Wednesday, September 17, 2014

Another Foxp2 article

Rich Hilliard sent me another report on the Foxp2 article that I brought to your attention yesterday. This one from the CBC, and being a very proud and smug Canadian I am bringing it to you attention as well. In addition, it has a very nice photo of a mouse with a re-engineered "humanized" Foxp2 gene (it really is adorable, btw).  It also gives a few more details of the experiment and the different kinds of information that humanized mice integrated better than "just" mice did. Here's the short version of the experiment as told by the CBC: The experimenters
"...trained mice to find chocolate in a maze. The animals had two options: use landmarks like lab equipment and furniture visible from the maze ("at the T-intersection, turn toward the chair") or by the feel of the floor ("smooth turn right, nubby turn left"). Mice with the human gene learned the route as well by seven days as regular mice did by 11….Surprsingly, however, when the scientists removed all the landmarks in the room, so mice could only learn by the feel-of-the-floor rule, the regular rodents did as well as the humanized ones. They also did just as well when the landmarks were present but the floor textiles were removed. It was only when mice culduse both learning techniques that those with the human brain gene excelled."
This is the basis for the speculation that Foxp2 helps with language, for Graybiel interprets the results to "suggest" that Foxp2 enhances the capacity to transition "from thinking about something consciously to doing it unconsciously." And this relates to language how?  Well when kids learn to speak they transition from consciously mimicking words they hear to speaking automatically. Really? This is the linking hypothesis? Am I alone in thinking that this gives speculation a bad name?  It doesn't even rise to the level of a just-so story.

Jerry Fodor is reputed to have said that neuroscience has taught us virtually nothing about the mind. I am not sure that I entirely agree, but I am pretty sure that this work tells us next to nothing about language. Look, I love mice. They sing, they are cute, they run mazes better than I can, they navigate well in the dark. I am even sort of interested that one can put a human Foxp2 gene into a mouse. But the results of this experiment are very modest and have nothing whatsoever to tell us about language. I assume the language link is just there to hype the work. Show business.

11 comments:

  1. Most people know the first law of hole digging: Once you dug yourself into a hole [for example by revealing your ignorance about work in biology or neuroscience] STOP digging. Not so fearless Norbert, undeterred he remains the most compulsive digger I've encountered.

    So something said on CBC [not exactly a source of pride for those of us who live in Canada] "gives speculation a bad name ... doesn't even rise to the level of a just-so story". It somehow follows that the great Jerry Fodor was almost entirely right to claim that "neuroscience has taught us virtually nothing about the mind ... [and] that this work tells us next to nothing about language".

    Just who, besides Norbert believes CBC is a reliable source of information on SCIENTIFIC work? And what are the GOOD kinds of speculations about the biology of language; those that should inform the field and the interested lay audience? Is it the Chomskyan evolution tale: Once upon a time a hominid walked across the hot savannah. Somehow he was hit by a miraculous mutation that installed Merge. Recognizing immediately the magnitude of this event he mated with every female of his group. They all had healthy Merge babies and lived happily ever after. This is how we got Language. The End. Or maybe the wonderful speculation that explains why we have an innate concept of psychic continuity? Chomsky's grand-children encountered it in fairytales describing the adventures of Sylvester the baby donkey that turns into a rock and back into a donkey [because it is a law of nature that fairytales always end happily]. There was not explanation in the tale yet the kids knew the concept anyhow. Hence it must have been innate. QED

    These and other speculations [for an overview see http://ling.auf.net/lingbuzz/001592] certainly have all the trimmings of good scientific hypotheses. And should one disagree one is told to ease off; after all Chomsky was addressing a non-scientific audience, you know; just making off the cuff remarks. And presumably for this reason we have to take 100% seriously what OTHERS say on CBC about their work...

    ReplyDelete
  2. Hi, Norbert, and thanks for posting this interesting commentary. I’m just wondering, does the same skepticism apply to declarative/procedural models of language acquisition? The suggestion that language acquisition starts out relying on declarative memory systems for figuring out the lexicon, and then shifts to procedural memory systems as grammatical rules are mastered, wouldn’t be a novel one in the cog neuro literature (eg as argued by Michael Ullman). This has always seemed to me a reasonable way of thinking about things even if we don’t yet have clear ideas about how exactly procedural learning might work for the most interesting stuff (ECP and so on).

    So the link to procedural learning in general, and even language acquisition, doesn’t seem so tenuous to me. I actually find these sorts of results really interesting – even if the take-away is (as Charles Yang noted in a comment on the other post) that FOXP2 is not and has never been a ‘language gene’. But I can see how things get dicier if one is skeptical about these memory systems’ involvement in language acquisition at all.

