California mice (Peromyscus californicus) adjust mouth movements for vocal production during early postnatal development

Abstract

Background

The order Rodentia is the largest group of mammals. Diversification of vocal communication has contributed to rodent radiation and allowed them to occupy diverse habitats and adopt different social systems. The mechanism by which efficient vocal sounds, which carry over surprisingly large distances, are generated is incompletely understood. Here we focused on the development and function of rhythmic mouth movements and laryngeal sound production. We studied spontaneously vocalizing California mice (Peromyscus californicus) through video and sound recordings. Mouth gape was estimated from video images and vocal characteristics were measured in synchronized sound recordings.

Results

California mice coordinated their mouth movements with laryngeal sound production but differently in two call types. In high-frequency whistles (“USV syllables”), mouth movements were present on postnatal day 1 but were reduced within the first 2 weeks of life. Mouth movements were prominently present during sustained vocalizations (“SV syllables”), and movements became more and more adjusted to syllable beginning and end. Maximum mouth gape was correlated with sound intensity and fundamental frequency of SV syllables. The effect on sound intensity was the strongest during postnatal development and most predictable when the mouth was closed by temporarily immobilizing the mandible in an elevated position.

Conclusions

This study demonstrates that rhythmic orofacial behavior not only plays a critical role in determining acoustic features of the vocal behavior of California mice but also shows remarkable adjustments during early development.

Background

Humans synchronize mouth movements with laryngeal sound production to generate distinct speech sounds [1,2,3]. This coordination can be disrupted in speech disorders such as dysarthria [4, 5]. The depression and elevation of the mandible also play a crucial role in vocal production in nonhuman primates [6,7,8], in bats [9], in domesticated cows and pigs [10, 11], in songbirds [12,13,14,15], and anurans [16, 17]. Rodents, being one of the most diverse mammalian groups [18], exhibit a wide range of vocal behaviors to regulate their social interactions [19,20,21,22]. However, the role of mouth movements in vocal production has remained largely unexplored. A preliminary survey of published video recordings suggested the presence of rhythmic orofacial movements during vocal production in numerous rodent species (Additional File 1 – Table S1). Hence, one of the aims of this study was to investigate the coordination between laryngeal sound production and mouth movements in a rodent.

Rodents produce communication sounds within the larynx through specific breathing and larynx movements corresponding to different call types [23,24,25]. Jaw movements are facilitated by various muscles, including those responsible for mandible depression (lateral pterigoid, digastric, and mylohyoid muscle) and mandible elevation (masseter, temporalis, medial and lateral pterygoid muscle) [26,27,28]. These mouth movements alter the upper airway geometry between the larynx and the mouth opening, potentially serving as a variable resonance filter [29]. Additionally, jaw movements affect mouth gape size (Fig. 1A), influencing the acoustic beam pattern, and, consequently, sound intensity around a vocalizing subject [30, 31]. However, neither of the two hypotheses has been tested in rodents. Particularly for small, primarily ground-dwelling rodents that communicate over relatively long distances [32, 33], the second mechanism appears crucial. Therefore, a primary objective of this study was to enhance our understanding of the role of orofacial movements in rodent vocal production.

Fig. 1

Mouth movements during vocal production in Peromyscus californicus neonates. A: Sound is generated at the larynx (VF, vocal folds). The sound travels through the vocal tract (VT) and is then radiated from the mouth (mouth gape, MG) (schematic modified after Titze, Palaparthi 2018). B and C: Still images of a PND 1 pup (B) and a PND 16 pup (C) during vocal production showing a closed mouth condition and a wide-open mouth condition. D: Amplitude (Level.; relative change in output voltage of microphone signal) and spectrogram of a call consisting of 3 SV syllables and 1 USV syllable. E: Corresponding mouth gape during call production shown in D. Note that 5 cyclic mandible movements are produced during the call. Only four mandible movements are associated with syllables. There is one additional unvoiced movement. F: Velocity trajectory of mandible movements during the call shown in D. G: Difference between the onset of mandible movement and the beginning of the associated syllable (N = 343 SV and USV syllables from 8 pups on PND 1). H: Difference between the offset of mandible movement and the end of the associated syllable (N = 343 SV and USV syllables from 8 pups on PND 1). I-N: Relationship between syllable and mouth movement repetition rates (I) between syllable duration and open-mouth phase duration (J), between mouth gape and mean fundamental frequency (K), mean sound intensity (L), relative amplitude of the second (M) and third harmonic (N). Parallel lines represent individual correlation lines or 8 pups on PND 1

This study was conducted in neonate California mice (Peromyscus californicus), a cricetid rodent species renowned for their monogamous social structure [34, 35]. Our investigation centered on two vocal types: audible vocalizations, referred to as ‘cries’ or ‘sustained vocalizations’ (hereafter ‘SV syllables’), and high-frequency whistles, known as ‘sweeps’ or ‘ultrasonic vocalizations” (hereafter ‘USV syllables’) [36,37,38]. When separated from their litter, California mouse neonates emit both SV and USV syllables, prompting parental retrieval into the nest [36, 39, 40]. Notably, SV and USV syllables are produced by airflow-induced vocal fold vibrations and by an aerodynamic whistle mechanism, respectively [38]. In Mus musculus it was shown that the motor control for the two types of vocalizations are also different [41]. However, it remains unknown whether these distinct vocal types are produced by call type-specific orofacial movement patterns. Therefore, another objective of this study was to explore differences in mouth movements between the two call types.

