HRskins lAborrllories SIRIus Report on Speech Research1992, SR-109/110, 13-26

Characteristics of Speech as a Motor Control System*

Vincent L. Gracco

The structural and functional organization of any biophysical system provides potentiallyimportant information on the underlying control structure. For speech, the anatomical andphysiological components of the vocal tract and the apparent functional nature of speechmotor actions suggest a characteristic control structure in which the. entire vocal tract can beviewed as the smallest functional unit. Sounds are coded as different relative vocal tractconfigurations generated from neuromuscular specifications of characteristic articulatoryactions. Sensorimotor processes are applied to the entire vocal tract to scale and sequencechanges in vocal tract states. Sensorimotor mechanisms are viewed as a means to predictivelyadjust speech motor output in the face of continuously changing peripheral conditions. Anunderlying oscillatory process is hypothesized as the basis for sequential speech movementadjustments in which a centrally-generated rhythm is modulated according to internal (task)requirements and the constantly changing configurational state of the vocal tract.

Speaking is a complex action involving anumber of levels of organization and repre­sentative processes. At a cognitive level, speakingrepresents the manipulation of abstract symbolsthrough a synthesis of associative processes ex­pressed through a sophisticated linguistic struc­ture. At a neuromotor level, at least seven articu­latory subsystems can be identified (respiratory,laryngeal, pharyngeal, lingual, velar, mandibular,and labial) which interact to produce coordinatedkinematic patterns within a complex and dynamicbiomechanical environment. At an acoustic level,characteristic patterns result from complex aero­dynamic manipulations of the vocal tract. Thecognitive, sensorimotor and acoustic processes ofspeech and their interactions are critical compo­nents to understanding this uniquely human be­havior. As the interface between the nervous sys­tem and the acoustic medium for speech produc­tion/perception, speech motor processes constitutea direct link between higher level neurophysiolog­ical processes and the resulting aerody­namic/acoustic events.

The author thanks E. V.·Bateson and C. Fowler for editorialcomments and to Y. Manning.Jones for word processing. Thewriting of this paper was supported by NIH grants DC-00121and DC-00594.

13

In the following chapter, characteristics of thespeech motor control process will be evaluatedfrom a functional perspective emphasizing thestructural and functional organization of the vocaltract and the timing characteristics associatedwith their continuous modulation. In contrast toperspectives which emphasize the large numbersof muscular/kinematic degrees of freedom, thecurrent perspective is one that assumes that theoverall vocal tract is the smallest unit of func­tional behavior. Sounds are encoded according tocharacteristic vocal tract shapes specified neuro­muscularly and modulated through sensorimotormechanisms to adapt to the constantly changingperipheral environment. Examination of thestructural components and their interaction isconsistent with this macroscopic organization asare a number of empirical observations. The func­tional organization is implemented by a limitednumber of sensorimotor control processes thatscale overall vocal tract actions spatiotemporallywithin a frequency-modulated rhythmic organiza­tion characteristic of more automatic, innate mo­tor behaviors.

Structural PropertiesIn order to describe speech from the perspective

of a motor control system, a necessary step is toidentify the components of the motor system to

14 GrtlCt:O

determine how their structural properties mayreflect on the overall functional organization. Thestructures of the vocal tract include the lungs, lar­ynx, pharynx, tongue, lips, jaw, and velum.Anatomically the vocal tract structures displayunique muscular architecture, muscular connec­tions, and muscular orientation that determinetheir potential contributions to the speech produc­tion process. For example, the orientation of themuscles of the pharynx, primarily the pharyngealconstrictors, is such that they generate a sphinc­teric action on the long axis of the vocal tract pro­ducing a change in the cross-sectional area andthe tension or compliance of the pharyngeal tis­sues. The muscles of the velum are oriented pri­marily to raise and lower the soft palate separat­ing the oral and nasal cavities. Perioral musclesare arranged such that various synergistic muscleactions result in a number of characteristicmovements such as opening and closing of the oralcavity and protruding and retracting the lips.Some of the components, such as the tongue andlarynx, can be subdivided into extrinsic and in­trinsic portions each of which appear to be in­volved in different functional actions. Intrinsictongue muscle fibers are oriented to allow finegrooving of the longitudinal axis of the tongue andtongue tip and lateral acijustments characteristicof liquid and continuant sounds. Extrinsic tonguemuscles are arranged predominantly to allowshaping of the tongue mass as well as elevation,depression and retraction of portions of thetongue. Intrinsic laryngeal muscles are arrangedto open and close the glottis reciprocally and ad­just the tension of the vibrating vocal folds, whileextrinsic laryngeal muscles are oriented to dis­place the entire laryngeal complex (thyroid carti­lage and associated intrinsic muscles and liga­ments). Generally, movements of the vocal tractcan be classified into two major categories; thosethat produce and release constrictions (valving)and those that modulate the shape or geometry ofthe vocal tract. The valving and shaping actionsare generally associated with the production ofconsonant and vowels sounds, respectively(Ohman, 1966; Perkell, 1969).

In addition to the structural arrangement of thevocal tract muscles for valving and shaping ac­tions, mechanical properties of individual vocaltract structures provide insight into the functionalorganization of the speech motor control system.The dynamic nature of the tissue load againstwhich the different vocal tract muscles contract isextremely heterogeneous. For some structuressuch as the lips and vocal folds, inertial considera-

tions are minimal, while for the jaw and respira­tory structures inertia is a significant considera­tion. The tongue and lips are soft tissue structuresthat undergo substantial viscoelastic deformationduring speech while the jaw and perhaps the lipsdisplay a degree of anisotropic tension (Lynn &Yemm, 1971). Even seemingly homogeneousstructures such as the upper and lower lips, dis­play different stiffness properties (Ho, Azar,Weinstein, & Bowley, 1982) possibly contributingto their differential movement patterns (Gracco &Abbs, 1986; Gracco, 1988; Kelso et aI., 1984).Considering the structural arrangement of the vo­cal tract, the different muscular orientations andthe vast interconnection of muscles, cartilages,and ligaments it is clear that complex biomechani­cal interactions among structures are the rule.Passive or reactive changes in the vocal tract dueto inherent mechanical coupling is a consequenceof almost any vocal tract action, with the relativesignificance varying according to the specificstructural components and conformational changeand the speed at which acijustments occur. As aresult, a single articulatory action may generateprimary as well as secondary effects throughoutthe vocal tract. The examination of individual ar­ticulatory actions are important to determinetheir contribution to the sound producing process.However, individual articulatory actions neverhave isolated effects. The combination of the vis­coelastic properties of the tissues, the differentbiomechanical properties of vocal tract structures,and the complex geometry of the vocal tract com­prise a complex biomechanical environment. Thekinematic and acoustic variability characteristicof speech production reflects in part the differen­tial filtering of neural control signals by the pe­ripheral biomechanics. Only through detailed bio­physical models of the vocal tract and considera­tions of potential biomechanical interaction asso­ciated with various phonetic environments can thecontrol principles of the speech motor control sys­tem be separated from structural or cogni­tive/linguistic influences.

