Altered auditory feedback studies using sudden, unexpected shifts have explored how participants respond when their auditory feedback is shifted multiple times during a sustained vocalization (Behroozmand et al., 2009), at a random point during sustained vocalizations (Larson et al., 2001; Franken et al., 2018), or during a random trial (Elman, 1981; Burnett et al., 1997; Purcell and Munhall, 2006a; Tourville et al., 2008). Participants in f₀ and formant manipulated auditory feedback studies have been found to typically produce a reflexive compensatory response in the opposite direction of the manipulation (Burnett et al., 1997; Hain et al., 2000). These responses are usually only partial compensations to the shift (Purcell and Munhall, 2006a; Chen et al., 2007). Although manipulations in f₀ studies typically range from ±25 to 600 cents (100 cents = 1 semitone), participants on average produce response magnitudes of less than 60 cents (Burnett et al., 1997; Larson et al., 2000; Natke and Kalveram, 2001; Burnett and Larson, 2002; Donath et al., 2002; Bauer and Larson, 2003; Natke et al., 2003; Sivasankar et al., 2005; Liu and Larson, 2007). In formant manipulation studies, participants produce compensatory responses that are on average less than 30% of the total shift (Purcell and Munhall, 2006a; Tourville et al., 2008; Mitsuya et al., 2015). A second response type, where vocal productions follow the same direction as the shift, called following responses, has also been observed (Burnett et al., 1997, 1998; Hain et al., 2000; Larson et al., 2007). It has been suggested that these following responses occur more frequently with large magnitude shifts (i.e., in f₀ perturbations; Burnett et al., 1998). Behroozmand et al. (2012) and Franken et al. (2018) however found that most individuals who produced opposing (compensatory) responses on average, tended to also produce following responses on some trials.

Both following and compensatory responses typically show an onset latency of approximately 100–150 ms, suggesting these responses are reflexive and automatic (Tourville et al., 2008). This has been supported by findings showing that participants produce compensatory productions even when instructed to ignore any manipulations (Burnett et al., 1997; Zarate and Zatorre, 2008; Patel et al., 2014; Hu et al., 2015). Hain et al. (2000), however, found that there appear to be two responses produced in f₀ manipulations: an early automatic response that can be modulated by instruction and a later response under voluntary control. Overall, these compensatory responses are thought to provide information about an individual’s auditory feedback control (Tourville et al., 2008; Cai et al., 2011). Larger response magnitudes opposing the direction of the shift are postulated to reflect greater reliance on auditory feedback, although the magnitude and direction of the applied perturbation in studies need to be taken into consideration (Heller Murray and Stepp, 2020).

In contrast, predictable, sustained auditory shifts are used to evaluate the updating of the feedforward system (feedforward control) through sensorimotor adaptation. In these paradigms, participants are presented with shifted auditory feedback stimuli over multiple successive trials, and gradually develop/learn an adaptive response to compensate for the perturbation (Houde and Jordan, 1998, 2002; Jones and Munhall, 2000, 2002; Purcell and Munhall, 2006b; Villacorta et al., 2007). In adults, these compensatory effects remain immediately following removal of the altered auditory feedback; such adaptation indicates a learned response in which stored motor programs have been updated (adapted) in response to the persistent compensatory productions made (Houde and Jordan, 1998, 2002; Jones and Munhall, 2000; Purcell and Munhall, 2006b). These studies typically consist of four phases: a baseline phase, where participants receive normal feedback; followed by a ramp phase, where the auditory feedback is incrementally shifted; then a hold phase, where the shifted stimuli is held at its maximum; and finally, sometimes, an end phase where the perturbations are removed. In these studies, two responses are examined: (a) how individuals’ responses during the hold phase differ from their average baseline phase productions (looking for compensation to shifts), and (b) how individuals’ productions at the end phase (when the perturbation is removed) differ from the baseline phase (looking for adaptation).

