Article

A Biomechanical Modeling Study of the Effects ofthe Orbicularis Oris Muscle and Jaw Posture on

Lip ShapeIan Stavness,a Mohammad Ali Nazari,b,c Pascal Perrier,b Didier Demolin,b and Yohan Payand

Purpose: The authors’ general aim is to use biomechanicalmodels of speech articulators to explore how possiblevariations in anatomical structure contribute to differences inarticulatory strategies and phone systems across humanpopulations. Specifically, they investigated 2 issues: (a) thelink between lip muscle anatomy and variability in lip gesturesand (b) the constraints of coupled lip/jaw biomechanics onjaw posture in labial sounds.Method: The authors used a model coupling the jaw, tongue,and face. First, the influence of the orbicularis oris (OO)anatomical implementation was analyzed by assessinghow changes in depth (from epidermis to the skull) andperipheralness (proximity to the lip horn center) affected lipshaping. Second, the capability of the lip/jaw system togenerate protrusion and rounding, or labial closure, wasevaluated for different jaw heights.

Results: Results showed that a peripheral and moderatelydeep OO implementation is most appropriate for protrusionand rounding; a superficial implementation facilitates closure;protrusion and rounding require a high jaw position; andclosure is achievable for various jaw heights.Conclusions:Models provide objective information regardingpossible links between anatomical and speech productionvariability across humans. Comparisons with experimentaldata will illustrate how motor control and cultural factors copewith these constraints.

Key Words: biomechanics, articulation, physiology, speechproduction, lip shape, orbicularis oris, jaw, face

Variations and regularities found in the soundsystems of human languages might be due, at leastin part, to the intrinsic properties of the orofacial

motor system. Variability in vocal tract anatomy acrosshuman populations could have initiated differences inarticulatory gestures across languages. Likewise, propertiesshared by all human orofacial motor systems could havebeen the basis for common articulatory and motor trendsobserved in a large number of languages. In this context,models of the orofacial motor system can be used to evaluatethe influence of variations in physiological and anatomicalproperties on articulatory speech gestures. Comparingpredictions made with models to data collected from

speakers of various languages permits a quantitative assess-ment of the physiological factors that have potentiallyinfluenced the emergence of sound system rules andvariability in the languages of the world.

Lip gestures are good candidates for investigatingpotential links between physiological variability in humansand variability in the sound systems of languages becausesignificant differences in facial muscle morphology areknown to exist across subjects. These anatomical variationscould explain differences in speech-specific lip gestures,such as lip protrusion and lip rounding. In a discussion ofanthropophonetic variations, Brosnahan (1961) and Catford(1977) quoted the studies of Huber (1931) and stated that therisorius muscle is found in about 20% of Australians andMelanesians, 60% of Africans, 75% to 80% of Europeans,and 80% to 100% of Chinese and Malays. More recently,Pessa et al. (1998) showed that as many as 22 of their 50cadaver specimens lacked the risorius muscle. Althoughbased on small samples of data, these studies suggest thatsuch genetic anatomical characteristics could be consistent,or even occur with increasing frequency, over successivesections of populations of the African-European-Asian landmass. Pessa et al. (1998) also found that, in 17 of their 50specimens, the zygomaticus major presented a bifid structurewith two insertions points. The two insertion points of this

aUniversity of Saskatchewan, Saskatoon, Saskatchewan, CanadabStendhal University and Grenoble University of Technology, Grenoble,

FrancecUniversity of Tehran, IrandJoseph Fourier University, La Tronche, France

Correspondence to Ian Stavness: stavness@gmail.com

Editor: Jody Kreiman

Associate Editor: Kate Bunton

Received June 25, 2012

Accepted October 15, 2012

DOI: 10.1044/1092-4388(2012/12-0200)

878 Journal of Speech, Language, and Hearing Research N Vol. 56 N 878–890 N June 2013 N � American Speech-Language-Hearing Association

muscle could cause the dimple in the cheeks that manypeople have when smiling (Schmidt & Cohn, 2001). Theseobservations regarding the structure of the zygomaticusmajor muscle confirm that significant interspeaker differ-ences exist in facial muscles. Such anatomical differences arelikely to determine variations in face shaping and orofacialgestures in facial expression and speech production.

Speech scientists are gradually accumulating data onpossible links between anatomical variability across humansand variations in articulatory and acoustical characteristicsof human languages. Ladefoged (1984) showed that differ-ences between the vowel systems of Yoruba, a Niger-Congolanguage spoken in West Africa, and Italian could have ananatomical basis. Ladefoged noted the existence of smalldifferences in formant values between Yoruba and Italian,which otherwise have very similar seven-vowel systems. Henoted that these differences are consistent with anatomicaldifferences generally observed between Africans andEuropeans. Ladefoged (1984) noted the following:

Some of the differences between the two languages aredue to the shapes of the lips of Italian as opposed toYoruba speakers […] With the exception of /i/ and toa lesser extent /e/, the second formant is lower forthe Italian vowels than for the Yoruba vowels. Thesedifferences are precisely those that one would expectif Yoruba speakers, on the whole, used a largermouth opening than that used by the Italian. […] Thepossibility of overall differences in mouth opening iscertainly compatible with the apparent facial differencesbetween speakers of Yoruba and Italian. (pp. 85–86)

More recently, Storto and Demolin (in press) foundthat none of their five subjects speaking Karitiana, a Tupilanguage spoken in Brazil, showed lip rounding andprotrusion while producing the vowel [o] (here, [o] denotesthe physical realization of the phoneme /o/). Measurementswere based on video data (Figure 1). The average first andsecond formant values for this vowel, measured on the fivespeakers, are, respectively, 459 Hz (n = 250) and 1056 Hz(n = 250), which correspond to F1/F2 values for mid-backor high-back vowels in the acoustic space. An auditoryperceptual test showed that Portuguese and French speakers,who also have the vowel [o] in their vowel systems, correctlyidentified the Karitiana [o] as the corresponding mid- tohigh-back vowel [o] in their languages. In addition, electro-myographic (EMG) recordings with surface electrodesplaced at the rim of the lips showed no activity whenKaritiana speakers produced the vowel [o]. In contrast,EMG measurements with the same electrode placement onPortuguese and French speakers, who have lip rounding andprotrusion (e.g., Figure 2), showed clear EMG activity forthe same vowel. The vowel [o] is the only back vowel of theKaritiana phonetic system, which lacks the high-back vowel[u]. Karitiana is not unique in this regard, as its vowel systemis similar to several other Tupi languages (Storto & Demolin,2012). The Yoruba and Karitiana data suggest that smallanatomical differences could create variations that influencethe shape of sound patterns found in the world’s languages.

