The riddle of speech

Abstract

The evolution of language, which is unique to humans, has been a riddle for scholars. Molecular biologists are now putting together a genetic theory of how speech and language developed.

The development of speech and languages has fascinated linguists, philosophers and biologists down the ages, given the enormous role they have played in social and cultural evolution of humans. This led to a proliferation of ideas and hypotheses, but with no conclusive evidence to back any of them, partly because language has not evolved in any nonhuman primates or other animals as parallels for study and partly because there is no associated fossil and archaeological evidence 1. Another key factor is that virtually no genes associated with linguistic processes have yet been discovered 1.

Once molecular biologists entered the fray, the FOXP2 gene attracted huge interest as a crucial candidate for the evolution of speech…

Once molecular biologists entered the fray, the FOXP2 gene attracted huge interest as a crucial candidate for the evolution of speech and it has dominated research for almost two decades. That is changing now as other candidate genes emerge as major players in the development of vocalization. It also led to greater understanding of the distinction between the vocalization of speech, which involves motor coordination of the larynx, and the emergence of complex language including grammar. Indeed, unravelling the genes involved in vocalization is a necessary precursor to developing a rigorous genetic basis for the evolution of speech and ultimately language, which involves even closer integration with neurological functions.

Unlike language, vocalization is shared with other species, notably songbirds, bats, dolphins and whales. This makes it easier to identify shared genes that either emerged in common ancestors or in parallel. There is also scope for studying the impact of deleterious mutations associated with speech both in humans and in transgenic mouse models—although those animals are not capable of learned vocalization, some of these genes affect general acquisition of motor skills for example.

The discovery of FOXP2

A breakthrough came with discovery of the FOXP2 gene in the late 1990s partly through serendipity. Indeed, the story of contemporary research on speech evolution could be said to date back 20 years when a group at Oxford University led by Simon Fisher identified a locus, or region, on chromosome 7 associated with a severe speech impediment called verbal dyspraxia through the study of one British family, in which the speech disorder was highly prevalent across at least three generations 2. The disorder is dominantly inherited and results in an inability to articulate basic sounds and syllables.

The DNA region contains about 70 genes, and the next step was to identify the one containing the deleterious mutation. “We performed genetic mapping in the family in the 1990s confirming monogenic inheritance and localizing the damaged gene to part of chromosome 7”, Fisher explained. “Then, after more work, in 2001 we discovered the gene that was responsible, FOXP2, identifying a missense mutation that disrupts this gene and causes the disorder in the KE family” 3. “In that study, we also identified an independent child who had a very similar speech/language disorder, caused by a different kind of mutation, a chromosomal translocation, disrupting FOXP2. These findings identified FOXP2 as the first gene to have ever been implicated in a speech/language disorder”.

This discovery triggered a lot of excitement among language researchers and prompted the tangential question of what role FOXP2 might play in the evolution of speech or language. The gene belongs to the subfamily P of the forkhead box (FOX) transcription factor family, whose members play important roles in the regulation of tissue‐ and cell‐specific gene transcription during development and adulthood. Several of these genes are involved in language development, because variants have been associated with deficiencies that prevent proper learning of vocalization as in FOXP2, according to Wolfgang Enard from Ludwig Maximilians University in Munich, Germany, who was involved in the discovery of FOXP2's link with language deficiencies in the KE family.

Unlike language, vocalization is shared with other species, notably songbirds, bats, dolphins and whales.

“I talk about genes being knobs you can tweak”, Enard explained. “That does not mean evolution has tweaked the knobs, but at least we have knobs that are candidates. For example, FOXP1 is also involved in speech, but the amino acid sequence (it codes for) is 100% identical in humans and chimps”. Since chimpanzees have no ability to acquire vocalization, FOXP1 cannot have been the subject of selection for speech. As Enard pointed out, it is still possible that the regulation of the gene changed during evolution, but that is very hard to study. Fisher and his colleagues including Enard thus turned their attention to the role of FOXP2 in the evolution of normal speech and identified two mutations leading to amino acid substitutions in the coded protein, which was then confirmed in a separate study 4.

Not a recent mutation

“These found two main things: first that the proteins encoded by human and chimpanzee versions of FOXP2 differed by two amino acid substitutions, and second that patterns of variation in that part of the FOXP2 gene locus in human populations were consistent with Darwinian selection in humans, occurring within the past 200,000 years or so”, Fisher said. This led both groups to hypothesize that the two amino acid substitutions had been subject to recent selection and that this had something to do with the evolution of language. They suggested that these two amino acids had been subject to a selective sweep that quickly took hold across the whole human population.

