A bird raised where it can hear no other bird still sings — and sings something no wild bird would recognise. The tune has to be learned.
Birds sing mainly to keep rivals out and to attract or hold a mate — both demonstrated by experiment, not assumed. But the sentence hides three things worth knowing: most of what a bird produces is calls rather than song, the song itself has to be learned from an adult, and in most songbird species the female sings too.
Song is one of the very few animal signals that is unmistakably learned, culturally transmitted, and available for anyone with a microphone to record and archive — which is why it has been studied harder than almost any other behaviour in biology, and why it is the best-evidenced case of culture in any animal. What that century of work established is narrower than the popular picture. Song and calls are different categories with different developmental origins; almost everything said about "birdsong" refers to the first. Song is not one thing an individual has by nature: a bird raised where it can hear no adult still sings, and sings something structurally wrong, so the specific pattern has to be acquired. And the two textbook functions — repelling rivals, attracting mates — were established by experiment in a small number of temperate species and are much better supported than the more specific claims layered on top of them about repertoire size and quality. The most important correction to the whole field arrived in 2014. Birdsong research grew up in temperate Europe and North America, where the conspicuous singer is usually a seasonal male, and generalised. A global survey found female song in around seventy per cent of songbird species and reconstructed it as the ancestral condition. The familiar version was never a fact about birds; it was a fact about where people were listening.
Song in the strict sense — the long, structured, usually learned performance — is mainly a songbird thing, and songbirds are about half of all bird species. Other birds vocalise, and a great deal of it is not song.
Often confused with: Calls, which are short, largely innate and used all year by both sexes; Sounds made mechanically rather than vocally, such as a snipe’s drumming tail or a hummingbird’s dive chirp
Quick facts
Song or call?
Different categories — song is long and learned, calls are short and mostly innate
Learned or inherited?
The structure is learned; a rough envelope is not
Do females sing?
In around 70% of songbird species surveyed — and the ancestor did
Best-evidenced animal culture
Songbird vocal traditions, ahead of any primate or cetacean case
The first distinction, and the one that decides whether anything else on this page applies to what you heard this morning.
The short answer
What is the difference between a bird’s song and its calls?
Song is long, structured, usually learned and mostly used in territory and mating. Calls are short, simple, largely innate and used by both sexes all year for alarm, contact, begging and keeping a flock together.
This matters because the two have almost entirely separate literatures, and popular writing collapses them. When someone says birdsong is learned, they mean song: a chaffinch raised in isolation produces abnormal song and perfectly normal alarm calls. When someone says birds sing to attract mates, they mean song again; the contact note a tit gives its flockmates in November is not doing that. The line is a convention rather than a law of nature, and there are sounds that sit awkwardly on it — some calls are partly learned, and Japanese tits combine two call types in an order that changes what listeners do, which is more than "simple" allows for. But as a first cut it is the single most useful thing to hold on to.
The Japanese tit result, which complicates the tidy version
Diagram
Song and calls, side by side
Schematic traces. They show the shape of the structure, not measured frequencies.
The same explanation in words
Two spectrogram panels. On the left, labelled song: four separate traces, each a repeated arch or zigzag, spread across the panel and each a few tenths of a second long — the picture of a structured, repeated phrase. This is the long performance, copied from an adult during a window early in life, and in most temperate species it is seasonal and produced by the male. On the right, labelled calls: six short marks, mostly single ticks or brief flat lines, with one small arch. These are alarm, contact, begging and flight sounds, used by both sexes throughout the year, and they mostly develop without any model to copy. Beneath both panels a note says the boundary is a convention rather than a law: some calls are partly learned, and Japanese tit calls combine in an order that changes how listeners behave, which the word "simple" does not cover.
The distinction, and where it breaks down.
Almost everything said about "birdsong" applies only to song, which is a long learned performance — not to calls, which are short, largely innate and used by both sexes all year.
