Duetting behavior occurs at both the individual and pair levels (Levin, 1996), while Logue (2006) studied duetting from an operational perspective in which two individuals establish a shared set of rules. This led to the notion of a “duet code”—a set of answering rules one individual uses to answer its mate’s song (Logue, 2006; Logue et al., 2008). While a duet is a pair-level property, a duet code is an individual attribute, and answers according to a duet code “adhere” to that code (Logue and Krupp, 2016). At its simplest, a single pairing rule, such as ‘‘answer F1 to M1,’’³ generates the cyclical duet [i-n(M1-F1)]⁴ produced by many songbirds (Levin, 1996; Rogers, 2005). A more complex duet code, such as “answer F1 to M1, F2 to M2, and F3 to M3,” generates a non-repeated duet [i-(M1-F1-M2-F2-M3-F3)], as produced by an African weaver bird endowed with such a large syllable repertoire that both partners constantly switch between syllable types (Voigt et al., 2006; Lemazina et al., 2021). Logue’s duet code concept opened up new avenues for measuring how code complexity and adherence vary across species (Logue and Krupp, 2016), whether duet codes are pair-specific (Mennill and Vehrencamp, 2005; Templeton et al., 2013), whether one sex or both adhere to these codes (Mann et al., 2003; Rivera-Cáceres, 2015), and whether duet codes emerge spontaneously in newly formed adult pairs or require vocal practice (Levin, 1996; Rivera-Cáceres et al., 2016). This begs the question: do young birds learn duet codes from their elders?

Evidence that duet codes are learned from adults comes from observations of juveniles singing alongside their parents (Farabaugh, 1982; Hall, 2009). Such collective singing presumably allows juveniles to gain duetting experience, which not only requires learning what to answer and when but also mastering the duet rhythm in coordination with breathing given the rapid alternation (2–5 Hz) of male and female syllables (Hoffmann et al., 2019; Coleman et al., 2021). For example, song coordination in juvenile canebrake wrens improves over time via parental influence and independently of maturational effects, indicating a learning process (Rivera-Cáceres et al., 2018). Whether song acquisition results from copying a same-sex parent or integrating auditory information from both parental “tutors” remains unknown. There may also be alternative modes of code development with age. For example, a code might be retained throughout life (“close-ended”), whereby phrase-pairing rules remain constant regardless of partner identity (Levin, 1996); alternatively, mature individuals might re-learn a code each time they acquire a new mate (“open-ended”; Wickler, 1980). In the case of canebrake wrens, different pairs have distinct duetting rules, suggesting that learning in adulthood is likely. Indeed, removing and translocating individuals of well-established pairs confirmed that adult wrens re-learn pair-specific duet codes after re-mating, with males showing more flexibility in phrase-pairing rules than females (Rivera-Cáceres et al., 2016). Consequently, Rivera-Cáceres et al. (2018) proposed a three-step model for duet learning: (1) memorizing song material from auditory exposure, (2) rehearsing duet songs with both parents, and (3) re-learning to coordinate songs with a breeding partner (Figure 1A). Whether these two latter forms of sensorimotor learning share the same neural connections is the subject of future research (Nieder and Mooney, 2019).

Many songbirds co-sing in rapid turn on a syllable-to-syllable basis with sub-second latencies (Mann et al., 2009; Fortune et al., 2011; Rivera-Cáceres, 2015). To achieve such tight coordination, individuals rely on sensory information originating from two sources of auditory feedback—one generated by the bird’s own voice (autogenous) and the other from its singing partner (heterogeneous). Owing to the velocity of sound, the longer the distance between the duetters the longer the delay for both receivers. Duetting songbirds adapt to these delays by altering the timing of their singing (Fortune et al., 2011) or by using visual cues in open habitats (Rek and Magrath, 2016, 2020). How, then, is auditory feedback encoded in the brain?

