Acoustic communication plays an essential role in social behaviors of many species. In tetrapod vertebrates (Figure 1), both the cries of infants and the songs used in courtship are the result of neural circuit activity that drives muscles interposed between the lungs and the mouth. Sensory, CNS and motor systems that support innate, species-specific vocal communication reflect heritable genetic differences over evolutionary times scales. For example, crying is an innate behavior in infants (deaf babies cry), a key component of social interactions in our species. As human listeners misinterpret distress levels conveyed by tempo and pitch in cries from other primate species (bonobo and chimpanzee, Kelly et al., 2017), baby Homo neanderthalis and sapiens cries were probably species-specific. Courtship songs in other tetrapods are also typically innate, with only a few exceptions, most notably songbirds (Jarvis, 2019). Producing and recognizing different innate vocalizations (e.g., call types in birds; crying, sighing, laughter in H. sapiens) is essential for social communication (Simonyan and Horwitz, 2011; Rose et al., 2022).

Acoustic communication is ancient in tetrapods and was maintained over long periods: (Chen and Wiens, 2020; Jorgewich-Cohen et al., 2022). How did the neural circuit architectures that support the production and perception of songs evolve? Figure 1 illustrates extant species that serve as models for the neural bases of vocal communication and their evolutionary histories. This review aims to compare CNS circuits that produce and respond to innate vocalizations in these models to identify conserved and divergent features across evolutionary time scales.

Anurans (frogs and toads) are among the most ancient acoustic communicators, appearing in the fossil record ~270 mya (Figure 1). Within the Anura, the terrestrial Neobatrachians (e.g. Ranids) emerged from a world-wide extinction event and underwent massive radiations at the KT boundary (~68 mya; Feng et al., 2017). The Archebatrachians, in contrast, neither became extinct, nor radiated massively, but instead gave rise to aquatic anurans, the Xenopodinae, exceptionally well-represented in the fossil record (Cannatella, 2015) as well by 29 extant species (Evans et al., 2015). Xenopus laevis from South Africa was adopted by biologists in the 1800s for studies in experimental ethology, development, endocrinology, cell biology, and neurobiology (Wallingford, 2022). Males in all 29 species communicate vocally (Tobias et al., 2011), as do females, although one species group (A) has lost the female release call (Tobias et al., 2014).

Xenopus are secondarily aquatic (derived from terrestrial ancestors) and vocalizations are produced by a larynx modified for underwater sound production to produce sounds without airflow (Kwong-Brown et al., 2019). The neural circuits that support sex- specific acoustic communication have been elucidated and hindbrain neurons responsible for species-specific song rhythms identified [reviewed in Kelley et al. (2020)]. Identifying the genetic basis of the production and reception of species-specific vocal signals in Xenopus (clawed frogs) is our current research focus.

Placental mammals are also ancient acoustic communicators (~80 mya; Figure 1). The Rodentia—having diverged from langiomorphs (rabbits) ~85 mya—include many vocal genera (Mus, Scotinomys, Rattus, Heterocephalus). Mice (Mus) have dominated investigations of rodent ultrasonic vocalizations (USVs), due in part to the genetic advantages of specific laboratory strains. The Chiroptera (bats)—another highly vocal group—diverged from other mammals ~75 mya (Agnarsson et al., 2011). In some species, bats use vocalizations in both social communication and prey capture (echolocation). Among Hominids, the ancestor of our own species (H. sapiens) is recent (~200–250 kya; Vidal et al., 2022). Homo sapiens is the sole extant species and we do not know whether, for example, H. neanderthalensis spoke or sang. In all mammals, the larynx is the major organ of vocal expression. Sounds are powered by expiration and shaped acoustically by the vocal tract (Milsom et al., 2022).

Birds evolved from Archosaurs (dinosaurs) ~240 MYA and bird species radiated ~60 mya, again reflecting the worldwide extinction event at the K-T boundary. All extant birds communicate vocally (Figure 1), suggesting that ancestral dinosaurs sang as well. Avian behaviors were a focus of early ethologists (e.g., Lorentz and Tinbergen) and the discovery of geographical dialects in some species, provided experimental model systems for vocal motor learning. The zebra finch (Figure 1, lower left), Taeniopygia guttata, is currently the “lab rat” for the study of bird song neural circuits. While song control nuclei in the forebrain are not homologs of mammalian cortical areas that participate in acoustic communication, they exhibit convergent neural circuit architectures including, for example, a role for dopamine in song learning (Gadagkar et al., 2016).

