EVOLUTION OF LANGUAGE

SPECIAL SECTION

NEWS with all the fingerprints wiped off,” says

brain scientist Terrence Deacon of the Uni-
The Origin of Speech versity of California, Berkeley.

The long view

How did the remarkable ability to communicate in words first evolve? Researchers Archaeologists have identif ied various
probing the neurological basis of language are focusing on seemingly unrelated milestones in human behavior in the 5-
abilities such as mimicry and movement million-year evolutionary void between
animal communication and human
In the 1860s, both the British Academy and speak with the tongues of angels, he says; speech, but there is no consensus on
the Société de Linguistique de Paris warned rather, it achieved a more modest state which achievements imply the capacity
their members not to discuss the origins of some researchers call “language readi- for language. For example, the first stone
language, because the topic was so seductive ness,” which opened the door to further tools date to 2.4 million years ago; some
—and so speculative—that it spawned end- advances in linguistic ability. researchers think this may indicate lin-
less, futile theorizing. More than a century Language origins are “certainly worth guistic facility, but others argue that tool-
later, Noam Chomsky, the most influential talking about now,” says Hurford, who in making is far removed from speech. An-
linguist of the last 50 years, wrote that lan- 1996 launched the first of a series of biennial other possible starting point is 2 million
guage evolution and the brain mechanisms conferences on language evolution* that years ago, when the hominid brain began
underlying it “appear to be beyond serious in- have grown steadily. Hurford’s Edinburgh a period of rapid expansion, including in
quiry at the moment.” colleague Simon Kirby has documented the the primary brain areas associated with

Downloaded from www.sciencemag.org on June 3, 2007

But the time now appears ripe for leap in interest with producing or processing language—
this endeavor. In the past decade, namely Broca’s area in the left frontal
an unprecedented number of re- cortex and Wernicke’s in the left temporal
searchers from many disci- lobe (see brain model, p. 1318).
plines have begun to tackle As for actually producing the sounds of
the origin of speech, spurred words, or phonemes, skeletal studies re-
by new techniques as well as veal that by about 300,000 years ago,
new ways of thinking. Among our ancestors had become more or
linguists, the question of lan- less “modern” anatomically, and
guage origins was long obscured by they possessed a larynx located at
the dominance of Chomsky, whose the top of the trachea, lower than in
theory of an innate “universal other primates (see diagram). This posi-
grammar” ignored the problem of tion increases the range of sounds humans
how this language ability arose. In can make, although it also makes it easier
1990, however, the wave of evolu- for food going down the esophagus to be
tionary thinking that had previously misdirected into the windpipe, leaving us
swept through biology finally struck more vulnerable than other mammals to
linguistics too: That year, Harvard cog- choking. Such anatomy could have de-
nitive scientist Steven Pinker and Yale veloped for no other purpose than
psychologist Paul Bloom published a speech, says Deacon.
long article in Behavioral and Brain Sci- Other possible milestones come
ences arguing that language must have from genetic studies. For example, re-
evolved by natural selection. The Pinker- searchers at the Max Planck Institute for
Bloom paper was “a kind of watershed,” Evolutionary Anthropology in Leipzig,
says linguist James Hurford the University Germany, reported last year that the
of Edinburgh, U.K. “Suddenly it was OK FOXP2 “speech gene,” which affects both
to talk about evolution of language in Dangerous talk. Side view of human vocal language and the ability to articulate (Sci-
Chomskyian circles.” tract shows that because of our lowered larynx, ence, 16 August 2002, p. 1105), was ap-
Meanwhile, advances in brain imaging, food and drink must pass over the trachea, risk- parently a target of natural selection. This
neuroscience, and genetics have enabled a ing a fall into the lungs if the epiglottis is open. gene may have undergone its final muta-
new contingent of researchers to go ever tion fewer than 100,000 years ago—and
deeper into our brains and our biological a citation search: The number of papers deal- no more than 200,000 years ago—perhaps
past. For a long time, researchers treated ing with both “language” and “evolution” laying the groundwork for a new level of
language ability as some sort of “miracle,” more than doubled from the 1980s to 1990s. linguistic fluency.
says neuroscientist Michael Arbib of the (See also Book Review, p. 1299.) Most researchers are inclined to the view
University of Southern California (USC) Yet despite all the activity, the new lines that language gradually emerged over per-
in Los Angeles. Now, he says, researchers of evidence remain indirect, leaving plenty haps a couple of hundred thousand years
ILLUSTRATION: C. SLAYDEN/SCIENCE

are breaking that miracle down into a se- of room for interpretation—and conflict. “If (Science, 20 November 1998, p. 1455). But
ries of smaller, more manageable “mira- you want a consensus, you won’t get it,” all we know for certain, says Pinker, is that
cles,” involving disparate capacities such says cognitive scientist Philip Lieberman of fully developed language was in place by at
as the ability to imitate facial expressions Brown University. With no fossils of speech, least 50,000 years ago, when humans in Eu-
or to string movements together. They’re the origin of language remains “a mystery rope were creating art and burying their
not fantasizing that the human brain at dead, symbolic behaviors that point un-
some point suddenly found that it could * www.ling.ed.ac.uk/evolang equivocally to fluent language.

