Articulatory and segmental performance in children with and without speech disorder: A multiple case pilot study

ABSTRACT

This multiple case pilot study explored how nonword imitation influences articulatory and segmental performance in children with and without speech disorder. Eight children, ages 4- to 8-years-old, participated, including two children with childhood apraxia of speech (CAS), four children with phonological disorder (PD), and two children with typical development (TD). Tokens included two complexity types and were presented in random order. Minimal feedback was provided and nonwords were never associated with a referent. Kinematic and transcription data were analysed to examine articulatory variability, segmental accuracy, and segmental variability in session 1 and session 5. Descriptive statistics, percent change, effect sizes, and Pearson correlations are reported. In session 1, the two participants with CAS showed high articulatory variability, low segmental accuracy, and high segmental variability compared to the participants with PD and TD. By session 5, both participants with CAS, two with PD, and one with TD showed increased articulatory variability in the lowest complexity nonword. Segmental accuracy remained low and variability remained high for the two participants with CAS in session 5, whereas several participants with PD and TD showed improved segmental performance. Articulatory and segmental variability were not significantly correlated. The results of this study suggest that motor practice with minimal feedback and no assignment of a lexical referent can instantiate positive changes to segmental performance for children without apraxia. Positive changes to segmental performance are not necessarily related to increased articulatory control; these two processing levels can show distinct and disparate learning trajectories.

KEYWORDS:

• Speech motor control
• speech motor learning
• segmental variability
• childhood apraxia of speech
• phonological disorder

Investigations concerning the development of speech motor control and its relationship with speech sound production are relevant to children with typical development (TD) as well as children with speech sound disorders. Speech motor control is not fully developed until late adolescence (Walsh & Smith, 2002), though most speech sounds are acquired by age five (Crowe & McLeod, 2020). Assessing motor speech, speech sound accuracy, and phonology informs differential diagnosis of childhood apraxia of speech (CAS), articulation disorder, or phonological disorder (PD) and guides intervention decisions (i.e. motor-based versus linguistic-based approach). To continue to improve differential diagnosis and treatment, we must develop a more nuanced understanding of how children with CAS acquire new speech targets compared to children with PD and children with TD. The purpose of this multiple case pilot study was to explore how independent motor practice of simple nonwords that are never assigned a referent influences articulatory (i.e. speech motor variability) and segmental (i.e. speech sound accuracy and variability) performance in a small number of children with CAS, children with PD, and children with TD.

Though researchers have historically studied speech motor control and language as separate phenomena, speech motor control develops for the express purpose of producing spoken language. Indeed, interactions between speech motor skills and language skills have been well-documented in the literature (e.g. Heisler et al., 2010; Maner et al., 2000; Vuolo & Goffman, 2018). Assessing articulatory and segmental performance is particularly relevant to children with childhood apraxia of speech (CAS) and children with phonological disorder (PD). Children with CAS and children with PD are differentially diagnosed as having a core impairment at either lower-level speech motor or higher-level linguistic levels of language production, respectively (though note the underlying impairment for each of these groups has been attributed to various processing levels, e.g. McNeill et al., 2009; Munson et al., 2005; Namasivayam et al., 2020). Children with CAS have a neurological motor speech disorder characterised by impaired spatiotemporal planning and programming (American Speech-Language-Hearing Association [ASHA], 2007). Speech features associated with CAS include, for example, inconsistent errors across multiple tokens (referred to henceforth as segmental variability), lengthened or disrupted coarticulatory transitions, prosodic disturbances, vowel errors, and voicing errors (e.g. ASHA, 2007; Shriberg et al., 2011). None of these features are necessary or sufficient for CAS diagnosis, and children with CAS represent a highly heterogeneous group. Further complicating accurate diagnosis, children with CAS can show concomitant phonological errors in addition to impaired speech motor control (e.g. Grigos et al., 2015; Vuolo & Goffman, 2018). Children with PD are also highly heterogeneous, but they show impairments in higher-level phonological processing skills such as phonological discrimination, encoding, and memory (Brosseau-Lapré & Rvachew, 2017; Shriberg et al., 2009) in the absence of motor speech involvement. Despite the heterogeneity observed within these diagnostic categories, children with CAS should universally show an overt impairment in speech motor control and children with PD should not.

Nonword repetition accuracy is influenced by how the nonwords are constructed and by an individual’s linguistic skills. The nonwords in the syllable repetition task (SRT; Shriberg et al., 2009) were designed to include only early developing sounds. This removes the confound that occurs when assessing nonword repetition accuracy in speakers who may not be able to produce all phonemes. In terms of linguistic skills, nonword repetition accuracy and receptive vocabulary skills are positively correlated at young ages, with this relationship decreasing over time (e.g. reviewed in Coady & Evans, 2008). Though many factors influence nonword repetition performance, nonwords that are never paired with a lexical referent or otherwise assigned meaning have been used in numerous kinematic studies. Nonwords are frequently chosen over real words to reduce the familiarity effects of real words and thus more directly examine speech motor planning/programming (e.g. Case & Grigos, 2020; Sasisekaran et al., 2019; Vuolo & Goffman, 2020; Walsh et al., 2006).

Principles of motor learning

Principles of motor learning research demonstrates that performance during acquisition of a new motor skill does not predict how well the skill will ultimately be learned (Maas et al., 2008; Schmidt et al., 2019). Rather, learning must be assessed following a period of practice. Examining motor performance at multiple time points is particularly relevant to children with CAS, who tend to require more intensive therapy than children with PD to achieve optimal outcomes (ASHA, 2007). A number of practice and feedback conditions can be manipulated to facilitate motor skill acquisition. Compared to a blocked practice schedule, in which the learner practices a single movement over multiple trials, a random practice schedule requires the participant to plan and programme a different movement sequence in each trial. Random practice increases the difficulty of the task and ultimately improves long-term retention of the motor skill (Maas et al., 2008). Knowledge of results, which provides information only about the adequacy of the response, also facilitates long-term retention by increasing the difficulty of the task and requiring the learner to self-evaluate (Maas et al., 2008; Schmidt et al., 2019). Several studies have demonstrated that incorporating principles of motor learning with speech practice can enhance motor learning and speech sound accuarcy for children with CAS (e.g. Case & Grigos, 2016; Maas et al., 2012; Maas & Farinella, 2012; Vuolo & Goffman, 2017).