I begin this review by examining neural circuit mechanisms that receive and generate species-specific Xenopus songs and then compare shared and divergent circuit motifs with other vocal vertebrates. As invertebrates are also prominent acoustic communicators. I conclude by comparing circuit motifs in tetrapods to acoustic communication in two invertebrates—fruit flies and Hawaiian crickets.

Understanding how the nervous system generates and responds to vocal signals and how circuit architectures diverge evolutionarily ideally requires a multispecies genus that communicates vocally, in which both the neural circuits that generate vocalizations—and those that respond to socially relevant sounds—can be mapped, characterized and compared electrophysiologically and anatomically, and the underlying genetic architectures of key neurons identified. Neural circuits are constructed developmentally and thus easy access to the nervous system at all developmental stages is advantageous. These features are all prominent in Xenopus, the focus of our experimental studies for many decades (see Kelley et al., 2020).

Each species of Xenopus can be identified definitively from the temporal and spectral features of male advertisement calls (Figure 2B; Tobias et al., 2011; Evans et al., 2015). A major experimental advantage of Xenopus are the ex vivo preparations: the vocal organ and the brain that “sing in the dish.” The ex vivo larynx generates fictive songs (sound pulse patterns) in response to stimulation of attached laryngeal nerves that mimics call patterns (Figure 2); spectral features are identical to actual calls (Tobias and Kelley, 1987). The ex vivo brain generates fictive songs, patterned laryngeal activity nerve that matches actual vocalizations, in response to application of the neuromodulator serotonin (Rhodes et al., 2007). Spectral features are thus created within the larynx without air flow while temporal features are generated by neural circuits within the CNS (Luksch et al., 1996; Kelley et al., 2020). Application of fluorescent dextran amines to the ex vivo brain allows visualization of neurons within specific brain nuclei that receive auditory input as well as the auditory vocal interface and components of the vocal motor pathways, including axonal trajectories (reviewed in Kelley et al., 2020). Sex-specific characteristics of these neural circuits and of the larynx are due to hormone-regulated development (Zornik and Kelley, 2011). These ex vivo preparations allow us to experimentally determine which hormonally regulated, sexually differentiated features are organ-autonomous and which reflect brain-larynx interactions.

Recordings from a South African pond across the breeding season together with laboratory studies (Tobias et al., 2004) reveal a rich vocal repertoire specific to social context and sex in Xenopus laevis, the most widely studied species (Figure 3).

In most tetrapods, sounds are powered by expiration of air from the lungs driving vibrations of the vocal folds (Ghazanfar and Rendall, 2008; Fitch and Suthers, 2016) or—in birds—of the internal tympaniform membranes (Elemans, 2014). In mice, for example, ultrasonic vocalizations are produced by high velocity air jets that power vocal fold vibrations (Håkansson et al., 2022). CNS respiratory and vocal circuitry must thus be closely linked.

However, the ability of the ex vivo Xenopus larynx to create sounds in the absence of air flow from the lungs, together with observations that these sounds are not shifted in frequency by heliox (Yager, 1982; Kwong-Brown et al., 2019), suggested that Xenopus do not use air flow to power their underwater songs. Instead, sounds are created by rapid separation of intra-laryngeal arytenoid cartilage disks (Figure 4), creating vibrations of the entire body (Kwong-Brown et al., 2019). Sounds are propagated effectively underwater because of impedance matching; the body is mostly water and the medium is water. In vivo, vibrations can be recorded from the entire body, including a single digit. The frog's body thus serves as a “loudspeaker.” This novel mechanism for anuran vocal production allowed Xenopus to retain ancestral, terrestrial frog vocal signaling (Feng et al., 2017) during underwater social interactions.

The ability to evoke sex- and species-typical sounds from the ex vivo Xenopus larynx (Figure 4) reveals that, unlike mammals and birds, in which respiration paces sound production and the CNS controls sound frequencies via the vocal tract (Matzinger and Fitch, 2021), the spectral features of Xenopus vocalizations are intrinsic to the larynx.

