On Tue, 24 Feb 2009 18:54:35 -0800, David J. Peterson <[log in to unmask]> wrote:

>On Feb 24, 2009, at 6°22 PM, Alex Fink wrote:
>> Leti has two classes of verbs.  In class II, which is unmarked, the
>> subject
>> prefixes are subject to the metathesis and apocope binding
>> processes that
>> take place at most word-internal juncture: so a verb like [sOpla]
>> 'sail'
>> with 1st singular prefix [u-] becomes [swOpla].  In class I, which
>> verbs can
>> get into for phonological or morphological or lexical (i.e. no)
>> reasons, the
>> binding processes are blocked: so [Beli] 'buy' has 1sg. [uBeli].
>> Voila,
>> infix and prefix as allomorphs.
>
>I don't think that's the best way to analyze that, but if you did,
>formally
>the two are now distinct, and have little if any relation to one
>another.

Just to be on the same page: what's "not the best way to analyze that" is
calling them allomorphs?  (I wasn't especially meaning to support a
morphemic analysis there, just to describe what would result from one.)

>> Well, if that's what disturbs you, why not say the masculine ending
>> is /-o/
>> in _Carlos_ as in every other word, and it just happens to have an
>> extra
>> morpheme /-s/ tacked on after that?  (Meaning, I dunno, '(masculine
>> proper
>> name)', but happens not to appear on most masculine proper names.)
>
>That is *precisely* what the formalism forces you to do: to posit
>some sort
>of meaning for the /-s/ there.  And doing that is patently absurd.

True.  

>I never said anything about WP, of course

Indeed; that was an interpolation on my part from memories of a previous
thread.  

>In order for it to be a
>workable linguistic
>framework, it must be constrained in some way, and before I left grad.
>school a few colleagues were coming up with some interesting ideas on
>just how to do this.  I don't think it will come to anything, but
>it's a start.

I'd be interested to see that, if their ideas panned out to any extent.  

>> I think it's too much to ask that one's theory of morphology rule
>> out such
>> absurdities by purely formal means.
>
>Unfortunately, that really is the goal of formal theoretical
>linguistics: to
>formally exclude that which can't exist, while formally explaining that
>which can.

Well, yeah.  And I expect there are many interesting insights to be had
along those lines, if you can get at how the human language faculty works;
and there are places where I think that goal is a perfectly good one.  But
I'm skeptical about trying to do it in too all-encompassing a fashion.  

For one, and this is something like what I was saying before, it's very
binary, either you exclude a given possible grammar or you implicitly bless
it by not having done so.  I'd like it more if it was something more like
describing a probability distribution on possible languages, to recognise
the fringe cases as fringe.  (E.g. say that, for some reason, the foonlitude
of a natural language is a (0,1)-normally distributed real variable, and so
all of the languages in our sample have foonlitude less than 4.  This
doesn't mean you should look for a reason foonlitude >=4 is impossible...)

I think there'll be some essentially hard constraints on human languages
that arise from properties of the underlying hardware and software, and then
a good deal of softer constraints that arise from other things, either
softer constraints on the language faculties ('you _can_ do X but it's a
PITA to process'), or probable patterns of historical change, or other
influences yet I don't foresee.  And if you have to formalise everything in
or out, well then what about factors of the latter sort?

Ah well, maybe I've been a little tainted, turned against noble ends, by the
amount of twist-the-world-to-fit-the-theory that goes on in practice :-(

>> I would suspect that for every one who simply accepts the idea of a
>> morpheme
>> and proceeds to do that, there's another three who haven't even
>> gotten that
>> far in linguistic background, and are doing that because it's the
>> easy thing
>> to think of and implement.  
>
>I agree it is simpler--for those who have a language like English as
>their only
>language resource.  This isn't the case for all conlangers.

Unless your alternative is something I hadn't thought of, I did mean
objectively simpler, in some sense that transcends whatever language or
culture we might happen to be using: I might restate this and say that it
requires fewer tokens of code to write a program that catenates "men" to the
end of a string than to do something with allomorphs etc. etc.  (Assuming
our bases are represented as character strings.  Maybe this is the product
of a cultural assumption, but then it would seem to be a more deep-seated
one...)

Native language bias is another thing, as is native grammatical tradition
bias.  Is English itself susceptible enough to a morphemic analysis that it
leads people into this sort of thing, do you think?  Or is it the way
English is taught?  Or both, or neither?  

What do folks do who have some other language as their only resource?

>Why on earth should regularity be confined to morphemes?

Oh, it shouldn't, not by any means; of course your example with the [batol]
and the [latob] was regular.  That was just an extension of my suggestion
that it's easy --- if you want _everything_ to be regular and easy to learn
as a design principle, well then saying the word for A and then saying the
marker for B if you mean A marked for category B is pretty easy.  

Alex