Anatomy and mechanisms of vocal production in harvest mice

Acoustic signals used by rodents for social communication are diverse in form (Dent et al., 2018; Fernández-Vargas et al., 2022). Understanding the processes that promote and constrain such diversification requires knowledge of underlying sound production mechanisms (Goller, 2022). Many rodents produce vocal signals over a broad frequency range between 0.5 and 120 kHz using two distinct production mechanisms (Pasch et al., 2017). Vocalizations between 0.5 and 25 kHz are produced using an airflow-induced vocal fold vibration mechanism, whereas those with fundamental frequencies above 17 kHz are produced by a whistle mechanism unique to rodents (Roberts, 1973; Riede et al., 2017; Riede and Pasch, 2020). The distribution of dual modes of vocal production remains understudied (Kubke and Wild, 2018; Riede et al., 2022). Cataloging the vocal repertoire, anatomy, and production mechanisms in diverse species can provide insight into the evolution of vocal communication systems.

Cricetid rodents in the subfamily Neotominae represent a vocally diverse group of muroid rodents (Fernández-Vargas et al., 2022; Kalcounis-Rueppell et al., 2018; Miller and Engstrom, 2007, 2010; Pasch et al., 2011, 2016). Members acoustically advertise their presence to conspecifics over long distances (>body length; Campbell et al., 2019; Miller and Engstrom, 2010, 2012; Pasch et al., 2011) with high amplitude vocalizations, or vocalize at much higher frequencies and lower sound amplitudes in close-distance interactions (Miller and Engstrom, 2010). However, the distribution of dual vocal production mechanisms across phylogeny remains unclear.

We studied western harvest mice [Reithrodontomys megalotis (Baird 1858)] to complement our understanding of vocal signal diversity and production mechanisms in Neotominae. Reithrodontomys is a sister taxon to the peromyscine genera Peromyscus and Onychomys, which use airflow-induced vocal fold vibrations to produce long-distance calls (Webster and Jones, 1982; Pasch et al., 2017), and to the baiomyine genera Baiomys and Scotinomys, which employ a whistle mechanism (Riede and Pasch, 2020). The acoustic repertoire of harvest mice includes long-distance calls (also known as ‘loud calls’, ‘songs’, ‘low-frequency calls’) that resemble baiomyine vocalizations in syllable repetition (Miller and Engstrom, 2007) and high-frequency quavers (also known as ‘ultrasonic vocalizations’) common to both baiomyines and peromyscines (Brito and Batallas, 2014; Miller and Engstrom, 2010). Long-distance calls are multi-syllable bouts with a fundamental frequency near 10 kHz and are similar in fundamental frequency to long-distance calls of Onychomys spp. and sustained vocalizations of Peromyscus spp. Quavers are single or multi-syllable vocalizations that have a modulated fundamental frequency above 60 kHz. Harvest mice travel large distances across their home range (Meserve, 1977; Clark et al., 1988; Smith et al., 2014) and may use long-distance calls to facilitate communication over large distances. We expected that long-distance calls are produced by airflow-induced tissue vibrations facilitated by multi-layer vocal folds and/or vocal membranes. Quavers are expected to be aerodynamic whistles facilitated by a ventral pouch with alar edge (Abhirami et al., 2023). Here, we studied these two call types and their production mechanisms using acoustic recordings in normal air and a light gas mixture. High-resolution micro-CT imaging and histology were used to characterize laryngeal anatomy. Additionally, long-distance calls were recorded from singly housed animals to assess the presence of nonlinear phenomena (NLP), an additional indicator of flow-induced sound production.