These adaptation paradigms provide key information about how speakers use auditory feedback, for calibration and maintenance (i.e., during hold phase) and to incorporate long-term changes in their speech production (i.e., during end phase; Houde and Jordan, 1998; Purcell and Munhall, 2006b; MacDonald et al., 2010). Similar to the reflexive responses to sudden perturbations, participants’ responses are typically in the opposite direction of the manipulation, with some responses following the perturbations (Houde and Jordan, 1998, 2002; Jones and Munhall, 2000, 2005; Purcell and Munhall, 2006b; Villacorta et al., 2007). These responses also only partially compensate for the total perturbation magnitude (Jones and Munhall, 2000, 2005; Houde and Jordan, 2002; Purcell and Munhall, 2006b; Villacorta et al., 2007; MacDonald et al., 2010). Katseff et al. (2012) suggested that based on findings that individuals showed greater compensation for small F1 perturbations than for larger perturbations, auditory feedback may play a larger role in small discrepancies. These responses have also been found to be automatic, occurring when participants are instructed to ignore manipulations (Munhall et al., 2009; Keough et al., 2013).

Across studies exploring responses to sudden shifts, there is consensus that these responses describe compensation. However, within the sensorimotor adaptation literature, inconsistencies persist. In some studies, responses produced when the perturbation is held at its maximum (hold phase trials) are described using the term compensation, while in others, these trials are referred to as adaptation. Although responses in these trials are thought to gradually reflect updating of motor commands and hence adaptation to the shift, within this article, the term compensation will be used to describe responses within the hold phase of a sensorimotor adaptation paradigm, as these productions represent both compensation and adaptation. Productions made following removal of shifts (during end phase trials) will be described as adaptation (also known in the literature as after-effects).

Examining responses to unexpected and sustained perturbations provides important information about feedback and feedforward control. Contrasting participants’ responses to sudden and sustained F1 manipulations, Franken et al. (2019) and Raharjo et al. (2021) found that individuals’ responses in the sudden vs. sustained conditions were not correlated with each other. In contrast, Lester-Smith et al. (2020) explored reflexive compensatory and adaptive responses to F1 and f₀ perturbation. Participants showed similarities in their reflexive and adaptive responses, where individuals with larger reflexive responses to sudden F1 perturbation also showed larger adaptive responses to predictable F1 manipulations. However, reflexive and adaptive responses to f₀ manipulated auditory feedback were not related. This highlights that differences may not only be evident in the mechanisms underlying responses in sudden (reflexive) and sustained (adaptive) perturbation paradigms, but also in control of articulatory and vocal settings. Although responses to fundamental and formant frequency altered auditory feedback have extensively been studied in adults, a contrastive look at responses in children has not previously been examined. Investigating how and if these responses differ developmentally will help shed light on underlying mechanisms and improve our understanding of speech motor control.

All ten of the studies exploring responses to f₀ altered auditory feedback involved unexpected (within trial) perturbations. Two studies also contrasted responses to unexpected perturbations with sustained (predictable) manipulations (Scheerer et al., 2016; Heller Murray and Stepp, 2020). In terms of manipulations applied, all ten of the studies included a negative manipulation of one semitone (−100 cents). Five of the studies also included a positive manipulation of one semitone (+100 cents), and two studies included additional magnitude manipulations either in negative (−50, −200 cents; Liu et al., 2013) or negative and positive directions (±50, 200, and 500 cents; Liu P. et al., 2010). Table 1 includes a summary of findings, sample size, age range and manipulations of f₀ manipulation studies.

All fourteen studies exploring responses to formant manipulated auditory feedback explored responses to sustained perturbations over several trials. Table 2 includes a summary of findings, sample size, age range, and formant manipulations, with more in-depth findings available in Supplementary Material 3 (see Footnote 1). Generally, studies included a baseline phase (ranging from 10–50 trials, M = 23.0) followed by a ramp phase where the perturbation was gradually introduced (ranging from 10–60 trials, M = 25.4), and a hold phase where the perturbation was held at its maximum (ranging from 18–120 trials, M = 41.0). Ten studies also included an end phase where the perturbation was removed (ranging from 10–40 trials, M = 20.15). One study additionally included a ramp-down phase where the perturbation was gradually removed following the hold phase (van den Bunt et al., 2018b).