We investigated the coordination between mouth movements and laryngeal sound production during the initial 16 days of life, a period marked by rapid development in rodents [42]. Jaw movements were analyzed by assessing individual image frames extracted from video recordings of spontaneously vocalizing animals. To approximate laryngeal and breathing movements, we utilized the sound of SV and USV syllables [25]. Additionally, we explored the function of mandible movements by examining the relationship between mouth gape and acoustic characteristics of SV syllables. Furthermore, we investigated the acoustic impact of temporary mandible immobilization. Finally, we compared vocal patterns in California mice with anticipated vocal tract filter frequencies and acoustic beam patterns derived from anatomical measurements such as vocal tract length and skull dimensions.

Results

Mandible movement and acoustic variables in one-day-old pups

Pups showed prominent mouth movements throughout the first 16 days as illustrated in Fig. 1B and C, and further demonstrated in Additional File 2—Video 1. Pups produced SV and USV syllables from day 1. The proportion of mixed calls (containing SV and USV syllables) was 55.7% (25.0, stdev) on day 1, 63.3% (35.5, stdev) on day 4, 59.8% (32.2, stdev) on day 8, 26.7% (22.9, stdev) on day 12, and 2.5% (4.6, stdev) on day 16. An example call shown in Fig. 1D consists of four syllables ( 3 SV syllables and 1 USV syllable) accompanied by five mouth movements (Fig. 1E) with peak opening velocities between 10 to 20 mm/s and closing velocities between 5 and 15 mm/s (Fig. 1F).

The majority of SV and USV syllables were associated with rhythmic mouth movements. Only 11 of 272 SV syllables (5.5%), and 4 of 71 USV syllables (5.3%) were produced with a closed mouth and no discernable mandible movement (Table S1).

Most SV and USV syllables were embedded within a mandible movement cycle, but although the syllable duration was shorter than the corresponding open-mouth phase both durations were associated (linear mixed effects model; 343 voiced syllables from 8 pups; t = 15.4, p < 0.001). The lowering of the mandible started on average 26.5 ± 51.3 ms (mean ± stdev) before the beginning of a syllable (intercept-only linear mixed effects model; t = -3.9, p = 0.0064) (Fig. 1G). The raising of the mandible ended on average 122.7 ± 90.8 ms after a syllable ended (intercept-only linear mixed effects model; t = 9.8, p < 0.001) (Fig. 1H). Syllable duration and open-mouth phase duration were correlated (rmcorr, r = 0.69 (95% CI: 0.63, 0.74), p < 0.001) (Fig. 1I). The repetition rates of both movements ranged between 2.5 and 3.6 Hz and were highly correlated (rmcorr, r = 0.73 (95% CI: 0.66, 0.78), p < 0.001) (Fig. 1J).

Maximum mouth gape was predictive for fundamental frequency and sound intensity in SV syllables. Mouth gape was larger in SV syllables with higher f0_mean (rmcorr, r = 0.48, (95% CI: 0.38, 0.57); p < 0.001), higher f0_max (rmcorr, r = 0.41 (95% CI: 0.30, 0.51), p < 0.001), and larger f0_BW (rmcorr, r = 0.13, (95% CI: 0.01, 0.25); p < 0.05). Mouth gape was also larger in SV syllables with higher mean sound intensity (rmcorr; r = 0.69 (95% CI: 0.62, 0.75); p < 0.001), and higher maximum sound intensity (rmcorr; r = 0.67 (95% CI: 0.60, 0.73); p < 0.001). It was negatively associated with the relative amplitude of the first harmonic (rmcorr; r = -0.23 (95% CI: -0.34, -0.11); p < 0.001), and it was not associated with the relative amplitude of the second harmonic (rmcorr; r = 0.03 (95% CI: -0.10, 0.15); p = 0.71) (Fig. 1K-N).

Figure 1D illustrates also one unvoiced among four voiced mouth movements. We found that 53 of 396 mouth movements (13.4%) were unvoiced. Voiced and unvoiced mouth movements were similar. Five performance variables were not different between voiced and unvoiced mandible movements (linear mixed effects model; opening velocity: t = 1.57, p = 0.12; closing velocity: t = 0.84, p = 0.99; mouth gape amplitude: t = 1.67, p = 0.99; open-mouth phase duration: t = 1.96, p = 0.051; mouth movement repetition rate: t = 0.001; p = 0.99).

