The Musilanguage Hypothesis

The gestural idea is older than Darwin's singing ancestor. In his Essay on the Origin of Human Knowledge (1746), the French philosopher Étienne Bonnot de Condillac opened his account of language with a "language of action," made of gestures and cries, and only then turned to articulate sounds.¹⁸ Princeton University Press still introduces Corballis's book as the first full theory of how language moved from gesture to speech, in a line of thought it traces back to Condillac.¹⁹ Hewes brought it back in 1973 by comparing what living primates could do with their hands and with their voices.¹⁶

What apes do with their hands. Two modern studies give the gestural case most of its comparative weight. Amy Pollick and Frans de Waal watched two groups of bonobos (13 animals) and two groups of chimpanzees (34 animals) and sorted their signals into 31 manual gestures and 18 facial and vocal signals.²⁰ Facial expressions and calls were tied to particular situations and looked much the same in every group. Gestures were looser. The same gesture turned up in several contexts, and repertoires differed between groups, as if each had a small gesture culture of its own.²⁰ Bonobos were the more flexible of the two species, and only in bonobos did adding a face or a call to a gesture change how the recipient responded.²⁰ The authors read this flexibility, the freedom of a signal from a fixed context, as the property language needed first.

Catherine Hobaiter and Richard Byrne went further in 2014. Working from video of wild chimpanzees in the Budongo Forest, Uganda, they analyzed more than 4,500 instances of gesturing and assigned meanings to most of the species' repertoire of 66 gestures. An arm raise asked for something; a hand fling or a slap on an object meant "move away."²¹ Several gestures often shared one meaning, so the flexibility lay mostly in having many ways to ask for the same thing. That is a lexicon in a weak sense, a set of signs with stable uses, and it lives in the hands.²¹

Arbib's ladder. Michael Arbib built the most detailed bridge from these abilities to language. His starting point is anatomical: the monkey premotor area F5, where mirror neurons were found, and Broca's area in humans are homologous regions, and both are active when an individual performs a grasp and when it watches one.²²²³ From there Arbib set out a sequence of stages: grasping, a mirror system for grasping, simple imitation, complex imitation, protosign and protospeech, and finally language. The key step is from simple to complex imitation. Chimpanzees can learn to copy object-oriented sequences only after long practice; humans can copy a new action quickly and break it into parts, and that ability supports pantomime.²³ Pantomime is expensive and ambiguous, so communities would have stripped it down into conventional signs, which is what Arbib means by protosign. In his account protosign never finished before speech began. It gave protospeech a scaffold, and the two grew together in what he calls an expanding spiral.²²²³ This is where the gestural and musical views come closest. A voice that grew up beside the hands could carry pitch and rhythm as well as reference.

A language made by children. Nicaraguan Sign Language shows what the manual channel can do when it is given a community. Before the 1970s deaf Nicaraguans lived mostly apart, using home signs with their families. In 1977 a special education center in the San Judas district of Managua took in about 50 deaf children; there were 100 by 1979, a vocational school for deaf adolescents opened in 1980, and by 1983 more than 400 deaf students attended the two schools.²⁴ The teachers taught lipreading and spoken Spanish, with little success, and took the children's signing for mime. In June 1986 the Ministry of Education called in Judy Kegl, a linguist from MIT, who found that the younger children had taken the older children's system and made it more complex, with verb agreement among other things.²⁴ Linguists now distinguish the first-stage system, Lenguaje de Signos Nicaragüense, from the fuller Idioma de Señas de Nicaragua of later cohorts.²⁴

Ann Senghas and Marie Coppola tracked one grammatical device, spatial modulation, which uses the direction of a sign to mark who did what to whom. Signers who started young used it far more than those who started older, and signers who entered after 1983 used it consistently where earlier signers had wavered.²⁴ The 2004 study of manner and path, described above, shows the same pattern.²⁵ Each new cohort of children made the system more regular, in about a decade.

What the evidence can and cannot show. Ape gesture shows that the primate hand is more flexible and more voluntary than the primate voice. Nicaragua shows that modern human brains will build full grammar in the manual channel within a generation or two. Neither shows that language began in the hands, since the Nicaraguan children already had brains shaped by hundreds of thousands of years of spoken language. The hardest question for the gestural view is the switch. If language began in the hands, nearly every human community now speaks. Corballis's answer is that speech took over gradually and became autonomous late, around 50,000 years ago in his estimate, as a cultural change.¹⁹ The multimodal view described below avoids the switch altogether by treating voice and hands as partners from early on.⁹