In terms of magnitude and direction of manipulations, eight studies manipulated F1 and F2 values of vowels, and five studies manipulated F1 only. F1 was manipulated in various ways, including: increased by 25% (Shiller and Rochon, 2014; van den Bunt et al., 2018a), 175 Hz (Shiller et al., 2010b), or 340 Hz, or decreased by 230 Hz (Coughler et al., 2021), or manipulated individually so the maximum perturbation represented a change from/ε/to/ae/ (Ohashi and Ostry, 2021). One study manipulated the frequency centroid of fricatives (decreased by 3 semitones; Shiller et al., 2010a). Manipulations of F1 and F2 were language dependent. In Dutch, this manipulation included an F1 increase of 25% and an F2 increase or decrease by 12.5% (Terband et al., 2014; van Brenk and Terband, 2020 respectively), in French F1 was increased by 27% and F2 decreased by 10% (Caudrelier et al., 2019), and in English F1 was increased by 200 Hz or 25% and F2 was decreased by 250 Hz or 12.5% (MacDonald et al., 2012; Daliri et al., 2018; Demopoulos et al., 2018). One study (van den Bunt et al., 2018b) individualized the manipulation so the maximum perturbation meant a change from/ı/to/ε/for each participant. Kim et al. (2020) manipulated F1 and F2 upward 335 cents (adult population received manipulation of ±250 cents).

Results across both fundamental and formant frequency manipulations found age-dependent developmental trajectories related to response latencies, magnitude of compensatory responses, variability and perceptual abilities, as well as relationships of compensation with literacy abilities. These findings as well as clinical and neurophysiological/neuroimaging findings are discussed below.

Four f₀ studies compared response latencies, the onset of the compensatory response to altered auditory feedback, across age groups and found that children consistently demonstrated longer response latencies to perturbations in auditory feedback than adult populations (Liu H. et al., 2010; Liu P. et al., 2010; Liu et al., 2013; Scheerer et al., 2013b). Two of these studies used multiple child age groups to explore potential age gradients within responses (Liu et al., 2013; Scheerer et al., 2013b). Scheerer et al. (2013b) found a main effect of age, where three of the four age groups under 18 (4–6, 7–10, and 11–13 years) independently demonstrated longer response latencies than the 18–30-year-olds (14–17-year-olds did not). Similarly, Liu et al. (2013) found their younger children (ages 10–12) had significantly longer response latencies than the adult group, however, older children (ages 13–15) did not differ from adults.

The finding of significantly longer response latencies for children compared to adults in the compensatory responses in pitch-shifted paradigms is consistent with developmental trends. Neurophysiological response latencies (i.e., event-related potential latencies) are considered an objective measure of the speed of neural integration and activity, reflecting the efficiency of information processing and the synaptic density in the auditory cortex, where shorter latencies reflect faster integration of auditory information (Eggermont, 1988; Kotecha et al., 2009). Vocal response latencies were similarly found to relate to maturational changes. Integration of rapid information in adult-like auditory processing may therefore be due to increased velocity and efficiency of neural conduction and intercortical communication in gray and white matter of the cortex.

All f₀ manipulation studies explored compensatory responses to pitch perturbations, and generally found children compensated in the opposite direction of the shift. Following responses were examined in three studies (Russo et al., 2008; Liu P. et al., 2010; Scheerer et al., 2020b), with two studies excluding participants who followed the perturbation (Scheerer et al., 2016; Demopoulos et al., 2018). Results exploring the magnitude of compensation responses to unexpected perturbations were mixed. When contrasting across age groups, children were found to have: (a) reduced magnitude responses compared to adults (Liu H. et al., 2010; Scheerer et al., 2016), (b) increased magnitude responses compared to adults (Liu P. et al., 2010; Heller Murray and Stepp, 2020), (c) increased responses that followed the manipulation compared to adults (i.e., following responses; Liu P. et al., 2010), and (d) no effect of age across childhood (Liu et al., 2013; Scheerer et al., 2013b,2020a) or compared to adults (Scheerer et al., 2013b; Heller Murray and Stepp, 2020). Heller Murray and Stepp (2020) found when analyzing opposing responses that only children with less sensitive pitch discrimination showed significantly larger responses, compared to adults and children with adult-like pitch discrimination. Liu et al. (2013) found an effect of sex, where male speakers produced larger response magnitudes than female speakers. Findings from Russo et al. (2008), Demopoulos et al. (2018), and Scheerer et al. (2020b) are described below in the Clinical Findings section.

