Speaking Rate and Speech Movement Velocity Profiles

Transcription

1 Journal of Speech and Hearing Research, Volume 36, 41-54, February 1993 Speaking Rate and Speech Movement Velocity Profiles Scott G. Adams The Toronto Hospital Toronto, Ontario, Canada Gary Weismer Raymond D. Kent University of Wisconsin-Madison The effects of speaking rate on the velocity profiles of movements of the lower lip and tongue tip during the production of stop consonants were examined using an x-ray microbeam system. Five young adults used a magnitude production task to produce five speaking rates that ranged from very fast to very slow. Results indicated that changes in speaking rate were associated with changes in the topology of the speech movement velocity-time function. Specifically, the velocity profile changed from a symmetrical, single-peaked function at the fast speaking rates to an asymmetrical and multi-peaked function at the slow speaking rates. This variation in velocity profile shape is interpreted as support for the view that alterations in speaking rate are associated with changes in motor control strategies. In particular, the control strategy for speech gestures produced at fast speaking rates appears to involve unitary movements that may be predominately preprogrammed, whereas gestures produced at slow speaking rates consist of multiple submovements that may be influenced by feedback mechanisms. KEY WORDS: speech rate, speech kinematics, velocity profile, magnitude production, x-ray microbeam, speech motor control A dominant issue in speech motor control research has been the identification of relatively invariant properties of movement. The emphasis on this theme has been based on the view that the delineation of movement invariants will reveal the basic motor control processes underlying speech production. A number of studies have indicated that the transitional movements of single speech articulators, into and out of occlusal consonants, may be relatively invariant across a variety of contexts and influences (Fujimura, 1986; Houde, 1967; Kent, 197). One such influence is rate of speech. Studies examining the effects of speaking rate on various parameters of speech movements have yielded inconsistent results. Although some investigations have indicated that the amplitudes of articulatory movements are unchanged as speaking rate is increased (Abbs, 1973; Benguerel & Cowan, 1974; Engstrand, 1988; Gay & Hirose, 1973; Hughes & Abbs, 1976), others have suggested that movement amplitudes are decreased (Flege, 1988; Kent & Moll, 1972; Lindblom, 1964). Similarly, articulatory movement velocities have been found to increase (Abbs, 1973; Flege, 1988; Gay & Hirose, 1973; Gay, Ushijima, Hirose, & Cooper, 1974), decrease (Gay et al., 1974; Kozhevnikov & Chistovich, 1965), or remain unchanged (Benguerel & Cowan, 1974; Gay & Hirose, 1973; Giles & Moll, 1975; Kent, 197; Lindblom, 1964) as speaking rate is increased. In addition, some studies have shown considerable variation across individuals in the behavior of these two movement variables as a function of speaking rate (Kuehn & Moll, 1976; Ostry & Munhall, 1985). For example, Ostry and Munhall (1985) found that, as subjects spoke more rapidly, some individuals increased the velocity but not the amplitude of their movements, whereas others decreased the amplitude but not the velocity of their movements. Thus, when movement amplitude and movement velocity are examined independently, neither appears to be invariant across changes in speaking rate. 1993, American Speech-Language-Hearing Association / $1./

2 42 Journal of Speech and Hearing Research Some researchers have suggested that these measures of single points on the amplitude and velocity time functions may be inadequate for determining the underlying processes of motor control (Ostry & Munhall, 1985). Instead, a more dynamic measure that characterizes the way in which a speech movement unfolds over time is considered to be a more appropriate method for examining the invariant aspects of motor control. For example, the shape of the velocity-time function or the velocity profile is one aspect of the movement trajectory that has been suggested to display invariance. For both rapid speech and limb movements, the velocity profile has consistently been observed to have a symmetric bell-like shape (Abend, Bizzi, & Morasso, 1982; Atkeson & Hollerbach, 1985; Georgopoulos, Kalaska & Massey, 1981; Kent, 1972; Morasso, 1981; Nelson, 1983; Soechting & Lacquaniti, 1981). In their attempts to quantify the consistency of the velocity profile, Munhall, Ostry, and Parush (1985) devised an index of the profile's geometry, which they referred to as parameter c. Parameter c was developed under the assumption that the triangle-like form of the velocity profile allows certain geometric methods, useful for measuring triangles, to be applied to the quantification of the velocity profile's shape. In particular, the peak velocity, maximum amplitude of movement, and movement duration are used to supply the height, area, and base of the triangular shape of the velocity profile, and these values are combined using the formula, (peak velocity/maximum amplitude) movement duration = parameter c, to give a value that serves as a metric of the velocity profile's shape. Munhall et al. (1985) used parameter c to characterize laryngeal movements at fast and habitual speaking rates and found that, by this criterion, the shape of the velocity profile remained constant. In a subsequent study, Ostry, Cooke, and Munhall (1987) used parameter c and a measure of profile symmetry (acceleration time/deceleration time) to characterize the velocity profiles of tongue dorsum lowering, jaw lowering, and elbow extension movements during subject-selected fast and slow rates of movement. For tongue and jaw movements, the value of parameter c and the amount of asymmetry increased with movement duration. For arm movements, asymmetry also increased with movement duration, but parameter c remained unchanged. Taken together, the studies by Munhall et al. (1985) and Ostry et al. (1987) raise some interesting questions about the invariance of the velocity profile across changes in speaking rate. The finding that laryngeal movements did not show a change in parameter c across fast and normal rates of speech but tongue dorsum and jaw movements did show a change in parameter c across fast and slow rates raises the following concerns. First, it suggests that a fairly large change in movement duration or speaking rate (i.e., from very fast to very slow) may be required before a change in velocity profile will be observed. It also suggests that there may be differences across structures (i.e., laryngeal versus mandibular) in terms of the extent to which a given rate of speech will be associated with a particular change in velocity profile. The finding that the value of parameter c remained constant across changes in arm movement speed, whereas the value of the symmetry of the profile changed, raises a concern about the utility of parameter c for future studies attempting to delineate variations in velocity profile shape. For example, it is possible to imagine a large set of velocity profiles that differ in the magnitude and direction of asymmetry but not in the value of parameter c. Similarly, it is possible to envision a variety of velocity profiles with the same parameter c value that show a wide range of shapes that deviate from the expected triangle-like form. Thus, although a change in the value of parameter c can indicate a change in velocity profile shape, it does not provide information about the nature of that change. Consequently, a variety of alternate measures must be developed to describe and quantify the shape of speech movement velocity profiles. The purpose of this study was to determine the effects of speaking rate on the shape of speech movement velocity profiles across a wide range of speaking rates. The present study attempted to extend previous work in this area in the following ways: 1. Five rather than two speaking rates were examined. This included two slower-than-normal and two faster-thannormal speaking rates. Slow speaking rates have often been ignored in previous rate control studies. 2. Subjects achieved changes in speaking rate via the use of an autophonic scaling procedure (Lane & Grosjean, 1973). In contrast to a self-selected rate manipulation procedure (Ostry et al., 1987), it was believed that this procedure would ensure that subjects produced a wide range of speaking rates; it also may provide new information about the psychophysical self-scaling of speaking rate. 3. The velocity profiles for the lower lip and tongue tip movements were described. Neither of these structures received attention in previous velocity profile studies. In addition, tongue tip movements have rarely been examined in rate control studies, although these are the most frequently used consonant articulations. 4. The velocity profile was quantified in terms of a number of parameters not previously examined. These included the time to peak velocity, the skewness and kurtosis of the velocity profile, and the number of velocity peaks per movement. Method Subjects The subjects were from 19 to 35 years of age and included one woman and four men. They had no history of speech, language, or hearing pathology and all were native speakers of Standard American English, having spent most of their lives in the midwestern United States. Each subject's speech articulation performance, hearing, and oral motor skills were judged to be within normal limits by an experienced speechlanguage pathologist. Subjects were each paid $3 dollars for their participation in the study. Speech Sample February 1993 Each subject produced the phrase tap a tad above 1 times at five different speaking rates, for a total of 5 utterances. Subjects were instructed to use a casual style of speech similar to that used in everyday conversation.

