If you look at a linguistic map of pre-Columbian North and South America, you don’t see a clean, color-coded block. You see a mosaic. From the icy coasts of the Arctic to the tip of Tierra del Fuego, the Americas were home to arguably the most diverse linguistic landscape on Earth. Linguists have spent decades effectively categorizing these into dozens of distinct language families, roughly comparable to “Indo-European” or “Sino-Tibetan.”
Then, in 1987, a Stanford linguist named Joseph Greenberg walked into this carefully curated museum of languages and tried to rearrange the entire collection.
Greenberg published a controversial book titled Language in the Americas, in which he claimed that the hundreds of distinct Native American languages actually belonged to just three mega-families. This proposal didn’t just cause a stir; it caused a scientific riot. It sparked one of the most heated debates in the history of the field—a battle known as “Lumpers vs. Splitters.”
The “Lumper” Philosophy: Greenberg’s Tripartite Scheme
Joseph Greenberg was not a fringe amateur; he was a titan in the field who had previously (and successfully) reclassified the languages of Africa. When he turned his gaze to the Americas, he applied a “lumping” philosophy. A “lumper” looks for broad similarities to group items into large categories, whereas a “splitter” focuses on differences and rigorous definitions to create precise, smaller categories.
Greenberg looked at the kaleidoscope of Native American languages and proposed they all fell into three super-groups:
- Eskimo-Aleut: Languages spoken in the extreme north (Inuit, Yupik, Aleut). This grouping was already largely accepted.
- Na-Dene: A family including Tlingit, Eyak, and the Athabaskan languages (like Navajo and Apache). This was also a recognized link, though Greenberg tried to expand it.
- Amerind: The bombshell. Greenberg claimed that every other language in North, Central, and South America—from Algonquin in Canada to Maya in Mexico and Quechua in Peru—belonged to a single, massive family he dubbed “Amerind.”
According to Greenberg, if you weren’t speaking an Eskimo or Na-Dene language, you were speaking Amerind. This hypothesis implied a single migration wave that populated almost two entire continents.
The Weapon of Choice: Mass Comparison
How did Greenberg arrive at a conclusion that contradicted decades of specialist work? He used a method called Mass Comparison (or multilateral comparison).
Standard historical linguistics relies on the Comparative Method. This is painstaking work. Linguists look for systematic sound correspondences (e.g., where Latin has a p, English consistently has an f, as in pater/father). They reconstruct proto-languages step-by-step to prove a genetic relationship.
Greenberg bypassed this. He believed the Comparative Method was too slow for deep time depths. Instead, he took huge lists of basic vocabulary (words for water, hand, mother, fire) from hundreds of different languages and laid them out in notebooks. He scanned the columns looking for resemblances.
If a word for “water” in a language in Oregon sounded vaguely like a word for “water” in a language in Brazil, Greenberg marked it as evidence of a connection. He argued that the sheer volume of “look-alikes” was statistically impossible to attribute to chance.
Why The “Splitters” Revolted
The specialists—the “splitters” who had dedicated their lives to studying specific language families like Uto-Aztecan or Iroquoian—were outraged. Their reaction wasn’t just protective territorialism; it was a defense of the scientific method in linguistics.
Here is why the linguistic community rejected the Amerind hypothesis:
1. The Problem of Chance and “Pan-Americanisms”
Human languages have a limited number of sounds. If you compare hundreds of languages, you will inevitably find words that sound synonymous by pure coincidence. For example, the word for “dog” in the Australian Aboriginal language Mbabaram is dog. There is zero historical connection to English; it is a complete accident.
Critics pointed out that Greenberg’s “Amerind” dictionary was full of these false positives. He often accepted loose definitions. If one language had tali for “tongue” and another had dala for “lick”, he might count them as a match. Once you loosen the strict criteria of meaning and sound, you can link almost any two languages on Earth.
2. Neglecting Borrowing
Languages borrow from neighbors constantly. English borrowed “taco” from Spanish, but that doesn’t make English a Romance language. In the Americas, extensive trade networks existed for thousands of years. Greenberg’s method had no way to distinguish between words that were inherited from a common ancestor 15,000 years ago and words that were simply borrowed between neighbors 500 years ago.
3. Data Errors
When specialists examined Greenberg’s data sources, they found errors. He relied on old, poorly transcribed word lists. In some cases, he listed words that didn’t exist, or mistook grammatical prefixes for parts of the root word. Lyle Campbell, a leading historical linguist, famously demonstrated that if you applied Greenberg’s loose method to Finnish and Basque (two completely unrelated European languages), you could “prove” they were related, too.
The Strangest Clue: The N/M Pronoun Pattern
Despite the overwhelming rejection of his methods, Greenberg did highlight one phenomenon that continues to intrigue linguists today: the “Amerind” pronoun pattern.
In a surprising number of Native American languages, the first-person pronoun (I/me) contains an ‘n’ sound, and the second-person pronoun (you) contains an ‘m’ sound. For example:
- Kintu (South America): ni (I) / mi (you)
- Mixe (Mexico): n- (I) / m- (you)
Greenberg argued this n/m pattern was a genetic marker—a linguistic fingerprint left by the original ancestral language. While mainstream linguists acknowledge this pattern is statistically higher in the Americas than chance would suggest, they argue it isn’t enough to prove a single family. It could be the result of ancient borrowing, or perhaps specific sound symbolism that is common to human psychology, much like how “mama” means mother in unrelated languages globally.
The Verdict: Where Do We Stand?
So, who won the war of Lumpers vs. Splitters? In the field of linguistics, the Splitters won the battle, but the war continues on different fronts.
Today, the vast majority of historical linguists reject “Amerind” as a valid language family. The evidence simply does not meet the standards of the Comparative Method. The consensus is that the Americas were likely populated by multiple waves of migration, or that the time depth is so great (15,000+ years) that linguistic methods cannot recover the original connections. After about 8,000 to 10,000 years, language changes so much that genetic links become invisible.
However, modern genetics has added a twist. DNA studies suggests that the indigenous populations of the Americas do descend largely from a small founding population from Siberia (with distinct separate waves for the Na-Dene and Eskimo-Aleut ancestors). While this genetic finding aligns eerily well with Greenberg’s three groups, genes are not languages. A population can share DNA but speak unrelated languages (think of Hungarians and their Austrian neighbors), or swap languages entirely.
Conclusion
The Greenberg controversy serves as a vital lesson for language lovers. It highlights the tension between the desire to see the “big picture” (Lumping) and the demand for rigorous, distinct proof (Splitting). While Greenberg’s “Amerind” failed the test of science, his audacity forced linguists to look closer, work harder, and document the incredibly rich, complex tapestry of the Americas with renewed vigor.
In the end, whether there was one original tongue or dozens, the diversity of Native American languages remains one of humanity’s greatest intellectual treasures—one that defies easy categorization.