Vowel Production in Children and Adults With Down Syndrome: Fundamental and Formant Frequencies of the Corner Vowels

Abstract

Purpose:

Atypical vowel production contributes to reduced speech intelligibility in children and adults with Down syndrome (DS). This study compares the acoustic data of the corner vowels /i/, /u/, /æ/, and /ɑ/ from speakers with DS against typically developing/developed (TD) speakers.

Method:

Measurements of the fundamental frequency (f _o) and first four formant frequencies (F1–F4) were obtained from single word recordings containing the target vowels from 81 participants with DS (ages 3–54 years) and 293 TD speakers (ages 4–92 years), all native speakers of English. The data were used to construct developmental trajectories and to determine interspeaker and intraspeaker variability.

Results:

Trajectories for DS differed from TD based on age and sex, but the groups were similar with the striking change in f _o and F1–F4 frequencies around age 10 years. Findings confirm higher f _o in DS, and vowel-specific differences between DS and TD in F1 and F2 frequencies, but not F3 and F4. The measure of F2 differences of front-versus-back vowels was more sensitive of compression than reduced vowel space area/centralization across age and sex. Low vowels had more pronounced F2 compression as related to reduced speech intelligibility. Intraspeaker variability was significantly greater for DS than TD for nearly all frequency values across age.

Discussion:

Vowel production differences between DS and TD are age- and sex-specific, which helps explain contradictory results in previous studies. Increased intraspeaker variability across age in DS confirms the presence of a persisting motor speech disorder. Atypical vowel production in DS is common and related to dysmorphology, delayed development, and disordered motor control.

Down syndrome (DS; Trisomy 21) is the most common form of intellectual disability, occurring in about one in 700–800 births (Sherman et al., 2007). It is one of the most complex genetic perturbations compatible with survival and often has multiple effects on body systems and organs (Roizen, 2010). Because of advances in health care, the life expectancy of people with DS has increased from less than 20 years to nearly 60 years over the last two generations (Bittles et al., 2007). Communication disorders are common in DS. In particular, speech intelligibility is compromised in many individuals and can be a lifelong problem that seriously reduces quality of life (Kent & Vorperian, 2013; Kumin, 1994; Wild et al., 2018). A survey of 228 parents of individuals with DS revealed that speech was identified as being in the top four of 16 areas needing further investigation and as the second area of greatest parental interest, behind only cognition (White et al., 2022). The speech disorder in DS is associated with disturbances across the subsystems of speech production—respiratory, phonatory, articulatory, resonatory, and prosodic (Jones et al., 2019; Kent & Vorperian, 2013). There does not appear to be a standardized approach to treatment of the speech disorder, although behavior analytic methods are promising (Neil & Jones, 2018). In their systematic review and meta-analysis of communication interventions in DS, Neil and Jones (2018) remarked that “Few interventions were tailored to the needs of the Down syndrome behavior phenotype” (p. 1). Approximately 90% of children aged birth to 5 years receive speech therapy services (King et al., 2022), but there may be a disparity between research and practice for such services (Frizelle et al., 2022). Promising behavioral treatments include intensive treatment with Lee Silverman Voice Treatment (LSVT LOUD; Boliek et al., 2022; Langlois et al., 2020; Mahler & Jones, 2012), biofeedback with ultrasound (Fawcett et al., 2008), electropalatography (Page & Johnson, 2021; Wood et al., 2019), or optical methods (Miyauchi et al., 2013).

Vowel Production in DS: Phonetic and Acoustic Studies

Impaired vowel production is one of the most frequently studied aspects of the speech disorder in DS. What follows is a summary of studies that examine the effects of DS on vowels at the supralaryngeal articulation level (vowel identity formant pattern) and on laryngeal voice/phonation level (perceptual and acoustic features), followed by a discussion of potential explanations for each.

A study of phonetic contrast errors in adults with DS showed that among the most frequent errors were vowel errors (e.g., long vs. short vowel, high vs. low vowel, front vs. back vowel; Bunton et al., 2007). The contribution of vowels to reduced speech intelligibility in DS was reported by Wild et al. (2018), who noted particular problems with the distinction between the low vowels /ɑ/ and /æ/. Acoustic studies have the potential to shed light on the nature of impaired vowel production in DS, but findings have been inconsistent. Atypical formant patterns reported include (a) a centralized or reduced acoustic vowel space area (VSA; Abolhasanizadeh & Olyiaiee, 2018, for Persian; Bunton & Leddy, 2011, for American English; Dar et al., 2020, for Kashmiri; Sorianello, 2015, for Italian), (b) atypical development of acoustic VSA and shape during early childhood (Whitworth & Bray, 2015, for British English), (c) overlapping F1–F2 regions of vowels (Carl et al., 2020, for American English; Novak, 1972, for American English), and (d) reduced difference of the F2 frequencies of vowels /i/ and /u/ (Moura et al., 2008, for Portuguese). However, it also has been reported that speakers with DS produce vowels that are similar to (Moran, 1986, for American English) or even more acoustically distinct than (Rochet-Capellan & Dohen, 2015, for French) those of typically developing (TD) speakers. Factors that might explain the differences among studies include language spoken, speaking task, method of analysis, and participant characteristics (e.g., age, sex, phenotypic variation).

Vowel Production in DS: Phonation

Atypical voice often is noted as a feature of speech production in DS (Kent & Vorperian, 2013). Properties of phonation are relevant to vowels because these speech sounds convey information on prosody and voice quality as well as phonetic identity. Acoustic studies of vocal fundamental frequency (f _o) in DS are not in complete agreement, with most studies reporting higher f _o in adults with DS than TD adults (Albertini et al., 2010; Corrales-Astorgano et al., 2018; M. T. Lee et al., 2009; Rochet-Capellan & Dohen, 2015), but findings in children are inconsistent with some studies reporting lower f _o in children with DS than TD children (Lyakso et al., 2021; Moura et al., 2008), while others reporting little or no difference in f _o between children with DS and TD children (Albertini et al., 2010; Michel & Carney, 1964; Weinberg & Zlatin, 1970). Krishnamurthy and Ramani (2020) concluded their review on voice by stating that there was a lack of standardized criteria to evaluate voice in children with DS, with limited findings confirming the presence of perceptual ratings of moderate strain in children with DS, along with reports on aerodynamic differences reflective of impairment of glottal valving (Pebbili et al., 2019). As reviewed by Kent and Vorperian (2013), studies of perceptual judgments of vocal pitch show a similar disagreement, a result that complicates any effort to characterize the nature of the voice quality in DS even though atypical voice quality is frequently mentioned in the scientific and lay literature.

Vowel Production in DS: Need for Further Research

The source of the differences among the studies cited in the above two sections on vowel production in DS is not clear but may relate to factors such as speech sample, language spoken, age and sex of participants, methods of data collection and analysis, phenotypic variation within the syndrome, and/or the effects of intervention for the speech disorder. Particularly lacking in published studies is the systematic investigation of both the acoustic and perceptual characteristics of the developmental pattern of vowels, including voice, in both males and females with DS. The focus of this study is on developmental changes in vowel acoustics, and it complements perceptual studies using the same words and participants as in the works of Wild et al. (2018) and Kent et al. (2021).

Syndromic Features Related to Vowel Production

Atypical speech production at both the supralaryngeal and laryngeal levels in DS is likely related to craniofacial dysmorphology and/or disordered motor control (including the likelihood of an underlying generalized hypotonia, defined as muscles having deficient tone or tension). Observed craniofacial dysmorphologies in the oral region in DS include relative macroglossia, shortened midface skeleton with smaller facial height, maxillary hypoplasia (including a short and narrow palate), mandibular hypoplasia, and dental malocclusion (Díaz-Quevedo et al., 2021; Kaczorowska et al., 2019; Sforza et al., 2012; Suri et al., 2010; Uong et al., 2001). Dysmorphologies and dysfunctions at the laryngeal level include laryngomalacia (malformation leading to collapse of supraglottic structures during inspiration, causing stridor), tracheal stenosis (Mitchell et al., 2003), and hypotonicity of the vocal musculature (ventricular as well as true vocal folds; Novak, 1972; Pebbili et al., 2019). In explaining the atypical formant patterns in DS, Moura et al. (2008) emphasized the characteristic dysmorphology of the syndrome. They suggested that the smaller ratio of the F2 frequencies for the high vowels /i/ and /u/ (which they named the “DS vocalic anatomical functional ratio” [DS-VR]), reflects maxillary hypoplasia (i.e., decreased oral cavity and pharyngeal space behind the incisors) with the related difficulty in the functional front/back movement of the tongue in the horizontal plane. They proposed that the DS-VR parameter could be useful in monitoring therapeutic effects in children with DS throughout the age range. Alternatively, atypical formant patterns could be the result of a motor speech disorder such as dysarthria or childhood apraxia of speech (CAS; Coêlho et al., 2020; Farpour et al., 2021; Kent & Vorperian, 2021; Kumin, 2006; Rupela et al., 2016; Wilson et al., 2019). For example, vowel centralization, a feature of some dysarthrias, is thought to result from a reduced range of tongue movement in vowel articulation (Kent & Rountrey, 2020) that is commonly quantified as a reduced F1–F2 acoustic VSA, defined as the quadrilateral formed by the four corner vowels. Very little information is available on the higher formants F3 and F4 in speakers with DS, but Lyakso et al. (2021) reported low values of F3 frequency. Palatal dimensions also have been correlated with errors in consonant production in DS (Ferraz & Ghirello-Pires, 2022). As for explaining the atypical voice quality in DS, commonly described to be strained, Pebbili et al. (2019) state that their findings of increased laryngeal aerodynamics (increased subglottal pressure and increased laryngeal airway resistance phonation threshold) in children with DS indicates that there is increased tension of the perilaryngeal muscles. This finding supports the hypothesis that increased tension is used to compensate for the presumed hypotonic musculature as a means to initiate and sustain phonation, resulting in hyperadduction of the ventricular and true vocal folds during phonation.

Because craniofacial dysmorphology and motor speech disorders (dysarthria and/or CAS), with underlying generalized hypotonia, appear to be common in DS, it is difficult to discern their relative effects and interactions. Insight into this problem might be achieved by examining the developmental pattern of vowel fundamental and formant frequencies, also, examining both interspeaker (between) and intraspeaker (within) variability in these frequencies. Presuming that intraspeaker variability reflects precision of motor control, it would be expected that speakers with DS would have greater intraspeaker variability than TD speakers. Furthermore, intraspeaker variability may vary across vowels. Wild et al. (2018) concluded that the perceptual rating of the low vowels /ɑ/ and /æ/ have greater variability than the high vowels /i/ and /u/ likely because the production of the former have decreased sensory feedback, decreased mechanical support (bracing), and/or restricted articulatory distance due to anatomic limitations (smaller mid and lower facial skeleton).

Goals of This Study

This study seeks to understand the nature of age-related changes in vowel production in DS and clarify the nature of disordered vowel production in persons with DS across their life span. The developmental patterns in vowel fundamental and formant frequencies are examined, and data are reported on the frequencies of the fundamental (f _o) and the first four formants (F1, F2, F3, and F4) of the corner vowels in children and adults of both sexes over the age range of 3.5–54 years. Using an identical methodology for data collection and acoustic analysis protocol as established in studies of speech production in TD children (Vorperian et al., 2019) and healthy adults (Eichhorn et al., 2018), we compare the data from speakers with DS by presenting it against the aggregated formant data from the TD speakers in the aforementioned papers (using an overlapping age range). More specifically, this report addresses the following three research questions (RQs) and related predictions:

RQ1. What is the developmental group trajectory of f _o and F1–F4 of the corner vowels in males and females with DS across the ages 3.5–54 years as compared to TD speakers? A general expectation is that the f _o and formant frequencies (F1–F4) are higher in speakers with DS given their smaller stature and therefore shorter vocal tract and smaller larynges. Additionally, we hypothesize that the trajectories for speakers with DS will show vowel-specific deviations as compared with TD speakers.

RQ2. Do speakers with DS show patterns of F2 compression and/or centralization of the acoustic VSA, as defined by the four corner vowels, across speaker-sex and age? Based on earlier reports of anatomic dysmorphology and disturbed speech motor control, we hypothesize that speakers with DS have smaller F1–F2 acoustic VSA (centralization) and that the F2 difference between front and back vowels will be smaller (compression) in speakers with DS than TD speakers regardless of tongue height. That is, the difference in F2 frequency between the high vowels /i/ and /u/ and between the low vowels /æ/ and /ɑ/ will be smaller in DS than TD.

RQ3. What is the pattern of interspeaker (between) and intraspeaker (within) variability in f _o and all four formants (F1–F4) frequencies of the corner vowels across development? Given phenotypic variation and variable patterns of speech motor control, we hypothesize that speakers with DS have greater interspeaker and intraspeaker variability in f _o and all four formants of the corner vowels. Furthermore, we hypothesize that differences in variability are vowel specific, in that both interspeaker and intraspeaker variability are greater for the low vowels /æ/ and /ɑ/ than for the high vowels /i/ and /u/ (given the decreased sensory feedback and decreased mechanical support [tongue bracing] during the production of low vowels).

