Melody of Vocants: Fixed Pattern or Shaped by Hearing?

Introduction: Vocants as infants’ first vocalic utterances are produced laryngeally while the vocal tract is maintained in a neutral position. These “primitive” sounds have sometimes been described as largely innate and, therefore, as sounding alike in both healthy and hearing-impaired young infants. Objective: The objective of this study is to compare melody features of vocants, recorded during face-to-face interaction, between infants (N = 8) with profound congenital sensorineural hearing loss (HI group) and age-matched (N = 18) controls (CO) group. The question was as follows: does a lack of auditory feedback have a noticeable effect on melodic features of vocants? Methods: The cooing database totalled 6,998 vocalizations (HI: N = 2,847; CO: N = 4,151), all of which had been recorded during the observation period of 60–181 days of age. Identification of the vocants (N = 1,148) was based on broadband spectrograms (KAY-CSL) and auditory impressions. Fundamental frequency (F0) analyses were performed (PRAAT) and the pattern of the F0 contour (melody) analysed using specific in-lab software (CDAP, pw-project). Generalized mixed linear models were used to perform group comparisons. Results: There was a clear predominance of a simple rising-falling pattern (single melody arcs) in vocants of both groups. Nonetheless, significantly more complex contours, particularly, double-arc structures, were found in vocants of the CO group. Moreover, vocants of the HI group were shorter than those uttered by the CO group, while the mean F0 did not significantly differ. Conclusion: Vocants are characterized by both, innate features, found in HI and CO groups, and features that additionally require a functioning auditory system. Even at an early pre-linguistic stage, somatosensory sensations cannot compensate for a lack of auditory feedback. Vocants might be relevant in the early diagnosis of hearing disorders and assessments of the effectiveness of, or adjustments required to, hearing aids.


Introduction
The auditory system of human infants starts to function intrauterine and is one of their most complex sensory systems [1].Auditory-evoked cortical activation has been confirmed with fMRI in the near-term foetus [2,3], and as early as 33 weeks of gestational age [4].The auditory system is specifically stimulated and imprinted in the last trimester of pregnancy, especially via bone conduction and virtually around the clock whenever the mother speaks, sings, walks, or plays music [5,6].Along with the rhythm of her heartbeat, the mother's voice is one of the most effective intrauterine stimuli [7].The precocious organization of the language-related areas in the brain enables the generation of auditory memory traces of the suprasegmental (melodic-rhythmic) properties of the surrounding language in utero [8].Perania et al. [8] demonstrated in two-day-old newborns that the "language-related neural substrate is fully active in both hemispheres with a preponderance in the right auditory cortex" (p.16,056).The "intensive training" of an already functioning auditory organ, which lasts for about 3 months in healthy foetuses, leaves traces in the brain that modulate not only auditory imprints but also the vocal production of the newborn [9][10][11].
The early maturity of the coupled auditory-vocal system raises a question about the effects of congenital hearing impairment on the earliest, still simple, comfort vocalizations, which are sometimes described as innate [12,13].These particularly include the vocants, which are very early vocalic utterances produced in a pleasurable comfort state and are sometimes considered to be the precursors of later vowels [14].Vocants as the first "primitive" vocalic comfort sounds are produced laryngeally while the vocal tract is maintained in a neutral position.Consequently, the specific resonance frequencies (formants) typical of vowels are not yet present at this stage.Like spontaneous natural crying, vocants are produced purely laryngeally, but they are recognized to be considerably shorter, more nasalized, and quieter [15].Vocants are the typical non-cry vocalizations uttered as part of the cooing repertoire, mainly in face-to-face situations [16].By focussing on vocants, supralaryngeal factors (consonant-like elements) were excluded.This resulted in a strong homogenization of the analysed infant sounds.
While vocalizations of older infants were found to be produced in relatively similar ways in normal hearing and hearing-impaired infants [17,18], a specific analysis of cooing or even vocants is still pending.However, a comparative study investigating cooing of normal-hearing three-month-old German and Cameroonian (Nso) infants in a face-to-face situation [19] observed differences with respect to some features.The cooing of all the Cameroonian infants regularly contained consonant-like elements (clicks) that not only imitated typical para-linguistic clicks of the parents but were also not observed in their German counterparts [19].Nevertheless, it may still be the case that at least pure vocalic cooing sounds (vocants) demonstrate innate features that are not or less influenced by auditory feedback.The rather "primitive" vocants could resemble in their simple acoustics without supralaryngeal activity rather fixed calls of non-human primates.If so, vocants should not differ in terms of their characteristics between healthy and hearing-impaired infants.We analysed standard measures as mean fundamental frequency (F 0 ) and sound duration but focused on properties of the F 0 contour (melody) and its development.In our experience, development metrics are extremely suitable to discover deviations from the typical developmental path towards language [20][21][22][23].Here, we analysed melody complexity, a measure that characterizes vocal control [22].The concept of melody development is based on the research of Wermke and Mende who summarized their findings in a melody development (MD) model [24][25][26].This model expresses early vocal development by an ordered sequence from initially prevailing simple (single-arc [SA]) melody contours to double-arc (DA) contours, followed by more complex (multiple-arc [MA]) contours.This development is assumed to be regular, unidirectional for each vocalization type, and universal in nature [24][25][26].We assumed vocants to reflect at least the first developmental step by exhibiting DA patterns.The role of auditory feedback upon the unfolding of the program has not yet been explored.
Hence, the study raises the question whether vocants are characterized by innate, fixed properties that are alike in normal-hearing and profoundly hearing-impaired infants.In case vocants will demonstrate features that are not invariant, but influenced by auditory feedback, respective cues can be used to diagnose deaf children at an early stage.To the best of our knowledge, this is the first study to analyse vocants quantitatively and compare their melodic and temporal features in healthy and profoundly hearing-impaired infants.

