Hallo conlangers!

On Friday 14 October 2011 09:01:31 R A Brown wrote:

> On 13/10/2011 21:34, Roman Rausch wrote:
> >> It *is* possible to start with *one* descendant of
> >> your protolanguage and work backwards from the
> >> descendant to the protolanguage, but that is definitely
> >> harder than working forward, and when you start with
> >> two or more "descendants", it is very likely that the
> >> roads refuse to meet.
> > 
> > To share a trick of the trade - you start out with a word
> > which appeals to you in any of your daughter languages,
> > reconstruct the proto-form, and derive cognates in the
> > other languages from it. I've learned that from Tolkien.

Yes, that works quite well, and I do this in my conlangs quite
often.  I invent a word in Old Albic, and reconstruct its Proto-
Albic and Proto-Hesperic ancestors.  Of course, to do that, you
must already know the sound changes for your languages, and the
phonology and the morphology of your protolanguage.

> Tolkien was a competent linguist and knew exactly what he
> was doing.

Surely he was.  And that's why we can learn so much from him.

> I think all generalizations in this area have to be treated
> with care, whether its "start with protolanguage and derive
> daughter(s)" or "do what Tolkien did" or some other
> strategy. It depends a great deal IMO on the individual.

Yes.  My focus in my work on the Hesperic family is not that
much on Proto-Hesperic but on Old Albic, which is just one
member of the whole family; but I already have a good idea
what Proto-Hesperic would have looked like, and how it
changed into Old Albic.  I still discover new details about
that development, though; only recently, I discovered that
Proto-Hesperic *s has become *d after *n, *l, *r in Albic,
for instance.

> [snip]
> 
> > In order to cross from one language into diachronic
> > conlanging, I've often heard the suggestion 'just make
> > your language the proto-language and derive new ones from
> > it'. But I don't think this is good advice at all. Who
> > would want to make a language close to one's heart into a
> > dead one?

The language close to my heart is not Proto-Hesperic but 
Old Albic, which is not "deader" than Latin - it is no longer
spoken as a native language, but well-attested, and has living
daughter languages.

What regards designing a protolanguage with future developments
in mind, it is of course perfectly legitimate and even advisable
to choose the protolanguage's phonology such that the intended
target phonology is reachable with a reasonable number of
plausible sound changes.  You wouldn't start with a Polynesian-
style phonology if your target is a language that could blend
well into the linguistic landscape of the Caucasus.  Sure, the
speakers of the protolanguage would be unaware about the sound
changes that would happen, from their viewpoint, in the far
future, but you as a conlanger can do what you want with your
conlangs, and set up the protolanguage such that the target
languages are easy to reach.

> Yes, but I was advising someone who was/is having problems
> deciding the phonology of the protolanguage.  My advice was
> to concentrate on pinning down that _phonology_ first and
> then worry about sound changes.

I don't know how much the OP "knows" yet about the conlangs
he is going to make.  If you have a good idea of where you
want to go, then, as I say above, you can (and even should)
fix the protolanguage to fit.  If not, you can start with
any reasonably naturalistic language you want, and decide
what sound changes happen to it later.
 
> There's no need whatever IMO to do the _whole_
> protolanguage, i.e. morphology, syntax & vocabulary first -
> though some idea of the sort of morphology (not necessarily
> the detail) may influence decisions you make about the of
> the phonology.

Very much so.  My protolang Proto-Hesperic is a very sketchy
matter.  All I have are a phonology, a few derivational and
inflectional suffixes, and a number of roots.  Most of the
stuff was actually reconstructed from Old Albic and PIE
(Hesperic is meant to be a sister family of Indo-European).
 
> > I have an adapted suggestion: Make your language only
> > little changed from the proto-language (as little as
> > would be realistic), and then derive descendants from the
> > latter.

Yes.  As for my project, Old Albic is a fairly conservative
descendant of Proto-Hesperic, and I can reconstruct Proto-
Hesperic from that quite well.
 
> It's a suggestion - but IMO there is no one "correct" way of
> going about this.  A lot must depend on an individual's
> linguistic competence and familiarity with the way languages
> do develop diachronically and also, not a little, on a
> person's own inclination.  The main thing is that a
> conlanger should feel comfortable with the approach s/he is
> taking.

AMEN!  Don't try something you don't understand.  Of course,
if you want to try it, go to a good library and check out some
textbooks on that matter.  There are plenty good textbooks on
historical linguistics; one of the best is written by Lyle
Campbell and simply titled _Historical Linguistics_.  Also
useful is a handbook of a major language family (such as
Benjamin W. Fortson IV, _Indo-European Language and Culture_)
to get an impression of how such a language family works out
in practice.  And of course, there are plenty of experienced
diachronic conlangers here on CONLANG; just ask us.
 
> >> If you want to design a language isolate and leave it
> >> at that, you need not worry about protolanguages and
> >> diachronics, though even in such a situation the
> >> diachronic method *will* add depth to your language.
> > 
> > I don't quite understand how one can avoid diachronics
> > even with one language. Synchronic irregularities within
> > a language mostly descend from regular diachronic
> > changes;
> 
> That's right. If you're after naturalness - and presumably
> you would be if you're concerned with protolanguage &
> daughter languages - then some irregularities are likely to
> appear - tho natlangs do vary very considerable in the
> amount of irregular form, some have very little and other
> seem to relish them. Compare the verb systems of Turkish and
> ancient Greek!
> 
> If you have only a modest number of irregularities, you may
> not need to go too far into diachronics.

Correct.  If you want irregularities in your conlang, go for
the diachronic approach, because irregularities are always
remnants of formerly productive patterns that were fouled up
by language change, especially sound change.  The English strong
verbs, for instance, resulted from a neatly regular pattern in
Proto-Indo-European that was more and more distorted by sound
changes until the forms became unpredictable.  If you want to
design a nicely regular language such as Turkish or Quechua,
you can indeed make do without much diachronics.

--
... brought to you by the Weeping Elf
http://www.joerg-rhiemeier.de/Conlang/index.html
"Bêsel asa Êm, a Êm atha cvanthal a cvanth atha Êmel." - SiM 1:1