Method

The University of Wisconsin–Madison Health Sciences Institutional Review Board approved this research.

Participants

The speech recordings in this study were from 374 (293 TD and 81 participants with DS). All were native speakers of English from the Midwest. The TD participants were ages 4–92 years, and participants with DS were ages 3.5–54 years, each with a similar proportion of male and female speakers (total: 193 females and 181 males; TD: 153 females and 140 males; DS: 40 females and 41 males). For additional details on participants with DS, see the work of Wild et al. (2018), and on TD participants, see the works of Eichhorn et al. (2018) and Vorperian et al. (2019). Briefly, the TD participants had no developmental or speech and hearing concerns and no notable observations of atypical orofacial structures or malfunctions. The participants with DS had no concomitant diagnosis; 92% of caregivers reported speech services between the ages 4 and 21 years; hearing assessment reflected a wide range from no hearing loss, to mild, to moderate, and two severe cases with hearing aids. Observations of the participants' orofacial structure and function revealed 37% open mouth posture, 32% underbite, 21% tongue thrust with swallow, and 3% drooling.

We also collected head circumference, neck circumference, body height, and body weight from 98.75% of speakers (seven speakers declined at least one of the measurements). Because individuals with DS typically have a smaller stature than their neurotypical peers, it is likely that they also have shorter vocal tracts and therefore higher formant frequencies. We therefore took the physical measures of body height, weight, head circumference, and neck circumference as proxy for vocal tract length (Fitch & Giedd, 1999) in our analysis to help in the interpretation of formant frequency differences between DS and TD speakers.

The 374 participants contributed 561 recordings where 308 participants were recorded once, and 66 participants were recorded multiple (2–8) times with a minimum of 1-year span between recordings. The number of recordings from participants by speaker type and sex are listed in Table 1, and by speaker recording and age-cohort for each group (DS, TD) in Table 2. The five age cohorts included adults (≥ 20;0 [years;months]) and the following four pubertal stage age cohorts: prepubertal (3;6–7;11), peripubertal (8;0–10;2), pubertal (10;3–14;5), and postpubertal (14;6–19;11), as described in Fitch and Giedd (1999) using the Tanner rating of pubertal stage. For clarity and ease of explanation, from here on, we refer to each recording session per participant as a “speaker recording.” So, the 23 participants who were recorded twice (see Table 1) and provided 46 recordings are counted as 46 speaker recordings in Table 2. To account for data dependence due to repeat recordings, we used mixed-effects statistical modeling as described in the analysis section below.

Table 1.

The number of repeat recordings from female and male TD participants and participants with DS.

	TD participants		Participants with DS		Total
No. of recordings	Female	Male	Female	Male
1	149	130	13	16	308
2	3	10	5	5	23
3	1		6	3	10
4			4	4	8
5			6	8	14
6			3	2	5
7			3	1	4
8				1	1
9				1	1
No. of participants	153	140	40	41	374

Note. TD = typically developing; DS = Down syndrome.

Table 2.

The number of speaker recordings from female and male speakers per four pubertal stage age-cohorts: prepubertal (3;6–7;11), peripubertal (8;0–10;2), pubertal (10;3–14;5), and postpubertal (14;6–19;11), as well as adults (≥ 20;0) for each group (DS, TD).

Age cohort	TD group		DS group		Total
	Female	Male	Female	Male
Prepubertal	23	26	11	21	81
Peripubertal	13	16	14	9	52
Pubertal	32	32	15	19	98
Postpubertal	34	30	26	14	104
Adults	56	46	60	64	226
Total	158	150	126	127	561

Note. DS = Down syndrome; TD = typically developing.

Speech Stimuli

The speech stimuli, also used in the works of Eichhorn et al. (2018), Vorperian et al. (2019), and Wild et al. (2018), consisted of the following five different monosyllabic American English words for each of the four corner vowels: /i/ (bead₂, bee, eat₂, sheep, and feet), /u/ (boo, boot₂, zoo, hoot₂, and shoe), /æ/ (bath, bat₂, cat, hat₂, and sad), and /ɑ/ (dot, hop, pot₂, top, and hot₂). Two of the five words for each vowel, identified with the subscript “2,” were presented and produced twice to assess intraspeaker variability. The decision to have only two productions for this purpose was to keep the task within the limits of the youngest participants and because there is precedent for the use of two productions in the study of speech development (S. Lee et al., 1999). The words were produced in isolation because a carrier phrase could impose additional cognitive and motor demands on the production task. Criteria in word selection were (a) familiarity to younger participants; (b) high phonological neighborhood density, which reportedly maximizes F1–F2 vowel space (Munson & Solomon, 2004); and (c) phonetic composition with no consonant clusters to facilitate production by young speakers and/or speakers with speech motor disorders.

Recording Protocol

Recording was done in a quiet room using a Shure SM48 microphone (Shure Inc.) mounted on a floor stand and adjusted to each participant's seated height at a 15-cm distance and 45° angle laterally from the mouth. The microphone was connected to a Marantz-PMD 660 digital audio recorder (Marantz Professional in Music Brands, Inc.) that digitizes at a rate of 48 kHz with a 16-bit resolution on a SanDisk Ultra II flashcard (SanDisk Western Digital Corporation). To optimize recording level, the Marantz recorder gain was adjusted to 6–12 dB below the maximum level. The stimuli were presented visually (picture and orthographic word) and aurally (recordings from an adult male—with a f _o of 110 Hz, from the Midwest, i.e., same regional dialect as where the participants were recruited from—were played through external speakers) using a laptop with the TOCS+ platform program (Hodge et al., 2009) for randomization. Speakers were instructed to repeat the speech stimuli (28 words total) at a normal loudness level, with two practice words at the beginning. This study used a combination of methods (visual and aural) for stimulus presentation that were originally designed to increase the likelihood of participation by young children with potential limitations in attention span, as well as potential limitations in cognitive, sensory, and motor functions. These procedures were used successfully in a study of speech intelligibility in children and adults with DS (Wild et al., 2018). Applying the same procedures with all participants, both children and adults, permits the comparability of data across speakers with and without developmental delay or disorder.

Acoustic Analysis and Measurements

The acoustic analysis procedures were the same as those used on TD participants by Eichhorn et al. (2018) and Vorperian et al. (2019), to permit comparison, and were based on a review of methods to optimize formant analysis (Kent & Vorperian, 2018). Briefly, the word recordings were uploaded to a computer, and the waveforms were segmented using Praat (Version 5.1.31, Boersma & Weenink, 2010) and saved as separate sound files. An upgraded version of TF32 (time–frequency analysis software for 32-bit Windows; Milenkovic, 2010) was used to measure the frequencies of f _o and formants F1–F4. TF32 does not degrade the signal through downsampling, has a linear predictive coding (LPC) formant-track overlaid on a grayscale spectrogram for visual inspection of formant patterns (along with a pitch track), and has a time-slice spectrum linked to the spectrogram that displays fast Fourier transform (FFT) and LPC spectral slice information (both of which were used to verify accuracy of formant frequency estimation as needed). The number of LPC coefficients was adjusted to achieve an optimum analysis of formants, and both wide-band and narrow-band spectrograms were inspected in cases of suspected formant–harmonic interaction, as may occur with voices of high f _o. Vowel-specific measurement point/inflection points, as defined in the works of Derdemezis et al. (2016) and Kent and Vorperian (2018), were as follows: vowel /i/, point of highest frequency of F2; vowel /u/, point of lowest frequency of F2; vowel /ɑ/, point of least separation between F1 and F2 frequencies; and vowel /æ/, point of most evenly spaced formants, while avoiding measurement at a point of decreasing F2–F1 difference (which reflects backing of the vowel). The frequencies of the first four formant frequencies were estimated by inspecting the spectrogram (with overlaid LPC formant tracks), FFT spectral slice (with zoom-in function to improve accuracy), and cepstrum. Parameter manipulations to optimize the spectrogram for acoustic analysis included adjustments of analysis bandwidth of FFT spectrograms, dynamic range, and number of coefficients on the time-slice LPC spectrum. Formant measurements that could not be reliably estimated or appeared to have extreme values (outliers) were reconsidered using a consensus analysis approach where two or three individuals experienced in the acoustic analysis of speech examined the spectrogram and time-slice spectrum displays to determine formant frequencies. If they could not come to an agreement or decided that the formant was not suitable for measurement, no measurement was taken (i.e., treated as missing data). The mean frequency (f _o and formant F1–F4) measures of the five words produced by each of the speakers per recording session was used for analysis and display. Given the ambitious goal of completing acoustic analysis on TD speakers (Eichhorn et al., 2018; Vorperian et al., 2019) and speakers with DS across the life span, a training protocol was completed by all raters to ensure that consistent methods and criteria were used across the seven raters involved throughout the 12-year period of data collection. This protocol was based on a subset of recordings from randomly selected participants that included male and female adults and children from TD speakers and speakers with DS. Training of the raters, who had completed basic coursework in speech science and speech acoustics, was done individually by a highly experienced speech scientist (R.D.K.) with 50 years of experience analyzing typical and atypical speech. Training materials included select readings and step-by-step instructions on use of the software for acoustic analysis along with detailed suggestions and examples on how to address challenges in formant estimation. Much of this information is detailed in the work of Kent and Vorperian (2018). The reliability of each rater was formally assessed prior to contributing to the DS and TD database measurements. In addition, reliability across raters was assessed using measurements from all raters of eight participant recordings (four male and four female) at different ages. Reliability was assessed by calculating the intraclass correlation coefficient (ICC) using the statistical package SPSS Version 25 (SPSS Inc.; with the value 1 being the maximal possible reproducibility of the frequency measurements). As expected, interrater reliability varied with vowel and formant frequency with an overall range of .8777–.986 ICC calculations. Reliability was excellent (ICC range: .90–.99) for nearly all vowel/formant combination, with the exception of F4 of /u/ (ICC = .88), and F4 of /æ/ (ICC = .88) where reliability remained good. Findings are consistent with those reported in the works of Eichhorn et al. (2018) and Vorperian et al. (2019). A likely reason for the poorer reliability of F4 measurements is that it is a harder formant to estimate, particularly for the vowel /æ/ that often was produced with shifting formant frequencies that made it difficult to identify a time point that was most representative of the overall formant pattern.

Statistical Analysis

Missing data rates were less than 1%–2% for f _o and F1–F3 values but up to 5% for some F4 values, which are not likely to affect results (Scheffer, 2002). In cases of missing data, we included observations in the trajectory model if there was at least one formant measurement for at least one word for a given corner vowel, and we included observations in the intraspeaker variability model if there was at least one pair of repeated formant measurements for at least one word.

To address RQ1, we used mixed-effects fractional polynomial regression (Royston & Altman, 1994) to model the developmental trajectories for each corner vowel f _o and formant (F1–F4) frequencies by sex and age while accounting for the repeat recordings. We modeled both first and second moments. The first-moment model was used to construct sex/age-indexed mean trajectories for data from the TD and DS groups. The second-moment model was fit to the TD data only (due to the limited sample size of the DS group since the stability of the higher order model is more sensitive to the sample size than the lower order model). This served the dual purpose of reweighting the first-moment model to account for heteroscedasticity and providing a standard deviation estimate for construction of 90% normal ranges for the TD data (Altman, 1993).

