A probe study on vocal development in two infants at risk for cerebral palsy

Abstract

The present work examined canonical babbling ratios longitudinally as a measure of onset and consolidation of canonical babbling in two infants at risk of cerebral palsy (CP) between 5–16 months. Ten typically developing infants were included for comparison at 6, 9, 12, and 16–19 months. Canonical babbling ratios (CBR) were calculated from five-minute segments and follow-up diagnostic outcomes were collected between 24 and 33 months. The two infants at risk demonstrated low CBR growth trajectories compared to the typical infant group, and slightly different patterns of consolidation. The two infants at risk were later diagnosed with different levels of CP and speech impairment severity. All infants demonstrated greater variability than expected. Studying canonical babbling and other prelinguistic milestones in this population may inform our perspective of the involvement of the motor system in the vocal domain. Additional implications on the analysis of canonical babbling using all-day home recordings are discussed.

Cerebral palsy (CP) is a group of neurodevelopmental disorders that affect fine and gross motor development in the fetal or infant brain.¹ Approximately 50–90% of children diagnosed with CP also present with speech impairments.^2,3 Speech impairment and its severity in children with CP can be highly variable and difficult to predict at early ages.^4–7 Research examining infrastructural patterns of early syllables—particularly well-formed consonant-vowel transitions—emerging in children with CP may help identify potential early indicators of speech impairment in this population.

The canonical babbling stage of infant vocal development is defined as the time period during which regular production of mature consonant-vowel syllables with well-formed transitions between the consonant and vowel (e.g., [baba] and [dada]) begins to occur.^8,9 The canonical babbling ratio (CBR) has been used in research since the 1970s to objectively measure the development of advanced vocal forms in infancy.^10–13 CBRs are calculated as the total number of canonical syllables divided by the total number of syllables (canonical and noncanonical) produced in an observation segment.^14,15 Previous research has suggested that parents’ report of their infants’ first canonical syllables often correspond to a CBR based on laboratory recordings of around 0.15;^14–17 however, other research has debated whether CBRs are comparable across laboratory and home settings.^11,18 A review of the history and utility of CBRs as a measurement tool is provided by Lee et al.¹⁰ Importantly, it is now possible to determine CBRs based on random sampling of segments from all-day recordings. ¹⁹ We believe this to be a more representative method to determine vocal patterns in infants, but it is not yet clear how CBRs analysed in all-day recordings will compare with prior values obtained in more short-term laboratory or home recordings.

In typically developing children, the onset of the canonical babbling stage is known to occur between ~6–10 months of age and late onset is widely understood to be a robust indicator of atypical speech development.^14,20 Studies have demonstrated late onsets using CBRs in children with a variety of developmental disorders,^18,21–23 including conditions involving neuromotor impairments such as Down syndrome¹⁶ and apraxia of speech.^24,25 Several cross-sectional studies have evaluated the development of canonical babbling and other prelinguistic features of vocalization in infants and young children with CP. Levin²⁶ found that the canonical babbling onset had not yet occurred in six out of eight infants observed by 11–12 months of age. Otapowicz et al.²⁷ retrospectively studied the medical history of children with CP at 3–16 years of age and found delayed “cooing” in 23/46 children; 22 of these 23 were later diagnosed with dysarthria. Lohmander et al.²⁸ found that 38 infants with CP between 9–21 months produced lower canonical babbling rates than a control group of 30 infants at 10–12 months without a CP diagnosis. Nyman & Lohmander¹² further observed that 5/18 infants in a “neurodevelopmental delay group”—including infants at 10–24 months with CP, Down syndrome, and other genetic syndromes—showed delays in canonical babbling using a simplified measure of utterance-level CBRs. To our knowledge, there are no studies observing longitudinal development of canonical babbling in infants at risk of CP. This type of evaluation may reveal abnormalities in the trajectory of canonical babbling development possibly indicative of future speech disorders.

