--- On Thu, 8/13/09, Jim Henry <[log in to unmask]> wrote:


> Hmm... there are several halfway decent ways to do
> cryptography with
> mental arithmetic and randomizers you might find ad-hoc
> wherever you
> might be, besides the really powerful cryptography you can
> do if you
> have a computer or even a calculator and pen and paper on
> hand; the
> advantage of this project, if it would be to have any
> advantage at
> all, would be to have more or less cryptographic properties
> while
> using only the language centers of the brain, not requiring
> the
> speaker to do conscious extralinguistic arithmetic or
> random number
> generation in order to say anything.
> 
> I suspect if we just put the multiple allomorphs of each
> common
> morpheme in free variation, though, each speaker at first,
> and the
> speaker community as a whole eventually, would tend to
> unconcsiously
> use one allomorph more often than others because it's
> easier to
> remember for whatever reason, or to develop at first
> subtle, then
> gradually more major differences in meaning and usage
> between them.
> 
> Writing the language with a syllabary rather than an
> alphabet might
> help in making all graphemes equally likely to occur in any
> position
> within a word or sentence.   That might make
> the phonological design
> of the language less constrained, as well, if you're
> primarily
> interested in flattening the statistical properties of text
> rather
> than speech.
> 

I'm think I'm confused about what you are trying to accomplish.

You don't want to force (strongly encourage?) users to avoid using the same word for a particular meaning? You don't want each consonant and vowel phoneme equally likely* to be used throughout your lexicon?

* Not the same as random, though that's what people tend to think of as random.

Lee