LINGUIST List 2.795

Sat 16 Nov 1991

Disc: Segments

Editor for this issue: <>

Directory

1. Vicki Fromkin, Re: 2.788 Are Segments Universal?
2. , Southeast Asian Languages and the Rhyme
3. Stephen P Spackman, Re: 2.788 (Segments) - Perception of your own language
4. Jacques Guy, Are segments universal?
5. Jacques Guy, Are Segments Universal?

Message 1: Re: 2.788 Are Segments Universal?

The question regarding the 'reality' of phonological segments can not be
answered by piece meal evidence regarding the difficulties 3 speakers
had in a task which does not necessarily reflect mental representations
nor the ability to analyze a language with or without segments or ...
this is an empirical question which requires converging evidence of all
kinds, including historical change (in all languages), phonological
'rules' and processes, perceptual evidence, child lng acquisition,
neurolingustic evidence from jargon aphasia, and other kinds of aphasic
language breakdown etc etc. Since a child can learn any language to
which it is exposed it would be rather strange if the phonological representa-
tion of lexical items were dramatically different from one language to the
next.
It is an important issue but if one is a 'God's truth linguistic' rather than
a hocus pocus linguist (such as Firth ((and that is not being critical of him
and his followers but stating what he himself believed to be the nature of
scientific theories))) it is important that anecdotal 'evidence' be evaluated
for what it is. It may provide interesting clues to be followed up in a
more substantive way.
Incidentally, speech errors in Taiwanese, Mandarin, Thai, and other languages
cited show segmental and feature errors similar to those in English, German,
Russian etc. These errors however are now being looked at in light of recent
theories of non-linear, autosegmental, and metrical phonology (and others).
Again, let me reiterate that I believe that all aspects of phonology must be
considered in our attempt to discover what phonological representation is.
Although I use speech error and aphasia data I would not want to base any
theoretical claims just on such data if there were other kinds of data which
appear to contradict these.

In response to David Gil's question about the phonological
structure of Southeast Asian languages, I reviewed a book
a few years ago in Language which dealt, in part, with this
issue. The reference is:
Cao, Xuan Hao. 1985. Phonologie et Line'arite': Re'flections
 sur critiques sur les postulats de la phonologie
 contemporaine. Socie'te' d'e'tudes linguistiques et
 anthropologiques de France. 18. Nume'ro spe'cial.
 Paris.
Cao's claim (among others) was that languages without inflectional
morphology whose canonical syllable structure was CV(C) (like
Vietnamese) would never force their speakers to `notice' individual
segments (particularly consonants), and that consequently their
phonologies would be radically different from languages with
consonant clusters or with `removable' (i.e. morphologically
separable) coda consonants. The claim seems rather reasonable to
me, but psycholinguistic explorations of monolinguals would seem
to be in order. Of course the fact that Vietnamese is written
with an alphabet might be a confounding variable, since I strongly
believe (and have published--CLS Parasession 1979) that orthographies
can remake underlying forms and strongly affect storage of forms
(actually these last two may be the same thing).
Anyway, I recommend the book for some interesting insights.
Geoff Nathan <ga3662siucvmb>

Message 3: Re: 2.788 (Segments) - Perception of your own language

Conversely, as a native speaker of english, I was in my teens before I
had any idea what the algorithm was for identifying syllables: I
always used to get 20 or 30% too many, could rarely give a precise
answer rather than a range, and often wanted to divide them in the
middle of orthographic letters. I remember this because on moving to
canada I was told to stop trying to use etymology in hyphenation and
just put the hyphens between the syllables. The results were strange
by usual standards. The idea of "syllable" is still a largely
theoretical concept for me, I think.
My naive intuition (or the theory I had developed by age 8, if you
prefer) is that "puts" in isolation has two *equal* parts, pu? tss,
and this changes when a following word shortens the s, as in "puts
back", pu? tsbak. Either way there's a boundary - and what I thought
was meant by a syllable boundary - in the middle of the orthographic
t.
-sps-

