Worms, voles and songbirds: why odd animals unlock human genetics

The Short Version

Some of the most important recent insights into human inheritance, social behavior and speech came from animals most people would never think about: a one-millimeter worm, several species of vole, rhesus monkeys living in outdoor colonies, and songbirds. Oded Rechavi's work in the worm C. elegans produced "clear cut proof" that acquired traits can be inherited through small RNAs [1]. Karen Parker's path ran from voles to monkeys to the first evidence that low vasopressin marks autism in children, and then to a small trial in which vasopressin improved social abilities [2]. Erich Jarvis showed that songbirds share not just speech-like brain circuits with humans but the same genes and even similar mutations, despite 300 million years of separation [3]. The pattern is consistent: the right odd animal makes something visible, fast or testable that is hidden, slow or off-limits in people, and the insight then travels back to human medicine [1][2][3]. Choosing the wrong model is also one of the main reasons promising drugs fail [2].

Table of contents

What They Said

Oded Rechavi: why a worm settled a century-old argument

Oded Rechavi, a Tel Aviv University neurobiologist, studies whether experiences can pass from parents to offspring, a question that has been taboo for over a century [1]. Huberman opens the model-organism part of their conversation by planting a flag: it was the first time the podcast had focused on one, and "I cannot emphasize enough the importance of model organisms and the incredible degree to which they've informed us about human health, especially when it comes to very basic functions and cells" [1]. He acknowledges exceptions, such as debate over whether telomere findings in mice carried over to humans, but concludes that "model organisms are absolutely critical and have been and basically inform most of what we understand about human health" [1].

A handful of organisms taught us most of biology. Rechavi explains that a model organism is one around which "there's a huge community of researchers that combine sources to create all the resources and the tools and the understanding that accumulates" [1]. "There is just a handful of model organisms in the short history of the field of biology," he says, and "we learned about every aspect of biology through them, including many important diseases, human diseases" [1]. His list includes E. coli bacteria, phage, flies, worms, zebrafish, mice, the plant Arabidopsis and non-human primates, with emerging models such as the regenerating planarian flatworm [1].

He gives two reasons they work. First, all life shares a common ancestor, so we share many functions and genes with these animals, and "they sometimes have some things that are much more apparent in them that we can study" [1]. His example: learning and memory research began largely in the sea snail Aplysia, because its big neurons are easy to study; "And yes, snails learn" [1]. Second, "you can actually experiment on them. We can't do these to humans, the things that we do to these animals" [1].

Why C. elegans in particular. C. elegans are round worms about a millimeter long that live mostly on rotting fruit [1]. Each adult has exactly 959 cells, 302 of them neurons, and each neuron is named; as Huberman notes, the community has "literally numbered and named each neuron, so that two laboratories on opposite sides of the world can publish papers on the same neuron, knowing that it's the same neuron in the two different laboratories. Something that is extremely hard to do in any mammalian model" [1]. Their full wiring diagram has been known since the 1980s, they are transparent so neurons can be watched and controlled with light, and theirs was the first animal genome sequenced [1].

Speed may be the biggest advantage. Huberman points out that waiting for a litter of mice can stretch a project by "three or four years" [1]. Rechavi agrees: a worm generation takes three days, "So you can do hundreds of worm generations in one PhD" [1]. Each mother produces about 250 nearly identical offspring, giving great statistics, and a controlled plate environment cleanly separates nature from nurture [1]. Rechavi adds an ethical point: "for me, it's much easier to work on worms. I don't have to feel bad about it" [1].

What the worm revealed. The foundation is RNA interference, discovered in worms by Andrew Fire and Craig Mello, who won the 2006 Nobel Prize [1]. Injecting double-stranded RNA silences the matching gene, the effect spreads through the body into the germ cells and offspring, and worms can even eat bacteria producing the RNA and get the same result [1]. It is conserved in many organisms including humans, and Rechavi says it is now routine, replicated "a million times" [1].

