Dave Embick and I just wrote a paper (for LCN) in which we speculate further about the possible relations between linguistics and neuroscience (as in Poeppel, D. and Embick, D. (2005). The relation between linguistics and
neuroscience. In A. Cutler (ed.), Twenty-First Century
Psycholinguistics: Four Cornerstones. Lawrence Erlbaum; and Poeppel, D. (2012). The maps problem and the mapping problem: Two
challenges for a cognitive neuroscience of speech and language. Cogn
Neuropsychol, 29(1-2):34-55. PDFs available on my site.) In particular, we discuss what we might aspire to, i.e. what the endgame might look like - or should look like. We are interested in reactions and advice of any kind. In some sense, we'd like to crowd-source the issue, i.e. collect examples of true success stories, spectacular failures, and so on ...
But: the bar is *high*. For example, a success might be something akin to the explanatory, mechanistic, causal understanding we have for sound localization in the barn owl (e.g. here). A failure might be akin to the case of C. elegans, a creature for which we know the genome and every neural ganglion and the entire wiring diagram but we cannot even figure out why the damn worm turns left or right. What, then, is a useful relation between computational-representational (CR) theories (as developed in linguistics, psycholinguistics, computer science, etc.) and neurobiological (NB) infrastructure?
In the review process, we got reactions across the spectrum, as per usual. One reviewer found the speculations reasonable, and in some places even helpful. Phew. Another reviewer found us relentlessly naive and misguided. Also phew.
Attached is a precis of the paper (which is, of course, available upon request). We welcome any advice, criticism, example, counterexample - either as comments here or messages to Embick (embick@BABEL.ling.upenn.edu) or me (david.poeppel@nyu.edu).
But: the bar is *high*. For example, a success might be something akin to the explanatory, mechanistic, causal understanding we have for sound localization in the barn owl (e.g. here). A failure might be akin to the case of C. elegans, a creature for which we know the genome and every neural ganglion and the entire wiring diagram but we cannot even figure out why the damn worm turns left or right. What, then, is a useful relation between computational-representational (CR) theories (as developed in linguistics, psycholinguistics, computer science, etc.) and neurobiological (NB) infrastructure?
In the review process, we got reactions across the spectrum, as per usual. One reviewer found the speculations reasonable, and in some places even helpful. Phew. Another reviewer found us relentlessly naive and misguided. Also phew.
Attached is a precis of the paper (which is, of course, available upon request). We welcome any advice, criticism, example, counterexample - either as comments here or messages to Embick (embick@BABEL.ling.upenn.edu) or me (david.poeppel@nyu.edu).
Towards a computational(ist) neurobiology of language:
Correlational, Integrated, and Explanatory neurolinguistics (***Précis***)
David Embick, University of Pennsylvania & David Poeppel, NYU and MPI
Abstract: We outline what an integrated approach to language research that connects experimental, theoretical, and neurobiological domains of inquiry would look like, and ask to what extent unification is possible across domains. At the center of the program is the idea that computational/representational (CR) theories of language must be used to investigate its neurobiological (NB) foundations. We consider different ways in which CR and NB might be connected. These are (1) A Correlational way, in which NB computation is correlated with the CR theory; (2) An Integrated way, in which NB data provide crucial evidence for choosing among CR theories; and (3) an Explanatory way, in which properties of NB explain why a CR theory is the way it is. We examine various questions concerning the prospects for Explanatory connections in particular, including to what extent it makes sense to say that NB could be specialized for particular computations.
Questions
(Q1) Basic Question: How does the brain execute the different computations that make up language?
(Q2) Advanced Question: Is the fact that human language is made up of certain computations (and not others) explained by the fact that these computations are executed in neurobiological structures that have certain properties (and not others)?
Possible Connections
Correlational Neurolinguistics: CR theories of language are used to investigate the NB foundations of language. Knowledge of how the brain computes is gained by capitalizing on CR knowledge of language.
Integrated Neurolinguistics: CR neurolinguistics plus the NB perspective provides crucial evidence that adjudicates among different CR theories. I.e., brain data enrich our understanding of language at the CR level.
