Shared and Separate Neuromuscular Underpinnings of Swallowing and Motor Speech Development in the School-Age Years

Abstract

Purpose:

Despite co-occurrence of swallowing and speech disorders in childhood, there is limited research on shared and separate neuromuscular underpinnings of these functions. The purpose of this study was to (a) compare neuromuscular control of swallowing and speech between younger and older children and (b) determine similarities and differences in neuromuscular control of swallowing and speech.

Method:

Twenty-six typically developing children (thirteen 7- to 8-year-olds and thirteen 11- to 12-year-olds) completed this cross-sectional study. Neuromuscular control was evaluated using surface electromyography of submental muscles and superior and inferior orbicularis oris muscles during parallel tasks of swallowing and speech. Outcome measures included normalized mean amplitude, burst duration, time to peak amplitude, and bilateral synchrony, which were examined using mixed-effects models.

Results:

For normalized mean amplitude, burst duration, and time to peak amplitude, there were significant two- and three-way interactions between muscle group, task, and age group, indicating that older and younger children demonstrated different muscle activation patterns, and these patterns varied by muscle and task. No differences were noted between groups for bilateral synchrony. For parallel tasks, children demonstrated different magnitudes of normalized mean amplitude and time to peak amplitude of speech and swallowing. However, they demonstrated a similar pattern: increases in magnitude as task complexity increased.

Conclusions:

Children continue to demonstrate refinement of their neuromuscular control of swallowing and speech between 7–8 and 11–12 years of age, and there are both shared and separate elements of neuromuscular control between these two vital functions. To improve generalizability of findings, future research should include longitudinal analysis of swallowing and speech development, as well as measures of central neurophysiology.

Supplemental Material:

https://doi.org/10.23641/asha.23796258

Swallowing and speech production are fundamental human behaviors that are, respectively, key to our survival and, crucially, relate to one another. Disorders of swallowing or speech can have serious consequences, such as undernutrition (Benfer et al., 2016) and decreased social connection (Azios et al., 2016), and are particularly devastating when they co-occur, which is relatively common across the lifespan. In pediatrics, clinical research reveals that these disorders frequently co-occur in children with cerebral palsy (CP; Motion et al., 2002), pediatric stroke (Sherman et al., 2021), and preterm birth (Adams-Chapman et al., 2013; Azios et al., 2016; Msall, 2006; Ream & Lehwald, 2018), among other conditions. Further, there is evidence that both systems may be impacted even in children who do not have neurodevelopmental diagnoses. For example, based on a retrospective chart review study, 48%–87% of older children with communication disorders and no history of preterm birth, anatomical, structural, neurodevelopmental, or cognitive disorders were found to have a history of feeding and swallowing difficulties (Malas et al., 2017). In addition to co-occurrence of swallowing and communication deficits, there is also evidence that early performance in one domain (i.e., feeding) may predict later communicative and/or neurodevelopmental outcomes (Adams-Chapman et al., 2013; Mizuno & Ueda, 2005).

Despite these clinical observations, there has been limited empirical research on the potential underlying relationship between swallowing and communicative control, and conclusions about their precise relationship have been mixed. Thus far, research in this area has primarily considered one element of communication: speech, and swallowing has been examined tangentially or separately. In the speech literature, two main views have been prominent: the task-dependent model, where speech is hypothesized to be controlled by a specialized system distinct from all other motor behaviors (Ziegler, 2003), and the integrative model, where speech motor control overlaps with other motor behaviors (Maas, 2017), which could include swallowing. The task-dependent model has predominated but is recently facing challenge (Lancheros et al., 2020; Maas, 2017). Most existing research comparing the peripheral control of speech and nonspeech movements largely falls under the task-dependent model, where speech is hypothesized to be separate from other domains (Green et al., 1997, 2000; Steeve & Moore, 2009). Two studies suggested task-dependent peripheral control of speech and feeding-related nonspeech movements (i.e., chewing), but neither directly evaluated swallowing (Ruark & Moore, 1997; Steeve & Moore, 2009). In these studies, speech is not compared directly to swallowing but instead to feeding-related movements, like chewing. This absence of research directly comparing speech and swallowing makes it difficult to draw conclusions about their shared or separate neural control. This research would be important because it could provide an underlying rationale for the clinical findings of deficit co-occurrence (Malas et al., 2017; Sherman et al., 2021) and could uncover new intervention targets with cross-system impact.

In order to determine shared and separate underpinnings of these two vitally important functions, it would be helpful to have experimental evidence, using carefully designed tasks, to compare and contrast activation patterns. This research should build on previous evidence on development within each individual domain. Peripheral neural control of swallowing development has been rarely studied, but there has been extensive exploration of the development of the peripheral control for speech. Specifically, the neuromuscular control of the speech motor system is known to be protracted throughout childhood, refining over time and becoming adult-like sometime after 14 years of age (yoa; Smith & Zelaznik, 2004; Vorperian et al., 2009). This developmental path is characterized by neural and anatomic changes. Anatomic development of structures and muscles has been quantified by magnetic resonance imaging, which reveals ongoing nonuniform growth of all oral and pharyngeal components of the vocal tract between birth and 19 yoa, including hard and soft tissue structures, such as the pharynx and larynx (Vorperian et al., 2005, 2009, p. 200). These anatomical changes are characterized by both pre- and postpubertal sex differences, and each individual's growth pattern directly influences acoustic development of speech (Vorperian & Kent, 2007). Anatomical changes are also paralleled by changes in movement patterns, such as increased laryngeal control, during spoken language (Green et al., 2002; Stathopoulos & Sapienza, 1997).

