bilingual fMRI

[Brian MacWhinney]

Dear FunkNet,
  Lise's account of the Kim, Relkin, Lee, and Hirsch paper sees reasonable.
 However, having just now gone over the article, I might add a few points.


1.  The task was covert subvocalization.  In other words, this is a full
language task that involves grammar and lexicon, but no clear input or
output phonology.  Subjects were describing what they had done earlier that
day.  However, they had formulated the descriptions earlier.  So this is a
replay of an earlier internalized narrative.

2.  There was no difference between late and early bilinguals in Wernicke's
area, only in Broca's.  Apparently, no other areas showed significant
activation in the task, which is a bit surprising.  Maybe they just didn't
report on other areas.

3.  The authors point to Kuhl's work on perceptual magnets in auditory
development as indicating that the region of cortex responsible for a
language task may change with development.  However, they then go on and
somehow suggest that Broca's area may be involved in phonetic processing.
Quite a non sequitur, I would say.

So the authors don't really present any account of their findings.   Lise's
mention of the fact that late bilinguals are "more bilateral" seems to be
relevant.  However, I worry about the fact that the Kosslyn/Damasio work is
about individual concrete nouns and this task was one involving connected
discourse.  I think that Lise is correct that Broca's has some activation
in the Kosslyn/Damasio studies, but I would need to double check all of
that before pursuing it as the account for these findings.

Let me try to float another explanation.  In a few recent papers, I have
argued that early bilinguals project the input linguistic data to a single
space, because that space is not yet saturated by weights on synaptic
connections and the two systems can be learned together in a
computationally reasonable sense without worrying about catastrophic
interference.  In adult L2 learning, the optimal area for an ability has
already been occupied and new learning must either use the old territory in
a new way or else coopt adjacent "new" territory.  In Broca's area, this
would involve particularly the use of adjacent cortical areas to control
alternative planning patterns for sentence production and the activation of
words in sequence.  The fact that Wernicke's shows a reuse of the older
area roughly matches up with a Kroll/DeGroot account of L2 conceptual
learning through parasitism on L1 conceptual structure.  However, I would
also expect to see some other cortical mismatches in other motor areas and
other auditory areas.  The fact that these were no detected in this study
may be attributable to the internalized nature of the task.

A lot of what is at issue here is exactly what the role of Broca's area in
language processing might be.  I am not pushing the grammar box theory, but
I would think that viewing Broca's as the seat of semantic organization
(not a term used in the original article, but in Suzette's message) is also
inadequate.  I would like to think of Broca's as controlling high level
sequential planning for language and related abilities.  The occasional use
of Broca's for object recognition would be in terms of matching the
affordances from objects to higher level action plans.

Suzette, hope this helps.  Comments invited.

--Brian MacWhinney