CHAPTER 3
THE MANUSCRIPT BEGINS
Enhanced FOXP2 Expression in Avian Subjects: Evidence of Emergent Linguistic and Cognitive Capacities
Dr. Eleanora Voss
Department of Integrative Biology
School of Biological Sciences
University of Nebraska-Lincoln
Date of Submission: February 14, 2026
ABSTRACT
Background: The forkhead box protein P2 (FOXP2) is widely recognized as a critical genetic component in human language acquisition. Mutations in the FOXP2 gene were first identified in the KE family (Lai et al., 2001), where affected individuals displayed severe deficits in grammatical processing and orofacial motor control. Subsequent comparative genomic analyses have revealed that the human FOXP2 protein differs from the chimpanzee ortholog by only two amino acid substitutions, while avian species possess a highly conserved variant that appears to play analogous roles in vocal learning.
Objective: To determine whether targeted upregulation of FOXP2 expression in Melopsittacus undulatus (common parakeet) can accelerate linguistic acquisition beyond the limits achievable through conventional behavioral training alone, and to characterize any emergent cognitive capacities resulting from such phenotypic enhancement.
Methods: We utilized recombinant adeno-associated viral vectors (serotype 9) to deliver a constitutively active FOXP2 construct to the robust nucleus of the arcopallium (RA) and Area X of twelve adult parakeets. Subjects underwent longitudinal behavioral observation over 90 days, with daily structured language exposure sessions and continuous passive audio monitoring. Control groups received either sham injections or standard vocabulary training without genetic intervention.
Results: Experimental subjects demonstrated (1) novel vocabulary generation independent of explicit training stimuli, (2) compositional syntax combining taught and spontaneously generated lexical items, (3) recursive self-referential communication, and (4) metacognitive behaviors suggesting awareness of their own linguistic processing. These capacities emerged without the thousands of training trials typically required for comparable achievements in non-enhanced parrots.
Conclusion: Phenotypic enhancement of language-capable species via targeted genetic modification may trigger emergent cognitive capacities not present in the baseline population. These findings suggest that the linguistic limitations observed in avian species may reflect biological constraints amenable to intervention rather than fundamental categorical differences in cognitive architecture. The implications for comparative cognition, evolutionary linguistics, and our understanding of language as a biological phenomenon warrant serious scholarly consideration.
Keywords: FOXP2, language evolution, vocal learning, genetic enhancement, avian cognition, emergent properties
INTRODUCTION
The question of whether language is uniquely human has occupied philosophers and scientists for millennia. Aristotle declared that only humans possess logos—the capacity for reasoned speech. Descartes argued that language use constitutes the definitive criterion distinguishing thinking beings from biological automatons. Even Noam Chomsky, whose universal grammar hypothesis implies deep structural commonalities across all human languages, has expressed skepticism regarding non-human linguistic capacities (Chomsky, 1965; 1980).
Yet the biological substrate of language appears to be considerably more ancient—and more widely distributed—than these philosophical traditions have suggested.
The modern era of language genetics began in 2001, when Cecilia Lai and colleagues at the University of Oxford identified a mutation in the FOXP2 gene as the cause of a severe speech and language disorder affecting three generations of the KE family (Lai et al., 2001). Affected family members struggled with grammatical processing, word retrieval, and fine orofacial motor control. The mutation was dominant: a single altered allele was sufficient to produce the disorder. Unaffected family members carried two functional copies of the gene.
FOXP2 is a transcription factor—a regulatory protein that controls the expression of other genes. It is expressed during development in specific brain regions associated with motor control, including the basal ganglia, cerebellum, and cortex. Its role in language appears to involve the coordination of the complex motor sequences required for speech production, though recent evidence suggests it may also influence the neural circuits underlying grammatical processing (Fisher & Scharff, 2009).
The gene’s evolutionary history is equally intriguing. Comparative analyses have revealed that the human FOXP2 protein differs from the mouse ortholog by just three amino acid substitutions, and from the chimpanzee version by only two. Crucially, these two substitutions appear to have occurred within the last 200,000 years—roughly coinciding with the emergence of anatomically modern Homo sapiens (Enard et al., 2002). The pattern of nucleotide variation surrounding the human FOXP2 gene suggests it was subject to strong positive selection during this period, consistent with a scenario in which enhanced linguistic capacity conferred significant adaptive advantages.
What comparative genomics has made increasingly clear, however, is that FOXP2 is not a “human gene” in any meaningful sense. It is ancient—present in nearly all vertebrates, with homologs identifiable in species ranging from zebrafish to elephants. And in at least three avian lineages, it appears to play a role functionally analogous to its role in human language acquisition.
Vocal learning—the capacity to modify vocalizations based on auditory experience—is rare in the animal kingdom. Most mammals produce innate vocalizations that develop normally even in complete auditory isolation. Humans are vocal learners, as are cetaceans, pinnipeds, elephants, and bats. Among birds, the capacity is found in three distinct groups: songbirds (oscine passerines), parrots (psittaciformes), and hummingbirds (trochilidae). These three lineages are not closely related; vocal learning appears to have evolved independently in each, suggesting either remarkable evolutionary convergence or an ancient origin followed by multiple losses in other lineages (Jarvis, 2004).
