HasJUn.. Laboratoria Statu. Report 00 Spe«h Ruearch1992, SR-109/110, 45-58

Analysis of Speech Movements: Practical Considerationsand Clinical Application

Vincent L. Gracco

The instrumental evaluation of speech movements is an important adjunct to theassessment and understanding of speech motor disorders. AIl the interface between thenervous system and aerodynamic modifications in the vocal tract, movement variablessuch as displacement, velocity, acceleration, and their time histories, can provide directinformation on speech motor disorders that can only be inferred from acoustic orperceptual evaluation. Impairment in various aspects of neuromotor functioning isreflected in the motion of individual articulators and their coordination, and may reflectearly signs of functional change due to disease or trauma. Within certain limits, movementanalysis can be used as an objective method for categorizing speech motor disorders andmonitoring change due to therapeutic intervention. Further, objective comparison oforofacial motor behavior during speech and nonspeech tasks may provide diagnosticinsight into underlying pathophysiological processes. A perspective on the potential utilityof speech movement analysis in the assessment, treatment, and understanding of speechmotor disorders is the focus of the present chapter. The limitations of speech movementanalysis and the need for clinically-relevant research will be presented.

INTRODUCTIONWith the increased availability of measurement

devices for transducing movements of the speecharticulators, computer software for automatedprocessing and analysis of data, and decreasedcost of computer hardware, instrumental evalua­tion of human vocal tract movements is becomingmore feasible for inclusion into the clinic. Analysisof upper and lower limb movements employingvarious instrumental tests have been used for thelast 40 years to aid in the evaluation and diagno­sis of various pathophysiological conditions and todetermine the outcome of clinical trials (see Potvin& Tourtellotte, 1985 for review). For speech,movement analysis is a potentially important ad­junct to more traditional acoustic and perceptualanalyses used routinely in the clinic. In addition,analysis of speech and nonspeech (orofacial)movements can be used to evaluate theconsequences of motor disorders that have not yet

This work was supported in part by NIDCD Grants DC·00594 and DC-00121.

45

developed to the point of significantly affecting thecommunicative process. The purpose of thepresent chapter is to outline some of the ways inwhich analysis of movement parameters andmovement patterns may be used clinically. Beforeproceeding, it may be helpful to reiterate a pointmade by Potvin and Tourtellotte (1985);

"To the extent that instrumented tests can bedeveloped for measuring fimctions, their selective usecan provide information that might not otherwise beavailable. However, investigators should be awarethat the ability to measure small differences reliablycan yield statistically significant differences that mayIlOt be of clinical importance."

In the following, the focus will be on measure­ments that may have specific functional utility interms of assessing speech production capabilities,detecting differences in neurologic function, andimproving understanding of speech motor perfor­mance. Because of the current limitation in nor­mative data and the wide range of inter- and in­trasubject variability, both qualitative and quanti­tative methods will be presented.

46 Grl2CCO

INTERPRETAnON OF MOVEMENT

The evaluation of movement can be approachedfrom a variety of perspectives. From a motorcontrol perspective, speech production is observedto be a sequential production of different vocaltract configurations that are coordinated in spaceand time and overlap to various degrees. Visualinspection of speech movements allows for aqualitative impression of overall motorfunctioning. Compare, for example, the lip andjaw movement signals presented in the left half ofFigure 1, obtained from a neurological normalsubject, with the movement signals in the righthalf of the figure, obtained from a subject withParkinson's disease (PD). Each subject isrepeating the same sentence and the scaling forthe two sets of signals is the same. Withoutknowing what is being said, and disregarding the

respective acoustic signals, it can be seen thatthere are marked differences in the two sets ofmovements. While there are some generalsimilarities in the overall movement patterns, theextent of articulator motion of both the upper lipand lower liP6aw movements for the PD subject isless than for the normal subject, consistent withthe clinical manifestations of hypokinesia.Movement velocities, displayed above and belowthe respective UL and LLJ displacements, areseverely reduced in magnitude for the PD subjectas well. Further insight can be gained into themanifestations of the disorder by evaluating theacoustic signal simultaneously with the movementsignals. The impoverished and slow movementsfrom the Parkinson's subject are accompanied by apoorly differentiated acoustic signal consistentwith the perceptual speech characteristics ofimprecise consonant production.

[

125

o mmlsec

·125

[

30

o mmlsec

-30

Velocity

Velocity

___ 15mm

Displacement

DisplacementI2.5 mm

500 msec

PO

UL

LLJ

SPEECH

Buy Bob by a Pop P Y

Figure 1. Upper lip (UL) and lower lip/jaw <LL» movemmt displacemmt and velocity from a neurologically normalsubject and a subject with Parldnaon'sdiHaae (PD). The subjects task was to repeat the u"erance "Buy bobby apoppy" at a comfortable rate and loudness with evm stress. Shown below each set of movement signals is therespective acoustic speech signal

