Proto-languages: what are the rules?
John Cowan
flame against the comparative method or even against its results.
Something that's always puzzled me: how do historical linguists decide
what form should be the standard reconstructed form? Consider the following
two cases:
In the reconstruction of Proto-Indo-European, we find that Germanic languages
have /f/ where other IE languages have /p/ in a large number of words.
Confidently, we reconstruct IE /p/ and postulate a sound-change that affected
Germanic only. Well and good.
But in the reconstruction of Proto-Romance, we find that Sard has /k/ where
other Romance languages have /s/ in a large number of words. In this case,
the reconstruction is /k/, and the postulation is that a sound-change
affected >all< the Romance languages except Sard. Apparent methodological
contradiction.
I realize, of course, that classical Latin provides a check on Proto-Romance.
Nevertheless, I suppose there must be other such cases that I have not
heard of. What general principle determines what is the Right Thing?
Daniel von Brighoff
>Something that's always puzzled me: how do historical linguists decide
>what form should be the standard reconstructed form? Consider the following
>
>[I]n the reconstruction of Proto-Romance, we find that Sard has /k/ where
>other Romance languages have /s/ in a large number of words. In this case,
>the reconstruction is /k/, and the postulation is that a sound-change
>affected >all< the Romance languages except Sard. Apparent methodological
>contradiction.
As a linguistics undergrad at the University of Chicago, I learned
that the principle balancing out "majority rules" is "phonetic plausibility."
Applying this to the Romance example, we notice that although there are
several documented cases of the change /k/ -> /s/,/sh/ (e.g. the division
of Proto-Indo-European into `centum' and s'atem languages, based on whether
inherited */k/ remains /k/ (as in Italic) or palatalises to /sh/ (as in
Slavic)), there is no (to my knowledge) example of a regular change in the
reverse direction, i.e. /s/ becoming /k/. Thus, we decide on a /k/ to /s/
change as being the only plausible explanation. So think of "plausiblity"
as being a more general "majority rules" principle which serves to limit the
first one.
Now, the downside: if this argument seems circular, it's because
(as faras this example goes), it is. Many of our guidelines for phonetic
plausiblity were developed through work on the Indo-European languages, in
large part because it afforded written sources old enough for us to check our
hypotheses (and, of course, Europeans were the first to use it extensively).
However, if certain phonetic changes common to I-E turn out to be _specific_
to it, we may become guilty of overgeneralisation in our attempts to see them
at work in other language families.
After all, like most sciences, historical reconstruction is still an
inexact one (which is why them *'s come in so darn handy : >).
--
Daniel "Da" von Brighoff (de...@midway.uchicago.edu) /\
951 E. 54th Place /__\
Chicago, IL 60637 /____\
Peter Christian
> As a linguistics undergrad at the University of Chicago, I learned
> that the principle balancing out "majority rules" is "phonetic plausibility."
.
.
> Now, the downside: if this argument seems circular, it's because
> (as faras this example goes), it is.
The phonetic plausibility argument is circular, but there are two
other sorts of plausibility argument which are not phonetic and are
therefore independent support for a hypothesis.
Firstly, a statistical argument. In fact the IE p > Gmc f is on one
of a whole series of differences between Gmc consonants and IE one.
It seems statistically unlikely that all the other consonant system
would change in more or less the same way leaving Gmc with the
original system. (Unless it could be shown that Gmc had broken away
and had missed out on innovations in all the other dialects.
Secondly, there's the question of the geographical distribution of
innovation. Say you have a range of dialects like this
(geographically)
X X X X X X A X X X
X X X X X X X X X X
X B X X X X X X X X
X X X X X X X C X X
What are the chances that a feature common to A,B & C but not found in
all the X dialects is an innovation? Well, it's possible, but clearly
much less likely than that they have simply not taken on an
innovation in the X dialects. In IE, the k is found in all the
Western dialects *and* the single most easterly, Tocharian. This
actually supports the k>s hypothesis without relying on majority vote
or phonetic plausibility and thus improves the credibility of the
whole hypothesis.
