Skip to content
NatureHQ

Senses and abilitiesability

Vocal learning

Learning what a sound means is common. Learning to make a sound you were not born able to make is rare enough to list — and humans are on a list with parrots and whales.

Learning to produce a sound you were not born able to make. It is rare — a handful of unrelated groups of birds and mammals, with humans among them — and each of those groups appears to have arrived at it independently, using the same anatomical shortcut: a direct line from the forebrain to the muscles of the vocal organ.

Almost every animal learns things about sounds. A dog learns what its name means; a monkey learns which alarm call to answer; a bird learns to ignore the neighbours it hears every day. None of that is vocal learning. Vocal production learning is the much rarer ability to change what you can *produce* on the basis of what you have heard — to end up making a sound you would never have made otherwise — and it is the difference between a system whose signals are fixed by development and one where new signals can be invented, copied and passed on. The distribution is the interesting part. It turns up in songbirds, parrots and hummingbirds among birds; in cetaceans, humans, some bats, some pinnipeds and elephants among mammals; and essentially nowhere else, including in our own closest relatives. Those groups are scattered across the tree in a pattern that only makes sense if the trait evolved several times over rather than once. And when the brains are compared, the same feature keeps appearing in each: a direct projection from the forebrain to the motor neurons that move the vocal organ, present in the learners and absent in their non-learning relatives. Where the picture gets untidy is at the edges. Several species modify otherwise innate calls to a limited degree, and whether they count depends on where a line is drawn rather than on any fact about the animal — which is why the field increasingly describes vocal learning as a continuum rather than a club.

Developed coverage · 80% complete · reviewed 2026-09-02

What this page covers

A capacity documented in three bird groups — songbirds, parrots, hummingbirds — and several mammal lineages including cetaceans, some bats, pinnipeds, elephants and humans. The evidence is extensive for the first group and thin for several of the others.

Often confused with: Learning what a sound means, which is a different and far more widespread ability; Learning when to use a call you already have, which most social animals can do

Quick facts

Bird groups that learn
Songbirds, parrots, hummingbirds — and no others
The shared anatomy
A direct forebrain line to the vocal muscles, absent in non-learners
Among primates
Documented in humans; not in any other primate
Open or closed?
Both — a zebra finch learns once, a canary relearns every year

Where this appears

Every organism below has been linked to this page because the evidence links them. Each one carries its own evidence, and its own limits.

Three things that all get called learning

Only one of them is rare, and the popular version of this subject runs them together.

The distinction the whole subject depends on.
KindWhat the animal learnsHow common
Auditory learningWhat a sound means, or who made itVery widespread — most animals with ears
Usage learningWhen and where to produce a call it already hasWidespread in social animals, including primates
Production learningHow to make a sound it could not otherwise makeRare — a handful of unrelated lineages

A vervet monkey giving the right alarm call for a leopard is doing the first two and not the third; the calls themselves develop without a model, and what the young monkey learns is what to attach them to. That is a real cognitive achievement and it is not vocal learning. The confusion matters because it is what makes headlines about talking animals possible: an animal that has learned to use a signal appropriately looks, from outside, exactly like an animal that invented it.

Learning to produce new sounds is rare. It turns up in a handful of unrelated groups — songbirds, parrots, hummingbirds, some whales and bats, humans — and each appears to have arrived at it separately.

Well supported

Good evidence backs this, though some details remain open.

Vocal production learning is documented in three avian clades (oscine passerines, psittaciformes, trochilidae) and several mammalian lineages (cetaceans, some chiropterans, pinnipeds, elephants, humans), distributed across the phylogeny in a pattern indicating multiple independent origins. Comparative neuroanatomy identifies a recurring correlate: a direct forebrain projection to the motor neurons controlling the vocal organ, absent in related non-learners.

Who this applies to
A comparative statement about the distribution of the trait, with sharply uneven evidence quality between the groups listed.
Studied in
Aves, Mammalia
Why we rate it this way, and what the caveats are
Well supportedHigh confidence

The phylogenetic pattern and the neuroanatomical correlate are both well established. The uncertainty is at the edges of the list rather than in its shape — several species show partial or contested abilities.

How far it can be extended

The behavioural demonstration is strong for songbirds, parrots, hummingbirds and humans, and rests on smaller or more indirect evidence for cetaceans, bats, pinnipeds and elephants.

Caveats

  • Vocal learning is better described as a graded continuum than a binary; several species modify innate calls to a limited degree and whether they "count" depends where a line is drawn.
  • Evidence quality is very uneven: extensive for songbirds, comparatively thin for bats, pinnipeds and elephants.
  • A shared neural feature across independent origins may reflect a shared developmental constraint rather than a shared solution.

Still unanswered

  • Whether limited call modification in some primates and other mammals belongs on the same continuum or is a different phenomenon.