Explanatory Neurolinguistics: (Correlational+Integrated Neurolinguistics) plus something about NB structure/function explains why the CR theory of language involves particular computations and representations (and not others).
Questions about specialization (crucial for Explanatory Neurolinguistics)
Specialization Question 1: Are there particular levels of NB organization that are to be privileged as candidates for CR specialization?
Specialization Question 2: Are there particular parts of the CR theory that are more likely to be candidates for Explanatory Neurolinguistic explanation than others?
Correlational, Integrated, and Explanatory neurolinguistics (***Précis***)
David Embick, University of Pennsylvania & David Poeppel, NYU and MPI
Abstract: We outline what an integrated approach to language research that connects experimental, theoretical, and neurobiological domains of inquiry would look like, and ask to what extent unification is possible across domains. At the center of the program is the idea that computational/representational (CR) theories of language must be used to investigate its neurobiological (NB) foundations. We consider different ways in which CR and NB might be connected. These are (1) A Correlational way, in which NB computation is correlated with the CR theory; (2) An Integrated way, in which NB data provide crucial evidence for choosing among CR theories; and (3) an Explanatory way, in which properties of NB explain why a CR theory is the way it is. We examine various questions concerning the prospects for Explanatory connections in particular, including to what extent it makes sense to say that NB could be specialized for particular computations.
Questions
(Q1) Basic Question: How does the brain execute the different computations that make up language?
(Q2) Advanced Question: Is the fact that human language is made up of certain computations (and not others) explained by the fact that these computations are executed in neurobiological structures that have certain properties (and not others)?
Possible Connections
Correlational Neurolinguistics: CR theories of language are used to investigate the NB foundations of language. Knowledge of how the brain computes is gained by capitalizing on CR knowledge of language.
Integrated Neurolinguistics: CR neurolinguistics plus the NB perspective provides crucial evidence that adjudicates among different CR theories. I.e., brain data enrich our understanding of language at the CR level.
Explanatory Neurolinguistics: (Correlational+Integrated Neurolinguistics) plus something about NB structure/function explains why the CR theory of language involves particular computations and representations (and not others).
Questions about specialization (crucial for Explanatory Neurolinguistics)
Specialization Question 1: Are there particular levels of NB organization that are to be privileged as candidates for CR specialization?
Specialization Question 2: Are there particular parts of the CR theory that are more likely to be candidates for Explanatory Neurolinguistic explanation than others?
4 comments:
I think the three questions you ask about CR and NB are valuable ones. But I'd like to know whether you believe that this (to quote): "CR neurolinguistics plus the NB perspective provides crucial evidence that adjudicates among different CR theories" has ever happened yet. If you do think it has, can you point to an example or two in the literature?
Max C
Max: In my own work I can say that the neuroscience of sensorimotor systems as studied in the visuomotor system of nonhuman primates and in speech has led me to rethink the representational models of the language system. Sensory perception models draw on neuroscience data quite a bit, and although I'm no expert, I'm sure a lot of memory work has been informed productively by neuroscience. Think back to the 80s when the bit question was whether or not perceptual system were automatic, bottom up or had a strong top down component. If we had looked to just the anatomy of the brain with its massive top down cocnnections, we could have resolved that debate immediately.
It seems to me that CR theories of language cannot be central to investigating the NB foundations. Rather, they must equally influence the other, with neither dominating. Approaches like Smolensky's Harmonic Grammar can, it seems, build a foundation where each discipline feeds the other, rather than one being the basis. Assuming that different disciplines observe a cause-and-effect relationship is, I think, an unhelpful direction to follow. What do you think?
Recently I have conducted an fMRI experiment that was indeed aimed at identifying neurobiological foundations of speech error detection to differentiate between theories of self-monitoring. And I do feel that the results of this study were able to do so. However, the conclusions that can be drawn on the basis of the neurobiological findings were severly limited due to the underspecification of these CR theories. So I believe that one important role of the NB findings is to force a more detailed specification of current CR theories (preferably in accordance with the NB findings).
Post a Comment