Complexity in language

[Isidore Dyen]

----------------------------Original message----------------------------
On Fri, 26 Jun 1998, Wouter Kusters wrote:
 
> ----------------------------Original message----------------------------
> At 15:26 24-06-98 -0400, Isidore Dyen wrote:
> >....I should add that, as I see it, theories can no be
> >tested--they can only be revised or replaced--but the hypotheses based on
> >a theory can be tested. Obviously a theory on which no hypothesis can be
> >based is not worth proposing.
>
> My use of the word 'theory' was indeed a bit old-fashioned Popperian, and
> in fact I completely agree with this Kuhnian idea.
>
> >Perhaps the greatest complication in measuring a language directly is the
> >apparent incommensurability of its parts. How can the inventory and
> >distribution of the phonemes, which a appear to be measurable, be measured
> >so that it is commensurable with the morphology, the syntax, the lexicon,
> >and/or the semantics and how are the latter four to be reduced to
> >commensurability. The theory of equicomplexity implies that these
> >structures, when measured in different languages, will somehow form an
> >equation.
> Two languages like Russian versus Navaho are indeed very difficult to
> measure on their difference in complexity. But why not taking two more
> closely related languages which differ only on one level, and in which the
> difference is obviously one of complexity (take Anem vs. Lusi in New
> Britain according to Thurston, or Shaba Swahili vs. Zanzibar Swahili). How
> does the theory of equicomplexity account for that?
> I think in these cases the theory of equicomplexity or at least its related
> hypotheses make the wrong predictions. In my opinion this theory can only
> make right statements, when you do not know what complexity actually
> comprises.
> The moment you define (a subpart of) complexity as e.g. 'irregularity in
> the morphology', (i.e. semantically intransparent relations between meaning
> and form), you can see that the intuitions of ordinary people, and the
> problems which arise in second language learning, and the structural
> differences between languages which have a status as lingua franca and
> languages which have a more 'ethnic' status, all point in the same
> direction, i.e. that there are differences in complexity between languages.
 
It appears to me that trying to compare the complexity of two
similar languages does not lead in the direction of a measure. Languages
do the same thing as means of oral communication.
The measure of complexity would indicate a difference of efficiency, that
is a difference in the amount of effort to achieve the same result. Since
the mechanism using a language (the human being) is taken to be the same
everywhere, the difference in efficiency could reasonably be identified as
a difference of complexity. Effectively then, if natural languages are
equiefficient, then they are equicomplex. All of them with the exception
of the creole langaguages are products of millennia of change. It would be
difficult to detect any change in their efficiency in the past, but the
same consideration would suggest that equiefficiency and thus
equicomplexity was the order of the day. It may be hard to believe that
this was always trueand therefore the way to test this assumption might be
to see in what way it could be true, i.e. what measure leads to an
equation.
 
> >At the same time it should be remembered that the ntuarl languages that we
> >are dealing with are the product of a long period of evolution that did
> >not produce better languages, as far as we can tell, or, for that matter,
> >worse languages.
> This may be true but has nothing to do with the complexity of languages.
> Bacteria are simpler than mammals, but this does not mean that the one is
> better or worse accommodated to the circumstances. In the world of language
> also, there exist different circumstances, under which languages prosper.
> The circumstances under which a 'contact language' grows are different from
> the circumstances of an 'ethnic' language.
> So, I think you can find differences in complexity, on the condition that
> you do not equal the complexity of a language with its 'value'. Measuring
> complexity may be hard, measuring the 'value' of a language is even harder.
It is rather easy to see differences in complexity among languages in
parts of languages. The problem concerns the totality of languages.
>
> Wouter Kusters
> University of Amsterdam.
>