Animal Communication

Abstract

Animal communication is a form of behavior: the transmission of information from a signaller to a receiver through an evolved signal that changes the receiver's behavior. This article examines what separates a signal from an ordinary cue, the debate over whether signals transmit information or manipulate receivers, and the problem of reliability that the handicap principle and costly-signalling theory address. It covers functionally referential signals such as the vervet monkey's predator-specific alarm calls and the honeybee waggle dance, and recent evidence for compositional syntax in bird calls. A worked example uses information theory to measure how much a noisy alarm-call system reduces a receiver's uncertainty about which predator is present, and three interactive demonstrations let the reader update a receiver's beliefs, tune signal cost to enforce honesty, and vary call reliability.

Keywords: animal communication, signalling, honest signals, referential signals, compositional syntax

Communication is what turns an animal from a solitary processor of its environment into a node in a social network. When one animal produces a sound, a posture, a colour, or a scent that alters what another animal does, information has crossed the gap between two nervous systems, and the behavior of each is now shaped by the other. The study of animal communication asks how such signals arise, what keeps them reliable when signaller and receiver have conflicting interests, and how far the resulting systems resemble human language.

This article treats animal communication as the branch of behavioral biology that studies evolved signals and the receivers that respond to them. It begins with the distinction between a signal and a mere cue, turns to the central theoretical argument over whether signals carry information or manipulate behavior, and then to the problem that dominates the modern field: why signals are honest often enough to be worth attending to. It closes with the cases that press hardest on the boundary with language, referential alarm calls and the first firm evidence of syntax in a non-human animal.

What Animal Communication Is

Animal communication is the transmission of information between animals by means of signals: traits or actions, produced by one animal, that have been shaped by natural selection to influence the behavior of another, where the receiver's response has itself been shaped by selection to depend on the signal. That double requirement is what separates communication from the ordinary flow of information through the world. A mouse rustling in dry leaves gives away its position to an owl, but the rustle is a cue, an incidental by-product; it was not shaped by selection to inform the owl, and the mouse would suppress it if it could. A peacock's train, by contrast, is a signal: costly, elaborate, and maintained by selection precisely because peahens attend to it (Bradbury & Vehrencamp, 2011). Many signals originate through ritualization, the evolutionary process by which an incidental movement or by-product—an intention movement made before flight, say—becomes exaggerated, stereotyped, and detached from its original context until selection has shaped it into a display whose present function is to inform.

The Medical Subject Headings vocabulary classifies animal communication as a form of animal behavior, placing it among the observable activities of non-human animals, and groups beneath it the sensory channels through which signals travel. Signals are built from almost every physical medium an animal can produce and another can sense: sound, as in birdsong and alarm calls; light and colour, as in the plumage of a bird or the flash of a firefly; chemicals, as in the pheromone trails of ants; posture and movement, as in the threat displays of primates; and even electric fields, as in the discharges of weakly electric fish. The field studies how these signals are produced, how they are received, and above all why they persist when the interests of signaller and receiver so often diverge (Searcy & Nowicki, 2005).

Types of Animal Communication

The Medical Subject Headings vocabulary indexes animal communication as a parent descriptor with two narrower descriptors directly beneath it. These narrower kinds divide the subject by sensory channel and mechanism rather than by a single principle, and they are far from an exhaustive partition: chemical, visual, tactile, and electrical signalling are all major modes of animal communication that the controlled vocabulary files elsewhere or not at all. The two descriptors below are therefore an indexing classification, a controlled vocabulary for retrieving literature, not a theory of how communication is organized, and a single communicative act, an echolocating bat's social call, for instance, may belong to both at once.

The two narrower descriptors MeSH files directly under animal communication are:

- Echolocation — the emission of sound pulses and the use of their returning echoes to locate objects, a sensory and navigational system in bats and toothed whales that also carries a communicative function between individuals. - Animal Vocalization — the production of sound by the vocal apparatus for communication, encompassing birdsong, primate calls, and the broad range of acoustic signalling across taxa.

Neither narrower descriptor is yet a separate article on this site; they are listed here to show how the parent topic is subdivided in the controlled vocabulary. The parent descriptor directly above animal communication in the hierarchy is animal behavior itself.

