Bite Force Estimation and the Fiber Architecture of Felid Masticatory Muscles



Increasingly, analyses of craniodental dietary adaptations take into account mechanical properties of foods. However, masticatory muscle fiber architecture has been described for relatively few lineages, even though an understanding of the scaling of this anatomy can yield important information about adaptations for stretch and strength in the masticatory system. Data on the mandibular adductors of 28 specimens from nine species of felids representing nearly the entire body size range of the family allow us to evaluate the influence of body size and diet on the masticatory apparatus within this lineage. Masticatory muscle masses scale isometrically, tending toward positive allometry, with body mass and jaw length. This allometry becomes significant when the independent variable is a geometric mean of cranial variables. For all three body size proxies, the physiological cross-sectional area and predicted bite forces scale with significant positive allometry. Average fiber lengths (FL) tend toward negative allometry though with wide confidence intervals resulting from substantial scatter. We believe that these FL residuals are affected by dietary signals within the sample; though the mechanical properties of felid diets are relatively similar across species, the most durophagous species in our sample (the jaguar) appears to have relatively higher force production capabilities. The more notable dietary trend in our sample is the relationship between FL and relative prey size: felid species that predominantly consume relatively small prey have short masticatory muscle fibers, and species that regularly consume relatively large prey have relatively long fibers. This suggests an adaptive signal related to gape. Anat Rec, 2012. © 2012 Wiley Periodicals, Inc.


The most important data for estimating the forces produced by the jaw adductors are (1) electromyographic (EMG) data on the activity of these muscles during relevant behaviors (e.g., ingestion and mastication), (2) physiological cross-sectional areas (PCSA) of the muscles, and (3) dimensions of the skull that approximate moment arms of the muscles and of the bite point.

Among felids, EMG data are available for only domestic cats (Gorniak and Gans, 1980; Gorniak, 1986; Thexton and McGarrick, 1994) and we are not in a position to collect these data on other species as many felid species are rare and unavailable for the invasive protocols required. Instead, we model masticatory forces by combining data on mandibular adductor anatomy and cranial geometry as if jaw adductor muscle activity is constant, uniform across the parts of the musculature, and maximal during bite events. We realize these are unrealistic assumptions but they are necessary given the state of the data. Furthermore, this type of estimate is common practice (e.g., Thomason, 1991; Wroe et al., 2005) and is considered to represent the bite force (BF) when the muscles are contracting maximally. We have also refined these estimates by incorporating EMG data on the working-side and balancing-side masticatory muscles (from Gorniak and Gans, 1980).

PCSA data are correlated with muscle force production and can be gathered from muscle tissue (Close, 1972; Weijs and Hillen, 1985; O'Connor et al., 2005; Anapol et al., 2008). PCSA is a function of muscle mass and fascicle length, and can be corrected based on estimates of pinnation as described in our prior protocol on jaw adductor PCSA in primates (Perry et al., 2011), which forms the basis for the protocol used here.

To approximate BFs, muscle PCSA must be scaled by muscle mechanical advantage. To do so requires measurements of the moment arm of the muscle and the moment arm of the bite. These can be measured directly from the skull using the muscle attachments as a guide and using realistic bite points along the tooth row. This method assumes that the jaw acts as a third-class lever as has been suggested by most authors (e.g., Turnbull, 1970; Picq et al., 1987; Hylander, 1992), and given the constraints on the movement of the mandibular condyle in felids (e.g., Turnbull, 1970), we will proceed under this common assumption.

In this study, we bring together data we have collected on felid jaw adductor dimensions and data on moment arms from the same individuals. First, we assess the scaling patterns of jaw adductor dimensions to better understand the influence of body size on the chewing system. Studies on primates have demonstrated that jaw adductor dimensions are generally isometric to body mass (Cachel, 1984; Anapol et al., 2008; Perry et al., 2011), but in some taxa they are positively allometric (Anapol et al., 2008).

Second, we assess any deviations from isometry in the context of feeding behavior. Interspecific differences in food material properties are likely to influence interspecific variation in jaw adductor PCSA: tougher and/or harder foods likely require more muscle force to achieve similar food breakdown. However, food mechanical properties vary little between felid diets (see Dietary Attribution section below). Furthermore, when felids consume vertebrate muscle generally there is very little true food breakdown (fracture into smaller particles). Prey size does differ between felid species—specifically prey size in relation to predator size is the variable of interest (see Dietary Attribution section below). This might influence muscle fascicle length as muscles need to stretch around a bite of food and still produce tension, and when all else is equal, killing a bigger prey item via the typically felid throat bite (Sunquist and Sunquist, 2002) implies a wider gape.

Last, we estimate moment arms of the jaw adductor muscles and of several bite points in our sample of felids. These are estimated from caliper measurements on skulls, using our studies of the locations and extents of the jaw adductor muscle attachments. We scale muscle PCSA to these moment arms to generate BF estimates at three points along the tooth row. Muscle data and osteological data in our sample were taken from the same individuals. Thus, we have avoided intraspecific variation within our calculations of BF. Since these estimates assume that all of the fascicles fire simultaneously, our results are only approximations of maximum BF. However, they represent taxa across nearly the entire felid body size range. We hope that these data will be useful in modeling BF and can be included meaningfully in finite element analyses and BF experiments.


Generally, scaling studies begin with the assumption of geometric similarity. Therefore, in the case of jaw adductor mass and scaling, we would hypothesize that:

H1a: Jaw adductor muscle mass scales isometrically with body mass (based on geometric similarity).

However, Hylander (1988) argued that, unlike limb muscles that must move and support the mass of the body, chewing muscles should not necessarily scale with isometry; rather, they should scale to food properties. Nevertheless, food properties do not necessarily lead to a scaling prediction—they do so if food properties themselves scale with body mass or some other mechanically relevant scalar (Lucas, 2004). Therefore without any a priori expectations about the scaling of food properties to body mass, we hypothesize that:

H1b: Jaw adductor muscle masses scale isometrically with body mass.

This is the null hypothesis with respect to geometric similarity, and if supported, would demonstrate that factors other than geometry determine jaw adductor muscle mass. One likely factor is food properties.

In the case of fiber length (FL), PCSA, and BF, our hypotheses are the same as above. For each variable, we hypothesize that it scales isometrically with body mass if there is geometric similarity. However, we consider the alternate hypothesis that the functional influences on this muscle will result in an allometric scaling relationship between muscle mass and body mass.