Analysis ofSpeech Movements: Practiall Considerations and Oinical Application 47

Other qualitative observations can be madefrom movement signals that are important for athorough understanding of the sensorimotorbreakdown and functional deficits associated withparticular speech disorders. Based on previous re­search it has been shown that multiple articula­tors engaged in the production of the same sounddisplay spatial and temporal patterns that reflecttheir cooperative behavior (Gracco, 1988, 1990;Gracco & Lofqvist, 1989). Individual speechmovements generally display smooth continuousmotion characterized by a unimodal velocity pro­file (Gracco & Abbs, 1986; Munhall, Ostry, &Parush, 1985; Nelson, 1983; Ostry, Cooke, &Munhall, 1987). Breakdown in the coordinativeaction of multiple articulators, a loss in the abilityto smoothly sequence concatenated vocal tractgestures, or multiple peaks in the velocity profileassociated with a single articulatory movementare observations that reflect qualitatively on theprocesses of speech motor control. From examina­1 '')n of discrete events associated with a single

~ech or nonspeech motor task, it is also possibleto functionally evaluate the neuromotor system atthe level that reflects on the net force applied toarticulators to produce individual movements. Inorder to generate movement a certain pattern ofexcitation and inhibition is produced in the ner­vous system and directed to the lower motor neu­rons. The action potentials generated by the inputsignals result in two distinct peripheral events;electrical responses in the muscle membranesproducing EMG's, and the generation of forcesoriginating from the contractile elements of themuscles. Movement reflects the summation of netactive and passive forces with a certain time his­tory filtered through the biomechanical propertiesof the structures being moved. If the structure isat least in part inertial, the initial acceleration ofthe load will be proportional to the initial contrac­tile force. Similarly, the peak velocity of a move­ment is generally proportional to the force magni­tude integrated over the movement time.Inspection of individual Inovement patterns canprovide heuristic information regarding the neu­romotor functioning of the patient and reflect onthe mechanical characteristics of particulararticulators.

BASIC KINEMATICSIn order to objectively and quantitatively evalu­

ate speech movements a measurement frameworkis required. Any description of movement relies onthe terminology of kinematics. A complete kine­matic description of any movement, especially of

the vocal tract, is geometrically complex. For mostpurposes, the motion of bodies can be reducedfrom irregular shaped masses to points, and themotion of such points can be described with kine­matic variables. The description of point motion isanalytically complex, requiring 15 data variableswhich change over time (Winter, 1979). For clini­cal purposes, the displacement (the distance froma starting to an ending position) and velocity (thedirectional speed) are the most useful for describ­ing articulatory motion. In order to keep track ofthe changing kinematic variables and maximizetheir descriptive usefulness it is important toadopt a reference convention and a coordinate sys­tem. Motion can be described relative to somestatic articulatory position, such as lip movementrelative to a rest position. An alternative that alsoprovides spatial information is to reference themovements to an immobile anatomical structure.The most frequently used spatial coordinate sys­tem involves three perpendicular axes represent­ing the sagittal, frontal, and transverse planes.Movements of articulators can then be describedwith respect to inferior-superior (Y), anterior-pos­terior (x), and lateral-medial (z) directions, re­spectively, relative to some anatomical reference.The most important consideration for clinical useis that a convention be established, one that isconsistent with respect to the purpose of the mea­surement and reproducible within and acrosssubjects.

As mentioned, the displacement of a point on anarticulator surface and the velocity at which thearticulator moves are two important kinematicvariables fundamental to the description andevaluation of motor disorders characterized byhypokinesia (reduction in movement extent),bradykinesia (slowness in movement; reduced ve­locity), and akinesia (slowness in movement initi­ation). Shown on the left in Figure 2 is a positiontime history of a single midsagittal point on thelower lip as it moves from opening for a vowel tooral closure for /pl. Under the displacement signalis the time history of the instantaneous velocitymathematically derived from the displacementsignal. From the displayed signals, the maximumdisplacement, calculated as the distance betweenonset position and offset position associated withthe movement, and the associated peak instanta­neous velocity, are easily obtained. Additionally,the duration of the movement, defined as the timefrom onset to completion can also be obtained. Asshown in the figure, the velocity profile can be fur­ther dissected to provide information on the accel­erative and decelerative phases of the movement.

48 GrllCCO

Ignoring gravity, the accelerative phase of amovement generally reflects the increase in netforce applied to the load (articulator) due to thecontraction of the muscles. In contrast, the decel­erative phase of a movement generally reflects thedecrease in net force acting on the load due to therelaxatiQIl of the contractile process and any an­tagonistic muscle actions. Displacement and

velocity measures provide the means to describeand quantify movement and also allow some in­ference on the properties of the muscular actionsthat caused the motion. In addition to measuringthe discrete components of a movement, the fre­quency and amplitude of repeated productions canalso be calculated as illustrated on the right sideof Figure 2.

LL

Velocit

Onset

Movemenl Time

Peak Velocity

Displacement

IPeak 10 Peak I

~alleY 10 valleyl

Frequency = 1/cycle duration

Peak position - onset position

Figure 2. Representation of the displacement and velocity of a point on the lower lip associated with a single oralclosing movement for !pI (left hand portion of the figure). Shown are some of the variables to be musured (see text forfurther details). The displacement and velocity of the same point on the lower lip during repetitive opening andclosing movements associated with repetition of Ipael. From repetitive syllabI., the frequency of production can bederived as shown.

