Nostratic
S.A.T. Haldane
I'm not a linguist, merely an interested amateur. I've heard the term
'nostratic' bandied about but I'm not sure what it means. Can anyone please
explain it to me? Is it, as I suspect, a common ancestor language to
language groups such as Indo-European and Hamito-Semitic? If anyone can
recommend me any good books on the topic (or anything relating to the
tracing back of language groups to proto-languages really since it's such a
fascinating topic) I'd be grateful.
Cheers,
Samuel
Mark Rosenfelder
In article <SAH.96Ju...@mscsv1.dl.ac.uk>,
This is in the FAQ; here's the relevant entry:
===============================================================================
22. What about Nostratic and Proto-World?
[--markrose]
In recent years some some linguists have attempted to reconstruct languages
far older than Indo-European.
Soviet linguists led by Vladislav Illich-Svitych, Aaron Dolgopolsky, and
Vitaly Shevoroshkin have done work on *Nostratic*, said to underlie the
Indo-European, Kartvelian (South Caucasion), Afro-Asiatic, Dravidian,
Uralic, Altaic, Chukchi-Kamchatkan, and Eskimo-Aleut families.
The methodology is the traditional comparative method, and over 600 roots
have been proposed. Most linguists remain skeptical, believing that chance
processes will have obscured any relationship at this level beyond
reconstruction, or question the accuracy of the derivations (a charge which
makes Nostraticists bristle). Others simply suspend judgment, especially
since much of the supporting material for Nostratic is available only in
Russian.
A good overview on Nostratic is Kaiser and Shevoroshkin, "Nostratic", in
the _Annual Review of Anthropology_, 17:309. Illich-Svitych's original
Russian article (from _Etymologia_, 1965) has been translated in
Shevoroshkin, ed., RECONSTRUCTING LANGUAGES AND CULTURES (1989).
Joseph Greenberg has proposed a grouping which covers much the same language
areas (omitting Afro-Asiastic and Dravidian, but adding Ainu and Gilyak),
called *Eurasiatic*. Greenberg's method of _mass comparison_ (which he has
also used to group together almost all Native American languages into one
superfamily, Amerind) basically consists of assembling huge lists of
common words and doing eyeball comparisons.
This methodology has been severely criticized by many historical linguists.
If 'mass comparison' were applied to the Indo-European languages, it would
be bedevilled by false positives (caused by borrowing or chance) and by
specious phonetic or semantic similarites. Greenberg's methods seem to
linguists to abandon the very methodological severity which has put
Indo-European linguistics on a scientific footing, and distinguished it from
the work of cranks. Relax the rules enough, and you can derive any language
from any other.
Greenberg replies that the patterns he has found are compelling enough to
make his classifications a compelling theory, and that he is merely
following in the footsteps of the originators of the comparative method:
linguists had to decide that the Indo-European languages were related before
attempting reconstructions.
The ultimate areal comparison would be *Proto-World*, the hypothetical
ancestor of all human languages. Greenberg has mentioned Proto-World, but
since he is not much interested in reconstruction, his proposal is not much
more than a statement of the monogenetic theory (a single origin for all
languages). Most linguists are skeptical that anything could be
reconstructed at this hypothetical time depth.
Greenberg's work on Amerind can be found in LANGUAGE IN THE AMERICAS (1987);
on Eurasiatic, in the forthcoming INDO-EUROPEAN AND ITS CLOSEST RELATIVES:
THE EURASIATIC LANGUAGE FAMILY. Introductions to the Nostratic and
Proto-World controversies were published in both _The Atlantic_ and
_Scientific American_ in April 1991. The essays in Lamb and Mitchell, eds.,
SPRUNG FROM SOME COMMON SOURCE (1991), are also relevant.
Loren Petrich maintains an annotated bibliography on Indo-European,
Nostratic, and Proto-World. I am also indebted to Peter Michalove for
citations used in this entry.
Richard M. Alderson III
In article <SAH.96Ju...@mscsv1.dl.ac.uk> s...@mscsv1.dl.ac.uk
(S.A.T. Haldane) writes:
>I'm not a linguist, merely an interested amateur. I've heard the term
>'nostratic' bandied about but I'm not sure what it means. Can anyone please
>explain it to me? Is it, as I suspect, a common ancestor language to language
>groups such as Indo-European and Hamito-Semitic? If anyone can recommend me
>any good books on the topic (or anything relating to the tracing back of
>language groups to proto-languages really since it's such a fascinating topic)
>I'd be grateful.
