¿ìè¶ÌÊÓƵ

PERHAPS IT IS because their business is words that linguists use them
with such devastating effect – often against each other. This, at least,
is how a spectator might view the current battle over how best to classify
the 600 or so American Indian languages. Do they fall into just three language
families, as one prominent linguist has recently proposed? Or are there
as many as 155 independent language families, as the opposing school argues?
This apparently arcane debate is closely tied to the question of when and
how humans first entered the Americas. Were there three separate migrations,
perhaps 12,000 years ago, as the most popular scenario has it? Or – a view
that is just beginning to gain support – did it all happen much earlier?
The idea of three language families, put forward recently by Joseph Greenberg
of Stanford University, in California, has been described variously as ‘irrelevant
nonsense’, ‘misguided and dangerous’, and ‘completely unscientific’. Lyle
Campbell, one of Greenberg’s most vocal opponents, wrote that the thesis
‘has a detrimental impact on the field’ and that it ‘should be shouted down
in order not to confuse nonspecialists or detract from the real contribution
linguistics can make to prehistory’. Greenberg’s riposte is equally blunt.
‘My critics are myopic and wedded to a technique of limited scope,’ he says
dismissively.

It is true that until recently most of Greenberg’s work has been on
linguistic problems outside of the Americas, so he could perhaps be viewed
as having ventured into unfamiliar territory. But it is also true that,
some 40 years ago, using the very same approach he is now applying to American
Indian languages, Greenberg produced a radical classification of African
languages that for the most part has stood the test of time. So, if Greenberg’s
approach worked so well in Africa, why should his conclusions about American
Indian languages not be taken seriously? Historical linguistics has a venerable
tradition. Its intellectual roots were firmly planted in the late 18th century
when the oriental scholar Sir William Jones, then a British judge in India,
recognised significant similarities between Sanskrit, Greek and Latin. These
similarities represent ‘a stronger affinity, both in the roots of verbs
and in the forms of grammar, than could have been produced by accident’,
he announced in a famous lecture in 1786. The affinity was ‘so strong that
no philosopher could examine Sanskrit, Greek and Latin without believing
them to have sprung from some common source, which, perhaps, no longer exists’.
The language family that Jones had identified has come to be known as Indo-European.

Here, then, was the core of historical linguistics: that, just as biological
species may originate from a single, common ancestor and diverge from each
other, so too may languages. The trick is to be able to identify those common
elements in two languages that indicate a common origin and to distinguish
them from elements whose resemblances have arisen for other reasons. For,
just as species might acquire the same traits independently and thus appear
deceptively to have come from a common ancestor, so it is with languages
– only more so.

For example, not only can separate languages independently invent common
elements of sound, vocabulary and grammatical structure, but they can also
borrow words from each other as a result of cultural contact or military
dominance. Ask a Frenchman, ‘Vas-tu manger le fast food ce weekend?’ and
you have two blatant examples. Many of the more subtle common features between
English and French are also borrowed as a result of extensive contact of
the past millennium, for although the two languages share a common root
in Indo-European, English belongs to the Germanic subgroup of languages
and French to the Romance subgroup.

In addition to shared features resulting from accident and borrowing,
onomatopoeic words, by their nature, tend to be rather similar in different
languages, leading the unwary to infer a common origin. These, then, are
the most obvious snares that historical linguists must avoid when a common
origin – a genetic relationship, as it is usually termed – is inferred for
two languages.

With Jones’s demonstration that many European and Eastern languages
have a common origin, scholars immersed themselves in the theory of language
evolution through time and in the practice of discerning clues to true genetic
relationships among languages. Work on Indo-European languages became a
growth industry and remains so today, making it the most extensively studied
language family in the world. There are at least a dozen language families
of similar status to Indo-European. Some scholars claim to be able to see
deeper links among these language families, eventually reaching back to
a single language, or Mother Tongue.

Out of this intensive activity has emerged the systematic approach of
comparative linguistics, in which ‘regular correspondence’ is the key tool.
Regular correspondence may involve common attributes in sounds, vocabularies
and grammar in the languages under comparison. More particularly the technique
identifies regular and consistent differences between them. For instance,
if a ‘p’ in one language is always represented by an ‘f’ in another, this
would constitute a regular sound correspondence – as in the following three
Latin/English pairs, pes/foot, piscis/fish, and pater/father. As Merritt
Ruhlen, an independent linguistic scholar from California, has pointed out:
‘Sound correspondences have come to epitomise what is good about contemporary
historical linguistics.’

