Functional morphology of the Alligator mississippiensis larynx with implications for vocal production

Crocodilian vocal repertoires consist of a diverse array of sound types (Herzog and Burghardt, 1977; Campbell, 1973; Garrick and Lang, 1977; Watanabe, 1980; Hunt and Watanabe, 1982; Vliet, 1989; Britton, 2001; Vergne et al., 2009; Wang et al., 2007; Todd, 2007; Wang et al., 2009; Sicuro et al., 2013) but little is known about how such variability is generated. Here, we focus on the morphology and performance of the structures that determine fundamental frequency (F₀) – a critical acoustic parameter that can reveal relevant information about the sender, and is therefore a prime selective target. To date, our understanding of the underlying mechanism of F₀ control in sauropsids is primarily limited to birds (Riede and Goller, 2010; Goller and Riede, 2013), whose syrinx-driven vocalization differs significantly from that of basal sauropsids. The American alligator is an ideal model system for elucidating the mechanism of vocal production in basal saurposids. It has a diverse vocal repertoire that, like mammals, is generated by the larynx. In mammals, laryngeal sound production results both from active control of laryngeal movement and breathing and the passive mechanical properties of vocal folds. Before voice onset, active positioning of the vocal folds close to the midline narrows the space between the folds (i.e. the glottis), thereby increasing the resistance and the magnitude of the transglottal pressure gradient. This produces a high-speed intra-glottal airflow with associated pressure changes on the vocal fold surfaces that, in turn, facilitate self-sustaining tissue oscillations (Titze, 1988). This was first demonstrated by blowing air through an excised alligator larynx (Müller, 1839) and confirmed by subglottal pressure and airflow measurements made in spontaneously calling animals (Riede et al., 2011). The rate of these oscillations depends on lung pressure (affecting transglottal pressure gradient), vocal fold geometry (length, thickness and depth), and vocal fold tension (similar to string oscillations in a string instrument). The vocal fold vibrations are induced by airflow and are self-sustained.

Alligators produce sounds with an F₀ between ∼50 and 1200 Hz. Lung pressure ranges between 0.5 and 6 kPa during vocalization (Riede et al., 2011). Vocal fold length scales isometrically with body mass (Riede et al., 2011), but actual mechanical properties of the vocal folds were unknown. To investigate mechanical properties, we used tensile tests to measure elasticity and stress relaxation. Because tissue mechanical properties depend on both the composition and organization of fibrillar proteins (i.e. collagen and elastin) and the content and distribution of amorphic substances (i.e. hyaluronan), we used several histological staining techniques and computer tomographic imaging to refine our understanding of the alligator's laryngeal morphology. Finally, a computer simulation of the alligator's larynx was used to integrate morphological and physiological data to explore boundaries of the alligator's available acoustic space.

The larynx consists of three cartilages (Fig. 1 and supplementary material Fig. S1). The cricoid cartilage forms a complete ring at the cranial end of the trachea. The arytenoid cartilages are two symmetric arches positioned rostrally from the cricoid ring. They are attached to the cricoid via one dorsal and one ventral joint. Each arch consists of a dorsal and a ventral branch. The basihyoid is an enlarged, cup-like structure housing all laryngeal elements.

The vocal fold epithelium consists of cuboid ciliated cells covering a thick layer of extracellular matrix (ECM) and muscles (Fig. 2A). Serous epithelial cells limited to the pharyngeal side of the vocal folds (Fig. 2B) remained intensely positively stained after hyaluronidase treatment and contain a range of different glucoseaminoglycans. The Alcian Blue stain before (Fig. 2C) and after (Fig. 2D) removal of hyaluronan by hyaluronidase digestion indicates that much of the positive staining in the vocal fold can be attributed to hyaluronan.

