minimal pairs (was: PIE e/o Ablaut)

Robert Whiting whiting at cc.helsinki.fi
Sun Oct 15 11:53:43 UTC 2000


On Fri, 28 Apr 2000 Ross Clark <r.clark at auckland.ac.nz> wrote:

<snip>

[Ross Clark]
>>> I am astonished that this discussion has proceeded for several
>>> days without anyone questioning the original statement about
>>> complementary distribution of [th] and [dh] in modern English,
>>> which is simply incorrect. Even if one does not have the
>>> pronunciation which makes "either" and "ether" a minimal pair,
>>> examples of [th] in voiced environments are not at all hard to
>>> find: pathology, authority, anathema, mathematics, Gothic,
>>> Arthur, etc etc.

>[to which Robert Whiting replied]

>> And I am astonished that anyone would present a list of
>> loanwords, however long, and claim that it has some bearing on
>> native English phonology.  Loan words do not necessarily follow
>> the phonological rules of the borrowing language.  In fact this
>> is usually one of the first clues that a word is a loan when it
>> doesn't obey the phonological rules.  This is how you can tell
>> that 'father' is a native (inherited) word and 'padre' is a loan.

>> I'm sorry if you got confused, but I thought it was clear that I
>> was speaking about native English words, not borrowings.  Perhaps
>> I should have been explicit, but I really thought that everyone
>> knows that when you are trying to establish the phonology of a
>> language you should deal with words that are native to that
>> language.  I'm surprised that you didn't include 'Athens' in your
>> list.  You can make a list of hundreds of words in English that
>> have [th] in voiced environments and every one of them will be a
>> loan.  There are a very few examples where the complementary
>> distribution of [th] and [dh] does break down, but you haven't
>> mentioned any of them.

>[to which I reply]

>I trust that we share the assumptions that (i) we are talking
>about the synchronic phonology of modern English, and (ii) the
>reality that we are trying to get at is what is in speakers'
>heads.

First, let me assure you that we are indeed talking about the
synchronic phonology of modern English.  About the second point,
I am much less sanguine.  I think it is possible to describe how
language works linguistically (some areas are easier than
others), but I don't think that we are in a position to say what
goes on in a speaker's head to produce language.  While it would
certainly be nice to know, I think that the cognitive processes
that produce language are beyond our reach at the moment.
Speakers themselves don't know how they produce language, so you
can't find out by just asking them.  So the only reality that we
can get at is the language that speakers produce.

Historical grammar, when we have a written record, is based on
empirically verifiable facts.  Synchronic grammar is based on a
hypothesis about how native speakers produce their language.  A
hypothesis is not a fact.  It is an explanation put forward to
account for observable facts.  People tend to forget this and
consider synchronic grammar a fact.  But whatever reality may
exist in the speaker's head just can't be gotten at at the
present time with our present knowledge.  In general, I agree
that what we are trying to get at is the reality in speakers'
heads, but it is a roundabout road that we have to take and we
have to have a realistic picture of language before we are likely
to get there.

>The rest of your post is entirely dependent on the further
>assumption that native speakers of modern English (in general,
>not just linguists) distinguish "foreign" from "native" words,
>and that the words I listed with /th/ in voiced environments are
>marked as "foreign". Since I don't share this assumption, I would
>like to know what evidence leads you to it. Do you have any such
>evidence, other than the fact that by excluding these hundreds of
>words you can arrive at a nice phonological generalization?

Let me answer this from back to front.

You ask what evidence I have that native speakers distinguish by
rule the [th] of loan words from the [dh] of native words other
than hundreds of examples and the fact that it produces a nice
phonological generalization.  This seems rather like asking what
evidence I have for Grimm's Law or Verner's Law other than
hundreds of examples and the fact that they provide a nice
phonological generalization.  What more can one ask of a theory
(or hypothesis or law) than that it account for the observed data
in a consistent and concise manner and that it provide a reliable
generalization?  Nothing, really, except, of course, that it have
a test for falsification.

But you can't say that things that native speakers are unaware of
can't affect phonology or phonemic analysis.  Most native
speakers are unaware in any conscious way of what phonemes are.
They may use them in every utterance they make, but they can't
tell you what they are (actually, linguists have a pretty hard
time telling you what they are :>).  If you ask a native speaker
what the phonemes of his language are, he won't be able to tell
you.  Determination of phonemes can only be done by recording
unselfconscious speech and seeing, from linguistic analysis,
what the phonemes are.  And different analysts may arrive at
different phonemic analyses.  So whether native speakers are
aware of the reasons for the way they use their language or not
is irrelevant to its analysis.

