INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

SPEECH PRODUCTION IS a complex motor activity that involves the coordination of the respiratory, laryngeal, and articulatory subsystems (4). The respiratory system provides the aerodynamic "power supply" for speech, driving the vibration of the vocal folds in the larynx, the voice source for speech production. The focus of the present investigation is the articulatory subsystem (the supralaryngeal airway), which acts as a dynamic acoustic filter on the energy supplied by the respiratory and laryngeal subsystems. Articulation refers to movement of the orofacial structures, such as the lips, tongue, and jaw (the articulators). Movements of the articulators dynamically alter the spatial configuration of the vocal tract. These dynamic changes in the configuration of the vocal tract cause changes in its acoustic resonant properties, and these modifications in vocal tract resonant properties result in the production of the various speech sounds of a language. In addition to its role as an acoustic filter, the articulatory subsystem also serves as a sound source for speech sounds that involve constrictions of the upper airway, for example, the frication noise in the "sh" segment or the burst that occurs when the lips are released to produce a labial consonant such as "p." Thus the goal of articulation in speech production is to reach varying vocal tract shapes to achieve specific speech sound goals.

In the speech production process, therefore, precise spatial and temporal control of articulator motion is necessary. The fine, rapid movements of the articulators are accomplished by the contraction of many orofacial muscles. How these fine movements are controlled and coordinated to produce the rapidly changing vocal tract configurations needed for speech production is not well understood. Furthermore, the changes in motor control processes that occur with development have not been explored in detail. Researchers studying the principles of speech movement control have considered a number of hypotheses derived from the general motor control literature, including central control of movement patterns and interarticulator coordination, as well as the role of sensory feedback (22).

One putative, simple principle of speech movement control has received very little attention: the principle of movement scaling according to the general size of the orofacial structures. The operation of such a principle might be particularly relevant in explaining changes in speech motor control that occur over childhood and adolescence. It is reasonable to assume, for example, that children move their articulators smaller distances than adults because of their smaller orofacial structure sizes. However, it is common practice in speech motor development research to report measurements of speech kinematic data, such as displacement and velocity, without taking the participant's orofacial structure size into consideration (16, 20, 27, 32). Direct kinematic investigations examining speech motor development generally have involved very few subjects. In addition, various articulators have been studied by using a variety of speech samples. As might be expected, the results of these studies are mixed. Therefore, no clear conclusion may be drawn concerning the question of whether speakers with smaller orofacial structures, e.g., young children and women, produce smaller speech movements (6, 16, 20, 23, 27, 32).

Studies of the control of motor behaviors such as locomotion, handwriting, and speech often reveal developmental trends of decreasing variability in both temporal and spatial parameters of movement and increasing velocity of movement with maturation (6, 8, 9, 14, 16, 18, 20, 23, 24, 27, 29, 31). The development of locomotion is also characterized by a physical scaling principle evidenced by a linear increase in step length as limb length increases (2, 18, 29, 30, 33). This scaling principle indicates that biomechanical factors, such as structure size, directly influence kinematic parameters in locomotion. Conversely, handwriting does not show a size-scaling principle in development. In fact, an investigation examining the development of the control of handwriting in a large number of children revealed that movement amplitude decreased with maturation (8). These results indicate that maturation of underlying neural processes may have a greater effect on kinematic parameters of handwriting and other fine motor behaviors (19) than biomechanical factors.

As noted above, earlier studies of speech kinematic parameters and their potential relationship to orofacial structure size provide conflicting results. For example, Kuehn and Moll (15) reported a positive relationship between orofacial structure size, displacement, and velocity in adults, such that individuals with larger oral structure sizes exhibited greater displacements and velocities than individuals with smaller structures. In contrast, Smith and McLean-Muse (28) reported that young children and adults had similar peak velocities and articulatory displacement. The goal of the present study is to determine in a larger sample of participants whether speech movement amplitude and velocity are scaled to orofacial structure size. Thus we compare young children with adults, and, within an age group, we assess whether the larger individuals of the group make larger oral movements for speech.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Subjects. Thirty 5-yr-old children and thirty adults between the ages of 20 and 22 yr (mean age = 21 yr) participated in this study. In both age groups, 15 of the participants were male and 15 were female. All participants spoke English as their first and primary language. All participants had no history of speech, language, or hearing disorders or neurological problems as determined by parent or self report. These subjects participated in data collection for multiple experimental protocols, as part of a larger project, that were not analyzed in the present investigation. Data from the present subjects are included in other papers that will be submitted in parallel with the present report.¹