The ECM is also rich in collagen fibers, and the area between the epithelium and the center of the vocal fold exhibits a gradient of decreasing collagen content. A dense layer of collagen below the epithelium (Fig. 2E,F) contained fibers oriented parallel to the epithelial surface with an average thickness of 2.4 μm (2.4±0.8 μm; mean±s.d.; measured in three juvenile alligators). In the center of the vocal fold, collagen fibers were less densely packed, had smaller average diameters (0.9±0.7 μm; mean±s.d.), and were oriented randomly.

The musculus cricoarytenoideus originates laterally on the cricoid cartilage (Fig. 3). It extends to the rostro-lateral aspect of the dorsal branch of the arytenoid and into the soft connective tissue of the vocal folds (Fig. 3, L2). Contraction of this cranial aspect of the muscle would pull on the rostral end of the arytenoid arch, causing movement laterad and opening the glottis (vocal fold abduction). There is also a caudal branch of the muscle attaching to the cricoid and to the medial side of the basihyoid (Fig. 4). This newly described attachment to the basihyoid confers this muscle's extrinsic properties. An isolated contraction of the caudal portion would pull the entire larynx caudally, exerting a strain on the vocal folds along their longitudinal (rostro-caudal) axis. This would generate tensile stress, a mechanism most important in facilitating variation in F₀, but previously thought to be absent in alligators (Riede et al., 2011).

The glottal adductor (musculus constrictor laryngis, Söller, 1931) consists of two branches, a pars ventralis and pars dorsalis. The left and right branches of the pars dorsalis meet in a raphe on the dorsal midline rostral to the cricoid. The muscle attaches laterally to the dorsal branch of the arytenoid (Fig. 3, L5). Contraction of the dorsal portion would adduct the dorsal aspects of the vocal folds. The pars ventralis runs between the cranial end of the dorsal arytenoid arch and a ventral raphe, where the left and right portions of this muscle are inserted. A contraction of the pars ventralis would adduct the ventral aspects of the vocal folds and reduce the size of the lateral ventricle (Fig. 3). Because the dorsal portion is postioned caudal and the ventral portion rostral, differential contraction of both portions is expected to cause differential tension in the dorso-ventral axis of the vocal fold.

Tensile tests were performed on the lamina propria of the vocal folds to determine elasticity and stress relaxation behavior. The tissue showed a linear stress–strain response in the low-strain region and a nonlinear relationship in the high-strain region during a cyclic loading–unloading test (Fig. 5A). Linear and nonlinear models (Table 1) reached regression coefficients of 0.97 and higher. Tissue stress ranged from 1 to 600 kPa for a vocal fold length of 1 cm at strains of up to 35%. F₀ values fell between 50 and ∼1200 Hz (Fig. 5B) at the same strain values, and correspond well with reported ranges in F₀ for natural calls of juvenile animals (e.g. Britton, 2001; Vergne et al., 2009). Hysteresis ranged from 31 to 55%. Stress relaxation ranged from 18 to 44%, and peak stress was reduced by 50% within the initial 205 to 315 ms (‘stress half life’) (Table 1). This initial stress relaxation likely has a large influence on modulation of F₀ in juvenile calls.

Fig. 6 presents heat maps of the interacting effects of subglottal pressure, glottal gap and vocal fold tension on F₀. White regions correspond to aphonia (no tissue oscillation). Fig. 6A shows how F₀ depends on the glottal gap and tissue stiffness. Higher tension (facilitated by larger strain) was associated with higher F₀ sounds. As glottal gap was reduced, adduction increased and so did F₀. However, there seems to be a ‘threshold tension’ at 0.05, below which F₀ changes minimally with varying glottal gap. Only above this threshold does variation in adduction (at a given tension) have large effects on F₀.

