Skip to content
NatureHQ

ability

Animal communication

Animals signal through sound, chemistry, vibration, electricity, movement and touch. A few use signals that point at specific things or specific individuals — and none of it is language.

Two questions get merged whenever animal communication is written about, and separating them makes almost everything clearer. The first is *how does the signal travel* — as sound, as a chemical, as a vibration through the ground, as a dance. The second is *what does it carry*. A honey bee’s waggle dance encodes direction and distance with real precision. A vervet monkey has different alarm calls for leopards, eagles and snakes, and other monkeys run into a tree, look up, or look down accordingly, with no predator present. A dolphin answers to a synthetic copy of its own signature whistle with the voice stripped out. Each of those is a genuine result about content. None of them is language, and the gap between "a signal that designates a category" and "a word" is where most overstatement about animal minds happens.

Developed record · 83% complete · reviewed 2026-08-09

What this page covers

A capability rather than an organism. Covers signalling wherever it has been directly tested — insects, birds, cetaceans, primates and elephants.

Quick facts

Kind
A capability page, assembled from evidence across species
Channels
Sound, chemical, visual, vibrational, electrical, tactile
Referential signals
Demonstrated in vervet monkeys, dolphins and elephants
Language
Not demonstrated in any species

Where this appears

Assembled from the knowledge graph. Each entry carries its own evidence and its own limits.

Two different questions

How a signal travels and what it carries are separate problems, and almost every confused claim about animal communication comes from treating them as one. "Elephants communicate through the ground" is a statement about a channel. It says nothing about what is being communicated, and the honest answer to that second question is usually much narrower.

Much animal communication happens in channels people cannot perceive at all

Established

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

Documented signalling channels include seismic vibration transmitted through the ground, electric fields detected by mechanosensory hairs, chemical trails, and dance encoding direction and distance — none of which is available to unaided human perception.

Who this applies to
the species in which each channel has been directly demonstrated
Studied in
Loxodonta africana, Bombus terrestris, Apis mellifera, Lasius niger
Why we rate it this way, and what the caveats are
EstablishedHigh confidence

Each channel is independently well evidenced — seismic playback in elephants, electric-field detection in bumblebees, the waggle dance confirmed by radar tracking, pheromone trails by direct manipulation.

How far it can be extended

Each channel rests on its own primary study in its own species. The claim is about the range of channels in use, built only from demonstrated cases.

Caveats

  • Detecting a signal is not the same as it being used to communicate; floral electric fields are a cue produced by the flower, not a message from another bee.
  • Channel and content are different questions — knowing how a signal travels says little about what it carries.

Still unanswered

  • How many channels remain undiscovered simply because nobody has looked?

Last reviewed 2026-08-09

The evidence (4 studies)

One caution belongs with the channel list. A bumblebee detecting a flower’s electric field is sensing a *cue* — the flower is not signalling to the bee, it is simply charged. Communication requires a signal produced for its effect on a receiver, and the difference matters when counting what animals talk to each other about.

Words used here
Signal
Something an animal produces because of its effect on another animal. Distinct from a cue, which is information leaking out whether or not anyone benefits.
Referential signal
A signal that reliably designates something specific — a predator type, an individual — rather than only conveying how the sender feels.

Signals that point at things

Some animal calls point at specific things in the world, not just at how the caller feels

Well supported

Good evidence backs this, though some details remain open.

Playback experiments show that certain calls elicit predator-appropriate responses in the absence of any predator, and that some species use learned individually-distinctive labels — indicating signals that designate categories or individuals rather than only expressing arousal.

Who this applies to
the species in which referential or label-like signalling has been directly tested
Studied in
Chlorocebus pygerythrus, Tursiops truncatus, Loxodonta africana
Why we rate it this way, and what the caveats are
Well supportedHigh confidence

The vervet result has been replicated and extended across taxa for four decades, and the playback logic is unusually clean: with no predator present, the response can only come from the call.

How far it can be extended

Assembled from primary playback studies in each species. The claim concerns which kinds of signal exist, so its strength comes from the independent demonstrations rather than from any one.

Caveats

  • A signal that designates a category is not a word, and none of this is language.
  • Whether callers intend to inform anyone is not established by playback designs.
  • Referential and emotional accounts are hard to separate completely; a call can do both.

