Animal Vocalization

Abstract

Animal vocalization is a form of animal communication: the production of sound signals by which animals convey information to others. This article treats vocal behavior as a window onto cognition. It explains how vocal signals are produced—a vibrating source, the larynx or the avian syrinx, filtered by the vocal tract so that formants become honest cues to body size; how songbirds learn their songs by matching their output to a memorized sensory template through a sensitive period, in a way that parallels human speech acquisition; and how some calls function referentially, labeling external events such as particular predators. A worked example converts vocal-tract length into formant spacing, and three interactive demonstrations let the reader shape a vocal tract, drive a juvenile song toward its template, and trigger referential alarm responses.

Keywords: animal vocalization, vocal learning, birdsong, referential signaling, source-filter theory

A chorus of frogs, the dawn song of a thrush, the alarm bark of a monkey, the whistle of a dolphin: animals fill the world with sound they make on purpose, and that sound carries information to whoever is listening. Vocalization is the most studied channel of animal communication because it is easy to record, it varies in ways that map onto what the animal is doing, and in a few remarkable groups it is learned—shaped during development against a remembered model in a process that looks startlingly like the way a human child learns to speak.

This article treats animal vocalization as a problem in cognitive science rather than acoustics alone. It begins with what counts as a vocalization and why the vocal channel is special, turns to the physics of how the sound is produced and what a listener can read from it, then to the learning that builds a songbird's song, and finally to the question of what calls mean—whether an alarm call labels a leopard the way a word does. Throughout, the animal's vocal behavior is used to infer the computations its nervous system must be performing.

What Animal Vocalization Is

A vocalization is a sound an animal produces with a dedicated organ—the mammalian larynx, the bird's syrinx, or an analogous structure—and uses as a signal. The Medical Subject Headings vocabulary places it under animal communication, and the grouping is natural: a vocalization is communication carried on the acoustic channel, which propagates around obstacles, works in darkness, and reaches many receivers at once, at the cost of being fleeting and easy to locate. What makes vocal behavior especially informative to a psychologist is that its form varies systematically with the sender's state, identity, and circumstance, so the signal is a readable trace of an internal process.

Most animals produce their calls without ever having heard a model: a dove reared in isolation still coos normally, because the motor pattern is specified developmentally rather than copied. A small set of lineages are different. Songbirds, parrots, and hummingbirds among birds, and cetaceans, pinnipeds, bats, elephants, and humans among mammals, acquire at least some of their vocal signals by imitation—a capacity called vocal production learning, and the rarest and most cognitively demanding form of vocal plasticity (Janik & Slater, 1997). The distribution of this ability across the tree of life, and whether it forms a sharp dichotomy or a graded continuum, is one of the organizing questions of the field (Jarvis, 2019). The founding demonstration that a vocalization could be culturally transmitted came from Peter Marler and Masakazu Konishi's generation: white-crowned sparrows raised in different regions sing distinct dialects that a young bird learns from the adults around it, not from its genes (Marler & Tamura, 1964).

How Vocal Signals Are Produced

The acoustics of nearly every vertebrate voice are captured by source-filter theory, carried into animal research from human phonetics (Fitch, 2000). A vibrating source—the vocal folds of the mammalian larynx, or the paired membranes of the bird's syrinx at the base of the trachea—chops the airstream into a periodic buzz whose repetition rate is the fundamental frequency, heard as the pitch of the call. That buzz then passes through the vocal tract, the tube of throat, mouth, and nasal passages, which resonates at certain frequencies and damps others. The resonances it imposes are the formants, and they, rather than the pitch, carry much of the information a listener can use.

The decisive point for cognition is that the formants depend on the length of the vocal tract, and vocal-tract length scales with body size. A longer tract resonates at lower and more closely spaced formants, so the spacing between formants—the formant dispersion—is an honest acoustic cue to how big the caller is, one that is hard to fake because it is anchored in anatomy (Taylor & Reby, 2010). Red deer stags lower their formants by retracting the larynx during roars, exaggerating apparent size; many species attend to formant spacing when judging a rival or a mate. The source and the filter are to a large degree independent—an animal can change pitch without changing formants and vice versa—which is exactly why the two convey different things, pitch tracking arousal and vocal effort while formants track size and identity. The birds complicate the picture with the syrinx, whose two independently controlled sides let a songbird sing two notes at once and switch between them faster than any human could (Nottebohm, Stokes, & Leonard, 1976). The demonstration lets the reader lengthen and shorten a vocal tract and watch the formants slide, reading off the body size a listener would infer.