Movement patterns underlying these changes have also been extensively examined using surface electromyography (sEMG) and kinematic methods. Kinematic studies reveal slower and more variable oral speech muscle activity in children compared to adults (Goffman & Smith, 1999; Green et al., 2000; Sharkey & Folkins, 1985; Smith & Goffman, 1998). Additionally, the relative contribution of muscles to speech control may change over childhood. For example, vertical movement of the jaw starts maturing earlier than independently controlled lip movement (Green et al., 2002) but may not fully reach the level of adult performance before the lips (Walsh & Smith, 2002). Further, development of speech motor control continues throughout childhood at a nonlinear pace: Children may demonstrate a plateau of coordinative development in late childhood (7–12 yoa), followed by a secondary period of maturation after the age of 12 years (Smith & Zelaznik, 2004). Overall, there is robust evidence that speech motor development continues through the teenage years, which is consistent with current theories of sensorimotor development, supporting protracted refinement of neuromuscular systems throughout childhood (Kunnen & van Geert, 2012; Goodway et al., 2019; Thelen, 2005).

According to these theories, feeding and swallowing motor control should also continue refining throughout childhood, but there are relatively few studies on the development of peripheral swallowing control, and findings are mixed. As a related feeding skill, chewing has been examined in young children, with evidence that children demonstrate improvements in mandibular control and muscle coordination for chewing between 9 and 36 months, and these changes are impacted by the structural differences in foods (Simione et al., 2018). Related specifically to swallowing, three studies provide some initial evidence of swallowing refinement throughout the school-age years, by demonstrating decreased swallowing duration in older compared to younger children when measured via swallowing acoustics (Hennessey et al., 2018) or sEMG (Green et al., 1997; Vaiman et al., 2004). Green et al. (1997) also found that cross-correlation, a measure of bilateral synchrony of muscle activity, increased, whereas variability decreased with age when 12-month-olds were compared to 48-month-old typically developing children (Green et al., 1997). In contrast, Ruark et al. examined amplitude of the perilabial, submental, and laryngeal strap muscles in 10 females in each of three age groups—5-year-olds, 8-year-olds, and young adults—and found no main effect for age on muscle activation patterns, though age differences in duration approached significance (Ruark et al., 2002).

These mixed results are difficult to interpret, which is further complicated by the small sample sizes, mixed ages, and methodological limitations of these studies. More specifically, in the study of Green et al. (1997), the investigators did not compare the same food consistencies across infants, whereas in the study of Vaiman et al. (2004), only one consistency (liquids) was tested. Also, the sEMG studies compared sEMG data between subjects without utilizing a maximum voluntary contraction (MVC) task to normalize the sEMG signal, which is vital in order to reliably compare sEMG amplitude values between subjects (Stepp, 2012).

In an effort to add to this literature and ensure reliable comparisons of neuromuscular control between domains, we recently developed a standardized sEMG research paradigm including tasks of swallowing and speech, increasing incrementally and in a parallel fashion in motor complexity (Hahn Arkenberg et al., in revision). In this previous study, we compared swallowing and speech in school-age children with unilateral CP and typically developing peers. We examined three electromyographic outcome measures that encompass components of the timing of muscle activation (time to peak amplitude), muscular effort (normalized mean amplitude), and coordination (bilateral synchrony). This design enabled reliable comparisons between these two functions across muscle groups, and we found that children with typical development demonstrated more similar muscular activation patterns for parallel tasks of swallowing and speech than children with unilateral CP (Hahn Arkenberg et al., in revision). We also observed preliminary patterns emerging that differentiated between the younger and older typically developing children, which informed the motivation for this study.

In order to systematically investigate these preliminary observations and the shared and separate neuromuscular control of swallowing and speech across typical development, we sought to recruit more typically developing children at two key points in neurodevelopment, 7–8 and 11–12 yoa. In addition to our earlier observations, there is compelling scientific rationale to examine these ages. Children demonstrate decreases in cortical gray matter volume and thickness throughout childhood and adolescence (Dima et al., 2021; Tamnes et al., 2017), which are primarily associated with increased myelination and general white matter maturation (Lebel & Deoni, 2018). However, there are periods of thickening throughout development that are relevant for the domains of swallowing and speech. Specifically, at 7–8 yoa, there is evidence of cortical thickening specifically in areas important for swallowing and speech control, such as the frontal operculum (Sowell et al., 2004), and 11–12 yoa encompasses the peak of gray matter volume (Giedd et al., 1999; Sowell et al., 2004).

The purpose of this study was to investigate the underlying neuromuscular control of swallowing and speech in these two periods of normal development. First, we aimed to determine if there are differences in neuromuscular control of swallowing and speech in older children (11–12 yoa) compared to younger children (7–8 yoa) and, second, to determine shared and separate patterns of activation between parallel tasks of swallowing and speech that increase gradually in motor complexity. We built upon the methodology in Hahn Arkenberg et al. (in revision) by examining the same three muscle groups with an additional outcome measure commonly used in other electromyographic studies across the lifespan: burst duration. The muscle groups were selected to encompass key areas necessary for bilabial speech tasks (upper and lower lips) and swallowing tasks (submental). Based on prior literature (Green et al., 1997; Smith & Zelaznik, 2004), we hypothesized that older children would demonstrate refinement of neuromuscular control for both swallowing and speech across all muscle groups (specifically exemplified by decreased burst sEMG duration, normalized mean amplitude, and time to peak amplitude and increased bilateral synchrony). Second, based on our prior work (Hahn Arkenberg et al., in revision; Malandraki et al., 2022), we hypothesized that children across age groups would demonstrate similar patterns of activation, timing, and coordination for swallowing and speech tasks, revealing aspects of shared neuromuscular control.

