Pygmy mouse songs reveal anatomical innovations underlying acoustic signal elaboration in rodents

Advertisement signals used in reproductive contexts are among the most diverse and elaborate traits in the animal kingdom (Bradbury and Vehrencamp, 2011). Understanding the mechanisms of signal production is central to understanding how signals diversify and the anatomical innovations that promote or constrain divergence. Indeed, recent work on traits used in social communication provides important insights into the processes underlying phenotypic evolution (e.g. Ord et al., 2013; Eliason et al., 2015, 2020).

Rodents produce diverse and complex vocal communication signals in a variety of social interactions (Sales and Pye, 1974; Brudzynski, 2018; Dent et al., 2018). Many muroid rodents are able to produce two distinct spectral ranges using divergent physical mechanisms (e.g. Pasch et al., 2017). Currently, sounds with spectral properties between 500 Hz and 16 kHz (hereafter ‘audible sounds’) are assumed to be produced by flow-induced vocal fold vibrations, whereas sounds with fundamental frequencies between 19 and 100 kHz (hereafter ‘ultrasonic vocalizations’, USV) are produced by a whistle mechanism (Riede, 2018). This distinction is based on data from laboratory mice (Mus musculus) and rats (Rattus rattus) and does not properly reflect the large acoustic diversity among non-traditional rodent models (e.g. Miller and Engstrom, 2007, 2010; Pasch et al., 2011, 2017). The spectral boundary between the two production mechanisms may not be as clear or mutually exclusive. Comparative data on vocal anatomy and underlying production mechanisms are thus needed to provide insight into the evolution of acoustic communication within the largest mammalian clade.

Current understanding of rodent vocal production mechanisms indicates that spectral content is determined by laryngeal anatomy (Riede, 2018). Ultrasonic whistle production appears to be dependent upon the presence of an intra-laryngeal air sac (ventral pouch) as well as a small cartilage (alar cartilage) that supports the ventral pouch entrance (Riede et al., 2017). The ventral pouch and the alar cartilage are located rostral from the vocal folds and have been identified in a few muroid rodents (laboratory mouse: Mus, laboratory rat: Rattus, and grasshopper mouse; Onychomys: Riede et al., 2017). However, both structures were absent in a heteromyid rodent (kangaroo rat: Dipodomys) that displays a limited vocal repertoire (Riede et al., 2017). Indeed, alterations of the ventral pouch cause changes in or disappearance of ultrasonic vocal capabilities (Riede et al., 2017). Furthermore, muscle activity can regulate the position of the alar cartilage and ventral pouch size to modulate spectral features of vocalization (Riede, 2013). Altogether, the ventral pouch appears to be critical for ultrasonic whistle production and its size is predicted to be negatively associated with the fundamental frequency of vocalizations (Riede et al., 2017).

Cricetid rodents in the subfamily Neotominae commonly produce audible (<20 kHz) vocalizations (Miller and Engstrom, 2007, 2010). Animals often assume an upright posture and open their mouths widely to generate loud vocalizations (Bailey, 1931; Blair, 1941; Packard, 1960; Hooper and Carleton, 1976), which are used to attract mates and repel rivals (Campbell et al., 2019; Pasch et al., 2011, 2013). Northern pygmy mice (Baiomys taylori) are crepuscular cricetid rodents that inhabit arid grasslands, coastal prairie mixed scrub, post oak savanna, and mesquite–cactus habitats from Texas to Mexico (Eshelman and Cameron, 1987; Light et al., 2016; Packard, 1960; Carleton, 1980). The staccato-like song of pygmy mice is described as a ‘high-pitched, barely audible squeal’ produced with the head ‘thrust forward and upward, stretching the throat’ (Blair, 1941). Miller and Engstrom (2007) reported that songs were produced by both sexes and were not as elaborate as those produced by sister taxa neotropical singing mice (Scotinomys; Pasch et al., 2011, 2013). How such frequency-modulated songs are produced is unknown but promises to provide important insight into the evolution of signal elaboration in the largest radiation of mammals.

In this study, we used acoustic recordings in normal air and a light gas mixture to determine the mode of vocal production of northern pygmy mouse songs. The use of gases that are lighter or heavier than ambient air has a long tradition in the study of vocal production mechanisms (e.g. Beil, 1962; Nowicki, 1987; Pasch et al., 2017). Sounds produced via a whistle mechanism demonstrate a characteristic increase in fundamental frequency in helium atmosphere that is distinct from the constancy of fundamental frequencies produced via flow-induced vocal fold vibration (Spencer and Titze, 2001). We then used contrast-enhanced micro-computed tomography (CT) imaging and histology to characterize the laryngeal anatomy subserving such unique sound production. Finally, we used geometric morphometrics to compare laryngeal differences within pygmy mice and among closely related species to assess the contribution of larynx and ventral pouch size and shape to acoustic diversity.

