The Uto-Aztecan tree

Proto-Uto-Aztecan is usually dated to around five thousand years ago, though estimates based on the amount of divergence are rough. Where it was spoken is one of the most argued homeland questions in the Americas, because the answer decides whether the family spread with farming or without it.

The older view looks north. Catherine Fowler's study of reconstructed plant and animal names in 1983 found words for plants and animals of upland and desert country, such as pines and several desert plants, and she placed the homeland in the region between southern California, the Great Basin and the Southwest. On this view the ancestors were foragers, and the southern branches moved down the western side of Mexico later, taking up maize farming as they went or on arrival.

In 2001 Jane Hill turned the picture round. She argued that a set of words for maize, grinding and cultivation could be reconstructed for the whole family, and that Proto-Uto-Aztecan speakers were therefore early cultivators in western or central Mexico. Their languages would then have spread north with maize, which reached the American Southwest by about 2000 BCE, in the way that farming is thought to have carried Bantu or Austronesian languages.

The linguistic evidence has been contested on its own terms. In 2009 William Merrill and colleagues reviewed the maize vocabulary and argued that it can be securely reconstructed only for Southern Uto-Aztecan, not for the family as a whole, which fits a northern homeland better. Archaeology dates the arrival of maize in the Southwest well enough, but cannot by itself say what language its first farmers spoke, and genetic studies of living populations have not produced a clear answer either. Most specialists now treat both scenarios as open.

Proto-Uto-Aztecan had a modest sound system: five vowels, including a high central vowel written *ï, and consonants with no contrast between voiced and voiceless stops. Much of the diversity of the modern languages comes from what happened to a few of these sounds. *p stayed in Hopi and Numic, softened to b in Yaqui and v in O'odham, and vanished at the start of words in Nahuatl. *y became d in O'odham. Numic languages add their own twist with consonant gradation and whispered, voiceless vowels at the ends of words.

The grammar is built mostly with suffixes and postpositions. Langacker reconstructed a verb-final word order for the proto-language, and Hopi, the Numic languages and Yaqui still put the verb last. O'odham and Nahuatl have moved away from this: O'odham word order is notably free, and Nahuatl tends to put the verb early. Many nouns carried a suffix, reconstructed as *-ta, that marked the plain, unpossessed form. In Nahuatl it survives as the familiar -tl, -tli or -li: calli 'house' but nocal 'my house'.

Two other habits recur across the family. Plurals are often made by repeating the first syllable, as in Classical Nahuatl coyōtl 'coyote', plural cōcoyoh. And several languages build nouns into the verb: Nahuatl nitlacua 'I eat (something)', with a general object prefix, beside ninacacua 'I eat meat', with naca- 'meat' incorporated.

None of the languages had a full writing system before Europeans arrived. The Aztecs used a system of pictures and signs that could spell out names by the rebus principle, but not whole sentences. Nahuatl was written in the Latin alphabet within a few years of the conquest, and practical alphabets for Hopi, O'odham, Yaqui and the Numic languages came only in the 20th century.

That the family exists is not in doubt. How its branches fit together is less certain. Most classifications recognise eight or so well-defined groups: Numic, Takic, Hopi and Tübatulabal in the north, and Tepiman, Taracahitan, Corachol and Aztecan in the south. Above that level, agreement fades. A Southern Uto-Aztecan branch is supported by shared words and sound changes and is widely accepted. A Northern Uto-Aztecan branch is more doubtful. Wick Miller's word counts grouped the northern languages together, but Lyle Campbell, Ronald Langacker and others have pointed out that shared features among neighbours can come from centuries of contact as well as from a common ancestor.

Within the south, Taracahitan may not be a single branch at all, and some scholars join Corachol (Cora and Huichol) with Aztecan in a Corachol-Aztecan group. The internal diversity of Nahuatl, which is a dialect continuum large enough to be counted as several languages, raises its own questions of classification.

The dating debate is tied to the homeland debate. A proto-language of the Mexican cultivators in Hill's model cannot be much older than the domestication of maize, while a northern forager homeland imposes no such limit. Neither camp can yet point to a decisive date.

The search for relatives beyond the family has been less successful. In 1937 Benjamin Whorf and George Trager proposed that Uto-Aztecan was related to the Kiowa-Tanoan languages of the Plains and the Rio Grande pueblos, forming an Aztec-Tanoan stock, and Sapir included it in his larger proposals of the 1920s. A few suggestive resemblances have been noted, but no regular sound correspondences have been demonstrated, and most specialists regard Aztec-Tanoan as unproven. Uto-Aztecan is therefore treated as an independent family.