find just a contemporarily dressed version of Jerry Fodor's 1981 [Some notes what linguistics is about] argument for why we should prefer the Right View [his capitals - though given his almost pathological modesty that hardly needs saying] over Platonism: "Platonism isn't incompatible with the Right View ... just deep down nobody is remotely interested in it. "Go ahead be a Platonist if you like. But the action is all at the other end of town". So there are no NEW arguments for preferring biolinguistics then? That seems a bit sobering. Now maybe you can answer a couple of questions to those who are not already on board:

    1. After roughly 30 years of action and fecundity, what are some of the results proponents of the Right View/biolinguistics have achieved that were not achievable by a Platonist? You seem to say the difference is here:

    "On the CP view, the aim is to find the cognitive mechanism behind the overt behavior. One studies linguistic properties as way stations to these cognitive powers."

    So can you name a couple of cognitive mechanisms that have been found based on the study of linguistic properties? Many of the actual results that for example David was referring to in his 2013 LSA keynote seemed to concern linguistic properties. So an example of one of those having lead to the discovery of a specific cognitive mechanism would be great.

    2. You write: "Locating linguistics within mathematics insulates it from the empirical concerns of CP. Here’s the comforting soliloquy: “What they find cannot in principle bear on what we find. After all, just as physics is incapable of upsetting mathematical truth so too work in brains, parsing, acquisition cannot bear on linguistic truth.” This is a cramped view. Why would anyone choose to endorse it?"

    I can't speak to the why question but want to make you aware of a couple of inaccuracies:
    1. Platonists do not locate linguistics WITHIN mathematics. Maybe you only read Katz' early work where he said such but in his later work no such claim is made and Postal certainly doesn't endorse it.
    2. It is not clear what you mean by 'linguistic truth'. If these are the kind of things Katz discusses it is not clear HOW work in brains or on acquisition could bear on them even on your view? Will modus ponens change depending on any acquisition data we may uncover? If by linguistic truth you mean facts about acquisition, then Platonists don't deny that work on brains is relevant to them [they specifically state so Katz&Postal, 1991 I really recommend reading it: http://ling.auf.net/lingbuzz/001607 ]

    3. You say: "biolinguistics is a program, a vision of language study and its fit in the wider scientific enterprise. It is possible to study language and its structure from a narrow perspective. In fact, for many questions and issues this perspective will be perfectly adequate (one reason that linguists tend not to care about the issue). However, it is hard to see what one gains by so restricting oneself to this perspective. It is possible to insulate one’s investigations from these wider concerns, but why would you want to?'

    It is not quite clear to me why focussing on a narrow area of research [that an individual can actually accomplish] has to be equated with 'insulating one's view from these wider concerns'. Why is it not like working on one piece of the puzzle that will eventually lead to a unified picture of the cognitive machinery involved in language acquisition and use. Just as others [say geneticists, neurophysilogists, etc.] focus on other pieces of the puzzle. It seems without such focus it would not be possible to discover phenomena that make David question the adequacy of Tomasello's account.
    [next point in new box]

    ReplyDelete
    Replies
    1. This is the final point which did not fit in the first box:

      4. about excitement [or lack thereof]: this is taken from Paul Postal's NYU page:

      "... the frontiers for someone interested in the mysteries of natural language are quite available everywhere; many are accessible even without leaving one's home. When one adds to the mysterious and uncharted nature of natural language the fact that language (and hence, linguistics) relates closely to a host of other areas, logic, formal studies, computer science, psychology and cognitive studies, the physics of sound, anthropology, to name only a few, this is a field which offers an immense range of exciting possibilities to students seeking a career in research and teaching."

      This potential for research and teaching seems quite exciting for my taste - so where do you disagree? Where is the cramping in this vision?

      Delete
    2. Clearly one is free to study whatever one chooses, in whatever manner one chooses. However, it seems to me the platonist has some tough sledding ahead in regard to delineating the object of study:

      What's natural about 'natural languages' if one cannot appeal to biology to delineate this? The Biolinguistics programme is clear in this regard.

      But if 'languages' are real, abstract objects what makes 'human language' a natural kind for study?

      Does the platonist linguist exclude or at least distinguish: the language of flowers, the language of neo-classical architecture, the languages admitted by monoids in a 3-category, the language of punishment, .. from 'human language'?

