The evolution of human language represents one of biology’s most outstanding puzzles, rooted in microscopic genetic differences that separate humans from other primates. A recent genetic analysis traces individual differences in modern spoken language abilities to specific non-coding regions of DNA that evolved rapidly in human ancestors. The findings were published in Science Advances.
The study examines human genetic variation to understand how complex spoken language emerged in hominins. The genetic sequences analyzed span millions of years, covering the divergence of humans from chimpanzees about six million years ago to the split from Neanderthals roughly 600,000 years ago. An open question in evolutionary biology revolves around the exact sequence of genetic events that gave rise to human speech.
Early genetic research highlighted a gene called FOXP2, based on the discovery that severe mutations within it cause specific difficulties in moving the mouth and face to form words. Subsequent studies revealed that common variations of FOXP2 do not explain typical differences in language skills among the broader population. Researchers shifted toward the model that language relies on thousands of regulatory elements across the genome, leaving it unclear exactly when these distributed networks evolved to support complex speech.
The paper was authored by co-first authors Lucas G. Casten and Tanner Koomar, alongside Jacob J. Michaelson and colleagues from the Department of Psychiatry at the University of Iowa. The team mapped cognitive traits across deep evolutionary time using a technique they developed called an evolutionary stratified polygenic score. A traditional polygenic score tallies up the tiny effects of thousands of genetic variants to predict a specific physical or cognitive trait. The new stratified approach acts as a chronological filter, separating these genetic variants based on when they originally appeared in the evolutionary timeline.
The researchers first analyzed a group of 350 children who underwent longitudinal language and cognitive testing from kindergarten through fourth grade. After finding genetic associations with language skills in this discovery group, they validated the results using large health databases encompassing more than 100,000 individuals. They applied their findings to a repository of ancient DNA spanning 20,000 years. This included remains from 3,244 ancient West Eurasians along with eight Neanderthals and two Denisovans.
To observe broader biological patterns, the team compared human sequence data with whole-genome alignments from 170 mammalian species. This comparison included 49 animals classified as vocal learners.
The analysis isolated stretches of DNA known as Human ancestor quickly evolved regions, referred to as HAQERs. These sequences make up less than 0.1 percent of the human genome. They changed extremely fast after the human lineage separated from chimpanzees, acquiring new functions before humans diverged from Neanderthals. The researchers found that sequence variations inside HAQERs strongly predict spoken language abilities, such as sentence repetition and vocabulary, in present-day populations.
Variations in these regions do not show an association with non-verbal intelligence. HAQERs function primarily as regulatory elements rather than direct protein-building genes. They act as molecular control dials that determine when and where specific genes are activated during brain development. The study notes that hominins evolved stronger binding affinities in HAQERs for certain regulatory proteins, including the transcription factor family containing FOXP2.
These ancient sequences direct the development of particular brain cells during early fetal growth. They heavily influence the growth of brain networks in the basal ganglia, an area deeply involved in vocal learning. While the core FOXP2 gene remains nearly identical across many species, the HAQER sequences determine exactly how strongly that gene functions in the developing human brain.
The team also uncovered an evolutionary balancing act shaped by reproductive constraints. By comparing polygenic scores across thousands of genomes spanning the last 20,000 years, they found that genetic variations for general intelligence increased over time as positive selection made them more common. The genetic scores tied strictly to language in HAQERs remained fixed at intermediate frequencies, suspended by an effect known as balancing selection.
The variants in HAQERs that improve language ability are also associated with larger head sizes at birth and a higher risk of birth complications. This relationship points to the obstetric dilemma, an evolutionary compromise where the physical risks of delivering large-headed infants through a narrow bipedal pelvis cap the ongoing enhancement of language traits. The mortality risk of birth complications naturally limits the most extreme language-enhancing genetic variants from saturating the human gene pool.
When looking beyond primates, the team found that HAQER-like sequences display convergent evolution in other mammals. Mammals that learn vocalizations directly show significantly higher similarities in these genetic regions than animals restricted to innate sounds. This parallel evolution hints that these specific regulatory sequences provide a fundamental biological pathway for developing sophisticated communication networks across mammalian lineages.
The researchers note several challenges in their analysis. Applying genetic prediction scores across diverse modern and ancient populations introduces statistical challenges, as genetic variants behave differently in distinct ancestries. The archaic DNA dataset relies on an extremely small sample of ten individuals. This means conclusions regarding the elevated HAQER scores found in Neanderthal and Denisovan genomes require conservative interpretation.
Phenotypic data across the replication datasets also varied widely. The initial 350-person group received extensive standardized testing, while the larger biobanks relied on abbreviated clinical measures or self-reported answers. Questions remain about how exact sequence mutations inside HAQERs direct specific cellular outcomes in the fetal brain. Future research will need to examine the morphological differences between archaic human species to see how Neanderthals might have accommodated large-headed infants safely.
The paper, “Ancient regulatory evolution shapes individual language abilities in present-day humans,” was authored by Lucas G. Casten, Tanner Koomar, Taylor R. Thomas, Jin-Young Koh, Dabney Hofammann, Savantha Thenuwara, Allison Momany, Marlea O’Brien, Jeffrey C. Murray, J. Bruce Tomblin, and Jacob J. Michaelson.