When males sing, neural activity that closely corresponds to actual male and female calls is recorded en passant from the laryngeal motor nerve (Yamaguchi and Kelley, 2000; Figure 5). Tightly synchronized Compound Action Potentials (CAPs) recorded from the nerve match the temporal pattern of simultaneously recorded underwater songs across sexes and species (fictive singing: Leininger and Kelley, 2013; Barkan et al., 2018). The temporal features of Xenopus songs are generated within the CNS.

Underwater sound waves produce vibrations of the Xenopus tympanic disk (Christensen-Dalsgaard and Elepfandt, 1995). The disk is located just behind the eye, under the skin (Figure 6). The stapes (a middle ear bone) inserts into the disk proximally and abuts the oval window distally (Mason et al., 2009). An air-filled cavity connects the two tympanic disks which thus function together as a pressure receiver.

As in other anurans, the inner ear includes an amphibian and a basilar papilla innervated by fibers of the eighth cranial nerve that arise from neuronal cell bodies in the acoustic ganglion (Homma et al., 2022), and whose terminals innervate post-synaptic neurons in the dorsal medullary nucleus (DMN, Figure 7; Kelley, 1980; Paton et al., 1982). Within the inferior colliculus of the midbrain (ICo), neurons in the laminar nucleus respond to calls (Elliott et al., 2011) and are rate-tuned to temporal properties of specific calls, as is the case for other anurans (Edwards et al., 2007). Song playbacks also activate the nucleus of the lateral line (NLL) and the principal nucleus (P) of the inferior colliculus (Kelley, 1980). Both project to the central nucleus of the thalamus (CT) nucleus (also illustrated below in Figure 7), suggesting that underwater sound waves are also detectable by the lateral line system.

When serotonin is applied to the isolated brain of males and females (Figure 7A), compound action potentials (CAPS) recorded from the laryngeal nerve (Figure 7A) match male- and female-specific vocal patterns (Rhodes et al., 2007). These patterns are called “fictive calling.” The fast trill portion of the fictive male advertisement call is driven by a rhythmic local field potential produced by neurons in the parabrachial nucleus (PB). The PB is a central pattern generator for advertisement calling.

Anterograde and retrograde mapping in ex vivo male and female brains—using fluorescent dextran amines thar travel both anterograde and retrograde—reveal components of the neural circuits that generate vocal patterns (Figure 8). Vocal motor neurons occupy caudal Nucleus Ambiguus, NA. Glottal motor neurons (the glottis is closed during calling), commissural interneurons, and neurons projecting bilaterally to the parabrachial nucleus (PB) occupy anterior Nucleus Ambiguus (antNA). Neurons in PB project throughout NA (shading in Figure 8B), both ipsilaterally and contralaterally, as well as reciprocally. Serotonergic neurons in the rostral Raphe, pars dorsalis (rRpd; Ra in Figure 8) project contralaterally to each other and ipsi- and contralaterally to vocal motor nuclei including the periaqueductal gray (PAG), PB, and NA (Brahic and Kelley, 2003; see also Figure 10C). Two forebrain nuclei (the Central nucleus of the Amygdala, CeA) and the Bed Nucleus of the Stria Terminalis (BNST) project to their contralateral counterparts as well as to Ra and PB. The resulting pattern of connectivity (Figure 8B) is highly recurrent and bilateral, insuring effective simultaneous contraction of laryngeal muscles required to produce a sound pulse (Figure 4).

In native ponds, the behaviors illustrated in Figure 3—calling and clasping—are seasonal and depend on reproductive state. A sexually reproductive state can be induced in the laboratory by injection of human chorionic gonadotropin (HCG). Embryos can thus be generated at any time of year, greatly facilitating discoveries in developmental, cell and molecular biology (reviewed in Wallingford, 2022) as well as powering discoveries in neurodevelopmental disorders (e.g. Willsey et al., 2021).