The two studies exploring sustained perturbation found children showed smaller compensatory responses compared to adults (Scheerer et al., 2016; Heller Murray and Stepp, 2020). Specifically, Heller Murray and Stepp (2020) found children with less sensitive pitch discrimination produced smaller compensatory responses compared to adults and children with adult-like pitch discrimination. The magnitude of responses produced to a sustained shift was negatively correlated with the magnitude of responses to a sudden shift (Heller Murray and Stepp, 2020). In contrast, Scheerer et al. (2016) found that children produced smaller compensatory responses compared to adults in both sudden and sustained pitch shift conditions, however, these responses were not examined for correlation. Scheerer et al. (2016) also explored adaptation in the end phase, finding magnitudes of responses following removal of pitch perturbation did not differ between children and adults. In general, these findings provide evidence supporting the DIVA model, where responses to sudden and sustained shifts are not considered separate processes (Guenther, 2006; Guenther and Vladusich, 2012).

All formant frequency manipulation studies explored compensatory responses, and overall found typically developing children generally showed compensation in the opposite direction of the perturbation in hold and end phases. Two studies examined following responses (Terband et al., 2014; van Brenk and Terband, 2020). Seven studies contrasted child and adult responses to formant manipulated altered auditory feedback (Shiller et al., 2010a; MacDonald et al., 2012; Daliri et al., 2018; Caudrelier et al., 2019; Kim et al., 2020; van Brenk and Terband, 2020; Ohashi and Ostry, 2021). Across the studies, children showed: (a) stronger compensation and adaptation responses in F1 than adults (van Brenk and Terband, 2020), (b) similar magnitude compensation to adults (Shiller et al., 2010a; MacDonald et al., 2012; Daliri et al., 2018; Kim et al., 2020; Ohashi and Ostry, 2021), and (c) no age effect in compensation or adaptation across childhood (Caudrelier et al., 2019). MacDonald et al. (2012) found young children showed similar compensation to adults, however, children under 4 years of age showed no compensatory response. Shiller and Rochon (2014) found after a period of perceptual training, children showed increased magnitude of compensation.

Based on the underlying mechanisms being examined, response magnitudes for unexpected and sustained shifts were analyzed separately. Unexpected shifts, used to explore an individual’s reliance on auditory feedback control, examined in f₀ manipulation studies, elicited mixed results, ranging from reduced magnitude to increased magnitude compensatory responses compared to adults to no age effect. One potential reason for these mixed findings could be due to proximity of shifted trials. As discussed in Cai et al. (2012), cross-trial adaptation effects have been found where participants’ early productions within a trial contain compensation responses to the perturbation of the previous trial.

By contrast, sustained shifts, used to explore the updating of feedforward control, generally showed no age effect after 4 years of age in formant manipulation studies. The lack of age effect suggests that children are using feedforward control similar to adults. In contrast, in f0 sustained shift studies, children exhibited smaller magnitude compensatory responses compared to adults. This smaller compensation response may indicate a greater reliance on sensory feedback, with reduced weighting on feedforward control.

Adaptive responses, when perturbations were removed, showed mixed results ranging from stronger adaptive responses compared to adults in formant manipulation studies, to no age effect across childhood in formant or f₀ manipulation paradigms. Although mixed, these findings of the presence of adaptation responses show that children used the altered auditory feedback to update their sensorimotor mappings for future vocalizations.

Contrasting pitch and formant manipulation paradigms, clear developmental differences are seen in the youngest ages where children appear to be using their auditory feedback to manipulate their vocal productions. Scheerer et al. (2020a) found children as young as 2 years of age compensated to f₀ shifted stimuli, whereas MacDonald et al. (2012) did not find compensatory responses in children under 4 years of age to formant shifted stimuli. This lack of compensation suggests that the ability to adaptively regulate measures of vocal control (i.e., f₀) arises before control over measures of articulatory settings (i.e., formants). However, further research is required to confirm this assumption, as the number of studies examining responses in children under 3 years of age is restricted to these two studies.

