Phonation Demonstrates Goal Dependence Under Unique Vocal Intensity and Aerobic Workload Conditions

Abstract

Purpose

This study investigated whether metabolic respiratory requirements (treadmill workload) affected glottal valving in phonation, based on aerodynamic measures, when a sound pressure level (vocal SPL) is dictated as a target goal. Consistent with a theory of action, we hypothesized that adjustments in glottal valving as measured by laryngeal airway resistance would be dependent upon vocal SPL level, even as workload increased, and loud vocal SPL would interfere more with respiratory homeostasis than spontaneous vocal SPL.

Method

Thirty-two women enrolled who were ages 18–35 years. A repeated-measures design was used with random assignment of workload and vocal SPL conditions. Aerodynamic and acoustic data were collected during phonation, as were gas volume and concentration data. Analyses were performed with generalized estimating equations.

Results

Laryngeal airway resistance at a low workload significantly increased when vocal SPL changed from spontaneous to loud. At a loud vocal SPL, laryngeal airway resistance decreased when workload changed from rest to either low or high. Regarding the respiratory system response, minute ventilation increased at a loud vocal SPL when workload changed from rest to either low or high. End-tidal CO₂ increased under low and high workloads relative to rest at loud and spontaneous vocal SPLs.

Conclusions

Mostly consistent with a theory of action, in which motor control is goal dependent (i.e., vocal SPL targets), speakers can achieve a loud vocal SPL despite increases in workload requirements. In contrast, laryngeal airway resistance stays relatively low when vocal SPL occurs spontaneously, suggesting glottal adjustments are made to improve gas exchange as metabolic respiratory requirements become prioritized. Metabolic respiratory requirements appear to be overcome by the overlay of motor control for voicing when a loud vocal SPL is targeted. The implication of goal-dependent phonation for clinicians is that real-world conditions (i.e., loud vocal SPL) matter in vocal testing and voice therapy.

The respiratory and laryngeal systems work together during voice production to regulate vocal intensity (vocal sound pressure level [SPL]; Finnegan, Luschei, & Hoffman, 2000; Hixon, Goldman, & Mead, 1973; Hixon, Mead, & Goldman, 1976; Stathopoulos & Sapienza, 1993). The two systems also interact during respiration to control blood gas concentrations and maintain homeostasis (Brancatisano, Collett, & Engel, 1983; Brancatisano, Dodd, & Engel, 1991; Butler, 2007; England & Bartlett, 1982). The laryngeal system, specifically the vocal folds, adducts in voice production as the respiratory system increases lung pressure to drive air through the glottis and entrain the vocal folds into self-sustained vibration (Titze, 1988, 2002). On the other hand, the vocal folds abduct to decrease airflow resistance and accommodate greater volumes of exhaled air as ventilation increases to maintain respiratory homeostasis. Thus, opposing laryngeal configurations and different physiologic strategies by the respiratory and laryngeal systems exist when breathing as during phonation—with closed vocal folds—versus breathing for purposes of metabolic homeostasis—with open vocal folds.

It is unknown whether the same physiologic strategies apply during conditions of increased ventilation (high respiratory drive) when the vocal folds abduct for increased ventilation (hyperpnea) while simultaneously approximating for voice production, for example, in those who provide vocal instruction during aerobic exercise and similar occupations. Specifically, does intentional communication with preferred vocal SPL targets in high–respiratory drive conditions result in vocal fold adjustments that allow for increases in airflow as metabolic needs increase, favoring those needs while sacrificing vocal output (Bailey & Hoit, 2002; Bunn & Mead, 1971; Doust & Patrick, 1981; Hoit, Lansing, & Perona, 2007; Otis & Clark, 1968)? This question is important because of the implications for glottal valving and the risk for vocal injury with high impact stress from tight vocal fold adduction and high subglottal pressures as often occurs with loud vocal SPL (Grillo & Verdolini, 2008; Titze, 2000; Verdolini, Druker, Palmer, & Samawi, 1998). On the other hand, the risk for speaking-related dyspnea is relevant for patients with pulmonary conditions who experience shortness of breath when talking (Lee, Friesen, Lambert, & Loudon, 1998) due to the obstructive effect of approximated vocal folds on respiration with voicing (Russell, Cerny, & Stathopoulos, 1998).

A range of studies have investigated the synergistic relationship between the mechanically coupled respiratory and laryngeal systems during voice production using laryngeal airway resistance (R_law). R_law is the ratio of subglottal pressure (P_s) to translaryngeal airflow (U) during phonation (Baken & Orlikoff, 2000). A noninvasive technique exists for estimating R_law during the production of /pa/ consonant–vowel combinations (Holmberg, Hillman, & Perkell, 1988; Holmberg, Hillman, Perkell, & Gress, 1994; Leeper & Graves, 1984; Smitheran & Hixon, 1981). In broad terms, R_law increases as vocal SPL increases in resting breathing conditions (Holmberg et al., 1988; Leeper & Graves, 1984; Wilson & Leeper, 1992). Increases in R_law with greater vocal SPL come primarily from higher P_s. Breathy voice quality causes R_law to decrease, which is primarily from an increase in U (Holmberg, 1980; Holmberg, Hillman, & Perkell, 1989). Airway resistance provided by the larynx during phonation has an obstructive effect on air expulsion as reflected in significantly higher CO₂ partial pressures in voicing than in silent breathing (Gillespie, Slivka, Atwood, & Verdolini Abbott, 2015). A related measure, phonation threshold pressure (i.e., the smallest subglottal pressure required for self-sustained vocal fold vibration), significantly increases after cycling (Sandage, Connor, & Pascoe, 2013). Likely due to increased respiratory rate with exercise (McArdle, Katch, & Katch, 2010), such changes occur in the setting of a meaningful decrease in pharyngeal temperature that can negatively impact muscle contractive properties of the vocal folds (Sandage et al., 2013). Changes in phonation threshold pressure may also reflect an increase in vocal fold viscosity from vocal fold tissue desiccation as occurs with oral breathing (Sivasankar & Erickson, 2009).

In addition to changes in R_law, other voice changes may be relevant to our concerns about vocal motor control under elevated levels of respiratory drive. As compared to normal respiratory drive conditions, voice is perceived as louder, breathier, and higher in pitch under high–respiratory drive conditions (Bailey & Hoit, 2002; Otis & Clark, 1968) that result in increases in ventilation from resting levels to tightly match the body's metabolic needs (i.e., O₂ and CO₂ partial pressures) to exercise demands (Butler, 2007; McArdle et al., 2010). Researchers have argued that metabolic needs begin to override linguistic demands as the intensity of high respiratory drive increases, as evidenced by changes in speech phrasing (Bailey & Hoit, 2002; Baker, Hipp, & Alessio, 2008; Otis & Clark, 1968) and difficulty with speaking tasks (Rotstein, Meckel, & Inbar, 2004). It has been reported that individuals provide a greater number of statements about increased concentration (i.e., mental effort) to breathe at higher intensities of high respiratory drive (Hoit et al., 2007) that may reflect a vulnerability of behavioral vocal motor control to the needs of automatic respiratory motor control. Taken together, findings to date suggest the voice reacts to—rather than acts in—conditions of high respiratory drive when constraints on vocal output are not in place (e.g., vocal SPL).

Action theory, which provides a framework for understanding possible competition between respiration and phonation, states that physiology of activity depends on functional goals (Reed, 1982). Under different initial conditions (e.g., different treadmill workloads), context-dependent tuning of coordinative structures such as the respiratory–laryngeal system takes place in attainment of an action goal (e.g., specified vocal SPL). The effect of a perturbation to a system (e.g., increased ventilation with exercise) should be negligible in terms of goal accomplishment, while other components adjust (Fowler, Rubin, Remez, & Turvey, 1980). This motor equivalence is possible because of heterarchical neural control that allows for goal attainment under varying, even extreme, initial conditions (e.g., high treadmill workload; Jürgens, 2009).

