LINGUIST List 3.560

Sat 11 Jul 1992

Disc: Phonology

Editor for this issue: <>


  1. "Richard L. Goerwitz", Re: 3.555 Comparatives, Phonology
  2. , dissimilation update
  3. , 3.551 Phonology
  4. , Phonology (Normal, Natural, Broad, Etc.)

Message 1: Re: 3.555 Comparatives, Phonology

Date: Tue, 7 Jul 92 15:07:45 CDTRe: 3.555 Comparatives, Phonology
From: "Richard L. Goerwitz" <>
Subject: Re: 3.555 Comparatives, Phonology

John Local writes:

> I agree [with John Coleman] wholeheartedly both with his
> observations concerning Natural Phonology and on his invitation to
> practitioners to demonstrate what they're claiming. Too much low
> grade observation and pre-theoretical formalisation has passed and
> still passes for 'phonology'. To some extent this can be seen in
> the way most 'theories' do indeed pick over the same old material.
> That's not to say that it's an illegitimate exercise (unless of
> course the picking over simply reworks some previous analysis
> without actually getting out there and listening/measuring to what
> happens!) rather that there is a concensus agenda which few
> practitioners are willing to challenge. I can find no evidence
> that Natural Phonologists have or are dealing with material which
> other frameworks have not essayed or find unaccountable.

On the face of things, it seems you are missing the point. Natural
Phonology was not necessarily meant to account for facts about
language that had never been accounted for. The SPE formalism had no
elegant mechanism for dealing with rules with obvious motivations in
the physical (and to some extent cognitive) mechanisms through with
speech is created. The SPE formalism, though, is powerful, and can
be extended to cover a lot of data. The question is one of natural-
ness and elegance as much as anything else.

Ironically, those very aspects of Natural Phonology that were
perceived as strengths by some were perceived as insuperable
weaknesses by others. As I recall, one of the big objections raised
almost immediately was that not all phonological rules are "natural."
If phonology is not "natural," then natural phonology loses its
predictive value, and becomes little more than yet another formalism.
For example, if you tell someone to pronounce "a big apple" as a
spoonerism ("an ig bapple" or whatever), the a/an choice is made after
the metathesis. This happens both in contrived examples, and in
actual speech. Now a/an is certainly not a reflection of some basic
phonological process. Children do not spontaneously add nasals
between vowels separated by a word boundary. Nor do languages in
general show this sort of process of nasal addition. So what's up?
A morphophonological alternation that applies very near to the
surface? Is this an oxymoron? Are other analyses possible? Are a/an
really as consistent as the literature critical of Stampe implies?

These are the sorts of objections that we have to talk about if we are
to confirm/refute Natural Phonology. I really don't see how inviting
Natural Phonologists to submit essays proving their case is going to
work. Let's talk about specific issues, and see what interested
parties have to say.

Most phonologists seem to earn their living creating new recondite
terms and formalisms. In most cases where I've seen theorists dip
into actual data in languages I study, the results have been less than
impressive. It is easy to be grateful for works like Patricia
Donegan's dissertation.

Mail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue

Message 2: dissimilation update

Date: Tue, 7 Jul 1992 15:36:01 -dissimilation update
From: <>
Subject: dissimilation update

Last week I posted a summary of the messages on dissimilation that we (= J.
Cole, C. Kisseberth & J.Hualde) had received. I indicated that from those
responses it appeared that true long distance dissimilations (i.e. between
segments in nonadjacent syllables) which behave as regular synchronic rules
were rather uncommon, if attested at all. Since then, we have received a
couple of messages reporting what appear to be some very interesting cases
of long distance dissimilation. David Odden points out that:
>	First, there are a fair number of tonal cases of Meeussen's Rule
>between nonadjacent syllables (many Lacustrine languages, and, outside of
>Bantu, a version exists in Arusa). There is also a labial dissimilation rule
>in Tashlhiyt Berber <Lisa Selkirk has been telling this tale> where /m/
>dissimilates to [n] when followed by a labial in the word -- any number of
>syllables can intervene. Finally -- and this is a recollection, which you'd
>clearly want to check up on -- in Kera, there is a constraint on coocurrences
>of voicing which leads to dissimilative alternations, which I recall can skip

In addition, Wayles Browne has informed us that Georgian has a liquid
dissimilation rule which is reminiscent of that of Latin, but which seems
to be truly productive regardless of the distance between trigger and

>Georgian seems to be an example of productive dissimilation. There is an
>adjective suffix -ur- as in phizikuri 'physical', kimiuri 'chemical',
>aktiuri 'active'. If
>there is an -r- even several syllables before it, the suffix appears as
>-ul-: phranguli 'French', germanuli 'German', reaktiuli 'reactive'.
>However if there is an -r- and later an -l-, the suffix is -ur- again:
>realuri 'real', terminaluri 'terminal'. References on Georgian: grammars by
>Howard Aronson and by Kita Tschenke'li.
>So it is similar to Latin but better.