Articulatory variability

Articulatory variability refers to the spatial and temporal movement variability present within multiple productions of the same syllable, word, phrase, or sentence. The spatiotemporal index (STI; Smith et al., 1995) is a measure used to quantify articulatory variability. Spatial and temporal parameters of articulatory movements are captured using physiologic motion tracking equipment and multiple productions of the same token are examined to determine how well these productions converge onto a stable underlying pattern of movement. Typical adult speakers produce ten repetitions of sentences such as ‘Buy Bobby a puppy’ with highly stable and patterned articulatory movements, reflecting mature speech motor control (Smith & Goffman, 1998). In contrast to adults, children with TD show more variable articulatory movements and thus a higher STI (e.g. Goffman & Smith, 1999).

A body of literature has investigated factors that influence articulatory variability in children with speech and language disorders, such as children who stutter (e.g. Sasisekaran et al., 2019) and children with DLD (e.g. Goffman et al., 2007; Heisler et al., 2010), compared to children with TD. Most relevant to the current investigation, children with CAS, a disorder characterised by impaired spatiotemporal planning and programming (ASHA, 2007), exhibit high articulatory variability relative to children with TD, children with PD, and children with DLD (e.g. Case & Grigos, 2016, 2020; Grigos et al., 2015; Terband et al., 2011; Vuolo & Goffman, 2017, 2018; though these differences are not always statistically significant). Task demands, phonetic complexity, and the specific articulator or combination of articulators studied (e.g. jaw variability versus lip aperture variability) differed across these studies, yet a common finding was that children with CAS exhibited poor spatiotemporal control relative to children without apraxia. In the current investigation, we predict the participants with CAS will show high articulatory variability compared to participants with PD and TD in session 1, reflecting the presence of a low-level motor speech impairment in children with CAS.

Speech motor control in children with CAS should be less amenable to change following long-term practice due to the enduring nature of their motor speech disorder. Two studies have documented a lack of improved articulatory control in children with CAS following long-term practice (Case & Grigos, 2016; Vuolo & Goffman, 2017). Case and Grigos (2016) taught children with CAS and children with TD two novel words, one with high and one with low phonetic complexity, over three practice sessions and did not observe changes to lip and jaw movements in the eight children with CAS. Vuolo and Goffman (2017) examined changes to articulatory variability as children with CAS, children with PD, and children with TD imitated and retrieved two-word phrases; the two children with CAS showed increased articulatory variability following five sessions of imitation practice. Importantly, in both studies the children with CAS showed gains at the segmental level, quantified as increased segmental accuracy and decreased segmental variability, following practice. Improved segmental performance was observed despite differences in task demands, types of stimuli, and types and amount of feedback provided across these two studies.

Articulatory control is affected by the complexity of nonwords participants are asked to imitate; this is observed across a variety of populations including children with TD, children with CAS, and adults (Case & Grigos, 2016, 2021; Maner et al., 2000; Walsh et al., 2006). Children with CAS show increased errors with increased planning and programming demands (Davis et al., 1998), and therefore the complexity of the nonwords they are asked to repeat is particularly likely to influence articulatory control. For example, Case and Grigos (2016) found that children with CAS and children with TD produced more stable articulatory movements in a low complexity (/bɑ.dɑ.bɑp./) versus a high complexity (‘madeepoom’) nonword. In the current study, we include nonwords with two different levels of complexity. Two ‘reduplicated’ nonwords, /mɑ.mɑ.mɑ./ and /bɑ.bɑ.bɑ/, are simple enough for even children with severe CAS to imitate successfully. Reduplication is characterised by alternating consonants and vowels that do not vary, emerges early in development, and requires minimal planning and programming demands (e.g. MacNeilage & Davis, 1990). Four ‘variegated’ nonwords, adapted from the SRT (Shriberg et al., 2009), contain a different consonant in every syllable with the same vowel (e.g. /mɑ.dɑ.bɑ/). We predict the two participants with CAS will produce reduplicated nonword strings with higher articulatory variability in session 1 compared to participants with PD and TD. However, in session 5 we predict that reduplicated nonwords will be more amenable to improved articulatory control compared to variegated nonwords. If children with CAS show decreased articulatory variability following the practice of reduplicated but not variegated nonwords, this would provide preliminary evidence that children with CAS can refine their articulatory control with practice when planning and programming demands are exceedingly low. We predict that participants with PD and TD will show few changes in articulatory variability in reduplicated or variegated nonwords; for these groups, relatively stable articulatory control at the outset will account for this lack of change.

Relationship between articulatory and segmental variability

The precise nature in which articulatory and segmental variability are interconnected has been a long-standing question (e.g. Goffman et al., 2007; Green, 2003; Kent, 1992). In typical development, there is evidence that three-year-old children produce highly variable segmental errors across multiple productions of a word, such as producing ‘ehwifun’, ‘elfint’, and ‘ehwidun’ for elephant (Macrae, 2013; McLeod & Hewett, 2008; Sosa, 2015; Van Haaften et al., 2019). However, considerable individual variation exists (Sosa, 2015) and some studies have documented quite stable productions at this age (e.g. Holm et al., 2007). One presumption is that variability in the speech motor system drives the variable realisation of segments (Kent, 1992). Several recent studies, however, have failed to document a clear relationship between articulatory and segmental variability in children with CAS, PD, DLD, or TD (Benham et al., 2018; Case & Grigos, 2016, 2021; Goffman et al., 2007; Heisler et al., 2010; Vuolo & Goffman, 2017, 2020). In other words, children’s productions can be perceptually accurate but produced with highly variable articulatory movements or can contain variable segmental errors but be produced with relatively low articulatory variability.

Specifying the relationship between articulatory and segmental variability has important diagnostic and intervention implications for children with motor-based versus linguistic-based speech disorders. For children with CAS in particular, an impairment in the ability to plan and programme spatial and temporal parameters of speech movements is thought to give rise to variable segmental errors (ASHA, 2007). In fact, children with CAS show high segmental variability relative to peers without apraxia across many different types of speech tasks, including syllable sequencing, production of monosyllabic and multisyllabic real words, production of ‘buy Bobby a puppy’, and in connected speech (Grigos et al., 2015; Iuzzini & Forrest, 2010; Iuzzini-Seigel, 2021; Iuzzini-Seigel et al., 2017; McNeill et al., 2009). However, as noted previously, empirical data has thus far not supported the notion that speech motor variability results in variable segmental errors (Case & Grigos, 2016, 2020; Goffman et al., 2007; Heisler et al., 2010; Vuolo & Goffman, 2017). It is also important to note that most of studies concerning children with CAS have investigated segmental variability in the context of real words. Though ecologically valid, the presence of high segmental variability in real words may be driven by linguistic factors such as shallow phonological or semantic representations (e.g. Forrest et al., 1997).

