On 11/13/2001 05:11:36 PM Lars Marius Garshol wrote:

>That I didn't understand. How can anyone claim that the term "segment"
>is independent of theory?

Segments are considered pretheoretic in the sense that they are more or
less undisputed data. There is a gestalt aspect to them in that we
naturally conceive of them as minimal units -- and it's natural that we
would do so since they are speech sounds that are minimally pronounceable.

>| Features of pronunciation are so small they can't be independently
>| articulated,

Yes, though Peter D's comments on simultaneity are also important.

>and if the characters of a script have graphic elements
>| that correlate with such tiny features of pronunciation they are
>| featural. Right?

Well, I'd say that there's a problem with this definition because "have
graphic elements that correlate with such tiny features of pronunciation"
is open to differing interpretations. That's what I'm wanting to see
pinned down.

>Aha! So I can't pronounce a feature on its own simply because a
>feature is just an aspect of a sound, and not an entire sound.

Bingo! You have just passed the first mid-term of phonology 101.

>Oh. So there isn't a closed set of features? There are different
>theories proposing different sets of features?

There are some differences. There are a number that would be undisputed
among phonologists today. I haven't kept current in phonology in over ten
years, but I've had the impression that the questions that have interested
phonologists have not been with pinning down the "right" set of features
but rather with most abstract issues having to do with relationships
between features (feature "geometry"), and with insights that can be
gained by allowing some representations at some levels to have incomplete
feature specifications, as well as other areas that are less directly
related to features.

>Right. Are features always properties of phonemes, meaning that a
>phoneme can be fully and uniquely characterized by its features?
>So {[+ voice], [+ labial], [- continuant]} == 'b', for example?

That's roughly the idea. There are complications we could go into, but not
by the first mid-term of phonology 101 (e.g. underspecification, the
validy of the notion of "phoneme", etc -- for purposes of our discussion
of scripts, this stuff isn't really relevant)

>Let me ask one test question to see if I understand this: is the the
>tall/deep distinction in Shavian featural?

I need a definition before I can answer that.

>Do you mean to say that types in a typology and prototypes (defined by
>means of an example) are the same thing?

I'll let Peter D give an explanation of "type" since he is more concerned
about the distinction between that and "class" than I.

"Prototype", which I introduced into this discussion refers (in this
particular technical usage) to a way of defining things that applies to
some concepts. There are some meanings that can be defined in terms of
clear-cut application or non-application of some criteria. For example,
"heifer" might be thought of as being defined by the semantic features [
bovine, female, young ]. But there are some meanings that simply are not
amenable to this type of analysis. Lakoff talks about various ways in
which things can be defined, what he called different "idealised
conceptual models". One of these is the prototype definition. The idea of
the prototype definition is that the set of object to which that term
applies does not have sharp boundaries. There are things that are clearly
in that category, and things that clearly are not. But in between there
are other things for which judgments are less obvious. For example, a
chair that has four legs, a seat, a back and arms is clearly an instance
of "chair". That is a good prototype that serves as a sample of the kind
of thing that clearly belongs. A bed clearly is not an instance of
"chair". Now, is a stool a kind of chair? Is a love seat a kind of chair?
My grandfather had a cane with a two-handle-like attachment that folded
over to provide a seat; is that a kind of chair? For these kinds of
objects it's less clear. We might be able to come up with a set of
semantic features (similar to what we did for "heifer") that clearly
applies to the prototypical objects and that provides a valid basis for
distinguishing those from clear non-cases. But those features don't all
have to apply, since there are less-clear (less prototypical) objects to
which the term *does* apply. For instance, if something had a seat, a very
narrow back, no arms and three legs, it would be an instance of "chair";
it just wouldn't be the best example, and not all of the features that
applied to the best examples apply to it.



- Peter


---------------------------------------------------------------------------
Peter Constable

Non-Roman Script Initiative, SIL International
7500 W. Camp Wisdom Rd., Dallas, TX 75236, USA
Tel: +1 972 708 7485
E-mail: <peter_constable@...>