Peter
----------------------------------------------------------------------------
Peter Christian
Dept of European Languages pe...@gold.ac.uk
Goldsmiths' College, London. pe...@cix.compulink.co.uk
----------------------------------------------------------------------------
Joerg Knappen
but say we have two sard words, one being k****, the other s****. These two
have cognates in french, which are both s**** (in pronouciation, they are
probably written differently). It is quite easy to explain that two formerly
different sounds merge into one, but that one sound in the _same environment_
splits cannot be explained.
--J"org Knappen.
Peter Christian
> It is quite easy to explain that two formerly
> different sounds merge into one, but that one sound in the _same environment_
> splits cannot be explained.
I agree that in classical comparative linguistics this would be
inexplicable, and one would have to search for unsuspected
environment differences (cf. Verner's Law). Or grammatical function
might be important - though some might argue that that is
environment, too.
But we now know, thanks to the sociolinguists, that although sound
changes may well end up without exceptions, they may proceed not only
environment by environment, but even word by word. Look at the Ripuarian German diale
cts, which have
unshifted Gmc t in `dat' and `wat' but all other Gmc t > s,
including, I think I'm right in saying, `etwas'. One might debate
about the reasons for this particular example, but in principle
either partial change or influence of neighbouring dialects would
account for this offence against neo-grammarian sound laws.
Rich Alderson
>Disclaimer: The following expresses a desire for an explanation, NOT a
>flame against the comparative method or even against its results.
>
>Something that's always puzzled me: how do historical linguists decide
>what form should be the standard reconstructed form? Consider the following
>two cases:
They're told in their classes in comparative method. :->
In general, appeals are made to plausibility of development, based on such
things as known histories of languages, proposed linguistic universals, etc.
>In the reconstruction of Proto-Indo-European, we find that Germanic languages
>have /f/ where other IE languages have /p/ in a large number of words.
>Confidently, we reconstruct IE /p/ and postulate a sound-change that affected
>Germanic only. Well and good.
However, recently there has been a suggestion that Germanic and Armenian (which
has a similar though not identical set of changes from standard PIE) may show
the original state of affairs, and the rest of Indo-European shows a shared
innovation. This grows out of work begun to bring our reconstructions into
line with Jakobson's purported universal, that no language has voiced aspirates
without also having voiceless aspirates.
(The other objection, that no language has aspirates without also having /h/ as
a phoneme can be answered by assigning /h/ as one of the reconstructed laryn-
geals, which *I* would do for other reasons in any case.)
>But in the reconstruction of Proto-Romance, we find that Sard has /k/ where
>other Romance languages have /s/ in a large number of words. In this case,
>the reconstruction is /k/, and the postulation is that a sound-change
>affected >all< the Romance languages except Sard. Apparent methodological
>contradiction.
Not really. After all, the methodology is to examine plausibility of proposed
changes.
There is, of course, an example of a change in the opposite direction within
Indo-European. In Albanian, one[1] outcome of *s is a voiced palatal stop,
written <gj>: The word for "snake," cognate with Latin serpens, Greek herpes,
Sanskrit sarpa-, is <gjarpr>.
[1] Unfortunately, I haven't studied Albanian myself, and have to rely on the
handbooks for information like this. I should have availed myself of the
chance when I was at Chicago.
>I realize, of course, that classical Latin provides a check on Proto-Romance.
>Nevertheless, I suppose there must be other such cases that I have not heard
>of. What general principle determines what is the Right Thing?
Occam's Razor.
So many more intermediate stages must be posited for a change from a dentalveo-
lar fricative to a palatovelar stop than vice versa that we find a change from
*k > s the more likely one.
--
Rich Alderson 'I wish life was not so short,' he thought. 'Languages take
such a time, and so do all the things one wants to know about.'
--J. R. R. Tolkien,
alde...@leland.stanford.edu _The Lost Road_
Rich Alderson
>As a linguistics undergrad at the University of Chicago, I learned that the
>principle balancing out "majority rules" is "phonetic plausibility." Applying
>this to the Romance example, we notice that although there are several
>documented cases of the change /k/ -> /s/,/sh/ (e.g. the division of
>Proto-Indo-European into `centum' and s'atem languages, based on whether
>inherited */k/ remains /k/ (as in Italic) or palatalises to /sh/ (as in
>Slavic)), there is no (to my knowledge) example of a regular change in the
>reverse direction, i.e. /s/ becoming /k/. Thus, we decide on a /k/ to /s/
>change as being the only plausible explanation. So think of "plausiblity" as
>being a more general "majority rules" principle which serves to limit the
>first one.