In the current investigation, high segmental variability was inclusionary for the two participants with CAS. The intention here was not to introduce circularity. Rather, defining children with CAS in a way that is consistent with many recent studies (e.g. Barrett et al., 2020; Grigos & Case, 2018; Hildebrand et al., 2020; Iuzzini-Seigel, 2021) allows us to specifically assess the relationship between segmental and articulatory variability during the practise of phonemically and motorically simple nonwords. In session 5, we expect to observe few relationships between segmental and articulatory performance in participants with CAS, PD, or TD (Case & Grigos, 2016, 2020; Sasisekaran et al., 2019; Vuolo & Goffman, 2017).

Current study

Many details regarding how task demands, stimuli complexity, practice schedule, and so forth influence articulatory and segmental performance in children with CAS compared to peers without apraxia have yet to be established. In the current study, children practised repeating nonwords words in the absence of prompts intended to improve speech accuracy or assignment of visual referents. The nonwords were modified from the SRT (Shriberg et al., 2009) so that only early developing sounds were used, which ensured that even children with significant CAS could complete the task. Because we asked participants to imitate quite simple nonwords we opted to make the task as challenging as possible beginning in session 1 by using a random presentation order and minimal knowledge of results feedback (only general information about accuracy) each session. Articulatory variability, segmental accuracy, and segmental variability were assessed in session 1 and session 5. The goal of this pilot work was to explore how imitation practice in the absence of direct word learning influences articulatory and segmental performance in children with motor-based versus linguistic-based speech disorders.

Method

Participants

Eight children, ranging in age from 4;8 (years;months) to 8;9, participated in this multiple case pilot study, including two children with CAS (two males), four children with PD (four males), and two children with TD (one male, one female). The Institutional Review Board at Purdue University approved all tasks and procedures, and parental consent and child assent were obtained prior to participation.

All subjects met the following inclusionary criteria: monolingual English speaker, normal or corrected-to-normal vision, passed a bilateral hearing screening (20 dB HL at 500, 1000, 2000 and 4000-Hz), scored within the average range on the structural component of the Robbins and Klee (1987) oral motor protocol (e.g. shows normal size and symmetry of the mandible and maxilla, normal symmetry of the velum, no tongue atrophy or fasciculations, etc.), and had no history of neurologic dysfunction, head injury, or autism spectrum disorder per parent report. Nonverbal cognitive ability is reported in Table 1 and was assessed using either the Columbia Mental Maturity Scale – Third Edition (Burgemeister et al., 1972), Primary Test of Nonverbal Intelligence (Ehrler & McGhee, 2008), or Wechsler Preschool and Primary Scale of Intelligence – Third Edition (administered by a licenced school psychologist; Wechsler, 1991). All participants with PD and participants with TD showed average receptive language abilities on the Test of Auditory Comprehension of Language – Third Edition (TACL-3; Carrow-Woolfolk, 1999); see, Table 1. For the two participants with CAS, recent language scores (within the last six months) are reported due to time constraints within each session. Both participants with CAS showed below average receptive/expressive language abilities; see, Table 1.

Table 1. Behavioural data by participant.

P	Age/Sex	NVBL IQ	Language	BBToP CI	DEAP Artic	DEAP WI	Mayo 10	Functional Oral Mech
CAS 1	5;10 M	52^a	66/50^d	66^h	55	60	5	> −3 SD
CAS 2	5;10 M	86^b	84^e	<55^i	55	56	5	> −3 SD
PD 1	5;0 M	112^c	113^f	76	70	28	0	Pass
PD 2	6;7 M	101^c	115^f	77	55	30	1	−1 SD
PD 3	7;11 M	108^c	109^f	82	55	8	1	Pass
PD 5	8;9 M	105^c	111^g	79	55	32	0	Pass
TD 1	4;8 F	100^c	117^f	90	80	20	2	> −1.5 SD
TD 2	7:4 M	106^c	119^f	93	95	0	0	Pass

P = participant; CAS = childhood apraxia of speech; PD = phonological disorder; TD = typical development; NVBL IQ = nonverbal intelligence; BBToP CI = Bankson-Bernthal Test of Phonology Consonant Inventory; DEAP Artic = Diagnostic Evaluation of Articulation and Phonology (DEAP) Articulation subtest; DEAP WI = DEAP Word Inconsistency subtest; Functional Oral Mech = Total Functional Score, Robbins and Klee (1987); a = Wechsler Preschool and Primary Scale of Intelligence – Third Edition; b = Primary Test of Nonverbal Intelligence; c = Columbia Mental Maturity Scale – Third Edition; d = Preschool Language Scales – Fifth Edition Auditory Comprehension/Expressive Communication standard score; e = Test of Auditory Comprehension of Language – Third Edition (TACL-3) grammatical morphemes subtest; f = TACL-3; g = Clinical Evaluation of Language Fundamentals – Fourth Edition core language score; h = Structured Photographic Articulation Test; i = Clinical Assessment of Articulation and Phonology (Secord and Donohue, 2002).

The two participants with CAS were recruited through Apraxia-Kids and had been previously diagnosed with CAS by a licenced and certified Speech-Language Pathologist. To confirm CAS diagnosis, these two participants were also required to score more than one standard deviation below the mean on the Bankson-Bernthal Test of Phonology (BBToP; Bankson & Bernthal, 1990) and the Diagnostic Evaluation of Articulation and Phonology (DEAP; Dodd et al. 2006) Articulation subtest, receive a DEAP Word Inconsistency (DEAP WI)DEAP score greater than 40%, show four or more CAS features on the Mayo 10 checklist (Shriberg et al., 2011), and score more than one standard deviation below the mean on the functional component of the oral motor protocol (Robbins & Klee, 1987). Mayo 10 features were assessed across the BBToP, DEAP, and oral motor protocol; a feature was required to occur in at least one task to be marked as present. Items on the functional component of the Robbins and Klee (1987) protocol include both oral functions (e.g. protrude lips, bite lower lip, elevate tongue to alveolar ridge, protrude tongue, place tongue between teeth) and speech functions (e.g. produce /u/, /f/, /t/, diadochokinetic rate testing). A cut-off of 40% on the DEAP WI, in conjunction with impaired oral motor skills, is used by some research groups to classify CAS (e.g. Dodd, 1996; McNeill et al., 2009).