Established
Specialists would state this without hedging. Multiple independent lines of evidence agree.
Avian vocalisations are conventionally divided into song — structurally complex, typically learned, usually produced in reproductive and territorial contexts — and calls, which are short, structurally simple, largely innate and used year-round for alarm, contact, begging and flight coordination. The developmental and functional literatures for the two categories are largely separate.
Who this applies to
A distinction used across birds, drawn most sharply in songbirds and blurrier in some other groups.
Studied in
Aves
Why we rate it this way, and what the caveats are
EstablishedHigh confidence
The categories are a convention rather than a discovery, and the substantive part — that song is typically learned and most calls are not — is supported by isolation and deafening experiments across many species.
How far it can be extended
The song/call division is applied throughout the ornithological literature and the developmental difference — song learned, most calls not — has been demonstrated separately in many songbird species.
Caveats
The boundary is a convention and there are intermediate vocalisations that sit awkwardly on either side of it.
Some calls are partly learned, and in a few species call learning is substantial.
Structural simplicity is not functional simplicity — alarm calls carry information about predator type and urgency.
Still unanswered
How much of the vocal repertoire outside song is learned in species that have not been tested, which is most of them.
Shows that calls are not the simple side of the division in every respect: tit calls combine according to an order rule that changes the response.
And a third category that gets counted as song by mistake.
Some of the loudest sounds birds make on purpose are not made with the voice at all. A male Anna’s hummingbird’s dive chirp is a tail feather fluttering in fast air.
Established
Specialists would state this without hedging. Multiple independent lines of evidence agree.
Sonation — non-vocal sound production by feathers or other structures — occurs across several bird groups. In Calypte anna, the display-dive sound is produced by aeroelastic flutter of the outer rectrices, demonstrated by feather removal and by wind-tunnel testing of isolated feathers.
Who this applies to
Demonstrated for one hummingbird’s dive sound; sonation occurs more widely and by varied mechanisms.Do not extend this beyond the taxa listed — the popular version over-reaches.
Studied in
Calypte anna
Why we rate it this way, and what the caveats are
EstablishedHigh confidence
The wind-tunnel test with a detached feather removes the animal from the explanation entirely, which is about as decisive as a mechanism demonstration gets.
How far it can be extended
Other hummingbird dive sounds are produced differently, and some are genuinely vocal, so the mechanism must be established species by species.
Caveats
The wind tunnel reproduces the airflow, not the intact tail, and feather interactions are harder to isolate.
What the sound is for, and whether females attend to it, is not tested here.
Sonation mechanisms differ between species and cannot be inferred from this one.
Still unanswered
How often sounds recorded in the field as vocalisations are in fact mechanical.
Clark and Feo, 2008 · Proceedings of the Royal Society B
Feather removal plus the isolated-feather wind-tunnel test.
How we know
Putting a single tail feather in a wind tunnel
Is the loud chirp at the bottom of a hummingbird’s display dive a sound the bird sings, or a sound its body makes?
Male Anna’s hummingbirds were filmed at high speed with synchronised sound as they performed their display dives, so the exact moment of the chirp could be located in the dive. Individual tail feathers were then removed or altered on some birds, and the dives repeated. Finally, isolated feathers — no bird attached — were mounted in a wind tunnel and exposed to the airspeed a bird reaches at the bottom of a dive.
What happened
Removing the outer tail feathers removed the chirp. An isolated outer tail feather held in fast-moving air produced the sound on its own.
What it shows
The sound is aeroelastic flutter of a feather, not a vocalisation — it is made by the bird’s tail rather than its voice box. That distinction matters well beyond hummingbirds: a loud, tonal, precisely timed sound produced during a display is exactly the kind of thing that gets counted as song without anybody checking where it comes from.
What it does not show
It says nothing about what the sound is for, or whether females attend to it. Other hummingbirds produce dive sounds in other ways and some of them are genuinely vocal, so this is not a general result about the family. And a detached feather in a wind tunnel reproduces the mechanism without reproducing the intact tail.