In addition to examining the onset of canonical babbling, another method of evaluating vocal development for potential abnormalities is to observe consolidation of canonical babbling. Consolidation can be characterized as the longitudinal trend for infants to produce a consistently high or increasing CBR, that is, a high or increasing proportion of syllables or utterances that can be classified as characterizing the canonical stage as infants develop greater control over the speech mechanism across time. The consolidation of the canonical babbling stage reflects robust biological maturation and provides the foundation for the development of more advanced capabilities such as production of first words.^29,30 On the other hand, lack of increase in canonical babbling across the stage may suggest a vulnerability in development as a result of damage to, or underdevelopment of, the infant brain as a result of developmental disorders.³¹ Large variations in rates of canonical syllables produced over time have indeed been observed in several conditions, including hearing impairment,²² Down syndrome,¹⁶ and William syndrome.²³ We propose that insufficient consolidation may occur in infants at risk of neurological disorders as a result of reduced motor control. Highly variable speech-motor control is characteristic of CP, even in older children with mild forms.^32,33 Thus, evaluating consolidation of the canonical babbling stage in this population may provide important insights about the integrity of the speech-motor system earlier in development.

The present study is a probe investigation of prelinguistic vocal development in two infants deemed at risk of CP based on perinatal complications. It is also one of the first to utilize all-day home recordings to observe mastery of prelinguistic vocal stages; thus, a secondary goal was to examine the utility of the CBR outside of the laboratory setting. We longitudinally evaluated canonical babbling emergence in the two at-risk infants compared with a group of typically developing (TD) infants whose data were available in the Origin of Language Laboratories (OLL) archives. Our hypotheses are based in the assumption that observing CBRs over time has the potential to identify atypical developmental patterns and thus early risk of speech or language disorders.

The onset of canonical babbling has been discussed as one of the earliest indicators of oral motor coordination and articulatory control.^31,34 Thus, we predicted that the neurological damage resulting in an at-risk CP status will result in atypical prelinguistic vocal patterns compared to typically developing infants. Specifically, we hypothesized that CBRs would reveal in the two infants at risk of CP:

1. Onset of canonical babbling beyond 12 months: Given the expected onset of canonical babbling between 7–11 months, we predicted that neither infant would have a ratio of at least 0.15 canonical syllables by 12 months of age.

2. A low degree of consolidation of canonical babbling across the four age points studied: We predicted that the two at-risk infants would show a protracted rate of canonical babbling growth compared to the TD sample.

Materials and Methods

IRB approval from the University of Memphis was obtained prior to data collection. Two female infants were recruited via word of mouth (CP1 and CP2). Both infants were at risk of CP based on their birth histories and early neurological assessments. Following consent, parents participated monthly across a 12-month period when the infants were between 5–16 months of age. Demographics and birth history information for both infants are presented in Table 1.

Table 1.

Demographics and birth history of participating infants

Infant	Gender	Birth status	Gestational age at birth	Birth weight	Conditions at birth	Hearing status
CP1	F	Full-term	41.5 weeks	2.72 kg	Respiratory distress, seizures, hypotonia, mild-moderate HIE (intermittent cord compression)	Normal
CP2	F	Full-term	38 weeks	3.22 kg	Moderate hypoxic ischemic encephalopathy (HIE, inconclusive underlying cause)	Normal

Archival data from 10 typically developing (TD) infants (5 male, 5 female) formed a comparison group. These infants had been recruited similarly and under the same kind of IRB approval from the University of Memphis in prior research in the OLL. Typically developing status was identified at the time of recruitment (during pregnancy or immediately after birth) as having no familial history of speech and language disorders and no birth complications.

Language Environmental Analysis (LENA) recording devices^19,35 were used to collect all-day recordings once a month for the 2 infants at risk from 5–16 months. All recordings were human coded as indicated below. The TD infants were recorded across the entire first year and through 19 months, although coding was available for appropriate comparison only at 6, 9, 12 and between 16–19 months using the same methods as the 2 infants at risk. The last recording for each of the TD infants occurred at either 16, 17, 18, or 19 months because the data collection protocol for these infants allowed for a single recording to be collected within this age range.