These variations may be one of the factors explainingso-called phonetic universals. Other factors include thecategorization of these variations and their cultural trans-mission across many generations.

Figure 1. Lateral and frontal views of the face of a Karitiana speakerproducing the vowel [o] in the word [koβot⌝], ‘‘sweet.’’ The frame waschosen from the corresponding acoustic recording. The positioncorresponds to the middle part of the first vowel in the word. It can beobserved that the lip shape does not match the classical patternsof protruded and rounded lips, as shown in Figure 2.

Figure 2. Lateral (left) and frontal (right) views of the face of aFrench speaker producing the French vowel [u]. The frame waschosen from the corresponding acoustic recording. A rounded andprotruded lip shape can be observed. (Courtesy of Pierre Badin;Badin et al., 2002.)

Stavness et al.: Effects of Orbicularis Oris and Jaw Posture on Lip Shape 879

On the other hand, regularities in phonetic realizationsof the world’s languages can also originate from physiolog-ical factors, including the muscle arrangements and theinterarticulatory interactions in the orofacial region. Afundamental mechanism of speech production is the couplingbetween the jaw and the tongue and lips, which determinesthe fine shaping of the vocal tract. This coupling is the basisfor reduplicative babbling, and its ontogenetic evolutionexplains variegated babbling (MacNeilage, 1998). Degreesof freedom and constraints in this coupling can influencepreferences in syllabic patterns in the world’s languages. Thecapability of the upper and lower lips to deform makes itpossible for a subject to produce bilabial stops, which requirea closed lip horn, for a range of jaw positions. Hence, thelips can stay in contact while the jaw moves downward,allowing speakers to use anticipatory strategies and coartic-ulation in speech production movements. For example,transitions from a bilabial stop toward subsequent openedsounds can be produced with a low jaw position withoutendangering the correct production of the stop. In a recentstudy, Rochet-Capellan and Schwartz (2007) investigated thecoordination among jaw, tongue tip, and lower lip duringrepetitions of labial-to-coronal (/pata/) and coronal-to-labial(/tapa/) CVCV sequences at increasing speaking rates. Theyfound that when the speaking rate increased, there was ageneral trend for the coronal-to-labial sequences to changetoward the labial-to-coronal sequences. From articulatorydata, they observed that, at slow speaking rates, both thebilabial and the coronal consonants were produced atthe end of upward movements of the jaw—that is, eachconsonantal closure was synchronized with the maximalelevation of the jaw. As the speaking rate increased, theyobserved that both consonants were produced during thesame upward movement of the jaw. The coronal closureremained synchronized with the maximal jaw elevation, butthe labial closure was produced during the upward move-ment from the vowel to the coronal consonant, that is,for a lower position of the jaw. This was made possible bythe capability of both lips to deform. In many languages,consonantal clusters with first a labial and then a coronalconsonant are significantly more frequent than clusters withfirst a coronal and then a labial consonant. The observationof labial-coronal ordering has been called the labial-coronaleffect (see MacNeilage & Davis, 2000). The mechanicalproperties of the lips could have contributed to theemergence of this and other patterns in languages.

The goal of our study was to demonstrate the effectof anatomical factors on lip shape using a biomechanicalmodel of the orofacial system. The model integrates the softtissues of the face, tongue, and lips with the hard structuresof the maxilla and jaw. Considering these hard structureswas important for our analysis because the underlying bonestructure and jaw position have a significant effect on theconfiguration of the lips. Whereas in previous biomechanicalstudies of lip protrusion researchers have used genericmodels (Kim & Gomi, 2007; Nazari, Perrier, Chabanas, &Payan, 2010), we adapted the morphology of our model to aparticular speaker. This decision was motivated by the fact

that we have a large set of experimental data for this speaker,which makes possible a quantitative assessment of thesimulations. To create our subject-specific model, we coupledthe Nazari et al. (2010) face model with the Stavness, Lloyd,Payan, and Fels (2011) jaw/tongue model. Thus, we reporthere an original biomechanical model that incorporatescoupling and contact effects among the face, the tongue, anda muscle-activated jaw.

We used the face model to investigate two questionsregarding variation and regularity in articulation: howvariation in lip musculature is associated with variation in lipgestures and how the properties of lip biomechanics imposeconstraints on jaw posture. In the first part of the study,we investigated the effect of orbicularis oris (OO) musclegeometry on simulated lip protrusion and rounding. Ourmodel-based analysis enables a quantitative assessment ofhow lip shaping is influenced by variations in the anatomicaldistribution of the marginalis and peripheralis parts of thismuscle. Comparing simulations with data, such as recordingsof the production of vowel [o] by Karitiana speakers, couldprovide evidence for links between anatomical and physiolo-gical variations and the diversity of the world’s languages. Inthe second part of the study, we assessed the extent to which asubject can move his/her jaw up and down while keeping abilabial closure. The model provides quantitative informationabout motor equivalence strategies likely used to producebilabial consonants in anticipation of subsequent syllables.

Method

We used a biomechanical face-jaw-tongue model tocreate simulations of lip gestures for a range of conditions onlip musculature and jaw posture. The resulting simulated lipshapes were compared using quantitative measurements.

Model

An assessment of the effect of variation in OOmorphology and jaw posture on lip protrusion and lip shaperequires a model that has consistent anatomy with a specificspeaker and that includes the underlying bony structures ofthe jaw and skull in addition to the soft tissues of the faceand lips. We created our model (see Figure 3) in the ArtiSynthbiomechanical modeling toolkit (www.artisynth.org; seealso Lloyd, Stavness, & Fels, 2012) by registering andintegrating two previously reported reference models (Nazariet al., 2010; Stavness et al., 2011).