Twenty harvest mice were captured near Rusty's RV Ranch, Rodeo, NM, USA, using Sherman live traps baited with sterilized bird seed. Mice were transferred to standard mouse cages and maintained in animal facilities at Northern Arizona University, Flagstaff, AZ, USA. Animals were housed individually in the vivarium, maintained on a 14 h:10 h dark:light cycle (21±2°C), and provided rodent chow and water ad libitum. Animals were captured with permits from the New Mexico Game and Fish Department (permit #3779). Four females were pregnant at the time of capture and gave birth in captivity (3–4 pups/litter; N=13 captive-born). Not all mice spontaneously vocalized in social isolation. In total, 18 mice (10 females, 8 males; 10 wild-caught, 8 captive-born) were recorded in solitary conditions at Northern Arizona University. Sixteen mice (8 males, 8 females) were then transferred to Midwestern University, Glendale, AZ, USA, for heliox experiments and anatomical analysis of the larynx. All procedures were performed in accordance with ethical standards and approval Northern Arizona University's Institutional Animal Care and Use Committee (19-006) and the Institutional Animal Care and Use Committee at Midwestern University (MWU#3011), as well as the guidelines of the American Society of Mammalogists (Sikes and Animal Care and Use Committee of the American Society of Mammalogists, 2016).

Individually housed mice in their home cage (N=18) were placed in semi-anechoic coolers lined with acoustic foam and recorded continuously over a period of 2–7 days. We used 1/4′′ microphones (Type 40BE, G.R.A.S., Holte, Denmark) connected to preamplifiers (Type 26CB, G.R.A.S.) to obtain long-distance recordings 33.3 cm above the center of the mouse cage. The microphone response was flat within ±1.5 dB from 10 Hz to 50 kHz, and the pre-amplifier response was flat within ±0.2 dB from 2 Hz to 200 kHz. Microphones were connected to a National Instruments Data Acquisition hardware (DAQ, USB 4431, Austin, TX, USA) sampling at 102.4 kHz to a desktop computer running a custom recording MATLAB software (v. 2018a) to record high amplitude long-distance vocalizations.

Long-distance calls were counted for each individual. Furthermore, we counted syllables with NLP categorized into subharmonic (SH), frequency jump (FJ), deterministic chaos (CH) or biphonation (BP). PRAAT software was used for acoustic analysis (v. 5.3.80, retrieved January 2014 from http://www.praat.org/). We identified NLP by visual inspection of a narrowband spectrogram of the signal (Herzel, 1993; Riede et al., 2000, 2004; Titze et al., 2008; Zollinger et al., 2008) and report the percentage occurrence of syllables with NLP and absolute numbers of syllables with SH, FJ, CH or BP.

To assess mechanisms of sound production, we randomly selected a subset of the most vocal animals to record in opposite-sex pairs (n=4 pairs) in air and heliox. Acoustic recordings in light gas atmosphere can be used to infer the sound production mechanism of both call types. The fundamental frequency (f₀) of aerodynamic whistles increases in light gas (Roberts, 1973; Riede, 2011; Pasch et al., 2017; Riede and Pasch, 2020), whereas the vibration rate of vocal folds in response to a passing airflow does not change (Titze et al., 2016). A pair was placed in an acrylic cage lined with fresh bedding. A 12 mm tube that delivered heliox gas (80% He, 20% O₂) at flow rates between 10 and 20 l min⁻¹ was placed into the cage after a brief (∼5 min) habituation period. A small dog whistle was built into the floor of the cage and blown periodically by a researcher to monitor the heliox concentration. The ratio of the frequency of the whistle tone in air and in heliox allowed estimation of the expected effect of light gas concentrations. Vocalizations were recorded with an ultrasonic microphone (Avisoft Bioacoustics, CM16/CMPA-5 V; frequency range: 2 to 200 kHz; approximate sensitivity is 500 mV/Pa) embedded into the cage top. Sounds were acquired through a National Instruments acquisition device (NI DAQ 6212) sampled at 300 kHz and saved as uncompressed .wav files using Avisoft Recorder software (v. 3.4.2, Avisoft-Bioacoustics, Berlin, Germany).

Long-distance calls and quavers were recorded in heliox and air. Mean fundamental frequency (f_0,mean) and syllable duration were extracted for each vocalization type. Fundamental frequency was quantified every 20 ms using the PRAAT pitch-tracking tool [1024-point Fast Fourier Transform (FFT), 75% frame size, Hann window, frequency resolution 100 Hz, temporal resolution 93.75%, 0.625 ms]. The PRAAT ‘extract pitch’ function tracks the visible fundamental frequency contour in the spectrogram. After visual confirmation that the tracking process was correct, the [time, frequency] values were imported to Excel (Microsoft) and further processed.