3 Adams et al.: Speaking Rate and Speech Movement Velocity Profiles 43 The five speaking rate conditions were achieved using a magnitude production procedure similar to that described by Lane and Grosjean (1973). Initially, each subject was told that his or her habitual conversational speaking rate was to be assigned a value of 1. The subject was then presented with numbers greater and less than 1 and asked to increase or decrease his or her speaking rate based on the ratio formed by the presented number and the number 1. For example, when the number 2 was presented, the subject was expected to produce a speaking rate that was twice as fast as the conversational speaking rate (1). The numbers 4, 2, 1, 5, and 2.5 were presented to each subject in randomized blocks of 1. This procedure resulted in five speaking rate conditions: a rate four times that of conversational speech (4), a rate two times that of conversational speech (2), a conversational rate of speech (1), a rate half that of conversational speech (5), and a rate one quarter that of conversational speech (2.5). The subject practiced the magnitude production task for approximately 1 minutes prior to beginning the experimental protocol. The subject was instructed to maintain a constant level of loudness and this was monitored throughout the experiment using a VU meter. Instrumentation The University of Wisconsin x-ray microbeam system was used to record the movements of small radiodense markers affixed to the tongue tip and lower lip during the various speaking tasks. Detailed descriptions of this system have been provided in previous reports (Adams, 199; Nadler, Abbs, & Fujimura, 1987; Westbury, 1991). Briefly, the x-ray microbeam is a computer-controlled system that uses a narrow beam of x-rays to localize and track the two-dimensional movements of small gold pellets (approximately 3 mm in diameter) attached to the various speech articulators. Using a series of high-speed calculations and dedicated digital circuits, the system is capable of tracking the movements of as many as 1 pellets with an aggregate sampling rate of up to 7 times per sec. Typically, the individual pellets are sampled at rates between 4 and 18 Hz. In the present experiment, 1 gold pellets were attached to the surfaces of various orofacial structures using dental cement (Ketac-bond). Figure 1 shows the approximate pellet positions on a midsagittal x-ray tracing of the vocal tract. The pellets attached to the tongue tip and lower lip were of primary interest in the present study. The tongue tip pellet was sampled at 18 samples/sec while the lower lip pellet was sampled at 9 samples/sec. The remaining eight pellets were attached to the tongue blade, tongue dorsum, upper lip, central maxillary incisor, maxillary molar (not shown in Figure 1), central mandibular incisor, mandibular molar, and bridge of the nose. All pellets, except the maxillary and mandibular molar pellets, were placed in the mid-sagittal position. The pellets attached to the central maxillary incisor and the maxillary molar were used to establish the maxillary occlusal plane. The central maxillary incisor pellet was also used with the pellet on the bridge of the nose to correct for changes in the subject's head position. The pellets attached to the REFERENCE PELLETS (MAXILLA) UPPER LIP PELLET TONGUE PELLETS LOWER LIP PELLET MANDIBLE PELLETS FIGURE 1. Mid-sagittal x-ray tracing of the head showing approximate pellet locations. central mandibular incisor and the mandibular molar were used to estimate the position of the jaw. Subjects were positioned perpendicular to the microbeam system allowing for movements along the inferior-superior (y) dimension and the anterior-posterior (x) dimension to be recorded. The acoustic speech signal was sensed using a Shure SM81 microphone placed approximately 25 cm in front of the subject's mouth and digitized on-line at a sampling rate of 2 khz. A SUN 2/12 data acquisition unit was used to acquire the acoustic and pellet movement data. Data were stored on two computers (SUN 2/17 and SUN 3/18) acting as file servers. Data were processed and analyzed using a SUN system workstation and a variety of software programs developed for the University of Wisconsin-Madison X-ray Microbeam Facility. Data Processing Movements of the lower lip and tongue tip were initially corrected for head movement artifacts in the sagittal plane. This was accomplished by expressing the lip and tongue movements relative to the maxillary reference pellets. In addition, the maxillary pellets were used to determine the coordinate space for the lip and tongue movements. The maxillary occlusal plane formed the horizontal axis and a line perpendicular to this axis and located at the maxillary incisor formed the vertical axis. Once this two-dimensional frame of reference was determined, the lip and tongue movements were examined visually using x-y displacement plots. For the present study, the closing and opening movements of the tongue tip and lower lip during the production of the second /t/ and the /b/ in the phrase tap a tad above were selected for analysis. Based on extensive visual examination of the data, it was felt that most of these movements could be accounted for by changes in position along a single dimension. To determine the single dimension that would account for greatest amount of change in movement, a principal component analysis was performed on the two-dimensional tongue and lower lip movements using the approach reported by Kiritani (1986). The first principal component determined the new axis (single dimension) along which movement characteristics would be calcu-