Still unanswered

  • Do any animals combine signals into structures with their own meaning?
  • How much of what looks referential is learned, and how much is developmentally fixed?

Last reviewed 2026-08-09

The evidence (3 studies)

The playback method is what makes these results trustworthy, and it is worth understanding once. Record a call. Play it back when the thing it refers to is absent. If the animals still respond appropriately, the response came from the call rather than from seeing a leopard. That single design underpins nearly everything known about referential communication.

Where these findings live

  • Elephants

    Individually specific name-like calls

  • Orca

    Pod-specific call repertoires passed down maternal lines

  • Honey bee

    Direction and distance encoded in a dance

Words used here
Playback experiment
Playing a recorded signal in the absence of what produced it, to test whether the signal alone drives the response.

Birdsong — the clearest case of a learned signal

Start with a distinction almost every article about birdsong skips. **Song** is the long, structured, usually seasonal performance — mostly by males in temperate species — used to hold a territory and attract a mate. **Calls** are short, largely innate, used by both sexes all year, for alarm, contact and begging. They develop differently and do different jobs, and hearing every bird sound as song is where most of the confusion begins.

Songbirds learn their song from other birds, during a window early in life

Established

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

White-crowned sparrows raised in acoustic isolation produce abnormal song retaining only broad species structure, while birds tutored during an early sensitive period reproduce the tutor dialect. Tutoring outside that window has little effect.

Who this applies to
white-crowned sparrows, with the pattern general across oscine songbirds
Studied in
Zonotrichia leucophrys, Passeriformes
Why we rate it this way, and what the caveats are
EstablishedHigh confidence

A clean isolation-and-tutoring design, replicated across species for five decades, and the foundation of the entire vocal-learning literature.

How far it can be extended

Vocal learning is established across oscine songbirds, parrots and hummingbirds. The sensitive-period experiment was done in one species; the balance of innate and learned components varies.

Caveats

  • Song and calls are different things: song is long, learned and usually seasonal; calls are short, largely innate and used year-round.
  • Most vocal-learning work covers temperate northern songbirds; female song and tropical duetting are under-studied.
  • Vocal learning is not language, and the two are routinely conflated.

Still unanswered

  • Why has vocal learning evolved in only a few bird groups?
  • How much does female song, long overlooked, change the standard account?

Last reviewed 2026-08-09

The evidence (2 studies)

The isolation experiment is what makes this more than an observation. A sparrow raised hearing nothing still sings — the rough species pattern is inherited — but it sings nothing like its neighbours. Give it a tutor recording during a window early in life and it reproduces that dialect as an adult. Regional song dialects in birds are therefore cultural: passed on by learning, drifting over generations, and different valley to valley.

One bird combines two calls in a fixed order, and the order changes what they mean

Emerging evidence

Real findings exist, but too few or too recent to be settled.

Japanese tits combine an alert note with a recruitment note in a consistent order. Playback in the natural order produced scanning followed by approach; the same notes reversed produced markedly weaker responses.

Who this applies to
wild Japanese titsDo not extend this beyond the taxa listed — the popular version over-reaches.
Studied in
Parus minor
Why we rate it this way, and what the caveats are
Emerging evidenceModerate confidence

The reversal control is strong — the same acoustic material in a different order does not work. Held at emerging because it is one species, one rule, and replication elsewhere has been difficult.

How far it can be extended

Explicitly not generalised. This is one system in one species, and attempts to demonstrate compositional effects elsewhere have had mixed results.

Caveats

  • Two elements and one ordering rule is not an open-ended grammar.
  • This is the strongest syntax candidate outside humans and remains a long way from language.
  • Wild playback offers limited control over the surrounding acoustic environment.

Still unanswered

  • Does compositional combination occur in any other species?

Last reviewed 2026-08-09

The evidence (1 study)

The Japanese tit result deserves its caution as much as its billing. Two calls combined by one ordering rule, where reversing the order weakens the response, is the strongest evidence for syntax outside humans — and it is two elements and one rule. Vocal learning and combination are the two properties birds share with language, and having two of them is not having language.