Vocal Learning and the Song Template

In the species that learn their songs, acquisition follows a developmental program with a shape that recurs across groups. A young songbird first passes through a sensory phase, during which it listens to adult song and stores a memory of it—the sensory template—without yet producing anything like it. The template can be acquired during a restricted window early in life, a sensitive period after which exposure has far less effect. The bird then enters a sensorimotor phase: it begins to vocalize, first with the rambling, variable subsong that resembles an infant's babbling, then with plastic song that gradually tightens, matching its own output to the stored template by ear until the song crystallizes into the stereotyped adult form (Marler, 1997).

The hinge of the whole process is auditory feedback. Masakazu Konishi showed that a young bird deafened before it sings never develops normal song, even though it had heard adult song and formed a template, because it cannot compare what it produces to what it remembers (Konishi, 1965). A bird deafened only after its song has crystallized is far less affected, which is how the sensitive period for production was first delineated. This architecture—memorize a target, then use self-hearing to drive motor output toward it—is the reason birdsong became the leading animal model for human speech acquisition, which shares the babbling phase, the dependence on hearing oneself, and a sensitive period (Doupe & Kuhl, 1999). The learning is supported by a dedicated forebrain circuit, discovered by Fernando Nottebohm, in which a premotor pathway drives the song while a separate basal-ganglia loop generates the variability that trial-and-error learning needs (Brainard & Doupe, 2002). The demonstration below lets the reader advance a juvenile through the developmental phases and watch its variable output converge on the template.

Referential Signaling

Does a call ever mean something in the way a word does? The question turned from philosophy to experiment with Dorothy Cheney, Robert Seyfarth, and Peter Marler's study of vervet monkeys, which give acoustically distinct alarm calls to leopards, eagles, and snakes. Playing a recorded call through a hidden speaker, with no predator present, elicited the response appropriate to that predator: the leopard call sent monkeys up into trees, the eagle call made them look up and scan the sky, the snake call made them stand and search the ground (Seyfarth, Cheney, & Marler, 1980). The call alone, divorced from the animal's fear and from the predator itself, was enough to specify a class of external event and the right action—the hallmark of a functionally referential signal, an alarm call that behaves less like an involuntary scream and more like a label.

Later work pressed on whether the monkeys respond to the acoustic form of the call or to what it denotes. Klaus Zuberbuhler, again with Cheney and Seyfarth, used habituation designs showing that when two different calls predict the same danger, hearing one call repeatedly dampens the response to the other, as if the listener tracks the underlying referent rather than the sound (Zuberbuhler, Cheney, & Seyfarth, 1999). Functional reference remains a carefully bounded claim: it describes what the signal does for sender and receiver, not a demonstration that the animal intends to inform or holds a concept in the human sense. But it establishes that a vocalization can carry information about the world and not merely about the caller's arousal, which is the minimal property any precursor to word meaning must have. The demonstration lets the reader issue each alarm call and watch the receivers select the matching escape.

Worked Example: Reading Body Size From Formants

Source-filter theory makes a quantitative prediction a listener could exploit. Modeling the vocal tract as a uniform tube closed at the vocal folds and open at the lips, its resonances fall at

F_n = (2n − 1) c / (4L),

where c is the speed of sound in the warm, humid air of the tract (about 350 metres per second), L is the vocal-tract length, and n indexes the formants. The spacing between successive formants—the formant dispersion—is therefore

ΔF = c / (2L),

a quantity that depends only on tract length, and so on body size, and not on the pitch the animal happens to be using. Table 1 evaluates it across a range of vocal-tract lengths.