I've received three direct enquiries about Sukhotin's algorithm. Even though I
have answered them individually, some references may still be of general
 interest.
Sukhotin, B.V. 1962. Eksperimental'noe vydelenie klassov bukv s pomoshch'ju EVM.
 Problemy strukturnoj lingvistiki. 234: 198-206.
 1966. Issledovanie jazyka deshifrovochnymi metodami. Russkii jazyk v
 shkole, No.6, pp.16ff.
 1973. Issledovanie struktury prostogo predlozhenija s pomoshch'ju EVM.
 Problemy strukturnoj lingvistiki. pp.429-488.
There's more, and in particular a 1974(?) paper entitled "Problemy mezhzvezdnoj
svjazi" (yes!), but I never was able to get copies of those.
For those who do not read Russian there is:
Sukhotin, B.V.
 1973. Methode de dechiffrage, outil de recherche en linguistique. T.A.
 Informations. 2:3-43. (translations from the Russian)
Boy, Joachim 1977. Dechiffrierungsalgorithmen zur phonetischen Identifikation
 von
 Buchstaben. Bochum: N.Brockmeyer (Ph.D. thesis on Sukhotin's
 consonant/vowel
 algorithm)
Guy, J.B.M. 1991. Vowel Identification: an Old (but Good) Algorithm.
 Cryptologia, vol. 15,
 no.3 (July).

H. Samuel Wang <onghiokling.nthu.edu.tw>
"When I started to learn English (at grade 7), I had a hard time
 trying to fight off the idea that the "n" in "not" is different
 from the "n" in "can"..."
Now it all comes back to me! I remember Professor Rygaloff telling us
that the "n" of, say, "shan1", was not the "n" of, say, "na2". I plumb
forgot the term he used in French, but as far as I can remember, it
boiled down to that syllable-initial n's were stops, syllable-final
ones approximants (I might misquote him here). It all went above our
heads of course, and we lived on happily ever after with our awful
foreign accents.
And:
"I tried to have my subjects break the syllables into segments and
 recombine the segments, only to find that most of my subjects could
 not manipulate segmentation of syllables."
I encountered the very same phenomenon with the northern dialect of Sakao,
an Austronesian language of Espiritu Santo. An orthodox explanation for
this was complex rules of regressive vowel harmony so that a speaker
could not break down a word into its component syllables. A word typically
contains only one stressed vowel (always the last vowel) and the phonetic
realization of the preceding vowels is conditioned by that stressed vowel,
and, to a lesser extent, by the rounding of the neighbouring consonants.
So if you try to break down a word into syllables, you cut off, as it were,
the chain that determines the realization of your unstressed vowels. Now that
is a phenomenon very different from what I think happens in Chinese. Still,
it's precisely the type of things which I had been indoctrinated to believe
just did not happen (I was force-fed Pike's phonemics then in an SIL crash
course in field methods. Had to unlearn it all).
And again:
"because the Chinese syllable structures are simple, and the number of
 possible syllables is rather limited (mostly under 500)[...] the native
 speakers do not have to spend the effort to further segment them as the
 number is cognitively manipulable."
Yes, very probably so. Tretiakoff has shown that the difference between
first and second-order character entropy is maximal in a phonemic
transcription where each symbol has been correctly replaced by V when
it denotes a vowel, by C when it denotes a consonant. In other words:
that H1-H2 provides an objective function of the correctness of a
particular phonological interpretation. Sukhotin had suggested earlier
that Shannon's entropy might provide the best objective functions for
his decipherment algorithms (the vowel/consonant one included). Now
Tretiakoff was unaware of Sukhotin's work. My guess is that an algorithm
to break down syllables into their components would produce a maximum
for Chinese before it reached the "phoneme" level. (Now that was very
awkwardly put, but since this line of research is rather unknown
in linguistics, I can't think of widely-accepted labels to stick on
those notions).