      And if so, how?

      Delete
    3. Interesting challenges. Since I am not a Platonist myself I can only guess how they would distinguish the language of flowers, the language of neo-classical architecture, or, the language of punishment from human language. Possibly the same way as Platonist mathematicians distinguish: the mathematics of flowers, the mathematics of neo-classical architecture, the mathematics of punishment from the boring kind of mathematics humans use? Maybe they would not care to draw such distinctions because humans brains determine what languages humans can learn [remember on the Platonist view there is a difference between language and knowledge of language.] Or they would ask you if there are any reasons to assume a human could not learn 'the language of flowers'. Are there? Are there currently any biolinguistic flower language studies establishing these differences? - As I said these are guesses, don't take my word for it but ask a Platonist. Or have a look at Paul Postal's linguistic publications. They suggest to me that whatever he's doing to distinguish human language from those others you list it works pretty darn well...

      Now you ask what makes 'human language' a natural kind for study? This is an excellent question. But before you give the obvious answer: a human language is a language acquired and used by a human' consider the following. [I am borrowing here one of Chomsky's intuition pumps]. One fine day a spaceship lands at the MIT front lawn. A group of Martian scientists emerges and one of them addresses the perplexed security staff in perfect [Bostonian accented] English explaining he wishes to work with the tireless popularizer of Martian science Noam Chomsky. And so they found an interplanetary research team. Imagine a couple years later one of the Martians dies of old age and when an autopsy is performed we find that his brain is fundamentally different from a human brain. Most importantly it does not have any structures implicated in human language. And when questioned the visitors reveal that the learned English by overhearing 'the Boston telephone exchange' [as Chomsky reader you know what I refer to]. Now the $100,000 question: Is the language spoken by the Martians one that is indistinguishable from the English of a person who grew up in Boston] a human language? I am not really interested in which answer you prefer but it seems 'natural kind' is maybe not quite as natural as one might think...

      Now that we had a bit of fun [so badly missing from cramped Platonist linguistics] maybe you can answer a couple of the questions I have posed. Most importantly: what ARE some of the cognitive mechanisms that have been found based on the biolinguistic study of linguistic properties?

      Delete
    4. I never understand what you are asking C, but at the risk of being pedantic (something I like actually) we have discovered transformations, binding islands etc. These involve rule types with certain properties. We have also discovered how grammar can allow us to bootstrap vocabulary (Gleitman et al) we have discovered that there is a good deal of transparency between operations of the grammar and the parser (Berwick, Phillips a.o.), we have discovered the parsing advantages of bounded left contexts which accompanies subjacency style accounts of syntactic generation, we have found that there are some generalizations that even very young kids are loathe to make (Crain and Nakayama), we have discovered that there is a double dissociation between linguistic capacity and other cognitive facilities, we have discovered that structure dependence is a central property of grammars. Now, I am sure that none of this will impress you. The cognitive mechanism that linguists have discovered are largely, well, linguistic. However, their interaction with other cognitive operations is getting a lot of interesting play and is shedding light on how linguistic knowledge is put to use. But, as I said, I doubt that this will satisfy you.

      Delete
    5. Thanks for this. I must have been unclear in my question. I was not asking whether bioliguistcs has ever discovered anything. Had i done that some items on your list might have indeed impressed me [and i am not sure why anyone would doubt that].

      My question was why someone [say my daughter thinking about studying linguistics] should prefer biolinguistics over linguistic Platonism. If I tell her because Platonism is 'conceptually cramped' and biolinguistics is 'intellectually expansive and potentially fecund' she may ask what that means [you know how teenagers are]. So I tell her that as biolinguist she can do stuff a Platonist can't do [or doesn't want to do - lets not get hung up on irrelevant details here]. She asks; Like what? So I was asking you to answer this "like what' question by providing a few examples of things biolinguists have discovered that Platonists could not have discovered [nothing like results to impress a teenager, right?]

      Now given that Platonism was first proposed in the early 1980s and is based on the premise that conceptualism [the metaphysical foundation for biolinguistics] is too limiting to account for all phenomena of natural languages, discoveries made by Chomsky before 1980 [great as these may be] can't be discoveries no Platonist could have made. [For the benefit of those in the audience who may not know the 'ancient history' of generative grammar: Paul Postal used to be a major proponent of it in the 1960s and his innovative work certainly helped to shape the field.] So the discovery of transformations or that structure dependence is a central property of grammars can't be what sets biolinguistics apart from Platonism. Same for much of the other discoveries you mention.

