LINGUIST List 2.122

Sunday, 7 Apr 1991

Disc: (Morpho)phonology

Editor for this issue: <>


Directory

  1. , Morphology; Evaluation Metric
  2. "Bruce E. Nevin", morphophonology/phonology boundary
  3. , RE: History of Phonology

Message 1: Morphology; Evaluation Metric

Date: Thu, 04 Apr 91 21:32:37 -0500
From: <bochnerdas.harvard.edu>
Subject: Morphology; Evaluation Metric
Phil Bralich writes:
>Is the Forris publication coming out soon?
>At the very least I would like to
>know more about your version of the Evaluation Metric.

The contract hasn't been signed yet, and we haven't discussed the production
schedule yet. So I don't know; I'll let you know when I know more.
I don't believe you mentioned whether your thesis would be available through
UMI; probably it would be easier to get a copy from you?

>Strictly speaking the thoery I propose does not actually posit internal bound
>aries or juncture of any sort. The theory I propose merely claims that 
>speakers are aware of the cateogies of the items involved in word formation. 
>However, in the framework I propose the categories are enhanced with the the
>bar levels proivided by the X-bar theory. The notion juncture falls out from
>the fact that speakers consider the categories and bar levels of the items
>involved in word foramtion. 

I agree that, in the context of a morpheme-based theory, this is a very
attractive move. I argue, however, that the morpheme-based model is
untenable for more basic reasons. This is the beginning of the
Evaluation Metric argument, and the easiest part to summarize, so here goes:

The usual, symbol-counting, Evaluation Metric is the implicit basis for
most arguments for morpheme-based theories as opposed to word-based
theories; cf. the arguments of Kiparsky(82). If you work out the
implications: you find a clear prediction that a morphologically complex
(= polymorphemic) word cannot have any properties that cannot be
predicted directly from it's morphemes; again, cf K(82, page 28). But
this prediction is easily falsified. The fact that it fails for cases of
opaque semantics has been discussed a fair bit; the fact that it fails
for routine cases of morpheme distribution ('potentiation' properties) has
gotten less attention. Consider:
perceive perceptible
conceive *conceptible

There's no semantic problem here: if *conceptible existed, it would mean
'conceivable'. So it clear that the ability to potentiate -ible can not
be attributed to ceive/cept. But it can't be attributed to the prefixes
either: *perversible shows that per- does not always enable -ible, while
comprehensible shows that com- does not always block it. Examples of this
sort can be multiplied at will.

So a basic prediction of the model turns out to be false. I omit discussion
of ways we might try to rescue it: I don't think any of them are compatible
with a strict interpretation of the symbol-counting Evaluation Metric. I
leave it to the proponents to work out an explict response to this problem.

So if we reject the symbol-counting EM, we open the door to word-based
theories. I take Jackendoff(75) as my starting point; my formalization of
the EM involves pattern matching applied to sets of related lexical entries.
Regardless of the formalization, however, the basic idea is that, instead
of equating simplicity with brevity, as in the symbol-counting EM, we need
to think of it as conformity with the patterns of the grammar.

> Also what phenomena do you account for?
My main concern is building up the morphological model; the
morphophonological concerns we share are just secondary issue for me,
and I don't do more than illustrate how the model works in this respect.
Once you get into the details, English morphology turns out to much more
interesting (and messy) that I expected!

-- Harry Bochner
-- bochnerdas.harvard.edu
Mail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue

Message 2: morphophonology/phonology boundary

Date: Fri, 5 Apr 91 08:20:49 EST
From: "Bruce E. Nevin" <bnevinccb.bbn.com>
Subject: morphophonology/phonology boundary
To Harry Bochner <bochnerdas.harvard.edu>, who in 0116 re
"Morphological Juncture; Morphophonology" agrees with Rick Wojcic that
SPE advanced the

>> mistaken position that there was no basic difference between 
>> morphonology and phonology.

and then suggests that Halle's original argument was

>an argument (convincing, I think) that Structuralist Phonemics placed
>the boundary between Phonology and Morphophonology in the wrong place.

and not about the existence of that boundary, that is, not about the
status of the phoneme.