Two lessons follow from the arithmetic. First, formant dispersion falls as tract length rises, so a listener who can measure the spacing of formants has a direct read on the caller's size—and because the relationship is fixed by the physics of a resonating tube, it is an honest signal that a small animal cannot easily counterfeit, which is why size-related formant cues are so widespread (Taylor & Reby, 2010). Second, the dispersion is independent of the fundamental frequency: a caller can raise or lower its pitch at will, but it cannot shrink the resonances of its vocal tract without physically shortening it, which is precisely why the deer's larynx-lowering trick to fake a longer tract is so striking (Fitch, 2000). The SourceFilterDemo above is built on exactly this equation: every formant it draws is (2n − 1) c / 4L for the tract length selected.

Discussion

The thread tying these sections together is that a vocalization is a behavior shaped by competing demands—on the sender to encode its state, size, or situation, and on the receiver to decode them—and that the acoustic form of the signal is the compromise. Source-filter production explains why some features of a call are honest and others free to vary; vocal learning explains how a signal can be transmitted culturally and diverge into dialects; functional reference explains how a call can come to stand for something in the world. Seen together, they show that the vocal channel supports a spectrum of signals, from the wholly innate and involuntary to the learned, flexible, and quasi-symbolic (Janik & Slater, 1997).

That spectrum is why vocalization has become a comparative probe into the evolution of language. Rather than ask whether any one species has language, researchers decompose the human faculty into component capacities—vocal production learning, a sensitive period, dedicated forebrain circuitry, referential signaling—and ask which other animals share each, and whether the shared traits arose once or evolved convergently. On this view the songbird's crystallized song, the vervet's alarm call, and the human word are not rungs on a single ladder but different combinations of separable traits, distributed across the tree of life in a pattern that itself is the evidence (Jarvis, 2019).

Cognitive Implications

For cognitive psychology, animal vocalization offers tractable models of processes that are hard to study directly in humans. Song learning is the clearest case: because the behavior is quantifiable, the circuitry accessible, and the developmental timetable fixed, it has become a reference system for sensorimotor learning in general—how a nervous system acquires a target by perception, generates variable output, and uses feedback to drive that output toward the target, all of which recur in human speech and skill acquisition (Doupe & Kuhl, 1999). The parallels are specific, not vague: a babbling-like practice phase, a dependence on hearing oneself, a critical window, and a division of labor between a circuit that executes the learned motor pattern and one that supplies exploratory variability.

Comparative neurobiology sharpens the lesson by distinguishing the kinds of vocalization a brain can produce. Innate calls, volitional control over whether and when to call, and genuine vocal production learning depend on partly different circuits, and mapping which taxa possess which reveals how the human capacity for learned, voluntary speech was assembled from older components (Nieder & Mooney, 2020). The study of animal calls thus feeds directly into questions about the representation of meaning, the control of action, and the developmental origins of a uniquely elaborated human ability, making vocalization a bridge between ethology and the cognitive science of language.

Current Directions

A major current move is to stop treating vocal learning as a single all-or-nothing trait. The modular view proposes that it is a bundle of separable components—the ability to imitate novel sounds, to modify call timing, to invent new signals, to learn across the lifespan—each of which can be present or absent and under separate genetic and neural control, which reconciles the messy comparative data far better than a binary learner/non-learner split (Wirthlin et al., 2019). This reframing is changing which species are worth studying and which comparisons are meaningful.

One beneficiary is the bat. Long known to echolocate, several bat species are now established as vocal production learners, and because they are mammals with rich social lives and tractable genetics they are becoming a model for the biology of vocal learning that complements the songbird, with projects linking genes such as FOXP2 to the vocal-motor circuitry (Vernes & Wilkinson, 2020). Chief among those genes is FOXP2, the first gene tied to a heritable human speech and language disorder: experimentally knocking down its expression in Area X, the song-learning basal-ganglia nucleus of juvenile zebra finches, yields incomplete and inaccurate imitation of the tutor song—direct evidence that a gene central to human speech also shapes how a bird learns to sing (Haesler et al., 2007). In parallel, the search for the genetic and circuit basis of learned vocalization across birds and mammals is identifying convergent molecular signatures in the forebrain regions that control song and speech, testing directly whether distant lineages solved the problem the same way (Jarvis, 2019; Nieder & Mooney, 2020). The open questions are no longer whether animals learn to vocalize but which components of the capacity each lineage has, and how those components are built.

