The dynamics behind diversity in suboscine songs Free

Birds can produce an impressive diversity of sounds, including tones, clicks, trills and screeches. In this Review, we will discuss some examples of how non-linear mechanisms contribute to the complexity of birdsong. We briefly touch on song production in oscine birds (i.e. songbirds; see Glossary), before focusing our attention on their sister clade, the suboscine birds (see Glossary). Both clades belong to the order Passeriformes. Suboscine birds include more than 1100 species (Tobias et al., 2012), whereas there are approximately 4000 oscine species. In contrast to oscines, which are vocal learners, suboscine passerines typically develop normal songs without learning. Several lines of evidence support the lack of vocal learning in suboscines: (1) suboscines display a general absence of vocal dialects (i.e. vocal variations across the geographical distribution of a species); (2) acoustic deprivation experiments show that their songs are largely innate (Kroodsma, 1984, 1985; Kroodsma and Konishi, 1991; Touchton et al., 2014); and (3) neuroanatomical research demonstrates that some species of suboscines lack the forebrain song nuclei used in vocal learning by oscine passerines (e.g. Gahr et al., 1993; Liu et al., 2013). However, there are some exceptions reported (e.g. Saranathan et al., 2007; Kroodsma et al., 2013). Thus, the suboscines provide the opportunity to study mechanisms of vocal differentiation that may occur without the vocal flexibility achieved through learning.

Birdsong produced by oscine birds is an ideal model for studying how a brain reconfigures itself during learning in order to achieve the complex motor tasks involved in vocal communication. The avian vocal organ, the syrinx (see Glossary), is quite similar across oscine species (Stein, 1968; Ames, 1971). The broad range of vocalizations displayed by these species is therefore believed to be associated with vocal learning and the complex and subtle neural control of the vocal organ achieved in the process. There has been extensive effort to unveil the nature of the neural coding of song in the oscine brain, where cortical and brainstem structures interact extensively in the generation of the instructions sent to the periphery in order to produce the song (e.g. Amador et al., 2017; Mooney, 2022; Sakata et al., 2020; Zeigler and Marler, 2008).

During avian vocalizations, the neural instructions engage the respiratory pathway in order to achieve the level of air sac pressure needed to generate sounds. Expiratory muscles compress the air sacs so air flows through the trachea and the vocal sources (see Fig. 1A). Oscine birds can produce morphologically rich pressure patterns, which are crucial in determining the tempo, timing, as well as – to some degree – the modulation of acoustic features (Hartley and Suthers, 1989; Goller and Suthers, 1996a; Mindlin et al., 2003; Amador and Margoliash, 2013). The neural instructions generated by the oscine brain during singing also reach the muscular apparatus acting on the syrinx. Fig. 1B shows a typical oscine syrinx with its complex set of muscles, allowing fine control of the sound source. Much research has focused on the action of the muscles determining the configuration of the syrinx and, therefore, the acoustic features of the sound (Goller and Suthers, 1996a,b; Suthers and Zollinger, 2004; Düring et al., 2013, 2017). This research suggests that much of the complexity in the song of oscine birds is rooted in the richness of the neuromuscular control (Amador et al., 2017).

For the suboscines, the situation is quite different. Unlike their sister group, the suboscines possess a less complex set of syringeal muscles but display remarkable morphological diversity of the syrinx (Ames, 1971; King, 1989; Goller et al., 2021). Recent studies on different suboscine species suggest that diversification of acoustic features in their vocal repertoires might have occurred through different morphological adaptations or adaptations of the way in which the elements of the vocal organ are used (Garcia et al., 2017; Goller et al., 2021). In the absence of vocal learning, such adaptations allow suboscines to generate a wide range of acoustic features. This is in contrast to oscine birds, where such diversity is mainly achieved through diverse and rich neuromuscular control as the vocal organ is highly conserved across oscine species. The comparative study of these phylogenetically close groups allows us to study the evolutionary impact of the presence or absence of vocal learning.

Here, we first discuss the mechanisms of vocal production in oscines, before considering some case studies in suboscine birds. When discussing suboscines, we focus on one specific acoustic feature of the vocalizations: slow soundwave modulations coexisting with the high-frequency oscillations that are typical of avian vocalizations (kilohertz range). As we will show, different suboscine species achieve this acoustic feature through a variety of anatomical strategies or by taking advantage of the non-linearity of the vocal source. The overall aim of this Review is to highlight how suboscine species use biomechanical adaptations to generate very rich and diverse songs.