The controls — what makes this evidence rather than a story
Removing specific feathers rather than all of them, so the sound can be traced to a particular structure.
The wind tunnel, which removes the bird entirely and leaves only the mechanical hypothesis to be tested.
High-speed film with synchronised audio, fixing the sound to a moment rather than to an impression.
The bird vocal organ, sitting where the windpipe divides rather than in the throat. Its two halves can be controlled separately, which is why some birds can sing two notes at once.
Sonation
A sound made deliberately by some part of a bird other than its voice — feathers, wings, a bill. Loud, purposeful, and not a vocalisation.
Two functions have been demonstrated by manipulation. Most of what is layered on top of them has not.
The standard answer — birds sing to defend a territory and to attract a mate — is right, and it is worth knowing how it was established, because "obviously that is what it is for" is not evidence. A bird holding a territory sings; that could be the singing doing the work, or it could be something a territory-holder happens to do. The way to tell them apart is to take the bird away and leave the song.
How we know
Removing the bird and leaving the song behind
Does song actually keep rivals out of a territory, or is singing just what a bird holding a territory happens to do?
Resident male great tits were removed from their territories in a woodland, creating a set of good, empty places for a new bird to move into. Some of those empty territories then had loudspeakers broadcasting recorded great tit song. Others had loudspeakers broadcasting a control sound. Others were left silent. Observers recorded how long each territory stayed empty before a new male settled in it.
What happened
Territories broadcasting great tit song were re-occupied considerably more slowly than silent or control-sound territories.
What it shows
The song alone deters. With the resident gone there is nothing else there to do it, so the delay is attributable to the sound and to nothing else. This is the design that converts the standard assumption — birds sing to hold territory — from a plausible story into a demonstration, and it is the same logic as any playback experiment, run at the scale of a wood.
What it does not show
A loudspeaker sings from one spot and never replies, and a real resident moves, answers and escalates, so the size of the effect here is a floor rather than a measure of what a live bird achieves. It tests the deterrent function only and says nothing about attracting a mate, which is the other half of what song is for. One temperate species with a small repertoire, in one habitat.
The controls — what makes this evidence rather than a story
Silent vacated territories, establishing how fast re-occupation happens with nothing at all.
A control sound, so any effect cannot simply be a response to a loudspeaker or to noise.
Territories assigned across treatments within the same wood and the same season.
Song holds a territory by itself. Remove the bird, leave a loudspeaker playing its song, and rivals stay away for longer.
Established
Specialists would state this without hedging. Multiple independent lines of evidence agree.
Playback of conspecific song into experimentally vacated territories delays re-occupation relative to silent and control-sound treatments, demonstrating that song functions in territorial exclusion independently of the presence of the signaller.
Who this applies to
Demonstrated experimentally in the great tit; the design has been repeated in a number of other territorial songbirds with comparable results.
Studied in
Parus major
Why we rate it this way, and what the caveats are
EstablishedHigh confidence
A field manipulation with the right controls, producing a clear effect on a behaviour that matters. This is the design that turned an assumption into a finding.
How far it can be extended
Territorial song and playback-induced exclusion are reported across many temperate songbirds, though the original loudspeaker-occupation design has been run in relatively few.
Caveats
A loudspeaker sings from one point and never replies, so the measured effect is a floor rather than an estimate of what a live resident achieves.
The design tests exclusion only, and says nothing about attracting a mate.
Territoriality varies enormously across birds; this is not a general fact about song.
Still unanswered
How much of the deterrent effect depends on recognising a particular neighbour rather than on hearing any conspecific.
Places the result in the wider evidence for territorial function and separates it from the correlational claims about repertoire size.