Related

Words used here
Song versus call
Song is long, learned, usually seasonal and usually male. Calls are short, largely innate, and used by both sexes year-round.
Sensitive period
A window early in life during which a young bird can learn a song. Outside it, the same tutoring has little effect.
Compositional syntax
Combining meaning-bearing signals so that the order changes the message. Demonstrated in one bird, with two elements.

A brain that grows a song and then lets it go

A canary’s song-control brain regions grow every spring and shrink again

Established

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

The forebrain nuclei controlling song production in male canaries change substantially in volume across the annual cycle, growing in the breeding season and regressing afterwards — a finding that contributed directly to the overturning of the view that adult vertebrate brains do not generate new neurons.

Who this applies to
canaries, which re-learn song each year; seasonal change is smaller in age-limited learners
Studied in
Serinus canaria
Why we rate it this way, and what the caveats are
EstablishedHigh confidence

Direct anatomical measurement, replicated widely, and followed by the demonstration of adult neurogenesis that it prompted. This is one of the better-established results in neuroscience.

How far it can be extended

Seasonal change in song nuclei has been documented in several songbird species, though its magnitude varies with whether the species re-learns song annually.

Caveats

  • Canaries are unusual; age-limited learners such as zebra finches show smaller changes.
  • Volume conflates neuron number, cell size and connectivity.
  • Captive photoperiod control exaggerates the seasonal transition.

Still unanswered

  • What is the functional benefit of regressing the circuit rather than maintaining it?
  • How far does adult neurogenesis extend in other vertebrate brains?

Last reviewed 2026-08-09

The evidence (3 studies)

A male canary’s song-control nuclei expand substantially every spring and regress afterwards. The finding mattered far beyond birdsong: the prevailing view was that adult vertebrate brains do not generate new neurons, and a brain region that doubles and halves annually is hard to reconcile with that.

The demonstration of adult neurogenesis in songbirds followed, and eventually reopened the question in mammals including ourselves. It is one of the clearest cases of a fact about an unfashionable animal overturning a general claim about brains.

The parallel to human speech is specific rather than loose — a memorised target, a critical period, a babbling phase, dependence on hearing your own voice — and it is a parallel about *acquisition*, not about meaning. Birdsong has no syntax in the linguistic sense and no compositional content.

Words used here
Neurogenesis
The generation of new neurons. Long thought impossible in adult vertebrate brains.

Why none of this is language

Human language has properties no animal system yet studied possesses: an open-ended vocabulary, combination of units into structures whose meaning depends on the arrangement, and the ability to refer to things absent in time as well as in space. Vervet alarm calls are a closed set. A dolphin signature whistle labels one individual. A waggle dance encodes two numbers.

That is not a way of diminishing them. A bee conveying a precise vector in the dark on a vertical comb is doing something no human can do unaided. The point is that "animal language" imports expectations the evidence cannot meet, and the resulting disappointment is unearned — the actual findings are stranger and better than the headline version.

  • Does any animal combine signals into structures with their own meaning?

    Why it matters: Combination is the property that separates a vocabulary from a language. Suggestive results exist in birds and primates and none is settled.

    What would settle it: Demonstrating that a combined signal means something neither component means alone, with the components independently controlled.

  • Do signalling animals intend to inform anyone?

    Why it matters: A call that reliably produces a useful response need not be produced *in order to* produce it. Almost every popular account assumes intent that no design has tested.

    What would settle it: Evidence that signalling varies with what the sender believes the receiver knows, rather than with the receiver’s mere presence.

  • How much are we missing for lack of the right instruments?

    Why it matters: Seismic elephant communication and bee electroreception were both invisible until someone measured the right channel.

    What would settle it: Systematic surveys of under-instrumented channels across taxa.

The research behind this page

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

You may also find this interesting

What this page is still missing

NatureHQ publishes its own gaps. This record is at 83% completeness against what we would call a finished subject.

  • 5 high-priority search intent(s) not yet covered
  • no popular claim about this subject has been checked yet
  • Birdsong — the largest and best-studied communication literature of all — is entirely absent.
  • Signal combination and syntax-like structure are discussed as open questions without primary research attached.
  • Chemical communication is represented only by ant trail pheromones.

Last reviewed 2026-08-09 · 15 claims · 11 search questions answered on this page