Five studies explored differences in vocal variability in f₀, contrasting baseline standard deviation in vocal productions across age groups (Russo et al., 2008; Scheerer et al., 2013b,2016, 2020a; Heller Murray and Stepp, 2020). Four of these studies contrasted children with adults, finding children consistently showed more variability than adults (Scheerer et al., 2013b,2016, 2020a). Heller Murray and Stepp (2020) found children with less sensitive pitch discrimination had significantly higher variability at baseline than both the children with adult-like pitch discrimination and adults. Children with less sensitive pitch discrimination also showed larger response magnitudes to unexpected pitch shifts and smaller responses to sustained pitch shifts compared with adults and children with adult-like pitch discrimination. Baseline vocal variability was also found to positively correlate with the magnitude of responses to unexpected perturbations, and negatively correlate with the magnitude of responses to sustained perturbations (Heller Murray and Stepp, 2020). Through regression analyses, Scheerer et al. (2013b) found that vocal variability accounted for a significant amount of the variance in the magnitude of the compensatory responses. Scheerer et al. (2016), however, did not find vocal variability correlated with the magnitude of compensatory responses. In further exploration of electrophysiological correlates, Scheerer et al. (2013b) found that age and vocal variability were significant predictors of N1 amplitude.

Six studies explored variability of baseline vocal productions related to articulatory control (Shiller et al., 2010a; MacDonald et al., 2012; van den Bunt et al., 2018a; van Brenk and Terband, 2020; Coughler et al., 2021; Ohashi and Ostry, 2021). Generally, children showed greater variability compared to adults (Shiller et al., 2010a; MacDonald et al., 2012; van Brenk and Terband, 2020; Ohashi and Ostry, 2021). MacDonald et al. (2012) found that variability decreased with age. Similarly, van den Bunt et al. (2018a) found literate children who read more non-words per minute showed less variation in vowel production. Coughler et al. (2021) found increased variability negatively correlated with the amount of compensation in TD children, whereas Ohashi and Ostry (2021) did not find variability correlated with the amount of compensation in children or adults.

Five studies additionally examined perceptual abilities related to articulatory control (i.e., discrimination; Shiller et al., 2010a,b; Shiller and Rochon, 2014; Terband et al., 2014; Coughler et al., 2021). Coughler et al. (2021) found F1 discrimination thresholds did not significantly correlate with percent compensation in the positive or negative condition, or with language scores in TD children or children with a specific deficit in language. Shiller and Rochon (2014) found that productions following an experimental block of relevant (to the formant frequency shift) auditory-perceptual training resulted in enhanced compensatory responses in children. Based on results from a phoneme identification test, Shiller et al. (2010a) found children had more imprecise perceptual boundaries than adults. Additionally, while adults demonstrated a significant shift in their perceptual boundaries for the perturbed sound contrast after testing, children did not reliably change these internal perceptual boundaries.

Studies involving f₀ and formant manipulated feedback consistently showed increased variability in child baseline productions compared to adults. Individuals with more variable vocal productions are thought to have less defined internal sensorimotor representations. These sensorimotor representations encode the relationship of stored motor commands (utilized for feedforward control) and their auditory and somatosensory consequences (utilized for feedback control). Early in development, it is hypothesized, that children must rely more on auditory feedback during vocalization to ensure that their speech is in line with their desired vocal output, resulting in unstable vocal productions (Scheerer and Jones, 2012). As exposure to speech increases, the reliability of internal sensorimotor representations increases and over-dependence on auditory feedback becomes unnecessary, with vocal output becoming more consistent, shifting their reliance to feedforward control (Scheerer and Jones, 2012). Feedback control, however, continues to be an integral part of speech motor control, as auditory and somatosensory feedback are used to inform and maintain feedforward control, updating and refining sensorimotor representations when mismatches occur between expected and actual output (Franken et al., 2019).

Perceptually, findings across f₀ and formant studies generally demonstrated reduced precision in vocal and articulatory control in children compared to adults. While baseline variability represents potentially different control mechanisms (i.e., articulatory or vocal), reduced perceptual discrimination, and increased vocal variability in younger children across paradigms aligns with the DIVA model where reliance is postulated to shift from feedback to feedforward control as sensorimotor targets are refined over multiple productions, with initial targets being larger and discrimination abilities less sensitive (Guenther, 2006; Tourville et al., 2008; Guenther and Vladusich, 2012). This was further supported by findings by MacDonald et al. (2012) where baseline F1 and F2 variability was found to decrease with age. This suggests that maturational changes occurring in the speech motor control system affect the extent to which auditory feedback is used to modify internal sensorimotor representations.

