Bodily motor imitation is generally believed to be a cognitively demanding form of social learning, as it “requires the cognitive system to know that the model’s and observer’s actions are similar from a third-party perspective; that it must be able to transform a sensory representation of the model’s body movements into a motor representation controlling the observer’s action” (pp R228 Heyes, 2021). This constitutes one of the main problems raised by imitation, that is, how these sensorimotor associations are established and by means of which neurocognitive mechanisms. This problem is known as the “correspondence problem” and it is still unresolved in the scientific community.

In this vivid “correspondence problem” debate, one of the distinctions that has attracted most attention is the difference between copying of “transparent” actions, when an observer performs the same movement, they will see the same spatial configuration. However, when performing “opaque” actions, the observer may see a different spatial configuration, or may not be able to see their own action at all, such as with facial expressions or whole-body movements like bowing or joining hands behind the back (as described by Heyes, 2021). It has been hypothesized that the former engages cognitive skills that can be less demanding than those required to match opaque actions, as in the latter what makes imitation a more difficult achievement is the difference between the information that is available to the observer’s senses when the body movement is performed by the model and when it is performed by the observer (Zentall, 2006; Heyes, 2021). Similarly, some researchers have argued that the copying of so-called transitive (object-oriented) actions is likely to engage cognitive skills that can be different from those required to match intransitive (body-oriented) actions (Heyes and Ray, 2000; Heyes, 2001). It has been suggested that intransitive imitation is limited to humans and the great apes (Miles et al., 1996). However, after several studies using this procedure in chimpanzees (Hayes and Hayes, 1952; Tomasello et al., 1993; Custance et al., 1995; Myowa-Yamakoshi and Matsuzawa, 1999) and orangutans (Call, 2001), it remains controversial whether great apes in general demonstrate a capacity for matching of their own body part actions to those of another ape (Whiten et al., 2009). In sum, comparative evidence has shown how difficult-but not impossible-it is for animals to copy pure movements that have no environmental effects compared with object-related actions.

Alternative and even narrower definitions have been proposed for imitation, for example arguing that the only form of true imitation that is found on higher-level cognitive skills, called “production learning,” is when the action of the model is entirely “novel” to the observer (Zentall, 2003, 2006; Hoppitt and Laland, 2013; Heyes, 2021). However, it is important to note that when defining novelty, there will always be some degree of similarity to what the observer has done before (Whiten et al., 1996) and that the imitation of novel actions could still be possible by the same learning mechanisms as familiar actions (see Heyes, 2021 and Michon et al., 2022 for review).

In any case, as can be seen in all the definitions mentioned above, the vast majority of them are oriented toward defining imitation in the “motor” or “gestural” domain (copying the topography of body movements), where as a result of observation of the model, the observer moves its body parts similarly to the model’s body parts (see Heyes, 2021). Moreover, it has been argued that the faithful reproduction of a demonstrator’s behavioral topography via imitation plays a key role in the emergence of human technology (Tennie et al., 2017) as it enables the accumulation of modifications over time (i.e., the ratchet effect) facilitating the evolution of culture together with other forms of social learning (Tomasello, 1999). Alternatively, it has been proposed that motor imitation’s crucial role relies on the cultural inheritance of human gestures, dialects, languages, and rituals (Heyes, 2021).

Vocal imitation (also referred to as vocal production learning) arises when an animal learns to modify its vocal output to match signals of other individuals. This includes the production of novel signals or modifications of signals that were already in an animal’s repertoire (Janik and Slater, 1997, 2000). Therefore, similar to bodily motor “productive” imitation, “true” or “complex” vocal learning is defined as the ability to copy sounds that are not part of the normal species-specific repertoire (Janik and Slater, 1997, 2000), which implies that the subject must learn a new acoustic template for the vocalization and then learn to develop a vocalization that matches the template (Tyack, 2020). Recent evidence have extended this definition as follows: vocal production learning is the production of modified or novel vocalizations, as a result of learning from the perceptual experience of the acoustic signals of others (Vernes et al., 2021).