Functional OrganizationIn order to characterize the speech motor control

system accurately, and pose the motor controlproblem correctly, it is important to determinehow the behavior is being regulated. That is, arethe individual sound-influencing elements beingindependently controlled or does the control struc­ture involve larger units of behavior, and if so,what is the organizational structure? For speech,the simple observation that even an isolated vowel

ChatrlCteristics ofSJ'f!t!Ch tIS II Motor Control System 15

sound requires activity in respiratory muscles,tension and adduction of the vocal folds a(ijust­menta in the compliance of the oropharyngealwalls, shaping of the tongue, positioning of thejaw, elevation of the velum, and some lip configu­ration is rather convincing evidence that speech isfunctionally organized at a level reflecting theoverall state of the vocal tract. It is the interactionof all the neuromuscular components that provideeach speech sound with its distinct character, notthe action of any single component. The often­cited fact that speech production involves over 70different muscular degrees of freedom, while per­haps anatomically factual, is a functional misrep­resentation of the motor control system organiza­tion. As early 88 the birth cry and through theearliest stages of speech development, the infantsvocalizations involve the cooperative action of res­piratory, laryngeal, and supralaryngeal muscles toproduce sounds. A similar observation can bemade for locomotion in that rhythmic steppingand other seemingly functional locomotion-likebehaviors can be elicited well before the infant

manifests upright walking (Thelen, 1985, 1986). Itappears that functional characteristics of manyhuman behaviors are present at birth or veryearly in the infants development suggesting thatthe "significant functional units of action"(Greene, 1972) may be innate properties of thenervous system. It is suggested that speech motordevelopment reflects the ability to make finer andmore varied a(ijustments of the vocal tract, not themastering of the articulatory or muscular degreesoffreedom.

As suggested above, the characteristics ofspeech 88 a motor control system include a controlstructure in which the smallest functional unit isthe entire vocal tract. Recent studies havedemonstrated examples of large scalemanipulation of vocal tract actions rather thanthe modulation of separate articulatory actions.As shown in Figure 1, movements of individualarticulators such 88 the upper lip, lower lip, andjaw demonstrate timing relations such thatadjustments in one structure are accompanied bya(ijustments in all functionally-related structures.

ULx I

LLX

JxFigure 1. Upper Up (UL), Lower Lip (LL), and Jaw 0) movemmt velocities assOCyted with the first Mp" closing inMsapapple." A. the preceding vowel duration changes, the timing of the UL, LL, and J change in a consistent andunitary manner (from Graceo, 1988). Gillbration ban are 50 mmlsec (vertical) and 100 rna (horizontal).

16 CrllCCO

The coordinative process reflects a constraint onarticulatory actions involved in the production of aspecific sound. Similar results can be observed forother more spatially remote, but functionallyrelated articulators. As shown in Figure 2, move­ments of the larynx and the lower lip demonstratea similar timing dependency for the production ofthe "f" in '"safety". In order to generate the friea­tion noise characteristic of the Iff, the glottalopening and labial constriction is appropriatelytimed. As the timing of one structure changes, thetiming of the other functionally-related articula­tory action also changes. Similarly, for movementsassociated with resonance producing vowel events,timing constraints can be observed betweenlaryngeal voicing and jaw opening associated withtongue positioning for a vowel (Figure 3). Here,the laryngeal action associated with phonationand the change in jaw posi'ming to assist thetongue in vowel production uemonstrate similarcoordinative interdependency. Some preliminaryevidence further suggests that certain physiologi­cal changes associated with the production of em-

phatic stress results in an increase in the actionsof all portions of the vocal tract rather than beingfocused on one specific articulator (Fowler, Gracco,& V.-Bateson, 1989). In the presence of a poten­tially disruptive mechanical disturbance appliedto one of the contributing articulators there is atendency for the timing of all articulators toreac:ijust (Gracco & Abbs, 1988). The timing of in­dividual articulators is apparently not adjustedsingularly but reflects a system level organization(see LOfqvist & Yoshioka, 1981; 1984; Tuller,Kelso, & Harris, 1982; for other examples). It isnot clear how general these observation are withregard to all speech sounds in all possiblecontexts. For example, the lipljaw and laryn­geallsupralaryngeal coordination observed inFigures 1 and 2 is modified when the sound is atthe beginning of a word apparently reflecting achange in the functional requirements of the task.The importance of these kinds of observations isnot the specific observable pattern but the pres­ence of characteristic patterns that are used fortime-dependent articulatory adjustments.

s a e y

100 inS

GLOTTIS

tOpen

Figure 2. Lower Up dosing and glottal opening movements for three repetitions df the word "safety." AB the lower lipclosing movement for "f'" varies, the timing of the glottal opening (devoicing) also varies (from GRCCO ~ Uifqvist,1989). Similar to Figure 1, the timing of the oral and laryngeal actions appear to be adjusted as a uniL

Chamcteristics ofSp«ch lIS II Motor Control System 17

JAW

s~'-..----._"p

Open

'--'"--

GLOTTI8'--....-

tOpen

Figure 3. Timing relations between the glottal closing and the jaw opening associated with the vowel in "sip." As theglottal opening/closing associated with the "5" and subsequent vowel varies, the jaw opening (noted by thedownward movement) also varies proportionally (from Gracco • Liifqvist, 1989).

Speech motor patterns reflect characteristicways of manipulating the vocal tract, in the pres­ence of a constant pressure source, to generaterecognizable and language-specific acousticsignals (Ohala, 1983). The process through whichsuch functional cooperation occurs has been de­scribed for many motor tasks in various contextswith the assumption that the control actions in­volve the assembly of functional units of the sys­tem organized into a larger systems known assynergies or coordinative structures (Bernstein,1967; Fowler, 1977; Fowler, Rubin, Remez, &Turvey, 1980; Gelfand, Gurfinkel, Tsetlin, & Shik,1971; Saltzman, 1979; 1986; Turvey, 1977; Kugler,Kelso, & Turvey, 1980; 1982). In keeping with theinteractive structural configuration outlined pre­viously and the apparent functional nature of thetask itself, a modification of this perspective is of­fered. Speaking appears to involve coordinativestructures (or synonymously motor programs; seeAbbs, Gracco, & Cole, 1984; Gracco, 1987)available for all characteristic vocal tract actions

associated with the sound inventory of thelanguage. It is not the case, however, that acoordinative structure or a motor program is aprocess but a set of sensorimotor specificationsidentifying the relative contribution of the vocaltract structures to the overall vocal tractconfiguration (see Abbs et al., 1984; Gracco, 1987).As such, coordinative structures may be morerigidly-specified than previously thought and thedistinction between a flexible coordinativestructure and a hard-wired motor programalgorithm may be more rhetorical than real (cf.Kelso, 1986 for discussion of differences). In thisregard, two observations are of note. When thecontribution of jaw movement is eliminated, byplacing a block between the teeth, jaw closingmuscle actions are still present (Folkins &Zimmermann, 1981). Further, in response to jawperturbation, both functionally-specific responsesand non-functional responses are observed such asupper lip muscle increases when the subjects arenot producing sounds requiring upper lip move-

18 GrlUXO

ment (Kelso Tuller, V.-Bateson, &. Fowler, 1984;Shaiman, 1989). Together, these observationsreflect on specific aspects of the speech motorcontrol process and suggest that speechproduction may rely to some degree on fixedneuromuscular specifications. The presence ofjawmuscle actions when the jaw movement iseliminated is consistent with the previoussuggestion that speech motor control is a wholisticprocess involving the entire vocal tract. Thepresence of upper lip muscle increases (albeitsmall) when the sound being produced does notinvolve the upper lip, reflects on the underlyingcontrol process. The interaction of the phasicstimulus (from the perturbation) with activatedmotoneurons will produce the functionally-specificcompensatory response. If the motoneurons areinactive, or slightly active, the phasic stimuliwould result in small increases in muscle activa­tion levels without any significant movementchanges. This is a much simpler control scheme inthat certain interactions and functionally-specificresponses are a consequence of the activation ofspecific muscles and the actual synapticinteractions of various vocal tract structures(Gracco, 1987). The advantage of this perspectiveis that certain properties of speech productionresult from the physiological organization andfocus the functional organization of the speechmotor control system on the neural coding ofspeech sounds and the characteristic sensorimotorprocesses that modulate and sequence vocal tractconfigurations.