We used a computed tomography (CT) data set of asingle male speaker as a means to adapt and coregister thesedisparate reference models. The speaker was chosen becausewe have an extensive set of experimental data on his speechmovements. The reference 3-D finite-element (FE) facemodel was originally built in the ANSYS simulation software(Nazari et al., 2010) and consists of 6,342 hexahedralelements arranged into three layers: superficial, middle,and deep. The model had been previously adapted to thespeaker’s CT data using a segmentation of the interior bonesurface and the exterior skin surface (Bucki, Nazari, &Payan, 2010). The reference jaw-tongue-hyoid bone model

880 Journal of Speech, Language, and Hearing Research N Vol. 56 N 878–890 N June 2013

(Stavness et al., 2011) combined and registered two referencemodels to the same CT data set: a 3-D rigid-body jaw-hyoidbone model (Hannam, Stavness, Lloyd, & Fels, 2008) and a3-D FE tongue model (Buchaillard, Perrier, & Payan, 2009;Gerard, Perrier, & Payan, 2006).

Although both the jaw-tongue-hyoid model and theface model had previously been adapted to the same CT dataset, the inner surface of the face and lips did not exactlyconform to the outer surfaces of the jaw and maxilla due toinaccuracies in the original data segmentation and modeladaptation. The interaction between the lips and underlyingbone surfaces is critical during lip movements as the jaw,maxilla, and dentition provide boundary conditions for lipmovements. To improve the fit between the two models,we used a contact-based morphing procedure and alsoperformed some minor manual editing of the face meshin order to improve the regularity of the elements afteradaptation.

The dynamics of the face, jaw, tongue, and hyoid bonemodels were coupled by defining attachment constraintsbetween the FE nodes of the face and tongue and the rigidbodies of the jaw and hyoid bone. Attachments between thetongue and jaw-hyoid models have been described previously(Stavness et al., 2011). For the face mesh, we attached anumber of inner-surface nodes to adjacent locations on jawand maxilla rigid-bodies, and left nodes in the region ofthe lips and cheeks unattached. We also attached adjacentsurfaces of the tongue and face models near the region of the

floor of the mouth. The attachment points are illustrated inFigure 3.

Contact between different articulators is anothercrucial component of speech production and includesdeformable–to–deformable-body contact (between the upperand lower lip) as well as deformable–to–rigid-body contact(between the lips and teeth, and between the tongue andpalate and teeth). ArtiSynth supports mesh-based collisiondetection and contact handling using dynamic constraints.Contact detection was enabled between the face and the jawand maxilla meshes (including the teeth). Subregions of anFE face mesh were defined for the upper and lower lips andused for contact between the lips.

In the reference face model, muscle forces wereapplied along serial line segments representing the muscle’sprincipal line of action (called cable elements; see Nazariet al., 2010). In the current model, muscle mechanics areincorporated with a transverse-isotropic FE material,whereby stress is increased in the fiber direction with muscleactivation (Weiss, Maker, & Govindjee, 1996). The OOmuscle was defined as a continuous loop of elements aroundthe lips, as shown in Figure 4. To vary the OO morphology,the size and location of this loop of elements was varied indifferent simulations, as described in the Simulations section.The fiber direction in an element associated with a particularmuscle represents the muscle’s principal line of action, andadditional stress is applied in that direction within theelement during muscle activation. The fiber directions foreach individual element associated with the OO muscle wereinterpolated from a canonical serial line segment representing

Figure 3. Upper panel: Sagittal cutaway (left) and posterior (right)views of the dynamic model that integrates the face, jaw, tongue,and hyoid bone. A closeup view of the lips (inset, left) shows thesuperficial (S), middle (M), and deep (D) layers of the three-layer,finite-element mesh. Bottom panel: Attachment points are shownbetween the face-skull (green points), face-jaw (red points), andface-tongue (blue points).

Figure 4. Frontal (left) and lateral (right) views of the face modelshowing the orbicularis oris (OO) muscle elements organized intodifferent peripheral loops from marginal to peripheral (1, 2, 3, 4) andinto S, M, and D depth layers in the mesh.

Stavness et al.: Effects of Orbicularis Oris and Jaw Posture on Lip Shape 881

the OO muscle fibers, as shown by the cyan lines in Figure 5.If an element lies within the region of the cyan loop, thenits fiber direction is interpolated from the nearby cyan linesegments. The interpolation is performed as a weightedaverage of line segment directions, where the weighting isinversely proportional to the distance between the elementand line segments. If an element lies outside of the region ofthe cyan loop, then its direction is set to the direction of theclosest line segment of the cyan loop.

The face model was implemented in ArtiSynth using alarge deformation FE simulation framework. The FE meshconsisted of eight-node, hexahedral elements arranged inthree layers from superficial to deep. The passive tissueproperties of the model were chosen to be consistent with theNazari et al. (2010) reference face model and included atissue density of 1,040 kg/m3 and an isotropic, nonlinear,hyperelastic material—a fifth-order Mooney-Rivlin material(Mooney, 1940; Rivlin, 1948) with coefficients of c10 = 2,500Pascals (Pa), c20 = 1,175 Pa, and c01 = c11 = c02 = 0 Pa. Themodel’s tissue was made incompressible with a constraint-based simulation (see Stavness et al., 2011, for further details).The model’s passive tissue properties included Rayleigh

damping, which is a viscous damping proportional to bothtissue stiffness (with coefficient β = 0.055 s) and tissue mass(with coefficient α = 19 s–1). A transverse-isotropic musclematerial—that is, a material with stiffness properties in thedirection along the muscle fiber that are different than in thedirections orthogonal to it—was superimposed on the passiveisotropic material based on the uncoupled strain energyformulation proposed by Weiss et al. (1996). Passive stressalong the fiber direction was made to increase exponentiallywith increasing fiber stretch (see Weiss et al., 1996, Equation7.2, p. 123). Parameters for this exponential passive fiberbehavior were chosen based on the muscle constitutiveequation provided by Blemker, Pinsky, and Delp (2005):l*= 1.4 (the long fiber stretch at which collagen fibers arestraightened); C3 = 0.05 (scales the exponential stresses),C4 = 6.6 (rate of uncrimping of the collagen fibers). Themaximum active fiber stress was 100,000 Pa.