The mate-attraction half is supported by a different kind of evidence — females approaching playback, unmated males singing more than mated ones, song rate dropping once a pair forms — and is solid at that level. Where the ground gets softer is the layer above: the widely repeated claims that a larger repertoire wins more mates, or that song complexity signals a male’s quality. Those rest largely on correlations, and correlations in this area admit several readings. Older males often have both larger repertoires and better territories.
The best answer to why song would be worth listening to at all — and the point at which it stops being demonstrated.
A bird that was underfed as a nestling sings a measurably poorer song as an adult, and cannot make it up later.
Well supported
Good evidence backs this, though some details remain open.
Nutritional stress during the sensitive period for song learning produces degraded song copying in adulthood and, in several species, reduced volume of song-control nuclei. The effect persists after the stressor is removed, which is the basis of the developmental stress hypothesis for song as an honest signal.
Who this applies to
Demonstrated in several songbird species by nutritional manipulation during development.
Studied in
Passeri
Why we rate it this way, and what the caveats are
Well supportedModerate confidence
The effect of early stress on song is well replicated. The further step — that receivers use song to assess developmental history — is much less well evidenced, and the popular version takes it for granted.
How far it can be extended
Independent nutritional-restriction experiments in different songbird species produce the same direction of effect on song quality.
Caveats
Evidence that stress degrades song is stronger than evidence that females attend to the difference.
Effect size and even direction vary between species and between the specific song features measured.
Laboratory nutritional restriction is a blunt stand-in for the range of stressors a wild nestling meets.
Still unanswered
Whether receivers discriminate on the features that early stress actually affects, rather than on features that happen to be easy to measure.
Beecher and Brenowitz, 2005 · Trends in Ecology & Evolution
Reviews the functional consequences of song learning in the field, where selective learning from future neighbours competes with signal-quality explanations.
The finding that made birdsong a subject rather than a pleasant noise.
Raise a chaffinch where it can hear no other chaffinch and it still sings. The song is roughly the right length, roughly the right pitch, and has none of the phrase structure or closing flourish that makes a chaffinch recognisable. That single result splits song into two parts with different origins — a rough inherited envelope, and a specific learned pattern — and everything else on this page and the next three depends on it.
How we know
Raising a bird somewhere there is nothing to copy
Is a bird’s song inherited, or does it have to hear another bird to produce it?
Chaffinches were hand-reared from before they could have memorised anything. Some were kept alone in sound-proofed rooms and heard no chaffinch at all. Some were kept alone but played recorded chaffinch song at various ages. Some were raised in groups of birds that had all been isolated together, so they could hear each other and nothing correct. When each bird began to sing as an adult, its song was recorded and put through a sound spectrograph — then a new instrument — so that the shapes could be laid side by side and compared rather than described from memory.
What happened
Isolated birds sang. Their songs were roughly the right length and in roughly the right frequency range, and had none of the phrase structure or the closing flourish of a wild chaffinch. Birds that heard recordings early reproduced the detail of what they had heard. The isolates raised together converged on a shared song that was wrong in the same way.
What it shows
Song has two parts with different origins. The rough envelope — how long, how high — comes without any model. The specific pattern that makes it recognisable as this species in this place does not, and has to be acquired from another bird. That distinction is the whole basis of vocal learning as a category, and it is why regional dialects can exist in birds at all.
What it does not show
It does not show that this is how wild chaffinches develop song, only what a bird can and cannot do when almost everything has been taken away. A soundproof box is a far poorer world than a hedge, and later work found that a live tutor produces learning at ages where a loudspeaker produces none — so the sensitive period this design measures is partly a fact about loudspeakers. It also says nothing about other species: the balance between inherited and learned components varies enormously across songbirds.
The controls — what makes this evidence rather than a story
Rearing from before any exposure was possible, so a bird cannot have learned something before the experiment started.
Recorded tutoring at different ages, which turns "does it learn?" into "when can it learn?".
The group of isolates, which separates hearing nothing at all from hearing something wrong.
A songbird raised where it can hear no other bird still sings — but it sings something structurally wrong, because the specific pattern has to be learned from an adult.
