Haskins Laboratories Status Report on Speech Research1994, SR-117/118, 49-65

Articulatory Organization of Mandibular, Labial, andVelar Movements During Speech*

H. Betty Kollia,t Vincent L. Gracco, and Katherine S. Harrist

It has been shown that articulator movements during speech are adjusted along a numberof spatiotemporal dimensions. For example, variations in the extent of lip, jaw, or tonguemotion are associated with proportional changes in the respective articulators' peakvelocity. Modifications in the timing of lip and jaw actions are apparently constrained,exhibiting relative timing covariation. Syllable prominence systematically affects somecombination of the articulator motion parameters, i.e., extent, speed, and duration. Thepresent investigation is an attempt to extend observations of the spatiotemporal propertiesof articulator movement to include the velum. Lip, jaw, and velar kinematics wererecorded optoelectronically and simultaneously with the acoustic signal duringproductions of the utterance Imabnab/. The spatial and temporal relations between thelips, the jaw, and the velum were examined and compared across articulators. Formovements associated with each syllable, the velum displayed scaling patternsqualitatively similar to those of the lips and jaw. Moreover, velocity-displacement relationswere more robust for the lowering than for the raising movements of the velum. There wasevidence of interarticulator coupling between the velum and the jaw, and between thevelum and the upper lip, although this coupling was not as strong as that observed amongthe oral articulators. Articulator specific differences in velocity-displacement correlationsand degree of interarticulator cohesion for the various movement phases may be related toa combination of aerodynamic and phonetic factors, such as the phonologically non­contrastive nature of nasalization in English.

INTRODUCTIONA number of investigations have shown that lip,

jaw, and tongue movements during speech aremodified systematically across segmental andsuprasegmental variations. One of the more ro­bust movement characteristics is the positive lin­ear relationship of an articulator's peak displace­ment with its associated peak velocity. Duringspeaking, reliable correlations have been observedin the kinematic parameters of various articula­tors, such as the jaw and the lip (Ohala et aI.,1968; Kelso et al., 1985), the vocal folds (Munhall& Ostry, 1985), and the tongue dorsum (Parush,Ostry, & Munhall, 1983; Ostry, Keller, & Parush,1983), to indicate that movement velocity is pro­portionally scaled to changes in movement extent.

This work is supported by NIH Grant DC-00121, and DC­00594 to Haskins Laboratories.

49

Further, this relationship between peak velocityand displacement is reliably observed regardlessof movement direction (i.e., raising vs. lowering;Kelso et al., 1985; Munhall, Ostry & Parush 1985;Parush et aI., 1983; Vatikiotis-Bateson, 1988;Vatikiotis-Bateson & Kelso, 1993). One of theaims of the present study was to examine the re­lationship between peak velocity and displace­ment in the velum.

No consistent direction-dependent trends havebeen noted in these relations in the literature. Ina kinematic study of tongue dorsum movement,the peak velocity-displacement correlation for theraising movement toward closure was about thesame as that for the lowering movement from thestop release (Parush et aI., 1983). For one of thethree subjects in that study, the raisingmovements were slightly faster than the loweringmovements, and for another subject the raisingmovements were slightly larger than the loweringmovements. In an analysis of speech kinematics of

50 Kollia et al.

the jaw and the lower lip, Kelso et aI. (1985) foundthat both opening and closing movements showedhigh correlations between peak velocity anddisplacement with a trend for the closingmovements to display higher (steeper) velocity­displacement slopes than the opening movements.The same finding was also reported by Vatikiotis­Bateson and Kelso (1993), who concluded that,when viewed as a second order dynamical system,the peak velocity-displacement relationship canprovide a sufficient spatial and temporaldescription of the overall movement behavior. Afurther aim of the present study was to exploredirection-dependent trends in the relationsbetween peak velocity and displacement in thevelum, as well as in the jaw and the upper lip.

Stress assignment has been found to producesystematic changes in kinematic variables.Movements associated with stressed syllables aregenerally of greater displacement, higher velocityand longer duration than their unstressed coun­terparts (Gay 1968; Kelso et aI., 1985; Kent &Netsell, 1971; MacNeilage, 1970; Ostry et aI.,1983; Ostry & Cooke, 1987; Vatikiotis-Bateson,1988; Vatikiotis-Bateson & Kelso, 1993). However,examination of the velocity-displacement relation­ships for various articulators reveals that the scal­ing of these two kinematic variables displays aconsistent pattern in stressed versus unstressedsyllables, that appears to reflect the varyingmovement durations. As the movement durationincreases for the stressed element compared to theunstressed, there is a notable tendency for the ve­locity-displacement ratio to decrease (Munhall &Ostry, 1985; Ostry & Cooke, 1987). This trend hasbeen observed for tongue dorsum movements(Ostry et aI., 1983; Ostry & Cooke, 1987) as well'as for jaw and lip movements (Edwards, Beckman,& Fletcher 1991; Kelso et aI., 1985; Stone, 1981;Vayra & Fowler, 1992).

In contrast to the numerous studies of the jaw,lip, and tongue movement, relatively little re­search has focused on the movement characteris­tics of the velum. It is not known, for example,whether peak displacement and peak velocityscale in a similar manner for the velum as for theoral and laryngeal articulators, or whether stressprominence effects are reflected on velar raisingmovements. There is, however, a body of empiricalevidence regarding certain determinants of velarposition and velar movement. For example, Bell­Berti et al. (1979) found velar position to be influ­enced by vowel quality during adjacent nasal andoral consonants in utterances such as lfipmipl orIfapmap/, and lfimpipl or Ifampap/. Velar position

was lower in the environment of the low, openvowel/al, than in that of the high, closed vowel Ii!.Further, after a nasal consonant the velum wasfound to rise sooner and faster for a high vowelthan for a low one. Similar findings have been re­ported by Moll (1962), Kent et aI. (1974), andClumeck (1976). Recently, Krakow (1993) reportedstress and rate effects on movements of the velum,indicating that the velum may have an active rolein the prosodic organization for speech. It was ofinterest, then, to cast a further glance on nonseg­mental variables, such as stress prominence andutterance position, to determine whether theirglobal effect on the kinematic parameters of theraising movements of the velum is comparable tothat on lip and jaw kinematics.

In terms of interarticulator cohesion, Krakow(1989) examined the patterns of activity of thelower lip raising and of velum lowering, varyingsyllable position and word affiliation. Sheobserved strong effects of syllable structure on thevelum, but not on the lower lip. Velar movementsshowed consistent effects of syllable affiliation,and were generally amplified in syllable-finalposition. For syllable-initial Iml the end of thelower lip raising movement preceded the end ofthe velar lowering movement, whereas forsyllable-final nasals, the beginning of the lower lipraising movement preceded the end of the velarlowering movement. Thus, Krakow (1989)presented evidence of coordination between themovements of the lower lip and the loweringmovements of the velum. McClean (1973) providedcomparable observations of velar movements atjunctural boundaries using cineradiography. Todate there has been no detailed kinematic timingexamination of the raising action of the velumwith respect to the movements of otherarticulators. It is not clear whether and to whatdegree the velar raising movement is coupled tothe movements of the jaw or the lips, when allarticulators are actively involved in producingsound sequences. The present study aims toinvestigate these interarticulator timing relations.

