From: cbmvax!uunet!math.ucla.edu!jimc
Message-Id: <9108080527.AA11680@luna.math.ucla.edu>
To: lojban-list@snark.thyrsus.com
Subject: Re: Conlangs: Languages or "Art" 
In-Reply-To: Your message of "Wed, 07 Aug 91 03:54:00 EDT."
             <m0k7iiR-00036RC@snark.thyrsus.com> 
Date: Wed, 07 Aug 91 22:27:38 -0700
Status: RO

> A conlang invented by a single person inherently must be arbitrary in
> assigning meanings.  If you invent a conlang, you know what the words
> mean because you decided the meanings.  Furthermore, you almost
> certainly define those meanings in terms of another language, usually
> English.  Thus most linguists claim that conlangs are nothing but
> encoded English.

Dave Matuszek says it well, that the attributed meanings from a "true"
language are equally arbitrary.  

More significant, few con-langs have a doctrine about definitions that
allows the definitions to rise above mere encoding of English or other
pre-existing languages.  JCB espoused the predicate logic kind of 
definition, in which a bridi is true or false depending on whether or
not its sumti are related as specified by the selbri.  I believe that
this doctrine continues in Lojban, though I haven't seen a whole lot of
enthusiasm recently for using it explicitly.  

This doctrine implies to me that the definition of a selbri IS a list
of sumti sets which are thus related (monads, pairs, triplets, etc.
depending on number of places, and neglecting complications about
modals).  Then to interpret a phrase the listener computes database
operations on definition lists.  No recourse is needed to pre-existing
languages; a language defined this way is in no way a code for another.

Humans have said to me "I don't have a SQL server in my brain", but I
don't claim to be modelling brain activity except in the most general
sense that the model gives (about) the same answers as live humans would,
yet the model can be understood theoretically and can be taught to humans
(and mechanicals) in simple terms and without mandatory contamination from
other languages.  

> From the start, we have stressed in Lojban that it is the speaker's
> responsibility to express things in terms understood by the listener
> rather than vice versa...  
>    ...the idea is that the meanings of words must be outside the
> control of individuals, or you have a code.  

Both points are important.  Yet the second seems neglected recently,
namely, the speaker must rely on the listener to interpret his words
according to standards SHARED by both.  Thus the speaker is not allowed
to emulate Humpty Dumpty and say "zdani" means whatever he wants, but
also is not required to ask the listener what he means by "zdani";
there is supposed to be something outside both of them that prescribes
what "zdani" means.  In a preliterate language tribal customs dictate
the meanings of words, whereas in a con-lang there is generally a
codebook issued by the language architect.  For a pure non-encoded
language like Lojban one can imagine a stack of videotapes where some
person holds up objects and says "ti cipni .i ti cribe", then goes on
to demonstrate multi-place gismu such as "citka".  (Although I doubt
this approach will really replace the code-file recently placed on the
langserv machine.)

Getting onto my favorite topic, the central authority of Lojban
prescribes the meanings of all the gismu, but central prescription is not
possible for lujvo both because of doctrine (which I enthusiastically
agree with) and because so many lujvo are possible.  Do we want the
meanings of field-built lujvo to be idiosyncratic with the speaker (and
differently idiosyncratic at the listener)?  If not, we need an
algorithmic procedure to derive the meaning of a lujvo from its rafsi /
gismu.

This is not to say that y'all have to adopt my diklujvo rules.  The
first mountain that has to be climbed over is to get people to
understand that rules are needed.  

		-- jimc