Talk to Eric Hamp about the Albanian change I mentioned in my followup to this
query.
>Now, the downside: if this argument seems circular, it's because (as faras
>this example goes), it is. Many of our guidelines for phonetic plausiblity
>were developed through work on the Indo-European languages, in large part
>because it afforded written sources old enough for us to check our hypotheses
>(and, of course, Europeans were the first to use it extensively). However, if
>certain phonetic changes common to I-E turn out to be _specific_ to it, we may
>become guilty of overgeneralisation in our attempts to see them at work in
>other language families.
With the amount of work that has been done in, for example, the languages of
the Americas, I think any overgeneralizations would have come to light 50 years
or more ago. _Pace_ some readers of this newsgroup, linguists *as a group* are
not so hide-bound as to reject evidence set before them in systematic fashion.
>After all, like most sciences, historical reconstruction is still an inexact
>one (which is why them *'s come in so darn handy : >).
The inexactitude lies not in the science, but in the historicity, of the
discipline.
But sometimes I wish that the asterisk was in the home row on my keyboard. :->
Andries Brouwer
::
::Something that's always puzzled me: how do historical linguists decide
::what form should be the standard reconstructed form?
Several reasonable answers have been given. Rich Alderson says:
:In general, appeals are made to plausibility of development, based on such
:things as known histories of languages, proposed linguistic universals, etc.
My favourite answer is the one given by Louis Hjelmslev:
When one compares the various IE languages, one finds certain
correspondences, e.g.,
Gothic f : Irish 0 : Latin p : Gr. p : Arm. h : OI p : Tokh. p
This correspondence gets a name, for instance *p.
This name is entirely conventional, and does not imply anything
about the phonetics of a parent language.
[Not even that it had as many distinct phonemes as we find correspondences.]
Similarly, de Saussure reconstructed a "quasi-sonante" *A without
trying to define its sound.
This point of view enables one to separate the more-or-less exact
science called IE comparative linguistics from the more speculative
theories that try to assign phonetic reality to formulas like *p
(or semantic reality to reconstructed roots).
Rich Alderson
>: John Cowan writes:
>::
>::Something that's always puzzled me: how do historical linguists decide
>::what form should be the standard reconstructed form?
>
>Several reasonable answers have been given. Rich Alderson says:
>
>:In general, appeals are made to plausibility of development, based on such
>:things as known histories of languages, proposed linguistic universals, etc.
>
>My favourite answer is the one given by Louis Hjelmslev:
>When one compares the various IE languages, one finds certain
>correspondences, e.g.,
> Gothic f : Irish 0 : Latin p : Gr. p : Arm. h : OI p : Tokh. p
>This correspondence gets a name, for instance *p.
>This name is entirely conventional, and does not imply anything
>about the phonetics of a parent language.
>[Not even that it had as many distinct phonemes as we find correspondences.]
There are a number of us who disagree, more or less strongly, with Hjelmslev's
programmatic statement. If I did not believe, for example, that the proto-
phoneme in question were not at least voiceless and labial, I would not write
either a *p or a *f.
Further, if we were to follow Hjelmslev, the entire notion of the typological
incongruity of the *t/*d/*dh system would be nought but a tempest in a teapot,
since anything we write as a symbol has no objective reality.
>Similarly, de Saussure reconstructed a "quasi-sonante" *A without
>trying to define its sound.
I disagree. He was very particular in his choice of symbols, precisely because
of the phonetics of the outcomes in the European languages. He was still,
after all, writing in a system that had as its basis a Sanskrit-like phonology.
He wanted to move away from <a1> and <a2> as the vowels of the protolanguage,
towards <e> and <o>, as Brugmann did very shortly afterwards.