Signals, Information, and Manipulation

The oldest theoretical argument in the field concerns what a signal fundamentally is. The information view, inherited from the mathematical theory of communication, treats a signal as something that reduces a receiver's uncertainty about the state of the world: before the alarm call the receiver did not know which predator was near, and after it the receiver does. On this reading communication is the transfer of information across a noisy channel, and its success is measured by how much the receiver's uncertainty falls. The waggle dance is the purest case: a returning honeybee performs a figure-eight run whose angle encodes the direction of a food source relative to the sun and whose duration encodes its distance, and hive-mates fly out accordingly, so the dance transmits a quantity of spatial information that can be measured (von Frisch, 1967).

A rival view, stated most sharply by Richard Dawkins and John Krebs, holds that framing signals as information transfer smuggles in a cooperative assumption that evolution does not guarantee (Dawkins & Krebs, 1978). On this reading a signal is better described as manipulation: the signaller acts to change the receiver's behavior in the signaller's own interest, and the receiver responds only because, on average, responding pays. Where the interests of the two conflict, as between a begging chick and a parent, or a courting male and a choosy female, the information a signal carries may be exactly what the signaller benefits from the receiver believing, true or not. The sharpest statement of the problem is that a signal has a function for the signaller and an effect on the receiver, and there is no guarantee the two coincide, so meaning in animal communication is not the shared, conventional meaning of human language (Font & Carazo, 2010). The first demonstration makes the information reading concrete: the reader sets how reliably a signal tracks the world and watches a receiver update its beliefs after hearing the signal.

Honest Signalling and the Handicap Principle

If a receiver responds to a signal, there is a standing temptation for a signaller to produce it dishonestly, to give the alarm call that scatters rivals from food, or to advertise a quality it does not have. Over evolutionary time such deception should make receivers stop attending, and the signal should collapse. That it does not collapse, that so many signals remain reliable enough to be worth heeding, is the problem of signal honesty, and it is the organizing question of the modern field (Searcy & Nowicki, 2005).

The most influential solution is Amotz Zahavi's handicap principle: a signal is kept honest by its cost (Zahavi, 1975). If a display is expensive to produce, and expensive in a way that a low-quality signaller can less afford than a high-quality one, then only high-quality individuals can bear the cost of an extreme signal, and the signal becomes an unfakeable guarantee of quality precisely because it is wasteful. The peacock's train is the textbook case: it is metabolically costly and makes its bearer conspicuous to predators, so a male that can grow and carry an enormous train while surviving thereby demonstrates his vigour. The idea was treated with suspicion for over a decade until Alan Grafen gave it a formal game-theoretic model showing that cost-based honest signalling is an evolutionarily stable strategy, a signalling equilibrium from which neither cheating signallers nor sceptical receivers can profitably deviate (Grafen, 1990). Cost is not the only enforcer of honesty, some signals are reliable because they are physically impossible to fake, as a deep roar that only a large body can produce, but the handicap principle remains the central theoretical result (Maynard Smith & Harper, 2003). Cost is also raised by eavesdropping: signals are routinely detected by receivers they were not directed at—a predator that homes in on a courting male's call, or a rival that monitors a contest between others—so the price a signaller pays includes the risk that unintended receivers exploit the signal. The second demonstration lets the reader adjust the cost of a signal and watch whether an honest equilibrium survives or collapses into cheating.

Referential Signals and the Approach to Language

The signals that press hardest on the boundary with human language are those that seem to refer: to stand for a specific external object or event rather than merely to express the signaller's internal state. The landmark evidence came from vervet monkeys, which give acoustically distinct alarm calls for their three main predators, a leopard, a martial eagle, and a python. Playback experiments by Robert Seyfarth, Dorothy Cheney, and Peter Marler showed that each call alone, in the absence of any predator, elicits the escape response appropriate to that predator: the leopard call sends vervets into trees, the eagle call makes them look up and dash into bushes, the snake call makes them stand and scan the ground (Seyfarth et al., 1980). Because the call controls behavior as if it designated a category of predator, it is said to be functionally referential, a behavioral criterion that brackets the question of what the monkey inwardly intends (Seyfarth et al., 1980).

Functional reference has since been documented across many taxa, and the comparative study of acoustic signalling has grown into a quantitative science that analyses the structure of call sequences with the same tools used for other time series (Kershenbaum et al., 2016). The deepest recent advance concerns syntax. Toshitaka Suzuki and colleagues showed that the Japanese tit combines two of its calls, an alert call and a recruitment call, into a composite sequence, and that the birds respond to the compound with a combination of the two separate responses, scanning for danger and approaching. Crucially, when the artificial order of the two elements is reversed, the response disappears, demonstrating that the meaning depends on the order of combination, the first experimental evidence of compositional syntax in a non-human animal (Suzuki et al., 2016). Whether such cases are true compositionality in the linguistic sense, or a simpler associative concatenation, is actively debated, and the comparison with human syntax must be drawn carefully (Townsend et al., 2018). The third demonstration quantifies the referential idea: the reader varies how reliably each alarm call tracks its predator and watches the information the call carries about which predator is present rise and fall.