Fig. 6B shows the influence of glottal gap size and subglottal pressure on F₀. It is well known that the subglottal pressure facilitates vocal fold vibration as long as a phonation threshold pressure (i.e. onset point of tissue vibrations) is exceeded. Dynamic stretch of the tissue during each vibration cycle is responsible for the dependence of F₀ on subglottal pressure. The borderline between aphonic and phonic regions corresponds to the phonation onset point, which decreases as adduction is increases (i.e. smaller glottal gap); in this way, pressure indirectly controls F₀. Subglottal pressure has an even more dramatic effect on increasing F₀ in the region where the glottal gap is very small – a phenomenon attributed to dynamic stretch and cubic nonlinearity of tissue properties.

Fig. 6C shows how both strain and subglottal pressure control F₀. The highest frequency of 930 Hz was achieved at a strain of 35% and subglottal pressure of 4 kPa. Subglottal pressure variations have maximal influence on F₀ when strain is large. At low strains, F₀ changes very little at a given pressure of 4 kPa.

Fig. 6D shows the time waveform, spectrogram and subglottal pressure pattern of a juvenile contact call (from Riede et al., 2011). F₀ starts at 900 Hz and is down-modulated to 100 Hz within 150 ms. The arrows in Fig. 6C indicate potential paths through the heat map that would produce this call. Note that multiple pathways are possible.

Alligator vocal folds have a complex morphology consisting of epithelium, lamina propria and muscle. The ECM of the lamina propria showed that collagen fibers are densely packed below the epithelium and less organized inside the vocal fold. The composition of the ECM is of central importance because of its crucial contributions to the mechanical properties of the vocal fold and the voice quality that results. Mechanical properties of the vibrating vocal folds set the boundaries for F₀. Tensile data suggest that small changes in length (up to 35% strain) of the rather stiff alligator vocal fold (compared with mammalian vocal folds) can facilitate F₀ values reported for juvenile alligator calls (50–1200 Hz). Complex, species-specific laryngeal morphologies (and associated sound-source properties) are widespread among mammals (e.g. Hirano, 1974; Riede, 2010; Riede et al., 2010; Frey and Riede, 2013; Klemuk et al., 2011) and have also been described for the labia, the avian analog of laryngeal vocal folds (Fee, 2002; Riede and Goller, 2014). In contrast to alligators, geckonid lizards (Moore et al., 1991; Rittenhouse et al., 1998; Russel et al., 2000) and tortoises (Sacchi et al., 2004) have higher numbers of elastic fibers in the subepithelial region of the vocal fold lamina propria. A purported vocal fold analog (‘fibrofatty laryngeal fold’: Fraher et al., 2010) has been described for leatherback sea turtles (Dermodyles coriacea). Thus, it is reasonable to expect species-specific morphologies in vocal folds/analogues as far back as basal reptiles.

The cricoarytenoid muscle possesses a ventral portion connecting the cricoid and basihyoid cartilage. Because the attachment is extrinsic, contraction of this muscle moves the larynx relative to the basihyoid. Vocal fold tissue bridges the arytenoid cartilage (at the caudal portion of vocal fold) and basihyoid (at the cranial portion of vocal fold). The tissue would be strained along the cranio-caudal axis by contraction of the ventral branch of the cricoarytenoid muscle. Indeed, Henle (1839) ascribed a ‘larynx retractor’ function to this muscle (‘dilator glottidis’) in some reptiles, upon finding an extrinsic laryngeal insertion – a morphology not noted for alligators (Henle, 1839, Göppert, 1899; Göppert, 1937; Söller, 1931). It remains to be tested how this retractor function is employed during vocal production by alligators.

Previously, we assumed that alligator vocal folds are only stretched dynamically by means of oscillation. However, two additional mechanisms are possible. Dynamic stretch is a passive mechanism occurring during each vocal fold vibration and is below 5% in humans or in excised dog larynges (Titze, 1989). Muscle activation could also initiate longitudinal (cranio-caudal) elongation along the vocal-fold axis, but would require neural control. This establishes three variables under active control: lung pressure, longitudinal stiffness (regulated by the cricoarytenoid muscle) and adduction, affecting the glottal gap (regulated by the constrictor larynges muscle). The heat-map representation for three pairwise associations of these actively controlled variables demonstrates how F₀ regulation could be achieved. Determining the net effect of subglottal pressure, vocal fold length and adduction on F₀ is of great importance because it would indicate complexity of motor control comparable with that of vocalizing mammals.