Analysis ofSpeech Movements: Practiatl Considerations and ClinicJll ApplicJltion 49

INSTRUMENTAnONPrior to presenting a protocol that we have been

using to instrumentally evaluate speech and non­speech movements, a brief discussion of themovement transduction devices and general oper­ating principles follows. Monitoring upper articu­lator movement can be accomplished using a va­riety of transduction techniques. In general, thesetechniques convert mechanical energy, repre­sented as movement of an articulator or group ofarticulators, to electrical energy, represented asan analog voltage. Many methods are available toconvert a physiological event to an electrical signaland generally involve direct or indirect variationin electrical quantities such as resistance, ca­pacitance, inductance, or the magnetic linkage be­tween coils. The four basic techniques currentlyavailable in different forms for use in the speechclinic involve strain gauge transduction, opticaltransduction (optoelectronic sensing devices),imaging (ultrasound), and electromagnetic trans­duction. The following will briefly review thetechniques and commercially available deviceswith respect to their basic principles of operation,clinical utility, and practical limitations. A moredetailed analysis can be obtained from varioussources such as Abbs and Watkin (1976), Baken(1987), and Geddes and Baker (1968).

Strain gauge transductionStrain gauges are resistive elements that are

mounted on a flexible, lightweight strip of metalanchored at one end and attached to a movingsurface on the other end. The voltage output froma gauge is proportional to the movement at theend of the mobile attachment. Strain gaugetransducers are used for monitoring externalarticulatory movements such as the lips and jaw.Initially, the technique was used in thetransduction of jaw and lip movements bySussman and Smith (1970a, b). Refinements of themethod of attachment have been reported by Abbsand Gilbert (1973) and Muller and Abbs (1979). Asignificant clinical development was reported byBarlow, Cole, and Abbs (1983) in which straingauge transducers were attached to a lightweightaluminum frame which could be mounted to asubjects head. This refinement allowed themonitoring of lip and jaw movement withoutrequiring stabilization of the subjects' head; formany neurological patients, head stabilization isan unacceptable condition. The cantilever beamscan be instrumented to sense motion in one or two(orthogonal) dimensions, although the twodimensional units and their attachments add

significantly to the overall weight and candecrease stability. The cantilever beams arecommonly attached to a point on the midsagittalplane (midpoint of the lips and chin) providinginferior-superior and anterior-posterior motionsensing. Strain gauge transducers provide acontinuous analog output that can faithfullyreproduce the fastest lip and jaw movements. Abridge amplifier is required for each direction ofmovement to supply an excitation voltage to theresistive elements and to amplify the signal priorto storage or analog-to-digital (AID) conversion.

Optical transductionThe most notable optical technique for tracking

human movement involves a position sensing de­vice and pulsed light-emitting diodes to trackpoints in a two or three dimensional coordinatesystem (Watsmart, Northern Digital, Inc., ofWaterloo, Ontario, Canada; Selspot, SelectiveElectronics, Inc., of Sweden). Devices that rely onthe sensing of LED's are limited in a similar man­ner to the strain gauge devices in that they canonly be used to monitor the external articulatorssuch as the lips and jaw. There are some photo­electric devices that rely on the sensing of lightreflection which can be used to monitor tonguemovement (Chuang & Wang, 1978; Fletcher,1982). However, such optical scanning systems fortongue motion require small LED light sourcesand photosensitive detectors arranged in an artifi­cial palate worn by the patient. In addition to thispractical limitation and the lack of commercialavailability, a distance dependent error has beenreported requiring a refinement in calibrationprocedures. (McCutcheon, Lakshminarayanan, &Fletcher, 1990). A final device, using charge cou­pled device (CCD) sensors eliminating reflectionerrors, is currently being marketed (Optotrak,Northern Digital, Inc.). Similar to the optoelectricdevices, the CCD device provides three dimen­sional information on the movement of visiblesensors .with 0.1 mm accuracy over a one cubicmeter volume. These commercial devices can alsobe purchased with customized software for analog­to-digital conversion, signal processing and auto­mated analysis. The most significant drawback tothese systems is the cost which may be as high as$50,000 to $60,000 for a complete three dimen­sional acquisition and analysis system.

ImagingThe most common imaging device having

potential clinical application is ultrasound (seeSonies, 1982 for review). An ultrasound signal is

50 GrlZCCO

passed into the body and the differential tissueproperties associated with different structurallayers provide different reflections to thegenerated sound. The ultrasound reflections are aseries of echoes that can then be detected by thetransducer. The longer the echoes take to bereflected, the further the tissue is away from thesource. Through a knowledge of the anatomy andthe different transmission times, the structureswithin the path of the ultrasound can bereconstructed. For the human vocal tract,ultrasound can be used to visualize and trackmotion of soft tissue structures such as the tongueand vocal folds. A number of research studies haveemployed ultrasound to evaluate the shape andmotion of the tongue (Sonies, Shawker, Hall, &Gerber, 1981; Stone, Morish, Sonies, & Shawker,1987; Stone, Shawker, Talbot, & Rich, 1988), themovement of the tongue dorsum during speech(Keller & Ostry, 1983), movement of the vocalfolds during devoicing (Munhall & Ostry, 1985)and tongue motion during swallowing (Stone &Shawker, 1986). While ultrasound devices arecommercially available they are costly and areoften not optimized for vocal tract use.

