Re: Harnad1 Word Origins

From: Stevan Harnad (harnad@coglit.ecs.soton.ac.uk)
Date: Thu Dec 09 1999 - 08:06:15 GMT


cg> Yesterday some of us questioned whether dance is a language or
cg> not. I think we should refer to our debate here. In my view, the principle
cg> feature of all languages is that they can express what we mean, e.g. our
cg> thoughts. It is for this reason that we have to mention thinking here.

Fine, but be careful about the direction of the expressivity:

(1) The focus was on "propositional content" -- that is, sentences
asserting that something is the case, something that can be, say, true
or false. (Questions, commands, etc. are just special cases.)

(2) I did not say that every feeling or experience could be expressed as
a proposition. (A picture is not only worth a thousand words, but an
infinite number of words: How do I even describe what "red" is to
someone who has not seen red?)

(3) So there may well be experiences that cannot be expressed in the
form of a proposition. Perhaps the feeling I have when I shrug my
shoulder with regret is not fully described by words. Maybe the act
itself, and seeing the act, always expresses more.

(4) But that was not the question! It was not whether there are other
forms of expression that can express things that cannot be expressed, or
cannot be fully expressed, in words. The question was whether dance,
music, etc. could express everything that can be expressed in words!

cg> One`s thoughts can easily be divided into groups, for example, judgements,
cg> sentiments, emotions, excitations, etc.(All right, this is an
cg> old-fashioned categorisation.) These are the thoughts one can will to
cg> express precisely. But we all know that there are different ways to
cg> express one`s thoughts. For example: the dance.
cg>
cg> Starting from this point of view one can see some difficulties
cg> concerning our definition about language, because this latter is based
cg> on intertranslatability. We can easily find some words (elements of
cg> language) that cannot be translated with a single word. In that case
cg> one can translate it with a sentence. But what obstructs me translating
cg> the elements of dance with a sentence? In my view: nothing.

You are perfectly right, but the direction of the translation is wrong!
The challenge is to express in dance what you can express in language.
It is not surprising (and it is to the credit of language, not dance)
that you can express (or at least approximate) in language what you
express through dance.

cg> We saw that
cg> ways of expression that differ from language (arts!) have their inner
cg> structure, nearly parallel to the grammar (Laban theory, theory of
cg> colours, etc.) The difference between language and dance (arts?) comes
cg> from nothing but from the difference between their objects.(When we
cg> want to express our philosophical thoughts, we will use language; when
cg> we want to express our love for example, we will chose dance (arts?).)

There are many forms of expression, but we were talking about the
expression of propositions (or rather the meaning of propositions).
Dance and music can express many things, but not propositions; porbably
propositions can approximate (poorly in some cases) most of the things
that can be expressed through dance or music.

What does "express" mean, after all? I express frustration if I shout.
When I shout, I feel I have expressed what I feel. And you have felt it
too, so you understand. That is the analog world of mimicry, and perhaps
"resonance" that we spoke about today.

But language is something different. It is a conventional notation
system whose symbols can be systematically intrepreted as expressing any
possible proposition -- any possible thought that can be put into words
(that is a little circular, but not viciously so).

Focus now on whether the translation can go into the direction
language --> dance.

cg> I think it also worths mentioning, and will be useful, to refer to L.
cg> Wittgenstein`s Tractatus Logico-Philosophicus 3.343.

Although you did not provide it, thanks to the Web, I found it at:

http://www.kfs.org/~jonathan/witt/t3343en.html

Welcome to the skywriting, Ludwig:

  lw> 3.343
  lw> Definitions are rules for the translation of one language into another.
  lw> Every correct symbolism must be translatable into every other according
  lw> to such rules. It is this which all have in common.

Correct. And how do we translate "the cube root of 27 is 3" into dance?

cg> I do not know what remedy can be found to treat (mistreat?) our definition
cg> of language. First, maybe we have to take attempt to answer the question:
cg> What is language?, instead of saying: English is a language.
cg> Well, I do not know what language is...