The first-moment model was as follows:

$${\overline{y}}_{\mathit{ij}}={\varphi }_1\left(\mathit{ag}{e}_{\mathit{ij}};{\mathit{\beta }}^{\mathit{age}},{\mathit{p}}^{\left(1\right)}\right)+g_i{\varphi }_1\left(\mathit{ag}{e}_{\mathit{ij}};{\mathit{\beta }}^{\mathit{dx}\mathit{age}},{\mathit{p}}^{\left(1\right)}\right)+\mathit{se}{x}_i{\varphi }_1\left(\mathit{ag}{e}_{\mathit{ij}};{\mathit{\beta }}^{\mathit{sex}\mathit{age}},{\mathit{p}}^{\left(1\right)}\right)+u_i^{\left(1\right)}+{ϵ}_{\mathit{ij}}^{\left(1\right)},$$ (1)

with fixed effects for each group $g_i$ (i.e., TD or DS), sex, age, and their pairwise interactions, and a per-speaker recording random effect $u_i^{\left(1\right)}$ . The response ${\overline{y}}_{\mathit{ij}}$ was the average f _o or formant frequency across all five words (with measurements from the first production of words produced twice) for the corner vowel assessed in speaker $i$ and repeat visit $j$ . The linkage function for age was a two-power piecewise fractional polynomial kernel:

$${\varphi }_1\left(\mathit{age};\mathit{\beta },{\mathit{p}}^{\left(\mathrm{1}\right)}\right)={\beta }_0+{\beta }_1\mathit{ag}{e}^{p_1^{\left(1\right)}}+{\beta }_2\mathit{ag}{e}^{p_2^{\left(1\right)}}+{\beta }_3{{\left(\mathit{age}-\mathit{ag}{e}_0\right)}_{+}}^{p_2^{\left(1\right)}},$$ (2)

where ${\left(\dots \right)}_{+}=max(\dots ,0$ ). We added the peri-to-pubertal knot term (breakpoint), ${\left(\dots \right)}_{+}$ , the higher power knot at $\mathit{ag}{e}_0=122.4$ months (the upper-end age of the peripubertal age-cohort, and at the cusp of transitioning to the pubertal age-cohort), to provide a more flexible fit that adapts to the rapid change occurring during the pubertal period. This was based on our examination of initial knot-free fits where we observed a sharp decrease in f _o and multiple formants across vowels at this time-point particularly for the DS data. The powers ${\mathit{p}}^{\left(k\right)}=\left\{p_1^{\left(k\right)},p_2^{\left(k\right)};p_1^{\left(k\right)}<p_2^{\left(k\right)}\right\}$ were selected from the candidate powers from the set $\mathrm{P}\mathrm{=}\left\{-2,-1,-\mathrm{0.5,0},\mathrm{0.5,1},2,3\right\}$ , with the convention that $x^{p=0}=log\left(x\right)$ . All pairs of powers were considered, with an initial unweighted model selected by minimizing deviance. Subsequently, we used an iterative reweighting scheme, whereby the normalized absolute residuals $s_{\mathit{ij}}=\sqrt{\pi /2}\left|{ϵ}_{\mathit{ij}}^{\left(1\right)}\right|$ from the first-moment model for TD speakers were fit to the second-moment fractional polynomial model:

$$s_{\mathit{ij}}={\varphi }_2\left(\mathit{ag}{e}_{\mathit{ij}};{\mathit{\beta }}^{\mathit{age}},{\mathit{p}}^{\left(2\right)}\right)+\mathit{se}{x}_i{\varphi }_2\left(\mathit{ag}{e}_{\mathit{ij}};{\mathit{\beta }}^{\mathit{sex}\mathit{age}},{\mathit{p}}^{\left(2\right)}\right)+u_i^{\left(2\right)}+{ϵ}_{\mathit{ij}}^{\left(2\right)},$$ (3)

also with powers ${\mathit{p}}^{\left(2\right)}$ selected by minimizing deviance across all possible pairs of candidate powers; for this model, the kernel ${\varphi }_2\left(\dots \right)$ was the same as ${\varphi }_1\left(\dots \right)$ with the peri-to-pubertal knot term, ${\varphi }_1{\left(\dots \right)}_{+}$ excluded. Fitted standard deviations from this model were used to reweight the mean model iteratively until the coefficients converged. Overall, the reweighting procedure had a qualitatively small effect on the final fit.

In addition, to visualize the frequency differences be?>tween TD and DS, we used the final fitted models to predict trajectories of first moments for the TD and DS groups as functions of sex and age, namely, $\hat{\mu }\left(\mathit{TD},\mathit{se}{x}_i,\mathit{ag}{e}_{\mathit{ij}}\right)$ and $\hat{\mu }\left(\mathit{DS},\mathit{se}{x}_i,\mathit{ag}{e}_{\mathit{ij}}\right)$ and the predicted trajectory of the second moment for the TD group, ${\hat{\sigma }}_{\mathit{TD}}\left(\mathit{sex},\mathit{age}\right)$ . These TD prediction functions were used to construct 90% normal ranges for the TD group:

$$\hat{\mu }\left(\mathit{TD},\mathit{se}{x}_i,\mathit{ag}{e}_{\mathit{ij}}\right)\pm 1.645{\hat{\sigma }}_{\mathit{TD}}\left(\mathit{sex},\mathit{age}\right),$$ (4)

and to calculate z scores for all speakers:

$$Z_{\mathit{ij}}=\frac{{\overline{y}}_{\mathit{ij}}-\hat{\mu }\left(\mathit{TD},\mathit{se}{x}_i,\mathit{ag}{e}_{\mathit{ij}}\right)}{{\hat{\sigma }}_{\mathit{TD}}\left(\mathit{se}{x}_i,\mathit{ag}{e}_{\mathit{ij}}\right)}.$$ (5)

We plotted the normal ranges, mean trajectories, and z scores by age and sex, for each corner vowel for f _o and all four formant combinations. Though we fit the models to data for all available ages, plotted trajectories and z scores were limited to the range 5–40 years, due to high uncertainty and poor fit at the extremes of the age range where data were limited. The resulting visuals are displayed in Figures 1 and 2.

Figure 1.

(a)–(d) Display of the vowel mean (first moment) frequency of each speaker recording for f _o and first-to-fourth formants (F1–F4) for each of the four corner vowels: (a) /i/, (b) /u/, (c) /æ/, and (d) /ɑ/. Male speakers are shown in the left panel in blue, and female speakers are shown in the right panel in red. Open bold symbols denote speakers with DS, and faint filled symbols TD speakers. For each frequency, we fit a linear mixed-effect fractional polynomial model, with knot (breakpoint where the trajectory changes) at age 10;2 (years;months; see text for explanation), that is displayed using dashed and solid lines for speakers with TD and DS speakers, respectively. The red (female) and blue (male) shaded regions reflect the 90% range for the TD group. Model fitted to data for all ages but plotted trajectories are limited to the age range 5–40 years (see text for explanation). Vertical lines delineate the four pubertal-stage cohorts (years;months): prepubertal (3;6–7;11), peripubertal (8;0–10;2), pubertal (10;3–14;5), and postpubertal including adults (14;6–19;11 and 20;00–54;0). f _o = fundamental frequency; DS = Down syndrome; TD = typically developing.

Figure 2.

(a)–(d) Display of the z score (standardized with fitted TD mean and variance trajectory) of speaker recordings for the DS (open bold circles) and TD (faint filled circles) speaker groups for f _o, F1 to F4 for each of the corner vowels: (a) /i/, (b) /u/, (c) /æ/, and (d) /ɑ/. The DS z-score trajectory with breakpoint at age 10;2 (see text for explanation) is displayed with a thick solid line against the TD trajectory, which, by definition, is the horizontal black line corresponding to z = 0. The trajectories are limited to the age range 5–40 years, but the model fitted to data for all available ages (see text for explanation). The shaded region covers 90% of the range of TD group data. Males (blue) are shown in the left panel, and females (red) are shown in the right panel. See Figure 1 caption for a description of the vertical lines. f _o = fundamental frequency; DS = Down syndrome; TD = typically developing.

To further examine RQ1, in terms of DS and TD differences in developmental first-moment trajectories when the physical characteristics of the DS phenotype are accounted for, we used an unweighted first-moment model similar to Equation (1) but with the linkage function, $\mathrm{\Phi }$ (…), including all powers in the candidate set, $\mathrm{P}$ to increase the model flexibility. The $p$ values were calculated by a likelihood ratio test (LRT) of $H_{0\left(1\right)}:{\mathit{\beta }}^{\mathit{dx}\mathit{age}}=\mathrm{0}$ . Tests were performed using both an unconditional and a conditional model with covariate adjustment for key physical measurements (height, weight, head circumference, and neck circumference). Bonferroni correction at $\alpha =\frac{0.05}{20}=0.0025$ was used to account for all combinations of four vowels and five frequencies for each tested model. The resulting LRT $p$ values are listed in Table 3.

Table 3.

The $p$ value of the likelihood ratio test of the mixed-effects model comparing the overall TD versus DS frequency developmental trajectories with and without accounting for the physical characteristics of the DS phenotype as specified in the text.

	$\mathit{p}$ values for differences between DS and TD speaker groups
	Adjusted for age and sex only				Adjusted for age, sex, and physical characteristics
Frequency	/i/	/u/	/æ/	/ɑ/	/i/	/u/	/æ/	/ɑ/
f o	.0000	.0000	.0007	.0009	.0030	.0006	.0118	.0036
F1	.0000	.0002	.0000	.0044	.0000	.0079	.0000	.0085
F2	.0326	.0000	.0000	.0000	.0111	.0000	.0000	.0000
F3	.0366	.4931	.0056	.0001	.1283	.3517	.0205	.0001
F4	.0024	.1561	.1792	.1223	.0226	.0550	.1486	.0547

Note. p values smaller than the Bonferroni-corrected significance value of .0025 are shown in bold. TD = typically developing; DS = Down syndrome; f _o = fundamental frequency.

To address RQ2, we first examined F2 compression, that is, the difference in F2 frequency measurements (raw data) between the high vowels and the low vowels. We also examined the three vowels in Table 3 with significant differences in F2 frequency trajectories between DS and TD. To determine whether the differences between the groups are present before and/or after the peri-to-pubertal knot, we conducted a series of likelihood tests using the model:

$${\overline{y}}_{\mathit{ij}}={\beta }^{g1}{g}_{i}I\left({\mathit{age}}_{\mathit{ij}}<{\mathit{age}}_0\right)+{\beta }^{g2}{g}_{i}I\left({\mathit{age}}_{\mathit{ij}}\ge {\mathit{age}}_0\right)+{\varphi }_2\left({\mathit{age}}_{\mathit{ij}};{\mathit{\beta }}^{\mathit{age}},{\mathit{p}}^{\left(3\right)}\right)+{\mathit{sex}}_i{\varphi }_2\left({\mathit{age}}_{\mathit{ij}};{\mathit{\beta }}^{\mathit{sex}\mathit{age}},{\mathit{p}}^{\left(3\right)}\right)+{u_i}^{\left(1.1\right)}+{{\in }_{\mathit{ij}}}^{\left(1.1\right)}\mathrm{.}$$ (6)

For each vowel, we compared the unconstrained model (the model without any constraints, i.e., ${\beta }^{g1}$ and ${\beta }^{g2}$ could be freely estimated) to the model with one of the three constraints, $H_{0\left(\mathrm{1.1,1}\right)}:{\beta }^{g1}=0$ , $H_{0\left(\mathrm{1.1,2}\right)}:{\beta }^{g2}=0$ , and $H_{0\left(\mathrm{1.1,3}\right)}:{\beta }^{g1}={\beta }^{g2}$ each at the α = 0.05 level using the LRT. Selection of a model with one of the three constraints would identify the period during which the trajectories between TD and DS are significantly different. Selection of the model with the first constraint would indicate that the differences occur after the peri-to-pubertal knot; the second constraint would indicate that the differences occur before the peri-to-pubertal knot; or the third constraint would indicate that the differences remain constant throughout the age period. We selected the constrained model that was not significantly different from the unconstrained model (with p > .05) since that implies that the fit is as good as the unconstrained model. If more than one model with constraint had p > .05, the constrained model with the greatest p value (i.e., least significant p value) was selected (an approach that is identical to choosing the one with the smallest Bayesian Information Criterion in our analyses with the same number of estimated parameters and sample size). In addition, to examine the relationship of F2 differences with age between the front/back high versus low vowels, we performed a similar model selection process using the dependent variable ${\overline{y}}_{\mathit{ij}{D}_v}$ , the absolute mean difference of the F2 frequency between the high (/i/ and /u/) or low (/æ/ and /ɑ/) vowels for speaker recording $i$ at repeat visit j. The resulting $p$ values are listed in Table 4.

Table 4.

The p values of the likelihood ratio test of the mixed-effects model that captures the age-period of TD/DS differences of the vowels with significant F2 (shown in bold in Table 3) with respect to the peri-to-pubertal knot at about age 10 years.

	TD/DS group differences pattern for F2
Group differences	/i/	/u/	/æ/	/ɑ/	|/i/−/u/|	|/æ/−/ɑ/|
Before	—	.0004	.8700	.0000	.2160	.7833
After	—	.0018	.0320	.0008	.0017	.0000
Constant	—	.9944	.0002	.9560	.6158	.0000

Note. The model selected is the one with the largest p value (per vowel F2 or vowel F2 differences), shown in bold, indicating when the TD/DS F2 differences are most evident with respect to the peri-to-pubertal knot (before, after, or constant/before and after). Dash indicates not tested. TD = typically developing; DS = Down syndrome.

Next, we examined differences in acoustic VSA between DS and TD for each of the five age-cohorts using the mixed-effects model to account for the repeated recordings. No control variables were included in the models. For each group, we first obtained the mean frequency and standard deviation per age-cohort for each vowel (summarized in Appendix A) and then calculated the acoustic VSA for each of the five age cohorts. Using the mean F1–F2 frequencies, planar area was calculated using the formula for the area of an irregular quadrilateral specified as follows:

$$\text{Area}=0.5\times /\mathrm{i}/\mathrm{F}2\times /\mathrm{}/\mathrm{F}1+/\mathrm{}/\mathrm{F}2\times /\mathrm{ɑ}/\mathrm{F}1+/\mathrm{ɑ}/\mathrm{F}2\times /\mathrm{u}/\mathrm{F}1+/\mathrm{u}/\mathrm{F}2\times /\mathrm{i}/\mathrm{F}1-/\mathrm{i}/\mathrm{F}1\times /\mathrm{}/\mathrm{F}2+/\mathrm{}/\mathrm{F}1\times /\mathrm{ɑ}/\mathrm{F}2+/\mathrm{ɑ}/\mathrm{F}1\times /\mathrm{u}/\mathrm{F}2+/\mathrm{u}/\mathrm{F}1\times /\mathrm{i}/\mathrm{F}2$$ (7)

where Fn = the formant number for the vowel symbol shown in the virgules; for example, /i/ F2 is the second formant for vowel /i/. The resulting acoustic VSA and $p$ values are listed in Table 5.

