Re: The Naturalist Manifesto revisited

Joerg wrote:

<<The recent discussion on whether Tolkien was a good conlanger,
and what constitutes a good conlang, brought back to my mind
the manifesto that Jesse Bangs posted two years ago
(yesterday was its second anniversary) under the subject line
"Lighting Some Flames: Towards conlang artistry".>>

Has it really been two years?!   Goodness...

<<snip everything>>

If this is the original post your referencing, you should recall that Jesse
intentionally overstated his claim, in order to foster discussion.   I think
what fell out of the discussion that followed was that:

(1) Yes, a language should be judged by the goals of its creator.   This was
an attempt to define the goals of the naturalist school (of which I'm a part,
if one can say there is such a thing.   I call chancellor!).   What was not
resolved was the feeling that some have had that a natural conlang might be
better than a nonnatural one.   This probably goes on a case-by-case basis, and
can truly never be resolved, because it's comparing apples and oranges.   Who
can say which is better amongst the two?   Well, the eater.   Though many might
prefer both (or neither).

(2) I believe the point of scale was discussed.   I can't remember what fell
out from that discussion, but it was probably something similar to the
discussion this time.   I think what makes this argument difficult is that, compared
to any other artform, it's hard to make a small conlang.   I mean, there are
haiku, minute-long songs, minimalist visual art, etc., but what is a small
conlang?   If you're actually creating a language, you have to eventually come to
terms with the fact that the language will need to be able to express
everything that's expressible in the given worldview you're operating under.   Since
there are no small worlds (and since the mind is infinite, in a way), it'd be
hard to imagine a language where you only needed like three words and no
compounds.   (Maybe the language of the little prince, when he was living on that
tiny planet?)   So, in one sense, yes, it is very true that quantity doesn't
equal quality.   On the other hand, though, languages are not small things:
They're very large things.   It's hard to quantify this.

(3) I'm also fairly certain that the "creativity" claim was hedged in much
the way it was with our discussion.   The point was to emphasize how someone who
codes over English, or whatever their L1 is, is not creating a good,
naturalist conlang.   Of course, it doesn't follow that you *have* to try to create a
language that's totally different.   In fact, I'd argue that it'd be just as
bad if your native language was English, but you went out and found a Dyirbal
grammar and basically coded *it* over.   It's the act of coding-over that he
wanted to single out   here.   Everything should have a reason behind it that's
not, "Well, that's the way I think it should be", when the result of "that's
the way I think it should be" turns out to be identical to the creator's L1.
This doesn't prohibit intuitions about the way the language works, or working
by "feel".   An example I can give is double case marking in Zhyler (there's
an actual term for this--a German term--but I can never remember what it is).
I was trying to decide how to deal with causative constructions, especially
where there's more than one causative marker.   So, in Zhyler, you have...

sajal = "to die"
sajasal = "to cause to die, or to kill"
sajasasal = "to cause to cause to die, or to hire a hitman to kill someone"

Many languages deal with such constructions by continually demoting the case
of the former direct object.   So, in Turkish if you were to say the following
(I don't know Turkish, so I'll do this in English):

The king-NOM died.
The man-NOM the king-ACC killed (=caused to die).
The woman-NOM the man-ACC the king-DAT hired to kill (=caused to cause to
die).

So, as you can see, the hierarchy is NOM>>ACC>>DAT and after that there's an
oblique postposition /tarafMndan/.   This is how most languages do it, and I
started to do something like this, but then I noticed two facts about Zhyler:

(1) /r/ > [z] /rV(C)_
(2) Accusative = -ar/-er/-r

It seemed to me like this was screaming for double accusative marking.   Why?
  Let's go back to Zhyler and see how these sentences work.

(1) petti sajlar.   (king-NOM. die-PAST) "The king died."
(2) sexa pettir sajaslar.   (man-NOM king-ACC die-CAUS-PAST) "The man killed
the king."
(3) ZijkM sexar pettirez sajasaslar (woman-NOM man-ACC king-ACC-ACC
die-CAUS-CAUS-PAST) "The woman hired the man to kill the king."

And there you go.   It's as if the phonology had been set up to license this
double accusative marking, because the accusative /r/ changes in exactly the
position where double-accusative marking happens.   This is a regular sound
change, and applies everywhere in the language, but I thought it was neat that it
worked out this way here.

So I don't think this claim is going against being able to use instinct in
conlanging.   In this case, I went with what I thought worked with the language;
what I thought would sound right.   It clearly wasn't based on English, and
also not on Turkish, so I wasn't coding anything over: I was just working with
the language as it presented itself.

Now, let's take a negative example from Megdevi (I pick on this language a
lot).   When working with complex structures, I needed to find a way to handle
ECM verbs, and the like.   So, in order to say something like "I want you to
eat an apple", I thought, "How could I do this?   Ah!   I know: Let's have a
particle that is inserted right before the imbedded clause."   So I ended up with
something like this:

?oj dZ@l@si lIZoj dZ@r@bi metsIlIm.
I     want-pres. COMP-you eat-pres. apple-ACC.

But then I thought, "It doesn't sound right that that second verb is in the
present tense."   This was probably based on the fact that in English you don't
say, "I want he eats the apple".   So, I changed it to a kind of infinitival:

?oj dZ@l@si lIZoj dZirejb@t lImetsIl
I     want-pres. COMP-you eating ACC.-apple

(The accusative changes when it's the object of a verbal noun.)

Now I   had something which sounded in English like, "I want you to eat the
apple", where you have an infinitival construction for the second verb.   But
notice what happens when you substitute "hope" for "want":

?oj h@l@fi lIZoj dZ@r@bi metsIlIm
I     hope-pres. COMP-you eat-pres. apple-acc.

So you have comparable sentences, but in one you get an infinitival, and the
other you get the present tense.   Coincidentally, this corresponds *exactly*
to English.   So you say:

"I want him to eat the apple", not *I want that he eats the apple.
"I hope that he eats the apple", not *I hope him to eat the apple.

Based on the grammar I developed for Megdevi, these two types of
constructions should have been treated identically.   But, as I was trying to work this
problem out, I thought it was really neat that I could use either a verbal noun
or a conjugated verb, and it would "sound right".   What this meant was, "It
sounds like English, and therefore must be right."

That's a bad example of intuition, and that's what Jesse was saying should be
avoided, and I agree.

All right, back to my OT final.

-David