The behavioral effects of HCG on male calling are due to gonadotropin itself, to direct effects on neurons expressing gonadotropin receptors in the CeA (Yang et al., 2007), as well as to evoking increased synthesis and release into the circulatory system of gonadal steroids (androgens and estrogens) that activate neurons in the CNS. Sexually unreceptive or ovariectomized females respond to male clasping with leg extension and ticking (Figure 3F) while gonadotropin-injected intact females respond with leg flexion (Figure 3E). Castration abolishes male clasping and calling, behaviors reinstated by androgen treatment (Kelley and Pfaff, 1976; Wetzel and Kelley, 1983). Androgen effects on calling include activating vocal motor neurons as well as their inputs from the parabrachial nucleus. On the auditory side, gonadal hormones effects include direct action on androgen receptor expressing neurons in the acoustic ganglion in the periphery (Kelley, 1981), in the auditory midbrain (Kelley, 1980) and in the CeA of the ventral forebrain, where auditory input and pre-motor output intersect (Hall et al., 2013). Similar patterns of hormone receptor expression are found across other vertebrates (see Figure 10).

In summary, innate acoustic communication in Xenopus is characterized by species-specificity, a large vocal repertoire, pronounced sexual differences due to secretion of gonadal hormones and male/female, male/male duetting. These features reflect the preeminence of acoustic signals in turbid aquatic habitats over the evolutionary time scales (~ 170 mya, Feng et al., 2017) since the Pipoidae diverged from terrestrial anurans.

Across phyla (Figure 1), acoustic communication is closely associated with nocturnal species (Chen and Wiens, 2020). Bats, for example, are among the most vocal species, using sound both for locating prey (echolocation) and for social interactions within the roost (Kanwal, 2009). Male sac winged bats (Saccopteryx bilineta) produce complex courtship songs directed toward the females in their harems at dawn and dusk, sandwiched between territorial songs (Behr and von Helversen, 2004). The time of day that vocal species are active can vary. Within rodents (~40% of mammalian species), mice (Mus) and rats (Rattus) are nocturnal while other genera (e.g., Scotinomys Neotropical mice), are diurnal (see below). Vocal communication is also prominent in subterranean genera such as naked mole rats (Heterocephalus glaber, Credner et al., 1997). Anurans (frogs and toads) are also nocturnal and vocal.

In contrast, birds and humans—perhaps the most highly vocal groups—are predominantly diurnal. Most birds sing as the sun first rises and throughout the day. Even night songsters -i.e., nightingales—join in the dawn chorus (Amrhein et al., 2002). Though humans are diurnal, for most of our evolutionary history, visual cues were not available at night; essential social cues (e.g., your own baby's cry) were vocal. Reflecting diurnal activity, in birds and humans social communication is multimodal; visual cues can shape auditory perception. In humans, watching sound production changes what is heard (the “McGurk effect,” Alsius et al., 2018). In songbirds (zebra finches: Taeniopygia guttata), visual signals from conspecifics influence the activity of auditory neurons (George et al., 2011).

As for vocal behaviors, the neural circuits that support acoustic communication in tetrapods leave no trace in the fossil record. We can however compare circuit architectures across vocal vertebrates to determine which features are shared and which are specific to a particular group. For this comparison I've chosen three mammals: a bat (Pteronotus parnelli) and two rodents: Alston's singing mice (Scotinomys teguina) as well as mice (Mus musculis), vocal species with well-characterized repertoires and CNS vocal circuits. As for humans, features of acoustic communication in some species of birds are learned (Jarvis, 2019). Zebra finches and related finches that also learn their songs provide the opportunity to compare circuit motifs across wide phylogenetic distances (Figure 1) as well as providing insight into how acoustic experience and feedback can modify brain circuitry more generally.

As discussed below, a vocal CPG that patterns mouse pup cries has recently been identified in the inferior reticular formation (Wei et al., 2022). In Xenopus, the parabrachial nucleus (PB) is a CPG for the male advertrisement call [reviewed in Kelley et al. (2020)]. When the ex vivo Xenopus brain is exposed to serotonin, fictive advertisement calling CAPs recorded from the laryngeal nerve coincide with a pronounced local field potential recorded from the PB (Rhodes et al., 2007). Transection at various levels of the CNS as well as cooling studies confirm the role of the PB as a vocal CPG. PB neurons retain their intrinsic rhythmicity in the ex vivo brain even when isolated synaptically (Barkan and Zornik, 2019). A CPG that drives slow trill has not yet been identified but might correspond to PiCO, a proposed inspiratory CPG in mice (Anderson et al., 2016). While Xenopus vocal production is independent of respiration (Figure 4), a rhythmically active neural circuit element that functions to gate the inspiratory/expiratory transition in rats (Dutschmann and Herbert, 2006) functions in Xenopus as a CPG controlling vocal rhythms, suggesting exaptation of a respiratory circuit element present in the common ancestor of tetrapods. As discussed below, one class of neurons in the PB, the FTNs, differ intrinsically in rhythmicity across related species, opening a window into genetic divergence that supports speciation (Baker et al., 2019).