Screening and speech protocol. Each participant passed age-appropriate speech/language screening tests. All participants passed a pure tone hearing screening at 20 dB hearing level at 500, 1,000, 2,000, 4,000, and 6,000 Hz for each ear. All participants scored within normal range on the Oral Speech Mechanism Screening Examination-Revised, which was administered to assess the integrity of the anatomical structures and physiological functioning of the oral mechanism.

Kinematic data collection. The Optotrak 3020 motion measurement system was used to obtain the kinematic data. The Optotrak system tracks the three-dimensional movement of plastic markers containing infrared-emitting diodes with a spatial resolution of 0.1 mm. The session was audio taped with a digital audio tape recorder. The session was also videotaped.

View larger version (95K): [in this window] [in a new window]   Fig. 1.   Child wearing the modified sport goggles and markers for oral movement tracking.

Anthropometric measurements. Anthropometric measures of the orofacial structures were made according to the procedures described by Farkas (5), and participants' heights and weights were recorded. A spreading caliper was used for measurements projecting linear distances between landmarks on different planes. A sliding caliper was used for measurements of linear projective distances between landmarks on the same plane. A soft measuring tape was used for measurements of tangential linear distances taken along the skin surface.

Data analysis. To remove head-motion artifacts, movements of the lower lip marker were referenced to the head coordinate system for each participant by using software provided with the Optotrak system. The lower lip superior/inferior displacement signals were then imported into Matlab software (Mathworks) and analyzed with a custom program. A Butterworth filter was used to digitally low-pass filter the signals in both forward and backward directions at 10 Hz. Velocity of the lower lip motion was computed from the filtered displacement records by using a three-point difference technique.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Reliability of measurements. For the 60 participants, there was a total of 660 anthropometric measurements. Of these 660 measurements, 18% were measured a third time.

Anthropometric measurements. Table 1 includes the means and standard deviations of five anthropometric measurements for both groups. It is apparent that children's orofacial structures and heights are significantly smaller than the adults'. Within the adult group, men are larger than women, but the 5-yr-old boys and girls have approximately equal values. On average, the four orofacial measurements in children were 83% of the adult measurements, in contrast to height, which was 67% of the adult value. A repeated-measures ANOVA was computed to determine effects of age and sex on the anthropometric measures. There was a significant effect of age [F(1,⁵⁶) = 382.6, P < 0.0001], a significant effect of sex [F(1,⁵⁶) = 26.71, P < 0.0001], and an interaction of age × sex [F(1,⁵⁶) = 7.63, P = 0.008].

Whole utterance measurements: duration, amplitude, and velocity. Presented in Table 2 are the means and standard deviations of the overall durations for the entire lower lip movement sequence for BBAP and MBPP. Children's total movement times were ~130% of the adult values. A repeated-measures ANOVA on total movement sequence duration indicated a highly significant effect of age [F(1,⁵⁶) = 63.91, P < 0.0001], but no effect of sex [F(1,⁵⁶) = 1.94, P = 0.17], and no interaction of age × sex [F(1,⁵⁶) = 1.56, P = 0.22].

View larger version (23K): [in this window] [in a new window]   Fig. 2.   Displacement (top) and velocity (bottom) trajectories for a typical adult speaker for the two sentences. The words are lined up approximately with the lower lip/jaw movements that would occur during their production. For both sentences, the first and last opening movements are truncated (e.g., the left plot starts with the first opening movement for "Buy" in midmovement) because the peak velocities of the first and last movement for the sentences were used to segment the data for analysis.

View larger version (23K): [in this window] [in a new window]   Fig. 3.   Displacement (top) and velocity (bottom) trajectories for a typical 5-yr-old subject for the two sentences. The words are lined up approximately with the lower lip/jaw movements that would occur during their production. For both sentences, the first and last opening movements are truncated (e.g., the left plot starts with the first opening movement for "Buy" in midmovement) because the peak velocities of the first and last movement for the sentences were used to segment the data for analysis.

