Does speech--that uniquely human trait--come from our genes, or is it learned? Luminaries such as linguist Noam Chomsky of the Massachusetts Institute of Technology have championed the role of evolutionary inheritance over that of culture. But for many years, proponents of this position could only look to languages themselves for evidence. They observed that many tongues share grammatical structures and other attributes, bolstering the argument that speech is innate. The suspicion that a "speech gene" might exist, however, remained unresolved. Then, in 1990, something extraordinary happened.

It could not have been a coincidence that a particularly large number of children from one family showed up at an English speech therapy school. The children mumbled almost unintelligibly and stumbled over grammar--they could not, for instance, describe events in a correct chronological order.

In studies led by Jane A. Hurst at Oxford Radcliffe Hospital in England, researchers found that the affected members of the family, dubbed the KEs, had a physically normal speech apparatus--lips, jaw, tongue and vocal chords. Their other fine-motor skills were normal, as were their hearing and IQ. For three generations about half the family members had suffered from the same speech defect. Clearly, the disorder had a genetic component and was specific enough in its effects to offer the hope that it was directly connected to that elusive speech gene.

When the gene responsible for the impediment was pinpointed just a few years later, it finally provided evidence that the ability to speak is indeed written in our DNA. But how exactly do genes regulate a complicated mental process such as speech? Studies of the gene in people and in our animal cousins--especially songbirds, whose vocal learning resembles that of people--could help explain why speech evolved in humans but not in any other species. They might also lead to therapies for speech impediments like the one plaguing the KE family.

Language in Our Genes
Geneticists led by Simon E. Fisher of the University of Oxford identified in the KE family a segment on chromosome 7 in which there must have been a mutation. But determining which of the dozens of genes in that segment was at fault promised to be a lengthy process of trial and error. The researchers got a lucky break when they found a boy from an unrelated family who had a similar speech problem. The boy had a visible defect on chromosome 7 in the same segment as the one that looked suspicious in the KE family. The chromosome was broken at a gene known as FOXP2, so the researchers started looking specifically at that gene. In 2001 they successfully identified FOXP2 mutations in KE family members with the speech defect, and the same defect was later confirmed in other people with similar speech impediments.

Ever since the first published reports about FOXP2, molecular geneticists and linguists have been engaged in a vociferous debate about how, precisely, the gene affects speech. Although the gene appears to be crucial for normal development, its specific role remains to be clarified. FOXP2 codes for a protein that affects hundreds or perhaps even thousands of other genes, and scientists have barely begun to understand its complex influence. Mutations in the FOXP2 gene appear to hinder the development of brain regions responsible for motor control as well as regions involved in language processing. Furthermore, the FOXP2 gene exists in a variety of species--from reptiles to mammals--so it must serve other functions besides facilitating speech.

Many researchers, including my team at the Max Planck Institute for Molecular Genetics in Berlin, are especially interested in FOXP2 in birds, because some songbirds learn their songs in a way that is strikingly similar to how children learn speech. By studying the role of FOXP2 in birdsong, we are revealing how it might affect the development of language in people.

Genes ensure that the brain develops normally in a number of ways. Specialized nerve cells need to be formed, they need to produce the correct connections to neighboring cells, and they need to be able to emit signals or conduct messages to other neurons. In addition, nerve cells in the brain must develop the capacity to process information so that they can "learn" things. Gene products, namely, proteins into which the genetic code is translated, are involved in all these processes.

FOXP2 codes for a transcription factor, a protein that binds to other segments of DNA, thereby affecting whether or not different genes are read and translated into their respective gene products. ("FOX" stands for "forkhead box" [see box above], which refers to the specific DNA sequence that codes for the part of the protein that latches onto other DNA molecules.)

As a transcription factor, the FOXP2 protein serves as an on-off switch for numerous target genes. Because all genetic material exists in duplicate (except in the case of the male Y chromosome), a FOXP2 mutation on one chromosome causes the body to produce only half as much of the transcription factor as it should. The resulting shortage somehow causes the speech defect found in the KE family.

Complex Interactions
To find out how the disturbed regulation of FOXP2 genes leads to speech disorders, we must first identify the regions of the brain in which FOXP2 is normally active. We can draw conclusions about the function of a particular gene based on when and where it is expressed, meaning when and where the cell produces a protein in accordance with the DNA blueprint.

The FOXP2 protein is produced very early on in the embryos developing brain, particularly in those regions that later become the cerebellum, the thalamus and the basal ganglia. Consistent with this pattern of expression, a structural analysis of the brains of patients with the telltale speech defect revealed that the volumes of their cerebellum and basal ganglia differed from those of people with unimpaired speech. In addition, when these patients spoke, parts of their basal ganglia were less active relative to those of normal subjects.

Both the basal ganglia and cerebellum control body movements. They are activated whenever complex motor skills are learned, such as those involved in playing a piano. Presumably these regions are also responsible for motor function during the formation of sounds. It seems plausible that the KE family's difficulty in articulating words is rooted in the malformation of these areas of the brain.