      Also, I am not interested here to question if all of these are 'legitimate discoveries' as some critics have done [e.g. providing evidence suggesting that 'binding islands' are purely processing effects (i.e.,t the grammar/parser 'transparency' exists because the grammar has nothing to do with it (Hofmeister et al.)) These kinds of disputes obviously exist in any field of active inquiry and while they might be a nice topic for a different blog they don't interest me right now.

      Now, lets look at a few candidates that might answer my daughter's question. You say "We have also discovered how grammar can allow us to bootstrap vocabulary" That sounds promising except in his Feb 26 post Cedric Boeckx writes [about Carey] "To "bootstrap" means, literally, to pull oneself up by one's own bootstraps---something that is clearly impossible." Well, if it's impossible, then, it's impossible." So maybe this was then not a 'discovery' but rather employing a metaphor? I am also sure you know that there is by now good evidence from psychology showing that the 'double dissociation between linguistic capacity and other cognitive facilities' can and does come apart - so 'discovery' may not be the right term here as well.

      In the last part of your reply you say: "The cognitive mechanism that linguists have discovered are largely, well, linguistic. However, their interaction with other cognitive operations is getting a lot of interesting play and is shedding light on how linguistic knowledge is put to use." Now this is EXACTLY the kind of thing I was asking for [so i have no idea why you would think talking about it would not satisfy me]. So what ARE some of the interactions of linguistic mechanisms with other cognitive operations [can you at least name some of these other operations - the teenager thing again]? And especially the part about how recent work has been shedding light on how linguistic knowledge is put to use - details about this certainly will satisfy me.

      Delete
    6. I wasn't singling you out, CB. I wonder whether anyone knows of platonist arguments for delineating NL as object of study. I went back (amazing the lengths one will go to to avoid refereeing duties) and looked at 3 Postal (including the one with Katz) papers, without success. These papers are basically attacks on Chomsky's naturalism, not defenses of platonism or realism as far as I can tell.

      Meanwhile, I do not think your platonist mathematics analogy quite holds. [separate message later]

      Delete
    7. As I said my analogy was a guess but I asked a platonist. Here is part of what he replied:
      "What is natural about natural languages is that people, natural objects, develop knowledge of them. We are interested in those collections of sentences which humans develop knowledge of, and then we hope to and sometimes do generalize to collections which various extraneous considerations would actually prevent people from learning, e.g. if every morpheme were two million phonemes long. I have
      never heard of a phonologist imposing a length limit on morphemes...but the trivial fact that there is no bound on morpheme length immediately shows that there are unlearnable languages and thus that the study of languages cannot be the study of linguistic knowledge.

      I hope this helps. You are correct that the papers are not intended as defence of Platonism. Such can be found in the books published by Katz since 1981.

      Delete
    8. Do we mean to say:

      PL studies the languages that humans can have knowledge of?

      That can't be right.

      Anyway, I do think there is a basic difference between math and science.
      Mathematicians start from definitions and explore their consequences; scientists start with (natural) phenomena, seeking explanations.

      In math, one is typically studying species of entities which have been delimited in some way. E.g., in group theory, groups are *ANYTHING* that can be characterized by a handful of axioms (identity, closed, associative multiplication, inverses). Mathematicians ask:
      Is X a group?
      How do we classify groups meeting the following additional conditions C?
      Similarly one might study : context free grammars, various systems of non-linear differential equations, topological spaces, probability theory, ...

      Mathematicians start with the rules and see where it takes them.
      Whereas scientists seek to discover the rules and principles for some set of phenomena/entities of interest. This doesn't mean mathematicians don't play with the rules, try alternate formulations, compare alternate definitions, look for analogies and insights from other parts of mathematics (or even insights from the sciences) -- I don't want to diminish the fact that much of what scientists and
      mathematicians do is very similar work, it seems there is this basic difference.

      So I would not expect a mathematician to tackle 'the mathematics of flowers' as an object of study, but a biologist might appeal to math (e.g., fractals) to attempt to understand principles of the growth or morphology of flowers.

      Anyway, this is all an aside to the main topics of this blog, so I will go back to reading latest posting!

      Delete