Neural Coding of Speech Motor ActionsThe coding of speech is viewed as the process by

which overall vocal tract states are "representedand transformed by the nervous system" (seePerkel &. Bullock, 1968). This coding is similar towhat has previously been identified as the selec­tion of muscular components associated with aspecific motor act (cf. Evarts, Bizzi, Burke,DeLong, &. Thach, 1972. In the following, theselection of characteristic vocal tract states will beevaluated with respect to two components of thehypothetical specifiCation process although theactual neural coding is viewed as a single processand is only presented separately for the purpose ofclarity. AB stated previously, the actions of thevocal tract are designed to either valve the airstream for different consonant sounds or to shapethe geometry of the vocal tract for different voweland vowel-like sounds. Considering the place ofarticulation for vowels and consonants naturallyresults in categorical distinctions which are

apparent acoustically and aerodynamically(Stevens, 1972). However, rather thandichotomizing these apparently discrepantprocesses, it is suggested that valving and shapingcan be conceptualized as a single physiologicalprocess. That is, speech sounds are coded accord­ing to overall vocal tract states which include pri­mary articulatory synergies. When the appropri­ate muscles are activated, the resulting force vec­tors create characteristic actions resulting in vocaltract states which act to valve the pressure orchange the geometry without creating turbulenceproducing constrictions. It is the orientation of theactivated muscle fibers, the activation of synergis­tic and antagonistic muscles, and the fixed bound­aries of the vocal tract (the immobile maxilla) thatresult in the achievement of characteristic shapesor constriction locations; certain muscular syner­gies can only result in certain vocal tract configu­rations. For example, selection of certain upperand lower lip muscles (orbicularis oris inferior andsuperior, depressor anguli oris, mentalis, depres­sor labii inferior) will always result in the approx­imation of the upper and lower lips for "p", "b", or"m". The magnitude or timing of the individualmuscle actions may vary, but bilabial closure willalways involve the activation of upper and lowerlip muscles; otherwise bilabial closure could not beattained. Similarly, changing the focus of neuralactivation to regions representing lower lip mus­cles (orbicularis oris inferior and mentalis withprimary focus in mentalis) results in movementsconsistent with labiodental constriction for "f' and"v" achieved against the immobile maxillary in­cisors (Folkins, 1976). Different relative contribu­tions of extrinsic and intrinsic tongue muscles re­sult in various shapes and movements on thetongue tip, blade and body resulting in character­istic constrictions or shapes as a consequence.Constriction location and constriction degree areuseful categories to describe different speechsounds because they specify what is distinctive toeach phonetic segment. Control over the vocaltract configuration through the development offiner control over the neuromuscular organizationprovides a more reasonable description of thespeech acquisition process because the entire vocaltract is manipulated not just the distinctive at­tributes for each sound. The neuromotor differ­ences in consonant and vowel sounds appear to bereflected in other characteristics of the controlprocess.

One such characteristic involves the compliantstates of the vocal tract consistent with the level oftension in the tissue walls. The importance of

ChIlnu:teristics ofSpeech lIS /I Motor Control System 19

tissue compliance can be inferred from a numberof observations. A major physical differencebetween voiced and voiceless consonants is in thelevel of air pressure associated with theirproduction. Voiceless sounds are generallyproduced with higher vocal tract pressures thantheir voiced counterparts. The pressure difference,which has significant aerodynamic and acousticconsequences, results from changes in the tensionin the pharyngeal and oral cavities as well as frompressure from the lungs (MUller &. Brown, 1980).For example, subjects engaged in producingspeech while simultaneously engaged in avalsalva maneuver (forceful closing the glottisthereby eliminating the lung contribution) wereable to maintain voiced/voiceless intraoralpressure differences apparently resulting fromchanges in the overall compliance of the vocaltract walls (Brown &. McGlone, 1979). Togetherwith experimental evidence that kinematic andelectromyographic characteristics of lip and jawmovements are insufficient to differentiate voicedand voiceless sounds (Lubker &. Parris, 1970;Harris, Lysaught, &. Schvey, 1965; Fromkin,1966), it appears that a major factor in generatingvoicing and voicelessness is the specification ofoverall vocal tract compliance. Two possiblecompliant states of the vocal tract are sufficient tocategorize most speech sounds; low complianceassociated with voiceless consonants and highcompliance associated with voiced consonants andvowels. Compliant states of the vocal tract areassociated with gross changes in the activity of atleast the pharyngeal constrictors as has beenobserved (Minifie, Abbs, Tarlow &. Kwaterski,1974; Perlman, Luschei, &. DuMond, 1989) andpossibly other portions of the walls of the vocaltract (intraoral cavity). The specification of lowcompliance (resulting in high vocal tractpressures) would be associated with increasedactivity in larygneal muscles to assist in thedevoicing gesture, and high compliance (resultingin low vocal tract pressures) would be associatedwith a relaxation of the muscle activity in thepharyngeal and oral cavities to allow cavityexpansion for voiced stops and continuants (Bell­Berti & Hirose, 1973; Westbury, 1983; Perkell,1969). Certain tense vowels may result from anintermediate level of compliance (between highand low) such that voicing is maintained butoverall compliance is slightly higher than for laxvowels. It is important to note that modification incompliance is a process that produces a relativelyslow change in the state of the vocal tract, withrelaxation (high compliance) a slower process than

constriction (low compliance). Together,specification of the compliant state of the vocaltract and selection of specific muscular actions isone means by which the vocal tract states may beneurally specified.

It should be noted, however, that the coding ofspeech motor actions is viewed primarily as astatic process in which characteristic states of thevocal tract are identified prior to their actual im­plementation. Considering some dynamic proper­ties of the speech motor control system providesome insight into the manner in which differentsounds may acquire further acoustic and kine­matic distinction. For example, lip closing move­ment associated with the voiceless bilabial stop"p- is generally but not consistently associatedwith a higher velocity than the voiced bilabial "b"or "In- (Chen, 1970; Gracco, submitted; Summers,1987; Sussman, MacNeilage, &. Hanson, 1973).Lip and jaw closing movements are initiated ear­lier relative to vowel onset for voiceless "p" thanfor voiced "b- or "In" (Gracco, submitted) resultingin shorter vowel durations. One possible explana­tion is that voiceless sounds are produced at ahigher rate or frequency than their voiced coun­terparts reflecting a different underlying fre­quency specification. Movement frequency is onedimension along which different speech soundscan be generally categorized. This hypotheticalfrequency modulation can be integrated with an­other dynamic property of the control system. Notonly are closing movements generally faster for avoiceless than for a voiced consonant, but the pre­ceding opening movement has also been observedto be faster (Gracco, submitted; Summers, 1987).It appears that not only may sounds be coded as afunction of the frequency of individual vocal tractacijustments but that the functional requirementsfor specific sounds may be distributed acrossmovement cycles rather than focused on a singlemovement phase. This observation suggests theoperation of a look-ahead mechanism (Henke,1966) similar to or identical with the mechanismunderlying anticipatory coarticulation which pre­dictively acijusts vocal tract actions. Speech motorcontrol is a dynamic neuromotor process in whichoverall vocal tract compliance, the location of pri­mary valving or shaping synergies, and frequency­modulated motor commands are specified by theimmediate and future acoustic/aerodynamic re­quirements.