1316 27 FEBRUARY 2004 VOL 303 SCIENCE www.sciencemag.org

 EVOLUTION OF LANGUAGE

SPECIAL SECTION
The motor route
Understanding when language emerged
will probably have to await better under-
standing of how it emerged. In recent
years many researchers have become in-
creasingly attracted to the notion that
changes in the brain’s motor areas were
crucial for language capability.
Although we tend to associate language
first with sound rather than movement,
speech may be better understood as a motor
activity, says Deacon. Like other fine motor
activities such as threading a needle or play-
ing the violin, speech demands extraordinar-
ily fine and rapid motor control. Elaborate From ape to human. Magnetic resonance images of a bonobo brain are warped onto the shape of a
movements of the larynx, mouth, face, human cortex, viewed from (left to right) the side, top, and front. Red and yellow areas in the tem-
tongue, and breath must be synchronized poral region (linked to language) and in the prefrontal and occipital regions had to be stretched the
with cognitive activity. most to reach the human configuration, whereas blue areas are similar in apes and humans.
Thus researchers are probing the links be-
tween language and areas in the brain that ease, which disrupts the basal ganglia, suffer touch below the neck due to a strange virus.

Downloaded from www.sciencemag.org on June 3, 2007

control gestures, either hand movements or erosion of syntactical abilities, as well as Although the man had to relearn the simplest
the articulatory gestures of mouth and problems with balance and movement. movements, using cognitive and visual feed-
tongue. Linguist Robert Kluender of the Uni- Pinker’s research, with cognitive scientist back to substitute for lost senses, he contin-
versity of California, San Diego, says explo- Michael Ullman of Georgetown University in ued to gesture automatically when he spoke,
rations of gestures, including sign language, Washington, D.C., lends weight to this view. even when researchers hid his hands from
CREDITS: (TOP TO BOTTOM) KARL ZILLES, PETER PIEPERHOFF, AND HARTMUT MOHLBERG/RESEARCH CENTER JÜLICH, INSTITUTE OF MEDICINE, GERMANY; NIGEL J. DENNIS/PHOTO RESEARCHERS INC.

offer glimpses into what might have been the They have shown that Parkinson’s patients his own and listeners’ view. “The hands are
“intermediate behavioral manifestations” be- with basal ganglia damage have more trouble really precisely linked to speech articula-
tween animal communication and speech. with regular verbs than with irregular ones. tion,” says McNeill. “Gesture is not a behav-
Many researchers think hand and face Conjugating a regular verb such as “walk,” ioral fossil that was superseded by language
gestures offer behavior that is more analo- Pinker explains, is a combinatorial, sequen- but an indispensable part of language.”
gous to speech than are animal vocaliza- tial task that calls for adding the “ed” for past But not everyone is ready to dismiss the
tions. In all other mammals, both breathing tense. But retrieving the past tense for an ir- meaningfulness of animal calls, with differ-
and articulation are directed by brain areas regular verb such as “come” simply calls on ing views often dependent on a scientist’s
quite separate from those associated with long-term memory. Such tasks require other specialty. Primatologist Marc Hauser of Har-
human speech, notes Pinker. Lieberman ar- brain areas as well, but Lieberman argues vard, for example, believes that primate calls
gues that nonhuman primates engage in “a that the basal ganglia are a common element are better candidates for speech precursors
limited number of stereotyped calls” such as in both movement and language disorders. than any gestures are. With chimp gestures,
alarm calls and that they don’t have the in- Indeed, although many other brain areas, “nothing gives a suggestion of anything
teractive or combinatorial quality of lan- including those responsible for articulation, referential”—that is, having an explicit asso-
guage. Apes’ anatomy is such that they hearing, planning, and memory, had to de- ciation with a concept or thing—he says.
“could produce a [phonetically] reduced velop to support language, there is abundant Primate alarm calls, in his view, “kind of
form of human speech,” adds Lieberman. behavioral evidence for an intimate connec- look like words.” For example, he cites work
“But they don’t.” They’re much better at tion between language and motor abilities, by psychologist Klaus Zuberbühler of the
signing, because apes’ motor behaviors have says Pinker. For example, psychologist University of St. Andrews, U.K., who has re-
more flexibility and are more involved David McNeill of the University of Chicago ported that African Diana monkeys can
in social interaction—through gaze, mouth cites the case of a man who lost all sense of modulate their alarm calls to indicate what
and facial movements, and type of animal (leopard or eagle) is threaten-
limb gestures—than their ing. Such sounds, says Hauser, “have a far
calls, Lieberman says. greater … connection to language than any
Lieberman argues that discovery on nonvocal signals.”
the crucial changes that Many linguists, too, are unmoved by
laid the groundwork for motor arguments, which they do not be-
language ability occurred lieve can explain how the brain developed
in brain circuits connected syntax. “Motor organs are for muscular
with the basal ganglia, sub- movements,” says Derek Bickerton of the
cortical structures involved University of Hawaii, Manoa, and that
in movement. In his view puts them at the “end of the pipeline” of
the basal ganglia is the “se- language production. “Whatever organiz-
quencing engine” that es motor movements is on a par with
makes combinations— what organizes throwing movements,”
both verbal and gestural— says Bickerton. “The purpose is to put
possible. As evidence he things in a regular invariant sequence.”
points to the fact that pa- Hand and mouth. Chimps gesture with both face and hands to That, he says, is very different from mak-
tients with Parkinson’s dis- help express themselves. ing sentences, which requires “putting