Table 5.

Acoustic vowel space area (VSA; Hz²) per age-cohort.

Age cohort	Speaker sex	DS	TD	p
Prepubertal	F	642891.9	987812.1	.0013
	M	555040.9	935368.3	.0029
Peripubertal	F	612018.7	829709.2	.0016
	M	633565.4	732428.1	.3237
Pubertal	F	441638.4	641909.2	.0003
	M	415498.8	586681.2	.0525
Postpubertal	F	437545.4	573960.2	.0055
	M	356716.5	366827.5	.9267
Adults	F	472906.1	636989.0	.0005
	M	387617.2	364886.9	.6074

Note. Pairs of DS and TD groups were examined using the mixed-effects model. VSA values that are significantly different (Bonferroni-corrected p < .005) are shown in bold. DS = Down syndrome; TD = typically developing; F = female; M = male.

Finally, to address RQ3 on variability, we first examined interspeaker (between) variability by comparing the variance of z scores between TD and DS using F tests. To adjust for multiple comparisons, a Bonferroni correction of $\alpha =\frac{0.05}{20}=0.0025$ was used. The resulting $p$ values are listed in Table 6. Next, we fit a coefficient of variation (CV) estimate to assess intraspeaker (within) variability by sex and age using similar mixed-effects fractional polynomial regressions for the first and second moments with an iterative reweighting procedure. In this case, the mean model excluding the peri-to-pubertal knot term was used to construct sex/age-indexed CV-trajectories; the variance model was used for reweighting only. For each corner vowel f _o or formant frequencies, the response used was the appropriately normalized mean observed difference between duplicate utterances of words corresponding to the target vowel:

$$c_{\mathit{ij}}=\frac{\sqrt{\pi }}{2}\frac{1}{n_w}{\sum }_{w=1}^{n_w}\frac{\left|y_{\mathit{ijw}1}-y_{\mathit{ijw}2}\right|}{{\overline{y}}_{\mathit{ijw}}},$$ (8)

for $y_{\mathit{\text{ijwk}}}$ the $k$ th utterance of the $w$ th word for speaker $i$ at repeat visit $j$ . Lastly, we performed the LRT to examine the presence of overall TD and DS differences. The resulting visuals are displayed in Figure 3 with embedded LRT $p$ values.

Table 6.

Variance of the z scores for the two groups of speakers, followed by the p value of the F test assessing interspeaker variability between the two groups.

Variance of z scores for TD and DS groups and p value of F test comparing the two groups
Frequency	/i/			/u/			/æ/			/ɑ/
	TD	DS	p	TD	DS	p	TD	DS	p	TD	DS	p
f o	0.96	1.04	.6191	1.04	1.07	.8688	1.01	1.26	.1834	1.00	0.94	.7231
F1	1.07	1.44	.0763	0.98	1.25	.1572	1.06	2.04	.0001	1.01	1.61	.0060
F2	0.99	1.77	.0006	1.07	1.64	.0112	1.01	1.24	.2383	0.98	1.69	.0013
F3	0.98	1.11	.4408	0.98	1.52	.0099	1.03	1.53	.0186	1.02	1.84	.0004
F4	0.95	1.01	.6696	1.02	0.96	.7493	1.03	1.16	.4851	0.91	0.97	.7266

Note. Pairs of significant variance differences between the two speaker groups (p < .0025) are shown in bold. TD = typically developing; DS = Down syndrome; f _o = fundamental frequency.

Figure 3.

(a)–(d). Display of coefficient of variation estimate to assess intraspeaker variability in speakers with DS (solid lines) as compared to TD speakers (dashed line) for each frequency (f _o, F1 to F4) of each corner vowels: (a) /i/, (b) /u/, (c) /æ/, and (d) /ɑ/. Male speakers are shown in the left panel (blue), and female speakers are shown in the right panel (red). The $p$ values of the overall likelihood ratio test comparing TD and DS differences are shown in the left panel with significant $p$ values denoted by an asterisk (*); p ≤ .0025 is the Bonferroni-corrected α = .05 level of significance. See Figure 1 caption for a description of the vertical lines. f _o = fundamental frequency; DS = Down syndrome; TD = typically developing.

Results

Results are presented in the sequence of the three RQs: (RQ1) developmental trajectories for f _o and formants F1–F4, (RQ2) F2 compression and vowel centralization, and (RQ3) interspeaker and intraspeaker variability in f _o and formants F1–F4. RQ2 is an in-depth exploration of select formant patterns, specifically the first two formants as related to articulatory working space. The results are presented sequentially below.

Developmental Trajectories in TD and DS Speaker Groups: f _o and Formant Frequencies

Figures 1a–1d display for each vowel the mean frequency (f _o and formants F1–F4) measure of the five words per speaker recording (DS [open bold symbols], TD [faint filled symbols]) as a function of age, with males displayed in the left panel and females in the right panel. The figures also display the fitted lines, with knot term, hereafter referred to as breakpoint and as described in the previous section, showing the TD and DS estimated mean trajectories by age from the first-moment model. The shaded areas show the 90% normal ranges for the TD group, based on the second-moment model as described in the previous section.

Figures 2a–2d provide a more direct vowel-specific comparison, using the z score of speaker recordings to display the DS (open bold symbols) against the TD (faint filled symbols). The plots display the z scores for f _o and formants F1–F4 for each of the corner vowels. The DS z-score trajectory is displayed with a thick solid line against the TD trajectory, which, by definition, is the horizontal line corresponding to z = 0, with the 90% range of the TD group represented by the upper and lower shaded regions.

Taken together, and as described in the following sections, Figures 1a–1d give an overview of sex- and age-related changes in f _o and the four formants F1–F4 for each corner vowel while capturing differences between TD and DS. In general, the display reveals the expected developmental trend of decreased f _o and formant frequencies from ages 5 to 40 years across all vowels in male and female speakers, for both TD and DS. The display also shows the expected differences in frequency between the two sexes including the higher formants F3 and F4 differences, and reveals the pronounced frequency drop during puberty in males for both TD and DS. Figures 2a–2d similarly display the differences between the two groups with greater clarity and detail for the f _o and formants F1–F4 (RQ1), while also displaying the increased interspeaker variability in DS (RQ3). As described in greater detail below, the displays provide insight on developmental differences in frequencies between the two groups with f _o findings confirming expectations but not formant frequencies across the 5 decades examined.

Developmental Trajectories: f _o

As noted above, the f _o decreased across the 5-decade age range studied for all vowels in both male and female speakers, across both TD and DS, with a pronounced peri-to-pubertal drop in males. The higher f _o for speakers with DS than TD speakers, as displayed in Figures 1 and 2, was confirmed to be significant using the LRT for all vowels. Table 3 displays the p values for the two comparisons implemented, as described in the Statistical Analysis section, where LRT was computed once while accounting for age and sex only, and again while accounting for age, sex, and physical characteristics of the DS phenotype (height, weight, head circumference, and neck circumference). Across all vowels and in both sexes, the f _o frequency was consistently higher for DS than TD across age (p < .001 for all corner vowels); this was expected given the smaller stature in speakers with DS and therefore shorter vocal tract and smaller larynges. As displayed in Figures 2a–2d, close examination of the f _o z scores and trajectories in the DS group (open bold circles, thick solid line) against TD (faint filled circles, with the horizontal black line corresponding to z = 0) revealed the DS f _o z-score trajectory to remain above the TD zero line across age in both males (blue, left panel) and females (red, right panel) for all four vowels. However, once the physical characteristics of the DS phenotype were accounted for in the LRT computation, the f _o difference dissipated and was no longer significant for three of the four vowels. The high vowel /u/ was an exception. Noteworthy of mention, as displayed in Figures 2a–2d, is that approximately between the ages 8 and 15 years (i.e., the peripubertal and pubertal age-cohorts), the differences between the TD and DS z-score trajectories were small, particularly in males. These plots also display the interspeaker variability as discussed below.

Developmental Trajectories: Formant Frequencies

Consistent with the expectations listed in RQ1, and similar to f _o, both speaker groups had a consistent decrease in all formant frequencies across age for all four vowels. Furthermore, the drop persisted and was steeper after the breakpoint particularly for the higher formants F3 and F4. Unlike f _o, there was not a consistent pattern of differences in all four formant frequencies between the two speaker groups. However, differences between the groups were apparent in the lower frequencies (F1 and F2) while differences in the higher formants were limited especially for F4. As seen in Figures 1 and 2, the size and direction of differences between the two groups varied across formant, vowel, sex, and age. Testing for TD and DS differences in overall developmental first-moment trajectories, using LRT, is summarized in Table 3 with Bonferroni-corrected significant $p$ values marked in bold. Findings indicated that TD and DS were significantly different in the F1 and F4 trajectories for the vowel /i/. The differences are displayed in Figures 1a and 2a where the F1 for DS was higher than TD across all ages. However, the F4 values of the youngest TD and DS were similar up to age 8 years, following which differences emerged during puberty and persisted throughout the age range examined with DS having higher fourth formant frequencies than TD. When controlling for the DS phenotype physical characteristics, the F1 difference remained significant but the F4 difference did not (p = .0226). For the vowel /u/, overall significant differences were present for F1 and F2 between the two speaker groups. As displayed in Figures 1b and 2b, the F1 and F2 values were higher in DS as compared to TD across all ages. When the physical characteristics were controlled, the F2 difference remained significant but the F1 difference did not (p = .0079). For the low vowel /æ/, the F1 and F2 trajectories of DS versus TD were significantly different. As displayed in Figures 1c and 2c, TD and DS had similar F1 values until age 10 years, following which differences emerged with DS having lower values than TD. In contrast, the F2 values of DS increased across age, with F2 values being lower before age 10 years but higher after age 10 years as compared to TD. These differences persisted when controlling for physical characteristics. Finally, for the low back vowel /ɑ/, the F2 and F3 trajectories of TD and DS were significantly different. As displayed in Figures 1d and 2d, DS had a higher F2 trajectory than TD across all age cohorts, a finding similar to the back vowel /u/. The F3 trajectory of DS was similarly higher than the TD before age 10 years but decreased abruptly after 10 years. These significant differences in F2 and F3 trajectories persisted even after accounting for the physical characteristics of the DS phenotype in the analysis.

In summary, although a consistent pattern of differences in formant frequencies was not evident between the two groups, examination of Figures 1 and 2 combined with the LRT results (summarized in Table 3 and described above) helped identify overall similarities and differences. Differences were most distinct in the lower formant frequencies F1 and F2. Specifically, overall vowel-specific formant differences were noted between the two groups where the F1 measurements were significantly higher in DS (as expected) but only for the high vowels /i/ and /u/ (with the significance for the difference in /i/ not dissipating when the physical characteristics of the DS phenotype were accounted for). As for F2 with significant LRT p values that persisted even after accounting for DS phenotype differences (see Table 3), closer examination of the F2 trajectories and z scores revealed vowel- and age-specific differences between the two groups that the overall LRT did not capture. For example, the F2 trajectory of DS was lower for the front vowel /æ/ but only during the first two age cohorts (i.e., before the breakpoint), unlike the F2 for the back vowels /u/ and /ɑ/ that were significantly higher across all age cohorts. This observation of F2 differences specific to select age-period warranted additional analyses and is addressed in the following section. As for the higher formants F3 and F4, once the DS phenotype physical characteristics was accounted for, no significant differences were present between the two groups except for F3 for the vowel /ɑ/ with a developmental trend of being higher than TD before breakpoint.

Formant Frequencies–F2 Compression and Vowel Centralization

F2 compression was assessed by examining F2 measurement differences between the high and low vowels for DS versus TD, as well as the three vowels with significant F2 differences between the two groups in Table 3. Findings from conducting a series of likelihood tests, using the model described in the Method section to determine whether differences between the two groups were present before and/or after the breakpoint, are shown in the last two columns of Table 4. The selected model representing the age period that TD and DS have differences are denoted in bold (the model with the least significant p value). As displayed in Figures 2c and 2d, respectively, prior to the breakpoint/puberty, speakers with DS produced /æ/ with a significantly lower F2; and /ɑ/, with a significantly higher F2 frequency as compared to TD speakers, with the effect of reducing the F2 distinction between these two vowels. After the breakpoint, the F2 frequency for vowel /æ/ was similar to (or even higher than) TD speakers, while F2 for vowel /ɑ/ remained elevated as compared to TD speakers. Similar to the back vowel /ɑ/, the back vowel /u/ also showed a consistent and significantly elevated F2 for DS as compared to TD across age (see Figures 2b and 2d). As expected, the absolute differences in F2 frequency measurements between the high vowels (|/i/−/u/|) and low vowels (|/æ/−/ɑ/|) exhibited different patterns between TD and DS, with DS exhibiting smaller differences in the formant frequencies between vowels in each pair, compared to TD. For the high vowels, this significant difference appeared to be constant before and after the breakpoint, with the estimated TD/DS F2 difference being 187 Hz. For the lower vowels, however, the difference was significant only before the breakpoint with an estimated TD/DS F2 difference of 253.5 Hz.

