Dirk Elzinga, On 13/10/2012 18:21:
> About 12 years ago, we talked about the "death of the phoneme". You can
> read the discussion here:
> In the time since that extended and enjoyable exchange (an enjoyment which
> has been renewed upon rereading),

A renewed enjoyment for me too!

> On Fri, Oct 12, 2012 at 11:47 PM, Alex Fink<[log in to unmask]>  wrote:
>> IMO the word "phoneme" and the // brackets are very much overexposed, and
>> people come to think they're the unmarked member of the opposition phone
>> vs. phoneme, when in fact the opposite is true.  It's possible to use the
>> [] brackets as narrowly or broadly as you like.  In an ideal world no-one
>> would encounter "phoneme" and // without being explicitly warned that to
>> use these is to accept a theoretical claim about the language in question,
>> and hopefully already have "phone" and [] committed to unconscious default
>> use by the time that happens...

This belongs in some sort of appendix to the Conlang-L FAQ, which should somehow be accessible only to readers who click "Yes! Hit me with your stern but O-so-just pedantry!".

>> I'm somewhat aware of alternatives to phonemes (not of Firth's in
>> particular; it'd be kind if someone gave a worked example of his prosodic
>> theory here, 'cause I'm not likely to have time to look soon).

Maybe try using Google books to look at pp 180--193 of Anderson's _Phonology in the Twentieth Century_? (It contains no worked exx, tho.)

>> Narrowly construed, I agree there are problems; certainly e.g. a
>> theory of phonemes that doesn't allow for multiple phonological
>> tiers is in trouble.

Featural tiers?

I don't think we can really start talking about what is and isn't an adequate theory or the best theory of phonology until we set some parameters:

1. Do you want a mentalist theory?
2. Do you want a theory of language (i.e. Universal) or only of a language (i.e. Parochial)?

There aren't right or wrong answers to these two questions (IMO): rather, they define the object of intellectual enquiry. Once you've answered those questions, ranking of theories becomes more meaningful. (My own answers to those questions are No and Parochial, which puts me into the least populous quadrant.) Further questions (to which there are righter and wronger answers, given the answers to the first two):

3. If phonology is what the form of sentences is built of, does phonology consist of phonetic material, or does phonology interface with a separate (psycho-)phonetic module?
4. Does phonology include all patterns and rules pertaining to form, or is there a separate allomorphy component of the lexicon? (This is an important question for English, but maybe not for all lgs.)

(My own answers: (3): There's a separate phonetic module; phonology is "substance-free". (4): There's a separate allomorphy component in the lexicon.)

And then we get to such questions as: What are the basic combinatorial and classificatory units of phonology? What are the syntagmatic relations of phonology? The trad phonemic analysis makes the segment the basic combinatorial unit, the phoneme or the feature the basic classificatory unit, and simple linear concatenation the basic syntagmatic unit. Theories since the mid-70s (but also Prosodic Phonology earlier, as mentioned by Ray) go for subsegmental and suprasegmental combinatorial units, and elaborate classification systems and syntagmatic relations. (Me: I go for units that are essentially segmental but whose phonetic expressions can be simultaneous, for a very simple classification system, and for a simple but pivotal inventory of syntagmatic relations and rather complex syntagmatic structures. (For English, this is.))

As for what all this means to the conlanger, I think it partly depends whether the conlanger aspires to conceive of their conlang as a preexisting language, which they try to describe in some ad hoc but efficacious way, in the way that most reference grammars are written, or whether they aspire to describe, or at least consciously know for their own purposes, the actual underlying mechanisms of the language, in which case the answers to the questions I listed, and relevant portions of the linguistics literature, would become very pertinent. Another interesting question for the conlanger is whether to invent the observable data and then try to discover the underyling rules as a linguistician would, or whether to directly invent the underlying rules.