Food mechanical properties have been found to influence felid electromyography (Gorniak and Gans, 1980; Gorniak, 1986), and we hypothesize that food mechanical and geometric properties will correlate with the fiber architecture as well. Namely, we hypothesize that:

H2a: species that consume relatively large prey will have relatively large masticatory muscles to produce greater BF, and species that consume relatively small prey will have relatively small masticatory muscles. This assumes that the material properties of prey are constant across scale and that prey items are not brittle—thus requiring greater BF with increased volume.

H2b: species that consume relatively obdurate (hard or tough) foods, such as bones and cartilage, will have relatively high PCSA to produce greater BF compared with species that consume relatively soft prey foods, for instance muscle and connective tissue.

H2c: species that consume relatively obdurate (hard or tough) prey will have relatively high estimated BFs (as reflected by combining their PCSA and leverage to estimate BF) compared with species that consume relatively soft prey elements.

H2d: species that consume relatively large prey or that have relatively long canines (i.e., N. nebulosa) will have relatively long masticatory muscle fascicles (high FL) to accommodate wider gape, and species that consume relatively small prey will have relatively small masticatory muscle fascicles (low FL). Thus, fascicle length should correlate with relative prey size or canine size, independent of predator size because FL is the variable that relates to a predator's gape ability.



We dissected the masticatory muscles of nine species of felids (Table 1), represented by a total of 28 specimens. All but two of the specimens were from Carolina Tiger Rescue (CTR; formerly the Carnivore Preservation Trust), a nonprofit rescue facility in Pittsboro, North Carolina. CTR feeds its animals whole-carcass diets; the small cats are fed predominantly rodents and chickens, and the large cats are fed predominantly chickens and deer. The foods provided to carnivorans at most captive facilities are primarily nutritionally supplemented ground meat—generally horse or cattle (Wack, 2003), a diet that mimics the nutritional qualities of wild cats, but ignores the mechanical properties of their food. Because this study focuses on masticatory anatomy, it is important that our sample mimic the mechanical properties of wild felid foods. One major advantage of the CTR sample is that these specimens are from the same facility, having been cared for using the same practices and fed the same diets.

Table 1. Sample
Species (common name)Abbrev.N dissected (male: female)BM (kg)aJL (mm)aGMaDPSbDMPc
  • a

    Of the individual studied for fiber architecture.

  • b

    See Sunquist and Sunquist (2002) for lists of prey species upon which these assessments were based.

  • c

    All of the species in this study have dietary material properties that are more or less average for a felid “Av.”—they consume whole small vertebrates and soft bones of larger vertebrates (e.g., ends of ribs)—except for the jaguar which is capable of using its jaws to puncture large mammal skulls and tortoise and turtle shells (Sunquist and Sunquist, 2002).

Caracal caracal (caracal)C.c.7 (5:2)16.5991.4442.653.5Av.
Leptailurus serval (serval)L.s.3 (1:2)13.9088.7339.091Av.
Leopardus pardalis (ocelot)L.p.2 (2:0)11.5990.5042.162Av.
Lynx rufus (bobcat)L.r.1 (1:0)15.5086.2839.702.5Av.
Neofelis nebulosa (clouded leopard)N.n.1 (1:0)24.00127.0052.282.5Av.
Panthera onca (jaguar)P.o.2 (2:0)100.00177.0076.412Hard
Panthera pardus (leopard)P.p.3 (1:2)34.10123.8057.592.5Av.
Panthera uncia (snow leopard)P.u.2 (1:1)38.64124.7556.982.5Av.
Panthera tigris (tiger)P.t.8 (4:4)200.00240.0095.992.5Av.

The two specimens not from CTR are the bobcat (Lynx rufus) and the clouded leopard (Neofelis nebulosa). The former was a wild specimen donated by a local taxidermist and added to the sample to increase the taxonomic diversity of the small felids, and the latter was a specimen from the Smithsonian National Zoological Park (NZP) and added to the sample because of the special interest in the cranial anatomy of this species. Namely, if there is a relationship between prey size and FL, that is, influenced by gape (e.g., hypothesis 2d), then this taxon should also have long masticatory FL to accommodate its long canines—the relatively longest of all felids.

None of these specimens were sacrificed for the purpose of this study. All of the captive specimens died of natural causes or were euthanized because of ailments unrelated to their masticatory system, and the bobcat was supplied as a byproduct of the taxidermy industry.

The cleaned skulls of all individuals examined in this study were used to complete the osteological components of our study. All are available for study upon request (to AHR) except the Neofelis specimen which was retained by the NZP.


The muscles examined in this study were the superficial masseter (SM), deep masseter (DM), zygomatico-mandibularis (ZM), zygomatic temporalis (ZT), superficial temporalis (ST), deep temporalis (DT), and medial pterygoid (MP). Although not a jaw adductor, the digastric (Dig) was included as well. The lateral pterygoid (LP) muscle in felids is exceptionally small (less than a gram in the largest tigers that we dissected) and is also likely not a mandibular adductor—its fibers are oriented entirely anteroposteriorly and insert very close to the condyle. Because of its small size and indeterminate function (see discussion), we do not include analyses of the lateral pterygoid. For the purpose of this study, some of the muscles were dissected in groups rather than as separate muscles. For instance, the superficial, deep and zygomatic temporalis generally were removed together. Likewise, for several of the specimens, superficial and deep masseter were removed together. The muscular divisions and fiber orientations of these muscles (Fig. 1) are simpler than those found in primates (Perry et al., 2011) but their origins and insertions are relatively elaborate (Figs. 2 and 3).

Figure 1.

The complexity of correcting for pinnation in the masticatory muscle of felids. (A) schematic representation of a coronal section of the masticatory muscle of a felid. Broken lines represent aponeuroses. The thickness of those lines roughly corresponds to the thickness of the aponeuroses. Thick black lines represent bones: cranium, mandible (M) and zygomatic arch (Z). Thin black lines represent nonfascial/nonaponeurotic muscle boundaries. Gray lines represent fascicular orientations. DT: deep temporalis; ST: superficial temporalis; ZT: zygomatic temporalis; ZM: zygomatico-mandibularis; DM: deep masseter; SM: superficial masseter; MP: medial pterygoid; LP: lateral pterygoid (included for schematic completeness though not analyzed in this study); Dig: digastric (mandibular abductor). Note that the fibers of LP and Dig are oriented anteroposteriorly and thus are represented in this coronal section as gray dots. (B) Shows the deep masseter considered alone. Using Anapol and Barry's (1996) method corrects PCSA relative to the line of action of the muscle (arrow). This is, however, oblique to the vertical BF. (C) shows that correcting PCSA to a vertical BF requires knowledge of the in situ orientation of the muscle—information generally lost in the absence of coronal sectioning. Compare this figure to Fig. 5 in Perry, Hartstone-Rose and Wall (2011).