Information about interarticulator cohesion haspotential importance for understanding correlatesof neuromotor organization for speech intheoretical questions that may be summarized asfollows. If the previously observed (Gracco & Abbs,1986; Gracco, 1988, 1994) consistency inarticulator timing among the lips and jaw is foundin movements of only that closely coupled group ofarticulators and did not include more distalarticulators such as the velum, this might suggestan articulator organization reflecting local

Articulatory Organization ofMandibular, Labial, and Velar Movements During Speech 51

constriction-producing events (i.e., oral, velar,laryngeal). Alternatively and more plausibly, ifvelar movements are found to demonstrate adegree of coupling with other articulators similarto the coupling observed between the lips and thejaw, that would suggest a control structurereflecting the functional demands of the systemfor speech production (Gracco, 1991, 1994).However, although a recent finding suggestingtightly coupled timing among the lips, jaw, andlarynx during oral opening and oral closingprovides some support for the latter position(Gracco & LOfqvist, in press), no comparablestudies extending to the velum currently exist.

In the present experiment we investigated thefunctional linkages among the jaw, the lips andthe velum, by examining the relative timing rela­tions among raising and lowering actions of thevelum, upper lip, and jaw. The hypothesis wasthat velar movement timing will be adjusted inconjunction with variations in lip and jaw move­ment timing. In sum, this study focused on thecharacteristics of velar movements within andacross syllables in comparison to concomitant lipand jaw movements, and on the degree of tempo­ral coupling among the velar and oral articulators.

Methods and ProceduresI. Subjects

The subjects were six females ranging in agefrom 25 to 50 years, with no known history ofneurologic, hearing, or speech disorder. Except forBK, the first author, all subjects were nativespeakers of English and naive to the exactpurposes of the experiment. AH, BK, CB, JP, andFE had had formal phonetic training, while LWhad no formal phonetic training, but is a fluentspeaker of a foreign language. The stimulusutterance obeys English phonotactic constraints.BK is a fluent speaker of English, and herproductions were judged appropriate byphonetically trained native speakers of English.

II. Speech stimuliThe stimulus utterance used, Imobnob/, was

selected because, for its production, the velum isexpected to reach extremes of its range ofmovement (low for the nasal consonants and highfor the stop consonant; Bell-Berti, 1973, 1976,Krakow, 1989). The velum begins rising from. alowered position for 1m! toward a high position forfbi, followed by rapid velar lowering for In/,and afinal rise for /hI. The vowelloJ was selected formaximal jaw and lip movements, as well as formaximal velar lowering since the coarticulatoryeffects of loJ On the nasal consonant result in a

lower velar position than do the coarticulatoryeffects of other vowels. That is, the velum is at alower position for 1m! in Imol than it is for 1m! inImil (Bell-Berti et al., 1979).

The utterance was spoken at a self-selectedconversational rate and subjects were instructedto place the primary stress of the utterance on thefirst syllable. The utterance was modeled for thesubjects by the experimenter. As the structure ofthe utterance was possibly conducive to placementof almost equal stress on both syllables, any of thesubjects' productions not conforming to theexperimental specifications were discarded. Thesubjects were then alerted to the required stresspattern and were asked to repeat the token.

III. InstrumentationIII.a. The VelotraceVelar movements were tracked using the

Velotrace, a device developed for this purpose byHoriguchi and Bell-Berti (1987). The Velotraceconsists of an internal lever connected to an ex­ternallever via a push-rod, which rides on a sup­port rod, and is encased in a stainless steel tube.Part of the support rod rests on the floor of thenose, and part extends outside the nose. The tip ofthe curved internal lever rides on the nasal sur­face of the velum, moving with it; the movementsof the internal lever are transmitted to the exter­nallever through the push-rod. The external leveris nearly twice as long as the internal lever, andtherefore the movements traced from the externallever are about twice as large as the actual velarmovements that they reflect. The absolute magni­tude of the Velotrace movements may vary acrossspeakers; anatomical differences among subjectsmay result in differentially optimal positioning ofthe internal lever of the Velotrace, and, conse­quently, in different absolute displacements of theexternal lever. Speakers may also differ in the ab­solute extent of velar movement. The Velotrace issensitive to even the very rapid movements ofconsecutive oral-nasal sequences.

III.b. Thejaw splintA custom fitted jaw splint was produced for each

subject. A mandibular dental impression wastaken and a plaster cast of the subject'smandibular dentition was made and used to moldthe jaw splint. An acrylic resin casing of the lowerteeth was made. The casing had two wide,stainless steel wires embedded in its outer edges.The wires were bent to exit the acrylic casingupward, at 45 degree angles, at the level of thecuspids, so as to not interfere with bilabialclosure. One of the wires was then bent in a Z­shape, close to the chin to allow monitoring of the

In the beginning and end of the experimentsustained lsI and Iml productions were obtained toview the maximum range of the velardisplacement in order to ascertain optimal andfunctional positioning of the Velotrace. Eachsubject's data Were analyzed separately.

v. Data AcquisitionThe data were recorded on a multichannel

instrumentation recorder at a speed of 3.75 inches

.jaw movement in the midsagittal plane. The otherwire was bent horizontally and cut short, in ordernot to obscure monitoring of the upper lipmovements. The jaw splint was kept in place withthe help of a commercial dental adhesive.

The experimental set-up consisted of anadjustable dental chair, enclosed in an aluminumtubular beam framework, onto which acephalostat was fixed. (For a more detaileddescription of this set-up see Kelsoet al. 1984, pp.814-815.) The Velotrace was then fastened to thestable parts of the cephalostat.

fII.c. The LEDsInfrared light emitting diodes (LEDs) were

attached to the subject's lips, jaw splint and theexternal lever of the Velotrace in the midsagittalplane, and were tracked optoelectronically using amodified Selspot system.

IV. Procedure

The Velotrace was coated with 2% Xylocaine(Lidocaine HCl, a topical anesthetic) gel, and thesubject's nasal cavity was sprayed with 4%Xylocaine spray. Even though no studiesaddressing the specific effect of topical anestheticon movements of the velum exist, studies of thenormal behavior of the larynx have revealed nodiscernible effect of topical anesthesia (Shipp,1968 and Zemlin, 1969) on laryngeal activity. (Seealso Hardcastle, 1975 for the effects of topicalanesthesia on the tongue.) The LEDs were placedon each subject at the following locations: thebridge of the nose (reference LED), the center ofthe vermilion border of the upper lip (upper lipLED), the tip of the external lever of the Velotrace(Velotrace LED), the external stable portion of thesupport rod of the Velotrace (Velotrace referenceLED), and the point on the arm of the jaw splintclosest to the subject's jaw (jaw LED).

The subjects repeated the stimulus utterance ata self-selected, comfortable speech rate following acueing tone. Each subject produced a minimum of50 tokens, as follows:

52

AH: 85,FE: 78,

BK: 69,JP: 137,

CB: 78,LW:50.

Koll ia et al.

per second for storage and subsequent analysis. Ahighly directional microphone (Sennheiser modelMKH816T) was used for audio recording duringthe experiment. After the experiment, allmovement signals were sampled from the tapeusing a laboratory computer with 12 bit resolutionat a rate of 500 Hz per channel; the acoustic signalwas low pass filtered at 4.8 kHz and sampledat 10 kHz. The horizontal and vertical positions ofthe two reference LEDs and the LEDs placedon the jaw splint and Velotrace were recordedalong with the vertical position of the upperlip. For one subject (BK) the signal from the upperlip LED could not be used because the LEDwas intermittently obscured by one of theouter steel wires of the jaw splint. As a result, noupper lip kinematics were obtained for thissubject.