Building on that, Rechavi asked whether worms could pass on virus resistance [1]. Worms lack T and B cells and fight viruses with small RNAs [1]. Using a fluorescent virus, his team bred descendants of infected worms that couldn't make their own small RNAs; they still destroyed the virus, using small RNAs inherited from their parents, for several generations [1]. His lab also found that small RNAs made only in the worm's brain changed descendants' ability to find food for three generations, that heat exposure made descendants mate more with males for three generations, and that genes he named MOTEK act as a clock limiting how long inherited effects last [1]. The worm even offers a detail that mammals lack: an enzyme that amplifies inherited small RNAs each generation so the signal isn't diluted [1].

The payoff is striking. "We now have very obvious and clear cut proof that there is inheritance of acquired traits," Rechavi says. "So much so that I don't think that anyone, pretty much in the epigenetic field, argues against it" [1].

A detour through a bad worm story. The same field shows how odd animals can also mislead. In the U.S., psychologist James McConnell claimed that trained planarian flatworms kept memories after their heads regrew, and that memory transferred when trained worms were ground up and fed to others [1]. Rechavi notes McConnell went further, claiming "the fraction that transmits the memory is the RNA. And this is very, very interesting because it was a long time before everything that we know about RNA today" [1]. The work became seen as pseudoscience, but Mike Levin's lab later used automated tracking to replicate some of the head-regrowth results, and Rechavi's lab is testing whether RNA explains McConnell's findings [1].

Where the worm stops. Rechavi is careful about translation. Mammals are not known to have the worm's RNA amplification system, about 90% of epigenetic marks are erased between generations in mammals, and separating nature from nurture in humans is extraordinarily hard [1]. In human famine studies, for example, a starved pregnant woman directly exposes her fetus and even its germ cells: "The baby is itself affected. It's a direct effect," which "doesn't require necessarily a new biology" of inheritance [1]. The worm proves the principle, but it cannot by itself say how much of it applies to people [1].

Karen Parker: from voles to monkeys to children

Karen Parker, a Stanford professor of psychiatry, describes a research path that ran through three species before reaching children with autism [2]. Her starting point is a frank critique of the usual model.

Why mice fall short for autism. The brain is hard to study in living people, especially children with autism, who often can't tolerate scanners [2]. So researchers turn to animals, but "we need to make sure that the species that we're modeling them in has features of control humans," Parker says. "So we need to have complex cognitive abilities. We need to have complex social skills. We need to have an organism that has vision as its primary sensory modality" [2]. "The tricky part, I think, until fairly recently, was that we were doing all of this work in mouse models. And, you know, the control mice just fundamentally lack many of the characteristics that are needed to model, you know, autism with fidelity" [2].

The cost of that mismatch is large. "When we look at drug development pipelines, about 50% of preclinical failures," meaning something works in an animal then fails in a human trial, "50% of those failures can be attributed to poorly selected animal models" [2]. Her answer is "developing sophisticated models as a sort of point of entry into being able to understand some of these things that are really difficult to study in people" [2]. Huberman agrees, adding "we've just seen over and over again that the incredibly small fraction of mouse models that lead to valid therapeutics and humans" [2].

Voles: a light switch for fatherhood. Parker's first insight came from voles. In the 1990s, researchers showed vasopressin was key to male social behavior in monogamous prairie voles, promoting pair bonding and fathering, in contrast to non-monogamous montane voles in which males provide no care [2]. As a graduate student hanging out in the animal rooms, Parker noticed something odd about meadow voles, normally non-monogamous: under winter-like day lengths, "the males were hanging out with the females. And when the female had a litter, he was like participating. And I was like, whoa, these are not supposed to be monogamous animals" [2]. When she infused vasopressin into the brains of these males, it was like flipping a light switch: they gathered pups into a nest and huddled over them, while placebo did nothing [2]. Huberman notes it shows latent circuits can be unmasked rather than built from scratch [2].

Monkeys: a natural model of social difficulty. To overcome the limits of mouse work, Parker built a model at the California National Primate Research Center using rhesus macaques, "an old world monkey," chosen because they are "some of the closest relatives to human that are used in biomedical research" [2]. Rather than engineering a disorder, she looked for monkeys in large outdoor colonies whose natural behavior resembled autism's social features [2]. Her team "back translates" a human rating scale, the Social Responsiveness Scale, for use in macaques [2]. The low-social monkeys "have diminished social motivations," she says; they groom less, make fewer social overtures, and "they don't lip smack back," ignoring the macaque's friendly gesture [2]. Huberman notes that primates show affiliative gaze and grooming and "ostracization of individuals in a troop," dynamics that "map so clearly on the human behavior and vice versa" [2].