Figure 2.

Coronal schematic including mandibular adductor origins (o) and insertions (i). Muscle abbreviations are the same as those in Fig. 1.

Figure 3.

Mandibular adductor origins (o) and insertions (i), lateral view. Muscle abbreviations are the same as those in Fig. 1. Note: from lateral view some origins (i.e., ZM, MP) and insertions (i.e., MP, DT) cannot be viewed. Male P. onca from study. Scale is in centimeters.

All of the specimens were dissected fresh/frozen (i.e., none had been preserved in formalin or ethanol) and none exhibited signs of excessive freezer damage. Little fiber shortening is expected because the specimens used were preserved whole with the muscles intact and attached to the skull. All specimens had similar degrees of gape, close to centric relation. Because of uniform gape among subjects, we did not normalize measured FLs to a standardized sarcomere length. Thus, FL and PCSA values reported here reflect FL and PCSA at near-minimum gape. A named muscle (e.g., superficial masseter) is a sheet of fibers with a similar orientation, bounded by fascial planes or by bony surfaces.

The masticatory muscles were dissected and muscle mass (MM) was recorded for 19 of the specimens (the other nine specimens were studied to confirm myology for that species) and one individual of each species was subjected to chemical dissection for muscle architecture analysis. To do this, we followed a protocol described previously (Perry and Wall, 2008; Perry et al., 2011), in which each muscle was cooked in a 10% sulfuric acid solution at 70-degree Celsius until enough connective tissue was dissolved to free muscle fascicles for measuring. These fascicles were then measured (hereafter we often refer to this fascicle length using the short-hand FL). The PCSA for each muscle group was calculated using a formula modified from Schumacher (1961):

equation image

Where q is PCSA, m is muscle mass, l is FL and p is the density constant of muscle (1.0564 g/cm3; Murphy and Beardsley, 1974). Previous studies (Perry and Wall, 2005; Perry et al., 2011) also calculated reduced PCSA (RPCSA or RPCA)—a variable adopted from Anapol and Barry (1996) that takes into account pinnation angles to estimate PCSA perpendicular to the pull of the muscle vector.

Previously we found that pinnation has only a small effect on the RPCSA equation given the actual pinnation values for most primate chewing muscles (Perry et al. 2011). Furthermore, the Anapol and Barry method for accounting for pinnation was developed using the relatively simple, bipinnate, limb flexor and extensor muscles. Those muscles are generally pinnate predominantly in a single plane (mediolateral) and their excursion is nearly entirely linear—bringing the insertion and origin closer together in the same plane as the muscle's orientation. Unfortunately, this method does little to capture the true complexity of fiber orientation in chewing muscles; the masticatory muscles are pinnate in multiple planes—not just the coronal plane (as corrected in most analyses that correct for chewing muscle pinnation) but also are generally fan shaped (as is the case for the superficial and deep temporalis) or have other anteroposterior inclinations (as is the case for the zygomatic temporalis and the masseter muscles; Fig. 2). Likewise, in addition to the complexity added by these pinnation angles in parasagittal planes, the masticatory muscles also likely pull in planes that are neither parallel nor perpendicular to the muscle's long axis. Thus, when their pinnation is “corrected” relative to this long axis (Fig. 1B), it should rather be adjusted relative to the orientation of pull (Fig. 1C)—a modification quite different from the one developed by Anapol and Barry. Furthermore, while all pinnate muscle fascicles likely change orientation throughout contraction, the angular excursion of the masticatory muscles add greatly to this complexity; unlike most limb muscles, the orientations of the masticatory muscles as whole units change—greatly affecting their moments. For these reasons, we have elected not to reduce PCSA here, but rather restrict all of our analyses to specimens of equivalent gapes (namely, near occlusion) and therefore make statements about the masticatory architecture and BF at that gape.

Total MM, Total PCSA (the sum of the masses and PCSAs, respectively, of the three adductor muscle groups), and Average FL were calculated (Table 2). Average FL is a weighted average of the mandibular adductors calculated according to the following formula which scales FL for each muscle to the muscle's weight:

equation image
Table 2. Masticatory muscle mass (g), average FL (cm), and PCSA (cm2)
SpeciesMasseter mass (g)Temporalis mass (g)Medial pterygoid mass (g)Total muscle mass (g)Average FL (cm)Total PCSA (cm2)
  • a

    Replaced missing values as described in the methods

Caracal caracal27.6051.917.9387.441.5662.81
Leptailurus serval11.1344.017.8963.031.0961.02
Leopardus pardalis28.1654.464.6887.311.4367.11
Lynx rufus24.2529.675.7559.671.3746.83
Neofelis nebulosa70.86101.39a9.00181.252.0195.37
Panthera onca268.28482.0059.00809.282.21388.40
Panthera pardus71.00146.0020.00237.001.92145.38
Panthera uncia65.85130.9419.24216.031.75134.80
Panthera tigris543.00921.00110.05a1574.052.76622.42

Here, FLx is the Average FL, FLMS, FLTMP, and FLPT are the average FL and mMS, mTMP, and mPT are the masses of masseter, temporalis and medial pterygoid, respectively.

BF at three different points along the tooth row (the tip of the canine, the primary cusp of the third maxillary premolar, and the upper carnassial notch) was then calculated by combining the PCSA data with a constant expressing the load that can be produced by a given cross-section of skeletal muscle (3 kg/cm2; Close, 1972) and the leverages of those muscles. The 3 kg/cm2 estimate is from the extensor digitorum longus of a rat at optimum length and is considered a reliable estimate by Close, although estimates for mammalian skeletal muscle vary widely, as much as from 1 to 5 kg/cm2. Leverage estimates were gathered from lateral photographs of the crania of each of the dissected specimens with the specimens posed in dental occlusion (Fig. 4). Lateral photographs of the separate crania and mandibles were used to visualize surfaces that overlap when the skull is in occlusion. The load arm is the measurement from the condyle to the bite points for each of the three chosen bite points. The lever arm for each of the muscle groups is the length of a perpendicular line that connects the condyle to the line of action for the muscle group. The line of action is a line connecting the centroid of the origin of the muscle group to the centroid of the insertion of that muscle group. This model assumes that the movement at the condyle is purely rotational and not translational—a fair assumption in felids (Turnbull, 1970; Herring and Herring, 1974; Herring, 1992). It also is also an oversimplification of these masticatory leverages because it does not account for the dynamism of these lever arms at different degrees of gape or the specific orientation of the BF (though we assume it to be perpendicular to the moment). However, since the masticatory muscles were dissected and their fibers measured at near occlusion, when combining these data and the leverage data gathered at occlusion, these estimated BFs can be assumed to represent BFs at occlusion. These BF estimates also assume that all muscles are maximally active at once.

