Decades of wear and tear haven’t been kind to the 2,713 wax cylinders in UC Berkeley’s Phoebe Hearst Museum of Anthropology, which linguists and anthropologists have used for over a century to study the languages and cultural practices of Native California. But a new project promises to revitalize these old, fragile recordings — the first of which was recorded by famed anthropologist Alfred Kroeber in 1901 — with cutting-edge optical scanning technology.
The prospect excites scholars and Native communities because the collection itself is a trove of Native American recordings, particularly from California — ceremonies, songs and traditional stories that capture slices of cultures that are in many cases endangered or extinct. Among its best known is Ishi’s retelling of the Story of Wood Duck, the only recording of the extinct Yahi language. Ishi was recorded between 1911 and 1914 by Berkeley anthropologist T.T. Waterman, who began translating the story but didn’t finish because the fuzzy sound quality made the words too difficult to discern.
“This is a record of native Californian speech and mythology and worldview that was very important to Ishi, the only recording of this that we know of,” says Hearst anthropologist Ira Jacknis, “and it’s still partially untranslated.” So Jacknis is excited about the potential of a remastered recording: With it, linguists can finally translate the whole thing. “It would let us get so much closer to what Ishi was saying,” he says.
Some of the other recordings are moldy, or have been worn down by repeated plays. They have been backed up before, with duplicates created on reel to reel tapes, cassette tapes, and most recently, on CD. But that iterative process made sound quality worse and worse. “Our recordings are really very difficult to listen to,” says Jacknis. “There’s a lot of distortion.”
The new technique, developed by Berkeley Lab physicist Carl Haber, goes back to the sound’s source: It takes high-res images of the wax cylinders’ ridges, which the scientists can then edit to get rid of the accumulated mold, dirt, and grime and create an audio file without ambient noise. Plus, the method doesn’t touch the delicate cylinders at all, capturing the data encoded in the ridges without destroying them. Haber, who earned a MacArthur “genius” grant for his work in 2013, has used this technique to preserve historic recordings all over the country.
The new Berkeley project is his largest undertaking yet.
The recordings of other groups will come into focus, too. “Oftentimes the cultural practices that were recorded are no longer known well or even known at all in the communities,” says Andrew Garrett, a Berkeley linguist and one of the project’s heads. “So these recordings are a way for communities to reconnect with those aspects of their cultural history.” Garrett sees it as a sort of cultural repatriation, part of a larger effort to restore these artifacts to their rightful owners.
Native communities have already used the collection to relearn songs and the languages of their great-grandparents, but clearer recordings would help those efforts along. “The structure of the language or the vocabulary or the way words and phrases are used — all of that is valuable for people who are trying to speak the language again,” Garrett says. So a member of, say, the Hupa tribe could use these recordings to piece together how Hupa was spoken back when the language was thriving.
Sometimes this experience is incredibly personal, as Garrett remembers from a conference he attended. “The great-granddaughter of someone who had been recorded by Kroeber in 1901 or 1902 had overheard her great-grandmother’s voice,” he says, “and she was so moved to be able to hear this and to try to learn what she had said.”
Higher-quality recordings would also be a boon for linguists, because they would capture hard-to-detect nuances in speech. As a specialist in Yurok, an endangered northern Californian language, Garrett is interested in tracking how the language has changed over time. “In the case of Yurok,” he says, “you have this window between the most recently recorded speakers in the 21st century telling stories, and speakers in the 1980s and the 1950s who were recorded by other linguists. Those speakers had in some cases grown up before white contact, so these recordings give you a window back to the first decade of the 20th century.” Garrett can compare the differences in the language over time using pronunciation, patterns like older verb forms, or certain kinds of stylistic elaboration to try and determine whether these differences stem from the natural evolution of language or the introduction of English in the 20th century.
But before anyone can benefit from these newly improved recordings, the cylinders need to be scanned, a time-consuming process. Currently, Haber and his team are constructing the machine, ordering equipment from Europe, and crafting custom parts. After the machine is built and calibrated in the next month, it’ll begin scanning the 2,713 cylinders day and night, working nearly nonstop for the next three years.
“It’s a great Berkeley story,” Jacknis says. The project would be impossible without Haber’s invention and technical expertise, without the linguistics department’s strength in language documentation and connections with Native communities, and without the early efforts of anthropologists like Kroeber. Haber is also enthused about the opportunities it presents to his students, who currently write software for the machine to collect data from the cylinders. “It’s an interesting problem from a technical perspective,” he says, as well as a chance for students in fields like applied math and physics to work on projects with immense cultural significance.
This machine is only the sixth of its kind in the world, Haber says — the first is at Lawrence Berkeley Lab, where it was initially constructed, and others are at the Library of Congress, in India, and at the Northeast Document Conservation Center in Massachusetts. After the new machine finishes scanning the entire Hearst collection, it’ll stay at the library, ready to render historic recordings from other collections into files that can be studied, shared, and played over and over again.