The authors are with the Speech Science Research Centre, Queen Margaret University, Musselburgh, UK.
Coarticulation as an Indicator of Speech Motor Control Development in Children:
An Ultrasound Study
Natalia Zharkova, Nigel Hewlett, William J. Hardcastle
There are still crucial gaps in our knowledge about developmental paths taken by children to adult-like speech motor control. Mature control of articulators during speaking is manifested in the appropriate extent of coarticulation (the articulatory overlap of speech sounds). This study compared lingual coarticulatory proper-ties of child and adult speech, using ultrasound tongue imaging. The participants were speakers of Standard Scottish English, ten adults and ten children aged 6–9 years. Consonant-vowel syllables were presented in a carrier phrase. Distances between tongue curves were used to quantify coarticulation. In both adults and children, vowel pairs /a/-/i/ and /a/-/u/ significantly affected the consonant, and the vowel pair /i/-/u/ did not. Extent of coarticulation was significantly greater in the children than in the adults, providing support for the notion that children’s speech production operates with larger units than adults’. More within-speaker variability was found in the children than in the adults.
Coarticulation and Motor ControlCoarticulation refers to the articulatory overlapping of adjacent sounds in speech. An example of coarticulation is the measurable and perceivable difference between two realizations of the consonant /s/, in the syllables /si/ and /su/, arising from the influence of the following vowel. Adults without speech disorders manifest a cer-tain extent of articulatory overlap, typical for a given language and acceptable to listeners. Appropriate coarticulation is indicative of mature control of articulators during speaking. A recent article on motor control in typical and disordered speech claims that lack of coarticulatory cohesion is related to problems in planning and programming of speech movements (Ziegler & Maassen 2007). Models of speech production, including models of coarticulation, have the same underlying challenges as models of nonspeech movement. One of the central questions is how patterns are generated through cooperation among different parts of a system (Kelso 1995). Coarticulation is a measurable manifestation of such cooperative activity in speech.
A number of models of speech production have been developed over a few decades (e.g., Öhman 1967; Keating 1988; Guenther 1995; Recasens et al. 1997; Perrier & Ma 2008). Common to all coarticulation models is an interest in under-
Coarticulation in Children: An Ultrasound Study 119
standing the control of the skilled activity of speech production, as well as the relation between discrete elements of phonological systems and their realizations in connected speech (Kühnert & Nolan 1999). Some models of coarticulation use very similar concepts to those used by theories and models of nonspeech perfor-mance (e.g., see Kandel et al., 2000, p. 817, for a description of general postural control that applies equally to speech). An important postulate of dynamical systems theory, which is aimed at explaining coordination in human movement systems, is that physical and biological systems are characterized by generic processes of self-organization, and movement patterns emerge through these processes (Glazier et al. 2003). This idea has also been applied to speech production models, nota-bly the coproduction theory based on the task-dynamic model (e.g., Saltzman & Kelso 1987; Saltzman & Munhall 1989; Saltzman 1991; Browman & Goldstein 1992; Saltzman & Byrd 2000). In the task-dynamic model, patterns of articulatory movement are represented as coordinated, dynamic gestures that overlap in time.
Studies of speech motor control have addressed coarticulatory processes occur-ring in various parts of the speech system: the larynx, velopharyngeal subsystem, tongue, jaw and lips (for a comprehensive review, see Hardcastle & Hewlett 1999). Particularly challenging for imaging and quantification is lingual coarticulation. The tongue is an internal articulator; its flexibility influences acoustic characteris-tics of the sounds produced during speech, and therefore correct perception by the listeners. A number of acoustic studies of lingual coarticulation are described in Recasens (1999). Articulatory studies that have analyzed lingual coarticulation have included the methods of electromyography (e.g., Bell-Berti & Harris 1974; Gay 1974; Gay et al. 1974; Fuchs et al. 2004); X-ray (e.g., Engstrand 1988; Lindblom et al. 2002); electropalatography (e.g., Farnetani & Recasens 1993; Engstrand et al. 1996; Gibbon & Nikolaidis 1999; Recasens 1999; Recasens & Pallarès 2001; Zharkova 2007); electromagnetic articulography (Fowler & Brancazio 2000; Fuchs et al. 2004; Recasens & Espinosa 2009); ultrasound (Vazquez Alvarez & Hewlett 2007; Wodzinski & Frisch 2006; Zharkova 2007; Zharkova & Hewlett 2009). The present study addresses the issue of the maturation of tongue control, by compar-ing anticipatory coarticulation in children and adults, using articulatory measures derived from ultrasound imaging.
Acquisition of Coarticulation by ChildrenInformation about developmental paths taken by children to adult-like motor control of speech is relevant to a number of disciplines. However, there are still crucial gaps in our knowledge. A recent literature review claims that there are very few laboratories working in the area of oral motor development for speech (Smith 2010). Some of the evidence from previous research on typical speech development sug-gests that the amount of gestural overlap in speech changes with age (e.g., Ziegler & Maassen 2007). Research on speech disorders has shown significant differences in coarticulation between children with typically developing and disordered speech. For example, speech production of children with developmental apraxia of speech was reported by Nijland et al. (2002; 2003a; 2003b) to be deviant compared with typical child productions (see also reviews in Hardcastle & Tjaden 2008). To establish to what extent deviant speech differs from the range of normal variability, and to undertake accurate diagnosis and subsequent successful treatment of speech
120 Zharkova, Hewlett, and Hardcastle
disorders, it is necessary to have more data on typical speech development. This study provides such data.
In the current study, we analyzed coarticulation in adults and children aged between six and nine years old. Ultrasound-based measures of tongue movement adapted from Zharkova & Hewlett (2009) were used to assess the differentiation of consonant curve sets across vowel environments. We used the consonant /∫/ followed by the vowel /i/, /a/ or /u/ in the study reported here. This study aimed to establish whether both children and adults show lingual coarticulation, and if so, whether children do so more than adults. Another aim was to find out whether children are more variable than adults. Predictions in this paper are situated within the theoreti-cal approach developed by Nittrouer et al. (1989; 1996). This approach is based on acoustic evidence showing that both children and adults demonstrate evidence of vowel-on-consonant coarticulation. Nittrouer et al. observed stronger vowel-on-consonant coarticulatory effects in children than in adults. Based on this finding, they postulated that in child speech, movements of individual articulators over a syllable are more “global” than in mature adult speech, resulting in syllable-sized units of production, as opposed by the more dynamic approach taken by adults, which affords the individual segments more autonomy. Each of the following sec-tions will briefly review previous studies and formulate our predictions in relation to Nittrouer’s theory.
Coarticulatory Effects
In Nittrouer et al. (1989; 1996), lingual coarticulatory effects of contrasting vocalic environments on the adjacent consonant were documented in typically developing American English speaking children aged three, five and seven years old, as well as in adults (see also Siren & Wilcox 1995, for similar results). Nittrouer et al. com-pared differences in spectral energy in the region of the second formant (F2) of /∫/ across different following vowel environments; the measurement was taken at 30 ms before the onset of the vowel. They did not measure lingual coarticulation any earlier in the consonant than this, and indeed they could not have done so, because F2 is not usually apparent any earlier. Ultrasound analysis gives us an opportunity of directly comparing tongue shapes at mid/∫/, when relevant information is absent from the acoustic signal. In Nittrouer et al. (1989), these comparisons were made using the vowel pair /i/-/u/; in Nittrouer et al. (1996), the vowel pair /a/-/i/ was used. In Scottish English, the lingual articulation of /u/ is more similar to that of /i/ than is the case in American English. On the other hand, there is almost certainly some difference between the lingual postures of /i/ and /u/, albeit a small one. This gave us grounds to expect a lingual coarticulatory effect on /∫/ from this vowel pair, which could be clearly revealed by direct articulatory analysis.