Because of the painstaking nature of seeking out and proving regular
correspondences, only small numbers of languages can be tackled at a time.
Bit by bit, this technique demonstrates genetic relationships. Eventually
a higher order structure – a classification of language families – may emerge.
It is a ‘bottom up’ approach to uncovering the history of today’s 5000 or
so living languages. Because the technique is so precise, it can reconstruct
the sounds, words and the structure of sentences of extinct languages.

But the demands of this same precision also mean that the signal quickly
fades as you reach back into the linguistic past. Unlike biological species,
languages change at an astonishing rate, as anyone who has struggled with
Chaucer will attest. As a result, most historical linguists agree that going
back more than 5000 to 7000 years is a futile enterprise. ‘The chance of
finding anything deeper than 7000 years in linguistic reconstruction is
somewhere between hopeless and zero,’ is how Johanna Nichols, of the University
of California at Berkeley, puts it. Seven thousand years just happens to
be the age given to proto-Indo-European, the ancestral tongue of the Indo-European
family of languages.

So, when Greenberg came along with his classification of American Indian
languages, which spanned at least 12,000 years, it was not surprising that
he raised some eyebrows. But what shocked Greenberg’s opponents most was
the way he had reached that classification. Greenberg uses a method called
multilateral comparison, which is a broad brush approach that does not involve
regular correspondences. ‘Essentially, it involves comparing many languages
at once, identifying similarities in sounds, words, and grammar and looking
for patterns among them,’ Greenberg explains. ‘It is a very powerful technique,
precisely because it involves large numbers of languages.’ The power derives
from the equivalent of the statistics of large numbers, and, it has to be
admitted, from Greenberg’s well-honed experience and intuition.

By identifying patterns among large numbers of languages, the analysis
tends to lead to a genetic classification – nested sets of families, eventually
going back to just one in the past. It is, therefore, a ‘top-down’ approach,
the opposite of standard practice among historical linguists. Unconstrained
by the tight rules of the standard methodology, multilateral comparison
can reach much further back into prehistory, says Greenberg. How much further,
he prefers not to say. Ruhlen, who works with Greenberg, believes that there
is essentially no limit: ‘Back to the first language, the Mother Tongue,’
he asserts. That could be 100,000 years or more. Most historical linguists
who champion Greenberg’s method – and there are not many of them – are less
sanguine, saying that there must come a point when the signal disappears
completely.

Greenberg published his revolutionary proposal a little more than two
years ago, and his critics have taken every opportunity to heap opprobrium
upon him – most recently, at a conference in Boulder, Colorado. ‘Most American
linguists believe Greenberg’s method is deeply flawed,’ declared Sarah Thomasen,
of the University of Pittsburgh, Pennsylvania. ‘It cannot escape the problem
of chance similarities.’

The Greenberg classification divides the 600 or so existing American
Indian languages into three independent family groups: Eskimo-Aleut, on
the northern fringe of North America; Na-Dene, in the northwest of North
America; and Amerind, covering the remainder of North America and the whole
of South America. Each of these groups would have been established by a
migration into the New World, giving three migrations, says Greenberg. Of
the three language families, Amerind is said to be the longest established,
perhaps for some 12,000 years. This is the date at which, until very recently,
many archaeologists would place mankind’s first entry into the Americas.

At the conference in Boulder, Nichols challenged Greenberg on the question
of the age of Amerind. Basing her calculation on the rate at which Indo
European languages are assumed to have diverged, she pointed out that the
Amerind languages would have taken longer to develop than people had been
on the continent. ‘The diversity of languages within Amerind requires more
than 12,000 years to develop. More like 20,000 or 30,000 years,’ she said.
This appears to be a telling blow against Greenberg’s analysis, but, as
we shall see later, new developments in the dating of migrations into the
Americas may deflect it. Moreover, although Greenberg has put an age of
12,000 years on Amerind, he is prepared to contemplate something much older.

Greenberg, meanwhile, is gaining more supporters. ‘Greenberg’s African
classification and the essence of the method he employed have become the
standard frame of reference for the entire field of African studies throughout
the world,’ claimed Paul Newman, of Indiana University. Campbell, on the
other hand, believes that ‘the African languages that Greenberg grouped
together are simply more demonstrably related than American languages are’.
Newman denies this vigorously. ‘It is commonplace in science that empirical
clues of all types become easy to see once someone has discovered the solution
to the problem. That’s what happened here.’