Simulations

Our primary aim was to determine the effect of OOmorphology on simulated lip protrusion and rounding.We were also interested in the effect of jaw posture on liprounding and protrusion, which we were able to analyze withour coupled face-jaw-tongue biomechanical model.

Simulations were performed in ArtiSynth, whichallows for fast-forward dynamics simulation with dynamiccoupling between rigid-body and FE models as well ascollision handling. Coupling and collision handling areimportant for modeling the interactions between the lips andthe underlying bony structures. ArtiSynth also providesgraphical user interface tools that were used to help withmodel registration and FE mesh editing.

Simulation speed is an important aspect of biomechan-ical models because faster simulations enable a moreextensive investigation of the model’s behavior over a rangeof input parameters. We achieved simulation times thatwere much faster than the reference face model in ANSYS.For the full model, with dynamic coupling and contact,each 500-ms simulation required approximately 225 s ofsimulation time on a 2.2 Ghz Intel Core i7 processor. Ithas been shown that ArtiSynth can be orders of magnitudefaster than ANSYS for similar simulation (Stavness et al.,2011).

Each simulation was 500 ms in duration: Muscleactivation for the OO muscle increased linearly over aduration of 400 ms and held the final activation for 100 ms.In all simulations, muscle activation was increased uniformlyfrom 0% to 50% of the maximum possible activation,which corresponds to an active muscle stress of 50 kPa. Thislevel of final activation was chosen to ensure numericalconvergence in all simulations while generating lip displace-ments of realistic amplitudes. Each simulation reached anequilibrium position by 500 ms.

Deepness and peripheralness. The OO muscle wasmodeled as a continuous loop of elements. In order to assessthe effect of deepness, simulations were performed with theOO muscle located only in the deep (D), or in the middle(M), or in the superficial (S) layer of the face mesh, where

Figure 5. Frontal (left) and lateral (right) views of the face modelshowing the OO muscle for the M3 configuration (depth = M,peripheralness = 3), including the elements (top panel), putative OOmuscle fiber direction (cyan line, middle panel), and muscle fiberdirection for each element (red lines, bottom panel). Grid spacing is10 mm.

882 Journal of Speech, Language, and Hearing Research N Vol. 56 N 878–890 N June 2013

deep is closer to the skull and superficial is closer to theskin surface (see Figure 3). In order to assess the effect ofperipheralness, we varied the radius of the OO muscle loop(centered in the middle of the lip horn) from smaller radius(more medial) to larger radius (more peripheral) in four sizes:1, 2, 3, and 4 (as shown in Figure 4).

Upper versus lower OO peripheralness. To investigatethe variation of OO size in more detail, simulations were alsoperformed for different peripheralness for the upper versusthe lower portion of the OO muscle. The structure/geometryof the FE face mesh is such that the lower portion of the OOmuscle is more peripheral to the lower lip than the upperportion of the OO muscle is to the upper lip (by inspection ofFigure 4). Therefore, we tested the different relative periph-eralness of the upper versus lower parts of the OO muscle. Inthese simulations, the OO muscle was located in all tissuelayers: deep, middle, and superficial.

Lip rounding with jaw lowering. To investigate the effectof jaw lowering on lip rounding and protrusion, simulationswere performed with synergistic activation of the OO andjaw lowering muscles (anterior belly of the digastric [ABD]and lateral pterygoid [LP] muscles). The jaw model wasdynamically coupled to the FE models of the tongue andface, and therefore we were able to simulate the biomechan-ical effect of the coupled system. We chose the OO muscleconfiguration from the above simulations that best matchedthe speaker’s lip protrusion (shown in Figure 2).

Lip closure with jaw lowering. In order to investigatethe extent to which lip closure is compatible in the modelwith jaw lowering, different configurations of OO musclewere evaluated. Based on the results of our deepness-versus-peripheralness simulations, we expected that lip closure couldbe achieved with the superficial portions of the OO muscle.We varied the different amount of peripheralness in OOneeded to achieve closure with two different degrees of jawlowering.

Lip Shape Metrics

In order to quantify the effect of OOmorphology on lipshape, we chose lip measurements similar to those proposedin previous studies (Abry & Boe, 1986; see also Figures 7and 8 in Nazari, Perrier, Chabanas, & Payan, 2011). Forwardlip protrusion was characterized by the anterior displacementof the most anterior point on the upper and lower lips,as illustrated in Figure 6 (left panel). We calculated thedisplacement as the difference in position between the mostanterior flesh point in the protruded posture from the mostanterior flesh point in the rest posture (for both the upperand lower lip). The anterior points in each posture may bedifferent flesh points because the lips may rotate duringprotrusion. Positive displacement corresponds to anteriorprotrusion relative to rest posture. The lip opening wascharacterized by the shape of the opening in an orthographicprojection of the frontal view of the model. The openingspace was segmented from the frontal projection image andits width, height, and area are measured as shown in Figure 6(right panel).

Results

Deepness and Peripheralness

Simulation results of lip protrusion for differentconfigurations of OO muscle geometry are plotted inFigure 7 for the same level of activation of the active muscleelements. Quantitative lip measurements for the simulationsare reported in Table 1. The results show that moreperipheral OO implementations are associated with largerprotrusion, independent of deepness, with one exception inthe superficial layer (see below). However, the degree ofdeepness influences the covariation of protrusion and liparea. For a deep OO implementation, peripheralness andprotrusion are systematically associated with larger lip widthand lip height, and therefore with larger lip area. For asuperficial implementation, peripheralness is also associatedwith larger lip area, mainly due to an increase in lip width.For a middle OO implementation, the influence of periph-eralness is different: We observed a nonlinear variation in lipheight, lip width, and lip area with peripheralness. FromPeripheralness 1 to Peripheralness 3, lip area increases,mainly because of the increase in lip height, but lip areadecreases from Peripheralness 3 to Peripheralness 4 due tothe combined decrease in lip height and lip width.