Most developmental studies have agreed with Nittrouer et al. at least to the extent of finding lingual coarticulation in typically developing children. Sereno et al. (1987) compared mean spectral peak values of the consonants /s/, /t/ and /d/ from consonant-vowel syllables with the vowels /i/ and /u/, produced by adults and children aged three to seven years. In both groups, there was a significant lowering of spectral energy for consonants preceding /u/ as opposed to /i/. Sereno & Lieberman (1987), in a study of 14 children aged between two and seven years old, analyzed spectral peaks at the release of the consonant in /ki/ and /ka/ syllables, and observed
Coarticulation in Children: An Ultrasound Study 121
lingual coarticulatory effects in 12 children. Katz et al. (1991) reported anticipatory lingual coarticulation in the syllables /si/ and /su/ in adults and children aged five and eight years old. In Nijland et al. (2002), in a study of six-year-old children, differences in F2 across vowel contexts were reported at the preceding consonant onset, suggesting the presence of vowel-on-consonant coarticulation. In our earlier ultrasound studies, we found evidence of coarticulation in children (Zharkova et al. 2008a; 2008b). However, no measurable coarticulatory effect was reported by Katz et al. (1991) for their youngest (three-year-old) age group.
Our predictions for this aspect of the study were as follows. For the vowel pairs /a/-/i/, /a/-/u/ and /i/-/u/ we hypothesized that we would observe a significant effect on /∫/, in children and in adults. We expected that the effect of the /i/-/u/ vowel pair would be smaller than those of the other two vowel pairs, due to the characteristics of Scottish English.
Extent of Coarticulation
The claim that children use more, not less, coarticulation than adults is the central claim of Nittrouer’s theory, and it was developed in opposition to the influential work of Kent in the 1980s. Kent (1983) claimed that overlapping of movements of different articulators over time is required for producing fluent and rapid speech. As an opposite to overlap, Kent referred to the principle “everything moves at once”, where different articulators are moving synchronously rather than overlapping, as being more common in developing or impaired speech. Kent (1983) provided several examples from child speech suggesting that young children coarticulate less extensively than adults. One example was the greater degree of undershoot of the vowel /u/ in adults than in four year old children, in the sentence “we saw you hit the cat”. Kent also referred to the study by Thompson & Hixon (1979), who found more anticipatory nasalisation in /ini/ sequences with increasing age (there were over a hundred participants, ranging between three year old children and adults, with four-month intervals). The interpretation of these age-related dif-ferences suggested by Kent is that the temporal ordering of speech segments must be mastered first, and the temporal structuring (or overlap) of these sequential segments is perfected later.
After the publication of Kent (1983), some studies reported data suggesting that child speech is not necessarily either more or less “segmental” than adult speech (e.g., Sereno et al. 1987; Katz et al. 1991). Nittrouer et al. (1989) set out to test the idea that children are more segmental than adults. In that study, as well as in Nittrouer et al. (1996), they used amount of spectral energy in the region of F2, in the same consonant across different vowel environments, as an indicator of degree of coarticulation. The children were aged three, five and seven, and there were ten children in each age group, and ten adults. The authors reported greater differences in F2 across vowel contexts in children than in adults. They interpreted these results to mean that in children, tongue contours in the consonant adapted more to the vowels, and that children initiated their vowel gestures earlier in the syllable than adults. They called this type of production “syllabic”, as opposed to “segmental”, meaning that the consonant adapted to the vowel within the same syllable more than in mature adult productions. Comparable results were presented in Nijland et al. (2002), in a study of six children aged five to seven years old. The
122 Zharkova, Hewlett, and Hardcastle
data were / CV/ sequences, with several Dutch consonants and the vowels /i/ and /u/. Nijland et al. found that at the consonant onset and midschwa, F2 differences across vowel contexts were significantly greater in children than in adults, suggest-ing greater coarticulation in children. Nittrouer et al. (1989; 1996) and Nijland et al. (2002) normalized the data for vocal tract size by computing ratios of F2 values across vowel contexts for each speaker.
Interestingly, Nittrouer et al. (1989) argued that one of Kent’s examples from child speech was consistent with their hypothesis. Kent (1983) reported a continu-ous rise of F2 throughout the vowel in the word “box” (demonstrating anticipation of the tongue position for the following velar consonant) produced by adults, as opposed to a relatively steady state F2 throughout the vowel in the same word produced by children. In Kent’s interpretation, these data mean that in children, there is less overlap between the tongue articulation for the vowel and the conso-nant, i.e., less coarticulation. Nittrouer et al. (1989) suggested that such F2 pattern could have been observed because children may have raised their tongue bodies in anticipation of /k/ at the onset of the vowel, thus demonstrating more coarticula-tion than adults. They claimed that this interpretation could be explained by their hypothesis that children’s speech is organized over a wider temporal domain than adult speech. They further argued that such “syllabic” organization was in fact what Kent (1983) referred to as the “everything moves at once” principle. Conflicting theoretical approaches can be brought about by different empirical findings or by different interpretations of the same data, but in any case the existence of conflict-ing points of view demonstrates that the question of coarticulation development needs to be addressed.
The studies by Zharkova et al. (2008a) and Zharkova et al. (2008b), using an ultrasound-based measure of coarticulation, found greater mean values for the extent of coarticulation in the child participants, although there was variation across pairs of contrasting vowel environments. The present study develops our previous research, using a more straightforward principle for the measurement of extent of coarticulation but a more careful and elaborate normalization procedure for the effect of the smaller vocal tracts of children. Quantifying the size of any coarticulatory effect found is obviously desirable, given the interest in the question of whether children coarticulate more or less than adults, or equally so. Comparing distances between tongue contours across age groups raises problems of normalization. The distance among consonant curves taken from two different vowel environments is a function not just of how much the speaker coarticulated the consonant with the following vowel but also partly of vocal tract size. We attempted to normalize for this factor, and we present and discuss the patterns observed before and after the normalization. The theoretical approach developed by Nittrouer et al. (1989; 1996) led us to hypothesize that we would observe more coarticulation in children than in adults.
Token-to-Token Variation
A number of developmental studies of motor control have shown that children are less consistent in their articulations than adults. Within-speaker variability in speech movements have been demonstrated to decrease with age (e.g., acoustic studies: Kent & Forner 1980; Sharkey & Folkins 1985; Nittrouer 1993; Lee et
Coarticulation in Children: An Ultrasound Study 123
al. 1999; Nijland et al. 2002; Nittrouer et al. 2005; articulatory studies of lip and jaw displacement: Smith & Goffman 1998; Walsh & Smith 2002; Riely & Smith 2003; articulatory studies of tongue displacement: Zharkova et al. 2008a; 2008b). Parameters on which greater inconsistency in children has been reported include timing, velocity and amplitude of articulatory movements (see, e.g., Walsh et al. 2006, and references cited there). It has been shown that motor control maturation takes a very protracted course, with speakers in mid teens still being significantly different from adults (e.g., Smith & Zelaznik 2004). The study reported here tested the hypothesis that children aged between six and nine years old are less consistent than adults in their positioning of the tongue over multiple repetitions of the same speech segment and therefore that the extent of token-to-token variation in tongue curves would be greater in children than in adults. If this hypothesis was supported, it would tell us that speech production skills in children are not as consistent as those of adults, regardless of whether the units of speech production are syllable-sized or segment-sized. In their studies of vowel-on-consonant coarticulation, Nittrouer et al. did not address the question of within-speaker variability, so the results from this study allowed us to contribute new information to their model.
