Selected Newsgroup Message

Dejanews Thread

From: "Sergio Navega" <snavega@attglobal.net>
Subject: Re: A Major Stumbling Block
Date: 05 Nov 1999 00:00:00 GMT
Message-ID: <3822b9df_4@news3.prserv.net>
References: <7vs91q$ss6$1@mochi.lava.net>
X-MimeOLE: Produced By Microsoft MimeOLE V4.71.1712.3
X-Complaints-To: abuse@prserv.net
X-Trace: 5 Nov 1999 11:05:03 GMT, 166.72.29.60
Organization: Intelliwise Research and Training
Newsgroups: comp.ai.nat-lang

Chaumont Devin wittily again wrote in message
<7vs91q$ss6$1@mochi.lava.net>...
>Fellow researchers,
>
>A major stumbling block in understanding the internal mechanisms of
>language has been the fact that we are unable to directly examine details
>of the brains of living subjects.  I was reading some old science fiction
>today and got an idea for a valuable alternative:
>
>In the brain we have neurons running all over the place in a huge tangle,
>but at the foramen magnum, the nerves going to the body are all arranged
>in parallel.
>
>My question is this:  Is there some way a collar could be built that would
>enable a machine to track the states of a single strand of nerve fiber?
>This would not be easy, because the soft tissue around the neck would make
>for imprecision, and the pulsing of blood through the jugular veins would
>add still more possibility of error.  But once in place, the device might
>easily be calibrated by asking the subject, say, to raise his left index
>finger.

This is done today, on a very, very limited scale. We already "know"
the kind of pulses (spikes) sent to motor systems in order to provoke
muscular action (however, we still don't know how information is coded in
those pulses). For instance, it is not difficult to "plug some wires"
to nerve efferents of the arm and link it to a computer in order to
produce some kinds of simple movements. This way, you can slap the
face of an enemy and say it was the computer.

>
>I do not know how such a device might be able to pick up nerve impulses,
>but I was thinking of something like the brilliant idea behind positron
>emmission brain scanning, which can pinpoint the source so precisely.
>

PET (Positron Emission Tomography) is, in reality, not very precise
besides requiring that the subject ingests a liquid with an isotope.
Modern methods such as fMRI are more precise and can be used to get
a more precise view of what's happening "in there". The results are,
for more than a decade, provoking a revolution in cognitive neuroscience,
with a lot of things being discovered. For instance, it was found that
that nouns and verbs are "processed" in different parts of the brain.
Also the past tense of regular verbs are processed in different parts
than irregular verbs.

>Anyway, if we could build such a device, we might be able to learn a great
>deal more about the workings of systems of nerves without actually getting
>inside any brains, which is something I would like to avoid.  It is true
>that I have thought a few times about the possibility of drilling very
>small holes through my skull and inserting tiny wires, but there is simply
>too much risk involved--especially for me, being blind, because my poor
>brain is really pretty much all that I have, besides the girls at Zippy's
>and my Vietnamese girl friend!
>

Although sometimes you may leave us with that desire (to puncture
your brain), I would not recommend such a procedure, not only for
the possible damage it may produce, but also because you will not
obtain much from it. There are so many neurons in a cubic millimeter
(about 50 thousand) that getting the signal from one of them would
require a microelectrode (which already exists). The data processing
of this signal, however, is hugely complex and what it informs is not
much, because most of the action (in information terms) is not in
the individual neuron level, but on the behavior of groups of neurons.

>I have also thought of drilling a few tiny holes in the skull of Sergeo
>Navega, which I cannot help thinking about since he is such a good fellow,
>and such an enlightened thinker, but I worry about what might happen
>if/when his wife were to hear the news.  I have been roundly cursed out in
>various languages, as all of you probably know by this time, but never yet
>in Portuguese!
>

Well, being bald as I am, puncturing my skull would be really easy.
But there's a better alternative, one which is less invasive and
allows one to obtain a good deal of information. It is a technique
known as ERP (Event Related Potential). It is sort of a derivative
of EEG (Electroencephalogram), a bunch of wires connected to one's
skull which shows electrical activity. Usually this activity is
interesting on a neurological level, with few applications to
cognitive research. However, when this signal is processed by
averaging amplifiers in a laboratory condition where it is possible
to repeat cognitive experiments, the results are interesting (I
can tell you more about how this averaging process works if you
want). The result are curves which show the response of the brain
to certain cognitive situations. Just to exemplify, if you say this:

The pizza was too hot to eat.

to a person linked to that equipment, you will obtain a certain
"waveshape", which is considered the "normal" processing of that
sentence. However, when you say this:

The pizza was too hot to cry.