The PD participants were recruited from local schools and clinics in the Greater Lafayette, IN area by mailing recruitment fliers to speech-language pathologists and by posting fliers in the M. D. Steer Speech, Language, and Swallowing Clinic at Purdue University. To confirm PD diagnosis, these participants were required to score greater than one standard deviation below the mean on the BBToP and the DEAP Articulation subtest, receive a DEAP WI score less than 40%, show fewer than four CAS features on the Mayo 10 checklist, and pass the functional component of the oral motor protocol. One participant with PD, PD 2, scored 1 SD below the mean on the functional component of the Robbins and Klee (1987) oral motor protocol.

The two participants with TD were recruited by posting research advertisements in Purdue Today, a faculty and staff newsletter, and by contacting appropriate families in the Speech, Language and Hearing Sciences research database. These two participants had no history of speech or language difficulties per parent report. Both participants scored within the average range on the BBToP, received a DEAP WI score of less than 40%, and showed fewer than four CAS features on the Mayo 10 checklist. TD 1 received a standard score of 80 on the DEAP Articulation subtest and a total functional score greater than 1.5 standard deviations below the mean on the functional component of the oral motor protocol. She did not consistently produce /s/ or the later developing sounds /ɹ/, /l/, /θ/, /tʃ/ and /dʒ/ on the BBToP and DEAP and showed emerging ability to produce oral and speech functions such as alternate pucker/smile and alternate /u/, /i/ on the functional component of the Robbins and Klee (1987) oral motor protocol. TD 1 was the youngest participant and there were no concerns from parents, teachers, or the test administrator (a certified Speech-Language Pathologist with over 10 years of clinical experience) regarding her speech and language development.

Procedure

Participants attended five individual sessions, each lasting approximately 60 minutes. Sessions were scheduled on separate days ideally no more than one week apart. The participants in the TD and CAS groups completed the study over an average of six and a half weeks (range: TD = 4–9; CAS = 6–7), and participants in the PD group completed the study over an average of three weeks (range: 1–4). The assessment battery was administered during the remainder of each session. CAS 1 did not complete the experimental tasks in session 4 due to equipment failure, and CAS 2 did not complete session 4 due to a scheduling conflict.

Stimuli and experimental task

The stimuli included six CVCVCV nonwords that were constructed to be amenable to kinematic analysis (described in more detail in the following section). Consonants included /b/, /m/, /d/, and /n/ and every vowel was a stressed /ɑ/. All nonwords were three syllables in length. Four of the nonwords contained a different consonant in each syllable position; these are referred to as variegated nonwords. Two of these variegated nonwords, [mɑ.dɑ.bɑ.] and [mɑ.nɑ.bɑ.], were selected from the Syllable Repetition Task (Shriberg et al., 2009). The first and last syllable of each of these two nonwords was then reversed to create the nonwords [bɑ.dɑ.mɑ.] and [bɑ.nɑ.mɑ.], respectively. Finally, two reduplicated nonwords, [mɑ.mɑ.mɑ.] and [bɑ.bɑ.bɑ.], were included to compare articulatory variability in nonwords that contained very simple reduplicated sequences versus more complex variegated sequences. Because all six nonwords were relatively simple we used a random presentation order in every session so that participants had to continuously generate new motor programmes; this increased the difficulty of the speech motor task. Audio stimuli were recorded in a sound booth by an adult native English female talker and equated for intensity in Praat (Boersma & Weenink, 2022).

In session 1 and session 5 participants imitated all six nonwords. In sessions 2, 3, and 4, participants practised imitating just three nonwords (two variegated and one reduplicated). The three nonwords that were practised in sessions 2, 3, and 4 are referred to as the ‘practised’ condition and the three nonwords that were only encountered in sessions 1 and 5 are referred to as the ‘generalization’ condition. The six nonwords were divided into randomisation A ([bɑ.nɑ.mɑ.], [mɑ.dɑ.bɑ.], and [mɑ.mɑ.mɑ.]) and B ([bɑ.dɑ.mɑ.], [mɑ.nɑ.bɑ.], and [bɑ.bɑ.bɑ.]) to determine which nonwords would be practised and which would be held for generalisation, and the randomisation assigned to each participant was counterbalanced within each group. Randomisation A included participants CAS 1 PD 1, PD 3, and TD 2, and randomisation B included participants CAS 2, PD 2, PD 5, and TD 1.

Children were seated approximately eight feet in front of a thirty-inch Dell computer monitor. The 3D Investigator (Northern Digital Inc., Waterloo, Ontario, Canada), used to collect the kinematic data, was mounted above the monitor. Each nonword was presented eight times in quasi-random order (no more than two occurrences of the same nonword in a row) in four blocks of 12 for a total of 48 presentations. A different abstract image was shown with each presentation of a nonword (48 images total) to maintain the child’s attention to the screen. Children were instructed ‘You are going to hear some pretend words. Try to say the word exactly how the lady says it’. In each session, knowledge of results was only provided the first time the child produced a particular nonword correctly (‘Yes, that was correct’). Children were never explicitly told that their production was incorrect or prompted to correct an error. No other feedback or prompts were provided except occasional verbal encouragement to continue the task.

Transcription and reliability

The author, a certified Speech-Language Pathologist, phonemically transcribed sessions 1 and 5 for each participant. Productions that contained disfluencies, laughter, whisper, or yelling were discarded. Segmental accuracy was calculated for each nonword as the percent phonemes correct (PPC), quantified as the number of phonemes produced correctly divided by the total number of phonemes and multiplied by 100. Because one of the aims the current study was to examine segmental variability across the three participant groups, additions were counted as errors. For example, one phoneme error occurred if /mɑ.dɑ.bɑ./ was produced as [mɑn.dɑ.bɑ.] or [mɑ.dɑ.bɑn.]. Segmental variability was indexed using the type-token ratio (TTR; Johnson, 1944), calculated as the total number of unique productions (i.e. differed by at least one phoneme) divided by the total number of productions and multiplied by 100. To avoid penalising a child for producing the phrase the same way each time (i.e. one divided by eight yields an TTR of 12.5%), the occurrence of no variability was set to zero by subtracting one from the numerator before dividing by the denominator. For example, a child who produced /mɑ.dɑ.bɑ./ as [mɑ.dɑ.bɑ.] eight times would receive a TTR score of (1–1)/8, or 0%. A child who produced /mɑ.dɑ.bɑ./ as [mɑ.dɑ.bɑ.] three times, [mɑn.dɑ.bɑ.] four times, and [bɑ.dɑ.bɑ.] one time would receive a TTR score of (3–1)/8, or 25%. Note that by setting the floor to zero the ceiling becomes 87.5% rather than 100% (i.e. (8–1)/8).

