Re: stress

#1 wrote:

<<
I'm now trying to find a good way to create the stress for my conlang

But I'd want it to be natural AND regular
 >>

Maybe Tom Wier can help me out on the details, but in my opinion,
the one and only thing optimality theory (OT) is good for is creating
regular,
interesting stress systems.  [Note: I don't mean that Tom should back
me up on my opinion, but that I'm far removed from OT, and certainly
don't want to look at it again, so what I'm about to explain might not
be right.]

According to OT, all stress systems are regular.  If there's what looks
like
an irregularity in the language, it should fall out from ranked
constraints,
even if there's only one exception.

Here's a crash course:

In OT, the way one pronounces a word is the most optimal way that word
can be pronounced.  Any other way is less optimal, for a variety of
reasons.
So, for example, if you want the lexeme "Canada", it would be extremely
inoptimal to pronounce it like the English word "dog".  In order for
this to
fall out, English would have a constraint that was something like "FAITH
IO", which means "what's in the input should be in the output".  In
this case,
the input is [k_h&n@R@], and the output is [dAg].  Thus, "FAITH IO" is
violated like six times.  Compare that to the form [k_h&n@R@].  That
violates
it no times, because what's in the input is in the output.  A form like
[k_h&nd@]
could be better, though, because it only violates it twice (I think).

You can summarize this as follows:

Input: [k_h&n@R@]   FAITH IO
[k_h&n@R@]
[k_h&nd@]                   **
[dAg]                            ******

The little asterisks are violations (I'm sure that won't line up).  So
that's basically how it works.

Now on to stress.

For stress there are a number of relevant constraints.  These are a few:

1.) STRESS TO WEIGHT: Violated if there's a stressed light syllable.
2.) WEIGHT TO STRESS: Violated if there's an unstressed heavy syllable.
3.) WORD HEAD RIGHT: The head foot of the word should be on the right
edge.
4.) WORD HEAD LEFT: The head foot of the word should be on the left
edge.
3.) FOOT HEAD RIGHT: The head of every foot should be on the right edge.
4.) FOOT HEAD LEFT: The head of every foot should be on the left edge.
5.) BUILD FEET RIGHT: Start building feet from the right edge.
6.) BUILD FEET LEFT: Start building feet from the left edge.
7.) BINARY FEET: Feet should have exactly two syllables.
8.) PARSE SYLLABLE: Build all syllables into feet.
9.) BIMORAIC FEET: A foot can have no more than two moras.

All right, this should give us enough constraints to play with.  Now,
here's
a form: /kan.da/.  It has no stress so far.  Let's assume that the word
is
built exactly like I say (so one heavy syllable followed by a light
syllable).
Stress can only be in one of two places: On the first syllable, and on
the
second.  By ranking the nine constraints above in various ways, stress
will be on different syllables for different reasons.  So, let's say
we're going
to build our feet from the left, but that the word head is on the
right, and
the foot head is on the left.  Now let's say PARSE SYLLABLE is highly
ranked, but that BIMORAIC FEET is ranked ahead of BINARY FEET.
That gives us the following:

(kan.)(da)

The parentheses indicate footing.  So this is a word that has two feet
and two syllables.  So, the word head is on the right.  This means that
the right most foot should be where the main stress falls.  That gives
us the following: /kan.'da/.  Now for more fun.

Given this bizarre stress system, let's say that STRESS TO WEIGHT
outranks WEIGHT TO STRESS.  Now let's say you have two com-
peting forms for the input /kan.da/:

a. /kan.'da/
b. /kan.'da:/

STRESS TO WEIGHT is violated if a light syllable is stressed.  Thus,
form (b) now looks better than form (a).  [Note: This will only hold
if there are other constraints which allow a vowel to be lengthened,
but prevent adding a coda [n], or something.  Assume those hold.]

So, what does this mean for your language?  This means that you
may have a bunch of words where two syllable forms that end in
a short vowel where stress is final and the vowels are lengthened.
This would be a bizarre language, but a feasible one, given the
ranking of the constraints.