Single movement measurements: amplitude and velocity. Plotted in Fig. 4 are the mean displacements of each of the four opening movements for the adults and 5-yr-old subjects. Clearly, there is a trend for smaller displacement for the 5-yr-old subjects in all four opening movements. However, this trend was not significant [F(1,⁵⁶) = 3.05, P = 0.09]. There was no effect of sex [F(1,⁵⁶) < 1] and no interaction of age × sex [F(1,⁵⁶) < 1]. A significant effect of syllable was noted [F(3,¹⁶⁸) = 94.9, P < 0.0001], as well as an interaction of syllable × age [F(3,¹⁶⁸) = 2.83, P = 0.04], indicating that the modulation of amplitude across the four opening movements was different between age groups.

View larger version (9K): [in this window] [in a new window]   Fig. 4.   Amplitude (means ± SE) for adults and children for the 4 single movements selected for analysis. Differences between groups were not significant (see RESULTS).

View larger version (9K): [in this window] [in a new window]   Fig. 5.   Peak velocities (means ± SE) for adults and children for the 4 single movements selected for analysis. These are oral opening movements so that velocity is negative, but for ease of comparison of the 2 groups, they are plotted as positive values. Adults clearly produce these speech movements with a much higher velocity than the young children.

Relationship between anthropometric measures and kinematic measures. For the children and adults, correlations between the displacement and velocity dynamic range measures and the anthropometric measures all failed to reach significance within each group. The correlations for the adults ranged from 0 to 0.33 and for the children ranged from 0 to 0.35. A correlation of 0.37 was required to reach significance at P = 0.05.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The results of this experiment clearly suggest that speech movement amplitude does not follow a size-scaling principle and that speech motor control is, therefore, more similar to handwriting than to locomotion (2, 8, 18). We examined this issue by using both between-group and within-group analyses. As expected, the between-group analyses indicate that young children's orofacial structures were smaller than the adult structures. If a scaling principle is applied to speech production, adults would move these larger structures significantly farther than children would. However, there was no evidence to support a size-scaling principle of these kinematic parameters according to the size of the effectors. Statistically, the children moved their articulators the same distances as adults. However, the 5-yr-old subjects made these movements at significantly lower velocities than adults. Thus, on the basis of between-group statistics, the children appear to have a large-amplitude, low-velocity speech movement strategy compared with young adults.

The between-group analyses also revealed a significant age × sex interaction for the anthropometric measures. The orofacial structures of adult women are on average smaller than those of adult men, whereas 5-yr-old boys and girls are equal in orofacial structure size. There were, however, no significant age × sex interactions in any of the kinematic measures. This was true for both the whole utterance measures (duration, displacement dynamic range, velocity dynamic range) and the single-movement measures. These results demonstrate that, although young adult women have smaller orofacial structures, they produce speech movements of equal amplitude and velocity to those of men. The results of the within-group correlations between the anthropometric measures and the kinematic measures are consistent with the findings of the between-group analyses. Almost all of the correlations failed to reach significance, which consequently provided no support for a size-scaling principle for displacement or velocity. Thus, in contrast to Kuehn and Moll (15), who examined a very small subject pool (n = 5 adults), we do not find evidence for a sex-related size-scaling principle for speech movements.

Our subject pools were relatively large for a speech kinematic study, but were the participants representative of the general population? Farkas (5) compiled normative anthropometric data on the head and face, height, and weight of 2,326 Caucasian North Americans ranging in age from newborn to young adult. The differences between the present measurements (mandibular arc, lower facial height, lower face depth, mandibular width, and stature) and Farkas' norms were no greater than 4%. Thus we assume that our samples of children and adults are representative of the general Caucasian North American population.

Another methodological issue concerns the rationale for our selection of the measures of displacement and velocity, particularly our use of 80% of the movement trajectory to determine dynamic ranges. In our initial approach, the whole utterance measures of displacement and velocity dynamic range were determined as the maximum minus the minimum point within the displacement and velocity trajectories for BBAP and MBPP. It became apparent that the location of the points selected with this method differed between adults and children and that the children displayed more variability in the location in the movement sequence at which maximum and minimum points were chosen. In other words, the specific movement components contributing to this dynamic range differed in children and adults, and within the child group were more variable between subjects. Therefore, comparisons of the two groups on this measure might be misleading. The dynamic range analysis was then modified to exclude the upper and lower 10% of points within a single trajectory, so that the measures would be comparable across the two subject groups. The single-movement measurements were chosen to complement the dynamic range analysis by providing movement amplitude and peak velocity information about specific movements that were the same for both age groups.