Importantly, most researchers consider vocal learning as less complex since the copy of a sound may not involve the correspondence problem, even when the sound is novel (Heyes, 2021). This argument usually relies on the concept of “equimodality,” that is, that when you hear someone producing a sound, you perceive it using the same hearing device used to perceive the sounds you produce yourself. On the other hand, the copy of motor actions, particularly facial gestures, requires “visual-tactile (and proprioceptive) cross-modal performance” (see Shettleworth, 2010). In other words, seeing an action being performed recruits a different set of sensory systems to those recruited to sense the actions being performed by your own body. However, we do not necessarily agree with this view as in the case of hand movements (and to some degree body movements), the observer may use the same visuo-motor circuits to monitor the observer’s and the own movements, that may be comparable to the audio-vocal circuits involved in vocal imitation; and in both cases proprioceptive signals can be relevant for successful copying (see Hickok, 2022). Moreover, this view of vocal imitation as “less complex” is somewhat surprising since this capacity is a hallmark of human spoken language, which, along with motor imitation and other advanced cognitive skills, has certainly fuelled the evolution of human culture. In addition, cross-modal interaction between gestures (especially facial gestures) and vocal communication is highly relevant in speech development and most likely it was so in the evolutionary origins of speech (Michon et al., 2022).

Finally, the capacity for vocal learning is not strictly a gift given to some animals, but as said there are different levels of vocal learning abilities in different species. Petkov and Jarvis (2012) categorized species on the basis of their vocal learning capacities, where animals like monkeys display a limited learning capacity, songbirds and parrots are relatively complex vocal learners and our species apparently exceeds all the others in the voluntary and fine control of phonation (Petkov and Jarvis, 2012).

In fact, and strikingly, comparative evidence has revealed that humans are an odd case among mammals, as the case for vocal production learning exists mainly in birds (see Catchpole and Slater, 2003), with only a few instances in mammals, with confirmed evidence for elephants (Stoeger et al., 2012), bats (Knörnschild et al., 2010), pinnipeds (Ralls et al., 1985; Schusterman, 2008; Reichmuth and Casey, 2014) and cetaceans (Tyack and Sayigh, 1997; Janik, 2014). Likewise, current evidence supports that vocal imitation is harder than motor imitation for primates, as Japanese monkeys seem to require greater effort in motor preparation for voluntary control of vocalization than for manual actions (Koda et al., 2018). Even after training, the experimental evidence for imitation is scarce, with reports for spontaneous imitation of human whistles and of human “wookies” in orangutans under human care (Wich et al., 2009; Lameira et al., 2016) and novel atonal-breathing utterances in Koko, the famous female enculturated gorilla (Perlman and Clark, 2015).

On the other hand, while there has been little evidence of vocal imitation in monkeys and apes in the wild (Hauser et al., 2002), more recent reports indicate that wild apes may have a greater degree of vocal plasticity than has been usually considered, with social-dependent vocalization patterns in orangutans and combinatorial vocal sequences in chimpanzees (Girard-Buttoz et al., 2022). Furthermore, the new-world marmoset monkeys have a complex social and vocal behavior, displaying reciprocal turn-taking vocalizations and capacity for vocal learning from adults during development (Margoliash and Tchernichovski, 2015; Pomberger et al., 2019; Gultekin et al., 2021; Varella and Ghazanfar, 2021); however, it is not clear yet whether this may qualify as imitation in the sense used in this article. Still, there is the apparent paradox of how can vocal imitation be difficult for many primates, if the correspondence problem is supposed not to exist in this modality? This fact has not only risen the open question of whether it is adequate to claim that vocal imitation is less complex than motor imitation, but also what kind of mechanism (motor, auditory or and/or cognitive) has driven its evolution (see Feenders et al., 2008; Aboitiz, 2017). For example, it has been suggested that the evolution of vocal learning is driven by modifications of the motor rather than the auditory system, which is more conserved across species, and partly relies on general-purpose mechanisms (see Aboitiz, 2017 Chapter 9; Feenders et al., 2008; Petkov and Jarvis, 2012).