Nostratic was first proposed by Holger Pedersen in 1903 as the hypothetical
ancestor of the Indo-European, Hamito-Semitic, Finno-Ugric/Uralic, and Altaic
language families. It was equated with Proto-Indo-European/Hamito-Semitic by
most scholars in the early part of the century, but some work was done on it by
Albert Cuny.
Interest was revived by Illich-Svitych in the 1960s; his work was continued by
a group of scholars in Moscow, most especially Aharon Dolgopolskij and Vitaly
Shevoroshkin. Other scholars whose names are associated with Nostratic studies
are Alexis Manaster-Ramer and Alan Bomhard; a similar proposal, Eurasiatic, has
been put forth by Joseph Greenberg.
The name of the proposed superfamily is dreived from the Latin _nostras_
"fellow countryman."
--
Rich Alderson You know the sort of thing that you can find in any dictionary
of a strange language, and which so excites the amateur philo-
logists, itching to derive one tongue from another that they
know better: a word that is nearly the same in form and meaning
as the corresponding word in English, or Latin, or Hebrew, or
what not.
--J. R. R. Tolkien,
alde...@netcom.com _The Notion Club Papers_
Jacques Guy
S.A.T. Haldane <s...@mscsv1.dl.ac.uk> wrote:
> I'm not a linguist, merely an interested amateur. I've heard the term
> 'nostratic' bandied about but I'm not sure what it means.
About the credibility of reconstructions (and proof of the
existence) of Nostratic, read Don Ringe's interview by Sidwell
which you will find at:
Harry Bowman
Jacques Guy wrote:
>
> S.A.T. Haldane <s...@mscsv1.dl.ac.uk> wrote:
> > I'm not a linguist, merely an interested amateur. I've heard the term
> > 'nostratic' bandied about but I'm not sure what it means.
>
> existence) of Nostratic, read Don Ringe's interview by Sidwell
> which you will find at:
>
> http://munkora.cs.mu.oz.au/%7Ensn/Work/ringe.html
If you take a look at this interview, some rather bad things
are said in it. In particular, it is suggested that reconstructions
at large time depths (10,000 years or so) are impossible because two
languages that split that long ago are only going to have a few
cognates on a list of common vocabulary terms. I would agree with
this, but the two lists could be lists of words from Bengali and
English. These two languages have a split date of 7,000 years or
so ago, but it can be clearly demonstrated that they are related.
However, the evidence is primarily found in the other 150 or so Indo-
European languages.
(For those wondering, a split date of 7000 years ago translates
to 20% cognate in standard glottochronology, and 10,000 years translates
into 10% cognate).
The other implausibility here is that these people seem to
argue that the maximum time depth for which we can construct a proto-
language is about 7000 years. This number sounds awfully suspicious.
This is the time depth of Indo-European, the first major family to be
identified. It is as if the inventors of the automobile immediately
produced a Ferrari Testarossa, and after 100 years of further effort
(or 200 in the case of language reconstruction), no better cars have
been built. In fact, there are well-established taxa with a time
depth in excess of this figure. The Afro-Asiatic languages (For those
who haven't looked at the FAQ, Hebrew, Arabic, Berber, Ancient Egyptian,
Somali, Hausa, and a couple of hundred others) have an estimated time
depth of someting like 9000 years. This is not far from what is
claimed for at least the Circumpolar group of Nostratic (This is what
Greenberg calls Eurasiatic- Indo-European, Uralic-Yukaghir, Altaic
(possibly with Korean, Ainu, and Japanese), Chukotko-Kamchatkan, the
Gilyak language, and Eskimo-Aleut). The Australian phylum is well
accepted, and some evidence suggests that it is far older than 10,000
years old.
Also, there is the fact that some words are very stable. I
admit that anything can be lost (note the near-nonexistence of pronouns
in Japanese), it is noticeable that most of the languages involved
in the Nostratic hypothesis have first-person pronouns with m- and
second-person pronouns with t-. (There is a suggested k- root for
an ergative or nominative pronoun which may even be found in Japanese
as the particle <ga>.) This seems unlikely to be a chance resemblance.
I would like to see some more statistics used in this field.
I am not an expert (I am an astronomy graduate student), but I know
that many things can be found statistically which seem impossible if
large data sets are not used (note recent measurements of nonuniformity
in the cosmic background radiation and the proper motions of the
Magellanic Clouds).
More on this controversy can be found (from the pro- deep
reconstruction standpoint) in Language in the Americas (Greenberg
1987) and in Classification of the World's Languages (Ruhlen 1987).