As for what evidence I have that native speakers are able to keep
track of foreign words (although not necessarily specifically as
"foreign" words), I have already quoted Hock 1986 elsewhere, but
for the sake of overkill, I will quote it here again:

   The importance of the Old Irish example is twofold.  First it
   suggests that 'nativization by adoption' is qualitatively
   different from the other nativization processes:  It does not
   really nativize at all, but merely 'admits' foreign words into
   the language without losing track of the fact that they are
   and remain foreign.  It is only after these words have been
   around for quite some time, have been used often enough, and
   in enough different and novel contexts, that speakers may
   slowly begin to lose the feeling that they are not native.
   Nativization by adoption, then, is not an immediate process,
   but one of slow, gradual, even grudging acceptance.

       Hock, _Principles of Historical Linguistics_ (1986),
       396-97

This is only one person's opinion, but I don't doubt that similar
statements could be found in other textbooks of historical
linguistics, so I will consider it as evidence that native
speakers can recognize foreign words, if they have not been
phonologically nativized, for a considerable time after they have
entered the language, by their phonological peculiarities.  If
not, how can one account for the un-Anglicized pronunciation of
'chanson' in English at least 400 years after it entered the
language (first attested in 1601 according to my dictionary).
But ultimately, it is unimportant whether speakers can still
recognize foreign words after several centuries or not. The
pronunciation rules are marked in the lexicon.  The native
speaker learns these rules and follows them.  It is these rules
that produce the pattern, not the speaker's perception of the
words as native or non-native.

For other evidence that native speakers distinguish by rule the
[th] of loan words from the [dh] of native words I offer the
pattern created by the presence of intervocalic [th] and [dh] in
English words.

But first it is necessary to show that a pattern actually exists.
If there is no pattern, which the theory relies on, then the
theory is falsified.  While the fact that the six typical English
words that you offered as examples (pathology, authority,
anathema, mathematics, Gothic, Arthur) are indeed all loan words
is suggestive, it is not really sufficient.

To see if there is a pattern, I have taken a lemmatized list of
English words based on the British National Corpus (BNC)
available on the web at

  http://www.itri.brighton.ac.uk/~Adam.Kilgarriff/bnc-readme.html

This list contains 6318 lemmata whose tokens have more than 800
occurrences in the BNC of over 6 million words.  From this I have
extracted lemmata with intervocalic <th>.  These were then marked
as words that can reasonably be considered native (already in OE,
[+n]) and as words that are non-native (came into the language in
ME or later [-n]).  I then determined the pronunciation of the
<th> segment ([th] or [dh]) using my dictionary.  The data are
arranged as follows:

   1. rank of frequency (based on 2.)
   2. total number of tokens of the lemma in the corpus
   3. lemma
   4. function
   5. native/non-native
   6. pronunciation of <th> segment
   7. any additional information

2558 3218 altogether adv        [+n] [dh]
159 60182 another det           [+n] [dh]
347 28321 anything pron         [+n] [th] [+compound]
1413 6852 author n              [-n] [th]
5647 968 authorise v            [-n] [th]
303 31231 authority n           [-n] [th]
2188 4048 bother v              [??] [dh]
864 11757 brother n             [+n] [dh]
3412 2072 cathedral n           [-n] [th]
3630 1892 clothing n            [+n] [dh]
656 15599 either adv            [+n] [dh]
840 12167 either det            [+n] [dh]
5143 1134 ethical a             [-n] [th]
6050 857 ethics n               [-n] [th]
436 23216 father n              [+n] [dh]
5302 1075 feather n             [+n] [dh]
1859 4986 gather v              [+n] [dh]
5096 1149 gathering n           [+n] [dh]
5008 1184 gothic a              [-n] [th]
4135 1568 hitherto adv          [+n] [dh]
3230 2264 hypothesis n          [-n] [th]
2939 2625 leather n             [+n] [dh]
6128 841 marathon n             [-n] [th]
4823 1248 mathematical a        [-n] [th]
3517 1969 mathematics n         [-n] [th]
575 18044 method n              [-n] [th]
5396 1041 methodology n         [-n] [th]
354 27784 mother n              [+n] [dh]
1798 5227 neither adv           [+n] [dh]
2687 3018 neither det           [+n] [dh]
277 34064 nothing pron          [+n] [th] [+compound]
75 135185 other a               [+n] [dh]
264 35164 other n               [+n] [dh]
685 14959 other pron            [+n] [dh]
1150 8798 otherwise adv         [+n] [dh]
219 42341 rather adv            [+n] [dh]
1988 4553 southern a            [+n] [dh]
4330 1473 sympathetic a         [-n] [th]
3113 2377 sympathy n            [-n] [th]
309 30960 together adv          [+n] [dh]
1623 5873 weather n             [+n] [dh]
260 36169 whether conj          [+n] [dh]
4875 1232 within adv            [+n] [dh] [th] [+compound]
207 45042 within prep           [+n] [dh] [th] [+compound]
204 45867 without prep          [+n] [dh] [th] [+compound]