The real problem is a clash of methodologies. If, as is built into the
accepted comparative method, reaching back beyond about 7000 years is simply
impossible, how can its proponents be expected to take seriously an alternative
method that does just that? Campbell, for instance, says in one breath that
he’s sympathetic to the idea that all Amerind languages go back to a single,
ancestral language, and in the next says that no methodology exists to demonstrate
it, least of all Greenberg’s. So, the argument doesn’t advance much.

The argument is going on against the backdrop of other sciences and
their interest in the peopling of the Americas, particularly archaeology
and molecular biology. In the past few months, evidence has begun to emerge
from these two disparate sources that lends some support to Greenberg: both
lines of evidence imply that the Americas were colonised earlier than most
anthropologists had come to believe.

The early archaeology of the New World is dominated by the Clovis people,
who, it was believed, exploded onto the scene a little more than 11,000
years ago, just as the last great glaciation began to retreat. Makers of
characteristic ‘fluted’ arrowheads and spear points, Clovis people have
always been thought of as accomplished big game hunters – simply because
the Clovis points were most frequently found in the midst of the remains
of the giant mammals that once lived in the Americas of the Ice Age. Modern
archaeologists also accepted that these people were ‘The First Americans’,
because most were sceptical about the validity of the many claims for sites
older than 11,000 years. But this picture is changing. Even the pre-eminence
of Clovis hunters is now in doubt.

‘We are beginning to see Clovis as a rather general hunter-gatherer
people, not the big game hunters that had often been imagined,’ said Dennis
Stanford of the Smithsonian Institution in Washington DC. The second point,
however, is crucial. ‘Some of the claims for pre-Clovis sites now look convincing,’
reported Stanford. Most prominent among these is a rock shelter at Meadowcroft,
near Pittsburgh, which really does seem to have been occupied 16,000 years
ago, as James Adavasio, the site’s principal researcher, has claimed for
years. ‘Even the most sceptical among us now accept this,’ says Stanford.

Of equal importance is Monte Verde, in south-central Chile, a site which
according to Tom Dillehay, of the University of Kentucky, was occupied 13,000
years ago. Dillehay has finally persuaded the sceptics that he has a genuine
and correctly dated site. ‘Now that the Clovis barrier has been broken,
I would expect to see more pre-11,000 year claims taken seriously,’ he says.
One of those claims will be for a very ancient occupation at Monte Verde,
dated at 33,000 years. ‘That would really put the entry time back,’ says
Stanford. ‘But it’s not out of the question.’

So, archaeology appears to offer a longer time for the Amerind languages
to diversify, if such extra time is indeed required. What help does molecular
biology provide? Ever since mitochondrial DNA (mtDNA) first made an appearance
in the world of anthropology, it has been controversial, not least in the
much publicised idea of ‘Mitochondrial Eve’, the mother of us all who lived
in Africa 150,000 years ago. If mtDNA is well suited to sorting out exactly
when and where modern humans originated, it is even more appropriate for
solving the problem of who were the First Americans.

Just imagine that a band of hunters entered the Americas as recently
as 12,000 years ago. Because they would be likely to be closely related,
the DNA in the mitochondria of the different individuals in the band would
essentially be of one type. As the millennia pass, mutations accumulate
in mtDNA, just as they do in the DNA in the nucleus, making the descendants
of the original migrants more and more genetically distinct from each other.
One key characteristic of mtDNA, however, is that the rate of accumulation
of mutations is about ten times as fast as those in the nucleus. This means
that genetic divergence will be detectable after a mere 12,000 years, the
minimum timescale that archaeologists use for the peopling of the Americas.
And, if you know the rate at which mtDNA accumulates mutations, you can
then look at the degree of genetic diversity in modern populations and calculate
how long it took for that diversity to develop: in other words, when the
founding migrants arrived. Nuclear DNA accumulates mutations too slowly
to detect divergence in such a short period with any accuracy.