Summary of Hypotheses
1. We would observe a significant effect on /∫/ from three vowel pairs, in children and in adults; the effect of the /i/-/u/ vowel pair would be smaller than those of the other two vowel pairs.
2. The extent of coarticulation would be greater in children than in adults.
3. The extent of token-to-token variation in tongue curves would be greater in children than in adults.
Method
Ultrasound
Synchronized ultrasound and acoustic data were collected using the Queen Marga-ret University ultrasound system (Articulate Instruments 2007, 2008; Wrench and Scobbie 2008). Ultrasound offers a direct representation of tongue movements in speech. It is a safe and noninvasive articulatory technique, and it provides informa-tion about the shape of most of the midsagittal tongue contour, including the root (e.g., Stone 2005; Davidson 2007; Vazquez Alvarez & Hewlett 2007; Gick et al. 2008; Kocjančič 2008; Lawson et al. 2008; Zharkova & Hewlett 2009).
The advantage of ultrasound over acoustic analysis is that acoustic analy-sis provides only indirect evidence of articulatory movements. Sometimes, the acoustic signal is not even present; for example, during closure of voiceless stops. In children’s speech, acoustic analysis is particularly problematic, because of the high fundamental frequency and consequent difficulties with formant tracking (e.g., Buder 1996; Assmann & Katz 2000). Electromagnetic articulography and electropalatography, other articulatory techniques that have been used to study child speech production (e.g., Katz & Bharadwaj 2001; Nijland et al. 2004; Timmins et al. 2008), are relatively invasive, and neither of them provides information on
124 Zharkova, Hewlett, and Hardcastle
the full tongue contour. Figure 1 is an example of an ultrasound image of a child’s tongue. The lower edge of the bright white curve is the surface of the tongue. The tongue tip is on the right and the black area beyond it is caused by the bone of the chin. The black area in the lower left of the figure is caused by the hyoid bone.
The figure demonstrates a well-known drawback of ultrasound, namely that it does not image internal articulators other than the tongue (e.g., the palate or the pharyngeal walls). This makes it problematic to compare lingual articulations across speakers, because of the lack of reference points that would allow normalization for differences in vocal tract sizes and shapes. Various solutions for imaging other internal articulators and/or introducing external reference points have been proposed (cf. Gick 2002; Stone 2005; Gick et al. 2006). In the current study, we normalized for across-speaker differences by using an approximate measure of tongue length for each speaker as a correcting factor in quantifying the extent of coarticulation (see Extent of coarticulation below).
A challenge in ultrasound speech analysis is the difficulty in controlling the transducer position relative to the head, which is required for statistical analysis based on multiple repetitions. This challenge has been dealt with in a number of ways. For instance, special support systems have been designed and applied (e.g., Stone & Davis 1995; Gick et al. 2005). The system used in the current study includes headgear for stabilizing the position of the transducer with respect to the head. It is described in detail in Scobbie et al. (2008). A photograph of a child participant wearing a transducer-stabilizing headset is presented in Figure 2.
Figure 1 — Ultrasound frame at the middle of the consonant /∫/ from the word “shah” produced by a child speaker. Distance is in cm and the origin of the scale is at bottom left. The line of brightness in the ultrasound image represents the tissue-air interface at the tongue surface.
Coarticulation in Children: An Ultrasound Study 125
The ultrasound frame rate in the recordings was 30 Hz. While relatively low frame rates may be a challenge in speech research, this did not present problems for our study, because all the analyzed segments (i.e., the consonant /∫/ and vowels) have a sufficiently long period of relatively stable articulation, making it unlikely that a relevant gesture would be absent from the record.
Participants and Data Collection
The participants, all native speakers of Standard Scottish English, were ten normally developing children and ten adults. Our aim was to recruit primary school children, the target age being seven years old. Actual age range was three years seven months (3;7), the youngest child was 6;3, the oldest child was 9;9. Mean child age was 7;7, and standard deviation was 1;4. Ten adults were recruited (mean age 33 years old; standard deviation six years).
The stimuli were the syllables /∫i/, /∫u/ and /∫a/, in the carrier phrase “It’s a … Pam”. (Consonant-vowel syllables with the consonant /s/ and the same three vowels were also collected, but in this paper only the data on /∫/ are reported.) The target syllables were spelled as “she”, “shoe” and “shah”, respectively. The sentences were shown to the participants on the computer screen, accompanied by images
Figure 2 — A photograph of a child participant wearing a purpose-designed headset for stabilizing the ultrasound transducer in relation to the head. A head-mounted microphone is attached to the headset.
126 Zharkova, Hewlett, and Hardcastle
corresponding to the target words. Every target was repeated ten times. The order of presentation was randomized.
Comparison of Tongue Curves
Ultrasound frames at two time points, the midpoint of the consonant and the mid-point of the vowel, were identified in each of the different consonant-vowel (CV) sequences, based on the acoustic data. At each time point, a cubic spline was auto-matically (with subsequent manual correction) fitted to the tongue surface contour. Each curve was then defined in terms of a series of closely spaced x-y coordinates, and these coordinates were used for comparing tongue curves.
Quantification of the difference between two tongue curves was carried out using mean nearest neighbor distance (Zharkova & Hewlett 2009). Mean nearest neighbor distance is calculated by taking each point (defined by its x-y coordinates) on one curve and finding the Euclidean distance from it to the nearest point on the comparison curve, using the Pythagoras’ theorem, a2 + b2 = c2, where c is the distance in question. The mean nearest neighbor distance is the mean of all these distances between the two curves. Figure 3 illustrates a single nearest neighbor distance.
For quantification of the difference in tongue position between realizations of /∫/ in different vowel contexts, mean nearest neighbor distances between two sets of curves were calculated. Each curve from one set of curves was compared with each
Figure 3 — Illustration of a nearest neighbor distance. The point q on curve B is the nearest neighbor of the point p on curve A. The two curves in Figure 3 are taken from the vowel midpoint for two different vowels: /a/ (curve A) and /i/ (curve B).
Coarticulation in Children: An Ultrasound Study 127
curve, in turn, of the other set of curves. For example, distances were calculated between each of the ten curves of /∫/ in /∫a/ and each of the ten curves, in turn, of /∫/ in /∫i/, producing 100 distances in all. These values, referred to as across-set (AS) distances, were calculated separately for each participant, between /∫/ in /∫a/ and /∫/ in /∫i/ (∫a—∫i); between /∫/ in /∫a/ and /∫/ in /∫u/ (∫a—∫u); and between /∫/ in /∫i/ and /∫/ in /∫u/ (∫i—∫u). AS distances were also calculated for three pairs of contrasting vowel environments (/a/-/i/, /a/-/u/ and /i/-/u/), for each participant.