Next, we examined vowel centralization, using acoustic VSA for males and females in each of the five age-cohorts. Findings, as summarized in Table 5, indicate that VSA differences between TD and DS depended on age and speaker sex. While significant differences in VSA were present between TD and DS in both males and females during the prepubertal period, the differences remained significant only for females across the remaining age cohorts with borderline significance during the postpubertal period with VSA smaller in DS than TD. The differences between the two groups in males were not significant (likely due to interspeaker variability in F1 and F2 as described in the following section), but as seen in Table 5, in general, the VSA in males was larger for TD than DS. Furthermore, the average male and female area was commonly larger for TD than DS, confirming a general trend of vowel centralization for DS.

Interspeaker and Intraspeaker Variability in TD and DS Speaker Groups: f _o and Formant Frequencies

Overall findings on RQ3 on variability, discussed in greater detail in the following subsections, revealed DS to have no significant interspeaker variability for f _o but highly significant intraspeaker variability as compared to TD. As for formant frequencies, DS generally had increased interspeaker and intraspeaker variability where the former (between) showed significant differences for select vowels and formant frequencies only, while differences in the latter (within) was highly significant for all vowels and all lower formants as well as almost all higher formants, as compared to the TD.

Interspeaker and Intraspeaker Variability: f _o

As displayed in Figure 2, in addition to the above noted higher f _o z scores in DS, the f _o variance values listed in Table 6 were larger for DS than TD. The p values of the F test (listed in the last column for each vowel) comparing the two groups, however, indicated that f _o interspeaker variability was not significant. LRT results comparing the two groups for intraspeaker (within) variability, on the other hand, confirmed that DS had significantly increased variability as compared to TD for all vowels (see LRT p values embedded in Figures 3a–3d), in both males and females across most vowel frequencies.

Figures 3a–3d, displays the fitted trajectories of intraspeaker variability, as measured by the CV, for DS (solid lines) and TD (dashed lines) as a function of age for each vowel in males (left panel, in blue) and females (right panel, in red). Findings show that f _o for DS displayed significantly increased intraspeaker variability compared to TD across all vowels, all ages, and both sexes. Figure 3 also shows that intraspeaker f _o variability was more before the onset of puberty for all vowels except for /u/ that displayed increased variability post puberty. Furthermore, f _o variability increased after age 25 years across all vowels.

Interspeaker and Intraspeaker Variability: Formant Frequencies

As described in the previous f _o section, Figures 2a–2d also display that for most vowel/formant combinations, DS had numerically larger interspeaker variability in formant z scores as compared to TD. This is summarized in Table 6, indicating that, in general, DS had greater formant variance for all vowels and formants but the p values of the F tests listed in the last column for each vowel indicated that differences in interspeaker variability were significant, at the Bonferroni-corrected α = .0025 level, only for F1 of vowel /æ/, F2 of vowel /i/ and /ɑ/, and F3 for vowel /ɑ/.

Figures 3a–3d include the p values of the LRT comparing TD and DS intraspeaker variation for each vowel/frequency combination. In general, the differences were statistically significant (p < .0025) as denoted by an asterisk, except for three instances: F3 for /i/ (p = .17), and F4 for /u/ (p = .003) and /æ/ (p = .06). As seen in Figure 3, for all vowels and formants, the CV decreased as age increased in both TD and DS, up to postpuberty and young adulthood; beyond that, the TD speakers' CVs continued to decrease, but the CV for DS increased for some vowels and formants such as F2 for the front low vowel /æ/.

Discussion

The nature of age- and sex-related changes in vowel production in DS across 5 decades is addressed below with respect to the following three RQs: (RQ1) What is the developmental trajectory of f _o and F1–F4 of the corner vowels in males and females with DS across the ages 3.5–54 years as compared to TD speakers? (RQ2) Is there F2 compression and/or centralization of the acoustic VSA across speaker age and sex? (RQ3) What is the pattern of interspeaker (between) and intraspeaker (within) variability in f _o and all four formants (F1–F4) of the corner vowels across development? The acoustic findings are then related to potential explanations and/or interpretations regarding (a) speech intelligibility, (b) anatomic dysmorphology, and (c) speech motor control.

Developmental Trajectories of f _o and F1–F4

As can be seen in Figures 1a–1d, for both groups, the f _o and F1–F4 trajectories of all four vowels showed the expected decrease in frequency with age, likely reflecting the increasing size of the larynx and the vocal tract. The f _o and F1–F4 trajectories also showed differences between males and females, with the former showing a more conspicuous breakpoint in the developmental curves at about 10 years of age, a feature likely related to changes in craniofacial anatomy, as discussed below.

The higher f _o trajectories for DS than TD across age support the general expectation given the smaller stature in DS and therefore shorter vocal tract and smaller larynges. This finding was further asserted when the physical characteristics of the DS phenotype were accounted for in the analysis and f _o changed in significance (see Table 3). The observation noted in the Results section that differences between the TD and DS z-score trajectories were small between the ages 8 and 15 years (i.e., the peripubertal and pubertal age-cohorts), particularly in males, provides a likely explanation of the conflicting findings of acoustic studies in children, where some studies reported lower f _o in children with DS than TD children (Lyakso et al., 2021; Moura et al., 2008), while others reported little or no difference in f _o between children with TD and DS children (Albertini et al., 2010; Michel & Carney, 1964; Weinberg & Zlatin, 1970). The notable f _o breakpoint at about age 10 years, in the trajectories of all vowels, particularly males, is likely reflective of the rapid anatomic changes that occur with the onset of puberty and include secondary descent of the larynx, lengthening of the pharynx, and increased size of the laryngeal structures.

The expectation of higher formant frequencies (F1–F4) in DS did not hold consistently. As detailed in the Results section, the DS formant trajectories for all four vowels were initially either above, below, or the same as TD. The pattern, however, switched course during the period following the breakpoint for most formants and all vowels. Scrutiny of the trajectories in Figures 1 and 2 combined with analysis results (see Table 3), particularly after accounting for the physical characteristics of the DS phenotype (see Table 3, right panel), revealed differences between the two groups to be predominantly for F2 as further discussed in the following section. The F1 differences were mostly in the expected direction. Such findings indicate that DS versus TD differences are due to factors other than the proximity index of phenotype physical characteristics controlled for in our model (such as the compromised oral cavity space limiting the front/back movement of the tongue [Moura et al., 2008]). Similarly, the absence of expected differences between the two groups (i.e., before accounting for DS phenotype characteristics; Table 3 left panel), as is mostly the case for F3 and F4, is also likely due to alternate factors such as other physical/anatomic characteristics in DS that are also developmental in nature. Potential anatomic factors including compromised growth in the oral region and growth trend differences in the oral versus pharyngeal regions are discussed in the designated section below.

F2 Compression and Vowel Centralization

The findings for RQ2 confirmed both F2 compression and vowel centralization for DS. F2 compression, defined as the absolute differences in F2 frequency between the high vowels (|/i/−/u/|) and low vowels (|/æ/−/ɑ/|), revealed developmental differences between TD and DS. Although the differences were smaller for DS as compared to TD for both vowel pairs, comparison of group differences revealed significant but smaller differences for the high vowels that were constant across age, whereas the differences for low vowels were larger and only significant before the breakpoint (last two columns in Table 4). Comparison of F2 differences for the individual vowels (see Table 4 and Figures 1c and 1d) revealed that the front low vowel /æ/, unlike back low vowel /ɑ/, changed course across age such that, after the breakpoint, the differences between TD and DS were no longer significant. The implication of this finding is that, after the breakpoint, the F2 frequencies for the low vowels /æ/ and /ɑ/ became more distinct and differentiated, with the likely explanation that anteroposterior tongue movement in DS was more restricted before the breakpoint. Such an interpretation is similar to that reached by Moura et al. (2008) who reported a smaller ratio of the F2 frequencies for vowels /i/ and /u/. In light of Wild et al.'s (2018) findings of difficulties in the perceptual distinction of these low vowels, we expected increased perceptual difficulties in the identification of the vowels /æ/ and /ɑ/ when their respective F2 values were less distinct, that is, before the breakpoint. We therefore explored the relation between intelligibility and F2 of the low vowels, discussed in a later section on Speech Intelligibility.

Acoustic VSA calculations encompass the F2 compression in DS addressed above—providing insight regarding a reduction of the range of tongue position in the anteroposterior dimension for both high and low vowels in speakers with DS—and F1 that presumably captures the dimension of tongue elevation or elevation of the tongue/mandible complex. Combined, the VSA calculation listed in Table 5, for the five age cohorts in males and females of both groups, show that, in general, TD have greater areas than DS (Appendix B shows the F1–F2 plots for males and females in both the TD and DS groups. Differences in VSA are apparent in these figures). Differences in acoustic VSA between TD and DS were significant for the children in both sexes but only for females in adults. Therefore, reduced VSA is a common but not universal feature of speech production in DS. As noted in the introduction, many but not all studies report reduced VSA in DS. The present data indicate that VSA is reduced in all cohorts for females but only for the younger cohorts in males. It is important to note that VSA calculations provide a lossy measure since the formula is based on the “mean” summary statistic of the individual vowels' F1 and F2, and therefore not sensitive to variations of the corner vowels in F1–F2 space (Kent & Vorperian, 2018). However, it continues to be used with disordered speech as a proxy to articulatory working space where decreased area is associated with a decrease in speech intelligibility.

Moura et al. (2008), who used F2 frequency ratio of the vowels /i/ and /u/, reported a smaller ratio in Portuguese-speaking children with DS (Portuguese does not have low vowels that contrast in anteroposterior position). They called this ratio the DS-VR, based on the assumption that the anatomy in DS explains the formant frequency pattern. Bunton and Leddy (2011) also reported a reduced range of F2 frequencies for the vowels /i/ and /u/. The results of this study similarly indicate a general contraction of the anteroposterior dimension of vowel production in children but not adults. However, it is not clear if the different results for children and adults are related to changes in anatomy, motor control, or both. Anatomic explanations include hyo-laryngeal descent and/or the disharmonious maxillary-mandibular growth in DS. Changes in speech motor control relate to learning to increase the range of tongue movement. We discuss each of these factors below in their respective sections.

Interspeaker and Intraspeaker Variability

Despite evident increase in interspeaker variability as displayed in Figure 2, likely reflective of anatomic growth differences between speakers with DS as well as differences in severity of speech motor control difficulties, only select formant frequencies for select vowels were significantly different between TD and DS (see Table 6). This is probably because the DS group was composed of many individuals who were recorded more than once, which would be expected to reduce the group variance. Significant formant differences of the front vowels are likely reflective of anatomic dysmorphology in DS as discussed in the designated section below.

Intraspeaker variability, however, was consistently greater for DS than TD for all vowels across age and most frequencies particularly during the first two age cohorts and after age 25 years for select formants (see Figure 3). Significant group differences in intraspeaker variability were present for f _o and most formants particularly the lower formants that are essential to convey vowel phonetic identity. This is interpreted to mean that individuals with DS have phonatory instability and less precise vowel articulations, findings that are respectively reflective of effort exerted for sustained phonation, and developmental changes in speech motor control as discussed in the allocated section below.

Findings as Related to Speech Intelligibility, Anatomic Dysmorphology, and Speech Motor Control

What follows is a discussion relating the most outstanding acoustic findings from the three RQs addressed in this study, to published reports on speech intelligibility, anatomic dysmorphology, and speech motor control.