Until the late 1990s, there was debate regarding the phonation mechanisms of songbirds (see Box 1). However, nowadays there is consensus on the idea that singing oscine birds use the same physical mechanism as humans when producing voiced sounds (Mindlin and Laje, 2006; Riede and Goller, 2010; Titze and Martin, 1994): the oscillation of soft tissues modulates the air flow, generating sound that is then filtered by the upper vocal tract. The syrinx of oscine birds contains two sets of labia, i.e. two pairs, each consisting of the lateral labium and medial labium (LL and ML in Fig. 1C,D). When the bird is breathing silently, the labia are so far apart that it is physically impossible to induce their oscillation with the expiratory airflow (Fig. 1C). In order to sing, the bird activates syringeal muscles that rotate the third cartilage in such a way as to be carried inward. In this way, the lateral labia are positioned in the bronchial lumen in a ‘pre-phonatory’ position (Goller and Larsen, 1997; Suthers et al., 1999; Larsen and Goller, 2002; Fig. 1D). The closeness of the labia allows auto-sustained oscillations to begin when the air flow reaches a sufficiently high pressure, thus generating sound (for more details, see Mindlin and Laje, 2006). In this way, sound is generated as the airflow is modulated by the labial oscillations. The pressure fluctuations are then filtered by the trachea, the oroesopharyngeal cavity (OEC; see Glossary) and the beak. Riede, Suthers and collaborators showed that the OEC is dynamically adjusted in order to emphasize the fundamental frequency (see Glossary) in some songbirds, thus generating more ‘tonal’ sounds (e.g. Riede et al., 2006; Riede and Suthers, 2009). Songbirds can very rapidly modulate the fundamental frequency of their sounds; thus, the underlying volume changes in the OEC need to occur very fast. This is observed for the tonal songs of northern cardinals (Cardinalis cardinalis) and white-throated sparrow (Zonotrichia albicollis), for example. In these birds, fine motor control of the syringeal muscles is coordinated with fine motor control of the upper vocal tract. An alternative type of song is generated by zebra finches (Taeniopygia guttata); these songs contain syllables that have a wide spectral content. For these birds, frequencies between 2.5 and 5 kHz are generally emphasized, presumably as a result of the filtering properties of the OEC (Riede et al., 2013).

The vocalizations of oscine species contain a wide range of timbres (see Glossary) that are compatible with the mechanisms described above. But as we explore outside the oscine world, we find a variety of strategies to control the vocal organ, and anatomical adaptations that further expand the timbrical landscape. Below, we describe three examples of sounds produced by suboscines that illustrate different control mechanisms used during song production, all of which involve oscillating membranes. These examples show similarities and differences when compared with song production by oscine birds. Because oscines and suboscines are so closely related, these comparisons may shed light on the evolutionary strategies used to achieve variability in birdsong production in the absence of learning.

Below, we review three examples of passerine suboscine vocalizations: (1) the white-tipped plantcutter (Phytotoma rutila), (2) the great kiskadee (Pitangus sulphuratus) and (3) the tracheophones, a group that includes the rufous hornero (Furnarius rufus). As discussed above, when considering suboscine vocalizations, we will focus on one specific acoustic feature: the appearance of sound modulations with frequencies between 100 and 200 Hz coexisting with high-frequency (i.e. of the order of kilohertz) components. This feature is easy to identify in a song, and is present in several suboscine species, making it ideal for comparing across species.

The white-tipped plantcutter (Fig. 2A) is a passerine bird belonging to the Cotingidae family. It is distributed throughout south-central South America, particularly from western Bolivia to northern and central Argentina, and also in some parts of Paraguay and Uruguay and extreme southwestern Brazil (Rodríguez-Cajarville et al., 2019). One of the most notable characteristics of this species is its vocalization, which consists of a rough, long note, with a sound comparable to that of a rusty hinge (see spectrogram in Fig. 2B; an example recording is available from www.xeno-canto.org/50074). Fig. 2C shows the initial fragment of the vocalization, which consists of a succession of short sound fragments of increasing speed. The production rate of these fragments is of the order of 100 Hz. A detailed inspection (Fig. 2D) shows that each sound fragment is made of fluctuations of a few kilohertz, the amplitudes of which decrease significantly before the start of the subsequent fragment. This observation is key to suggesting the mechanism behind this vocalization.

