Re: On the design of an ideal language

On 5/2/06, And Rosta <and.rosta@...> wrote:
> > Jim Henry, On 01/05/2006 23:52:
> >> Can you elaborate on how your conlang(s) manage
> >> noise resistance at the level of phonetic realization?
> >
> > Simply by making sure that the phonetic realizations of contrasting
> > phonemes or phonological strings are sufficiently distinct acoustically
> > in the mouth of an averagely lazy speaker. I base this on personal
> > judgement here, not on facts about phonetics.
> > For example, in my strategy I might have /i/ and /E/ but no /e/. In your
> > strategy, you might have all three, but have a rule that /e/ blocks /i/
> > and /E/, and /i/ and /E/ block /e/. Potentially the two strategies end
> > up with comparable levels of redundancy.
Yes, probably.

An early draft of this engelang I'm working on
had a phoneme inventory where no two phonemes
differed by less than two distinctive features, and
voicing was not distinctive (all were voiced in
shouting-over-background-noise mode and all
unvoiced in whisper-mode).  But
it was both cacaphonous and verbose, and didn't
have much room to grow the vocabulary with
reasonably short words, so I abandoned that
phonology early on.


> > I would advocate relying on just the contrast in at least two
> > distinctive features, and applying it just to consonants.
So, how and why apply it only to consonants?
Do you mean that /pi/ and /po/ should
not be considered sufficiently distinct because
both of the differing distinctive features (height
and backness) are in the vowel, while /pi/ and
/mi/ would be distinct enough (manner of articulation
and voicing)?  What about /pi/ and /be/?  (voicing
of the consonant and height of the vowel) or
/bi/ and /bo:~/ (differing by height, backness,
roundness, length and nasality)?

It's possible that this principle would lead you
to ensure that words are highly distinct
vis-a-vis their consonants, while the vowels
are phonemic schwa that's realized as this
or that phonetic vowel depending on the surrounding
consonants.  That seems wasteful, though.

> >>> >> 8. Principle of Concision.
> >>> >> The language should be as concise as possible *on average*. As a
> >> I suspect that my current engelang may evolve in that direction
> >> from its isolating grammar in phase 1.  The most common
> >> two-word phrases being replaced by new monosyllabic
> >> words in the next phase, the isolated grammatical particles
> >> would "fuse" with the words they occur most frequently
> >> in connection with (though still having a stand-alone form
> >> for use with less common words).
> > My conlanging experience is that such 'string substitution' devices are
> > less effective than alternatives.
Alternatives such as...?  Fusional marking
of the most common categories?  Fusionality
does lead to concision, but the problem
is, I'm not sure how to combine fusional morphology
with the noise resistance criteria.  I could have,
for instance, root morphemes like

kok
tak
pik
...

and fusional suffixes like
im
on
aN
...

each of which codes for two different
grammatical categories, say evidentiality
and mood on a verb; but the noise resistance
criterion *decreases* the number of morphemes
available at any given length, and fusional
morphology *increases* the number
of short morphemes you require as compared
with isolating or agglutinative morphology,
so they seem to be in conflict.  Maybe the
ideal balance would be to fuse the most
frequently marked categories and mark
others with isolating particles or agglutinative
suffixes.  Or you could get better use of a
small set of available morpheme-shapes
by fusing the categories that don't have
full combinatorial possibilties -- e.g.
evidentiality and mood (maybe only in
indicative mood would you distinguish
different evidentialities, while in imperative
or interrogative mood you would not
mark evidentiality, or at least not need
as many distinct evidentialities),
but leave combinatorial categories isolating
or agglutinative (e.g., any tense can
be used with any aspect).


On 5/3/06, Jackson Moore <jacksonmoore@...> wrote:
> On May 3, 2006, at 1:08 AM, Jackson Moore wrote:
>
> > On the other hand, maximizing redundancy within a distributional
> > category, and minimizing it between them, is a great idea if you
> > want to make the grammar itself audible (acoustically salient).  As
> > natural languages are most definitely not cut out for this task,
> > this is a great reason to construct one (it's certainly one of my
> > primary motivations).
>
> Sorry, this is ambiguous depending on whether you're talking about
> redundant features or redundant distinctions - I meant to say that in
> this case you'd want to minimize redundant distinctions (maximize
> shared features) within a distributional category and maximize
> redundant distinctions (minimize shared features) between them.
>
> At any rate, the inverse relationship of lexical ambiguity and
> categorical ambiguity vis-a-vis phonology is interesting.
Yes.  For instance, if you have part of speech
marking as in Esperanto or Ilomi, that increases
the length of words without adding any
noise resistance.  Really maximizing the
phonological distinctions within distributional
categories requires giving up part of speech
marking and thus increasing the burden
of memorization, and it also requires
giving up self-segregating morphology,
which requires a Livagian-style lexicon
design  (no morpheme is a prefix or suffix
substring match for another morpheme)
if you want to avoid parsing ambiguity.

I am inclined to balance part of speech
and semantic category marking against
noise resistance and aim for an optimal
mix of their good qualities, rather than
maximizing one at the expense of the
other.  One possibility would be to
make sure that all of the part of speech
markers differ from each other by
at least two distinctive features or
two phonemes or whatever
E.g., you might have (if two distinctive
features is the criterion)

-a	transitive verb
-i	intransitive verb
-u	noun
-em	modifier
-oj	preposition

--
Jim Henry
http://www.pobox.com/~jimhenry/gzb/gzb.htm

Replies