Recording
Digital (48 kHz Fs, 16 Bit) recordings of comfort vocalizations (no crying) were made using a TASCAM DAT recorder (DR-100) and an Earthworks microphone (TC20).The recording took place within a sound booth or noise-reduced room, with the distance between the microphone and the infant's mouth being approximately 15 cm.All the comfort vocalizations an infant produced during joyful interactions with their mother in the ENT department of our hospital and partially at home (longitudinal data) were recorded.The average length of a recording session was about 10-15 min.

Participants
Our sample comprised a total of 26 infants: eight (three female) with profound congenital sensorineural hearing loss (HI group) and 18 who were age-matched (eight female) and had normal hearing (control [CO] group).The infants came from German families (bilinguals: CO: 39%; HI: 13%) that were closely comparable in terms of their socio-economic status and their engagement in fostering their child's progress.Each parent had experienced at least a high school education, and the monthly family income reflected a middle-class standard of living.The psychomotor and cognitive development experienced by all the infants was normal.Cranial magnetic resonance imaging examinations of those in the HI group revealed no abnormalities.The audiological data obtained are reported in Table 1.
Four of the eight infants in the final HI group were recorded repeatedly for the purposes of the study, the goal of which was to improve the combined audiological/pre-speech treatment program for HI infants referred to the audiology service of the ENT and plastic surgery departments at the University Hospital Wuerzburg (cf.Table 1).The data for three infants of the CO group were also recorded repeatedly as part of the early language development program referred to above.All audio files were taken from the sound archive at the Center for Pre-Speech Development and Developmental Disorders at University Hospital Wuerzburg.Four of the eight infants in our HI group had been given hearing devices at the end of the audiological/ pre-speech treatment program's observation period, although these were of limited value because the hearing loss was profound.All of these infants subsequently received or will receive cochlear implants.
In total, then, audio files were analysed of 26 infants aged from 60 to 181 days.All of them had (1) experienced an inconspicuous pregnancy; (2) age-appropriate somatic development at birth; (3) an inconspicuous perinatal adjustment (APGAR score of ≥8 at 5 min and ≥9 at 10 min); and (4) metabolic indices in the normal range.The mean birth weight was 3,375 g (range: 2,310-4,455 g) in the HI group and 3,416 g (range: 2,800-4,610 g) in the CO group.The mean gestational age for the former was 38.6 weeks (range: 34-41 weeks; including a monozygotic preterm twin pair) and 39.3 weeks (37-41) for the latter.

