Science Notes Logo more Science Notes Logo
Contents Robofly Killer Surf Tracking the Bloom This won’t hurt a bit Echoes from the Past A Ride on the Wild Side Tongue Twister KC and the Ground Sludge Band Twinkle, Twinkle, Collapsing Star One If By Land, One if by Sea Text-only Version Science Notes Home more Science Notes Logo more Science Notes Logo Echoes
from the Past

A maverick linguist has devised a new way to scope out what our ancestors were up to 50,000 years ago.

By: Bob Adler
Illustrations by: Simon Lo

Splashy illustration

Language’s Hidden Voice

What people say tells us a lot about them. By the time someone says a dozen words, we know whether they’re young or old, with-it or out-of-it, from halfway around the block or halfway around the world. Language doesn’t just say what we want it to—it tattles about our history, whispers about where we come from and where we’ve been.

That hidden, other voice of language captivated Johanna Nichols and has dominated her life for the past 35 years. Nichols is a linguist—a scientist who studies language and languages. She’s trekked to the far reaches of the former Soviet Union—to Chechnya, Ingushetia and Makhachkala—to describe and preserve native languages. These days, however, she spends her time in a quiet, sixth-floor office with a panoramic view of the U. C. Berkeley campus. Though she often gazes out the window, Nichols is not looking at the towering eucalyptus trees, the rolling green lawns, or the students hurrying by. She is preoccupied with faraway places and ancient times. Voices from hundreds of languages are clamoring for her attention. What can languages today, she wonders, tell us about a great wave of exploration and migration that began 50,000 years ago and eventually circled the Pacific Ocean? What do they reveal about the earliest seafaring people and about who first discovered and populated the New World? Does the babel of modern tongues hide a linguistic clock that can date the birth of human language itself?

A Linguistic Time Machine

To answer these questions, Nichols has spent the past 15 years creating a kind of time machine—a unique new way of using languages to listen to the fading echoes of human events thousands of years in the past. Just as a radio telescope lets astronomers catch and decode faint signals from far away and long ago, the novel, and controversial, method that Nichols has developed lets her detect signals from the past hidden within the languages of the present. Through those messages Nichols believes she can chart the migrations of our ancestors tens of thousands of years ago.

To use language to discern the shape of such ancient events, Nichols had to break out of a mindset that still holds most other linguists back. She set aside the successful, tried-and-true study of language family trees, and developed a new approach that focuses on identifying and mapping unique grammatical building blocks, even when the languages that carry those building blocks are not related. She bases her studies on similarities between languages that other linguists have refused to interpret because the patterns did not make sense to them. Just as the infamous ozone hole lurked undetected for years because scientists had pre-programmed their computers to reject readings outside the range they expected, language patterns the size of the Pacific Ocean remained invisible until Nichols discovered them, took them seriously, and began to ask what they meant.

Family Trees

For nearly two centuries, linguists have painstakingly worked out family trees for languages, and they’ve succeeded remarkably well. As if slowly assembling a gigantic puzzle, linguists have sorted out most of the 6,000 languages spoken today, along with many that have died out. When two languages share enough similarities, linguists can be sure that they are related. For example, French, Spanish, Italian and Portuguese are sister languages, children of the same famous parent, Latin. Together they form a language family. English and Dutch are siblings in another family, sharing Low German as their parent. Even when an ancestral language has disappeared, linguists can recreate much of its vocabulary and grammar from its offspring.

Linguists have even found relationships between the remote ancestors of many of today’s languages. For example, English and Dutch share an extinct great-grandparent, Germanic, with Swedish, Danish and Norwegian. Often, a whole family tree turns out to be a branch of an even bigger, and older, tree. Ancient Greek, Latin, and seven other language families including our own, are all parts of a very large, very old family called Indo-European.

But linguists constructing language family trees eventually hit a wall. They could trace many language families back about 6,000 years, and a very few close to 10,000 years. Beyond that blurry horizon, random changes in words swamp any genuine similarities. “Everything changes over time in languages,” says Nichols, “and even the most durable signs of similarity eventually fade out.”

Rather than discovering one great family tree with all the world’s languages and language families on it, linguists found themselves wandering in a forest. They had identified 200 to 300 separate language families, called stocks. Some stocks were big and bushy, bearing dozens of related languages, while some were skinny, with just a few branches. Some, like Basque, a language with no known relatives that is spoken only in the Pyrenees of France and Spain, stood alone. But few stocks could be traced back much beyond 6,000 years, and none of them was provably related to any other.