The intersection of learned phonation mediated by higher cortical pathways and automatic breathing at the medullary reticular formation in the brainstem sets the stage for competition between phonatory and respiratory functions (Blitzer, Brin, & Ramig, 2009). Higher brain centers during special acts of breathing (e.g., voice) transmit signals to the medulla that influence respiratory activity. The integration of information from cortical areas permits an interruption of the normal rhythmic breathing pattern generation. If this dual control system does in fact exist, final integration of neuronal signals would occur at the level of the motor neuron. The timing and strength of the final signal reaching respiratory muscle motor neurons would ultimately depend on brainstem (i.e., medullary) and higher cortical inputs as well as other input via networks of interneurons (Butler, 2007). However, up to this date, research that has addressed voice under conditions of high respiratory drive has only done so under conditions in which the voice was free to vary, paying too little attention to glottal valving under real-world conditions that require loud vocal SPL.

Objective

The objective of this study was to investigate whether metabolic respiratory requirements affect glottal valving in phonation, based on aerodynamic measures, when a specific vocal SPL is dictated as a target goal and coupled with increased respiratory demands. We hypothesized that adjustments in R_law would be dependent upon vocal SPL target, reflecting deference to a phonatory goal (SPL) in a constrained voice condition (i.e., loud vocal SPL), even as workload increased, and would interfere with respiratory homeostasis (minute ventilation [Ve] and end-tidal carbon dioxide [P_ETCO₂]) through an obstructive effect. Such findings would be consistent with a theory of action (see Figure 1).

Figure 1.

Conceptual schematic of goal-dependent model of motor control for voice. VO₂ = oxygen consumption; SPL = sound pressure level; R_law = laryngeal airway resistance; Ve = minute ventilation; P_ETCO₂ = end-tidal carbon dioxide.

Method

Study Design

This research utilized a prospective, repeated-measures experimental design with random assignment of counterbalanced sequences of conditions to minimize order effects. Reporting of this study was consistent with the Strengthening of Reporting of Observational Studies in Epidemiology statement (von Elm et al., 2008).

Setting

Participant recruitment and data collection sessions across all participants occurred over a period of 2 months, and all sessions for a given participant were completed within a 3-week time frame. Screening and preexperimental training occurred at the University of Pittsburgh Voice Center. The subsequent experimental protocol was completed at the University of Pittsburgh Physical Activity and Weight Management Research Center where environmental conditions (i.e., relative humidity and temperature) met recommended standards for fitness facilities (American College of Sports Medicine, 2012). Participants were instructed to maintain normal fluid and dietary intake prior to all sessions and to wear similar clothing suitable for exercise to all sessions. To allow for metabolic recovery, participants returned no less than 24 hr after a previous session.

Participants

Research activities were approved by the University of Pittsburgh Institutional Review Board (IRB# 12030732). Participants were ages 18–35 years and recruited from the Pittsburgh metropolitan region using institutional review board–approved flyers distributed in physical and electronic formats. Based on power analyses described shortly, 32 women completed the full experimental protocol. A small group of men (n = 4) was also included to gather data on potential gender effects (Harms, 2006; Hunter, Tanner, & Smith, 2011; Sheel & Guenette, 2008); their data were explored descriptively only and are presented in Tables 2 and 3. After passing an online screening, participants were invited for an in-clinic screening if determined to have a perceptually normal vocal quality over the telephone by a speech-language pathologist.

Table 2.

Participant physical and voice characteristics.

Characteristic	Female (n = 32)	Male (n = 4)
	M (SD)	M (SD)
Age (years)	23.0 (3.4)	24.8 (3.4)
Body weight (kg)	59.8 (7.4)	69.5 (13.0)
Body height (cm)	165.5 (6.5)	178.4 (11.0)
BMI	21.8 (2.2)	21.8 (1.7)
Heart rate (bpm)	69.4 (9.2)	65.3 (8.6)
Systolic blood pressure (mmHg)	99.6 (7.4)	110.8 (5.1)
Diastolic blood pressure (mmHg)	64.9 (5.4)	66.5 (6.4)
FVC a (L), % predicted	4.0 (0.5), 98%	5.3 (1.2), 95%
FEV1 a (L), % predicted	3.4 (0.3), 97%	4.4 (0.8), 95%
CAPE-V rating (mm)	3.4 (2.3)	2.5 (1.9)
VHI-10 score	1.9 (2.0)	3.5 (3.3)

Note. BMI = body mass index; FVC = forced vital capacity; FEV1 = forced expiratory volume in the first second; CAPE-V = Consensus Auditory–Perceptual Evaluation of Voice; VHI-10 Voice Handicap Index-10.

^a % Predicted based on normative data from Crapo et al. (1981).

Table 3.

Unadjusted group means (standard deviations) of phonatory and respiratory measures for women and men.

Condition by sex	Primary phonatory	Secondary phonatory		Primary respiratory		Covariates
	Rlaw (cmH2O/[L/s])	Ps (cmH2O)	U (L/s)	Ve (L/min)	PETCO2 (mm Hg)	f0 (Hz)	SPL (dB)
	M (SD)	M (SD)	M (SD)	M (SD)	M (SD)	M (SD)	M (SD)
Women
Rest–standing
Quiet/silent				11.9 (1.7)	32.1 (2.5)
Spontaneous	62.6 (20.3)	9.5 (2.3)	0.16 (0.05)	9.7 (2.9)	34.0 (2.2)	217.9 (21.0)	86.5 (3.3)
Loud	120.5 (45.8)	16.5 (4.2)	0.15 (0.05)	10.6 (2.7)	33.1 (2.0)	240.7 (23.9)	95.1 (2.7)
Low workload
Quiet/silent				36.1 (6.8)	43.7 (3.0)
Spontaneous	69.8 (33.3)	12.0 (3.4)	0.20 (0.07)	15.3 (4.5)	54.5 (4.5)	229.4 (29.2)	91.3 (4.4)
Loud	104.1 (40.5)	17.0 (4.1)	0.18 (0.06)	16.4 (8.7)	53.9 (4.1)	255.2 (30.1)	97.3 (3.6)
High workload
Quiet/silent				46.6 (8.9)	45.2 (3.7)
Spontaneous	69.3 (28.6)	12.7 (3.4)	0.21 (0.07)	18.0 (4.0)	59.5 (5.3)	237.2 (35.2)	91.6 (3.9)
Loud	92.7 (42.4)	16.8 (5.6)	0.20 (0.07)	19.3 (13.7)	59.1 (4.5)	265.4 (36.0)	97.3 (2.7)
Men
Rest–standing
Quiet/silent				12.7 (1.4)	34.9 (1.5)
Spontaneous	40.6 (15.6)	7.2 (0.8)	0.19 (0.05)	8.9 (3.1)	36.7 (2.2)	122.9 (14.5)	94.0 (3.0)
Loud	53.8 (25.7)	14.5 (5.9)	0.28 (0.02)	8.5 (0.5)	34.9 (2.4)	162.5 (17.2)	100.8 (1.4)
Low workload
Quiet/silent				43.0 (11.1)	47.1 (4.3)
Spontaneous	28.0 (9.3)	8.6 (1.0)	0.33 (0.10)	12.2 (3.2)	58.3 (6.3)	136.3 (26.5)	97.6 (1.8)
Loud	53.4 (25.0)	18.4 (4.8)	0.37 (0.09)	13.4 (4.5)	56.7 (5.0)	172.3 (22.3)	102.5 (2.1)
High workload
Quiet/silent				57.8 (12.4)	52.2 (6.8)
Spontaneous	24.6 (0.5)	9.8 (3.2)	0.33 (0.01)	16.8 (3.1)	62.9 (4.1)	129.7 (13.6)	97.4 (3.5)
Loud	45.3 (32.2)	15.8 (6.3)	0.41 (0.14)	22.4 (7.7)	60.6 (4.6)	193.6 (39.3)	106.2 (0.6)

Note. R_law = laryngeal airway resistance; P_s = estimated subglottal pressure; U = translaryngeal airflow; Ve = minute ventilation; P_ETCO₂ = end-tidal carbon dioxide; f0 = fundamental frequency; SPL = sound pressure level.