Established
Specialists would state this without hedging. Multiple independent lines of evidence agree.
Acoustic isolation during development produces abnormal song in songbirds tested: broad species-typical features such as duration and frequency range are retained while phrase structure, syllable morphology and terminal elements are degraded or absent. Normal song development requires exposure to conspecific song during a sensitive period.
Who this applies to
Songbirds. Demonstrated in a number of species; the balance of inherited and learned components varies considerably between them.
Studied in
Passeri
Why we rate it this way, and what the caveats are
EstablishedHigh confidence
Multiple independent laboratories, multiple species, a large effect, and a clean manipulation — removing the model — with a consistent result over sixty years.
How far it can be extended
Isolation-rearing experiments have been run independently in chaffinches, white-crowned sparrows, swamp sparrows, zebra finches and others, with the same qualitative outcome.
Caveats
Acoustic isolation is a far poorer environment than anything wild, so this bounds what is necessary rather than describing normal development.
A live tutor produces learning at ages where a loudspeaker does not, so sensitive-period estimates from tape tutoring are narrower than the bird’s actual capacity.
How much structure survives isolation differs sharply between species.
Still unanswered
What the inherited component actually consists of, given that the errors isolate birds make are biased rather than random.
The correction that reversed a default, and the reason it took until 2014.
The short answer
Do female birds sing?
Yes — in around seventy per cent of the songbird species that have been surveyed, and the common ancestor of songbirds almost certainly did. Male-only song is a repeated loss rather than the starting point.
The reason "only males sing" became a textbook fact is not that anybody tested it and got that answer. Song research grew up in temperate Europe and North America, where male-only seasonal song genuinely is common, and generalised from the birds outside the window. Most songbirds live in the tropics, where female song is ordinary and where hardly any of the research was done. When somebody finally scored a thousand species from around the world and mapped the trait onto the family tree, the default flipped. It is a clean example of how a sampling bias becomes a fact — and the residual error almost certainly runs the same way, since observers who did not expect female song were the least likely to write it down.
A thousand species, and an ancestral reconstruction
Several textbook facts about birdsong are facts about temperate northern songbirds, because that is where nearly all of the research was done.
Established
Specialists would state this without hedging. Multiple independent lines of evidence agree.
The foundational birdsong literature is heavily concentrated on temperate Northern Hemisphere passerines, in which seasonal male-only song is common. Generalisations about the sex, seasonality and function of song derived from this sample do not hold across the songbird radiation, most of which is tropical.
Who this applies to
A statement about the evidence base for birdsong research rather than about birds themselves.
Studied in
Passeri
Why we rate it this way, and what the caveats are
EstablishedHigh confidence
The geographic concentration of the literature is a matter of record, and the specific generalisations it produced have been directly contradicted by global sampling.
How far it can be extended
The bias is documented in reviews of the field and is the explicit motivation for the global female-song survey and the duetting literature.
Caveats
Naming the bias does not make the temperate findings wrong within their scope — the isolation and playback results are robust.
Tropical song research remains much thinner, so what replaces the old generalisations is often "it varies" rather than a new rule.
Still unanswered
How many other song generalisations would fail if the same global sampling were applied to seasonality, repertoire size and learning mode.
Not a primate, not a whale. A sparrow — and mostly because song leaves a recording.
The strongest case for culture in any animal is not a primate or a whale. It is a sparrow.
Well supported
Good evidence backs this, though some details remain open.
Songbird vocal traditions satisfy the criteria for cultural transmission more completely than any other documented animal tradition: social acquisition is demonstrated experimentally, geographic structure is mapped, historical change is traceable in recordings, and the transmission process itself has been manipulated in the laboratory and in the wild.
Who this applies to
A comparative judgement about the strength of the evidence, not a claim that birds are more cultural than other animals.