Invariance, Redundancy, and PrecisionBefore presenting some of the specific processes

of the speech motor control system that are used

20 GrQCQl

,0 modulate overall vocal tract organization, twoaportant 8} related issues should be addressed;iVariance and precision. The search forlvariance has a long and generally unsuccessful

nistory in investigations of speech production withthe obvious conclusion that invariance is not adirectly observable event (alternatively, theappropriate metric has not been identified). Fromthe perspective of speech 88 a motor control

'stem, a more fundamental issue is the precision'th which any quantity, variable, or vocal tractnfiguration is regulated. The presence ofJstantial acoustic, kinematic, electromyo­aphic, and aerodynamic variability suggestsat the speech motor control process operates atss than maximal precision (or within ratherroad tolerance limits). The achievement oflaracteristic vocal tract configurations or

udividual articulatory actions is accomplished bya synthesis of general activation of most vocaltract structures (setting of overall vocal tractcompliance) and focused activation of the relevantmuscular synergies. This is consistent withneurophysiological evidence demonstrated in thestudies of Kots (Kots, 1975) in which voluntarymovement is seen as a synthesis of diffuseexcitation (pretuning), a more fixed and discreteincrease in motoneuron excitability (tuning) andthe final "triggering" process. Similarly, brain p0­

tentials prior to the onset of muscle activity dis­play rather diffuse activation over multiple corti­cal areas for discrete, finger and toe movements(Boschert, Hink, & Deecke, 1983; Deecke, Scheid,& Komhuber, 1969) and involve larger regions forproduction of speech. (Curry, Peters, & Weinberg,1978; Larsen, Skinh.j, & Lassen, 1978). Onepl :msible perspective is that the nervous systemudulates the focus of primary activation but thatthis process is not punctate. That is, activationand deactivation of cortical and perhaps subcorti­cal cells involve diffuse and slow changes in acti­vation or deactivation which result in distributedtonic and phasic muscle activity. Specification ofvocal tract configurations for specific sounds mayinvolve characteristic patterns of activation andinhibition in all vocal tract muscles with only'1htly greater focus on critical articulators in­

ved in the more dominant or sound-criticalvements. In some cases muscles may betially activated just because of the proximity ofjr motoneurons to other activated motoneu-

as. One conclusion is that the neural processes.mderlying speech motor control are broadly speci­fied and that the functional speech productiongoals (and the requisite perceptual properties) are

only categorically invariant. As suggested by theapparent quantal nature of speech (Stevens,1972), as long as the articulatory patterns arewithin a certain range (have not made a categorychange), the corresponding phonetic propertieswill be perceived, with kinematic variations pro­ducing very little perceptual effect. Perhapsspeech perception and production should be ap­propriately represented as stochastic processesbased on probability statements implementedthrough an adequate but imprecise control system.Strict determinism, invariance, and precision aremost likely relegated to man-made machinesworking under rigid tolerance limits or simplifiedspecifications, not to complex biological systems.

Sensorimotor Control ProcessesSimilar to the temporal organization for speech,

spatial interactions are evident that reflectmultiarticulate manipulations to achir·vecharacteristic vocal tract states. The des ,texamples of cooperative and functionally-relev .,.itspatial interactions are observed when onearticulator, such as the lip or jaw, is disturbedduring speaking. Following the application of adynamic perturbation impeding the articulatorymovement, a compensatory adjustment isobserved in the articulator being perturbed as wellas ther functionally-related, spatially-distantar uaton (Abbs & Gracco, 1984; Folkins &Ai ,1975; Gracco & Abbe, 1988; Kelso et al.,1984; Shaiman, 1989) reflecting the presence ofafferent dependent mechanisms in the control ofspeech movements. The distributed compensatoryresponse to external perturbations is a directreflection of the overall functional organization ofthe speech motor control process and iscomparable to other sensorimotor actions observedfor other motor behaviors such as posturalacijustments (Marsden, Merton, & Morton, 1981;Nashner & Cordo, 1981; Nashner, Woollacott, &Tuma, 1979), eye-head interactions (Bizzi, Kalil, &Taglaisco, 1971; Morasso, Bizzi, & Dichgans,1973), wrist-thumb actions (Traub, Rothwell, &Marsden, 1980), and thumb-finger coordination(Cole, Gracco, & Abbs, 1984). Changing the size ofthe oral cavity with the placement of a blockbetween the teeth similarly results incompensatory changes in articulatory actionsresulting in perceptually-acceptable vowel sounds(Lindblom, Lubker, & Gay, 1979; Fowler &Turvey, 1980). It appears that the speech motorcontrol system is designed to achieve functionalbehaviors through interaction of ascendingsensory signals with descending motor commands.

OJImIcteristics 0[ Speech lIS II MotM' Control System 21

Human and nonhuman studies have shown thatsensory recepton located throughout the vocaltract are sufficient to provide a range of dynamicand static information which can be used to signalposition, speed, and location ofphysiological struc­tures on a movement to movement basis (d.Munger &: Halata, 1983; Dubner, Sessle, &: Storey,1978; Kubota, Nakamura, &: Schumacher, 1980;Landgren &: Olsson, 1982 for reviews). Studiesutilizing perturbation of speech motor output indi­cate that the rich supply of orofacial somatic sen­sory afferents have the requisite properties to in­teract with central motor operations to yield theflexible speech motor patterns associated with oralcommunication (Abbs &: Gracco, 1984; Gracco &:Abbs, 1985; Gracco &: Abba, 1988; Kelso et al.,1984). Because of the constantly changing periph­eral conditions during speaking, the absolute posi­tion of vocal tract structures can vary widely de­pending on the surrounding phonetic environ­ment. The speech motor control system apparentlyadjusts for these movement to movement varia­tions by incorporating somatic sensory informa­tion from the various muscle and mechanorecep­tors located throughout the vocal tract.Considerations outlined elsewhere (Gracco, 1987;Gracco &: Abbs, 1987) suggest that the speech mo­tor control system appears to use somatic sensoryinformation in two distinct ways; in a comparativemanner to feed back information on the attain­ment of a speech goal and to predictively parame­terize or adjust upcoming control actions.Structurally, there is strong evidence for the in­teraction of sensory information from recepton lo­cated within the vocal tract with speech motoroutput at many if not all levels of the neuraxis (d.Gracco, 1987; Gracco &: Abbs, 1987 for a summaryof the vocal tract representation in multiple corti­cal and subcortical sensory and motor regions).Further, brain stem organization, evidenced byreflex studies, demonstrate a range of complex in­teractions in which sensory input from one struc­ture such as the jaw or face is potentially able tomodify motor output from lip and tongue as wellas jaw muscles (Bratzlavsky, 1976; Dubner et al.,1978; Smith, Moore, Weber, McFarland, &: Moon,1985; Weber &: Smith, 1987). It appean that thereare multiple synaptic interactions possiblethroughout the neural system controlling the vo­cal tract, with the specific interaction dependenton how the system is actively configured.