As bird songs are coordinated with respiration, one approach to finding a vocal CPG in birds is to identify the respiratory CPG. Wild (1997) described neurons in nucleus retroambiguus (Ram: Figure 11B) projecting to respiratory motor neurons in pigeon and songbirds. RAm efferents were also observed in the PB, rostroventral lateral medulla (RVL), caudal pons and in XIIts. Both RAm and PB have been considered candidate vocal pattern generating nuclei in songbirds. A recent review (Mooney, 2020) suggests instead that the songbird vocal CPG is located in a reticular nucleus, RVL. RVL drives activity of syringeal motor neurons but is gated by neurons in the caudolateral PAG. An alternative suggestion is that RVL coordinates activity of vocal motor neurons (as suggested for LRF in Scotinomys) while the homolog of PB contributes controls vocal patterning. If so, the origin of the PB as a vocal CPG could be evolutionarily ancient.

A recent study (Wei et al., 2022) sought to identify a vocal CPG in infant mouse pups by examining the neural circuits that generate USVs. In pups, a single large breath can be associated with either one or multiple cries. For multiples, each cry is accompanied by a smaller increases or decreases in airflow (resembling the “minibreaths” in canary and zebra finch songs, Hartley and Suthers, 1989; Wild et al., 1998). The authors predicted that this vocal pattern is generated by an intrinsically faster CPG that coordinates with the overall breathing pattern. Previous studies in mice have established that breathing is patterned by a inspiratory CPG that includes neurons in the preBotzinger nucleus (PBC). Blocking the activity of laryngeal TA and CT prevented cry production but not the minibreath pattern, suggesting separate CPGs for cry production and minibreaths. Interneurons innervating TA and CT motor neurons form three groups: rv-iRF (glutamatergic neurons), Botzinger and preBotzinger nuclei (gabaergic) and Nucleus Retroambiguus (mixed). Interneurons innervating both tongue motor neurons and TA motor neurons were also found in rv-iRF. Inactivating rv-iRF disrupted the interval between cry bouts as well as intervals within a bout, but not basal breathing. Brief optogentic stimulation of the rv-iRF produced cry bouts throughout the longer breath. Comparing activity patterns in brain slices that included the rv-iRF and the pre-Botzinger nucleus revealed a faster oscillation (every 6s as compared to 23s) in the former. These experiments provide strong evidence that the rv-iRF generates the pattern of pup cries. This vocal CPG provides input to preBotzinger neurons to drive inspiration (triggering minibreaths) and coordinates activity in the laryngeal TA and CT muscles that control glottal opening.

The IRO is also a candidate participant in patterning of the more complex courtship vocalizations of adult mice. At the behavioral level, the overall spectro-temporal features of male mouse USVs develop continuously from pup calls, stabilizing about 4 weeks later. A shared CPG might represent the “common biological mechanism” suggested by Castellucci et al. (2018). Regardless, the Wei et al. study provides an experimental blueprint for identifying candidate CPGs in adult mice as well as other rodents, such as rats and Scotinomys. Identification of IRO as a CPG for pup calls does not preclude the participation of other CPGs in patterning vocalizations. In cats, for example, neurons in the parabrachial nucleus (specifically the Kolliker-Fuse nucleus) are rhythmically active during inspiration, post-inspiration and expiration (Dick et al., 1993), providing a candidate vocal CPG (see also Hage, 2010 for discussion of primate vocal CPGs).

Another highly conserved feature of CNS vocal circuitry is the expression of receptors for gonadal steroids (typically androgen in males and estrogen in females) in auditory and vocal neurons (shown for the androgen receptor in motor components of the vocal circuit in Scotinomys (nuclei in yellow: Figure 10B) and for Xenopus (nuclei in yellow and green, Figure 10C).