Three studies collected articulation information, with two using the Goldman-Fristoe Test of Articulation 2 (GFTA-2; Shiller et al., 2010b; Daliri et al., 2018). Five studies collected receptive and expressive language information with the most common test used being the Clinical Evaluation of Language Fundamentals (CELF; Russo et al., 2008; Shiller et al., 2010b; Coughler et al., 2021). Five studies collected information about cognitive abilities with the most common test used being the Wechsler Abbreviated Scale of Intelligence (WASI; Russo et al., 2008; Daliri et al., 2018; Scheerer et al., 2020b; Coughler et al., 2021). In these studies results included: (a) no significant correlation found between speech and language tests and compensation (Daliri et al., 2018; Scheerer et al., 2020b; Coughler et al., 2021); (b) a significant correlation of response magnitudes with core, receptive and expressive language scores, where decreased magnitude responses were related to higher language scores (Russo et al., 2008); and (c) a significant positive correlation of compensation with performance on non-word repetition (Terband et al., 2014). Scheerer et al. (2020b) also found average response latency significantly predicted Multidimensional Social Competence Scale scores (MSCS).

Phonological awareness measures were collected in three studies (van den Bunt et al., 2018a,b; Caudrelier et al., 2019). Reading measures were additionally collected in two of these studies (van den Bunt et al., 2018a,b). Better rapid automatized naming was found to correlate with better compensation (van den Bunt et al., 2018a), as well as correlated with weaker deviation from the baseline during the ramp-up phase and stronger de-adaptation during the ramp-down phase (van den Bunt et al., 2018b).

Average phonological awareness scores were significantly higher in children who compensated than to non-adapting children (Caudrelier et al., 2019), and was associated with stronger compensation during ramp-up and hold phase, and weaker de-adaptation in the ramp-down and end phases (van den Bunt et al., 2018b). Phonological awareness, rapid naming, and letter knowledge correlated significantly with compensation in preliterate children, whereas reading correlated with compensation in literate children (van den Bunt et al., 2018a). Overall, literacy was also found to play a role in compensatory response magnitude, where significantly stronger compensation in hold and end phases were found for literate children relative to preliterate children (van den Bunt et al., 2018a).

In general, findings examining the relationship of speech, language, and cognitive measures in relation to compensation magnitude were limited. A few studies found no significant relationship of speech and language abilities with compensation, while one study found a significant negative relationship with language abilities. One possibility for these differences could be a result of differences in the assessment tools used to assess speech, and language. As well, although several studies collected information on speech, language and cognitive abilities, the relationship of these abilities with compensation magnitude were not examined.

A clear relationship however was evident for literacy in relation to the developmental trajectory of auditory feedback control. Reading and preliteracy skills (e.g., phonological awareness) significantly correlated with compensation ability. In particular, phonological awareness scores, a strong predictor of later reading ability, were significantly higher in children who compensated compared to those who showed no compensatory response (Caudrelier et al., 2019), and in those who showed greater compensation (van den Bunt et al., 2018b). These results suggest an interplay among auditory-integration, speech motor control, and reading, supporting theories that impaired phonological representations (essentially sensorimotor representations) may underlie reading deficiencies (Ramus and Szenkovits, 2008), although further exploration is needed to understand other potential factors that may influence this relationship.

Three studies exploring clinical population responses to f₀ manipulated altered auditory feedback included children with autism spectrum disorder (ASD; Russo et al., 2008; Scheerer et al., 2020b) and children who are 16p11.2 deletion carriers (Demopoulos et al., 2018). Deletion at 16p11.2 is commonly observed in individuals with diagnoses of developmental coordination disorder, phonological processing disorder, language disorders, and ASD (Demopoulos et al., 2018). Russo et al. (2008) had mixed findings, where some children with ASD demonstrated smaller mean magnitude compensatory responses, while others created atypically large compensatory responses compared to TD children. Scheerer et al. (2020b) found children with ASD had shorter response latencies than TD children, but showed similar compensatory responses to TD peers. In contrast, Demopoulos et al. (2018) consistently found children with 16p11.2 deletion showed larger pitch compensation compared to controls.