If we don' know what a lnaguge is and isn't, then we don't of WHAT we are
ask about the origins, or of WHAT we are asking about the adaptive
value.

  sh> It is also a rather vague conjecture that a language as a whole is
  sh> open to multiple interpretations -- say, English as it is, versus
  sh> "Fenglish," in which the meanings of "true" and "false" are swapped
  sh> and all other meanings are suitably adjusted so that everything
  sh> remains coherent: If one said "`That is a rabbit' is true" in
  sh> Fenglish, "true" would mean what false means in English, but only
  sh> because Fenglish "is" means what English "isn't" means, "rabbit" means
  sh> "non-rabbit," and so on. Yet in order to have English and Fenglish
  sh> speakers continue to discourse with one another coherently, in the
  sh> same world of objects and events, without ever suspecting that their
  sh> words don't mean the same thing, so many adjustments seem to be needed
  sh> that to conjecture that the deception is even possible may be
  sh> equivalent to assuming that formal "duals" of meaning exist (like the
  sh> duals of logic and mathematics, where it can be proved that certain
  sh> formal operations can be systematically swapped under a transformation
  sh> in such a way as to yield coherent dual interpretations)[2]. Such a
  sh> strong conjecture calls for a proof, and as far as I know, no one has
  sh> offered a proof of the existence of semantic duals in language.
cg>
cg> Perhaps, there is no proof of semantic duals, but not all multiple
cg> interpretation is semantic dual. For example, when I use the word
cg> `category`, I have in mind Kant`s concept; I suppose every
cg> other interpretation of this word will be more or less different from mine,
cg> even Kant`s interpretation.

We were not speaking of multiple interpretations of a WORD but of an
entire LANGUAGE. That means at least two (radically different) ways to
interpret every possible sentence in a language.

But this is a rather abstruse point, so let us pass on.

  sh> Apart from the absence of a formal proof, another reason for
  sh> suspecting that coherent dual interpretations of languages may not be
  sh> possible is that the systematic adjustments would have to go beyond
  sh> linguistic meaning. They would have to encompass perception too, and
  sh> would thereby inherit the problems of the "inverted spectrum"
  sh> conjectures (e.g., Cole 1990): Could you and I be walking around the
  sh> same world speaking and behaving identically, even though I see the
  sh> sky as blue and the earth as green, whereas you see the sky as green
  sh> and the earth as blue? Again, if our classifications are always
  sh> approximate, it may be a long time before we discover the
  sh> difference[3]. But if there is ever a difference, it will disambiguate
  sh> us forever. And until then there's no difference between identity and
  sh> approximate identity -- or at least no difference that makes a
  sh> difference -- no uncertainty on which any actual outcome depends.
cg>
cg> Just a short note. All muslims and christians have the same god: Allah/God
cg> (gavagai/rabbit!). Maybe these differences make no linguistic difference,
cg> but in reality...

To an approximation Allah/God mean the same thing. Of course among
people the differences can outweight the commonalities, but that is a
matter of value, not meaning.

But this point is abstruse, so perhaps we should not dwell on it too
much.

The reason I introduced the idea of duals was just to point out that
finding even ONE coherent interpretation of all the symbols and symbol
strings in a language is a highly nontrivial task (in cryptology);
finding several is even more daunting...

  sh> Whorf's original example concerned the Hopi language, which
  sh> apparently lacks a future tense.
cg>
cg> On the ground of this theory there is no future tense in Hungarian. But we
cg> know there is.

Hat akkor ezt biztos nem fogod megerteni: Penteken delbe gyulunk a
Godorbe.

(But we agree about this.)

cg> The point here is that one can have the `idea` of future even if in there
cg> is no future tense in the language. Language is a way of expressing our
cg> thoughts. We can express nearly all of them; so, we are able to define
cg> some grammatical structure in order to make our language to be capable of
cg> expressing our thoughts.

Even simpler: you don't need a formal future tense to express time or
futurity in words.

  sh> Let us recall, at this midpoint of our discussion, the informal
  sh> criteria we adopted for language at the outset: A language is an
  sh> approximately intertranslatable system for approximately categorizing
  sh> the world. We have spoken a little about categorization and
  sh> approximation. What about intertranslatability? Names of perceptual
  sh> categories are trivially intertranslatable. All one need do is coin a
  sh> gloss: "rabbit" = "gavagai."
cg>
cg> In my view it works only if both languages `know` rabbit/gavagi. One can
cg> even describe a rabbit to another person, but it seems a bit doubtful that
cg> the latter will `see` (understand totally and properly) what a rabbit is.
cg> Just try to describe a colour to one who have never seen that.