4 44 Journal of Speech and Hearing Research February ACOUSTIC TAP A TAD ABOVE * O TONGUE TIP MOVEMENT X COORDNATE (mm) FIGURE 2. First principal component for the tongue tip pellet movements (TT) produced by Subject 2. Movements are from all of the subject's opening and closing gestures produced during the second t/ In the utterance tap a tad above. Associated movements of mandibular Incisor (J2) and molar (J1) pellets are also shown. Palate refers to the mid-sagittal trace of the subject's palate. The reference planes Included the maxillary occlusal plane and the central maxillary Incisor plane (CMI). lated. All of the data were subsequently rotated and reexpressed relative to this new axis. The average percent of variance in the movements accounted for by the first principal component was 91.3 (range %). Figure 2 shows an example of the first principal component that was determined for 1 subject's tongue tip movements. The movement data were then low-pass filtered at 1 Hz using a bidirectional, third-order Butterworth filter. The velocity and acceleration time functions were created by calculating the first and second derivatives of the filtered movement data respectively. Derivatives were calculated using the method of determining the central differences spanning three points. The acoustic speech signal was low-pass filtered at 1 khz at the time of data collection. Measurements In order to examine the effects of the magnitude production task on speaking rate, total utterance durations were determined from the speech acoustic signal (amplitude-time function). Total utterance duration was measured as the time from the onset of the burst for the first alveolar plosive to the time of the last glottal pulse in the utterance. Prior to making any kinematic measures, the movements of interest were located and indexed. To accomplish this, the acoustic, displacement, and velocity time functions were simultaneously displayed and the onset and termination points for the opening and closing gestures of the tongue tip and lower lip movements were determined. The acoustic time function was used to locate the speech segments of interest. Next, the points of maximum and minimum displacement associated with this acoustic segment were located on the displacement time function. Finally, the points of zero velocity that were closest to these displacement maxima and minima were located on the velocity time function. These To VELOCITY NJ GESTURE V7 OPENNG GESTURE /V \ 4 1 r 7 1J m1 1 msec FIGURE 3. An example of the onset and termination Indices for the tongue tip closing and opening gestures in the utterance tap a tad above produced by Subject 2. zero-velocity crossings were used to index the gesture's onset and termination points. Figure 3 shows an example of the acoustic, displacement, and velocity time functions and the indices that were determined for the tongue tip's opening and closing gestures. For each opening and closing movement, the maximum change in displacement (D), the maximum (peak) velocity (PV), and the total movement time (T) were determined. These measures were used to calculate a parameter c value for each gesture (see previously described formula). The percent of total movement time taken to reach peak velocity (TTP) was determined to provide a measure of the symmetry of the velocity profile [(time from movement onset to peak velocity/t) 1]. The number of velocity peaks in each profile was determined by counting the number of zero crossings on each gesture's acceleration time function. Finally, the moments for the velocity profile were determined by treating the instantaneous velocity at each sample point as the frequency value for a discrete frequency distribution. An example of a velocity profile and the resulting frequency distribution (histogram) are shown in Figure 4. The data from the velocity profile frequency distribution were then used to calculate the coefficients of the third (skewness) and fourth (kurtosis) moments about the mean. The method for determining these coefficients has been described in other reports (Forrest, Weismer, Milenkovic, & Dougall, 1988; Newell & Hancock, 1984). Briefly, the coefficient of skewness is formulated by determining the average deviation of scores from the mean raised to the third power, divided by the standard deviation raised to the third power. For the symmetrical "normal" distribution the coefficient of skewness is zero. For asymmetrical distributions the coefficient of skewness can take on a negative or positive value that reflects the size and direction of the asymmetry. The coefficient of kurtosis is determined by the average deviation of scores from the mean raised to the fourth power, divided by the standard deviation raised to the fourth power, which is then subtracted by 3. The bell-shaped "normal" distribution has a coefficient of kurtosis value of zero. For distributions that are more

5 Adams et al.: Speaking Rate and Speech Movement Velocity Profiles 45 VELOCITY Z 5 4 o=s1 i =S2 a =S3 A=S4 O =S5 o i. II FL II111 1Was..,1W I I I I111,J TIME FIGURE 4. An example of a velocity time function and the corresponding histogram used for the moments analysis. peaked than the normal distribution (leptokurtic) the kurtosis value is greater than zero. Distributions flatter than the normal distribution (platykurtic) have negative kurtosis values. It was believed that these skewness and kurtosis values could provide a means of quantifying changes in the symmetry and peakedness of the velocity profile across the various experimental conditions. Statistical Analysis A separate univariate statistical analysis was performed on each subject's kinematic and acoustic measures (withinsubject design). This analysis involved a series of planned comparisons using the Mann-Whitney statistic. A nonparametric statistic was chosen because of the non-normal distributional characteristics of ratio data (Atchley, Gaskins, & Anderson, 1976). A Type I error rate (alpha level) of p <.5 was allotted to the set of rate comparisons performed on each subject's kinematic and acoustic variables. This p <.5 alpha level was shared by each of the planned comparisons using the Dunn-Bonferoni procedure (Kirk, 1982). Results Acoustic Analysis I Figure 5 shows the median durations of the utterance tap a tad above produced at each of the five intended speaking rates for each of the 5 subjects. This figure indicates that all subjects adjusted their speaking rates in the intended directions. On average, subjects increased their speaking rates by approximately 57% (4-7%) and decreased their rates by about 7% (62-8%), relative to their habitual speaking rates (rate 1). Results of the statistical comparisons of utterance durations across speaking rate conditions are given in Table 1. Note that, in all but two cases, the median utterance durations associated with the slower speaking rates were o 3-1- Z ' o i I INTENDED RATE (LOG SCALE) I FIGURE 5. Subjects' median utterance durations for each of the Intended speech rate conditions. significantly longer than the utterance durations of the faster speaking rates (p <.25). These data indicate that the magnitude production task can be an effective procedure for the examination of a wide range of speaking rates. Analysis of Movement Duration o i I I I I Figure 6 shows the median durations of the closing and opening movements of the lower lip and tongue tip produced at each of the five speaking rates for subjects S1-S5. Results of the corresponding statistical comparisons are given in Table 2. Both the qualitative and statistical analyses indicated that movement duration increased as speaking rate decreased. However, the magnitude of the changes in movement duration across speaking rates varied across structures, gestures, and speakers. For example, at the slower TABLE 1. Comparisons of utterance durations across speaking rates. The durations of the utterance tap a tad above were compared across each of the five speaking rates using the Mann-Whitney statistic (alpha =.25). S1 S2 S3 S4 S5 R1 vr R2 v R R3v R R4v R R1v R R2vR R3vR R1v R R2 v R R1v R Note. The presence of a + or a - indicates a significant comparison. The + indicates that the first rate of the comparison had the higher value, whereas the - means that the second rate had the higher value (R1 = rate 2.5, R2 = rate 5, R3 = rate 1, R4 = rate 2, R5 = rate 4, S = subject).