Last reviewed 2026-09-02

The evidence (3 studies)
Words used here
Production learning
Changing the sounds you are able to make on the basis of what you have heard. The narrow sense of vocal learning, and the rare one.
Usage learning
Learning when to use a call that develops without any model. Much more common, and often mistaken for production learning.

Which animals learn to make new sounds

A short list, scattered across the tree — which is what tells you it evolved more than once.

Diagram

Where vocal production learning turns up

Every dot is labelled, because a colour is not information. "Partial" means the evidence is real and thin.

Which animals learn to produce new soundsScattered across the tree — so it evolved several times over.SongbirdsdemonstratedWhere nearly all the evidence isParrotsdemonstratedCross-fostering shows learned callsHummingbirdsdemonstratedOne keeps learning as an adultHumansdemonstratedThe case the comparison exists forCetaceansdemonstratedHumpback song; dolphin whistlesBatspartialReal, and much thinner evidencePinnipedspartialA few individuals, famouslyElephantspartialImitation in known individualsMost other birdsnot shownCalls develop without a modelMost other mammalsnot shownIncluding our closest relativesEvery dot is labelled, because colour alone is not information.“Partial” means the evidence is real and thin — not that the animal half-learns.
The same explanation in words

A list of ten groups, each marked as demonstrated, partial or contested, or not demonstrated. Demonstrated: songbirds, where almost all the evidence comes from; parrots, where cross-fostering shows learned signatures; hummingbirds, where one species keeps learning as an adult; humans, the case the whole comparison exists for; and cetaceans, with humpback song sweeps and dolphin signature whistles. Partial or contested: bats, pinnipeds and elephants, where the evidence is real and much thinner — a few individuals in some cases. Not demonstrated: most other birds, whose calls develop without a model, and most other mammals, including our closest relatives. The pattern is scattered across the tree, which means the ability evolved several times over rather than once.

Two things about that list are worth holding onto. The first is that the strong entries and the weak ones are not equally strong: songbird vocal learning rests on sixty years of tutoring, isolation and deafening experiments, while the elephant and pinniped entries rest on documented imitation by a small number of famous individuals. The second is that the list has an edge rather than a boundary. A handful of species modify otherwise innate calls slightly — enough to be measurable, not enough to build a new signal — and whether they are on the list depends on where somebody draws a line.

Hummingbirds are the third group of birds that learn their sounds — and at least one species carries on changing its song throughout adult life to match its neighbours.

Well supported

Good evidence backs this, though some details remain open.

Long-billed hermits at leks modify song after the first year in ways that track changes in the locally shared song type, indicating open-ended vocal production learning in a trochilid.

Who this applies to
One lekking hummingbird; the family is large and vocal behaviour varies widely across it.
Studied in
Phaethornis longirostris
Why we rate it this way, and what the caveats are
Well supportedModerate confidence

Marked individuals followed across years, with changes tracking the neighbourhood rather than drifting idiosyncratically. Observational, and one species.

Caveats

  • Observational — no tutoring, isolation or removal experiment has been run in this species.
  • Marked-individual field studies give modest sample sizes.
  • Lekking, with stable acoustic neighbours across years, is an unusual social arrangement.

Still unanswered

  • How widespread open-ended learning is across the hummingbird family, most of which has never been examined.

Last reviewed 2026-09-02

The evidence (2 studies)

A wild parrot chick develops an individually distinctive call that resembles the calls of whoever raised it — and swapping eggs between nests is what shows it is learned rather than inherited.

Well supported

Good evidence backs this, though some details remain open.

In green-rumped parrotlets, cross-fostering demonstrates that individually distinctive contact-call signatures develop with the acoustic features of foster parents rather than genetic parents. Parents produce distinct calls directed at individual offspring prior to the offspring producing calls of their own.

Who this applies to
One small neotropical parrot with an unusual nesting arrangement; parrots are a large and diverse group.
Studied in
Forpus passerinus
Why we rate it this way, and what the caveats are
Well supportedHigh confidence

Cross-fostering in the wild is the manipulation that fully separates inheritance from upbringing, and it is the design the equivalent dolphin work has never been able to run.

Caveats

  • A learned individually distinctive label is not a name: nothing here shows one parrot using another’s call to refer to it.
  • Sample sizes for cross-fostered wild broods are necessarily small.
  • What the signature is used for beyond individual recognition is not established.

Still unanswered

  • Whether parrots use another individual’s signature to address that individual, as dolphins appear to.

Last reviewed 2026-09-02

The evidence (2 studies)

The mammal cases, in detail

The one feature they all have

Five lineages that are not close relatives, converging on the same wiring.

Diagram

The direct line, and the loop around it

Schematic. Not anatomy.