The movement signals were smoothed using a42 ms triangular window. The reference channelswere subtracted from the appropriate kinematicchannels to correct for any vertical headmovement during the experiment. The resultingupper lip, jaw, and velum signals were thendifferentiated using a central difference algorithmto obtain the corresponding instantaneousvelocities, which were subsequently smoothedusing the same 42 ms triangular window.

VI. Data AnalysisThe local maxima and minima of the movement

and velocity signals for each token were markedautomatically and individually inspected. Theonset and offset of the movements were markedbased on the zero-crossing values of thecorresponding velocity signals. The coordinativetiming of the velum, jaw, and lips were alsoexamined. For these measures the time ofoccurrence for each articulator's peak velocity wasidentified for each movement phase andreferenced to a common line-up point dependingon the specific movement phase. For the first oralclosing movement timing was referenced to thevowel onset in the first syllable; for thesubsequent velar loweringloral opening theoccurrence of peak velocity was referenced to thepeak jaw closing associated with the first fbI inImabnab/; for the final oral closing the oralarticulators were referenced to the second jawopening peak velocity associated with the onset ofthe vowel in the second syllable. Shown in Figure1 are the acoustic signal for the utterance, alongwith the kinematic traces for the velum, the upperlip, the lower lip, and the jaw. The differentmovement phases are indicated by the shadedareas, and the vowel onset is marked.

Articulatory Organization ofMandibular, Labial, and Velar Movements During Speech 53

Jaw

Audio

Velum

I I200 msec

~mm

Figure 1. Acoustic signal for Imabnabl, with the corresponding velar and oral movement kinematics. Thevelopharyngeal port is open during production of the nasal consonants 1m! and InI, and is closed for the production ofthe bilabial stop fbi. For nasal sounds the velum is at a low position, while for oral sounds the velum is elevated. Thefirst vowel onset is marked (thin vertical line), and shown are the corresponding averaged kinematic traces of thevelum, the upper lip, the lower lip, and the jaw for tokens from subject AH. The shaded areas 1 and 2 indicate the firstand second closing movements, and shaded area 3 indicates the intervening opening movements.

RESULTSIn the present study, we examined the lip, jaw

and velum movement characteristics associatedwith two contiguous syllables for a group of sixsubjects.

I. Movement characteristics:Velocity-Displacement relations

La. First closingThe velar movement for the first syllable was

morphologically different from that of the lips andjaw, and different from that of the velar move­ment for the second syllable. As shown in Figure1, the velar motion for the first syllable was aunidirectional movement (raising from a low to ahigh position, from 1m! to fbi), whereas the lip andjaw movements were composed of two distinct

phases (an oral opening for the vowel, 1m! to 10/,and an oral closing for the consonant, 101 to fbi).For the jaw and the lips, each opening and closingmovement had a single associated peak velocity.However, this was not the case for some instancesof the velar closing movement. Specifically, eventhough the velum was raised from 1m! to fbi, thevelocity trace sometimes presented two distinctpeaks. For all subjects some instances of two dis­tinct velar raising movements were observed; oneassociated with the raising from the nasal to thevowel 10/, and a subsequent raising for the conso­nant fbi. These multi-step movements were incon­sistently present and generally occurred on thelonger syllables, i.e., on the slower tokens. Similarmulti-step movements were observed for velumlowering by Bell-Berti and Krakow (1991; see alsoBoyce et aI., 1990).

54 Kollia et al.

Each velar raising step had its associated peakvelocity. Since the first raising movement and itsassociated peak velocity correspond to the raisingof the velum from the 1m! to the 10/, and were notidentifiable in all of the tokens analyzed, theywere not used in the analyses presented here.Instead, the peak velocity associated with theraising of the velum from 10/ to /hI was used. Thissecond velocity peak was comparable to thevelocity peak of the jaw and lip movements for /hIclosure.

However, in these single-step velar raisingmovements the single velocity peak did not alwaysoccur at the same time in the velocity trace as thesecond velocity peak of the two-step jaw raisingmovement. Consequently, the temporal intervalcontaining the velar peak velocity was longer thanits oral counterpart, and, in fact, longer than theanalogous interval for the second velar, raisingmovement, which was invariably achieved in onestep.

The first analysis focused on lip and jaw closingand velar raising for /hI. The velocity-displacementcharacteristics of the jaw closing and velar raisingmovements in the first stressed syllable wereexamined. On the left side of Figure 2 the jawpeak closing velocity and associated maximumclosing displacement are presented for two of thesix subjects. It can be seen that the peak velocityand displacement covary systematically. The velarpeak velocity-displacement relations from thesame context are presented on the right side ofFigure 2 for the same two subjects. It is evidentthat, although the general relationship betweenvelar raising movement velocity and displacementwas comparable to that of the jaw, these variableswere not as highly correlated for the velum asthey were for the jaw. The same was true for theupper lip: movement velocity-displacementcorrelations were somewhat lower for the velumthan for the upper lip, whose kinematic relationswere comparable to the jaw's.

Jaw Velum

200 200, ~

150 ,:. 150

100 100

~... ,~ .

A ·50 • 50

---- ..u AH AHQ)00 0 0

...........0 10 20 30S 0 10 20 30

S.........

~......250u 2000...... •Q)

> •200 '. 150 •, c:.150 100 .~..100 50 1 _:4'

S6 • ,..... .. :t\;• JP50 "I. • JP0

0 10 20 30 10 20 30 40

Displacement (rom)

Figure 2. Peak closing velocity - maximum displacement correlations for the 1st closing movement (/mab/), for the jawand the velum for two of the six subjects. Peak closing velocity is plotted as a function of the displacement of theclosing movement. Movement displacement is in mm and peak velocity in mm/sec.

Articulatory Organization ofMandibular, Labial, and Velar Movements During Speech 55

The increased variability in the velocity­displacement relations observed in the velum maybe specific to the morphology of the first velarraising movement as discussed above. The trendfor more variable velar velocity-displacementrelations for the velum compared to the lip or jawwas seen in the data for all subjects. Presented inTable 1 are the velocity-displacement correlationsfor the upper lip, jaw, and velum for all subjects.The correlations for the upper lip and jaw rangedfrom r = .734 to r = .940. For velar raising, thecorrelations between peak velocity and peakdisplacement for the six subjects ranged from r =.138 to r = .773. All lip, jaw, and velar correlationswere significant at the .01 level with the exceptionof the correlations in the velar movement forsubject BK (see Table 1).

Table 1. Velocity - displacement correlations for the1st closing movement. (p<. 01)

SUBJECT VELUM JAW UPPER LIP

AH .625 .827 .734BK .138 NS .881CB .350 .770 .800FE .773 .891 .746JP .494 .748 .940LW .305 .925 .753

mean .485 .853 .815

Jaw200

150 •

~f100

50.-.. AHu

Q)0rFl

-...... 0 10 20 30SS-:>,..... 200.....u0

I-Q) 150>

100

50JP

00 10 20 30

Lb. Second closingWe next examined the correlation between the

peak velocity and displacement for the secondclosing movement of the jaw, the velum, and theupper lip. The second syllable and, hence, closingmovement differed from the first in terms of ut­terance position, phonetic context (lmabl vs. Inabl),and stress prominence. Unlike the velum move­ments of the first syllable, the kinematic profile ofthe velum for the second syllable included both alowering and a raising movement, thus renderingits morphology comparable to that of the jaw andlip. Moreover, in contrast with the first syllable,the velum raising movement in the second syllablewas always achieved in a single step movement,even though this movement too involved raising ofthe velum for two separate "targets": from 1m! tolui, and from lui to fbI. However, as a result of thesingle-step movement, only one evident velocitypeak was associated with velar raising.

