Selected Newsgroup Message
From: "Sergio Navega" <snavega@ibm.net>
Subject: Re: NN formats
Date: 18 May 1999 00:00:00 GMT
Message-ID: <37416784@news3.us.ibm.net>
References: <3739A199.9E2B8209@erols.com> <3739B919.B4A6AF36@clickshop.com>
<3739C62F.527BAF91@erols.com> <3739E816.4BF79A4B@clickshop.com>
<373A143C.6A467BFF@erols.com> <373C1480.7482A5B5@tig.com.au>
<373C58B7.E700B8F6@erols.com> <373EC872.80D8C02D@tig.com.au>
<3740344b@news3.us.ibm.net> <37409FED.488F8B89@clickshop.com>
X-Notice: should be reported to postmaster@ibm.net
X-MimeOLE: Produced By Microsoft MimeOLE V4.71.1712.3
X-Complaints-To: postmaster@ibm.net
X-Trace: 18 May 1999 13:13:40 GMT, 166.72.21.120
Organization: Intelliwise Research and Training
Newsgroups: comp.ai.nat-lang
Seth Russell wrote in message <37409FED.488F8B89@clickshop.com>...
>Sergio Navega wrote:
>
>> Anton Funk Tism Trees wrote in message <373EC872.80D8C02D@tig.com.au>...
>> >[snip]
>> >Additionally, human brains seem to come with a certain amount of
>> >structure built in thanks to evolution. Language abilities are
>> >foremost in the list of such inherited structure, but there are
>> >also a whole range of things like reflexes (smiling, crying,
>> >sucking), and recognition of human faces.
>> >
>>
>> Anton, I enjoyed the remaining of your post, but in this particular
>> paragraph you touched a matter that, in my vision, could be
>> interpreted differently. In particular, native language mechanisms,
>> although assumed to be present by the majority of the cognitive
>> scientists of today, is finding serious opposition from recent
>> neurobiological evidences. In my vision, the question of innateness
>> of language is open, and in this regard I'd stay on the side
>> of those who say that language is learned, not innate.
>
>Recently I heard a talk by Dr. Patricia Kuhl of the University of
>Washington (see below). The essence of her findings was that babies
>start out being able to hear all vowel sounds, but by six months some
>innate mechanism has trained their hearing such that they only hear
>the vowel sounds in their native language (sorry for my paraphrase
>of the research, perhaps you could get Dr Kuhl to send you copies
>of her papers). To me this is clear evidence that there are innate
>processes involved in learning language.
>
>CC: Dr. Patricia Kuhl's Research
>http://weber.u.washington.edu/~sphsc/research/pkkresearch.htm
>
Lucky you! I'd like to have heard to that talk too. Dr. Kuhl's
results are indeed important, but not for the innateness of
language. Quite the opposite, I would say.
The fact that parents use "parentese" (the babbling used by
parents) to talk to their babies may have some explaination
in terms of evolutionary aspects. The effect on the baby
appears to be a progressive specialization of their phonological
speech processing abilities. After some months of exposure
to this "parentese", the babies become more "specialized" in
the exquisite differences that make up their native language.
This is what later contributes to define the differences in
hearing ability of an english person from those of a russian.
This effect is a sign that the babies' brain could be "empty"
from the beginning, exactly the opposite of what is proposed
by innatists.
Yes, there is an innate mechanism that is able to acquire these
speech clues in order to become more specialized. What is
interesting to note is that there aren't conclusive evidences
that show that this mechanism is *specific* for language.
It all appears to be a *general mechanism*, that is capable
of similar learning performance in other areas of our
cognition (vision, etc).
In particular, some connectionist models are able to perform
similar feats starting from a completely "empty" network.
For a good example, see:
http://www.siu.edu/~psycho/faculty/morten/statlearn.html
Christiansen proposes a model based on Simple Recurrent
Networks (SRN) that is impressively able to segment speech.
Regards,
Sergio Navega.
From: "Sergio Navega" <snavega@ibm.net>
Subject: Re: NN formats
Date: 18 May 1999 00:00:00 GMT
Message-ID: <3741b5ac@news3.us.ibm.net>
References: <3739A199.9E2B8209@erols.com> <3739B919.B4A6AF36@clickshop.com>
<3739C62F.527BAF91@erols.com> <3739E816.4BF79A4B@clickshop.com>
<373A143C.6A467BFF@erols.com> <373C1480.7482A5B5@tig.com.au>
<373C58B7.E700B8F6@erols.com> <373EC872.80D8C02D@tig.com.au>
<3740344b@news3.us.ibm.net> <37409FED.488F8B89@clickshop.com>
<37416784@news3.us.ibm.net> <37417C34.E43F3FDC@clickshop.com>
X-Notice: should be reported to postmaster@ibm.net
X-MimeOLE: Produced By Microsoft MimeOLE V4.71.1712.3
X-Complaints-To: postmaster@ibm.net
X-Trace: 18 May 1999 18:47:08 GMT, 200.229.243.90
Organization: Intelliwise Research and Training
Newsgroups: comp.ai.nat-lang
Seth Russell wrote in message <37417C34.E43F3FDC@clickshop.com>...
>Sergio Navega wrote:
>
>> Yes, there is an innate mechanism that is able to acquire these
>> speech clues in order to become more specialized.
>
>But that is all that I am saying. Obviously it is that "innate
mechanism"
that
>you seek. It is obviously not specific to any given language. It would
seem
>to be targeted at some specific type of repeating patterns. I don't think
it
>is a coincidence that all languages are built on instances of this specific
>type of repeating patterns. Perhaps everything that we humans cognize is
made
>up of these specific type of repeating patterns. Can you further
characterize
>this "specific type" ?
>
It seems to be a semantic question now. When we're talking about
innate mechanisms, I have not much to question, after all, our
brains are physical things that, one way or other, have several
characteristics determined genetically. But the problem appears
when we talk about innate *knowledge*. Lets take a look at
this subtle difference.
Instances of innate mechanisms are our perception of color, the
improvement of motion detection of the receptive fields of the
surround of our vision, etc. One way or another, these
characteristics are coded in our genes, because they
conferred an advantage to us.
What Chomsky proposed was innate knowledge of grammatical structures.