A trained master’s student in Speech-Language Pathology, who was blind to study aims and group assignment, completed the transcription reliability for this study. Reliability was conducted on 25% of the data (four of the 16 total sessions). Two of the four sessions were randomly selected from the children with PD because they represent half of the participants in the study, and one session was randomly selected from the participants with CAS and with TD. Phoneme-by-phoneme percentages of agreement were obtained. These were judged as narrow agreement that the two transcribers transcribed the same phoneme for the target. Reliability averaged 96.6%.

Kinematic data collection and processing

Three infrared light emitting diodes (IREDs) recorded the lip and jaw movements: one placed at midline on the vermillion border of the upper lip, one at midline on the lower lip and one on a small splint attached underneath the jaw with medical adhesive. Five additional IREDs served as a reference frame to subtract head movement from the lip and jaw movements: one placed on the forehead at midline and four placed on a pair of sports goggles worn by the child. Kinematic data were collected at a sampling rate of 250 Hz. A Butterworth filter was used to low pass filter displacement data with a cut-off frequency of 10 Hz (forward and backward). A time-locked acoustic signal was collected at a sampling rate of 16,000 Hz.

The movements associated with each nonword were extracted from the continuous speech data files in MATLAB ® (MathWorks, 2019). Movement onsets and offsets were initially selected via visual inspection of the lower lip displacement record. An algorithm determined the peak velocity within a 25-point (100-ms) window of the examiner selected point. As illustrated in the middle panel of Figure 1, the peak velocity of the first opening movement marked the onset of the nonword (e.g. in [mɑ.dɑ.bɑ.], which corresponds to the release of the initial [m] to the [ɑ]). The peak velocity of the last opening movement marked the offset of the nonword (e.g. in [mɑ.dɑ.bɑ.], which corresponds to the release of the final [b] to the [ɑ]). Productions that contained missing signals from the lip or jaw IREDs, disfluencies, laughter, whisper, yelling, or did not begin or end with a [b], [p], or [m] were excluded from analysis. Consistent with other kinematic studies (Case & Grigos, 2016; Vuolo & Goffman, 2018, 2020), productions that contained segmental errors were included if the place of articulation aligned with the target labial or alveolar. For example, for the target [mɑ.dɑ.bɑ.], [bɑ.dɑ.bɑ.] was included but [mɑ.nɑ.dɑ.] or [mɑ.bɑ.mɑ.] was not. The minimum number of records that were extracted for the kinematic analysis was four. The STI (Smith et al., 1995) was calculated from lip aperture. Lip aperture is calculated for each extracted record as the sample-by-sample subtraction of the lower lip displacement signal from the upper lip displacement signal. The lip aperture trajectories were amplitude- and time-normalised to eliminate differences in speech rate and loudness, respectively, which reveals the underlying spatiotemporal patterning. Standard deviations were computed at 2% intervals across the normalised records. The sum of these 50 standard deviations yields the STI, which quantifies articulatory variability over multiple productions of a nonword. The bottom panel of Figure 1 shows seven normalised productions of [mɑ.dɑ.bɑ.] produced by PD 1 in session 1 and the associated STI. In the current study, the STIs for the two variegated nonwords in the practiced condition and the two variegated nonwords in the generalisation condition were averaged separately.

Figure 1. Example of the kinematic data extraction. The top and middle panels show the lower lip displacement and velocity, respectively, from a child with PD producing [mɑdɑbɑ] in session 1. The bottom panel shows seven normalised productions and the associated STI.

Statistical analyses

Because only one reduplicated nonword was included in the practiced and generalisation conditions we used percent change to capture changes to the STI from session 1 to session 5. Percent change was calculated as the session 5 mean minus the session 1 mean divided by the session 1 mean. For the variegated nonwords, effect sizes (ES) were used to measure change from session 1 to session 5 for the PPC, TTR, and STI measures. We calculated ES as the session 5 mean minus the session 1 mean divided by the weighted (S1 + S5) standard deviation (SD). We used the weighted SD instead of the session 1 SD because there were several instances of zero variability in session 1, particularly in the PPC and TTR measures. Following Maas and Farinella (2012) and Maas et al. (2012), we interpreted an ES > 1 or < −1 as significant (i.e. the magnitude of change from session 1 to session 5 exceeded the SD). A positive percent change score/ES indicates the STI, PPC, or TTR increased from session 1 to session 5; a negative percent change score/ES indicates the STI, PPC, or TTR decreased. Finally, to explore relationships between articulatory and segmental variability, Pearson correlations were calculated between all STI and TTR values in session 1 and session 5 for the CAS participants compared to the PD and TD participants. We first aligned the TTR values to the number of productions that were amenable to kinematic analyses, then excluded all pairs with a TTR of zero.

Results

Articulatory variability

In session 1, the percentage of productions amenable to kinematic analysis was 57% (55/96) for the two participants with CAS, 92% (177/192) for the four participants with PD, and 95% (91/96) for the two participants with TD. In session 5, the percentage of usable productions was 69% (66/96) for the two participants with CAS, 97% (187/192) for the four participants with PD, and 99% (95/96) for the two participants with TD. The CAS participants had fewer usable productions due to the presence of many initial and final [d] and [t] substitution errors, which are not amenable to kinematic analysis.

Articulatory variability in session 1

Table 2 shows the mean STI for all six nonwords and for the reduplicated and variegated nonwords separately by participant in session 1. In session 1, the two participants with CAS showed the highest articulatory variability compared to participants with PD and participants with TD. TD 1, the youngest participant, also showed relatively high STI values in the variegated nonwords.

Table 2. Articulatory variability (STI), segmental accuracy (PPC), and segmental variability (TTR) means and SDs in session 1 by particpant.