This study and previous work (Riede et al., 2011) provide information on vocal production mechanisms in American alligators that may help illuminate broad evolutionary patterns. Most excitingly, the data suggest that alligators can employ complex motor control, and that basic design patterns described for the vocal folds of mammals and birds (Hirano, 1974; Riede and Goller, 2014) also apply to non-avian reptiles. Additional mechanical properties of cartilage framework and laryngeal muscles, as well as the possibility of active motor control remain to be explored. The potential for the downstream modulation (e.g. vocal tract filtering) seen in mammals and birds may exist, but this is unknown.

The data complement the picture of the alligator larynx as a myoelastic-aerodynamic sound source, a concept developed for human phonation (van den Berg, 1958; Titze, 2006) and applicable to the avian syrinx (e.g. Riede and Goller, 2010). Muscles actively position and stretch the oscillating tissue. Vocal sound output in alligators is therefore a result of the integration of active motor control of the larynx and the respiratory system, as well as the viscoelastic properties of the vocal folds. The airflow-induced (passive) oscillations of the vocal folds depend largely on the magnitude of the forces produced at the sound source. As in turtles, lizards, mammals and birds, alligator vocal folds are multilayered, suggesting convergent design that is functionally driven. Layered tissue confers anisotropic properties at the gross structural level, generating specific responses to different types of deformations (e.g. stretch and adduction). It follows that any change in lung pressure, vocal fold tension and/or adduction, will alter the effective stiffness of this multi-layered system – a parameter that changes how folds vibrate and thereby convert aerodynamic energy into acoustic energy. The computer simulation illustrates how these three variables dictate F₀. The simulation generates a large possible acoustic space, and suggests that multiple alternative variable combinations can generate the same acoustic output. Plotting an actual alligator call (Fig. 6) into the simulated acoustic space indicates that many of these variable combinations are biologically and physically possible.

The model considers three physiological variables (subglottal pressure, glottal adduction force and vocal fold tension). The exact measurements, however, exist only for subglottal pressure. The adduction force was therefore only estimated for the alligator larynx. The vocal fold abduction and adduction are nonetheless important, not only for swallowing and closure of the airway, but also for voice quality. In general, adduction first decreases the glottal gap to zero and then pushes the vocal folds against each other with increasing force. It is critical that the pre-phonatory positioning of vocal folds – the glottal gap – reduces the minimal subglottal pressure that is needed to sustain vocal folds oscillations (Titze, 1988). In humans, vocal fold posturing affects intensity and pitch of the voice (e.g. Murry et al., 1998; Honda, 1983), where adduction of the vocal folds is a key to optimize airflow–tissue interaction at voice onset (Titze et al., 1995; Chan and Titze, 2006). For the relatively low subglottal pressures measured in alligators (0.5 to 2 kPa at phonation onset) and for the given tissue properties, the assumption of a very small glottal gap seems justified. Variable adductive force had been suggested to play a role in control of fundamental frequency in alligators (Riede et al., 2011), but remains to be measured.

Finally, we also considered how sound production has evolved (Brazaitis and Watanabe, 2011) because of its implications for sexual selection and parental care. How the Permian and Triassic landscapes sounded and the mechanisms whereby non-avian dinosaurs and other extinct amniotes vocalized will remain speculative without a sufficient fossil record of vocal organs (e.g. Senter, 2008). The early steps will remain elusive if we only study the young stem groups. This study contributes to a critical phylogenetic bracket. Unifying control and design principles among the ancestral, derived and more distantly related groups (mammals) indicate functional constraints that may make it possible to discern vocal capabilities in extinct groups (e.g. Bass et al., 2008). This study contributes to these efforts because birds and crocodilians are sister taxa that bracket pterosaurs and non-avian dinosaurs, and thus these results shed some light on possible vocal fold morphology and function in the ancestral archosaur.