Figure 3 shows the peak closing jaw andvelar velocity plotted as a function of thedisplacement of the closing movement for the fbIin the syllable (/nabl) for two subjects. As shown inthe figure and summarized for all subjects inTable 2, the velocity-displacement correlations forthe velum were higher for the secondclosing than for the first closing. All correla­tions were significant (p<.01) except as noted.

Velum200

150 •100

~50AH

00 10 20 30

200 •

11'150 . ~

100 -- .' ..50

JP•0

0 10 20 30

Displacement (mm)Figure 3. Peak closing velocity - maximum displacement correlations for the 2nd closing movement (/nab!), for the jawand the velum for two of the six subjects. Movement displacement is in mm and peak velocity in mmlsec.

56 Kollia et al.

Table 2. Velocity - displacement correlations for the Table 3. Velocity - displacement correlations for the2nd closing movement. (p<.Ol except where marked). opening movement /bna/. (p<.Ol for all correlation

coefficients. )

SUBJECT VELUM JAW UPPER LIP SUBJECT VELUM JAW UPPER LIP

AH .627 .772 .835 AH .942 .884 .462BK .308 p<.02 .916 BK .695 .906 .734CB .634 .761 .753 CB .851 .762 .781FE .885 .916 .601 FE .934 .914 .775JP .621 .857 .900 JP .891 .866 .611LW .875 .887 .341 p<.02 LW .895 .945 .565

mean .707 .865 .735 mean .885 .890 .670

The upper lip and jaw velocity-displacementcorrelations for the second oral closing movementwere comparable to those for the first closingmovement except for LW's upper lip secondclosing movement. In this respect, and unlike thevelum, the upper lip and the jaw showed similarvelocity-displacement relations for the two closingmovements, regardless of utterance position,phonetic context, or stress prominence.

I.e. Comparisons of the two closing movements.Consistent with previous studies, in all articula­

tors the closing movements for the stressed firstsyllable were larger compared to those of the sec­ond syllable. Figure 4 shows the average dis­placements for the first and second closing move­ments. For five of the six subjects, the movementdisplacements for the first syllable were signifi­cantly greater than those for the second syllablefor the jaw and the velum (p<.0l). For the upperlip, the results were less consistent. As shown inFigure 4 , the peak velocities also differed betweenthe first and second closing movements for the jawand the upper lip, and less consistently for thevelum. The jaw and upper lip closing peak veloci­ties were generally higher for the first syllable.For the velum, the trend was for the raising veloc­ity to be equal or higher in the second syllable.

J.d. Opening movementAs mentioned above, for the first syllable (lmabl)

the utterance began with the velum at a lowposition. For this reason we could examine onlyone velar lowering movement flanked by the firstand second raising movements. Figure 5 presentsplots of the peak velocity-peak displacement forthe jaw and the velum lowering movements fortwo subjects. The jaw and velum loweringmovements behave quite similarly as evidenced bythe magnitude of their respective correlations. Thepeak opening velocity-displacement correlationsfor the jaw, the velum, and the upper lip for allsubjects are presented in Table 3. All correlationswere significant (p<.0l).

To evaluate the possibility that the high peakvelocity-displacement correlations observed in thevelar lowering movement might reflect a mechani­cal effect related to the mass of the Velotracerather than an active lowering action, we obtainedthe coefficients of variation for the loweringmovement of the jaw and the velum. The reason­ing was as follows: if the mass of the Velotracewere the major determinant of the loweringmovement, then the characteristics of thatmovement (duration or velocity) would beextremely consistent, and, unlike the jaw loweringmovement, would exhibit only a small degree ofvariability. We found, however, that thecoefficients of variation (CV) for the loweringmovements of the two articulators were relativelylarge. Moreover, the duration and velocity of thevelar lowering movement was more variable thanthat of the jaw. The CV for the opening movementdisplacement ranged from 9.5 to 17.2 mm (mean13.1) for the jaw, and 13.2 to 23.4 mm (mean 18.3)for the velum; peak velocity CV's ranged from 15.5to 33.8 mm (mean 26.4) for the jaw, and 12.4 to40.3 mm (mean 27.4) for the velum. It appearsthen, that the systematicity observed in thelowering movement characteristics of the velum isnot an artifact of the presence of the transductiondevice.

I.e. Comparisons ofmovements in sequenceThe velocity-displacement relations of the two

closing movements were next compared to thevelocity-displacement relations for the interveningopening action. Regression slopes were obtainedfrom a simple least squares analysis relating thepeak velocity and the displacement for each of themovement phases for the jaw, the velum, and theupper lip. In a simple sense, the slopemeasurements can be viewed as indicators of eacharticulator's movement frequency (see also Kelsoet aI., 1985). As such, it was possible to examinethe global effects of variables such as utteranceposition, stress prominence, and movement

Articulatory Organization ofMandibular, Labial, and Velar Movements During Speech 57

direction by comparing the slope changes withinand across articulators. Table 4 presents the slopesobtained from the regression analysis. The velumshowed the most consistent pattern across subjects,with the first closing movement having the lowestfrequency, and the opening movement having thehighest. The second closing movement was fasterthan the first, but slower than the opening.

For the jaw the opposite was true, the firstclosing was the fastest, and the opening

Displacement (mm)

6-.---.,....--------......,Upper lip

movement was the slowest; the slope for thesecond closing tended to be lower than for the firstclosing, but this was not consistent acrosssubjects. For the upper lip the results were not assystematic across subjects as they were for the jawand the velum. For three of the five subjects theslopes of the closing movements were reducedfrom the first to the second syllable, while for theremaining two subjects (AH, JP) the slopesincreased in the second syllable.

Velocity (mm/sec)

160 ..,.,..--,.,.------------.Upper lip

4

2

o

120

20 "1""'=-----------.....,Jaw

15

10

5

o

• FjrstClosing MoyementIi Second Closing Movement

40 ~~-------------.....,Velum

30

20

10

oAH BK CB FE JP LW

Suqect

160 ...,---:--------r---...,

120

III

40

o

160Velum

120

III

40

0AH BK CB FE JP LW

SUbject

Figure 4. Mean fbI closing displacements and velocities of articulator movements for the two syllables of Imabnabl forthe six subjects. The darker bars indicate the values for the first syllable, and the lighter bars the values for the second.The upper lip movement values are on the top panels, the jaw values in the middle, and the velum ones on the bottompanels. For the jaw, the first syllable is Imabl and the second Inab/, while for the velum, the first syllable begins at theacoustic onset of the first vowel (lab!), and the second begins at the offset of the stop closure (i.e., onset of the nasal:/bnab!). Displacements are in nun and velocities are in mm1sec.

58 Kallia et a/.

Jaw Velum

-300 -300

-200 • -200,

"#',

-u -100Q) • • -100til........ ...S AH AHS 0 0'-' 0 10 20 30 0 10 20 30>...........u0- -])0 -300Q) /

> .) .."",

-200 -200 ~f{'• .,~

• •• ,--..-100 ,,,: ·100 ~ ..., ...• LW LW

0 00 10 20 30 0 10 J) 30

Displacement (mm)

Figure 5. Peak lowering velocity - maximum displacement correlations for the opening movement for the jaw (lbna/)and the velum (Ibn/) for two of the six subjects. Movement displacement is in mm and peak velocity in mmlsec.