The prototypical characteristic of the protrusion/rounding gesture, as observed in rounded vowels such as /u/or /o/, is a significant amount of lip protrusion associatedwith a small lip opening area. In the conditions of oursimulations, the protrusion/rounding gesture is most effec-tively produced for the most peripheral implementationof the OO muscle in the middle layer of the face tissues(especially for the lower lip). It is interesting to note that animplementation in the superficial layer cancels the influenceof peripheralness on protrusion for the upper lip (see Table 1,upper lip protrusion). Hence, a superficial implementation

Figure 6. Lip protrusion and shapingmetrics (inspired by Abry & Boe,1986). Frontal view (left): lip opening width (W), height (H), and area(A). Lateral view (right): upper/lower lip anterior protrusion (Pu/Pl).

Stavness et al.: Effects of Orbicularis Oris and Jaw Posture on Lip Shape 883

of the OO seems to be particularly inappropriate to thegeneration of protrusion and rounding.

The results also show that, for a given degree ofperipheralness, a superficial location enables efficient closure,and a deep location reduces the impact of OO activation onclosure. Hence, in the absence of a protrusion requirement,a superficial implementation facilitates lip closing gestures.

Except in one case (middle and most marginalimplementation of the OO), the lower lip shows larger pro-trusion than the upper lip. This may be a subject-specific

property due either to the subject lip volume (i.e., a lowerlip volume larger than the upper lip volume) or to the factthat the lower OO muscle is more peripheral than the upperOO muscle in the FE mesh, as shown in Figure 4. This issueis investigated below.

Upper Versus Lower OO Peripheralness

The results of the previous section show differences inprotrusion amplitude between the upper and the lower lips.

Figure 7. Simulation results for different OO muscle deepness and peripheralness. Superficial placement of OO resulted in more closure,whereas deep placement resulted in more opening. Peripheral placement of OO resulted, in general, in more protrusion and more aperture thanmarginal placement.

Figure 8. Simulation results for different peripheralness of the upper versus the lower portion of the OO muscle.

884 Journal of Speech, Language, and Hearing Research N Vol. 56 N 878–890 N June 2013

This might be due to differences in the peripheralness ofthe OO implementation. Simulation results for lip shapesobtained with differential upper versus lower OO peripher-alness are plotted in Figure 8, and quantitative measuresare reported in Table 2. We assumed an implementation ofthe OO in all layers (superficial, middle, and deep) together.We chose to test more peripheral implementations for theupper portion of OO (Rings 3 and 4, as compared to Rings 2and 3 for the lower portion) because the structure/geometryof the FE face mesh is such that the lower portion of theOO muscle is more peripheral to the lower lip than the upperportion of the OO muscle is to the upper lip (by inspectionof Figure 4).

In general, the protrusion amplitudes are much larger(between 1 mm and 3 mm) for both parts of the lips than inthe previous section. This is due to the fact that here all layersof the OO were activated together, whereas in the deepnessand peripheralness simulations, each layer was activatedseparately. Apart from this side effect, the results demon-strate that differential protrusion of the upper versus lowerlip is determined by the peripheralness of the upper versuslower OO muscle fibers. Consistent with the results of theprevious section, the protrusion of the lower lip is system-atically larger than the protrusion of the upper lip.

As expected from the results of the previous section, amore peripheral implementation of the OO muscle in onepart of the lips increases the protrusion of the same part.However, a more peripheral implementation of the upperpart reduces the protrusion of the lower part. A conse-quence of this phenomenon is that the smallest difference inprotrusion between the upper and lower lips is obtained fora peripheral implementation of the OO in the upper lip(fourth radius) and a marginal implementation in the lower

lip (second radius), whereas the largest difference isobtained for a marginal implementation in the upper lip(third radius) and peripheral implementation in the lowerlip (third radius).

Ideally, lip rounding is associated with a small widthand a reasonably small area (between 20 mm2 and 30 mm2).Smallest widths are obtained for the more marginalimplementation of the OO in the lower lip (second radius).For this lower radius, the more appropriate lip area isobtained for the less peripheral implementation in the upperlip (third radius). This configuration (second lower radiusand third upper radius) provided the best tradeoff betweenlip rounding and protrusion. We also found that thisconfiguration provided the best qualitative match to thesubject’s data (Figure 2). This best-case OO configurationwas chosen for simulations below on lip protrusion duringjaw lowering.

Table 1. Quantitative measurements for deepness-versus-peripheralness simulations.

Variable

Peripheral radius

1 2 3 4

Upper lip protrusion (mm)Deep 0.6 1.9 2.9 3.9Middle 0.3 1.4 2.0 2.7Superficial 0.3 0.0 0.1 0.2

Lower lip protrusion (mm)Deep 1.6 3.3 4.4 4.8Middle 0.0 2.5 4.5 4.8Superficial 0.6 0.9 1.5 1.5

Lip opening width (mm)Deep 14.8 21.9 25.7 27.1Middle 10.9 9.6 14.4 12.9Superficial 0.0 0.0 8.4 9.3

Lip opening height (mm)Deep 2.3 3.0 4.1 4.9Middle 1.0 1.6 3.2 2.6Superficial 0.0 0.0 0.8 0.7

Lip opening area (mm2)Deep 22.0 33.1 59.5 74.6Middle 6.3 7.7 26.5 18.0Superficial 0.0 0.0 3.3 3.6

Table 2. Quantitative measurements for upper-lip–versus–lower-lipperipheralness simulations.

Upper lip protrusion (mm)

Lower OO peripheralradius

Upper OO peripheral radius

3 4

2 3.8 4.33 4.3 5.4

Lower lip protrusion (mm)

Lower OO peripheralradius

Upper OO peripheral radius

3 4

2 6.6 6.13 7.8 7.6

Lip opening width (mm)

Lower OO peripheralradius

Upper OO peripheral radius

3 4

2 13.8 11.63 16.4 16.4

Lip opening height (mm)

Lower OO peripheralradius

Upper OO peripheral radius

3 4

2 3.4 2.33 5.1 3.9

Lip opening area (mm2)

Lower OO peripheralradius

Upper OO peripheral radius

3 4

2 25.3 11.93 51.4 31.4

Note. OO = orbicularis oris muscle.