This is equivalent of asking "Do humans have a portion of their
genes coding grammatical knowledge"? This is a very strong question
that can be reasonably answered only with the presentation of
solid evidences. It seems that we're still missing them.
To suggest that this is true, Chomsky, Fodor and several others
say that it must be so, because there's no other way to explain
the following (among several other topics):
a) That children appear to capture grammatical structures with
very few exemplars (this is related to the "poverty of stimulus").
b) That all human languages, although very different in surface
aspects, share a common interior, deep structure.
c) That only humans (and no other animal) developed language,
so it is in our genes but not in theirs.
d) More, but these are enough for now.
You may see a "common structure" in all these questions: that
they don't know how things really operate (neither the
anti-innnatists, for that matter) and, because of that,
their hypothesis must be true. I think this can be seen as
a new kind of fallacy, the one of "having nothing better to
propose".
I'm not proposing nothing here, I just want to analyze
critically what they claim.
a) Chomskians have "poverty of stimulus" backwards. It is not
"poverty", it is "excess" of stimulus. Babies are exposed to
a hectic world full of stimulus and full of potential forms
of interaction. It is not reasonable to ignore all this and
say that "language should emerge independent of these aspects".
Several activities that babies do can be seen to interfere
in future language acquisition, from simple concepts like
determining the limits of one's body (and associating this
with "I", "you") to action verbs that imply movement
("push",
"take", "give"). There is a strong regularity in all this, and
Chomskians appear to ignore the semantic influences of
these aspects in language acquisition.
b) Well, I can find quite a few things that all humans have in
common, like the way they walk, the ability with hands, the
mechanics of saccadic eye movements, the way they sneeze.
Now, is it reasonable to see all these things in common
and don't admit that they could influence the way we construct
our grammar? After all, languages were developed also to
satisfy the task of communicating ideas. Then, it is
reasonable that all languages naturally share something,
because all languages share *the same* fundamental purposes
in human communities.
c) This is not true anymore, we have several examples of
birds, dolphins, apes, etc. that are able to communicate up
to some primitive level. What's missing in these cases is a
refined, recursive grammar, able to assemble sentences
referring to third persons, refer to abstract concepts, etc.
This is what can be assigned to our larger cortex: we can
"see" deeper structures. To see how this can affect
communication, suppose we were visited by aliens that have
twice the size of our cortex. Maybe we would not understand
what they say, even if they learned to speak english, because
they would use phrases such as this:
"I thought you have believed that Mary told John
that our suspicion was that Rick found amazing the
doubt of Carter on Julie's belief".
For these aliens, that could be an "informal", easy to parse
phrase, for us it is a nightmare, although it is a grammatical
phrase. The aliens would find us "boring" like dolphins.
And just to finish, start your Dejanews and look for my
post:
Re: Do we have a data structure for mind or not? (or RDF is SVO)
23 Sept 1998
where I list some additional reasons against innateness (of
knowledge).
Regards,
Sergio Navega.
From: "Sergio Navega" <snavega@ibm.net>
Subject: Re: NN formats
Date: 19 May 1999 00:00:00 GMT
Message-ID: <3742ad2f@news3.us.ibm.net>
References: <3739A199.9E2B8209@erols.com> <3739B919.B4A6AF36@clickshop.com>
<3739C62F.527BAF91@erols.com> <3739E816.4BF79A4B@clickshop.com>
<373A143C.6A467BFF@erols.com> <373C1480.7482A5B5@tig.com.au>
<373C58B7.E700B8F6@erols.com> <373EC872.80D8C02D@tig.com.au>
<3740344b@news3.us.ibm.net> <37409FED.488F8B89@clickshop.com>
<37416784@news3.us.ibm.net> <37417C34.E43F3FDC@clickshop.com>
<3741b5ac@news3.us.ibm.net> <3741F014.9B0FF4D@clickshop.com>
X-Notice: should be reported to postmaster@ibm.net
X-MimeOLE: Produced By Microsoft MimeOLE V4.71.1712.3
X-Complaints-To: postmaster@ibm.net
X-Trace: 19 May 1999 12:23:11 GMT, 200.229.240.225
Organization: Intelliwise Research and Training
Newsgroups: comp.ai.nat-lang
Seth Russell wrote in message <3741F014.9B0FF4D@clickshop.com>...
>Sergio Navega wrote:
>
>> It seems to be a semantic question now. When we're talking about
>> innate mechanisms, I have not much to question, after all, our
>> brains are physical things that, one way or other, have several
>> characteristics determined genetically. But the problem appears
>> when we talk about innate *knowledge*.
>
>Innate *knowledge*? What would that be? You didn't hear that
>particular concatenation of concepts from me. But there *are*
>innate abilities such as the ability to rapidly learn to filter for certain
>sounds produced by air expelled over a shaped tong past teeth
>through shaped lips.
I didn't hear from you, but I thought you was questioning this.
As I said, there's no doubt we have some sort of innate mechanisms
(you may recall that I'm even more flexible than, for instance,
Bill Modlin, because I find a place for "innate feature detectors").
The big question is not only what Chomsky hypothesized (that our
brain contains knowledge of essential structures of grammar) but
also that this ability was the result of natural selection (as
proposed by Pinker and others). I have arguments against the former
and serious doubts about the latter.
>
>> What Chomsky proposed was innate knowledge of grammatical structures.
>> This is equivalent of asking "Do humans have a portion of their
>> genes coding grammatical knowledge"? This is a very strong question
>> that can be reasonably answered only with the presentation of
>> solid evidences. It seems that we're still missing them.
>
>I agree, and also I agree that rhetoric will not make us any smarter
>here, instead we need solid evidence like Kuhl 1996. I wonder if
>someone has started with these vowels and asked: What is the
>next filter that is learned?
>
Kuhl's evidence shows that brains of babies specialize their speech
recognition mechanisms after a few months of exposure to "parentese".
It is the analysis of this evidence that we seem to be discussing.
For me, this means that the initial mechanism was general (closer
to "empty") and that, due to scarce resources, had to specialize
on the recognition of the native language. This appears to be
explainable by connectionist mechanisms (Seidenberg, Christiansen,
Plaut, McClelland and others).
>> "I thought you have believed that Mary told John
>> that our suspicion was that Rick found amazing the
>> doubt of Carter on Julie's belief".