Mean nearest neighbor distances between different tokens of /∫/ in the same vowel environment were calculated (Zharkova & Hewlett 2009), separately for each vowel environment and in each participant. Mean nearest neighbor distances between different tokens of each of the three vowels were also calculated, separately for each vowel and in each participant. All these values, referred to as within-set (WS) distances, were used in the analysis, together with AS distances.
Coarticulatory Effects
Analyses were carried out to find out whether the change in tongue position at midconsonant determined by the vowel environment was significantly greater than the variation across ten repetitions of the consonant within the same vowel environ-ment. The method used for these calculations was described in Zharkova & Hewlett (2009), and it involves comparing AS distances and WS distances. If AS distances (for example, for ∫a—∫i) were significantly greater than both WS distances (in this example, for ∫a and for ∫i), then it was concluded that the two vowel environments produced a significant effect on the consonant.
ANOVAs were carried out separately for children and adults, and separately for each vowel pair. The independent variable was called “Distance Type”, and comprised of three distances: AS distances and two sets of WS distances. For example, in an ANOVA for ∫a-∫i in children, AS distances for ∫a-∫i were compared with WS distances for ∫a and WS distances for ∫i. For each speaker in every ANOVA, there were 100 AS distances, 45 WS distances for one vowel context, and 45 WS distances for the other vowel context. The three types of distances were pooled together across speakers, so in every test there were 1000 AS distances, 450 WS distances for one set of curves and 450 WS distances for the other set of curves. If the main effect of Distance Type was significant, and the Post-Hoc test showed that AS distances were significantly greater than both WS distances, it was concluded that the vowel pair produced a significant coarticulatory effect on the consonant. Tukey HSD Post Hoc tests were used when variances of the three distances were equal, and Games-Howell Post Hoc tests were used when variances were unequal. Bonferroni correction for six tests was applied.
Extent of Coarticulation
Comparing extent of coarticulation across participants was based on the AS dis-tances. Some method of normalization for size of vocal tract was required. The normalization we devised was based on a measurement of the length of the imaged tongue contour during the first repetition of /a/, in each participant. The vowel /a/ was chosen because the tongue tip is low during the production of this vowel, thus reducing the risk of loss of image from an air gap beneath the tongue tip and
128 Zharkova, Hewlett, and Hardcastle
because the larynx is also comparatively low and therefore the shadow of the hyoid bone is less likely to foreshorten the imaged contour at the back of the tongue. The calculation was done by summing the distances between all pairs of adjacent x-y points on the contour. It is important to emphasize the obvious assumptions involved: that length of tongue correlates with the overall size of the vocal tract and that the proportion of tongue imaged in the ultrasound display is similar across participants. While we believe that both assumptions are reasonable, we also acknowledge the approximate nature of the resulting normalization.
The first part of the procedure involved calculating tongue length, in mm, for all speakers. Adult 3 had the greatest length of imaged tongue surface. Tongue length values for all participants were expressed as a proportion of this length. Normalization of the AS distances by tongue length values was then carried out. The AS distances were divided by the tongue length value, on a participant-by-participant basis. These values will be referred to as AS distances normalized for tongue length (ASTL). ANOVAs were carried out, separately for AS values and for ASTL values, across age groups and vowel pairs. A Bonferroni correction for the two tests was applied.
Token-to-Token Variation
The WS distances for any given participant provide direct measures of variability in tongue placement for that participant. However, a comparison of WS distances between children and adults must take account of the fact that, other things being equal, greater WS distances would be expected between the tongue contours of the adults than those of the children, because of the tendency for adults to have larger vocal tracts. It was decided that if a nondirectional t test demonstrated that the children showed significantly greater raw WS distances than the adults, despite smaller tongue lengths, then the normalization for tongue length would not be applied; we would, in other words, be content to demonstrate greater variability on the part of the children, without attempting a quantification.
Results
Coarticulatory Effects
Mean AS distances between /∫/ in three pairs of vowel contexts for each speaker are presented in Table 1, along with the mean WS distances. The table shows that within each age group, mean AS distances for each vowel pair context are greater than mean WS distances. In Figure 4, example tongue contours for ∫a and ∫i (first row), ∫a and ∫u (second row), and ∫i and ∫u (third row) are displayed for child (left column) and adult (right column) speakers. Each plot in this and the following figure corresponds to a different speaker.
Results of six ANOVAs are presented in Table 2. The table contains F values, together with actual probability values for the main effect and for Post Hoc tests. In each analysis (i.e., in each cell of the table), three probability values need to be asterisked to conclude that a significant coarticulatory effect has taken place: a probability value for the main effect of Distance Type, and two probability values for the Post Hoc test: AS versus WS for one vowel context, and AS versus WS
Coarticulation in Children: An Ultrasound Study 129
Table 1 Mean across-set distances, in mm, for ∫a—∫i, ∫a—∫u and ∫i—∫u, for each participant; mean within-set distances, in mm, for each target segment and for each participant.
for the other vowel context. Table 2 shows that in ∫a-∫i and ∫a-∫u, both adults and children had a significant main effect of Distance Type, and the AS distance was significantly greater than both WS distances. We conclude that adults had significant vowel-on-consonant coarticulatory effects in ∫a-∫i and ∫a-∫u (at a 0.01 significance level, following Bonferroni correction), and that children also had significant vowel-on-consonant coarticulatory effects in ∫a-∫i and ∫a-∫u (at a significance level of 0.05, following Bonferroni correction). In ∫i-∫u, neither adults nor children had a significant coarticulatory effect. For children, the main effect of Type of Distance was not significant. For adults, the main effect was significant, and the difference between AS and WS in ∫u was significant. However, the difference between AS and
130 Zharkova, Hewlett, and Hardcastle
Figure 4 — Tongue contours for the consonant /∫/ in contrasting vowel environments. Each plot corresponds to a different speaker. First row: ∫a (solid lines) and ∫i (dashed-dotted lines); Child 2 on the left and Adult 3 on the right. Second row: ∫a (solid lines) and ∫u (dashed-dotted lines); Child 9 on the left and Adult 2 on the right. Third row: ∫i (solid lines) and ∫u (dashed-dotted lines); Child 1 on the left and Adult 5 on the right.
WS in ∫i did not reach significance. Therefore the criterion of achieving a significant difference from both WS distances was not met in this case.
Mean AS and WS distances for the three vowel environments are presented in Table 3. Example tongue contours for pairs of contrasting vowels can be found in Figure 5: /a/ and /i/ (first row), /a/ and /u/ (second row), and /i/ and /u/ (third row) are displayed for child (left column) and adult (right column) speakers. It is clear from Table 3 and Figure 5 that in both age groups, separation between sets of tongue contours for the vowels is noticeably greater than between sets of tongue contours for /∫/ across vowel environments.
Coarticulation in Children: An Ultrasound Study 131
Table 2 Results of the statistical testing for the presence of coarticulatory effects. “AS” is across-set distance; “WSa” is within-set distance in the context of /a/; “WSi” is within-set distance in the context of /i/; “WSu” is within-wet distance in the context of /u/. The table is based on the output from SPSS, so in cases where the output was “.000”, the table entry is “p < 0.001”. All p values that are less than 0.01 following Bonferroni correction have two asterisks against them. All p values that are less than 0.05 following Bonferroni correction have one asterisk against them.