Method

This study was a cross-sectional design, with data collected at Purdue University, approved by the Purdue Institutional Review Board (No. 1410015417; IRB-2021-216). Data collection was completed from 2015 to 2021, with a pause from March 2020 to March 2021 due to the COVID-19 pandemic.

Recruitment and Screening

Inclusion criteria were age (7–8 or 11–12 yoa), being native speakers of American English, and scoring within the normative range on language subtests of the Clinical Evaluation of Language Fundamentals–Fifth Edition (CELF-5; Wiig et al., 2013) and on nonverbal intelligence screening (Test of Nonverbal Intelligence–Fourth Edition [TONI-4]; Fopiano, 2013). Exclusion criteria were history of neurological, speech, language, hearing, or swallowing disorders. Children were recruited for participation from the Greater Lafayette, Indiana area.

For the majority of participants (15/26), all screenings, including a hearing screening, the TONI-4; (Fopiano, 2013), the CELF-5 subtests (Coret & McCrimmon, 2015), and a cranial nerve examination, were completed in person before a temporary study pause due to the COVID-19 pandemic. For the remaining children (11/26), all screenings were conducted online using secure video-conferencing, as part of a comprehensive effort to reduce in-person contact due to the COVID-19 pandemic, and hearing status was evaluated by parent report.

Data Collection and Experimental Tasks

Our procedure for sEMG data collection has been described in detail in Hahn Arkenberg et al. (in revision). In short, double differential wireless electrodes (Trigno, Delsys) were placed on the left and right superior and inferior orbicularis oris and on the left and right submental area (under the chin; see Supplemental Material S1). These areas enabled comparison between swallowing and speech with muscles active during both functions (Hahn Arkenberg et al., in revision). Superior perilabial electrodes were placed between the midline and the corner of the mouth with grounds on the forehead (Walsh & Smith, 2013). Inferior perilabial electrodes were positioned horizontally below the vermillion border of the lower lip, between the midline and corner of the mouth with grounds on zygomatic bones. Submental electrodes were placed below the chin, parallel to the mandibular line with grounds on mastoid process. Sensors had an inter-electrode distance of 8 mm for the perilabial muscles and 16 mm for the submental area.

For experimental tasks, participants sat in a chair with back and foot support to facilitate a relaxed, supported posture (see Supplemental Material S1). They were familiarized with the equipment, and then researchers cleaned their skin with an alcohol wipe and tape (Stepp, 2012), to ensure maximum conductivity of electrodes. Following the methodology described in our prior work (Hahn Arkenberg et al., in revision; Kantarcigil et al., 2020), all participants completed criterion reference tasks that consisted of MVCs for each muscle group. For the perilabial muscles, participants completed three trials of an exaggerated lip pursing (Weber & Smith, 1987), and for the submental muscles, children completed three repetitions of maximal anterior lingual pressure (measured using the Iowa Oral Performance Instrument [IOPI; IOPI Medical]) to ensure maximum contraction (Kantarcigil et al., 2020). Then they completed the experimental tasks. These included three self-fed sips or bites of each of four consistencies and eight repetitions of four speech tasks (repeating prerecorded words and sentences; see Table 1). Participants were verbally instructed to swallow each sip or bite, and each word/sentence was played from prerecorded audio. For each domain (swallow and speech), the simplest tasks were small sips of water and repetition of two-syllable nonwords heavy in bilabial sounds, respectively; the next level of tasks were bites of pudding and repetition of four-syllable nonwords heavy in bilabial sounds, respectively; and the most complex tasks were bites from pretzel rods and sentence repetition also heavy in bilabial sounds (see Table 1 for details). These tasks were parallel between swallowing and speech domains, as they increased similarly and gradually in motor complexity (see Table 1).

Table 1.

Parallel experimental tasks.

Complexity	Swallow tasks	Speech tasks (heavy in bilabial sounds)
Level 1	5- and 10-ml thin liquid cup sip	2-syllable nonsense words (baba, mama)
Level 2	5 cc pudding	4-syllable nonsense words (babababa, mamamama)
Level 3	Large intact pretzel rod (7.5 in.)	Short sentences (Mom pats the puppy. Monkey drives the purple and yellow boat.)

Note. Adapted protocol from the work of Hahn Arkenberg et al. (under revision).

Surface EMG data were collected with a 2-kHz sampling rate and a 550-μV range in conjunction with four confirmatory measures (see Supplemental Material S1). Measures used to confirm swallowing (with sampling rate and ranges listed respectively in parentheses) were respiratory inductance plethysmography (2 kHz, 100 mV), accelerometry (on the thyroid notch; 2 kHz, 500 mV), and a manual user trigger (button push), and measures used for speech were audio recording (20 kHz, 20 mV) and the manual user trigger (Hahn Arkenberg et al., in revision). Measures were recorded simultaneously with PowerLab bio-amp system (PowerLab 8/35, ADInstruments, Inc.) and data acquisition software (LabChart 8).

Data Processing

Raw data, with maintained sampling rate and range, were exported to MATLAB R2017a (MATLAB Inc.). A custom-written MATLAB script was utilized to complete pre- and postprocessing EMG analyses (Smith et al., 1987, 1996; Stepp, 2012; Walsh & Smith, 2013). Artifact, spikes in signal amplitude that could not be attributed to muscular activity or variations in slope width time, was identified and removed automatically and reviewed by a trained experimenter to ensure accuracy (Kantarcigil et al., 2020; S. S. Mitchell, 2021). After artifact removal, the signal was bandpass filtered (between 20 and 300 Hz), notch filtered (60 Hz), and full wave rectified and smoothed with an 80-ms window (Hermens et al., 2000; Stepp, 2012).