    Heidi

    ReplyDelete
    Replies
    1. @heidi
      I'm not sure that I understand the procedural vs declarative distinction. It seem more like a routinized vs non routinized distinction. I also did not see why using landmarks was declarative while using texture of pads was not. This may be a failing on my part, but it seemed very breezy to me.

      As for what this shows: it would be nice to have tried to hook it up to something semi real. What I saw looked more like mushy assertion than anything else, with language tacked on so that it had some sex appeal. Like I said, I like speculation, especially informed surmise. But this seemed like a real stretch, something that were we to make similar airy claims would be dismissed out of hand.

      Delete
    2. In the sense I'm familiar with, the procedural/declarative distinction hinges on the type of knowledge being learned - essentially facts (declarative) vs rules (procedural). Motor learning is a classic procedural-memory function so that might be what they're going for: seems like the crucial distinction was whether the association was motoric or not (get a reward for going to a certain landmark vs. get a reward for turning right).* Motor is definitely not grammar but nevertheless in humans there are parallels, particularly in the brain structures involved, such that motor and grammar deficits often co-occur (as in the KE family). For me that's enough to make these foxp2/motor learning associations interesting.

      Now, whether the authors themselves have any idea how their findings actually relate to language... that's a separate issue and here I'm with you in thinking they don't. ('Conscious mimicking' is a pretty clear indicator of that.) So you may be right about their motives.

      *Obligatory caveat - haven't read the study so it's quite possible this was not their intent after all. Also, these terms may be used somewhat differently in the animal lit, in which case I wouldn't be familiar with their meanings.

      Delete
  3. OK, since I seemed to have instigated this thread, I'll offer a wild speculation on how this finding might be connected to other speculations about Merge, etc. The "humanized" mice perform better only in the condition where both landmarks and textures are available. Remove either one of these and their performance drops to be on par with the "non-humanized" mice. One interpretation of this could be that the extra ability that the "humanized" mice have is being able to conjoin features across domains as necessary. This would be similar to the "hardwired" versus "on-demand" distinction drawn by VanRullen (http://www.tandfonline.com/doi/abs/10.1080/13506280802196451?journalCode=pvis20#preview).

    ReplyDelete
    Replies
    1. Yes, that's one way of interpreting this experiments --- one can take that line of thought quite a long way until you get to this commentary which is further than I would go. But if FOXP2 is implicated in language learning, then it's not that much of a stretch.

      Delete
    2. If I understood the blog entry you linked to, it is implicated in all kinds of learning, language being included. I confess that I did not understand what the blogger meant by operant conditioning here. the Chomsky critique of Skinner, which he refers to in both posts I looked at, argued that the notion was either false or vacuous in the case Skinner proposed. I always thought the vacuity claim was the right way to go myself. I think that the current idea that Ellen (below) and Bill (above) are pointing to is that Foxp2 might be implicated in the capacity to coordinate information across modules. This is an interesting idea (though we don't really have a good analytic idea of what this means: what does the capacity to link across modules rest on?). It is also interesting to speculate how, if at all, this relates to recursivity (I don't see that it does actually). In fact, it would be nice to know exactly what grammatical facility is implicated by this. The best I can come up with is that it has something to do with a syntactic categorization (i.e. things forms groupings independently of what they "mean"). At any rate, it seems that I was, once again, wrong (big surprise). This result might have more legs than I gave it credit for.

      Delete
    3. I'm not a blogger, I'm a tenured professor at a German university who also blogs :-)

      And no, FoxP genes are linked to the motor capacity of controlling your tongue, lips, vocal chords, all of which you need to articulate. The way we learn this coordination is done by operant conditioning (see post linked to in my comment below). The coordination itself is first flexible and full of errors (think babbling) and only after much practice does speech become automatic. FoxP seems to be involved in this automatization of articulation.

      Björn Brembs

      Delete
  4. Or like Liz Spelke's experiments on language being necessary for cross-domain feature conjunction.

    ReplyDelete
    Replies
    1. See how fruitful speculation can be when you try to relate it to what we already know. Let's hear it for mice!

      Delete
  5. Actually, the humanized mice did not perform 'better' - they were faster in forming automated behaviors. Making your behaviors less flexible sooner than you normally would is not necessarily a good thing in my books.
    Second, most, if not all, researchers in this field emphasize that this work concerns mainly the speech component of language and the relation of (operant) motor learning to this process has long been acknowledged, not the least by Skinner, who's work is essentially rehabilitated (and Chomsky's review falsified):
    http://bjoern.brembs.net/2014/06/the-drosophila-foxp-gene-is-necessary-for-operant-self-learning/

    Björn Brembs (for some reason my account doesn't seem go through)

    ReplyDelete