www.sciencemag.org SCIENCE VOL 303 27 FEBRUARY 2004 1317

 EVOLUTION OF LANGUAGE
SPECIAL SECTION

things into an extremely plastic order de- tecture for imitation”—in people. He com- with an intuitive sense of how their body
termined by your conceptual structure.” bined the results of single-cell brain record- parts correspond with those of others. Thus a
ings in monkeys with functional magnetic small child knows how to raise its hand in re-
Mirror, mirror resonance imaging in humans while they sponse to a parental wave. “There’s obviously
Despite such caveats, the motor-language watched or imitated finger movements or fa- a direct representation of your body in its
connection continues to draw attention, in cial expressions. Iacoboni says that in addi- body,” says Studdert-Kennedy.
part because of a 1996 discovery that many tion to Broca’s, the circuit comprises an area The theory developed new life when
see as the first hard data in years to bolster in the superior temporal cortex (which over- Studdert-Kennedy brought it to bear on
the theory. This is the so-called mirror neu- laps with Wernicke’s and has neurons that re- questions of language evolution. Mirror
ron system found in monkeys’ brains. spond to face and body movements) and one neurons, he says, “for the first time provide
Mirror neurons’ link to language depends in the parietal cortex, the homolog to the an example of a direct physiological
on imitation, a skill largely unique to humans macaque area called PF, which combines vi- hookup between input and output”: the ob-
and considered vital to language. Although sual and bodily information. “The neural ar- servation of an action and its imitation. In-
parrots and dolphins can do vocal mimicry, chitecture for imitation … overlaps very well deed, Rizzolatti’s group recently reported
imitation is not as a rule a that the macaque has “audiovisual” mirror
mammalian attribute: Even neurons: Some of the cells in F5 fire not
nonhuman primates do it only when a macaque watches a meaning-
poorly (contrary to the im- ful grasping action, but when it hears the
plication of the term “to sound of one, such as the sound of breaking
ape”). But imitation is the peanuts (Science, 2 August 2002, p. 846).