Vocalization Repertoire and Identification of Vocants
In the first step, all the recorded audio sequences (WAV files) were manually annotated into single events using PRAAT v. 6.0.373[27].These single events encompassed all types of egressive vocalizations of the infant, breath sounds, silent intervals, speech, and background noise.The egressive vocalizations were initially identified, marked using manually set cursors, and then automatically saved as single audio signals (Fig. 1).An egressive vocalization was defined as the vocal output occurring in a single cycle of expiratory breathing (cf.Fig. 1).The final database totalled 6,998 comfort vocalizations (cooing sounds), 2,847 vocalizations from the HI group, and 4,151 from the CO group, all of which had been obtained during the observation period of 60-181 days of age.As typical for cooing, the recorded comfort sounds comprised not only vocants but also isolated closants or vocant-closant combinations (cf. the raspberry sound in Fig. 1).Closants are primitive consonant-like elements, similar to /k/ and /g/ [18].In this paper, only vocants were assessed.The criteria finally retained for the labelling of vocants was auditory impression first, which was checked by broadband spectrograms then (Fig. 2).As vocants are produced with a neutral vocal tract, vocants' spectrograms displayed stable ("fixed") resonance frequencies related only to vocal tract geometry (supralaryngeal inactivity).
Table 2 reports the vocalizations analysed per infant and the number of identified vocants.As can be seen in Table 2, there was high inter-individual variability in the number of comfort sounds uttered per session.This was mainly because the infants included both very "talkative" and "less talkative" vocalizers.This was considered in the statistical analysis by applying mixed models.

Analysis of Vocant Features
As described elsewhere in more detail [28], the F 0 was automatically analysed and carefully verified using PRAAT v. 6.0.373.The analysis was then repeated manually by two of the authors (FC and DB) in cases where there were obvious F 0 -tracking problems in the automatic routine.After transferring these F 0 data to a further software system that was developed specifically for the purpose of pre-speech analyses (CDAP; pw-project, Germany), we produced frequency-time diagrams with a logarithmic scale for the F 0 and a quarter-tone grid (Fig. 3a).These diagrams were used to analyse temporal features (duration: DV, DA1, DA2), and F 0 features automatically based on time intervals that were marked manually by cursors set at the beginning and end of the F 0 contour (melody).We analysed three F 0 features: (1) mean F 0 per vocant (Fig. 3a), (2) F 0 contour in terms of melody pattern (cf.section Analysis of Vocants' Melody Pattern), and (3) tnorm(F 0 max): time normalized F 0 maximum per melody arc to specifically characterize the F 0 contour in DA melodies (Fig. 3b).In a previous work, the latter feature was introduced in order to quantitatively evaluate F 0 contour of SA melodies [11].Here, we used the method to characterize contour characteristics of DA vocants to describe precursors of prosodic organization in DA vocants (cf.section

Analysis of Vocants' Melody Pattern
Infants' vocalizations demonstrate melody patterns (F 0 contours) that can be assigned to exhibit an either 'simple' (SA) or "complex" (MA) pattern [15,28,29].For this pattern analysis, all F 0 contours (melodies) were post-processed, with low-pass filtering with a cut-off frequency of about 40 Hz applied (CDAP; pw-project, Germany).This eliminated high-frequency modulation noise to filter out short-term fluctuations in F 0 that are not relevant to the intended pattern analysis of the melody pattern [20,26].The analysis of melody patterns was based on a dif-ferentiation between vocants exhibiting a simple, SA or a complex, MA melody pattern [26,28,30].A melody arc was defined as being longer than 150 ms and as exhibiting a frequency modulation amplitude (FM amplitude: F 0 max -F 0 min) of at least two semitones (cf.Fig. 3a).Vocants as shorter as 150 ms were excluded as vocalizations were containing laryngeal constrictions [31].Using the melody diagrams in CDAP with a logarithmic scale for F 0 and a quarter-tone grid, vocant melodies were objectively subdivided into those with a SA pattern and those with a MA pattern (melody with ≥2 arcs), based on the arc criteria referred to above (cf.Fig. 3 exemplifying DA melodies).DA patterns prevailed among all complex melodies of vocants and were specifically analysed.Only a few vocants with a melody exhibiting more than two melodic arcs were observed.