∫a-∫i ∫a-∫u ∫i-∫u
Children F (2, 1897) = 13.424 F (2, 1897) = 9.709 F (2, 1897) = 2.013
p < .001 ** p < .001 ** p = .134
Tukey HSD Post Hoc Tukey HSD Post Hoc Tukey HSD Post Hoc
AS-WSa: p < .001 ** AS-WSa: p < .001 ** AS-WSi: p = .610
AS-WSi: p = .005 * AS-WSu: p = .004 * AS-WSu: p = .119
Adults F (2, 1897) = 265.399 F (2, 1897) = 160.778 F (2, 1897) = 4.795
p < .001 ** p < .001 ** p = .008 *
Games-Howell Post Hoc
Games-Howell Post Hoc
Tukey HSD Post Hoc
AS-WSa: p < .001 ** AS-WSa: p < .001 ** AS-WSi: p = .292
AS-WSi: p < .001 ** AS-WSu: p < .001 ** AS-WSu: p = .007 *
Tongue Length
The results of the tongue length calculations are presented in Table 4. Actual length, to the nearest mm, is given in brackets.
Extent of Coarticulation
Since no significant coarticulatory effects on /∫/ were found from the vowel pair /i/-/u/ in either children or adults, ∫i—∫u data were not included in calculations of the extent of coarticulation. Table 1 shows that mean AS distances (with all speakers in an age group pooled together) were greater in children than in adults, in all vowel contexts. A univariate ANOVA revealed a significant effect of age and a significant vowel pair effect. The interaction of vowel pair and age was also significant. The results of the ANOVA are presented in Table 5. The table contains F values and actual probability values. Thus, even without normalization, children showed a greater extent of coarticulation than adults.
A final estimation of extent of coarticulation was made by calculating ASTL values, i.e., dividing the AS values by the tongue length values given in Table 4, on a participant-by-participant basis. Every AS value for each participant, in each of the two vowel pair contexts, was divided by the relevant tongue length value (see Table 4). The calculations produced 100 ASTL values per speaker per vowel context. For the vowel pair ∫a—∫i, average ASTL values were 3.52 mm in children
132 Zharkova, Hewlett, and Hardcastle
Table 3 Mean across-set distances, in mm, for /a/—/i/, /a/—/u/ and /i/—/u/, for each participant; mean within-set distances, in mm, for each vowel and for each participant.
Speaker Across-set distances Within-set distancesa—i a—u i—u a i u
Child 1 8.88 8.57 2.44 2.87 2.41 2.41
Child 2 8.10 4.05 5.18 1.39 1.69 1.51
Child 3 8.90 8.86 1.74 2.35 1.75 1.70
Child 4 5.67 5.91 3.93 1.72 1.72 3.70
Child 5 8.30 4.17 5.30 1.56 3.32 2.24
Child 6 5.86 4.34 3.00 2.62 2.36 1.95
Child 7 8.57 5.94 3.40 0.94 2.31 1.83
Child 8 6.85 4.20 2.98 1.61 1.08 1.20
Child 9 7.90 4.97 4.02 1.63 1.27 1.02
Child 10 6.84 3.91 3.89 1.21 1.67 1.52
Mean 7.59 5.49 3.59 1.79 1.96 1.91
Adult 1 8.68 6.39 3.02 1.15 1.23 1.23
Adult 2 9.93 7.90 3.19 0.95 1.25 1.21
Adult 3 9.88 8.64 2.35 0.75 1.07 1.40
Adult 4 8.37 5.04 3.52 1.43 1.35 0.91
Adult 5 9.08 7.57 2.22 0.86 0.91 1.02
Adult 6 6.52 4.77 2.78 0.96 1.38 1.15
Adult 7 7.17 6.83 1.66 1.60 1.85 1.32
Adult 8 11.80 7.98 4.26 0.95 1.10 0.66
Adult 9 12.90 11.72 2.23 1.52 1.42 1.24
Adult 10 12.41 11.61 1.78 1.10 1.44 1.19
Mean 9.67 7.85 2.70 1.13 1.30 1.13
and 2.45 mm in adults. For the vowel pair ∫a—∫u, average ASTL values were 3.42 mm in children and 2.17 mm in adults.
Adult ASTL values were compared with child values, in a univariate ANOVA, with vowel pair being the other independent variable. The results of the ANOVA are presented in Table 5. A significant effect of age was found. As would be expected, the effect of normalization was to increase the difference between children and adults, with respect to extent of coarticulation. The vowel pair effect also remained significant. The interaction of vowel pair and age was not significant after Bonfer-roni correction.
Token-to-Token Variation
WS distances for the consonant can be found in Table 1, and those for the vowels are presented in Table 3. A nondirectional t test which compared all children’s WS
Coarticulation in Children: An Ultrasound Study 133
distances with all adults’ WS distances (2700 WS distance values in each group) demonstrated significantly greater WS distances in the child data. The average WS distance for adults was 1.18 mm, and the average WS distance for children was 1.82 mm. The difference was significant at p < .001. Because of the direction of the difference, it can be concluded, without the need to normalize for vocal tract size, that the children’s articulations were indeed more variable than the adults’.
Figure 5 — Tongue contours for the three vowels. Each plot corresponds to a different speaker. First row: /a/ (solid lines) and /i/ (dashed-dotted lines); Child 10 on the left and Adult 1 on the right. Second row: /a/ (solid lines) and /u/ (dashed-dotted lines); Child 7 on the left and Adult 4 on the right. Third row: /i/ (solid lines) and /u/ (dashed-dotted lines); Child 8 on the left and Adult 8 on the right.
134 Zharkova, Hewlett, and Hardcastle
Table 4 Tongue length in children and adults. Adult 3 had the greatest length of imaged tongue surface and the values for all participants are given as a proportion of this length. Actual length, to the nearest mm, is given in brackets.
Speaker number Children Adults
1 0.60 (61 mm) 0.74 (75 mm)
2 0.59 (60 mm) 0.87 (88 mm)
3 0.57 (57 mm) 1.00 (101 mm)
4 0.60 (60 mm) 0.71 (71 mm)
5 0.50 (50 mm) 0.93 (93 mm)
6 0.47 (47 mm) 0.62 (62 mm)
7 0.55 (55 mm) 0.55 (56 mm)
8 0.61 (62 mm) 0.78 (79 mm)
9 0.64 (64 mm) 0.78 (78 mm)
10 0.51 (51 mm) 0.97 (97 mm)
Mean 0.56 (57 mm) 0.80 (80 mm)
Table 5 Results of the statistical testing for the extent of coarticulatory effects. The second column presents the results of the ANOVA on across-set (AS) distance values. The third column presents the results of the ANOVA on across-set distance values corrected for tongue length (ASTL). As in table 1, in cases where the SPSS output was “.000”, the table entry is “p < 0.001”. All p values that are less than 0.01 following Bonferroni correction have two asterisks against them.