Neuroanatomical studies of duetting songbirds reveal the presence of well-developed brain nuclei dedicated to song production learning in both sexes, which contrasts with the females of species in which only males sing (Nottebohm and Arnold, 1976; Brenowitz and Arnold, 1986; Deng et al., 2001; but see Lobato et al., 2015). Research into the neural underpinnings of antiphonal duetting targets the HVC (used as a proper name), a high-order forebrain song nucleus involved in sensorimotor learning (Nieder and Mooney, 2019). Contrary to neurophysiological data obtained for songbirds in which only males sing, extracellular recordings in the HVC of anesthetized wrens show strong responses to auditory presentations of both male and female song when played in isolation (Fortune et al., 2011). Furthermore, experimental manipulation of song stimuli shows a sensitivity of HVC neurons to inter-syllable intervals. Importantly, the response strength of HVC neurons to duet stimuli exceeds the sum of neuronal responses to each individual’s song. This suggests that each participant not only knows what to sing but also develops an internal representation of the pair-specific duet (Fortune et al., 2011). Groundbreaking work in free-ranging African weavers further demonstrates the alternation of neuronal activity in each partner’s HVC, with bursts temporally locked to syllable onsets (Hoffmann et al., 2019). This “on-off” pattern appears to be regulated by heterogeneous auditory feedback that reciprocally inhibits HVC premotor activity (Coleman et al., 2021). Such brain-to-brain coupling mechanisms ensure precise timing of dyadic vocal interactions, most likely through gamma-aminobutyric acid-ergic inhibition (Benichov and Vallentin, 2020). For comprehensive reviews on this topic, see Elie et al. (2019) and Rivera-Cáceres and Templeton (2019).

Neighboring groups of singing primates often call antiphonally (Kinzey et al., 1977; Marler and Tenaza, 1977; Raemaekers and Raemaekers, 1985) and counter-sung solos and duets are longer than solos and duets sung alone (Tenaza, 1976; Mitani, 1985). In support of the flexible timing of vocal output, active counter-singing and singing motivation have been experimentally corroborated (Chivers and MacKinnon, 1977; Mitani, 1988; Dooley and Judge, 2007). Studies of communication networks showing that siamangs are sensitive to their neighbors’ group disruption (Morino et al., 2021) are likely to unveil further instances of vocal flexibility in the future.

Individual gibbons flexibly time their contributions relative to their mates’ during the great-call sequence. Guided by subtle changes in female introductory notes that signal an impending great call, the male suspends phonation; cued by her post-climax descending notes, he resumes singing with a coda phrase according to a precise turn-taking pattern (Terleph et al., 2018a). Flexibility is needed given individual variability in the female great call (Terleph et al., 2015, 2016). Adjustment made by hylobatids in response to a mate’s vocal “mistakes” are termed “repairs,” a universal principle of human conversation (Schegloff et al., 1977; Dingemanse et al., 2015). Repairs have been scrutinized for self-corrected, stalled, and aborted great calls (Haimoff, 1988; Haraway and Maples, 1998; Terleph et al., 2018a). Such studies confirm the existence of duet codes and answering rules in lesser apes. Non-adherence to the duet code (e.g., production of atypical notes or unexpected call timing) may result in duet interruption and song reset by the mate.

Coordinated singing and rhythm dynamics are not necessarily tied (Ravignani et al., 2014). For example, inter-onset call intervals extracted at each level of the indri’s song organization (i.e., units and phrases), reveal music-like categorical rhythmicity (De Gregorio et al., 2021a). Both in adults and young individuals, females exhibit more flexibility than males, with a sensitivity to chorus size (Gamba et al., 2016; De Gregorio et al., 2019, 2021b). Sex-related “divergence” in indri song rhythm contrasts with titis and tarsiers. In a cross-sectional study of duetting pairs of titi monkeys, partners were found to adjust pulse rate and phrase duration to one another, showing call “convergence” (Clink et al., 2019). A longitudinal study with newly formed pairs of titis might establish whether vocal learning is involved through convergence in the spectral features of calls, as reported in marmosets (Elowson and Snowdon, 1994; Snowdon and Elowson, 1999; Zürcher et al., 2021). Likewise, male and female tarsiers flexibly adjust call rhythm relative to their partner through simultaneous accelerations and decelerations (Clink et al., 2020). Within-pair convergence in duet tempo might be achieved by entrainment, i.e., spontaneous responsiveness to a perceived rhythmic signal (de Reus et al., 2021).