Studied in
Passeri
Why we rate it this way, and what the caveats are
Well supportedHigh confidence
This is a statement about evidence rather than about animals, and the asymmetry is stark: song can be recorded, archived, measured and experimentally manipulated in ways no primate tool tradition can.
How far it can be extended
The experimental base spans many songbird species and several independent methods — isolation, tutoring, cross-fostering, seeded traditions and longitudinal recording.
Caveats
Song is over-represented partly because it is easy to record, so the comparison is partly about tractability rather than about the animals.
The criteria for calling a behaviour cultural are contested; a different set would rank the cases differently.
A signal is a narrower thing than a technology, and birdsong culture does not accumulate complexity the way human traditions do.
Still unanswered
Whether non-vocal avian traditions would look comparably strong if they were as easy to record.
Culture manipulated directly: a tradition destroyed and rebuilt across generations under controlled conditions.
How we know
Starting a colony with birds that sing the wrong song
If a population’s song is destroyed, does it drift away for good, or does it come back?
Zebra finches raised in isolation sing abnormally. Those isolates were used as the tutors for a second generation, whose birds tutored a third, and so on down several generations — both in one-to-one tutoring, where a single pupil had a single wrong model, and in a small breeding colony founded entirely by isolates and left to run. At every generation the song was measured against the wild-type song on a set of acoustic features.
What happened
Song moved towards wild-type at every generation. Within three to four generations the colony was singing song that could not be statistically distinguished from wild-type on the features measured.
What it shows
Copying errors are not random. Learners systematically fail to reproduce certain features of a wrong model, and they fail in a consistent direction, so running the culture forward reassembles the species song from a degraded start. Biology and culture are not competing explanations here: the culture carries the song and an inherited bias decides where it settles.
What it does not show
"Indistinguishable on the measured features" is not identity, and the features chosen bound the claim. It also happened in a cage with no predators, no competition and no female choice, in a domesticated species with an unusually stereotyped song. Whether an open-ended learner with a large repertoire would reconstruct anything comparable is untested.
The controls — what makes this evidence rather than a story
Founding stock with a known, measured deficit rather than an assumed one.
Two designs run in parallel — controlled one-to-one tutoring and a free colony — so the result is not an artefact of either.
Wild-type song measured on the same features throughout, as the fixed comparison.
Two questions with unusually clear popular answers and unusually murky real ones.
Two things people notice about song without needing any biology: it is loudest at dawn, and it sounds different in a city. Both have well-known explanations in circulation, and both explanations are more confident than the evidence behind them. They have pages of their own because the honest answers take more than a sentence.
Six decades in which song went from an instinct, to a learned skill, to a piece of culture — with two of the reversals coming from noticing what nobody had looked at.
1958
Landmark experiment
A chaffinch raised alone sings the wrong song
Chaffinches hand-reared in acoustic isolation produced song of roughly normal length and pitch with none of the phrase structure of wild song, while birds played recordings early in life reproduced their detail. The sound spectrograph — new at the time — is what made the comparison possible at all.
Song varies by valley, and the variation is learned
White-crowned sparrow populations a few kilometres apart around San Francisco Bay were found to sing consistently different variants, sharp enough to place a recording — and birds raised without local song did not develop them. The word "dialect" entered animal behaviour here.
Sparrows deafened after hearing a tutor but before singing produced song as abnormal as birds that had heard nothing. Deafening after the song had settled left it intact. Memorising the model and matching your own output to it are separate processes.
Changes how the 1958 result reads
The isolation result had been read as "a bird needs to hear an adult". This split that requirement in two: it needs to hear an adult, and then it needs to hear itself — and failing either produces the same broken song.
Lesion and tracing work in canaries identified an interconnected network of forebrain nuclei required for normal song, with a pathway running down to the muscles of the syrinx. It made birdsong a model system for the whole of motor learning.
And those regions turn out to change size every year
The song-control nuclei were measured across the annual cycle and found to be substantially larger in the singing season. The adult vertebrate brain was supposed to be structurally fixed; this is the result that opened adult neurogenesis as a field.