It seems that the dispute is often rather nasty. They certainly
don't get along as well as us astronomers.
PS: For those of you who don't believe in Amerind, would you like
to suggest evidence that the Americas were inhabited more than 12,000
years ago? Certainly a sample of 600 languages should contain
enough information to show relationship.
Richard M. Alderson III
>If you take a look at this interview, some rather bad things are said in it.
>In particular, it is suggested that reconstructions at large time depths
>(10,000 years or so) are impossible because two languages that split that long
>ago are only going to have a few cognates on a list of common vocabulary
>terms. I would agree with this, but the two lists could be lists of words
>from Bengali and English. These two languages have a split date of 7,000
>years or so ago, but it can be clearly demonstrated that they are related.
>However, the evidence is primarily found in the other 150 or so Indo- European
>languages.
That's Ringe's point: You *can't* prove English and Bengali are related by
using *only* English and Bengali as a starting point.
The problem with Ringe's straw man is that no real historical linguist compares
only two languages at a time. Bloomfield, for example, advised the use of four
languages in a presumed family; this allows a reasonable chance for gleaning at
least the preliminaries for phonological and morphological correspondences but
does not require a great deal of cross-indexing of citations.
Ringe seems to have bought into Greenberg's version of non-massive-comparison,
which was a bogus straw man in Greenberg's work, as well.
>(For those wondering, a split date of 7000 years ago translates to 20% cognate
>in standard glottochronology, and 10,000 years translates into 10% cognate).
Oh? Is there a standard for glottochronology? Never seemed to be one when I
was in school.
>The other implausibility here is that these people seem to argue that the
>maximum time depth for which we can construct a proto-language is about 7000
>years. This number sounds awfully suspicious. This is the time depth of
>Indo-European, the first major family to be identified.
I agree, the setting of an absolute limit around the hypothetical time-depth of
PIE is rather disingenuous. The proofs I've seen offered for such a limitation
have struck me as so much Kantian tail-chasing.
By the bye, you're wrong about Indo-European. Honour of place goes to Finno-
Ugric, in a monograph demonstrating their incontrovertible relationship that
appeared in 1752. Second place I would give to Dr. Johnson, for his demonstra-
tion of the Germanic family, in the 1760s or early 1770s, as I recall. His
good friend Sir Wm. Jones, then, comes in third with Indo-European, in 1786.
I have nothing to say about Afro-Asiatic--it is too clearly a good counter-
example--or about Eurasiatic--Greenberg has wasted enough of my time already in
this life.
>The Australian phylum is well accepted, and some evidence suggests that it is
>far older than 10,000 years old.
Is it? By whom? The folks in the field from whom I've heard seem to think
it's *likely*, but it's not established in any rigourous sense--and I have
heard a lot of hand-waving about "We can't do it the way you Indo-Europeanists
do, there's too much borrowing and the like." I don't buy it.
>Also, there is the fact that some words are very stable.
That's not a fact. That's *not* a fact. That's NOT A FACT!
It's an hypothesis, and a damned silly one at that. Leads moderately talented
historical linguists up the garden path. Makes for bad linguistics.
It's a bloody crutch, dreamed up by folks who think they can get around the
damned hard work of research by *assuming* that this *must* be true because
they want so much that it *be* true.
>I would like to see some more statistics used in this field. I am not an
>expert (I am an astronomy graduate student), but I know that many things can
>be found statistically which seem impossible if large data sets are not used
I *am* an expert. I've spent more than 25 years learning this stuff. I even
have a passing acquaintance with statistics, having done some real hard science
when I thought I might like to be a phoneticist: Acoustical physics, analysis
of variance on various things. Never really fell in love with it--not enough
to give up Indo-European studies.
So tell me. What would you like to see quantified? Because using statistics
will require quantification. Have you seen Nichol's book? As far as I can
tell, this is the most rigourour application of statistical methods to the
issues of historical and comparative linguistics yet written--and all she was
able to do was to demonstrate that her biases were borne out by her statistics.
I could do as much, I think.
>More on this controversy can be found (from the pro- deep reconstruction
>standpoint) in Language in the Americas (Greenberg 1987) and in Classification
>of the World's Languages (Ruhlen 1987). It seems that the dispute is often
>rather nasty. They certainly don't get along as well as us astronomers.
The dispute is nasty because these two have abandoned rigourous research for
"feel good" pseudo-linguistics.
>PS: For those of you who don't believe in Amerind, would you like to suggest
>evidence that the Americas were inhabited more than 12,000 years ago?
