Re: USAGE: Circumfixes

Mark J. Reed said:
> On Mon, May 10, 2004 at 07:34:33PM -0500, Mark P. Line wrote:
>> Orthographic words are yet another kind of word, and they also don't
>> match
>> up perfectly with phonological and morphosyntactic words in natlangs.
>
> Ok.  Depending on the orthography, orthographic words can be pretty
> easy to identify.  But what are the criteria for isolating
> morphosyntactic and phonological words?
The short answer is that the discovery procedures for delimiting
morphosyntactic words and phonological words are not theory-neutral, and
you should consult your resident theoretician for guidance.

That said, I'll just pretend to be the resident theoretician for a bit
(definitely the long answer...):

A phonological word is a phonotactic unit that extends from one
phonological word juncture to the next. (Duh.) Languages have all sorts of
ways to mark phonological word junctures, but when you've seen all the
ways, you can usually catch on to how it works in a particular language
without too much trouble. No language is so linguist-friendly as to put a
special segment between words (although that's how we often represent the
juncture), so phonological word boundaries are actually marked indirectly
by suprasegmental or allosegmental phenomena.

In English, it's lexical stress patterns. In French, it's lexical stress
patterns on drugs. In Turkish (and most of UA, AFAIK), it's vowel harmony
and phonological stress patterns. In Japanese, it's lexical pitch patterns
as well as a number of word-final allophonic and allomorphophonemic
alternations. There are languages that are word-timed (just as English is
phrase-timed, Spanish is syllable-timed and Japanese is mora-timed), and
the timing serves to indirectly mark word juncture (can't think of an
example as I sit here). Other languages have word-internal sandhi
processes (segmental or tonal) that show where the word junctures have to
be. Etc. etc.

There are also some new-fangled neuropsychological tests that can be done
to identify phonological words, because phonological words are the most
salient units in speech.

A morphosyntactic word is a member of a class of morphological forms that
are either unchanging or which are made up of a combination of morphemes
in some coherent fashion and which distribute syntactically as a class in
some coherent fashion. It's generally the morphological processes (i.e.
processes by which bound morphemes are attached to other morphemes) -- or
lack thereof -- that wind up delineating the morphosyntactic words of a
language for you.

Swahili nouns and verbs are recognizable as morphosyntactic words because
they have a typical, relatively consistent morphological structure and
because, whatever their internal structure, they tend to distribute
syntactically as whole units. Vietnamese words are recognizable
morphosyntactically because there's very little morphology and nearly
everything has to be a word anyway. Morphosyntactic words are recognizable
in Inuktitut because of the complex morphology: certain types of morphemes
combine with certain other types of morphemes in a definitely non-random
fashion, and any useful analysis of this complex behavior will break the
language down into a number of word classes (each of which has its own
morphological processes).

The question is never *whether* the words are there, but how to build a
model of morphosyntax that helps you understand (and explain, and predict)
how the language works.


-- Mark

Reply