Comparative evidence exploring both modalities, vocal and gestural learning, is needed to clarify if the same happens in other vertebrates and specifically in other mammal species. Moreover, these results evidence that if vocal imitation is infrequent in non-human mammals, the capacity for them to imitate in both domains, bodily and acoustic, may be even scarcer. The evidence regarding diversity in manual and vocal control abilities in different primates supports that in the human lineage, vocal evolution probably evolved in close coevolution with hand control, associated to tool making and gesture behavior (Aboitiz, 2017, 2018; Koda et al., 2018). Importantly, recent neurological evidence in baboons has shown a dissociation of handedness’ types between manipulative action and communicative gestures, with only the latter being associated with a frontal cortex lateralization signature which might reflect a phylogenetical continuity with language-related Broca lateralization in humans (Becker et al., 2022). Perhaps the conjunction of vocal and motor communication systems in early Hominins drove the development of multimodal imitation capacities leading to the origin of a rudimentary proto-language, using both vocal and gestural signals. In this line, perhaps marmoset monkeys’ vocal abilities might better resemble the early vocal behavior of ancestral hominins than the macaques’. Basic vocal learning capacities used in turn-taking vocal interactions, perhaps comparable to those of marmoset monkeys (Hage et al., 2016; Takahashi et al., 2016), could have evolved early in our human ancestors, which complies with the apparent antiqueness of the human-specific mutations of the speech-related gene FOXP2 in our lineage (Aboitiz, 2018; Atkinson et al., 2018). A degree of voluntary control of vocalization for social purposes, together with skilled hand control associated with early tool-making abilities, and more specifically, for gestural expression, may have provided the feedstock for the evolution of visual-gestural and auditory-vocal imitation capacities that distinguish our species.

Although the existence of social traditions does not prove imitative learning per se, in cetaceans there have been observational reports of specific behavioral tactics in several species in the wild, which exhibit substantial behavioral diversity between sympatric groups in terms of the motor and acoustic features of their behavioral repertoires (foraging tactics, songs, calls, or dialects among others; see Rendell and Whitehead, 2001 for review).

With regards to baleen cetaceans the most documented case belongs to the humpback whale, with motor behaviors candidates to be learned by imitation. For example, some foraging methods consist of complex cooperative herding that includes bubble-cloud feeding combined with trumpet-like calls and “lobtail feeding” (see Whitehead and Rendell, 2014 for review). However, the most well-known imitation candidates are the whale’s songs, which possess several features that make them one of the most complex vocal displays in the animal kingdom: (a) they display hierarchical structures where sequences of units (discrete sounds) conform phrases that are repeated several times, each set of repetitions of a phrase type conforming a theme and a sequence of different themes conforming a song (units → phrases → themes → songs; Payne and McVay, 1971); (b) they are long, lasting from 6 to 35 min and can be repeated continuously during a song session (e.g., 22 h); (c) they undergo constant evolutionary change. This last characteristic is most remarkable as it constitutes a coordinated change over time, where all males in a population tend to share the same song, even while it changes gradually throughout the reproductive season (which results in that the shared song at the start and at the end of the season is not the same). This shared song that slowly drifts in its structure over time is almost unique among non-human mammals and the most accepted hypothesis is that other whales might learn novel variations introduced by other individual, however alternative explanations exists (see Iii, 2018).

Toothed cetaceans, also known as odontocetes, include species such as bottlenose dolphins and orcas that are thought to have their own unique and potentially culturally transmitted traditions. This could be the result of social learning through mechanisms like motor and vocal imitation. Similarly, these reports include foraging tactics as bottlenose sponging, shelling fishing tactics, mud-bank ring feeding (or mud plume fishing) and herding cooperative feeding (see Krützen et al., 2014). Additional examples are the orcas’ intentional beaching (Lopez and Lopez, 1985; Guinet, 1991; Guinet and Bouvier, 1995), the “carousel feeding” technique (Similä and Ugarte, 1993), or the “cooperative wave-washing behaviour” to take seals off the ice floe (Visser et al., 2008; Pitman and Durban, 2012) among others.

Although most of the authors are inclined to believe that these behaviors, reported in the wild, could be largely scaffolded by social learning processes including motor and vocal imitation (Ford, 1991; Deecke et al., 2000; Yurk et al., 2002; Barrett-Lennard and Heise, 2006; Weiß et al., 2011), other scholars have argued that the evidence for imitation is weak at best and that their ‘cultural traditions’ can be explained by other less complex learning processes (Caro and Hauser, 1992; Galef, 2001). Therefore, to elucidate the social learning mechanisms that underpin these group-specific behaviors, we will focus our review on experimental studies realized in controlled conditions that rule out alternative explanations.