AS analysis ASTL analysis
Age F (1, 3996) = 33.962 p < .001 ** F (1, 3996) = 652.826 p < .001 **
Vowel F (1, 3996) = 23.338 p < .001 ** F (1, 3996) = 16.690 p < .001 **
Age x vowel pair F (1, 3996) = 9.471 p = .002 ** F (1, 3996) = 3.928 p = .048
DiscussionTo a large extent, the hypotheses concerning the pattern of coarticulatory effects on /∫/ were supported. The children exhibited coarticulation in the same vowel contexts as adults: they showed an effect on the tongue position of /∫/ according to whether the following vowel was /a/, on the one hand, or either /i/ or /u/ on the other. This is a direct and robust demonstration, using articulatory data, of the presence of vowel-on-consonant coarticulation in the speech of six to nine year old children. The results did however fail to show the expected effect on the tongue position of
Coarticulation in Children: An Ultrasound Study 135
/∫/ from the following vowel being /i/ as opposed to /u/, in either the adults or the children. So far as the articulation of these vowels themselves was concerned, the expected pattern emerged, namely that their tongue positions were clearly differ-ent, though much less so than the differences between /i/ and /a/ and between /a/ and /u/, as shown in Table 3. However, this difference between /i/ and /u/ seems to have been insufficient to exert a consistent influence on the tongue position of the preceding /∫/. This was true of both children and adults, though it is perhaps worth pointing out that the adult group came closer to showing an effect, since their AS distances for ∫i—∫u did just reach significance when matched against the WS distances for ∫u, before Bonferroni correction. In traditional terminology, the results might be described as revealing two allophones of /∫/, one that occurs before /a/ and another that occurs before /i/ and /u/. However, this characterization would place the difference, probably wrongly, in the realm of phonology rather than phonetics. The difference is nothing like absolute but more a matter of statistical probability. This in turn suggests that the phenomenon is more likely to originate within the motor control system for the implementation of the speech code than it is to originate from any distinction in the speech code itself.
The hypothesis about more coarticulation in children was supported. Our results suggest that in six to nine year old children the consonant adapts to the vowel more than in adults. This provides further evidence in favor of Nittrouer’s idea that children produce a CV syllable as a syllable-sized unit. Our previous studies (Zharkova et al. 2008a; 2008b) showed that mean values for the extent of coarticulation were significantly greater in children, although there was variation across vowel contexts. In those studies we did not normalize for size of vocal tract using a direct measurement of tongue length. We therefore regard the current study as providing a more accurate comparison of extent of coarticulation in children and adults. The application of normalization to speech data, however, is problematic. In particular, the vocal tract is a complex space and its growth is not merely a matter of an increase in size. Uncertainty concerning the precision of the normalization process for vocal tract size is a reason for exercising caution in the interpretation of this aspect of our results.
The children performed significantly differently from adults on the measure of variability. The hypothesis that the children’s tongue positions would be more variable than those of the adults was unambiguously supported. We were fortunate that this greater variability on the part of the children was demonstrable without the need for normalization (factoring in the tongue length values from Table 4 would merely have reduced the probability value further); it also testifies to the strength of the difference between child and adult on this measure. Reduction in unconditioned variation is a well-established feature of increasing maturity of speech motor control and this result suggests that children under ten are unlikely to approach adult-like capability.
However it should be noted that most of the children in the current study were within the age range where important developmental changes in the oral cavity are taking place. A period of rapid growth of vocal tract length and the lower face has been documented between seven and ten years old (Temple et al. 2002). Inconsistency in lingual articulations reported in this study could be partly ascribed to articulatory adjustments due to rapid physiological changes, and not only to immature speech motor control. The interrelationship of vocal tract growth and
136 Zharkova, Hewlett, and Hardcastle
development of speech motor control will undoubtedly prove to be complex. Smith (2010) claims that adolescents also demonstrate differences from adult speech motor control. She suggests that adolescents achieve faster speech rates than younger children at the expense of smaller displacement of lips and jaw, younger children having relatively larger articulator displacements accompanied by slow speech rates. It would be interesting to have results from a single study of tongue control in children, adolescents and adults but we do not know of any such study to date.
Our data are “static”, in the sense that we only compared /∫/ tongue contours at a single time point (midconsonant). This has the virtue of simplicity and is per-fectly valid for the first two of the three aspects of child and adult pronunciations of CV syllables that were addressed in this study. Unfortunately, however, it raises a further complication in relation to the comparison of extent of coarticulation. Children tend to have slower speech rates than adults, the implication being that their speech segments tend to be longer. Therefore at the point at which the /∫/ tongue contour is sampled, the conditioning factor for coarticulation, namely the tongue position of the following vowel, is probably further away in time in the case of the children than in the adults. It is unlikely that any simple correction for this would be valid. Some evidence from other studies suggests that tongue dynamics can differ between children and adults. For example, Katz & Bharadwaj (2001), in an articulatory study of CV syllables using electromagnetic articulography, reported dynamic coarticulatory patterns in five to seven year old children and adults. Children demonstrated earlier onset of lingual anticipatory coarticulation for /sV/ syllables, but similar patterns to adults for /∫V/ syllables.
An analysis of coarticulatory changes in a CV syllable over time would be an alternative way of comparing child and adult productions. If control over timing in CV syllables is not developed fully in children, we might observe earlier (or later) onset and offset of the consonant-vowel transition compared with those of adults. Measures of the dynamics of tongue movement over CV syllables, such as the tongue trajectory and the rate of tongue displacement, would need to be analyzed to address this issue. We are planning to undertake an ultrasound analysis of dynamic lingual coarticulation in /sV/ and /∫V/ syllables produced by children and adults, using such dynamic measures. We would predict that the results of such a study would support the present findings with respect to /∫/. Finally, it is worth mention-ing that past reports showing less coarticulation in children than in adults analyzed consonant-on-vowel coarticulatory effects (e.g., Thompson & Hixon 1979; Kent 1983), while most of the other literature was concerned with vowel-on-consonant coarticulation. In the future, both vowel-on-consonant and consonant-on-vowel influence need to be addressed, preferably using the same methodology.
Acknowledgments
We are grateful to Alan Wrench for his technical and other advice, to Stephen Cowen for assistance with instrumentation, and to Robert Rush for his crucial help with the statistical analysis. We also wish to thank the editor and two anonymous reviewers for helpful com-ments. This work was supported by an ESRC research grant RES-000-22-2833.
Coarticulation in Children: An Ultrasound Study 137
ReferencesArticulate Instruments Ltd (2007). Articulate Assistant Advanced User Guide: Version 2.07.
sion 1.13. Edinburgh, UK: Articulate Instruments Ltd.Assmann, P.F., & Katz, W.F. (2000). Time-varying spectral change in the vowels of children
and adults. The Journal of the Acoustical Society of America, 108, 1856–1866.Bell-Berti, F., & Harris, K.S. (1974). More on the motor organization of speech gestures.
Haskins Laboratories Status Report on Speech Research, SR-37(38), 73–77.Browman, C.P., & Goldstein, L. (1992). Articulatory phonology: an overview. Phonetica,
measures from child speech. In B. Bernhardt, J. Gilbert, & D. Ingram (Eds), Proceed-ings of the UBC International Conference on Phonological Acquisition. Somerville: Cascadilla Press. Pp. 254-265.
Davidson, L. (2007). Coarticulation in contrasting Russian stop sequences. Proceedings of ICPhS 2007, Saarbrücken, 6-10 August 2007. Pp. 417-420.
Engstrand, O. (1988). Articulatory correlates of stress and speaking rate in Swedish VCV utterances. The Journal of the Acoustical Society of America, 83, 1863–1875.
Engstrand, O., McAllister, R., & Lindblom, B. (1996). Investigating the “trough”: vowel dynamics and aerodynamics. The Journal of the Acoustical Society of America, 100, 2659–2660.
Farnetani, E., & Recasens, D. (1993). Anticipatory consonant-to-vowel coarticulation in the production of VCV sequences in Italian. Language and Speech, 36, 279–302.
Fowler, C.A., & Brancazio, L. (2000). Coarticulation resistance of American consonants and its effects on transconsonantal vowel-to-vowel coarticulation. Language and Speech, 43, 1–41.
