walking around in the brain

Brian MacWhinney macw at CMU.EDU
Sun Sep 20 22:27:28 UTC 1998


Colin and FunkNet,
  As Lise Menn noted, several aspects of your comments have already been
run through the email mill on info-childes at andrew.cmu.edu.  On that list,
Ann Peters initially pointed to Jaeger et al. as a potential challenge to
connectionist/emergentist models of language development and processing.  I
pointed out that there is at least one fully implemented neural network
model that fully predicts results of the type found by Jaeger et al. --
this being the model developed in articles in JML and Connection Science by
Alan Kawamoto.  So, it is very hard to see how this type of differential
activation can be used  to support modularity above connectionism.
   Like Lise, Joe Stemberger, and Liz Bates, I pointed to various stimulus
and tasks differences involved in the comparison between regulars and
irregulars.  However, in that discussion, I don't believe that anyone
managed to come up with the excellent rival hypothesis that you are
suggesting.  It seems to me that one of the best potential uses of the fMRI
 technique which is now supplanting the PET technique used by Jaeger et al.
is to check out precisely the type of hypothesis you have suggested.  As
you note, people like Mishkin, Ungerleider, Schneider, and Damasio have
shown that activation of motor concepts leads to activation of parts of
motor cortex in extremely systematic ways.  The nice think about your
hypothesis is that it can be tested so easily.
   If someone hasn't already run this study, I would be sorely tempted to
run it myself.  Thanks for the great point.

--Brian MacWhinney



More information about the Funknet mailing list