P	S1 STI All	S1 STI Reduplicated	S1 STI Variegated	S1 PPC All	S1 TTR All
CAS 1	19.59 (3.97)	16.87 (2.35)	22.32 (3.46)	81.09 (16.06)	25.79 (30.73)
CAS 2	18.04 (1.34)	18.65 (1.44)	17.64 (1.40)	74.75 (16.24)	41.67 (22.35)
PD 1	15.39 (2.75)	15.21 (1.20)	15.47 (3.48)	95.83 (4.75)	12.50 (15.81)
PD 2	13.72 (5.34)	14.83 (8.35)	13.16 (4.81)	95.49 (6.77)	6.25 (6.85)
PD 3	13.62 (2.70)	14.70 (3.54)	13.08 (2.60)	90.58 (8.49)	27.98 (24.81)
PD 5	13.82 (3.03)	13.55 (1.79)	13.96 (3.76)	95.83 (4.17)	16.67 (15.14)
TD 1	16.52 (4.14)	14.29 (2.32)	17.63 (4.67)	95.84 (6.04)	13.89 (17.41)
TD 2	15.03 (4.12)	12.02 (1.87)	16.53 (4.26)	97.92 (3.23)	5.00 (11.18)

P = participant; CAS = childhood apraxia of speech; PD = phonological disorder; TD = typical development; S1 = session 1; STI = spatiotemporal index; Reduplicated = same C and V ([bɑ.bɑ.bɑ.]); Variegated = different C same V ([mɑ.dɑ.bɑ.]); All = mean of all 6 nonwords, PPC = percent phonemes correct; TTR = type-token ratio. SDs are shown in parentheses.

Articulatory variability in session 1 versus session 5

Figure 2 shows the STI in session 1 versus session 5 for the reduplicated and variegated nonwords in practice and generalisation. Table 3 shows the mean STI values in session 1 and session 5 and percent change for the reduplicated nonwords (left panel) and effects sizes for the variegated nonwords (right panel) in the practiced and generalisation conditions.

Figure 2. STI in session 1 versus session 5 for the reduplicated and variegated nonwords in practice and generalisation.

Table 3. Mean articulatory variability (STI) in session 1 and session 5, percent change for the reduplicated nonwords, and effect sizes for the variegated nonwords in the practised and generalisation conditions.

Reduplicated Nonwords							Variegated Nonwords
Practised			Generalisation				Practised			Generalisation
P	S1	S5	%change	S1	S5	%change	S1	S5	ES	S1	S5	ES
CAS 1	15.20	16.85	10.86	18.53	17.10	−7.72	19.87*	20.26*	[1.96]	24.76*	23.69*	[−4.32]
CAS 2	17.63	21.36	21.16	19.66	11.87	−39.62	19.16*	15.33	−11.07^+	16.88	21.12	1.30
PD 1	14.36	9.90	−31.06	16.06	16.27	1.31	13.81	14.97	0.19	17.14	14.91	−0.18
PD 2	20.73	10.46	−49.54	8.92	11.74	31.61	10.35	14.14	−0.06	15.98	13.49	−0.37
PD 3	12.20	21.93	79.75	17.20	11.50	−33.14	12.41	15.38	0.43	13.75	14.72	0.33
PD 5	12.28	15.98	30.13	14.81	12.33	−16.75	11.83	13.62	0.41	16.08	15.69	−0.06
TD 1	15.93	15.23	−4.39	12.65	16.53	30.67	14.98	18.36	0.93	20.28	24.29	0.42
TD 2	10.69	12.14	13.56	13.34	18.81	41.00	19.35	11.06	−1.27	13.72	13.39	−0.23

Reduplicated = same C and V ([bɑ.bɑ.bɑ]); Variegated = different C same V ([mɑ.dɑ.bɑ]); P = participant; CAS = childhood apraxia of speech; PD = phonological disorder; TD = typical development; S1 = session 1; S5 = session 5; %change = percent change from session 1 to session 5. ES = effect size. A positive %change/ES indicates that articulatory variability increased from session 1 to session 5; a negative %change/ES indicates that articulatory variability decreased. Bolded ES are > 1 or < −1 and are judged to be significant. *Only one of the variegated nonwords had enough usable productions to include in the STI; therefore, %change was calculated for these productions and is reported in brackets. ⁺ES was calculated using only the S5 SD.

Reduplicated nonwords

CAS participants

In the practiced condition, both participants with CAS showed increased articulatory variability (CAS 1 = 10.86%; CAS 2 = 21.16%) from session 1 to session 5. In the generalisation condition, both participants with CAS showed decreased articulatory variability (CAS 1 = −7.72%; CAS 2 = −39.62%).

PD and TD participants

For the participants with PD and TD, mixed changes were observed in the practiced condition. PD 3 (79.75%), PD 5 (30.13%), and TD 2 (13.56%) showed increased articulatory variability whereas PD 1 (−31.06%), PD 2 (−49.54%), and TD 1 (−4.39%) showed decreased articulatory variability.

Mixed changes were also observed in the generalisation condition. PD 1 (1.31%), PD 2 (31.61%), TD 1 (30.67%), and TD 2 (41.00%) showed increased articulatory variability whereas PD 3 (−33.14%) and PD 5 (−16.75%) showed decreased articulatory variability.

Variegated nonwords

CAS participants

In the practiced condition, the two participants with CAS showed different patterns of change from session 1 to session 5. CAS 1 showed a significant increase (ES = 1.41) and CAS 2 showed a significant decrease (ES = −11.07) in articulatory variability. In the generalisation condition, CAS 2 (ES = 1.30) showed a significant increase in articulatory variability.

PD and TD participants

Only one participant with PD or TD showed a significant effect size in the practiced condition. TD 2 (ES = −1.27) showed decreased articulatory variability following practice. No participant with PD or TD showed a significant ES in the variegated nonwords in the generalisation condition.

Segmental accuracy and variability

In session 1, the percentage of productions amenable to phonemic transcription was 86% (83/96) for the two participants with CAS, 99% (191/192) for the four participants with PD, and 98% (94/96) for the two participants with TD. In session 5, the percentage of productions amenable to phonemic transcription was 99% (95/96) for the two participants with CAS, 99% (190/192) for the four participants with PD, and 98% (94/96) for the two participants with TD.

Segmental accuracy and segmental variability in session 1

Table 2 shows the mean PPC and TTR for all six nonwords by participant in session 1. Standard deviations are high because segmental accuracy was close to or at ceiling and segmental variability was close to or at floor for the two reduplicated nonwords but the four variegated nonwords were more challenging. In session 1, both participants with CAS showed low segmental accuracy relative to the other participants. The two participants with CAS also showed high segmental variability in session 1, though PD 3 showed segmental variability levels comparable to the two participants with CAS.

Segmental accuracy and segmental variability in session 1 versus session 5

Table 4 shows the mean PPC (left panel) and TTR (right panel) in session 1 and session 5 and effect sizes for the variegated nonwords in the practised and generalisation conditions. The reduplicated nonwords are not considered here because performance was at ceiling (100%) for PPC and floor (0%) for TTR for most participants.