>
>Well cheeze, Sergio, aren't you molded :=) ... I would have
>thought you should have believed that Mary told John about
>our suspicions of Rick's amazement of Carter's doubt, but I
>now know Mary, and she would not tell of suspect
>amazements period, and whereas you certainly know that of
>me now, I knew (ha ha) you would still think badly of my
>old beliefs and would call them forth towards me because you
>are just down right chiggy http://www.clickshop.com/chiggy/
>... so it all comes down to that in the end ... doesn't it?
Geez! Seth Russel, you're an alien! And I didn't notice! :-)
That mouse on the page you pointed looks like me, when I
see the amount of things I have to read (nah, it doesn't
look as surprised as I am).
>
>> Re: Do we have a data structure for mind or not? (or RDF is SVO)
>> 23 Sept 1998
>> where I list some additional reasons against innateness (of
>> knowledge).
>
>If you are arguing against innateness of knowledge, then you
>are not auguing against my ideas. If, on the other hand, you are
>ignoring the importance of the patterns that exist in the external
>environment of the culture and language in which we grew up,
>then I suggest you are ignoring too much.
>
No, I'm not ignoring the patterns in the environment, because these
are the essential "food" for our brains. I'm just questioning the
"easy way out" of assuming that we already have knowledge in our
brains since we're born. That does not seem to be so.
Regards,
Sergio Navega.
From: "Sergio Navega" <snavega@ibm.net>
Subject: Re: Languages as the result of Natural Selection process (was: NN formats)
Date: 25 May 1999 00:00:00 GMT
Message-ID: <374ad742@news3.us.ibm.net>
References: <3739A199.9E2B8209@erols.com> <3739B919.B4A6AF36@clickshop.com>
<3739C62F.527BAF91@erols.com> <3739E816.4BF79A4B@clickshop.com>
<373A143C.6A467BFF@erols.com> <373C1480.7482A5B5@tig.com.au>
<373C58B7.E700B8F6@erols.com> <373EC872.80D8C02D@tig.com.au>
<3740344b@news3.us.ibm.net> <37409FED.488F8B89@clickshop.com>
<37416784@news3.us.ibm.net> <37417C34.E43F3FDC@clickshop.com>
<3741b5ac@news3.us.ibm.net> <3741F014.9B0FF4D@clickshop.com>
<3742ad2f@news3.us.ibm.net> <3742E4C2.A5100FB9@clickshop.com>
<7idv2l$jtn$1@nnrp1.deja.com>
X-Notice: should be reported to postmaster@ibm.net
X-MimeOLE: Produced By Microsoft MimeOLE V4.71.1712.3
X-Complaints-To: postmaster@ibm.net
X-Trace: 25 May 1999 17:00:50 GMT, 129.37.182.61
Organization: Intelliwise Research and Training
Newsgroups: comp.ai.nat-lang,comp.ai.alife,comp.ai.genetic,comp.ai.neural-nets
Jiri Donat wrote in message <7idv2l$jtn$1@nnrp1.deja.com>...
>I was motivated to start this discussion by the following statement:
>
>In article <3742E4C2.A5100FB9@clickshop.com>,
> Seth Russell <sethruss@clickshop.com> wrote:
>> Sergio Navega wrote:
>>
>> > No, I'm not ignoring the patterns in the environment, because these
>> > are the essential "food" for our brains. I'm just questioning the
>> > "easy way out" of assuming that we already have knowledge in our
>> > brains since we're born. That does not seem to be so.
>>
>> Ok we agree: There is no knowledge in our brains when we are born.
>
>
>This is certainly a very difficult statement to accept. Without wanting
>to slide to a "label discussion" (I mean discussion about different
>possible definitions of terms used in the sentence) I must argue that,
>for example, the construction of the brain alone is a very significant
>information by itself. This information implies abilities of our brain
>to process information and thus it heavily influences its future
>developments and even the information it can ever contain (e.g. native
>language objects). As you see, we are naturally coming to one of the
>basic philosophical dilemma - determinism vs. voluntarism.
>[snip]
>
>In these days we can see different natural languages competing each
>with other (like English vs. French) for their importance (for example
>on Internet and thus for the amount of resources they eventually win
>for their future development). This competition follows the universal
>rules of natural selection. According to the layered model of our world
>these languages should form something like a "bigger organism", e.g.
>the "family of languages", which would contain all possible forms of
>communication, incl. languages of animals and body language of animals
>and people.
>Which are the common rules fulfilled by all the members of this family?
>Again: Robustness? Redundace? Context-dependence?
>Any other suggestions?
>
I thought that you would try to defend the position of some researchers
that hypothesise the presence of some kind of knowledge in the brains
of babies. This is the point that I'd like to challenge.
It is obvious that language is subject to natural selection.
Throughout time words, syntax and semantics suffer alterations because
all the process is dynamic and collaborative. However, what is often
a point of disagreement is *where* these alterations are inscribed.
The nativists say that some of these alterations are such that make
their way into our genes, which allows the birth of "language organs",
innate grammars.
There is no neuroscientific evidence today that our brain possess
such an "organ", let alone that it is determined genetically. Those
who say that language-like organs appeared in brains because of
natural selection (competitive advantage) such as Steven Pinker
may have a hard time trying to find out the reason this natural
selection affects language organs and not other cognitive
abilities. What appears more reasonable is that brains evolved in
generic abilities (3D object representation in 2D drawings, speach
segmentation, analogical reasoning, creativity, abstraction, etc),
enough to also support impressive language learning abilities.
Language evolution appears to be situated in the context of a
civilization, and that each brain, as soon as it is born, must
learn its structures and peculiarities. The universal grammar
that Chomsky proposes is as universal as the learning abilities
of humans: something which appears to be related to our
sensorimotor activities in the world, which is really
common among all of us.
Regards,
Sergio Navega.