“A Stroke of Genius”

While most linguists continue to push that 6,000-year wall back a bit at a time, Nichols vaulted over it. She stopped trying to refine or connect family trees. Instead she focused on certain language features—she calls them grammatical building blocks—that allow her to ask questions about those 200-plus deeply rooted language stocks without trying to pin down how they might be related. “Until Nichols came along, historical linguists had talked themselves into a straitjacket, in which they saw their major enterprise to be the reconstruction of proto-languages,” says John Moore, an anthropologist at the University of Florida in Gainesville. “She loosened that up to show that there are other important projects to be accomplished. I’m trying to be restrained, but I think it was a stroke of genius.”

Nichols made her next advance when she selected a representative language from each stock and plotted those that used particular building blocks on a map of the world. To her surprise, striking geographic patterns appeared; groups of languages thousands of miles apart contained clusters of identical building blocks. While other linguists attributed such shared features to chance, Nichols puzzled over them. She became convinced that many of the patterns that appeared on her maps could not be explained by chance. Nichols has spent a decade asking just what these patterns mean.

Building Blocks

Today, Nichols uses several dozen grammatical features to sort languages into different types. Many of the grammatical building blocks she uses are familiar—for example, whether a language favors prefixes or suffixes, whether it puts verbs at the beginning, middle or end of a sentence, and how it indicates possession. For the most part, she uses grammatical features rather than words or sounds because, as the skeleton of a language, grammar tends to change more slowly than words.

Still, Nichols has included several useful sound features in her set of linguistic building blocks. She classifies languages according to how many of their pronouns—words like “I,” “you,” “he,” or “she" —start with an “m” or “n” sound. She also notes if a language uses tones, as in music, to change word meanings. Chinese, Thai and Navajo do this, to name a few.

One important grammatical feature that Nichols considers is whether a language uses “numeral classifiers,” a feature English lacks. In English and most related languages, we can link a number and what it’s counting without any frills. “Six beagles,” we say, or “nine daisies.” But many languages require a kind of fastener to join the number and the object together. Most Asian languages, including Mandarin Chinese, Japanese and Korean, require this verbal Velcro. In them, you’d say something like “seven-classifier-duck” or “three-classifier-drum.” These special joining words are called classifiers because they specify the shape of the thing being counted. To translate completely, you’d have to say something like “five long-skinny-classifier pencil” or “a dozen round-classifier balloon.” And because things come in many shapes, languages that need numeral classifiers use lots of them. Yurok, spoken by a Native American tribe, has 15, Korean 26, and Mandarin Chinese an impressive 51.

It was Nichols’ great idea to make grammatical building blocks the focus of her study, rather than languages and their degree of relatedness. By giving up trying to figure out just how languages are related, she freed herself to start asking other, more important questions, such as where the different linguistic markers were distributed around the globe, and how they might have gotten there.

The Pacific Rim “Necklace”

In Nichols’ mind, the picture is clear. An enormous and sustained wave of human migration started about 50,000 years ago somewhere in Southeast Asia. Over thousands of years, successive bands of people spread out from the region. They could move relatively quickly because they were coastally adapted—they knew how to make simple boats and make a living from the sea. Over thousands of years, some carried their languages south and west through coastal New Guinea and into northern Australia, while others moved clockwise up the coast of Asia, across the Bering Strait into Alaska, then down the west coast of North and South America.

The evidence for this slow-motion human tsunami appears on Nichols’ world maps, where strands of languages sharing particular features ring the Pacific Ocean. Languages that start their pronouns with “m” and “n” sounds, languages that put their verbs first, and languages that use numeral classifiers form a pattern that circles the Pacific Ocean like a necklace. Languages sharing these features dot the islands of New Guinea, bead the coast of Asia from the southeast to the northwest, and trail the length of the Pacific coast of North and South America.

Map of research languages
Languages using numeral classifiers—chance distribution or echoes of a great migration?

“Numeral classifiers are endemic, ubiquitous, frequent and striking in the languages of Asia—Chinese or Japanese or Korean or Thai,” Nichols said at a recent scientific meeting. “They are not infrequent in Melanesia and New Guinea. And they’re found up and down the West Coast of the Americas, and nowhere else. This is one feature that genuinely seems to be found nowhere else on earth but in these areas.”

Nichols uses standard statistical techniques to calculate the probability that this necklace around the Pacific might show up on her maps by chance. The odds are vanishingly small that so many of the languages with these key features would cling to the Pacific Rim, while so few appear in the vast areas of Asia, Africa, and inland America.

Nichols is frequently asked how the languages that form the Pacific Rim necklace can share these grammatical building blocks if they are not related to each other. She explains that the stocks they represent may have originated in the same geographic region; neighboring but unrelated languages often share a significant number of traits. Or, as groups of people interacted over time, they may have borrowed language features from one another or from cultures that had arrived earlier, a sort of cross-fertilization. Language stocks with several identical grammatical markers clearly share some “ancient affinities,” Nichols says, but it’s not possible, or necessary, to figure out just what those affinities are.