6 46 Journal of Speech and Hearing Research February 1993 LOWER LIP CLOSING GESTURE 6 -S o-51 5 * -S2 A-S3 A -S4 4 o S5 LOWER LIP OPENING GESTURE 65~~~- o-s2 4 *-S5 3oo 1oo. S o o i i to o o 2 4 z 3 o wr TONGUE TIP CLOSING GESTURE oo-s1 5o IOA ~5.~ -S2 A~ AS3 A A -S4 4 s5 3oo 2 A 1oo. * * 2. 5 I lo 2 4 4oo TONGUE TIP OPENING GESTURE g o -si -S2. A - S3 S4 2 A S B so S5 2 INTENDED RATE FIGURE 6. Subjects' median movement durations for the tongue tip and lower lip movements produced across the Intended speaking rate conditions. Panels correspond to (a) lower lip closing gestures, (b) lower lip opening gesture, (c) tongue tip closing gesture, (d) tongue tip opening gesture. rates of speech (2.5 and 5), the median durations of the tongue tip movements were much larger than those of the lower lip movements. Further, there was a wider range of tongue tip movement durations than lower lip movement durations. In addition, the effects of rate on movement duration were more consistent across subjects for the opening than the closing gestures. In fact, the tongue tip opening gestures showed the most consistent and extensive changes in movement duration across the 5 subjects and the five speaking rate conditions. Based on these findings, all subsequent kinematic analyses were applied to data from the tongue tip opening gestures only. TABLE 2. Comparisons of movement duration across speaking rates. The durations of selected closing (cl) and opening (op) gestures were determined for movements of the lower lip and tongue tip during 1 productions of the utterance tap a tad above. The movement durations for each structure were compared across each of the five speaking rates using the Mann-Whitney statistic (alpha =.25). Subject 1 Subject 2 Subject 3 Subject 4 Subject 5 lip tongue lip tongue lip tongue lip tongue tip tongue cl op cl op cl op cl op cl op cl op cl op cl op cl op cl op R1v R R2 v R R3 v R R4 v R R1v R R2 v R R3 v R R1v R R2 v R R1v R Note. A + or a - indicates that a significant difference was found for the listed rate comparison. The + indicates that the first rate of the comparison had the higher value, whereas the - means that the second rate had the higher value (R1 = rate 2.5, R2 = rate 5, R3 = rate 1, R4 = rate 2, and R5 = rate 4).

7 Adams et a.: Speaking Rate and Speech Movement Velocity Profiles 47 S1 S2 S3 S4 S5 RATE , K - - -I k_ I 1 I hr_ r -I-- I y - a. L,*O- --- il-_ RATE 5 RATE 1 RATE 2 k RATE 4 TM FIGURE 7. Velocity profiles for the tongue tip opening gestures produced across the Intended speaking rate conditions. Velocity Profile Analysis Velocity profiles corresponding to each subject's tongue tip opening gestures produced at each of the five speaking rates are shown in Figure 7. Note that, within subjects, the velocity profiles for movements produced under the faster speaking rate conditions were qualitatively different from those produced under the slower speaking rate conditions. Velocity profiles from the fast rates typically were symmetrical and characterized by one relatively large velocity peak, whereas the slow rate velocity profiles generally were asymmetrical and had multiple velocity peaks. Selected results of the quantitative analyses of velocity profile shape are illustrated in Figure 8 for Subject 2. This figure shows the values for (a) parameter c, (b) time to peak velocity, (c) coefficient of skewness, (d) coefficient of kurtosis, and (e) the number of velocity peaks for the tongue tip opening movements at each of the five speaking rates. Each measure of the velocity profile is plotted as a function of the movement duration of the tongue tip opening gesture. In addition, the intended rates are indicated by the different symbols. Figure 8 shows that as speaking rate decreased (and movement duration became longer) there was a tendency for increases in (a) the value of parameter c, (b) the value of the coefficient of skewness, and (c) the number of velocity peaks per profile. In contrast, values for the time to peak velocity and coefficient of kurtosis decreased as speaking rate became slower. Figure 9 shows the subjects' median values for parameter c, the time to peak velocity, the coefficient of skewness, and the coefficient of kurtosis at each of the five speaking rates. Results of the statistical comparisons for each of the velocity profile measures across the speaking rate conditions are summarized in Table 3. Both Table 3 and Figure 9 indicate that parameter c and the coefficient of skewness tended to increase as speaking rate became slower. It should be noted, however, that for Subjects 1 and 5, none of the comparisons of skewness reached significance. The values for the time to peak velocity were observed to decrease as the rate of speech was slowed. An inconsistent pattern of results was observed across subjects for the coefficient of kurtosis. The results of the statistical comparisons of the number of velocity peaks are shown in Table 3. In addition, Figure 1 shows the frequency of occurrence of velocity profiles with one, two, three or more than three velocity peaks. For all subjects, the velocity profiles from the slower rates of speech (2.5 and 5) had significantly greater numbers of velocity peaks than the fast rate velocity profiles (Table 3). Discussion Speaking Rate and Movement Duration The magnitude production task employed in the present investigation appears to be a useful procedure in the study of speaking rate. In particular, it yields a number of subjectselected speaking rates that cover a wide range of the speech rate continuum. With regard to the resulting movement durations, the lower lip and tongue tip opening gestures showed more consistent changes in movement duration across changes in speaking rate than the closing gestures. This was observed both

8 48 Journal of Speech and Hearing Research February E 3 we 2 E E 1 t.,. Subject 2 a.'f~~ - MOVEMENT DURATION (msec) o o Rate 2.5- o ~~Rate 5 = Rate 1 -e Rate 2 - Rate 4 - U I l li _. 5E ir E a. 2,;S S MOVEMENT DURATION (meec) - 5 U, z a U, a U- 3 U Z U, To Y z - ar L W LU U MOVEMENT DURATION (msec) MOVEMENT DURATION (msec) 7 Subject a a v LU o z, 2 5. o 4.- co a) 3.- a * o o 2. - * A a e e Rate 2.5= Rate 5 = Rate 1 =- 1 O- n as Rate 2 =A Rate 4 =. I U ZUU 4uU UU ouu MOVEMENT DURATION (msec) FIGURE 8. Values for the velocity profile measures obtained from Subject 2's tongue tip opening gestures at each of the five speaking rates. Velocity profile measures are plotted as a function of the movement duration and Intended rates are Indicated by the symbols In the keys. Panels correspond to the following measures; (a) parameter c, (b) time to peak velocity, (c) coefficient of skewness, (d) coefficient of kurtosis, and (e) the number of velocity peaks. across and within subjects and may be related to certain perceptual-acoustic or linguistic influences on motor control strategies. In a study by Amerman and Parnell (1981), consonant-vowel (CV) transitions obtained from utterances spoken at different speaking rates were recognized more accurately than vowel-consonant (VC) transitions suggesting that CV transitions are more perceptually robust in the face of changes in speaking rate. The greater perceptual resilience of the CV transitions may be related to the greater consistency in the movement durations for opening gestures observed in the present study. If, as Marslen-Wilson (198) has proposed, the identification of word-initial consonants is critical to speech perception, then speech motor control strategies may have developed to minimize the variability of the opening gestures at different speaking rates. The notion that speech motor control may be structured, at least in part, to produce good word onsets has been suggested in one recent model of speech motor control (Weismer & Liss, 1991). Alternately, it is possible that minimal variability in production creates "islands of reliability" that are exploited by