>This point of view enables one to separate the more-or-less exact
>science called IE comparative linguistics from the more speculative
>theories that try to assign phonetic reality to formulas like *p
>(or semantic reality to reconstructed roots).
I deny that attempts to assign phonetic reality to our reconstructions makes
them any less scientific. I'm hardly alone in this.
JA...@ukcc.uky.edu
Sard /kentu/ French /sa~/ Algheres /sent/
/sento/ /sa~/ /sent/
these two words meaning 'hundred' and 'senses (vb) respectively. Add to
these dozens more (and examples from other romance languages: cf Italian
/cento/ and /sento/, wherein I am using /c/ for a sound like that at the
margins of English 'church'), and we begin to feel that the Proto-Language
had two different sounds. Since everyone agrees on /s/ for the one set, we
need something else for the initial consonant of the other. We cannot go
much furthere at the moment, since we need to a) examing these consonants
in the pattern of other consonants in each language, b) check for items
with intriguing alternations (Italian /diko/ 'I say' vs /dice/ 'says'),
c) have a grounding in the types of change and alternations which are
common (unmarked?) in languages. Sorry if this is not elegantly stated:
in the past three weeks I've been to meetings on both coasts, my furnace
pegged out, and so did my (formerly) faithful old Tercel car, and the
local access to Usenet is playing tricks and has to be accessed differently,
and is suffering a personality change. Like they say, Murphy was an optimist.
A. Brouwer elsewhere has stated the situation nicely with respect to Indo-
European: although it's comforting to note that Sard remains a kentum
language, while the other Romance languages have defected (no a) to satem.
JA...@ukcc.uky.edu
alde...@cisco.com (Rich Alderson) writes:
>
>In article <86...@charon.cwi.nl>, aeb@cwi (Andries Brouwer) writes:
>>: John Cowan writes:
>>
>>My favourite answer is the one given by Louis Hjelmslev:
>>When one compares the various IE languages, one finds certain
>>correspondences, e.g.,
>> Gothic f : Irish 0 : Latin p : Gr. p : Arm. h : OI p : Tokh. p
>>This correspondence gets a name, for instance *p.
>>This name is entirely conventional, and does not imply anything
>>about the phonetics of a parent language.
>
>programmatic statement. If I did not believe, for example, that the proto-
>phoneme in question were not at least voiceless and labial, I would not write
>either a *p or a *f.
>
>Further, if we were to follow Hjelmslev, the entire notion of the typological
>incongruity of the *t/*d/*dh system would be nought but a tempest in a teapot,
>since anything we write as a symbol has no objective reality.
>
>>Similarly, de Saussure reconstructed a "quasi-sonante" *A without
>>trying to define its sound.
>
>I disagree. He was very particular in his choice of symbols, precisely because
>of the phonetics of the outcomes in the European languages. He was still,
>after all, writing in a system that had as its basis a Sanskrit-like phonology.
>He wanted to move away from <a1> and <a2> as the vowels of the protolanguage,
>towards <e> and <o>, as Brugmann did very shortly afterwards.
>
>>This point of view enables one to separate the more-or-less exact
>>science called IE comparative linguistics from the more speculative
>>theories that try to assign phonetic reality to formulas like *p
>>(or semantic reality to reconstructed roots).
>
>I deny that attempts to assign phonetic reality to our reconstructions makes
>them any less scientific. I'm hardly alone in this.
>--
Bravo! If we are to do reconstruction of phonology, what shall we reconstruct
if not phonology, even though we may allow that there is room for a relatively
SMALL area of doubt on DETAIL. Indeed, we continue to require of ourselves
that the reconstruction be likely phonologically, and reject unlikely and
impossible inventories for proto-languages.
Andries Brouwer
:::
:::My favourite answer is the one given by Louis Hjelmslev:
:::When one compares the various IE languages, one finds certain
:::correspondences, e.g.,
::: Gothic f : Irish 0 : Latin p : Gr. p : Arm. h : OI p : Tokh. p
:::This correspondence gets a name, for instance *p.
:::This name is entirely conventional, and does not imply anything
:::about the phonetics of a parent language.
