More LRM

In my presentation, I didn't make any mention of Item and Process
morphology, which came after IA, but has been pretty much
abandoned.  I did this for time and simplicity's sake.  An anonymous
lurker asked me about it off-list, though, so for the sake of
completeness,
I thought I'd put the relevant bits on-list (on the condition that the
writer remain anonymous).  Everything in <<...>> is to be attributed
to Anonymous:

<<
By the way, what about Item-and-Process linguistics?
 >>

In a general introduction to WP morphology, one always is
introduced first to Item and Arrangement, then to Item and
Process, and then to WP.  In my presentation, I elected not
to talk about Item and Process because it's essentially a
notational variant of either IA or early WP.

<<
What about Item-and-Process linguistics is in common with Word-and-
Paradigm that makes them both different from Item-and-Arrangement?
 >>

IP and early IA are nearly identical.  Instead of saying that a
plural morpheme /-s/ is attached to a noun, it'd say a noun
undergoes a pluralization process, the result of which is the
attachment of a plural /-s/.  IP and early WP are different in
whether or not that /-s/ was called a morpheme (in IP it was,
in WP it wasn't).

Both IP and early WP were different from LRM in that it felt
it had to deal with ordering.  So, for example, in Turkish you
have:

el "hand, sg."
el-ler "hand, plu."
el-ler-i "hand, plu., acc."

One has to be able to explain why you get /elleri/ and not /eliler/.
LRM simply states the relationship:

X  <->  Xi
N         N
nom.   acc.

And...

X  <->  Xler
N         N
sg.       plu.

And finally...

X  <->   Xler   <->  Xleri
N          N               N
nomsg nomplu    accplu

These are just patterns of the language.  So if you see a word like
/elleri/, you know that it's an accusative plural, and it can't be
/eliler/, because that pattern doesn't exist in the language (though
that could be the plural of a noun /eli/).  LRM treats these ordering
generalizations as something that simply must be learned, like
learning the word for "dog" is [dAg] and not [blIk].  Both IA, IP,
early WP and syntacto-centric frameworks like Distributed
Morphology (the latest instantiation of IA) want to make ordering
fall out from their analyses.

In IA, to do this you have to give a diacritic to each suffix.  For
example, /-ler/ would be a type 1 suffix, and /-i/ would be a
type 2 suffix.  You'd then have to state that a type 1 suffix can
only attach to a root, and a type 2 suffix can attach to a root or
any suffix that's a lower number than itself (or something similar).

Both IP and WP did this with rule blocks that it tried to force into
universal constraints.  So, for an IP/early WP analysis, you'd
say that the pluralization rule is in Rule Block A, and the case
rules are in Rule Block B.  To form a noun, you start off with
an abstract notion that has no phonological form:

{HAND, N, +accusative, +plural}

Then you go through the rule blocks.  There'd be a rule block
at the beginning that usually isn't stated which gives you the
base (presumably coming from memory):

{HAND, N, +accusative, + plural}, X -> el

Then you'd go through the rule blocks in order.  Because plurality
is more universally salient than case (by what logic, I can't say),
you first go through Rule Block A.  There are two rules in this
block:

{N, +plural} X -> Xler
{N, -plural} X -> X

The second rule is a rewrite rule which says that if you don't have
the specified features, you just return what you started with.
Frameworks differ as to whether you need the "-plural" in there
or not, and that leads to the whole big argument about privative
features vs. binary features which is present in any linguistic
theory with features that we really don't need to go into.  Let's
just say there is "-plural" for simplicity's sake.  Anyway, to continue,
the noun now looks like this:

{HAND, N, +accusative, + plural}, el -> eller

Next you go through Rule Block B which has rules that look like
this:

{N, +accusative} X -> Xi
{N, +genitive} X -> Xin
{N, +locative} X -> Xde
{N, +dative} X -> Xe
{N, +ablative} X -> Xden
{N, +nominative} X -> X

You check which feature you get, and then the accusativization
process applies, and you get:

{HAND, N, +accusative, +plural} eller -> elleri

At the end of the journey, you (hopefully) get a fully formed
noun with the correct affixes which apply to its feature set.

The main problem with IP is the ordering problem.  Since that's
a problem with IA anyway, and not a problem with IP, there's
pretty much no reason to do IP anymore, and that's why no
one does it (except for some people who think they're doing
WP but are really doing IP).  IP solutions for languages like
Georgian just left one wondering why the heck all these rule
blocks were necessary, how they were supposedly natural
and universal, and how the analysis was preferable to a simple
IA treatment, or a more sophisticated WP treatment.  I actually
have a handout I did on an early WP analysis of Georgian if
you'd like to take a look at it.  It's wildly complex.

<<
Is there anything at all that Item-and-Arrangement and Word-and-
Paradigm have in common with each other that makes them both
different from Item-and-Process?
 >>

I can't think of anything off-hand...  I'll have to think about it.

Oh, I know.  Both IA and WP treat affix ordering as something
you have to deal with.  IA does it by just positing diacritics; WP
does it by stating the ordering as something that must be learned.
IP is the one that tried to posit a universal explanation for any
and all affix ordering.  That probably was not a good idea.  There
are universal tendencies, sure, but saying that they're rules, or
that there's anything cognitive behind it is not the way to go, in
my opinion.  Unfortunately, Distributed Morphology has picked
this bad habit back up.  The DM analyses I've seen posit an
underlying tree structure for words with a fixed ordering of
affixes, and if the affixes ever appear in a different order than
that order, a movement operation must take place so that, on
the surface, the ordering is correct.  Some of the analyses are
quite..."interesting" (see Marit Julien's analysis of Northern Sami
causatives).

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

-Jim Morrison

http://dedalvs.free.fr/