In order to use this idea to create a stress system, you don't need
to actually do OT and use the constraints, and all that.  It can be
helpful, and it can help you understand what implications your
stress system has.  (I.e., given that ranking, what would happen
to /kan.dan/ and /ka.dan/ and /ka.pa.ta/, and /ka.pan.ta/, etc.)
But it is by no means necessary.

For Gweydr's stress system, I employed some basic notions I'd
learned from OT.  For example, there's something called a "default
to left" stress system.  It does exist in the world's languages.
What it does is it puts the stress on the leftmost syllable, unless
there's some reason not to.  And, of course, a given language
will define what those reasons are.  I did this for Gweydr.

For Gweydr, I came up with some basic rules I wanted to see
happen:

(1) I wanted a lot of multisyllabic words to be stressed initially.
(2) I wanted some words to be stressed penultimately, if they
"sounded right" (i.e., if they were heavy enough).
(3) I didn't want basic weight to rule the system.

So, this is what I did.  I created three basic ordered rules:

(1) Stress the ultimate syllable if it's superheavy.
(2) Stress the penultimate syllable, if it's heavy.
(3) Otherwise, stress the left-most syllable.

Where the irregularity comes is in how I define "superheavy".
Heavy is bimoraic.  So, since Gweydr doesn't have long vowels,
it means a vowel plus a coda.  That's heavy.  Superheavy,
however, is a vowel plus two coda consonants, but *not*
when the two consonants are a voiceless stop followed by /s/.
This was crucial, because the plural marker in Gweydr is /-ks/,
and if it counted as superheavy, all plurals would be stressed
finally (same thing with Spanish, which makes an exception
for a word ending in /-s/).  This gives you the following:

(a) /dZanu/ > ['dZa.nu] "leg"
(b) /dZanu-ks/ > ['dZa.nuks] "legs" (not *[dZa.'nuks])
(c) /kelin&/ > ['ke.li.n&] "sister"
(d) /hagoltS/ > [ha.'goltS] "warrior"
(e) /kibentiks/ > [ki.'ben.tiks] "eyes (terminative case)"

This was the system I wanted, but with one flaw.  I had
words that I really wanted that didn't work.  For example,
I really wanted the phonological form of the word "sword"
(for whatever reason) to be [e.'zi.ni], with stress on the
penultimate syllable.  According to my system, stress
should be on the first.  What to do?

This is where you can play around to make exceptions.
Another thing I wanted to have was syllabic liquids (as
evidenced by the name Gweydr, which is pronounced
['gwej.dr=]).  If one could have syllabic liquids, why not
syllabic glides?

Back to the Gweydr word for "sword", then, I decided
that the *phonological* form would be /ezinj/, with a
coda palatal glide.  And, when preceded by a consonant
in word-final position, glides become syllabic, so /j/ >
[i] and /w/ > [u].  For the purposes of stress, though,
it sees a disyllabic word with a superheavy final syllable.
Thus, it puts stress on the vowel of the final syllable, which
is the [i].  Then the coda glide becomes a vowel, and you
get what I wanted: [e.'zi.ni].

Note, though, that the stress system has remained regular.
Now, should a stress system be regular?  No, not necessarily.
That's how you get languages with lexical stress.  There are
always tendencies, but then you get "pervert" the noun vs.
"pervert" the verb (though this does belong to a general
paradigm of noun/verb alternations).  Right now we're
looking at a language in my field methods class called Moro
which we're pretty sure is a stress language, but which is
so wildly unpredictable that I doubt OT could handle it.
Nevertheless, there are tendencies.  Give me a Moro word,
and I'll have a good shot at guessing where the stress is.
(Of course, give me a seven syllable word, and I'll have to
guess *how many* primary stresses there will be.  It's no
tone language, but it tries its hardest to be!)  Nevertheless,
methods like these can give you a naturalistic sounding
stress system that's still regular, if you want a regular
stress system.

-David