:::This point of view enables one to separate the more-or-less exact
:::science called IE comparative linguistics from the more speculative
:::theories that try to assign phonetic reality to formulas like *p
:::(or semantic reality to reconstructed roots).
Rich Alderson answered:
::I deny that attempts to assign phonetic reality to our reconstructions makes
::them any less scientific. I'm hardly alone in this.
I answered Rich Alderson by email, but maybe we should continue
the discussion here. Let us distinguish two stages. The first is
almost mechanical (the donkey work, as someone remarked in recent post):
take the vocabularies (and collections of morphemes in various paradigms)
of all IE languages and deduce the regular correspondences of sounds
(or maybe only of letters, in cases where the pronounciation is doubtful).
The second stage is highly speculative: from the results of the first
stage try to reconstruct a PIE language.
From a theoretical or methodological point of view, both stages
have their problems. For example, what precisely does one do
in order to "deduce the regular correspondences"?
Roughly speaking, the algorithm is: 1. collect sets of obvious
cognates [words look similar, have a similar meaning]
2. deduce the regular correspondences by seeing what happens
in the overwhelming majority of cases
3. go back to 1, where now "look similar" means "look similar
after applying the known transformations".
So, in the first iteration of the algorithm everyone will think that
German "haben" and Latin "habere" (to have) are obvious cognates,
but after deducing the correspondence German h : Latin c one is
forced to change one's mind, and rather consider "haben" and
"capere" as possible cognates.
Similarly, in the first iteration of the algorithm my computer
thought that Dutch "woud" (forest) and English "wood" were
obvious cognates, but after deducing the correspondence
Dutch oud : English old it had to change its mind and relate
"woud" to "wold".
A priori there is no reason why the outcome of this iterative
process should be independent of the individual linguist's
judgements, but in practice it is, to a high degree.
To a first approximation, Stage One is immensely successful:
a very large amount of data is explained by a relatively small
number of correspondences. However, the better one looks the
more special rules are required to explain smaller and smaller
groups of exceptions and exceptions to the exceptions.
The outcome of Stage One is a set of correspondences, like
Gothic f : Irish 0 : Latin p : Gr. p : Arm. h : OI p : Tokh. p
possibly provided with a context.
How can one explain the existence of such correspondences?
By assuming that these various languages are in some way
continuations of one single language. But we do not *know*
a single thing about this language, if it ever existed.
Thus, in Stage Two one uses heuristics, and plausibility, and so on.
:alde...@cisco.com (Rich Alderson) writes:
::
::There are a number of us who disagree, more or less strongly, with
::Hjelmslev's programmatic statement. If I did not believe, for example,
::that the proto-phoneme in question were not at least voiceless and labial,
::I would not write either a *p or a *f.
::
::Further, if we were to follow Hjelmslev, the entire notion of the
::typological incongruity of the *t/*d/*dh system would be nought but
::a tempest in a teapot, since anything we write as a symbol has no
::objective reality.
But really there are many assumptions built into these remarks.
From the data we have, we reconstruct a hypothetical language, PIE,
spoken, say, 3000BC. But why was it a single language at that time?
Maybe there was a large amount of dialectal variation, and a more
uniform parent language, X, spoken, say, 10000BC. In such a case
"the proto-phoneme in question" would not be well-defined for PIE,
and the time gap between X and PIE would allow any number of intervening
sound changes so that it would be impossible to say something about
the phonetics of X.
Or, just to make a completely different objection, why
"voiceless and labial"? In current Danish the b/p distinction
is not voiced/voiceless, but unaspirated/aspirated. I cannot
see how one might conclude that "voiceless" was a distinctive
characteristic of a PIE *p.
JA...@UKCC.UKY.EDU writes:
:Bravo! If we are to do reconstruction of phonology, what shall we reconstruct
:if not phonology, even though we may allow that there is room for a relatively
:SMALL area of doubt on DETAIL. Indeed, we continue to require of ourselves
:that the reconstruction be likely phonologically, and reject unlikely and
:impossible inventories for proto-languages.
The "relatively SMALL area of doubt on DETAIL" might apply to the
outcome of Stage One, although for example the precise distribution
of laryngeals (or even their number) is still under debate.