Common Misconceptions

All animals learn their calls the way birds learn song.

No. Vocal production learning—acquiring a signal by imitation—is rare, confined to a few groups such as songbirds, parrots, hummingbirds, cetaceans, bats, and humans. Most species develop their calls on a fixed schedule with no model, as isolation-reared animals demonstrate (Janik & Slater, 1997).

The pitch of a call is its most informative feature.

Often it is not. Pitch, set by the source, tracks arousal and effort and can be varied at will, whereas the formants, set by vocal-tract length, give a harder-to-fake cue to body size and identity; the two are largely independent (Fitch, 2000; Taylor & Reby, 2010).

A referential alarm call means the animal has a word for the predator.

Functional reference is a deliberately bounded claim: the call reliably specifies a class of event and elicits the matching response, but this does not establish that the animal intends to inform or holds a concept as a human does (Seyfarth, Cheney, & Marler, 1980; Zuberbuhler, Cheney, & Seyfarth, 1999).

A songbird raised in silence will still sing normally.

It will not. A bird needs to hear adult song to form its sensory template and needs to hear itself to match its output to that template; deprived of either, through isolation or deafening, it develops abnormal song (Konishi, 1965; Marler, 1997).

Glossary

Alarm call.

A vocalization given in response to a predator; in some species it is functionally referential, specifying a class of predator and eliciting the matching escape.

Auditory feedback.

The sound of its own voice that an animal hears and uses to compare its vocal output with a remembered target; essential to song learning.

Crystallized song.

The stable, stereotyped adult form a learned song takes once the sensorimotor phase is complete and the song has stopped changing.

Formant dispersion.

The average spacing between successive formants, equal to c divided by twice the vocal-tract length; an honest acoustic cue to body size.

Formant.

A resonance of the vocal tract that shapes the source sound into a peak in the output spectrum; formants carry information about tract length and identity.

FOXP2.

A transcription-factor gene, the first linked to a heritable human speech and language disorder, that is also required for normal vocal learning in songbirds; a molecular bridge between birdsong and speech.

Functional reference.

The property of a signal that reliably denotes a class of external events and elicits an appropriate response, without implying human-like intention or concepts.

Fundamental frequency.

The repetition rate of the source vibration, heard as the pitch of a call; set by the larynx or syrinx and largely independent of the formants.

Larynx.

The mammalian vocal source, a cartilaginous organ whose vibrating folds chop the airstream into the buzz that the vocal tract then filters.

Sensitive period.

A restricted developmental window during which vocal learning proceeds readily and after which the same experience has much less effect.

Sensory template.

The stored memory of adult song that a young bird acquires during the sensory phase and later uses as the target for its own developing song.

Song dialect.

A regionally distinct version of a species' song, transmitted culturally when young birds learn from local adults rather than from their genes.

Song-control system.

The dedicated forebrain circuit of songbirds, including a premotor pathway for song production and a basal-ganglia loop supplying the variability that learning requires.

Source-filter theory.

The account of vocal production in which a vibrating source sets the pitch and the vocal tract filters it into formants; carried from human phonetics into animal research.

Subsong.

The rambling, variable early vocalizations of a juvenile songbird, analogous to human infant babbling, that precede plastic and crystallized song.

Syrinx.

The avian vocal source at the base of the trachea, whose two independently controlled sides let a songbird produce two notes at once.

Vocal learning.

The modification of vocal signals through experience; its demanding form, vocal production learning, is the acquisition of new sounds by imitation.

Vocal production learning.

The capacity to imitate and produce novel vocalizations learned from others, found in only a few animal groups and central to the evolution of speech.