From: "Sergio Navega" <snavega@ibm.net>
Subject: Re: Languages as the result of Natural Selection process
Date: 27 May 1999 00:00:00 GMT
Message-ID: <374dc46a@news3.us.ibm.net>
References: <3739A199.9E2B8209@erols.com>
<3739E816.4BF79A4B@clickshop.com> <373A143C.6A467BFF@erols.com>
<373C1480.7482A5B5@tig.com.au> <373C58B7.E700B8F6@erols.com>
<373EC872.80D8C02D@tig.com.au> <3740344b@news3.us.ibm.net>
<37409FED.488F8B89@clickshop.com> <37416784@news3.us.ibm.net>
<37417C34.E43F3FDC@clickshop.com> <3741b5ac@news3.us.ibm.net>
<3741F014.9B0FF4D@clickshop.com> <3742ad2f@news3.us.ibm.net>
<3742E4C2.A5100FB9@clickshop.com> <7idv2l$jtn$1@nnrp1.deja.com>
<374ad742@news3.us.ibm.net> <927743027.249264@BITS.bris.ac.uk>
X-Notice: should be reported to postmaster@ibm.net
X-MimeOLE: Produced By Microsoft MimeOLE V4.71.1712.3
X-Complaints-To: postmaster@ibm.net
X-Trace: 27 May 1999 22:17:14 GMT, 200.229.240.213
Organization: Intelliwise Research and Training
Newsgroups: comp.ai.nat-lang,comp.ai.genetic,comp.ai.neural-nets
Tim Tyler wrote in message <927743027.249264@BITS.bris.ac.uk>...
>In a number of newsgroups, Sergio Navega <snavega@ibm.net> wrote:
>:>> Sergio Navega wrote:
>:>>
>:>> > No, I'm not ignoring the patterns in the environment, because these
>:>> > are the essential "food" for our brains. I'm just questioning
the
>:>> > "easy way out" of assuming that we already have knowledge in
our
>:>> > brains since we're born. That does not seem to be so.
>
>[snip criticism]
>
>: I thought that you would try to defend the position of some researchers
>: that hypothesise the presence of some kind of knowledge in the brains
>: of babies. This is the point that I'd like to challenge.
>
>I'd have thought that position was well enough established to not require
>extensive defense these days ;-)
>
That position (that babies are born with some innate knowledge) is
the one assumed by the majority of the cognitive scientists, among
them Pinker and others. It is that point that I criticize (obviously,
I'm doing this because I follow the steps of other respectable
researchers, such as Jeff Elman and Elisabeth Bates, who grouped
a lot of indications indicating the opposite of the majority).
>: It is obvious that language is subject to natural selection.
>: Throughout time words, syntax and semantics suffer alterations because
>: all the process is dynamic and collaborative. However, what is often
>: a point of disagreement is *where* these alterations are inscribed.
>: The nativists say that some of these alterations are such that make
>: their way into our genes, which allows the birth of "language organs",
>: innate grammars.
>
>: There is no neuroscientific evidence today that our brain possess
>: such an "organ", let alone that it is determined genetically.
>
>If by "neuroscientific" you include the effects of brain lesions, then
>there appears to be plenty of such evidence, as surveyed recently by
>Steven Pinker.
The evidences of Broca's and Wernicke's areas are suggestive, but
are not convincing and can be interpreted differently. Lesions in
these areas do produce significant impairment in speech generation
(Broca) or semantic understanding (Werkicke). This demonstrates
that these areas are the modules responsible for at least an
important part of these abilities. However, these evidences don't
say nothing about the imperative participation of other areas of
the brain during these tasks. More importantly, the evidences
*do not* suggest that these areas are present *when the baby is
born*, but just that they are there in adults. This is one of the
details that they seem to be disregarding.
>
>: Those who say that language-like organs appeared in brains because of
>: natural selection (competitive advantage) such as Steven Pinker
>: may have a hard time trying to find out the reason this natural
>: selection affects language organs and not other cognitive
>: abilities.
>
>If this were their thesis they may well have a hard time. Mental
>characteristics from intelligence, through affective states to
>preferences and instincts are all examples of cognitive abilities
>with heritable components - i.e. language is only one example of a
>mental competence with a genetic basis among many.
>
>The most obvious argument for things like intelligence exibiting
>heritable variation, is that they /must/ have done this over much of
>our recent evolutionary history, in order for these traits to have
>evolved into the state of development which they have reached in man.
>
When someone proposes that language is a selective advantage and so
this could slowly influence the genetic traits that "win" over time,
we can say that, again, that this is not conclusive.
a) Language is a fast moving target. Evolution adapts the organism
rapidly but only during infrequent changes. A climatic change can
render 90% of a population unable to survive. Given enough time,
this population will develop in a manner as to produce enough
mutations to survive another, future change. However, if another
climatic change happens *within a short period of time* from the
first, we would have another 1/10 decrease in the already
decreased population. Sequential and frequent changes would
extinguish a population *very fast*. Language change very fast
and frequently and what remains is only a very deep set of
cognitive abilities that is also shared by other parts of our
cognition (there is some aspects of object recognition that have
certain recursively embedded features, just like deep grammars).
b) It is more reasonable to suppose that the result of selective
pressures can influence generic aspects of intelligence. We
often take for granted some of our more impressive abilities
and look for language as if it was the only aspect that
differentiates us from other animals (this is typical of
Pinker).
Take 3D object recognition. Most mammals are able to do it
quite well, because it represents a strong competitive
advantage (to identify predators). But when Homo Sapiens made
some drawings in caves more than 40.000 years ago, they were
using *another* cognitive ability that no other animal seem
possess: the ability to "decode" a 2D drawing into a 3D
mental image of the scene (that is, to mentally add depth
in drawings). This is the result, in my opinion, of a larger
and better developed neocortex, the kind of thing that *can*
affect not only our ability to recognize and generate deep
and embedded structures in language but *also* a huge
collection of other cognitive abilities.
>: What appears more reasonable is that brains evolved in
>: generic abilities (3D object representation in 2D drawings, speach
>: segmentation, analogical reasoning, creativity, abstraction, etc),
>: enough to also support impressive language learning abilities.
>
>Steven Pinker explicitly rejects this view - on the grounds that there
>are individuals with sub-standard reasoning abilities in virtually all
>other areas, and yet who talk fluently; and there are also people with
>/severe/ speech disorders, whose competence on all virtually other tasks
>is good.
>--
This is a faulty argument. The fact that you can find people
with problems in other cognitive abilities and not in language
can only assert that there is a certain degree of independence of
these abilities in relation with language. This does not "prove"
that there is a language organ with innate knowledge, but only
that our brain is modular and that this modularity may be
corrupted in certain points without affecting others.
The big question is that our brain is an astonishing
self-organizing mechanism and that the critical period of the
development of this modularization is centered around initial
years of childhood.