Figure 4.

Leverage measurements for the (I) lever arms and working-side load arms and (II) the balancing-side load arms. The red lines represent the line of action of mandibular adductors (solid = temporalis, dashed = masseter, dotted = medial pterygoid). These lines connect the centroids (red circles) of the origin areas to the centroids of the insertion areas. The lever arm (yellow lines) for each mandibular adductor are the lines connecting the temporomandibular joint perpendicularly to the red lines. These lines differ in length and orientation at different gapes (Christiansen, 2011), but for the purposes of this study, we measured them in occlusion, nearly the same gape state as the fiber architecture was dissected. Thus these BF estimates should be considered BF at occlusion. The load arms (blue lines) were measured, for the working-side (Ia–c.) from the temporomandibular joint to each of three bite points: upper canine tip (a), upper third premolar tip (b), and upper carnassial notch (c). The load arms for the balancing-side were the sum of the distance from the temporomandibular joint to the superio-anterior corner of the mandibular symphysis (green line) and the distance from that point to each bite point (IIa–c.). We follow Davis (1955) in assuming that the load arms are the direct distance between the joint and the bite points, assuming the jaw can only open in an arc centered on the joint. Male P. onca from study. Scale is in centimeters.

The working-side BF estimates were then calculated using the following equations:

equation image
equation image
equation image

where WBFCA, WBFPM, and WBFCM, are the working-side BFs at the canine, premolar and notch of the lower carnassial (M1), respectively, c is the force constant of muscle cross-section (3 kg/cm2; Close, 1972), qMS, qTMP, and qPT, are the PCSA of the masseter, temporalis and medial pterygoid, respectively, LMS, LTMP, and LPT, are the lengths of the moment arms of the masseter, temporalis and medial pterygoid, respectively, and LCA, LPM, and LCM, are the lengths of the bite moment arms at the maxillary canine, third premolar and carnassial notch, respectively.

Because felids have relatively fused mandibles and their interdigitated symphyses likely transmit large amounts of force from the balancing side muscles, in an effort to more fully estimate total BFs at each bite point, we added an estimate of the contribution to this BF of the balancing-side muscles. This was calculated by combining EMG data with balancing-side leverages. Gorniak and Gans (1980) reported EMG amplitudes for both working- and balancing-side muscles during mastication by a domestic cat. Assuming that all felids fire their muscles according to this pattern (a necessary assumption given the lack of EMG data on other species), and that the reported working-side amplitudes represent maximal working-side BFs, we created a ratio of the working-side and balancing-side amplitudes for each muscle and multiplied these by the PCSA and muscle moment arms for each respective muscle (which are the same as the working-side moment arms), and combined this with the bite point moment arms to the balancing-side condyle—a measure of the distance from the balancing-side mandibular condyle to the mandibular symphysis plus the distance from the mandibular symphysis to the bite point on the working side (Fig. 4). Thus, the balancing-side contribution to the BF was calculated as:

equation image
equation image
equation image

Where BBFCA, BBFPM, and BBFCM, are the BF contributions of the balancing-side muscles at the canine, premolar and notch of the lower carnassial (M1), respectively, EMS, ETMP, and EPT are the ratios of balancing-side to working-side EMG amplitudes taken from Gorniak and Gans (1980) for the masseter, temporalis and medial pterygoid, respectively, L'CA, L'PM, and L'CM, are the balancing-side bite moment arms at the maxillary canine, third premolar and carnassial notch, respectively, and all of the other variables are the same as those used to calculate the working-side BF. Total BF is the simple sum of these two metrics for each bite point. Thus:

equation image
equation image
equation image

Three of the 27 muscle groups (the N. nebulosa temporalis, and the C. caracal and P. tigris medial pterygoid) were damaged before dissection. These muscles were omitted from the individual analyses of muscle mass scaling, but their values (weight, FL, and the PCSA derived from those two variables) were estimated to include them in the prediction of Total MM, Average FL, Total PCSA, and BF for all species. The estimates were calculated in SPSS (version 19.0) following a similar approach to that used by Yokley and Churchill (2002) to replace missing data using multiple regression. Missing values for mass and FL were predicted from the least-squares equation (created using the beta weights from the MRA analyses) based on the observed values of mass and FL, respectively.

These variables (BF, Average FL, Total MM and Total PCSA) were regressed in JMP (version 9.0; SAS) against Body Mass (BM), a Geometric Mean (GM) of eight craniomandibular measurements (Tables 1 and 3) and Jaw Length (JL) to study overall scaling effects. BM is the weight taken at the last antemortem physical exam for all captive specimens; a weight estimate was provided by the supplier of the bobcat. JL (described in Table 3) is often used as the independent variable in studies of masticatory scaling (Hylander, 1979; Hylander, 1985; Vinyard et al., 2003; Perry et al., 2011). Before regression, certain variables were reduced to linear variables to simplify allometric evaluations—standardizing the predicted slopes to 1; thus, the cube root of BM and MM and the square root of PCSA and the BF measures were taken and logged. We assessed BF as having units of area according to the procedures of Huber et al. (2006) because BF is essentially an area measurement multiplied by a constant (the force produced per area of PCSA) and dimensionless ratio (a lever arm divided by a lever arm, efficiency ratio). Species were categorized according to diet (see section below) both in terms of Dietary Prey Size (DPS) and Dietary Mechanical Properties (DMP) to evaluate the hypotheses relating dietary variation and masticatory muscle architecture.

Table 3. Measurements that make up the geometric mean
  • a

    This landmark is defined as the posterior edge of the mandibular condyle.

  • b

    Since only the most senescent felids (a group excluded from this study) have complete postorbital bars, this measure might not pass through the centroid of the orbit but is measured in the superior-most projection of the inferior (zygomatic) post orbital process.