Development of mouth movement and acoustic variables between PND 1 and 16

There was no detectable trend in the proportion of SV syllables without noticeable mandible movement. The proportion varied between 4.2 and 10.9% among 8 pups on the 5 days (Fig. 2A) (linear mixed effect model; t = 1.9, p = 0.062). However, the average proportion of USV syllables without mandible movement increased from 5.3% on PND 1 to 97.7% on PND 16 (linear mixed effect model; t = 8.4, p < 0.001), and the proportion of unvoiced mouth movements decreased to almost 0% on PND 16 (linear mixed effect model; t = -5.3, p < 0.001) (Fig. 2A; Table S1).

Fig. 2

Development of mandible performance and vocal sounds. A The proportion of SV syllables that were not associated with a mandible movement did not change over 16 days. USV syllables that were not associated with a mandible movement increased to almost 100%, and mandible movements that were in the vicinity of an SV call but were not associated with sound decreased to zero. B Maximum mouth gape during syllable production. C Time difference between the onset of mandible movements and syllable beginning as well as between the offset of mandible movements and syllable end. D Mouth movement duration decreased over 16 days more than SV syllable duration. E Velocity of mandible movement at the beginning of a syllable (mandible depression) and at the end of a syllable (mandible elevation). F Repetition rates of mandible movements and syllable production. G Mean fundamental frequency (f0). H Fundamental frequency bandwidth. I Relative amplitude of the second (2F0-F0) and third (3F0-F0) harmonic. A-H Each data point represents the average from 8 and 9 pups, respectively (mean±stdev). J-M Linear mixed effect models for mouth gape and mean fundamental frequency (J), and maximum intensity (K), and second (L) and third relative harmonic (M)

Maximum mouth gape increased with age from 1 mm on PND 1 to more than 3 mm on PND 16 (Fig. 2B). Onset and offset differences became smaller over time (Fig. 2C). On PND 16 mandible movement preceded syllable onset by 7.5 ms ± 2.9 ms (intercept-only linear mixed effect model; t = -2.9; p < 0.05). Overall, the onset difference decreased by 0.0016 s/day (intercept-only linear mixed effect model; 95% CI: 0.0011, 0.0021; t = 6.4, p < 0.001). The offset difference decreased from 122.7 ± 90.8 ms on PND 1 to 37.6 ± 58.3 ms on PND 16, which amounts to an average decrease of -0.0054 s/day (intercept-only linear mixed effect model; 95% CI: -0.0061, -0.0047; t = -14.3, p < 0.001).

The reduction in on- and offset differences was achieved only by mouth movement adjustments not by changing syllable duration. SV syllable duration did not change during the first 16 days (linear mixed effects model; 95% CI: -0.0002, 0.0012; t = 1.5, p = 0.14). The open mouth phase duration decreased by -3.8 ms/day (linear mixed effects model; 95% CI: -4.7, -3.0; t = -8.5, p < 0.001) (Fig. 2D). Mouth movement became faster. The mouth opening velocity increased by 1.4 mm/s/day (linear mixed effects model; 95% CI: 1.3, 1.5; t = 28.8, p < 0.001), and the mouth closing velocity increased by 1.3 mm/s/day (linear mixed effects model; 95% CI: 1.2, 1.4; t = 36.0, p < 0.001) (Fig. 2E). The difference between syllable duration and the open mouth phase duration decreased from 0.12 s on PND 1 to 0.04 s on PND 16 (linear mixed effects model; coefficient = -0.0058 s/day; 95% CI: -0.007, -0.005; t = -15.7, p < 0.001). Syllable and mouth movement repetition rates increased both by similar rates (linear mixed effects model; syllables: 0.054 Hz/day, 95% CI: 0.047, 0.061, t = 14.6, p < 0.001; mouth movements: 0.053 Hz/day, 95% CI: 0.041, 0.065, t = 8.3, p < 0.001) (Fig. 2F) and therefore the difference in the repetition rates of the two movements did not change significantly with age (linear mixed effects model; coefficient = -0.0034 Hz/day, 95% CI: -0.007, 0.013, t = 0.67, p = 0.50). These results confirm that larynx sound production and mandible movements are coordinated, but adjustments were made predominantly to the opening and closing of the mouth.

SV syllable call characteristics, mouth gape, and closed-mouth vocalization

If the degree of mouth gape determines spectral features of the SV calls, we expected (a) that the relationship between mouth gape and acoustic parameters would strengthen with age, and (b) that disturbing mandible movement would cause changes in acoustic variables.