II. Interarticulator timing: Jaw-Velum-Upper Lipcoordination

The second analysis focused on the relativetiming relations among the velum, the jaw, andthe upper lip for the first and second closingmovements, as well as for the intervening openingmovement.

II.a. Articulatory orderingIt is known that articulators cooperating in the

same motor task demonstrate a high degree oftemporal coupling to each other. As a first step toestablishing whether there is any degree of tem­poral coupling between the velum and the jaw, welooked at the order with which the upper lip, thelower lip, the jaw, and the velum achieve peakposition for closure. Generally the lips attainedpeak closing position first and the jaw and thevelum followed. Beyond that, however, a morespecific or stable ordering of the closure eventswas not observed. Regarding the order with whichthe same articulators arrived at peak position inthe subsequent (/bnal) opening movement, theonly consistent observation was that the jawreached peak opening position last, while the

other articulators showed no systematic orderingpattern.

II.b. Relative timing: Comparisons betweenuelar-oral and oral-oral articulator pairs

Previous studies have revealed a number of con­sistent kinematic relations among the movementsof functionally related articulators. It has beenshown, for example, that the lips and the jaw arecoupled in their relative timing for oral closing(Gracco & Abbs, 1986; Gracco, 1988; McClean,Kroll, & Loftus, 1990; cf. also DeNil & Abbs,1991). By analogy, we expected that, in thepresent study, the upper lip and jaw woulddemonstrate consistent relative timing of theirkinematic landmarks and that the velum mightassume a consistent relationship to them. Thus, inorder to explore interarticulator cohesion, weexamined the patterns of covariation exhibited inthe kinematic behavior of the three articulators.The specific variables we examined were the timesof attainment of peak velocity and peak positionfor the jaw, the upper lip, and the velum for thedifferent movement phases. The degree ofcorrelation between these variables was

Articulatory Organization ofMandibular, Labial, and Velar Movements During Speech 59

considered an indicator of interarticulatorcohesion.

lI.b.I. First closing movementFor the first fbi closing movement, the times of

arrival at peak velocity and peak position weremeasured relative to the acoustic onset of firstvowel in Imabnab/. As expected, a high degree ofcorrelation between the timing of the jaw and theupper lip for the attainment of peak velocity wasobserved (see Table 5). The relation between thecorresponding intervals for the velum and jawshowed greater variability than between theupper lip and jaw. One possibility for theapparently reduced cohesion among the velumand the jaw may relate to the particular velocitytrace of the first raising movement, as wasdiscussed in section La., earlier. A secondpossibility is that the relative timing forattainment of peak velocity for closure is not an

important coordinating variable. Instead, adifferent temporal variable for the threearticulators might demonstrate greater stabilityas an indicator of interarticulator coordination.

In order to examine this latter possibility, therelative timing of peak jaw, upper lip and velarpositions were obtained for the first closingmovement. Overall, the relative times of peakposition rather than the times of peak velocityshowed a higher correlation for the velum-jaw andthe velum-upper lip closing movements.Interestingly, this was not the case for the upperlip-jaw relations. The relative timing results usingpeak position and peak velocity are presented inTable 5. In general, it seems that the relativetiming of the closing events for the velum with thejaw and with the upper lip are related, but not asconsistently as for the oral articulators with eachother (i.e., the upper lip with the jaw).

Table 4. Slopes of the regression equation of x on y, x=Displacement and y=Peak Velocity. Regressions weresignificant at the .01 level unless marked otherwise.

Closing 1 Opening 1 Closing 2SUBJECT Velum Jaw Upper Lip Velum Jaw Upper Lip Velum Jaw Upper Lip

AH 3.7 8.8 9.7 10.4 6.4 6.5 4.4 7.9 10.5BK 1.1 NS 10.2 13.6 8.3 4.4 6.7CB 1.5 6.3 11.1 11.5 7.1 10.4 6.4 4.5 8.0FE 5.9 10.4 9.0 12.2 9.5 14.7 5.5 10.7 6.1JP 3.1 8.6 19.4 16.0 9.9 8.5 5.2 9.8 20.4LW 1.3 12.4 11.4 13.2 6.2 9.4 6.0 9.1 5.3mean 2.8 9.5 12.1 12.7 7.9 9.9 5.3 8.1 10.1

Table 5. Times ofpeak position and peak velocity for the first fbi closing movement: Correlations for Velum-Jaw andUpper Lip-Jaw. Correlations were significant at the .01 level. The fbi closing time is measuredfrom the time ofthe 1stvowel onset.

Time of peak position Time of peak velocitySUBJECT Upper Lip- Velum - Upper Upper Lip- Velum - Upper

Velum - Jaw Jaw Lip Velum- Jaw Jaw Lip

AH .85 .62 .78 .32 .97 .30BK .78 .54CB .77 .87 .71 .13 NS .94 .13 NSFE .59 .74 .69 -.01 NS .91 .03 NSJP .36 .67 .51 .43 .91 .50LW .65 .89 .73 .37 .95 .41mean .695 .781 .694 .306 .941 .284

60 Kollia et aI.

Il.b.2. Second closing movement.For the second closing movement, the referent

used for measuring the times to peak velocity andpeak position was the time of jaw peak velocity at10/ opening for Inab/. Examination of the closingmovements for the second syllable showed a highcorrelation between the times of the attainment ofpeak closing velocity for the jaw and the upper lip(see Table 6). We found that the relative timingbased on the attainment of peak position wassomewhat better for the velum~upperlip pair thanfor the velum-jaw pair.

For subjects AH, BK, and FE the relative timesof peak position for closure between velum-jawand velum-upper lip were found to be more highlycorrelated than the relative times of theirrespective peak velocities. This was also true forsubject CB in the timing relations between velumand upper lip. However, in the case of the velum­jaw relative timing the opposite was true for CB.Similarly, subject JP exhibited a higher degree ofcorrelation in the relative timing betweenarticulator peak velocities than between peakpositions for velum-jaw and for velum-upper lip.For subject LW the timing of peak position for thevelum did not correlate with either the time ofpeak position of the jaw, or with that of the upperlip. The relative timing results for peak positionand peak velocity are presented in Table 6.

Overall, it seems that the relative timing of thevelum for the second closing movement demon­strates some degree of temporal coupling with theoral articulators, particularly with the upper lip.For the second /hI closing movement, the degree of

coupling (as indicated by the correlations in thepeak velocity times) among velar-oral articulatorswas not as high as that observed among the oralarticulators (upper lip and jaw). However, the de­gree of coupling in the peak position times amongvelar-oral articulators was comparable to that ob­served among the oral articulators.

lI.b.3. Opening movementThe relative timing between the velum and the

jaw for the oral and velar opening movement inthe second syllable was examined next. The inter­vals to peak position and peak velocity were mea­sured from the peak jaw position at /hI closure inImab/. As noted earlier, velar lowering appears tobe an actively controlled gesture. Therefore, thetime of the peak velar lowering movement from /hIto In! should covary with that of the jaw loweringfrom /hI to 10/ in the Inabl context. Here, it was thetiming of peak velocity across articulators (ratherthan of peak position as in the fIrst syllable) thatresulted in higher correlations. Moreover, thevelum and jaw displayed the most consistent re­lations across the six subjects with all correlationsbeing signifIcant (p<.Ol). The correlation of thetime of peak velocity between the velum and thejaw averaged r=.61 across subjects, ranging fromr=.40 to r=.83. However, only two upper lip-velumcorrelations reached our signifIcance criterion(p<.Ol) and another approximated it (p<.02). Thetimes of peak velocity for the jaw and the upperlip were quite variable, with correlation coeffI­cients ranging from r=-.12 to r=.67 across subjects(see Table 7). Similar fIndings have been reportedpreviously (e.g., Gracco, 1988).