Thus, the majority of the entries in an etymological dictionary
from 1920 agree to a large extent with those in one from 1990.
On the other hand, no year passes without the publication of yet
another proposal for the inventory of phonemes in PIE. This part
of the theory cannot be regarded as firmly established.
(And what are "impossible inventories for proto-languages"?)
Rich Alderson
>:::
>:::My favourite answer is the one given by Louis Hjelmslev:
>:::When one compares the various IE languages, one finds certain
>:::correspondences, e.g.,
>::: Gothic f : Irish 0 : Latin p : Gr. p : Arm. h : OI p : Tokh. p
>:::This correspondence gets a name, for instance *p.
>:::This name is entirely conventional, and does not imply anything
>:::about the phonetics of a parent language.
>
>:::This point of view enables one to separate the more-or-less exact
>:::science called IE comparative linguistics from the more speculative
>:::theories that try to assign phonetic reality to formulas like *p
>:::(or semantic reality to reconstructed roots).
>
>Rich Alderson answered:
>::I deny that attempts to assign phonetic reality to our reconstructions makes
>::them any less scientific. I'm hardly alone in this.
I (over?)reacted to the word "speculative" in the post quoting Hjelmslev. I've
always taken his statement as limiting what we can, or should, do as linguists
to an uninteresting collection of correspondences without ever allowing our-
selves to go beyond the mechanical.
>I answered Rich Alderson by email, but maybe we should continue
>the discussion here. Let us distinguish two stages. The first is
>almost mechanical (the donkey work, as someone remarked in recent post):
>take the vocabularies (and collections of morphemes in various paradigms)
>of all IE languages and deduce the regular correspondences of sounds
>(or maybe only of letters, in cases where the pronounciation is doubtful).
>The second stage is highly speculative: from the results of the first
>stage try to reconstruct a PIE language.
I asked for persmission to post the e-mail reply mentioned, and received same,
before seeing this posting. They cover much the same ground, though a couple
of comments from the e-mail may be of interest. I will include it at the end
of this followup for completeness.
To quote Lucretius, "ex nihilo nihil." I think we as linguists would agree
that the language must have existed, though not *exactly* as we reconstruct it.
>:alde...@cisco.com (Rich Alderson) writes:
>::
>::There are a number of us who disagree, more or less strongly, with
>::Hjelmslev's programmatic statement. If I did not believe, for example, that
>::the proto-phoneme in question were not at least voiceless and labial, I
>::would not write either a *p or a *f.
>::
>::Further, if we were to follow Hjelmslev, the entire notion of the
>::typological incongruity of the *t/*d/*dh system would be nought but a
>::tempest in a teapot, since anything we write as a symbol has no objective
>::reality.
>
>But really there are many assumptions built into these remarks.
True, but without those assumptions (or some set of assumptions) to guide the
work at hand, there would be nothing interesting to say after the correspon-
dences were noted in the first place.
A difference between me and a number of colleagues and predecessors is that I
explicitly state that I *am* making claims when I write one symbol rather than
another for a reconstructed phoneme. This is not to denigrate their work, but
to clarify my opposition to Hjelmslev, who, if taken literally, would be no
less happy to write <t> or <R> to symbolize the correspondence set noted above
than to write <p> or <f>.
>From the data we have, we reconstruct a hypothetical language, PIE, spoken,
>say, 3000BC. But why was it a single language at that time? Maybe there was a
>large amount of dialectal variation, and a more uniform parent language, X,
>spoken, say, 10000BC. In such a case "the proto-phoneme in question" would not
>be well-defined for PIE, and the time gap between X and PIE would allow any
>number of intervening sound changes so that it would be impossible to say
>something about the phonetics of X.
Clearly there was dialectal variation, but just as clearly, given the kind of
agreement between the languages in question we can be to a large degree certain
about smoe of our reconstructions.
And just as clearly, there are reconstructions that must be reconsidered in the
light of changes (advances) in linguistic theory.
But if we follow Hjelmslev, there is nothing to be considered, because any
symbol is as good as any other for our correspondence sets.