Vocant Analysis
Prosodic Precursors in Vocants with a DA Melodic Pattern Prosodic features such as melody, intensity, and duration are essential for an infant acquiring language.Meanwhile, there is mounting evidence that infants are sensitive to prosodic features of their native language from very early on, long before first syllables in canonical babbling occur [9-11, 32, 33].In comparison to perception, the beginning of prosodic organization in sound production seems to emerge several months later [34][35][36][37].Here, we investigated prosodic precursors in terms of time normalized F 0 maxima and arc duration in DA vocants (Fig. 3a, b).This approach mimics previous analyses of stress in disyllabic babbling in older infants [34,38].Note that, no intentionality is assumed here, instead vocally available coordinative abilities to the later task of language acquisition are assessed.
Applying a method to quantify the contour of melodic arcs introduced by Mampe et al. [11], we investigated DA contour characteristics by normalizing the arc duration to 1 and determining the normalized time (tnorm[F 0 max]) corresponding to F 0 max for each arc (see Fig. 3b).Values of tnorm(F 0 max) <0.4 represent "falling contours," tnorm(F 0 max) values >0.6 represent "rising contours," while values between 0.4 and 0.6 characterized "symmetrically" rising then falling melodic arcs.The occurrence of different "contour combinations" in DA vocants was statistically assessed.
Additionally, the duration ratio of the two arcs in DA was expressed by the quotient of the duration of the first arc and the total duration of the vocant (DA1:DV; cf.Fig. 3a).This yielded a standardized value between 0 and 1 that can easily be interpreted.

Statistical Analysis
Due to the hierarchical structure in our data, a generalized linear mixed model method was used to investigate group differences.Our study included longitudinal data with repeated measurements, and the data were subject to both fixed and random factors.Descriptive information for all measures was also reported.All these statistical analyses were performed using SPSS version 26 (IBM Corp., Armonk, NY, USA).

Prosodic Precursor Analysis
A calculation of effect size measures with significance or confidence intervals directly within statistical analyses of DA vocants is only possible, when the hierarchical nature of the data is taken into account.As we have multiple measures per child, the data are not independent.We therefore investigated the effect size for the dependency between contour category of first and second arc within each group by calculating Cramer's V, Wermke/Clad/Blum/Cebulla/Shehata-Dieler but without reporting the p values or confidence intervals.Hence, Cramer's V is interpreted here only as a descriptive measure.
For being able to report inferential statistics for the prosodic precursor analysis, a further generalized linear mixed model was calculated in contour category of the second arc as dependent variable, group, and contour category of first arc as predictors.In the framework of the mixed model, it is possible to take the dependency of the measures from the same person into account and thus also inferential statistics are allowed to be reported here.In this model, the interaction between group and contour category of first arc would show that the dependency between contour category of first and second arc differs between the groups.So, it would provide a p value to the descriptive comparison of the effects via Cramer's Vs noted above.Additionally to the p value, the odds ratios (ORs) with 95% confidence intervals will be reported for further investigation of the effect and potential use in meta-analyses.The Akaike information criterion was used to assess or compare different models and to consider the best model fit.The analyses (Cramer's V and generalized linear mixed model for prosodic precursor) were conducted using the software R version 4.2.2 (R Core Team [2022]).

Results
In the analysis of the melodic pattern of vocants across the observation age of 60-181 days, MA and SA contours were observed in both groups.However, SA pattern prevailed in both groups.Among MA pattern, only DA regularly occurred, while pattern that were more complex did rarely (1%) occur (CO: four triple-arc and one fourarc melody; HI: three triple-arc melodies).We observed that vocants with DA contour occurred twice as often in the CO group (Fig. 4).
A statistical analysis of group differences was conducted by applying a generalized mixed model to the dichotomous variable of SA ("0") versus MA ("1") pattern in relation to all 1,148 vocants within the observation age range.This yielded significant outcomes, with more MA contours (≥ two arcs) in the vocants of the CO group (Table 3).The table shows the evaluations of the mixed model with regard to melody complexity.With a p value of 0.012, the groups differ significantly with respect to the binary variable of melody complexity.The OR, with a value of 0.286, indicates that the probability of an SA pattern is 28.6% higher for group HI compared to group CO.There is also a significant age effect with a p value of 0.031.The effect size of 1.007 indicates that the probability of an MA pattern increases by 0.7% daily.
Table 4 shows the descriptive analysis for the vocant duration and their mean F 0 .We only considered SA vocants as prevailing pattern.It should be noted that vocants were only classified as such if they had a length of at least 150 ms (cf.Methods).Table 5 reports the results of the generalized mixed model.A significant difference was identified, with the vocants of the infants in the HI group being shorter than those uttered by the healthy controls.No significant group differences were found for mean F 0 .
Figure 4 above demonstrates that the proportion of DA melodies varied between the groups.Against this background, the melodic-(pre)prosodic properties are only reported for vocants with DA melody.For each arc, three contour categories were distinguished: "falling," "symmetric," and "rising."The frequencies of the specific combination patterns observed are reported in Table 6.The patterns are sketched graphically to aid understanding.The first arc of the DA vocants was predominantly falling.This was very common in the vocants of the infants in the CO group (69%).In the HI group, however, a falling first arc was identified in only 52% of the DA vocants (Table 6).The opposite picture emerged in relation to the contours of the second arc in the DA vocants: there was an accentuation of the end of the arc, i.e., a rising contour was predominant (Table 6).The most frequent combination in the vocants of the HI group was a symmetric-symmetric pattern (26%).However, the falling symmetric and falling-rising patterns were almost as common (both 24%).The DA vocants of the infants in the CO group predominantly had a falling-rising pattern (34%).Despite the small sample size, we report the statistical differences in DA combination types in order to conduct future research, including meta-analysis: With respect to effect size, the statistical analyses provided a Cramer's V of 0.333 in the HI group compared to a value of 0.182 in the CO group, we see a stronger dependency in the HI group.For being able to report inferential statistics (p value, odds ratio, confidence interval), a further generalized linear mixed model was calculated for contour combinations in a DA mixed model (Table 7).As the predictor contour of first arc has three categories, the variable is used dummy, coded with the category falling as reference category.We therefore got the results for each combination of the other two categories (rising and symmetric).As reported in Table 7 (two last lines: interaction), results show that none of the interaction terms is significant, but the p values are small (0.115 and 0.091), indicating a tendency.The ORs show a similar effect for both interaction categories (0.074 and 0.071).

