A letter to the current issue of Nature has caused a stir among those interested in the evolution of language. It looks at the FOXP2 gene in more detail than any paper has ever done before. It also inspires at least as many questions as it answers, but now at least we have better questions. Also it has dealt yet another blow to the theory that language depends on distinct cognitive modules that permit internal thought and that later interface with motor modules (vocalizing or signing) for “externalizing” what you are thinking. If anything is becoming apparent from FOXP2, it is that language and motor activities are deeply entangled. It also provides more reason to doubt the original recent date ascribed to the gene's mutations.
The letter, titled, “Human-specific transcriptional regulation of C[entral] N[ervous] S[ystem] development genes by FOXP2,” was written by a large team represented by Genevieve Konopka and Daniel Geschwind (abstract here). The journal also had a summary article on the study, “The Importance of Being Human,” by Martin H. Dominguez and Pasko Rakic (abstract here).
The story so far: The FOXP2 gene is a highly stable gene, changing very little over millions of years, but it has changed twice since the last common ancestor with chimpanzees and bonobos, and it turns out to be important to the proper development of language. It cannot be directly responsible for speech, however, because it works by regulating the activity of other genes. (See: The Human FOXP2 Gene) The gene’s role in vocalization has been strengthened by evidence that FOXP2 is crucial in songbirds for enabling young birds to learn how to imitate the songs of their elders, (see: Birds Also Use FOXP2) and making echolocation sounds in bats (see: The Latest on the FOXP2 Gene). A year ago British researchers identified a “downstream” gene (CNTNAP2) regulated by FOXP2. Their report ended by saying that their work was a first step in understanding molecular networks affecting language. (See: A Second Gene Supports Language). Now it is time for another step.
What’s new: The letter reports a series of experiments. One set (in vitro) used human neurons in lab dishes. Their normal FOXP2 genes were replaced by chimpanzee versions of the gene. The other set of experiments (in vivo) used living mice who had their mouse FOXP2 replaced with human genes. The great difficulty of FOXP2 is that it regulates the operation of other genes. A full understanding of the gene thus requires learning what those downstream genes are, how they are regulated by FOXP2, how the changes to human FOXP2 have altered their operation, and how the downstream genes themselves have changed.
Briefly, the in vitro experiments established that chimp and human versions of FOXP2 have distinguishable downstream effects; the human gene makes some other genes react more strongly, others less strongly, while a third set of genes react the same way to both human and chimp versions of the gene. It sounds pretty banal when reduced to a sentence, but of course knowing which genes react in different ways amounts to a mass of valuable technical data. We now have a list of over 100 genes that act differently in response to human FOXP2 than to the chimp version. This work ends the determined skeptic’s argument that although we can see that the human FOXP2 is slightly different from chimpanzee versions, there is no proof that the differences are functional. The differences are functional, full stop.
The investigators also established that five downstream genes have altered in the human version and “met the standard criteria … for positive selection on the human lineage” [p. 216]. In other words, not only has the FOXP2 gene changed, but so have some of the genes it regulates* and the changes were the result of selection rather than drift.
This co-evolution of FOXP2 and its downstream genes suggests (to my ears) that the recent date for the human version of the gene that originally given must be wrong. The first attempt at dating the gene placed it at less than 200 thousand years ago, more or less contemporary with Homo sapiens, but I have previously reported another study placing it at over 1.5 million years ago. And then there is the controversial finding that Neanderthals also had the human FOXP2 gene. (See: FOXP2 Gene Over 1.5 Million Years Old). If FOXP2 was a solitary gene that accomplished many things at once, then a recent date would at least seem possible. But a gene that must co-evolve with many other genes is another story. All that takes time.
The in vivo experiments used human FOXP2 in mouse brains to see which areas were altered. The letter summed up their results:
this study reveals enrichment of differential FOXP2 targets with known involvement in cerebellar motor function, craniofacial formation, and cartilage and connective tissue formation, suggesting an important role for human FOXP2 in establishing both the neural circuitry and physical structures needed for spoken language. 
More will be coming for many years..
*For the record, the genes in question are: AMT, C6orf48, MAGEA10, PHACTRA2, and SH3PXD2B.
UPDATE; Simon Fisher (the man who first isolated the FOXP2) comments, "The Geschwind team did not do any work at all with mice in this paper. What they did was to compare the in vitro well-based data to general expression differences found in dissected tissue from normal humans and chimps - not a specific test of FOXP2 function in vivo, but one that can provide some tangential support for the relevance of the in vitro findings. The new paper does not actually show functional evidence of "involvement in cerebellar motor function, craniofacial formation, and cartilage and connective tissue formation", this interpretation is based on looking at the list of genes affected in vitro and asking what is already known/suspected about their functions from other work. My feeling is that the expression differences highlighted by the screening in this paper give some intriguing candidates requiring further solid functional validation to see how relevant they are to aspects of language evolution. We are still some way off from describing how differences in FOXP2 alter the properties and behaviour of neurons in the living brain. What we need now is the integration of expression profiling data like those in this new paper, with other approaches (state-of-the-art neurobiology, model systems and so on) which should help us move closer to a functional understanding of the roles of FOXP2 in the neural bases of human speech and language."