There are reports of adult left hemispherectomy (removal of
entire left hemisphere) with severe damage to language
comprehension and generation. This could be seen as a
justification of the language organ hypothesis. However, when
this hemispherectomy is done in early age children (these
surgeries are unfortunately necessary mostly because of
uncontrollable epilepsy), these children are able to develop
language almost to a point of normality!.
It is also shown that the early the hemispherectomy is done
the better the recovery of language abilities, demonstrating
that this "language organ" does not exist since birth, but
is "fabricated" along the way.
What recent studies are trying to indicate is that our brain
evolves significantly during the initial months of life and
that a lot of our perceptual abilities are also solidified
during that time. Brain plasticity continues in children well
over 5-6 years and during that time a lot of our cognition
is developed.
My belief is, then, that language is the result of that kind
of generic learning ability that is marvelously developed
because of our interactions with the world, the kind of
interaction that also allows us to think in spatial terms,
represent the mental model of another person, think
abstractly, and a lot of other capacities that we take for
granted.
Regards,
Sergio Navega.
From: "Sergio Navega" <snavega@ibm.net>
Subject: Re: Languages as the result of Natural Selection process
Date: 28 May 1999 00:00:00 GMT
Message-ID: <374ef92e@news3.us.ibm.net>
References: <3739A199.9E2B8209@erols.com>
<373C1480.7482A5B5@tig.com.au> <373C58B7.E700B8F6@erols.com>
<373EC872.80D8C02D@tig.com.au> <3740344b@news3.us.ibm.net>
<37409FED.488F8B89@clickshop.com> <37416784@news3.us.ibm.net>
<37417C34.E43F3FDC@clickshop.com> <3741b5ac@news3.us.ibm.net>
<3741F014.9B0FF4D@clickshop.com> <3742ad2f@news3.us.ibm.net>
<3742E4C2.A5100FB9@clickshop.com> <7idv2l$jtn$1@nnrp1.deja.com>
<374ad742@news3.us.ibm.net> <927743027.249264@BITS.bris.ac.uk>
<374dc46a@news3.us.ibm.net> <927906710.752556@BITS.bris.ac.uk>
X-Notice: should be reported to postmaster@ibm.net
X-MimeOLE: Produced By Microsoft MimeOLE V4.71.1712.3
X-Complaints-To: postmaster@ibm.net
X-Trace: 28 May 1999 20:14:38 GMT, 166.72.21.185
Organization: Intelliwise Research and Training
Newsgroups: comp.ai.nat-lang,comp.ai.genetic,comp.ai.neural-nets
Tim Tyler wrote in message <927906710.752556@BITS.bris.ac.uk>...
>In comp.ai.genetic Sergio Navega <snavega@ibm.net> wrote:
>
>: That position (that babies are born with some innate knowledge) is
>: the one assumed by the majority of the cognitive scientists, among
>: them Pinker and others.
>
>: It is that point that I criticize (obviously, I'm doing this because
>: I follow the steps of other respectable researchers, such as Jeff Elman
>: and Elisabeth Bates, who grouped a lot of indications indicating the
>: opposite of the majority).
>
>I'm not familiar with their work.
>
Jeff Elman is one of the connectionists which suspected that Chomsky's
argument of the poverty of the stimulus was wrong. He have a very
interesting paper, "Language as a dynamical system", but I'd highly
recommend his book "Rethinking Innateness" (writen in conjunction
with Kim Plunket, Annette Karmilof-Smith and others). Elman
devised a kind of neural net known as Simple Recurrent Network
(SRN) which is able of learning like children, without any
initial knowledge.
Another good reference is Terrence Deacon's book "The Symbolic
Species". Deacon is a neuroscientist and an evolutionary
anthropologist and he addresses the issue quite convincingly.
>
>: The evidences of Broca's and Wernicke's areas are suggestive, but
>: are not convincing and can be interpreted differently. Lesions in
>: these areas do produce significant impairment in speech generation
>: (Broca) or semantic understanding (Werkicke). This demonstrates
>: that these areas are the modules responsible for at least an
>: important part of these abilities. However, these evidences don't
>: say nothing about the imperative participation of other areas of
>: the brain during these tasks. More importantly, the evidences
>: *do not* suggest that these areas are present *when the baby is
>: born*, but just that they are there in adults. This is one of the
>: details that they seem to be disregarding.
>
>I don't think that they are making any claim that the areas themselves
>are present when the baby is born. Their claims would correspond to
>the statement that there exists, within the DNA inside each neuron,
>a developmental predisposition for these "speech areas" to develop in
>the way in which they commonly appear in the adult form.
>
This position is clearly more reasonable than that of Chomsky and
Fodor, which admits there is "knowledge" of grammar embedded in
the DNA. But even then, this position is not conclusive, because
it could be explained by self-organizing, learning driven aspects
of the brain.
>: a) Language is a fast moving target. [...] Language change very fast
>: and frequently and what remains is only a very deep set of
>: cognitive abilities that is also shared by other parts of our
>: cognition (there is some aspects of object recognition that have
>: certain recursively embedded features, just like deep grammars).
>
>I don't agree. I'd say that phoneme recognition in infants is different
>in humans and chimpanzees, in a manner that predisposes the former to be
>better equipped to learn language. I doubt that the human ability to
>parse speech-like sounds is entirely a side effect of advances in other
>parts of cognition.
>
>If anything I'd support the reverse thesis, that other aspects of
>cognition can now benefit from association with the 'new' language organ -
>and consequently people can now bolster their spatial intuitions by
>performing serial sequential mathematical operations (for example) as a
>side effect of having learned a verbal language.
>
What you say appears to be reasonable but can also be seen under
a different optic. The question at stake is the ability of an
organism to perceive redundancies deeply embedded in sensory
inputs. Speech processing demands not only auditory processing
but also syllable and word segmentation and also word
disambiguation. This is a process which extends hierarchically
through several levels. To do this, it is necessary to have
processing power. Simple brains like that of dogs and
cats and birds cannot do speech segmentation with that level
of sophistication. But if they had more cortex, probably they
could be able to get better results. Monkeys do that and the
case of Kanzi (a bonobo ape which understands a good deal of
spoken language) is an evidence.