Downloaded from www.sciencemag.org on June 3, 2007

way babies learn their first Arbib believes that mirror systems probably
words. And it’s the only exist in other parts of the brain for many
way a common meaning other behaviors.
can emerge for an abstract He and others feel that mirror neurons
symbol, a phenomenon that offer the first concrete neurological evi-
linguists call “parity.” “Imi- dence of abilities crucial to language, but it’s
tation is the common thread a long way from a few firing neurons to
for people writing about speech. Some scientists think the potential
language origins,” says significance of mirror neurons may be exag-
neuroscientist Marco gerated. Macaques, after all, can’t speak and
Iacoboni of the University they can’t imitate either, notes Pinker. In his
of California, Los Angeles. view, mirror neurons’ “relevance to lan-
So researchers were ex- Wired for imitation? Classic language areas—Broca’s and Wernicke’s guage is still pretty fuzzy.”
cited when a team led by (yellow)—overlap (orange) with areas critical for imitation (red).
Giacomo Rizzolatti of the The first syntax: words or waves?
University of Parma, Italy, found what they with well-known language areas in the hu- Despite such drawbacks, mirror neuron the-
considered a plausible antecedent for the man brain,” says Iacoboni, who concludes ory is being invoked by both sides in the
human ability to imitate in the brains of that the dual-use nature of Broca’s area in schism over whether the earliest language—
monkeys. The team recorded electrical ac- particular “suggests an evolutionary continu- that is, symbolic sounds or gestures con-
tivity in macaques from 532 neurons in an ity between action recognition, imitation, nected by some sort of rules of syntax—
area called F5, which is homologous to and language.” used the voice or the hands.
Broca’s area in humans. Neurons in F5 Mirror neurons provide the “neural miss- Those who favor gestural origins, such as
are known to fire during monkeys’ “goal- ing link” between movement and speech con- psychologist Michael Corballis of the Uni-
directed” hand and mouth movements—for trol, says Arbib. They also fit well with an old versity of Auckland, New Zealand, point out
example, when reaching for food. theory, the “motor theory of speech percep- that mirror neurons are found in brain areas
What intrigued the researchers is that a tion,” developed in the 1950s by the late responsible for grasping. “I think it’s ex-
subset of these neurons, which they dubbed Alvin Liberman of Yale University’s Haskins tremely likely that language evolved in our
mirror neurons, also became active when a Laboratories. Psychologist Michael Studdert- early ancestors as a manual system, not as a
monkey merely watched another monkey (or Kennedy of Haskins Labs explains that when vocal one,” as far back as a million years ago,
a human) perform the action. This finding children imitate their first words, experiments says Corballis. He notes that when robbed of
CREDIT: MARCO JACOBIONI, JOHN BACHELLER, AND ARTHUR TOGA/UCLA

“opened a whole new approach to the lan- have shown that they (unlike another imitator, speech, people quickly develop sign lan-
guage evolution story,” says Arbib of USC. the parrot) are guided by the “gestural” fea- guage, as has been shown by the case of a
“What would a mirror system for grasping be tures of the sound—that is, by the actions of community of deaf Nicaraguans who created
doing in the speech area of the brain?” The the mouth rather than by a sound’s acoustic their own language.
researchers concluded that these mirror cells features. A well-known trick to demonstrate Given the strong role of manual and facial
form a system for matching the observation this is known as the McGurk effect: If you gesture in speech and the relatively recent fi-
and execution of mouth and hand actions— watch someone pronounce the syllable “ga” nal mutation of the FOXP2 gene, Corballis
the first steps toward imitation. while listening to a recording of someone argues that “autonomous” speech may not
So far, mirror neurons have been found in saying “ba,” you will likely hear “da,” a have become fully developed until the cultur-
only two brain areas in macaques, and the sound anatomically between the other two. al explosion beginning 50,000 years ago. The
single-cell brain recording technique that re- This means “you perceive speech by re- mirror system, he believes, reinforces his the-
vealed the macaque neurons isn’t done on ferring the sounds you hear to your own pro- ory, because it apparently evolved first for
humans. But Iacoboni believes he has identi- duction mechanism,” says Studdert-Kennedy. manual control. It “probably picked up vocal
fied a similar circuit—“a core neural archi- Humans, unlike other animals, are equipped and facial control quite late in hominid evolu-

1318 27 FEBRUARY 2004 VOL 303 SCIENCE www.sciencemag.org

 EVOLUTION OF LANGUAGE

SPECIAL SECTION
tion,” he says, as speech became the pre- them will be through ever-finer brain imag- mals, will help yield a better “taxonomy” of
ferred modality for communication for vari- ing technology so they can, as Bickerton the language conundrum, especially if bol-
ous reasons, such as the need to free the puts it, “find out the flow chart for a sen- stered by computational modeling. But an-
hands for work or to talk in the dark. tence in the brain.” Harvard’s Hauser and swers won’t come all at once. “I see this as
But others believe equally strongly that colleagues believe that research in animals a process of gradual convergence. The
even if movement and language are insepara- may identify behavioral analogs for “recur- problem space is shrinking” at long last,
ble, language is primarily an oral, not manual, sion”: the ability to string words together in says Bickerton. “It will be solved when that
behavior. Psychologist Peter MacNeilage of infinite hierarchical combinations. Arbib space goes to zero, not when someone
the University of Texas, Austin, has devel- predicts that the discovery of other types of comes up with the killer solution.”
oped a theory that monkey oral behaviors mirror systems, in both humans and ani- –CONSTANCE HOLDEN
(not vocalizations) are precursors of human
syllables, and he argues that the mirror neu-
ron system—especially the recent discovery NEWS
of neurons that respond to lip smacking and
nut cracking—bolsters his ideas.
MacNeilage suggests that the brain’s sup-
plementary motor area (an area adjacent to
The First Language?
the primary motor cortex that is important for Genetic and linguistic data indicate—but can’t quite prove—that our ancient
motor memory and sequential movements) ancestors spoke with strange clicking noises
controls the physical constraints on vocal ex-