Discussion
Unlike vocalizations that are based on neurophysiological control that involves auditory feedback, a hearing impairment would not affect sound properties of innate vocalic sounds.Therefore, the study raised the question: does a lack of auditory feedback have a noticeable effect on melody features of vocants?
To answer this question, we analysed several prosodyrelevant features: mean F 0 , duration, melody complexity, and specific contour features in DA vocants.The vocants examined in this study had both similarities and differences between the hearing and profoundly hearing-impaired infant groups.There was similarity in the fact that the vocants had a relatively short duration (on average, about 400 ms) and a mean F 0 of about 340-360 Hz.In this respect, this finding confirms earlier descriptions of vocal sounds [15].The prevailing short duration of vocants does probably explain why the average F 0 variability (standard deviation) was similar between the groups with vocants of the CO group with only a 10 Hz higher variability in average.F 0 variability is related to later language development [39] and was found in crying to be lower in hearing-impaired infants [40].This is in agreement with our finding that the vocants of the HI group had a little lower standard deviation (not significant).However, in our experience, this measure is less reliable in short vocalizations, like vocants, compared for instance to the much longer character of natural cry utterances.
With respect to the original question, the new and most important findings are related to vocants' melody pattern and their development.There was similarity in the fact that the vocants predominantly had a simple pattern consisting of an SA in the F 0 contour (melody) in both groups.There was also similarity in the fact that melody complexity significantly increased with age in both groups.Further similarity was in the fact that DA contours were the second most common pattern in both groups.This result is in agreement with the MD model by Wermke and Mende [24,26]: "the simplest form of complexification in this sense is a doubling of shape identical melody arcs.[. ..]Further complexification of (cry) melody is characterized by multiple repetitions or concatenations of single melody arcs or arc-like patterns."[24] (p.636) In a recent longitudinal study, a significant age-dependent developmental pattern towards more melody complexity was demonstrated in cry and non-cry vocalizations over the first 6 weeks of life using fractional   polynomial multi-level mixed effects logistic regression models [28].A further longitudinal study confirmed the model [29]: The study of how often different F 0 contour patterns (melody patterns) occur in various vocalizations in the first 6 months found that about 15-20% of the subjects' crying, cooing, and babbling was composed of DA melodic patterns [29].Furthermore, an early study by Kent et al. [15] involving infants aged between three and 6 months identified that 9-16% of the vocants comprised DA F 0 patterns and 2-6% a more complex pattern.(Note that vocalic sounds of the study by Kent et al. [15] involved vocants and vowels).This means that the developmental trajectory of the melodic patterns of vocants belong to the same developmental trajectory as the other types of early vocalizations [26,28,30].The theoretical concept of melody  Our data show the opposite: in the HI group, the vocants exhibited significantly less melodic complexity.This was due to the prevailing DA pattern among all complex patterns.The percentage of DA in the CO group was twice as high as that in the HI group.
Furthermore, a specific analysis of DA patterns was performed to evaluate vocal flexibility in terms of variations of F 0 contour and duration between the two arcs.Analyzing the duration of each melodic arc (DA1 and DA2) in DA melodies provided no systematic "final lengthening:" in vocants of the CO group, the same number of longer first arcs or longer second arcs was observed.In DA vocants of the HI group prevailed a longer first arc.This means there is a higher respiratorylaryngeal flexibility in the CO group.Influences of own sound production and the surrounding language could have played a role as these were absent in the HI group.Just a certain somatosensory input could be responsible for vocants' feature variation in the HI group.However, systematic studies are still pending here.
Although the investigations of arc-specific contour pattern in DA vocants are only pilot in nature, they probably delivered the most important result for future research from a clinical perspective.We found differences in the occurrence of specific contour combination patterns between groups.The statistical analysis delivered a tendency for the interaction between "contour of arc 1" and infant group to "contour of arc 2." This means that the relationship between arc 1 and arc 2 is different for the two groups, with a weaker relationship in the CO group (Cramer's V lower).Therefore, it was found that the contour of arc 2 seems to predict the contour of arc one in the HI group, whereas the two contours are more independent in DA vocants of the CO group.We tentatively interpret this result to mean that auditory feedback seems to provoke more trial and error and contour variation (vocal play) in vocant production of the CO group.The production of melody contour in vocants of the HI group depends probably more on pure physiological mechanisms.Whether the differences within the CO group are possibly due to the different language influences needs to be investigated in further studies.In any case, the playful rehearsals described above will later enable the prosodic patterns that are typical of the respective ambient language to predominate.