Key Researchers

Dorothy L. Cheney

(University of Pennsylvania; 1950–2018). American primatologist who, with Robert Seyfarth, ran the long-term vervet-monkey studies that established functionally referential alarm calls and the primacy of meaning over acoustic form in primate vocal communication. Wikipedia · Wikidata

Allison J. Doupe

(University of California, San Francisco; 1954–2014). Canadian-American neuroscientist whose work on zebra-finch song linked song learning to human speech acquisition and to the basal-ganglia circuitry of sensorimotor learning. Wikipedia

W. Tecumseh Fitch

(University of Vienna). American evolutionary biologist and cognitive scientist who brought source-filter theory into animal research and studies the evolution of vocal production, communication, and language through the comparative method. ORCID · Wikipedia

Erich D. Jarvis

(Rockefeller University). American neurobiologist whose work on the molecular and circuit basis of vocal learning in songbirds, and its convergent evolution with spoken language, reshaped the field's taxonomy of vocal learners. ORCID · Wikipedia

Masakazu Konishi

(California Institute of Technology; 1933–2020). Japanese neurobiologist who established that songbirds require auditory feedback to develop normal song, the founding evidence for the sensory-template model of song learning. Wikipedia · Wikidata

Peter Marler

(University of California, Davis; 1928–2014). British-American ethologist who founded the modern study of birdsong learning—song dialects and the sensory-template model—and, with Cheney and Seyfarth, of referential communication in primates. Wikipedia · Wikidata

Fernando Nottebohm

(Rockefeller University). Argentine-American neuroscientist who mapped the songbird song-control system and discovered adult neurogenesis in the avian brain, founding the neuroethology of vocal production. Wikipedia

Sonja C. Vernes

(University of St Andrews). British neurogeneticist who co-founded the Bat1K genome project and established bats as a mammalian model of vocal production learning, linking genes such as FOXP2 to vocal-motor circuits. ORCID · Wikipedia

Frequently Asked Questions

What is animal vocalization?

It is the production of sound signals by animals using a dedicated organ—the mammalian larynx, the bird's syrinx, or an analogue—to communicate. The Medical Subject Headings vocabulary indexes it under animal communication, as the acoustic channel of signaling (Marler & Tamura, 1964).

How do animals produce vocal sounds?

By a source-filter process: a vibrating source, the larynx or syrinx, sets the fundamental frequency, and the vocal tract filters that sound into resonances called formants. Pitch comes from the source, formants from the shape and length of the tract (Fitch, 2000).

What can a listener learn from the formants of a call?

Because formant spacing depends on vocal-tract length, and tract length scales with body size, formant dispersion is an honest cue to how large the caller is—one difficult to fake because it is fixed by anatomy (Taylor & Reby, 2010).

Which animals learn their vocalizations?

Vocal production learning—acquiring new sounds by imitation—is found in only a few groups: songbirds, parrots, and hummingbirds among birds, and cetaceans, pinnipeds, bats, elephants, and humans among mammals. Most other species develop their calls without a model (Janik & Slater, 1997; Jarvis, 2019).

How does a songbird learn its song?

It first memorizes adult song as a sensory template during a sensitive period, then practices—producing variable subsong, then plastic song—using auditory feedback to match its output to the template until the song crystallizes into the adult form (Marler, 1997; Konishi, 1965).

Why is birdsong a model for human speech?

The two share a babbling-like practice phase, a dependence on hearing oneself, a sensitive period for learning, and dedicated forebrain circuitry that separates execution of the learned pattern from the variability needed to acquire it (Doupe & Kuhl, 1999; Brainard & Doupe, 2002).

Do animal calls have meaning like words?

Some are functionally referential: vervet monkey alarm calls specify a class of predator and elicit the right escape even when no predator is present. This shows a call can carry information about the world, though it does not prove the animal intends to inform or holds a concept as humans do (Seyfarth, Cheney, & Marler, 1980; Zuberbuhler, Cheney, & Seyfarth, 1999).

What is the current thinking on vocal learning?

Researchers increasingly treat vocal learning not as a single ability but as a bundle of separable components—imitation, timing control, invention, lifelong plasticity—distributed differently across species, with bats emerging as a key mammalian model (Wirthlin et al., 2019; Vernes & Wilkinson, 2020).

References