Intriguingly, evolutionary biologists have recently discovered genetic evidence suggesting that Nichols has uncovered something more substantial than mere linguistic echoes. Ted Schurr, part of a team of geneticists at Emory University in Atlanta, has spent years comparing mitochondrial DNA, a kind of genetic material that is passed only from mother to child, from groups of people around the world. He discovered a genetic marker that shows up in approximately the same Pacific Rim pattern Nichols found.

Nichols interprets this match-up between her findings and genetics cautiously. She simply suggests that the genetic mutation Schurr found may have started in the same gene pool as the Pacific Rim languages. Recent research in China also supports her conclusions. Genetic studies there indicate that after leaving Africa, early modern humans colonized Asia’s southern coast before they spread north.

The Earliest Americans

The Pacific Rim pattern that Nichols discovered leads her to think that some of the earliest immigrants to the New World were seafaring people who paddled from island to island in the Aleutians and hugged the Pacific coast of Alaska. Since the earliest parts of this migration took place during the Ice Age, the groups that made it to the New World must have known how to make a living from the sea and protect themselves from the cold. They may have found shelter in refugia, coastal areas that archaeologists think may have remained ice-free throughout the Ice Age.

The waxing and waning ice sheets also play a role in Nichols’ dating of the colonization of the New World. For 60 years, archaeologists believed that pioneers from Asia first crossed the land bridge that linked Siberia and Alaska no more than 11,500 years ago, toward the end of the Ice Age. These intrepid hunters, it was thought, swept from the edge of the ice sheets to the tip of South America in search of mammoth and bison. Their fluted stone projectile points, first found near Clovis, New Mexico, turn up throughout North and South America. The consistent age of these so-called Clovis sites, around 11,000 years, as well as the absence of convincing evidence for any earlier inhabitants of the New World, led archaeologists to conclude that these hunters were the first Americans. The Clovis worldview solidified into an entrenched archaeological edifice. Generations of students memorized the mantra “Clovis first.”

This view changed dramatically in January 1997. That’s when Tom Dillehay of the University of Kentucky in Lexington invited nine of his archaeological peers to Monte Verde, Chile, to examine a site he had been excavating 20 years. The “Clovis Police,” as they were nicknamed, came, saw, and were convinced. The tent stakes, digging sticks and footprints that Dillehay had found proved, for the first time, that people lived in the New World 12,500 years ago, a thousand years before the dawn of the Clovis culture.

Nichols was not surprised. She had previously developed two different language-based dating methods that led her to believe, and suggest to archaeologists, that people had entered the Americas many thousands of years before the Clovis time line.

The Language Clock

Nichols based one of her dates for the discovery of the Americas on a striking linguistic feature of the New World, the vast diversity of Native American language families.  Native American languages comprise close to 150 independent stocks, half of all the world’s language families.

Nichols studied all known language stocks to determine how often they have branched off to create new languages.  Like human families, some language stocks—Latin for example—produce lots of offspring.  But most languages spawn just a few, some survive, but sire no offspring, and some lines die out entirely.  Nichols found that, on average, one-and-a-half new languages developed per stock every six thousand years.  In effect, Nichols created a kind of linguistic clock, ticking once every 6,000 years.

Based on her calculations, it would have taken 20,000 to 30,000 years, at the very least, for even multiple waves of prehistoric immigrants to produce the abundance of languages found in North and South America.

David Meltzer, an anthropologist at Southern Methodist University, lauds Nichols for advancing this finding.  “When most linguists were arguing for a short time span that would fit with a Clovis chronology,” he says, “Nichols was arguing that the linguistic evidence suggested much greater antiquity, from the sheer diversity of language families.”

Nichols based her second estimate not on the linguistic birthrate, but on how language families have moved across the globe. She studied historical and archaeological records to determine the rates at which languages have spread across different kinds of geographic areas. She used those rates to estimate how long it would have taken an expanding family of languages to cross the unpopulated regions from the edge of the great ice sheets to the sites of early human habitation in the Americas. She calculated, for example, that it would take at least 7,000 years for a language to become established deep in South America. With Monte Verde’s 12,500-year-old artifacts in mind, Nichols reasoned that immigrants must have entered the New World at least 7,000 years earlier, or 19,500 years ago.

But once again the Ice Age figures in.  “That was the very height of glaciation,” she says, “when it was probably impossible to get in.  That suffices to tell us that people got in before the very height of the glaciation, certainly before 22- or 24,000 years ago.”