[Side note: I would place the period of commonality at around 6000 years
before present, which gives us approximately 2000 years before the earliest
texts, and the period of breakup at around 5000 years ago, that is, your
figure. Certainly, I think 12000 years before present is much too long a time
scale--one of my objections to Renfrew's claims.]
>Or, just to make a completely different objection, why "voiceless and labial"?
>In current Danish the b/p distinction is not voiced/voiceless, but
>unaspirated/aspirated. I cannot see how one might conclude that "voiceless"
>was a distinctive characteristic of a PIE *p.
Why voiceless? Because there is an entire *system* of correspondences, which
clearly show a three-way opposition among the non-sibilant buccal obstruents.
(If the laryngealists are correct and voiceless aspirates arise from voiceless
plain + laryngeal; otherwise, it's *four*-way.) Sets such as the one you list
contrast with sets in which the members are voiced and with sets in which some
embers are voiced but with some additional feature that makes them contrast
with the other voiced ones. (I know this is simplified, but I don't want to
get into the entire set of correspondences right now.)
Why labial? Because the preponderance of correspondences across the languages
are labial.
I hope you noticed that I didn't insist that this particular set is a stop.
>JA...@UKCC.UKY.EDU writes:
>
>:Bravo! If we are to do reconstruction of phonology, what shall we
>:reconstruct if not phonology, even though we may allow that there is room for
>:a relatively SMALL area of doubt on DETAIL. Indeed, we continue to require
>:of ourselves that the reconstruction be likely phonologically, and reject
>:unlikely and impossible inventories for proto-languages.
>
>The "relatively SMALL area of doubt on DETAIL" might apply to the outcome of
>Stage One, although for example the precise distribution of laryngeals (or
>even their number) is still under debate. Thus, the majority of the entries
>in an etymological dictionary from 1920 agree to a large extent with those in
>one from 1990. On the other hand, no year passes without the publication of
>yet another proposal for the inventory of phonemes in PIE. This part of the
>theory cannot be regarded as firmly established. (And what are "impossible
>inventories for proto-languages"?)
There is detail and there is detail: I'm happy enough with a stop *p, although
I wouldn't be unhappy with *f, but I don't think we can say whether it was an
aspirated or an unaspirated stop. I think we *can* say that it wasn't an
ejective or implosive.
(Note for non-Indo-Europeanists: In the following, understand the triplet of
symbols t/d/dh to refer to the entire buccal obstruent system, and not just the
dentals.)
As for impossible phoneme inventories, many Indo-Europeanists are unhappy with
the (implicit) claims made about the three-way distinction which we reconstruct
by the symbols we traditionally use for it--the t/d/dh that I mentioned in my
disavowal of Hjelmslev. According to Jakobson, it violates a phonological
universal, that /dh/ does not occur in a system that does not contain /th/.
I personally have worked on the issue of the phonological likelihood of the
vowel system reconstructed by the laryngealists: A single mid vowel alternat-
ing between front unrounded and back rounded allophones, with distinctive
length in certain morphological settings and a zero allophone in others. At
one time I argued for making it a low vowel by typological comparison with
Kabardian; I would still make it a low vowel, but I would reconstruct a larger
inventory of vowels as phonemes rather than as allophones of consonantal
resonants.
We won't settle the issue of the laryngeals here, but given the Anatolian
evidence, along with a revised reconstruction of the vowel system, I think we
can safely posit only 4, and have a good shot at assigning them symbols that
make phonemic claims with which we can live.
Below I have included Andries Brouwer's e-mail to me, with his permission:
>Received: from dirt.cisco.com by last-call.cisco.com with TCP; Fri, 15 Jan 93
> 06:27:32 -0800
>Received: from charon.cwi.nl by dirt.cisco.com with TCP; Fri, 15 Jan 93
> 06:27:45 -0800
>Received: from papegaai.cwi.nl by charon.cwi.nl with SMTP
> id AA29149 (5.65b/3.6/CWI-Amsterdam); Fri, 15 Jan 1993 15:27:38 +0100
>Received: by papegaai.cwi.nl
> id AA18425 (5.65b/3.5/CWI-Amsterdam); Fri, 15 Jan 1993 15:27:37 +0100
>Date: Fri, 15 Jan 1993 15:27:37 +0100
>From: Andries...@cwi.nl
>Message-Id: <9301151427....@papegaai.cwi.nl>
>To: alde...@cisco.com
>Subject: Re: Proto-languages: what are the rules?