9 to U3 E 4..o O=S * =S2 A =S3 3. o- * A =S4 =S A.- I I I I INTENDED RATE Adams et al.: Speaking Rate and Speech Movement Velocity Profiles 49 (o g E LI- C O 2 a at 1 o=s1 =-S2 75- a=s INTENDED RATE I i E 1. /1.V.5. _-.5 o 5i - LI z _ -1.U A INTENDED RATE O=S1 *=S2 A A =S3 6 =S4 *~ ~O=S5 aa n a l I I I vn C1 Y I- Q L. oli o U W ;i 2. O=S1 * =S2 -.5 o A =S3 A o=s4 Ob =S5-2. I I I I I INTENDED RATE FIGURE 9. Subjects' median values for (a) parameter c, (b) the time to peak velocity, (c) the coefficient of skewness, and (d) the coefficient of kurtosis, during each of the five Intended speaking rate conditions. speech perception processes. Another potential influence may have been the particular stress pattern of the sentence tap a tad above. The opening gestures in this sentence were associated with stressed vowels, whereas the closing gestures were associated with unstressed vowels. Thus, it is possible that when speakers make adjustments in their rates of speech, more consistent changes in movement duration occurred for the movements associated with stressed vowels TABLE 3. Comparisons of the measures of velocity profile shape across speaking rates. Measures Included parameter c (C), percent of time to peak velocity (TT), coefficient of skewness (SK), coefficient of kurtosis (KU), and the number of velocity peaks (PK). The values for each measure were compared across each of the five speaking rates using the Mann-Whitney statistic (alpha =.25). Subject 1 Subject 2 Subject 3 Subject 4 Subject 5 C TT SK KUPK CTTSKKUPK C TT SKKUPK CTTSKKU PK C TT SK KU PK R1v R R2 v R R3 v R4 + + R4 v R5 R1v R R2v R R3vR R1v R R2v R R1v R Note. A + or a - indicates that a significant difference was found for the listed rate comparison. The + indicates that the first rate of the comparison had the higher value, whereas the - means that the second rate had the higher value (R1 = rate 2.5, R2 = rate 5, R3 = rate 1, R4 = rate 2, and R5 = rate 4).

10 5 Journal of Speech and Hearing Research February 993 S1 S2 S3 S4 S5 w C., zo 1 6W1 LL l L I RATE 4 2 o7o O C.) oill a, ;- L I L 1 z 5 2 L 1 L NUMBER OF VELOCITY PEAKS FIGURE 1. Frequency of occurrence of velocity profiles with one, two, three, or more than three velocity peaks at each of the Intended speaking rates. than unstressed vowels. A similar explanation could be made with regard to the influence of vowel tension on these gestures, as recent evidence suggests that the durations of tense and lax vowels can be differentially affected by speaking rate (Gopal, 199). Another potential influence is the relationship of the opening and closing gestures to the syllabic pattern of the sentence. For both the lower lip and tongue tip, the closing gestures occurred across a syllabic juncture whereas the opening gestures occurred within a syllable. It is possible that, across changes in speaking rate, there was greater variability in the timing of movements that occurred across syllabic junctures than for movements that occurred within syllables. These linguistic-phonetic factors need to be examined in future studies of speech rate control. The tongue tip gestures were found to have much longer movement durations than the lower lip gestures, especially during the slow speaking rate conditions. This may be related to differences in linguistic context. The differences in lower lip and tongue tip movement durations may be a consequence of the movements being associated with different word positions in the same sentence. In the present study, the lower lip movements were taken from the last word in the sentences, whereas the tongue tip movements were from the second-to-last word in the sentences. To determine whether the stress pattern of these sentences may have contributed to the differences in lip and tongue movement durations, additional studies are required that control for the effects of linguistic context. Biomechanical differences, such as the size, shape, and interconnections of muscle fibers in the two structures also may have contributed to differences in the lower lip and tongue tip movement durations. In addition, the lower lip and tongue tip movements to and from consonantal sounds may be differentially affected by the adjacent tongue movements required for vowel production. The consonantal movements of the lower lip, unlike the consonantal tongue tip movements, are not physically linked to the tongue movements required for vowel production. Thus, any effects of changes in speaking rate on vowel movements may have less impact on consonantal lower lip movements than consonantal tongue tip movements. In summary, the results of this study indicated that the effects of changes in speaking rate on movement duration were different for (a) opening and closing gestures, and (b) lower lip and tongue tip gestures. These findings suggest that kinematic descriptions that are limited to only one gesture type or one single articulator may be inadequate for characterizing general organizational principles in speech production.