Electromagnetic transductionUsing alternating magnetic fields it is possible

to track point movement of small transducersplaced on the tongue, lips, velum, and jaw in themidsagittal plane. The basic device employs a si­nusoidal signal driving a transmitter coil whichproduces lines of magnetic flux. Small receivercoils, or transducers, moving through the mag­netic field are induced with a signal that is pro­portional to the effective cross-sectional area ofthe receiver coil and the flux density. If thetransmitter and receiver axes are parallel, themagnitude of the induced signal is a measure ofthe distance between the transmitter and receiver.Recently, a commercially available electromag­netic system for tracking movements of the upperarticulators has been developed and marketed un­der the name of the Articulograph AG100(Carstens Medizinelektronik, Gottingen, WestGermany). This system allows the tracking of upto five small receiver coils placed on varioussupraglottal articulatory structures in the mid­sagittal plane. The transmitter assembly is placedon the subjects head and secured in a mannersimilar to the head mounted movement systemdeveloped by Barlow et a1. (1983). Although thesystem is commercially available, developmentand refinement is continuing (see Tuller, Shao,and Kelso, 1990 for initial evaluation of system

performance). The system requires a microcom­puter to calculate the x-y positions of each trans­ducer in real time and stores the data on the com­puter disk. Software routines are provided fordata display and analysis. Cost of the system, in­cluding a microcomputer, is approximately$42,000. Other magnetic devices are commerciallyavailable to record positions and movements of themandible and the interested reader is referred toan article by Michler, Bakke, and M.ller (1987)for further information.

There are a variety of commercial devices for thetransduction of speech movements, each with cer­tain strengths and weaknesses. The optoelectricdevices are capable of three dimensional motiontracking and provide sophisticated software foranalysis; the major limitation is the cost. Theheadmounted movement system is a low cost al­ternative that can be used with children andadults. The system can be configured to allowtransduction in two dimensions although someproblems may arise due to the extra weight of thetransducer unit. Ultrasound and theArticulograph are the only devices available thatallow transduction of tongue movements. Similarto the optoelectric devices, the cost of the respec­tive equipment is high. For all devices, a certainamount of technical sophistication and a basic un­derstanding of the operating principles is re­quired. A final consideration is the transduction oflower lip and jaw movement. The movementtransduced at the lower lip is actually a combina­tion of lower lip and jaw movement. In order toevaluate the separate lower lip and jaw actionsduring speech or nonspeech movements, both thejaw and lower lip and jaw movements are ac­quired. The jaw signal is then subtracted from thelower lip/jaw signal yielding net lower lip move­ment. Using the magnetic device, a transducer coilplaced on the midpoint between the lower centralincisors, can be used as a reflection of "true- jawmotion. For the optical devices, a custom fittedjaw splint can be used with an additional lightemitting diode used to track jaw motion. While itis possible to obtain jaw movement from a sensingdevice placed on the chin, such placement may re­sult in skin movement artifact (see Kuehn, Reich,& Jordan, 1980). For most clinical applications,the combined movement of the lower lip and jawmay suffice, eliminating the need to factor out thecontributions of the two articulators.

(ijherconsiderationsOnce obtained, the data must be stored in some

form for analysis. The storage device may be an

Atullysis ofSpeech Movements: PracticJll Considerations and Clinical ApPlication 51

oscillographic recorder with a paper medium, anFM tape recorder, or the signals may be digitizeddirectly to computer disk. Data converted fromanalog to digital form requires anti-aliasingfiltering prior to conversion. The general functionof anti-aliasing is to insure that false frequenciesnot present in the original signal are not intro­duced into the digitized signal. In order to avoidaliasing, the analog signal must be filtered andthen digitized at a rate that is at least twice thecutoff frequency of the anti-aliasing filter. Theminimum sampling rate is known as the Nyquistrate and is calculated by doubling the highest fre­quency contained in the signal of interest. Sincespeech movements contain mostly low frequencies(generally below 15 Hz), the Nyquist rate could beas low as 30 Hz with the anti-aliasing (low pass)filter having a cut off frequency at 15 Hz.However, a 30 Hz sampling rate provides a poorquality time display with a point sampled only ev­ery 33.3 ms. (A movement that lasts approxi­mately 120 ms would be represented by only 4points.) In order to improve the temporal quality,also important when deriving the velocity of themovement, higher sampling rates are often used.An additional consideration is that hardwarefilters create phase delays in the signal whichvary as a function of the cut off frequency.Therefore, it is generally desirable to use an antialiasing filter with as high a cut off frequency aspossible. Once digitized, the movement signalsmay be further smoothed in software to eliminateany noise in the signal. Using digital filters timedelays can be eliminated and the signal can befiltered at a much lower frequency. Similarly,software differentiation (central difference algo­rithm) is the preferred method of obtaining firstand second derivatives since it does not introducetime distortions to the signal.