Speech motor actions involve the activation orinactivation of various muscles of the vocal tractwhich are acijusted based on the peripheral condi­tions and the specific phonetic requirements. An

important question related to the neural represen­tation for speech is the character of the underlyingactivation process for different articulatoryactions. A number of recent studies, evaluatingthe kinematic characteristics of differentarticulators, are consistent with a singlesensorimotor process to generate a variety ofarticulatory actions. One method for evaluatingthe similarity in the underlying representation formultiple speech sounds and their associatedmovement dynamics is to compare the geometric(normalized) form of velocity profiles. A change invelocity profile shape accompanying experimentalmanipulation of phonetic context suggests achange in the movement dynamics, and byinference a change in the underlying neuralrepresentation. Conversely, a demonstration oftrajectory invariance or scalar equivalence for avariety of movements suggests that differentmovements can be produced from the sameunderlying dynamics (Atkeson & Hollerbach,1985; Hollerbach &: Flash, 1982). That is, in orderto produce movement variations appropriate toperipheral conditions and task requirements, itmay be necessary only to scale the parameters of asingle underlying dynamical relation; a muchsimpler task and, by inference, a simpler neuralprocess. For movements of the vocal folds, tongue,lips, and jaw during speech it has been shown thatchanges in movement duration and to a lesser ex·tent movement amplitude reflect a scaling of abase velocity profile (Gracco, submitted; Munhall,Ostry, &: Parush, 1985; Ostry & Cooke, 1987;Ostry, Cooke, &: Munhall, 1987; Ostry & Munhall,1985). A scalar relation across a class of speechsounds involving the same articulators main­tained for different initial conditions (differentvowel contexts) suggests that the neural represen·tation has been maximized and such a representa­tion might reflect a basic component of speechproduction. That is, all speech movements mayinvolve a simple scaling of a single characteristicdynamic (force-time) relationship (Kelso & Tuller,1984) with the kinematic variations reflecting theinfluence of biomechanical and timing specifi.cations. In addition, specification of control signalsin terms of dynamics eliminates the need to spec­ify individual movement trajectories since thepath taken by any articulator is a consequence ofthe dynamics rather than being explicitly specified(see Kelso et al., 1984; Saltzman, 1986; Saltzman&: Munhall, 1989). The scaling of individualactions appears to be another characteristicprocess that eliminates the need to store allpossible phonetic variations explicitly. Rather, the

22 Gr/lCCO

control process is a acaling of characteristic motorpatterns adjusted for endogenous conditions(speaking rate, emphasis, upcoming functionalrequirements) and the surrounding phoneticenvironment (sensorimotor adjustments). Theclassic central-peripheral, motor program-reflexperspectives have given way to more reasonableand realistic issues including when and howsensory information may be used and how thedifferent representations are coded for thegeneration of all possible speech movements.

Movement sequencing

A significant characteristic of many motor be­haviors such as speech, locomotion, chewing, andtyping is the production of sequential movements.Observations that interarticulator timing is not

disrupted following perturbation (Gracco & Abbs,1988), that speech rate can be modulated bychanges in sensory input (Gracco & Abbs, 1989),and that perturbation induces minimal changes inspeech movement duration (Gracco & Abbs, 1988;Lindblom et a1., 1987) are consistent with an un­derlying oscillatory mechanism for speech.Further, somatic sensory-induced changes in thetiming of oral closing action (due to lower lip per­turbation) is consistent with an underlyingoscillatory process (Gracco & Abbs, 1988; 1989).Qualitative observations of temporal consistencyof sequential movements are also consistent withan underlying oscillatory or rhythm generatingmechanism. Presented in Figure 4 are 24 super­imposed movements of the upper lip, lower lip,and jaw for the sentence "Buy Bobby a Poppy."

UL

LL

JAW~

Speech

6mm

B uy B 0 bb y a Po PP Y

Figure 4. Superimposed upper lip (UL), lower lip (LL) and jaw 0) movements associated with 24 repetitions of thesentence "Buy Bobby a Poppy'"; the paUems are remarkably similar displaying liule spatiotemporal variation. Onlythe acoustic signal from a single repetition is shown.

Clulracteristics ofSpeech lIS II Motor Control System 23

These repetitions were produced as part of alarger study and were produced at different timesduring the experiment. The subject produced onerepetition per breath and each repetition wasproduced at a comfortable subject-defined rate. Ascan be seen there is a consistency to therepetitions that suggests an underlying periodicityindicative of a rhythmic process. A few studies, at­tempting to address the periodicity and apparentrhythmicity of speech have demonstrated thepresence of some form of underlying frequencygenerating mechanism. Ohala (1975) recordedover 10,000 jaw movements within a 1.5 hourperiod of oral reading and was able to identifyfrequencies ranging from 2-6 Hz with significantdurational variability. Kelso et al. (1985) usingreiterant productions of the syllable "ba- or "Ina­demonstrated a rather strong periodicity at ap­proximately 5-6 Hz with minimal durational vari­ability. The findings of the Kelso et al., (1985) areconsistent with an underlying oscillatory process.In contrast, the range of frequencies found byOhala (1975) may reflect the frequencymodulation associated with the sounds of thelanguage, a factor minimized in the Kelso et al.(1985) study. The modulation of frequency,dependent on specific aerodynamic properties ofthe specific sounds and surrounding articulatoryenvironment may be a mechanism underlying thespeech movement sequencing (see also Saltzman& Munhall, 1989 for further discussion of serialdynamics). That fact that the frequency valuesreported by Kelso et al. (1985) were similar for"ba- and "Ina- suggest that vowels may be a majorfactor in determining the local periodicity.However, it is the case that the individualmovements or movement cycles are not the same;local frequencies are different depending on thephonetic context.

In addition, speech production involves many ofthe same muscles as such automatic behaviors asbreathing, chewing, sucking, and swallowing. Ithas been suggested that the mechanisms underly­ing speech may incorporate, to some degree, thesame mechanisms as more automatic motorbehaviors but adapted for the specialized functionof communication (Evarts, 1982 Gracco &: Abbs,1988; Grillner, 1982; Kelso, Tuller, & Harris,1983; Lund, Appenteng, & Seguin 1982). Fewstudies have focused specifically on the similarityof speech with more innate, rhythmic motorbehaviors (Moore, Smith, & Ringel, 1988; Ostry &Flanagan, 1989) with mixed interpretations.Recent experiments and theoretical perspectiveson the organization of central pattern generators

for rhythmic behaviors such as locomotion,respiration and mastication suggest a moreflexible conceptualization of the possiblebehavioral outputs than has previously beenenvisioned for the neural control of rhythmicbehaviors (see Cohen, Rossignol, & Grillner, 1988;Getting, 1989 for reviews). For example, in vitroresults suggest that the central pattern generatorfor respiration may more appropriately beconsidered as two separate but interrelatedfunctions; one generating the rhythm and onegenerating the motor pattern (Feldman, Smith,McCrimmon, Ellenberger, & Speck, 1988). Theimplication for other rhythmic and quasi-rhythmicbehaviors such as speech, is that each functioncan be modulated independently thus generalizingthe concept of a central pattern generator to awider range of behaviors. Recently, Patla (1988)has suggested that nonlinear conservativeoscillators are the most plausible class ofbiological oscillators to model central patterngenerators in that they provide the necessarytime-keeping function as well as independentshaping of the output (see also Kelso & Tuller,1984). The recent demonstration by Moore et al.(1988) that mandibular muscle actions for speechare fundamentally different than for chewingimggests that the patterning for each behavior isdifferent. That is, speech and chewing may sharethe same generator but have different patterningor, conversely~ rely on different generators and Ii

patterns. Conceptually and theoretically, afundamental frequency oscillator and staticnonlinear shaping function can generate a numberof complex patterns. While speculative, somecurrent CPG models have the necessarycomplexity to be tentatively applied andrigorously tested as to their appropriateness forspeech motor control.