The LENA device is a digital recorder that weighs 2 ounces and collects up to 16 hours of audio with a 16-bit channel audio quality and 16 kHz sample rate.^36,37 On the recording day, parents placed the infant in an appropriately sized vest that securely housed the recording device. Parents were instructed to complete recordings on a single day of the month that they were at home with the infant, with normal home activity (i.e., days that were absent of environmental activity that may deviate from a typical day at home). After each day of recording, the LENA recording device was returned to the laboratory and audio files were uploaded to the LENA software system. Twenty-four 5-minute segments were randomly selected and extracted from each recording for coding analysis for the infants deemed at risk at all ages. The same was true of the ten TD infants for all ages except the last one between 16–19 months. For this last recording, 12 five-minute segments were selected at random.

Coding procedures

These segments were human coded in real-time for infant canonical and non-canonical syllables by individual listeners yielding a total count of canonical and noncanonical syllables per segment and recording. As it relates to infant vocalization, a syllable was defined as the production of a quasi-resonant or fully resonant nuclear element (vowel) that may contain a consonant element with an adult-like or slow formant transition.^38,39 Syllables are considered the minimal rhythmic unit of both adult speech and speech-like vocalizations in infancy. CP1’s segments were coded by a single primary listener (first author). CP2’s segments were split equally between two listeners for coding (first author and a trained graduate student). Both listeners had previously participated in the same training on the identification of syllables as canonical (i.e., well-formed, adult-like formant transition between the consonant and vowel) or noncanonical (vocalic element without a consonant or with a slow formant transition between a consonant-like element and the nucleus). Specifically, a syllable was coded canonical if it was identified as an adult-like production with a well-formed transition between the consonant and vowel. All other syllables (quasi-resonant vowels, fully resonant vowels, and immature or “marginal” CV syllables with slow formant transitions) were coded as noncanonical.

The coding of the recordings of the TD infants was conducted similarly by a team of twelve graduate students trained via the very same protocol at an earlier point in time. Agreement data were collected by assigning seven of the graduate students to recode semi-randomly selected segments that had been coded by other individuals in the group.

Canonical babbling ratios (CBRs)

A canonical babbling ratio (CBR) was calculated as the ratio of the total number of canonical syllables to the total number of all syllables (canonical and noncanonical) from each infant’s recording yielding an age-level CBR per child. A CBR of 0.15 is widely accepted as the level at which infants can be judged to have reached the onset of the canonical babbling stage, and this level is expected to be reached by typically developing children between 6–10 months of age.^8,11,18,40 Age-level CBRs for each infant were calculated and plotted across time points to observe consolidation, defined as a consistently high or increasing CBR over time.

Five-minute segments with fewer than ten total syllables were excluded from analysis to reduce the chance of unusual ratios based on few syllables (e.g., a recording with 2 syllables—1 canonical and 1 noncanonical—would yield a CBR of 0.50). An average of ~14 segments were included in each recording using this criterion (range: 6–22) for the at-risk infant recordings. The TD infant recordings at 6, 9, and 12 months included an average of ~12 segments (range: 4–20) across the ages studied; their recordings between 16–19 months included an average of ~8 segments (range: 5–11).

Follow-up diagnostic data

All infants completed the MacArthur-Bates Communicative Development Inventories- Words and Sentences⁴¹ at 24 months (both infants at risk) or 29 months (ten TD infants). CP1 additionally completed the Arizona Articulation Proficiency Scale-3⁴² and the Preschool Language Scale-5⁴³ at 33 months of age. CP2 was unable to complete formal or informal speech testing for follow-up. The ten TD infants also completed cognitive and motor profiles of the Bayley at 29 months to confirm TD status.

Coding reliability

We randomly selected 25% of segments from all recordings (5–16 months) of the infants at risk for recoding by a second listener for reliability analysis. Because CP1 had one primary coder, a single listener recoded the selected reliability segments. CP2 had two listeners involved in the primary coding; each of those listeners coded the other listener’s selected reliability segments. The secondary reliability coder for CP1 was a different graduate assistant than the second primary coder for CP2. For these reasons, each infant’s reliability data was analysed separately. Pearson correlations were calculated for three different values across segments: noncanonical syllable counts, canonical syllable counts, and canonical syllable ratios. We used Cohen (1988)’s interpretation of correlation effect sizes.⁴⁴

For CP1, there were large, positive correlations between coders for total noncanonical syllables (r = .908, p < .001), canonical syllables (r = .828, p < .001), and calculated CBRs (r = .758, p < .001). For CP2, there were large, positive correlations between coders for total noncanonical syllables (r = .873, p < .001), canonical syllables (r = .893, p < .001), and CBRs (r = .632, p < .001).