The waveshape will be interestingly different, with a distinct
negative "pulse" appearing about 400 miliseconds after the event.
This appears so regularly that it was called N400 pulse and always
happens in this situation. In another situation, when the subject
listens to a phrase like this:

* The cats won't EATING their food.

Another waveshape develops, this time a positive going pulse about
600 miliseconds from the start of the event (utterance). Again this
is so regular that it was named P600 pulse and it appears when
there are syntactic problems in the phrase. N400 appears on
semantic problems. Interesting results, no?

There's another interesting phenomenon called MMN (mismatch
negativity) but I'll let this for another time.

Regards,
Sergio Navega.

From: "Sergio Navega" <snavega@attglobal.net>
Subject: Re: A Major Stumbling Block
Date: 05 Nov 1999 00:00:00 GMT
Message-ID: <38233b65_4@news3.prserv.net>
References: <7vs91q$ss6$1@mochi.lava.net> <3822b9df_4@news3.prserv.net> <38230CD6.60085607@yahoo.com>
X-MimeOLE: Produced By Microsoft MimeOLE V4.71.1712.3
X-Complaints-To: abuse@prserv.net
X-Trace: 5 Nov 1999 20:17:41 GMT, 32.100.0.169
Organization: Intelliwise Research and Training
Newsgroups: comp.ai.nat-lang

Rick Harker wrote in message <38230CD6.60085607@yahoo.com>...
>Sergio Navega wrote:
>> Well, being bald as I am, puncturing my skull would be really easy.
>> But there's a better alternative, one which is less invasive and
>> allows one to obtain a good deal of information. It is a technique
>> known as ERP (Event Related Potential). It is sort of a derivative
>> of EEG (Electroencephalogram), a bunch of wires connected to one's
>> skull which shows electrical activity.
>
>Sergio,
>Can you please post some more on this?
>This sounds way cool!
>
>(I know it's not quite nat-lang, but still closely related :) )
>--

I recently gave a talk about cognitive neuroscience. Some of
the slides were about ERP. Take a look at:

http://www.intelliwise.com/seminars/traneuro.htm

Have fun,
Sergio Navega.

From: "Sergio Navega" <snavega@attglobal.net>
Subject: Re: A Major Stumbling Block
Date: 05 Nov 1999 00:00:00 GMT
Message-ID: <38232171_3@news3.prserv.net>
References: <7vuosd$ct3$1@mochi.lava.net>
X-MimeOLE: Produced By Microsoft MimeOLE V4.71.1712.3
X-Complaints-To: abuse@prserv.net
X-Trace: 5 Nov 1999 18:26:57 GMT, slip-32-102-102-104.sp.br.prserv.net
Organization: Intelliwise Research and Training
Newsgroups: comp.ai.nat-lang

Chaumont Devin wrote in message <7vuosd$ct3$1@mochi.lava.net>...
>Dear Sergeo,
>
>Thanks for the interesting info you posted on recent neurological
>research.
>
>On Fri, 5 Nov 1999 07:30:28 -0200, "Sergio Navega" <snavega@attglobal.net>
>wrote:
>
>>For instance, it was found that
>>nouns and verbs are "processed" in different parts of the brain.
>>Also the past tense of regular verbs are processed in different parts
>>than irregular verbs.
>
>Aha, and what about prepositions, subordinating conjunctions, adverbs, and
>adjectives?
>

I owe you something about these aspects. I don't have them right
now. But I'm about to receive Steven Pinker's recently published
book 'Words and Rules' and I guess he will address some of these
issues there. I promise I'll post something about this as soon as
I sink my teeth on it.

>>The result are curves which show the response of the brain
>>to certain cognitive situations. Just to exemplify, if you say this:
>
>>The pizza was too hot to eat.
>
>>to a person linked to that equipment, you will obtain a certain
>>"waveshape", which is considered the "normal" processing of that
>>sentence. However, when you say this:
>
>>The pizza was too hot to cry.
>
>>The waveshape will be interestingly different, with a distinct
>>negative "pulse" appearing about 400 miliseconds after the event.
>
>Aha!  Double take!  Does not compute!
>

I suspected that you would like this result. Yes, there's no
doubt that there are very different things happening inside our
brain when one utters phrases with and without semantic coherence.
I'll not be surprised if you'll use this as a "proof" that there
is a semantic plane as you've been proposing for eons.