Speech Intelligibility

A single-word speech intelligibility study in speakers with DS conducted by Wild et al. (2018) used the same participants and words as reported here, providing us with a unique opportunity to compare acoustic and intelligibility data for vowel production in DS. An important finding by Wild et al. was that high vowels were more intelligible and developed earlier than low vowels for both male and female individuals with DS. High vowels generally developed before the age of 4 years, whereas intelligibility of low vowels continued to increase through age 12 years for vowel /æ/ and through age 16 years for vowel /ɑ/. In other words, the period of reduced contrast between the low-front and low-back vowels was protracted and had an adverse effect on speech intelligibility. The present acoustic findings of TD versus DS F2 compression differences being larger for the low vowels than the high vowels seem to provide an acoustic explanation/dimension for the reduced intelligibility noted in the Wild et al. study. Given the RQ2 finding that the F2 pattern for the vowel /æ/ is significantly different between TD and DS before breakpoint (see Table 4), and also, the display of the F2 pattern in Figures 2c and 2d for the vowels /æ/ and /ɑ/ before breakpoint where the F2 distinction between these two vowels was reduced, decreased intelligibility of the low vowels was expected. Figure 4 is an exploratory display of the low vowel F2 pattern in speakers with DS against their intelligibility scores (from Wild et al.). Indeed, Figure 4 shows that as the distinction between the F2 of the vowels /ɑ/ and /æ/ increased, displayed in red and blue, respectively, single-word intelligibility score also increased. Vowel errors also have been noted in other auditory-perceptual studies of speech in DS (Bunton et al., 2007; Carl et al., 2020; Jones et al., 2019; Kent et al., 2021). The acoustic data reported here indicate that the intelligibility of vowels in DS may be compromised in two ways. First, the vowel formant patterns in DS are less distinct than in TD speech, often with F2 compression. Second, vowels are produced with greater variability in DS than TD, which conceivably could hinder vowel perception by listeners. Therefore, speakers with DS have both atypical formant patterns and a greater variability in vowel production.

Figure 4.

Intelligibility scores of the speakers with DS (as calculated in Wild et al., 2018) as a function of their formant frequencies F1–F4 for the low vowels /æ/ (blue) and /ɑ/ (red). Fit lines are linear regressions on a B-spline basis, fitted separately for each formant/vowel combination. F2 panel is emphasized to draw attention to the most obvious formant differences between /ɑ/ (red) and /æ/ (blue) with intelligibility.

Anatomic Dysmorphology: Potential Explanations of Acoustic Findings

As seen in Figure 2 and described in the Method section, in our fit model, we added a knot term (breakpoint) at about age 10 years (peripubertal to pubertal phase transition) to provide a more flexible fit that adapts to the sharp decrease in f _o and multiple formants that we had observed at this time point particularly for the DS data. This outstanding acoustic finding from RQ1 that f _o and all four formants displayed a pronounced drop in frequencies during puberty (after the breakpoint), particularly in males, is likely related to a number of concurrent anatomic changes that are conducive to a rather abrupt growth in both laryngeal and vocal tract length dimensions. It is well known that the growth of the vocal tract is nonuniform (Fant, 1966). Using the oral-pharyngeal two-tube model analogy (Fant, 1980), its heterogeneous growth pattern has been characterized to have a neural growth type in the oral/horizontal plane, and a general or somatic growth type in the vertical plane (Vorperian et al., 2009). While the initial rapid growth pattern during early childhood is similar for both growth types, a critical difference (as defined by Scammon, 1930) is that, by age 6 years, neural growth achieves 80% of the adult size, whereas somatic growth achieves 25%–40% of adult size. Following this rapid growth phase during early childhood, both growth types display ongoing slow growth until maturity, except that the somatic growth type undergoes a second period of accelerated growth during puberty. It is noteworthy that the majority of documented anatomic dysmorphologies characterized by arrested growth in DS, as listed in the introduction and discussed in greater detail below, are in the region of the anterior facial skeleton, that is, in the oral region excluding the tongue/soft tissue (with corroborating acoustic finding of F2 as discussed below). Growth in the pharyngeal region appears to be spared since, aside from the expected smaller pharyngeal dimensions, no anatomic dysmorphologies have been reported, though airway, feeding, and swallowing problems are common in DS (Spender et al., 1996). Laryngeal dysmorphology, specifically higher incidence of laryngomalacia, has been reported (e.g. Bertrand et al., 2003), however. Hamilton et al. (2016) noted that the use of thorough diagnostic airway examination (airway endoscopy under general anesthesia) is necessary to accurately report on the nature of the higher incidence of respiratory problems in children with DS. They reported that 13.8% of the 239 children with DS they tested had respiratory problems where the majority had either tracheal problems (8.79%) or subglottic stenosis (5.86%), but the incidence of laryngomalacia was small (0.83%).

Based on the acoustic observation of abrupt and synchronized decrease in f _o and all four formant frequencies (particularly in males) at about the same age as the onset of the second phase of accelerated somatic growth, it is reasonable to postulate that the growth of vocal tract structures in the vertical plane in DS (including pharyngeal growth with hyo-laryngeal descent) and laryngeal development appear to follow a growth trend that is very similar to TD. Such a proposition is further supported by the RQ1 finding that F4 (the resonance of the laryngeal cavity [Takemoto et al., 2006]) displayed no significant differences between the two groups (though Figure 1a shows some differences post breakpoint for the vowel /i/), and also that f _o differences between TD and DS were minimal between the ages of 8 and 15 years. In other words, the discrepant growth differences between the two groups, with arrested neural growth in the horizontal plane in DS but with a common second phase of rapid somatic growth in the vertical plane for both groups, results in a relatively more abrupt change in vocal tract length in the vertical dimension for DS at the onset of puberty.

A second outstanding acoustic finding was that, of the four formants examined, F2 was the most different between DS and TD with a developmental pattern. F2 was lower in DS than TD for the front vowels before age 10 years (see Figures 1a, 1c, 2a, and 2c), but changed developmentally, particularly the front low vowel /æ/ (see Table 4), implying the front vowels were initially produced with a more posterior tongue position. Furthermore, findings revealed F2 compression for both the high and low vowels, but more so for the low vowels.

The present findings are in line with Moura et al.'s (2008) anatomic interpretation that implicates maxillary hypoplasia for the restricted anteroposterior tongue movement. In other words, the likely anatomic origin of restricted tongue movement in the anteroposterior dimension is a short palate (formed by the palatine processes of the maxilla for its anterior three quarters plus a horizontal plate of the palatine bone where its midsagittal length ends at the level of the posterior nasal spine at the juncture of the hard and soft palates). Shapiro et al. (1967), who obtained metric information on height, width, and length (ages 6 years to adulthood), determined that, compared to the typical, palatal vault/height is not necessarily higher in DS, but palatal width is narrower and palatal length is dramatically shorter. In fact, they noted that palatal length “separates the group with Down syndrome from the normal subjects more distinctly than any other reported physical trait” (p. 1463). Furthermore, they noted that the high or “steeple” palate was not a consistent finding but that the narrower width gives the impression of “high,” a conclusion similar to that of Redman et al. (1965) who noted that a short and narrow palate creates the illusory impression that the palate is high. Indeed, the nature of the palatal dysmorphology in DS, as described in more recent literature, reporting the presence of a “high arched palate” or “gothic palate” in a large percentage of individuals with DS are from visual impressionistic examinations (e.g., 75% of 53 individuals ages 7–16 years in Al Sakarna & Othman, 2010; 89% of 27 individuals ages 10–14 years in Bhowate & Dubey, 2005; 84% of 77 individuals ages 6–40 years in Shukla et al., 2014). So, although the literature up to the present attests to the persisting belief that a high arched or gothic palate is a highly identifiable and important feature of DS, published studies that present quantitative data on palatal morphology, as summarized in Appendix C, show that there is general agreement that the palate is reduced in one or more dimensions in DS. The most consistent features are short and/or narrow palate, with the results for height being mixed. This conclusion is consistent with a systematic reviews and meta-analyses in which it was concluded that the maxilla is smaller in DS than control groups (Díaz-Quevedo et al., 2021; Vicente et al., 2020). In summary, the preponderance of the evidence points to palatal dysmorphology in the form of a short and narrow palate, but not a gothic palate as conventionally defined. Therefore, palatal length and width are better descriptors of the dysmorphology in DS. Although motoric contributions/involvement cannot be discounted particularly given the greater F2 compression and reduced intelligibility of the low vowels as discussed in the previous section, palatal dysmorphology is considered to be important in explaining vowel formant differences. Palatal width also has been reported to be correlated with errors in consonant production (Ferraz & Ghirello-Pires, 2022).

In addition to the palatal dysmorphology, mandibular dimensions are also compromised in DS and variously described as a smaller/scaled down version of TD (Guimaraes et al., 2008; Korbmacher et al., 2005; van Marrewijk et al., 2016), shorter in height (Ferrario et al., 2004; Jesuino & Valladares-Neto, 2013), narrower (Jayaratne et al., 2017; Sforza et al., 2012 ), and wider (Uong et al., 2001). In a systematic review, Díaz-Quevedo et al. (2021) concluded that the dimensions of the maxillary bone, mandible, and skull base are reduced in DS with more frequent presentation of Class III malocclusion in DS than TD. In other words, despite contradictory findings on mandibular growth, the general consensus supports the absence of synchronized growth of the maxillo-mandibular complex, which would affect tongue movement and articulatory contact. For example, a narrow palate and/or a wider mandible could be another anatomic explanation as to why front vowels were produced with a more posterior tongue position before age 10 years. Unfortunately, there are few sources of data, either cross-sectional or longitudinal, on craniofacial and upper airway development in children with DS. Allareddy et al. (2016) conducted a cross-sectional study of 27 individuals with DS ages 3–25 years that included a dental exam and lateral cephalometric radiograph with a comprehensive set of measurements including maxillary, mandibular, maxillo-mandibular relational measurements (ANB angle, Wits appraisal), occlusion, and airway measurements. Their findings revealed a flexion point in growth curves at about age 10–11 years of age for most of those measures. This flexion point roughly corresponds to the age of onset of the second period of accelerated growth of the somatic growth, described earlier, that the facial skeleton follows (Ekström, 1982). However, it is also the age when the various craniofacial measurements corresponding to anatomic dysmorphologies, specifically maxillo-mandibular disharmonies including Class II and Class III malocclusion, open-bite, and cross-bite become more pronounced (Guyer et al., 1986). Shukla et al. (2014) reported a prevalence of 97% Class III malocclusion in individuals with DS ages 6–40 years, and Achmad et al. (2021) reported 92% from a systematic review. Such high incidence of Class III in DS was further confirmed with moderate strength by a meta-analysis of eight studies by Doriguêtto et al. (2019), who used the important criterion of including studies with both a dental exam and cephalometric measurements to determine occlusion classification and severity of dentofacial disharmony. Indeed, Allareddy and colleagues conclude by stating that all their study participants demonstrated a Class III skeletal pattern that was more pronounced in the older age group as compared to the younger age groups; they also reported that 12/27 participants had maxillary hyperplasia and none had mandibular hyperplasia. Such specificity on the nature of the malocclusion, as well as origin of malocclusion, is critical because it has implications on treatment indicated.

Guyer et al. (1986) clarified that mandibular prognathism is not synonymous with Class III malocclusion as the morphology of the craniofacial complex can vary. Their study findings, using lateral cephalographs of 144 Class III children ages 5–15 years, revealed the incidence of maxillary retrusion with a normal mandible to be 25%, normal maxilla with mandibular protrusion 18.7%, maxillary retrusion with mandibular prognathism 22.2%, and so forth. In other words, Class III is not a single diagnostic entity; it simply refers to a skeletal facial deformity characterized by a mandibular position that is positioned ahead of the maxilla or cranial base. It is therefore not surprising that, although Class III is associated with articulation disorders (Lathrop-Marshall et al., 2022), it is not clear how it affects vowels. The limited number of studies that have examined vowel acoustics in young adult males with Class III occlusion have reported different findings. Xue et al. (2011), who used Acoustic Reflection technology with Class III participants who had greater oral length and oral volume, reported smaller VSA in Class III primarily because of the very high F1 for the vowel /u/. Ahn et al. (2015) reported their Class III participants to have larger F1–F2 acoustic VSA. Whereas, Pravitharangul et al. (2022) reported no differences in their Class III participants except for the vowel /o/ (back and rounded vowel) that was produced with a significantly higher F2 and lower F4. It is therefore not feasible to reach any conclusions on how Class III could affect acoustic VSA given the multiple ways in which maxilla-mandibular disharmony, each with a range in severity, can result in this type of occlusion during the course of development. Suri et al. (2010) report that their participants with DS (M _age = 15.1 years) displayed mandibular hypoplasia that was less severe than maxillary hypoplasia, and Allareddy et al. report that 12 of 27 participants (ages 3–25 years) had maxillary hyperplasia and none had mandibular hyperplasia. Additional developmental information in three dimension is needed to determine, as compared to TD, the presence of regional dysmorphologies in DS for each structure at birth and/or the timing of their emergence, including the emergence of disharmonious maxillary-mandibular development, to better understand malocclusion (since it is secondarily acquired later in life) and the effect of occlusion on vowel production. Based on the findings of vowel studies on speakers with Class III occlusion, our finding of the absence of significant differences in the VSA between DS and TD (see Table 5) is less surprising. It may be that F2 differences of front–back vowels for high versus low vowels can be a better index to assess severity of antero-posterior tongue movement as related to speech intelligibility. Data-driven mean frequencies per age cohort for DS and TD used in this study can be found in Appendix A.

