Re: Language superiority, improvement, etc.
From: | Matt Pearson <mpearson@...> |
Date: | Thursday, October 15, 1998, 18:28 |
Joshua Shinavier wrote:
>> >There *are*, however, properties of languages which *can* be compared quite
>> >effectively, logic and complexity being two of the easiest to define a
>> >means of measurement for --
>>
>> I'm not so sure about that. The last time this issue was discussed on
>> Conlang, we tried to figure out some objective criteria for measuring
>> relative logic and complexity, without much success (or at least, *I* was
>> unconvinced). The only real quantifiable feature that people could come
>> up with was relative number of morphologically irregular forms.
[snip]
>What I had in mind (i.e. my particular conception of "the complexity of a
>language") is more along the lines of the amount of syntactical information
>neccessary to speak the language -- the complexity of the "mental
>programming"
>which correct use of the language demands. My choice of that particular
>"definition" is probably influenced by my A.I. work (I've had to actually
>*program a human language* into a machine; details of complexity become very
>obvious here!), but in the abscence of any other more concrete definition
>(what
>is "irregularity"?), at least the more "logical" languages like my Danoven or
>the various loglangs become measurable, as they have fixed, visible rules
>to be
>followed.
I guess I can see the applicability of this definition to Artificial
Intelligence work, where you're comparing the ability of natlangs and conlangs
to perform a highly specific task - namely, be computer-programable. But
(as per your comments below) your definition is not applicable to comparing
one natlang with another, given that we still know essentially nothing
about how natural languages are 'programmed' into the mind. (And
evaluation of natlangs
is what we've been discussing. At least, I *think* that's what we've been
discussing, or have I gotten lost along the way? These long threads can
get very confusing... :-) )
I mean, I suppose we could go ahead and *define* complexity in terms of
the amount of 'mental programming' required to use a language. It's just that,
given how little we know about the neurology of language, we're not yet at
a stage where we can apply that definition in a useful way (at least to
natlangs).
The observation that all natlangs are acquired by children at the same rate
*suggests* that all natlangs require roughly the same amount of 'mental
programming', but we'll have to know a whole lot more about the brain before
we can be sure of that.
>For natural languages things become much less clear, as the speakers
>of the language are themselves not fully aware of the rules or influences
>that
>go into their choice of a certain pattern in speech
Agreed!
>The "logical" nature of a language, I would say, can be described (and
>quantified, if desired) fairly well by the degrees to which it relies on
>>clearly describable, discrete constructs
What are these constructs, and how are they described?
>I don't know if anyone else has any suggestions, but that is in any case
>my idea of logic and complexity in language. To figure "irregularity" in,
>I'd
>take the *simplest possible description* of any given set of grammatical
>rules
>(and exceptions, which are also rules), and call "irregular" (if you want
>to >get some sort of use out of the word) those branches with especially
>limited >fields of applicability.
Sounds reasonable to me.
>Irregularities do increase the complexity of a grammar, but they are
>>*definitely not* the sole determining factor.
Agreed! In fact, I'd go so far as to say that they're a trivial factor -
hence my comment earlier that irregularity is not a suitable criterion for
defining or judging relative complexity.
>Try programming a computer to speak English some time; you'll gain, if not
>a >computer which speaks English, an aggravatingly clear view of
>linguistic >"logic" and "complexity" in a practical sense ;-)
Well, you don't have to convince me of the difficulty of programming a
computer to speak English! :-) But I don't see what that has to do with
the question of whether objective criteria of logic/complexity can be used
to compare one natlang with another natlang. It may be true that it's
massively difficult to program a computer to speak English, but I'm sure
it's equally difficult to program one to speak Mandarin, or Ilocano,
or Lebanese Arabic.
I grant you that such an exercise would give you insight into *one particular
approach* to defining logic and complexity. I'm just saying that that
approach is not useful for evaluating natlangs against each other. At
least not yet.
Perhaps when we know more...
Matt.
------------------------------------
Matt Pearson
mpearson@ucla.edu
UCLA Linguistics Department
405 Hilgard Avenue
Los Angeles, CA 90095-1543
------------------------------------