Immature individuals singing jointly with their elders have long sparked research attention (Deputte, 1982; Raemaekers et al., 1984; Pollock, 1986; Reichard, 2003). A longitudinal study of mother-daughter vocal interactions in gibbons revealed the acquisition of correct note sequencing over time (5–30 months; Merker and Cox, 1999). In a cross-sectional study of free-ranging family groups, an inverse relationship was found between mother-daughter co-singing rates and call synchronization; less proficient daughters co-sang at higher rates. Interestingly, mothers adjusted their song to a more stereotyped pattern when co-singing than when singing alone, suggesting a “teaching role” of mothers (Koda et al., 2013). While sexually mature females sing an adult-like maternal song (Brockelman and Schilling, 1984; Merker and Cox, 1999; Koda et al., 2013), males master the multi-part coda phrase years later (Hradec et al., 2021) via an intriguing developmental trajectory (Figure 1B).

Spontaneous production of a female-like great call by immature males has been reported in several gibbon species (Koda et al., 2014; Hradec et al., 2016, 2017, 2021). A triggering role of the maternal call in young males, possibly associated with low androgen levels, has also been invoked (Koda et al., 2014). Immature individuals producing male calls potentially face aggression from the father (Hradec et al., 2021) and there is evidence that the stress hormone cortisol may negatively interact with testosterone in influencing the expression of secondary sexual traits (Puts et al., 2016). Close monitoring of hormone levels would be worthwhile in order to determine the impact of parent–offspring relationships on gibbon song development (Burns and Judge, 2016).

To reproduce outside their natal groups, mature individuals must coordinate their song with a prospective mate “having both different genetic parentage and a different history of developmental experience than their own” (Haraway and Maples, 1998). In indris, spectral-temporal features of descending phrases correlate with genetic distance in males, whereas females are less constrained (Torti et al., 2017). Thus, indri choruses may inform conspecifics about individuals’ genetic relatedness. Such an effect is less apparent in titi duets (Clink et al., 2022). Consistent with vocal flexibility and duet code learning, the duets of long-term mates are better coordinated than those of newly formed pairs (Geissmann, 1986, 1999; Maples et al., 1989; Müller and Anzenberger, 2002).

Robust conditioned responses are obtained in lesser apes via reinforcement and extinction procedures in which song presentation is contingent upon an individual’s own vocalization (Haraway et al., 1981; Maples and Haraway, 1982; Maples et al., 1988). Moreover, both in lemurs and gibbons, phonation can be brought under volitional control in response to an arbitrary visual signal (Wilson, 1975; Koda et al., 2007), thus demonstrating voluntary control over call timing.

Captive siamangs can alter their calls using various “tricks,” including the production of hand- modulated and echoing calls (Badraun et al., 1998). Geissmann (2009) observed one female gibbon who amplified her duet contribution by slamming the sliding door of her sleeping quarters at the climax of her great call.

As renowned “soprano singers” (Koda et al., 2012), gibbons produce pure-tonal melodious song that requires appropriate hormonal and neural machinery for pitch control. Contrasting with humans, however, higher androgen levels result in calls with a higher pitch (Barelli et al., 2013; Puts et al., 2016). Experiments in a helium-oxygen atmosphere revealed that the unshifted call fundamental frequency is strongly attenuated and the first harmonic is emphasized, suggesting that the sound source (larynx) operates independently of the supralaryngeal vocal tract (Koda et al., 2012). Thus, call flexibility can be achieved by controlling laryngeal function and/or the resonance filter configuration (Gamba et al., 2011, 2017; Fitch et al., 2016), but the challenge is to account for the larynx development (Zhang et al., 2020). Importantly, bipolar excitation in the inferior portion of the precentral gyrus in the left hemisphere yields adduction of the vocal folds (Mott et al., 1911). This suggests that, in the gibbon brain, a direct pathway exists from the laryngeal representation in the primary motor cortex to the laryngeal motoneurons of the nucleus ambiguus, which controls the muscles of the larynx for vocal production (Simonyan, 2014). This might explain why gibbons can be trained to call on command (Koda et al., 2007; but see Hage and Nieder, 2013).