In order to explore the possible mechanisms used to generate this sound, dynamical systems models are very helpful. A dynamical systems model can be used to capture the most relevant features of the behavior in a given biomechanical process; it can therefore be used to analyze the viability of any proposed mechanism. Within this framework, a dynamical system consistent with a given biomechanical process can be used to synthesize artificial sounds that can be compared with the sound under study. For example, Fig. 2E shows a sequence of pulses generated by a biomechanical model representing the sound source of the white-tipped plantcutter vocalization. The vocal production of low-rate pulsating sounds requires simple motor gestures, such as coordinating respiratory muscles that will interact with the tissue membranes of the vocal source (i.e. the syringeal labia in this case). The physical mechanism used to generate pulse-like sound consists of holding the syringeal labia together while increasing the pressure in the air sac system through the activation of the expiratory muscles. In this way, sub-syringeal pressure builds up until the force exerted on the labia can overcome the force exerted by the syringeal muscles that hold the labia together. When the air flow is established, the pressure between the labia decreases (according to Bernoulli's principle; see Glossary), the labia collide against each other, and the process begins again (for further details; see Gardner et al., 2001; Amador and Mindlin, 2008; Mindlin, 2017; Perl et al., 2011). This mechanism has been directly observed in birds during sound generation (Jensen et al., 2007; Goller and Riede, 2013). Note that there is a big difference between a pulse sequence generated by the sound source (Fig. 2E) and the sound displayed in the vocalization (Fig. 2D). This is because of the effect of the OEC (see Fig. 1A), which filters the sound signal generated by the syringeal labia. This effect is illustrated in Fig. 2F, which shows the result of filtering the series of pulses in Fig. 2E with a biomechanical model using a Helmholtz resonator (see Glossary) to represent the OEC. The details of the dynamical systems model used to synthesize the pulses and the acoustic filters of the vocal cavities are described in Uribarri et al. (2020).

Note that it would be possible to produce a vocalization similar to that of the white-tipped plantcutter by an alternative mechanism. In fact, there are numerous species that can induce oscillations in the labia of the order of kilohertz, as well as an oscillation of less than 200 Hz. This pattern of vocalization can result from the action of ultra-fast muscles (Elemans et al., 2008). However, this mechanism is unable to account for the specific decay in the amplitude of each of the sound segments that compose the vocalization of the white-tipped plantcutter. In fact, indirect evidence supports the hypothesis that the fast oscillatory component is due to the effect of a passive sound filter (see Glossary). This evidence has been obtained from the analysis of songs from individuals recorded in different habitats. If the OEC is the main determinant of the passive filtering, variations in animal size would be reflected in changes in the vocalization frequency, because commensurate changes in OEC size would modify its resonant frequency (see ‘acoustic resonance’ in Glossary). In Uribarri et al. (2020), the frequencies in the kilohertz range were analyzed from different individuals from the xeno-canto database (https://www.xeno-canto.org/) and the Macaulay Library at the Cornell Lab of Ornithology (https://www.macaulaylibrary.org/). For each vocalization, the mean peak frequency was noted, along with the location where the vocalization was recorded. A simple analysis showed a linear correlation between mean peak frequency (range: 3.5–4.1 kHz) and habitat altitude (range: 0–3500 m above sea level). This is significant because of Bergmann's rule (Meiri and Dayan, 2003), which states that populations of larger size are found in colder environments.

To estimate body size, in Uribarri et al. (2020), lengths of tarsometatarsus bones from museum specimens were analyzed. The tarsometatarsus (often referred to as ‘tarsus’) is a bone that is found in the lower leg of birds and some non-avian dinosaurs (Proctor et al., 1993), and its size correlates with animal size. In Uribarri et al. (2020), it was found that the tarsi lengths correlated with the altitudes at which the specimens were collected. For animals captured between 0 and 3500 m above sea level, the change in size was approximately 13%. This result is consistent with Bergmann's rule (Rodríguez-Cajarville et al., 2019). Then, the frequencies of vocalizations were analyzed, and a linear correlation was found between the resonant frequency and altitude. The change in the whole range of the mean peak frequency was 13.5%. Beyond building confidence in the biomechanical model and in the fact that the high frequencies in these vocalizations are the result of filtering signals from a sound source with a pulsatile structure, this result suggests that this species is capable of reliably transmitting information about body size through acoustic properties. This may not be the case for many oscine species, for which the correlation between acoustic features and size tends to be weak and variable (e.g. Cardoso, 2012; Linhart and Fuchs, 2015; Liu et al., 2017). Moreover, songbirds use their ability to adjust filter properties dynamically, e.g. by changing the OEC volume during singing (e.g. Riede et al., 2006), which obscures the relationship between vocalization frequency and size.