Changes how the 1976 result reads
The 1976 anatomy had been read as a fixed circuit underlying a fixed behaviour. Measuring the same regions across a year showed the circuit itself is seasonal, which changed what kind of thing the song system was taken to be.
The sensitive period was partly an artefact of loudspeakers
Every published estimate of the window for song learning had been made by playing tapes. With a live bird as tutor instead, young sparrows learned at ages where tapes produced nothing — and some copied the wrong species entirely.
Changes how the 1965 result reads
The deprivation experiments had established what a bird cannot do without a model, and their sensitive-period boundaries were being read as developmental facts. A live tutor widened the window and weakened the species filter, which makes those boundaries a measurement of the stimulus as much as of the bird.
Adult zebra finches given a slightly delayed echo of their own song lost it over weeks, and largely rebuilt it once normal hearing returned. Corrupting the feedback proved far more destructive than removing it.
Changes how the 1965 result reads
That a deafened adult keeps singing had been read as evidence that crystallised song becomes a fixed motor programme. It does not: it is actively maintained against a stored target, and giving the comparison a wrong answer takes the song apart.
A song culture rebuilds itself from birds that never heard one
Zebra finches raised in isolation were used to tutor a second generation, and that generation a third. Within three or four generations the colony was singing wild-type song. Copying errors are not random; they lean towards the species song.
Changes how the 1958 result reads
An isolate’s abnormal song had been the evidence that the pattern is cultural rather than inherited. Running the culture forward showed the inherited component is doing far more than setting a length — it biases the errors, and the bias reassembles the song.
Female song turns out to be the ancestral condition
Over a thousand songbird species were scored for female song from sources worldwide and mapped onto the family tree. Female song appeared in around seventy per cent, and in the common ancestor. Male-only song is a repeated loss.
Changes how the 1964 result reads
The founding work, and nearly everything after it, was done on temperate northern species where the conspicuous singer is a seasonal male — and generalised. Global sampling inverted the default: what the field had treated as the normal case is the exception, and its explanations for song function were built on it.
A song variant crosses a continent, tracked as it happens
Twenty years of recordings, many made by amateur birders and deposited in public archives, showed a new ending to white-throated sparrow song spreading east across Canada and replacing the traditional one across most of the range.
What does the inherited part of song actually consist of, given that isolate birds make errors that lean in a consistent direction rather than randomly?
Why it matters: It is the difference between "birds inherit a rough template" as a metaphor and as a mechanism. The de novo culture experiment shows the bias exists and does not say what it is.
What would settle it: Identifying the perceptual or motor constraints that make particular acoustic features easier to produce or to memorise, and predicting isolate error from them.
How much of what is known about song learning is really about the zebra finch?
Why it matters: Most of the mechanism has been worked out in one domesticated species with a short song learned once, kept in a cage. Species with large repertoires and open-ended learning may work differently in ways nobody has looked for.
What would settle it: Running the same developmental and feedback manipulations in an open-ended learner with a large repertoire.
Do listeners actually use the song features that early-life stress affects?
Why it matters: The developmental stress hypothesis is the best account of why song would be worth attending to. The half that is well evidenced is that stress degrades song; the half that is assumed is that anybody is reading it.
What would settle it: Playback and choice experiments using songs that differ specifically on the features early stress affects, rather than on features that are convenient to measure.
How complete this page is, and what it is still missing
NatureHQ publishes its own gaps. This page is at 95% completeness against what we would call a finished subject, and was last reviewed on 2026-09-02. It carries 15 claims and answers 23 mapped search questions.
Repertoire size and song complexity as signals are summarised rather than treated properly; the correlational literature is large and deserves its own treatment.
Non-songbird vocalisation — the calls of seabirds, waterfowl and raptors — is barely covered here.
The syrinx itself, and how two-voice production works, is named and not explained.