So what appears to be important is the size of the cortex,
the amount and quality of the neurons that are used to
process the signals.
>: b) It is more reasonable to suppose that the result of selective
>: pressures can influence generic aspects of intelligence. [...]
>
>This sounds very reasonable. However to argue that there are /no/
>neurological adaptions for speech seems perverse to me. There are
>adaptations for speaking in the vocal apparatus. As there has been
>enough time for such evolution to occur, I can see no reasons for not
>expecting to find such adaptations in the mental apparatus as well.
>
In fact, the adaptations in the speech production of humans is
claimed by some anthropologists to be the detail that drove humans
to develop language. Homo sapiens existed before 100.000 years
ago, and without evidences of the development of language at
that time. However, around 40.000 years ago, a subtle change in
the vocal tract of humans occurred, giving us the ability to
generate a *much more* sophisticated range of sounds. It was at
that time that the first evidences of language-related interactions
were found. This can be more than a simple coincidence.
It may indicate that our brain was "prepared" for language
for a long time.
Theories of dynamic and spontaneous emergence of language in a
population of competing agents (with some successful simulations
in computers) are more likely to explain the birth of language.
And this don't seem to have much to do with language organs,
but just with our ability of manipulating a lot of recursively
embedded patterns in a dynamic environment.
>: We often take for granted some of our more impressive abilities
>: and look for language as if it was the only aspect that
>: differentiates us from other animals (this is typical of
>: Pinker).
>
>Well, speech /is/ an important difference between us and other primates.
>As it may have come early, it may even have been a pivotal one.
>
Yes, as I said, I think that our abilities in speech were the
decisive factor. But we have more abilities, like our precision with
hands.
>: Take 3D object recognition. Most mammals are able to do it
>: quite well, because it represents a strong competitive
>: advantage (to identify predators). But when Homo Sapiens made
>: some drawings in caves more than 40.000 years ago, they were
>: using *another* cognitive ability that no other animal seem
>: possess: the ability to "decode" a 2D drawing into a 3D
>: mental image of the scene (that is, to mentally add depth
>: in drawings). This is the result, in my opinion, of a larger
>: and better developed neocortex, the kind of thing that *can*
>: affect not only our ability to recognize and generate deep
>: and embedded structures in language but *also* a huge
>: collection of other cognitive abilities.
>
>I wouldn't argue that 2D -> 3D mapping has developed much if anything
>in the way of specialised neural circuitry to deal with it (at least to
>the extent that it has developed in modern humans). I'd be more
>sympathetic to the idea of adaptations favouring reading ;-)
>
Or that our writing was influenced by what we had in terms
of visual abilities, which appears to be more reasonable (after all,
we invented writing to suit our needs).
>Speech is a different kettle of fish, though. Very likely quite ancient,
>and, from an examiniation of deaf peole, /extremely/ important, I can
>imagine that significant selection pressures were present and acted to
>select our more articulate ancestors. In the light of this, a 'speech
>recognition organ' appears highly plausible.
>
I agree with the importance of the speech, but not to support
a speech recognition organ. We have a powerful speech producing organ,
that appears to be the determining factor in our success in
handling language but we don't have strong evidences of a
speech organ.
Quite the contrary, in later deaf adults it was verified that
the brain *rewired* itself in such a way as to use the parts
of auditory cortex for some functions of hand movement (sign
language) and vision. This is an indication that most of the
things that happen in our brain are done in a "just in time"
fashion.
>:>Steven Pinker explicitly rejects this view - on the grounds that there
>:>are individuals with sub-standard reasoning abilities in virtually all
>:>other areas, and yet who talk fluently; and there are also people with
>:>/severe/ speech disorders, whose competence on all virtually other tasks
>:>is good.
>
>: This is a faulty argument. The fact that you can find people
>: with problems in other cognitive abilities and not in language
>: can only assert that there is a certain degree of independence of
>: these abilities in relation with language. This does not "prove"
>: that there is a language organ with innate knowledge, but only
>: that our brain is modular and that this modularity may be
>: corrupted in certain points without affecting others.
>
>If the "modules" are interchangable, then there is no reason for speech
>to prefer to reside in one of them rather than another, and disorders
>in speech with mental origins should not be heritable - they should
>rather translate randomly into disorders in other mental systems.
>
>If the modules are /not/ interchangable, and one of them (usually) winds
>up supporting language, then it would appear to be a good candidate
>for being a language organ.
>
>I don't recall Pinker presenting much evidence relating to heritable
>generic disorders influencing grammar, etc - though he /does/ cover a very
>large volume of evidence in "The Language Instinct", so it may well be in
>there somewhere.
>
What I have clear is that, based on recent neuroscientific papers,
what we find is that unused areas of the cortex often end up
being "requested" by other areas.
>: The big question is that our brain is an astonishing
>: self-organizing mechanism and that the critical period of the
>: development of this modularization is centered around initial
>: years of childhood.
>
>I seem to have missed the big question...? ;-)
>
If that's true, then you're in good company, most of the
cognitive researchers think that way. But it is time to
rethink all this question.
>: There are reports of adult left hemispherectomy (removal of
>: entire left hemisphere) with severe damage to language
>: comprehension and generation. This could be seen as a
>: justification of the language organ hypothesis. However, when
>: this hemispherectomy is done in early age children (these
>: surgeries are unfortunately necessary mostly because of
>: uncontrollable epilepsy), these children are able to develop
>: language almost to a point of normality!.
>
>Well, language doesn't /always/ reside in the left hemisphere anyway.
>
>Genes in the right hemisphere can obviously code for the "language organ"
>almost as well as those in the left ;-)
>
Nice bending of evidences you have done here :-)
This happens even with some people that have normal left hemispheres
but are lefthanded: language uses the right hemisphere. What is
common between all these cases is that the language area is close
to areas of sensorimotor activities and auditory processing, which
is one way of seeing how the problem may be solved: that language
is a process linked to motor activities, something that has been
proposed decades ago by Piaget.
>: It is also shown that the early the hemispherectomy is done
>: the better the recovery of language abilities, demonstrating
>: that this "language organ" does not exist since birth, but
>: is "fabricated" along the way.
>
>I'd agree that the organ is constructed during development.
>
>Sex organs are also constructed during development. Does the absence
>of secondary sex characteristics in children mean that these do
>not qualify as evolved adaptations? No - they are present in the
>children's genes, but not expressed in their bodies until puberty.