It is noteworthy that the high frequencies of the white-tipped plantcutter call do not show significant modulations during vocalization. In other words, the filter is kept constant while a smooth modulation occurs in the rate of generation of sound segments; that is, of the pulses that will be filtered by the cavity. In other suboscine species, this modulation is generated by varying the pressure of the air sacs (Amador et al., 2008), which is consistent with the observation that in the white-tipped plantcutter, the first and last pulses are those generated at a lower firing rate. In this way, Uribarri et al. (2020) show that in the white-tipped plantcutter, the low frequency is associated with the series of pulses generated by the syringeal labia, produced at a very low rate that correlates with the air sac pressure; by contrast, the high-frequency component is due to the passive filtering of these pulses by the OEC.

The great kiskadee (Fig. 3A) is another example of a suboscine species that produces songs in which low-frequency components coexist with frequencies of the order of kilohertz (example recordings available from www.xeno-canto.org/272848). Here, we will analyze its song and the dynamical origin of these frequency components.

The great kiskadee is a species of passerine bird belonging to the Tyrannidae family. It is native to tropical America (Neotropics), and widely distributed from southern USA to central Argentina. In 2008, a rather surprising result was published: the modulation of the fundamental frequency (see Glossary) of the great kiskadee song does not seem to require the activation of syringeal muscles (Amador et al., 2008). Indeed, after the syringeal muscles were inactivated, by sectioning both branches of the tracheosyringeal nerve, the birds did not show drastic changes in their songs. Specifically, the fundamental frequency modulations of the vocalizations remained unchanged. The result was surprising, because in oscine birds, the syringeal muscles play an important role in controlling the phonology of the vocalizations. In a foundational piece of work, Goller and Suthers (1996b) showed that electromyographic (EMG) activity in the muscle syringealis ventralis (vS in Fig. 1B), the largest syringeal muscle, increased exponentially with the fundamental frequency of the generated sound and also correlated with the frequency modulation. Moreover, if syringeal muscles are inactivated in oscine birds, the songs suffer major acoustic changes and birds may have difficulty breathing (Goller and Cooper, 2004; Bhama et al., 2011). Although the functions of the different muscles seem to be more complex than indicated in the seminal studies (e.g. Düring et al., 2013), the activity of the ventral syringeal muscles correlates with the fundamental frequency in oscine birds (Goller and Suthers, 1996a,b; Suthers et al., 1999). By contrast, the work of Amador and collaborators (2008) with great kiskadees found something very different: a linear correlation between the modulation of the fundamental frequency and the modulation of the air sac pressure measured during singing. This suggested a novel mechanism for pitch modulation during song production in birds. Some years later, the same mechanism was found in zebra finches (Amador and Margoliash, 2013), which, in comparison with great kiskadees, have a much more complex syrinx musculature (see Fig. 1B) involved in the phonology of the vocalizations. Moreover, the mechanism for frequency modulation through pressure is also found in humans (Titze, 1989). Thus, pitch modulation through modulation of the respiratory pressure potentially represents a general mechanism for frequency modulation when sound is generated by oscillating membranes. The linear correlation between air sac pressure and the fundamental frequency of the vocalizations explains how pitch could be modulated with no activation of the syringeal muscles. This result is particularly intriguing as great kiskadees possess large ventral syringeal muscles. What, then, is their role?

To answer this question, Döppler et al. (2020) recorded vocalizations simultaneously with EMG activity in the syringeal muscles of great kiskadees (more precisely in the obliquus ventralis muscle; ovm). Fig. 3B–D shows three repetitions of the great kiskadee song. Fig. 3B shows a soundwave of the song recorded with a microphone, Fig. 3C shows its spectrogram, and Fig. 3D shows the corresponding electrical activity in the ovm. The most noticeable muscle activity (Fig. 3D) appears simultaneously with a subtle acoustic property: a ‘roughness’ in the fundamental frequency (see syllable 1 in Fig. 3C). This acoustic property results from a low-frequency modulation of the amplitude of the sound signal, as shown in Fig. 4. The simultaneous study of the EMG activity shows a pattern of activity occurring at the same rate as the modulation of the sound envelope. However, the action of the ovm cannot be solely responsible for this acoustic feature: after resection of the tracheosyringeal nerve (which inactivates the ovm), the amplitude of the modulations decreases, but this feature does not disappear (see details in Döppler et al., 2020).