Nicholas Evans of the University of Melbourne, however, is not convinced that Nichols’ linguistic clock keeps good time.  His research on Australian languages suggests that the clock has ticked more slowly there.  Australia has been populated for perhaps 50,000 years, but has far fewer language families than the Americas.  He suspects that something turned the clock’s hands faster in the Americas, explaining its large number of language families without having to assume immigration before the peak of the Ice Age.

Still, studies of human genes appear to support Nichols’ 24,000-year timeline.  Douglas Wallace, one of the pioneers of human genetics, estimated that humans first crossed from Siberia into America 20,000 to 40,000 years ago.  More recently, Antonio Torroni, a colleague of Ted Schurr at Emory University, used mitochondrial DNA to narrow that range to 22,000 to 29,000 years ago.

The First Four Discoveries of America

Nichols believes that she can detect the linguistic echoes of four distinct “discoveries” of the Americas through detailed mapping of language features. The very first group, she thinks, arrived more than 22,000 years ago, and spread throughout the habitable regions of North and South America during the Ice Age. Next, as the ice sheets melted, groups from South and Central America, perhaps carrying the Clovis toolkit, moved north to repopulate North America. They eventually met and interacted with new coastal immigrants working their way down the Pacific Rim and gradually moving inland. And finally, about 5,000 years ago, the Eskimo-Aleut peoples, with known Siberian roots, entered and occupied the arctic and subarctic regions.

The Birth of Language

Nichols has also consulted her linguistic clock to address the question of the birth of human language. The great diversity among the world’s language stocks, she found, can be accounted for only by pushing the dawn of language back 100,000 to 132,000 years. Even then, it is necessary to assume that language must have sprung up at about the same time in 10 or more human groups spread out across the East African cradle of modern humanity. Nichols’ dates fall neatly between the estimates of some biologists who believe that language evolved gradually over several hundred thousand years and some archaeologists who think that fully modern language flowered along with cave paintings, sculpture and other symbolic activities as recently as 50,000 years ago.

Conflict and Controversy

To say that Nichols is bold to use linguistic tools to attempt to recreate events 100,000 years ago is an enormous understatement. No one doubts that she is a trailblazer. She is the first, and so far the only, linguist to use languages to go so far back in time. But like most pioneers, she’s drawn plenty of criticism. Lyle Campbell, a highly respected linguist at New Zealand’s University of Canterbury, writes, “While in other areas I think she is one of the smartest and most independent and astute of living linguists, in this area I think she is very, very, very wrong.” Although most of Campbell’s criticisms are highly technical, his conclusion is clear—that the geographic patterns Nichols has identified are either accidental or the product of “parallel but unconnected development.” They do not, he asserts, support historical conclusions about the peopling of the Americas or elsewhere.

Evans, the Australian linguist, is somewhat less critical, though he too is not convinced that the grammatical building blocks Nichols has identified are solid enough to support her conclusions. “There is something out there that requires explanation,” he says. “I just don’t accept the explanation that she is giving. It’s a very interesting area, but it’s going too far too fast. The interpretations have gotten ahead of the data.”

Others, however, support Nichols’ approach. “One problem in evaluating her work is that she is the only one using her methodology,” says University of Florida anthropologist John Moore. “Most scholars are hopeful that she’s right, that the envelope can be pushed back thousands of years by using the methodology she’s developed. The problem is to have her work corroborated by other scientists and by data from other parts of the world.”

That corroboration may have to wait, since Nichols remains the only linguist generating these kinds of data. “This is a territory that other respectable, competent linguists haven’t dared to explore,” says Victor Golla, a linguist who works with Native American languages. “So she’s out on a limb, exploring territory that not very many are willing to follow her into.”

As for Nichols, she’s turned her linguistic radio telescope back to the Americas. At this year’s meeting of the American Association of Physical Anthropology, Nichols described a new geographic pattern that sheds light on the second-oldest cluster of languages in the Americas and reinforces her views on the antiquity of the first Americans.

What keeps Nichols going, out on that limb, still alone, and often in the face of harsh criticism?

"We can trace language prehistory back extremely far," she says, choosing her words carefully. “I can get you times, places and directions of movement, answering the same kind of questions geneticists and archaeologists ask about the origins and migrations of people. That’s very interesting to me.”


WRITER Robert Adler
B.S., physics and mathematics, University of New Mexico; Ph.D., psychology, University of California, Berkeley.
Internship: New Scientist, London office.
B.A., art, UCSC, 1999.

Back to top | Contact Information | Science Notes Home

Text © 1999 Bob Adler
Illustrations © 1999 Simon Lo