Seven studies explored compensatory responses to formant frequency manipulations of children with speech and language difficulties. Two of these studies examined children with SSD (Shiller et al., 2010b; Terband et al., 2014), two examined responses of children who stutter (Daliri et al., 2018; Kim et al., 2020), one of children with dyslexia (van den Bunt et al., 2018b), one with children who are 16p11.2 deletion carriers (Demopoulos et al., 2018), and one with children with developmental language disorders (DLD; Coughler et al., 2021). No consistent findings were seen across these clinical populations. Children who stutter were found in one study (Daliri et al., 2018) to show greater compensation in F1 than adults who stutter, however they did not differ from children who do not stutter. However, Kim et al. (2020) found children who stutter showed no compensatory response. Conflicting results were similarly found for children with SSD. Terband et al. (2014) found children with SSD followed the perturbation in F1 during hold and end phases, whereas Shiller et al. (2010b) found compensation as well as an adaptation response. It is important to note that Shiller et al. (2010b) only included one participant. Similar to Shiller et al. (2010b), van den Bunt et al. (2018b) found all children with dyslexia compensated in the opposite direction of the perturbation, with the only difference from typically reading peers being a weaker return to baseline during the ramp-down phase. In contrast, children with 16p11.2 deletion showed significantly weaker compensation than their TD peers (Demopoulos et al., 2018). Coughler et al. (2021) found a unique pattern where children with a specific deficit in language (DLD) demonstrated differential compensation in positive and negative shift conditions. Children with DLD showed larger compensation in the positive shift condition and compensated less in the negative shift condition compared to typically developing peers.

Clinically, across formant and f₀ manipulation studies, a broad range of disorder areas were examined, from ASD, 16p11.2 deletion, SSD, dyslexia, fluency, to DLD. All of these disorders have been linked to impairments in or closely linked to auditory processing. Although there was a lack of methodological consistency, several studies found aberrant responses in some of the clinical populations compared with typically developing children. This included increased following responses and larger or smaller compensation responses compared to typically developing peers.

Results involving children who stutter further support developmental sensorimotor control changes into adulthood. In Daliri et al. (2018), children who stutter produced greater compensation than adults who stuttered, however, they did not differ compared to children who do not stutter. This suggests some shift in reliance in adults, which is further supported by the finding that adults who stutter did not show any adaptation, whereas children who stutter did successfully adapt in their F1 productions. Adults who stutter were no longer updating their stored motor programs through feedforward control unlike children who stutter. Kim et al. (2020) found very different findings, children who stutter showed no significant compensatory response, although similarly, adults showed a reduced compensation compared to TD adults. Of note here, compensatory responses significantly correlated with age, where greater compensation occurred with increased age. Potential differences between these two studies may be due to differences in shifts applied, where Daliri et al. (2018) shifted the phonemic category of the vowels, and Kim et al. (2020) did not.

Neuroanatomically, the atypical mixed compensation responses seen in children with ASD in f₀ manipulations studies (some in the typical range and others overcompensating) may relate to findings that children with ASD have weaker white matter connections between left ventral premotor cortex, a key area in speech motor planning, and other cortical regions involved in speech production (Russo et al., 2008; Peeva et al., 2013). Although not collected in their studies (Russo et al., 2008; Scheerer et al., 2020b), weaker white matter connections could be associated with the overcompensation profile found in some children. Scheerer et al. (2020b) additionally found shorter response latencies in children with autism, indicating more research is needed to further understand what may be driving differences.

Similarly, the mixed findings found for children with SSD, with some showing typical responses (Shiller et al., 2010b) and others showing increased following responses (Terband et al., 2014), could be related to gray and white matter volume differences. Previous studies have found abnormal gray and white matter volume in areas relating to speech motor control, which is thought to be related to delays in synaptic pruning (Preston et al., 2014).

Of interest, the type of manipulation was shown to affect the direction of aberrant responses. Demopoulos et al. (2018) found children who were carriers of 16p11.2 deletion showed overcompensation compared to controls in response to pitch perturbation (unexpected shift), but undercompensation compared to controls in response to formant manipulated feedback (sustained shift). This further supports theories that these vocal and articulatory controls develop at differing rates and time points. However, it is important to take into consideration that different experimental designs (i.e., sudden vs. sustained shift) may be the driving factor resulting in these differing compensation responses. Further research is needed contrasting responses to sudden and sustained shifts using consistent manipulations.