We captured northern pygmy mice, Baiomys taylori (Thomas 1887), near Rodeo, NM, USA, using Sherman live-traps baited with sterilized bird seed. Mice were transferred in standard mouse cages to animal facilities at Northern Arizona University, Flagstaff, AZ, USA, maintained on a 14 h light:10 h dark cycle (21±2°C) and provided with rodent chow, bird seed and water ad libitum. A subset of animals was transferred to Midwestern University, Glendale, AZ, USA, for heliox experiments and morphological analysis. All procedures in studies involving animals were performed in accordance with the ethical standards and approval of the Institutional Animal Care and Use Committee at Midwestern University (MWU#2852) and Northern Arizona University (16-001 and 19-006) and guidelines of the American Society of Mammalogists (Sikes et al., 2016). Animals were captured with permits from the New Mexico Department of Game and Fish (3562) and the Arizona Game and Fish Department (607608).

We recorded the mass and vocalizations of 10 wild-captured mice (n=5 per sex) over 10 days. Singly housed mice in their home cage were placed in semi-anechoic coolers lined with acoustic foam. We used 1/4 inch microphones (Type 40BE, G.R.A.S.) connected to preamplifiers (Type 26 CB, G.R.A.S.) to obtain recordings 33.3 cm above the center of the focal mouse cage. Microphone response was flat within ±1.5 dB from 10 Hz to 50 kHz, and pre-amplifier response was flat within ±0.2 dB from 2 Hz to 200 kHz. Microphones were connected to a National Instruments DAQ (USB 4431) sampling at 102.4 kHz and to a desktop computer running MATLAB (version 2018a).

We used Avisoft SASLab Pro [version 5.2.13, Avisoft Bioacoustics; 1024-point Fast Fourier Transform (FFT), 75% frame size, Hann window, frequency resolution 100 Hz, temporal resolution 93.75%, 0.625 ms] to automatically extract temporal (song duration) and spectral [maximum and minimum fundamental frequency (F₀), and frequency bandwidth averaged across all notes in the song] parameters from each bandpass-filtered (17.5–40.5 kHz) recording.

Student's t-tests were used to compare body mass, number of songs and acoustic parameters between females and males in JMP Pro (version 14.1.0, SAS Institute, Inc., Cary, NC, USA). We used a Bonferroni correction to control for multiple (n=5) comparisons (corrected α=0.01). We also assessed whether repeatability of acoustic parameters differed between the sexes by calculating intraclass correlation coefficients (ICC; Wolak et al., 2012) using the ICC R package (version 2.3.0; Wolak et al., 2012) in R version 3.6.1 (http://www.R-project.org/). Acoustic parameters were considered repeatable if the 95% confidence interval (CI) of ICC values excluded zero, and similar between sexes if CIs overlapped one another. Values are reported as means±s.d. in the text.

Songs of six additional animals (3 songs per sex per treatment) were recorded in normal air and in a heliox atmosphere. Individual mice were placed in a standard mouse cage. The cage was equipped with bedding, food and water. Heliox gas (80% He, 20% O₂) was injected into the cage at flow rates between 20 and 40 l min⁻¹ through a 12 mm wide tube placed into the cage wall near the floor. Predicted effects of light gas concentrations were estimated with a small whistle placed at the floor of the cage and connected externally by a silastic tube. The whistle was blown and recorded at regular intervals in order to monitor the heliox concentration. The ratio of the frequency of the whistle in air and in heliox allowed an estimation of the expected effect for any given heliox concentration.

Three randomly selected syllables (from the first, second and last third of the song) from each song were analyzed for minimum and maximum F₀, syllable duration, and relative amplitude of the first and second harmonic of the F₀. While conducting the experiments, we noticed that the temporal characteristics of songs changed in heliox. Therefore, we measured syllable repetition rate, syllable duration and intersyllable interval (i.e. the duration between two adjacent syllables) for three songs in normal air and three in heliox. Syllable repetition rate was calculated by dividing the number of syllables within a song by song duration. Syllable duration and intersyllable interval were averaged over all syllables of a song. Acoustic differences between normal air and heliox songs were assessed with paired t-tests.