Understanding the origin of anatomic dysmorphologies in DS, including palatal shape and disharmonious maxillary-mandibular growth, can be foundational to treatment. Klingel et al. (2017) report DS to have a typically shaped palate up to age 6 months, but from 6 to 9 months onward, the growth pattern decreases irregularly. As for the mandible, Guyer et al. (1986) report that mandibular differences begin to emerge by age 6–7 years, and malocclusion is confirmed by age 10 years. Functionally, Allareddy et al. (2016) report increases in airway dimensions with age as malocclusion develops. Thus, assessment of anatomic dysmorphologies in DS and/or the timing of their emergence should be done within the context of daily living including environmental factors since altered function and/or impaired motor control, in the absence of intervention, can be a contributing factor in the exacerbation of congenital dysmorphologies or the emergence of new ones (Cheng et al., 2002; Klingel et al., 2017). For example, a commonly reported “at rest” orofacial description of individuals with DS includes an open mouth posture and a low, protruding passive tongue position, both of which are typically attributed to generalized hypotonia. Open mouth posture with low tongue position is reported highly correlated to malocclusion and vaulted-palatal shape (Oliveira et al., 2008). Several studies report success in correcting tongue position and oral seal (lip approximation) following orofacial regulation therapy at an early age (2 months to 4 years) that includes use of a stimulating palatal plate along with stimulation of neuromotor face trigger points, a treatment approach originally developed by the Argentinian rehabilitation physician Rodolfo Castillo Morales (1941–2011) that is carried out while communicating/interacting with the child as he/she gets older (Carlstedt et al., 2003; Javed et al., 2018; Limbrock et al., 1991, 1993; Zavaglia et al., 2003). Findings also reflect the facilitation of speech development (Hohoff & Ehmer, 1997).

Speech Motor Control in DS

Increased intraspeaker variability in vowel production in DS indicates phonatory instability and spatial imprecision in articulation. Voice instability in DS has been linked to several factors, including (a) increased subglottal pressure and increased laryngeal airway resistance and phonation threshold (Pebbili et al., 2019); (b) increased muscular activation to overcome hypotonia (Pryce, 1994); and asymmetric vibration of the vocal folds and inadequate medial contact of the folds (Hidalgo-De la Guía et al., 2021). Any one or combination of these factors could account for the present significant f _o intraspeaker variability findings. Spatial imprecision in articulation may be a result of faulty motor programming and/or inaccurate motor execution. The present formant-frequency data, together with data indicating temporal imprecision in speech production (Brown-Sweeney & Smith, 1997), points to the possibility of a general spatiotemporal motor difficulty. Motor problems in DS have been noted in several studies. In their review of the literature, Carvalho and Vasconcelos (2011) noted that, in addition to a developmental delay, individuals with DS exhibit motor impairments in the form of slow and variable movements (attributed to factors such as low muscle tone). It appears that speech movements are no exception to this conclusion. Several studies have shown that syllabic diadochokinesis in individuals with DS has a slow rate and is variable in a temporal pattern (Brown-Sweeney & Smith, 1997; Hamilton, 1993; Rupela et al., 2016; Rupela & Manjula, 2010; Zarzo-Benlloch et al., 2017). Spatiotemporal instability could contribute to reduced speech intelligibility if the instability leads to ambiguity in the acoustic cues for phonetic segments.

In addition to structural contributions and developmental delay, the speech disorder in DS may be associated with dysarthria, CAS, or both of these. Chenausky and Tager-Flusberg (2022) commented that many of the characteristics of different dysarthrias resemble the core impairments in CAS. They also note a complication in that dysarthria may co-occur with CAS, which makes diagnosis even more difficult. This circumstance applies in DS given the likelihood of hypotonia and other motor disturbances that likely interfere with the development and accurate execution of speech motor commands. Among the most frequently occurring features of CAS are inconsistencies and vowel errors (Sayahi & Jalaie, 2016), both of which are common in DS. It is a diagnostic challenge to distinguish CAS and dysarthria in the presence of generalized motor dysfunction and craniofacial dysmorphologies. Resolution likely will depend on further studies that examine the details of movement patterns for phonetic sequences and the assessment of somatosensory information (Terband et al., 2009). As further discussed below in the implications section, there is the distinct possibility of a disorder-specific motor development profile.

Sensory factors may also have a role in understanding speech production in DS. Hearing loss is common. The incidence of Hearing loss in neonates and infants with DS is between 15% and 30%, between 25% and 85% in children and adolescents, and between 50%and 75% in adults (Hendrix et al., 2020). The hearing loss is typically associated with otitis media with effusion, but mixed conductive-sensorineural loss is also common and may be secondary to inner ear malformations. There also are indications of impaired orosensory function, as in the task of oral stereognosis (Shemshadi et al., 2013). Individuals with DS may contend with sensory dysfunctions that contribute to speech difficulties, but it is not clear how these dysfunctions relate to the overall pattern of speech disorder.

Limitations and Implications for Clinical Treatment and Future Research

Limitations

The results of this study should be interpreted with caution given the complexity of the speech disorder in DS, the phenotypic variation within the sample of participants including severity of reduced intelligibility, and limited knowledge on craniofacial and motoric development in the syndrome. The participants in this study represented a wide range of ages for both sexes, but the proportion of males and females at a given age was approximately equal. A further limitation is that the present results pertain only to steady-state measurements of the corner vowels produced in single words. Speech production in DS involves several different aspects that must be considered in an overall understanding of reduced intelligibility. In addition to difficulties in vowel production, speakers with DS often have atypical production of consonants (Nyman et al., 2021) and prosody (Corrales-Astorgano et al., 2021).

Implications for Clinical Treatment and Future Research

In line with clinical implications discussed in the work of Wild et al. (2018), vowels contribute to speech intelligibility and are central to the formation of syllables (Kent & Rountrey, 2020; Wild et al., 2018). Therefore, it is important to consider vowels, especially the low vowels, in clinical assessment and as part of a treatment plan. Furthermore, given the underlying generalized hypotonia that likely affects motor control development (Lott, 2012; Santoro et al., 2021) and a rapidly changing anatomy particularly during childhood, it is advisable to incorporate specialized interventions to establish continuous nasal breathing and facilitate the development of critical oral functions (feeding and swallowing) to help promote the typical, healthy development of the craniofacial, aerodigestive, and vocal tract structures, and also to minimize the exacerbation of congenital dsymorphologies (such as a small, short and narrow, palate), or the emergence of new ones (such as malocclusion; Torre & Guilleminault, 2018). Also promising are interventions that, in addition to facilitating the typical growth of structures, promote the development of motor control—particularly during early childhood—such as the treatment for orofacial dysfunction using the Castillo Morales stimulating orthodontic palatal plate as described in the section above on anatomic dysmorphology (Carlstedt et al., 2003; Limbrock et al., 1991, 1993; Zavaglia et al., 2003). In a systematic review of eight studies, Javed et al. (2018) stated that “Most of the included studies suggest that palatal plate therapy in combination with physiotherapy/orofacial regulation therapy according to Castillo Morales/speech and language intervention seems to be effective in improving orofacial disorders in children with DS” (p. 20). Chouchene et al. (2021) reached a similar conclusion, from a systematic review of six studies, while emphasizing the importance of early intervention involving a multidisciplinary team. Javed et al. (2018), however, also recommended “Longitudinal trials with standardized evaluation methods, age of children at treatment initiation, treatment duration and standard orofacial outcomes…” to better quantify such findings. Intervention for guided motor training/practice that takes account of sensory strength (e.g., visual learning, receptive vocabulary) and deficits (e.g., sensorimotor control including somatosensory and proprioceptive deficits) is also promising (Carvalho & Vasconcelos, 2011; Horvat et al., 2016).

This study focused on formant-frequency estimates of the steady-states of the corner vowels. A more complete picture of vowel production would provide data on additional vowels, vowels produced in multisyllabic utterances, and vowel-inherent spectral change. The additional data would be informative about timing and dynamic aspects of vowel production, to include measures of vowel duration (e.g., tense vs. lax vowels) and vowel-inherent spectral change (Morrison & Assmann, 2013). A combination of spectral and temporal information on vowel production could lead to a better understanding of motor control issues.

Speech is an area of high interest to parents of individuals with DS (White et al., 2022) and could very well be linked to cognition, the deficit of highest interest. Although studies to date have documented the severity of the speech impairment and its roots in multisystem involvement, several fundamental questions relevant to clinical intervention remain unanswered. (a) What is the relative contribution of motoric disturbances versus craniofacial and laryngeal dysmorphologies? Specifically, does a highly prevalent condition such as maxillary dysplasia place hard limits on the nature of speech articulation, or can it be overcome by articulatory compensations? (b) Can dysmorphologies be prevented or mitigated by interventions? It has been proposed that some aspects of craniofacial dysmorphology are secondarily acquired during postnatal development (Contaldo et al., 2021; Klingel et al., 2017; Macho et al., 2014), which invites the possibility of effective early intervention with methods such as orthodontic palate plate therapy (Limbrock et al., 1991, 1993) or rapid maxillary expansion (Moura et al., 2005). (c) What is the nature of the motor involvement in DS? Possibly, this involvement can be satisfactorily described as a variable combination of dysarthria and CAS, but it could be a condition characteristic of DS that is yet to be identified. Hypotonia may be involved, but hypotonia is poorly understood and difficult to assess clinically (Dan, 2022; Latash et al., 2008), and its role in speech production in DS is unclear (Chu & Barlow, 2016). With respect to basic motor skills, it has been suggested that children with DS have a disorder-specific motor development profile that is related to the combined effects of hypotonia, joint hypermobility, and reduced postural reactions (Lauteslager et al., 1998, 2020). The presence of unique motor problems has similarly been noted by Block (1991) who, in addition to these factors, lists evidence of the additional challenges for motor development including abnormal reflex development, instability, obesity, medical and health problems (congenital heart defects, atlantoaxial subluxation, sensory-motor problems), as well as social and cognitive abilities. Similarly, the speech disorder in DS may be a disorder-specific profile related to the combined effects of hypotonia, dysarthria, disordered motor programming, phonological disorder, and craniofacial dysmorphology. (d) Given that there is a wide range of speech intelligibility in DS, are there anatomic, motoric, or somatosensory differences between individuals with high versus low intelligibility? Better understanding may come through a research initiative that objectively addresses the interactions among anatomic, biomechanical, and acoustic features of speech production, as related to severity of reduced speech intelligibility, through the life span in persons with DS.

Data Availability Statement

The recordings obtained and analyzed and data sets generated during the current study are not publicly available due to ethical considerations but could be available from the corresponding author on reasonable request with institutional review board approval.

Appendix A

Data-Derived Mean Fundamental and Formant Frequencies, and Standard Deviation (SD) for Each of the Five Age Cohorts