Stavness et al.: Effects of Orbicularis Oris and Jaw Posture on Lip Shape 885

Lip Rounding With Jaw Lowering

We evaluated the degree to which lip rounding iscompatible with jaw lowering using simulations withincreasing degrees of jaw lowering during OO activation. Inthis experiment, as explained in the Method section, theselection of lip muscles activation was based on the results ofthe previous simulations that evaluated the impact of the OOdeepness and peripheralness on the protrusion/roundinggesture. We used the configuration that provided the bestgesture: second lower radius and third upper radius withan implementation of the OO in the three layers together(superficial, middle, and deep). Jaw lowering distance wasmeasured as the distance between the lower and upper mid-incisor points.

Results are plotted in Figure 9 and quantitative lipmeasurements reported in Table 3. Increase in jaw loweringmuscle activation lowers the jaw and, obviously, causesincreased lip opening. The backward movement of theanterior part of the jaw, associated with jaw lowering, movesthe lower lip backwards. The upper lip position also varieswith jaw position but is affected less than the lower lipposition. These simulation results show that the rounding/protrusion gesture is rather sensitive to variation in jawheight and suggest that having a high jaw position is arequirement for the achievement of a correct protrusion androunding lip gesture.

Lip Closure With Jaw Lowering

By activating both the peripheral and marginalportions of the OO muscle in the superficial layer (S1 + S2 +S3 + S4), we could achieve lip closure, similar to lip shapesin bilabial consonants /b/ or /p/, for a low jaw posture.The results are plotted in Figure 10 and quantitative lipmeasurements reported in Table 4. The additional recruit-ment of middle, marginal portion (M1) achieves lip closurewith a very low jaw posture. The peripheral OO activationprovided the required closure of the lips by downwardmovement of the upper lip and upward movement of thelower lip. Notably, we also observed coupling effects betweenthe face and jaw: Activation of OO to achieve lip closureinduces slight jaw closure. These simulations demonstratethat, contrary to lip rounding, lip closure is compatible withvariable jaw heights.

Discussion

The properties and structure of sounds in humanlanguages, including diachronic evolution, arrangementinto sequences, (co)articulation, and variability of acousticand articulatory correlates, are the results of a complexcombination of influences. These influences arise fromvarious factors, including the intrinsic physical properties ofthe speech production system, basic motor control principlesof human skilled movements, the intrinsic properties ofthe auditory and visual perception systems, memorizationcapabilities in humans, social factors, environmental factors,and communication efficiency principles. A major limitation

Figure 9. Simulated lip rounding and protrusion for different levelsof jaw lowering. Jaw lowering increases lip opening and reduces lipprotrusion.

886 Journal of Speech, Language, and Hearing Research N Vol. 56 N 878–890 N June 2013

of experimental studies that aim to explain how soundsystems in the world’s languages are structured and how theyvaried and evolved is the difficulty of disentangling thesedifferent influences in the signals recorded from humanspeakers.

In this context, using computational models torepresent the various processes that contribute to languagestructure is a potentially fruitful approach. Models aredesigned to be a simplified representation of reality. Inaddition, modeling isolated subsystems of a complex process,without accounting for their interactions, provides only apartial view into the whole process. However, these modelscan give clear pictures of the constraints that each subsystemexerts on the whole process. The goals of the present studywere to assess the constraints that orofacial biomechanicsexert on speech production gestures and to provide aninterpretation of these constraints in the context of regularityand variability of the sound systems in the world’s languages.We focused specifically on the lip gestures as a prototypicalcase. Toward this aim, we used a sophisticated realistic3-D biomechanical model of the whole orofacial motorapparatus—that is, the peripheral part of the motor systemincluding the jaw, tongue, and face, and accounting formuscle commands and muscle mechanics (Buchaillard et al.,2009; Hannam et al., 2008; Nazari et al., 2010; Stavness et al.,2011).

Given that variability in orofacial muscle anatomyhas been found across humans, we tested the influence ofplausible variability in the anatomical implementation of theOO muscle on lip shapes. This muscle plays a central role inthe production of the lip protrusion and rounding gestures(Nazari et al., 2011). These gestures are essential for roundedvowels such as /u/, /o/, or /y/, and for closing labial gestures,which represent the key articulatory feature of bilabial stopssuch as /b/ and /p/. Our approach consisted of measuringthe variability of key parameters of the lip shape gesturewhen the OO anatomical implementation was systematicallyvaried. A potentially confounding factor is that motorcontrol strategies can be adapted across speakers, or forthe same speaker across conditions, in order to achieve thesame motor goal with different configurations of the motorapparatus (see, e.g., Hughes & Abbs, 1976). However, weintentionally did not consider this possibility in our studyand used the same level of muscle activation for all the testedanatomical OO implementations. This limited scope allowedus to evaluate the intrinsic influence of anatomy, independentof any possible adaptation of motor control strategies.

Variability was tested in terms of depth of the OOanatomical implementation, from superficial (i.e., close tothe skin) to deep (i.e., close to the maxillary bones), andin terms of distance from the center of the lip horn, frommarginal (i.e., close to the center) to peripheral. Differenceswere also considered in the implementation between theupper and the lower lips. Not surprisingly, it was found that

Table 3. Quantitative measurements for OO activation with varying degrees of jaw lowering.

Jaw lowering (mm)Upper lip

protrusion (mm)Lower lip

protrusion (mm)Lip openingwidth (mm)

Lip openingheight (mm)

Lip openingarea (mm2)

10 4.3 8.3 14.4 3.0 19.116 3.0 4.8 18.5 5.6 63.421 3.7 4.1 22.4 8.6 122.2

Figure 10. Simulated lip closure with different OO configurations fortwo different levels of jaw lowering: 10% activation of jaw openersresulted in approximately 10 mm jaw lowering (left panels), whereas20% activation achieved approximately 16 mm lowering (rightpanels). Lip closure was achieved by activating the superficial layerof the OO muscle.