Eligibility criteria included the following: (a) by self-report, recreationally active (90–150 min per week in aerobic exercise or sports activity); negative history of voice problems, voice treatment, or more than 1 year of voice training with the exception of high school choir; negative history of smoking in the past 5 years; not currently pregnant or within the last 6 months (confirmed by negative result on rapid urine test conducted at the screening visit); negative history of medication use during the month prior to participation, with the exception of medications that regulate menstrual cycle; regular monthly menstruation as of the most recent cycle or amenorrhea due to menstrual-regulating medications at the time of data collection; and no contraindications to participate in exercise as outlined by the American College of Sports Medicine (2009); (b) by clinician judgment, sufficient English comprehension to provide consent and follow instructions, ability to match and sustain a pitch (men: C₃ [130 Hz] ± 1 semitone [ST]; female: A₃ [220 Hz] ± 1 ST) for 3 s when given a reference tone, and free of signs (or symptoms, per self-report) of active allergies or upper respiratory infection on days of participation; (c) no indication of voice handicap (score of > 11 on the Voice Handicap Index-10; Arffa, Krishna, Gartner-Schmidt, & Rosen, 2012; Rosen, Lee, Osborne, Zullo, & Murry, 2004); and (d) normal voice as judged by a speech-language pathologist, with an overall severity score ≤ 10 mm on the Consensus Auditory Perceptual Evaluation of Voice (Kempster, Gerratt, Verdolini Abbott, Barkmeier-Kraemer, & Hillman, 2009); normal hearing with pure-tone audiometry conducted by a speech-language pathologist using a Maico Diagnostics Audiometer at 30 dB at 250 and 500 Hz as well as 1, 2, 4, and 8 kHz in the better ear; normal larynx as judged by a fellowship-trained laryngologist during flexible laryngeal endoscopy with stroboscopy after administration of both lidocaine and neosynephrine; normal body mass index (18.5–24.9 kg/m² [National Heart, Lung and Blood Institute, Obesity Education Initiative, Expert Panel, 1998] ± 1.0 kg/m²); normal resting blood pressure (90/60–120/80 mmHg; Chobanian et al., 2003); normal resting heart rate (50–80 beats per minute [bpm]; American College of Sports Medicine, 2009);and normal pulmonary function (forced vital capacity and forced expiratory volume in 1 s [FVCand FEV1, respectively] values at 80% or above predicted for age, height, and gender; Crapo, Morris, & Gardner, 1981) as assessed with spirometry by a speech-language pathologist (Miller et al., 2005) using a Koko spirometer (Grace Medical). Participants who met the criteria were enrolled into the full experimental protocol.

Study Size

An a priori power analysis indicated 30 participants was required to achieve 80% statistical power using a repeated-measures design with an alpha of .05 and an anticipated moderate effect size for R_law and Ve. Counterbalancing of vocal SPL and workload conditions described shortly required the total participant number be a multiple of 4, and therefore, 32 female participants completed the protocol. Four men also participated, and a tabular summary of their data is presented separately from the data of female participants.

Procedure and Equipment

Screening and Exercise Testing

Volunteers satisfying the criteria were enrolled. Participants engaged in preexperimental procedures to determine low and high treadmill workloads for achieving target heart rates of the systematically manipulated experimental treadmill workloads, with the exception of the rest condition. Experimental procedures required exercise on a Biodex Gait Trainer 2 motorized treadmill (Biodex Medical Systems, Inc.) using a ramping protocol as heart rate was continuously monitored with a Polar T31 Transmitter Heart Rate Sensor (Polar Electro, Inc.). The protocol involved a systematic increase in treadmill grade and speed following every minute of testing, until participants achieved an age-predicted heart rate for the high-workload condition as determined by the Karvonen formula (target heart rate [HR] = [max HR, i.e., 220 – age] – resting HR * 0.70] + resting HR; Kelly, 2006).

Training of Voice Task

Instrumental setup utilized the Phonatory Aerodynamic System 6600 (PAS 6600), and calibration was carried out according to the manufacturer's instructions (KayPENTAX, 2009). Participants stood on a treadmill and held the PAS 6600 mask unit to their face to cover the mouth and nose while two straps from a neoprene mask secured the PAS 6600 in place. Participants maintained their grip on the PAS 6600 handles throughout testing. Before commencing R_law training, participants' spontaneous vocal SPL and fundamental frequency (f0) were determined using well-vetted procedures (Russell et al., 1998) while reading the Rainbow Passage (Fairbanks, 1960) for 5 min having been instructed to “speak normally.” The middle 3 min of the passage were analyzed for spontaneous vocal SPL and f0.

Then, participants learned procedures for R_law data collection. Training required participants to produce a string of seven repeated /pa:/ syllables at a rate of 1.5 syllables per second in both spontaneous and loud voices until produced consistently at an individually determined f0 (± 1 ST) as established by a certified voice-specialized clinician's auditory–perception and PAS 6600 output. A reference pitch was provided with a pitch pipe, and a Matrix MR500 Metronome (manufactured in South Korea) maintained a recurring sound at 88 bpm. Participants who had difficulty matching the reference pitch received vocal models of the pitch in the appropriate octave until they successfully achieved the reference pitch.

The target vocal intensity for loud vocal SPL was the average spontaneous vocal SPL derived from reading plus 10 dB (± 2 dB). This loud vocal SPL target ensured an adequate change of ventilatory stress (Russell et al., 1998) and provided a distinctly different loudness compared to spontaneous vocal SPL (Hoffman-Ruddy, Lehman, Crandell, Ingram, & Sapienza, 2001). Most participants demonstrated loud vocal SPL without difficulty; for some, coaching was provided with hand signals. Training ceased after both principal investigator and participant expressed confidence in correct execution of the R_law task.

Equipment and Setup

Instrumental setup (see Figure 2) utilized components of the PAS 6600 and the CareFusion Encore Metabolic Cart (CareFusion Corp). The PAS 6600 was used for measurement of intraoral pressure and airflow rate, as well as recording the voice signal. PAS 6600 hardware consists of a pneumotach coupled to a facemask with an integral intraoral pressure tube and an external microphone. The external microphone on the PAS 6600 was repositioned to accommodate the serial attachment of the CareFusion Encore mass flow sensor behind the PAS 6600 pneumotach. The new position of the external microphone behind the CareFusion Encore mass flow sensor placed it further from the mouth, requiring manual calibration before data collection.

Figure 2.

Diagram of setup for experimental data collection. IOP = intra-oral pressure.

Calibration of the PAS 6600 pneumotach was carried out by depressing a 1.0-L syringe for 2–4 s (proper value: 1.0 L ± 1%–2%). The PAS 6600 system automatically calibrates intraoral pressure and was completed before data collection for each experimental condition for each participant. The repositioned external microphone was manually calibrated using published procedures before each testing session since the distance from the mouth increased with the new microphone location (Russell et al., 1998). To calibrate, the PAS 6600 was attached to a Styrofoam display head that housed a speaker within the mouth. A 200-Hz tone played for 30 s while the PAS 6600 microphone recorded. At the same level as the external microphone, a Larson Davis System 824 Sound Level Meter (Larson Davis Inc.) set to slow setting and C weighting collected the 200-Hz tone's SPL in dB to provide a correction factor. This correction factor was added to the SPL value calculated by the PAS 6600 at data analysis. The CareFusion Encore mass flow sensor was calibrated by depressing a 3.0-L syringe for 2–4 s (proper value: 3.0 L ± 1%–2%). Gas concentrations were calibrated prior to each participant session using two-span gases: (a) 26% O₂, 0% CO₂, and the balance in nitrogen and (b) 16% O₂, 0% CO₂, and the balance in nitrogen.

Experimental Procedure

An overview of experimental procedures is shown in Table 1. First, participants stood on a SensorMedics 2000 Treadmill (SensorMedics Corp.) wearing a heart rate sensor and breathed quietly into the PAS 6600 for 5 min of tidal breathing. Concerns are negligible for rising CO₂ levels with the use of a face mask over the duration of the experimental protocol (Huber, Stathopoulos, Bormann, & Johnson, 1998). Facemask still in place, participants produced a seven-syllable /pa:/ sequence at rest while standing, using the predetermined f0 and spontaneous vocal SPL and then the predetermined f0 and loud vocal SPL (+ 10 dB above spontaneous SPL, ± 2 dB). The instruction for the non–goal-oriented spontaneous vocal SPL condition was, “Speak three sets of seven syllables of /pa:/ with your typical voice until instructed to stop at the following pitch and speaking rate.” The instruction for the goal-oriented loud vocal SPL condition was, “Speak three sets of seven syllables of /pa:/ continuously, with a loud and clear voice until told to stop at the following pitch and speaking rate.”

Table 1.

Template for experimental procedures.

Condition	Activity	Time (min)
First experimental session
Rest	Breathe quietly with no voicing (rest no-voice baseline)	5
	Produce a consonant–vowel syllable string (spontaneous [spon] voice baseline)	½
	Breathe quietly with no voicing to return to rest baseline	1
	Produce a consonant–vowel syllable string (loud voice)	½
Rest-to-work transition	Walk on treadmill to achieve a low workload	3
50% of age-predicted maximal heart rate (low)	Breathe quietly with no voicing at a low workload (50% no voice)	1
	Produce a consonant–vowel syllable string (spon 50%)	½
	Breathe quietly with no voicing to return to 50% baseline	1
	Produce a consonant–vowel syllable string (loud 50%)	½
Recovery period	Walk on treadmill until resting heart rate is achieved	3
Second experimental session
Rest	Breathe quietly with no voicing (rest no-voice baseline)	5
	Produce a consonant–vowel syllable string (spon baseline)	½
	Breathe quietly with no voicing to return to rest baseline	1
	Produce a consonant–vowel syllable string (loud voice)	½
Rest-to-work transition	Run on treadmill to achieve a high workload	3
70% of age-predicted maximal heart rate (high)	Breathe quietly with no voicing at a high workload (70% no voice)	1
	Produce a consonant–vowel syllable string (spon 70%)	½
	Breathe quietly with no voicing to return to 70% baseline	1
	Produce a consonant–vowel syllable string (loud 70%)	½
Recovery period	Walk on treadmill until resting heart rate is achieved	3

During loud vocal SPL, participants received a hand gesture from the experimenter to increase or decrease vocal intensity depending on visual inspection of the PAS 6600 output screen displaying two reference lines for the target vocal intensity range. Some participants completed additional sets if inspection determined loud vocal SPL had not been achieved. Clarity was included in the instruction as voice quality interacts with loudness, and it is a valid ecological communication goal for intelligibility purposes. However, clarity was not an inclusion criterion for using data during analysis; the instruction served to motivate participants in a high-effort voice production. Attainment of vocal clarity was not determined to be a problem during collection of data as determined informally through auditory perception.

Following completion of the first set of R_law tasks, participants observed a 1-min rest period. Subsequently, participants completed a rest-to-work transition on the treadmill. The appropriate workload was set per participant-specific speed and grade adjustments to achieve workloads of either 50% or 70% of age-predicted maximal heart rate. Participants maintained that workload for 2 min to achieve steady state in oxygen consumption (VO₂; McArdle et al., 2010), establishing new steady-state baselines for each workload. The same R_law tasks were repeated as previously, using the same vocal SPL conditions, after workload steady state had been achieved. Finally, participants completed a recovery period until heart rate responses resolved to baseline. These procedures were repeated during the second session after participants reviewed R_law task procedures. The experimental protocol at the second session required participants to engage in the workload they had not yet encountered (50% or 70%). Within participants, the same order of spontaneous versus loud vocal SPL was used for the first session.

Variables

Independent variables were vocal SPL condition (no voice [quiet breathing], spontaneous, and loud vocal SPL) and workload (standing [rest], walking at 50% of age-predicted maximal heart rate [low], and walking at 70% of age-predicted maximal heart rate [high]). Dependent variables were R_law (in cmH₂O/[L/s]), Ve (in L/min), and P_ETCO₂ (in mmHg). Covariates were f0 (in Hz) and SPL (in dB).

Data Sources and Measurement Methods

R_law, mean P_s, and mean U were derived from the PAS 6600 signal during phonation. The middle five pressure peaks generated during /p/ of the respective signals and their corresponding airflow signal during the vowel portion were manually selected and examined for acceptable morphology, intensity, and frequency based on visual inspection (Helou & Solomon, 2011; Holmberg et al., 1988; Smitheran & Hixon, 1981). For acceptable tokens, f0 during all voice conditions could vary by ± 1 ST, while vocal SPL during the loud condition could fluctuate by ± 2 dB as determined during data analysis using the PAS 6600 output. A digital tone generator (http://www.seventhstring.com/tuningfork/tuningfork.html; Limited, 2009) confirmed f0 values miscalculated by the PAS 6600 system at data analysis, using the nearest frequency of the tone produced by the online generator that best corresponded to the participant's pitch. Criteria for selection were determined a priori and applied rigorously; samples not meeting the criteria were discarded. Values for P_s, U, f0, and SPL were calculated using the PAS 6600 software and retained for statistical analysis.

The CareFusion Encore data acquisition system software used breath-by-breath analysis to capture respiratory data. A research assistant recorded times for when participants started and stopped an R_law task to ensure respiratory and phonatory data were time locked. Respiratory data were continuously recorded during all voice conditions of the R_law task, all periods of silence for each workload, and the rest condition when participants stood on the treadmill. Baseline values of Ve and P_ETCO₂ were calculated by averaging data of voicing (spontaneous vocal SPL) while standing (at rest). For voiced segments, Ve data were calculated from the entire period during which the participant phonated (i.e., from beginning to end of the R_law task). The last 30 s of each quiet breathing period before phonating were used as respiratory data for the nonphonated conditions at each workload level (i.e., Ve and VO₂). Data for P_ETCO₂ were extracted from a 15-s segment following completion of an R_law task. A heart rate sensor provided pulse data. Covariates were order of workload and vocal SPL conditions, attainment of f0 and SPL targets, and their respective values.

Bias

A counterbalancing sequence of conditions was randomly assigned to participants at the first experimental session, with one exception described shortly (see Table 5). Online software (Haahr, 1998) generated a randomized list of counterbalanced conditions. The same ordering of spontaneous vocal SPL versus loud vocal SPL was used, within participants, for R_law tasks across the experimental protocol.

Table 5.

R_law contrasts: pairwise comparisons of marginal linear predictions (women only).

Reference condition	Comparison condition	exp(b) Comp/Ref	SE	p	95% Confidence interval
Norm_Rest	Norm_Low	1.01	0.11	.951	[0.82, 1.24]
	Norm_High	0.95	0.10	.656	[0.78, 1.16]
	Loud_Rest	1.53	0.19	.001	[1.19, 1.95]
	Loud_Low	1.26	0.19	.131	[0.93, 1.70]
	Loud_High	1.12	0.17	.457	[0.83, 1.50]
Norm_Low	Norm_High	0.95	0.06	.417	[0.84, 1.07]
	Loud_Rest	1.52	0.10	< .001	[1.33, 1.73]
	Loud_Low	1.25	0.11	.010	[1.05, 1.49]
	Loud_High	1.11	0.10	.242	[0.93, 1.32]
Norm_High	Loud_Rest	1.60	0.14	< .001	[1.34, 1.90]
	Loud_Low	1.32	0.15	.014	[1.06, 1.64]
	Loud_High	1.17	0.13	.148	[0.94, 1.45]
Loud_Rest	Loud_Low	0.83	0.04	< .001	[0.74, 0.92]
	Loud_High	0.73	0.05	< .001	[0.64, 0.84]
Loud_Low	Loud_High	0.89	0.05	.046	[0.79, 1.00]

Note. Bolded p values (e.g., < .001) represent statistically significant results. R_law = laryngeal airway resistance.

Statistical Analyses

Confirmation of Workloads

Paired-samples t test was performed to determine whether target and observed heart rates differed at low and high workloads. A one-way repeated-measures analysis of variance (ANOVA-RM) was performed on observed heart rates during quiet breathing as a function of workload. Post hoc comparisons were completed using a Bonferroni adjustment.

Generalized Estimating Equations

Generalized estimating equations (GEEs; Zeger & Liang, 1986; Zeger, Liang, & Albert, 1988) with robust standard error estimates (Huber/White Sandwich estimators) were used to evaluate associations, adjusting for confounders. We specified a Gaussian (i.e., normal) distributional family for unratioed continuous dependent variables (i.e., Ve, VO₂, and P_ETCO₂) and a gamma family for R_law, which is ratioed, due to its strict positivity and excess skew in the positive direction. For both families, a log link function was selected with an exchangeable correlation structure (Ballinger, 2004). Data for f0 and SPL were centered. Indicators were created for not meeting vocal SPL and f0 targets; this allowed for interpretation of model constants as typical responses for the average participant who achieved vocal SPL and f0 targets. Analyses required fitting two models: one for silent conditions (no adjustments except sequence order) and one for voice conditions (with f0 and vocal SPL adjustments; StataCorp, 2017).

The results for the GEE models are presented as ratios representing relative percentage changes. Ratios > 1 are interpreted as an increase in the level of an outcome, and ratios < 1 are interpreted as a decrease in the level of an outcome. Ratios sufficiently close to 1 are not significantly associated with either an increase or a decrease. To assess the relevance of all potential predictors of each outcome, 95% confidence intervals were calculated. Statistical significance was set at p < .05, and all reported p values are two-sided.

Results

Participant Characteristics

One hundred ninety-nine individuals completed an online prescreening form to determine initial eligibility. Of those, 32 recreationally active and vocally untrained women met criteria for participation and completed all experimental procedures. One female participant who experienced an unanticipated adverse event (i.e., vasovagal response) during data collection was replaced to preserve the target sample size for statistical analyses. Four men also completed the experimental protocol, but analysis of their data was done only descriptively. Participant characteristics are shown in Table 2.

Target Heart Rate and Workload Exposure

The experiment imposed heart rates that were greater than resting and that were achieved by treadmill exercise. Treadmill workloads included standing (rest) and walking at 50% of age-predicted maximal heart rate (low) and 70% of age-predicted maximal heart rate (high). The following summary describes relative attainment of different treadmill workloads as assessed by heart rate data.

Using a paired-samples t test, no differences were observed between target and observed heart rates at low (133.1 ± 5.0 vs. 134.1 ± 5.8 bpm) and high (158.6 ± 3.7 vs. 159.1 ± 5.5 bpm) workloads, t(31) = 1.599, p = .120, d = 0.28 and t(31) = 0.758, p = .454, d = 0.13, respectively. A one-way ANOVA-RM was performed on observed heart rates as a function of treadmill workload to ensure that heart rates were significantly different among treadmill workload conditions. The assumption of sphericity had been violated, Mauchly's W = .661, χ²(2) = 12.426, p = .002, and a Greenhouse–Geisser adjustment was performed. The one-way ANOVA-RM revealed a main effect of workload, F(1.494, 46.299) = 2866.643, p < .001, η² = .989. Post hoc analyses revealed mean heart rate was significantly different for all comparisons, in predicted directions (rest [69.4 ± 9.6 bpm] vs. low [134.1 ± 5.8 bpm], p < .001; rest vs. high [159.1 ± 5.5 bpm], p < .001; and low vs. high, p < .001).

Voice Goal Attainment

The experiment constrained f0 across all conditions with voice and constrained vocal SPL in loud vocal SPL conditions with vocal SPL free to vary in the spontaneous vocal SPL conditions. f0 Criteria were not met in 45% of opportunities during the experiment, and f0 violations primarily occurred in the loud vocal SPL conditions as commonly occurs with increases in vocal intensity. The lack of f0 attainment was considered of minimal concern given it lacks a clear influence over R_law (Holmberg et al., 1989). In the loud vocal SPL condition, participants were expected to attain a vocal intensity 10 dB or greater than their spontaneous speaking vocal intensity. Some participants—in loud vocal SPL conditions—exhibited vocal intensities that were greater than the expected vocal SPL target, but participants were never below. Further specifics on f0 and vocal SPL are described subsequently (see Table 3).

f0

One participant had missing f0 data due to a technical issue. Means and standard deviations of f0 for women and men are shown in Table 3. Data on f0 were not excluded from statistical analyses when participants did not attain the target f0 given its lack of a clear relationship with R_law (Holmberg et al., 1989). A 2 × 3 ANOVA-RM was performed on f0 as a function of vocal SPL and workload to determine differences in f0 across conditions. A violation of the assumption of sphericity existed for the workload effect, Mauchly's W = .745, χ²(2) = 7.371, p = .025. The 2 × 3 ANOVA-RM revealed main effects of vocal SPL, F(1, 26) = 53.233, p < .001, partial η² = .672, and workload, F(1.680, 43.693) = 15.246, p < .001, partial η² = .370. Post hoc comparisons revealed significantly higher f0 during loud vocal SPL (M = 253.7, SE = 5.226) than spontaneous vocal SPL (M = 228.1, SE = 5.038). Participants had significantly higher f0s in low (M = 242.3, SE = 5.179) and high (M = 251.3, SE = 6.510) workloads compared to rest (M = 229.3, SE = 4.083), p = .001 and p < .001, respectively.

Vocal Intensity (SPL)

Means and standard deviations of vocal SPL conditions for women and men are reported in Table 3. Follow-up analyses were performed on data from participants who met the loud vocal SPL target. A 2 × 3 ANOVA-RM was performed on SPL as a function of vocal SPL and workload. The ANOVA-RM revealed an interaction effect of vocal SPL and workload, F(2, 50) = 11.347, p < .001, partial η² = .312. The simple main effect of workload for spontaneous vocal SPL, F(2, 50) = 44.811, p < .001, partial η² = .642, and loud vocal SPL, F(2, 52) = 14.026, p < .001, partial η² = .350, was significant. Post hoc comparisons revealed significantly higher vocal SPL during spontaneous vocal SPL in low and high workloads compared to rest, p < .001 and p < .001, respectively. Participants had significantly higher vocal SPLs during loud vocal SPL in low and high workloads compared to rest, p < .001 and p < .001, respectively. The simple main effect of vocal SPL for rest, low, and high workloads was significant. At each workload, participants had a higher vocal intensity in loud vocal SPL compared to spontaneous vocal SPL: rest, t(31) = 15.684, p < .001; low, t(27) = 7.138, p < .001; and high, t(29) = 9.089, p < .001.

GEEs

GEE analyses were conducted to investigate differences among conditions, some which imposed vocal intensity and heart rate targets and other conditions during which vocal intensity was free to vary. GEE analyses with robust standard error estimates were used, specifying a Gaussian family for unratioed data and a gamma family for ratioed data, log link function, and an exchangeable correlation structure. Confounder adjustments, as dictated by model type, included (a) continuous variables (f0 and SPL values), mean centered (subtracting sample means); (b) indictors for meeting vocal SPL and f0, coded as categorical variables (met f0 [0/1] and met vocal SPL [0/1]); and (c) counterbalancing sequence. This summary highlights results of phonatory and respiratory responder variables to different vocal SPL and workload conditions.

Aerobic Demands of Workload

Results of the GEE model for VO₂ are shown in Table 4. For participants who met vocal SPL and f0 targets, participants in quiet breathing during the rest workload had an average VO₂ of 4.63 L/(kg/min). In quiet breathing, VO₂ increased significantly for all workload comparisons (low vs. rest, high vs. rest, and high vs. low; ps < .001).

Table 4.

VO₂ contrasts: pairwise comparisons of marginal linear predictions (women only).

Reference condition	Comparison condition	exp(b) Comp/Ref	SE	p	95% Confidence interval
Quiet_Rest	Quiet_Low	5.22	0.12	< .001	[4.99, 5.46]
	Quiet_High	6.67	0.17	< .001	[6.35, 7.01]
Quiet_Low	Quiet_High	1.28	0.02	< .001	[1.24, 1.32]

Note. Bolded p values (e.g., < .001) represent statistically significant results. VO₂ = oxygen consumption.

R_law

Unadjusted means and standard deviations for R_law and its component parts of P_s and transglottal U are reported in Table 3. Unadjusted individual and group data on R_law are plotted in Figure 3a. For women, 14% (37/256) of the total number of cells containing R_law data were excluded from analysis for not meeting morphology criteria. Missing data accounted for 9% and 14% in spontaneous vocal SPL and loud vocal SPL at rest, 19% for both vocal SLP conditions at a low workload, and 19% and 13% in spontaneous vocal SPL and loud vocal SPL at a high workload, respectively.

Figure 3.

(a) Plot of unadjusted individual and group data on laryngeal airway resistance by vocal sound pressure level (SPL) and workload condition. (b) Plot of unadjusted individual and group data on minute ventilation by vocal SPL and workload condition.

Results of the GEE models for R_law are shown in Table 5. For participants who met vocal SPL and f0 targets, adjusting for actual f0 and SPL values, participants at rest with a spontaneous vocal SPL had an average R_law of 72.6 cm H₂O/(L/s). Changing spontaneous vocal SPL to loud while maintaining rest workload, R_law significantly increased by an estimated 53% relative to the response at spontaneous vocal SPL (95% confidence interval [19.2%, 95.5%], p = .001). R_law significantly increased when changing spontaneous vocal SPL to loud at a low workload (+ 25.3%, 95% confidence interval [5.5%, 48.7%], p = .010). When requiring loud vocal SPL, R_law decreased significantly under low and high workloads as compared to rest (low vs. rest, p < .001; high vs. rest, p < .001).

Ve

Unadjusted means and standard deviations for respiratory variables are reported in Table 3. Unadjusted individual and group data on Ve are plotted in Figure 3b. Results of the GEE models for Ve are shown in Table 6. For participants who met vocal SPL and f0 targets, participants in quiet breathing during the rest workload had an average Ve of 12.2 L/min. In quiet breathing, Ve increased significantly for all workload comparisons (low vs. rest, high vs. rest, and high vs. low; ps < .001). For participants who met vocal SPL and f0 targets, adjusting for actual f0 and SPL values, participants in the rest workload with spontaneous vocal SPL had an average Ve of 10.6 L/min. Changing rest workload to low or high while maintaining spontaneous vocal SPL, Ve significantly increased relative to the response at rest (ps < .001). Changing rest workload to low or high while maintaining loud vocal SPL, Ve significantly increased relative to the response at rest (ps < .001). Ve significantly increased when changing spontaneous vocal SPL to loud at low and high workloads, p = .001 and p < .001, respectively.

Table 6.

Ve contrasts: pairwise comparisons of marginal linear predictions (women only).

Reference condition	Comparison condition	exp(b) Comp/Ref	SE	p	95% Confidence interval
Quiet_Rest	Quiet_Low	3.04	0.08	< .001	[2.88, 3.20]
	Quiet_High	3.92	0.11	< .001	[3.71, 4.15]
Quiet_Low	Quiet_High	1.29	0.02	< .001	[1.24, 1.34]
Norm_Rest	Norm_Low	1.86	0.33	< .001	[1.31, 2.64]
	Norm_High	2.37	0.26	< .001	[1.91, 2.94]
	Loud_Rest	1.20	0.22	.311	[0.84, 1.72]
	Loud_Low	3.83	1.03	< .001	[2.25, 6.50]
	Loud_High	5.80	1.70	< .001	[3.27, 10.29]
Norm_Low	Norm_High	1.27	0.17	.072	[0.98, 1.65]
	Loud_Rest	0.65	0.10	.004	[0.48, 0.87]
	Loud_Low	2.06	0.45	.001	[1.34, 3.16]
	Loud_High	3.11	0.62	< .001	[2.10, 4.61]
Norm_High	Loud_Rest	0.51	0.07	< .001	[0.38, 0.68]
	Loud_Low	1.62	0.36	.032	[1.04, 2.51]
	Loud_High	2.45	0.49	< .001	[1.65, 3.63]
Loud_Rest	Loud_Low	3.18	0.68	< .001	[2.09, 4.83]
	Loud_High	4.81	1.14	< .001	[3.03, 7.65]
Loud_Low	Loud_High	1.51	0.32	.048	[1.00, 2.29]

Note. Bolded p values (e.g., < .001) represent statistically significant results. Ve = minute ventilation.

P_ETCO₂

Results of the GEE models for P_ETCO₂ are shown in Table 7. For participants who met vocal SPL and f0 targets, participants in quiet breathing during the rest workload had an average P_ETCO₂ of 32.17 mmHg. In quiet breathing, P_ETCO₂ increased significantly for all workload comparisons (low vs. rest, high vs. rest, and high vs. low; ps < .001). For participants who met vocal SPL and f0 targets, adjusting for actual f0 and SPL values, participants at rest workload with spontaneous vocal SPL had an average P_ETCO₂ of 33.53 mmHg. Changing rest workload to low or high while maintaining spontaneous vocal SPL, P_ETCO₂ significantly increased relative to the response at rest (ps < .001). Changing rest workload to low or high while maintaining loud vocal SPL, P_ETCO₂ significantly increased relative to the response at rest (ps < .001).

Table 7.

P_ETCO₂ contrasts: pairwise comparisons of marginal linear predictions (women only).

Reference condition	Comparison condition	exp(b) Comp/Ref	SE	p	95% Confidence interval
Quiet_Rest	Quiet_Low	1.36	0.02	< .001	[1.32, 1.40]
	Quiet_High	1.41	0.02	< .001	[1.36, 1.46]
Quiet_Low	Quiet_High	1.03	0.01	< .001	[1.01, 1.05]
Norm_Rest	Norm_Low	1.60	0.03	< .001	[1.55, 1.66]
	Norm_High	1.74	0.03	< .001	[1.67, 1.80]
	Loud_Rest	0.99	0.02	.606	[0.95, 1.03]
	Loud_Low	1.60	0.04	< .001	[1.53, 1.68]
	Loud_High	1.76	0.05	< .001	[1.66, 1.86]
Norm_Low	Norm_High	1.08	0.02	< .001	[1.05, 1.11]
	Loud_Rest	0.62	0.01	< .001	[0.59, 0.64]
	Loud_Low	1.00	0.01	.902	[0.98, 1.02]
	Loud_High	1.09	0.02	< .001	[1.05, 1.14]
Norm_High	Loud_Rest	0.57	0.01	< .001	[0.55, 0.59]
	Loud_Low	0.92	0.01	< .001	[0.90, 0.95]
	Loud_High	1.01	0.02	.574	[0.97, 1.05]
Loud_Rest	Loud_Low	1.62	0.02	< .001	[1.58, 1.67]
	Loud_High	1.77	0.03	< .001	[1.72, 1.83]
Loud_Low	Loud_High	1.09	0.01	< .001	[1.07, 1.12]

Note. Bolded p values (e.g., < .001) represent statistically significant results. P_ETCO₂ = end-tidal carbon dioxide.

Discussion

Loud voice produces an acoustic signal of sufficient discriminative power that contrasts from interfering signals and facilitates perceptual distinctiveness (Bourne & Garnier, 2012; DeLeo LeBorgne, Lee, Stemple, & Bush, 2010; Stone, Cleveland, Sundberg, & Prokop, 2003). Aspects of voice such as vocal SPL for some professional voice users are relatively fixed due to occupational demands (i.e., loud speech; Hoffman-Ruddy et al., 2001). As for patients with respiratory conditions, the obstructive effect of the larynx on ventilation when speaking exacerbates breathing symptoms (Lee et al., 1998). Using aerodynamic outcome measures, this study investigated whether phonatory glottal valving in the face of different workload conditions is dependent on SPL targets. The hypothesis was that adjustments in glottal valving as measured by R_law and respiratory response as measured by Ve would be dependent upon vocal SPL targets. Specifically, we hypothesized R_law would be greater in loud vocal SPL than in the spontaneous vocal SPL condition that was free to vary, regardless of treadmill workload, due to glottal valving adjustments that favor voicing in loud vocal SPL but that reduce Ve and result in hypercapnia. We further hypothesized that the difference in R_law between loud and spontaneous vocal SPLs would widen as treadmill workload increased from rest to low and high due to glottal valving adjustments in spontaneous vocal SPL that favor the body's metabolic respiratory requirements to increase Ve for greater CO₂ expulsion. Results were largely congruent with hypotheses, partially validating a theory of action in vocal motor control. Consistent with a theory of action, phonatory goals such as vocal SPL seemed important when specified, and left unconstrained, glottal valving appeared to favor the body's metabolic needs as evidenced by lower R_law values. Discussion of results is limited to women who met their loud vocal SPL target.

Key Results

R_law

Consistent with hypotheses, in women who met their loud vocal SPL target, R_law values at rest were significantly higher in loud vocal SPL than during spontaneous vocal SPL. The higher R_law in loud vocal SPL than in spontaneous vocal SPL also was evident at a low workload, although the magnitude of difference diminished as participants went from standing to walking on the treadmill at a low workload intensity. R_law at a high workload was not significantly different between loud and spontaneous vocal SPLs. Contrary to hypotheses, R_law decreased slightly under low and high treadmill workloads when compared to rest, but decreases were greater in loud vocal SPL than in spontaneous vocal SPL.

These findings suggest loud voice can “win” using physiology specific to a vocal SPL target in the face of high–respiratory drive conditions, at least in the short term. That is, metabolic needs increased to accomplish treadmill walking in the low-workload condition, yet individuals still met vocal SPL targets in the loud condition. In fact, regardless of workload, individuals met vocal SPL targets in the loud condition, and some even exceeded their loud vocal SPL target. R_law decreased in loud vocal SPL as workload intensity increased, but the loud vocal SPL target was still attained.

The lack of R_law differences between loud vocal SPL and spontaneous vocal SPL conditions at a high workload appears to be due to increases in P_sub in the spontaneous vocal SPL condition. Though SPL was free to vary in the spontaneous vocal SPL condition, SPL increased systematically in the spontaneous condition as workload intensity increased. This increase in SPL may be due to larger lung pressures as the respiratory system increases ventilation to meet metabolic demands for a greater intensity of exercise. Despite the lack of differences between the vocal SPL conditions, R_law in the loud vocal SPL condition was still greater at a high workload than at rest. Together, these findings provide evidence that vocal motor control is goal dependent and that speakers demonstrate a physiology of vocal activity despite initial conditions that perturb the respiratory system.

Respiratory Variables

As predicted, Ve increased significantly at low and high workloads from a baseline of quiet breathing at rest, indicating hyperpnea. During spontaneous and loud vocal SPL, increases in workload to low and high significantly increased Ve from baseline values. Ve did not significantly differ between loud vocal SPL and spontaneous vocal SPL at rest. Consistent with hypotheses, loud vocal SPL produced significantly greater Ve as compared to spontaneous vocal SPL at low and high workloads.

Increasing the intensity of a physiologic stressor typically increases ventilation, while increases in airway resistance from phonation generally decreases ventilation (Bunn & Mead, 1971; Doust & Patrick, 1981). Changes in ventilation are known to potentially impact gas exchange. Consistent with hypotheses, P_ETCO₂ increased significantly at low and high workloads from quiet breathing at rest. An increase in workload significantly affected P_ETCO₂ within each vocal SPL condition. Both spontaneous vocal SPL and loud vocal SPL demonstrated significantly greater P_ETCO₂ at low and high workloads compared to rest. Inconsistent with hypotheses, P_ETCO₂ did not differ between spontaneous vocal SPL and loud vocal SPL at each workload despite differences in R_law between those voice types at rest and low workload. This lack of differences is likely due to the short duration of the R_law task.

Interpretation of the Data

Presumably, increases in airflow and breathy voice quality occur with relatively open vocal folds (Peterson, Verdolini-Marston, Barkmeier, & Hoffman, 1994). Although airflow increases during speech in high respiratory drive compared with speech under eupnea (Bailey & Hoit, 2002; Bunn & Mead, 1971; Doust & Patrick, 1981; Hoit et al., 2007), this increase is mainly linked to nonvoiced expirations, frequently at ends of phrases. In contrast, voiced expirations demonstrate only slight increases in airflow compared to spontaneous speech under normal drive. If loud voice is associated with increased vocal fold adduction, then consequently, exhalatory airflow in phonation should be impeded and blood gas levels simultaneously negatively impacted from decreased ventilation (Bailey & Hoit, 2002; Bunn & Mead, 1971; Doust & Patrick, 1981; Hoit et al., 2007; Otis & Clark, 1968). The following expands on findings that participants achieved loud vocal SPL targets under different workload conditions in line with a goal-independent model of vocal motor control.

Phonatory Outcomes

The physiology of the respiratory–laryngeal systems consists of multiple degrees of freedom (Hixon, Weismer, & Hoit, 2008; Pennestri, Cavacece, & Vita, 2005; Sellars & Sellars, 1983; Smith & Mead, 1986; von Leden & Moore, 1961). A theory of action states that motor behavior is goal oriented and functionally specific even in changing initial conditions. Goal-oriented loud voice resulted in a greater change in R_law from baseline compared to spontaneous voice to meet a loud acoustic goal, at least over a short duration, at rest and at a low workload. This finding is consistent with research on R_law in conditions of normal respiratory drive, in which increases in SPL cause R_law to increase (Holmberg et al., 1988; Wilson & Leeper, 1992). Breath holding demonstrates similar goal-oriented control, in which the termination point can be prolonged with distraction or motivation, even when partial pressure of CO₂ rises to supposedly dangerous levels (Parkes, 2006). The participants in this study demonstrated a similar ability to override metabolic need in attainment of loud voice, at least for a short duration. Participants somewhat maintained R_law even in spontaneous voice, which was supposed to allow deference toward respiratory goals.

Changes in R_law from baseline decreased as workload increased, and this finding was especially true in loud voice. When the goal was respiration, the pulmonary–laryngeal coordinative structure served gas exchange to meet somatic goals, as evidenced by expected changes in R_law relative to loud vocal SPL at a high compared to low workload. This study is the first to investigate R_law changes as a function of physical activity. Gillespie et al. (2015) found in their female participants that R_law and its component parts of P_s and U in spontaneous voice (i.e., unconstrained by instructions or other experimental maneuvers) did not differ between eupnea and hypercapnia induced via inspiration of CO₂ despite a significant increase in Ve from eupnea. That is, R_law exhibited robustness against challenges to homeostasis in individuals free to vary dimensions of voice such as f0 and vocal SPL during data collection. They suggested R_law may be a control parameter for voice that individuals maintain at a relatively constant level, despite respiratory gas exchange. In contrast, we found that R_law changes depending on initial conditions, but despite these alterations and the specific R_law value, participants achieved specified vocal intensity targets.

Differences in findings between that study and the current one may be due to ventilation requirements. Participants in the current study increased Ve in spontaneous voice at a low workload by approximately 24 L/min over baseline quiet breathing at rest and in spontaneous voice at a high workload by 35 L/min over baseline quiet breathing at rest. Participants in Gillespie's study only increased Ve by 12 L/min from eupnea to hyperpnea. Airway dilation is a known consequence of increased respiratory drive during exercise, including the larynx (Farrell, Joyner, & Caiozzo, 2012). The current study suggests that R_law is susceptible to changes in ventilation, although not to the point at which ventilatory requirements prevent individuals from varying SPL. Participants continued to maintain a greater increase in R_law in loud voice compared to spontaneous voice, in relation to baseline, across rest and low workloads. In light of this difference, phonatory goals seem important when specified, but left unconstrained, phonation favors metabolic need as evidenced by lower R_law.

Research on neural mechanisms of breathing support a motor control organization in which two pathways exist for breathing, that is, a cortical pathway that serves volitional acts of breathing and exerts influence on the medullary respiratory center and a bulbar pathway for central pattern generation of breathing for life (Jürgens, 2009; Ludlow, 2005, 2011). The finding on R_law provides new evidence that participants make attempts to meet phonatory goals, at least over a short duration, even when metabolic need is high. This evidence is in line with overcoming increasing metabolic pressures of other volitional acts of breathing (Parkes, 2006), suggesting cortical processes exert influence on lower breathing centers.

Respiratory Outcomes

The study also investigated Ve as a measure of adequate and efficient gas exchange to understand how the body achieves somatic goals in the face of voice production. This study found increases in workload induced a systematic increase in Ve from baseline during breathing and in spontaneous and loud voices, providing evidence of a hyperpneic response due to increases in respiratory drive (McArdle et al., 2010). Gillespie et al. (2015) found participants significantly increased Ve in response to inspired CO₂. This finding likely reflects the response of peripheral chemoreceptors that react to changes in partial pressures of CO₂ to maintain respiratory homeostasis. However, unlike Gillespie et al. (2015), the hyperpneic response observed during treadmill exercise in this study most likely reflects several ventilator control mechanisms, including locomotor-linked stimuli, namely, central command (feedforward) and muscle receptors (feedback), as well as matching to metabolic needs via partial pressures of O₂ and CO₂ (feedback; Farrell et al., 2012).

The obstructive nature of the larynx during phonation appeared to manifest differently when considering different SPLs at low and high workloads but not at rest. This finding indicates that changes in R_law that occur with loud voice, compared to spontaneous voice, impact ventilation. Taken together, evidence suggests that participants strive to meet SPL targets for a short duration despite sacrificing somatic goals to some extent. Moreover, implicit somatic goals only impact acoustic goals, whether spontaneous or loud, when they substantially increase in strength (Baker et al., 2008; Meckel, Rotstein, & Inbar, 2002). That finding reflects a motor control of respiration that favors the ability of the individual to override ventilatory needs when prioritizing SPL targets, even when somatic goals are strong.

Intense exercise results in more CO₂ production than moderate exercise (McArdle et al., 2010). That participants exhibited a significantly larger increase in P_ETCO₂ from baseline in the high workload compared to low is not surprising. Research shows that exercise results in increases in P_ETCO₂ and reflects the importance of lung perfusion and alveolar ventilation to eliminate CO₂, restore partial pressures of CO₂, and maintain acid–base balance (McArdle et al., 2010). More CO₂ in exhaled air reflects normal respiratory function. Gillespie et al. (2015) found that mean P_ETCO₂ levels during voiced segments were significantly greater in hypercapnia than eupnea, and P_ETCO₂ was significantly greater in voiced segments compared to nonvoiced ones during eupnea and hypercapnia. Unexpectedly, P_ETCO₂ did not differ between normal and loud voices at all workloads. This finding indicates that both SPLs provide a similar ventilatory limitation and hypoventilatory response, at least in the short term. In contrast, one study found that loud voice in normal drive results in less P_ETCO₂ than spontaneous speech, presumably due to the increased ventilation observed in loud speech (Russell et al., 1998). The difference between that study and the current one likely lies in the nature of the ventilatory response from speech. At rest, speech results in hyperventilation, but during exercise, speech results in hypoventilation. Another difference between studies may be task duration, where a longer task may be required to result in differences in P_ETCO₂.

Generalizability

This study provided a window into phonatory and respiratory function during simultaneous phonation and physical activity. Data indicate acoustic goals matter, even in the face of significant respiratory perturbations. This new knowledge will guide development of evaluation and treatment approaches for voice users who experience respiratory perturbations.

Limitations

Several limitations deserve mention. Although heart rates increase linearly with workload changes, using maximal oxygen consumption to establish workloads may have decreased response variability. Differences in aerobic fitness may have led to variability in the response to a physiologic stressor, for which future studies should control by determining maximal oxygen consumption. The short duration of the voice task likely did not allow metabolic variables to attain steady state when voicing, and future research could require a longer voice task. Participants exhibited difficulty matching pitch with loud voice, and some participants overshot their loud voice goal. Future studies should test ability to match pitch and loudness in all voice conditions, including in high respiratory drive. Finally, participants exhibited some pressure peaks that did not meet morphology requirements, possibly from jarring with treadmill walking. Despite this data loss, results were in line with what was predicted, and it would be unlikely that the “nonpredicted” data were preferentially lost. Future studies may consider a cycle ergometer as an alternative to a treadmill.

Conclusions

This study set out to test a theory of action in investigating the function of the respiratory–laryngeal coordinative structure under different respiratory and phonatory goals. Individuals who focus on an acoustic goal such as loud voice demonstrate R_law values that differ from spontaneous voice, despite increasing metabolic need. Participants accomplished goal-directed behavior for loud and, to some extent, spontaneous voice production, despite respiratory perturbations. These results demonstrate greater flexibility in physiology to achieve loud voice, and to some extent spontaneous voice, than previously thought. In relation to respiratory goals, the increased flow restriction that phonation places on airflow during loud voice does appear to impact ventilation requirements and metabolic needs, but for a short duration, acoustic goals may supersede respiratory goals. Spontaneous voice allows for less interference with somatic needs than loud voice, and the ventilatory response better serves metabolic needs when respiratory goals take precedence.

Funding Statement

This research was supported with awards to Ziegler that included a F31 Fellowship Grant (F31 DC012960) from the National Institute on Deafness and Other Communication Disorders, Research Development Funds from the University of Pittsburgh's School of Health and Rehabilitation Sciences, and the New Century Scholars Doctoral Scholarship from the American Speech-Language-Hearing Foundation.