While there are several procedures to study imitation in animals, the most used (or the gold standard) in cetaceans in controlled conditions is the “Do as I Do” method. This paradigm, initially used by Hayes and Hayes (1952) in a study of motor imitation in a home-raised chimpanzee, involves copying another’s action under a specific signal (“Do this!”), without any other scaffolding information (e.g., results-based cues). This specific method consists of training the observer to respond to a visible gesture-based command “copy” (“Do that”) given by the trainer. The sign itself could be gestural (usually hands) or vocal. This generic “copy” sign command made up for this purpose is the same over all presentations or trials, regardless of the behavior from the model that must be copied by the observer. Some authors have argued that to solve this task the observers need to “understand” that he/she is required to imitate what the model is performing, that is, that the animal subjects need to have some kind of concept of imitation, as the method depends on the generalization of a trained signal to a conceptual order that is “copy what I am (or what the other is) doing” (Whiten, 2000; Herman, 2002, 2006, 2010). This training technique has been successfully used in several species of great apes (Tomasello et al., 1993; Custance et al., 1995; Call, 2001), dogs (Topál et al., 2006; Fugazza et al., 2016), cats (Fugazza et al., 2021) and cetaceans: dolphins (Xitco, 1988; Bauer and Johnson, 1994; Herman, 2002), orcas (Abramson et al., 2013, 2018) and belugas (Abramson et al., 2017).

Cetaceans have developed several anatomical and biochemical hearing and vocal specializations that allow them to hear and emit a much wider range of sound frequencies (into the ultrasonic range) than humans. Accordingly, the ability of cetaceans to explore and interpret their world via a sophisticated echolocation or biosonar should be carefully considered. Toothed cetaceans and bats emit short-broadband ultrasonic clicks vocalizations (“nasalizations” would apply better in this case) to localize prey and to explore their surroundings (Au, 1993). The information is obtained via analysis of echoes generated from objects by sound reception through a region of the lower jaw and transmission into the ear. It has been documented that dolphins change the temporal and spectral features of their echolocation signals in relation to distance, size, shape and acoustic properties of the focused target (Au, 2004; Jensen et al., 2009). That is, odontocetes and possibly mysticetes (see humpback whale songs long-range sonar hypothesis by Iii, 2018) can associate information that they receive from vision with information that they obtain from echolocation when this information concerns stationary objects, but they are even able to integrate dynamic information about movement across echolocation and visual senses. This means that echolocation per-se is a cross modal perceptual system as it is based on “active production and perception of an acoustic beam transformed to a visual cross-modal representation.”

Moreover, it has been suggested that echolocation in cetaceans not only plays a physical function of sensing the environment for one individual but also plays a social function in communication with other individuals. Experimental evidence shows that it is possible for an eavesdropping dolphin to discern object information from the returning echoes generated by the echolocation signals of conspecifics (Gregg et al., 2007; Arranz et al., 2016). All these results in cetaceans parallel the evidence of a convergent evolution of echolocation and vocal learning in bats. In addition, in bats, echolocation also facilitates communication in various forms, such as eavesdropping, territorial defense and courtship (Knörnschild et al., 2012).

But echolocation is not the only sophisticated adaptation we have to consider for the presence of multimodal imitation in aquatic animals. Terrestrial animals rely upon voice cues to encode individual identity with their vocalizations. Whereas the recipient must learn to recognize the voice of each individual, the speaker is not necessarily constrained to learn to produce voice cues. Speaker-specific cues stem from natural differences in the air-filled vocal tracts of each person. Diving mammals are unable to rely upon such vocal cues for individual recognition because, as they dive, the volume of gases in their vocal tract is halved with each additional atmosphere of pressure: this change in the vocal tract renders voice cues unreliable (Tyack, 2000). Therefore, diving mammals that rely upon individual-specific social relationships must learn to control their vocal apparatus to produce individually distinctive vocal signature signals with an individually distinctive frequency pattern like the signature whistle above mentioned. This may be another significant reason behind cetacean capacity for vocal learning.