Ten mice (5 per sex) were euthanized with isoflurane and then transcardially perfused with saline solution followed by 10% buffered formalin. Larynges were dissected and placed in 10% buffered formalin phosphate (SF100-4; Fisher Scientific) for 2 days.

Four specimens (2 per sex) were used to prepare histological sections. Mid-membraneous coronal sections (5 mm thick) were stained with hematoxylin–eosin for a general overview. We also attempted to use Masson's Trichrome (TRI) for collagen fiber stain and Elastica van Gieson (EVG) for elastic fiber stain. However, pygmy mouse vocal folds are small (see below), and we were unsuccessful in collecting a sufficient sample size of mid-membraneous sections for different stains in each individual. Sections were scanned with an Aperio CS 2 slide scanner and processed with Imagescope software (version 8.2.5.1263; Aperio Tech.).

Two specimens (1 per sex) were whole-body scanned at 50 µm resolution. Larynges from these two individuals and four additional mice (3 per sex) were also X-rayed at 5 µm resolution. First, tissues were transferred from the formalin solution to 99% ethanol. Tissues were then stained in 1% phosphotungstic acid (PTA; Sigma-Aldrich, 79690) in 70% ethanol. After 5 days, the staining solution was renewed and the tissue was stained for another 5 days. After staining, specimens were placed in a custom-made acrylic tube and scanned in air. CT scanning was done using a Skyscan 1172 (Bruker). Reconstructed image stacks were then imported into AVIZO software (version Lite 9.0.1). Laryngeal cartilages and the border between the airway and soft tissues of the larynx in the CT scans were traced manually. This approach provided outlines of the cartilaginous framework and the airway. Derived 3D surfaces of all six specimens have been archived at Morphobank (http://www.morphobank.org, project # 3638).

We investigated whether the laryngeal size or shape of B. taylori mice was sexually dimorphic using a geometric morphometric (GM) approach. 3D surfaces of three male and three female laryngeal cartilages were used to quantify shape using curve and surface landmarks. Landmarks were placed on surface renderings using the ‘geomorph’ package, version 3.0.5 (https://cran.r-project.org/package=geomorph) for the R software package (http://www.R-project.org/). Fixed landmarks were placed along the cartilage border and supplemented by 100 surface semi-landmarks, which were placed with the help of an interactive function to build a template of 3D surface semi-landmarks. In order to compare cartilage shape, a generalized Procrustes analysis was used to remove variation related to position, size and orientation. Next, we employed a series of principal components analyses (PCA) to summarize the main patterns of shape variance in the data. The procedure used here has been previously established (Borgard et al., 2019).

Centroid size was used to estimate overall size for each cartilage. Centroid size is the square root of the sum of squared distances of each landmark from the center of the cartilage. The location is obtained by averaging the x, y and z coordinates of all landmarks (Zelditch et al., 2004).

For comparative analyses, we used data from laboratory mice (Mus musculus), laboratory rats (Rattus rattus) and grasshopper mice (Onychomys spp.) published previously by Borgard et al. (2019). To assess whether laryngeal cartilage shape exhibited sex differences in pygmy mice, we performed a multivariate analysis of variance (MANOVA) on the PC1 and PC2 scores for each cartilage.

Long-distance songs consisted of repeated frequency-modulated syllables (Fig. 1A). All song parameters were repeatable within individuals and between sexes (Table 1). We found no sex differences in mass (females: 9.95±1.1 g; males: 9.22±0.8 g; t₈=−1.23, P=0.25) nor number of songs produced (females: 53±52.3 songs; males: 162.8±134.6 songs; t₈=1.69, P=0.13; Table 1) despite males showing higher variation in song rate (range: 84–401 songs) than females (range: 5–132 songs). Males produced longer songs with slightly lower minimum F₀ and larger frequency bandwidths compared with females (Table 1).

Both maximum and minimum F₀ increased in heliox (paired t-test, maximum F₀: t=−10.0, P<0.001; minimum F₀: t=−14.1, P<0.001) compared with those in normal air (Fig. 1B–D). However, the amplitude of higher harmonics (2F₀) in normal air and in heliox relative to F₀ at the center of syllables did not differ (paired t-test, t=−1.3, P=0.261; Fig. 1E).