Bizygomatic breadth: instrumentally determined maximum breadth of the skull at outside edge of bone
Length of cranium: from inion to interdentale (Bass, 1995)
Height of the cranium: measured at highest point on the frontal bones at the post orbital processes to the posterior edge of the palate in the midsagittal plane
Jaw length (JL): length of mandible from condylarea to infradentale (Bass, 1995)
Jaw height: from the carnassial notch to the inferior border of the mandible in the coronal plane (Hartstone-Rose, 2008)
Jaw width: width of the mandible inferior to the carnassials, perpendicular to Jaw Height (Hartstone-Rose, 2008)
Orbital height: maximum instrumentally determined measure of the internal margin of the orbit
Orbital width: width of bony orbit measured perpendicular to Orbital Heightb

Dietary Attribution

Species were sorted into categories based on the relative dietary prey size (DPS) and dietary mechanical properties (DMP) (Table 1). Sorting was based on published dietary descriptions (Hemmer, 1972; Mazak, 1981; Seymour, 1989; Larivier and Walton, 1997; Murray and Gardner, 1997; Sunquist and Sunquist, 2002; Wilson et al., 2009). DPS was determined as follows: a felid species was assigned to DPS Category 1 if it consumes mostly prey, that is, smaller than one order of magnitude below its own body mass. Category 2 represents species that mostly consume prey species that are between its own body size and one order of magnitude smaller than its own body size. Category 3 represents species that regularly consume prey items larger than its own body size. Finally, species were assigned to Category 4 if the majority of its prey items are larger than itself. No species of felid in this study consumes solely species smaller than one order of magnitude of its body mass or larger than its body mass. For most of the species in our sample, some study locations would indicate one dietary category attribution while at least one other study location would indicate another dietary category. In these cases, the felid species is assigned to both DPS categories (in Table 1) and the arithmetic mean of those two scores is used in the analyses (e.g., of the 12 field studies for Panthera pardus, 10 indicated a score of 2, and two indicated a score of 3 (Sunquist and Sunquist, 2002), so that species is assigned a DPS score of 2.5 in our analyses). Since different numbers of field studies have been conducted on different species and different standards were used in the various field studies, no effort was made to assign finer scores than half-integers.

The mechanical properties of the diet of the felids in this study (DMP) were also assessed from the same literature. For each species, the hardest food included in the diet was noted and a hardness score was assigned based on this (Table 1). Unfortunately, our sample does not include any of the truly “hypercarnivorous” (species that consume only flesh; sensu Holliday, 2001; Holliday and Steppan, 2004) felids such as the cheetah, Acinonyx jubatus (Morseback, 1987; Van Valkenburgh, 1996; Sunquist and Sunquist, 2002). Instead it is almost entirely made up of generalized felids that consume whole carcasses of their smallest or softest prey and probably also consume the softest parts of the skeleton (e.g., the ends of the ribs and costal cartilages) of their larger prey. Though no felid is truly durophagous (i.e., compared to hyaenids; Werdelin, 1983; Werdelin and Solounias, 1991; Hartstone-Rose and Wahl, 2008; Hartstone-Rose, 2011), our sample does include what is likely the most durophagous felid: the jaguar (Panthera onca), a species known to consume tortoises and turtles by breaking their shells and known for killing large mammalian prey by puncturing their skulls via canine biting (Sunquist and Sunquist [2002] and references therein). Thus our sample mainly focuses on species with diets of moderate mechanical properties for felids; the jaguar is expected to be an outlier in terms of its ability to generate force (i.e., high PCSA).


Scaling of Muscle Mass

The values for MM, FL, PCSA and BF are presented in Tables 2 and 3. Mass for each of the jaw adductor groups scales isometrically tending toward positive allometry and have high correlations with body mass, and jaw length (Figs. 5 and 6 and Tables 4 and 5). At an alpha of 0.05, the slope of isometry (equal to 1) is barely contained within the slope confidence limits for each of the masticatory muscles. When the cranial geometric mean is used as the independent variable, the correlations increase, narrowing the slope confidence interval and raising the lower confidence limit for the slope such that it is above isometry (again, for all but the medial pterygoid). Thus, the masticatory muscle masses scale with slight positive allometry, but they do so in a statistically significant manner only when regressed against the cranial geometric mean. In this respect, larger cats appear to have relatively slightly larger masticatory muscles.

Figure 5.

RMA regressions masticatory muscle mass against body mass. Temporalis (squares and long broken line), masseter (diamonds and short broken line), medial pterygoid (triangles and dotted line). The mandibular abductor digastric (circles and solid line) is included for comparison. See Table 1 for abbreviations of species names, and Table 5 for descriptive statistics.

Figure 6.

RMA regressions of total muscle mass against body mass (A), jaw length (B), and the cranial geometric mean (C). See Table 1 for abbreviations of species names, and Table 5 for descriptive statistics.

Table 4. Bite force estimates at three bite points (in kilograms)
SpeciesWorking-side BFCAWorking-side BFPMWorking-side BFCMBFCAaBFPMaBFCMa
  • a

    Combined balancing-side and working-side.

Caracal caracal47.7960.5878.5977.8785.90100.89
Leptailurus serval43.1956.3472.2268.0577.2490.96
Leopardus pardalis47.6261.9482.2273.3983.83101.70
Lynx rufus32.9843.8356.5951.5159.3070.49
Neofelis nebulosa69.61102.03132.88108.95133.37161.34
Panthera onca314.15457.19568.83503.57606.62705.79
Panthera pardus102.59139.10180.46166.25190.83226.61
Panthera uncia94.86127.82165.78148.59172.35205.78
Panthera tigris428.01594.08707.63703.74813.63910.12
Table 5. Descriptive statistics for RMA regressions against the cube root of body mass
y-variableaSlope (β)by-interceptrLower β CLcUpper β CLc
  • a

    All variables were converted to base 10 logarithms for regression.

  • b

    Because the cube root and square root of all of the cubic and square variables (respectively) were taken prior to regression, the expected slope of isometry for all of these regressions is 1, and where 1 is outside of the confidence limits (i.e., “Lower β CL”>1), then the slope of the line is statistically allometric.

  • c

    Alpha = 0.05

Log masseter mass (g) ⁁1/31.27−1.310.960.971.67
Log temp mass (g) ⁁1/31.19−1.100.980.951.49
Log MP mass (g) ⁁1/31.17−1.360.980.961.44
Log digastric mass (g) ⁁1/31.23−1.430.970.961.57
Log total adductor mass ⁁1/31.18−1.010.980.991.41
Log average FL0.87−1.060.890.531.43
Log total PCSA ⁁1/21.38−1.020.981.151.65
Log BFCA ⁁1/21.41−1.050.981.141.75
Log BFPM ⁁1/21.44−1.050.981.161.78
Log BFCM ⁁1/21.41−0.970.981.131.77

The correlations between the individual muscle masses and body size are so high (Tables 46), with almost no scatter around the line of fit, that no dietary signal can be discerned beyond this general scaling pattern. Contrary to hypotheses 1b and 2a there does not appear to be any difference in muscle mass emphases based on variation in diet—no species falls particularly far from the line of muscle mass regression in any way that appears to relate to diet.