While on one hand, the dark and opaque underwater environment may be a good reason to think of cetaceans as predominantly vocal creatures (with their echolocation sense being one of their more sophisticated adaptations), on the other hand we are rarely aware of the importance of their sense of touch and kinesthetic in their communicative systems in such environment (see Kremers et al., 2016). It has been suggested that cetaceans utilize a scale-dependent, multimodal sensory system to assess and increase prey encounters (Torres, 2017). Moreover, from its early years to the present day, research in the field of cetacean communication have evidenced how both, motor behavior and sound, correlate in several contexts (see Herzing, 2015 for review).

Regarding their sense of sight, while it is true that the marine environment limits visual perception, especially as depth increases, in clear waters light can penetrate up to 200 m and depending on turbidity, the range of visibility may vary from centimeters to tens of meters (see Lammers and Oswald, 2015). In fact, cetacean sense of sight is pretty good and more developed (or conserved from their terrestrial ancestors) than commonly thought (Kremers et al., 2016). Accordingly, they use it in a variety of contexts, mediating prey capture and social interactions (see Kremers et al., 2016 for review). Regarding short-range communication dynamics, cetaceans use a variety of visual displays where postures are thought to signal intent and demeanor of the signal emitter (Würsig et al., 1990; Dudzinski, 1996) for example in reproductive contexts when dolphins present their genital region to sexually attract a mating partner (Tyack, 2000) or as indication of the affiliative bonding between individuals reflected by proximity and synchronous movements (Connor, 2007). They also use vision for the inspection of objects above water (Madsen and Herman, 1980), sometimes surfacing vertically and lifting the head out of the water to do it (spyhopping; e.g., Ford, 1984; Whitehead and Weilgart, 1991). Similarly, dolphins have shown understanding of human pointing gestures produced above the water (Herman et al., 1999; Xitco et al., 2001). It is important to note that most of the motor imitation studies reviewed, use conspecific or human actions in which most of the topography (or even all of it in the case of human models), has been demonstrated above the surface, which clearly requires good vision (Abramson et al., 2013, 2017, 2018; Herman, 2002. Bender et al., 2009).

Altogether, these observations suggest that social learning, particularly imitation, and synchrony are important communication and cooperative coordination signaling mechanisms for many odontocetes. As mentioned by Stephanie King and McGregor (2016), cetacean vocalizations facilitate social bonding and group synchrony, which makes it comparable to early human communication and other forms of primate vocal communication (like marmoset monkeys turn-taking), and similar to birdsong (see Hage et al., 2016; King and McGregor, 2016; Takahashi et al., 2016; Aboitiz, 2017).

A very interesting example of odontocete multimodal and synchronous social communication in orcas in controlled environments is given by Bowles et al. (2016). She and her collages described in two adult female killer whales two types of synchronous multimodal behaviors that consisted in bouts of stereotyped pulsed calls that occurred concurrently with bubble formations and nodding. As synchronous bubbling behaviors were linked with a functional vocalization type in both subjects, the authors suggested that elements of the vocal repertoire are marked for emphasis, for instance, by a bubble stream. Differential use of synchronous behaviors has a function in call learning. Vocalization rates in other odontocetes increase immediately before parturition and in the presence of calves, after parturition (Mann and Smuts, 1998; Mello and Amundin, 2005; Fripp and Tyack, 2008). This has been interpreted as a behavior that facilitates learning. They suggest that vocalizations are important to calf survival, such as the mother’s signature whistle or primary elements of a group dialect, could be emphasized with synchronous behaviors. This is a very interesting hypothesis that deserves to be explored in the future, not only in the linking between bubbling or other synchronous behaviors with functional vocalization types in the orcas but in other cetaceans as well (Bowles et al., 2016). Supporting these observations, recent experimental evidence suggests that during synchronous behaviors, dolphins use acoustic cues, and more particularly click trains, to coordinate their movements; possibly by eavesdropping on the clicks or echoes produced by one individual leading the navigation (Jaakkola et al., 2018; King et al., 2021; Marulanda et al., 2021).