The one feature every vocal learner has, and non-learners do notForebrainsong nucleihearingsyrinx musclesThe direct lineA projection running straight fromforebrain to the motor neurons thatmove the voice box.The loop that makes it learningThe bird hears its own output andcompares it with what it memorised.Break the loop and the song comes apart.Non-learnersHave the vocal organ and the hearing.They do not have the direct line.Schematic. Not anatomy.
The same explanation in words

A schematic of a bird’s head. Inside a box labelled forebrain sit two ovals: one labelled song nuclei, one labelled hearing, with an arrow from hearing to the song nuclei. A thick arrow runs straight down from the song nuclei, out of the forebrain, to a box at the bottom labelled syrinx muscles. Three notes explain it. The direct line is a projection running straight from forebrain to the motor neurons that move the voice box. The loop that makes it learning is the bird hearing its own output and comparing it with what it memorised — break the loop and the song comes apart. And non-learners have the vocal organ and the hearing; what they do not have is the direct line.

Learned song runs on a specific, mappable set of brain regions with a direct line to the muscles of the voice box — and the birds that cannot learn song do not have that line.

Established

Specialists would state this without hedging. Multiple independent lines of evidence agree.

Song production and learning in songbirds depend on a discrete forebrain network with a descending motor pathway to the syringeal motor neurons. Lesions to specific nuclei abolish or degrade song. A comparable direct forebrain-to-motor-neuron projection is present in other vocal-learning lineages and absent in related non-learners.

Who this applies to
Characterised in detail in songbirds; the comparative claim spans the avian vocal-learning clades and, more loosely, mammals.
Studied in
Passeri, Psittaciformes, Trochilidae
Why we rate it this way, and what the caveats are
EstablishedHigh confidence

Lesion, tracing and comparative anatomy agree, and the presence/absence correlation with vocal learning is the strongest structure-behaviour link of its kind in any vertebrate.

How far it can be extended

The pathway has been described independently in songbirds, parrots and hummingbirds, and its absence in related non-learners is the comparative test.

Caveats

  • A lesion establishes that a region is necessary, not what it computes.
  • The anatomy is far better characterised than the function, and remained so for decades.
  • Independent origins sharing a feature may reflect a shared developmental constraint rather than a shared solution.

Still unanswered

  • What the equivalent circuit is in mammalian vocal learners, where the anatomy is much less clear than in birds.

Last reviewed 2026-09-02

The evidence (3 studies)

It is worth being careful about what a correspondence like that establishes. Lesioning a region and losing the song shows the region is necessary; it does not show what the region does, and the anatomy of the song system was far better characterised than its computation for several decades. The comparative claim is stronger than the mechanistic one: the presence of the direct projection in learners and its absence in their non-learning relatives is one of the cleanest structure-behaviour correspondences in any vertebrate.

And one consequence that reshaped a different field entirely.

In some songbirds the brain regions that control song grow substantially in the singing season and shrink again afterwards — in adults.

Established

Specialists would state this without hedging. Multiple independent lines of evidence agree.

Song-control nuclei in seasonally breeding songbirds undergo large, reversible changes in volume across the annual cycle, associated with seasonal changes in song production and mediated in part by the addition and loss of neurons in adulthood.

Who this applies to
Demonstrated most dramatically in canaries; the magnitude varies widely and is small or absent in non-seasonal species.Do not extend this beyond the taxa listed — the popular version over-reaches.
Studied in
Serinus canaria, Passeri
Why we rate it this way, and what the caveats are
EstablishedHigh confidence

Replicated across laboratories and species, and it opened adult neurogenesis as a field — the finding is secure even though its magnitude in any given species is not.

How far it can be extended

Seasonal plasticity tracks seasonal singing, so species that sing year-round show little of it; treating the canary figure as typical overstates the effect substantially.

Caveats

  • Volume is a coarse measure; neuron number, cell size, spacing and connectivity all contribute and are not separated by it.
  • The canary effect is at the extreme end. Popular accounts describing birds growing a new brain every spring are describing something else.
  • Correlating regional volume with singing does not by itself establish which drives which.

Still unanswered

  • What functional difference the seasonal change actually makes to what a bird can sing.

Last reviewed 2026-09-02

The evidence (2 studies)

Why a rare ability keeps coming up in arguments about language

Vocal learning is a prerequisite for speech and nothing like sufficient for it.

Speech requires being able to produce sounds you were not born making — no vocal learning, no words — which is why the comparison keeps being made. But every non-human vocal learner has the capacity and none has language, which tells you how much of the gap the capacity actually closes. A humpback learns a song of extraordinary structural complexity and there is no evidence any part of it refers to anything. A parrot can reproduce human speech precisely and the evidence that it means anything by it comes down to one intensively trained individual.

Where this argument is had properly

What is still unsettled

  • Is limited call modification in some primates and other mammals the same phenomenon in a weaker form, or a different thing altogether?