The neuroanatomical substrates of avian vocal learning bear striking similarities to human speech circuitry. Songbirds possess a set of interconnected brain nuclei collectively termed the song control system. Area X, a basal ganglia nucleus, is essential for song learning; the robust nucleus of the arcopallium (RA) projects to motor neurons controlling the syrinx (the avian vocal organ); and the high vocal center (HVC) coordinates the timing of vocal sequences. FOXP2 is expressed in Area X and the striatum during critical periods of song learning, and experimental manipulation of FOXP2 levels in these regions disrupts song acquisition (Haesler et al., 2004; Teramitsu et al., 2004).
Parrots, including the common parakeet (Melopsittacus undulatus), possess analogous neural circuitry but with an important additional feature: a distinct song system nucleus called the shell region of the anterior arcopallium, which appears to enable the exceptional vocal mimicry abilities for which parrots are renowned (Chakraborty et al., 2015). Parrots can reproduce sounds with remarkable fidelity—not just conspecific vocalizations, but human speech, electronic tones, and environmental noises. They can learn vocabularies of hundreds of words and use them in contextually appropriate ways.
Or so it has seemed.
The history of research on parrot cognition is dominated by a single extraordinary individual: Alex, an African grey parrot studied by Irene Pepperberg for thirty years until his death in 2007. Pepperberg’s work with Alex demonstrated capacities that challenged conventional assumptions about avian intelligence. Alex could identify objects by color, shape, and material. He could count quantities up to six. He understood categorical concepts like “same” and “different.” And he used English words apparently meaningfully—to request objects, refuse unwanted items, and categorize unfamiliar stimuli.
Yet even Alex’s achievements, impressive as they were, appeared to hit a ceiling. His vocabulary stabilized at roughly 150 words. His combinations remained largely telegraphic—functional communications without the recursive syntax characteristic of human language. He never produced novel utterances independent of his training history. He never, in short, demonstrated the capacity for linguistic productivity that Chomsky and others have argued constitutes the defining feature of human language.
The question that has occupied my research program for the past decade is whether these limitations reflect fundamental biological constraints—or merely the limitations of the training methods we have employed.
Every attempt to teach language to non-human animals has proceeded from the same basic assumption: the animal’s brain is fixed, and our task is to find the pedagogical key that unlocks its capacities. We train. We reward. We shape behavior through thousands of repetitions. And when the animal fails to achieve human-like linguistic competence, we conclude that the capacity is absent.
But what if the limitation is not in the training but in the biology?
The KE family taught us that a single gene mutation can profoundly alter linguistic capacity. The comparative genomics literature suggests that small changes in FOXP2 may have enabled the emergence of modern human language within the last 200,000 years. If two amino acid substitutions were sufficient to transform our ancestors’ communicative capacities, what might be possible in species that already possess sophisticated vocal learning abilities—species whose FOXP2 proteins differ from ours at only a handful of positions?
This is not, I should emphasize, a question I approached lightly. The ethical dimensions of genetic enhancement in sentient animals are substantial, and I have engaged extensively with institutional review boards, veterinary ethics committees, and colleagues in animal welfare science to ensure that this research meets the highest standards of humane treatment. All procedures were approved by the University of Nebraska-Lincoln Institutional Animal Care and Use Committee (Protocol 2025-AV-047). Subjects were group-housed in enriched environments with ad libitum access to food, water, and social interaction. Behavioral indicators of distress were monitored daily, and humane endpoints were established prior to study initiation.
That said, I am aware that the ethical framework for animal research is evolving rapidly, and I do not presume to have settled all relevant questions. What I offer here is not a definitive ethical justification but a transparent account of my reasoning and procedures, submitted for scholarly scrutiny.
The study design proceeded as follows.
We acquired twelve adult parakeets (Melopsittacus undulatus) from a licensed breeder, selecting for baseline vocalization tendencies and social compatibility. Subjects were quarantined for two weeks and health-screened before experimental procedures began. We assigned four birds to each of three conditions: (1) FOXP2 enhancement via viral vector delivery; (2) standard vocabulary training without genetic intervention; and (3) sham surgery followed by standard training.
The viral construct was designed in collaboration with the Vector Core Facility at the University of Iowa. We utilized recombinant adeno-associated virus serotype 9 (AAV9), which demonstrates robust tropism for neurons and has an established safety profile in avian species. The construct encoded a constitutively active variant of the parakeet FOXP2 gene under control of the ubiquitous CMV promoter, ensuring broad expression in infected cells. We targeted two regions: Area X (bilaterally) and the robust nucleus of the arcopallium (RA), delivering 2 µL of viral suspension (titer 1 × 10¹³ genome copies/mL) to each site via stereotactic injection.