SUMMARYFrom the present perspective, the speech motor

control system is viewed as a biophysical structurewith unique configurational characteristics. Thestructure does not constrain the systems'operation but significantly affects the observablebehavior and hence the resulting acousticmanifestations. Consideration of the structuralorganization and the potential contributions frombiomechanical interactions are suggested aspotential explanations for some speech motorvariability. Sensorimotor mechanisms wereimplicated as the means by which adjustments incharacteristic vocal tract shapes can bedynamically and predictively modified to

24 Graa:o

accommodate the changing peripheral conditions.From the perspective of the vocal tract as thecontrolled system, the consistent coordinativetiming relationships reflect the functionalmodification of all the control elements orarticulatory structures. Rather than describingsound production as the modulation or assemblyof discrete units of action, the current functionalperspective suggests that entire vocal tract actionsare modulated to regulate acoustirJaerodynamicoutput parameters. The different parameters arerealized by manipulation of the frequency of theforcing function applied uniformly to the controlelements of the system. Rather than a parametricforcing in which some parameter such as stiffnessis viewed as a regulated variable, it ishypothesized that the system is extrinsicallyforced by manipulation of the frequency of neuraloutput consistent with the spatial requirements(e.g. movement extent) of the task. The frequency­modulated neuromotor actions are then filteredthrough a complex peripheral biomechanicalenvironment resulting in elaborate kinematicpatterns. Speech motor control is viewed as ahierarchically organized control structure inwhich peripheral somatic sensory informationinteracts with central motor representations. Thecontrol scheme is viewed as hierarchical from thestandpoint that the motor adjustments areembedded within a number of levels of orga­nization reflecting the overall goal of the motoract, communication. Modifications in the controlsignals reflect the parallel processing of multiplebrain regions to scale and sequence changes inoverall vocal tract states (Gracco &; Abbs, 1987).The organizational characteristics of speech as amotor control system are fundamentally similar toother sequential motor actions and are felt toinvolve a limited number of general sensorimotorcontrol processes.

REFERENCESAbbs, J. H., &. Gracco, V. L. (1984). Control of complex motor

gestures: 0r0facW muscles resporees to loacl perturbation ofthe lip during speech.]oumIIl ofNeuropIrysiology, 51, ~723.

Abbs, J. H., Gracco, V. L., .. Cole, K. J. (19M). Control of multi­movement coordination: Sensorimotor mechanisms inspeech motor programming. fotmurl ofMotor BriImor, 16, 195­232.

Atkeson. C. G., .. Hollerbach. J. M. (1985). Kinematic features ofunrestrained WItical arm movements. ]ounwlof~, 5,2318-2330.

Bell-Berti, F., .. Hirose, H. (1913). Stop comonant voicing andpharyngeal cavity size. ]ounud of the ....coustiad Society of"-n0l, 53, 295-315.

Bernstein, N. (1967). The co-ortlirt,lticm.rtd regvlJItiort of-ts.New Yorlc Pergamon Pre8&..

Bizzi, E., Kalil. R., .. Tag1iasco, V. (1971). Eye-head coordinationin monkeys: Evidence for centrally patterned organization.Sciert«, 173, ~2-4S4.

Boschert. J., Hink. R. F., .. Deecke, L. (1983). Finger Ver'SUS toemovement-related potentials: Further evidence forsupplementary motor area (SMA) participation prior tovohmtary action. F.zperimmtlll B";II Resaudt, 52, 73-80.

Bratzlavsky, M. (1976). Human brainstem reilexes. In M. Shahani(Ed.), The motor system: Neuroplrysiology .rtd muscle mechtlllisms.New Yorlc Elsevier.

Brown, W. 5., &.: McGlone, R. E. (1979). Supraglottic air pressureduring a valsalva maneuver. In J. J. Wolf &.: D. H. Klatt, (Eds.),Sp«elt CmmnuniClltion Pllpers: 97tlt Meetillg of the ....coustiCilISocietyof....mmc., New York.

Chen,. M. (1970). Vowel length variation u a function of thevoicing of the consonant environment. PItortdiCII, 22, 129-159.

Cohen, A. H., ROIIlignol, 5., .. Grillner, S. (1988). NellTld comrol ofriryt1rmic 1IIIJmIImts in M'ttbrIlta. New York: Wiley.

Cole, K. J., Gracco, V. L., &. Abbs, J. H. (1984). Autogenic andnonautogenic sensorimotor actions in the control ofmultiarticulate hand movements. Ezptrimmtlll B,..ill Restrzrclt,56, 582-585.

Curry, S. H., Peters, J. F., &. Weinberg, H. (1978). Choice of activeelectrode site and recording montage as variables affectingCNV amplitude preceding speech. In D. A. Otto (Ed).,MultidiBtiplill/lry pnspectifltS ill twIIt-mllted b,..ill pottlltilllrtStIIrclt (pp. 275-279). Washington, DC: US EnvironmentalProtection Agency.

Deecke, L, Scheid, P., &. Kombuber, H. H. (1969). Distribution ofreadiness potential. pre-motion positivity, and motor potentialof the human cerebral cortex preceding voluntary fingermovements. ExptrinreIItIIl B,..in Restrzrclt, 7, 158-168.

Dubner, Ro, Sessle, B. J., &.: Storey, A T. (1978). The 1IeU,..1 bIlsis ofcnllllld jIciiIl function. New York: Plenum Press.

Evarts, E. V. (1982). Analogies between central motor programsfor speech and for limb movements. In S. Grillner, B. Undblom.J. Lubkei', &.: A Persson (Eds.), Sp«elt motor amtrol (pp. 19-41).0xf0I'd: Pergammon Pre8&..

Evarts, Eo V., Bizzi. E., Burke, R. E., DeLong, M., &. Thach, W. T.,(Eds.) (1972). NftlI"OBCieIIuB~It SymJlOBilllfJ Sll1II1IIIIM, Vol.6: Cmtrlll COfItrol of_t(pp. 1-170).

Feldman, J. L., Smith, J. C., McCrimmon. D. R., Ellenberger', H. H.,&. Speck, D. F. (1988). Generation of respiratory patterns inmammals. In A. H. Cohen, S. Rossignol, &. S. Grillner (Eds.),NtIl1II1 comrol of riryt/rmk _ts in tItrltl:trlltes (pp. 73-100).New Yorlc Wiley and Sons.

Folkins, J. W. (1976). Multidi1nertsionlll lower lip dispu,ammtresultillgfrum IlditIrItiort ofintlitJiduIIl UrbiIll muscles: DtwIopmmt ofII st.tic model. Doctoral diIIIlertation. University ofWashington.

FoIkins, J. W., &. Zimmermann, G. N. (1981). Jaw-muscle activityduring speech with the mandible fixed. ]0W'IIIl1 of the ....coustiCIIlSocietyof....mmc., 69, 1441-1«5.

Folkim, J. W., &. Abbs, J. H. (1975). Up and jaw motor controlduring speech:R~ to resistive loacling of the jaw.]oumIll ofSpetelt.rtd Htarirtg RaelrdI, 18,207-220.

Fowler, C. A. (1977). Timing amtrol in Ip«dI produdicm. lndianaUniversity Unguistics Oub, BloominpJn. IN.

Fowler, C. A., Rubin, P., Remez, R. E., .. Turvey, M. T. (1980).Implications for speech production of a general theory ofaction. In B. Butterworth (Ed.), Umgtlllge production (pp. 373­'10). New Yorlc Academic Pre8&..

Fowler, C. A, &. Turvey, M. T. (1980). Immediate compensation inbite-block speech. ]oumIll ofPItortdia, 37, 3Q6.326.

Fowler, C. A, Gracco, V. L., .. V.-Bat8on, E., (1989). Remote andlocal effects of stress within and among articulatory subsys­\Il!mS. ]oumIl1 ofthe ....CDUItiCIIl Society of....1IItriaI, 86(1), 5115.

OrImIcteristics 0[ Speech lIS " Motor Control System 25

Fromkin. V. A. (1966). Neur<HI\UliCUlar specification of linguisticunits. r..gwge& Sperch, 9, 1~199.

Gelfand, L M., GurfinkeI. V. 5., TlIetIin, M. L., • SWk, M. L. (1971).Some problems in the mlI1ysis of movements. In L M. Gelfand.V. S. GurfinkeI, 5. V. Jlomin. &t M. L. Tsetlin (Eds.), ModdsciftheltnIdunU-fv1IctioNIl orgrr1lizJltion cif t:erlJIin bidogial systems (pp.329-345). Cambridge, Mus: MIT PreIIi.