Turn-taking in other primates. In 2013 Daniel Takahashi, Darshana Narayanan and Asif Ghazanfar recorded pairs of common marmosets separated by an opaque curtain. The monkeys exchanged calls in long alternating sequences, and when one partner sped up or slowed down, the other followed. The authors modeled the exchange as two coupled oscillators, the same kind of model proposed for human conversation.³⁰ Not everyone was convinced. A Language Log post argued the data did not show turn-taking in the full human sense.³⁰ Gibbon duets show a tighter coordination. In white-handed gibbons the female's great call is answered by the male's coda, and males time their replies to the end of her phrases.¹⁰

The great apes take turns too, mostly with their hands. Marlen Fröhlich and colleagues filmed wild mothers and infants at two bonobo sites in the Democratic Republic of Congo and two chimpanzee sites in Uganda and Côte d'Ivoire, and applied the tools of conversation analysis to the gestures they used to start traveling together.³¹ Both species ran exchanges that looked like cooperative turn-taking, in which a signal was handed to a partner who was expected to answer. The two species had different styles. Bonobos usually caught the partner's eye before signaling and often answered before the gesture had finished: 47 percent of bonobo responses overlapped the signal, against 32 percent for chimpanzees. Chimpanzees waited longer and negotiated more, with more than a third of their responses coming two seconds or more after the gesture.³¹ The timing is far slower than human talk, but the shape of the exchange is already there.

The human gap. Tanya Stivers and colleagues studied conversations in ten languages, from small indigenous communities to major world languages, and published the results in 2009. The languages were Danish, Dutch, English, Italian, Japanese, Korean, Lao, ǂĀkhoe Haiǁom of Namibia, Tzeltal of Mexico and Yélî Dnye of Papua New Guinea.³² To compare like with like, the team took 350 questions from informal video recordings in each language and kept only yes-or-no questions, then measured the time from the end of each question to the start of its answer. The most common gap in the whole sample was zero, the mean was 208 milliseconds, and speakers avoided both overlap and long silence. Average gaps differed between languages, but by no more than about 250 milliseconds from the overall mean: Japanese speakers answered fastest, with a mean of 7 milliseconds, and Danish speakers slowest, at 469.³² The same things sped answers up everywhere. Nods and other visible responses came faster than spoken ones, confirmations came about 200 milliseconds faster than disconfirmations, and answers came sooner when the questioner was looking at the answerer.³²

Stephen Levinson later put the typical gap at about 200 milliseconds, while planning even a single word takes 600 milliseconds or more, so listeners must prepare their reply while the other person is still talking.³³ Levinson suggested the human turn-taking system is inherited from primates, began in gesture like the turn-taking of great apes, and became mainly vocal perhaps a million years ago.³³ On this view the timing of exchange is part of an older "interaction engine" that language later filled with content.

The first conversations. Babies take turns before they have words. Mary Catherine Bateson, filming mothers with infants of a few months in the 1970s, called the exchanges "proto-conversations" because the two partners seldom overlapped. Colwyn Trevarthen described the same face-to-face exchanges as primary intersubjectivity, a mutual tuning of emotion and timing.³⁴ The musicologist Stephen Malloch and Trevarthen later analyzed the sound of these exchanges and found a shared pulse, a shaping of vocal quality, and a narrative arc across the exchange. They called the capacity behind it communicative musicality.³⁵

Malloch's first case was a recording Trevarthen had made years earlier of a mother and her six-week-old daughter. Each utterance, the mother's or the baby's, lasted about a second and a half, with little variation, and the exchange fell into short stories of no more than about 30 seconds. In one of them the mother began with low phrases, prompted the baby for several seconds, reached a peak of rapid back-and-forth, and returned to her starting pitch, a shape Malloch heard as introduction, development, climax and resolution.³⁵ Malloch has said he coined the term after noticing his own foot tapping along to the tape. Trevarthen's further claim, that babies deliberately start these musical exchanges, is disputed by some psychologists and musicologists.³⁵

Measured precisely, infant timing follows a curious path. Elma Hilbrink, Merideth Gattis and Levinson followed twelve mothers and infants from 3 to 18 months. At three to five months the babies answered quickly, and just over a third of their turns overlapped the mother's, more often than the mothers overlapped them. Around nine months, earlier than the authors expected, the babies' replies slowed, while the mothers' timing stayed steady across the whole period. By 18 months the infants overlapped about as often as their mothers.³⁶ Levinson notes that infant turn-taking appears early and then slows for a time as the child's language grows more complex.³³ The rhythm of exchange comes first, and the words, when they arrive, at first get in its way.

For the musilanguage debate this matters in a specific way. Turn-taking is neither music nor language. It is a timing system shared by both, present in other primates and in infants before speech. Any song-like protolanguage would have been sung in turns.