Stavness et al.: Effects of Orbicularis Oris and Jaw Posture on Lip Shape 887

anatomical variability has a noticeable impact on lip shaping.When we considered similar implementations in the upperand the lower lips, we observed general trends such that asuperficial location facilitates closure, a deep location actsagainst closure, and a peripheral location causes protrusionand aperture. However, some nonmonotonous relations wereobserved for the most peripheral OO implementations.Interactions were also found between deepness and periph-eralness, such that a largely peripheral implementation withan intermediate deepness is the most appropriate imple-mentation for the efficient achievement of the protrusion androunding gesture.

The evaluation of some differences in the OOimplementation between the upper and the lower lip revealedthat a very peripheral implementation in the upper lipassociated with an intermediate implementation of the lowerlip generates the best protrusion and rounding gesture.Conversely, a superficial implementation is not well adaptedfor the production of this gesture, because it does notfacilitate protrusion.

In the context of the study of sound systems in theworld’s languages, these results can be interpreted as follows.If anatomical differences in the OO anatomy exist acrossgroups of humans, which is consistent with the hypothesesunderlying our work, it can be expected that more roundedand protruded sounds exist in groups of humans where theimplementation is more peripheral and reasonably deepin both lips. Because languages are under the influence ofnumerous factors, we do not assert that all the languagesin these groups of humans would have these characteristics.However, these findings could explain a general trend inthese languages as concerns protruded and rounded vowels.Considering the example of Karitiana speakers, our modeling

results suggest that a potential explanation for the limitedamount of protrusion and the relatively large aperture of thelip horn during /o/ could lie in an OO implementation thatis more marginal and/or deep. Future EMG studies areplanned to clarify the location of the muscle fibers that areactivated during the production of the vowel /o/ in Karitiana.

The second part of our study aimed to evaluate theextent to which achieving correct lip gestures, includingprotrusion/rounding and labial closure, is compatible withvariations in jaw height. In the majority of the languages,protruded and rounded labial vowels, as well as bilabialstops, are associated with high jaw positions. We aimed toassess whether a high jaw position is a strict biomechanicalrequirement for these sounds or whether there exists freedomto lower the jaw without endangering production of thecorrect lip gesture. Our simulations suggest that jaw height isindeed a strong requirement for the achievement of a correctprotrusion and rounding gesture. This result has beenobtained for a specific OO activation, and its generalizationshould be considered with caution. However, because thisactivation was shown to be very well suited, in the model, forthe achievement of the protrusion and rounding gesture,we think that it could explain the fact that the majority ofrounded vowels are high vowels.

As concerns the achievement of bilabial closure, oursimulations tend to show that high jaw positions are not arequirement. In our model, a reasonable activation of thesuperficial layer of the OO in its marginal and peripheral partsenables lip closure even if the distance between the upper andthe lower incisors is as large as 1 cm. Such a large freedomin jaw positioning can certainly be used to plan sequencesof speech gestures in order to find the most appropriate jawtrajectories to achieve (a) a sequence of articulatory goals

Table 4. Quantitative measurements for 10% and 20% jaw lowering muscle activation with different OO configurations.

10% jaw lowering muscle activation (,10 mm jaw lowering)

OO musclesUpper lip

protrusion (mm)Lower lip

protrusion (mm)Lip openingwidth (mm)

Lip openingheight (mm)

Lip openingarea (mm2)

Rest 0.0 –1.6 37.2 7.6 175.6S1 0.3 –0.9 33.7 5.5 95.2S12 0.3 –0.1 21.8 3.6 32.4S123* 0.3 1.0 0.0 0.0 0.0S1234* 0.5 1.9 0.0 0.0 0.0S1234+M1* 1.0 2.6 0.0 0.0 0.0

20% jaw lowering muscle activation (,16 mm jaw lowering)

OO musclesUpper lip

protrusion (mm)Lower lip

protrusion (mm)Lip openingwidth (mm)

Lip openingheight (mm)

Lip openingarea (mm2)

Rest 0.0 –2.0 37.9 10.3 258.9S1 0.3 –1.3 35.9 7.8 166.8S12 0.3 –0.5 33.7 6.0 88.5S123 0.3 0.6 20.3 3.6 27.4S1234 0.2 1.3 9.0 1.3 3.4S1234+M1* 0.6 2.1 0.0 0.0 0.0

*Denotes a case where lip closure is achieved.

888 Journal of Speech, Language, and Hearing Research N Vol. 56 N 878–890 N June 2013

within a short time interval or (b) anticipated articulatorygoals with coarticulation. The simulated freedom in jawposture during bilabial stops is consistent with the observedpreference for the labial-coronal order in sequences ofconsonants in the world’s languages. Rochet-Capellan andSchwartz (2007; see their Figure 5B) suggested that this effectarises when the bilabial stop is produced for a low jawposition during the jaw upwardmovement from the precedingvowel to the subsequent coronal stop. Our simulationsconfirm that this is indeed possible, even for relatively low jawpositions and plausible magnitude of muscle activations.

Conclusions

Our study demonstrates the utility of a realistic modelof orofacial biomechanics in the analysis of variation andregularity in articulatory patterns. We found evidence forpotential links between variability in the anatomy of the lipsand variability in articulatory and acoustical characteristicsof speech sounds. An association between the biomechanicalconstraints of the lips and regularities in articulatory gestureswas also found. We demonstrated that the deformationcapabilities and muscle activations of the lips allow for acertain amount of freedom in jaw positioning during bilabialstops. This freedom could be used for coarticulationplanning in bilabial-stop–vowel–coronal-stop–vowelsequences at fast speaking rates. Conversely, we alsodemonstrated that protrusion and rounding of the lipsrequires high jaw positions.

These results contribute to a better understandingof the influences under which the languages of the worldstructured themselves and varied diachronically. We haveshown that physical differences and regularities of thespeech motor apparatus are factors that may influencethe emergence and evolution of speech sounds. Using abiomechanical model permitted an objective evaluation ofthese phenomena, independent of any influence of motorcontrol strategies and cultural factors. In future studies,the comparison of these model-based predictions with datarecorded in various languages from speakers with variousorigins should shed light on the way motor control andcultural factors cope with speaker-specific physical con-straints in speech production.