While song duration and intersyllable interval did not differ between normal air and heliox (paired t-test, song duration: t=0.70, P=0.513; intersyllable interval: t=0.61, P=0.566; Fig. 2A,B), other temporal features of the song changed in heliox. Syllable duration decreased significantly (paired t-test, t=5.5, P<0.01) and, as intersyllable interval was unaltered, syllable repetition rate increased (paired t-test, t=−4.6, P<0.01) (Fig. 2C,D). Fig. 2E,F shows individual syllable durations over the course of three songs in normal air and in heliox, for a male and a female mouse. Syllable duration decreased by 15±6.7% and syllable repetition rate increased 10.5±4.4%, yet the number of syllables per song did not change significantly (number of syllables in air: 28.9±4.6, in heliox: 29.5±6.9; t=0.88, P=0.422; Table 1).

The membraneous portion of a vocal fold (captured by six sequential sections in Fig. 3A–F) refers to the soft tissue that starts at the vocal process of the arytenoid cartilage and ends where the vocal fold attaches to the interior of the thyroid cartilage. The ventral section of the membraneous portion of the vocal fold is reinforced with cartilaginous tissue and is continuous with the alar cartilage and epiglottis. Consequently, the cartilaginous-free, membraneous portion of the vocal fold is very short (<150±24 µm; n=4). The membraneous portion of the vocal fold is composed of thyroarytenoid muscle, lamina propria and epithelium. The lamina propria of the membraneous portion is composed of a thick layer (up to 110±1.5 µm) of collagen and elastic fibers (Fig. 3G).

The ventral pouch of B. taylori is large with a medially positioned portion rostral from the vocal folds (Fig. 4A). Additionally, pockets connected to the medial portion extend laterally from each vocal fold, causing a unique morphology of the ventral portion of the vocal fold unprecedented in the mammalian larynx (e.g. Negus, 1949; Schneider, 1964). The vocal fold is very thin and membranous (Fig. 3C–F). The alar cartilage forms the edge of the entrance into the ventral pouch. Fig. 3H,I shows representative images of the cartilage. Histological sections demonstrate that the cartilage is associated with the epiglottis (Fig. 3I).

The thyroid cartilage consists of a left and right lamina, each with a rostral and caudal horn (Fig. 5A–D). The rostral horn is narrow and pointed. The rostro-ventral margin of the thyroid bends dorsally, toward the laryngeal lumen forming a bulla thyroidea (Fig. 5D). The ventral pouch is embedded in this bulla. A small cartilaginous protuberance is present on the medial surface of the thyroid cartilage in a caudal midsagittal position (Fig. 5B). The structure consists of highly mineralized cartilage that protrudes into the laryngeal lumen. Vocal folds attach to the thyroid cartilage via this protuberance.

The cricoid cartilage (Fig. 5E–H) forms a complete ring with a broad plate dorsally and a narrow band laterally and ventrally. The arytenoid cartilages form triangular shaped structures with a vocal process, a short muscular process and a long and narrow dorsal process (aka apex) (Fig. 5I–L). The epiglottis (Fig. 5M–P) forms a small sheet bending towards the airway. Most notable is the structure at the caudal end of the epiglottis, which is described as alar cartilage in other species. In M. musculus, R. rattus and Onychomys spp., the alar cartilage was a separate structure without connection to another laryngeal cartilage (Riede et al., 2017). In B. taylori, the alar cartilage supports the entrance into the ventral pouch and is moved by fibers from the thyroarytenoid muscle. Unlike in M. musculus, R. rattus and Onychomys spp., the alar cartilage is tightly connected with the epiglottis in B. taylori.

The length of the oral and pharyngeal cavity (vocal tract length measured between the vocal fold and tip of incisivi) measured 16 and 17 mm in a female and male specimen, respectively, for which the whole body was scanned before the larynx was excised for high-resolution scanning.

Laryngeal size represents an important source of spectral differences in animal vocalizations. We therefore investigated the size and shape of laryngeal cartilages as well as the ventral pouch. First, we tested for morphological differences between sexes in B. taylori, and then compared larynx size of B. taylori with that of M. musculus, R. rattus and Onychomys spp. in order to explain the fundamental frequency of B. taylori songs.

Neither the size of the four laryngeal cartilages nor any of the ventral pouch dimensions demonstrated sexual dimorphism in B. taylori (Table 2). Cartilage shape was also not different between the sexes. None of the four cartilages exhibited separation of their shape along the first or second principal component (PC1 and PC2) (thyroid cartilage: F_1,3=0.88, P=0.21; Wilks’ Λ=0.82; cricoid cartilage: F_1,3=0.89, P=0.21; Wilks’ Λ=0.82; arytenoid cartilage: F_1,3=0.85, P=26; Wilks’ Λ=0.78; epiglottis: F_1,3=0.64, P=0.84; Wilks’ Λ=0.51) (Fig. 6A–D).