The list contains only 45 words and many of these are
phonologically identical with functional differences or are
derivations from the same root.  The pattern presented by these
45 words, however, is overwhelmingly consistent.  Of the native
words, five have a [th] pronunciation, and these are all
transparent compounds where the <th> occurs at a morpheme
boundary (moreover, three of these five can have either
pronunciation); the rest have the [dh] pronunciation.  Of the
non-native words, *all* have the [th] pronunciation.  As far as I
can see, there are no close calls on native vs. non-native.  The
only word whose origin is questionable is 'bother'.

One thing that can be noticed is that there are only 15 lemmata
with intervocalic [th] that can be marked as non-native (and 5 of
these are clearly derivative).  You say that there are hundreds
of these words in the language (and so there are, if not
thousands).  This makes one point quite clear.  While there may
be quite a large number of loan words in English, the core
vocabulary -- the most commonly used words -- remains primarily
native.  Thus a search of 6318 lemmata with over 800 occurrences
in a corpus of over 6 million words turns up only 15 (10 if you
eliminate derivatives) of these lemmata.

This further suggests that the knowledge of many of the lemmata
with intervocalic [th] is a function of vocabulary size.  The
larger one's vocabulary the more such lemmata (not simply as an
absolute, but as a percentage of the total) one is likely to
know.  Vocabulary size is also correlated with educational level.
Hence by the time that one has acquired a large number of such
lemmata, one is likely to be sufficiently educated to realize
that such words are loans.  Poorly educated native speakers are
not likely to realize that these words are loans.  They may even
mispronounce them; but then, dictionaries are not written (nor
often consulted) by native speakers of this educational level.
Thus my test of the pronunciation of these words may not be
entirely accurate, since I am relying on my own pronunciation and
on the pronunciation given by the dictionary.  As a turnabout,
if you have evidence that native speakers regularly mispronounce
these words because they don't know that they are loans, that
would be germane.

The best evidence that the pattern is created by rule by the
speakers of the language is the fact that the pattern exists.
For if there is a pattern in a language, synchronically it must
be created by its speakers because that is the only place that
language comes from.  If the pattern is not created by rule, then
it is simple coincidence.  If it can be shown that the pattern
created by intervocalic [th] in loan words and intervocalic [dh]
in native words is below the level of coincidence then the theory
is falsified.  Even without applying any statistical tests to the
pattern, I think coincidence can be ruled out.

Now if there were no pattern, then intervocalic [th] and [dh]
should appear scattered through the lexicon in a random fashion.
As a percentage of the total, the number of [th]'s and [dh]'s
should be about the same in both loans and native words.  In this
case, [th] and [dh] would be phonemically distinct because there
would be no way to predict which one would occur in a voiced
environment.  However, the fact that *all* the loans have [th]
suggests that this is not coincidence.  As more and more
examples of intervocalic [th] in English are investigated, I
believe that the pattern will become overwhelmingly clear, far
beyond the limits of coincidence.  Intervocalic [th] will appear
regularly in loan words, principally classical neologisms or
borrowings from French; intervocalic [dh] will appear in native
words. There will be intervocalic [th] in some native words, but
these will form patterns of their own, occurring in words where
the sound change rule did not operate or where [th] has been
restored through analogy. There will also be a significant number
of native words where either [th] or [dh] can appear as free
variants.

Thus far, I think I have shown that there is a pattern of
intervocalic [th] in non-native words and intervocalic [dh] in
native words, and that this pattern is not likely to be the
result of coincidence.

You said above that

   The rest of your post is entirely dependent on the further
   assumption that native speakers of modern English (in general,
   not just linguists) distinguish "foreign" from "native" words,
   and that the words I listed with /th/ in voiced environments
   are marked as "foreign". Since I don't share this assumption,
   ...