>
>It is the same with the "language organ": it is not initially present
>in a direct form in the brain - though it is present in the genes in the
>neurons there - but it subsequently develops in response to hearing speech
>sounds in certain contexts.
>
Interesting point you have here. But there is a sensible difference:
sex organs develop by themselves, independent of the activity of
the individual (unless one happens to get married with
Lorena Bobbit :-). Language demands interaction. When a child
learns a language, she does this very fast. When an adult
learns a language, the process is difficult. There are some
research showing that there is a critical period for the
language acquisition by children, past of which the thing
goes more hardly. Interestingly, this period is coincident with the
greatest plasticity of the brain, from 0 to 5 or 6 years. From
6 years to adult age, the individual loses its ability to
adapt to new languages.
This suggests that language is something that evolves dynamically
with the children, probably because it requires a strong
adaptation of the brain to handle its generative structure.
At that time, essential semantic structures (derived from
sensorimotor interactions) also appear to be acquired.
What could be said is that, because of our learning of language,
other cognitive tasks may be eased, by using some of the
recursive structures that language introduced.
>
>: My belief is, then, that language is the result of that kind
>: of generic learning ability that is marvelously developed
>: because of our interactions with the world, the kind of
>: interaction that also allows us to think in spatial terms,
>: represent the mental model of another person, think
>: abstractly, and a lot of other capacities that we take for
>: granted.
>
>I don't think that the thesis that language is /solely/ a result of
>generic learning abilities has much going for it. Unless language
>competence is /entirely/ a cognitive side effect, there must be a
>"language organ" of /some/ kind, however weak.
>
>If pressed, personally I'd put figures of 90% specific adaptation, 10%
>generic learning capability on the ability to speak, to indicate the
>relative importance I give to adaptations specifically directed towards
>speech.
I think that this is still an open issue, although I put myself
exactly in the opposite corner than the one you are referring to.
I'm skeptical that there is a language organ. The burden of
proof is in the hands of those who claim that this organ exists.
As I said, there are some connectionist models that are able to
learn speech segmentation and rule-like word sequences that mimick
the learning process of children without any innate knowledge.
This is a proof that it is possible to develop language without
native structures. So it is up to the ones which claim otherwise
to present strong indications of the existence of that organ.
Regards,
Sergio Navega.
From: "Sergio Navega" <snavega@ibm.net>
Subject: Re: Languages as the result of Natural Selection process
Date: 31 May 1999 00:00:00 GMT
Message-ID: <37529fce@news3.us.ibm.net>
References: <3739A199.9E2B8209@erols.com>
<373EC872.80D8C02D@tig.com.au> <3740344b@news3.us.ibm.net>
<37409FED.488F8B89@clickshop.com> <37416784@news3.us.ibm.net>
<37417C34.E43F3FDC@clickshop.com> <3741b5ac@news3.us.ibm.net>
<3741F014.9B0FF4D@clickshop.com> <3742ad2f@news3.us.ibm.net>
<3742E4C2.A5100FB9@clickshop.com> <7idv2l$jtn$1@nnrp1.deja.com>
<374ad742@news3.us.ibm.net> <927743027.249264@BITS.bris.ac.uk>
<374dc46a@news3.us.ibm.net> <927906710.752556@BITS.bris.ac.uk>
<374ef92e@news3.us.ibm.net> <927999570.637258@BITS.bris.ac.uk>
X-Notice: should be reported to postmaster@ibm.net
X-MimeOLE: Produced By Microsoft MimeOLE V4.71.1712.3
X-Complaints-To: postmaster@ibm.net
X-Trace: 31 May 1999 14:42:22 GMT, 129.37.182.70
Organization: Intelliwise Research and Training
Newsgroups: comp.ai.nat-lang,comp.ai.genetic,comp.ai.neural-nets
Tim Tyler wrote in message <927999570.637258@BITS.bris.ac.uk>...
>In comp.ai.genetic Sergio Navega <snavega@ibm.net> wrote:
>
>: Elman devised a kind of neural net known as Simple Recurrent Network
>: (SRN) which is able of learning like children, without any
>: initial knowledge.
>
>The idea that a neural network with no language knowlegde could learn
>a modern language at a reasonable rate from the inputs available to a
>child seems plausible to me. However the possibility that language could
>in principle be worked out from scratch reasonably well does not mean that
>it actually is. As almost all children learn a language I'd expect that
>the Baldwin effect would have kicked in, and allowed more rapid language
>acquisition to take place.
>
I agree. In particular, the Baldwin effect is, in my opinion, the great
responsible for the state of our society. I use this to say that
language knowledge is not in the brain as Pinker claims, but on the
world, in our society. Our brains are machines of learning and we
acquire all the habits and processes that we are exposed to. Children
capture all this since babyhood and this give the appearance that they
"had" this since birth.
>
>: So what appears to be important is the size of the cortex,
>: the amount and quality of the neurons that are used to
>: process the signals.
>
>Well, this /is/ important, although it doesn't look like the only factor.
>
>Pinker surveys the level of speech understanding of chimps - which have
>fairly large brains - and rates it very poorly.
>
It may be poor if compared to humans, but it may be good if compared
to other mammals.
>: Homo sapiens existed before 100.000 years
>: ago, and without evidences of the development of language at
>: that time. However, around 40.000 years ago, a subtle change in
>: the vocal tract of humans occurred, giving us the ability to
>: generate a *much more* sophisticated range of sounds. It was at
>: that time that the first evidences of language-related interactions
>: were found. This can be more than a simple coincidence. [...]
>
>Pinker argues that speech is even more ancient, saying that a limited
>range of vowel sounds places few restrictions on the range of sentences
>which may be used. He casts doubt on the 'language-related' interactions,
>saying these are typically things which indicate symbolic abstract
>thought - such as cave paintings - and bear no logical relation to
>speech production at all.
>
>While identifying the birth of language from fossils may not be easy - as
>many of the changes in the structure of the larynx, etc. are confounded by
>other concurrent changes, including man's upright gait - it should be
>possible.
>
>Was the "subtle change" you refer to the descent of the larynx?
>
Exactly. It is hard for us to "see" what may happen given this ability,
but the fossil records appear to indicate that this is the turning
point.