The capacity for synthesizing estrogen arose before the evolution of the ancestral ER (Eick and Thornton, 2011). Steroid hormone receptors are also evolutionarily ancient; derived from a single ancestral receptor that diverged from the nuclear receptor superfamily early in vertebrate evolution. These receptors diversified in the chordates; amphioxus has two: an ER and a member of the AR/PR/GR/MR family. The pipoidae (ancestral to modern pipids including Xenopus) emerged during the Jurassic ~170 mya and the genus Xenopus ~50 mya (Feng et al., 2017). The Rodentia diverged from a common ancestor with the Lagomorpa ~65 mya (Romanenko et al., 2012) and Scotinomys perhaps 7 mya (Fernández-Vargas et al., 2021). While characters related to reproductive signaling (such as sexual dimorphism) can be lost as well as gained evolutionarily (e.g., Leininger and Kelley, 2013), similarities in the distribution of androgen receptors in two evolutionarily very distant species (Xenopus and Scotinomys, see below) suggest an ancient role in the coordination of vocal signaling during reproduction.

Xenopus In X. laevis, androgen (acting synergistically with gonadotropin) controls male clasping (Kelley and Pfaff, 1976) and estrogen (acting synergistically with LHRH and gonadotropin) controls female receptivity (Kelley, 1982). Gonadectomy abolishes adult reproductive behaviors in both sexes. These hormones also participate in the control of vocal communication. Gonadal steroid receptors are expressed in brain regions implicated in acoustic communication; from the acoustic ganglion through to the larynx during both development and adulthood including laryngeal motor neurons (Kelley, 1980).

In females, but not in males, preferential auditory evoked potential responses to each species' dominant frequencies are abolished by ovariectomy and reinstated by androgen (Hall et al., 2013). Testosterone is the major circulating gonadal steroid in female Xenopus (Lutz et al., 2001) but can be converted to estrogen in situ by aromatase. Neurons within the CeA express estrogen (Figure 10C) and gonadotropin receptors (Yang et al., 2007). Gonadotropin synergizes with androgen to restore calling to castrated makes (Wetzel and Kelley, 1983). The CeA receives auditory input and is required for males to produce socially appropriate responses to female calls (Hall et al., 2013). Laryngeal motor neurons in NA express androgen receptor (Kelley, 1980). The vocal pattern generator (PB) includes neurons expressing androgen but not estrogen receptor (Figure 10C).

Bats as Kanwal points out: “There is a deep connection between hormones, the perception and production of social vocalizations, and behavior. Hormones-to-circuits-to-perception or production is a bi-directional process… hormones can modulate and set up either transient or long-lasting neural circuits for the processing, perception, and production of sounds, particularly those having social consequences” (Kanwal, 2021, p. 239). Neurons in the DSCF (Doppler-shifted constant frequency) region of P. parnelli respond both to echolocation and to social vocalizations (Washington and Kanwal, 2008) and processing is lateralized in males (but not females) with more responsive neurons in the left hemisphere (Kanwal, 2021), suggesting a sex difference likely to be driven by gonadal hormones. While the locations of gonadal hormone receptor expressing neurons have not yet been mapped, the bat CeA most likely shares this common vertebrate circuit motif. Current research on bat social communication is shifting to Carollia perspicillata, as this species is more readily maintained in breeding colonies and uses complex vocal interactions to communicate. Individual C. perspicillata have distinctive vocal signatures. Distress calls have been shown to activate neurons in the amygdala (Hechavarría et al., 2020). Mapping gonadal steroid hormone receptor distributions in C. perspicillata will be a useful test of evolutionary hypotheses.

Singing mice despite evolutionary divergence (Figure 1) circuit motifs for acoustic communication and AR expression in S. teguina share multiple features with other tetrapods, inclding Xenopus (Figure 11). Notably, in both species, vocal motor neurons in nucleus ambiguus and pre-motor neurons in the parabrachial nucleus express androgen receptor (yellow B; yellow and green, C). Neurons in the inferior colliculus of both species also express AR (not illustrated in B). While estrogen receptor expression has not been mapped in Scotinomys, ER is expressed in inferior colliculus and CeA of laboratory mice (Charitidi and Canlon, 2010) and is likely to also be expressed in auditory nuclei of singing mice.