Worked Example: Measuring the Information in an Alarm Call

The information view of communication can be made quantitative with Claude Shannon's measure of uncertainty. Suppose a vervet faces three predators whose background probabilities of being the current threat are leopard 0.5, eagle 0.3, and snake 0.2. The receiver's uncertainty about which predator is present, before any call, is the Shannon entropy

H(X) = −Σ p(x) log₂ p(x) = −(0.5 log₂0.5 + 0.3 log₂0.3 + 0.2 log₂0.2) ≈ 1.485 bits.

Now suppose the colony has three alarm calls, one meant for each predator, but the system is noisy: a caller gives the correct call with reliability r, and otherwise gives one of the two wrong calls with equal probability. After hearing a call the receiver updates its belief about the predator, and its remaining uncertainty is the conditional entropy H(X | call). The value of the call is the amount by which it cuts uncertainty, the mutual information

I(X; C) = H(X) − H(X | C).

Table 1 evaluates this for the three-predator distribution above as the reliability r varies from a perfect system down to the point where a call is worthless.

The pattern is the point. A perfectly reliable call removes all 1.485 bits of uncertainty: hearing it, the receiver knows exactly which predator is present. As reliability falls, the information the call carries falls faster than linearly, and at r = 1/3, when a call is no more likely to be correct than any other, it conveys nothing and the receiver is as uncertain after the call as before. At the realistic middle value r = 0.8, hearing the leopard call (the bark) shifts the receiver's belief from a 0.5 prior on leopard to a 0.89 posterior, and the call delivers 0.614 bits, about 41% of the total uncertainty. This is why a noisy but better-than-chance signal is still worth attending to, and why selection on receivers to respond, and on signallers to be reliable enough that responding pays, can sustain a communication system well short of perfect honesty.

Discussion

The study of animal communication is unified by a single tension: signaller and receiver are separate evolutionary agents whose interests overlap but rarely coincide. Every major theme in the field is a consequence of that tension. The information-versus-manipulation debate is a disagreement about whose perspective to privilege, the receiver gaining knowledge or the signaller altering behavior. The problem of honesty arises because conflicting interests tempt signallers to cheat, and the handicap principle resolves it by finding the one currency, cost, that cheats cannot counterfeit. Even the question of reference is, at bottom, a question about what the receiver can reliably extract from a signal that the signaller has its own reasons to give.

Holding these themes together guards against two opposite errors. The first is romantic over-reading: treating every structured animal signal as a word, every call sequence as a sentence, and animal communication as language in miniature. The second is deflationary dismissal: treating all animal signals as mere reflexes empty of content. The field's mature position is between the two. Vervet alarm calls are functionally referential without being words; the Japanese tit's call combinations show order-dependent meaning without demonstrating the open-ended recursion of human syntax. Specifying exactly which language-like property a given system does and does not have, rather than asking the yes-or-no question of whether animals ‘have language’, is what has turned the comparison into a tractable science (Fitch, 2020).

Cognitive Implications

For cognitive psychology, animal communication is the comparative baseline against which the human language faculty is defined. Every claim about what makes human language special, its open-ended productivity, its use of syntax to build complex meanings from simple parts, its detachment from immediate context, is implicitly a claim about what other animals cannot do, and that claim can only be tested by looking. The discovery of order-dependent call combination in a bird does not show that tits have language, but it does show that at least one component once thought uniquely human, combining signals so that order changes meaning, has an independent evolutionary instance, which sharpens the question of what remains distinctive (Townsend et al., 2018).

The communication literature also supplies a hard lesson about inferring minds from behavior. A functionally referential alarm call is defined by what it makes receivers do, not by what the caller is assumed to intend, precisely because the inner states of another creature are not directly observable and the intuitive reading of a signal is so often wrong. This behavioral discipline, describing the information a signal carries and the response it controls before reaching for a mentalistic interpretation, is the same caution that governs the study of animal cognition generally, and it is a useful corrective in any domain where observers are tempted to read rich intention into structured behavior (Fitch, 2020).