MOVEMENT ANALYSISMost movement disorders result in a reduction

in movement extent (hypokinesia), speed(bradykinesia), a slowness in initiation (akinesia),or become generally dyscoordinated. Each of theseclinical signs can be evaluated kinematically andsubsequently quantified for intrasubject compar­isons. We have recently been using a limitedspeech and oral motor inventory with subjectshaving various movement disorders focusing onmovements of the lips and jaw. Subjects are re­quested to produce syllables and nonspeech ges­tures at two rates; a comfortable (preferred) andmaximal rate. Words and sentences are also re­peated at a comfortable rate and are used for both

qualitative and quantitative examination.Nonspeech movements are used to evaluate theorofacial motor system to determine the extent ofneuromuscular involvement. It is felt that thisprotocol provides the minimal amount of informa­tion necessary to understand the functional andstructural changes accompanying many motordisorders. In the following, movement data for aportion of the protocol will be presented from twosubjects, both with PD, who have different degreesof speech motor impairment. Subject one (SI) hasminimal speech motor involvement while subjecttwo (S2) has a moderately severe dysarthria char­acterized by imprecise consonants. Motion of theupper lip and lower lipljaw were transduced usinga head mounted movement system (Barlow et al.,1983) instrumented with strain gauges aligned fortwo dimensional sensing. The head mountedframe was oriented such that inferior-superiorand anterior-posterior movements were referencedto the Frankfort plane.

An initial step in the analysis involvesexamination of some of the data in twodimensional space. Shown in Figure 3 is the pathof the jaw in x-y space, with anterior-posteriormovements represented on the x axis and inferior­superior movements represented on the y axis, fora series of speech and nonspeech opening andclosing movements. The subject produced repet­itive opening and closing movements of the lip/jawand repeated the syllable /sa! for approximately 5seconds the two rates; a comfortable (preferred)and fast (maximal) rate. A number of observationscan be made from the x-y representation. First,~he increase in speed required for the fast ratesresults in a general reduction in the movementextent for each task. Second, the extent ofmovement for /sa! repetitions is less than that foropening and closing the mouth and the /sa!repetitions are produced in the middle two thirdsof the space occupied by the opening and closingnonspeech movements. Finally, the path taken bythe jaw in both tasks and conditions is essentiallystraight and smooth. These observations from anindividual with Parkinson's disease are qualita­tively similar to those made for normal subjects.Figure 4, in contrast, displays similar dataobtained from S2. As mentioned, this subjects'speech motor skills are more severely affectedthan the previous subject. From the x-y represen­tations of the speech and nonspeech movementsit can be seen that the lip/jaw movementsare reduced in extent, less smooth, and morevariable than was observed in the previous figure(note the different scales for the two figures).

52 Gracco

10 20

Fast Rate

Inferior

Postorior Anterior Posterior Anlelior

Figure 3. Two d~ensionalmovemmt of the lower lip/jaw for repetitive productions of oral opening/closing(nonspeech) and /uel for 51 (see text). Movement directiON as indicated.

Inferior

Superior

I~ Open/Close 9ff

~6

3Inferior

5 10

Normal Rate 82 Fast Rate

Superior

Isae·sael

Posterior Anlerlor Posterior /", 'Jrior

Figure 4. Two dimensional movement of the lower lipljaw for repetitive productions of oral opening/closing(nonspeech) and lsael for 52 (see text). Movement directions as indicated.

Anillysis ofSpeech Movements: Practiall Considerations and Oinical ApPlication 53

In order to evaluate such data quantitatively thetime histories of the movements must bedisplayed. Presented in Figure 5 are examplesfrom the two subjects of continuous openingand closing inferior-superior movements(nonspeech) of the LLJ. The well defined peaksand valleys in the displacement trace provides away of automatically identifying the differentmovement phases (opening-closing) and cal­culating the displacement and frequency ofrepetition. Below each trace is the summary of asoftware routine which identifies the peaksand valleys in the displacement trace andcalculates the frequency of repetition (FO), and theaverage displacement (mm) of the sequentialmovements.

Shown in Figure 6 are the upper lip and lowerlip/jaw movements in the x and y dimensionsassociated with repeated production of the syllable/pae/. It can be seen that the upper and lower lipsmove in both a superior-inferior and anterior­posterior direction. The movements are generallysmooth and regular, and the upper lip moves lessin extent than the lower lip. Shown in the nextfigure (Figure 7) are examples from the twosubjects illustrating the results of the automated

analysis routine applied to the displacementtraces. Average movement displacement and thefrequency of production at each rate wascalculated from the inferior-superior movement ofthe lower lip/jaw. Subjects repeated the syllablesat a comfortable or preferred rate and as fast aspossible for approximately six seconds. The peaksand valleys in the displacement signals areindicated by the vertical ticks above the tracesand the summary measures were calculated asshown under each trace. From these results it can

.be seen that the lower lip/jaw movement for themore severe subject (S2) displays a smallermovement displacement compared to the lessimpaired subject (Sl) although the preferred rateof repetition is approximately equivalent (2.9 vs.2.8 Hz). At the fast rate the less severe subject(Sl) is able to increase the frequency of production(2.8 to 5.4 Hz; 93% increase) with a concomitantreduction in the movement displacement (8.0 to5.6 mm). In contrast, S2 is unable to increase thefrequency of syllable repetitions to the samedegree (2.9 to 3.5 Hz; 20% increase). In addition tomeasuring the movement displacement andfrequency, similar measures can be made on thederived velocity time histories.