The effects of a faulty FOXP2 gene do not stop there, however. Brain changes also appear in two well-known cortical speech centers: the Wernicke speech area and Broca's region. Neurolinguists have long suspected that Wernicke's area controls the understanding of speech, whereas Broca's region is involved in the production of speech. We know now, however, that this strict division is a little too cut-and-dried because a number of other areas of the brain are involved in both understanding and producing speech. The human brain probably processes spoken information in many areas of the brain simultaneously.

The idea that the brain uses parallel processing to understand and control speech is consistent with another observation: in the KEs, parts of the brain that are normally not involved in speech are active. This activity could be a direct consequence of the FOXP2 defect--a proper amount of the transcription factor would have made those areas behave normally--or it could be a sign that the brain is attempting to compensate for one of the other flaws caused by the mutation.

Differentiating between the direct and indirect effects of FOXP2 is no simple matter. For example, because the gene is active during the embryonic stage, its dysfunction could disturb brain development. The brain could be "wired" incorrectly, or certain specialized nerve cells could fail to form. On the opposite end of the spectrum of possibility, the brain might develop normally but run into problems with information processing later on--for example, during the phase when children learn to speak. The real effects of the FOXP2 mutation probably lie somewhere in between these two extremes.

Bird Babble
To probe further FOXP2's effects on cognitive development, researchers are turning to animals for clues. The FOXP2 gene has been identified in primates, whales, birds and even crocodiles; it is highly likely that all vertebrates have it. The sequence of the gene in these animals is almost identical to that in humans. For example, only three of the 715 amino acids in the mouse FOXP2 gene product differ from those in the human version. The timing and location of the genes expression in the brains of other species are also very similar. So what is the FOXP2 gene doing in the brains of these animals, none of which is capable of speech?

Although most animals have vocalizations that seem to be innate, a few species--among them songbirds, parrots, hummingbirds, some marine mammals and bats--do learn vocal patterns by imitating their parents. To some extent, this process is similar to that of a human infant making his or her first efforts at learning speech. At first, baby sparrows can imitate only minute elements of their future song, for instance. This type of vocalization is referred to as subsong, and it is similar to infant babbling. When the young animal hears an example of what is correct, it adapts its vocal output.

Through intensive practice, young songbirds increasingly come to sound like their role models, mastering the repertoire by the onset of sexual maturity. As is the case with humans, songbirds are dependent on what they hear to develop normal vocalization. If songbirds are subjected to loud noises, if they become deaf or if the feedback from their "teacher" is interrupted, they never learn to sing properly.

The similarities between learned avian song and human speech run even deeper. Both humans and songbirds have developed neuronal structures that specialize in the perception and production of sounds. Compared with humans, whose brains use parallel processing to comprehend speech, songbirds have a rather more modularly constructed brain in which various centers assume specific roles. In the avian brain, auditory stimuli reach the high vocal center, which controls the muscular movements of the vocal organ via the motor center; damage to this region prevents singing.

Another important data path in bird brains extends from the high vocal center via area X--a song-learning center in the basal ganglia--to the thalamus and from there back to the cortex. This so-called corticobasal ganglia loop also exists in the brains of mammals, including humans, where it is vital for learning. In young birds, lesions in area X lead to abnormal twittering, whereas such lesions seem to have no effect in the adults of most songbird species--until they try to learn a new song. Apparently the corticobasal network is important for learning songs but not necessarily for singing them. In humans also, FOXP2 proteins are produced in large quantities in the basal ganglia, which is where the structural and functional anomalies occur in patients with FOXP2-related speech defects.

Genetic Songwriter
In the brains of zebra finches, area X contains more FOXP2 during the song-learning phase than during infancy or adulthood. Likewise in the canary, which changes its melody once a year after breeding season, FOXP2 is expressed particularly strongly in area X during this learning phase [see illustration on opposite page].

Accordingly, FOXP2 may well be involved in song plasticity—the ability to learn new songs.
To explore this possibility, our team genetically reduced the amount of FOXP2 in area X in zebra finches to artificially induce a situation similar to a FOXP2 mutation in humans. The crucial question is, What happens to a melody if less FOXP2 is expressed in area X while it is being learned?

Our initial results show that the zebra finches have difficulty learning their songs when they
have less than the normal amount of the FOXP2 transcription factor. We concluded that this protein is necessary for zebra fi nches to learn a song, but it is less important for motor functions overall. Thus, a FOXP2 mutation does not simply cause the brain to develop abnormally. The defect continues to play a role once the brain has become fully developed—an important clue about what happens in humans who have the speech problems displayed by the KE family.

Careful analogy of the zebra finch’s impaired song learning with the problems seen in the KE family leads us to believe that affected family members may have diffi culty imitating the sounds made by their parents. They are unable to harmonize their own speech with that of others. Should this suspicion be borne out, it would mean that the similarity between avian song learning and human speech acquisition extends all the way to the molecular level.

The logical implication is that the evolution of language is not a unique feature of the human lineage. Many species share the structure and molecular makeup of the brain that was already in place as our ancestors began to speak. Only as existing genes and neuronal systems continued to develop was the path cleared for the uniquely human capacity for speech.