Table 6. Descriptive statistics for RMA regressions against jaw length
y-variableaSlope (β)by-interceptRLower β CLcUpper β CLc
  • a

    All variables were converted to base 10 logarithms for regression.

  • b

    The expected slope of isometry for all of these regressions is 1.

  • c

    Alpha = 0.05.

Log masseter mass (g) ⁁1/31.16−1.820.970.941.43
Log temp mass (g) ⁁1/31.08−1.560.990.941.24
Log MP mass (g) ⁁1/31.09−1.870.920.691.73
Log digastric mass (g) ⁁1/31.12−1.920.970.871.44
Log total adductor mass ⁁1/31.08−1.490.990.961.21
Log average FL0.80−1.420.930.551.16
Log total PCSA ⁁1/21.26−1.580.981.051.51
Log BFCA ⁁1/21.29−1.600.991.101.51
Log BFPM ⁁1/21.31−1.610.991.111.54
Log BFCM ⁁1/21.29−1.530.991.081.53

Scaling of FL

There does appear to be a dietary signal in the scaling of masticatory FL (Fig. 7). FL has the weakest correlation with the body size proxies among all of the dependent variables (Tables 57). Thus, although the slopes are indistinguishable from isometry, there is substantial scatter around the line. Importantly, this scatter fits the pattern predicted based on diets (supporting hypothesis 2b). As predicted in hypothesis 2b, the clouded leopard (N. nebulosa) has the highest residuals, suggesting that its muscles are adapted for stretching to accommodate a wider gape to allow food to pass beyond the relatively long canines.

Figure 7.

RMA regressions of average FL against body mass (A), jaw length (B), and the cranial geometric mean (C). See Table 1 for abbreviations of species names, and Table 5 for descriptive statistics.

Table 7. Descriptive statistics for RMA regressions against cranial geometric mean
y-variableaSlope (β)by-interceptrLower β CLcUpper β CLc
  • a

    All variables were converted to base 10 logarithms for regression.

  • b

    The expected slope of isometry for all of these regressions is 1.

  • c

    Alpha = 0.05

Log masseter mass (g) ⁁1/31.31−1.670.981.101.56
Log temp mass (g) ⁁1/31.22−1.430.991.081.38
Log MP mass (g) ⁁1/31.18−1.650.940.821.71
Log digastric mass (g) ⁁1/31.27−1.790.981.031.56
Log total adductor mass ⁁1/31.22−1.351.001.151.30
Log average FL0.90−1.320.930.631.29
Log total PCSA ⁁ 1/21.43−1.420.991.261.62
Log BFCA ⁁1/21.46−1.440.991.251.69
Log BFPM ⁁1/21.48−1.450.991.271.72
Log BFCN ⁁1/21.45−1.370.991.241.70

The cat species that eats the relatively largest prey (C. caracal) also has relatively high residuals and the species that eats the relatively smallest prey (Lep. serval) has the relatively lowest residuals (Fig. 8). Unfortunately, the sample sizes (especially within dietary groups) are so small that these trends cannot be assessed using conventional statistics (i.e., with N = 1 in each of the highest and lowest DPS groups, ANOVA comparisons are not appropriate), but given that these residuals conform so well to our initial predictions, they are worth noting. When fit with OLS regression lines, the whole sample (regressed against body mass) has an r2 value of 0.36 and when N. nebulosa is excluded from the regression (because its FL is likely influenced by the gape necessitated by its long canines and not its DPS) the coefficient of determination is 0.49. In short, though the sample sizes are insufficient for statistical separation based on dietary categorization, it does appear that felids that require larger gape do have longer FL and those that consume the relatively smallest prey do have the smallest FL.

Figure 8.

Residuals of the RMA regression of average FL against the cube root of body mass, by dietary prey size. OLS regression lines for the whole sample (solid) and sample excluding N. nebulosa (dotted) with r2 of 0.36 and 0.49, respectively. See Table 1 for abbreviations of species names.

Scaling of PCSA

PCSA correlates tightly with the body size proxies and displays significantly positive allometry (Fig. 9 and Tables 57). The most durophagous species in the sample is P. onca (jaguar), with all other species representing a standard felid diet in terms of dietary mechanical properties (Table 1). Though the jaguar plots, as predicted, above the regression line (Fig. 9), the residual is not very high. By some measures, it is not as high a residual as that found in Leo. pardalis—a species with no reported need for exceptional durophagous abilities.

Figure 9.

RMA regressions of PCSA against body mass (A), jaw length (B), and the cranial geometric mean (C). See Table 1 for abbreviations of species names, and Table 5 for descriptive statistics.

Scaling of BF

Our BF estimates are the product of our PCSA findings and leverage measurements for both working-side and balancing-side masticatory muscles (Fig. 10 and Tables 38). Given that cranial shape is obviously correlated to leverage and that it is very conservative in felids across body sizes, it is not surprising to find that BF scales nearly identically to PCSA. BF is highly correlated with and scales with positive allometry with respect to body size proxies (Tables 46). However, by this measure, the jaguar (P. onca) does have a more substantial positive residual which results from the product of the somewhat high PCSA and cranial adaptations that increase the bite leverage in that taxon (Fig. 10).

Figure 10.

RMA regressions of BF at the canine (diamonds and solid line), third premolar (squares and dotted line), and carnassial notch (triangles and dashed line) against body mass. See Table 1 for abbreviations of species names, and Table 5 for descriptive statistics.

Table 8. Total bite force (kg)
SpeciesCanineP3Carnassial notch
Caracal caracal77.8785.90100.89
Leptailurus serval68.0577.2490.96
Leopardus pardalis73.3983.83101.70
Lynx rufus51.5159.3070.49
Neofelis nebulosa108.95133.37161.34
Panthera onca503.57606.62705.79
Panthera pardus166.25190.83226.61
Panthera uncia148.59172.35205.78
Panthera tigris703.74813.63910.12

The other two apparent outliers in the RMA regression of BF against body mass (Leo. pardalis and Ly. rufus) probably can be explained best as variation in the body mass measurements and are probably not dietary signals. These taxa are outliers only in the body mass regression (Fig. 10) and fall close to the line in the regressions of JL and GM (see discussion).