>: Theories of dynamic and spontaneous emergence of language in a
>: population of competing agents (with some successful simulations
>: in computers) are more likely to explain the birth of language.
>
>In this instance I'd give computer simulations roughly zero weight.
>If language /can/ arise in a complex brain with no "language organ" that
>doesn't mean that it actually does so in nature.
>
I agree. These computer simulations only prove that systems without
initial knowledge may be able to develop grammar by themselves.
This is an argument against Chomsky's poverty of stimulus.
>[much snip]
>
>: [...] We have a powerful speech producing organ,
>: that appears to be the determining factor in our success in
>: handling language but we don't have strong evidences of a
>: speech organ.
>
>: Quite the contrary, in later deaf adults it was verified that
>: the brain *rewired* itself in such a way as to use the parts
>: of auditory cortex for some functions of hand movement (sign
>: language) and vision. This is an indication that most of the
>: things that happen in our brain are done in a "just in time"
>: fashion.
>
>I don't see the ability of the brain to rewire itself in this way as
>any sort of evidence, one way or the other for an organ of language.
>
Again, this is really not an evidence of nonexistence, but it
is an evidence that it is possible to explain language without
resorting to innate organs. The position here is that the guys
that propose native organs say that it couldn't be otherwise.
The skeptics of innateness say that this argument is wrong,
because a brain that can rewire itself can give origin to
strong structural modifications that can account for the language
aspect without knowledge. Nobody is here really solving the
puzzle, but the position of the latter seems more tenable.
>
>: This happens even with some people that have normal left hemispheres
>: but are lefthanded: language uses the right hemisphere. What is
>: common between all these cases is that the language area is close
>: to areas of sensorimotor activities and auditory processing, which
>: is one way of seeing how the problem may be solved: that language
>: is a process linked to motor activities, something that has been
>: proposed decades ago by Piaget.
>
>Language recognition and auditory processing, I can understand.
>
>Language and motor control machinery implies speech production rather
>than speech recognition - and that the processes are closely connected.
>
>Evidence for children (and adults) speaking written passages under
>their breaths may bear on this. Without more information it's hard
>to see what, if any, the significance of this might be.
>
This is probably the most important part of our discussion.
What some propose to explain language and also the semantics is
the emergence of these aspects as a result of the sensorimotor
interactions of the child in the world. The patterns that the
child perceives of motor actions are hypothesized as being
fundamental for the semantic ground of languages.
Chomsky is known for steping aside any analysis of semantics.
His ideas were focused mainly on syntax. But what if there are
important points behind the interaction of syntax and semantics?
Some of the proposed alternatives are semantic constraints
(Jean-Loup, are you there?) which can be used to "route" our
understanding. The often quoted passage is that one due
to Chomsky:
"colorless green ideas sleep furiously"
This is syntactically perfect, but meaningless. What is
proposed by some researchers is that the child is able to
notice "deeper patterns" that stand over sensory snippets and
that constitute part of her knowledge of the world. In this
way, these inner patterns are able to direct the child's
understanding and generation of language, and the origin
of these patterns appear to be traceable to the most
elementar interactions with the world (such as grabbing
an object, throwing a ball, giving a toy to a friend, etc).
>:>
>:>It is the same with the "language organ": it is not initially present
>:>in a direct form in the brain - though it is present in the genes in the
>:>neurons there - but it subsequently develops in response to hearing
speech
>:>sounds in certain contexts.
>
>: Interesting point you have here. But there is a sensible difference:
>: sex organs develop by themselves, independent of the activity of
>: the individual (unless one happens to get married with
>: Lorena Bobbit :-). Language demands interaction. [...]
>
>Well, this doesn't seem to make much difference to the point I was
>trying to make: the absence of the language organ in a fully-formed state
>in the brain of newborn babies does not seem very relevant to whether or
>not such an organ exists in the adult - because other organs also develop
>significantly after birth.
>
>Legs strong enough for walking on do not exist in the newborn baby -
>yet few would dispute that walking is an evolved characteristic, with
>special reflexes in the foot 'designed' to assist its development.
>
>Strong legs - perhaps unlike secondary sex charactietics - don't develop
>"by themselves"; if a child is prevented from standing up for most of its
>youth it is unlikely to recover the ability to do so properly.
>
I'd agree with this line of reasoning if it were possible to physically
distinguish an area in the Broca's area from other in the frontal
lobe. Put on a microscope neurons from both areas: they are only
neurons with an axon and a huge amount of dentrites. Where is language?
Our knowledge is supposed to reside in the number, connection and
strength of our synapses, that join neurons. Recent investigations
discovered that neurons from the hippocampus of mouses subjected to
enriched environments (lots of things to do, explore, activities, etc)
were much more connected (more dendritic branches, greater number
of synapses) than the ones found in mouses that were in poor
environments. There are strong evidences that the more enriched
the environment is, the more the brain develops. This fits better
in the model of learned language, not innate.
>:>If pressed, personally I'd put figures of 90% specific adaptation, 10%
>:>generic learning capability on the ability to speak, to indicate the
>:>relative importance I give to adaptations specifically directed towards
>:>speech.
>
>: I think that this is still an open issue, although I put myself
>: exactly in the opposite corner than the one you are referring to.
>: I'm skeptical that there is a language organ. The burden of
>: proof is in the hands of those who claim that this organ exists.
>
>Whenever this "burden of proof" phrase arises, I /always/ seem to be
>on the side with the burden ;-)
:-) I know what you mean, I often put myself in that situation too.
Remember, we can't prove a negative. I can't prove that Santa Claus does
*not* exist. However, I can prove that he exists: by showing an exemplar.
>: So it is up to the ones which claim otherwise to present strong
>: indications of the existence of that organ.
>
>I'm inclined to agree; however, I'm under the impression that there
>are now sufficient strong indications - whereas you seem to see the
>indications as rather flimsy.
>
>I expect we'll have to agree to disagree on the point.
That's ok, after all this is still an open issue. But the implications
of this subject for artificial intelligence, that is my primary concern,
are significant and I'm ready to consider the evidences of innate
organs, as long as they are as good or better than the ones I have
saying that this doesn't exist.
Regards,
Sergio Navega.
Back to Menu of Messages Sergio's Homepage
Any Comments? snavega@attglobal.net