Table 6. Times ofpeak position and peak velocity for the second fbi closing movement (for Inabl). Correlations forVelum-Jaw, Upper Lip-Jaw and Velum-Upper Lip. fbi closing time is with respect to the time ofJaw peak velocity forIa! opening in Inabl. Correlations are significant at the .01 level, unless otherwise indicated.

Time of peak position Time of peak velocitySUBJECT Upper Lip - Velum- Upper Lip- Velum - Upper

Velum - Jaw Jaw Upper Lip Velum -Jaw Jaw Lip

AH .47 .40 .86 .35 .92 .30BK .63 .56CB .33 .54 .72 .46 .91 .44FE .29 p<.05 .61 .64 .28 p<.05 .95 .30JP .51 .70 .76 .73 .90 .82LW .13 NS .67 .24 NS .69 .73 .48mean .405 .592 .685 .531 .900 .505

Articulatory Organization ofMandibular, Labial, and Velar Movements During Speech 61

Table 7. Times ofpeak position and peak velocity for the Ibnal opening movement: Correlations for Jaw- Velum andJaw-Upper Lip. Ibnal opening re: time of Jaw peak position for Ibl closing for Imab!. p<.Ol except where markedotherwise.

Time of peak position Time of peak velocitySUBJECT Velum - Jaw Upper Lip- Velum - Velum - Jaw Upper Lip- Velum -

Jaw Upper Lip Jaw Upper Lip

AH .74 .45 .53 .83 .67 .62BK .51 .59CB .42 .06 NS .10 NS .40 .13 NS .05 NSFE .13 NS -.14 NS .12 NS .42 -.12 NS .27 p<02JP .53 .64 .56 .78 .37 .47LW .39 .40 .16 NS .46 .36 .21 NSmean .475 .306 .310 .614 .306 .340

SUMMARY AND DISCUSSIONThe purpose of this investigation was to

examine the movement characteristics of thevelum in a specific phonetic environment, acrosssyllables and across concurrently active oralarticulators, and to assess oral-velarinterarticulator cohesion. Overall, the movementand relative timing characteristics of the velumwere found to be similar to those of the jaw andthe upper lip, although some consistentdifferences Were seen in the velar raising motion.As has been previously observed for jaw and lipmovement, the correlations between peak velocityand peak displacement were consistently high.For the velum, the peak velocity-displacementcorrelations for the two raising movements werestatistically significant, but lower than thecorrelations observed in the oral articulators. Incontrast to the rais'ing'movement, the loweringmovement of the velum displayed higher velocity­displacement correlations than did thecorresponding lowering movements of the lip andjaw. In terms of interarticulator cohesion, lip andjaw timing were found to covary as in previousstudies of oral articulator coordination. In terms ofremote articulators' intergestural coheSion, oral­velar closing movements showed a differentpattern than oral-velar opening movements. Therelative timing between the velar· raisingmovements and the lip or jaw closing movementsindicated less tight coupling than either oral-oralclosing movements or oral-velar openingmovements. The velar lowering movement, whichdisplayed the most robust velocity-displacementcorrelations,also showed tight temporal couplingto the jaw lowering movement. The significance ofthese findings is discussed below.

1. Movement characteristics

1.a. Experimental considerationsBefore addressing the results of the present

study certain issues regarding the transductiontechnique and the methods used in the presentstudy should be discussed. One of the findings ofthis investigation indicates that the movementand relative timing of the velum can be consideredqualitatively similar to those of oral articulatorssuch as the lips and jaw. In contrast to previousstudies of lip and jaw movement and of relativetiming, the movement characteristics of the velumwere less consistent. One possibility accountingfor the observed difference is that the presence ofthe device (the Velotrace) modified the velarmovement patterns in a significant manner.Previous investigation of the characteristics of thedevice, however,suggests otherwise (Horiguchi &Bell-Berti, 1987). Horiguchi and Bell-Berti (1987)showed, for example, that the movements of thevelum obtained with the Velotrace are qualita­tively similar to comparable data obtained endo­scopically from various subjects using the samespeech material and that movement data obtainedfrom simultaneous Velotrace and cineradiographicrecordings are quantitatively similar. Specifically,they compared vertical velar movement measure­ments obtained cineradiographically to movementmeasurements of the tip of the internal level ofthe Velotrace. These movements were highly cor­related (1'=.90 and higher).

Another possibility is that the device, whichonly transduces velar motion in a plane, may notbe capturing the complexity of the changes invelopharyngeal port area. While linear motiondoes not allow for accurate extrapolation to areameasures due to the nonlinear relationship be-

62 Kollia et aI.

tween planar movement and changes in area, it isimprobable that measures of velar movement andarea measures are unrelated. In fact, the dis­placement of the velum and fiberoptic measures ofchanges in port size have been shown to be highlycorrelated (Zimmermann, Dalston, Brown,Folkins, Linville, & Seaver, 1987). Specifically,Zimmermann et al. (1987) found correlations ofr=.78 and r=.89 between photodetector output, in­dicating changes in port size, and displacement ofthe velum from the pharyngeal wall, measuredcineradiographically. By analogy, a single pointtransduced at the midline of the lips or jaw doesnot allow for direct information on the position ofall parts of either structure. Admittedly, reducedkinematic descriptions have certain limitations,yet results based on such simplified descriptionsmay, nonetheless, provide important informationon articulator control principles.

l.b. Context effectsDespite the observed intra- and inter-subject

variability, several generally consistent trendswere evident in the data. Perhaps one of the mostsignificant factors pertaining to velar and oral ar­ticulator differences is related to the morphologyof the observed movements. For the oral articula­tors, the pattern of movement alternated as afunction of the phonetic composition of the utter­ance. That is, the jaw started at a relatively highposition for 1m!, lowered for the first vowel, raisedfor the fbi and In!, lowered for the second voweland raised for the final fbi. As detailed in the pre­vious section, the velum started low for the 1m!and raised continuously through the first vowel,reached a peak for fbi, lowered for the In! andraised again for the secon9. vowel and the finalconsonant. For the lip and jaw movements, theoral closing motion was always achieved in a sin­gle step, and was associated with a single phoneticsegment. In contrast, the velar motion during thesame interval progressed through two contiguousphonetic units.

The movement characteristics associated withvelar raising were more variable than thoseobserved for the lip and jaw, since the velarraising movement reflected a compound motioncombining two velar gestures associated with twophonetic segments. Each segment was associatedwith a distinct velar position, with the position forthe bilabial stop being higher than the position forthe preceding vowel. This explanation isconsistent with data and interpretations offeredby previous investigators (Bell-Berti, 1976; Bell­Berti & Hirose, 1975; Bell-Berti, Baer, Harris, &Niimi, 1979; Boyce et al., 1990; Fritzell, 1969;

Kent, Carney, & Severeid, 1974; Lubker, 1968;Krakow, 1989) that position of the velum is notspecified simply in a binary manner (open vs.closed port), but in a continuous one, with theintermediate positions between maximally open(low) and maximally closed (high) beingdependent on phonetic identity. Moreover, for thevelar lowering movement associated with thesingle phonetic segment Inl, the velocity­displacement relations of the velum werecomparable to those ofthe oral articulators. Thesefindings underscore the contributions ofcontextual manipulations to investigations ofarticulatory characteristics.