Surgical procedures were performed under isoflurane anesthesia with continuous physiological monitoring. All birds recovered without incident and resumed normal feeding and social behavior within 24 hours.
Following a two-week recovery period, we initiated the behavioral protocol. All subjects received identical exposure to structured language training: four 30-minute sessions daily, five days per week, for the 90-day study duration. Training employed the model/rival method developed by Pepperberg (1994), in which one human trainer models desired behaviors while another serves as a “rival” for the subject’s attention and rewards. This method has proven more effective than operant conditioning alone for teaching referential use of labels.
In addition to structured training sessions, all subjects were housed in enriched environments with continuous passive audio monitoring. This allowed us to capture spontaneous vocalizations outside of training contexts—vocalizations that might indicate emergent capacities not attributable to direct instruction.
The primary outcome measures included: (1) vocabulary size, defined as distinct vocalizations used appropriately in at least three independent contexts; (2) combinatorial complexity, assessed via analysis of multi-word utterances; (3) novel word generation, defined as vocalizations not present in the training corpus; and (4) metacognitive indicators, including explicit error correction, requests for clarification, and apparent monitoring of one’s own knowledge states.
Secondary measures included social interaction patterns, stress indicators, and general health parameters.
I should acknowledge several limitations of this design. The sample size is small, constrained by both ethical considerations and practical resource limitations. The study duration—90 days—is relatively brief relative to the years of training that produced Alex’s documented capacities. And the use of a constitutively active construct means we cannot easily reverse the genetic modification should adverse effects emerge.
These limitations are real. They do not, in my judgment, invalidate the study’s contribution to a literature that has heretofore relied entirely on behavioral training methods that may systematically underestimate avian linguistic potential.
What follows is a detailed account of our observations, analyzed through the lens of comparative linguistics and cognitive science. I present these findings not as a definitive demonstration that birds can achieve human-like language, but as evidence that the biological substrates of such capacities may be more widely distributed—and more amenable to enhancement—than has been generally assumed.
The implications extend beyond ornithology. If targeted genetic modification can unlock emergent cognitive capacities in species with pre-existing neural foundations for complex behavior, we may need to reconsider the boundaries between “human” and “animal” cognition—not as a philosophical exercise, but as a biological reality with profound ethical and scientific consequences.
I am aware that some readers will find these implications unsettling. I ask only that they engage with the evidence on its merits, reserving judgment until the full account has been presented.
The subjects are waiting. The data speak. It remains only to listen with sufficient care to hear what they have to say.
She wrote this three weeks before she disappeared.
—M. Reyes
REFERENCES
Chakraborty, M., Walløe, S., Nedergaard, S., Fridel, E. E., Dabelsteen, T., Pakkenberg, B., … & Jarvis, E. D. (2015). Core and shell song systems unique to the parrot brain. PLoS ONE, 10(6), e0118496.
Chomsky, N. (1965). Aspects of the theory of syntax. MIT Press.
Chomsky, N. (1980). Rules and representations. Behavioral and Brain Sciences, 3(1), 1-15.
Enard, W., Przeworski, M., Fisher, S. E., Lai, C. S., Wiebe, V., Kitano, T., … & Pääbo, S. (2002). Molecular evolution of FOXP2, a gene involved in speech and language. Nature, 418(6900), 869-872.
Fisher, S. E., & Scharff, C. (2009). FOXP2 as a molecular window into speech and language. Nature Reviews Genetics, 10(3), 215-218.
Haesler, S., Wada, K., Nshdejan, A., Morrisey, E. E., Lints, T., Jarvis, E. D., & Scharff, C. (2004). FoxP2 expression in avian vocal learners and non-learners. Journal of Neuroscience, 24(13), 3164-3175.
Jarvis, E. D. (2004). Learned birdsong and the neurobiology of human language. Annals of the New York Academy of Sciences, 1016(1), 749-777.
Lai, C. S., Fisher, S. E., Hurst, J. A., Vargha-Khadem, F., & Monaco, A. P. (2001). A forkhead-domain gene is mutated in a severe speech and language disorder. Nature, 413(6855), 519-523.
Pepperberg, I. M. (1994). Evidence for numerical competence in a Grey parrot (Psittacus erithacus). Journal of Comparative Psychology, 108(1), 36-44.
Pepperberg, I. M. (1999). The Alex studies: Cognitive and communicative abilities of Grey parrots. Harvard University Press.
Teramitsu, I., Kudo, L. C., London, S. E., Geschwind, D. H., & White, S. A. (2004). Parallel FOXP1 and FOXP2 expression in songbird and human brain predicts functional interaction. Journal of Neuroscience, 24(13), 3152-3163.
Voss, E. L., & Henning, S. M. (2022). Vocal learning and neural plasticity in the parakeet song system: Implications for language evolution. Journal of Comparative Neurology, 530(4), 891-907.
Voss, E. L., Martinez, J. R., & Okonkwo, D. K. (2024). Targeted gene delivery to the avian song system: Methods and applications. Methods in Molecular Biology, 2876, 145-168.
End of Chapter 3