LLJ

OPENING/CLOSING--PREFERRED RATE

PEAK-CENTERED CYCLESN= 5 F0 = 1.098CYCLE DURATION MEAN = 0.911CYCLE SIZE MEAN = 14.997

51

LLJ r52

PEAK-CENTERED CYCLESN = 3 F0 = 0.638CYCLE DURATION MEAN = 1.567CYCLE SIZE MEAN = 6.055

Figure 5. Opening and dosing lower lip/jaw movements in the inferioJ'oSuperior direction for St and S2. Peak centeredinformation is displayed under each trace.

54 Grllcco

ULx

1A

ULy P

LLJx

LLJy

1S

SPEECH

Ipae-pae-pae ...1

Figure 6. Upper lip and lower lip/jaw movemmt in the anterior-posterior (x) and inferior-nperior (y) directions forrepetition of the syllable /paeJ. M .hoWD, UL and LLJ movement for the opening and closing involve movement inboth x and y directions.

Ipael REPETITION

PREFERRED RATE FAST RATE

iii •• iii iii Iii i • iii iii iii' I Ii iii iii iii II

PEAK-CENTERED CYCLESN = 15 FO = 2.822CYCLE DURATION MEAN = 0.354CYCLE SIZE MEAN = 8.004

PEAK-CENTERED CYCLESN = 24 FO = 5.364CYCLE DURATION MEAN = 0.186CYCLE SIZE MEAN = 5.617

I'J',~\.J'\.'~---82

iii iii iii iii iii iii iii iii i • ii,

PEAK-CENTERED CYCLESN = 12 FO = 2.959CYCLE DURATION MEAN = 0.338CYCLE SIZE MEAN = 3.017

PEAK-CENTERED CYCLESN = 15 FO = 3.529CYCLE DURATION MEAN = 0.283CYCLE SIZE MEAN = 2.521

Figure 7. Output from an automated analysil routine that picks peaks and valley. in the diaplacemmt signal (indicatedby the vertical ticks above and below the respective signals) and calculates the number of pub (valleys), thefrequency of production (FO in Hz), mean cycle duration (period), and the mean cycle size (nun). Shown are data fromtwo subjects with different degrees of speech motor impaiDnmt secondary to PD. Subjects repeated the syllables at apreferred rate and as fast as possible.

Analysis ofSpeech Movements: Practiall Considerations and Cliniall ApplialtiDn 55

A comparison of speech and nonspeechmovement tasks is presented in the next twofigures (Figures 8 and 9). These data wereobtained from the same two subjects presented inthe previous figure. In each case the subject's taskwas to purse and retract the lips and to repeat thevowel sequence "uu"-"ee," at comfortable and fastrates. Because these movements are predom­inantly produced with anterior-posteriormovements of the lips, only the anterior-posteriormovements were measured. For Sl (Figure 8),both lips appear to be moving together (in phase)for all tasks. The consistency of the timingrelations can be easily calculated using crosscorrelation. The nonspeech task (purse-retract) isnot constrained by phonetic requirements andallows a more detailed evaluation of orofacialmobility. For this subject the nonspeech task is

accomplished by equivalent contributions of theupper and lower lips. In contrast, "uuJee"repetitions predominantly involve lower lip action.The frequency of both the speech and nonspeechtasks increase in the fast rate condition, althoughthe nonspeech tasks demonstrates a greaterdegree of change. Results from the more severelyinvolved subject (S2) are presented in Figure 9.For this subject, the rate changes are much lessnoticeable with the nonspeech task demonstratinga greater degree of impairment than was noted inthe speech task. In addition, the nonspeech taskwas apparently difficult for S2 who demonstratesslow and labored protrusion and retraction ofthe lips. There is also some indication of adyscoordination of the upper and lower lipmovements at the faster rate during the speechtask.

Purse/Retract UU/EE

ULX Preferred

Rate

LLJX Anterior

15 mm

1Frequency = 1.1 Hz Frequency = 1.5 Hz

Posterior

ULX

FastRate

LLJX

1 sec

Frequency = 3.3 Hz 81 Frequency = 2.0 Hz

Figure 8. Two different repetitive tasks involving predominantly anterior-posterior (x) motion of the UL and LLJ forS1. Shown are the position time histories for altemating and continuous puning and retracting and altemating vowelproduction "uu-ee" at preferred and maximally fast rates. Below each panel is the average frequency of production.