Outcome Variables

sEMG outcome variables were selected to reflect a range of subcomponents of neuromuscular control including muscular effort, coordination, and temporal control. Each of these components gives insight into a particular element of neuromuscular control, and they may reveal differential patterns of refinement or stability across development.

Burst duration of sEMG activity is the total activation time and provides broad insight into the duration of activation (Crary et al., 2006). Time to peak amplitude indicates how quickly muscles reach their maximum level of activation (Crary & Baldwin, 1997; Crary et al., 2006). To assess coordination, we calculated bilateral synchrony of the left and right sides of each muscle pair (Walsh & Smith, 2013). Our final outcome measure was normalized mean amplitude (i.e., area under the curve [AUC]), which is calculated as a percentage of the MVC tasks. Normalized mean amplitude is not a direct measure of strength but is an indication of muscular effort and represents extent of muscular contraction (Oh, 2016).

Burst duration. To calculate burst duration of smoothed sEMG activity during tasks of interest, research assistants visually identified each event (swallow or speech) by examining the six sEMG channels one by one in conjunction with the simultaneously recorded confirmatory measures. The exact onsets and offsets of each event were defined by an automatic algorithm in the custom MATLAB script, which identified onsets and offsets as a change greater than 2 SDs from the baseline of the sEMG signal within the user-identified window (Kantarcigil et al., 2020; S. S. Mitchell, 2021) for each channel. If the algorithm did not automatically detect the onset or offset of an event, manual markers were placed by trained research assistants. Burst duration was then calculated by subtracting the offset from the onset time.

Normalized mean amplitude (AUC). Normalized mean amplitude was computed by the custom-written MATLAB script, which calculated the AUC between the previously defined events' onsets and offsets (Kantarcigil et al., 2020; Smith et al., 1987). The script also generated the AUC value for the three trials of maximal contraction. Then, the AUC of each event was normalized against the highest AUC value of the MVC tasks (exaggerated kiss for perilabial muscles, tongue press with IOPI for submental); that is, all reported mean amplitude values are a percentage of the MVC tasks to allow for comparison across time points and subjects (Smith et al., 1987; Weber & Smith, 1987).

Time to peak amplitude. Time to peak amplitude was calculated via the custom MATLAB script by recording the latency between the defined event onset and the peak smoothed sEMG signal amplitude occurring in the user-identified time window (Crary & Baldwin, 1997).

Bilateral synchrony. Bilateral synchrony was computed by calculating the pairwise zero-lag cross-correlation coefficients between muscle pairs (Walsh & Smith, 2013). We specifically investigated the cross-correlation of each left/right muscle pair (upper lip, lower lip, submental).

Statistical Analysis

Sample size was determined a priori; power analyses were based on preliminary data available through our prior work (normalized mean amplitude; Hahn Arkenberg et al., in revision). The standardized effect size (Cohen's d) for normalized mean amplitude in this preliminary work was 1.29; therefore, the minimum sample size needed with this effect size was revealed to be nine participants per group at an α < 0.05 with 0.8 power. Thus, the obtained sample size of 13 is more than adequate to test our hypotheses. For comparison of demographic variables, we used t tests with Bonferroni correction for multiple comparisons.

To explore our two aims, we used linear mixed-effects modeling using the MIXED procedure in SAS (SAS Institute), as these models allow us to assess main effects and interactions while accounting for multiple observations from each subject. Each outcome was log-transformed, as these data were left-skewed. For Aim 1, in order to examine differences between younger and older children, we examined each of the four sEMG outcome variables in a separate model, resulting in four models for swallowing and four models for speech tasks. Fixed effects were a between-subject factor age group (young and old) and within-subject factors of muscle location (superior perilabial, inferior perilabial, submental) and task (liquid, pudding, pretzel, two-syllable words, four-syllable words, sentences). We also considered a subject random effect and its interactions with the within-subject factors. We used Bayesian information criterion values to choose the best correlation structure, also allowing the variances to differ. We used similar mixed-effects models for Aim 2, comparing cross-domain relationships with separate models for each outcome variable. In order to avoid any differences due to cognitive load for swallowing and speech tasks, for the levels that included nonwords (speech Levels 1 and 2), we only included the simplest two- and four-syllable nonwords from each level in analysis: baba/mama, babababa/mamamama. Fixed effects in these models were task (swallowing and speech), complexity (Level 1, Level 2, Level 3) and age group (younger, older).

Results

Demographics

Thirteen younger (seven girls, six boys) and thirteen older (seven girls, six boys) children participated in this study. The younger children ranged from 7;0 (years;months) to 8;10 (M = 8.06, SD = 0.62), and older children ranged from 11;1 to 12;9 (M = 11.78, SD = 0.52). Data for 10 of these children (four younger, six older) were collected as part of our prior larger study on swallowing and speech control in CP and typical development (Hahn Arkenberg et al., in revision; Malandraki et al., 2022). Detailed demographic data of these children have been described in Malandraki et al. (2022), and a portion of their data have been separately analyzed in Hahn Arkenberg et al. (in revision). All analyses reported here are new and not previously reported. The remaining 16 participants were recruited solely for this study (aggregate demographic data in Table 2). There were no significant differences between groups in socioeconomic status (determined by mother's education), nonverbal intelligence, or language.

Table 2.

Participant demographic information.