In order to test the hypothesis that roughness is actively controlled by the ovm during song production in kiskadees, a dynamical systems model was developed (Döppler et al., 2020). Tyrannid suboscines have two sound sources, each one at the end of the bronchi, where they meet to form the trachea (Ames, 1971). Each sound source contains membranes (labia) that can perform auto-sustained oscillations if enough energy is fed into the system by the air flow of an expiratory pulse. Consistent with previous studies, it was assumed that each sound source could be modelled as a non-linear oscillator (see Glossary; for detailed explanations and equations, see Amador et al., 2008; Mindlin and Laje, 2006). The properties of the modelled oscillators allow slow modulation of the air sac pressure to be transduced into low-frequency modulation of the song (see Döppler et al., 2020, for further details). The model also assumed (1) a coupling between the two sound sources, as the medial labia are supported by a common bone structure, and (2) slightly different elastic properties for each of the two labia. Thus, the physical architecture of the kiskadee vocal source shows important similarities with the oscine syrinx. Yet, its control is quite unique. Döppler et al. (2020) tested the hypothesis that it was possible to reproduce all the acoustic and physiological properties of the kiskadee song by assuming two out-of-tune sound sources driven at a specific frequency (see Doppler et al., 2020, for modeling details).

The biomechanical model presented in Döppler et al. (2020) allowed several observations to be accounted for. If the parameters under which each of the sources operated were slightly different, the model could explain why it is possible that the modulations appear in the first syllable, as it is this syllable that is generated with the highest values of air sac pressure. In this dynamical system, the frequencies of the oscillators depend on the pressure: the higher the pressure, the greater the difference between the frequencies of each oscillator. When two oscillators are coupled, the smaller the frequency difference between them, the easier it is to keep them synchronized. Thus, based on this model, the first syllable, and a small segment of the second syllable, are likely to show an amplitude modulation, because of the difficulty in synchronizing two oscillators with very different natural frequencies. These modulations, a product of the lack of synchrony between the sound sources, are emphasized by the action of the ovm. In this way, the results from experiments and mathematical simulations suggest that the ovm produces a controlled detuning during the execution of the first syllable of the song.

Within the suboscines, a group of birds called the tracheophones possess a vocal organ characterized by having a pair of membranes, one ventral and one dorsal, in the trachea, just above where the bronchi meet to merge into the trachea (Ames, 1971; Garcia et al., 2017). These tracheal membranes, the membrana trachealis (MT; shown in Fig. 5), were thought to constitute the sound source of their vocalizations, mostly for anatomical reasons (Rüppell, 1933). However, we now know that the mechanism underlying the vocalizations of tracheophones involves more complex dynamics. Garcia et al. (2017) showed that tracheophones actually have three sound sources: in addition to the MTs, these birds have two pairs of labia at the junction of the bronchi and trachea (see the two pairs of bronchial labia in Fig. 5; BL). These three pairs of oscillating membranes anatomically constitute three sound sources. Garcia et al. (2017) designed a set of experiments to test the hypothesis that all three sound sources are active during vocalizations. They used a fiberscope for direct visualization of the membranes while air was injected into the respiratory system to induce phonation (Fig. 5B). Under certain flow conditions, oscillations were induced in the MTs. In addition to these membranes, the BL were thought to also participate in phonation. To test this hypothesis, Garcia et al. (2017) used two strategies. Firstly, while injecting air into the respiratory system (at a flow level high enough to induce phonations), the fiberscope was positioned in such a way to obstruct the oscillations of the MTs, while at the same time allowing observation of the BL. The BL maintained their movement as the sound in these induced phonations continued. Secondly, the elastic properties of the MTs were altered by applying tissue adhesive. This manipulation did not prevent vocalizations; however, it did cause changes in the acoustic properties of the vocalizations, validating the idea that the three sound sources interact to generate the vocalizations of the species.

In all six species of suboscines studied in Garcia et al. (2017), the intact syrinx produced low-frequency sounds, characterized by pulse-like signals (see Fig. 6A, MT intact, ‘pulsations’). Here, we show as an example some of the results for the rufous hornero (Fig. 5D), a medium-sized ovenbird belonging to Furnariidae tracheophones (typical duet between male and female rufous hornero available from www.xeno-canto.org/748509). When the MTs of these birds were non-functional, there were three important changes: (1) the ‘pulsatile’ characteristic of the oscillations was lost, (2) the sound amplitude decreased, and (3) the frequency of the oscillations increased (see Fig. 6). It is noteworthy that the presence of pulsatile sounds is not only obtained by inducing vocalizations in anesthetized animals when air is injected into the respiratory system. The spontaneous distress vocalizations of the several species studied in Garcia et al. (2017) also contain important modulations in the low-frequency range (similar to the pulsation modulations shown in Fig. 6A, MT intact), and these disappear when the tracheal oscillations are prevented by surgical manipulations.