Table 4. Mean segmental accuracy (PPC) and segmental variability (TTR) in session 1 and session 5 and effect sizes for the variegated nonwords in the practised and generalisation conditions.

PPC							TTR
Practised			Generalisation				Practised			Generalisation
P	S1	S5	ES	S1	S5	ES	S1	S5	ES	S1	S5	ES
CAS 1	73.81	75.00	0.05	69.45	66.67	−0.18	35.72	31.25	−0.08	41.67	25.00	−1.41
CAS 2	66.97	53.13	−1.29	62.50	61.46	−0.08	40.18	44.65	0.10	66.07	37.50	−2.26
PD 1	91.67	100.00	2.83	95.84	95.83	0.00	25.00	0.00	−1.41	12.50	18.75	0.24
PD 2	95.84	100.00	0.71	90.63	92.71	0.18	6.25	0.00	−0.71	12.50	12.50	[0.00]
PD 3	81.10	96.88	9.33	90.63	87.80	−1.49	52.68	12.50	−2.89	31.25	18.75	−0.71
PD 5	96.88	100.00	0.71	91.67	100.00	2.83	12.50	0.00	−0.71	31.25	0.00	−3.54
TD 1	98.96	92.71	−0.53	88.55	81.25	−0.99	6.25	6.25	0.00	35.42	18.75	−0.57
TD 2	93.75	100.00	[6.67]	100.00	92.71	−4.96	25.00	0.00	[−25.00]	0.00	18.75	2.12

P = participant; CAS = childhood apraxia of speech; PD = phonological disorder; TD = typical development; PPC = percent phonemes correct; TTR = type-token ratio; S1 = session 1; S5 = session 5; ES = effect size. A positive ES indicates that segmental accuracy/segmental variability increased from session 1 to session 5; a negative ES indicates that segmental accuracy/segmental variability decreased. Bolded ES are > 1 or < −1 and are judged to be significant. Values reported in brackets are %change scores because the SD was zero in session 1 and session 5.

Segmental accuracy

CAS participants

In the practised condition, CAS 1 showed no significant change and CAS 2 showed a significant decrease (ES = −1.29) in segmental accuracy from session 1 to session 5. Overall, by session 5 segmental accuracy remained quite low for the two participants with CAS, ranging from 53–75%, compared to participants with PD and TD, whose accuracy ranged from 92–100%.

In the generalisation condition, neither participant with CAS showed a significant change in segmental accuracy from session 1 to session 5.

PD and TD participants

In the practiced condition, PD 1 (ES = 2.83) and PD 3 (ES = 9.33), showed significant ES that reflected increased accuracy. No other significant ES were observed in the participants with PD or TD.

In the generalisation condition, PD 3 (ES = −1.49) showed decreased segmental accuracy and PD 5 (ES = 2.83) showed increased segmental accuracy. TD 2 (ES = −4.96) was the only participant with TD with significant a ES in the generalisation condition; he showed decreased segmental accuracy.

Segmental variability

CAS participants

In the practised condition, the two participants with CAS continued to show high segmental variability in session 5 with no significant ES observed. In the generalisation condition, CAS 1 (ES = −1.41) and CAS 2 (−2.26) both showed a significant decrease in segmental variability from session 1 to session 5.

PD and TD participants

In the practiced condition, two participants with PD showed significant ES. Segmental variability decreased for PD 1 (ES = −1.41) and PD 3 (ES = −2.89); note that these same participants also showed significant increases to segmental accuracy. TD 2 (−25.00%) also showed decreased segmental variability; his segmental accuracy score reached ceiling in session 5 (6.67%).

Two signifcant ES were observed in the participants with PD and TD in the generalisation condition. PD 5 (ES = −3.54) showed decreased segmental variability and TD 2 (ES = 2.12) showed increased segmental variability.

Relationship between articulatory and segmental variability

Pearson correlations showed no significant relationship between the STI and TTR measures for the participants with CAS (r = −0.431, p = 0.186) or the participants with PD and TD (r = 0.255, p = 0.190).

Discussion

This pilot study examined articulatory variability, segmental accuracy, and segmental variability in children with CAS, children with PD, and children with TD using a multiple case design. Specifically, we examined nonword repetition performance in session 1 and session 5. The present findings, which will be discussed in greater detail below, revealed different profiles of performance for the participants with CAS compared to participants with PD and TD. Clear dissociations between articulatory and segmental performance were observed irrespective of diagnostic category.

Articulatory and segmental performance

In session 1, our prediction that the two participants with CAS would show higher articulatory variability compared to participants with PD and TD was supported. Even these simple nonword constructions were able to capture the motor planning and programming difficulties experienced by the children with CAS. These data are in line with other studies which have shown that children with CAS exhibit high articulatory variability compared to non-apraxic peers in tasks that necessitate production of relatively simple novel words and real words (Case & Grigos, 2016; Vuolo & Goffman, 2017). A growing body of evidence demonstrates that CAS is characterised by an impaired spatiotemporal planning and programming. Impaired planning and programming in children with CAS was inferred initially based on perceptual analyses (Goffman et al., 2007; Green, 2003) but kinematic studies are now providing empirical evidence to support this claim.

We anticipated that the two participants with CAS would be able to refine their articulatory control following multiple sessions of practice of reduplicated nonwords. We based this expectation on the fact that in typical development, reduplicated syllable sequences emerge early in development and require minimal motor planning and programming demands (MacNeilage & Davis, 1990). However, contrary to our prediction, both participants with CAS showed increased articulatory variability in the reduplicated nonwords in session 5. Two participants with PD and one with TD also showed this effect. Rote imitation of simple nonwords disrupted articulatory control in the majority (60%) of participants. Vuolo and Goffman (2017) also documented significantly increased articulatory variability in five out of nine participants (56%), across participants with CAS, PD, and TD, following multiple practice sessions imitating two-word phrases. Taken together, we show that rote imitation without a linguistic goal disrupts articulatory performance in nonwords and real words, at least over five practice sessions. One possibility is that we are observing only part of a U-shaped learning trajectory (Gershkoff-Stowe & Thelen, 2004). In other words, articulatory variability may increase initially as children explore the movement parameters associated with novel words. However, additional practice could allow children to refine their spatiotemporal control, which would result in decreased articulatory variability. The clinical implications of these findings are discussed in more detail in the clinical implications section below.