Characteristic	Young	Older
Sex	7 female, 6 male	7 female, 6 male
Age range (years;months)	7;0–8;10	11;1–12;9
Race & ethnicity	10 White, not of Hispanic origin	11 White, not of Hispanic origin
	1 Black & White, not of Hispanic origin	1 Black & White, not of Hispanic origin
	1 Black, not of Hispanic origin	1 Asian, not of Hispanic origin
	1 Hispanic origin
TONI scores – Range (M)	89–128 (105.75)	97–130 (110.61)
CELF-5 RS – Range (Mean)	9–19 (12.4)	10–18 (12.92)
Mother's education (Mean rating)	5.42	6.27

Note. TONI = Test of Nonverbal Intelligence; CELF-5 RS = Clinical Evaluation of Language Fundamentals–Fifth Edition, Recalling Sentences (no significant differences).

Reliability

All sEMG data analysis was completed by one rater trained to > 85% accuracy with the lab's gold-standard sEMG rater. To ensure continued intra- and interrater reliability, 20% of data were reanalyzed by the primary rater and gold-standard rater. Reliability analysis for normalized mean amplitude (which encompasses timing and amplitude measures) of data revealed an intraclass correlation coefficient of .92 for intrarater reliability and .85 for interrater reliability.

Aim 1: Comparisons Between Younger and Older Children in Each Domain

Linear mixed-effects models were used to investigate if there was an interaction effect or main effect for age group (younger vs. older children). All tests were completed using degrees of freedom determined by the Kenward–Roger method (Kenward & Roger, 1997).

Swallowing

Burst duration. For burst duration of the smoothed sEMG signal during swallowing tasks, the linear mixed-effect model revealed a significant, F(4, 92) = 3.24, p < .016, three-way interaction between age group, muscle location, and task. As depicted in Figure 1, younger and older children demonstrated different patterns of burst duration of the smoothed sEMG signal during swallows across muscles and tasks. Overall, sEMG burst duration of swallows was shorter in older children for all tasks, but this varied in magnitude by muscle location and task. For example, the burst duration of the submental muscles (yellow) was similar in both age groups for liquids but on average differed by almost half a second between younger and older groups for the most complex task—solid (pretzel). There was also a significant two-way interaction between muscle location and task, F(4, 92) = 5.58, p < .0005, indicating generally shorter burst sEMG duration of the submental muscles' activity compared to perilabial muscles' activity (gray and blue), and this pattern was more consistent in younger children.

Figure 1.

Interaction plots showing burst duration of the smoothed surface electromyography (sEMG) signal during swallow tasks for the younger and older children.

Normalized mean amplitude (AUC). Our analyses for normalized mean amplitude of the smoothed sEMG signal revealed a significant two-way interaction between age group and muscle location, F(2, 46.4) = 6.94, p < .0023 (see Figure 2). In both groups, the lower lip muscles (gray) demonstrated higher normalized mean amplitude than the other muscle groups across tasks, but in younger children, the upper lip muscles (blue) exhibited higher amplitude than the submentals (yellow), whereas the opposite was true for older children. There was also a two-way interaction between muscle location and swallowing tasks, F(4, 90.1) = 7.48, p < .0001, with the submental muscle activity (yellow) steadily increasing in normalized mean amplitude across more complex swallowing tasks for both groups, whereas both groups demonstrated highest amplitude for the medium-complexity task (pudding) for the upper lip muscles (blue) and different patterns in younger and older children for the lower lips (blue).

Figure 2.

Interaction plots showing normalized mean amplitude (%MVC) of the smoothed sEMG signal during swallow tasks for the younger and older children. MVC = maximum voluntary contraction; sEMG = surface electromyography.

Time to peak amplitude. For time to peak amplitude during swallows, we also found a significant two-way interaction between age group and muscle location, F(2, 46) = 7.37, p < .0017 (see Figure 3). Specifically, older children demonstrated shorter time to peak amplitude compared to younger children for the upper lip muscles (blue) and longer time to peak amplitude for the lower lip (gray) and submental area muscles (yellow) during swallow tasks. Time to peak for the lower lip muscles was longest overall for both groups. There was also a significant two-way interaction between muscle location and swallow task, F(4, 92) = 9.05, p < .0001. Submental (yellow) and lower lip muscles (gray) displayed similar time to peak across tasks for both younger and older children (i.e., longer time to peak for the two more complex/Level 3 tasks than the Level 1 task and longest time to peak for the medium-complexity/Level 2 task). The upper lip muscles for the older children demonstrated the opposite pattern: decreasing time to peak amplitude in the more complex tasks.

Figure 3.

Interaction plots showing time to peak surface electromyography (sEMG) amplitude during swallow tasks for the younger and older children.

Bilateral synchrony. There were no significant age interactions or main effects for bilateral synchrony of swallowing tasks (see Supplemental Material S2).

Speech

Burst duration. For burst duration of the smoothed sEMG signal of the speech tasks, we also saw a significant three-way interaction between age group, muscle location, and tasks, F(4, 90.9) = 3.40, p < .012. As expected, and depicted in Figure 4, burst duration increased with increased task complexity for both age groups. Burst duration was generally shorter in older children than younger children, except for the submental muscle group (yellow line). We also observed a significant two-way interaction between muscle location and age group, F(2, 46.3) = 4.34, p < .019. Specifically, the submental muscles' burst durations (yellow) of younger children were the shortest across all speech tasks compared to the burst durations of the perilabials (blue and gray; see Figure 4) but the longest for older children.

Figure 4.

Interaction plots showing burst duration of the smoothed surface electromyography (sEMG) signal during speech tasks for the younger and older children.

Normalized mean amplitude. For normalized mean amplitude of the smoothed sEMG signal during speech tasks, there were no significant interactions with age group. Instead, there were strong main effects for location and task, as shown in Figure 5 (p < .0001).