Fuchs, S., Hoole, P., Brunner, J., & Inoue, M. (2004). The trough effect – an aerodynamic phenomenon? In J. Slifka, S. Manuel & M. Matthies (Eds), Proceedings of the Confer-ence “From Sound to Sense: 50+ Years of Discoveries in Speech Communication”. Massachusetts Institute of Technology. Pp. C-25–C-30.
Gay, T. (1974). Some electromyographic measures of coarticulation in VCV utterances. Haskins Laboratories Status Report on Speech Research, SR-44, 137–145.
Gay, T., Ushijima, T., Hirose, H., & Cooper, F.S. (1974). Effect of speaking rate on labial consonant-vowel articulation. Journal of Phonetics, 2, 46–63.
Gibbon, F., & Nikolaidis, K. (1999). Palatography. In W. Hardcastle & N. Hewlett (Eds.), Coarticulation: Theory, Data and Techniques (pp. 229–245). Cambridge: Cambridge University Press.
Gick, B. (2002). The use of ultrasound for linguistic phonetic fieldwork. Journal of the International Phonetic Association, 32, 113–122.
Gick, B., Bacsfalvi, P., Bernhardt, B.M., Oh, S., Stolar, S., & Wilson, I. (2008). A motor dif-ferentiation model for liquid substitutions: English /r/ variants in normal and disordered acquisition. Proceedings of Meetings on Acoustics, 1(060003), 1–9.
Gick, B., Bird, S., & Wilson, I. (2005). Techniques for field application of lingual ultrasound imaging. Clinical Linguistics & Phonetics, 19, 503–514.
Gick, B., Campbell, F., Oh, S., & Tamburri-Watt, L. (2006). Toward universals in the gestural organization of syllables: a cross-linguistic study of liquids. Journal of Phonetics, 34, 49–72.
138 Zharkova, Hewlett, and Hardcastle
Glazier, P.S., Davids, K. & Bartlett, R.M. (2003). Dynamical systems theory: a relevant frame-work for performance-oriented sports biomechanics research. Sportscience, Feb, 1-8.
Guenther, F.H. (1995). A modelling framework for speech motor development and kinematic articulator control. Proceedings of the XIIIth International Congress of Phonetic Sci-ences, vol. 2. Stockholm, Sweden: KTH and Stockholm University. Pp. 92-99.
Hardcastle, W., & Hewlett, N. (Eds.). (1999). Coarticulation: Theory, Data and Techniques. Cambridge: Cambridge University Press.
Hardcastle, W., & Tjaden, K. (2008). Coarticulation and speech impairment. In M.J. Ball, M.R. Perkins, N. Müller, & S. Howard (Eds.), The Handbook of Clinical Linguistics (pp. 506–524). Oxford: Blackwell.
Kandel, E.R., Schwartz, J.H., & Jessell, T.M. (Eds.). (2000). Principles of Neural Science (4th ed.). New York, London: McGraw-Hill.
Katz, W.F., & Bharadwaj, S. (2001). Coarticulation in fricative-vowel syllables produced by children and adults: a preliminary report. Clinical Linguistics & Phonetics, 15, 139–143.
Katz, W.F., Kripke, C., & Tallal, P. (1991). Anticipatory coarticulation in the speech of adults and young children: acoustic, perceptual, and video data. Journal of Speech and Hearing Research, 34, 1222–1232.
Keating, P.A. (1988). Underspecification in phonetics. Phonology, 5(2), 275–292.Kelso, J.A.S. (1995). Dynamic Patterns: The Self-Organization of Brain and Behaviour.
Cambridge, London: MIT Press.Kent, R.D. (1983). The segmental organization of speech. In P.F. MacNeilage (Ed.), The
Production of Speech (pp. 57–89). New York: Springer-Verlag.Kent, R.D., & Forner, L.L. (1980). Speech segment durations in sentence recitations by
children and adults. Journal of Phonetics, 8, 157–168.Kocjančič, T. (2008). Ultrasound investigation of tongue movements in syllables with dif-
ferent onset structure. In R. Sock, S. Fuchs & Y. Laprie (Eds), Proceedings of the 8th International Seminar on Speech Production 2008, Strasbourg, France, 8-12 December 2008. Pp. 237-240.
Kühnert, B., & Nolan, F. (1999). The origin of coarticulation. In W. Hardcastle & N. Hewlett (Eds.), Coarticulation: Theory, Data and Techniques (pp. 7–30). Cambridge: Cambridge University Press.
Lawson, E., Stuart-Smith, J., & Scobbie, J.M. (2008). Articulatory insights into language variation and change: preliminary findings from an ultrasound study of derhoticization in Scottish English. In K. Gorman (Ed.), University of Pennsylvania Working Papers in Linguistics, 14.2: Papers from New Ways of Analyzing Variation 36, 102-110.
Lee, S., Potamianos, A., & Narayanan, S. (1999). Acoustics of children’s speech: devel-opmental changes of temporal and spectral parameters. The Journal of the Acoustical Society of America, 105, 1455–1468.
Lindblom, B., Sussman, H.M., Modarresi, G., & Burlingame, E. (2002). The trough effect: implications for speech motor programming. Phonetica, 59, 245–262.
Nijland, L., Maassen, B., Hulstijn, W. & Peters, H. (2004). Speech motor coordination in Dutch-speaking children with DAS studied with EMMA. Journal of Multilingual Communication Disorders, 2, 50–60.
Nijland, L., Maassen, B., Van der Meulen, S., Gabrieëls, F., Kraaimaat, F.W., & Schreuder, R. (2002). Coarticulation patterns in children with developmental apraxia of speech. Clinical Linguistics & Phonetics, 16, 461–483.
Nijland, L., Maassen, B., Van der Meulen, S., Gabrieëls, F., Kraaimaat, F.W., & Schreuder, R. (2003a). Planning of syllables in children with developmental apraxia of speech. Clinical Linguistics & Phonetics, 17, 1–24.
Nijland, L., Maassen, B., & Van der Meulen, S. (2003b). Evidence of motor programming deficits in children diagnosed with DAS. Journal of Speech, Language, and Hearing Research: JSLHR, 46, 437–450.
Coarticulation in Children: An Ultrasound Study 139
Nittrouer, S. (1993). The emergence of mature gestural patterns is not uniform: evidence from an acoustic study. Journal of Speech and Hearing Research, 36, 959–972.
Nittrouer, S., Estee, S., Lowenstein, J.H., & Smith, J. (2005). The emergence of mature gestural patterns in the production of voiceless and voiced word-final stops. The Journal of the Acoustical Society of America, 117, 351–364.
Nittrouer, S., Studdert-Kennedy, M., & McGowan, R.S. (1989). The emergence of phonetic segments: evidence from the spectral structure of fricative-vowel syllables spoken by children and adults. Journal of Speech and Hearing Research, 32, 120–132.
Nittrouer, S., Studdert-Kennedy, M., & Neely, S.T. (1996). How children learn to organize their speech gestures: further evidence from fricative-vowel syllables. Journal of Speech and Hearing Research, 39, 379–389.
Öhman, S.E.G. (1967). Numerical model of coarticulation. The Journal of the Acoustical Society of America, 41, 310–319.