>Newsgroups: sci.lang
>
>In sci.lang you write:
>
>:In article <86...@charon.cwi.nl>, aeb@cwi (Andries Brouwer) writes:
>::: John Cowan writes:
>::::
>::::Something that's always puzzled me: how do historical linguists decide
>::::what form should be the standard reconstructed form?
>::
>::Several reasonable answers have been given. Rich Alderson says:
>::
>:::In general, appeals are made to plausibility of development, based on such
>:::things as known histories of languages, proposed linguistic universals, etc.
>::
>::My favourite answer is the one given by Louis Hjelmslev:
>::When one compares the various IE languages, one finds certain
>::correspondences, e.g.,
>:: Gothic f : Irish 0 : Latin p : Gr. p : Arm. h : OI p : Tokh. p
>::This correspondence gets a name, for instance *p.
>::This name is entirely conventional, and does not imply anything
>::about the phonetics of a parent language.
>::[Not even that it had as many distinct phonemes as we find correspondences.]
>
>
>:There are a number of us who disagree, more or less strongly, with Hjelmslev's
>:programmatic statement. If I did not believe, for example, that the proto-
>:phoneme in question were not at least voiceless and labial, I would not write
>:either a *p or a *f.
>
>:Further, if we were to follow Hjelmslev, the entire notion of the typological
>:incongruity of the *t/*d/*dh system would be nought but a tempest in a teapot,
>:since anything we write as a symbol has no objective reality.
>
>
>::This point of view enables one to separate the more-or-less exact
>::science called IE comparative linguistics from the more speculative
>::theories that try to assign phonetic reality to formulas like *p
>::(or semantic reality to reconstructed roots).
>
>:I deny that attempts to assign phonetic reality to our reconstructions makes
>:them any less scientific. I'm hardly alone in this.
>
>
>No, indeed I agree with you. But you oppose a claim that I did not make.
>I only said that this second branch of science is of a more speculative
>nature than the first. It is possible to `prove' a correspondence like
>Gothic f : Latin p by giving a thousand examples - we have facts here,
>established beyond reasonable doubt.
>However, I cannot see how anyone could convince me of
>any particular realisation of *p in PIE. You write above
> If I did not believe, for example, that the proto-
> phoneme in question were not at least voiceless and labial, ...
>and maybe I share your belief, but there are really many objections:
>- maybe voiced/voiceless was not a significant distinction in PIE;
> for example, in current Danish the b/p distinction is not voiced/voiceless,
> but unaspirated/aspirated, so that the English `pudding' was borrowed
> as `budding'.
>- maybe there never was a uniform PIE (at the stage we are reconstructing,
> say 3000BC or perhaps 5000BC), but, say, a common predecessor 10000BC.
> In such a case there would be quite a lot of `dialectal' variation,
> and no single realisation of *p. And if we go back a few thousand years
> to a time where the language was uniform (if indeed there was such a time),
> then we have ample room for many intervening sound changes.
>Etc. I do not deny that it is meaningful to think about the phonemics of
>PIE, but stated only that it involves a high degree of speculation,
>while finding correspondences is almost a mechanical activity.
>[Although it is very difficult to give this activity a precise
>theoretical foundation. You know the procedure: start with a collection
>of `obvious' cognates, deduce sound laws, revise your idea about what
>are obvious cognates, etc. So I start with Latin habere, German haben
>`to have' and Dutch woud, English wood `forest' and am very disappointed
>that I cannot call these cognates anymore after the first stage.
>There is no fundamental reason why there could not be more than one
>final stage, depending on the order in which the evidence was examined.]
>
>Hope I am not too incoherent - best regards - Andries Brouwer
I'm not entirely sure what is meant by "more than one final stage, depending on
the order in which the evidence was examined." Are you talking about a final
stage of the reconstruction *process*, or about successive "final" stages of
the *output* of the reconstruction process as more data becomes available?