Re: backwards conlanging

On Tue, 28 Nov 2000 15:34:27 +0100, daniel andreasson
<daniel.andreasson@...> wrote:

>Yoon Ha wrote:
>
>> I have a stupid boring /i/ /e/ /a/ /o/ /u/ vowel system,
>> with two diphthongs.  I guess I'll have to figure out
>> something more complicated that could've simplified down
>> to the 5-vowel system.
>
>Or why not the other way around? The protolang might have
>had only /a/ /u/ /i/, and then /u/ and /i/ became [o] and
>[e] in some contexts. Then, in the future, Chevraqis (or some
>dialect) could develop front rounded vowels, or length
>distinctions, or nasality, or... well you get the point.
Yes, indeed.

But there was a reason why I pointed to that possibility first.

It seems that phonemes mostly multiply through various syntagmatic
changes (that is, changes that depend on specific sound environment).

But then the paradigmatic changes come into play: symmetrization,
elimination of less functionally loaded phonemes, and the like. They
mostly lead to a major simplification.

For me, both types form a kind of cycle.

My point was that when you need to *decide* on what changes you want,
it's easier to proceed from the paradigmatic changes: they are more
directly connected with the target state of the system, and allow you to
choose the types of syntagmatic changes you need for the previous stage.

Another cool thing with paradigmatic changes is that they can produce
weird correspondences.

To take an example resembling yours: suppose your target system is the
5-vowel standard, and you wish to derive it from the simple a-i-u triangle.
You are thinking about some syntagmatic change making each proto-vowel to
split into two phonemes, e. g. one more closed and the other more open.
So you get an intermediary system:

  a
e @ o
i   u

You need to eliminate one element, and the simplest solution is to
get rid of what is absent in the target system, the [@]. But if you
consider its position in the system, you'll immediately see
that there are several ways to do so. Moreover, they can be
applied simultaneously: for example, @ > I > i in some
environments but @ > o in others.

However, you can think of something more tricky. For example, you've
noticed that e and o are least frequent vowels, so merging them will
cause least trouble. Then you can consider e. g. the following:

@ > a ('cause it's natural for central vowels to get more open)
a > A > O (to preserve the contrast with former @) > o (to restore the
symmetry)
o > u (to keep it distinct from the narrowing O)
u > I (since it's not former o that you want it to merge with) > e

Now you have two closely related dialects with rather untrivial
correspondences between them in certain words: [i] ~ [a], [u] ~ [e].

I like playing with sound changes like these!


Basilius