It is no use trying to describe something if there is no shared
vocabulary of prior categories of which it can be seen as a combination.
I can describe a rabbit to someone who has not seen one before because I
can describe its parts, which he has seen. A color or taste that is
totally unlike one we already know (if that is possible) could not be
described in words. It would have to be seen, and then given a
ground-level category name.

  sh> This three-level representational system is grounded bottom-up in
  sh> psychophysical categories. "Top-down" influences occur through CP:
  sh> Similarity judgments are not mediated purely by iconic representations
  sh> (or perhaps iconic representations are not pure): Belonging to the
  sh> same category -- i.e., having the same name -- makes things look more
  sh> similar, and belonging to a different category (different name) makes
  sh> them look more distinct. Most of the symbolic component consists of
  sh> internal translation: "An X is a Y," "A Y is a Z," etc. It is the
  sh> primitive symbols, which are grounded in nonsymbolic representations
  sh> -- iconic and categorical ones -- that prevent this symbolic circle
  sh> from being vicious. [11] Two different systems of grounded symbolic
  sh> representation are in principle intertranslatable and their respective
  sh> groundings can be tested against ostensive experience in the real
  sh> world.
cg>
cg> I agree with these statements. Nevertheless, I have to ask whether all
cg> representation is in this close connection to language/speaking, or not.
cg> While thinking, everybody have representations that can be expressed
cg> sometimes by words and sentences, but there are some that we do not want
cg> to express, or even that cannot be expressed by using our finite language.
cg> Maybe there are categories of thinking behind our three-level system.

Can you give me some examples? (Be careful not use any words!)

  sh> So my hypothesis about the origin of words is really a hypothesis
  sh> about the origin of symbolic categories: They originate in sensory
  sh> categories, and are grounded in the iconic and categorical
  sh> representations that make it possible for you to pick out those
  sh> sensory categories.
cg>
cg> Our reasoning seems to be culminated at a speculative, purely theoretical
cg> hypothesis about the origin of words. In this sense it describes (only)
cg> theoretically the genese of words. And if we add that behind the quoted
cg> representations one can also find representations (indipendents of
cg> speaking), one shall say that we must go further and analyse the
cg> connections between our three-level system and other representational
cg> system(s?) of our thinking, in order to show how these latter one(s) make
cg> our first system possible.

It is possible to go beyond this vague speculation with actual working
models. Here are some examples:

    Harnad, S. Hanson, S.J. & Lubin, J. (1995) Learned Categorical
    Perception in Neural Nets: Implications for Symbol Grounding. In:
    V. Honavar & L. Uhr (eds) Symbol Processors and Connectionist
    Network Models in Artificial Intelligence and Cognitive Modelling:
    Steps Toward Principled Integration. Academic Press. pp. 191-206.
    http://www.cogsci.soton.ac.uk/~harnad/Papers/Harnad/harnad95.cpnets.html
     
    Tijsseling, A. & Harnad, S. (1997) Warping Similarity Space in
    Category Learning by Backprop Nets. In: Ramscar, M., Hahn, U.,
    Cambouropolos, E. & Pain, H. (Eds.) Proceedings of SimCat 1997:
    Interdisciplinary Workshop on Similarity and Categorization.
    Department of Artificial Intelligence, Edinburgh University: 263 -
    269.
    http://www.cogsci.soton.ac.uk/~harnad/Papers/Harnad/harnad97.cpnets.html
    
    Greco, A., Cangelosi, A., & Harnad, S. (2000) A Connectionist Model
    for Categorical Perception and Symbol Grounding. Connection
    Science. URL TO COME

cg> What about the origin of thinking?

Mental states, including planning, recognizing, remembering, expecting
-- all came before language. But perhaps reasoning, theorizing, and all
the other off-line things we can do with symbols did not.

--------------------------------------------------------------------
Stevan Harnad harnad@cogsci.soton.ac.uk
Professor of Cognitive Science harnad@princeton.edu
Department of Electronics and phone: +44 23-80 592-582
Computer Science fax: +44 23-80 592-865
University of Southampton http://www.cogsci.soton.ac.uk/~harnad/
Highfield, Southampton http://www.princeton.edu/~harnad/
SO17 1BJ UNITED KINGDOM



This archive was generated by hypermail 2b30 : Tue Feb 13 2001 - 16:23:06 GMT