11 Adams et al.: Speaking Rate and Speech Movement Velocity Profiles 51 Velocity Profile Shape and Speaking Rate Previous research has suggested that the velocity profiles from speech and nonspeech movements have a relatively invariant triangular shape and that this reflects the operation of a specific strategy of motor control (Abend, Bizzi & Morasso, 1982; Atkeson & Hollerbach, 1985; Beggs & Howarth, 1972; Morasso, 1981; Munhall, 1984; Ostry & Munhall, 1985). The present study attempted to measure the stability of the velocity profile's topology across a wide range of changes in speaking rate. Measures of velocity profile shape included parameter c, the percent of time to peak velocity, the coefficient of skewness, the coefficient of kurtosis, and the number of velocity peaks. Parameter c. The results of the present study indicated that the value of parameter c varied across speaking rates. Specifically, parameter c gradually increased as speaking rate became slower (and movement duration increased). This result is in contrast to the suggestion advanced by Munhall et al. (1985) that parameter c and the velocity profile are invariant across speaking rates. Munhall et al. (1985) based this notion on a study of tongue dorsum movements in which they found a constant value for parameter c across two speaking rate conditions (a normal rate and a fast rate). On the other hand, the present results are in agreement with those reported in a more recent study of tongue dorsum and jaw movements by Ostry, Cooke, and Munhall (1987). These authors found differences in parameter c values across slow and fast speaking rate conditions. For these results, and those of the present study, it appears that when large changes in speaking rate are examined, changes in the value of parameter c will be observed. Such changes in parameter c values indicate that the shape of the velocity profile is not invariant across large changes in speaking rate. Symmetry. Asymmetries in velocity profiles may occur frequently in speech, even at a constant speaking rate. Kent (1986) reported that the rate of movement for an articulation varied with its location and function in a phonetic sequence. In particular, consonant movements occurring during a period when the vocal tract is obstructed for another consonant can have a markedly slower velocity than when the vocal tract is not so obstructed. Speakers seem to have ready and flexible control over at least some aspects of articulatory velocity. Two measures of velocity profile symmetry were examined in the present study, the percent of total movement time taken to reach peak velocity (TTP%) and the velocity profile's skewness coefficient. Both the TTP% and the coefficient of skewness were found to change across the various speaking rate conditions. For the tongue tip opening gestures that were examined in this study, the TTP% values decreased and the skewness values increased (i.e., they became more positively skewed) as speaking rate became slower. These results for the tongue tip opening gestures are in agreement with those reported by Ostry et al. (1987) in a study of tongue dorsum opening gestures. They found increases in the proportion of time spent in the deceleratory phase of the movement as speaking rate decreased. However, these results are in contrast to previous studies of closing gestures. In two previous studies (Adams, 199; Connor, Abbs, Nychis-Florence, & Caligiuri, 1988) of jaw, lower lip, and tongue tip closing gestures it was found that the proportion of time spent in the deceleratory phase of movement decreased as speaking rate became slower. Taken together, these results indicate that, for a variety of speech articulators, changes in speaking rate have differential effects on the symmetry of the velocity profile depending on whether the movement is an opening or closing gesture. These differential effects of rate changes on the symmetry of the velocity profile for opening and closing movements may indicate that opening and closing movements are controlled differently. This sentiment was expressed by Gracco (1988) who suggested "that the opening movements and closing movements may reflect separate 'synergies,' implicating each action as a basic component of speech production" (p. 4637). The finding that the symmetry of the velocity profile changed across speaking rate is consistent with the results of a study of aimed hand movements reported by Zelaznik, Schmidt, and Gielen (1986). They found that, as movement duration increased, a greater proportion of the total movement time was spent in deceleration. Similar results were found for opening movements of the tongue tip in the present study. The fact that similar results are observed for hand and tongue movements suggests that the changes in velocity profile shape may reflect some common strategies in the control of movement duration across various motor systems. Changes in the symmetry of the velocity profile across changes in speaking rate and movement duration are contrary to the predictions of generalized motor program theories (Schmidt, 1976, 1982). Motor program theory predicts that when there is a change in the intended duration of a movement but the distance to be moved remains constant, the movement will be "time-rescaled" and this will be reflected in the movement's velocity profile. That is, although the scale of the velocity profile would be affected by a change in movement duration, its shape would remain constant. The results of the present study, like those of Zelaznik et al. (1986), suggest that the concept of time rescalability needs to be reformulated in generalized motor program theory. Possibly the concepts of reorganization or phase transitions (Jeka & Kelso, 1989) that predict qualitative changes in the behavior at critical movement durations need to be incorporated into any general theories of motor control. Bullock and Grossberg (1988) recently have proposed a model of motor control that they suggest accounts for the occurrence of velocity profile asymmetries at longer movement durations. Their vector integration to endpoint (VITE) model was proposed as an alternative to the numerous theories that posit the explicit computation and internal representation of invariant velocity profiles (Abend, Bizzi, & Morasso, 1982; Atkeson & Hollerbach, 1985; Flash & Hogan, 1985; Hasan, 1986; Hogan, 1984; Morasso, 1981). Within the VITE model, speed-dependent asymmetry is the result of feedback mechanisms that provide information about the current position of a structure and allow for ongoing corrective movements as a target position is approached. As movement speed decreases, the contribution of these feedback mechanisms to movement increases with the result that a less symmetrical profile is produced. Implicit to the VITE

12 52 Journal of Speech and Hearing Research model is the notion that, as movement speed decreases, a critical speed is reached at which certain feedback processes are invoked. With continued reductions in speed, these feedback processes eventually result in the production of asymmetrical velocity profiles. In general terms, the dichotomy between the VITE model and models that propose preplanned invariant velocity profiles appears to be an elaborate redressing of the traditional controversy between closed-loop versus open-loop feedback mechanisms in motor control (Abbs & Gracco, 1984; Gracco & Abbs, 1987). For example, the VITE model is similar to the notion of closed-loop feedback, whereas models proposing an invariant velocity profile are suggestive of a system that is based on open-looped control. One possible explanation for the change in velocity profile shape that was observed at slower movement speeds in the present study is that the motor control system shifts from a strategy that is predominantly open-looped at fast movement speeds to a strategy that is predominately closed-looped at slow movement speeds. This explanation could be considered as just one of many possible versions of a reorganization hypothesis for speech rate control. Kurtosis. The results from the present study suggested that the effects of changes in speaking rate on the velocity profile kurtosis values were inconsistent across subjects, gestures, and structures. This variation in kurtosis values may reflect the diversity of profile peakedness across individuals and gestures. However, it is also possible that the variation across subjects and gestures was due to factors other than variations in the profiles peakedness per se. For example, it has been suggested that kurtosis values are difficult to interpret when the population distributions they are describing are asymmetrical or multimodal distributions (Darlington, 197; Micceri, 1989; Wetherill, 1972). In the present study, many of the velocity profiles, especially from the slower speaking rate conditions, were either asymmetrical or multimodal. Thus, the potential effects of asymmetrical and multiple peaked profiles on the kurtosis values make it difficult to interpret the extent to which the peakedness of the velocity profiles was affected by changes in speaking rate. Some authors have recommended measures of tail weight as alternatives to measures of kurtosis for describing the shape of population distributions (Elashoff & Elashoff, 1978; Micceri, 1989). In particular, Q statistics (ratios of outer means) and C statistics (ratios of outer percentile points) have been recommended as two robust measures of tail weight (Elashoff & Elashoff, 1978; Micceri, 1989). In future studies, it may be useful to examine these measures of tail weight as alternate methods for quantifying changes in velocity profile shape. Number of velocity peaks. In the present study, the number of velocity peaks per gesture increased as speaking rate became slower. This finding is consistent with a study of the effects of speaking rate on jaw movements reported by Wieneke, Janssen, and Belderbos (1987). These authors suggested that the increase in velocity peaks at the slower speaking rates showed a change in the movement pattern that "may be the effect of reorganization of the motor commands related to speaking rate" (p. 124). The present results extend these findings to movements of the lower lip February 993 and tongue tip at slow speaking rates. Thus, it is suggested that the occurrence of multiple velocity peaks associated with slow speaking rates may be a general characteristic of all speech articulators. Furthermore, the results of Wieneke et al. (1987) were derived from Dutch speakers whereas those of the present study were obtained from speakers of American English. Taken together, these findings suggest that there may be certain universal mechanisms of rate control in speech production. The present results regarding the effects of changes in speaking rate on the number of velocity peaks suggest the need for a reinterpretation of some previous studies of speech kinematics in disordered speakers. In a study of lower lip speech gestures by McNeil, Caliguiri, and Rosenbek (1989), a group of subjects with apraxia of speech were found to have significantly greater numbers of velocity peaks per gesture than a group of age-matched neurologically normal subjects. It was suggested that the larger number of velocity peaks reflected a symptom of the speech movement disorder in apraxia of speech that was termed dysmetria. In a study of subjects with Parkinson's disease, Connor and Abbs (1991) reported the presence of multiple velocity peaks in novel, visually guided, jaw-opening movements and interpreted this to be an "expression of classical Parkinson's disease motor impairments." In both of these studies, however, the neurologically impaired subjects' movement durations were much longer than those of the control subjects. Given the finding of multiple velocity peaks during slow speech in the present study, it is possible that the occurrence of multiple velocity peaks in these disordered subjects may have been a natural consequence of their longer movement durations rather than an independent indication of neuropathology. The present study highlights the need to control for the potential effects of speaking rate in any study aimed at describing the speech movement correlates of neuropathology. One means of doing so would be to collect speech data from neurologically impaired subjects at a variety of rates and examine the resulting velocity profiles for evidence of changes in the number of velocity peaks. Multiple velocity peaks also have been observed in hand movements produced at slow rates (Milner, 1989; Milner & Ijaz, 199). In his error correction model of motor control, Milner (1989) suggests that these multiple peaks reflect a sequence of overlapping submovements that are used to make spatial and temporal adjustments over the course of a movement. Thus, for movements in which the spatial accuracy demands are constant, the production of multiple submovements is seen as a motor control strategy used to accomplish accurate increases in movement duration at slow movement speeds. Milner and Ijaz (199) suggest that this submovement strategy is used in slow movements because "it may be difficult to generate a single motor command with a duration as long as 1, msec" and that this may be related to "the higher frequency periodic behavior within the central nervous system" (p. 374). An extension of this interpretation may be that, as one changes from fast to slow movements, a point is reached at which there is a shift from a motor control strategy based on one movement per discrete gesture to a strategy in which multiple submovements are combined in order to produce a gesture. This change