For the TD infants, seven coders were each assigned to recode ~30 of the 5-min segments from the study. A different individual had done the original coding in each case. There were 212 agreement segments in total, and each coder’s agreement assignment included segments from all the infants and all the ages in the study, randomly selected except for the requirements for balancing across ages and infants.^45,46 The agreement results showed large, positive correlations between coders for total noncanonical syllables (r = .876, p < .001), canonical syllables (r = .923, p < .001), and CBRs (r = .879, p < .001).

Results

Onset and consolidation at four time points

Figure 1 presents the CBR of the two infants at-risk at 6, 9, 12, and 16 months at the four age points available for TD group comparison (error bars: ±1 SD). The TD infant recordings at the 16-month age point occurred at either 16, 17, 18, or 19 months because the data collection protocol for these infants allowed for a single recording to be collected within this age range. The dotted line represents the traditional 0.15 criterion for mastery of the canonical babbling stage.

Figure 1.

Canonical babbling ratios of 2 infants at risk of CP and 10 TD infants at four time points

CP1 did not reach the traditional criterion for onset of canonical babbling at any age observed but approached the 0.15 criterion by 16 months. Relative to the TD group growth trajectory, CP1 showed a protracted, but consistent rate of consolidation of CBRs across the observed ages available for TD comparison. CP2 approached the criterion for mastery of canonical babbling at 9 months and did not reach the 0.15 criterion until 16 months. Compared to the TD group and CP1, her CBRs showed unstable growth over time; specifically, her CBRs decreased between 9 and 12 months, then increased again between 12 and 16 months—suggesting reduced consolidation of the canonical babbling stage across the observed ages.

The aggregated CBRs for the TD infants across the four age points suggest the anticipated consolidation, i.e., monotonic growth over time, expected in typical development. It is important to highlight that their CBRs were lower than expected based on the 0.15 criterion despite growing consistently as a group. Specifically, the TD infant data hint that they did not reach 0.15 at 9 months, and the error bars indicate that even at 12 months not all the TD infants met the 0.15 criterion.

Individual CBR trajectories within groups

Although Figure 1 might suggest monotonic growth of CBR over time across both groups, this was not observed for all infants at the individual-level. Figure 2 presents the monthly CBRs of the two infants at risk between 5–16 months alongside the CBRs of each of the ten TD infants. Table S1 in the supplementary material presents the individual CBR data points for all participants.

Figure 2.

Individual canonical babbling ratios of 2 infants at risk of CP and 10 typically developing infants

The consecutive CBRs across all 12 months of the two infants at risk show unexpected variability than what was observed in Figure 1, which showed only the four ages with available TD infant comparison data. Figure 2 instead shows that CP1 reached the 0.15 criterion for onset at 11 months, then showed a decrease in CBR at 12 months followed by an increase over time after that age. The CBRs of CP2 approached the 0.15 criterion between 8–9 months and reached criterion at 11 months after a decrease at 10 months. Her CBRs decreased again at 12 months with an overall increasing pattern but greater variability overall compared to CP1.

The TD infant CBRs also show unexpected variability, even across the four time points available. Specifically, three TD infants had a lower CBR at 12 months than their CBR at 9 months, and two TD infants had a lower CBR between 16–19 months than their CBR at 12 months. The range of CBRs within the TD group at 9, 12, and 16–19 months, but not at 6 months, highlight the unexpected variability in the consolidation of CBRs, even in typical infants.