… a recent study seems to have finally confirmed that those two mutations in FOXP2 did indeed evolve before the emergence of modern humans.

However, as Fisher admitted, this view had to be revised in the light of more DNA sequences both from current humans and from ancient hominid samples. “For example, through advances in sequencing of extinct archaic hominin DNA, Krause et al 5 showed that the amino acid substitutions in FOXP2 protein coding sequence were in fact already present in our common ancestor with Neanderthals. So those changes could not have arisen as recently as 200,000 years ago”, Fisher explained.

This in turn spurred other studies to look for non‐coding DNA variants, such as regulatory sequences, that could affect FOXP2 expression and been targets of a selective sweep, but as Fisher noted without any conclusive findings. Then, a recent study seems to have finally confirmed that those two mutations in FOXP2 did indeed evolve before the emergence of modern humans 6. This study undertook a comprehensive reanalysis of FOXP2 data based on the larger number of human genomes now sequenced and concluded that there was no evidence of positive or balancing selection for those two amino acids and that the original signal was a statistical artefact resulting from a small sample size.

“We could look at the entire sequence across the entire FOXP2 gene”, said Brenna Henn from the University of California, Davis, and a lead author on the paper. “In the original paper they were only able to sequence three introns surrounding this particular exon where the two amino acid mutations occurred. So they had a very tiny sequencing set, about 6,000 base pairs, as opposed to 600,000 base pairs. When we applied the same sort of statistical test, we came up with this conundrum that if we pooled our samples together, we were able to replicate the original signal but then if we parsed our sample into European and African populations, the signal disappeared. We realized they had been pooling individuals from around the world, Europeans and Asians, with just 6 or 7 Africans in the original data set”.

The situation was summed up by another senior author on that paper, Elizabeth Atkinson from the Broad Institute of Harvard and MIT Department of Epidemiology in the USA. “Our findings demonstrate that a sweep targeting a couple positions in FOXP2 was neither necessary nor sufficient for the evolution of spoken language in modern humans”.

FOXP2 and its role in speech development

However, although these recent developments have reaffirmed that FOXP2 was not after all subject to strong selection during evolution of modern humans, they have also confirmed that FOXP2 does have evolutionary significance for the development of speech. As a result, many researchers continue to study that gene almost exclusively, in songbirds and bats as well as humans.

The current status is that those two amino acid changes were important for the subsequent development of learned vocalization, but occurred earlier in evolutionary history. A notable point is that the FOXP2 protein is exceptionally well conserved across all other mammalian species and it is therefore highly likely that, at some point, it came under selection in humans, according to Svante Pääbo, Director at the Max Planck Institute for Evolutionary Anthropology in Leipzig, and involved in the early discoveries that the gene is implicated in language. “The FOXP2 protein in humans carries two amino acid changes relative to other primates. That is quite an extreme finding given how the protein is very conserved and only carries one additional change compared to the mouse”, he said. “This finding has been confirmed by a lot of genome sequences produced later. So yes, there is no convincing evidence that there is selection within the last 100,000 years or so on FOXP2. But there are these older changes in the encoded protein for which there is evidence that they may be important”.

In fact, the recent US work refuting the idea that FOXP2 has undergone a recent selective sweep in humans does seem to confirm that it was subject to earlier selection. The main finding was that a non‐coding, or intronic, region within the FOXP2 coding sequence close to those two amino acids has been conserved through evolution, or “constrained”, but is much more variable or polymorphic in humans. “This ROI (Region of Interest) is tightly constrained across vertebrate species in terms of its DNA sequence, which is interpreted as playing a significant functional role that maintains it under strong purifying selection, selecting out any mutations that arose here”, Atkinson commented. “In humans, however, we see a bunch of polymorphisms at common minor allele frequencies, which suggests that the purifying selection has been relaxed in the human line, unlike in the other species we looked at which still did not have any variation in this region. This in turn can be interpreted that whatever functional role this area used to have in the common ancestors of hominins and chimpanzees, it is either no longer doing that or it is not as important in the human line”.

Atkinson added that this is not conclusive, but that the ROI has been found to have many properties common to elements that enhance gene expression. It could therefore have played a role modulating expression of key speech genes. “We also find it expressed in low amounts in human brain tissue, so it could historically have been acting in this tissue of particular interest”, she said.

While Atkinson and Henn consider there has been too much emphasis on just FOXP2 in human speech/language evolution, they do not dispute that it plays a key role in vocalization. Indeed, the thinking now is that unravelling speech is essential for making progress on the greater challenge of language evolution, which is a highly complex phenotype that must involve multiple genes in combination with the development of motor vocalization skills. The evolution of vocalization and complex language must have proceeded in tandem generating interactive selective pressures for both domains.

