Re: Fw: Intro to natural language processing
Rob,
I wrote that previous note without going back to
the book to gather up examples about syntax.
Re Williams Syndrome: See excerpt at end.
The most relevant information Deacon gives about
syntax is in Chapter 10 "Locating Language".
The evidence comes from several sources:
1. Lesions. The oldest evidence comes from studies
of the language deficits caused by damage to the
brain from strokes, bullet holes, or other trauma.
(Wars are good sources of experimental subjects.)
The two oldest regions to be identified are Broca's
area (from the 1860s), which seems to disturb grammar,
and Wernicke's area (from the 1870s), which seems to
disturb semantics. Both of these areas have been
studied intensively.
2. Stimulation. "In the 1950s, a neurosurgeon by the
name of Wilder Penfield perfected a method for
assessing functional location by electrical stimulation
of awake, locally anesthetized patients... By probing
many areas, Penfield and others mapped the extent of
cortical areas where electrical stimulationg interfered
with language functions."
3. Brain scans. Three techniques: "regional cerebral
blood flow (rCBF); positron emission tomography (PET);
and functional magnetic resonance imaging (fMRI)."
Method #1 is like cutting wires in a computer to see what
happens; cutting the wire to the printer makes your printer
stop, but it doesn't prove that the wire does the printing.
Method #2 is like adding noise to a telephone line; it
disrupts the conversation; but it doesn't show where the
speakers live. Method #3 is the least invasive, and it is
more likely to show active regions, but it doesn't indicate
what the regions are doing. All three provide complementary
information, which requires extensive corroboration with
linguistic and psychological data about the effects.
Unlike the lesions, which point to two rather narrow areas,
methods #2 and #3 show that regions all over the brain are
involved in language generation and interpretation. Furthermore,
different kinds of tasks affect different regions. Noun-noun
word association tests, for example, light up different areas
in brain scans than noun-verb tests.
One signicant point is that operations that require the utmost
speed in analysis tend to be located the closest to the auditory
areas, but operations that are less time-critical tend to be
located farther away.
1. From stimulation studies (method #2), phoneme identification
is closest to the auditory centers, naming and reading tend
to be farther away, and short-term memory tends to be the
farthest away. This pattern is correlated with time delays
on the order of 10 milliseconds, 100 ms, or seconds.
2. Images of the working brain "show that the classic language
areas [Broca's and Wernicke's] are not unitary modules, but
rather complicated clusters of areas, each with different
component functions. This is not consistent with the view
that there is one self-contained "language organ" in the
brain, even one dedicated only to grammar."
3. "If a grammar module exists, it is not localized to one
cortical region, since subfunctions associated with
grammatical and syntactic processes are found in both
anterior and posterior regions."
There is one significant piece of evidence that threatens
to demolish any hypothesis about a single module that could
support Chomsky's Universal Grammar (UG):
"Even more serious challenges have been offered by
comparisons of agrammatism in patients who are native
speakers of very different languages. Broca's aphasics
who speak languages that are highly inflected, for example,
do not seem to be as agrammatic as their English speaking
counterparts... [In inflected languages] Wernicke's aphasics,
who also show disturbances of semantic analysis but not
speech fluency, are more impaired in producing and analyzing
the corresponding grmmatical transformation than are patients
with Broca's aphasia. So if there is a grammar module,
then the parts of this module map in very different ways to
different grammatical operations, depending on the relative
importance of positional or inflectional tricks for cuing
grammatical decisions in different languages."
This is interesting support for the idea that case relations
or thematic roles are closer to semantics than to syntax.
Marking such relations by inflections is more closely involved
with Wernicke's area, which seems to be more involved with
semantic procesing. Positional marking, as in English, seems
to be more purely syntactic. But as Deacon emphasizes, the
neural regions do not show as clear a separation between
syntax and semantics as some linguists claim.
Jackendoff (2002) mentioned Deacon's book in several short
passages and dismissed it. But evidence of this kind cannot
be ignored. Either Jackendoff didn't read Chapter 10, or he
has selective amnesia.
Another well-known feature is the "lateralization" of language
functions, usually in the left-side of the brain, but sometimes
in the right for left-handed people or those with early injuries
to the left side of the brain. Deacon claims that lateralization
is not a prerequisite for language, but merely the result of
keeping all functions that depend on high-speed interactions
on the same side. There is undoubtedly a slight bias for the
left side, but that bias can be overcome.
One significant point is that people who have been trained
to do simultaneous translation (e.g., listening to language X
at the U.N. while simultaneously speaking the translation in
a different language Y) tend to devote different hemispheres
to each language. This ability to use both hemispheres
is not acquired in infancy, but in young adults during an
intensive training period. The people who take such training
undoubtedly have considerable talent for languages, but it's
significant that a hemisphere that had not previously been
dedicated to language can be recruited for that function
as late as the early adult stage.
In my opinion, evidence of this kind makes the Chomsky-Pinker-
Jackendoff hypothesis of an innate "language module" dubious,
to say the least. Deacon agrees that the human brain has a
predisposition toward language and it has evolved to facilitate
the learning of language. But I find his summary compelling:
"If we keep in mind that primate brains very much like our
own have been around for tens of millions of years and that
the mammal brain plan which our brain follows has probably
been around for over 100 million years, it becomes evident
that the logic of language is probably highly constrained
to fit an ape brain logic... The patterns we observe
probably reflect, in a very indirect sort of way, the
processing problems produced by mapping a symbolic reference
system encoded in a serially presented modality [speech]
onto the processing logic of an ape brain."
John Sowa
_____________________________________________________________
RF> William's syndrome huh? Does it have to be me who is the
> first one to mention the parallel with gender stereotypes?
Which stereotype do you have in mind? Garrulous old men?
Busy-body old ladies? Talkative chicks who are trying
to fend off their dates?
See the following description from the NIH:
Source: http://www.ninds.nih.gov/disorders/williams/williams.htm
Williams syndrome is a rare, congenital (present at birth) disorder
characterized by physical and developmental problems including an
impulsive and outgoing (excessively social) personality, limited spatial
skills and motor control, and intellectual disability (i.e.,
developmental delay, learning disabilities, mental retardation, or
attention deficit disorder). Other features include characteristic
"elfin-like" facial features, heart and blood vessel problems,
hypercalcemia (elevated blood calcium levels), low birth weight, slow
weight gain, feeding problems, irritability during infancy, dental and
kidney abnormalities, hyperacusis (sensitive hearing), and
musculoskeletal problems. Symptoms vary among patients. Although
individuals with Williams syndrome may show competence in areas such as
language, music, and interpersonal relations, their IQs are usually
below average, and they are considered moderately to mildly retarded.
Scientists have learned that most individuals with Williams syndrome
have a deletion of genetic material on chromosome 7. This probably
causes the physical and developmental problems experienced by patients.