Compared with three other species, B. taylori possesses a very large and unusually shaped ventral pouch (Fig. 7A–D). Overall larynx size remains linked to body size (Fig. 7E–H). All four cartilages scaled with negative allometry against body mass. Interestingly, the size of the ventral pouch, a structure which is located inside the larynx, is not linked to larynx size or body size (Fig. 7I–J). However, ventral pouch dimensions, in particular its latero-lateral width, appear to be associated with F₀ among the four species (Fig. 7K, Table 3).

Our findings indicate that the elaborate singing behavior of baiomyine mice arises from a unique laryngeal morphology that facilitates aerodynamic whistle production. In particular, a large air sac termed the ventral pouch and a robust alar cartilage are associated with the production of F₀ through an aerodynamic whistle. In rodents, the presence of a ventral pouch and alar cartilage are associated with ultrasonic whistle production, and damage to the alar cartilage and ventral pouch compromises whistle fidelity (Riede et al., 2017). Our current findings further support the association between the presence of a ventral pouch and alar cartilage and the ability to produce ultrasonic vocalizations by an aerodynamic whistle mechanism. More specifically, comparative analyses indicate that ventral pouch size is associated with F₀, allowing relatively small species to produce low-frequency songs independent of body size (Table 3). Our findings provide important insight into the anatomical innovations that facilitate signal elaboration.

The relatively low F₀ of pygmy mouse songs are remarkable because they are among the smallest rodents (8–12 g; Wilson and Reeder, 2005). The songs extend to the lowest aerodynamic whistle frequency recorded among four species for which anatomical information and light gas experiments have informed our understanding of the vocal production mechanisms. From one perspective, pygmy mouse ultrasonic whistles overlap the audible calls of grasshopper mice, a genus that is larger in mass (32–38 g) and produces long-distance vocalizations using vocal fold vibrations (Pasch et al., 2017). Conversely, pygmy mouse frequencies also overlap with the lowest ultrasonic frequencies reported for laboratory rats (19 kHz; Brudzynski et al., 1993; Wright et al., 2010), which are 10–20 times larger than pygmy mice (250–500 g; Wilson and Reeder, 2005). In addition, pygmy mouse song frequencies are much lower than those of laboratory mouse (M. musculus) songs (Holy and Guo, 2005), a species that is slightly larger (25–50 g; Wilson and Reeder, 2005), but whose ventral pouch is much smaller than that of Baiomys (Riede et al., 2017; herein). Thus, while the overall size of the vocal apparatus appears to scale linearly with body size (Fig. 7E–H), the dimensions of the ventral pouch uncouple the relationship between body size and size-dependent acoustic features (Fig. 7I–K).

The fundamental frequency of USVs depends on glottal airflow velocity, the distance between the glottal and alar edge, and ventral pouch volume (Riede et al., 2017). Our data suggest that laryngeal cartilage size scaled allometrically with body size but was not associated with vocal frequencies. In contrast, ventral pouch size was inversely related to maximum and minimum F₀ across species and likely explains the low frequency range of Baiomys songs. While the ventral pouch did not explain sexual dimorphism of acoustic features in the songs of pygmy mice, the slight difference (2 kHz) in spectral content between the sexes may be too minimal to detect with our methodology. Alternatively, sex differences in muscle tension may contribute to laryngeal shape and size differences that mediate vocal output (Riede, 2013).

Side branches or cavities branching off the main upper airway above and below the laryngeal valve are common in mammals and birds (e.g. King and McLelland, 1984; Riede et al., 2008). Such rigid or inflatable cavities serve to amplify sound levels, modify existing resonance properties or introduce an additional resonance frequency (Riede et al., 2008; deBoer, 2009). In rodents, the air sac or ventral pouch is different from air sacs in non-human primates, cervids and birds. Whereas non-rodent air sacs can branch off the airway above or below the glottis, the rodent ventral pouch is always positioned rostral to the glottis, embedded inside the thyroid cartilage lumen and supported by cartilaginous structures. The alar cartilage or modifications of the epiglottis reinforce the entrance and lateral wall of the ventral pouch. Riede et al. (2017) suggested that the alar edge of the reinforced entrance acts like the blade or labium of the air hole of a recorder. The glottal airflow is guided over the ventral pouch entrance and hits the alar edge which opposes the glottis, generating pressure fluctuations that presumably excite resonances inside the ventral pouch. In addition, the alar cartilage is associated with a small body of muscle fibers branching off from the thyroarytenoid muscle, and activity of this muscle is precisely associated with F₀ features of vocalizations (Riede, 2013). Together, our anatomical findings suggest that the dimensions of the rodent ventral pouch, unlike those of other air sacs, are precisely controlled.