Parker insists on the limits: "I would never say they have autism... they have features that resemble human autism and that allow us to model this" [2]. Then she asked a question outsiders to a field sometimes ask: "why are we looking in blood?" [2]. Measuring many neurochemical systems in blood and spinal fluid, a statistical classifier sorted monkeys into low- and high-social groups "with 93% accuracy," and the key driver was vasopressin in cerebrospinal fluid, not blood [2]. Levels were stable within individuals and tracked grooming, and the finding replicated in a second group [2].

Children: the translation test. "Monkeys don't have autism," Parker says. "So then the question becomes, does this have what's called translational value?" [2]. She went looking in human spinal fluid, placing hot pink stickers on lumbar puncture trays so she'd be called whenever children had spinal taps for medical reasons [2]. In seven children with autism and seven without, spinal fluid vasopressin alone classified 13 of 14 correctly; oxytocin showed no difference [2]. A replication with NIH samples, now including girls, found lower vasopressin in autism in both sexes, with lower levels predicting greater social symptom severity [2]. Infants later diagnosed with autism already had low vasopressin [2]. And in a small randomized, double-blind, placebo-controlled trial, intranasal vasopressin twice a day for four weeks improved social abilities in children aged six to twelve, with 17 on the drug and 13 on placebo [2]. She stresses it was a small pilot and a larger trial is underway [2].

When the wrong animal leads a drug astray. Parker's story also shows the cost of misreading animal data. Roche tested balovaptan, which blocks the V1A vasopressin receptor; it missed its primary outcome, and a later trial was stopped for futility [2]. Parker thinks the rationale leaned on animal evidence of vasopressin-driven aggression from solitary hamsters and territorial prairie vole fathers, species and settings where aggression is appropriate [2]. The animal was fine; the lesson drawn from it didn't transfer.

Erich Jarvis: songbirds as a mirror for human speech

Erich Jarvis, a Rockefeller University neuroscientist and Howard Hughes Medical Institute investigator, studies speech through birds [3]. His starting point is that humans are not as special as we like to think. "I think we humans overrate ourselves when compared to other species," he cautions [3].

Why birds and not apes. The rare ability is not understanding speech but producing learned sounds [3]. Hearing pathways are widespread, while the speech production pathway is special to humans, parrots and songbirds [3]. "Dogs can understand several hundred human speech words, great aids, you can teach them for several thousand. But they can't say a word," Jarvis says [3]. Among our closest relatives, the other primates, none are advanced vocal learners [3]. Birds fill the gap: researchers going back to the 1950s found that songbirds, parrots and hummingbirds, "just three of them out of the 40-something bird groups out there on the planet orders," can "imitate sounds like we do" [3]. "They had this kind of behavior that's more similar to us than chimpanzees have with us," he says [3].

Behavioral parallels. Then researchers "discovered even more similarities, these critical periods" [3]. A child raised without human language who reaches puberty finds it hard to learn one, "Well, the birds undergo these same thing" [3]. And when humans go deaf, "our speech starts to deteriorate," while if "a chicken becomes deaf, their vocalizations don't deteriorate, very little at least" [3]. Following Peter Marler's work, young songbirds learn their own species' song best but can learn others, which Marler called "the innate predisposition to learn," roughly the equivalent of universal grammar in linguistics [3].

Brain and gene parallels. Jarvis's former advisor Fernando Nottebohm found song brain regions absent in non-imitating birds, and Jarvis's own work showed these circuits have parallel functions with human regions like Broca's area and the laryngeal motor cortex [3]. More recently his lab found the genes specially expressed in these regions are similar across humans, songbirds and parrots, "and now we're finding the specific mutations are also similar, not always identical but similar, which indicates remarkable convergence for a so-called complex behavior in species separated by 300 million years from the common ancestor" [3]. Crucially, "mutations in these genes that cause speech deficits in humans, like in Fox P2, if you put those same mutations or similar type of deficits in these vocal learning birds, you get similar deficits" [3]. His summary: "convergence of the behavior is associated with similar genetic disorders of the behavior" [3].