    Why it matters: It decides whether vocal learning is a threshold our lineage crossed or a dial our lineage turned up, which is a different story about how speech became possible.

    What would settle it: Comparable measurement protocols applied across species, instead of each case being assessed against its own literature.

  • What is the mammalian equivalent of the songbird song-control pathway?

    Why it matters: The avian anatomy is well mapped and the mammalian side is not, so the strongest comparative evidence for convergence rests mostly on one class.

    What would settle it: Tracing studies in a mammalian vocal learner that is tractable in the laboratory — bats being the most likely candidate.

  • Does the shared circuit reflect a shared solution, or a shared constraint on what brains can build?

    Why it matters: Convergence on the same design is usually read as evidence that the design is the answer. It can equally mean it was the only route available from a common starting plan.

    What would settle it: Developmental work on how the projection forms in each lineage, which would show whether the same route is being taken for the same reason.

Claims about this, checked

Things people have heard, and what the evidence actually supports.

The research behind this page

13 studies, newest first. Each one has a page explaining what it found and what it could not show.

2019Science

Evolution of vocal learning and spoken language

Vocal production learning is documented in a small number of distantly related lineages, and the independently evolved brain circuits supporting it share a recurring feature: a direct projection from forebrain to the motor neurons controlling the vocal organ, which is absent in non-learners.

2013Biology Letters

Open-ended song learning in a hummingbird

Adult birds changed their songs after their first year, and the changes tracked shifts in the local shared song type rather than being idiosyncratic drift.

2012Proceedings of the Royal Society B

Vertical transmission of learned signatures in a wild parrot

Each chick developed an individually distinctive contact call whose features resembled those of the parents that raised it rather than its genetic parents, and parents used distinct calls when addressing different chicks before the chicks called at all.

2009Nature

De novo establishment of wild-type song culture in the zebra finch

Song moved towards wild-type structure at every generation, and within three to four generations the colony was singing song statistically indistinguishable from wild-type on the measured features.

2001Science

Dynamics of the vocal imitation process: how a zebra finch learns its song

Song development proceeds from highly variable, unstructured subsong through progressive differentiation of individual syllables, with different syllables converging on their targets at different times and rates rather than the song improving as a whole.

1999Nature

Decrystallization of adult birdsong by perturbation of auditory feedback

Song progressively degraded — syllables were lost, added and distorted — over weeks of perturbed feedback, and largely recovered towards the original song once normal feedback was restored.

1987Animal Learning & Behavior

Acquisition of the same/different concept by an African Grey parrot (Psittacus erithacus): learning with respect to categories of color, shape, and material

The parrot answered correctly well above chance with novel objects, and could report which category the objects differed in rather than only that they differed.

1981Science

A brain for all seasons: cyclical anatomical changes in song control nuclei of the canary brain

Song control nuclei grew substantially in spring and shrank afterwards, changing volume by a large fraction across the annual cycle in an adult brain.

1979Science

Can an ape create a sentence?

The great majority of the chimpanzee's signs were prompted: they reproduced signs a teacher had made moments earlier, or were produced while the teacher was signing.

1976Journal of Comparative Neurology

Central control of song in the canary, Serinus canarius

A specific interconnected set of forebrain regions was found to be necessary for normal song, with lesions producing loss or severe degradation of song, and the pathway traced from forebrain to the motor neurons controlling the syrinx.

1970Journal of Comparative and Physiological Psychology

A comparative approach to vocal learning: song development in white-crowned sparrows

Birds tutored during a sensitive period early in life reproduced that dialect as adults.

1965Zeitschrift für Tierpsychologie

The role of auditory feedback in the control of vocalization in the white-crowned sparrow

Birds deafened before singing produced highly abnormal, unstructured song even when they had already heard a tutor.

1958Ibis

The learning of song patterns by birds, with especial reference to the song of the chaffinch Fringilla coelebs

Isolated birds produced song of roughly normal length and pitch range but without the phrase structure and terminal flourish of wild song.

This page is a stop on a longer route

A guided journey reads several subjects in a deliberate order, with an argument for why one follows another. You can join in the middle.

Where to go from here

Each of these follows from something on this page — a relationship in the evidence, a claim people ask about, or the next mechanism along.

How complete this page is, and what it is still missing

NatureHQ publishes its own gaps. This page is at 80% completeness against what we would call a finished subject, and was last reviewed on 2026-09-02. It carries 9 claims and answers 2 mapped search questions.

  • 4 high-priority search intent(s) not yet covered
  • no research from the last few years is attached — check for newer work
  • The mammalian cases are summarised and linked rather than developed; bats and pinnipeds in particular deserve more than a line in a table.
  • The genomic side — convergent gene expression in vocal-learning brains — is not covered.
  • Vocal usage learning in primates is treated only as a contrast, and is a substantial literature in its own right.