13 Adams et al.: Speaking Rate and Speech Movement Velocity Profiles 53 from a single to a multiple submovement strategy could be interpreted as yet another version of the reorganization hypothesis of rate control originally proposed by Gay (1981). Acknowledgments This research was supported by NIH-NINCDS Grant #NS We wish to express our appreciation to Ruth Martin and John Westbury for their assistance with various aspects of this study. We also express our appreciation to Anne Smith and Vincent Gracco for their helpful comments and suggestions. References Abbe, J. H. (1973). The influence of the gamma motor system on the jaw movements during speech: A theoretical framework and some preliminary observations. Journal of Speech and Hearing Research, 16, Abbs, J. H., & Gracco, V. L. (1984). Control of complex motor gestures: Orofacial muscle responses to load perturbations of the lip during speech. Journal of Neurophysiology, 52, Abend, W., Bizzi, E., & Morasso, P. (1982). Human arm trajectory formation. Brain, 15, Adams, S. G. (199). Rate and clarity of speech: An x-ray microbeam study. Unpublished doctoral dissertation, University of Wisconsin-Madison. Amerman, J. D., & Parnell, M. M. (1981). Influence of the context and rate of speech on stop-consonant recognition. Journal of Phonetics, 9, Atchley, W. R., Gaskins, C. T., & Anderson, D. (1976). Statistical properties of ratios.. Empirical results. Systematic Zoology, 25, Atkeson, C. G., & Hollerbach, J. M. (1985). Kinematic features of unrestrained vertical arm movements. The Journal of Neuroscience, 5, Beggs, W. D. A., & Howarth, C.. (1972). The movement of the hand towards a target. Quarterly Journal of Experimental Psychology, 24, Benguerel, A., & Cowan, H. A. (1974). Coarticulation of upper lip protrusion in French. Phonetica, 3, Bullock, D., & Grossberg, S. (1988). Neural dynamics of planned arm movements: Emergent invariants and speed-accuracy properties during trajectory formation. Psychological Review, 95, Connor, N. P., Abbs, J. H., Nychls-Florence, C., & Caliglurl, M. P. (1988). Alterations in lip and jaw kinematics as a function of speech rate. Paper presented at the 115th meeting of the Acoustical Society of America, Seattle. Connor, N. P., & Abbs, J. H. (1991). Task-dependent variations in Parkinsonian motor impairments. Brain, 114, Darllngton, R. B. (197). Is kurtosis really "peakedness?" The American Statistician, 24, Elashoff, J. D., & Elashoff, R. M. (1978). Effects of errors in statistical assumptions. In W. H. Kruskal and J. M. Tanur (Eds.), International encyclopedia of statistics. New York: Free Press. Engstrand,. (1988). Articulatory correlates of stress and speaking rate in swedish VCV utterances. Journal of the Acoustical Society of America, 83, Flash, T., & Hogan, N. (1985). The coordination of arm movements: An experimentally confirmed mathematical model. The Journal of Neuroscience, 5, Flege, J. E. (1988). Effects of speaking rate on tongue position and velocity of movement in vowel production. Journal of the Acoustical Society of America, 84, Forrest, K., Welsmer, G., Mllenkovlc, P., & Dougall, R. (1988). Statistical analysis of word-initial voiceless obstruents: Preliminary data. Journal of the Acoustical Society of America, 84, Fulmura,. (1986). Relative invariance of articulatory movements: An iceberg model. In J. S. Perkell and D. H. Klatt (Eds.), Invariance and variability in speech processes. Hillsdale: Lawrence Erlbaum. Gay, T. (1981). Mechanisms in the control of speech rate. Phonetica, 38, Gay, T., & Hlrose, H. (1973). Effect of speaking rate on labial consonant production. Phonetica, 27, Gay, T., Ushillma, T., Hlrose, H., & Cooper, F. S. (1974). Effect of speaking rate on labial consonant-vowel articulation. Journal of Phonetics, 2, Georgopoulos, A. P., Kalaska, J.F., & Massey, J.T. (1981). Spatial trajectories and reaction times of aimed movements: Effects of practice, uncertainty and change in target location. Journal of Neurophysiology, 46, Glles, S. B., & Moll, K. L. (1975). Cinefluorographic study of selected allophones of English /1l. Phonetica, 31, Gopal, H. S. (199). Effects of speaking rate on the behavior of tense and lax vowel durations. Journal of Phonetics, 18, Gracco, V. L. (1988). Timing factors in the coordination of speech movements. The Journal of Neuroscience, 8, Gracco, V. L, & Abbs, J. H. (1987). Programming and execution processes of speech movement control: Potential neural correlates. In E. Keller and M. Gopnik (Eds.), Motor and sensory processes of language. Hillsdale, NJ: Lawrence Erlbaum. Hasan, Z. (1986). Optimized movement trajectories and joint stiffness in unperturbed, inertially loaded movements. Biological Cybernetics, 53, Hogan, N. (1984). An organizing principle for a class of voluntary movements. Journal of Neuroscience, 4, Houde, R. (1967). A study of tongue body movement during selected speech sounds. Unpublished doctoral dissertation, University of Michigan. Hughes, O. M., & Abbe, J. H. (1976). Labial-mandibular coordination in the production of speech: Implications for the operation of motor equivalence. Phonetica, 33, Jeka, J. J., & Kelso, J. A. S. (1989). The dynamic pattern approach to coordinated behavior: A tutorial review. In S. A. Wallace (Ed.), Perspectives on the coordination of movement. Amsterdam: North-Holland. Kent, R. D. (197). A cinefluorographic-spectrographic investigation of the component gestures in lingual articulation. Unpublished doctoral dissertation, University of Iowa, Iowa City. Kent, R. D. (1972). Some considerations in the cinefluorographic analysis of tongue movements during speech. Phonetica, 26, Kent, R. D. (1986). The iceberg hypothesis: The temporal assembly of speech movements. In J. S. Perkell and D. H. Klatt (Eds.), Invariance and variability in speech processes. Hillsdale, NJ: Lawrence Erlbaum. Kent, R. D., & Moll, K. L. (1972). Cinefluorographic analyses of selected lingual consonants. Journal of Speech and Hearing Research, 15, Kirk, R. E. (1982). Experimental design: Procedures for the behavioral sciences (2nd ed.). Belmont, CA: Brooks/Cole Publishing. Klritani, S. (1986). X-ray microbeam method for measurement of articulatory dynamics: Techniques and results. Speech Communication, 5, Kozhevnlkov, V. A., & Chlstovich, L. A. (1965). Speech: Articulation and perception. Translation from the U.S. Department of Commerce Clearinghouse for Federal Scientific and Technical Information, Washington, D.C. Kuehn, D. P., & Moll, K. L. (1976). A cineradiographic study of VC and CV articulatory velocities. Journal of Phonetics, 4, Lane, H., & Groslean, F. (1973). Perception of reading rate by speakers and listeners. Journal of Experimental Psychology, 97, Undblom, B. (1964). Articulatory activity in vowels. Speech Transmission Laboratory Quarterly Progress and Status Report, Royal Institute of Technology, Stockholm, Sweden, 2, Marslen-Wilson, W. D. (198). Speech understanding as a psychological process. In J. C. Simon (Ed.), Spoken language generation and understanding. Proceedings of the NATO Advanced Study Institute (pp ). Dordrecht: D. Reidel.