In relation to David Stampe's comments on our earlier summary, there is a
small point about the Latin case that we would like to clarify. The process
was morphologically conditioned. In fact, it was restricted to words
containing the suffix /-alis/. There is thus nothing remarkable about its
non-application in /mellis/ and similar words; or, for that matter, about
the lack of dissimilation in /li:lium/, where the sequence is
morpheme-internal or in /ferr-a:rius/, which contains a different suffix.
Since liquid dissimilation was furthermore restricted for the most part to
words where the two liquids were in adjacent syllables, we must agree with
Browne that the Georgian rule (which is also restricted to one suffix)
appears to be better (but, we still haven't checked the facts).
Richard Goerwitz has informed us about a case in Tiberian Hebrew where
there is dissimilation between two vowels separated by certain consonants.
We will be happy to send the complete file of received messages to anyone
requesting it.
Jose Ignacio Hualde
Dept. of Spanish, Italian and Portuguese
4040 FLB, Univ. of Illinois, Urbana, IL 61801
Mail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue

Message 3: 3.551 Phonology

Date: Wed, 8 Jul 92 19:35:00 EDT3.551 Phonology
From: <>
Subject: 3.551 Phonology

In a recent posting David Stampe claims that

"Dissimilations, like other fortitions, apply in
narrower prosodic domains only if they apply in wider domains."


"This is the opposite of assimilations, which, like all lenitions,
apply in wider prosodic domains only if they apply in narrower ones."

I am not sure what a prosodic domain is exactly, but it would
seem that the fact that Grassmann's Law in Sanskrit applies within
(root) morphemes but not, for example, across a compound boundary
would be a counterexample. Or am I misreading the claim?

In the case of assimilations, I think that Polish voicing assimilation
facts are a counterexample. In Polish all obstruent clusters agree in
voicing, but the direction of assimilation varies. For clusters whose
second element is a stop or affricate, assimilation is always regressive.
Likewise, for clusters whose second element is a voiceless fricative.
However, for clusters whose second element is a voiced fricative, it
appears that within morphemes assimilation is progressive, whereas
across boundaries it is regressive.

Now, these claims are somewhat difficult to substantiate in terms
of synchronic alternations within the language, but they seem
to apply in situations where Polish speakers borrow foreign words
as well as in made-up words, speaking foreign languages, etc.

Actually, the regressive assimilations are amply documented across
word boundaries and compound boundaries in native forms. However,
it is difficult to find native prefixes that end in voiceless obstruents
so it is convenient to see what Poles do with the Russian prefix ot.
In forms where this is a live prefix, there is always regressive assimilation,
e.g., ot-vyazat', ot-vratit', etc., whereas in otvetit' (which I
guess is etymologically analyzable but not synchronically), there is always
progressive assimilation.

Something similar occurs with German forms.
For example, German Ausweis and Luftwaffe come out /awzvays/ and
/luvdvafa/, where somehow the German boundaries are respected,
whereas Auschwitz comes out /aus^fic/ (s^ is used to represent
the orthographic 'sz', phonetically an almost retroflex shibilant).

Another nice morpheme-internal example I observed was in the
speech of a Hebrew-speaking Polish native speaker. Hebrew is
helpful because it has a lot of relevant alternations. The following
occurred in spontaneous conversation: Hebrew 'titrakzi' 'you (fem. sg.)
will concentrate' came out as 'titraksi'. This verb has forms such
as 'mitrakezet '(fem. sg.) is concentrating, which clearly shows the
/k/ and the /z/ as the underlying forms.
Also, I find it common for Polish speakers of Hebrew to say
to assimilate progressively in 'hitvakeax' 'he argued', where the
prefix hit- is not perceived as a prefix, but to assimilate
regressively in forms where the hit- is a productive reflexive/
stative/passive prefix.

I should add that I do not have compelling data for the case
of stem+suffix.
Mail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue

Message 4: Phonology (Normal, Natural, Broad, Etc.)

Date: Wed, 8 Jul 92 20:53:20 EDTPhonology (Normal, Natural, Broad, Etc.)
From: <>
Subject: Phonology (Normal, Natural, Broad, Etc.)

Perhaps it may not be amiss to point out that some of me has/have
working for some years now on an approach to phonology which, while
originally inspired in part by NP, is distinct from it as well as
from the supposedly dominant approach originally called generative
phonology, but now often referred to simply as 'phonology' or
'phonological theory' (hence I refer to as 'normal').

I have referred to this approach as 'broad' or 'horizontal'
phonology. Its main tenet is that phonology is not deep
(that is, characterized by levels, derivations, etc.) but
'broad' (that is, characterized by interactions between
different and partly independent components, all of them
quite shallow). Thus, many of the phenomena which both normal
phonology and natural phonology explain by assuming various
kinds of representations deeper and/or more abstract than
phonetic are better explained, I suggest, by, for example,
the interaction of phonetic and orthographic components or
the constraints on the interaction of two phonetic components,
such as, fast speech and slow speech.

Methodologically, my emphasis has been on two points. First,
on getting the facts right. I have argued that an incredible
proportion of the crucial examples in phonological theory are
simple wrong, undermining a number of the most widely held
assumptions in both natural phonology and in normal phonology,

Second, I have devoted a lot of attention to the question of
how you choose the best explanation of a phenomenon, focusing
on such things as the use of control cases in phonological
(and other linguistic) reasoning.

The kind of thing that I delight in finding is the fact that
the particular concept of the phoneme which American structuralists
ended up adopting in the 1930's and which Halle demolished in
the 1950's was introduced in the early 30's by Henryk Ulaszyn
on the basis of Polish facts essentially identical to the
Russian facts which Halle used to criticize the same concept.
And that these facts, if you read Halle carefully, pose a
very similar for anyone who believes in underlying representations
as for anyone who believes in phonemics.

Indeed, this is what first led to my attempts to develop a broad
approach in which neither underlying nor phonemic representations
were required and other explanations are offered for the facts
which seem to support these.
Mail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue