Aerodynamic Demand Responses to the Fluid Interval Test for Voice: A Laryngeal Diadochokinetic Vocal Loading Task

Abstract

Introduction:

Vocal loading tasks are useful for studying vocal demand responses—the unique sets of biomechanical, aerodynamic, and acoustic adaptations that speakers employ to meet the vocal demands of communication. We discuss aerodynamic demand responses to the Fluid Interval Test for Voice (FIT-V), a vocal loading task founded on resisted laryngeal diadochokinesis (LDDK), to determine how aerodynamic factors may influence LDDK performance.

Method:

Participants (N = 30) produced loud abductory LDDK (/hʌ hʌ hʌ/) for 30 min, alternating 30-s intervals of exercise and rest. Depending on the task condition, fluid back pressure of 0 or 5 cm H₂O was imposed to modulate respiratory demand. Airflow data were captured continuously and analyzed using a two-way analysis of variance (Task × Time) to characterize aerodynamic adaptations in both LDDK pulses and whole breath groups.

Results:

Instantaneous LDDK rate did not change significantly within trials or intervals—unlike the aerodynamic variables, which exhibited task- and time-dependent changes. Large increases in mean airflow and expiratory volume within trials were characteristic of the resisted FIT-V₅ task, whereas the nonresisted FIT-V₀ task exhibited declining mean airflow and expiratory volume. Irrespective of task, all breath groups grew shorter within trials, and all aerodynamic variables decreased dramatically within intervals.

Conclusions:

Voice users actively modulated their aerodynamic output at multiple levels throughout the FIT-V tasks, although it is unclear whether these modulations were responsible for participants' ability to maintain rapid LDDK rates. Future work will examine how different vocal demand responses influence perceived fatigue, performance fatigability, and laryngeal pathology risk.

When approaching communication scenarios, speakers have access to thousands of unique permutations of aerodynamic, acoustic, and biomechanical adaptations that would enable them to achieve their communication goals. In voice science, these permutations have been termed vocal demand responses (Calvache Mora et al., 2024; Hunter et al., 2020): If communicative success in a given scenario were represented by the summit of a mountain, then vocal demand responses would correspond to the individual paths available to reach that summit. The choice of path, whether made deliberately or inadvertently, is nontrivial, both for speech-language pathologists, who make their livings as expert mountaineers, and for voice users, who must avoid paths that may place them at greater risk of hazards such as perceived vocal fatigue or vocal effort, phonotrauma, and chronic voice impairment. Because the terrain of a speaker's vocal demands changes constantly, vocal demand responses are presumed to be contextual, variable across speakers, and fluid in time (Hunter & Berardi, 2025; Hunter et al., 2020).

Although contextual shifts in vocal demand response would ideally be studied in ecological communication scenarios, naturally occurring speech is highly variable. Thus, one key purpose of vocal loading tasks (VLTs)—exercise tasks with known vocal demand parameters—is to characterize vocal demand responses under controlled conditions. Typical vocal loading models include loud oral reading, singing, or sustained vowel phonation tasks that range in duration from 15 min to 3.75 hr and target high sound pressure levels (Fujiki & Sivasankar, 2017). Dependent measures include ratings of perceived vocal effort or perceived vocal fatigue (Berardi & Hunter, 2024; Chang & Karnell, 2004), aerodynamic measures such as phonation threshold pressure (Chang & Karnell, 2004; Sivasankar & Erickson-Levendoski, 2012; Solomon & DiMattia, 2000; Whitling et al., 2015), and acoustic measures such as fundamental frequency (F0), jitter, shimmer, noise-to-harmonic ratio, and cepstral peak prominence (De Bodt et al., 1998; Fujiki et al., 2017; Gorham-Rowan et al., 2016; Remacle et al., 2012). Only perceived vocal effort and a closely linked construct, perceived vocal fatigue, have consistently changed during VLT administration (Fujiki & Sivasankar, 2017).

Related to this observation, a second key purpose of VLTs is to identify mechanisms underlying vocal fatigue. Fatigue, as it is understood in neuromuscular physiology (Enoka & Duchateau, 2016; Kluger et al., 2013) and voice science (Hunter et al., 2020), consists of two dimensions. The first dimension, perceived fatigue, measures subjective sensations of tiredness, as well as affective responses to these sensations (Venhorst et al., 2018a, 2018b). As a short-term state, perceived vocal fatigue is inferred via ratings of perceived effort or exertion (Borg & Noble, 1974; Ford Baldner et al., 2015); as a long-term trait, it has been measured using scales such as the Vocal Fatigue Index (VFI; Nanjundeswaran et al., 2015). In both cases, perceived fatigue must be measured by patient report. The second dimension, performance fatigability, quantifies declines in task performance over time—declines in running or walking speed, jumping height, resistance exercise force or power, and so forth (Enoka & Duchateau, 2016; Kluger et al., 2013). Where declines are observed, physiological measures (electromyography, electroencephalography, near-infrared spectroscopy, phosphorus-31 nuclear magnetic resonance spectroscopy) are often used to identify their potential mechanisms.

Although perceived fatigue and performance fatigability often mirror one another, disruptions in muscle afferent feedback can alter this relationship (Amann et al., 2020). In a study of exhaustive cycling, for example, Blain et al. (2016) found significant differences in a range of neuromuscular and biomechanical performance variables between control cyclists and cyclists who had undergone fentanyl blockade of Group III/IV muscle afferent fibers, even though no between-groups differences were observed in perceived exertion. Voice science has long acknowledged that sensorimotor deficits may contribute to the development of hyperfunctional voice disorders such as nodules, muscle tension dysphonia, and chronic vocal fatigue (Hillman et al., 2020; Shembel & Nanjundeswaran, 2025). It is therefore essential that VLT models distinguish between perceived fatigue and performance fatigability—a difficult task in speech-based VLTs, where linguistic or prosodic variability may obscure biomechanical or physiological variability.

We have previously proposed that laryngeal diadochokinesis (LDDK; Apfelbach et al., 2024a; Kent et al., 2022; Lombard & Solomon, 2020; Snell et al., 2020) may complement speech and sustained vowel production as a vocal–motor gesture that both elicits and measures performance fatigability in the intrinsic muscles of the larynx. Relative to speech, LDDK has no linguistic or prosodic variability and imposes lower vocal fold vibratory doses (Apfelbach et al., 2024b); the latter feature may be valuable for delineating vibratory sources of vocal fatigue (Titze, 1994; Welham & Maclagan, 2003) from neuromuscular sources. Finally, because the aerodynamic and acoustic signals produced during LDDK are quasiperiodic, variables such as LDDK rate, strength, and rhythmicity can be quantified on a pulse-by-pulse basis to track changes in task performance, whether positive (faster, stronger, or more rhythmic pulses) or negative (slower, weaker, or less rhythmic pulses).

Our previous work resulted in the creation of the Fluid Interval Test for Voice (FIT-V), a laryngeal diadochokinetic VLT (Apfelbach et al., 2024a, 2024b). The FIT-V has three essential design features: (a) loud LDDK performed as quickly as possible, (b) intervallic rest and exercise, and (c) resistive fluid back pressure. Task-specific demand parameters such as the type of LDDK (adductory /ʔʌ/ vs. abductory /hʌ/), the duration or duty cycle of rest and exercise intervals, and the source or magnitude of resistive back pressure can all be modulated separately, as can demand parameters shared with speech-based VLTs (target F0 and sound pressure, VLT duration, etc.). Two abductory variants of the FIT-V task (/hʌ/) were investigated in the current study: a back pressure–resisted FIT-V task (FIT-V₅, 5 cm H₂O back pressure) and a nonresisted FIT-V task (FIT-V₀, 0 cm H₂O). Task demand parameters and justifications are provided in the Procedure section (see Table 1).

Table 1.

Demand parameters for the FIT-V0 and FIT-V5 vocal loading tasks.

Parameter	FIT-V0	FIT-V5	Justification
LDDK type	Abductory (/hʌ hʌ hʌ/)	Abductory (/hʌ hʌ hʌ/)	Abductory LDDK was initially selected to maximize the displacement (and mechanical work) of the arytenoid cartilages during LDDK. Adductory LDDK will be investigated in future, however, as we acknowledge that vocal fold adduction is a key feature of hyperfunctional voice disorders.
Rest interval duration (s)	30	30	Drawn from research on high-intensity interval training (HIIT; MacInnis & Gibala, 2017).
Exercise interval duration (s)	30	30
Duty cycle	50%	50%
Back pressure source	EMST device	EMST device	Back pressure magnitudes were drawn from research on semi-occluded vocal tract exercise (SOVTE; Apfelbach & Guzmán, 2021) and feasibility pilot testing (Apfelbach et al., 2024a). The FIT-V0 condition assesses the effect of LDDK as the target vocal motor gesture, relative to speech; the FIT-V5 condition assesses the additional effect of resistive back pressure.
Back pressure magnitude (cm H2O)	0	5
Target F0 (Hz)	Habitual, as determined by participants	Habitual, as determined by participants	Drawn from reviews of vocal loading task designs (Fujiki & Sivasankar, 2017).
Target sound pressure (dB)	75–80 dB at 45 cm	75–80 dB at 45 cm
VLT duration (min)	30	30

Note. FIT-V = Fluid Interval Test for Voice; LDDK = laryngeal diadochokinesis; EMST = expiratory muscle strength training; F0 = fundamental frequency.

This study examined changes in LDDK rate, airflow, and breath group measures—all potential measures of performance fatigability or adaptation in the vocal demand response—across tasks (FIT-V₀ vs. FIT-V₅), time points (first vs. middle vs. last 10% of sampled data points), and levels of analysis (within trials vs. within intervals). On the principle that exercise performance generally declines faster as exercise intensity increases (Hill, 1993; Poole et al., 2016), we hypothesized that LDDK rate would decline within trials and intervals in both the nonresisted FIT-V₀ and the back pressure–resisted FIT-V₅ tasks, but that declines would be more pronounced in the FIT-V₅ task. Based on the high levels of perceived respiratory exertion that participants reported in Apfelbach et al. (2024b), we also hypothesized that mean airflow for pulses and breath groups would increase within trials in both tasks, though to a greater extent in the FIT-V₅ task.

Method

Participants

All study procedures were reviewed and approved by the University of Delaware Institutional Review Board (ID #1776237-7). Participants were recruited from the University of Delaware campus and Newark, DE, community using physical flyers, mass e-mail, and word of mouth from February 2022 until November 2022. Of the 103 prospective participants, four did not consent to participate, 42 were eliminated due to the study's inclusion and exclusion criteria, and 27 either withdrew for personal reasons or were removed from the study after they failed to attend data collection sessions.

The 30 remaining participants who completed the study were equally sex-balanced. All but one participant identified as cisgender. Participants consisted of 21 White (10 female, 11 male), three Black (two female, one nonbinary AFAB), four Asian (one female, three male), one Pacific Islander (one female), and one multiracial (one male), none of whom identified as Hispanic or Latino. Participants were compensated $80 in Amazon gift cards after completing the four in-person data collection sessions.

Participants were enrolled in the study if (a) they were between 18 and 40 years old and (b) they denied clinically significant symptoms of vocal fatigue, as judged by VFI scores ≥ 24 (Factor 1), ≥ 7 (Factor 2), or ≤ 7 (Factor 3; Nanjundeswaran et al., 2015). Additional exclusion criteria included (a) a history within the past 5 years of speech, language, voice, or hearing disorders; (b) a diagnosis of a pulmonary, neuromuscular, cardiovascular, autoimmune, or endocrine disorder that might reasonably affect vocal function; (c) any history of smoking, vaping, or tobacco use; (d) a diagnosis of head or neck cancer; (e) a diagnosis of gastroesophageal or laryngopharyngeal reflux; (f) consistent use of drying medications (e.g., antihistamines, diuretics, ACE inhibitors, Accutane); (g) pregnancy; (h) a history of vocal training, operationally defined as formal training or participation in amateur or professional singing/acting groups within the past 5 years; (i) irregular, aperiodic laryngeal diadochokinetic airflow signals after training (see Figure 1); or (j) invalid or unreliable forced vital capacity (FVC) maneuvers after training, per American Thoracic Society guidelines (Graham et al., 2019).

Figure 1.

(A) Irregular, aperiodic LDDK. Distinct airflow pulses are rarely present. (B) Regular, periodic LDDK. Airflow pulses remain distinct throughout the breath group. LDDK = laryngeal diadochokinesis.

Materials and Equipment

Audio Recording

An American National Standards Institute Type II sound-level meter accurate to within ±1.4 dB (Model R8080, Reed Instruments) was used to monitor participants' sound pressure level throughout the vocal exercise task. Audio recordings of task performance were captured in Audacity using a cardioid microphone (SM48-LC; Shure, Inc.). The microphone and sound-level meter were each placed 45 cm away from the participant at a 30° offset.

Spirometry

FVC, forced expiratory volume in the first second (FEV₁), and peak expiratory flow rate (PEFR) were captured prior to data collection using a Vitalograph Pneumotrac spirometer (Model 6800; Vitalograph, Ltd.).

FIT-V Apparatus

To minimize oral airflow leakage while imposing resistive back pressure, a Phonatory Aerodynamic System (PAS Model 6600; Pentax Medical) was outfitted with 3D-printed couplings, flexible PVC piping (24 in. long × 7/8 in. inner diameter), and disposable silicone diving bite mouthpieces (Silicone Gear/Casco Bay Molding; see Figure 2). Oral airflow passed from the mouthpiece through the PAS pneumotachometer, where it was sampled at 200 Hz. Two expiratory muscle strength training (EMST) devices (EMST 75 Lite; Aspire Products LLC) with the spring and plate removed (FIT-V₀) or set to 5 cm H₂O (FIT-V₅) were mounted to the pneumotachometer to generate the different resistive loads in the FIT-V₀ and FIT-V₅ conditions. Prior investigations of different EMST devices found that the EMST 75 Lite has a trigger pressure of 4.79 ± 0.48 cm H₂O when set at 5 cm H₂O (Dietsch et al., 2024). All participants wore nose clips to minimize nasal airflow leakage.

Figure 2.

The Fluid Interval Test for Voice apparatus. The mouthpiece, couplings, and flexible PVC tubing are attached anteriorly to the Phonatory Aerodynamic System. The EMST 75 Lite, in green, is attached posteriorly to the pneumotachometer.

Analysis and Data Visualization

Signal processing, statistical analysis, and data visualization were conducted using JASP 0.18.3.0, GraphPad Prism 10, and Python 3.13.

Design

IV₁: Level of Analysis

Each 30-min FIT-V trial was subdivided into 1-min intervals with a 50% duty cycle (see Figure 3). The dependent variables were analyzed both within trials (i.e., sampled at the beginning, middle, and end of each 30-min trial) and within intervals (i.e., sampled at the beginning, middle, and end of each 30-s interval). Results are split by level of analysis, but no between-levels comparisons were made.

Figure 3.

Design of the Fluid Interval Test for Voice (FIT-V) tasks. The FIT-V₀ and FIT-V₅ tasks differed only in the level of resistive back pressure imposed by the expiratory muscle strength training device. LDDK = laryngeal diadochokinesis; F0 = fundamental frequency.

IV₂: Task

The study used a three-period, three-treatment crossover design with Latin squares and a 2-day washout period between tasks (see Figure 4; the loud oral reading task will not be discussed in the current article, but is described in Apfelbach et al., 2024b). The FIT-V₀ and FIT-V₅ task conditions differed only in the magnitude of back pressure imposed by the EMST device (0 cm H₂O for the FIT-V₀ vs. 5 cm H₂O for the FIT-V₅).

Figure 4.

Flow between study sessions. Note that the loud oral reading task, shown in green, is not discussed in this article, as aerodynamic measures were only collected during the two Fluid Interval Test for Voice (FIT-V) tasks.

IV₃: Time

Ten percent of the data were sampled at each level of analysis: first, middle, and final. Within trials, 10% typically corresponded to three intervals (0:00–3:00, 13:30–16:30, and 27:00–30:00) unless participants failed to complete a full 30-min FIT-V trial. Within intervals, 10% corresponded to 3 s (0:00–0:03, 0:13.5–0:16.5, and 0:27–0:30, aggregated across all intervals).

DV₁: Alternating Current Airflow (Q_AC)

Local airflow minima and maxima were grouped into pairs called pulses. For each LDDK pulse, the alternating current (AC) airflow Q_{AC i} was calculated as the difference between the local airflow maximum Q_maxi and its preceding minimum Q_mini (see Figure 5, in red):

Figure 5.

Alternating current (AC) and direct current (DC) airflow after Savitzky–Golay smoothing of the raw airflow trace. DC airflow, shown as blue points, was the local airflow minimum in the oscillating laryngeal diadochokinesis signal. AC airflow, shown as red lines, was the airflow difference between the local airflow maximum and its preceding minimum.

$$Q_{\mathrm{AC}i}=Q_{maxi}-Q_{mini}$$ (1)

DV₂: Direct Current Airflow (Q_DC)

Direct current (DC) airflow Q_{DC i} was identical to the local airflow minimum Q_mini (see Figure 5, in blue):

$$Q_{\mathrm{DC}i}=Q_{mini}$$ (2)

DV₃: Exhaled Volume (V_ex)

Sequences of LDDK pulses were aggregated into breath groups. For each breath group, the exhaled volume V_ex was calculated by integrating the airflow trace Q from the onset of the breath group b_on to its offset b_off:

$$V_{\mathrm{ex}}=\sum _{i=b_{\mathrm{on}}}^{b_{\mathrm{off}}}{Q}_i$$ (3)

DV₄: Expiratory Duration (t_ex)

The expiratory duration t_ex of each breath group was calculated by subtracting the timestamp of the offset of the breath group b_off from that of the breath group onset b_on:

$$t_{\mathrm{ex}}=b_{\mathrm{off}}-b_{\mathrm{on}}$$ (4)

DV₅: Instantaneous LDDK Rate (ν_inst)

Because time-averaged LDDK rates can obscure variability in pulse timing, we calculated the duration between successive airflow maxima as the interpulse interval (IPI):

$$\mathrm{IPI}=s_{\mathrm{\text{Qmax}}i+1}-s_{\mathrm{\text{Qmax}}i},$$ (5)

where s _{Qmax i + 1} is the timestamp of one local airflow maximum and s_{Qmax i} is the timestamp of the local airflow maximum immediately preceding it (see Figure 6). The rate ν_inst can then be calculated as the inverse of the interpulse interval IPI:

Figure 6.

Interpulse intervals shown over the first half of a breath group. Instantaneous laryngeal diadochokinesis rate v_inst was calculated as the inverse of the interpulse interval.

$${\nu }_{\mathrm{\text{inst}}}=\frac{1}{\mathrm{IPI}}$$ (6)

These calculations mirror calculations used to quantify neuronal activity (see, e.g., Gerstner et al., 2014, Section 7.2.1).

Procedure

The study was advertised via flyer, mass e-mail, and word of mouth to residents of Newark, DE, and University of Delaware students. Interested prospective participants scanned a QR code that directed them to the study's REDCap database, where they answered eligibility questions anonymously. Participants who were eligible for inclusion were invited to provide their contact information, whereupon the first author arranged to document informed consent remotely over Zoom. After agreeing to participate, participants completed the VFI (Nanjundeswaran et al., 2015), Voice Handicap Index-10 (Rosen et al., 2004), and the Reflux Symptom Index (RSI; Belafsky et al., 2002) on REDCap. Participants who scored above the threshold criteria for either the RSI (≥ 14) or any VFI factor (≥ 24, Factor 1; ≥ 7, Factor 2; ≤ 7, Factor 3) were dismissed from the study.

Participants next took part in a training session that consisted of (a) spirometry, (b) LDDK training, and (c) practice FIT-V trials. Spirometric measures (including FVC, FEV₁, and PEFR) were collected during an FVC task; the best of three valid trials was used, per American Thoracic Society guidelines (Graham et al., 2019). Participants were also coached to produce fast, perceptually strong LDDK pulse trains, oriented to the Borg CentiMax scale used to measure perceived exertion (see Apfelbach et al., 2024b, for details), and finally asked to perform one mock FIT-V trial at each resistance level.

Data from the FIT-V₀ and FIT-V₅ tasks were gathered in counterbalanced 30-min data collection sessions separated by a washout period of at least 48 hr (see Figure 4). Participants provided clean-catch urine samples at the start of each session to account for the effects of urine-specific gravity (ACON Biotech), an index of hydration status, on perceived vocal exertion and vocal performance (Levendoski et al., 2014; Sivasankar et al., 2008; Verdolini et al., 1994; Wu & Zhang, 2017, 2024). Temperature and relative humidity were also recorded using a digital psychrometer (SAM990DW, General Tools).

The target sound pressure level during task performance was 75–80 dB SPL (45-cm microphone distance, 30^° microphone angle). The first author monitored participants' sound pressure level continuously using a sound-level meter (Model R8080, Reed Instruments) and cued participants to increase their sound pressure level if they dropped below the target. Participants remained seated during all training and data collection sessions and were outfitted with nose plugs during task performance to minimize nasal airflow.

FIT-V Tasks (FIT-V₀ and FIT-V₅)

The demand parameters of the FIT-V₀ and FIT-V₅ tasks are described below (see Table 1). In both task conditions, exercise continued until (a) 30 min elapsed, marking a completed trial; (b) Borg CentiMax ratings exceeded 95 in two consecutive intervals, indicating near-maximal perceived exertion; (c) LDDK rate dropped below 4 Hz for more than 10 s despite vigorous encouragement; or (d) the participant voluntarily stopped the vocal loading protocol.

Signal Processing

F0, SPL, and airflow signals from the PAS were sampled at 200 Hz and imported into Python. The airflow trace was smoothed using a fourth-order Savitzky–Golay filter with a window length of 25 frames; smoothed airflow, not raw airflow, was used for all subsequent analyses. Next, all airflow extrema (minima Q_min or maxima Q_max) were identified and filtered: Extrema with low airflow rates (Q_max ≤ .25 L/s, Q_min ≤ −.50 L/s) or small peak prominences (Q_prom ≤ .15 L/s) were eliminated. Cutoff values were selected by the first author after visual inspection of the data set. Extrema above or below the listed thresholds often occurred during inhalation, momentary airflow perturbations, and so forth that did not reflect valid LDDK pulses.

After filtering, extrema were grouped into Q_min → Q_max pairs, each collectively called a pulse. Pulses with low AC airflow values (Q_AC ≤ .10 L/s), long or short durations (Duration_pulse ≥ 0.25 s or ≤ 0.025 s), and high or low instantaneous LDDK rates (ν_inst ≥ 12 Hz or ≤ 2.5 Hz) were eliminated to prevent pulses that met extrema thresholds—but did not reflect valid LDDK—from influencing statistical analyses. Finally, breath groups were segmented by calculating the gap between successive airflow maxima. Although segmentation would ideally have been performed based on zero-crossings of the smoothed airflow trace, the EMST device used in the FIT-V₅ task prevented retrograde airflow through the pneumotachometer during inspiration, making zero-crossings unreliable. Thus, an alternative method was used: Airflow maxima with sufficiently lengthy gaps between them (≥ 0.40 s) were flagged as new breath groups; those with short durations (Duration_breath ≤ 0.50 s) were eliminated. Finally, all airflow and volume measures were normalized to participants' FVCs to reduce the influence of height, sex, age, and so forth on statistical analyses.

Statistical Analysis

Two-way analysis of variance (ANOVA) models with robust standard errors were created to examine the effects of task (FIT-V₀, FIT-V₅), time (first 10%, middle 10%, final 10%), and Task × Time interactions on each of the five dependent variables listed in the Design section. Separate models were created for within-trial and within-intervals analyses. Effect sizes for the omnibus ANOVA models were calculated using partial epsilon-squared ( ${\epsilon }_p^2$ ), a less biased measure of effect size than partial eta-squared or partial omega-squared (Okada, 2013). For all analyses, ${\epsilon }_p^2$ < .02 is considered very small, $.02\le {\epsilon }_p^2<.13$ is considered small, $.13\le {\epsilon }_p^2<.26$ is considered medium, and ${\epsilon }_p^2\ge .26$ is considered large (Cohen, 1992).

Pairwise comparisons using Tukey's honestly significant difference were performed (a) between tasks at equivalent time points and (b) between time points within tasks whenever the omnibus Task, Time, or Task × Time ANOVA models were significant (α = .05). Because variance and sample size varied widely across groups and time points, Glass' δ was used to estimate pairwise effect size, where the FIT-V₀ was the control task and the control time points were the first 10% in the first–middle and first–final pairs and the middle 10% in the middle–final pair. Pairwise comparisons are compiled in the Appendix for brevity.

Results

Five dependent measures were analyzed: (a) airflow (Q_AC), (b) DC airflow (Q_DC), (c) exhaled volume per breath group (V_ex), (d) expiratory duration per breath group (t_ex), and (e) instantaneous LDDK rate (ν_inst). Main effects of Task (unresisted FIT-V₀ vs. resisted FIT-V₅) and Time (first 10% vs. middle 10% vs. final 10%) were analyzed within trials and intervals, as were select Task × Time interaction effects.

Pulse Airflow Measures (Q_AC and Q_DC)

Q_AC and Q_DC represent the oscillating, alternating current and nonoscillating, DC components of the diadochokinetic airflow signal, respectively. The omnibus ANOVA models revealed significant Task, Time, and Task × Time effects within trials and intervals for both Q_AC and Q_DC (see Table 2). Task had a medium-sized effect on Q_AC ( ${\epsilon }_p^2\ge .13$ ) and a small effect on Q_DC ( ${\epsilon }_p^2\ge .02$ ). All other effects were negligible.

Table 2.

Results of two-way analysis of variance examining Task, Time, and Task × Time effects on Q_AC and Q_DC.

QAC
Variable	Sum of squares	df	M square	F	p	${\epsilon }_p^2$
Time: within trials
Task	248	1	248	12590	< .001***	.201**
Time	2.35	2	1.18	59.7	< .001***	< .02
Task × Time	7.52	2	3.76	191	< .001***	< .02
Residuals	985	49956	0.0197
Time: within intervals
Task	277	1	277	13867	< .001***	.215**
Time	16.0	2	8.01	401	< .001***	< .02
Task × Time	2.26	2	1.13	56.6	< .001***	< .02
Residuals	1,011	50633	0.0200
Q DC
Time: within trials
Task	251	1	251	4576	< .001***	.0839*
Time	31.0	2	15.5	283	< .001***	< .02
Task × Time	30.6	2	15.3	280	< .001***	< .02
Residuals	2,737	49956	0.0548
Time: within intervals
Task	168	1	168	2679	< .001***	.0502*
Time	17.1	2	8.55	137	< .001***	< .02
Task × Time	2.45	2	1.22	19.6	< .001***	< .02
Residuals	3,170	50633	0.0626

Note. All values were rounded to three significant figures. For p values, p < .001***, p < .01**, and p < .05*. For partial epsilon-squared measures of effect size ( ${\epsilon }_p^2$ ), ${\epsilon }_p^2<.02$ = very small, $.02\le {\epsilon }_p^2<.13$ = small*, $.13\le {\epsilon }_p^2<.26$ = medium**, and ${\epsilon }_p^2\ge .26$ = large*** (Cohen, 1992).

Task comparisons showed that the unresisted FIT-V₀ condition elicited Q_AC values of 12%–15% FVC/s, while the resisted FIT-V₅ condition averaged 25%–30% FVC/s—roughly double that of the FIT-V₀. This trend was reversed for Q_DC; however, the FIT-V₀ task showed mean Q_DC values of 18%–23% FVC/s, compared with 10%–13% FVC/s in the FIT-V₅ task (see Table 3). The ratios of AC and DC flow between tasks remained largely stable over time, although signs of divergence and convergence are visible within trials (see Figures 7 and 8).

Table 3.

Summary statistics for Q_AC and Q_DC, expressed as percentage of forced vital capacity, by task and time conditions.

QAC
Variable	Task: FIT-V0				Task: FIT-V5
	Count	M	SD	COV	Count	M	SD	COV
Time: within trials
Beginning (0%–10%)	8,400	0.152	0.101	0.661	9,664	0.261	0.159	0.610
Middle (45%–55%)	7,686	0.132	0.0872	0.659	8,266	0.276	0.173	0.626
End (90%–100%)	7,464	0.125	0.0825	0.662	8,482	0.290	0.188	0.647
Time: within intervals
Beginning (0%–10%)	9,791	0.154	0.0959	0.622	10,889	0.313	0.183	0.586
Middle (45%–55%)	7,224	0.134	0.0924	0.692	8,213	0.269	0.166	0.617
End (90%–100%)	6,439	0.128	0.0869	0.680	8,083	0.257	0.165	0.640
Q DC
Time: within trials
Beginning (0%–10%)	8,400	0.223	0.222	0.993	9,664	0.020	0.233	11.40
Middle (45%–55%)	7,686	0.227	0.197	0.867	8,266	0.114	0.263	2.30
End (90%–100%)	7,464	0.228	0.201	0.881	8,482	0.135	0.273	2.02
Time: within intervals
Beginning (0%–10%)	9,791	0.260	0.233	0.894	10,889	0.134	0.311	2.33
Middle (45%–55%)	7,224	0.220	0.199	0.905	8,213	0.106	0.257	2.42
End (90%–100%)	6,439	0.209	0.179	0.856	8,083	0.116	0.261	2.25

Note. Counts, means, standard deviations, and coefficients of variation are given by task, time point, and level of analysis (within trials or within intervals). FIT-V = Fluid Interval Test for Voice; COV = coefficient of variation.

Figure 7.

Q_AC, in %FVC/s, by Task and Time conditions. Results for the FIT-V₀ and FIT-V₅ are shown in blue and red, respectively. FVC = forced vital capacity; FIT-V = Fluid Interval Test for Voice.

Figure 8.

Q_DC, in %FVC/s, by Task and Time conditions. Results for the FIT-V₀ and FIT-V₅ are shown in blue and red, respectively. FVC = forced vital capacity; FIT-V = Fluid Interval Test for Voice.

Breath Group Measures (V_ex and t_ex)

The omnibus ANOVA models for exhaled volume per breath group V_ex and expiratory duration t_ex (see Table 4) showed significant yet small effects of Task within trials and larger effects of Task and Time within intervals.

Table 4.

Results of two-way analysis of variance examining Task, Time, and Task × Time effects on V_ex and t_ex.

Vex
Variable	Sum of squares	df	M square	F	p	${\epsilon }_p^2$
Time: within trials
Task	16.8	1	16.8	255	< .001***	.0773*
Time	0.763	2	0.381	5.79	.00310**	< .02
Task × Time	4.80	2	2.40	36.4	< .001***	.02280*
Residuals	200	3035	0.0659
Time: within intervals
Task	48.0	1	48.0	790	< .001***	.126*
Time	46.2	2	23.1	380	< .001***	.122*
Task × Time	1.93	2	0.966	15.9	< .001***	< .02
Residuals	332	5459	0.0607
t ex
Time: within trials
Task	258	1	258	87.1	< .001***	.0276*
Time	148	2	74.0	25.0	< .001***	< .02
Task × Time	3.64	2	1.82	0.616	.540	< .02
Residuals	8,982	3,035	2.96
Time: within intervals
Task	512	1	512	159	< .001***	.0280*
Time	2,434	2	1217	377	< .001***	.121*
Task × Time	24.1	2	12.0	3.73	.0240*	< .02
Residuals	17,618	5,459	3.23

Note. All values were rounded to three significant figures. For p values, p < .001***, p < .01**, and p < .05*. For partial epsilon-squared measures of effect size ( ${\epsilon }_p^2$ ), ${\epsilon }_p^2<.02$ = very small, $.02\le {\epsilon }_p^2<.13$ = small*, $.13\le {\epsilon }_p^2<.26$ = medium**, and ${\epsilon }_p^2\ge .26$ = large*** (Cohen, 1992).

Both FIT-V tasks exhibited within-trial declines of 15%–20% in expiratory duration t_ex—declines that heightened to 38% within intervals (see Table 5). In the FIT-V₀ condition, these declines in t_ex were matched by concomitant decreases in exhaled volume per breath group V_ex, indicating progressively smaller breath groups with no substantial changes in mean airflow rate (see Figures 9 and 10). In the FIT-V₅ condition, however, declining expiratory durations were associated with progressively higher exhaled volumes per breath group, suggesting rising mean airflow rates.

Table 5.

Summary statistics for V_ex, expressed as percentage of forced vital capacity, by task and time conditions.

Vex
Variable	Task: FIT-V0				Task: FIT-V5
	Count	M	SD	COV	Count	M	SD	COV
Time: within trials
Beginning (0%–10%)	512	0.495	0.255	0.515	481	0.544	0.232	0.427
Middle (45%–55%)	519	0.441	0.224	0.509	478	0.621	0.279	0.449
End (90%–100%)	556	0.386	0.229	0.593	495	0.625	0.314	0.501
Time: within intervals
Beginning (0%–10%)	820	0.558	0.233	0.418	819	0.797	0.280	0.351
Middle (45%–55%)	1011	0.461	0.217	0.471	1,019	0.651	0.266	0.410
End (90%–100%)	885	0.367	0.207	0.563	911	0.511	0.268	0.525
t ex
Time: within trials
Beginning (0%–10%)	512	3.30	1.66	0.505	481	3.95	1.93	0.489
Middle (45%–55%)	519	3.12	1.63	0.520	478	3.62	1.69	0.468
End (90%–100%)	556	2.77	1.61	0.581	495	3.39	1.80	0.533
Time: within intervals
Beginning (0%–10%)	820	4.09	1.95	0.477	819	4.87	2.08	0.428
Middle (45%–55%)	1011	3.41	1.63	0.477	1,019	4.06	1.87	0.460
End (90%–100%)	885	2.57	1.53	0.595	911	3.02	1.70	0.562

Note. Counts, means, standard deviations, and coefficients of variation are given by task, time point, and level of analysis (within trials or within intervals). FIT-V = Fluid Interval Test for Voice; COV = coefficient of variation.

Figure 9.

V_ex, in %FVC, by Task and Time conditions. Results for the FIT-V₀ and FIT-V₅ are shown in blue and red, respectively. FVC = forced vital capacity; FIT-V = Fluid Interval Test for Voice.

Figure 10.

t_ex, in seconds, by Task and Time conditions. Results for the FIT-V₀ and FIT-V₅ are shown in blue and red, respectively. FIT-V = Fluid Interval Test for Voice.

Instantaneous LDDK Rate (ν_inst)

Instantaneous LDDK rate (ν_inst) was calculated as the inverse of the duration between successive airflow maxima. A two-way ANOVA model with robust standard errors showed no substantive effects $\left({\epsilon }_p^2<.02\right)$ of Task, Time, or Task × Time interactions on ν_inst at any level of analysis (see Table 6). Instantaneous LDDK rate was both stable in time and equivalent between task conditions (see Table 7): Within trials, declines in mean ν_inst were effectively nonexistent; within intervals, on the order of 5% (see Figure 11).

Table 6.

Results of two-way analysis of variance examining Task, Time, and Task × Time effects on ν_inst.

Variable	Sum of squares	df	M square	F	p	${\epsilon }_p^2$
Time: within trials
Task	623	1	623	544	< .001***	< .02
Time	15.7	2	7.84	6.85	.00106**	< .02
Task × Time	11.2	2	5.62	4.90	.00744**	< .02
Residuals	57,228	49956	1.15
Time: within intervals
Task	565	1	565	503	< .001***	< .02
Time	758	2	379	337	< .001***	< .02
Task × Time	9.99	2	4.99	4.44	.0118*	< .02
Residuals	56912	50633	1.12

Note. All values were rounded to three significant figures. For p values, p < .001***, p < .01**, and p < .05*. For partial epsilon-squared measures of effect size ( ${\epsilon }_p^2$ ), ${\epsilon }_p^2<.02$ = very small, $.02\le {\epsilon }_p^2<.13$ = small*, $.13\le {\epsilon }_p^2<.26$ = medium**, and ${\epsilon }_p^2\ge .26$ = large*** (Cohen, 1992).

Table 7.

Summary statistics in hertz for ν_inst by task and time conditions.

Variable	Task: FIT-V0				Task: FIT-V5
	Count	M	SD	COV	Count	M	SD	COV
Time: within trials
Beginning (0%–10%)	8,400	5.46	0.994	0.182	9,664	5.72	1.12	0.195
Middle (45%–55%)	7,686	5.44	1.01	0.186	8,266	5.65	1.12	0.198
End (90%–100%)	7,464	5.49	1.08	0.197	8,482	5.68	1.08	0.190
Time: within intervals
Beginning (0%–10%)	9,791	5.64	0.947	0.168	10,889	5.82	1.06	0.183
Middle (45%–55%)	7,224	5.42	0.999	0.184	8,213	5.65	1.11	0.196
End (90%–100%)	6,439	5.32	1.08	0.203	8,083	5.57	1.17	0.210

Note. Counts, means, standard deviations, and coefficients of variation are given by task, time point, and level of analysis (within trials or within intervals). FIT-V = Fluid Interval Test for Voice; COV = coefficient of variation.

Figure 11.

ν_inst, in hertz, by Task and Time conditions. Results for the FIT-V₀ and FIT-V₅ are shown in blue and red, respectively. FIT-V = Fluid Interval Test for Voice.

Discussion

The current study quantified adaptations in instantaneous LDDK rate (ν_inst) and several aerodynamic variables during a 30-min VLT, the FIT-V. These adaptations constitute one part of the vocal demand response—the sum of the aerodynamic, acoustic, biomechanical, and cognitive responses to imposed vocal demands. Elucidating the critical components of economical and noneconomical vocal demand responses may advance the treatment of hyperfunctional voice disorders (e.g., muscle tension dysphonia, vocal fold nodules), conditions in which the vocal demand responses that voice users employ to overcome transient decrements in vocal performance can yield either virtuous or vicious cycles of endurance or fatigue, ease or stiffness, or comfort or discomfort (Hillman et al., 2020; Hunter et al., 2020).

Aerodynamic Adaptations

Across all levels of analysis, trial, and interval, the results of the current study demonstrate that participants continuously modulated their aerodynamic output to meet the demands of rapid, high-SPL LDDK. The effects of time and task on aerodynamic measures were often modestly sized; however, these effects may have been a function of the high variability observed across participants. Traditional statistical methods that examine group means may prove to be poorly suited to studying vocal demand responses, which are inherently variable across speakers, time points, and vocal demand parameters (Berardi et al., 2025; Hunter & Berardi, 2025; Hunter et al., 2020).

Adaptations were task specific. Participants performing the FIT-V₀ task decreased their mean AC airflow Q_AC, exhaled volume per breath group V_ex, and expiratory time t_ex as the trial progressed while holding DC airflow Q_DC stable. These results suggest that participants ended FIT-V₀ trials using shorter breath groups and slightly lower mean airflow rates (largely driven by reductions in Q_AC) than when they started. Connecting these results with participant interviews, we speculate that the absence of resistive back pressure in the FIT-V₀ task may have encouraged participants to “overblow” in the early intervals before eventually moderating their airflow rates, just as runners who sprint off the starting block too quickly eventually settle into sustainable racing paces.

Whereas participants adapted to the FIT-V₀ task by decreasing overall aerodynamic output, they adapted to the FIT-V₅ task by dramatically increasing it. For example, expiratory duration t_ex decreased in both FIT-V₀ and FIT-V₅ trials by roughly 15%; rather than a proportional reduction in exhaled volume per breath group V_ex, as observed in the FIT-V₀, however, the FIT-V₅ elicited a 15% increase in V_ex. This pronounced increase in mean airflow rate per breath group was driven in large part by six- to sevenfold increases in DC airflow Q_DC. One interpretation of these data is that participants increasingly used elastic recoil mechanisms in the lungs and chest wall to overcome the FIT-V₅'s resistive back pressure, but ultimately “wasted” greater proportions of their vital capacity on nonoscillating DC airflow as the task progressed. In our previous work on the FIT-V, participants reported that their perceived respiratory exertion (i.e., lightheadedness, plus exertion localized to the thorax and abdomen) increased as back pressure increased (Apfelbach et al., 2024b); the aerodynamic differences between the FIT-V₀ and FIT-V₅ task variants could explain these perceptual differences, although more granular analyses are needed.

Task-specific performance adaptations were less evident within intervals, where both the FIT-V₀ and FIT-V₅ tasks produced 15%–20% declines in airflow measures Q_AC and Q_DC and 35%–40% declines in breath group measures V_ex and t_ex. We are guarded in our interpretation of the breath group measures, as a breath group produced at the start of an interval after a full, unhurried inhalation may scarcely resemble a breath group produced at the end of an interval, especially if the final breath group was prematurely terminated. Our confidence in the pulse-level measures (Q_AC, Q_DC) is stronger, as we would expect the final 10% of an interval's duration to have captured far more LDDK pulses than breath groups. Regardless, participants tended to decrease their aerodynamic output substantially within intervals—although whether this decrement represents short-term, reversible fatigue, an adaptive pacing strategy, or another process is beyond the scope of this article.

Instantaneous LDDK Rate

Decrements in instantaneous LDDK rate (ν_inst) within trials and intervals were functionally nonexistent. We offer four potential interpretations of these data: First, many FIT-V₀ trials (15/30) and FIT-V₅ trials (10/30) elicited within-trial increases in ν_inst, a result not predicted by models of neuromuscular fatigue. Increases in ν_inst in some participants may have statistically offset decreases in ν_inst in other participants. If true, these increases could reflect warm-up effects, learning effects, or submaximal efforts that were not adequately mitigated during the training session.

Second, it is possible that maximal ν_inst cannot become meaningfully impaired in the healthy, young people recruited for the current study. Although a discussion of the factors regulating neuromuscular performance fatigability is beyond the scope of this article (for further reading, see Allen et al., 2008; Enoka & Duchateau, 2016; Gandevia, 2001; Kent-Braun et al., 2012), diverse sets of genetic, embryological, physiological, and biomechanical data support the assertion that the craniofacial and limb skeletal muscles may exhibit strikingly different contractile properties, including fatigue resistance (Hoh, 2005, 2010; Kent, 2004; Yahya et al., 2022). Because age (Lombard & Solomon, 2020) and neuromuscular pathology (Kent et al., 2022; Verdolini Abbott & Palmer, 1997) adversely influence ν_inst, however, laryngeal diadochokinetic performance can clearly be impacted below an unspecified threshold of neuromuscular integrity.

A third possibility is that the demands of the FIT-V tasks were not strenuous enough to elicit ν_inst performance fatigability. The type of LDDK (adductory vs. abductory), target F0 or dB SPL, level of resistive back pressure, task duration, and task duty cycle (i.e., the ratio of exercise to rest time in each interval) could all be modulated in future studies to increase vocal demand and, theoretically, decrements in ν_inst.

A fourth interpretation of the ν_inst data presented in the current study is that participants mitigated decrements in ν_inst by modulating their aerodynamic output. By this logic, strategic allocation of respiratory resources—that is, an economical vocal demand response—allowed participants to compensate for declining laryngeal performance and maintain ν_inst. Participants with impaired respiratory kinematics or respiratory pathology would be expected under this paradigm to exhibit more severe ν_inst declines during prolonged LDDK.

Limitations

We can identify three main limitations of the current study. First, laryngeal diadochokinetic performance was highly variable across participants. Although every effort was made to standardize performance by training participants before data collection, normalizing airflow and volume measures to vital capacity, enforcing a target SPL, and so forth, group-based statistical methods remain imperfect tools for analyzing vocal demand responses, which are idiosyncratic and mutable. In the future, we may default to cluster analyses or other high-resolution analytical techniques to highlight rather than efface individual variability (Berardi et al., 2025; Hunter & Berardi, 2025); every path to the communicative summit may have unique advantages and disadvantages for speakers.

Second, no physiological or kinematic measures were gathered in the current study. We therefore cannot conclusively attribute any of the results discussed in the preceding two sections to processes such as fatigue, warm-up, motor learning, or conscious pacing strategies. Future work may use electromyography, respiratory inductance plethysmography, motion capture, and so forth to elucidate the physiological and kinematic correlates of the aerodynamic and diadochokinetic behaviors reported here.

Third, the FIT-V as a vocal loading platform is still in its infancy, and many technical refinements are needed. Calculating LDDK efficiency measures (Q_AC/Q_DC) was infeasible, for example, because the design of the EMST device often produced artifactual, negative Q_DC values in the FIT-V₅ task. Similarly, breath groups were segmented using the duration between successive Q_max values, rather than via zero-crossings in the airflow trace, because the EMST's resistive plate obstructed inspiratory airflow in the FIT-V₅ task. Alternative sources of resistive back pressure (Tesla valves, e.g.; Zhang et al., 2023) may be required to address these challenges.

Future Directions

Much work remains to improve the FIT-V's utility as a platform for vocal loading and investigation of the vocal demand response. In the immediate term, refinement of the FIT-V apparatus' design (the type and magnitude of back pressure, the diameter of the tube connecting the participant's mouthpiece with the pneumotachometer, etc.), analytic pipeline (individualized thresholds for valid extrema, pulses, and breaths), and demand parameters (duration, duty cycle, target F0 and SPL, adductory vs. abductory LDDK, etc.) are top priorities. Once the experimental paradigm has been developed further, the FIT-V could be used to study how LDDK performance responds to a range of physiological and biomechanical perturbations, how the physiology of phonation changes during strenuous vocal exercise, and how resisted LDDK functions as an exercise stimulus for neuromuscular diseases that implicate the corticobulbar tract.

Conclusions

The current study examined two variations of a novel VLT, the FIT-V. The FIT-V tasks incorporate three essential elements: (a) LDDK, a vocal–motor gesture that combines pulses of transglottal airflow with rapid adduction and abduction of the vocal folds; (b) intervallic exercise and rest; and (c) resistive back pressure. Our results indicate that laryngeal diadochokinetic rate—the behavioral target of the FIT-V tasks—was unchanged across 30 min of high-intensity, intervallic vocal exercise, both in the presence of resistive back pressure (FIT-V₅) and in its absence (FIT-V₀). However, we noted substantial changes in airflow, breath group size, and breath group duration within trials and individual intervals. We postulate that these aerodynamic adaptations may have helped voice users to maintain their diadochokinetic performance target in the face of strenuous vocal demands. If true, these findings would have major implications for communication disorders in which laryngeal or respiratory biomechanics are disrupted, including chronic vocal fatigue, muscle tension dysphonia, laryngeal dystonia, and motor speech disorders.

Author Contributions

Christopher S. Apfelbach: Conceptualization, Data curation, Formal analysis, Investigation, Project administration, Software, Validation, Visualization, Writing – original draft, Funding acquisition, Methodology, Writing – review & editing. Mary J. Sandage: Supervision, Conceptualization, Funding acquisition, Methodology, Writing – review & editing. Katherine Verdolini Abbott: Supervision, Resources, Conceptualization, Funding acquisition, Methodology, Writing – review & editing.

Data Availability Statement

The data sets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request..

Appendix

Pairwise Comparisons

Table A1.

Results of Tukey's honestly significant difference pairwise comparisons of select Task × Time interaction effects on Q_AC.

Variable 1	Variable 2	p	M difference	Glass' δ
Time: within trials
FIT-V0 Beginning	FIT-V0 Middle	< .001***	−0.0199	−0.197
FIT-V0 Middle	FIT-V0 End	.00960**	−0.00771	−0.0884
FIT-V0 Beginning	FIT-V0 End	< .001***	−0.0276	−0.274
FIT-V5 Beginning	FIT-V5 Middle	< .001***	0.0157	0.0985
FIT-V5 Middle	FIT-V5 End	< .001***	0.0135	0.0783
FIT-V5 Beginning	FIT-V5 End	< .001***	0.0292	0.184
FIT-V0 Beginning	FIT-V5 Beginning	< .001***	0.108	1.08
FIT-V0 Middle	FIT-V5 Middle	< .001***	0.144	1.65
FIT-V0 End	FIT-V5 End	< .001***	0.165	2.00
Time: within intervals
FIT-V0 Beginning	FIT-V0 Middle	< .001***	−0.0206	−0.215
FIT-V0 Middle	FIT-V0 End	< .001***	−0.00574	−0.0621
FIT-V0 Beginning	FIT-V0 End	< .001***	−0.0264	−0.275
FIT-V5 Beginning	FIT-V5 Middle	< .001***	−0.0443	−0.241
FIT-V5 Middle	FIT-V5 End	< .001***	−0.0111	−0.0670
FIT-V5 Beginning	FIT-V5 End	< .001***	−0.0554	−0.302
FIT-V0 Beginning	FIT-V5 Beginning	< .001***	0.159	1.65
FIT-V0 Middle	FIT-V5 Middle	< .001***	0.135	1.46
FIT-V0 End	FIT-V5 End	< .001***	0.130	1.49

Note. FIT-V = Fluid Interval Test for Voice.

^** p < .001.

^*** p < .001.

Table A2.

Results of Tukey's honestly significant difference pairwise comparisons of select Task × Time interaction effects on Q_DC.