Downloaded from www.sciencemag.org on June 3, 2007

pression. The actions of chewing, sucking, In the 1980 movie The Gods Must Be ancestral to all living humans lived in the sa-
and licking took on communicative Crazy, a soda bottle falls out of the sky and vanna and used clicks,” says vertebrate sys-
content—a job for Broca’s predecessor— lands among some strange-sounding tematist Alec Knight of Stanford University.
in the form of lip smacks, tongue smacks, Africans. Their excited chatter, punctuated Knight estimates that today only about
and teeth chatters. The next stage, says by rapid-fire sucking and clicking noises, 120,000 people rely on these odd sounds.
MacNeilage, was to give voice to these be- sounded intriguing but alien to audiences Even so, they are providing new insights into
haviors by bringing the larynx into play. This around the world. But a handful of studies how humans evolved the gift of gab, particu-
theory fits well with the fact that the unique of this seemingly esoteric language suggest larly when researchers add up the results of
sounds of click languages, which some spec- that our early ancestors depended on such different kinds of data. “There’s a lot of
ulate may have been the original mother clicks to communicate. The latest linguistic mileage to be gained by cross-referencing
tongue (see next story), do not use the work points to clicks as having deep roots, linguistic, genetic, and archaeological data
larynx. Once the larynx was involved, a originating at the limits of linguistic analy- and theories,” says Nigel Crawhall, a gradu-
phonology—a set of sounds that could be sis sometime earlier than 10,000 years ago, ate student studying click languages at the
combined in endless ways to form a large and genetic data suggest that click-speaking University of Cape Town, South Africa.
vocabulary—developed, and this in turn populations go back to a common ancestor
paved the way for the emergence of syntax. perhaps 50,000 or more years ago. Clicks in context
“I don’t believe manual gestural commu- Although the idea is far from proven, “it Today clicks are part of typical conversation
nication got to the point of the combinatorial seems plausible that the population that was for about 30 groups of people, most from
phonology that I’m talking about, because if Botswana, Namibia, South Africa, and
it did we’d still have it,” says MacNeilage. In nearby. The only recognized non-
his view, if sign language had become that African click language is Damin, an ex-
complex, there would have been no reason tinct Australian aboriginal language
strong enough—the desire to talk in the used only during manhood initiation
dark notwithstanding—to cause a transition ceremonies. Among African click
to vocal speech. “Nobody who argues that speakers, daily conversations can be
we went from sign to speech has given us an dominated by clicks, and sometimes
adequate translation theory,” he says. verbal sounds drop out completely.
Others say the “which came first” debate Adept tongue and inward air move-
is beside the point. “Evolution selected the ments distinguish clicks from other
ability to combine speech and gesture under nonverbal utterances. They are really
a meaning,” says McNeill. “The combina- just very strongly pronounced conso-
tion was the essential property”; neither ges- nants, says Amanda Miller-Ockhuizen,
ture nor speech could have evolved without a linguist at Cornell University in Itha-
the other, he says. It doesn’t matter which ca, New York. Click speakers have
came first, agrees Zuberbühler: “Once an click sounds in common, but they have
individual reaches a certain threshold in its different words and therefore very dif-
cognitive sophistication, it will inevitably ferent languages.* Some researchers ar-
express itself in a sophisticated way,” gue that click languages are far more
through any means at its command, he says. different from each other than English
CREDIT: NIGEL CRAWHALL

The deepest questions—such as how hu- is from Japanese.

mans became symbolic thinkers and devel- But that diversity is only now being
oped “theory of mind,” or awareness of oth- * To hear click sounds, go to hctv.humnet.
ers’ thought processes—remain far from re- All alone. Researchers ponder why the Hadzabe live ucla.edu/departments/linguistics/
solved. Researchers say one way to tackle so far from other click speakers. VowelsandConsonants/index.html

www.sciencemag.org SCIENCE VOL 303 27 FEBRUARY 2004 1319