As well as indicating how long people have been in the Americas, data
from mtDNA should also reveal how many groups migrated there. The reason
is that the individuals in each separate migration would carry a type of
mtDNA peculiar to that group. Mutations would accumulate within that type
in different members of the population, but the type as such would still
be identifiable. So, if you find just three basic types of mtDNA in modern
native Americans, that would indicate that the modern population had descended
from just three separate migrations. Four types, four migrations, and so
on. This makes molecular biology a potentially powerful tool, one that many
archaeologists are anxious to see put to work. Were Greenberg’s three separate
language groups each established from a single migration: one for Eskimo-Aleut,
one for Na-Dene and one for Amerind? Half a dozen laboratories around the
US are working on various aspects of this problem, including Rebecca Cann’s
at the University of Hawaii and Svante Paabo’s at Berkeley. Reporting on
the latest evidence from both labs, Cann had participants at the conference
in Boulder holding their breath.

If there was a single migration that established the Amerind language
group, then it must have occurred between 40,000 and 60,000 years ago,’
she said. ‘If there had been a series of migrations, each genetically distinct
as far as its mtDNA is concerned, then the date could be closer to 20,000
years ago.’

Even the most recent of these dates would have been sneered at by archaeologists
just a few months ago. But now that many are beginning to accept that Clovis
people were not the first colonists, these dates no longer seem so absurd.
Nevertheless, even the most ardent proponent of pre-Clovis occupation might
find 60,000 years a little too ancient, although it is feasible in terms
of when glaciation would have allowed passage from northeast Asia across
the Bering Straits into Alaska and the Yukon.

If Greenberg is correct, and Amerind is a cohesive language family,
then its founders would be expected to have shared a common mtDNA. In that
case, according to the data Cann presented, 60,000 years might just have
to be taken seriously. The alternative, multiple migrations of genetically
distinct populations from, say, 20,000 years ago, would almost certainly
have involved distinct language groups, too, in which case Amerind would
not be a genuine language family. Is there any biological evidence to suggest
that Amerind is indeed a coherent family, and that archaeologists should
be thinking in terms of 60,000 years? Christy Turner, of Arizona State University,
suggests that there is. For some time, he has been arguing that evidence
from dental anatomy indicates three population groups among American Indians,
and these coincide closely with Greenberg’s three major language groups.
In addition, Luigi Luca Cavalli-Sforza, a geneticist at Stanford, has recently
published data from a wide array of nuclear genes that also support Greenberg’s
three groups. The confluence of archaeological, genetic and linguistic data
is tantalising, yet each one has significant uncertainties that remain to
be resolved.

Meanwhile, Greenberg, approaching 76, is becoming a little impatient
with the bickering among his linguistic colleagues. I don’t want to wait
30 or 40 years, to get posthumous vindication on this one,’ he said, only
half jokingly, at Boulder. Not surprisingly, he believes his track record
in Africa should be taken seriously. Equally predictably, Campbell and other
critics charge that success in one arena does not guarantee success in another,
and what is needed is proof. For Newman, the resolution is clear: ‘If you
were forced to bet, you’d just have to bet on Greenberg. He’s been right
so often.’

* * *

Linguists have the first word

IF modern Homo sapiens originated as a small, discrete population relatively
recently in human prehistory, say, 150,000 years ago, then it is also likely
that human languages as we know them had a single, discrete origin. Known
as monogenesis, this notion implies that all modern languages are genetic
descendants of that original language, or Mother Tongue.

In theory, therefore, the world’s living languages might bear traces
of their common origin. This would permit a limited degree of reconstruction
of the long-extinct Mother Tongue. It is upon such a quest that a small
band of historical linguists is currently embarked.

‘Recovering ancestral languages spoken in the remote past will give
us the key to many central issues about the origin, development and diffusion
of peoples, cultures, and humankind itself,’ says Vitalij Shevoroshkin,
a Russian linguist at the University of Michigan, and a keen proponent of
the technique known as deep reconstruction, way back in time. The philosophy
here is that a people’s language reflects its social and practical concerns,
a mirror of what is most important to its members. So, reconstruct the language
of an extinct society, and you gain an insight into the lives of its members
that conventional archaeology cannot match. That is the hope, but how far
has it been achieved?

‘So far we have been able to reconstruct about 1000 words of Nostratic,
a proto-language that was spoken somewhere in the Near East some 15,000
years ago,’ says Shevoroshkin. Nostratic is said to be the root stock that
gave rise to half a dozen of the major language families of Eurasia, including
Indo-European, of which English is a member.