Fundamental frequency increased with age (linear mixed effect model; f0_mean: slope = 268.3 Hz/day, 95% CI: 245.0, 291.5, t = 22.6, p < 0.001; f0_max: slope = 213.5 Hz/day, 95% CI: 188.6, 238.4, t = 16.8, p < 0.001). Fundamental frequency bandwidth decreased with age (f0_BW: slope = -122.6 Hz/day, 95% CI: -147.6, -98.0, t = -9.7, p < 0.001). The relative amplitude of the second (linear mixed effect model; slope = 0.65 dB/day, 95% CI: 0.56—0.74,), t = 14.6, p < 0.001) and third harmonic increased with age (slope = 1.37 dB/day, 95% CI: 1.26—1.49, t = 23.4, p < 0.0001) (Fig. 2G-I).

The association with mouth gape increased with age for f0_mean (linear mixed effect model; t = 6.3, p < 0.001), f0_max (linear mixed effect model; t = 4.4, p < 0.001), I_mean (linear mixed effect model; t = 7.3, p < 0.001), I_max (linear mixed effect model; t = 10.5 p < 0.001). Mouth gape also co-varied with the amplitude of both the second (linear mixed effect model; t = 3.4 p < 0.001) and the third harmonic (linear mixed effect model; t = 3.9 p < 0.001) (Fig. 2J-M). Mouth gape did not covary with f0_BW (linear mixed effect model; t = 1.4 p = 0.173).

Differences in acoustic parameters of syllables produced with a closed mouth by immobilizing the mandible in an elevated position provided additional evidence for a role of mouth movements in determining the acoustic characteristics of SV syllables. Sound intensity decreased during the immobilization (paired t-test; t = 5.45, P = 0.002) (Fig. 3A). Neither mean fundamental frequency (paired t-test; t = -0.28, P = 0.79), syllable duration (paired t-test; t = 1.64, P = 0.16), nor the relative amplitude of the second (paired t-test; t = -0.07, P = 0.95) and third harmonic (paired t-test; t = 1.34, P = 0.22) changed systematically during the immobilization (Fig. 3B-E). Interestingly, syllable repetition rate decreased in all 6 pups during the immobilization (paired t-test; t = 4.06, P = 0.009) (Fig. 3 F) suggesting that the two rhythmic movements are coupled and a disturbance of one movement is associated with slowing of the other.

Fig. 3

Mandible immobilization causes vocal changes in SV calls of six pups (PND 5). A Mean syllable intensity. B Mean fundamental frequency. C Syllable duration. Relative amplitude of the second (D) and third harmonic (E). F Syllable repetition rate (SRR). Each data point represents the average from 6 pups (mean±stdev; N=6 pups)

Estimating vocal tract resonances and sound radiation efficiency

The distance between glottis and rostral end of the mouth was 5 mm in a one-day-old pup and 15 mm in a 50-day-old mouse. Utilizing these measurements with Eq. 1, we find that the first resonance of a one-day-old pup’s vocal tract is expected at approximately 16 kHz. The mean fundamental frequency range is 18 to 24 kHz for pups. The opening of the mouth effectively shortens the vocal tract potentially aligning the first formant and fundamental frequency (Fig. 4A).

Fig. 4

Predicting formant frequencies and structural wavelengths. A: Relationship between vocal tract length and first (blue line) and second (orange line) formant frequency. Mean fundamental frequency of SV syllables is near the first formant in one-day-old pups but approaches the second formant in older animals. B: Estimating sound and structural wavelengths. Acoustic wavelength decreases with increasing fundamental frequency. The structural wavelength (gray horizontal bar) was estimated from the ear-to-ear diameter of a day-old pup (14 mm) and a 50 day old mouse (18 mm). (PND, postnatal day)

Mean fundamental frequency in SV syllables ranges between 18 and 24 kHz, indicative of acoustic wavelengths between 16 and 28 mm (according to Eq. 2). The head diameter of a pup measured 14 mm and 18 mm in an adult mouse. We observed a close match between structural and acoustic wavelengths (Fig. 4B).

Discussion

Our experimental findings inform all three of our initial aims. Firstly, we observed that mouth movements occur during vocal production in many rodents, indicating the ubiquity of this behavior among mammals [3, 8, 9, 43, 44]. In neonate California mice, we noted that jaw movements were synchronized with the onset and offset of SV syllables, and this coordination was adjusted during the first two weeks of life. Secondly, we found that mouth movements in California mice were consistently linked to sound intensity, and sound intensity decreased in all pups when the mouth was temporarily closed by immobilizing the mandible in an elevated position. Moreover, theoretical considerations suggest that mouth movements during SV syllables likely serve to optimize both sound radiation from the mouth and vocal tract filtering. There were alignments between structural factors (such as vocal tract length and skull dimensions) and acoustic wavelengths. Thirdly, we observed distinct differences in mouth movements between call types. Specifically, jaw movements ceased during the production of USV syllables within the first two weeks of life, while mouth movements during SV syllable production became increasingly consistent over time.