Current Directions

One active frontier is multimodality. Most research has isolated a single channel, usually sound, but animals signal with combinations of vocalization, gesture, facial expression, and posture at once, and the integration of these channels may be where the evolutionary roots of language actually lie. Studies of great-ape gesture in particular have been drawn together with vocal research into an argument that language evolved from a multimodal communicative system rather than from voice alone, reframing the search for language origins around how channels combine (Fröhlich et al., 2019).

A second frontier is the rigorous comparative analysis of sequence structure. As recordings accumulate across species, researchers increasingly treat call sequences as formal strings and ask what kind of grammar could generate them, borrowing the hierarchy of formal-language classes from computer science to place each species' system on a common scale (Kershenbaum et al., 2016). The aim is to replace the binary question of whether a species ‘has syntax’ with a graded measure of combinatorial complexity, so that the Japanese tit, the Campbell's monkey, and the human can be located on one axis and the genuine discontinuities identified against a background of shared structure (Townsend et al., 2018).

Common Misconceptions

Any informative stimulus is a signal.

No. A signal is a trait shaped by selection because it informs a receiver, and the receiver's response is shaped to depend on it; an incidental by-product that happens to leak information, such as the sound of prey moving, is a cue, not a signal, and the distinction is foundational to the field (Bradbury & Vehrencamp, 2011).

Communication means the signaller is cooperatively sharing true information.

Not necessarily. Because signaller and receiver are distinct evolutionary agents with partly conflicting interests, a signal can equally be described as the signaller manipulating the receiver's behavior to its own advantage; honesty is an outcome that must be explained, not an assumption (Font & Carazo, 2010; Searcy & Nowicki, 2005).

A referential alarm call is a word.

Only in a limited, behavioral sense. Vervet calls are functionally referential because each controls the escape response appropriate to a predator, but this criterion is defined by effect on the receiver and does not establish that the call carries the conventional, intended meaning of a human word (Seyfarth et al., 1980; Fitch, 2020).

Only humans combine signals into ordered, meaningful sequences.

No longer tenable as stated. The Japanese tit combines two calls into a sequence whose meaning depends on their order, and reversing the order abolishes the response, giving experimental evidence of compositional syntax in a non-human animal, though whether it matches human syntax in full remains debated (Suzuki et al., 2016; Townsend et al., 2018).

Glossary

Alarm call.

A vocal signal given in response to a predator that alters the behavior of receivers, sometimes in a predator-specific, functionally referential way.

Animal communication.

The transmission of information between animals through evolved signals, where both the signal and the receiver's response have been shaped by natural selection.

Compositional syntax.

The combination of discrete signal elements into a sequence whose meaning depends on the order of combination, demonstrated experimentally in the Japanese tit.

Costly signalling.

The theory that a signal remains honest because producing it is expensive in a way only high-quality signallers can afford, formalized as an evolutionarily stable strategy.

Cue.

A stimulus that incidentally carries information about the world but was not shaped by selection to inform a receiver, as distinct from a signal.

Deception.

The production of a signal that misrepresents the state of the world or the signaller to the signaller's advantage, the possibility that makes honesty a problem to explain.

Eavesdropping.

The extraction of information from a signal by a receiver other than the intended one, such as a predator homing in on a courtship call.

Functionally referential signal.

A signal, such as a predator-specific alarm call, that controls receiver behavior as if it designated an external category, defined behaviorally without a claim about intent.

Handicap principle.

Zahavi's proposal that the cost of a signal guarantees its honesty, because only a high-quality individual can bear the cost of an extreme display.

Honest signal.

A signal whose value reliably correlates with the quality or state it advertises, so that attending to it pays the receiver on average.

Mutual information.

The reduction in a receiver's uncertainty about the world produced by a signal, measured in bits as the difference between prior and conditional entropy.

Receiver.

The animal whose behavior a signal is shaped to influence, and whose response to the signal has itself been shaped by selection.

Ritualization.

The evolutionary process by which an incidental movement or cue becomes exaggerated and stereotyped into a dedicated signal.

Signal.

A trait or action produced by one animal, shaped by selection to influence the behavior of another, as opposed to an incidental cue.

Signaller.

The animal that produces a signal, for which the signal has a function in altering a receiver's behavior.

Waggle dance.

The figure-eight movement by which a returning honeybee encodes the direction and distance of a food source, decoded by Karl von Frisch.