To quantify laryngeal anatomy, 16 (8 males, 8 females) mice were euthanized with ketamine/xylazine and transcardially perfused first with saline solution followed by 10% buffered formalin. Larynges were dissected and placed in 10% buffered formalin phosphate (SF100-4; Fisher Scientific) for 24 h. 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 at 5 µm resolution. CT scanning was done using a Skyscan 1172 (Bruker). Reconstructed image stacks were then imported into AVIZO software (v. 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. Reconstruction was used for qualitative descriptions of larynx morphology and an estimation of vocal fold length measured as the distance between the vocal process of the arytenoid cartilage and the vocal fold attachment to the interior of the thyroid cartilage, and the volume of the ventral pouch. The length of the right femur was used as proxy for body size to assess if laryngeal anatomy scales allometrically (Darwaiz et al., 2022).

To qualitatively describe vocal fold morphology, four specimens (2 males, 2 females) were used to prepare histological sections. Mid-membraneous coronal sections (5 mm thick) were stained with Haematoxylin and Eosin for a general overview, Masson's Trichrome (TRI) for collagen fiber stain and Elastica–Van Gieson (EVG) for elastic fiber stain. Sections were scanned with an Aperio CS 2 slide scanner and processed with Imagescope software (v. 8.2.5.1263; Aperio Tech.).

We used non-parametric tests, Wilcoxon sign rank test and Friedman repeated measures analysis of variance on ranks, to evaluate differences in the number of calls, the number of syllables, NLP occurrence and NLP duration between the sexes. We used paired t-tests to assess acoustic differences between normal air and heliox vocal sounds. Analyses were performed with SPSS (IBM SPSS software). Significance levels were set P=0.05.

Twenty-one mice produced a total of 1241 long-distance calls in social isolation. Eleven female mice produced 505 long-distance calls (call range: 1–320; syllable range per call: 5–20) and ten males produced 736 long-distance calls (call range: 1–425; syllable range per call: 7–19). Female and male harvest mice produced long-distance calls at similar rates (U=4.5, P=0.52) with similar numbers of syllables per long-distance call (F_2,39=2.02, P=0.17). Harvest mice long-distance calls typically contained 11 syllables (females: 11.2±2.4; males: 11.5±2.0), with only 2 animals (1 male, 1 female) producing bouts with more than 19 syllables (Table S1).

Harvest mice produced vocalizations by two distinct mechanisms. Long-distance calls were produced by airflow-induced vocal fold vibrations. In contrast, a whistle mechanism mediates quaver calls with fundamental frequencies between 50 and 120 kHz. Fig. 1A,B shows spectrographic representations of both call types produced in air and in heliox. Neither mean fundamental frequency (t=0.12, P=0.45) nor call duration (t=0.59, P=0.29) of long-distance calls changed in heliox compared with normal air (Fig. 1C,D). Mean fundamental frequency of quavers increased in heliox (paired t-tests, t=−19.7, P<0.001) compared with normal air, whereas duration of quavers did not change (t=−1.56, P=0.11) (Fig. 1E,F).

Long-distance calls showed various NLP. Irregular vocal fold vibrations can produce NLP, and thus their presence can serve as indicators for the mode of sound production (Herzel, 1993). Individuals varied in the percentage of their vocalizations that contained at least one type (range: females, 0–46%; males, 0–87%). SH accounted for the most common NLP. Across all long-distance vocalizations (n=754, n=18 mice; males and females combined) containing NLP, 295 (39%) had subharmonics, 18 (2.3%) had frequency jumps, 1 (0.1%) had biphonation, 2 (0.3%) had deterministic chaos, and 19 (2.5%) contained multiple NLP (Table S1). The exact mechanisms that cause nonlinear phenomena in rodents are unknown but candidates include the presence of vocal membranes (Mergell et al., 1999), asymmetries in vocal fold size (e.g. Tokuda et al., 2007), or nonlinear interactions between the larynx and vocal tract filter (e.g. Titze et al., 2008).