56 Gracco

Purse/Retract UU/EE

ULX Preferre

Rate

LLJX

Ismm Anterior

IFrequency =0.85 Hz

Poslerior

ULX

FastRate

LLJX

1 sec

82 Frequency = 1.3 Hz

Figure 9. The same repetitive tub as in Figure 7 involving predominantly anterior-posterior (x) motion of the UL andLLJ for S2. Shown are the position time histories for alternating and continuous pursing and retracting and alternatingand continuous vowel production "uu-ee" at preferred and maximally fast rates. Below each panel is the averagefrequency of production except for the purse/retrad task because of the slowness of production.

Other applicationsThere are additional applications in which

movement transduction and analysis can used inthe clinical evaluation of movement disorders.Instrumental tests can be used to provideinformation on the reaction time, speed, andvisuomotor integrative abilities of the patient. Insimple reaction time, the delay from thepresentation of an auditory or visual stimulus tothe onset of some response is measured. If theresponse involves movement to a target, such asclosing the lips, the movement time can also bemeasured. Reaction time and movement time canbe differentially affected in certain disorders suchas Parkinsonism (Evarts, Teraviiinnen, & Calne,1981) and provide a means to objectively assessakinesia and bradykinesia, respectively during anonspeech task. Tracking tests require the subjectto follow a moving target with the output of atransducer attached to one of the articulators (seeMcClean, Beukelman, & Yorkston, 1987 forapJ ation to components of the speech motor

system). Clinical applications usually involvescoring techniques which reflect the magnitude ofthe error between the target and the patientsoutput. Such tests have been useful in evaluatingataxia or characterizing the impairment inproducing smooth continuous motion of aneffector. While not directly applicable to theperceptual deficits associated with speech motordisorders, these nonspeech results may proveuseful in understanding the neurological condi­tion, aspects of which may be masked by compen­satory behavior of the patient. These novel tech­niques have been used in evaluating limb impair­ments associated with a variety of neurologicaldisorders. The reader is referred to Potvin andTourtellotte (1985) for an extensive compilation ofmeasures and references.

RESEARCH NEEDSIn attempting to provide a quantitative basis for

the evaluation of speech movements, two needsare obvious; the need for standardization and

ArJQlysis ofSpeech Movements: Practical Considerations and Clinical ApPlication 57

normative data bases. All measures that havebeen described or implicated can be used to objec­tively monitor subject performance and evaluatedisease progression or improvement due to thera­peutic intervention. However, diagnostically, suchmeasures have limited utility due to; 1) the lack ofnorms currently available, 2) solid correlationalstudies which attempt to relate kinematic charac­teristics with disease states or severity of in­volvement, and 3) technical standardization to al·low valid intersubject comparisons. However, itmay be the case that norms, while useful, mayprove to be relatively uninformative or even mis­leading due to the range of variability in the nor­mal population. This is not to suggest that norma­tive data are not necessary. Rather, it may bemore important to realize that speech movementdata should not be evaluated in isolation withoutconsidering concomitant acoustic and perceptualcharacteristics of the disordered speech as well asoverall motor and sensory performance levels.Only through a synthesis of observations can wehope to understand the communicative breakdownthat is often interleaved with a more general sen­sorimotor deficit due to damage to the nervoussystem or modifications in nervous systemoperation.

CONCLUSIONSMovement analysis is an objective and quantita­

tive method of describing the behavior of the oro­facial system during speech and nonspeech tasks.Evaluation of speech movement characteristicsverify, refine, and extend, inferences and observa­tions based on acoustic, aerodynamic, or auditoryperceptual analyses. Both speech and nonspeechmovements provide important information on theneuromotor functioning of the patient and facili­tate assessment of disease states. Further, infor­mation related to the movement impairment canbe easily assimilated by members of an interdisci­plinary rehabilitation team. Quantitatively,movement transduction provides a reliable esti­mate of motor performance and can objectivelymonitor changes in performance associated withvarious forms of therapeutic intervention orchanges in disease state. An improved under­standing of movement deficits that underlie aspecific motor disorder may lead to the develop­ment of novel treatment approaches that mightnot otherwise be considered.

REFERENCESAbbs, J. H. &: Gilbert, B. N. (1973). A strain gage transduction

system for lip and jaw motion in two dimensions: Design

criteria and calibration data. Jou"",1 of Speech lind HellringResermh, 19, 19-35. .

Abbs, J. H.... Watkin, K. L (1976). Instrumentation for the studyof speech physiology. In N. J. Lass (Ed), Contemporrzry issues inaperimnrtll1 phonetics (pp. 41-75). New York: Academic Press.

Baken, R. J. (1987). Qinicll1 merlSUTement of speech lind ooia. Boston:Uttle, Brown, ... Co.

Barlow, S. M., Cole, K. J., "', Abbs, J. H. (1983). A new head­mounted lip-jaw movement transduction system for the studyof motor speech disorders. JOUTnllI of Speech lind HearingResermh, 26, 283-288.

Chuang, C. K., ... Wang, W. S.-Y. (1978). Use of optical distancesensing to track tongue motion. Jou"",1 of Speech and HearingResermh, 21, 482-496.

Evarts, E. V., Teriviiinnen, H. T., ... Caine, D. B. (1981). Reactiontime in parkinson's disease. Bnlin, 104,167-186.

Fletcher, S. G. (1982). Seeing speech in real time. IEEE Spectrum,April, 42-45.

