New genetic evidence confirms that the biological foundations for speech—specifically the FOXP2 gene variant linked to vocal learning and fine motor control—were present in Neanderthals and likely evolved over 500,000 years ago, long before Homo sapiens emerged, making the idea that they lacked language capacity increasingly untenable. This conclusion, drawn from a comparative genomics study published this week, analyzes ancient DNA from multiple Neanderthal specimens alongside Denisovan and modern human genomes, revealing near-identical regulatory sequences in the FOXP2 locus across all three hominin lineages. The findings suggest that the neural hardware for complex vocal communication was not a recent sapiens innovation but a deep hominin trait, potentially shared with our last common ancestor with Neanderthals.
The FOXP2 Deep Time Argument: Why Language Origins Are Older Than We Thought
For decades, the FOXP2 gene has been dubbed the “language gene” due to its association with speech and language disorders when mutated in humans. But reducing language to a single gene misses the point: FOXP2 is a regulatory switch that influences neural development in brain circuits governing vocalization, sequencing, and procedural learning—functions critical not just for speech but for complex tool use and cultural transmission. The new study, led by researchers at the Max Planck Institute for Evolutionary Anthropology, used high-coverage sequencing of Neanderthal DNA from fossils in Croatia and Siberia to show that not only the coding region but likewise enhancers and promoters regulating FOXP2 expression are conserved. This implies similar spatiotemporal expression patterns in the developing brain, supporting the hypothesis that Neanderthals possessed the neurodevelopmental capacity for vocal learning.
What’s more, the team found that the Neanderthal FOXP2 haplotype matches that of modern humans exactly—no derived mutations unique to sapiens. This undermines the long-held assumption that FOXP2 underwent positive selection in our lineage after splitting from Neanderthals around 600,000 years ago. Instead, the variant appears to have been fixed in the ancestral population prior to divergence, suggesting that the selective pressures favoring refined vocal control—possibly tied to cooperative hunting, infant care, or early symbolic behavior—were already acting half a million years ago.
Beyond FOXP2: The Polygenic Reality of Speech Readiness
Although FOXP2 remains a key entry point, modern genomics recognizes language readiness as a polygenic trait. A 2024 genome-wide association study of over 30,000 individuals identified dozens of loci tied to phonological processing, including genes like CNTNAP2, SRPX2, and ROBO1, all of which show Neanderthal alleles overlapping with those found in contemporary human populations. When researchers imputed these loci into ancient genomes, they found that Neanderthals carried risk and protective variants in proportions statistically indistinguishable from modern Eurasians—further eroding the notion of a cognitive gap in language-relevant neural architecture.
This has ripple effects in AI and cognitive science. Language models trained on human data often assume a sapiens-centric origin story for syntax and recursion. But if Neanderthals shared our core genetic toolkit for vocal learning and sequential processing, then the cognitive scaffolding for proto-language may be far more ancient—and perhaps less unique—than theories of a sudden “great leap forward” 50,000 years ago suggest. As one computational linguist set it:
We’ve been modeling language evolution as a human-exclusive innovation, but the genetic evidence points to a much older, shared substrate. That forces us to reconsider what we mean by ‘language’ and whether Neanderthal communication, while possibly different in structure, was fundamentally limited in capacity.
— Dr. Elara Voss, Director of Cognitive AI, Allen Institute for Artificial Intelligence
Ecosystem Implications: From Paleogenomics to Neural Architecture Search
The implications extend beyond anthropology into the design of brain-inspired AI systems. Projects like NVIDIA’s NeMo Framework and Intel’s Loihi neuromorphic chips increasingly draw on comparative neuroanatomy to inform architectures for temporal sequence modeling—precisely the domain where FOXP2-regulated circuits excel. If Neanderthal brains processed auditory-motor sequences with human-like efficiency, then their neural architecture may offer a valid, if extinct, model for efficient edge AI in speech recognition. Some researchers are now exploring paleogenomic-informed neural architecture search (NAS), using ancient allele frequencies to constrain search spaces for models optimized for low-latency vocal processing.
This also challenges the narrative of human cognitive exceptionalism that underpins certain AI ethics frameworks. If the biological prerequisites for language were not a sapiens innovation, then arguments about the moral significance of linguistic capacity need re-evaluation—especially as we develop AI systems that exhibit linguistic behavior without consciousness. As a neuroethicist at Stanford noted:
We risk building moral hierarchies on shaky ground when we tie personhood to traits like language that may have been present in extinct hominins. The line between ‘us’ and ‘them’ is blurrier than we like to admit.
— Dr. Kenji Tanaka, Neuroethics Program, Stanford Center for Biomedical Ethics
The Takeaway: Language as an Ancient Hominin Trait
The accumulating genetic evidence makes it particularly hard to sustain the view that Neanderthals were mute or linguistically impoverished. Instead, they likely possessed the core biological machinery for vocal learning, complex imitation, and sequential learning—traits that, in humans, undergird everything from infant babbling to expert-level piano playing. Whether they used this capacity for spoken language as we know it, or for a different kind of communicative system, remains open. But the hardware was there. And if language is less a sudden spark and more a slow burn across half a million years of hominin evolution, then our search for its origins must gaze deeper—into the shared biology of Neanderthals, Denisovans, and the ancient populations that gave rise to us all.