In addition to the experimental work discussed above, the plausibility of the proposed mechanisms of vocalization in tracheophones was explored using a biomechanical model of the vocal organ. All of the experimental observations were reproduced using a dynamical systems model (Fig. 6B). The model was designed to generate synthetic sound with two different types of oscillators representing the BL and the MTs. This model assumed that the two bronchial sources were synchronized, and therefore the two sets of BL were represented using one oscillator. This is possible because, if two identical sound sources are synchronized, they generate the same airflow modulation at the same time; therefore, the result is indistinguishable from having one or two oscillators in the biomechanical model. This was the simplest biomechanical model that allowed the data to be reproduced. It is possible that for other vocalizations, the two pairs of BL are out of synchrony or are controlled independently, but this was not observed for the distress call in the studied species in Garcia et al. (2017). In this model for sound production, if the frequencies of the two oscillators (representing BL and MTs) were sufficiently different, the lower frequency of the MTs modulated the higher frequency of the BL sound source (Fig. 6B, MT intact, ‘pulsations’). Indeed, this is seen in the pronounced amplitude modulation in the spontaneous calls of the rufous hornero (see Garcia et al., 2017), and in the induced vocalizations (Fig. 6A, MT intact). By contrast, if the frequencies were similar to each other, the two sources become ‘locked’, and the resulting vibrations were of lower frequency than if the labial source were to vibrate on its own (Fig. 6B, MT intact, ‘locking’). This is consistent with the observed shift to higher frequencies after the MTs were disabled across the different species studied, as well as with the reduced pulse-like quality of the vibrations (Garcia et al., 2017). The pulsatile oscillations arise from the non-linear interaction of two oscillators representing the MTs and the BL. If the oscillator representing the MT is removed from the system, the pulsations and locking disappear (Fig. 6B, MT disabled).

In this way, the interactions between the BL oscillations and those of the MTs constitute a morphological solution to the generation of intense sounds, with modulated amplitude, as well as pulse-like properties. In oscine species, many of these properties are achieved by neuromuscular control of their sound sources (Beckers, 2011; Goller and Riede, 2013; Suthers and Zollinger, 2008). The suboscines do not have the same level of muscular control, but work on tracheophones shows that they are capable of generating similar acoustic properties using specific morphological features.

In recent years, much progress has been made in understanding the physical mechanisms and dynamics involved in the generation of song by oscine birds. In parallel, a lot of effort has been applied to the study of the neuronal substrate that controls the vocal apparatus. Both the neuronal architecture underlying the control of the vocal apparatus and the anatomy of the syrinx are highly conserved across oscine species.

The picture is quite different when it comes to the sister clade of oscines, the suboscine birds, which are thought to be, in their great majority, non-vocal learners. These birds display a remarkable anatomical diversity, and the mechanisms used in the generation of their songs are extremely varied. To illustrate this phenomenon, here, we have concentrated on the study of a particular acoustic property: the existence of slow modulations (100–200 Hz) coexisting with fast oscillations (kHz range). This is an acoustic feature that is clearly identifiable and is present in several suboscine species. We have discussed the dynamical origin of this feature in three different cases: the screeching sound of the white-tipped plantcutter, the carefully controlled detuning between the two sound sources in the kiskadee's first syllable, and the sounds generated by the complex anatomical structure of the tracheophones. In all three examples, some set of properties in the sound signals suggested a starting point for building a non-linear dynamical model, which then generated quantitative predictions that could be used to test the assumed hypotheses regarding the mechanisms underlying the vocalizations. Taken together, these three stories – which integrate biological experiments and biomechanical modeling using non-linear dynamical systems – illustrate how a morphological or dynamical adaptation can produce a complex acoustic property without the need for complex neuromuscular control.

An emerging hypothesis posits that it is through their different adaptations that suboscines are able to achieve a diverse range of acoustic properties in their repertoires. It is possible that it is precisely because of the lack of vocal learning that the suboscines must rely on morphological adaptations to produce acoustic characteristics that oscines can achieve through more complex neuromuscular control.

We thank Franz Goller for continuous discussions and common work on this subject.