The findings complement our understanding of production and diversification of vocal signals in neotomine rodents. Most species in this group studied thus far produce long-distance signals with a fundamental frequency in the 8–30 kHz range, consisting either of one syllable (Onychomys spp.) or multiple stereotyped syllables repeated at rates of 3–8 Hz (Baiomys, Scotinomys, Peromyscus, Reithrodontomys spp.). In Reithrodontomyini, syllables are generated by airflow-induced vocal fold vibrations (Pasch et al., 2017; Riede et al., 2022; this study). However, in Baiomyini (Baiomys spp.; Riede and Pasch, 2020) the syllables of the long-distance calls are whistled.

The larynx consisted of thyroid, cricoid, arytenoid cartilages, epiglottis and an alar cartilage (Fig. 2A–C). The alar cartilage is connected with the ventral aspect of the epiglottis. Vocal fold length ranged between 240 and 390 µm. Femur length did not scale significantly on body mass (r²=0.25; P=0.21) (Fig. 2H). Neither vocal fold length (r²=0.005; P=0.93) nor cartilage size (thyroid cartilage: r²=0.061; P=0.55; cricoid cartilage: r²=0.015; P=0.77; arytenoid cartilage: r²=0.14; P=0.37) were associated with femur length (Fig. 2D–I).

Vocal folds consisted of a large thyroarytenoid muscle (∼350 µm in latero-lateral diameter), lamina propria (35 µm) and epithelium. The lamina propria contains cellular (fibroblasts) and noncellular components (collagen and elastin fibers) (Fig. 2J–M). In general, a layered organization of the vocal folds (i.e. epithelium, lamina propria) facilitates production of a larger range of frequencies of vocalizations generated by airflow-induced vocal fold vibrations (Hirano, 1981; Titze et al., 2016), and variation in the relative size and position of layers contributes to species-specific call properties. In particular, the presence of vocal membranes is hypothesized to contribute to the expansion of the fundamental frequency range and/or the increase in glottal efficiency (Pasch et al., 2017; Kanaya et al., 2022). However, although common in Onychomys spp. (Pasch et al., 2017), vocal membranes were present only in a subset of Peromyscus spp. (Riede et al., 2022) and absent in harvest mice (this study) and pygmy mice (Riede and Pasch, 2020). Such findings indicate that vocal membranes may arise from mechanical stress associated with use and form later in life, as is found in marmosets (Zhang et al., 2019). Correlational studies that assess variation in vocal membrane presence and morphology with call rates and experiments that manipulate call rate will provide important insight. Such studies will also help clarify the cause and potential function of nonlinear phenomena found in the long-distance calls of harvest mice (herein) and Peromyscus spp. (Riede et al., 2022).

A ventral pouch was present in all 8 individuals investigated. The ventral pouch is positioned medially and rostral from the vocal folds (Fig. 2A). Its volume was not related to femur length (r²=0.22; P=0.26) (Fig. 2D). All rodent larynges investigated to date which produce high-frequency whistles possess a ventral pouch and an alar cartilage. Both structures appear critical to the generation of ultrasonic whistles (Riede et al., 2017; Abhirami et al., 2023). The ventral pouches of R. megalotis (this study), Onychomys spp. (Pasch et al., 2017), Peromyscus spp. (Riede et al., 2022), Rattus norvegicus (Riede et al., 2017) and Mus musculus (Darwaiz et al., 2022) are shaped like a sphere. The ventral pouch in Baiomys spp. is larger than in all members of Reithrodontomyinae. Such large size was interpreted as causing the unusually low spectral range of their long-distance call syllables (18–40 kHz) that overlap in frequency with syllables produced by Reithrodontomyinae (Riede and Pasch, 2020). However, a systematic analysis of the relationship between ventral pouch design and acoustic properties of high-frequency whistles is somewhat limited. Abhirami et al. (2023) studied the role of two aspects of the ventral pouch in determining fundamental frequency variation: (1) glottal airflow rates over the pouch and (2) changes in the glottal–alar edge distance. Both variables are critical for the generation of high-frequency whistles and changes of both parameters cause fundamental frequency changes in high-frequency whistles (Abhirami et al., 2023). An effect of different volume size of the ventral pouch on frequency parameters while controlling for glottal airflow rates and glottal–alar edge distance, has not been studied so far. However, an ontogenetic study in M. musculus suggested that ventral pouch volume remains rather constant during postnatal development which was interpreted as cause for the relatively small shift in fundamental frequency range of high-frequency whistles with age despite a large change in body size (Darwaiz et al., 2022).