You are saying that this pattern can't be based on the fact that
words in English with intervocalic [th] are overwhelmingly loan
words because the speakers of the language are unaware that these
are loans.  As I mentioned above, this is not just a weak
argument, it is a spurious one.  It may be true that most
speakers do not know that these are loans, but that is irrelevant
to the phonemic analysis of the language.  Were it not, then most
linguistic analyses would have no validity because speakers of a
language are in general woefully ignorant of the linguistic
mechanisms of their language.

What native speakers know is how to use their language, not how
it works.  Now it is quite possible to be able to use a system
without knowing in detail how that system works.  Think about it
the next time you switch on your computer or your TV set or get
in your car and drive off.  With all these systems you may know
what actions on your part cause the system to do certain things,
but chances are that you don't know why the system reacts that
way or how it accomplishes that reaction.  So it is with native
speakers of a language.

The fact that naive native speakers know how to use their
language but don't know how it works has two natural
side-effects:  (1) Naive native speakers are not a reliable
source of grammatical information about their language.  If asked
if something is grammatically correct (acceptable usage), they
will be able to say "yes" or "no" or "it depends"; but if asked
why something is or is not grammatically correct, the response is
likely to be "I don't know, it just is." (2) It is possible for a
foreigner who has studied the language linguistically to know
more about how the language works than a naive native speaker
does, while still not being able to speak the language like a
native speaker.

What is important about all this is to keep in mind that it is
possible to use some system even if one does not know in detail
how it works, and that it is possible to know how a system works
without being able to use it.  It is even possible to talk about
how a system works if the casual user of the system doesn't know
what you are talking about.

Admittedly, cars and TVs and computers are different from
language.  The former are designed by human beings who know
exactly how they work.  They are designed to be used by people
who don't know how they work.  Often the people who design them
don't even want the user to know how they work.  With language,
nobody really knows how it works.  Users can use it and linguists
have studied many aspects of it to find out how it works, but
basically, linguists don't generally try to tell users how
language works.  They (the linguists) are too busy studying yet
more aspects of language in yet more detail and trying to reach
agreement on how it works.  And the users, for the most part,
don't care how language works.  Grammar (in the sense of
prescriptive rules) is something that they have to study in
school.  When they don't have to study it, they forget it
quickly.  I think the attitude to grammar of most native speakers
can be summed up by paraphrasing one of my favorite lines from
_The Magnificent Seven_:  "Grammar? -- We don't need no steenking
grammar!"  If speakers of a language really, passionately, want
to know how their language works, then they become linguists.

You are trying to maintain that it is not possible to discuss
language in linguistic terms if naive native speakers do not
realize that the linguistic terms exist or what they mean.  More
simply you maintain that linguistic descriptions of the
functioning of a language cannot be based on distinctions that
naive native speakers are unaware of.  As a reductio ad absurdum,
this would mean that 'feet' cannot possibly be the plural of
'foot' because the naive native speaker does not know what umlaut
is.  The different vowels in 'foot' and 'feet' cannot be
morphophonemic alternation because the naive native speaker does
not know what morphophonemics is.  The different pronunciation of
intervocalic [th] and [dh] cannot be attributed to a difference
between foreign and native words because naive native speakers do
not know the difference between foreign words and native words
(truly naive native speakers do not even know that there are
foreign words or, indeed, that there are other languages than
their own).

So there is a synchronic rule in English that makes 'feet' the
plural of 'foot'.  There has to be such a rule because otherwise
the plural of 'foot' would be 'foots'.  The fact that the native
speaker is unaware of the historical reason for this rule is
irrelevant.

How, then, is it possible for linguists to know more about how
language works than native speakers do?  The answer is, of
course, that linguists make it their business to know how
language works.  They study various languages to find out what is
common to them in an effort to discover how language works.
Native speakers do not know (and most of the time, they don't
care) how language works; they only know how to use a language
(note the distinction between "language" and "a language") and
they can't explain how they know this.  Knowing how to use a
system is not the same thing as knowing how the system works.

Part of the answer lies in the fact that linguists and native
speakers come about their knowledge of language in different
ways.  Native speakers learn their language from hearing it and
then imitating it, guessing at new forms by deduction, induction,
and abduction.  In doing so, they make a lot of mistakes.  Often
these mistakes are corrected at the time they are made, but
sometimes they are self-corrected after further analysis or
better imitation (and sometimes they are never corrected; this
results in change [or at least variation] in the language).
Children will often say 'foots' for 'feet'; when corrected, they
will then often use 'feets' instead of 'feet' because, although
they now know that the correct plural form is 'feet', they still
want to apply the productive rule which adds /s/ to form the
plural.