Conclusion
For the first time, vocant features related to later speech prosody were quantitatively analysed.Although the study has pilot character, it suggest vocants' potential as diagnostic tools for hearing disorders, emphasizing auditory feedback's importance in vocal development and hearing aid evaluation.This study has provided new insights into the properties of very early "primitive" comfort vocalizations.The study supports the hypothesis that there is a universal, innate developmental program for the F 0 contour (melody) right at the start of vocal production.The finding that infants with profound hearing loss had significant differences in their prosody-related features of seemingly primitive vocants demonstrates the importance of a wellfunctioning auditory system.Indeed, somatosensory sensations cannot compensate for a lack of auditory feedback, even at an early pre-speech age.In this respect, the study completes our understanding of early vocal development.In addition to analyses of other vocalizations, like natural crying or canonical babbling, we believe that vocants are also relevant in the early diagnosis of hearing disorders and assessments of the effectiveness of, or adjustments required to, hearing aids.

Fig. 1 .
Fig. 1.Example of an annotated-recorded cooing sequence in PRAAT.The upper section of the PRAAT output window displays the amplitude of the recorded sequence.The middle section shows the frequency spectrogram (frequency range linear 0-4 kHz).The lower area of the output window shows the annotation variables of the individual events.Here, the sequence consists of a background noise (bn), vocant, silent pause (s), breathing, raspberry (articulated sound).

Fig. 2 .
Fig. 2. Examples of typical broadband spectrograms (frequency vs. time) of two vocants.(KAY/Pentax CSL).The F 0 and the two lowest resonance bands of the vocal tract can be easily identified along the dark horizontal bands.Both images correspond with the typical resonance frequency image of a vocant.Resonance frequencies are very constant in both examples (no movement) and reflect mainly the geometry of the vocal tract (articulatory inactivity).

Fig. 3 .
Fig. 3. a Frequency-time (melody) diagram with a logarithmic scale for the F 0 and a quarter-tone grid.The melody of a vocant exhibiting two melodic arcs and their metrics are displayed.b Frequency-normalized time diagram of a DA melody.The diagram explains the determination of SA contours.These categories were analysed by normalizing the arc duration to 1 and determining the normalized time (tnorm[F 0 max]) corresponding to the maximum F 0 (F 0 max).The value of tnorm(F 0 max) = 0.24 represents a "falling contour" of arc 1, while tnorm(F 0 max) = 0.59 represents a "rising" contour of arc 2.

Fig. 4 .
Fig. 4. Relative proportion of single arcs (SAs), double arcs (DAs), and melodies with more than two arcs (>DA) of the vocants of the HI group (left) and the CO group (right).

Table 3 .
Mixed Model analysis of structural complexity differences between the HI group and CO group

Table 4 .
Duration and mean F 0 of SA vocants of the HI and CO groups

Table 6 .
Combinations of first and second arc contour in DA vocants Does This Mean Auditory Feedback Is Not Required for the Unfolding of the Inborn MD Program?