My question: which Structuralist Phonemics? The historical situation
was richer in variety than is supposed in the traditionally recited
strawman arguments. For example, Harris's more abstract phonemes
allowed him to place the phonemics/morphophonemics boundary differently
than Bloch and others were able to do with their less abstract (more
"natural"?) phonemes.

The salient distinction: for Bloch, two phones had to share a
characteristic phonetic feature. He treated phonetic identity as the
primitive term and from it (together with distributional criteria)
derived contrast and its complement, repetition. Harris treated
repetition (and therefore contrast) as primitive. This is the
theoretical significance Harris's innovation, the pair test.

Because of this move, Harrisian phonological theory can treat phonetic
likeness not as a requirement but as just one optionally useful
criterion among several. For Harris, distributional considerations may
override phonetic likeness. Problematic cases of overlapping are the
usual example, as in his treatment of the problem of vowel height (not
length, though that is relevant in some dialects of English) in e.g.
"writer/rider" by phonemicizing the sequence as a whole rather than one
segment at a time. For Bloch, this rephonemicization step would not be
possible because all his allophones of /d/ have to share the same
characteristic phonetic feature, and all allophones of /t/ have to
share a different phonetic feature, whatever the other phonetic
differences between the allophones of either.

Overlapping forces Bloch to accept awkward conclusions, and it is those
conclusions that were usually the crux of attacks on "taxonomic
phonemics." For similar reasons, Bloch distrusted the treatment of
junctures as phonemic entities [relevant to his "pod:pa'd (go if he
could)" example].

There are obvious parallels to the natural/abstract debate. A
difference is that for Harris the arbitrariness of language that sets
abstract phonological entities and rules off from natural ones is of a
piece with the arbitrariness of all social convention, rather than being
due to a language organ, which is to say "natural" in some neurological
sense. In his (Naturalist) view, the language system is immanent in the
speech community rather than in the biological endowment of individual
language users, though doubtless constrained by neurological factors
much as phonology is constrained by various kinds of phonetic factors.
After a long reductio argument between two perspectives that seem to
divide the universe between them it can be refreshing to include a
third.

 Bruce Nevin
 bnbbn.com
Mail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue

Message 3: RE: History of Phonology

Date: Thu, 4 Apr 91 16:26:31 PST
From: <rwojcikatc.boeing.com>
Subject: RE: History of Phonology
John Coleman writes:
>Whether a phenomenon such as this is morphophonological or just an
>automatic alternation depends on other considerations, especially the
>form of morphophonological representations and the theory of phonetic 
>interpretation which is being assumed. Wojcik's comment seems to me to 
>assume that both of these are settled, and that there can therefore be
>no redrawing of the lines between phonetics and phonology, even if such
>realignment allows data previously regarded as independent to be handled
>by generalisations already proposed for some other set of examples.

We are talking at cross purposes. I never said anything about redrawing the
lines between phonetics and phonology. I restricted my comments exclusively
to the distinction between phonology and morphonology, the historical roots of
which most modern linguists have little understanding. Vowel reduction
usually refers to a purely phonological phenomenon--what happens when speakers
try to pronounce vowels in unstressed environments. Purely phonological
processes such as vowel reduction participate in phonological foreign accent.
English speakers apply vowel reduction to foreign languages as well as
their own. The 'title/titular' alternation is of a different nature. It does
not bear on how we pronounce sounds and plays no concrete role in the
mispronunciation of foreign words. (I am proposing phonological accent as a
kind of litmus test, although I believe that there are rare cases where
morphology influences accent--e.g. when English speakers pronounce Spanish
plurals with a /z/ instead of an /s/.)