So children learn the synchronic rules of grammar by trial and
error. They don't learn rules like [+foreign] or [-native] (in
fact, features like these are probably no longer used by
linguists either, but are relics from my time in grad school;
this doesn't really matter either, since, even though the
terminology may have changed, the results will be the same).
When faced with words like 'mother' and 'mathematics' they don't
learn them as [+native] and [-native]; rather they learn them as
[+voicing rule] and [-voicing rule].  In the normal way of things
children don't learn the reasons for the rules or the reasons why
the rules don't apply in some cases.  They only learn them as
[+rule] and [-rule] or sometimes as [+rule1] / environment1 and
[+rule2] / environment2.  It remains for a linguist to look at all
the examples of [-voicing rule] and determine that the voicing
rule operated at some time in the past and that any word that
came into the language after this is not subject to it.  Native
speakers never learn this (unless they are really good at pattern
recognition) and simply know that the voicing rule applies to
some words and not to others.

So even though the native speaker may not know that the reason
for the rule that prevents voicing is based (for the most part)
on [-native], he will still know that there is some rule that
operates.  He may refer to these words as "fancy words" or as
"high-falutin' words" or as "book-learnin' words" but he will
have some way of categorizing them.  Furthermore, he will
contrast them with "plain words" or "everyday words" or "words of
one or two syllables" in a way that will remarkably parallel the
linguistic idea of [+native] and [-native] (it won't match it
exactly, because many loans, particularly short ones that match
native English phonology will not be considered "fancy words").

We have already noted that there are a remarkable number of loan
words in English, but that the core vocabulary, the most
frequently used everyday words, remains Germanic.  People who do
not recognize Greek and Latin neologisms for what they are are
not likely to have many of them in their vocabulary.  Most of
these words will fall into categories that such speakers will
have little use for.  Note also that mono- or bisyllabic loans
from other Germanic or even Romance languages are much less
likely to be recognized as loan words than polysyllabic words
from Romance or Greek.  Most speakers would not recognize 'sky'
or 'air' as loan words (or consider them "fancy" words), whereas
I feel confident that most speakers would recognize 'atmosphere'
as a loan (neologism) word (or consider it a "fancy" word).  Most
native speakers would never doubt that 'curse' is a native (or a
"plain") word, whereas 'anathema', if it is in the speaker's
vocabulary, would surely be considered foreign (or "fancy").  And
one reason that speakers will have for considering a word a
"fancy" word is that it doesn't follow the same phonological
rules that "plain" words do.  I simply maintain that native
speakers will make some kind of distinction between words like
'air' and 'atmosphere' or between 'curse' and 'anathema'.

Of course this is getting dangerously close to looking like a
circular argument:  Speakers know that words belong to this
category because they are pronounced differently and the words
are pronounced differently because they belong to this category.
But is not so circular as it may seem.  At first the speaker may
not know why (or even *that*) the words are pronounced
differently.  But as he collects more and more of them, they will
eventually form a category (whatever he may call it) in his mind
and he will know that any word that has this pronunciation
anomaly is likely to belong to this category and that any word
that he considers should belong to this category is likely to
have the pronunciation anomaly.  Rather than being circular, this
is known as positive feedback.  And in the long run, the words
are not pronounced differently because they belong to this
category; the rules for pronouncing these words are marked in the
dictionary.  It just turns out on analysis that words that are
pronounced this way are overwhelmingly likely to belong to this
category.

Linguists know that there is likely to be a difference in the
phonology of native words and foreign words.  In fact many
languages have a layer of phonology devoted to loan words.
Linguists know this because they have checked this
cross-linguistically.  Native speakers don't usually get a chance
to do this.  For example, in recent borrowings in Finnish (within
the past 200 years or so) consonant gradation does not take place
(please don't ask me to explain consonant gradation :>).  Thus
<auto> 'automobile', <auton> 'of an automobile' (compare <katu>
'street' <kadun> 'of a street') or <muki> 'mug', <mukin> 'of a
mug' (compare <mäki> 'hill' <mäen> 'of a hill').