Song birds As for anurans and rodents, androgen receptor expression is widespread in bird vocal control nuclei (Figure 11B); estrogen receptor however is limited to HVc (Frankl-Vilches and Gahr, 2018). In female white-crowned sparrows, circulating estrogen during the breeding season increases responses of auditory neurons in Field L (Figure 10A) (Caras et al., 2012) as well as immediate early gene expression in the social behavior network (Maney et al., 2008). Auditory responses to song have also been recorded in the bird homolog of the amygdala, nucleus taenia (Fujii et al., 2016). Hormonal regulation of both sensory and motor neural circuits that participate in vocal courtship in vertebrates appears ancient.

The central nucleus of the amygdala (CeA) has been described as the “autonomic” amygdala because of its role in respiration and heart rate. Given the prominence of expiration for vocal expression across vertebrates, CeA involvement in vocal communication makes sense. In Pteronotus CeA stimulation evokes agonistic vocalizations (Ma and Kanwal, 2014). Neurons in the CeA also respond to social vocalizations, especially those associated with aggression. In primates, a baby's cry activates the parents' amygdala (Riem et al., 2021). Autonomic rhythms pace vocalizations of marmosets (Zhang and Ghazanfar, 2016). Output from the CeA (central-medial boundary) to the PAG transiently suppress vocalization in mouse pups (Tschida et al., 2019). In adult mice, activating neurons in the preoptic area of the hypothalamus (POA) that express estrogen receptor in adults inhibits inhibitory PAG neurons allowing USV expression as well as scaling the duration and persistence of bouts (Chen et al., 2021).

In Xenopus, a species that uncoupled breathing from calling many millions of years ago, the CeA matches acoustic stimuli to vocal expression. Lesions of the CeA in Xenopus result in socially inappropriate responses of males to song playbacks (Hall et al., 2013). Lesioned males respond to broadcasts of rapping and even an actual rapping female (Figure 3G) with prolonged vocal suppression (the response normally elicited by a vocally dominant male) rather than answer calling (the socially appropriate response; Figure 3). In Xenopus, the inhibitory output from CeA to a putative PAG homolog is conserved. However, unlike mammals, the Xenopus APOA does not project to PAG directly, instead innervating and receiving input from rRpd (Brahic and Kelley, 2003). rRpd, a serotonergic nucleus, is reciprocally connected to APOA, PB and NA and also projects to PAG. Thus, while in mice POA ER-expressing neurons have direct access to the PAG—inhibiting inhibitory neurons and promoting vocalization—in Xenopus, APOA neurons may influence vocalization via serotonergic innervation of PB and NA.

The persistence of species depends on successful reproduction: the production and survival of offspring that go on to reproduce and survive themselves (Darwin, 1872). Because hybrid offspring can be disadvantaged in development, survival and/or mating, identifying a potential mate of the same species is a primary imperative (Lemmon and Lemmon, 2010). As there are 29 extant species of Xenopus, in all of which males produce advertisement calls (Tobias et al., 2011), evolutionary conservation of neural circuitry across the genus can be explored. For example, the role of the PB in generating vocal patterns in Xenopus was evaluated recently by cross-species comparison between X. laevis and X. petersii, members of the L species subgroup that diverged ~8 mya (Figure 12A).

A specific class of rhythmically active neurons (Fast Trill Neurons or FTNs) in the PB was identified electrophysiologically in both species (Figure 12F). When synaptically- isolated by blocking sodium channels and stimulated by application of a glutamate agonist (NMDA), the membrane potential of FTNs oscillates at the species-specific rate and rhythm. This inter-species observation strengthens the identification of the PB as the vocal CPG (Rhodes et al., 2007) and implicates a specific class of rhythmically active neurons in divergence of vocal signaling across the L subgroup.

To drive the beginning phases of speciation that resulted, for example, in the different advertisement calls of the L subgroup, divergence in male courtship songs across populations must have co-evolved with female sensitivity to—or preference for -acoustic features of those songs (or vice versa). In Xenopus, each sound pulse includes two dominant frequencies (Figure 1) that differ across species. In the L subgroup, the DF2/DF1 ratio is 1.22, except for X. laevis in which the ratio is 1.14 (Kwong-Brown et al., 2019). Auditory evoked potentials reveal that females are preferentially acoustically sensitive to species-specific DFs at the species-specific ratio (Hall et al., 2013), suggesting that this spectral feature is salient for same species recognition by females in L subgroup species. Both temporal and spectral features of male and female calls determine vocal responses in X. laevis (Vignal and Kelley, 2007). As described above, in most vocal vertebrates the CNS controls both the temporal features of songs (via respiratory/vocal CPGs) and song spectral features (via hypoglossal control of the vocal tract). In Xenopus however the brain controls only the temporal features while spectral features are inherent to the larynx. This separation simplifies the genetic analysis of song divergence during speciation.