Getting, P. A. (1989). Emerging principles governing theoperation of neural networks. AnnlMl1 .Rellifto cifNetI1'OBCiena, U,ll1S-2<M.

Gracco, V. L. (submitted). Some organizational principles forspeech movement control. Jou17UII of Speech lind HetlringRestrmh.

Gracro, V. L. (1988). Timing factors in the coordination of speechmovements. Joumtd cifNetI1'06Cinta, 8,~

Gracco, V. L. (1987). A multilevel. control model. for speech motoractivity. In H. Peters &t W. HuIstijn (Eds.), Sp«tCh motor dyrumtiGin stuttering (pp. 57-76). Wien: Springer-Verlag.

Gracco, V. L., &t Abba, ,. H. (1985). Dynamic control of the perioralsystem during speech: Kinematic lIII81yses of autogenic andnonautogenic sensorimotor processes. founull of Neurophysics,54, .18-432.

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

Gracco, V. L., &: Abbs, ,. H. (1988). Central patterning of speechmovements. E:rperimentIII B,."in Restrlrch, 71, 5150526.

Gracco, V. L., &: Abba, J. H. (1989). SensorimolDr characteristics ofspeech motor sequences. E:rpmmentlll B,."in Restrlrch, 75, 5~598.

Gracco, V. L., &: Abbs, ,. H. (1987). Programming and executionprocesses of speech movement control: Potential neuralrorreI.ates. In E. Keller &: M. Gopnik (Eds.), Symposium on motorlind sensory 1I111p11ge processes (pp. 163-201). Hillsdale, NJ:Lawrence Erlbaum Associates,

Gracco, V. L., &: L6fqvist (1989). Speech movement coordination:Oral-laryngeal interactions. founuil of the AcoustiCIII Society cifAmmClln, 86, S11•.

Greene, P. H. (1971). Problems of organization of motor systems.In R. Rosen &: F. SneI.l (Eds.), ProgTtss in tlttomiClll biology (pp.303-338). New York: Academic Press. .

Grillner, S. (1982). P066ibie lIII810gies in the control of innatemotor acts and the production of sound in speech. In S.Gri1lner, B. Lindblom, ,. Lubkei', &: A PerIllIOI\ (Eds.), Speechmotor control (pp. 217-230). Oxford: Pergammon Press.

Harris, K. S., Lysaught, G. F., &: Schvey, M. M. (1965). Someaspects of the production of oral and nasal labial stops.ul1Ipllgund Sperch, 8, 135-1.7.

HeNce, W. L. (1966). Dytumric IIrlicuJatory model cifsp«ch productionlUling annpu.tn' smuu.tion. Unpublished doctoral dissertation,Massachusetts Institu., ofTechnology.

Ho, T. P., Az.ar, 1(., WeiJwtein. S. &: Bowley, W. W. (1982). Physicalproperties of human lips: EXperimental and theoreticalanalysis. fount.IIl cifBiomec"nU:s, 15, 859-866.

Hollerbach, ,. M. &t Flam, T. (1982). Dynamic interactionsbetween limb segments dUring planar arm movement.BiologiClIl Cybernetics, 44, 67-71.

Kelso, ,. A. S., • Tuller, B. (19a.). Converging evidence in supportof common dynamical principles for speech and movementcoordination. Amtriarn Joumtd cifPhysiology, 15, R928-R935.

Kelso, J. A. S., Tuller, B., &: Harris, K. S. (1983). A 'dynamicpattern' perspective on the control and coordination ofmovement. In P. MacNeilage (Ed), The production cifsp«dI (pp.137-173). New York: Springer-Verlag.

Kelso, ,. A. S., Tuller, B, V.-Bateson, E., &: Fowler, C. (19M).Functionally specific articulatory cooperation following jaw

perturbations during speech: Evidence for coordinativestructures. founuIl cifE:rpmmmtIJl Psychology: Htmll/n Perceptionand Perfor-nce, 10, 8U-832

Kelso, ,. A. S. (1986). Pattern formation in speech and limbmovements involving many degrees of freedom. In, H. Heuer&: C. Fromm (Eds.), Gmertltion lind modullltion ofIIction pIIHerns

(pp. lOS-US). Bertin: Springer-Verlag.Kots, Ya. M. (1975). The orgllniZAtion cif tIOluntllry movement. New

YoU: Plenum Press.Kubota, K., Nakamura, Y, &: Schumacher, G. H. (1980). Jaw

pofiition lind jaw mcmmrent. Berlin: Verlag Volk und GesundeilKugler, P. N., Kelso, J. A. S., &: Turvey, M. T. (1980). On the

concept of coordinative structures as dissipative structures: 1.Theoretical lines of convergence. In G. E. Stelmach (Ed.),Tutorillis in motor behlltlior (pp. 3-47). Amsterdam:ElsevierI North Holland.

Kugler, P. N., Kelso, ,. A. S., &: Turvey, M. T. (1982). On thecontrol and coordination of naturally developing systems. In J.A. S. Kelso &: ,. E. Clark (Eds.), The deoelopment of mcmmrentamtrolllnd coonlinIItiorl (pp. 5-78). New York: Wiley.

Landgren, S., &: Olsson, 1(. A. (1982). Oral mechanoreceptors. In S.Gri1lner, B. Undblom, J. Lubker, &: A Persson (Eds.), Speechmotor control (pp. 129-140). Oxford: Pergammon Press.

Larsen, B., Skinbsj, E., &: Lassen, N. A. (1978). Variations inregional cortical blood flow in the right and left hemispheresduring automatic speech. B,."in, 101,193-209.

Undblom, B., Lubker, J, &: Gay, T. (1979). Formant frequencies ofsome fixed-mandible vowels and a model of speech motorprogramming by predictive simulation. Jou17Ul1 ofPhonetics, 7,1.7-161.

Lindblom, B., Lubker, J., Gay, T., Lyberg, P., Branderal, P., &:Holgren, 1(. (1987). The concept of target and speech timing. InR. ChaJmon • L. Shockery (Eds.), In honor of Dse Le#tiste (pp.161-181). Dord:recht, The Netherlands: Foris Publicalions.

L6fqvist, A., &: Yoshioka, H. (1981). Interarticulator programmingin obstruent production. PhonetiCll, 38, 21-~.

L6fqvist, A., &: Yoshioka, H. (1984). Intrasegmental timing:Laryngeal-oral coordination in voicel._ consonant production.Sp«c:h CommuniaItion, 3, 27'9-289.

Lubker, ,. F.,. Parris, P. ,. (1970). Simultaneous measurements ofintraoral pressure, force of labial contact, and labiale1.ectromyographic activity during production of the stopconsonant cognates Ipl and Ib/. Jou17Ul1 of the AcoustiCIIISociety cifAmnia, 47, 625-633.

Lund. J. P., Appenteng, 1(., &: Sequin,. ,. ,. (1982). Analogies andcommon features in the speech and masticating systems.In S. GrilIner, B. Lindblom, ,. Lubker, &: A Persson(Eds.), Speech motor amtrol (pp. 231-245). Oxford: PergammonPress.

Lynn, A. M. J., &: Yemm, R. (1971). External forces required tomove the mandible of relaxed human subjects. Archiws ofOrvlBiology, 16, 1~1"7.

Marsden,. C., Merton, P., &: Morton, H. (1981). Human posturalresponses. BrtIin, 104, 513-534.

Minifie, F., Abba, ,. H., Tarlow, A., &: Kwatersld, M. (1974). EMGactivity within the pharynx during speech production. fount.IIlcifSpeech Htm1ng Restrlrch, 17, .97-504.