However, both Figures 1 and 2 suggest that, at least by 12 months, the two infants at risk of CP have CBRs hovering at the low end of normal. Compared to the ten TD infants, CP1 had a CBR lower than only five of the TD infants at 6 months, but at 9 and 12 months, CP1 had a CBR lower than nine of the TD infants. At 6 months, CP2 had a CBR lower than three of the TD infants. But at 9- and 12-months, her CBR was lower than four and then nine of the TD infants, respectively. These data suggest that despite the unexpected variability in canonical babbling consolidation, the two infants at risk appear increasingly delayed beyond expected ages of mastery relative to the TD sample.

Table 2 presents the canonical syllable types reported by the parents of the two infants at risk of CP across the ages observed. Although consonant data were not collected for the ten TD infants, canonical syllables of typically developing infants are well-documented in previous literature to include alveolar (/t, d/), labial (/p, b/), and velar (/k, g/) stops and nasals (/m, n/) by 12 months,⁴⁷ although velar stops have been reported to appear later than other stops.⁴⁸ By 9 months, CP1 was consistently using four canonical syllable types (including nasals, labial, and alveolar stop consonants), and by 16 months she was also producing syllables with two velar stop consonants. By 12 months, CP2 was only producing two canonical syllable types (voiced labial and alveolar stops) and was not yet producing any other types by 16 months.

Table 2.

Canonical syllable types of the two infants at risk for cerebral palsy. Types are listed at the earliest age point reported by parents.

Age (months)	CP1	CP2
6	[ma]
9	[da, ba, na]	[da]
12	[ga]	[ba]
16	[ka]

Follow-up outcomes

Table 3 provides follow-up diagnostic information for the two infants at risk of CP. Both infants at risk were diagnosed with different levels of severity for CP and speech impairments in toddlerhood.

Table 3.

Follow-up diagnostic information of the two infants at risk of CP.

	Cerebral palsy (CP)		Speech impairment
	Age of diagnosis	Diagnosis	Age of diagnosis	Diagnosis
CP1	24 months	Hypotonic	33 months	Mild speech impairment
CP2	27 months	Severe spastic quadriplegia	27 months	Nonverbal

By 24 months, CP1 was diagnosed with mild hypotonic CP, and per parent report, she demonstrated mildly delayed gross motor milestones and walked with ankle foot orthosis braces. Also at 24 months, CP1 had an MB-CDI raw word score of 32, placing her below the 1^st percentile for her age. At 33 months, CP1 completed a formal speech and language assessment and was diagnosed with a mild articulation disorder and reduced overall speech intelligibility. She obtained a standard score of 75 (age equivalency < 20 months) on the Arizona-3 and presented with multiple phonological processes including fronting, stopping, final consonant deletion, and assimilation. CP1 also presented with oral motor hypotonia, tongue protrusion, and difficulty coordinating articulatory movements for speech production. Verbal dyspraxia was suspected but not formally diagnosed. CP1 performed within normal limits on the PLS-5 (standard scores: auditory comprehension- 104, expressive communication-103, total language-104), indicating age-appropriate receptive and expressive language abilities.

At 24 months, CP2 obtained an MB-CDI raw word score of 0, indicating no functional expressive words. Parent report at this time also indicated that CP2 was able to recognize her name but did not combine different syllables or use any words. At 27 months, CP2 was diagnosed with severe spastic CP and quadriplegia. By 36 months, CP2 remained nonverbal and was still vocalizing using canonical syllable singletons only (two canonical syllables reported: [ba] and [da]). Per parent and SLP report at this age, CP2 demonstrated significant cognitive impairments, and as a result no formal speech or language assessment data is available. All areas of motor function in CP2 were limited and she had no means of independent movement.

For the ten comparison infants, typically developing status was confirmed at 29 months of age using the MB-CDI and Bayley assessments. The group had a mean MB-CDI raw word score of 288 (SD = 112, ~21^st percentile for males and ~13^th percentile for females). On the Bayley, the TD group had a mean cognitive composite score of 113 (range: 90–145) and a mean motor composite score of 114 (range: 79–142).