Most heliox studies in bioacoustics focus on spectral analysis because resultant patterns are indicative of underlying production mechanisms. We hypothesize that heliox experiments can also inform our understanding of vocal–respiratory coordination. In particular, we found that pygmy mouse syllables were shorter in duration in heliox than in air. Our findings are in accordance with heliox-induced shortening of vocalizations in grasshopper mice (Onychomys; Pasch et al., 2017) and common marmosets (Callithrix jacchus; Zhang and Ghazanfar, 2018) that both produce vocalizations by airflow-induced vocal fold vibration. The magnitude of the effect was similar among species (B. taylori, 15%; Onychomys, 19%; C. jacchus, 11%). We speculate that the lower density of inhaled heliox affects laryngeal aerodynamics by an eased gas flow through the constricted glottal valve during vocal production (Sundberg, 1981). Consequently, a slightly reduced air volume is available during expiration to decrease syllable duration. Confirmation of this hypothesis will require further experimentation with varying concentrations of light gas and/or use of heavy gas.

Our findings provide further insight into vocal–respiratory coordination when contextualized in relation to recent work in another baiomyine rodent (Alston's singing mouse; Scotinomys teguina) that produces a similar repetitive song (e.g. Campbell et al., 2010; Pasch et al., 2011). Okobi et al. (2019) found that cooling of the orofacial motor cortex decreased the rate of change of syllable duration without affecting the duration of individual syllables. While not explicitly stated, cooling appears to do so by increasing the duration of the inhalation phase (i.e. intersyllable interval), which directly contrasts our heliox findings, which show an impact on the expiratory phase (syllable duration) without changing intersyllable interval. Together, available data suggest that songs emerge from a motor pattern coded by a central pattern generator in the brainstem (Tschida et al., 2019) whose timing is modulated by cortical control of breathing (specifically the inhalation phase) (Okobi et al., 2019). Differences in the acoustic output, however, can also simply be the result of mechanical constraints of the peripheral organ; for example, laryngeal aerodynamics that are sensitive to airflow (this study). Integration of laryngeal and breathing movements coupled with a detailed understanding of vocal organ morphology and laryngeal aerodynamics provides a fuller understanding of the nature of rodent acoustic signals and their evolution.

Ultimately, understanding how acoustic characters evolve requires knowledge of how signals are produced. Our findings indicate that diversity in the size of a unique laryngeal structure termed the vocal pouch is associated with extreme divergence in F₀ of ultrasonic whistles. Thus, the traditional approach of categorizing vocal signals by spectral range into ‘audible’ (up to 16 kHz) and ‘ultrasonic’ (above 19 kHz) vocalizations requires revision. Indeed, the challenge of differentiating ‘audible’ and ‘ultrasonic’ vocalizations in rodents has been noted (e.g. Grimsley et al., 2016; Kalcounis-Rueppell, 2010, 2018; Miller and Engstrom, 200,7, 2010, 2012). A production-based definition that distinguishes between flow-induced vocal fold vibration and aerodynamic whistles provides a more robust framework to explore diversification of rodent voices. Heliox experiments confirmed that the ultrasonic songs in pygmy mice are aerodynamic whistles and can reach low into the ‘sonic’ frequency range. Other species may have adaptations that extend the spectral range of whistles to low frequencies (this study) or the range of flow-induced vocal fold vibrations into exceptionally high spectral regions (e.g. Titze et al., 2016). The present study also informs our understanding of the vocal production mechanism, which requires precise coordination between three systems (larynx, breathing and vocal tract). Aerodynamics at the vocal folds (vibrating or not) are determined not only by the coordination of the three systems but also by the laryngeal morphology and biomechanical properties.

Altogether, the current study demonstrates that rodents are a promising system for understanding how mechanisms of complex behaviors like vocal production may promote or constrain divergence.

We thank Nathaniel Mull for assistance in trapping animals in the field and Bryan Carlin for assistance with recording animals in the laboratory.