>Certainly a sample of 600 languages should contain enough information to show
>relationship.
It certainly should--but there has to be some rigour in the examination of the
data, and Greenberg simply did not do the work. (I suspect that if he did not
know, due to his predecessors, that Na-Dene was relatively recently come into
North America, *that* would have been part of "Amerind" as well.) You can say
anything, make any outrageous claim you wish--but you'd damned well be able to
back it up with theoretical rigour.
Jacques Guy
This is tedious and I am absolutely fed right up to my eyeteeth
with these ever-reborn ignorant, mathematically and linguistically
unwashed arguments.
Harry Bowman <hl...@cornell.edu> writes:
<<If you take a look at this interview, some rather bad things are said in it.
In particular, it is suggested that reconstructions at large time depths
(10,000 years or so) are impossible because two languages that split that long
ago are only going to have a few cognates on a list of common vocabulary
terms. I would agree with this, but the two lists could be lists of words
from Bengali and English. These two languages have a split date of 7,000
years or so ago, but it can be clearly demonstrated that they are related.>>
Thanks to Sanskrit, and Latin, and Greek, all of which take us a good
2500 years and more back in time. In other words, there is historical,
documentary evidence. So, the recognition that Bengali and English are
related is made possible by documents in those *mesolanguages*: Sanskrit
(and Pali), Latin, Classical Greek. It is therefore false to extrapolate
saying that two languages having split 7000 years ago would equally
be demonstrably related in the absence of similar historical documentary
evidence.
Alderson:
> The problem with Ringe's straw man is that no real historical linguist compares
> only two languages at a time.
It does not matter how many languages you take into account. What matters is
the amount of evidence for relatedness that remains between the two most
*phenetically* distant languages. Ringe is not setting up a straw man, period.
Bowman:
<<(For those wondering, a split date of 7000 years ago translates to 20% cognate
in standard glottochronology, and 10,000 years translates into 10% cognate)>>
Alderson:
<<Oh? Is there a standard for glottochronology? Never seemed to be one when I
was in school.>>
Yes, there is: INTELLECTUAL DISHONESTY. Proponents of glot.chron. keep harping
at the same old trite tripe, again, and again, and again, however much their
claims are proved false. They just wait a bit, and regurgitate the same claptrap.
The latest I read here (or, rather, it must have been on LINGUIST) is that
on the 100-item Swadesh list no language was known to have less that 86%
retention per 1000 years. When you adduce evidence to the contrary, the answer
is the same old tired tripe: the list needs refining.
And now this is typical of the sorry state of historical linguistics, drawing
attention to the trees in order to hide the forest:
<<By the bye, you're wrong about Indo-European. Honour of place goes to Finno-
Ugric, in a monograph demonstrating their incontrovertible relationship that
appeared in 1752. Second place I would give to Dr. Johnson, for his demonstra-
tion of the Germanic family, in the 1760s or early 1770s, as I recall. His
good friend Sir Wm. Jones, then, comes in third with Indo-European, in 1786.>>
Ands this is absolute claptrap, aka proof by assertion:
<<The Australian phylum is well accepted, and some evidence suggests that it is
far older than 10,000 years old.>>
And here is a fine example of begging the question:
<<Also, there is the fact that some words are very stable.>>
I wrote earlier on, didn't I? That's it. Word replacement is *random*,
by which is meant that it is no more predictable than the fair roll
of a die. Looking back, however, you can say: "Oh, this word ("snot" for
instance) is remarkably stable!". Given a large enough number of languages
you will always find remarkably stable words. Just like, give a large
number of spins of a roulette, you will find long sequences of the same
colour turning up.
And now, for a bit of a change: j'en ai plein le cul, ras-le-bol, mais
alors la, jusque par-dessu la tete et au-dela de voir ce fatras de conneries
ressortir continuellement etre faire la une de soi-disant revues
scientifiques comme Scientific American, et maintenant Nature, et New
Scientist. Quand on voit les aneries que ca publie en linguistique
on commence a douter serieusement de la valeur de ce que ca publie en
science.
Et merde!
Jacques Guy
Harry Bowman wrote:
> (For those wondering, a split date of 7000 years ago translates
> to 20% cognate in standard glottochronology, and 10,000 years translates
> into 10% cognate).
