LINGUIST List 2.244

Monday, 27 May 1991

Disc: Phonology and Orthography (Part 1)

Editor for this issue: <>

Directory

1. Hurch, phonology and orthography
2. Jim Scobbie, Rules and representations
3. John Coleman, Putative syllabaries in Semitic

Thanks to John Coleman for explaining me that there might be more than
one phonemic analysis for a given language. Maybe the Anonymous Icelanderr
did not know that in the Middle Ages and thus he just adapted the Latin
alphabet to his own language. But could you also explain to me why this
should not be a phonemic analysis then.
Moreover, maybe we should open a query on the use of the definite article
in English (which is, as anybody might guess, my mother tongue) but still
I think that my text "the phonemic analysis was taken out quite detailed..."
 ^^^^^^^^^^^^^
(emphasis John Coleman's) does not imply that other phonemic analyses of
Old/Middle Icelandic are excluded/impossible or whatever. How definite is
the definite article then in English, John Coleman?
Bernhard

Rick Wojcik writes:**************************(rwojcikatc.boeing.com)********
The issue between Harry Bochner and Mike Hammond on listing vs. rule-governed
behavior recalls Langacker's warnings about the "rule/list" fallacy that ties
linguists in knots from time to time (cf. p. 42 of Foundations_of_Cognitive_
Grammar). Langacker points out that there is no reason why speakers can't
memorize some plural forms as fixed units and still have a rule that derives
plurals. Why not have both lists and rules?
****************************************************************************

In constraint-based approaches to grammar, the rule/representation
distinction is irrelevant. A highly specific constraint might describe
a single lexical entry --- it therefore expresses no generalisations 
over the lexicon of the language. A highly general constraint might force
every representation to have a certain characteristic --- a universal
principle would take this form. In the middle, of course, there are a 
very large number of partial generalisations.

This means speakers need not learn two types of ontological category. They
must assign only the appropriate degree of a single category.

>From: John Coleman <JSC1vaxb.york.ac.uk>*********************************
Mike Hammond statement that 
> Rules are rules. What else is there? 
is too glib. There are different kinds of rules and rule-systems, of 
different generative power. "Rules" in phonology no longer means
simply "productive generalisation" or "licensor of a step in
a derivation". It means something more like the Hallean kind of
A --> B / C ___ E, even in theories like Autosegmental Phonology.
 D

There are types of rules that are much less powerful than this,
such as structure-building and redundancy rules, which, being
monotonic, need not be extrinsically ordered.
**********************************************************************

So, in a 'declarative' or 'constraint-based' approach to phonology,[like C's]
*some familar rules* from procedural, derivation-based phonology
are permitted, recast as conditional constraints on possible representations.
The generative rule A-->B/C is allowed as a logical constraint meaning: 'if
some structure is C+A then it must also be B', just in case it is monotonic.
C+A must not be altered. Moreover, the order in which the rule is applied must
not be significant. The rule will more typically be a redundancy than a
rewrite rule, though in SPE at least there is no formal distinction between em.

The original worry and its solution were that:
phonologists (linguists in general) get het up over the rule/representation
choice as the means to describe data when it is perfectly possible for
both rules *and* representations to be used in an analysis of a phenomenon.
An alternative solution is that:
it is perfectly possible to do without the rule/representation altogether.

Sane-sanity can be learned for what it is: a highly restricted yet relatively
regular phenomenon. 

> The notion that West Semitic writing systems such as Hebrew and Arabic 
> are syllabaries has been persuasively refuted. The syllabary idea was 
> proposed by Gelb in 1952. 
Perhaps Firth's (1948) discussion (in "Sounds and Prosodies") is more
convincing.
> Since then the idea has been taken as authoritative by 
> popularizers and some linguists, though, I think, mainly by those whose 
> knowledge of these systems is second-hand. 
I don't think you could accuse Firth of that.

> My own opinion is that these systems are in fact a kind of alphabet that 
> happens not to represent some or most of the vowels 
By the same logic you could argue that Japanese kana represent CVC syllables,
but happen to leave off the final C!

To Rick Wojcik:

> Historically, linguists have assumed that alphabetic orthographies were 
> based on the phonemic analysis of a dialect of a language. 
> The reverse hypothesis is hardly a fact. 

If the "reverse hypothesis" (that phonemic analysis developed out of 
the study of alphabetic orthography) is not "a fact", then a number of
things need to be explained, like how come phonemic analysis is alphabetic
(concatenative, segmental ...), why were the originators of phonemic analysis
also often into spelling reform, why did phonemic phonology originate in
alphabetic cultures etc. On the basis of such considerations, I maintain
that phonemic analysis is a product of alphabetic literacy, not that the
existence of alphabets is evidence for the reality of phonemes (or, more
generally, segmental phonology.)

> John Coleman's point about there not being a unique phonemic analysis for a
> language is correct but irrelevant. 
It is relevant if alphabet-phoneme correspondences are proposed as support
for segmental phonology, as a number of LINGUIST contributors have done.

> The fact is that constructing an alphabet is a practical matter
Isn't phonemic analysis also a practical matter? What point are you making?

> about the efficiency of alphabetic vs. morabetic writing. ... You can't
> compare the number of symbols needed for Hebrew with the number of symbols
> needed for Greek, and claim that that tells you anything about the efficiency
> of morabetic vs. alphabetic writing. 
It does if you want to make the sweeping claim that mora-based writing systems
are (all) less efficient than alphabetic writing systems. I would suggest that
writing systems of different kinds are suitable for phonological systems of 
different kinds i.e. there is nothing distinguished about alphabetic writing
or segmental phonological analysis.

> For any given language, the number of
> morabetic symbols required to represent that language will always be greater
> than the number of alphabetic symbols needed. 
The number of symbolic distinctions (bits of information) needed to encode a 
set of CV moras using mora-symbols may be less than if alphabetic symbols are 
used. There are several reasons for this. Firstly, non-occuring CV combinations
are simply not listed in the mora-set, whereas the alphabetic representation
encodes non-occuring combinations as readily as occuring combinations.
Another reason is that the order of C and V elements is usually redundant
information, which is encoded by an alphabetic representation, but not
in a mora-based representation.

> In a pure CV language you need enough symbols to cover all possible CV 
> combinations. 
In such a language the use of separate consonant and vowel symbols would
be best. But that would still not make it an alphabet, necessarily, because
the most efficient representation would be one in which the order of C and 
V were non-distinctive. C-over-V or V-over-C notation might be most suitable 
in such cases. Would this be an alphabet? It depends on your definition. 
Whatever your definition, though, such a system would not be an alphabet of 
the *usual* sort (not concatenative, for instance).

--- John Coleman

[End Linguist List, Vol. 2, No. 244]