Alligator (Alligator mississippiensis Daudin 1802) larynges were provided by the Alligator Tissue Bank at UC Irvine (courtesy of Tomasz Owerkowicz and Jim Hicks), by Ruth Elsey (Rockefeller Wildlife Refuge, Los Angeles, CA), by John V. Price (Insta-Gator, Ranch and Hatchery, Los Angeles, CA) and by the University of Utah Vertebrate Evolution lab.

Larynges of four females and two males were processed in order to investigate tissue composition of the vocal folds. The animals ranged in size between 0.05 kg (snout–vent length, SVL: 12 cm) and 12 kg (SVL: 81 cm). The tissue was fixed immediately after harvest in 10% neutral buffered formalin (SF100-4 Fisher Scientific) for 2 weeks and subsequently decalcified for 8 h (Surgipath Decalcifier 1), embedded in paraffin, and sectioned from rostral to caudal at 5 μm thickness at five subsequent levels which were 1 to 2 mm apart. Sections were exposed to hematoxylin and eosin (H&E) for general overview, elastica van Gieson stain (EVG) for elastin fibers, Masson's trichrome stain (TRI) for collagen fibers or Alcian Blue stain (AB) (pH 2.5) for mucopolysaccharides. We also performed a digestion procedure with bovine testicular hyaluronidase to test for relative HA compostion; if hyaluronan is a major component of the mucosubstances, incubation with hyaluronidase will destroy Alcian Blue positivity). Slides were first deparaffinized, then incubated in hyaluronidase at 37°C for 1 h, washed in running water for 5 min and finally stained with Alcian Blue. All stains were performed in conjunction with control stains (artery for EVG; liver for TRI; umbilical cord for AB). Images were digitized with a ScanScope CS2 scanner (Aperio, Vista, CA) and analyzed with Imagescope (Vista, CA) and ImageJ (NIH, open source).

Computer tomographic (CT) imaging was used to investigate the morphology of soft and hard tissue components. Two specimens (1 male, 3 years, SVL: 45 cm; 1 male 7 years, SVL: 96 cm) were prepared at the Texas Natural History Collection (TNHC) and scanned at the University of Texas at Austin High-Resolution X-ray Computed Tomography Facility (UTCT). The tongue body with the intact larynx was fixed in 10% neutral-buffered formalin solution for 2 weeks and then transferred into ethanol (100%) to dehydrate for 8 days. The specimen was stained for 4 days in an iodine solution (I₂E, 1% w/v, after Metscher, 2009). Images were obtained at a resolution of 1024×1024 dpi and stored as 16-bit Tiff format files. The entire larynx region was investigated in 975 slices. Scan parameters were 180 kV, 0.32 mA, slice thickness 0.06969 mm, interslice spacing of 0.06969 mm and a field of reconstruction of 66 mm. CT images were imported into Avizo 6.3 (FEI Visualization Sciences Group) for segmentation of the laryngeal cartilages, muscular tissues and soft connective tissue. SurfaceGen was used to render the 3-D structures.

Tensile tests were used to investigate stress response to longitudinal deformation. Vocal folds from four specimens (1 kg body mass, 1 cm vocal fold length, SVL: 33 cm) were studied. Connective tissue of a vocal fold was dissected by removing muscle tissue but keeping epithelium intact. The tissue was mounted vertically in a chamber containing Ringer's solution. The chamber was immersed in a water bath maintained at 38°C (see Riede, 2010, for an illustration of the testing apparatus). The length between fixation points was measured with a caliper (±0.1 mm accuracy). The tissue was pre-conditioned by straining it by 0.5 mm in three cycles at a rate of 1 Hz. Force and displacement data were obtained by 15 cycles of 1 Hz sinusoidal stretch-and-release with a dual-mode servo-control lever system (Aurora Scientific Model 305B, Aurora, ON, Canada; resolution 1 μm and 0.3 mN). Signals were recorded with a 16-bit analog-to-digital acquisition board (Windaq Model DI722, DATAQ Instruments; Akron, OH) at 3 kHz sampling frequency. After 2 min rest, stress relaxation was measured. The tissue was strained to 40% for 30 s.

