Thursday, January 17, 2013

A Teachable Moment

There is an interesting discussion going on in the comments section of the “Darwin’s Problem” post (here).  In fact, I believe that it offers a useful teachable moment, one that I would like to exploit.  Doing so requires stepping back from the argument and considering its logic; what premises drive what argumentative moves to what conclusions and why.  I believe that the revealed dialectic is instructive for it shows that the “debate,” such as it is, is less about theoretically and empirically difficult substantive issues than about a certain skepticism about the facts.  Let me explain.[1]

Step 1: For a linguist, Darwin’s Problem starts with the identification of the explanadum (that which needs explaining).[2] And qua linguist I have a pretty good idea what that is, viz. the structure of UG.  What I want to know is how UG arose in the species, i.e. how it got there? Moreover, qua linguist, I am confident that I have a pretty good idea what UG looks like. I may not know everything, but I know a goodly amount about UGs basic features. Standard versions of GB theory provide a reasonable compendium of these. Though for current purposes, I would not be parochial. If you prefer alternative accounts, GPSG, LFG, HPSG, RG etc. that’s fine. For present purposes these all make the same relevant distinctions, viz. they have roughly the same hierarchy and locality licensing conditions on movement and construal, the same basic endocentric phrase structure, the same basic ideas about crossover etc.  In other words, broadly speaking, most grammarians agree about the general lay of the grammatical land.

Why such consensus? Because the facts dictate it.  Linguists have done a lot of work over the last 60 years and looked at a lot of languages and the general conclusion is that all grammars embody these properties (all specific Gs conform to UG). Thus, Island effects are real, as are Principle A, B and C effects, Weak and Strong Cross Over effects, Fixed Subject effects, Tensed S effects etc.  Theories differ not on the whether but on the how/why. So, for linguists what the basic features of grammars are is not really a wide open question and that’s why Darwin’s Problem for a generative linguist, starts with a description of the target of explanation being something like a UG as described along roughly GB lines.

Step 2: Given this, Darwin’s problem is to explain how it arose. Now other additional facts serve as added boundary conditions. These are not facts that linguists have discovered, but ones found by anthropologists and biologists. These include those that the BFCB paper discuss; the two biggies being that language arose in the species about 100,000 years ago, that whatever arose was fixed in the species before the move out of Africa about 80,000 years ago.  Another biggie is that what humans do linguistically is unique. No other animals (no fish, no fowl, no beasts, no bugs) do it even remotely like we do it. These three facts require that whatever took place was effectively unique and happened very rapidly as measured on evolutionary time scales. 

Step 3: Combining steps 1 and 2 we have the following problem: explain how something with roughly the properties of GB could have arisen in about 20,000 years about 100,000 years ago. That’s Darwin’s Problem for a linguist.  Note, if you think that this arose through natural selection (NS) by modifying some non-linguistic cognitive capacities other animals have/had then the job is to show how something with roughly the properties of GB could have arisen from these other capacities by the sort of modifications NS favors in the time period identified. So, if e.g. you think that GB arose by modification of the capacity to plan as exhibited in non-linguistic animals then starting with a description of planning capacities show how NS would have modified these in response to specified environmental “problems” to get something with roughly the properties of GB, viz. it’s kind of locality conditions, it’s kind of phrase structure rules, it’s kind of hierarchical dependencies, it’s kinds of displacement, etc.  If this can be done, great.  Really that would be terrific. What is not great and very unterrific (at least given these background conditions) is to start from the assumption that we really don’t know much about UG for this does not answer the question posed but answers a different question. Again, let me elaborate.

Many people reject the question posed above because they already know what kind of explanation is required. So, many think that NS is the only viable evolutionary explanation for complex capacities.  And, because NS accounts have had (at least till now) no notable success in accounting for the emergence of something with the GBish properties of UG, they conclude that language cannot have these properties.[3] In other words, they deny what I would call, simply, the facts. For my money, this makes such individuals the linguistic equivalents of climate change deniers. There are some out there that either deny that the climate is getting warmer and/or that human activity is a prime cause of this warming.  Similarly, there are those that deny that language has hierarchical structure, is governed by structure dependent locality conditions and has structure dependent processes.  Such people do exist. From where I sit, they have nothing to add to the present discussion for the same reason that climate change deniers have nothing to add to the climate change discussions. To play the game, you need to acknowledge the facts. And that’s where the minimalist program is so interesting: it tries to provide a scenario that respects these facts. It includes a mutation or two (aka “miracles”), because there appears to be no way to get to what we know to be the properties of UG without them. It tries to unify the various GB modules for this reduces the number of “miracles” needed.  If you can dispense with these assumptions and get to the same place (i.e. show how a GBish UG could arise), that would be really wonderful. There would be no objection from generativists. I’ll personally plan a little parade in your honor.  But (you knew this was coming didn’t you?) no GBish UG as end product, then no parade.  No Rosie Ruizing here: you win the prize for running the course not for taking the bus.