AcknowledgmentsThis work was partly supported by the French Agence

Nationale de la Recherche (Project SKULLSPEECH, ANR-08-BLAN-0272). We thank the ArtiSynth team at the University ofBritish Columbia for making the simulation software available andPierre Badin at Gipsa-lab for providing the CT data used to adaptthe model, as well as for Figure 2.

ReferencesAbry, C., & Boe, L. (1986). Laws for lips. Speech Communication, 5,

97–104.

Badin, P., Bailly, G., Reveret, L., Baciu, M., Segebarth, C., &Savariaux, C. (2002). Three-dimensional linear articulatory

modeling of tongue, lips and face, based on MRI and videoimages. Journal of Phonetics, 30, 533–553.

Blemker, S., Pinsky, P., & Delp, S. (2005). A 3D model of musclereveals the causes of nonuniform strains in the biceps brachii.Journal of Biomechanics, 38, 657–665.

Brosnahan, L. F. (1961). Sounds of language: An inquiry into the roleof genetic factors in the development of sound systems.Cambridge, United Kingdom: W. Heffer & Sons.

Buchaillard, S., Perrier, P., & Payan, Y. (2009). A biomechanicalmodel of cardinal vowel production: Muscle activations and theimpact of gravity on tongue positioning. The Journal of theAcoustical Society of America, 126, 2033–2051.

Bucki, M., Nazari, M., & Payan, Y. (2010). Finite element speaker-specific face model generation for the study of speech production.Computer Methods in Biomechanics and Biomedical Engineering,13, 459–467.

Catford, I. (1977). Fundamental problems in phonetics. Edinburgh,Scotland: Edinburgh University Press.

Gerard, J. M., Perrier, P., & Payan, Y. (2006). 3D biomechanicaltongue modelling to study speech production In J. Harrington &M. Tabain (Eds.), Speech production: Models, phonetic processes,and techniques (pp. 85–102). New York, NY: Psychology Press.

Hannam, A. G., Stavness, I., Lloyd, J. E., & Fels, S. (2008). Adynamic model of jaw and hyoid biomechanics during chewing.Journal of Biomechanics, 41, 1069–1076.

Huber, E. (1931). Evolution of facial musculature and facialexpression. Baltimore, MD: The Johns Hopkins Press.

Hughes, M. O., & Abbs, J. A. (1976). Labial-mandibularcoordination in the production of speech: Implications forthe operation of motor equivalence. Phonetica, 33, 119–221.

Kim, K., & Gomi, H. (2007). Model-based investigation of controland dynamics in human articulatory motion. Journal of SystemDesign and Dynamics, 1, 558–569.

Ladefoged, P. (1984). ‘‘Out of chaos comes order’’: Physical,biological, and structural patterns in phonetics. Proceedings ofthe Tenth International Congress of Phonetic Sciences, 83–95.

Lloyd, J. E., Stavness, I., & Fels, S. (2012). ArtiSynth: A fastinteractive biomechanical modeling toolkit combining multibodyand finite element simulation. In Y. Payan (Ed.), Soft tissuebiomechanical modeling for computer assisted surgery (pp. 355–394). Berlin, Germany: Springer.

MacNeilage, P. F. (1998). The frame/content theory of evolutionof speech production. Behavioral and Brain Sciences, 21,499–546.

MacNeilage, P. F., & Davis, B. L. (2000, April 21). On the origins ofinternal structure of word forms. Science, 288, 527–531.

Mooney, M. (1940). A theory of large elastic deformation. Journal ofApplied Physics, 11, 582–592.

Nazari, M. A., Perrier, P., Chabanas, M., & Payan, Y. (2010).Simulation of dynamic orofacial movements using aconstitutive law varying with muscle activation. ComputerMethods in Biomechanics and Biomedical Engineering, 13,469–482.

Nazari, M. A., Perrier, P., Chabanas, M., & Payan, Y. (2011).Shaping by stiffening: A modeling study for lips. Motor Control,15, 141–168.

Pessa J., Zadoo, V., Adrian, E. J., Yuan, C., Aydelotte, J., & Garza,J. (1998). Variability of the midfacial muscles: Analysis of 50hemifacial cadaver dissections. Plastic and ReconstructiveSurgery, 102, 1888–1893.

Rivlin, R. S. (1948). Large elastic deformations of isotropicmaterials: IV. Further developments of the general theory.Philosophical Transactions of the Royal Society of London. SeriesA, Mathematical and Physical Sciences, 241, 379–397.

Stavness et al.: Effects of Orbicularis Oris and Jaw Posture on Lip Shape 889

Rochet-Capellan, A., & Schwartz, J. L. (2007). An articulatory basisfor the labial-to-coronal effect: /pata/ seems a more stablearticulatory pattern than /tapa/. The Journal of the AcousticalSociety of America, 121, 3740–3754.

Schmidt, K. L., & Cohn, J. F. (2001). Human facial expressions asadaptations: Evolutionary questions in facial expressionresearch. Yearbook of Physical Anthropology, 44, 3–24.

Stavness, I., Lloyd, J., Payan, Y., & Fels, S. (2011). Coupled hard-soft tissue simulation with contact and constraints applied tojaw-tongue-hyoid dynamics. International Journal of NumericalMethods in Biomedical Engineering, 27, 367–390.

Storto, L., & Demolin, D. (2012). The phonetics and phonology ofSouth American languages. In L. Campbell & V. Grondona(Eds.), The indigenous languages of South America: A comprehen-sive guide (pp. 331–390). Berlin, Germany: De Gruyter Mouton.

Storto, L., & Demolin, D. (in press). Phonetics and phonology ofKaritiana. Memoires de l’Academie Royale des Sciences d’outremer de Belgique.

Weiss, J. A., Maker, B. N., & Govindjee, S. (1996). Finite elementimplementation of incompressible, transversely isotropichyperelasticity. Computer Methods in Applied Mechanics andEngineering, 135, 107–128.

890 Journal of Speech, Language, and Hearing Research N Vol. 56 N 878–890 N June 2013

http://jslhr.asha.org/cgi/content/full/56/3/878