A computational model of the alligator vocal folds had been constructed previously (Riede et al., 2011). Here, we employed this model to study how measurements of the elastic property and the dimension of the larynx contribute to the alligator call. Our interest was to clarify the dependence of F₀ on the three parameters: (1) vocal fold tension, which was measured in this study; (2) subglottal pressure, which ranges between 0.5 and 6 kPa (Riede et al., 2011); (3) vocal fold adduction. Alligators can alter the degree of glottal opening (i.e. adduction strength). Visual inspection in two vocalizing alligators (our unpublished data) suggested that vocal folds are adducted prior to vocalization. However, adduction strength has not yet been measured and was estimated in this study.

Anatomical investigations presented here indicate that alligators possess mechanisms (points 1 and 2 above) to control the tension of the vocal fold by modulating strain. Using the computational model, we studied F₀ ranges that can be produced and compared the computational outcome with previously recorded alligator vocalizations.

The two-mass model (Ishizaka and Flanagan, 1972; Steinecke and Herzel, 1995) describes the vocal fold tissue as a set of two masses coupled by springs (see Riede et al., 2011 for an illustration of the model). The phase shift of the two masses, representing, respectively, the upper and lower parts of the vocal fold, efficiently transfers the glottal flow energy to the vocal fold (Titze, 1988), generating vibrations.

 Under the assumption that the flow inside the glottis obeys Bernoulli's Principle below the narrowest part of the glottis, the pressures that act on individual masses were calculated as P₁=P_s{1−Θ(a_min)(a_min/a₁)²}Θ(a₁) and P₂=0, where P_s represents the subglottal pressure and a_min=min(a₁, a₂). The glottal flow was given as U=(2P_s/ρ_air)^1/2 a_minΘ(a_min) using the air density constant (ρ_air=0.00113 g cm⁻²).

The parameter values were set in accordance with our measurements. The length, width and depth of the vocal fold were set as l=1 cm, w=0.4 cm, and d=d₁+d₂=0.48 cm (d₁=0.4 cm, d₂=0.08 cm), respectively, corresponding to the vocal folds of a 1 kg alligator (45 cm snout–vent length). The masses were given by m₁=ρlwd₁ and m₂=ρlwd₂. Using the Young's modulus, which was estimated from the stress-strain curve as E=σ′=abe^bε, the stiffness was computed as k₁=wd₁E/l, k₂=wd₂E/l. Following the standard setting, the nonlinear stiffness, the mutual coupling, and the collision constants were set as η=100, k_c=k₁/3, c₁=3k₁, c₂=3k₂. The damping constants were set as r_i=2ζ(m_ik_i)^1/2 (i=1,2) using a damping ratio of ζ=0.05. Once the parameter values were fixed, Eqn 5 was integrated by the Euler method with a time step of Δt=1/3200 s. The glottal flow U was generated and analyzed by the FFT (Fast Fourier Transform) to compute F₀.

The main variables controlling phonation were strain ε (3 to 35%), subglottal pressure P_s (1 to 4 kPa), and glottal gap x₀. Strain determines vocal fold stiffness. A small glottal gap value represents a high level of adduction. Glottal gap was varied between 0 cm and 0.1 cm.

We are grateful for tissue donations from the Alligator Tissue Bank at UC Irvine (courtesy of Tomasz Owerkowicz and Jim Hicks; NSF grant IOS 0922756) and Ruth M. Elsey from the Rockefeller Wildlife Refuge, Louisiana Department of Wildlife and Fisheries, Grand Chenier, Louisiana.