Children learning the language, however, are unaware that these
are loanwords, and unless or until they are corrected will use
'audon'.  Once corrected, they will use 'auton' and the word will
be marked [-consonant gradation].  Chances are that the speaker
will never know why it is marked [-rule].  It is sufficient for
his purposes that it is so marked.  Some day a linguist may tell
him that it is [-rule] because it is [-native] (or with enough
education or experience he may figure it out for himself), but it
won't make any difference to the way he uses the language.  He
doesn't need to know the reason for the rule, he just needs to
know the rule.  It just makes it easier for him if he can apply
the rule to a category of words rather to each word individually.
Once he finds out that one of these words that follow this rule
is a recent loan, he is likely to consider that they all are.
And so eventually he may generalize the rule so that any word
that is perceived as a recent loan will be without consonant
gradation and any word without consonant gradation will be
considered a recent loan.

So the pattern exists; its regularity makes it extremely unlikely
to be coincidence.  The pattern can be easily accounted for on
historical grounds.  The fact that synchronic grammar is unable
to account for the pattern is not grounds for saying that the
pattern does not exist or for saying that the pattern is not
significant.  Rather, it points to a deficiency in synchronic
grammar, a blind spot where the rules that speakers use to create
this pattern cannot be reconstructed.  But synchronically the
pattern must be generated by the speakers of the language;
otherwise, it wouldn't be there.  It can't be a phonemic
distinction because it is predictable.

To get to the heart of the matter, the real crux here is the
difference between diachronic (historical) grammar and synchronic
(descriptive) grammar.  At various times it has been claimed that
synchronic grammar liberates languages and linguistics from its
historical shackles, that it makes historical linguistics
obsolete, that it severs the connection between a language's
history and the way it is used.  But as time wears on, it has
become obvious that there is really not all that much difference
between synchronic grammar and historical linguistics.  And
especially that there is virtually no difference in the results
of the two methods.  When one stops to think about it, there
can't be much difference between the results, because the results
are the same:  the modern language as it is presently spoken.  In
one case, the language is the result of the historical processes
that produced it.  In the other, it is the result of the
processes that its native speakers use to produce it.  But it is
still the same language.

So historical grammar and synchronic grammar simply focus on two
different aspects of the same thing.  And this should be a clue
that they are in many ways complementary.  Historical rules often
make it easier (or even possible) to understand synchronic rules.
But synchronic rules have to operate without reference to
historical rules because the speakers of a language are, for the
most part, blissfully unaware of the history of the grammar that
they use.  But synchronic rules and historical rules have to have
the same result (even though, for practical reasons, one may be
expressed in opposite terms from the other or many historical
rules may be compressed into a single synchronic rule) because
they both describe the same language.  Furthermore, there are
those, like Hjelmslev, who maintain that internal reconstruction
using a single language can never be historical, because anything
that can be extracted about the language in this way must be in
the language synchronically.

Therefore, if there are morphophonemic alternations in the
language, there must be synchronic rules that produce them or
else they wouldn't be there.  Using features like [-native] to
block certain rules is doubtless out of vogue as a synchronic
reconstruction since it can be argued that native speakers can't
tell the difference between native words and foreign words (and I
would agree, in principle, that the average native speaker would
be unaware of this distinction when stated this way), but there
must be some feature that acts to block the voicing rules
synchronically or else the unvoiced spirants wouldn't be there in
such a predictable pattern.

But I can't see that inventing dummy phonemes that block the
operation of rules is an improvement.  And I find it hard to say
that the difference between intervocalic [th] and intervocalic
[dh] is a phonemic distinction when all Greek and Latin loanwords
or neologisms with intervocalic [th] have [th] in English and the
only contrast that can be pointed out is between [i:th at r] (a
Greek loanword) and [i:dh at r] (a native word).  The pronunciation
of [th] in 'ether' is required by rule (as indicated by an
overwhelming pattern of occurrences even if the rule can't be
formulated synchronically) and the pronunciation of [dh] in
'either' is required by rule (ditto).  If the sounds that occur
in these words is predictable by rule, then the distinction isn't
phonemic.

So the question becomes:  Just how much *do* speakers know about
their language?  This is a question that is somehow tied up with
that of how cognition and language processing work, and, to my
knowledge, it doesn't have an answer as of now.  But it should
not be argued that linguistic descriptions of linguistic
phenomena are invalid because native speakers are unaware of
them.  Historical grammar and synchronic grammar must have the
same result because they both describe the same language.  In
this way, they are rather like the wave models and particle
models in particle physics.  While the bases of these models may
be contradictory, they must ultimately be complementary because
they both accurately describe different aspects of the same
physical reality.

Bob Whiting
whiting at cc.helsinki.fi



More information about the Indo-european mailing list