The basic anatomy of felid chewing muscles has been known for over a century. Toldt produced the definitive work on the subject (1904–1905), having performed a detailed dissection of the chewing musculature of a domestic cat. The resulting description was sufficiently detailed and accurate that Turnbull translated it into English and quoted it in its entirety in his monograph of the chewing apparatus of mammals (Turnbull, 1970).

Toldt (1904- 1905) described the masseter, zygomatico-mandibularis, and medial (internal) pterygoid. To this, Turnbull (1970) added layered illustrations and a description of the temporalis musculature. To summarize these contributions, the masseter was found to have two layers (superficial and deep), the zygomatico-mandibularis is a single layer of fascicles deep to the masseter, and the temporalis was found to have two layers (superficial and deep). Stark (1933) and Fiedler (1953) found the zygomatico-mandibularis to have its own separate innervation, and thus considered it a muscle distinct from the masseter. Schumacher (1961) described the anatomy of felids based on one adult female Puma concolor, one adult female and one juvenile male Panthera leo, and one adult Pa. pardus. He reported muscle weights as well as cross-sectional areas. Schumacher described several fascial planes for the masseter, dividing the muscle into at least six layers. The most superficial layer corresponds to the layer typically referred to as the superficial masseter, whereas the deepest layer corresponds to the zygomatico-mandibularis. Thus, the other layers correspond to the deep masseter, a relatively small part of the masseter musculature.

It is difficult to sort out the classification of the masseter group. A survey of the classical literature (Saban, 1968) reveals that between two and six layers have been described for the mammalian masseter. In some cases, a layer is not entirely distinct from adjacent layers. Furthermore, authors disagree on the classification, even for the same taxon. Here we have opted for a simple classification into three layers: superficial masseter, deep masseter, and zygomatico-mandibularis (see anatomical description below). Though there is disagreement about the nomenclature for the individual jaw adductors muscles (Toldt, 1904- 1905; Edgeworth, 1935; Fiedler, 1953; Saban, 1968; Turnbull, 1970; Gaspard et al., 1973a–c; Cordell, 1991), we have essentially followed the scheme of muscular divisions provided by Turnbull (1970). However, as we have chemically dissected all muscles, the fascicles we measured represent a sampling of the entire musculature.

Our findings entirely corroborate those of Turnbull (1970) regarding the masseter. We observed in every specimen that an anterior part of the superficial masseter directly inserts on a raphe shared with the medial pterygoid muscle. This corresponds to the anterior-most fleshy lobe described by Toldt (1904- 1905). Like Turnbull (1970), however, we found that in most cases some part of the superficial masseter inserted on bone. Like Turnbull (1970), we also did not observe a clear separation of the superficial masseter into four lobes in the manner Toldt (1904- 1905) described. However, like Toldt, but unlike Turnbull, we found the zygomatico-mandibularis to be readily separable from the overlying deep masseter. A thick aponeurosis covers the zygomatico-mandibularis on its superficial surface and there is a distinct plane of separation between the two muscles in all taxa studied. The zygomatico-mandibularis in felids is unusual among mammals in having a strong posterior inclination (origin posterior to insertion) to its fibers. The muscle likely adducts and retracts the mandible. However, the fibers converge toward the origin, passing very close to the temporomandibular joint, and thus close to the point of rotation of the chewing system. This suggests that the leverage of the muscle at the joint is very poor (as for the lateral pterygoid). The role of the zygomatico-mandibularis of felids might be to prevent dislocation of the joint by resisting ventral and anterior movement of the condyle with respect to the glenoid fossa.

For the most part, authors agree about the classification of the temporalis (see Saban, 1968). The temporalis consists of the zygomatic temporalis, superficial temporalis, and deep temporalis. Our findings confirm those of Turnbull (1970) with respect to the temporalis. In many cases it was hard to completely divide the zygomatic temporalis from the superficial temporalis.

There is little disagreement about the medial pterygoid. Whereas the primate medial pterygoid has very complex fiber orientation (Perry et al., 2011), the felid medial pterygoid appears to be unipinnate or nearly so. Furthermore, we did not observe the strong division into two parts described by Toldt (1904- 1905). However, we did not dissect the muscle apart completely as we needed to preserve intact fibers for calculating PCSA.

Toldt's description of the lateral pterygoid differs substantially from that of Turnbull. Toldt reported that this muscle is extremely small and consists of a single belly of straight fibers running from bony origin to bony insertion. By contrast, Turnbull described this muscle as two bellies spiraling around each other, the first with a fleshy origin and a tendinous insertion, the second with a tendinous origin and a fleshy insertion. As suggested by Toldt, we found that the lateral pterygoid is very small in felids. Even in the largest tiger we dissected, the lateral pterygoid consisted of only a few fascicles and weighed less than 1 g. We did observe a small tendon within the muscle, as described by Turnbull (1970). We did not include this muscle in our quantitative analyses because it probably performs little function during mastication. The lateral pterygoid is most likely primarily responsible for anterior translation of the mandibular condyle in primates (Grant, 1973; McNamara, 1973). However, the small size of the lateral pterygoid in felids, along with the osteologically constrained temporomandibular joint, suggest that anterior translation is not an important feature of felid mastication.

In many mammals, including primates, the digastric is generally comprised of two bellies (an anterior and a posterior) connected by a tendon (Standring, 2008). The felid digastric is a very simple muscle with one belly and straight fibers running nearly the entire length. We detected neither the typical long, thick digastric tendon, nor the thin and barely distinct tendon suggested by Toldt (1904- 1905).

Though three muscles of the temporalis group and three muscles of the masseter group can be identified in felids, the fascial planes separating adjoining muscles are generally incomplete and muscle fibers from successive layers interdigitate. This might reflect a lesser degree of functional compartmentalization in these muscles in felids relative to some other mammalian groups. Felids have relatively restricted jaw movement, partly due to a fairly restricted temporomandibular joint. Felids mainly adduct and abduct their jaws with little translation, and therefore do not require as many separate muscles (Turnbull, 1970; Christiansen, 2011).

Comparison of Independent Variables

In our examinations of masticatory scaling (e.g., Hartstone-Rose and Perry, 2007a, b; Perry et al., 2011), we usually prefer to regress our dependent variables against measured body mass because that variable should be the best proxy for an overall body size and it has such well-established ecological correlates in and of itself (Clutton-Brock and Harvey, 1983). However, BM is often measured inconsistently and even when measured consistently, it varies so much throughout an animal's life (Glander et al., 2006; Hartstone-Rose and Perry, 2011) that it can be difficult to hypothesize about correlations. For instance, should masticatory variables scale to an animal's maximum lifetime body mass, its average adult body mass or the body mass measured closest to the time of study (Hartstone-Rose and Perry, 2011)? Despite these difficult questions, it is useful to look at scaling in relation to overall body mass because the latter is key to understanding energetics (via well-established mass/metabolic rate relationships), activity budgets, and interspecific ecological relationships (e.g., predator-prey relationships).