We expected the two participants with CAS would continue to exhibit high articulatory variability in the variegated nonwords even after five practice sessions. We predicted this because variegated nonwords are more articulatorily complex than reduplicated nonwords. Articulatory control should be resistant to change in individuals with a motor planning/programming impairment, particularly in the context of independent practice. Surprisingly, one participant with CAS showed a significant reduction in articulatory variability. For this participant with CAS, rote imitation improved articulatory control of variegated but not reduplicated nonwords. The only other participant to show this same pattern of performance was the oldest participant with TD. Though we observed improved articulatory control in only a single participant with CAS, we suggest that this finding was unlikely to be spurious. Rather, there are several possible reasons why some children, such as CAS 2 and TD 2 in the current work, show improved articulatory control in higher complexity nonwords and disrupted articulatory control in lower complexity nonwords. Case and Grigos (2020), in their framework of motor complexity, posit that nonwords containing little variation from syllable to syllable require highly precise speech motor control. Though these authors applied this interpretation to the variegated nonwords in their stimuli, it is possible that when stimuli only contain reduplicated and variegated nonwords the motor demands of moving across different consonants in variegated nonwords better reflect the experiences children have producing multisyllabic words than more ‘artificial’ and tightly constrained reduplicated strings. Furthermore, there are likely individual characteristics that can be systematically explored in future work to better understand the specific conditions under which some children with CAS are able to improve articulatory control. Individual characteristics might include, for example, CAS severity, speech perception skills, receptive language skills, et cetera.

Regarding segmental performance, the two participants with CAS did not show positive changes to segmental accuracy or segmental variability following practice. One explanation for this finding is that both children with CAS showed below average receptive language skills, which is known to affect nonword repetition accuracy (e.g. Coady & Evans, 2008). Cychosz and colleagues (Cychosz et al., 2021) propose that children with larger receptive vocabularies are better able to abstract segmental representations to novel words. The two children with CAS may not have benefitted from this bootstrapping effect due to the presence of poor receptive language skills. However, there is at least one caveat to this interpretation. Five out of ten of the participants with suspected CAS in the Rvachew and Matthews (2017) SRT study showed below average receptive language skills yet nine out of ten showed below average segmental accuracy scores for their age. These authors only examined nonword repetition at a single time point, so we do not know whether long-term practice influences segmental performance differently in children with CAS and typical receptive language abilities versus those with receptive language impairment. However, it is unlikely that receptive language skills alone account for the lack of improved segmental performance observed in the current investigation. Speech perception, phonological awareness, and other linguistic skills also influence nonword repetition accuracy (Coady & Evans, 2008; Rvachew & Matthews, 2017).

Relationship between articulatory and segmental variability

The present data highlight clear dissociations between articulatory and segmental levels. Pearson correlations between the STI and TTR measures were not significant for participants with CAS or participants with PD and TD, and descriptive data support these results. For example, in the variegated nonwords that were practiced, CAS 2 showed decreased articulatory variability (reflecting increased speech motor stability) but also a decrease in segmental accuracy (reflecting degraded performance) and no change to segmental variability (which remained quite high at 45%). These findings support the notion that speech motor stability, or variability, cannot be inferred from perceptual measures. Several studies have shown that children can sacrifice articulatory stability to achieve more perceptually accurate speech, or they can implement stable articulatory movements despite the presence of variable speech errors (Benham et al., 2018; Case & Grigos, 2016, 2021; Goffman et al., 2007; Heisler et al., 2010; Vuolo & Goffman, 2017, 2020). These tradeoffs are subconscious and influenced by complex factors. There remains much to learn about how language production levels influence and interact with one another, but empirical evidence shows that perceptual and physiologic measures assess different levels of language production (Benham et al., 2018; Goffman et al., 2007; Heisler et al., 2010; Vuolo & Goffman, 2020).

Clinical implications

In contrast to our findings, nonword imitation has been used in formalised treatment programmes such as Rapid Syllable Transition Treatment (ReST, McCabe et al., 2020) with good success. Nonwords are used in ReST so that children can practice generating novel motor plans with less interference from prior linguistic knowledge (McCabe et al., 2020). Practice is designed to address the three core features of CAS, accuracy and consistency, coarticulation, and prosody, by helping the child focus on the ‘sounds’, ‘beats’, and ‘smoothness’ of their productions. Critically, each session begins with a short teaching phase to develop the child’s internal reference of correctness, which makes the child an active participant in their own learning. Here we see that nonwords have clinical utility if they are utilised in conjunction with training, practice, and explicit prompts and feedback. Currently, ReST shows good efficacy and generalisation to untreated real words in children with CAS (Murray et al., 2015). This improvement is thought to stem from improved motor planning/programming (McCabe et al., 2020). We do not yet know whether children with CAS show reduced articulatory variability following ReST treatment; as discussed earlier, this cannot be presumed based on improved perceptual accuracy.

Limitations and future directions

Every study has limitations. The most obvious limitation in the current study is that a small and heterogeneous sample of children participated. The two participants with CAS also received one less practice session than the other participants, which may have affected the results in session 5. In future work it will be important to explore, in a larger number of participants, the conditions under which children with CAS can increase their speech accuracy and movement stability through independent practice. Age, cognition, and receptive language skills in particular should be more carefully controlled in future work. In the current study, we intentionally did not include knowledge of performance feedback. However, a study design in which children were provided knowledge of performance feedback on only half of the nonwords would have allowed for better assessment of how children with CAS are able to improve segmental performance independently versus with cues. Likewise, children with CAS may be able to make greater gains through independent practice if nonwords are paired with a referent, which would be a more ecologically valid task. In future work, it will be important to continue to examine how task complexity, structure of practice, and feedback schedules influence segmental and motor performance in children with CAS compared to peers without apraxia.

Conclusions

In this multiple case pilot study, we found additional support for the notion that children with CAS show a low-level motor planning and programming disorder characterised by highly variable articulatory control; this was observed in a simple nonword imitation task. Long-term practice resulted in heterogenous changes to articulatory control, with half of the participants showing increased articulatory variability in the reduplicated nonwords in session 5. Articulatory and segmental levels of performance do not necessarily align; this finding will be important to consider in relation to current therapeutic approaches for children with CAS.

Acknowledgments

We thank Allison Gladfelter, Meredith Saletta, Janna Berlin, and Allison Kinross for their assistance with data collection and processing, and the families and children who participated in this study.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This work was supported by the National Institute on Deafness and Other Communication Disorders Grants F31 DC015176 (Vuolo) and R01 DC04826 (Goffman).