Perrier, P., & Ma, L. (2008). Speech planning for V1CV2 sequences: influence of the planned sequence. In R. Sock, S. Fuchs & Y. Laprie (Eds), Proceedings of the 8th International Seminar on Speech Production 2008, Strasbourg, France, 8-12 December 2008. Pp. 69-72.
Recasens, D. (1999). Lingual coarticulation. In W. Hardcastle & N. Hewlett (Eds.), Coarticulation: Theory, Data and Techniques (pp. 80–104). Cambridge: Cambridge University Press.
Recasens, D. & Espinosa, A. (2009). An articulatory investigation of lingual coarticulatory resistance and aggressiveness for consonants and vowels in Catalan. Journal of the Acoustical Society of America, 125, 2288-2298.
Recasens, D., & Pallarès, M.D. (2001). Coarticulation, assimilation and blending in Catalan consonant clusters. Journal of Phonetics, 29, 273–301.
Recasens, D., Pallarès, M.D., & Fontdevila, J. (1997). A model of lingual coarticulation based on articulatory constraints. The Journal of the Acoustical Society of America, 102, 544–561.
Riely, R.R., & Smith, A. (2003). Speech movements do not scale by orofacial structure size. Journal of Applied Physiology, 94, 2119–2126.
Saltzman, E.L. (1991). The task dynamic model in speech production. In H.F.M. Peters, W. Hulstijn, & C.W. Starkweather (Eds.), Speech Motor Control and Stuttering (pp. 37–52). Amsterdam: Elsevier Science Publishers.
Saltzman, E., & Byrd, D. (2000). Task-dynamics of gestural timing: phase windows and multifrequency rhythms. Human Movement Science, 19, 499–526.
Saltzman, E., & Kelso, J.A.S. (1987). Skilled actions: a task dynamic approach. Psychologi-cal Review, 94, 84–106.
Saltzman, E.L., & Munhall, K.G. (1989). A dynamical approach to gestural patterning in speech production. Ecological Psychology, 1, 333–382.
Scobbie, J., Wrench, A., & van der Linden, M. (2008). Head-probe stabilisation in ultrasound tongue imaging using a headset to permit natural head movement. In R. Sock, S. Fuchs & Y. Laprie (Eds), Proceedings of the 8th International Seminar on Speech Production 2008, Strasbourg, France, 8-12 December 2008. Pp. 373-376.
Sereno, J.A., Baum, A.R., Cameron Marean, G., & Lieberman, P. (1987). Acoustic analyses and perceptual data on anticipatory labial coarticulation in adults and children. The Journal of the Acoustical Society of America, 81, 512–519.
Sereno, J.A., & Lieberman, P. (1987). Developmental aspects of lingual coarticulation. Journal of Phonetics, 15, 247–257.
Sharkey, S., & Folkins, J. (1985). Variability of lip and jaw movements in children and adults: implications for the development of speech motor control. Journal of Speech and Hearing Research, 28, 3–15.
140 Zharkova, Hewlett, and Hardcastle
Siren, K.A., & Wilcox, K.A. (1995). Effects of lexical meaning and practiced productions on coarticulation in children’s and adults’ speech. Journal of Speech and Hearing Research, 38, 351–359.
Smith, A. (2010). Development of neural control of orofacial movements for speech. In W.J. Hardcastle, J. Laver, & F.E. Gibbon (Eds.), The Handbook of Phonetic Sciences (2nd ed., pp. 251–296).
Smith, A., & Goffman, L. (1998). Stability and patterning of speech movement sequences in children and adults. Journal of Speech and Hearing Research, 41, 18–30.
Smith, A., & Zelaznik, H.N. (2004). Development of functional synergies for speech motor coordination in childhood and adolescence. Developmental Psychobiology, 45, 22–33.
Stone, M. (2005). A guide to analyzing tongue motion from ultrasound images. Clinical Linguistics & Phonetics, 19, 455–502.
Stone, M., & Davis, E.P. (1995). A head and transducer support system for making ultrasound images of tongue/jaw movement. The Journal of the Acoustical Society of America, 6, 3107–3112.
Temple, E.C., Hutchinson, I., Laing, D.G. & Jinks, A.L. (2002). Taste development: dif-ferential growth rates of tongue regions in humans. Brain Research. Developmental Brain Research, 135, 65–70.
Thompson, A.E., & Hixon, T.J. (1979). Nasal air flow during normal speech production. The Cleft Palate Journal, 16, 412–420.
Timmins, C., Hardcastle, W., McCann, J., Wood, S., & Wishart, J. (2008).Coarticulation in children with Down’s syndrome: an electropalatographic analysis. In R. Sock, S. Fuchs & Y. Laprie (Eds), Proceedings of the 8th International Seminar on Speech Production 2008, Strasbourg, France, 8-12 December 2008. Pp. 273-276.
Vazquez Alvarez, Y., & Hewlett, N. (2007). The trough effect: an ultrasound study. Pho-netica, 65, 105–121.
Walsh, B., & Smith, A. (2002). Articulatory movements in adolescents: evidence for pro-tracted development of speech motor control processes. Journal of Speech, Language, and Hearing Research: JSLHR, 45, 1119–1133.
Walsh, B., Smith, A., & Weber-Fox, C. (2006). Short-term plasticity in children’s speech motor systems. Developmental Psychobiology, 48, 660–674.
Wodzinski, S., & Frisch, S. (2006). Ultrasound study of velar-vowel coarticulation. The Journal of the Acoustical Society of America, 120, 3373–3374.
Wrench, A., & Scobbie, J. (2008). High-speed cineloop ultrasound vs. video ultrasound tongue imaging: comparison of front and back lingual gesture location and relative timing. In R. Sock, S. Fuchs & Y. Laprie (Eds), Proceedings of the 8th International Semi-nar on Speech Production 2008, Strasbourg, France, 8-12 December 2008. Pp. 57-60.
Zharkova, N. (2007). Quantification of coarticulatory effects in several Scottish English phonemes using ultrasound. QMU Speech Science Research Centre Working Papers, WP-13, 1-19.
Zharkova, N., & Hewlett, N. (2009). Measuring lingual coarticulation from midsagittal tongue contours: description and example calculations using English /t / and /�/. Journal of Phonetics, 37, 248–256.
Zharkova, N., Hewlett, N., & Hardcastle, W. (2008a). Analysing coarticulation in Scottish English children and adults: an ultrasound study. Canadian Acoustics, 36. Proceedings of the Acoustics Week in Canada, 2008, 158–159.
Zharkova, N., Hewlett, N., & Hardcastle, W. (2008b). An ultrasound study of lingual coar-ticulation in children and adults. In R. Sock, S. Fuchs & Y. Laprie (Eds), Proceedings of the 8th International Seminar on Speech Production 2008, Strasbourg, France, 8-12 December 2008. Pp. 161-164.
Ziegler, W., & Maassen, B. (2007). The role of the syllable in disorders of spoken language communication. In B. Maassen, R.D. Kent, H.F.M. Peters, P. van Lieshout, & W. Hul-stijn (Eds.), Speech Motor Control in Normal and Disordered Speech (pp. 415–448). Oxford: Oxford University Press.
![Nasal Aerodynamics and Coarticulation in Bininj Kunwok ......keep phonemes phonologically contrastive coarticulation needs to be limited [4], [5]. In Australian languages, due to the](https://static.documents.pub/doc/80x56/603ea53cb704da0962121412/nasal-aerodynamics-and-coarticulation-in-bininj-kunwok-keep-phonemes-phonologically.jpg)