Figure 5.

Interaction plots showing normalized mean amplitude (%MVC) of the smoothed sEMG signal during speech tasks for the younger and older children. MVC = maximum voluntary contraction; sEMG = surface electromyography.

Time to peak amplitude. For time to peak amplitude during speech tasks, there was a significant three-way interaction, F(4, 138) = 2.91, p < .024, between age group, muscle location, and task, followed by a significant two-way interaction between age and task, F(2, 138) = 4.27, p < .016. As shown in Figure 6, time to peak amplitude during speech tasks for the submental muscles (yellow) was shorter than time to peak amplitude during the same tasks for the perilabial muscles (gray and blue), with time to peak for the lower lip (gray) being longer than both. Also, time to peak amplitude for tasks of higher motor complexity was generally shorter than for tasks of lower motor complexity (i.e., sentences had shorter time to peak amplitude than words), and younger children needed less time than older children to reach their peak sEMG amplitude of their submental muscles but similar time for their perilabial muscles.

Figure 6.

Interaction plots showing time to reach peak surface electromyography (sEMG) amplitude during speech tasks for the younger and older children.

Bilateral synchrony. There were no significant age interactions or main effects for bilateral synchrony of speech tasks (see Supplemental Material S3).

Aim 2: Comparison Between Parallel Tasks of Swallowing and Speech—Relationship Between the Two Domains

Linear mixed-effects models were tested for three outcomes measures—normalized mean amplitude, time to peak, and bilateral synchrony—to examine complexity-based interactions or main effects between parallel tasks across the two domains (swallowing and speech). For the purpose of examining cross-system effects, age groups and muscle groups were all included in this analysis, with age and muscle groups as fixed factors. Burst duration was not examined because the duration of words and sentences is categorically substantially longer than that of single swallow events.

Normalized mean amplitude (AUC). Mixed-effects models of normalized mean amplitude values revealed an interaction between domain and complexity, F(2, 245) = 4.07, p < .0186 (see Figure 7). As shown in Figure 7, there was a larger difference in mean amplitude between swallowing and speech for the more complex tasks than for the simplest tasks. There was also a significant main effect for domain, F(1, 723) = 395.17, p < .0001, indicating that children demonstrated different normalized mean amplitude for swallowing versus speech tasks. Lastly, there was a significant main effect for complexity, F(2, 723) = 563.48, p < .0001, with increased normalized mean amplitude for more complex tasks across both domains. Though children demonstrated different normalized mean amplitude for tasks of speech and swallowing, they demonstrated a common pattern: increasing mean amplitude as complexity increased.

Figure 7.

Interaction plots depicting normalized mean surface electromyography (sEMG) amplitude, time to reach peak sEMG amplitude, and bilateral synchrony, for parallel swallowing and speech tasks at three levels of complexity, with standard error (see Table 1 for reminder of levels).

Time to peak amplitude. For time to peak amplitude, we also found a significant main effect for domain, F(1, 368) = 107.12, p < .0001, indicating that children demonstrated different time to reach their peak amplitudes for swallowing versus speech tasks across muscle groups. As shown in Figure 7, time to peak amplitude was longer for speech tasks than swallowing. There was also a significant main effect for complexity, F(2, 252) = 8.85, p < .0012, with increased time to peak amplitude for more complex tasks across both domains. Similar to the normalized mean amplitude results, though children required different times to reach their peak amplitude for swallowing and speech tasks, time to reach peak amplitude for both domains increased with increased complexity, revealing a common pattern.

Bilateral synchrony. No interactions or main effects for domain were found for bilateral synchrony, indicating that our sample displayed similar bilateral synchrony for swallowing and speech tasks, F(2, 299) = 1.18, p < .308. As shown in Figure 7, bilateral synchrony was relatively constant across tasks for both domains.

Discussion

In this study, we examined the peripheral neuromuscular development of swallowing and speech in a cross-sectional sample of typically developing children. Children in two age groups (7–8 yoa and 11–12 yoa) completed tasks of swallowing and speech during an sEMG paradigm. The goals were to determine (a) potential age differences in neuromuscular control of swallowing and speech separately between these two age groups and (b) shared and separate patterns of activation between parallel tasks of swallowing and speech. We hypothesized that (a) older children would demonstrate refinement of muscle activation (specifically exemplified by decreased burst sEMG duration, normalized mean amplitude, and time to peak amplitude and increased bilateral synchrony) compared to younger children and that (b) children of both age groups would demonstrate similar patterns of activation for the swallowing and speech tasks of parallel motor complexity, revealing aspects of shared neuromuscular control across domains.

Development of the Neuromuscular Control of Swallowing and Speech

Consistent with our initial hypothesis, and as shown by the multiple significant interactions we found, older children demonstrated some evidence for refinement of neuromuscular control for both swallowing and speech for most outcome measures examined. For swallowing, older children generally displayed shorter burst duration across muscle groups, possibly indicating higher orofacial neuromuscular efficiency than the younger group. Further, older children had longer duration of submental than perilabial muscle activity during swallows. As submental muscles are the primary muscles found to be involved in the pharyngeal triggering of the swallow (Palmer et al., 1999; Perlman et al., 1999), this may indicate higher specificity in the older group. This slightly varied by task and location, as also likely expected in normal development (Williams & Castro, 1997).