Examination of effects such as utteranceposition and stress prominence on the articulators'closing movements indicated that they affectednot only lip and jaw movements, as was expected,but also velar raising movements. As shown inTable 4, the slopes of the velocity-displacementcorrelations for the velar raising movementsdiffered across syllables, apparently reflecting thedegree to which the two steps of the raisingmovement are uncovered. In the first syllable,movement durations were somewhat longer,uncovering the two raising steps and resulting inweaker velocity-displacement correlations. In thesecond syllable, the raising movement wasinvariably achieved in one step, and the velocity­displacement correlations were higher than thoseseen in the first raising movement.

Certain direction-dependent trends wereobserved in the relations between peak velocityand displacement in the velum, as well as in thejaw and the upper lip. The movement frequency,as indexed by the slope of the velocity­displacement correlation, was higher in theclosing movements than in the opening one for thejaw. The opposite was true for the velum: themovement frequency was higher in the openingmovement than in the closing ones. It is possiblethat the difference observed between the jaw andvelum trends is related to the morphology of thevelar raising movements outlined earlier. Asnoted, the velocity-displacement correlations forvelar lowering were generally higher than for thevelar raising. While it seems possible that thevelum motion was influenced by the masS of thelever of the Velotrace, subsequent analysesindicate that a mechanical interpretation of thefmdings is not tenable. The conclusion is, then,that velar lowering, like velar raising, is the resultof controlled neuromuscular action, purposefullyadjusted to the requirements of the phoneticenvironment and velar activity timing is regulated

ArtiClllatory Organization ofMandibular, Labial, and Velar Movements During Speech 63

along with other concomitant articulatory actions.This is consistent with interpretations offered byBell-Berti and Krakow (1991).

II. Interarticulator timing: Upper Lip-Jaw,Velum-Jaw

The relative timing between upper lip and jawmovements for oral closure was highly consistent.This result is similar to findings of previousstudies of lip and jaw temporal coordination(Gracco & Abbs, 1986; Gracco, 1994). Thevariations observed in vowel duration during thefirst and second syllables effected proportionalchanges in the timing of the two oral articulators.These changes, apparently aimed at maintaininga high degree of interarticulator cohesion, wereobserved only in the closing movement; theopening movement displayed no evidence of stronginterarticulator coupling for lip and jaw timing(see also Gracco, 1988; 1994). Several findingsregarding the coordination of the velum with theoral articulators are discussed below.

The relative timing relations of the velum witheither of the oral articulators (lip or jaw) weresomewhat weaker than those among the oralarticulators. As outlined earlier, for the raisingmovements of the velum, the low velar-oralinterarticulator correlations most likely reflect thedifferences in the movement pattern for thespecific phonetic configuration. Given that theraising movement was a blending of two adjacentmovements, the resultant velocity profile was notnecessarily associated with a single segment.Instead, the velocity profile reflected the combinedmovement trajectory. Perhaps as a result, thetime of peak velocity showed greater variabilitythan the time of peak position, reducing thestrength of the interarticulator correlations.Support for this explanation comes from theobservation that, for the velar raising movements,higher correlations were found for the attainmentof peak positions (rather thlln peak velocities)across the contributing articulators. In the Cllse ofpeak positions, the intervals of interest wereassociated with the same phonetic target.

Overall, the velar-oral (velum-upper lip, as wellas velum-jaw) correlations were somewhat lowerthan those found among the lip and jaw in thisand previous studies. Comparisons oftime of peakposition vs. time of peak velocity across articula­tors demonstrated certain articulator- and param­eter-specific differences. Velar-oral comparisonsshowed higher correlations for.the timing of peakposition than for the timing of the peak velocity.In contrast, for oral-oral (upper lip-jaw) C01llpar-

isons higher correlations were found in the timingof peak velocities than in the timing ofpeak posi­tions. Within the oral-oral closure framework, oneexplanation that may account for the relaxed oral­oral peak position timing relations compared tothe peak velocity timing relations is that, for localconstriction-producing events, like oral closing,the timing of lip and jaw movement toward clo­sure (better indexed by the timing of peak veloc­ity) is more critical than the timing of movementoffset (or release, better indexed by the timing ofpeak position). In that case, the time of peak ve­locity is the measure more critically associatedwith the coordination of multiple articulators. Forphonetic events such as velar-oral closure-giventhe morphological differences of the articulatorymovements-the relative timing conditions mayallow for a broader critical time frame of actionconstraint than for local events, thus providing anexplanation ofthe observed patterns.

Consistent with this notion are the results fromthe oral-velar opening interactions for Ibno/; thestop closure is followed by a nasal consonant andlow vowel requiring velum lowering and oralopening. It is reasonable to assume that the velarlowering for the nasal and the jaw lowering for thevowel should be tightly coupled in this context.For this portion of the second syllable the oral­velar interactions were indeed very strong. Thiswas further evidenced in the higher correlationsfound in the times of peak lowering velocitycompared to the times of peak lowering position.Recent findings from a study of the relative timingof the lips and larynx provide additional supportfor this notion (Gracco & LOfqvist, in press).Therefore, as interarticulator timing becomesmore or less critical, the parameters that areindicative of coupling may also adjust: peakvelocity relations may be more illustrative ofcritical or constrained timing relations while peakposition timing relations may reflect a lessconstrained coordinative coupling.

In terms of contextual influences, a possibleaccount for the somewhat relaxed peak positiontiming across velar-oral articulators is that vowelnasalization is not phonologically contrastive forany of the languages spoken by the subjects. Hadvowel nasalization been controlled as aphonologically contrastive variable, the temporalinteractions between velar and oral articulatorswould have been more highly constrained in orderto maintain the contrast.

Examining these effects across syllables, itappears that oral-oral timing decreases from thefirst to the second closing movement, to the

64 Kollia et al.

intervening opening movement, with respect topeak position and peak velocity times. The oral­velar correlation pattern in peak position wasnearly the reverse from that seen in peak velocity.Oral-velar peak velocity timing was less tightlyconstrained for the first than for the secondclosure, where the velum-lip and velum-jawrelative timing relations were more robust. Again,the reduced oral-velar peak velocity temporalcohesion for the first syllable is most likely relatedto the difference in the movement pattern for thefirst compared to the second syllable. For theopening movement, the oral-velar timingcorrelations were rather variable, but they were atleast as high as the oral-oral timing correlations.

The general direction-dependent trends ofinterarticulator cohesion found in the dataindicated that for closure events oral-oral cohesionwas higher than oral-velar cohesion. For openingactions, however, oral-velar temporal coordinationwas better than oral-oral coherence.

In summary, the present investigationdemonstrated that the kinematic characteristics ofthe velum are similar to those of the lips and thejaw. Velar velocity was scaled with velardisplacement and the relative timing of velaractions showed adherence to the actions of the lipsand the jaw. While there was a tendency for thetiming covariation in the kinematic variables forthe velar closing action to be less robust than thatobserved for the lips and jaw, the significance ofthis difference should be further explored invarying phonetic contexts.

REFERENCESBell-Berti, F. (1973). The velopharyngeal mechanism: An

electromyographic study. Unpublished doctoral dissertation, CityUniversity of New York, New York.

Bell-Berti, F. (1976). An electromyographic study ofvelopharyngeal function. Journal of Speech and Hearing Research,19,225-240.

Bell-Berti, E, & Hirose, H. (1975). Palatal actiVity in voicingdistinctions: A simultaneous fiberoptic and electromyographicstudy. Journal of Phonetics, 3, 69-74.