Within the L subgroup, advertisement calls are species-specific. What differences in gene expression between FTNs in the hindbrain and the vocal organ of different species contribute to species specific vocal signlling? Unusually, in Xenopus interspecific hybrids between extant species can produce fertile F1 and F2 offspring of both sexes (Evans, 2008). Genetic candidates for speciation-associated divergence in song temporal features are loci encoding or modulating FTN ion channels (Barkan et al., 2018). Candidates for species divergence in song spectral features are loci that contribute to the laryngeal cartilage components that support production of the two dominant frequencies (Kwong-Brown et al., 2019). To function in species divergence, female vocal perception and preference and diverging male songs must co-evolve. Recent research in two invertebrates, Hawaiian crickets and fruit flies, is revealing genetic architectures that support co-ordination of evolution of acoustic signaling in the sexes.

Acoustic signaling is ancient and phylogenetically associated with extant species that are nocturnal or especially vocal at dawn or dusk (Figure 1, Chen and Wiens, 2020). Birds and humans are the major exceptions. Humans display the “McGurk effect” in which visual stimuli from the face of a speaker influence the acoustic identification of a syllable (McGurk and MacDonald, 1976). In starlings, conspecific visual stimuli also modify responses of neurons in the primary forebrain projection area—Field L—to familiar and unfamiliar songs: familiar songs suppress responses and unfamiliar songs enhance responses (George et al., 2011). Multimodal sensory integration is thus also likely to have shaped the evolution of vocal communication in diurnal species such as primates and birds.

Because speciation reflects both sexual selection (success in attracting mates) and natural selection (survival), sex and speciation are linked at many levels. In vertebrates, sexual differentiation is governed by pituitary and gonadal hormones. Patterns of AR and ER expression—from sensory receptors through to neural circuits for muscle effectors—are targets for evolutionary selection. Broder et al. (2021) argue that “it may be easier than assumed to evolve new sexual signals because sexual signals may be arbitrary, sexual conflict is common and receivers are capable of perceiving much more of the world than just existing sexual signals.” Part of this argument is based on the idea that an arbitrary sensory stimulus associated with a positive experience (for example a red food source) can be co-opted to shape behaviors, such as approach (sensory exploitation). However, such co-opted sensory stimuli are not necessarily useful finding or selecting reproductive partner of the same species for reproduction. This is particularly for true females whose gametes are finite and provide resources for the embryo, unlike those of males.

Robert and the late Dorothy Cheyney argued (Seyfarth and Cheney, 2014)—using multi-year field data on vocal communication in baboons—that the origin of human language might lie in social cognition. Baboons have a matrilineal dominance hierarchy; each female has a distinctive “grunt” vocalization. Seyfarth and Cheyney recorded vocalizations during social interactions between all female pairs in Year 1. In Year 2, they observed female A/female B interactions and then played back A's call to B to determine whether B's response to the playback reflected what had happened (grooming, for example, or biting) during that specific interaction. Did B stay put (grooming: positive interaction) or move away (biting: negative interaction)? They reported that B's response was triggered specifically by A's grunt and matched the social valence of their recent interaction. If indeed the substrate for language evolution, we have much more to learn about the neurobiology of vocal communication across species.

While humans do not actually bite each other during arguments (at least as adults), we do use biting language. We also devote considerable attention to decoding how people feel about us from cues in voice to construct a socially appropriate response. Areas of the human brain involved in language production and perception must (at the very least) access other areas that identify social context-driven voice cues regulated by the endocrine and neuromodulatory systems described in this review. Advances in fMRI now allow imaging of entire brains in response to conspecific and heterospecific vocal sounds in other animals (Van Ruijssevelt et al., 2013; Gábor et al., 2020). Imaging whole brain activity during vocal communication across species over the next few years will drive additional discoveries in this scientific arena.

The original contributions presented in the study are included in the article/supplementary material, further inquiries can be directed to the corresponding author.

DBK wrote this manuscript and adapted from published work or created the figures.

The author declares that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.