For 20% cognates to remain after 7,000 years the mean retention rate
would have to be 89.14% per 1000 years (0.8914^14 =0.20. Now if by
"standard glottochronology" is meant that in Swadesh's original IJAL
1950 article, the retention rate it supposes is 81% calculated on
his 200-item list. 0.81^14 = 0.0523, i.e. 5%. So, on the 200-item
list you are likely to find 10 cognates. My contention is that,
unless the sound correspondences are quite transparent, you will
never find them. Further, if you allow for semantics at all, then
the number of chance resemblances will drown the little evidence
for cognation. Using a longer list will change nothing.
If by "standard glottochronology" is meant the figures in Lees 1953,
then the mean retention rate is even lower than that and the proportions
of remaining cognates drops according.
> Also, there is the fact that some words are very stable.
That, as I have already written, is known with hindsight. Further,
the very stable words of one language family (say, "good" in
Romance) are not the same in other ("good" is extremely unstable
in the languages of Vanuatu). Stability is thus an artifact of
the choice of a standard wordlist. Indeed, since the purpose
of eliciting a standard wordlist in putatively related languages
is to work out sound correspondences (displicet Ruhlenio) and
find cognate, the items of such wordlists are selected for
stability and "universality" (by which is meant, the notions
ought to exist in all languages, or at least, as many as possible).
Evidence that this is so is that retention rates calculated using
the 100-item Swadesh list are generally higher than using the
200-item list.
> I would like to see some more statistics used in this field
> [chance resemblances].
The formulas involved become quickly completely intractable. That is why
I used Monte-Carlo simulations in my Anthropos article to arrive at
estimates of chance resemblances.
> More on this controversy can be found (from the pro- deep
> reconstruction standpoint) in Language in the Americas (Greenberg
> 1987) and in Classification of the World's Languages (Ruhlen 1987).
> It seems that the dispute is often rather nasty. They certainly
> don't get along as well as us astronomers.
The nastiness is due to the fact that linguistics is as much of a
science as astronomy was when it was still astrology. More than half
of linguistics, and growing every day, is nothing but hocus pocus,
all too often masquerading under a cloak of (pseudo) scientific rigour.
On that topic I can do no better than quote Metalleus again (and, no
I am emphatically not the person who is hiding under that pseudonym,
but how I wish I was!):
HOW TO MAKE A LINGUISTIC THEORY*
*This manuscript was found in an empty xerox-paper box at Harvard
University. Within the history of linguistic science we believe it
dates from the early medieval period, but we do not really care much.
Assemble a judicious amount of grammar, preferably English
grammar since you're aiming at readers of English. (If you feel
there might be a market for linguistic theories written in Cebuano,
by all means, give it your best shot.) Be sure to include passive
constructions, accusative-with-infinitive constructions, and
constructions with front-shifting. Leave everything else to future
research (don't worry, you'll never have to actually do it).
Set up two levels of linguistic representation; call them
Level 1 and Level 2, or even better, Level Alpha and Level Beta.
This is to divide your explicanda into two conceptual domains so
you can let one explain the other. Leave these levels and all
constructs supporting them undefined; these will be your
Theoretical Primes. Define everything else, however, not only as
rigorously as possible but using as many symbols from the predicate
calculus as you can understand.
Be sure to leave undefined the notion "mu." Now make "mu" a
unit at both undefined levels. For each "mu" use ordinary English
spelling, but in upper case letters on one level, and in lower case
letters on the other. Use abbreviations with upper case; for
example ERG, PRO, +ITAL for "ergative," "pronominal," "borrowed
from Italian."
From this point on you need a graphics expert. Draw guitar
strings (don't call them that, of course) from units on one level
to units on the other level. Count and classify the various
arrangements of strings you need for the amount of grammar you
began with; then pronounce all other logically possible
arrangements of strings forbidden by Universal Constraints.
Give each constraint a handy name, such as "The Adjustable Bridge
Constraint," "The Open-String Pull-Off Constraint." Always
capitalize and use "the" with constraints.
At this point it will be proper, though not absolutely
necessary, to bung in a bit of data from other languages. Since
ultimately theories like yours can be constructed only by trained
linguists who speak natively the languages they are examining,
frankly, the Second Coming will be upon us well before you'll
really have to think seriously about other languages. Besides, you
have this neat argument:
Premiss 1: If my theory won't account for English,
then it won't account for all languages.
Premiss 2: My theory won't account for English.
Conclusion: Bingo.
With regard to marketing your theory, this is a cinch because
of the way the academic world works. Your theory won't work, even
for English, right? That's a foregone conclusion. But for twenty
or thirty years, other people will make such a good living patching
it up that they'll praise you as a genius even while they're
bashing the daylights out of you, since without you, where would
they be?
Make occasional references to Kuhn.
- Metalleus