Key Researchers

Dorothy L. Cheney

(University of Pennsylvania; 1950–2018). American primatologist whose long-term field studies of vervet monkeys and baboons, conducted with Robert Seyfarth, used playback experiments to establish predator-specific alarm calls and the social knowledge behind primate communication. Wikipedia · Wikidata

W. Tecumseh Fitch

(University of Vienna, Professor of Cognitive Biology). American cognitive biologist whose comparative bioacoustics distinguishes the communicative from the cognitive in the evolution of speech and language, and who insists the language comparison be drawn property by property. ORCID · Scholar · Wikipedia

Karl von Frisch

(University of Munich; 1886–1982). Austrian ethologist who decoded the honeybee waggle dance, showing that an invertebrate encodes the distance and direction of a food source in a symbolic signal; shared the 1973 Nobel Prize. Wikipedia

Peter Marler

(University of California, Davis; 1928–2014). British-American ethologist who founded the modern study of birdsong, discovering song dialects and the sensitive-period template for vocal learning, and co-authored the vervet alarm-call studies. Wikipedia

Robert M. Seyfarth

(University of Pennsylvania, Emeritus Professor of Psychology). American psychologist whose vervet and baboon field experiments, with Dorothy Cheney, provided the canonical evidence for functionally referential alarm calls and the social cognition behind primate communication. Scholar · Faculty · Wikipedia

Toshitaka N. Suzuki

(University of Tokyo, Research Center for Advanced Science and Technology). Japanese behavioral ecologist whose field experiments on the Japanese tit provided the first firm evidence for compositional syntax in a non-human animal. ORCID · Faculty

Amotz Zahavi

(Tel Aviv University; 1928–2017). Israeli evolutionary biologist who proposed the handicap principle, arguing that costly signals are evolutionarily stable precisely because their cost guarantees honesty. Wikipedia

Klaus Zuberbühler

(University of Neuchâtel and University of St Andrews). Swiss primatologist who extended functionally referential and combinatorial analyses of primate vocalizations, bridging animal communication and the evolution of language. ORCID · Scholar · Faculty

Frequently Asked Questions

What is the difference between a signal and a cue?

A signal is a trait or action shaped by natural selection specifically because it influences a receiver's behavior, and the receiver's response has evolved to depend on it. A cue carries information only incidentally, as the sound of a moving animal reveals its position; it was not shaped to inform, and the animal would suppress it if it could (Bradbury & Vehrencamp, 2011).

Do animal signals transmit information or manipulate the receiver?

Both descriptions capture part of the truth, and the disagreement is long-standing. The information view treats a signal as reducing a receiver's uncertainty; the manipulation view stresses that a signaller acts in its own interest and the receiver responds only because responding usually pays. Where interests conflict, the two readings come apart (Font & Carazo, 2010).

Why don't animals simply cheat by giving dishonest signals?

Some do, but cheating is held in check. The handicap principle shows that a signal costly enough that only high-quality individuals can afford its extreme form is unfakeable, and formal models confirm that cost-based honest signalling is an evolutionarily stable strategy; other signals are honest because they are physically impossible to fake (Zahavi, 1975; Grafen, 1990).

What are functionally referential signals?

They are signals that control a receiver's behavior as if they stood for a specific external category. Vervet monkeys give acoustically distinct alarm calls for leopards, eagles, and snakes, and each call alone elicits the escape response appropriate to that predator, meeting a behavioral criterion for reference without a claim about what the caller intends (Seyfarth et al., 1980; Seyfarth et al., 1980).

Is there syntax in animal communication?

There is at least one firm case. The Japanese tit combines an alert call and a recruitment call into an ordered sequence, and the birds respond to the compound with both behaviors; reversing the order abolishes the response, showing that meaning depends on order. Whether this matches human syntax in full is debated (Suzuki et al., 2016; Townsend et al., 2018).

How is the honeybee waggle dance a form of communication?

A forager returning to the hive performs a figure-eight run in which the angle of the straight portion encodes the direction of a food source relative to the sun and its duration encodes the distance. Hive-mates read the dance and fly out accordingly, making it a symbolic signal that transmits measurable spatial information (von Frisch, 1967).

Can the information in a signal be measured?

Yes, with Shannon's entropy. The information a signal conveys is the mutual information between the signal and the state of the world, the amount by which hearing the signal reduces the receiver's uncertainty, measured in bits. A perfectly reliable alarm call removes all uncertainty about the predator; a chance-level call removes none (Kershenbaum et al., 2016).

Does animal communication tell us anything about human language?

It supplies the comparative baseline that defines what is special about language. By locating which language-like properties, reference, order-dependent combination, multimodal integration, appear in other species and which do not, the comparison turns the vague question of whether animals ‘have language’ into specific, testable claims about individual components (Fitch, 2020; Fröhlich et al., 2019).

References