The genes turned out to have understandable jobs. One group controls neural wiring; surprisingly, many axon guidance genes were turned off, because they normally repel connections, so "by turning it off, you got to gain a function for speech" [3]. Another handles calcium buffering and neural protection, which made sense once Jarvis learned the larynx has the fastest-firing muscles in the body [3].

Stuttering, found by accident. Birds also shed light on a human disorder. When Jarvis's lab damaged the striatum, part of the basal ganglia, in the song pathway, birds "started to stutter as the brain region recovered" [3]. Unlike humans, they recovered after three or four months, "Because bird brains undergoes new neurogenesis in a way that human or mammal brains don't," and it was the new neurons, not yet firing properly, that caused the stutter [3]. In humans, basal ganglia damage causes what is called neurogenic stuttering, and people born with a stutter often show basal ganglia disruption [3].

Dancing parrots and a theory of speech's origin. Work by Ani Patel and others showed that only vocal-learning species can learn to dance; parrots like Snowball the cockatoo bob and stomp to the beat [3]. Jarvis's lab found that vocal learning pathways in songbirds, humans and parrots are embedded within circuits for learning movement, leading to the motor theory of vocal learning origin: speech pathways arose from a duplication of surrounding motor circuits [3]. The tight link between hearing and moving the larynx then spread to neighboring regions, letting humans and parrots synchronize the whole body to sound; as Huberman puts it, "we're speaking with our bodies when we dance" [3].

Nature's own experiments. Jarvis now helps lead the Vertebrate Genomes Project, aiming to sequence all 70,000 vertebrate species [3]. Comparing genomes of species that evolved the same trait independently, like vocal learning, acts like nature's own experiment [3]. In the formerly missing "dark matter" of better genomes, his teams are finding regulatory regions specialized in vocal learners [3].

Where They Agree, and Where They Don't

All three rely on shared ancestry. Rechavi says we can learn about humans from these animals because we share common ancestry, functions and genes [1]. Parker chooses macaques because they are among our closest relatives in biomedical research [2]. Jarvis goes further, showing that even distant species can converge on the same genes for the same trait [3].

All three pick the animal that makes the question visible. Rechavi uses a worm with named neurons, three-day generations and transparent bodies [1]; Parker uses voles with switchable fathering and monkeys with natural social variation [2]; Jarvis uses birds because apes can't do the thing he wants to study [3]. In each case the choice is driven by what the question needs, not by closeness to humans.

They agree the model must fit the question. Parker's criticism of mouse models for autism, that they lack complex social cognition and vision-first perception, is the sharpest statement of this [2]. Jarvis makes the same point in reverse: for speech, chimpanzees are closer relatives but worse models than songbirds [3]. Rechavi notes that each model organism has different advantages that serve different questions [1].

They differ on how directly results transfer. Rechavi is the most cautious. He calls inherited acquired traits settled in worms but says mammals lack key machinery and the human evidence is hard to interpret [1]. Parker built translation into her design: monkey, then human spinal fluid, then infants, then a trial, while repeating that monkeys don't have autism [2]. Jarvis shows the most direct transfer, with the same genes and similar mutations producing similar deficits in birds and humans [3].

They also show that animals can mislead. McConnell's flatworm claims went off the rails for decades [1]. Roche's balovaptan may have leaned on aggression findings in hamsters and prairie voles that didn't fit autism [2]. Half of preclinical drug failures trace to poorly selected animal models [2]. Huberman notes that some mouse findings, such as those on telomeres, have been debated [1]. The odd animal is powerful, but only when matched carefully to the human question.

Why It Works

The sources describe several reasons unglamorous organisms keep paying off.

Shared biology. Common ancestry means shared genes and functions, so a mechanism found in one species often exists in another [1]. RNA interference, first seen in worms, is conserved in humans and now used in drugs [1]. Vasopressin and oxytocin are hundreds of millions of years old and conserved across species for social behavior [2].