>To demonstrate my point about the theory-internal nature of the dividing
>line between (morpho)phonology and (physio)phonetics...

I have little to say about most of what John wrote in the paragraph I excerpt
this from, but I don't understand the parentheses around 'morpho' here. The
correct historical relationship between Baudouin's terminology and
Trubetzkoy's is the following: psychophonetics = morpho(pho)nology and
physiophonetics = phonology. It was no accident that Trubetzkoy lumped
archiphonemics and allophonics together under the rubric of 'phonology' in his
Fundamentals. That was because the alternations covering those two phenomena
in his theory corresponded to Baudouin's physiophonetic domain. Trubetzkoy
mentioned morphonology briefly in an appendix to his text, since it wasn't
properly a topic of discussion in a phonology text. (And I should mention
that the term 'morphophoneme', first coined by Baudouin's student Ulaszyn, was
historically misused by structuralists to represent psychophonetic
alternations.)

John further responds to my claim that Sapir, the Moscow School, and Stampean
Natural Phonology retain the original Baudouinian dichotomy between phonology
and morphonology:

>There are two others I know of. The Copenhagen school, I believe still 
>has adherents, and also Firthian Prosodic Analysis, which is still very
>much alive. That's my background, which is why I found Wojcik's
>piece a bit ironic.

I'm not sure that either school has anything to do with the original
dichotomy, but I plead only a superficial grasp of both schools. I do know
that British phoneticians were influenced by Shcherba (Leningrad), who 
redefined the phoneme as a purely perceptual unit. Shcherba, in effect, was an 
original proponent of the so-called "once a phoneme, always a phoneme" view.
Reformatiskii (Moscow School) jokingly referred to his redefinition of the
concept as the "Shcherbeme". The shcherbeme is essentially the classical
phoneme as defined in the West. (Russian readers can look at A.A.
Reformatskii's 1970 Iz_istorii_otechestvennoi_fonologii for a revealing
Moscow-based retrospective on phonological history.)

Here is the basic alternational dichotomy. Baudouin distinguished two-phoneme
(psycophonetic) alternations from one-phoneme (physiophonetic) alternations.
An example of a PHYSIOphonetic alternation is the word-final devoicing in the
Russian word 'rod' [rot]/'roda' [roda] ('type, kind' nom./gen.). For
Baudouin, there was only the single phoneme /d/ underlying the alternation,
but for Shcherba there had to be two distinct phonemes /t/ and /d/. Baudouin
believed that phonemes represented the phonetic segments that speakers
attempted to pronounce. So he believed that Russians were intending to
pronounce voiced obstruents at the ends of some words, but that they ended up
pronouncing them as voiceless. To him, this was a 'physical' fact--not under
the conscious control of the speaker. On the other hand, the /ay/~/I/
alternation in 'type/typical' is a PSYCHOphonetic alternation--an alternation
that the speaker can control volitionally. You can say 't[ay]pIcal' or
't[I]pical' with no feeling of articulatory difficulty. Hence, two distinct
phonemes are alternating here. (BTW, Shcherba was Baudouin's student in
St. Petersburg and came to dominate the 'Leningrad School' after Baudouin 
left.)

Sapir (cf. Language p. 62 in the Harcourt, Brace paperback) used 
'book[s]~bag[z]' and 'hou[s]e~hou[z]es' to illustrate the dichotomy. He said 
"The two alternations belong...to entirely different psychological 
categories." Sapir was not unique or original in his view of the phoneme. In
fact, he was one of the last advocates of the original concept of phonemics
(=phonology) before the dark days of shcherbeme-dominated structuralism
rolled over everything.

(BTW, I purposely avoid the term "automatic alternation" for physiophonetic 
alternations, because the psychophonetic ones can be quite automatic as well.)

 -Rick Wojcik (rwojcikatc.boeing.com)

[End Linguist List, Vol. 2, No. 122]
Mail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue