Re: HELP: Adding irregularities to conlangs?

Carsten wrote:

<<I'd be grateful for tips & tricks.>>

It wasn't too long before I became bored by regulariy (in a language
that was supposed to be *natural*.   Naturally [excuse the pun], if the
language is supposed to be an IAL, or a logical language, irregularity
might not be something to be hoped for).   At the time, though, all I could
do to introduce irregularity was add things at random or imitate the
irregularity of other languages.   And that's not a bad way to start:
After all, other languages are irregular in specific ways, and, to some
extent, irregularity is random.   Recently, though, I was introduced to
what I think is a better way to go about it.

If you're introduced to morphology as an undergraduate (and that's
a big if, I've learned), you're introduced to this kind of morphology:

elleri = "hands (acc.)" in Turkish, and is composed of:

el- = "hand"
-ler = plural
-i = accusative

This works out rather nicely for Turkish (90% of the time), but it doesn't
for a lot of languages.   There's a really nice example of why this doesn't
work using an example from Tundra Nenets.   It's extremely complex,
though, and I don't have my handouts ready, so I'll describe the phenomenon:
Tundra Nenets has eleven noun cases, I believe, but every noun had one
thing in common: There was no way to predict what the accusative
plural would be.   Some ended in /-o/; some in any number of consonants;
some were accompanied by a vowel change; some by fortition of a previous
consonant.   At most, you could group one or two nouns into a class and say,
"For nouns of class 1204884, you voice the last consonant and add /-a/", but
then you'd have hundreds of classes.   It seems like a pretty odd way to
build
a language.

However, if you look at the system as a whole, you'll see that no matter what
the accusative plural is, every oblique plural case is built off it.   So
let's say you
had some case endings like:

-l = illative
-? = translative
-N = adessive

Then all you'd need to know to form the plurals of these is a simple rule:

(1) Take the accusative plural and add the case ending.

What the language learner has to do, then, is learn three things: (1) The
nominative
form of the noun; (2) the genitive singular form of the noun (there are other
reasons
for this that I didn't explain); and (3) the accusative plural.   If you
simply memorize
those three forms, you can predict the entire paradigm for every single noun.

At the outset, though, just glancing at the Tundra Nenets case system makes
it look
totally unpredictable (especially before you shave off all the phonological
changes),
and totally irregular.

There are two things in this: (1) There are some irregularities that simply
can't be
explained without history (e.g., why the plural of "child" is "children"),
but there are
others that only look irregular superficially--really irregular--and this is
what one
wants to try to caputre; and (2) knowing this, how does one imitate it?

I've found that the principles of Word and Paradigm (WP) Morphology can
really help
in this area.

The main thing (from a language creator's perspective) to remember about WP
is this:
Each what has a paradigm it needs to fill.   This paradigm is composed of
every possible
expression this word--or, rather, the lexeme, the meaning associated with
that concept--
can be used in.   Every language needs to fill every cell of that paradigm
(since all languages
can express every concept equally well [has this been proven?]).   The
question is: How
will your language fill that paradigm?

There's no question that affixation is rampant throughout language.   But
rather than
thinking, "What will the plural affix be in my language?", think, rather,
"How can plurality
be expressed in my language?"

My first attempt at putting WP to work for conlanging was with my revamped
language
Gweydr.   It was the second language I ever created, but it was really,
really bad.   So I just
took the name, the script, and some of the vocabulary and built a whole new
language
with it.   So back to plurality.   Well, before I got to plurality (I'd
already decided that I only
wanted singular and plural to be expressed synthetically.   This means that
in the rest of the
number paradigm, the cells will be filled with paraphrastic expressions, like
in English, where
the dual of "chair", is "two chairs", and the quadral is "four chairs", sexal
"six chairs", etc.), I
had to deal with noun cases.   I decided on what noun cases I wanted to use
(not yet on how
to use them), and decided that my core cases would be:

Nominative
Terminative
Partitive
Genitive

(Terminative is a kind of Accusative plus Dative plus...uh, Terminative [kind
of a goal-related
case].)

The rest of the cases were oblique cases.

One thing languages do is they tend to (not always) like to make a
distinction between the core
cases and the oblique cases.   I thought one way to do this could have to do
with plurality.   Borrowing
an idea from German, I decided that some words would have an umlauted stem.
So,
for the word /roTl/, "hawk", for example, there are two stems: [rOTl] and
[r9Tl].

In addition to an umlauted stem, I also wanted to have a plural suffix: A
holdover
from the old language, /-ks/.   This suffix now displays some phonological
alternation,
but I won't get into that.   Anyway, it was then that I decided to combine
these three
bits to form the noun case system.   Now there are (basically) three types of
nouns:

-Type 1: These words are regular, so to speak.   They have a single stem and
take
the suffix /-ks/ in the plural.

-Type 2: These words are the umlauted words.   They have two stems: A
singular
stem, to which case prefixes are added to form the singular cases, and an
umlauted
stem, to which case prefixes are added to form the plural cases.

-Type 3: These words are mixed.   They have a regular stem and umlauted stem,
like type 2 words, but they're used differently.   In the singular core
cases, the
regular stem is used plus the core case prefixes.   In the singular oblique
cases, the
umlauted stem is used plus the oblique case prefixes.   In the plural core
cases, the
umlauted stem is used plus the core case prefixes.   In the plural oblique
cases, the
umlauted stem is used plus the oblique case prefixes *plus* the plural
suffix.

If you tried to analyze this system in an Item an Arrangement way (the oldest
form of morphology, still used today, in varied forms), it would look like
Type 3
nouns are getting double plural marking in the plural oblique cases, since
umlaut-
ing is a kind of plural marking, as is the plural suffix.   This isn't so,
though.   What's
happening is that the speakers of Gweydr have to fill this paradigm somehow.
Since some words have two different stems, they've used them in two different
ways.   After all, each cell is distinct from each other cell.

The point of this is to get away from a compositional approach to language
creation.   Rather than thinking, "What piece can I make to stand for this
meaning, or this conglomeration of meanings", one thinks, "I have to express
this set of features (let's say, "house", plural, accusative): How can I do
it?"

Anyway, learning about WP helped me to add irregularity to my languages
that I was satisfied with that didn't end up looking like irregularity for
the
sake of irregularity.

An interesting question you posed, Carsten:

<<How do
irregularities evolve naturally, anyway?>>

I suppose it depends on the specific phenomenon.   It can happen in tons of
ways.   I think the common theme is that...no, I don't know if I can even pin
it down.   The speakers of the language change.   How about that?

Nik wrote:

<<One way is to figure out the proto-language, particularly phonemes that
may have merged.>>

That's probably the best way to go about it.   Not just for phonology, but
for
everything.   As things change, though, keep paradigm leveling in mind (the
artificial regularization of a paradigm that has become irregular.   It
happens).
A good introduction would be to look at Lyle Campbell's book Historical
Linguistics.   It lists all the different types of sound changes there are,
and
gives examples; goes into the histories of various languages; gives likely
candidates for lexicalization (i.e., what kind of things turn into what kind
of
things over time), etc.   If you start from scratch (i.e., a proto-language)
and
have a good explanation for every change, though, you'll probably find the
result highly satisfactory and very realistic.

-David
*******************************************************************
"sunly eleSkarez ygralleryf ydZZixelje je ox2mejze."
"No eternal reward will forgive us now for wasting the dawn."

-Jim Morrison

http://dedalvs.free.fr/

Reply