Two studies explored EEG responses to f₀ altered auditory feedback (Liu et al., 2013; Scheerer et al., 2013b). Both found that P1 and N1 latency, and P1 amplitude decreased with age (Liu et al., 2013; Scheerer et al., 2013b). Scheerer et al. (2013b) also found that N1 amplitude increased with age, however, Liu et al. (2013) did not find a significant effect of age on N1 amplitude. P2 amplitude showed greater variability, but generally was shown to increase with age (Liu et al., 2013; Scheerer et al., 2013b). Notably, Liu et al. (2013) also found significant interactions between sex and age in the N1 and P2 potentials. Male participants generally produced larger N1 amplitudes, and within the group of older children, females demonstrated significantly shorter N1 latencies than males. Among the young children, males had significantly larger P2 amplitudes than females, and young females had significantly larger P2 amplitudes than older females (Liu et al., 2013). In terms of P2 response latencies, P2 latency was found to be age-dependent for males only, however, within the group of older children, females were found to have significantly shorter P2 response latencies compared to males (Liu et al., 2013). The variable age-sex interactions found by Liu et al. (2013) speaks to the complexity of factors influencing neural responses to auditory feedback.

Two studies (van den Bunt et al., 2018b; Ohashi and Ostry, 2021) examined neural activation in relation to formant manipulation sensorimotor control studies. van den Bunt et al. (2018b) used fractional anisotropy to measure connectivity of the arcuate fasciculus/superior longitudinal fasciculus (AF/SLF). Fractional anisotropy of the AF did not directly relate to altered auditory feedback responses, but did correlate strongly with phonological awareness scores. When phonological awareness was controlled, higher fractional anisotropy was found to be associated with less adaptation during altered auditory feedback (van den Bunt et al., 2018b). Ohashi and Ostry (2021) found children and adults had distinct patterns of functional connectivity. In adults, compensation to altered auditory feedback was positively correlated with activation in the right inferior frontal gyrus (area 44) and associative sensory regions. In children, compensation was positively correlated with functional connectivity of the primary somatosensory cortex (S1)/primary motor cortex (M1) and posterior rostral cingulate zone (RCZ) and left anterior insular cortex. When contrasting younger and older children (over 9 years), older children showed an increasingly adult-like pattern of connectivity.

Neurophysiological and neuroimaging studies involved examining evoked potentials, diffusion-weighted imaging and resting-state functional connectivity. Evoked potentials have a well-established history of being an objective measure of maturation of the nervous system, which can increase our understanding of neurophysiological changes that underlie behavioral responses to sensory input (Eggermont, 1988). Due to small sample sizes and differing languages (English in Scheerer et al., 2013b, and Mandarin in Liu et al., 2013), age-dependent conclusions are guarded, although both utilized similar f₀ shifted paradigms. Developmental trends observed in P1-N1-P2 amplitudes and latencies mirrored general developmental trends found during passive listening tasks (Ponton et al., 2000; Wunderlich and Cone-Wesson, 2006; Fitzroy et al., 2015). These similarities support the existence of a developmental gradient in auditory integration.

The significant decreases in latency observed with age across both the P1 and the N1 components of the P1-N1-P2 complex (Liu et al., 2013; Scheerer et al., 2013b), alongside the decreases in behavioral response latencies, together provide significant support for the existence of age-related changes in the efficiency of auditory integration in the cortex. This in turn suggests that the efficiency of information processing in cortical areas supporting sensory function influences the developmental trajectory of speech motor control. Findings of consistent decreases in P1 amplitude (Liu et al., 2013; Scheerer et al., 2013b) and increases in N1 amplitude (Scheerer et al., 2013b) across age during the processing of altered auditory feedback provides more evidence for the age-dependent shift from reliance on feedback to reliance on feedforward control identified through the analysis of response magnitudes.

Neuroanatomically, although findings examining fractional anisotropy of the left arcuate fasciculus were not found to directly relate to compensation responses (van den Bunt et al., 2018b), resting-state functional connectivity showed distinct patterns of connectivity which significantly related to compensation in speech sensorimotor adaptation tasks (Ohashi and Ostry, 2021). This finding supports the hypothesis that protracted neural plasticity during development relates to differences in performance in speech motor learning, demonstrating that speech motor adaptation abilities relate to cortical remodeling and reorganization occurring across development.