Two last points:

First, as BFCB rightly emphasize, aside from the grammatical uniqueness of humans, we also have a second distinctive linguistic quality: our cognitive atoms (words, basic concepts) are entirely unlike those we find in other animals.  The uniqueness of the lexicon thus also requires explanation and is thus more grist for the linguistic version of Darwin’s Problem.  Once again, this difference between our basic concepts and those of our furry, scaly and feathered friends is an empirical conclusion that seems to have more than a wee bit of evidence sitting behind it. For a good accessible review see Spelke’s discussion in "What makes us smart" (here). She notes that there is little evidence that humans differ from other mammals in what she dubs “core knowledge.” We differ in being speakers of natural langauges. This unique capacity, she notes, allows for the emergence of a qualitatively different set of cognitive capacities when added to the core competencies that we share with other mammals.  So, there are two problems to address, grammar is unique and so are the “words”/concepts we deploy. It would be nice to show that these belong together, but there is currently no good story of how this might be so.

Second, it’s taken for granted that all the discussion is being carried on at a level of abstraction that may compromise the whole enterprise. The abstraction is doubly removed from the real evolutionary playing field. First, The speculations are conducted at the level of minds, not brains nor genomes. Genomes build brains and brains have mental properties. Evolution, via tinkering with genes must have resulted in some new feature of brains ( leading to, e.g. some rewiring or the addition of a new kind of circuit), which led to novel mental capacities. The stories I’ve been offering assume that a small cognitive/mental addition (e.g. merge or label) corresponds to a small brain change which corresponds to a small genetic change. This, of course, need not be true. From what I have been told (thanks Randy, David P), we really don’t have many good ideas concerning the brain correlates of our cognitive primitives (this will be the subject of a future post) in any domain, nor, I suspect do we really know how genes grow brains. Indeed, Randy Gallistel believes that neuroscience as currently pursued is heading off in the wrong direction if the aim is to understand how to brainly embody the right cognitive operations.  So, given this, I have been making the (no doubt stupid? heroic?) assumption that simple cognitive circuits correspond to simple brain circuits that correspond to simple genetic changes.  I doubt that this is true, but right now it is the best we can do and is a game that, I believe, everyone is playing. 

To conclude: Arguments are interesting where there are a sufficient number of shared premises.  The premises that make considering Darwin’s Problem interesting for a linguist include a description of the explanadum as roughly having the properties as described by GB or its kissing cousins.  Absent these, there is no debate or even discussion, simply a denial of the relevant facts. There is nothing one can do about climate change deniers (there is no law against denying climate change, nor should there be), and there is nothing one can do about UG deniers. However, there is no interesting debate with deniers of either kind. The sad fact is that some debates will be interesting, some not. It pays to know which are which.  

[1] For not the first time, Paul Pietroski made these points ahead of me and better. But since I had already written these up I thought I’d post them anyway.  Never too much of a good thing, in my view.
[2] Berwick, Friederici, Chomsky and Bolhuis make this point and emphasize its importance to the logic of the argument.
[3] NS enthusiasts are not alone in this move. It’s a game favored by neuroscientists as well, especially those impressed with connectionism and its attendant associationsim (see Gallistel & King for extensive discussion).


  1. An analogy that Chomsky has offered more than once: 19th century chemistry wasn't bad; and if a 19th century physicist had insisted that the chemists needed to get with the program established by the "more basic" science, then such a physicist would have gotten the theory he deserved--namely, one that made a mystery of stable molecules and the periodic table.

  2. So does Boeckx count as a climate change denier?
    e.g. this quote "I think that minimalist guidelines
    suggest an architecture of grammar that is more plausible biologically speaking that a
    fully specified, highly specific UG – especially considering the very little time nature had
    to evolve this remarkable ability that defines our species. If syntax is at the heart of what
    had to evolve de novo, syntactic parameters would have to have been part of this very late
    evolutionary addition. Although I confess that our intuitions pertaining to what could
    have evolved very rapidly are not as robust as one would like, I think that Darwin’s
    Problem (the logical problem of language evolution) becomes very hard to approach if a
    GB-style architecture is assumed."

    Isn't the whole point of the MP an attempt to minimise UG in some sense so that it is more plausible from an evolutionary perspective?

    I think perhaps my genuine bafflement here is because of a shift in the use of UG: does it mean roughly
    a) shared properties of all adult I-languages
    b) the initial state of the LAD/FL?

    Boeckx is perhaps referring to b) when he says a GB architecture and you are using it to refer to a)?
    But what evolved is b) so surely that is the root of Darwin's problem.