Four years after Frank Seifart started documenting endangered dialects in Colombia, the guerillas came. In 2004, soldiers from the Revolutionary Armed Forces of Colombia swept past the Amazonian village where he did most of his fieldwork. The linguist reluctantly left for another village, south of the Peruvian border.
When he got there, the chief was away. In the central roundhouse, an old man beat out a rhythm on two enormous drums: “A stranger has arrived. Come home.” And the chief did. It was the first time Seifart, now at the University of Cologne and the French National Center for Scientific Research in Lyon, had heard the traditional drums not just making music, but sending a message.
Now, he and his colleagues have published the first in-depth study of how the drummers do it: Tiny variations in the time between beats match how words in the spoken language are vocalized. The finding, reported today in Royal Society Open Science, reveals how the group known as the Bora can create complex drummed messages. It may also help explain how the rest of us “get” what others are saying at loud cocktail parties, by detecting those tiny variations in time even when other sounds are drowned out.
“It is quite innovative,” says descriptive linguist Katarzyna Wojtylak, a postdoctoral research fellow at James Cook University in Cairns, Australia, who has studied the language and drumming systems of the Witoto, a related group. “Nobody has ever done such an extensive and detailed analysis of rhythm in a drummed language.”
For centuries, people from the forests of West Africa, Asia, and the Amazon have been using drums to send long-distance messages. The instruments’ rumbles can travel up to 10 times farther than singing or shouting because of their low pitch and high volume. But most drummed modes aren’t simple signals or Morse code–style sequences; instead, they are reduced versions of spoken languages, without consonants or vowels but with enough connection to the original language that speakers can reliably interpret what they mean. In that sense, drummed speech is similar to speaking underwater or through the static of a bad phone line.
What carries the message? One answer is tone. All but one of the 20 or so drummed speech systems come from tonal languages, including Yoruba in Nigeria, Banda-Linda in the Central African Republic, and Chin in Myanmar. Spoken Bora has two tones, which are recreated using two different drums made from hollowed logs, called manguaré. The thinner “male” has a higher tone, and the thicker “female” has a lower one.
But tone alone isn’t enough to distinguish all the words a drummer might want to say. So Seifart and his colleagues looked at what he calls a “neglected” quality in linguistics—rhythm. The team started by collecting recordings of drummed Bora. It wasn’t easy. One hundred years ago, nearly every man in every village would have known how to use the drums. Now, such skills are rare, and only about 20 manguaré drums are in active use in the Bora region.
Seifart asked five drummers to play scores of messages, including the waking up message, the message of return, and the message that went out like clockwork every afternoon asking for coca leaves for the chief.
As predicted, the tones of the 169 drummed messages matched the high and low tones of spoken Bora. Words appeared in a formulaic order, and nouns and verbs were always followed by a special marker. Each message ended with the line, “Now don’t say that I am a liar.” Roughly translated, Seifart says, it’s like a mother warning her child: “Don’t say you didn’t hear me!”
When the team compared the drumbeats to the words they represented, they found a second pattern: The intervals between beats changed in length depending on the sounds that followed each vowel. If a sound segment consisted of just one vowel, the time after the beat was quite short. But if that vowel was followed by a consonant, the time after the beat went up an average of 80 milliseconds. Two vowels followed by a consonant added another 40 milliseconds. And a vowel followed by two consonants added a final 30 milliseconds.
These short durations are enough to distinguish the drummed messages for “go fishing” and “bring firewood,” which are identical in tone, but not in their ordering of consonants and vowels. That means, the researchers write, that rhythm plays a crucial expressive role in drummed languages. If the same holds true for spoken languages, Seifert says, tiny delays after certain sound combinations—such as consonant clusters—could help us distinguish words in noisy settings.
University of Edinburgh linguist Bob Ladd isn’t so sure that that the finding will solve what is commonly referred to as the cocktail party problem. But he is impressed by how clearly the work shows our “incredible sensitivity” to small differences in timing. He says it could even push some linguists to redefine how they analyze rhythm, from focusing on regular, metered divisions (think the time signature of a song) to irregular variations instead.
“I’m amazed that these tiny milliseconds are doing the job,” says Onur Güntürkün, a biopsychologist at the Institute of Cognitive Neuroscience in Bochum, Germany, who has studied how whistled languages are processed in the brain. He says he would love to see an analysis of how the two hemispheres of the brain process drummed speech, in particular the timing cues.
That’s not on the table for now. What is, Seifert says, is a controlled experiment that will present Bora listeners with drummed messages that are identical except for slight timing variations. Just a few milliseconds, he thinks, could make all the difference.