Conserved networks synchronize laryngeal sound production and mouth movements

Four characteristics suggest conserved networks integrating laryngeal and mandible motor patterns for vocal production across mammalian clades. First, the mandible movement repetition rates in California mice (2–5 Hz) closely resembled rates observed in human speech (2–8 Hz) [45] and calling marmoset monkeys (4 – 8 Hz) [8].

Second, jaw movements exhibit vocal type specificity. Humans produce distinct speech sounds through specific articulation patterns [44, 46]. Similarly, California mice demonstrate distinct mouth movements, one for SV syllables and a different one for USV syllables.

Third, California mice exhibited numerous non-vocalized mouth movements resembling voiced movements, which diminished by day 16. Similar phenomena are observed in human infants during canonical babbling, aiding in motor coordination learning [47, 48]. Songbirds also exhibit non-vocalized ‘vocal movements’, interpreted as rehearsal to consolidate vocal motor patterns [49]. The role of non-vocalized mandible movements in neonate California mice warrants further investigation, possibly indicating a complex maturation process of orofacial motor patterns.

A fourth characteristic is the apparent coupling between laryngeal sound production and mouth movement, suggesting a role of somatosensory or auditory feedback in coordinating these actions akin to human speech [50] and other orofacial behaviors in rodents [51, 52]. Age- or noise -induced hearing loss and concomitant vocal changes are known for a related cricetid rodent [53]. Here, we observed that a disruption of jaw somatosensory input during immobilization was associated with a slower syllable repetition rate (Fig. 3F), suggesting consistent effects on coupled but independent rhythmic behavior. The coordination of multiple orofacial movements and the role of somatosensory feedback remains an active area of research, not only in vocal production [e.g., [54,55,56,57,58,59,60].

Finally, the step-wise recruitment of orofacial movements during speech development in humans suggests that jaw movements may be the oldest pattern associated with laryngeal sound production. In human speech, vowel and consonant sounds are linked to specific movements of the jaw, tongue, and lips, with jaw control appearing earliest among these speech-related motor patterns [47, 61].

Postnatal development of movement goals

Two of the vocal types produced by California mouse pups are SV and USV syllables [36, 39, 40]. These syllable types are typically organized into multi-syllabic calls. Notably, the occurrence of mixed calls – those containing both SV and USV syllables—was initially high but decreased on PND 12 and 16. Previous research in Mus musculus has indicated that lower frequency calls (SV calls in Peromyscus) and high frequency whistles (USV syllables in Peromyscus) are not only produced by different laryngeal mechanisms but also engage distinct neural circuits in the brain [22, 41]. Here, we show that these two vocal types (SV and USV syllables) are further characterized by distinct kinematic endpoints of jaw movements. Specifically, the onset and offset of the mouth movements aligned with the beginning and the end of the SV syllable throughout the first 16 postnatal days, with more pronounced adjustment observed at the end of the syllable.

Similar to human infants acquiring speech [2, 62, 63], California mice displayed characteristic increases in mandible performance markers, including velocity, repetition rate, and amplitude. However, the specific contributions of peripheral biomechanics (e.g., larynx size, mandible size, distribution of muscle fiber types) and neural control to changes in mandible performance remain to be studied [64,65,66,67,68,69,70,71].

Forebrain structures in rodents have been implicated in the timing of vocal signal production [72] and the generation of different types of rhythmic jaw movements [73,74,75], suggesting their potential involvement in achieving precise movement coordination in both temporal and spatial domain. Future studies may investigate whether the vocal behavior of the California mouse challenges the hypothesis of a limited role of forebrain structures in vocal motor control [76].

Functions of mouth movements for vocal production

The current findings provide support for two potential functions of mouth movements in sound production: (1) vocal tract filtering and (2) optimizing sound radiation.

California mice produce SV syllables through airflow-induced vocal fold vibrations and USV syllables through a whistle mechanism [38, 77]. The oral cavity plays a crucial role as part of the vocal tract filter, i.e., the geometry of the vocal tract determines how sound is shaped as it passes through the oral vocal tract. An adjustable vocal tract filter is essential for human speech [29], nonhuman primate vocal communication [78], for the characteristic sound of low fundamental frequency calls in various non-passerine species [79], and the tonal sound quality of songs produced by songbirds [15]. Songbirds and nonhuman primates achieve an effective tonality and sound intensity by matching fundamental frequency with a formant [14, 80]. We found that the fundamental frequency of SV syllables and the first two vocal tract filter frequencies overlap (Fig. 4A). A large mouth gape shortens the vocal tract tube at its rostral end, potentially aiding in matching fundamental frequency and the first filter frequency. We observed that filter frequencies changed during mandible immobilization but differently among pups (Fig. 3D, E). The diverse responses among pups could reflect immature coordination between laryngeal sound production and mouth movements. They could also reflect individual-specific responses to the immobilization procedure.