Morasso, P., Bizzi, E., &: Dichgans, J. (1973). Adjustments ofsaccade characteristics during head movements. ErperimentalB1IIin Restrlrch, 16, .92-500.

Moore, C. A., Smith, A., &: Ringel, R. L. (1988). Task-specificorganization of activity in human jaw muscles. fount.IIl cifSpeechHaring Restrlrch, 31, 670-680.

Munger, B. L., &: Halata, Z. (1983). The sensory innervation ofprimate facial skin. I. Hairy skin. Bnlin Restrlrch Rer1iews, 5, 45­SO.

26 GrQCQl

MillIer, E. M., 6: Brown, W. S. (1980). Vuiations in thesupraglottal air pre8UJ'e waveform and their articulatoryinterpretation. In N. L.- (EeL), Sp«dt and 1IIIIpage.4.mmea illIuic-mtand"mite (Vol 4.pp. 317-389). Academic~

Munhall. K. C., D. J. 0Itry, 6: Parush. A (198S). Charac:teriatics oivelocity profiles of speech movements. Jau1'ul uf EzpmmentalPsydtology: HlmWn Perceptioft and PerJor-rta, 11, 45747•.

Nashner, L. M., 6: Cordo, P. J. (1981). Relation of automaticpostural resporwes and racaon-time voluntary movements oihuman leg muscles. E:qInimentll1 Brain Raardt, 43, 395-4ai.

Nashner, L. M., Woollacott. M., 6: TUllIA. C. (1979). Organizationof rapid resporwes to postural and loccmotor-lilte perturbationof standing man.~ Brain ResalrdI,36, 463-476.

0ha1a, J. J. (1975). The temporal regulation oi speech. In C. Fant &rM. A A. Tatham, (Eds.), AJUlitory mIIlysis and per«ptiort ufsp«ch (pp. .a1~). London: ACIldemic, .

Ohala. J. J. (1983). The origin of sound patterns in vocal tractconstraints. In P. F. MacNeilage, (EeL), TIte production ufspeteh(pp.189-216). New York: Springer-Verlag.

Ohman, S. E. C. (1966). Coarticulation in VCV utterances:Spectrographic measurements. JourruIl of tlte Acoustiall Society ufAmmOl,39,151-168.

Ostry, D. J., &r Flanagan. J. R. (1989). Human jaw movement inmastication and speech. ArcJritll!S ofOral Biology, 34, 685-693.

Ostry, D. J., &r J. D. Cooke (1987). Kinematic patterns in speechand limb movements. In,. E. Keller &r M. Copnik, (Eds.), Motorlind sensory processes uf /t",gwtge (pp. 223-235). Hillsdale, NJ:Lawrence Erlbaum.

Ostry, D. J., Cooke, J. D., &r Munhall, K. C. (1987). Velocity curvesof human arm and speech movements. F.%perimnatlll BI'lIinResalrch,74, 1-10.

Ostry, D. J., &r Munhall, K. C. (1985). Control oi rate and durationoispeech movements. Jourul oftlte AtDUStial Soddy ufAmerictI,77,~.

Patla, A. E. (1988). Analytic approaches to the study oi outputs&om central pattern generators. In A. H. Cohen. S., Ro&&ignol,&r Crillner, S. (Eds.), Ne1awJ COfItrol uf rlrytlmric IIIOMIteItts intJertebrrIUs (pp.~).New York: Wiley.

Perlcell, J. 5. (1969). Physiology oi speech production: Resulls andimplications oi a quantitative cineradiographic study <R-rchMonograph No. 53). Cambridge, MA: MIT Prell&.

Perlce1, D. H., &r Bulloc1c, T. H. (1968). Neural coding. Neut'C16dencesResalrch Progm;s Bulletin, 6(3), 223-3K

Perlman, A, Luschei, E., &r DuMond, C. E. (1989). Electricalactivity from the superior pharyngeal constrictor duringreflexive and nonreftexive tasks. Jau1'ul uf Sp«dI and HearingResalrch, 32, 7~7S'.

Saltzman, E. L. (1979). Levels of sensorimotor representation.]ountIIl ufMiltItemlltiaI/ Psycltology, 20, 91-163.

Saltzman, E. L. (1986). TlIlIk dynamic coordination of the speecharticulators: a preliminary model. In H. Heuer &r C. Fromm

(Eds.), GclsmltiorI and moduliltion ufactiort fMlterns (pp. 129-1+&).Berlin: Springer-Verlag.

Saltzman,. E. L. &r Munhall, K. (1989). A dynamical approach togestural pattenUIg in speech production. EcdogiOlI Psydtology,1,333-382.

Shaiman. S. (1989). Kinematic and electromyographic responsesto perturbation oi the jaw. ]ounlld of the Acoustictll Society ufAmmaz, 86, 78-87.

Smith, A., Moore, C. A., Weber, C. M., Mcfarland, D. H., &r Moon,J. B. (1985). Re6ex responses of the human jaw-cl08ing systemdepend on the locus of intraoral mechanical stimulation.E.rpmmentll1 Neurology, 90, ~9-509.

Stevens, K. N. (1972). The quantal nature of speech: Evidencefrom articulatory-acoustic data. In E. E. David &r P. B. Denes(Eds.), HllmIIn anrrmtIIIiaIticm: A unified tIiew (pp. 51-66). NewYork: McCraw-Hill.

Summers, W. V. (1987). Effects of stress and final-consonantvoicing on vowel production: Articulatory and acousticanalyses. ]oumal uftlte AcoustictIl Society ofA7MricII, 82, 847-863.

SuIlsman,. H. Mo, MacNeilage, P. F., &r Hanson, R. J. (1973). Labialand mandibular dynamiCS during the production of bilabialconsonants: Preliminary observations. ]ourtttIl of Speech lindHaring ResalrdI, 16,397-420.

Thelen, E. (1985). Developmental origins of motor coordination:Leg movement in human infants. IXMopmentlll Psychology, 18,1-22.

Thelen, E. (1986). Treadmill-elicited stepping in seven-month-oldinfants. Child Dtmopmmt, 57, 1.98-1506.

Traub, M. M., Rothwell, J. C., &r Marsden, C. D. (1980). A grabre6ex in the human hand. Brain, 103, 869-884.

Tuller, B., • Kelso, J. A. S. (1984). The timing of articulatorygestures: Evidence for relational invuiants. ]ourtttll of theAautial1 Society ofAmmaz, 76, 1~1<B6.

Tuller, B., Kelso, J. A. 5., &r Harris, K. S. (1982). Interarticulatorphasing as an index oi temporal regularity in speech. Jau1'ul ofE.rpmmentll1 Psycltology, 8,~m.

Turvey, M. T. (1977). Preliminaries to a theory of action withreference to vision. In R. Shaw &r J. Bransford (Ed.), PerceitMg,IICtiIIg and latowing: T_rrl.n eeologiaI1 psychology (pp.211-265).HilI&dale, NJ: Lawrence Erlbaum AIBOCiates, Inc.

Weber, C. M., &r Smith, A. (1987). Reflex responses in human jaw,lip, and tongue muscles elicited by mechanical stimulation.]olmr4l ufSp«clt and Hellring ResalrdI, 30, 70-79.

Westbury, J. R. (1983). Enlargement oi the supraglottal cavity andits relation to stop consonant voicing. ]ourrlld uf tlte AcoustictIlSociety ofA7MricII, 73, 1322-1336.

FOOTNOTE'"To IIpIlea' in G. HaJIIIIIOIId (Ed.), Cerebral cOlllrol of&P«cJl aNJ liMb~. Amst«dam: ElMvi«.