Finally, although behavioral synchrony can be achieved by other social cognitive mechanisms such as response facilitation by priming (Byrne, 2009), cetaceans also display the capacity to reproduce a demonstrator’s multimodal imitation of novel motor actions and sounds. As mentioned earlier, several odontocetes have synchronous motor behaviors while in parallel learn their distinctive individual and/or group identification calls largely by social learning. This produces several motor and vocal idiosyncratic behaviors from early life that are supported by a multimodal social learning capacity in at least these two domains. Moreover, they also show vocal synchrony in some collaborative contexts. Which instances of motor and vocal social learning are done in a synchronic way remains an open question that certainly deserves attention in the near future.

Human language and music are multimodal processes involving not only vocalizations but also facial, hand and body gestures (Arbib, 2011; Cross et al., 2013; Aboitiz, 2017). This supports the notion that there was a close gestural-vocal coevolution in the human lineage (Aboitiz, 2017). There is increasing evidence that communication is multimodal in humans, apes and monkeys. It has been documented that apes can use different channels of communication in an opportunistic way (e.g., Leavens et al., 2004), and it may be that the multimodality that characterizes speech (and singing) has its origins in such capacities (Cross et al., 2013; Aboitiz, 2017). In chimpanzees, Taglialatela et al. (2008, 2011) have shown activation of Broca’s area homolog with the production of both manual and vocal communicative signals. Likewise, Willems and Hagoort (2007) have shown a similar situation in humans. Body gesturing not only accompanies speech, singing (and of course instrumental music), it also helps communication in a similar way as prosody supports speech. It has been pointed out that human communication is opportunistic and multimodal, and that our species uses any possible way to convey messages if we cannot speak (Aboitiz, 2013). Furthermore, speech rhythm is perceived multimodally, using both acoustic and visual cues, from observing lip movements (Peña et al., 2016; Michon et al., 2019, 2020). Importantly, the main point in human speech evolution it is not whether pantomime or vocal behavior was first but that it is only in the human lineage where a combination of learned gestures and vocal plasticity has taken place to develop a symbolic system, possibly fuelled by increasing imitative behavior in both systems.

Supporting this perspective, it has been argued that while birdsong focuses on mate attraction and territorial defense, vocal learning in other animals like parrots and specially cetaceans, promotes social bonds and behavioral coordination, a condition that may be closer to developing a primitive referential system (King and McGregor, 2016). Toothed echolocating cetaceans, probably like our ancestors, live in fission-fusion groups that come together and separate during foraging (Chereskin et al., 2022; Hersh et al., 2022). Their tonal sounds play a vital role in social communication and the sociality of these species has impacted the evolution of tonal sound complexity, as there is a correlation between social and whistle complexity (May-Collado et al., 2007). In large groups, where individuals are far from each other, these species engage in reciprocal vocal exchanges of whistles to establish social bonds, as these sounds replace body contact. Studies have shown that there is a significant correlation between increased tonal sound modulation and group size, and that changes in tonal sound complexity are closely tied to the different social branches of cetaceans (May-Collado et al., 2007).

The complexity that these tonal calls can reach is reflected in the idiosyncratic vocalizations from several cetaceans, which could be seen as proto referential signals that serve to identify group membership and even specific individuals (see Janik and Sayigh, 2013 for review). Field studies of orcas have shown that their call structure (usually referred as dialects) reflects relatedness and social affiliation (Deecke et al., 2010) and sperm whales’, codas may cue individual, unit and vocal clan identity (Schulz et al., 2008; Gero et al., 2016; Oliveira et al., 2016). It has recently been proposed that these “identity codas” may even function as “symbolic” markers of cultural identity among Pacific Ocean sperm whale clans, resembling human ethnolinguistic groups (Hersh et al., 2022). A special and, as far as we know, unique case of referential communication in the non-human animal kingdom, are the dolphins “signature whistles,” learned whistle distinctive to each dolphin, which allows them to recognize one another (see Janik and Sayigh, 2013). However, this signature call is not categorically different from other calls shared by all members of the group but is rather a variant of these. Recent research suggests that dolphin mothers start singing to their babies before they are born, apparently teaching them their signature whistle, a process that continues over the first weeks after birth (King et al., 2016). The signature call of dolphins appears to serve to maintain group cohesion, as individuals separated from the group emit their signature whistles while the others respond with their own signatures until they come together again. Moreover, it has been documented that they transmit identity information independent of the caller’s voice location and even after all voice features have been removed from the whistle signal, something only known to be done by humans (Janik et al., 2006). This is supported by the evidence in the laboratory that bottlenose dolphins can be trained to use novel, learned signals to label objects (Richards et al., 1984) and that they can remember these signature whistles for over 20 years (Bruck, 2013).

