Uto-Aztecan Language Research

PETER WHITELEY (Curator, Division of Anthropology): The Hopi account of their origins involves the migration in of multiple clans, from all directions. Most clans are associated with emergence from the world below. There’s a whole group of other clans, however, who migrated up from the South.

The Hopi language is- is one of a group of languages that we call Uto-Aztecan. Linguists have had their own ideas many ideas about where Uto-Aztecan languages arose.

But the exact nature of the relationships and what story it tells historically—that has been argued over ceaselessly for at least half a century. 

[SHELF LIFE TITLE SEQUENCE]

WHITELEY: My name is Peter Whiteley and I’m the curator of North American ethnology. 

WARD WHEELER (Curator-in-Charge, Science Computing Facility and Curator, Division of Invertebrate Zoology): My name is Ward Wheeler. I’m Curator of Invertebrate Zoology and Curator-In-Charge of computational sciences.

I’ve always done molecular systematics, but I always had a hankering to work on languages.

If we can think of words as sound sequences, each word is sort of like a gene.

WHITELEY: I became fascinated by the idea of this as a means for explaining evolution of languages.

The Uto-Aztecan languages stretch geographically from Idaho, all the way down to Panama, and they encompass a large number of different types of culture. All the way from very small-scale hunter-gatherer bands, like the Southern Paiute, to the Aztec society, which was a full-fledged state. 

There have been multiple hypotheses about where Uto-Aztecan languages originate. There are some in recent years // that have suggested a southern origin. 

That’s what we generally refer to as the farmer hypothesis. Languages originated in an area close to the origins of the domestication of maize, beans, and squash. And then radiated out from that. 

Surely if it’s to be a knowledge that we would classify as scientific, there has to be a way of testing that.

WHEELER: Since I’ve been working so long with DNA sequences it was very easy for me to think about how we might generalize this to linguistic data.

So, instead of A-C-G-T, you have, you know, all the sounds people can make.   

In the same way that we would look at variations in DNA sequences when we want to build a tree of organisms, we would look at variations in the sound sequences to build a tree of languages. 

We took the existing algorithms that we had developed for DNA sequence analysis and were able to generalize them so that we could then use them for sound sequences—words. 

In the same way that we need to use homologous genes when we compare organisms, so we want to compare apples to apples, we want to use homologous words, also. And the set of homologous words we use is called the Swadesh list.

WHITELEY: In the 1950s, Morris Swadesh, who was a very smart linguist determined on a set of 100 words that seemed to be most resistant to change by influence from other languages.

In English, this goes louse, two, water, ear, die, I, liver, eye, hand, hear, tree.

In Orayvi Hopi: [speaks Hopi]

So, we used the Swadesh List from 37 Uto-Aztecan languages. The words in the 100- word list came from missionary accounts, 19^th century grammars and dictionaries, all the way on up to contemporary recording of languages that are still actively spoken. 

All of the original sources had different ways of writing those languages. So, how do you compare them rigorously? You need to render them all into a single format—the International Phonetic Alphabet.

WHEELER: To understand the relationships among the Uto-Aztecan languages, we look to variations in the sound sequences in the words and from that, we can reconstruct evolutionary trees of the languages themselves.

We have to assign values to transformations between different sorts of sounds.

WHITELEY: The Hopi word for dog – [Hopi word]. In Shoshone— [Shoshone word]. And in another northern language, is something like [Numic word].

WHEELER: What is the cost of editing one word into another? Because that’s how you say one tree is better or worse than another tree.

You want to summarize all of those edits that transform all of those sequences into all those other sequences on a tree. And you add that up. That’s the cost of the tree.

So, when we have all the trees, which imply different sets of changes, we try to choose the lowest cost, the best interpretive framework that helps us reconstruct past events.

It’s among the most complex problems in computer science.

WHITELEY: The results of the analysis are confirming some perspectives that have existed for a long time, and challenging others.

We took this tree—the most likely relationships among the languages that we were working with—and then mapped it onto the landscape.  It does tend to support the hypothesis for a southern origin. 

WHEELER: We know things were happening first in the South. And then things were happening secondly in the North. And so, that meant there had to be a linguistic migration, moving north.

Plus, we developed a technique to analyze languages in a way that we feel is more objective and that’s a cool thing, too.