Future investigations will include older animals and explore the function of mouth movements in vocal communication. As the mice mature, the fundamental frequency of SV syllables increases (Fig. 2), and a longer vocal tract produces lower resonance frequencies, possibly resulting in a more precise alignment between filter and fundamental frequency. Additionally, behavioral research suggests the presence of distinct subtypes of SV calls [37], prompting further inquiry into whether these acoustic variations correlate with different patterns of mouth movements.

Mouth gape plays also a crucial role in the beam pattern formation around the vocalizing animal. Cricetid rodents not only position themselves in locations that maximize sound transmission [32] but also adjust their mouth opening to increase sound source energy. We found that the relationship between mouth gape and maximum sound intensity was significant in all age groups (Fig. 2H), that mandible immobilization at PND 5 resulted in decreased sound intensity in all six mice tested (Fig. 3A), and mouth movements are likely to assist in the matching between acoustic and structural wavelengths (Fig. 4B).

With a closed mouth, pup vocalizations appeared ‘mumbled’, i.e. all syllables were quieter. When the mouth of a pup was forcefully closed by immobilizing the mandible in an elevated position, sound intensity, and syllable repetition rates were lower, but changes in syllable duration, fundamental frequency, and relative harmonics were variable. We cannot exclude the possibility that the manipulation itself caused a change in emotional state of the pup which then contributed to some of the observed vocal changes. An association between orofacial movements and the emotional state of an animal was suggested in domesticated animals [10, 11]. In general, changes in arousal or emotional state, as noted by Mendl et al. [81], can cause vocal changes. In nonhuman mammals other than rodents, increased arousal was associated with higher sound intensity, faster repetition patterns, and utterances with longer duration [82,83,84]. While arousal or valence was not quantified here, no increased bodily activity was observed in the mouse pups. We believe the observed vocal changes were predominantly due to changes in vocal tract acoustics and beam pattern acoustics. However, future work may further explore the role of orofacial movements in shaping nuances of emotions onto acoustic parameters of a vocal signal.

Conclusions

Many biomedical research studies use USV whistling in Rattus rattus and Mus musculus mice. The California mouse seems to be an interesting model system with great potential for voice research. Mouth movements play a crucial role in vocal production. Initially, USV syllables in California mice are associated with large mouth movements, but by approximately postnatal week 2, mouth gape remains minimal. In contrast, SV syllables, produced through airflow-induced vocal fold vibration [38], require coordinated breathing, laryngeal, and mandibular movements. Acoustic features of SV syllables can therefore likely serve as indicators of aberrant movements or structural injury to the vocal organ, both the sound source and the vocal tract filter and their respective neural control circuits.

The widespread occurrence of rhythmic mouth movements during vocal production in rodents and other vertebrates suggests that this association with vocal production is phylogenetically old. It is plausible that the absence of mouth movements in USV whistling is a derived feature, while the coordination between laryngeal sound production and mandible movement is ancestral. In Mus musculus and Rattus rattus, mouth movements are absent during USV whistling [23], and both species are not known for conspicuous advertisement signals in the lower fundamental frequency range. Exploring orofacial movements in non-traditional rodent models in the context of vocal production holds promise for voice research.

Methods

Experimental setup and recording

All mice were subjected to standard animal care procedures following guidelines established by the American Society of Mammalogists [85]. Experiments were reviewed and approved by the Institutional Animal Care and Use Committee of Midwestern University (protocol 4196). Peromyscus californicus was obtained from the Peromyscus Genetic Stock Center at the University of South Carolina and bred at Midwestern University. Mice were housed at 22ºC, fed with a standard rodent diet (Teklad 2918), and ad libitum water access on a 12 h-light and 12 h-dark cycle. A total of 15 California mice were used in two experiments.

Nine mice (5 males, 4 females) were used to video record mouth movements and audio record associated vocal behavior. Video and audio recordings were made on postnatal days (PND) 1, 4, 8, 12, and 16. Pups were removed from their litter, and placed in a small cup with woodchip bedding for up to 5 min. Pups on PND 12 and 16 were mobile and were kept in a small cage for up to 20 min. From each pup up to 10 calls were analyzed. This resulted in 9 to 50 syllables from each pup on each day (Additional File 3—Table 2, Additional File 4 – Data).

Video imaging was performed with a GoPro Camera (Hero 7) at 60 frames per second and true 1,000 × 1,000 imaging resolution. Additional magnification was achieved with a mountable 10 × lens. The camera also recorded a sound with an internal microphone (approximate frequency range 20 Hz to 20 kHz). The camera was positioned 10 cm above or in front of the mouse. Sound was additionally recorded with an Avisoft ultrasound microphone (Avisoft-Bioacoustics, CM16/CMPA-5 V; frequency range 2 to 200 kHz; approximate sensitivity 500 mV Pa-1). The CM16 microphone was placed 25 cm above the mouse. The uncompressed sound signal was digitized with an Avisoft acquisition board (Ultrasoundgate 816H) and then saved using Avisoft Recorder software at a 200 kHz sampling rate. The camera microphone signal was used to synchronize the video signal with the Avisoft/CM16 sound signal.