Bell-Berti, F. & Krakow, R. A. (1991). Anticipatory velar lowering:A coproduction account. Journal of the Acoustical Society ofAmerica, 90, 112-123.

Bell-Berti, F., Baer, T., Harris, K. S., & Niimi S. (1979).Coarticulatory effects of vowel quality on velar elevation.Phonetica, 36,187-193.

Boyce, S. E., Krakow, R. A., Bell-Berti, F., & Gelfer, C. E. (1990).Converging sources of evidence for dissecting articulatorymovements into core gestures. Journal ofPhonetics, 18, 173-188.

Clumeck, H. (1976). Patterns of soft palate movements in sixlanguages. Journal ofPhonetics, 4, 337-351.

DeNil, 1., & Abbs, J. (1991). Influence of speaking rate on upperlip, lower lip, and jaw peak velocity sequencing during bilabialclosing movements. Journal of the Acoustical Society of America,89,845-849.

Edwards, J., Beckman, M. E., & Fletcher, J. (1991). The articulatorykinematics of final lengthening. Journal of the Acoustical SocietyofAmerica, 89, 369-382.

Fritzell, B. (1969). The velopharyngeal muscles in speech: Anelectromyographic and cineradiographic study. ActaOtolaryngologica, suppi. 250.

Gay, T. J. (1968). Effect of speaking rate on diphthong formantmovements. Journal of the Acoustical Society ofAmerica, 44, 1570­1573.

Gracco, V. L. (1991). Sensorimotor mechanisms in speech motorcontrol. In H. Peters, W. Hultsijn, & C. W. Starkweather (Eds.)Speech motor control and stuttering (pp. 53-78). North HollandElsevier.

Gracco, V. L. (1994). Some organizational characteristics of speechmovement control. Journal of Speech and Hearing Research, 37, 4­27.

Gracco, V. L. (1988). Timing factors in the coordination of speechmovements. Journal ofNeuroscience, 8,4628-4639.

Gracco, V. 1.., & Abbs, J. H. (1986). Variant and invariantcharacteristics of speech movements. Experimental BrainResearch, 65, 156-166.

Gracco, V. L., & Lofqvist, A. (in press). Speech motor coordinationand control: Evidence from lip, jaw, and laryngeal movements.Journal ofNeuroscience.

Hardcastle, W. J. (1975). Some aspects of speech production undercontrolled conditions of oral anaesthesia and auditory masking.Journal ofPhonetics, 3,197-214.

Horiguchi, S., & Bell-Berti, F. (1987). The Velotrace: A device formonitoring velar position. Cleft Palate Journal, 24, 104-111.

Kelso, J. A. S., Tuller, B., Vatikiotis-Bateson, E., & Fowler, C.A.(1984). Functionally specific articulatory cooperation follOWingjaw perturbations during speech: Evidence for coordinativestructures. Journal of Experimental Psychology: Human Perceptionand Performance, 10, 812-832.

Kelso, J. A. S., Vatikiotis-Bateson, E., Saltzman, E., & Kay, B.(1985). A qualitative dynamic analysis of reiterant speechproduction: Phase portraits, kinematics, and dynamicmodeling. Journal of the Acoustical Society ofAmerica, 77,266-280.

Kent, R. D., & Netsell, R. (1971). Effects of stress contrasts oncertain articulatory parameters. Phonetica, 24, 23-44.

Kent, R. D., Carney, P. J., & Severeid, L. R. (1974). Velarmovement and timing: Evaluation of a model for binarycontrol. Journal ofSpeech and Hearing Research, 17,470-488.

Krakow, R. A. (1989). The articulatory organization in syllables: Akinematic analysis of labial and velar gestures. Unpublisheddoctoral dissertation, Yale University, New Haven, CT.

Krakow, R. A. (1993). Nonsegmental influences on the velummovement patterns: Syllables, sentences, stress, and speakingrate. In Huffman, M. K. & Krakow, R. A. (Eds.), Nasals,nasalization, and the velum (87-118). (Phonetics & Phonology V).New York: Academic Press,

Lubker, J. (1968). An electromyographic-cineradiographicinvestigation of velar function during normal speechproduction. Cleft Palate Journal, 5, 1-18.

MacNeilage, P. F. (1970). Motor control of serial ordering ofspeech. Psychological Review, 77, 182-196.

McClean, M. D., Kroll, R. M., & Loftus, S. N. (1990). Kinematicanalysis of lip closure in stutterers' fluent speech. Journal ofSpeech and Hearing Research, 33, 755-760.

McClean, M. (1973). Forward coarticulation of velar movement atmarked junctural boundaries. Journal of Speech and HearingResearch, 16,286-296.

Moll, K. L. (1962). Velopharyngeal closure of vowels. Journal ofSpeech and Hearing Research, 5, 30-77.

Munhall, K. G., & Ostry, D. J. (1985). Ultrasonic measurement oflaryngeal kinematics. In 1. R. Titze & R. C. SCherer (Eds.), Vocal

Articulatory Organization of Mandibular, Labial, and Velar Movements During Speech 65

fold physiology (pp. 145-162). Denver. Denver Center forPerforming Arts.

Munhall, K. G., Ostry, D. J., & Parush, A. (1985). Characteristics ofvelocity profiles of speech movements. Journal of ExperimentalPsychology: Human Perception and Performance, 11, 457-474.

Ohala, J. J., Hiki, S., Hubler, S., & Harshman, R. (1968).Photoelectric methods of transducing lip and jaw movementsin speech. UCLA Working papers in Phonetics, 10, 135-144.

Ostry, D. J. & Cooke, J. D. (1987). Kinematic patterns in speechand limb movements. In E. Keller and M. Gopnik (Eds.): Motorand sensory processes of language (pp. 223-235). Hillsdale, NJ:Lawrence Erlbaum Associates Inc..

Ostry, D. J., Keller, E., & Parush, A. (1983). Similarities in thecontrol of speech articulators and the limbs: Kinematics oftongue dorsum movement in speech. Journal of ExperimentalPsychology: Human Perception and Performance, 9, 622-636.

Parush, A., Ostry, D. J., & Munhall, K. G. (1983). A kinematicstudy of lingual coarticulation in VCV sequences. Journal of theAcoustical Society ofAmerica, 74,1115-1125.

Shipp, T. (1968). A technique for examination of laryngealmuscles during phonation. Proceedings of the 1st InternationalCongress of Electromyographic Kinesiology, pp. 21-26, Montreal,Canada, August 1968.

Stone, M. (1981). Evidence for a rhythm pattern in speechproduction: Observations of jaw movement. Journal ofPhonetics,9,109-120.

Vatikiotis-Bateson, E. (1988). Linguistic structure and articulatorydynamics. Bloomington, IN: Indiana University LinguisticsClub.

Vatikiotis-Bateson, E. & Kelso, J. A. S. (1993). Rhythm type andarticulatory dynamics in English, French and Japanese. Journalof Phonetics, 21, 231-265.

Vayra, M. & Fowler, c.A. (1992). Declination of supra laryngealgestures in spoken Italian. Phonetica, 49,48-60.

Zemlin, W. R. (1969). The effect of topical anesthesia on internallaryngeal behavior. Acta Otolaryngologica, 68,176-196.

Zimmermann, G., Dalston, R. M., Brown, c., Folkins, J. W.,Linville, R. N., & Seaver, E. J. (1987). Comparison ofcineradiographic and photodetection techniques for assessingvelopharyngeal function during speech. Journal of Speech andHearing Research, 30, 564-569.

FOOTNOTEt Also City University of New York Graduate Center.