These facts support the hypothesis that whistles are referential signals for the representation of conspecifics, that is, they are vocal labels that could be used in a similar way as the use of names in humans, either addressing individuals or referring to them. Furthermore, a recent work has just demonstrated cross-modal, individual recognition not only by sound, but also by taste in bottlenose dolphins (Bruck et al., 2022). In this remarkable study the authors showed that dolphins recognize perception of identity of other individuals by gustation alone and then can integrate information from acoustic and taste inputs to identify specific individuals, indicating a modality independent, labeled concept for known conspecifics. These results show that dolphins form persistent modality-independent multimodal representations that have learned labels similar as in human concept formation (Bruck et al., 2022).

So far, the evidence suggests that vocal learning in cetaceans serves multiple purposes. For instance, the main four proposed functions of humpback song are attracting females to individual singers, determining or facilitating male–male interactions, drawing females to a male group within a lekking system (Herman, 2017), and as a long-range sonar (Frazer and Mercado, 2000; Iii, 2018). Toothed cetaceans primarily use learned vocal signals for maintaining social relationships, group cooperation, and coordination. They also use these signals to identify group membership (such as sperm whale codas and orca dialects) and specific individuals (dolphin signature whistles; see Janik, 2014; Whitehead and Rendell, 2014).

In humans, imitation is also likely to play a major role in the cultural inheritance of communicative and ritualistic behaviors that have an important effect on cooperation within groups (Cross et al., 2013; Heyes, 2021). Nonverbal communicative and ritualistic human behaviors may be targets of cumulative cultural evolution as they may promote group cohesion, supporting forms of cooperation such as long-term teaching and division of labor (Bender et al., 2009). More than grasping behavior, vocal behavior is what has made our lineage different from that of other apes, and probably our vocal system underwent rapid evolution in early stages of our lineage to support social cohesion and behavioral coordination like toolmaking or foraging over long distances. Accordingly, it has been suggested that common proximal factors for humans to communicate through language and music are language’s practical value in the organization of collective behavior and music’s significant role in eliciting and managing prosocial attitudes (Cross et al., 2013). The majority of human speech takes place in back-and-forth conversations. Dunbar (1998) suggests that mutual vocal exchanges in nonhuman primates are akin to grooming behavior, as individuals help each other remove parasites, thereby strengthening their bonds. It is believed that music originally served a similar purpose for human social bonding when group sizes became too large for grooming alone. Our sense of musical tonality may have evolved to recognize and process conspecific vocalizations, which are the most important tonal sound signals in the human environment (Purves, 2017). In this way, music served as a bridge to language as human speech became the primary means of group bonding and cohesion (Dunbar, 2012). This musical protolanguage (Darwin, 1871), which surely included melodic and rhythmic signals expressed through music and dance, was used for social cohesion and behavioral coordination. The melodic contours of early human vocalizations likely emerged as a form of prosody, with changes in pitch and intensity used to convey emotional signals, capture the listener’s attention, and enhance group coordination and cooperation (see Aboitiz, 2017).

Certainly, multimodal imitation in cetaceans convergently serve both motivational functions as well. In fact, similar assumptions have been made for cetacean dialects and other forms of idiosyncratic behaviors. Researchers have highlighted the strong tendency to imitate the actions of members of their own group in several cetacean species, which may fuel intergroup differentiation and intra-group identity (see Byrne, 2009; Whitehead and Rendell, 2014). For instance, it is believed that the dietary choices and foraging behaviors of orcas have a significant cultural component and that social learning plays a role in their development. This learning may involve the use of multimodal imitation for differentiating between different groups and establishing a sense of identity within each group (Barrett-Lennard and Heise, 2006). Certainly, orcas are remarkable conformists about the type of food they eat and the conspecifics with which they interact and reproduce (Baird and Whitehead, 2000; Barrett-Lennard and Heise, 2006). Moreover, evidence of teaching in foraging contexts has been reported in orcas and in Atlantic spotted dolphins (Guinet, 1991; Bender et al., 2009).