14 54 Journal of Speech and Heanng Research McNeil, M. R., Caliguirl, M., & Rosenbek, J. (1989). A comparison of speech kinematics among apraxic, conduction aphasic, ataxic dysarthric and normal geriatric speakers. In T. E. Prescott (Ed.), Clinical aphasiology. Boston: College-Hill Press. Micceri, T. (1989). The unicorn, the normal curve, and other improbable creatures. Psychological Bulletin, 15, Milner, T. E. (1989). A model for discrete error correction applied to three-dimensional movement kinematics. Unpublished manuscript. Milner, T. E., & Ijaz, M. (199). The effect of accuracy constraints on three-dimensional movement kinematics. Neuroscience, 35, Morasso, P. (1981). Spatial control of arm movements. Experimental Brain Research, 42, Munhall, K. G. (1984). Temporal adjustment in speech motor control: Evidence from laryngeal kinematics. Unpublished doctoral dissertation, McGill University, Montreal. Munhall, K. G., Ostry, D. J., & Parush, A. (1985). Characteristics of velocity profiles of speech movements. Journal of Experimental Psychology: Human Perception and Performance, 11, Nadler, R. D., Abbe, J. H., & Fujlimura,. (1987). Speech movement research using the new x-ray microbeam system. Proceedings of the Eleventh International Congress of Phonetic Sciences, Tallinn, Estonia, USSR, Paper Se Nelson, W. L. (1983). Physical principles for economies of skilled movements. Biological Cybernetics, 46, Newell, K. M., & Hancock, P. A. (1984). Forgotten moments: A note on skewness and kurtosis as influential factors in inferences extrapolated from response distributions. Journal of Motor Behavior, 16, Ostry, D. J., & Munhall, K. G. (1985). Control of rate and duration of speech movements. Journal of the Acoustical Society of America, 77, Ostry, D. J., Cooke, J. D., & Munhall, K. G. (1987). Velocity curves of human arm and speech movements. Experimental Brain Research, 68, Schmidt, R. A. (1976). Control processes in motor skills. Exercise and Sport Sciences Reviews, 4, Schmidt, R. A. (1982). Motor control and learning: A behavioral emphasis. Champaign, IL: Human Kinetics. Soechting, J. F., & Lacquaniti, F. (1981). Invariant characteristics of pointing movement in man. Journal of Neuroscience, 1, Welsmer, G., & Liss, J. (1991). Reductionism is a dead end in speech research. In C. A. Moore, K. M. Yorkston and D. R. Beukelman (Eds.), Dysarthria and apraxia of speech: Perspectives on management. Baltimore, MD: P. H. Brookes. Westbury, J. (1991). The significance and measurement of head position during speech production experiments using the x-ray microbeam system. Journal of the Acoustical Society of America, 89, Wetherill, G. B. (1972). Elementary statistical methods. London: Chapman & Hall. Wleneke, G., Janssen, P., & Belderbos, H. (1987). The influence of speaking rate on the duration of jaw movements. Journal of Phonetics, 15, Zelaznik, H. N., Schmidt, R. A., & Glelen, S. (1986). Kinematic properties of rapid aimed hand movements. Journal of Motor Behavior, 18, Received November 21, 1991 Accepted July 14, February 993 Contact author: Scott G. Adams, PhD, Speech and Swallowing Lab, , The Toronto Hospital, 399 Bathurst Street, Toronto, Ontario, Canada, M5P 2S8.