The results of the dynamic range and the single movement measurements provide a consistent picture across both utterances of a relatively large-amplitude, low-velocity strategy of speech production for children compared with young adults. Previous direct kinematic studies provided limited support for the idea that children use a low-velocity, large-amplitude movement strategy of speech production (23, 27). The data from the present study indicate that, although young children move their articulators comparable distances to adults during speech production, they make these movements at a significantly slower speed. An increase in movement amplitude coupled with a decrease in velocity results in a slower rate of speech. In fact, our measures of overall movement trajectory duration (mean durations for the 2 sentences were 1.24 and 1.38 s for the 5-yr-old subjects, and 0.87 and 0.96 s for the adults; Table 2) suggest that the children's rates must increase by 30% to reach the adult value. A recent paper from our laboratory (31) demonstrates that this slower speech rate persists into adolescence. With more time to produce speech, one possibility is that children are allowed additional time to organize and perform the cognitive and linguistic aspects of speech production that may be more automatic for adults. Also, the premotor planning processes might be slower in children.

Another possible benefit of an increased duration in speech production resulting from a low-velocity strategy is that children take advantage of the increased amount of time to use position-sensitive feedback during speech production. Adams et al. (1) examined the effect of speaking rate on the velocity profile of the lower lip and tongue tip movement in five adults. They found that a slow speaking rate resulted in less symmetrical, multipeaked velocity profiles, whereas faster speaking rates resulted in symmetrical, single peaked profiles. Using Bullock and Grossberg's (3) model of motor control, Adams et al. suggested that, as the speed of movement decreases, the potential contribution of information provided by feedback increases. In terms of the classic speed-accuracy trade off (34), this finding is consistent with a movement control strategy in young children in which accuracy is weighted more heavily than speed. Earlier studies have clearly established that, compared with adults, children are less consistent from trial to trial in generating speech movement trajectories (6, 23, 31). If the motor system of the young child is inherently less consistent, he/she may use a slower movement speed to preserve accuracy. To our knowledge, studies of changes in speech motor control in children as a function of changing speech rate have not been done. Children may use a low-velocity, closed-loop strategy for speech production to make use of feedback on the position of their articulators and to decrease the probability of errors, whereas adults use a more rapid, open-loop strategy. That is, speech movements may be more automatic for adults, because the nature in which target positions are achieved has been well established by years of practice.

The goal of speaking is to produce an acoustic signal that can be interpreted by a listener. As in handwriting (8), the size of the effectors does not seem to have a significant effect on the amplitude of oral movement for speech. It could be speculated that there is an acoustic advantage for a relatively large oral opening for young children. This larger oral opening would contribute to increased intensity of the speech acoustic output. Increased overall intensity of the speech signal could be advantageous in offsetting the higher fundamental frequency and formant frequencies produced by young children with small vocal folds and relatively short vocal tracts. The same speculation could be made for adult females. Although we did not measure the intensity of the speech produced by our participants, studies of the acoustic properties of speech produced by adults and young children have shown that young children produce speech of equal loudness compared with adult subjects in a laboratory testing environment (11).

In conclusion, the maturation of the neural systems underlying language processing and the production of speech follow a protracted time course of cortical, dendritic, and synaptic development (10, 12, 17) continuing into the mid-teen years. This pattern of neural development suggests that adult-like motor control over fine movements may not be possible for young children, and a recent study (31) suggests that some aspects of oral motor control for speech are not mature even at an age of 16 yr. Therefore, children learning to produce speech and to write are less able to precisely control their motor output. This lack of precise control of movement coordination may explain why relatively large hand movements are seen in children learning to write (8) and relatively large movements of the orofacial structures are observed in children during speech production. The nonfluent writing strategy of young children is also partly attributed to a closed-loop strategy, such that sensory feedback (i.e., visual guidance) is needed to produce and correct the movements (8). Young children developing speech motor control processes may rely more on sensory feedback to control orofacial movements and may need more time for language formulation and planning; thus they adopt a slow, large movement strategy in speech production.