However, we must be cautious with any conclusion. What ERP reveals
is the result of the processing of millions (if not hundred of
millions) of neurons. I guess that a lot of distinct theories
could be used to explain what is happening there equally well.
More experiment is clearly necessary. But there are some things
we can think based on these results.

>>This appears so regularly that it was called N400 pulse and always
>>happens in this situation. In another situation, when the subject
>>listens to a phrase like this:
>
>>* The cats won't EATING their food.
>
>Do you mean with the word, "eat" replaced by "eating"?
>

Yes. That develops a different response, not related to semantics,
but to the grammatical aspect of the phrase. What amazes me is that
this response (besides being positive, and not negative as the
former) occurs 600 milliseconds after the event, which means,
200 milliseconds *after* the previous. This is interesting, it
suggests (though not conclusively) that gramatical processing is
done *after* semantic processing. How could that be? I can only
speculate.

We know that a lot of language processing is done unconsciously
(consciously only if one is a novice in a foreign language). Upon
reception of an utterance, there are lots of things happening
in parallel. The first thing are auditory and phonological
processing, with the identification of phonemes, the sound of
vowels, the disambiguation (to a limit) of consonants, etc. From
this level, it appears that a lot of other parallel processes
are fired.

Some of these processes will try to see some kind of association
with semantic values. When one utters "apple", the image of
an apple may be recalled independently of the grammatical
category of the word.

ERP appears to be revealing that this semantic spread of
activations runs faster (or has less things to do) than the
corresponding grammatical fronts (another hypothesis is that
grammatical processing is started by semantic processing, and
then appears to run later). This will not be so surprising
when one recalls that sensorimotor, visual and auditory areas
of our cortex are requested to help in the conceptual
processing of the phrase. I can find this as a slight
suggestion that semantics is more important than grammar and
that semantics is closer (in other words, more efficient)
because it is also processed by areas of the cortex related
with our physical presence in the world. When we're babies,
we don't have language, we have only sensory experiences and
this is what we use to support all the abstract things that
we learn later. So it is not unreasonable to think that
first we "understand" what it means, and later we "parse" the
sentence (at least in non-ambiguous utterances).

>>Another waveshape develops, this time a positive going pulse about
>>600 miliseconds from the start of the event (utterance). Again this
>>is so regular that it was named P600 pulse and it appears when
>>there are syntactic problems in the phrase. N400 appears on
>>semantic problems. Interesting results, no?
>
>More than interesting.
>
>What amazes me is that even after all of this precise regularity has been
>discovered, people still have the audacity to come round telling us that
>universal grammar does not exist.  They claim that although we all have
>two arms, two legs, ten fingers, etc., we somehow all have to process
>language differently.  What a shame!
>

I'm sure Alex Gross will know how to answer better to this comment
of yours. Needless to say, I'm one of the guys who don't believe
in innately determined universal grammars. But that's not the theme
here.

>>There's another interesting phenomenon called MMN (mismatch
>>negativity) but I'll let this for another time.
>
>How about some time pretty soon?
>

Ok, this will give me the opportunity to exercise this newly
acquired knowledge.

Mismatch Negativity, a very specific field of investigation, is
a "feature" that appears in ERP that is correlated with 'novelty'.
It was that characteristic that sparked my interest.

When subjects are presented to a continuous tone, their ERPs show
a constant waveshape. However, when in the middle of this constant
auditory stimulus a new kind of signal is inserted (something that
is clearly different from the preceding signal, such as a slight
alteration of frequency or duration of tones), the corresponding
ERP is altered and this alteration is called MMN.

MMN is, then, the result of a neural processing which detects
changes between incoming signals and previous experiences of
that signal. It is sensitive to this comparison, as if the brain
had been woke up by the "unexpectedness". This is a very important
result for me, because it suggests how important is the mechanism
of detection of novelty. Another result that research in MMN
brings is that often this detection is pre-attentive, which
means, does not depend on conscious attention. It is an
automatic processing that runs by itself.

Regards,
Sergio Navega.


Back to Menu of Messages               Sergio's Homepage

Any Comments? snavega@attglobal.net