… the thinking now is that unravelling speech is essential for making progress on the greater challenge of language evolution…

More candidate genes for speech evolution

On this front, significant advances have been made identifying other candidate genes for speech and language, notably by Fisher and his group. “We've been using multiple novel screening approaches to identify additional genes that might be implicated in speech and language impairments, given that FOXP2 mutations only account for a small proportion of such cases”, he said. “For example, we recently used next‐generation DNA sequencing to sequence whole genomes of a cohort of children with childhood apraxia of speech and identified novel rare mutations in a set of regulatory genes that are co‐expressed in the developing human brain 7”. Apraxia is a motor speech disorder whereby children find it hard to articulate words even though the vocal muscles are normal, because the signals from the brain are not getting through properly.

This paper has set the stage for a new chapter in the study of language evolution, according to Enard. “Simon's paper is a remarkable step forward because it gives a bunch of new knobs that might be relevant”, he said. In cases where DNA was also available from parents unaffected by apraxia, the group was able to discover de novo mutations in various genes including CHD3, SETD1A and WDR5 that play roles in epigenetic regulation of gene expression. In other cases, novel loss‐of‐function variants of various regulatory genes linked to neurodevelopment were identified. A lot of these genes are highly expressed together during early human brain development, suggesting that regulatory pathways in the developing brain contribute to acquisition of proficient speech and to loss of function when variants occur.

A lack of animal models

But this discovery can be seen just as a preliminary advance towards the long‐term goal of developing a genetic theory of speech and language evolution. The next question is to determine whether they were subject to selection during evolution, which, as Enard pointed out, is more difficult to determine. “The biggest problem we have is we don't have a very good model system, or assay, to test these genes”, he explained. The major established animal model whose genes can be manipulated by insertion of human analogues is the mouse, but that animal lacks any vocalization learning capability. Even so, it has enabled us to study the impact of FOXP2 mutations given their effect on learned behaviour. The introduction of the two human FOXP2 amino acid changes into mice profoundly affected learning and striatal neuroplasticity involved in motor functions. This led to the hypothesis that human FOXP2 evolution was associated with declarative and procedural learning and contributed to adapting the human brain for speech and language acquisition 8.

But the mouse model cannot be used to assess some of the newly discovered genes involved in speech, since their impact may not show up in mice if they play just regulatory roles in the development of vocalization during early childhood. The songbird would be a candidate, but cannot yet be manipulated genetically in the same way and is considerably more distant from humans in terms of evolutionary space. For this reason, the bat, being a mammal, might emerge as a stronger model. However, there is a possibility of developing an assay for mice that allows assessment of human speech genes more directly. It could also be possible to develop a genetically manipulatable model of the zebra finch that is already used as a model for studying development of vocalization as well as other aspects of neurobehavioral research.

At any rate, there is still scope for more insights from songbird research, including those related to FOX2P. “Our birdsong research suggests that the active regulation of FOXP2 in basal ganglia brain regions by the production of learned vocalizations is key to the quality of those communication signals”, commented Stephanie White, head of a laboratory focusing on social influences on the neural basis of vocal learning at the University of California at Los Angeles. “Our group and Constance Scharff's have shown this in birds. By extension, I predict it is critical in humans as well”.

This, unlike most work on speech evolution so far, has potential application for treating speech disorders in humans. “Our work strongly supports clinical efforts that involve intense behaviour therapy”, White explained. “What is now needed is to identify pathways that can be harnessed to improve the efficiency of these behavioural interventions since it is so time‐consuming and labour‐intensive to provide one‐on‐one interactions in the clinic that help to improve human communication disorders”. White concluded though by emphasizing that for all the recent attention paid to downplaying the role of FOXP2, it remains the only gene associated with speech disorders on a predictable heritable basis.

The biggest problem we have is we don't have a very good model system, or assay, to test these genes.

Yet, as Enard pointed out, even if a complete genetic theory of speech vocalization were developed, that would still not resolve the riddle of how language evolved. Since language in its full syntactical complexity and expressive richness is unique to humans, there are no animal models to test candidate genes and their alleles. However, recent research has shown that genes such as FOXP2 intimately related to vocalization also play broader roles in neurological development that impinge on language. A fuller understanding of the genes and their interactions involved in speech and vocalization could provide a foundation for a more comprehensive genetic theory of language.

EMBO Reports (2019) 20: e47618