Mouth gape estimation

A call consists of one or multiple syllables. A multi-syllable call can contain both SV and USV syllables. We calculated the ratio between mixed calls (containing SV and USV syllables) and SV-calls (containing only SV syllables).

Most syllables were associated with mouth movement (‘voiced mandible movement’). Some syllables showed no identifiable mouth movements (‘syllable without mouth movement’). Numerous mandible movements were not associated with sound production (‘unvoiced mandible movement’) but generated within the context of call production. All mandible movements within a multisyllabic call as well as mandible movements less than 500 ms before the first syllable or after the last syllable were included in the analysis of a call.

Mouth gape was determined on individual video frames by measuring the distance between one marker (small ink dot) on the upper and one on the lower lip. The landmarks for each marker were placed manually and their coordinates were digitized and exported as text file using MaxTraq Lite software (Innovision, Inc. 2022). The mouth gape measurement was corrected by subtracting the marker distance at baseline. Baseline marker distance was estimated over 10 to 20 video frames when the animal’s mouth was closed, and the animal rested.

During a single mouth movement, mouth gape first increases towards a maximum and then decreases towards baseline. The onset of a mouth movement was measured when the marker distance increased 0.15 mm from the baseline (= movement onset). A return to within 0.15 mm of baseline was considered the end of mandible movement (= movement offset). In between syllables (SV-calls and mixed calls), mouth gape did not return to or start at baseline. Therefore, the smallest value of mouth gape during a cyclic mandible movement (“opening-closing-opening”) was considered the end of a movement cycle and the start of the next movement cycle, respectively.

Only videos with a frontal or near-frontal view were used. Video frames in which markers could not be identified were discarded. A reference of known length positioned at the mid-sagittal level of the mouse allowed calibration of distance measurement.

Intraobserver error rate of mouth gape estimation

To estimate the intra-observer error rate of manually placing two landmarks on individual video frames, landmark placing was repeated for one call sequence that consisted of 5 SV syllables. The sequence was recorded on 115 video frames. The distance measurement on corresponding video frames differed on average by 0.11 mm (stdev.: 0.083 mm).

Mouth gape and mandible movement parameters

Five variables were derived to characterize mandible movement. They include the mouth-open phase duration (time of movement offset minus time at movement onset), absolute mouth gape, mandible movement repetition rate (MM RR = 1 / movement onset to movement onset interval), movement onset velocity, and movement offset velocity. Jaw movement velocity was estimated as the slope of mouth gape between movement onset or offset, respectively, and the maximum mouth gape.

Sound analysis

Five acoustic variables are measured in SV syllables: mean fundamental frequency (f0_mean), fundamental frequency bandwidth (f0_BW), mean sound intensity (I_mean), syllable duration (Dur), and syllable repetition rate (SRR). USV syllables were analyzed for syllable duration and syllable repetition rate. Acoustic measurements were performed by using the sound analysis software Praat (version 6.3.09; www.praat.org). PRAAT’s pitch and intensity tracking tool was utilized to measure fundamental frequency and intensity. F0 tracking results were visually confirmed by overlaying the measurement onto the spectrogram. The distance between pup and microphone was maintained constant allowing for within-pup comparisons.

Closed-mouth vocalization by immobilizing the mandible in an elevated position

To investigate the significance of mandibular movements for acoustic parameters in California mice, the mandible was immobilized in an elevated position, a technique also used in humans [86, 87] and songbirds [13, 88] for vocal production studies. Mandibular movement was temporarily restrained by placing a custom-made rubber band (about 2 mm wide) around the mandible and frontal/nasal bone. Six additional pups on postnatal day 5 (PND 5) were selected for this study, as they are robust enough to be handled for the placement of the rubber ring over the snout, yet still calm enough to tolerate the device while exhibiting robust vocal behavior. Older pups were observed to remove the ring with forelimb movement promptly. The pups' behavior during mandible immobilization did not raise concerns regarding compromised breathing or elevated stress levels. Bodily activity appeared not elevated during the immobilization. However, no formal measurement of stress levels was made.

Video images confirmed that the rubber band effectively prevented mouth opening. For analysis, SV syllables from 10 calls without immobilization and 10 with immobilization were analyzed. Parameters such as mean fundamental frequency, fundamental frequency bandwidth, mean sound intensity, syllable duration, syllable repetition rate, relative amplitudes of the second and third harmonic, were compared within individuals between vocalizations with and without immobilization.

Statistical analysis

Three separate investigations were conducted. The coordination between laryngeal sound production and mouth gape was first studied on postnatal day 1 (PND 1). On PND 1, a total of 343 syllables, and 396 cyclic mandible movements arranged into 80 calls from 8 pups were available for statistical analysis. Next, the postnatal development of sound production and mouth movements was investigated by including data from PND 1, 4, 8, 12, and 16. A total of 1735 syllables, and 1604 cyclic mouth movements from 9 pups on 5 postnatal days (PND 1, 4, 8, 12, 16) were available. One pup did not vocalize on PND 1 and a second pup did not vocalize on PND 16. Finally, the role of mouth movements for determining acoustic variables was investigated by exploring (a) the relationships between mouth gape and acoustic variables and (b) the effect of temporary mandibular immobilization.