Further, though sEMG burst duration was descriptively shorter in older children for all tasks, the duration difference varied in magnitude by muscle location and task. For example, younger children's upper lip muscle activation was on average almost half a second longer than the same activation of older children for liquid swallows, but only 0.2 s longer for swallows of pretzel. Green et al. (1997) and Vaiman et al. (2004) both also found age differences in swallowing duration, that is, burst duration decreasing with age. Ruark et al. (2002), on the other hand, found no age differences in swallowing duration between 5-year-olds and young adults; however, they examined different muscles and food consistencies. Additionally, it is important to note that pace of developmental change likely differs in preschool compared to older children, such as those included here. We found that in these older children, swallowing burst duration differs by muscle location and food/liquid type; therefore, it is unsurprising that when using variable foods/liquids and measuring the activity of different muscles, findings from the previous literature may be mixed.

In addition, for both outcome variables of normalized mean amplitude (a measure of muscular effort) and time to peak amplitude (a measure of neuromuscular reaction time), our statistical models also revealed different patterns between the age groups. First, the submental muscles had higher normalized mean amplitude during older children's swallows than the upper lip muscles, whereas younger children demonstrated the opposite pattern. Similarly, the submental muscles of older children required more time to reach their peak amplitude than the upper lips. This may further indicate more specificity and maturity in the older group, given the contribution of the submental muscles to the pharyngeal swallow (Huckabee & Steele, 2006; Palmer et al., 1999; Perlman et al., 1999). Specification is a mark of neuromuscular refinement across the development of other body systems, which supports the protracted refinement hypothesis (Piek, 2002; Piek et al., 2008). However, the lower lip showed higher amplitude and longer time to peak amplitude than all other muscle groups for both age groups. Although this was not expected, it could be related more to methodological considerations, such as slight variation in the adherence of sensors to the different muscle locations. Overall, for swallowing, our results support the protracted refinement hypothesis for three of four outcome measures.

Similarly, for speech, we found evidence of protracted refinement across the two age groups. Older children demonstrated shorter duration than younger children, more complex speech tasks generally required longer burst duration, and these patterns varied slightly by muscle location. Prior work has shown that kinematic duration during speech is longer in younger children (e.g., Goffman & Smith, 1999; Green et al., 2000; Sharkey & Folkins, 1985; Smith & Goffman, 1998). We extended this work by further showing that task and muscle location have a significant impact on the timing and extent of muscle activation. For example, children, in general, demonstrated longer duration for more complex utterances, but this was not a linear relationship for both ages. Younger children consistently demonstrated longer labial activation, whereas older children demonstrated similar activation across muscle groups, particularly at the simplest level. Increased labial activation in younger children could be related to the importance of these muscles for labial consonants, which the younger children may have emphasized as part of developing motor control. It is established that articulatory movement variability decreases longitudinally (Grigos, 2009), and it is possible that we observed more labial activation in younger children as a byproduct of this variability.

For time to peak amplitude during the speech tasks, across muscle groups, we found significant differences (interactions and main effects) in our two age groups, with older children requiring shorter or similar time to peak amplitude for the same tasks. This was particularly evident in sentences, the most complex speech task, where older children required about 20% less time to reach peak amplitude than younger counterparts. This is consistent with prior kinematic work showing that older children require less muscular effort and time for speech compared to younger children (Smith & Goffman, 1998; Smith & Zelaznik, 2004; Walsh & Smith, 2002). It is additionally consistent with theoretical models of development proposing continued refinement of all body systems throughout childhood (Kunnen & van Geert, 2012; Goodway et al., 2019; Thelen, 2005).

Finally, we did not find significant age interactions or main effects for bilateral synchrony during swallow or speech tasks. Bilateral synchrony is a measure of coordination between the left and right sides of the measured muscles (Oh, 2016). This finding indicates that, in our sample, 7- to 8-year-old children and 11- to 12-year-old children displayed similar coordination. This further validates our prior research in this area, which indicated that bilateral synchrony of both swallowing and speech is well developed by school age and even in children with unilateral CP (Hahn Arkenberg et al., in revision). Further, our findings are consistent with the refinement of coordination of chewing in the early years (Green et al., 1997; Simione et al., 2018), with the speech literature that reports relatively early maturation of coordination (Moore & Ruark, 1996; Ruark & Moore, 1997), and an observed plateau of coordination in the motor development of school-age children (Smith & Zelaznik, 2004). The period of 7–12 yoa is a period of significant growth of craniofacial structures that impact both swallowing and speech, including oral, pharyngeal, and laryngeal structures (Vorperian et al., 2005, 2009). In Smith and Zelaznik's study of developmental change in speech motor variability, they concluded that the observed plateau in coordination was likely related to neural adaptation to this changing anatomical and neural landscape (Smith & Zelaznik, 2004). This could also explain the similarities in bilateral coordination we observed in the children in our study between 7 and 12 yoa.

The Relationship Between Swallowing and Speech

The second aim of this study was to systematically investigate the shared and separate neuromuscular control of swallowing and speech across typical development. Swallowing and speech share anatomic structures and some overlapping neurophysiology (Babaei et al., 2013; Eickhoff et al., 2009; Malandraki et al., 2011; Price, 2012). Investigating similarities and differences in the control of these two systems is clinically relevant because it is still unclear why swallowing and speech deficits may co-occur in children, which limits the potential for developing cross-system interventions.

Cross-system or multimodal interventions have the potential to maximize efficacy and efficiency and are already being proposed and investigated for gross motor and cognitive systems, with positive results (Boyd et al., 2013; James et al., 2015; Reid et al., 2015). Understanding the typical shared and separate control mechanisms for swallowing and speech could inform these potential cross-system interventions in the future. Based on our prior research (Hahn Arkenberg et al., in revision), we hypothesized that there would be similarities in underlying muscle activation for tasks of swallowing and speech in typically developing children, which would indicate some level of shared neuromuscular control between these two domains. This hypothesis was partially supported, as there were both shared and separate patterns of peripheral neuromuscular control for swallowing and speech.