Although ultimate factors underlying the origin and diversification of rodent vocalizations are debated (Hofer, 2010; Brudzynski, 2014), our study provides a relevant mechanistic perspective on the functional consequences. We infer that the most recent common ancestor in the Neotominae group was capable of both vocal production mechanisms, but that Reithrodontomyini and Baiomyini evolved different mechanisms to achieve communication over longer distances (Fig. 3). Emerging evidence indicates that amplitudes of long-distance vocalizations differ between flow-induced vocalizations in Reithrodontomyini compared with whistling Baiomyini. For example, reithrodontomyine grasshopper mice (Onychomys spp.) produce calls at 85.7 dB SPL at 1 m (Green et al., 2020) with an active space (communication distance) of 50 m and pinyon mice (P. truei) produce calls at 72.1 dB with a communication distance of 12.4 m (Brzozowski et al., 2023). In contrast, the amplitudes of baiomyine whistles appear much lower in amplitude (59.6 dB at 1 m in Scotinomys spp.; Pasch et al., 2011) with an estimated communication distance of 5 m (Miller and Engstrom, 2007). Thus, >20 dB difference in sound pressure levels may translate into 10-fold differences in communication distance. The mechanistic explanation for this difference is intrinsic to the sound production mechanisms themselves.

Whereas sound pressure amplitudes produced by airflow-induced vocal fold vibrations are determined predominantly by subglottal pressure (see Thomson et al., 2005; Zhang, 2016), precise interactions between subglottal pressure (Fletcher, 1976) and airflow properties downstream from the glottis shape whistle amplitudes (Riede et al., 2017; Abhirami et al., 2023). In rats, subglottal pressures of ultrasonic whistles (22 kHz and 50 kHz calls) range between 0.5 and 2.9 kPa (Riede, 2011). In contrast, audible calls produced by airflow-induced vocal fold vibrations are generated with much higher subglottal pressures (>6 kPa; T.R., unpublished data) similar to those found in humans and in excised larynx experiments studying larynges from various nonhuman mammals (up to 15 kPa; Holmberg et al., 1988; Riede and Brown, 2013; Lagier et al., 2017). We hypothesize that rodent whistles depend on low and narrowly controlled subglottal pressures and therefore cannot achieve comparably high sound pressures of their airflow-induced counterparts. In turn, high-amplitude vocalizations permit larger communication distances (Larsen and Wahlberg, 2017) corresponding to larger home ranges and territories in Reithrodontomyini (e.g. ∼17,000–40,000 m² in Onychomys spp.; Stapp, 1999; Kraft and Stapp, 2013) compared with Baiomyini (e.g. 200–1620 m² in Scotinomys spp.; Blondel et al., 2009, Ribble and Rathbun, 2018). The ability to use a wide subglottal pressure range to facilitate a wide range of sound amplitudes is common within and among species of echolocating bats to facilitate detection of prey at long (aerial hawkers) and short (gleaners) distances (Hackett et al., 2014) and may manifest across lineages in rodents to subserve social communication. However, in contrast to rodents, all bats generate sound by airflow-induced vibrations of intralaryngeal tissues (vocal folds or vocal membranes; Schnitzler, 1970, 1973; Roberts, 1973; Hartley and Suthers, 1988), raising important distinctions for future comparative studies.

In summary, the most recent common ancestor in the Neotominae group was likely capable of both vocal production mechanisms. Reithrodontomyini and Baiomyini employ different mechanisms for the long-distance vocalizations, likely contributing to large differences in communication distance. The structural variability of vocal folds among Neotominae on the one hand, and similarity of vocal production mechanism with human speech production on the other, make cricetid rodents a promising system to investigate the relationship between vocal fold structure and function.

We thank Ryan Brzozowski and Mariah E. Letowt for assistance with animal husbandry, and David Hendershott for assistance with trapping animals in the field.