West Africa and the Americas. In many African traditions call and response is a pattern of shared participation that extends from public discussion and ritual to music. A leader calls, a group answers, and the answer can be varied. In drumming a master drummer sets a pattern and others reply; in oral performance a griot's audience responds at set moments.³⁷ Enslaved Africans carried the practice to the Americas, where it shaped work songs, the ring shout, spirituals, blues, gospel and jazz, and later soul, funk and hip hop.³⁷³⁸ The first published collection of African American music of any kind, Slave Songs of the United States (1867), gathered 136 songs, many of them from newly freed Gullah Geechee people on Saint Helena Island, South Carolina, during the Civil War.³⁹

The ring shout, a counterclockwise dance with handclapping and call-and-response singing, is often described as an ancestor of later African American vocal genres.³⁸ It survives in one community on the Georgia coast. In Bolden, near Eulonia in McIntosh County, a song leader sets each song, a group of singers called basers answers the leader's lines, and a stick-man beside the leader beats time on the wooden floor with a broom or stick while the basers clap and pat their feet. The shouters move counterclockwise in a hitching shuffle, and their feet never cross; crossing them would make it dancing, which the tradition counts as unholy.⁴⁰ The stick stands in for drums, which coastal planters banned for fear they could signal a revolt. Outsiders thought the ring shout had died out until 1980, when folklorists learned that Bolden still kept it on Watch Night, New Year's Eve. The McIntosh County Shouters formed that year and received a National Heritage Fellowship in 1993.⁴⁰

Lining out. Church practice met it from another direction. Lining out, in which a leader sings or chants each line of a hymn before the congregation repeats it, was endorsed by the Westminster Assembly in 1644 for congregations where many could not read. It survives in Gaelic psalm singing on the Isle of Lewis, among the Old Regular Baptists of Appalachia, and in many African American congregations.⁴¹ The leader, the precentor or clerk, gives the line, and the congregation answers slowly, each singer ornamenting the tune a little differently so that the voices spread around the melody instead of moving in unison. Old Regular Baptist leaders sometimes line out only a verse or two and then call "Sing on!"⁴¹ How much the Scottish and the African traditions fed each other in America is disputed. Highland Presbyterian settlers in the Carolinas may have brought lining out to some enslaved Africans, but English and lowland Scottish colonists were far more numerous, and a 2009 article by the ethnomusicologist Terry Miller calls the Gaelic-origin story a myth in the making.⁴¹

The party. In the Bronx in 1973, at a back-to-school party DJ Kool Herc played for his sister Cindy Campbell, his friend Coke La Rock took the microphone and called out to people in the crowd. He is often named as hip hop's first MC, and the crowd shout-out became a fixture of the form.⁴² A DJ who drops a record and watches the floor answer is part of the same loop.

Antiphony and the psalm. Christian worship inherited two kinds of exchange. In responsorial psalmody a soloist sang the verses and the congregation answered with a short refrain, often "Amen" or "Alleluia," a practice taken over from Jewish liturgy. In antiphonal psalmody two half choirs sang alternate lines or half lines.⁴³ The Catholic Encyclopedia credits Ambrose, bishop of Milan from 374, with bringing antiphonal psalm singing to the Latin West by dividing his congregation into two choirs.⁴⁴ Augustine, who was in Milan at the time, describes the church taking up hymns and psalms "after the manner of the Eastern Church" while Ambrose and his people kept watch in the basilica during the Empress Justina's persecution.⁴⁵ He gives the reason as keeping the people from wearing down under their sorrow, one of the oldest jobs of singing together.

Tibetan ritual. Tibetan Buddhist ritual has its own exchange of voice and instrument. Monastic chant is led by a chant master, the umdze, who also plays the cymbals that mark the structure of the rite.⁴⁶ A full ensemble adds bells, the hand drum damaru, the large frame drum nga, thighbone and long trumpets, and two kinds of cymbals.⁴⁷⁴⁶ The ethnomusicologist Ter Ellingson, working with the instrumental ensemble (rol mo) of a Kagyu monastery at Swayambhu in Nepal, showed how the leader on the cymbals gives each piece its form. The basic cymbal stroke can be split into smaller units, strung into series or stretched out, and the music is organized by sequences of time units that unfold one after another instead of repeating in cycles.⁴⁸ The ensemble follows the leader's strokes the way a congregation follows a precentor.

Some chant traditions are written in a notation called yang yig, which draws the rise and fall of the voice as curving lines; Walter Kaufmann's Tibetan Buddhist Chant studies one such songbook.⁴⁹ At Gyuto, the Tantric college whose monks recorded with Huston Smith in 1967, each chanter can sound a low fundamental with audible overtones, so a single voice carries a chord.⁴⁷

What the forms share. Across these traditions the leader carries the variation and the group carries the frame. The shout leader, the precentor, the cantor and the umdze each set the pace, and the answer confirms that everyone is still together. It is the same arrangement that conversation analysts find in talk and that Fröhlich found in apes, scaled up from two partners to a crowd.³²³¹