First, children demonstrated similar high temporal bilateral coordination of muscle activity between the two domains. These findings are not surprising given the partially shared underlying bilateral muscular and central innervation of both systems (Babaei et al., 2013; Eickhoff et al., 2009; Malandraki et al., 2011; Price, 2012). However, they are in contrast to the task-specific theory of development explored for these two domains in the 1990s and 2000s, which indicated separate control mechanisms for swallowing and speech (Moore & Ruark, 1996; Ruark & Moore, 1997; Steeve & Moore, 2009; Steeve et al., 2008). It has to be noted that these earlier studies focused on infants and toddlers (9–24 months), and they compared speech tasks with tasks that are relevant to feeding more broadly than swallowing (i.e., chewing and lip protrusion). The specific tasks compared in earlier work were as follows: babbling and chewing at 9–22 months (Steeve & Moore, 2009); syllable repetition, chewing, lip protrusion, and speech at 2 years (Ruark & Moore, 1997); sucking, chewing, and babble at 9 months (Steeve et al., 2008); and chewing, suckling, babbling, and speech at 15 months (Moore & Ruark, 1996). Our study instead focused on neuromuscular activation at the moment of the swallow and included school-age children. It is possible that neuromuscular coordination of swallowing and speech develop independently at the beginning of life and converge later in development. Alternatively, or in addition, our experimental design (tasks specifically designed to increase in motor complexity gradually and including actual swallows and speech events) provided more robust comparisons between domains than previous studies.

Measures of time to peak amplitude and normalized mean amplitude revealed different magnitudes for tasks of swallowing and speech but similar patterns. The differences in magnitude are consistent with previous research (Ruark & Moore, 1997, Ruark et al., 2002; Steeve & Moore, 2009); however, the similar pattern is a new finding. Normalized mean amplitude was similar for swallowing and speech at the simplest level, but this diverged as tasks became complex. Children took longer to reach peak amplitude for speech tasks, and their time to peak amplitude increased as complexity increased in both domains. Together, these indicate that children use largely task-dependent times to peak activation and muscular effort, but across both domains, children increase muscular effort and time to peak activation relative to task complexity.

To summarize, in contrast to previous research, which hypothesized that swallowing and speech are controlled by largely separate mechanisms (Green et al., 1997, p. 12; Green et al., 2000; Ruark & Moore, 1997; Steeve & Moore, 2009), our carefully designed methods and tasks allowed us to demonstrate that though there are some significant differences in neuromuscular control between these two tasks, there indeed are some overlapping patterns of activation, specifically in coordination and parallel changes with increased task complexity. Further longitudinal study of neuromuscular development of swallowing and speech could provide further clarification on the pathway of development, along with shared and separate elements of control, and how those may converge or diverge throughout time. This approach sets the stage for future basic and clinical research on cross-system interactions.

This initial evidence of some shared patterns between the peripheral neuromuscular control of swallowing and speech in tasks of similar complexity adds to the theoretical rationale for the exploration of the cross-system impact of interventions in each of these two domains and the potential for future development of multimodal interventions. An expert review published in Nature Reviews: Neurology proposed that the next frontier in improving therapies will be the use of multimodal intervention (Reid et al., 2015). They posited that there was strong neurobiological rationale for multimodal interventions and referenced several recent studies (Boyd et al., 2013; James et al., 2015; L. Mitchell et al., 2012) documenting increased positive change in therapies targeting more than one domain (i.e., cognition in combination with motor skills). There remain multiple questions about how this could be developed and implemented for speech and swallowing, but this early developmental work also supports that future research should consider cross-system function and multimodal possibilities for swallowing and speech as well.

This study had some limitations, including a relatively small sample size of limited racial and ethnic diversity. Our sample was cross-sectional in nature, which introduces confounding elements, such as the influence of genetics and differing environments, that could have influenced our interpretation of the results. Though we ensured that our groups were matched on age and sex, age of onset of puberty is variable (Livadas & Chrousos, 2016), and it is possible that there were pubertal influences on vocal tract development in our older group that could have influenced our conclusions. This should be further considered in future research, with particular attention to laryngeal development. Additionally, children were speaking and swallowing with multiple small sensors on their face, which could have led to slight alterations from their typical eating and speaking patterns. Finally, this study was conducted with the assumption that the swallowing and speech tasks were parallel in complexity, based on previous work (Hahn Arkenberg et al., in revision). Exactly how parallel these tasks are should be further studied in future work. Despite these limitations, this study was adequately powered to detect differences between these two groups of children and provides valuable information to direct future efforts.

Conclusions

This cross-sectional study of 13 younger and 13 older school-age children enabled comparison of the neuromuscular control of swallowing and speech between these age groups. We observed that older children demonstrate refinement in neuromuscular activation for both functions. Also, we found evidence that the neuromuscular control of these two systems has both common and separate neuromuscular underpinnings. Future research should include longitudinal analysis of swallowing and speech development, as well as measures of central neurophysiology, to further elucidate the relationship between these two vital functions.

Data Availability Statement

The data are available on request from the corresponding author. De-identified data will be available to public repository upon completion of the larger NIH study.

Funding Statement

Research reported here was partially supported by the National Institute on Deafness and Other Communication Disorders (Award R21DC015867; principal investigator [PI]: Georgia A. Malandraki), by the National Institutes of Health (NIH) T32 Training Grant 2T32DC000030-31, and by the Indiana Lions Speech & Hearing Grant (PI: Rachel E. Hahn Arkenberg).