Discussion

The present paper offers an evaluation of the canonical babbling ratio (CBR), a commonly used tool in vocal development research, to examine canonical babbling onset and its consolidation in two infants at risk of CP relative to ten typically developing (TD) infants. We hypothesized the at-risk infants would demonstrate a delayed canonical babbling onset and reduced consolidation (i.e., limited increase in CBR over time) compared to the TD group at the ages available for comparison. More specifically, we predicted that the CBRs of the two infants at risk of CP would not reach the traditional 0.15 criterion by 12 months, and that their CBRs would be reduced relative to the trajectories of the TD infants. Our findings revealed low CBRs compared to the TD infant sample but different CBR growth patterns in the two infants at risk. Unexpected variability was also observed in both groups, although group-level TD averages showed consistently increasing consolidation over the period sampled.

Our key hypothesis was that delayed onset and reduced consolidation of canonical babbling would be associated with delays in motor speech development characteristic of CP. The low CBRs of the two at-risk infants suggest weak support for the hypothesis. The variability across the monthly data for the at-risk infants, and even across the age points observed in the TD infants, may represent day-to-day variation similar to that occurring in hearing infants. The Dynamic Systems Theory^34,49,50 presumes that variability is necessary to support the mastery of new skills in any healthy system and to allow for adaptation to changes in the environment; perhaps the variability we observed across recordings and across infants is a manifestation of necessary day-to-day variation.⁵¹ Still, too much variability may be a sign of disorder. The CBR trajectories of CP2 showed more increases and decreases over time than those of CP1, who showed consistent but protracted rates of consolidation relative to the TD sample. Interestingly, the follow-up data revealed only mild speech impairments and mild CP in CP1 but significant speech and language impairments and severe CP in CP2. One might speculate that the neurological damage resulting in severe CP in CP2 may have produced greater CBR variability, perhaps reflecting a disturbed process of motor development.⁵²

Another important consideration is that relative to CP2, CP1 was producing a larger number of canonical syllable types by 12 and even 16 months, which may be associated with differences between the two infants’ later speech outcomes. These findings are consistent with earlier work suggesting limited consonant inventories by 12 months in CP²⁶ and other neurodevelopmental disorders.^53–55 Additional research is needed to further examine the rate and frequency of canonical syllable types used across reduplicated and variegated sequences in CP to determine the predictive nature of syllable types produced over time.

It is important to note that both infants at risk of CP showed apparent CBR regression between 11–12 months, as observed in Figure 2. Of course, the result could simply reflect sampling error. On the other hand, it is tempting to speculate that the regression might reflect early signs of impairment in their ability to use canonical syllables at the beginning of word learning, which is clearly common in TD infants at that age. Perhaps producing canonical syllables while trying to approximate words requires a particularly high level of motoric control in order to coordinate the production of well-formed syllables while focusing on the target forms. Future research to evaluate the speculation would require analysis of both word and babble productions across TD and at-risk infants in the age range we have studied here.

Several other factors may have influenced the two at-risk infants’ patterns. Both infants were diagnosed with hypoxic ischemic encephalopathy (HIE) around the time of birth; however, we only know the underlying cause of HIE in CP1 (intermittent cord compression) and do not know the extent of brain damage of either infant around the time of birth. Neuroimaging information can offer additional insight on this topic and has the potential to greatly inform early indicators of speech impairments. Environmental factors such as number of siblings and frequency of interaction eliciting advanced vocal forms may also play a role in speech outcomes. It is important to highlight that CP1 had one sibling and her mother was a speech-language pathologist. CP2 had three siblings and her mother was a stay-at-home parent.

Measuring canonical babbling onset in all-day home recordings

We expected the all-day recording method with random sampling of segments to offer a stable assessment given that the samples can be argued to be more representative than laboratory or short-term home recordings that have been used in the past.^11,14,20 But the study did not produce results suggesting clear monotonic growth in CBRs of all individual TD infants as might have been expected. Instead, day-to-day variability in CBRs was common, and sampling error may have contributed to much of the variability for both the at-risk and TD infants. Also, infant sickness, levels of fatigue, and daily environmental differences (e.g., interaction partners and activities) may also have resulted in canonical babbling differences across individual time points. Another notable observation in our study is the overall variability in the range of CBRs within the TD sample after 6 months. Perhaps with denser sampling across all infants (e.g., with multiple all-day recordings within each age range or collapsing multiple monthly data points across age bands), results would show more stable patterns, a possibility that is suggested for the TD infants in Figure 1. Additionally, averaged CBRs for multiple consecutive time points may better reflect developmental trajectories compared to observing CBRs across widely spaced recordings. The observed variability suggests a challenge for using the CBR as an indicator of infant vocal development at the individual level. A more thorough comparison of CBRs calculated across laboratory, home, and all-day home recordings is clearly warranted.