Mandible movement and acoustic variables in one-day-old pups

The coordination between laryngeal and mandible movements was tested by comparing the difference between mandible movement onset and syllable beginning (= “onset difference”) as well as the difference between mandible movement termination and syllable end (= “offset difference”). Both differences were compared to 0 using an intercept-only linear mixed effects model (pup identity = random effect). Syllable duration and open-mouth duration were compared with a linear mixed effects model including syllable type (SV syllable; USV syllable) as fixed effect and pup ID as random effect.

Motor performance of voiced and unvoiced mouth movements were compared by five variables (opening velocity, closing velocity, open-mouth duration, maximum mouth gape, and mouth-movement repetition rate) using a linear mixed effects model. Pup ID and voicing were included as crossed random effects.

The relationships between (1) syllable duration and open-mouth duration, (2) syllable repetition rate and mouth movement repetition rates, and (3) mouth gape and six acoustic variables (mean and maximum fundamental frequency; mean and maximum sound intensity; relative second and relative third harmonic) were studied with repeated measure correlations (rmcorr). The procedure allows for determining within-individual association for paired measures assessed on two or more occasions for multiple individuals. Nonindependence among observations is accounted for by rmcorr by using analysis of covariance to adjust for interindividual variability. This removes measured variance between pups. Parallel regression lines with identical slopes and varying intercepts provide the best linear fit for each participant. The rmcorr coefficient (r_rm) is bounded by − 1 to 1 and represents the strength of the linear association between two variables.

Development of mandible movements and acoustic variables between PND 1 and 16. Potential trends in three different proportions (proportion of SV syllables without mouth movement; the proportion of USV syllables without mouth movements; the proportion of mouth movements without sound) were investigated with a linear mixed effect model (PND = fixed effect; pup ID = random effect). Onset and offset differences were compared to 0 with intercept-only linear mixed effect models for each PND. Potential trends in mouth movement variables (opening velocity, closing velocity, open-mouth duration, mouth-movement repetition rate) and syllable repetition rate were investigated with a linear mixed effects model (PND = fixed effect; pup ID = random effect).

We also studied whether syllable repetition rate and mouth movement repetition rate changed with age. If the two repetition rates changed differently, the difference between syllable repetition rate and mouth movement repetition rate is expected to change with age. This was tested with a linear mixed effect model (PND = fixed effect; pup ID = random effect). The procedure was repeated for the difference between syllable duration and open-mouth phase duration. The slope for every 1-day change was reported with the associated 95% confidence interval.

SV syllable call characteristics, mouth gape and mandible immobilization

The relationships between age and four acoustic variables (mean and maximum fundamental frequency; relative second and third harmonic) were also tested with linear mixed effects models (PND as fixed effect; pup ID as random effect). Again, the slope for every 1-day change was reported with the associated 95% confidence interval.

Vocal changes during temporary mandible immobilization were compared with paired t-tests.

Statistical analyses were performed using R (version 3.5.1). In all performed tests, P values < 0.05 were considered significant.

Estimating vocal tract resonances and sound radiation efficiency.

The opening and closing of the mouth alters the geometry of the airway above the larynx. The supra-laryngeal airway acts as a resonance filter and its geometry determines acoustic filter characteristics [29]. In a first approximation, the vocal tract filter can be modeled as uniform tube that is closed at one end (larynx side) and open at the other (mouth or nares). The resonance frequencies (also known as formants) can be estimated by Eq. 1.

$$\text{F(n)}=\frac{\left(2n-1\right)*c}{4*VTL}$$

(1)

, where n is the number of formant, c is speed of sound (350 m/s) and VTL is vocal tract length.

Mouth gape size can be associated with the sound power radiated around a vocalizing animal or human [31, 89]. Fundamental frequency is important for the relationship between mouth gape and intensity [30, 90]. The open mouth favors sound with a fundamental frequency whose wavelength matches the size of the radiating mouth/head. If the head diameter (= ‘structural wavelength’) is the same or smaller than the sound wavelength, Titze and Palaparthi [31] estimated energy losses at the mouth caused by back-reflection into the oral cavity to be near zero. In other words, radiation efficiency is maximized when the structural (i.e., head diameter) and acoustic wavelengths (wavelength of sound with a fundamental frequency) are approximately equal. Acoustic wavelength is calculated with Eq. 2:

$$\uplambda = \frac{c}{f0}$$

(2)

, where $\uplambda$ is the acoustic wavelength, c is speed of sound (350 m/s) and f0 is the fundamental frequency of the sound.