Variable 1	Variable 2	p	M difference	Glass' δ
Time: within trials
FIT-V0 Beginning	FIT-V0 Middle	.9278	0.00358	0.0162
FIT-V0 Middle	FIT-V0 End	.9996	0.00119	0.0060
FIT-V0 Beginning	FIT-V0 End	.796	0.00477	0.0215
FIT-V5 Beginning	FIT-V5 Middle	< .001***	0.0938	0.403
FIT-V5 Middle	FIT-V5 End	< .001***	0.0213	0.0810
FIT-V5 Beginning	FIT-V5 End	< .001***	0.115	0.494
FIT-V0 Beginning	FIT-V5 Beginning	< .001***	−0.203	−0.915
FIT-V0 Middle	FIT-V5 Middle	< .001***	−0.113	−0.572
FIT-V0 End	FIT-V5 End	< .001***	−0.0924	−0.460
Time: within intervals
FIT-V0 Beginning	FIT-V0 Middle	< .001***	−0.0402	−0.173
FIT-V0 Middle	FIT-V0 End	.149	−0.0104	−0.0522
FIT-V0 Beginning	FIT-V0 End	< .001***	−0.0506	−0.218
FIT-V5 Beginning	FIT-V5 Middle	< .001***	−0.0275	−0.0885
FIT-V5 Middle	FIT-V5 End	.143	0.00956	0.0372
FIT-V5 Beginning	FIT-V5 End	< .001***	−0.0180	−0.0578
FIT-V0 Beginning	FIT-V5 Beginning	< .001***	−0.126	−0.543
FIT-V0 Middle	FIT-V5 Middle	< .001***	−0.114	−0.572
FIT-V0 End	FIT-V5 End	< .001***	−0.0938	−0.523

Note. FIT-V = Fluid Interval Test for Voice.

^*** p < .001.

Table A3.

Results of Tukey's honestly significant difference pairwise comparisons of select Task × Time interaction effects on V_ex.

Variable 1	Variable 2	p	M difference	Glass' δ
Time: within trials
FIT-V0 Beginning	FIT-V0 Middle	.0102*	−0.0538	−0.211
FIT-V0 Middle	FIT-V0 End	.00650**	−0.0547	−0.244
FIT-V0 Beginning	FIT-V0 End	< .001***	−0.108	−0.426
FIT-V5 Beginning	FIT-V5 Middle	< .001***	0.0771	0.332
FIT-V5 Middle	FIT-V5 End	1.00	0.00421	0.0151
FIT-V5 Beginning	FIT-V5 End	.0323*	0.0814	0.350
FIT-V0 Beginning	FIT-V5 Beginning	< .001***	0.0489	0.192
FIT-V0 Middle	FIT-V5 Middle	< .001***	0.180	0.802
FIT-V0 End	FIT-V5 End	< .001***	0.239	1.04
Time: within intervals
FIT-V0 Beginning	FIT-V0 Middle	< .001***	−0.0966	−0.414
FIT-V0 Middle	FIT-V0 End	< .001***	−0.0943	−0.434
FIT-V0 Beginning	FIT-V0 End	< .001***	−0.191	−0.818
FIT-V5 Beginning	FIT-V5 Middle	< .001***	−0.147	−0.525
FIT-V5 Middle	FIT-V5 End	< .001***	−0.140	−0.525
FIT-V5 Beginning	FIT-V5 End	< .001***	−0.287	−1.03
FIT-V0 Beginning	FIT-V5 Beginning	< .001***	0.239	1.03
FIT-V0 Middle	FIT-V5 Middle	< .001***	0.189	0.870
FIT-V0 End	FIT-V5 End	< .001***	0.143	0.694

Note. FIT-V = Fluid Interval Test for Voice.

^* p < .05.

^** p < .01.

^*** p < .001.

Table A4.

Results of Tukey's honestly significant difference Pairwise Comparisons of Select Task × Time interaction effects on t_ex.

Variable 1	Variable 2	p	M difference	Glass' δ
Time: within trials
FIT-V0 Beginning	FIT-V0 Middle	.593	−0.172	−0.104
FIT-V0 Middle	FIT-V0 End	.0100**	−0.354	−0.218
FIT-V0 Beginning	FIT-V0 End	< .001***	−0.526	−0.316
FIT-V5 Beginning	FIT-V5 Middle	.0315*	−0.334	−0.173
FIT-V5 Middle	FIT-V5 End	.302	−0.229	−0.135
FIT-V5 Beginning	FIT-V5 End	< .001***	−0.563	−0.291
FIT-V0 Beginning	FIT-V5 Beginning	< .001***	0.655	0.394
FIT-V0 Middle	FIT-V5 Middle	< .001***	0.493	0.303
FIT-V0 End	FIT-V5 End	< .001***	0.618	0.384
Time: within intervals
FIT-V0 Beginning	FIT-V0 Middle	< .001***	−0.680	−0.348
FIT-V0 Middle	FIT-V0 End	< .001***	−0.841	−0.517
FIT-V0 Beginning	FIT-V0 End	< .001***	−1.52	−0.779
FIT-V5 Beginning	FIT-V5 Middle	< .001***	−0.812	−0.390
FIT-V5 Middle	FIT-V5 End	< .001***	−1.04	−0.555
FIT-V5 Beginning	FIT-V5 End	< .001***	−1.85	−0.888
FIT-V0 Beginning	FIT-V5 Beginning	< .001***	0.782	0.400
FIT-V0 Middle	FIT-V5 Middle	< .001***	0.650	0.399
FIT-V0 End	FIT-V5 End	< .001***	0.452	0.295

Note. FIT-V = Fluid Interval Test for Voice.

^* p < .05.

^** p < .01.

^*** p < .001.

Table A5.

Results of Tukey's honestly significant difference pairwise comparisons of select Task × Time interaction effects on ν_inst.

Variable 1	Variable 2	p	M difference	Glass' δ
Time: within trials
FIT-V0 Beginning	FIT-V0 Middle	.328	−0.0179	−0.0180
FIT-V0 Middle	FIT-V0 End	.897	0.0524	0.0518
FIT-V0 Beginning	FIT-V0 End	< .001***	0.0345	0.0347
FIT-V5 Beginning	FIT-V5 Middle	.0311*	−0.0615	−0.0551
FIT-V5 Middle	FIT-V5 End	< .001***	0.0242	0.0216
FIT-V5 Beginning	FIT-V5 End	< .001***	−0.0373	−0.0334
FIT-V0 Beginning	FIT-V5 Beginning	.177	0.259	0.261
FIT-V0 Middle	FIT-V5 Middle	.00170**	0.215	0.213
FIT-V0 End	FIT-V5 End	.687	0.187	0.173
Time: within intervals
FIT-V0 Beginning	FIT-V0 Middle	< .001***	−0.217	−0.229
FIT-V0 Middle	FIT-V0 End	< .001***	−0.0984	−0.0984
FIT-V0 Beginning	FIT-V0 End	< .001***	−0.315	−0.333
FIT-V5 Beginning	FIT-V5 Middle	< .001***	−0.170	−0.160
FIT-V5 Middle	FIT-V5 End	< .001***	−0.0808	−0.0730
FIT-V5 Beginning	FIT-V5 End	< .001***	−0.251	−0.236
FIT-V0 Beginning	FIT-V5 Beginning	< .001***	0.179	0.189
FIT-V0 Middle	FIT-V5 Middle	< .001***	0.226	0.226
FIT-V0 End	FIT-V5 End	< .001***	0.244	0.225

Note. FIT-V = Fluid Interval Test for Voice.

^* p < .05.

^** p < .01.

^*** p < .001.

Funding Statement

The authors gratefully acknowledge the National Institute on Deafness and Other Communication Disorders (NIDCD), which sponsored this work (NIDCD F31DC020362; principal investigator: Christopher S. Apfelbach).