The results also suggest that the 0.15 CBR criterion for onset of the canonical stage may be too high for use with all-day recordings. By 9 months, the vast majority of TD infants are expected to be in the canonical stage by parent report and prior laboratory recording results, but our results suggest fewer than half of TD infants may show a 0.15 CBR based on random sampling from all-day recordings. It is important to note that earlier research used a 0.20 criterion to measure canonical syllable mastery¹⁵, and recent research evaluating the validity of this criterion suggests that 0.14 may in fact be more appropriate for laboratory recordings of parent-infant interactions.¹⁷ Even still, perhaps the traditional short-term laboratory or home recordings with all infants guaranteed to be awake and alert make it possible for parents to elicit a larger than usual amount of canonical babbling. At present, there is no empirically determined criterion for the judgment of canonical babbling onset in all-day home recording. In all-day recordings we have no guarantee that infants are alert (although they are deemed awake by the coders), and often it is clear that they are not engaged in any kind of vocal interaction. Still, if we wish to obtain representative samples, the all-day method may be preferable. The present results suggest more research is needed to determine optimal recording and sampling methods for assessing infant canonical status.

Limitations

Future studies should include larger sample sizes to enable generalization to the broader population of both infants at risk of CP and TD infants given the apparent variability of individual CBRs measured across all-day home recordings. Second, the selection of an average of only a handful of 5-minute segments per recording represents a small proportion of the infant’s daily vocal activity. It might be worthwhile to compare the current approach with purposive sampling of segments identified as having high vocalization counts to reflect infant performance during periods with high vocal activity. Above all, we emphasize that the variability across monthly CBRs may be a result of sampling error from recording only a single day per month. Anecdotal reports from parents suggest that infant vocalization patterns can vary based on wellness, interactivity levels, and environmental changes. In this study, parents were asked to record their infants on days they were home and engaging in their “typical” environment; however, internal and external factors can still impact performance of infants on any given day. Follow-up research should include several recording points per month to account for some of this possible sampling variability.

An important consideration in the future study of infants at risk for motor speech disorders is the coding procedure. Although the binary coding of canonical and noncanonical syllables is a commonly used method in the field of vocal development, alternative perceptual coding methods may further enhance our study of vocal behaviors in this population. The Stark Assessment of Early Vocal Development-Revised (SAEVD-R),²⁰ for example, was specifically developed using a motoric perspective and categorizes infant utterances across five developmental levels according to the articulatory complexity of sounds.³⁸ This more fine-grained level of analysis has great potential to offer additional perspective on the development of articulatory characteristics of infant vocalization to support the identification of vocal biomarkers for speech impairment in CP.

Conclusions

The findings from the present study offer preliminary support for future research directions in prelinguistic vocal development in CP. The infant with mild CP and mild speech impairment outcomes demonstrated delayed but consistent CBR consolidation with generally age-expected consonant inventories. The infant with severe CP and nonverbal speech outcomes presented with delayed and slightly reduced CBR consolidation at the ages available for TD comparison as well as a reduced consonant inventory by the latest age observed. Our results suggest variability in the emergence of canonical babbling may be common in both typical and atypical development. These observations warrant additional investigation with larger sample sizes in all-day home recordings. An important goal is to determine optimal methods of observation of prelinguistic vocal stage emergence in this population to help flag children at risk of speech impairments alongside clinical assessment to facilitate earlier referral for speech pathology services.

Data availability statement

Due to the nature of this research, participants of this study did not agree for their data to be shared publicly, so recording data is not available.