LINGUIST List 5.834

Thu 21 Jul 1994

Disc: Williams Syndrome & Modularity

Editor for this issue: <>

Directory

1. Paul Deane, Williams Syndrome & Modularity

Vicki Fromkin's most recent contribution to the thread
which grew out of my "linguist-bashing" post illustrates how
difficult it can be to communicate across a theoretical di-
vide. We are, apparently, pointing to exactly the same set of
facts and arguing from them to (apparently) diametrically
opposite conclusions. It's worth taking a look at why.

(By the way: I should warn LINGUIST subscribers that a proper response
 to the issues that Fromkin raises must go on at some length, so those
 uninterested in the subject would do well to zap this post NOW.)

 Fromkin's argument appears to run as follows:

 1. Some parts of language are preserved in Williams
 syndrome (e.g., syntax). Other parts are
 compromised.

 2. Some parts of spatial cognition are preserved in
 Williams syndrome, and others are compromised.

 3. This is evidence for modularity since otherwise we
 wouldn't expect brain pathology to affect some
 mental processes (e.g. syntax) more than others.

The problem isn't in (1) and (2), which are as far as I can
tell are undisputed facts. The problem is the stark either/
or alternative presupposed in (3): Fromkin appears to be
assuming that there are only two options: either language
and mind are modular (with all the theoretical baggage that
come with the term), or else they are non-modular, which
Fromkin appears to interpret as entailing a complete absence
of any division of computational labor within the brain. But
that is a straw man. In neuroscience, such an utterly anti-modular
position would correspond to Lashley's (now-discredited) theory of
"mass action" which claimed in effect that every neuron in the brain
was involved in every mental process. I certainly don't hold that
position and I doubt seriously that anybody in linguistics does.
 I think it is abundantly clear (at least from the neurosciences
literature I have read) that the brain is extremely efficient at factoring
the information it receives into lots and lots of different pieces spread
out in many different parts of the brain. But that does not by itself prove
that the brain (and hence language) is modular in Fromkin's sense. As I see
it, the modularist position entails at least the following additional assump-
tions:

 1. Information distributed to different parts of the
 brain is highly segregated, with little interaction
 bringing it back together except through very narrow
 channels that are only able to look at the output of
 the "highest" representation within a "module".

 2. Each "module" has a piece of neural anatomy DEDICATED
 to processing information solely WITHIN the module.
 This would appear to entail that such functions do not
 involve SHARING of neuroanatomy between different
 modules (at least for those functions that are central
 to the module).

These assumptions entail that we can safely study
neurological problems involving one phenomenon (say,
language) without worrying too much about the fact that the
very same patients may also display unusual patterns in
another cognitive domain (say, space). And it is precisely
this entailment that Fromkin is presupposing when she argues
that the facts in Williams Syndrome support modularism. They
would, if the partial disruption of spatial cognition had
nothing to do with the partial disruption of language. But
that is precisely the issue about which Fromkin and I
disagree.
 The real issue is whether there is any way to show that
the selective preservation of syntax has anything to do with
the selective preservation of certain parts of spatial
cognition. A connection could exist for at least the
following reasons:

 (1) There might be a shared piece of neuroanatomy,
 perhaps because that part of the brain is efficient
 at performing a computational task necessary for both
 functions.

 (2) There might be a shared resource, such as attention,
 whose disruption might affect some computations more
 seriously than others.

 (3) The apparent difference between the two functions
 might turn out to be illusory.

When we are talking about linguistic and spatial cognition,
(3) is rather unlikely. (2) is possible and there is rather
suggestive evidence indicating that (1) could be correct.
 To begin with, there is evidence that spatial cognition
and language occupy overlapping territories in the brain.
The relevant reference is:

 Robertson, Lynn C., and Marvin R. Lamb. 1991.
 Neuropsychological contributions to theories of
 part/whole organization. Cognitive Psychology
 23: 299-330.

Robertson and Lamb argue, drawing on a variety of
neuropsychological studies, that visual perception of
part/whole relations is localized in the following areas of
the brain: (a) the inferior parietal lobe, and (b) adjacent
areas of the temporal lobe (i.e., the superior part of the
posterior temporal lobe, in both left and right
hemispheres). Now, it just so happens that the posterior
language areas of the brain comprise Wernicke's area
(occupying the left superior posterior temporal lobe) plus
the left inferior parietal lobe. So there is a direct over-
lap between the parts of the brain that handle visual
perception of part/whole structure and the parts of the
brain that handle language.
 In addition, Robertson and Lamb argue that visual
part/whole perception involves the following division of
labor: (i) the right temporal region is most efficient at
perceiving global spatial patterns; (ii) the left temporal
region is most efficient at perceiving spatial details;
(iii) the inferior parietal lobe handles the division of
attention between global properties and local detail.
Evidence for this conclusion comes from a variety of
sources. Among the most interesting involves figures like
the following:

 S S
 S S
 S S
 SSSSSSSSSS
 S S
 S S
 S S

In the studies Robertson & Lamb cite, patients with left
hemisphere damage to the posterior temporal lobe recognize
the presence of lots of S's but fail to recognize the global
H-pattern. Patients with posterior temporal lobe damage on
the opposite side of the brain recognize the H but fail to
discriminate the individual S's. The evidence, then, is that
a critical language area, Wernicke's area, is also crucially
implicated in the perception of spatial details.
 This is why I find Bellugi's observation of preserved
spatial abilities in Williams Syndrome patients so striking.
She specifically notes that they seem to pay more attention
to local details than to global patterns. What does this
suggest? That the preservation of grammar in Williams
Syndrome is associated with attention to spatial detail
because both functions are associated with left hemisphere
"language" areas in the brain.
 But once one sees a connection between the two, there is
an obvious relationship between visual detail and grammar:
both involve automatic analysis of larger "wholes" into hierarchies
of "parts". Or to put it another way: we could suggest that constituency
in grammar and part/whole structure in vision are processed by the same
brain areas precisely because they ARE the same sort of phenomenon (abstract
part/whole structures). It is in fact suggestive that just as right-brain
vision involves perception of global gestalts, the right brain has been
associated with such relatively global gestalt-like phenomena as discourse
structure, intonation patterns, and metaphor.
 So I do not see a phenomenon like Williams' syndrome as
supporting Fromkins' view of modularity. At best, it is
consistent with it (though of course there could be evidence favoring
modularity of which I am unaware), but I think a view which seeks connections
across domains is much more promising. In my experience, whenever I have
taken a close look at phenomena supposedly supporting a modular theory, I
have found that a lot of interpretation has taken place to build the argument
--interpretive moves that tend to presuppose a modularist viewpoint to
start out with.