		DS								TD
Prepubertal		/i/		/u/		/æ/		/ɑ/		/i/		/u/		/æ/		/ɑ/
		M	(SD)	M	(SD)	M	(SD)	M	(SD)	M	(SD)	M	(SD)	M	(SD)	M	(SD)
f o	F	287.93	(28.56)	276.24	(35.63)	270.90	(27.84)	271.80	(23.89)	250.79	(24.97)	235.83	(24.85)	239.19	(14.72)	237.15	(19.28)
	M	274.56	(37.34)	268.42	(36.47)	260.85	(35.26)	254.43	(41.46)	250.12	(37.34)	234.12	(32.11)	238.09	(29.54)	239.17	(34.80)
F1	F	485.95	(83.90)	520.49	(83.07)	1022.35	(167.50)	1089.88	(132.25)	405.93	(55.02)	454.28	(69.30)	1008.79	(155.80)	1210.46	(99.36)
	M	472.90	(83.48)	469.55	(64.63)	992.13	(130.02)	1096.71	(147.22)	397.94	(45.94)	430.75	(54.53)	957.75	(131.44)	1146.54	(153.17)
F2	F	3316.06	(149.82)	1405.14	(182.10)	2437.32	(201.91)	2029.65	(123.60)	3487.59	(230.43)	1321.07	(119.02)	2620.19	(214.07)	1839.58	(198.73)
	M	3130.43	(209.52)	1479.13	(244.15)	2246.85	(157.19)	1972.88	(193.12)	3350.12	(182.74)	1201.77	(175.32)	2521.97	(219.92)	1732.86	(198.15)
F3	F	3992.45	(199.55)	3558.38	(161.73)	3857.95	(275.91)	3637.23	(415.72)	4192.12	(331.45)	3614.07	(305.70)	3713.52	(248.34)	3509.67	(294.54)
	M	3889.25	(223.46)	3620.98	(215.70)	3755.86	(267.36)	3667.42	(417.75)	4017.05	(280.24)	3456.09	(287.40)	3625.48	(261.25)	3366.96	(288.50)
F4	F	5176.23	(218.65)	4829.42	(198.03)	4970.61	(338.40)	4728.43	(194.09)	5184.88	(378.18)	4799.06	(426.00)	4970.57	(371.90)	4625.33	(329.53)
	M	4962.83	(316.67)	4664.98	(265.51)	4837.16	(242.42)	4662.43	(300.81)	5032.13	(351.49)	4616.61	(309.56)	4730.68	(311.03)	4459.43	(295.38)
Peripubertal		/i/		/u/		/æ/		/ɑ/		/i/		/u/		/æ/		/ɑ/
		M	(SD)	M	(SD)	M	(SD)	M	(SD)	M	(SD)	M	(SD)	M	(SD)	M	(SD)
f o	F	251.51	(35.55)	236.37	(30.47)	237.18	(37.29)	235.65	(27.07)	230.37	(9.91)	216.04	(15.49)	222.92	(15.92)	216.83	(13.99)
	M	243.84	(27.04)	223.15	(29.51)	213.17	(36.98)	227.96	(25.75)	231.23	(21.49)	222.56	(25.99)	222.42	(22.58)	222.44	(21.06)
F1	F	435.50	(43.97)	438.06	(75.50)	944.14	(123.49)	1076.34	(119.52)	398.75	(29.71)	432.72	(35.65)	987.75	(111.87)	1152.32	(104.44)
	M	419.90	(66.61)	391.59	(91.23)	964.96	(180.42)	1024.62	(125.30)	378.06	(45.95)	399.56	(35.38)	919.70	(113.55)	1028.45	(83.30)
F2	F	3149.05	(142.01)	1401.36	(204.56)	2326.52	(192.00)	1870.90	(156.03)	3254.26	(216.86)	1274.06	(108.88)	2336.23	(152.32)	1712.71	(163.50)
	M	2967.81	(223.13)	1397.64	(326.52)	2241.13	(223.22)	1698.38	(197.36)	3068.71	(193.80)	1147.44	(144.59)	2158.60	(151.46)	1566.53	(84.53)
F3	F	3712.00	(170.31)	3311.01	(215.04)	3484.32	(132.29)	3390.04	(253.76)	3912.62	(295.75)	3266.26	(180.27)	3381.77	(211.40)	3269.11	(181.60)
	M	3689.79	(203.57)	3330.40	(267.26)	3548.51	(313.82)	3349.91	(343.30)	3640.00	(204.68)	3137.21	(172.94)	3232.56	(143.72)	3088.67	(275.04)
F4	F	4757.37	(206.29)	4467.18	(170.74)	4478.29	(234.25)	4327.69	(259.92)	4823.37	(280.44)	4483.78	(321.08)	4575.84	(296.19)	4371.13	(206.11)
	M	4766.75	(359.39)	4362.18	(296.89)	4636.58	(335.73)	4375.14	(266.94)	4578.56	(212.15)	4339.34	(209.38)	4323.60	(208.84)	4199.44	(277.11)
Pubertal		/i/		/u/		/æ/		/ɑ/		/i/		/u/		/æ/		/ɑ/
		M	(SD)	M	(SD)	M	(SD)	M	(SD)	M	(SD)	M	(SD)	M	(SD)	M	(SD)
f o	F	228.27	(33.44)	220.86	(26.15)	220.88	(20.57)	218.82	(26.47)	222.05	(23.09)	220.76	(22.85)	215.68	(24.05)	214.96	(24.43)
	M	177.93	(56.47)	170.73	(48.98)	169.22	(53.05)	166.86	(52.36)	195.40	(47.45)	187.62	(46.96)	187.90	(45.49)	190.34	(49.48)
F1	F	405.64	(51.86)	423.68	(69.28)	832.98	(155.36)	910.61	(111.10)	376.31	(45.18)	401.42	(55.97)	924.55	(89.62)	1041.24	(82.76)
	M	370.19	(41.42)	379.71	(46.32)	771.66	(127.07)	841.53	(139.21)	347.36	(54.54)	359.34	(49.67)	826.23	(122.26)	940.15	(126.49)
F2	F	2778.63	(333.16)	1304.01	(180.04)	2090.43	(219.79)	1585.17	(137.87)	2923.11	(199.06)	1229.12	(148.79)	2115.68	(161.41)	1609.45	(152.05)
	M	2626.56	(365.00)	1227.76	(185.93)	2004.23	(334.73)	1512.78	(252.30)	2757.28	(255.00)	1088.72	(123.78)	2008.89	(192.32)	1465.67	(144.85)
F3	F	3371.61	(373.36)	2832.64	(255.23)	2988.27	(244.36)	2793.27	(315.64)	3527.65	(253.43)	2957.52	(233.93)	3121.02	(178.36)	3056.55	(232.33)
	M	3232.55	(399.37)	2881.10	(525.09)	3101.08	(417.70)	2858.14	(417.97)	3369.41	(285.38)	2778.28	(282.33)	2946.10	(246.93)	2800.59	(214.19)
F4	F	4432.69	(467.69)	3993.02	(400.55)	4104.04	(266.82)	3989.14	(245.82)	4414.59	(289.80)	4114.01	(336.77)	4267.26	(232.83)	4122.61	(222.37)
	M	4208.64	(434.04)	3937.04	(444.62)	4133.44	(478.17)	3984.66	(369.91)	4262.56	(331.07)	3987.10	(378.19)	4070.87	(268.38)	3960.70	(292.95)
Postpubertal		/i/		/u/		/æ/		/ɑ/		/i/		/u/		/æ/		/ɑ/
		M	(SD)	M	(SD)	M	(SD)	M	(SD)	M	(SD)	M	(SD)	M	(SD)	M	(SD)
f o	F	215.24	(22.99)	211.52	(25.99)	213.44	(31.57)	207.34	(25.13)	213.30	(21.26)	215.62	(19.31)	211.79	(20.51)	207.83	(21.91)
	M	143.48	(28.50)	137.37	(27.53)	134.01	(25.49)	136.67	(28.90)	105.21	(20.48)	105.20	(21.70)	101.25	(21.16)	102.18	(21.60)
F1	F	380.85	(27.23)	414.32	(48.87)	829.75	(144.55)	973.80	(114.62)	356.94	(32.62)	380.04	(41.96)	871.66	(64.43)	972.37	(69.95)
	M	305.44	(50.06)	313.21	(49.67)	687.57	(111.99)	764.89	(125.43)	274.03	(27.25)	316.48	(34.86)	684.39	(57.01)	767.26	(65.23)
F2	F	2669.74	(275.34)	1310.36	(216.46)	2020.82	(157.71)	1601.86	(187.38)	2789.25	(166.50)	1193.66	(165.54)	2011.11	(128.02)	1502.76	(95.92)
	M	2352.41	(257.86)	1064.46	(124.58)	1791.79	(205.65)	1378.73	(168.28)	2289.90	(189.31)	1021.64	(120.91)	1701.83	(122.30)	1239.86	(89.04)
F3	F	3265.62	(332.39)	2748.14	(273.87)	2957.78	(268.63)	2715.60	(285.79)	3360.65	(220.00)	2749.70	(138.50)	2928.28	(157.47)	2829.59	(159.49)
	M	2945.00	(254.08)	2400.98	(211.76)	2611.45	(218.46)	2450.21	(247.59)	2959.85	(218.42)	2368.70	(158.83)	2564.89	(129.01)	2570.92	(165.52)
F4	F	4213.58	(414.89)	3735.10	(282.56)	4034.78	(286.59)	3950.36	(255.33)	4242.77	(265.62)	3948.86	(262.22)	4159.10	(218.64)	3966.19	(189.51)
	M	3799.80	(309.09)	3473.87	(267.33)	3738.24	(215.54)	3658.98	(241.96)	3566.47	(239.61)	3355.57	(243.43)	3635.50	(240.78)	3575.67	(186.34)
Adults		/i/		/u/		/æ/		/ɑ/		/i/		/u/		/æ/		/ɑ/
		M	(SD)	M	(SD)	M	(SD)	M	(SD)	M	(SD)	M	(SD)	M	(SD)	M	(SD)
f o	F	211.14	(28.02)	198.50	(23.50)	200.25	(18.16)	197.13	(21.95)	193.07	(28.35)	192.34	(28.99)	181.92	(29.61)	180.39	(28.68)
	M	146.15	(25.67)	143.15	(22.75)	137.66	(24.55)	135.63	(24.63)	114.55	(18.87)	115.50	(19.14)	109.69	(19.07)	108.27	(17.8)
F1	F	371.12	(44.71)	398.96	(42.62)	748.19	(115.42)	962.06	(109.69)	340.84	(40.13)	352.21	(39.92)	800.14	(101.25)	977.07	(98.87)
	M	311.16	(46.48)	321.28	(49.07)	679.56	(78.42)	806.30	(113.77)	278.85	(23.96)	310.02	(30.93)	637.54	(52.57)	773.03	(70.34)
F2	F	2654.71	(258.10)	1151.93	(169.08)	2065.27	(145.91)	1553.02	(127.92)	2770.05	(175.27)	1056.56	(123.00)	2117.72	(173.02)	1484.25	(115.02)
	M	2442.35	(714.49)	1065.66	(153.21)	1825.75	(144.50)	1385.67	(174.04)	2261.91	(156.74)	940.95	(109.74)	1752.73	(125.59)	1291.83	(77.60)
F3	F	3269.53	(266.28)	2728.45	(261.52)	2961.93	(212.98)	2765.57	(268.86)	3350.72	(202.87)	2751.47	(173.11)	2943.98	(158.22)	2830.57	(183.48)
	M	2895.11	(221.62)	2473.85	(183.78)	2606.12	(134.56)	2512.67	(168.85)	2902.03	(157.88)	2325.81	(149.81)	2513.65	(137.88)	2546.28	(143.69)
F4	F	4250.44	(339.06)	3853.15	(344.49)	4078.75	(268.14)	3938.86	(283.27)	4191.41	(253.38)	3869.40	(251.56)	4107.42	(258.44)	3908.52	(281.99)
	M	3734.37	(252.31)	3418.86	(223.11)	3621.09	(270.32)	3582.63	(213.72)	3583.99	(211.61)	3266.38	(215.82)	3515.15	(232.54)	3516.50	(194.33)

Note. DS = Down syndrome; TD = typically developing; f _o = fundamental frequency; F = female; M = male.

Appendix B

Mean Vowel Space Area in DS and TD

Display of VSA of the mean F1 and F2 for each of the corner vowels of all five age cohorts (denoted by the vowel symbol), as well as the mean F1 and F2 for each of the five Age Cohort I (prepubertal: ages 3;6–7;11 [years;months]), II (peripubertal: ages 8;0–10;2), III (pubertal: ages 10;3–14;5), IV (postpubertal: ages 14;6–19;11), and adults (≥ 20;0) in males (top panel) and females (lower panel). DS is denoted with a solid line; and TD, with dotted lines.

Appendix C

Palatal Dimensions in DS Compared to Neurotypical Controls, as Reported in Studies Listed in the Left-Most Column

Source	No. of participants (males, females, if available)	Ages (mos = months, yrs = years, gw = gestation weeks)	Length	Height	Width
Abeleira et al. (2015)	30 (25, 15)	10–40 yrs	—	—	<
Alió et al. (2011)	47 (25, 22)	8–28 yrs	<	<	—
Al Darwish & Farh (2019)	15	8–10 yrs	<	<	<
Al-Shawaf & Al-Faleh (2011)	30		—	>	<
Austin et al. (1969)	10 (5, 5)	Newborns	—	<	—
Bhagyalakshmi et al. (2007)	88 (46, 42)	6–16 yrs	—	>	—
Cicero et al. (2004)	88	11–14 gw	<	—	—
Dellavia et al. (2007)	47 (32,15)	20–45 yrs	<	>	—
Fansa et al. (2019)	50 (30, 20)	4–14 yrs	<	<	<
Fischer-Brandies (1988)	1896 (1062, 834)	0–14 yrs	<	—	—
Fischer-Brandies et al. (1986)	90 (42, 48)	2–16 mos	<	—	—
Ghaib (2003)	50 (25, 25)	14–18 yrs	—	—	<
Hashimoto et al. (2014)	9	Adults	<	—	<
Jesuino & Valladares-Neto (2013)	30	6–11 yrs	<	—	—
Klingel et al. (2017)	40 (20, 20)	Infants	<	<	<
Panchon-Ruiz et al. (2000)	57 (38, 19)	18–36 yrs	<	<	<
Redman et al. (1965)	48 (20, 28)	17+ yrs	<	<	<
Takizawa et al. (2021)	12 (7, 5)	M 9.72 (SD: 1.48)	<	—	—
Uong et al. (2001)	11	2–4 yrs	>	—	<
Westerman et al. (1975)	40 (19, 21)	16–29 yrs	<	<	<

Note. This table includes reference to source, number of participants with DS, ages studied (years [yrs], months [mos], gestational weeks [gs]), and palatal dimensions of length, height, and width. A less than sign (<) indicates a smaller value in DS, and a greater than sign (>) indicates a larger value in DS. Dash indicates dimension not examined or reported. DS = Down syndrome.

Funding Statement

Research reported in this publication was supported by awards from Grants R01DC006282 and 3R01DC006282-15S1 (Vorperian, PI) from the National Institute on Deafness and Other Communication Disorders, and Waisman Center Core Grant U54 HD090256 (Chang, PI) from the National Institute of Child Health and Human Development.