‘Computers will provide tremendous assistance in building the ‘proto-proto-language’
from which Nostratic and other equivalent major language groups derived,’
says Shevoroshkin. ‘How far back we will be able to go, I wouldn’t like
to say. But it is further than many historical linguists currently believe.’

Ancestral languages are reconstructed by comparing common elements in
words between different, related languages. These common elements then essentially
become the reconstructed word in the presumed proto-language.

‘If we compare, say, English me, Spanish me, Sanskrit -ma (where -a
originates from -e) and many other Indo-European languages, we can reconstruct
Indo European me,’ explains Shevoroshkin. ‘No Indo-European individual ever
wrote this word, but we are now confident that it was spoken around 6000
µþ°ä.’

Most reconstructions are more complex than this, and often include strange
looking notations and symbols for vowels and consonants whose sound has
been inferred. The result is a script that is unreadable, except to the
initiated. Nevertheless, starting with the most stable of words (usually
pronouns, parts of the body and fundamental components of the environment),
lexicons of proto-languages are steadily built up, in the case of Nostratic
reaching about 1000 words. Once you go from using living languages to reconstructing
the precursors of ancestral languages, your reconstruction is based on words
that are already the products of reconstruction. And the deeper in time
you go, the further you stray from any kind of known reality. It is partly
this tenuousness that makes many historical linguists cautious about going
back beyond about 7000 years.

Not so Shevoroshkin, who confidently infers aspects of Nostratic life
from the reconstructed lexicon. ‘Since there were no words for cultivated
plants in Nostratic, one may suppose that the ‘Nostrates’ gathered plants
and used them for food but did not cultivate any,’ he says. ‘Nostratic also
has no words for domesticated animals. Thus, in the Upper Old Stone Age
(around 15,000 BC), the Nostrates were apparently hunters, not cattlemen.’

Shevoroshkin also notes that Nostratic has the same word for dog and
wolf, indicating that these people ‘were beginning to domesticate wolves,
gradually transforming these wild animals into dogs, making friends of foes.’

The intellectual drive towards this kind of deep reconstruction stems
from the Soviet Union in the late 1950s and early 1960s. At this time two
linguists, Vladislav Markovic Illic-Svityc and Aaron Dolgopolsky, were working
inde pendently in Moscow on the ancestral protolanguage of Indo-European
and equivalent language families in Eurasia.

They eventually became aware of each other, and discovered a remarkable
agreement in the lexicons they had independently reconstructed. Had there
been little or no agreement, the whole business of deep reconstruction would
have collapsed at that point.

Illic-Svityc was killed in a road accident in 1966, at the age of 31.
Dolgopolsky was left to carry on the quest, and was later joined by Shevoroshkin.
The two linguists abandoned their native Russia in the mid-1970s, Dolgopolsky
went to Israel, Shevoroshkin to the US. They spread the idea of deep reconstruction
beyond the Soviet Union, but that country remained the centre of such studies.

Only in 1988 did this brand of historical linguists first come together
as an international group, at a conference organised by Shervoroshkin in
Ann Arbor, Michigan. That meeting, says Roger Wescott, of the University
of Tennessee, ‘gave us a sense of no longer being isolated’.

Ever since the turn of the century, the notion that human languages
all stem from one ancestral language – monogenesis – has been under an intel
lectual cloud. This antipathy to deep reconstruction stemmed from the inability
of linguists, using conventional techniques of comparative linguistics,
to link Indo-European to any other major language family. ‘I believe the
general rejection of attempts to connect Indo-European with other families
has effectively blocked consideration of the question of monogenesis,’ says
Merritt Ruhlen, an independent linguistic scholar in Palo Alto, California.
‘Monogenesis became almost a dirty word among traditional linguists.’

If, as many molecular biologists believe, modern Homo sapiens arose
as a discrete evolutionary event some 150,000 years ago, monogenesis of
language is likely. Recent evidence produced by Luigi Luca Cavalli-Sforza,
a geneticist at Stanford University, seems to support this idea. He and
his colleagues have uncovered an uncanny geographic coincidence between
the language families identified in the modern world and genetically discrete
populations. Such a coincidence could have come about only if early populations
diverged from a single origin.

But even if monogenesis of human languages were an historical fact,
Illic-Svityc and his intellectual descendants have so far reached back only
a short distance towards it. And there are many who believe that going all
the way will always remain a frustrating and unattainable dream.