Geddes, L. A., ... Baker, L E. (1968). Principles ofapplied I1iomedicll1instnnrlenUtiof'I. New York: Wiley.

Gracco, V. L (1990). Characteristics of speech as a motor controlsystem. In G. Hammond (Ed), Cerebnll amtrol ofspeech and limbMm1mrents (pp. 1-28). Amsterdam: Elsevier.

Gracco, V. L (1988). Timing factors in the coordination of speechmovements. JourTIIIl ofNeuroscience, 8,462~.

Gracco, V. L., &: L6fqvist, A. (1989). Speech movementcoordination: Oral-laryngeal interactions. Journal of theAcoustiall Society ofAmerial, 86, 5114.

Gracco, V. L. &: Abbs, J. H. (1986). Variant and invariantcharacteristics of speech movements. Experimentlll BrllinResearch,65, 156-166.

Keller, E., ... Ostry, D. (1983). Computer measurement of tonguedorsum movement with pulsed echo ultrasound. Journal of theAcoustiall Society ofAmerial, 73, 1309-1315.

Kuehn, D. P., Reich, A. R., ... Jordan, J. E. (1980). Acineradiographic study of chin marker positioning:Implications for the strain gauge transduction of jawmovemenl Journal of the Acoustial Society ofAmerica, 67, 1825­1827.

McClean, M. D., BeukeIman, D. R., &: Yorkston, K. M. (1987).Speech-muscle visuomotor tracking in dysarthric andnonimpaired speakers. Jou"",1 of Speech lind Hellring Resetlrch,30,276-282.

McCutcheon, M. J., Lakshminarayanan, A. V., &: Fletcher, S. G.(1990). A comparison of gIossometric measurements of tongueposition with magnetic resonance images of the vocal tract.J0lU7III1 ofthe Acoustiall Society ofAmeriCII, 87, 5122.

Michler, L, Bakke, M., ... Meller, E. (1987). Graphic assessment ofnatural mandibular movements. JOUTnIII ofCnlniofacilll Disorders:Facitll Sf Onrl Pain, 1, 97-114-

Miiller, E. M., &: Abbs, J. H. (1979). Strain gauge transduction oflip and jaw motion in the midsagittal plane: Refinement of aprototype system. JOUTnIII ofthe AcoustiCIIl Society ofAmeriCII, 65,481-486.

Munhall, K. G., &: Ostry, D. J. (1985). Ultrasonic measurement ofIaryngeal kinematics. In I. R. Titze'" R. C. Scherer (Eds.), Vaallfold physiology: Biomeclwnics, acoustics and phonatory control (pp.145-162). Denver: Denver Center for the Performing Arts.

Munhall, K. G., Ostry, D. J., ... Parusb. A. (1985). Characteristics ofvelocity profiles of speech movements. Jou"",1 of ExperimentalPsychology: [Human Perception and Perjimnance], 11, 457-474.

Nelson, W. L (1983). Physical principles for economies of skilledmovements. BiologiClll Cybernetics, 46, 135-147.

Ostry, D. L Cooke, J. D., ... Munhall, K. G. (1987). Velocity curvesof human arm and speech movements. Experimental BnlinResearch, 68, 37-46.

58 Gracco

Potvin, A. R., &r Tourtellotte, W. W. (1985). QuantitatitleExamination of neurologic functions. Volume I: Scientific basis anddesign ofinstrvmerlted tests. Boca Raton: CRC Press.

Sonies, B. C. (1982). Oral imaging systems: A review and clinicalapplications. Tou"",1 of the National Student Speech LanguRgeHeRring Associlltion, 10,30-43.

Sonies, B. C. , Shawker, T. H., Hall, T. E., Gerber, 1. H., &rLeighton, S. B. (1981). Ultrasonic visualization of tonguemotion during speech. Journal of the Acoustical Society ofAmerica, 70, 683-686.

Stone, M., &r Shawker, T. (1986). An ultrasound examination oftongue movement during swallowing. Dysphagia, 1, 78-83.

Stone, M., Morish, 1(., Sonies, B., &: Shawker, T. (1987). Tonguecurvature: A model of shape during vowel production. FoliaPhoniatrica,39, 302-315.

Stone, M., Shawker, T., Talbot, T., &r Rich, A. (1988). Cross­sectional tongue shape during vowel production. Jou"",1 of theAt:mIStiad Society ofAmeriCII, 83, 1586-1596.

Sussman, H. M., &r Smith, K. U. (1970a). Transducer formeasuring mandibular movements. Jou"",1 of the AcousticalSoddy ofAmericA, 48, 857-858.

Sussman, H. M., &r Smith, K. U. (1970b). Transducer formeasuring lip movements during speech. Jou"",1 of theAt:mIStiad Society ofAmeriCII, 48, 858-860.

Tuller, B., Shao, S, &r Kelso, J. A. S. (1990). An evaluation of an al­ternating magnetic field device for monitoring tonguemovements. Journal of the Acoustiall Soddy ofAmerica, 88, 674­679.

Winter, D. A. (1979). Biomechanics of human mmJement. New York:Wiley.