Convergence. Even without close ancestry, species that evolve the same trait can do it with the same genes [3]. That lets birds serve as a genetic model for human speech, with mutations in genes like FOXP2 producing similar deficits in both [3].

Visibility. Some animals make a process easier to see. Aplysia's big neurons made memory research possible [1]; worms are transparent with named neurons [1]; low-social monkeys show clear, measurable social behavior [2]; birds stutter and recover in ways that reveal the role of new neurons [3].

Speed and scale. Three-day generations and hundreds of genetically near-identical offspring let a single PhD test hundreds of generations with strong statistics [1]. That is what made multi-generation inheritance experiments feasible [1].

Experimental access. Researchers can change genes, infuse molecules into the brain and control environments in animals in ways impossible in people [1][2]. Parker's vasopressin infusion in voles and Jarvis's lesions in birds are examples [2][3].

Natural variation. Parker's monkeys weren't engineered; they varied naturally, as humans do, which let her find a biomarker that then held up in children [2]. Comparing genomes of species that evolved traits independently also uses nature's own variation as an experiment [3].

Where It Applies, and Where It Doesn't

It applies to basic mechanisms. How genes are silenced, how RNA can carry information, how social peptides act, and how speech circuits are wired all came from or were clarified by non-human animals [1][2][3].

It applies when the model shares the relevant feature. Songbirds share vocal learning, macaques share complex social behavior and vision-first perception, and worms share RNA interference machinery [1][2][3].

It doesn't apply automatically to humans. The worm's RNA amplification system has no known mammalian equivalent, so inherited effects may not last in people as they do in worms [1]. Parker's monkeys model features of autism but don't have it [2]. Birds regrow neurons in ways humans can't, which is why they recover from stuttering and people often don't [3].

It fails when the model is chosen for convenience rather than fit. Parker's figure, that about half of preclinical failures come from poorly chosen animal models, is the clearest warning [2]. Mice may simply lack the features a human condition depends on [2].

It requires rigor. McConnell's worm results were long dismissed as pseudoscience before partial replication with better methods [1]. Parker calls her own positive trial a small pilot awaiting a larger one [2]. Odd animals generate insights; careful human studies confirm them.

If you are considering treatments based on animal research, such as vasopressin, which Parker warns people against obtaining on their own because of its blood pressure effects, talk to a physician first [2].

What To Do With It

1. When you read a health headline, ask which animal it came from. Findings in worms, flies or mice can reveal real mechanisms, but Rechavi and Parker both separate what is proven in animals from what is known in humans [1][2].

2. Ask whether the animal fits the question. Parker's checklist for modeling autism, complex cognition, complex social skills and vision as the main sense, is a useful test of whether a model can say anything about a human condition [2].

3. Don't dismiss research on strange animals as wasteful. Worm RNA research led to drugs, vole research pointed Parker toward vasopressin, and songbird research revealed genes and mutations behind human speech [1][2][3].

4. Look for replication across species. Parker's vasopressin finding moved from voles to monkeys to children to infants to a trial [2]; Jarvis's genes show up in humans, songbirds and parrots [3]. A result that holds across species is more likely to matter for people.

5. Treat big claims about inherited trauma or memory with care. Inheritance of acquired traits is settled in worms, but human evidence is complicated by direct effects in the womb and by mammals lacking the worm's amplification machinery [1].

6. Support the tools that make comparison possible. Genome projects like the Vertebrate Genomes Project turn every species into a potential natural experiment [3], and model-organism communities share resources that let labs worldwide build on each other's work [1].

7. Remember that humans are one species among many. As Jarvis warns, we tend to overrate ourselves compared with other species [3]. Some of the best clues about human health are hiding in worms, voles, monkeys and birds.

Sources

1. Dr Oded Rechavi Genes The Inheritance Of Memories Across Generations · Huberman Lab
2. Dr. Karen Parker: The Causes & Treatments for Autism · Huberman Lab
3. Dr. Erich Jarvis: The Neuroscience of Speech, Language & Music · Huberman Lab