One alternative is to scale masticatory variables to a functionally meaningful anatomical measurement. Mandible length is often used as a scaling variable because it is a rough proxy for the load arm of an anterior bite (Hylander, 1979; Daegling, 2001; Vinyard et al., 2003; Vinyard and Hanna, 2005). If posterior bites are of greater interest, then the distance between the mandibular condyle and a molar could be used (Taylor et al., 2012). Instead of scaling to these variables, we have explicitly included them in the estimate of BF by measuring different hypothetical load arms at different bite points. This allows us to examine BF as a single variable and consider its scaling in relation to body size.

To sidestep the problems of periodic body mass fluctuations, we also use (e.g., in Perry, 2008; Perry and Hartstone-Rose, 2011; Perry et al., 2011) a cranial geometric mean as a proxy for overall body size. BM, JL and GM all yield very similar (statistically indistinguishable) regression results across all of the taxa that we have studied, though this is especially true for the morphologically homogenous Felidae.

The dependent variables appear to scale with higher correlation to the cranial geometric mean than they do to body mass measurements (e.g., this study and Perry et al., 2011). In the case of the current study, this appears to be evident particularly with the ocelot and bobcat (P. pardalis and Ly. rufus, respectively) specimens; consistently (Figs. 5–10), the ocelot plots significantly above the line and the bobcat plots significantly below the line when BM is used as the independent variable. However, in the regressions including JL and GM as the independent variables, those two taxa plot much closer to the line. While it is possible that the ocelot specimen studied has a relatively large head and the bobcat has a relatively small one (affecting both GM and JL), it is possible that the mass of each of the specimens was mismeasured or for some unknown reason is a poor estimate of body size. For instance, the weight value we used for the ocelot was taken shortly before death and it may be that that individual lost significant weight due to illness before its death. In this respect, we encourage researchers to report findings in terms of all three measures.

The Relationship Between Masticatory Fiber Architecture Outliers and Diet

Though Felidae as a family represents fairly narrow dietary diversity (remarkably, across their substantial size range, felids consume relatively similar foods) variation in the fiber architecture of their masticatory muscles does seem to indicate diet related adaptations. Namely, the most durophagous species in our sample (the jaguar) does plot above the regression line for both PCSA and predicted BF. The dietary signal appears to be stronger regarding the relationship between masticatory FLs and prey size.

As predicted, the jaguar plots above the regression lines for PCSA and BF. However, the residual is not very high. This would seem to suggest that the jaguar is not adapted for producing exceptionally high forces with its soft tissue masticatory anatomy. However, the positive allometry in PCSA and BF means that larger felids in general have relatively larger masticatory cross-sectional areas. This trend suggests that the largest felids (P. onca and P. tigris) are adapted to produce relatively higher forces. Thus, though the field reports (cited in Sunquist and Sunquist, 2002) do not indicate a particularly obdurate diet in tigers, by virtue of this unexpected positive allometry, their soft tissue masticatory capabilities seem to indicate an ability to produce relatively strong adduction compared to other felids. Interestingly, it is the combination of two subtle trends (one soft-tissue—masticatory PCSA—and one hard-tissue—jaw leverages) that combine to increase the force capability in the jaguar. Unfortunately, the sample sizes examined in this study, particularly in relation to the dietary homogeneity of the lineage, are too small to confirm this finding statistically.

Likewise, the relationship between masticatory FL and dietary prey size is also difficult to support statistically with this small sample size. However, the apparent relationship is striking: the felid that consistently consumes the relatively smallest prey (the serval—Lep. serval) does indeed have the relatively shortest masticatory muscle fibers (Fig. 8). Its sister taxon (the caracal—C. caracal) consistently consumes prey larger than its own body mass, and this species has amongst the relatively longest FL suggesting adaptations related to the gape necessities of diet. This is confirmed by the other apparently gape adapted species—the clouded leopard (N. nebulosa)—which, as predicted, has the relatively longest FL, presumably to accommodate its relatively long canines.

Exploration of these dietary trends will benefit from the addition of other carnivoran taxa from more dietarily diverse families. Not only will this allow the statistical evaluation of the influence of dietary mechanical properties on masticatory adaptations, but including families with greater cranial shape variation (e.g., long and short faced canids, mustelids and ursids) will allow greater evaluation of the influence of jaw length on gape and leverage on BF production.


Our results demonstrate that masticatory muscle masses scale isometrically tending toward positive allometry when regressed against body mass and jaw length. The slope is statistically greater than one for all muscles (except the highly variable medial pterygoid) when regressed against a geometric mean of cranial variables. PCSA, and BF scale with significant positive allometry for all three scaling variables. This significance is driven by the exceptionally high correlations of these variables with all three independent variables in this morphologically conservative lineage. Scaling of FL tends towards negative allometry, but there is great scatter in the data and isometry cannot be ruled out statistically. However, FL is slightly less well-correlated with body size, resulting in higher absolute residuals, suggesting the potential for a dietary signal. Indeed a strong signal suggests a relationship between FL and relative prey size. Thus, in the family Felidae, where food material properties vary little, food geometric properties appear to select for muscle architectural properties. In at least one instance (the jaguar), estimated BF appears to signal food material properties—namely this species is capable of consuming more obdurate foods. Though the sample encompasses nearly a quarter of the felid species and nearly the entire body size range of the family, it lacks some dietary diversity (e.g., the inclusion of the hypercarnivorous cheetah would have allowed us to better explore the food mechanical properties correlates)—a reality incumbent in dietary studies of this relatively dietarily homogenous lineage.

Future Directions

We have begun expanding this research in two directions: (1) exploration of the masticatory architecture in other families of carnivorans and (2) soft-tissue and BF reconstruction in fossil felids.

We found dietary signals in the masticatory muscle architecture of the felids in this study—particularly as the architecture relates to relative prey size. However, because felids are so dietarily homogenous (i.e., the mechanical properties of their diets, with few exceptions, are nearly indistinguishable), adding other carnivorans to the sample will allow elaboration of our understanding of dietary adaptation of the masticatory muscle architecture. Specifically, we are exploring these properties in broad intrafamilial samples of canids, mustelids and ursids. These groups contain species that span broader dietary ranges including hypercarnivores and durophages. To this, we are also adding a broad range of species representing the other carnivoran families including durophagous hyaenids and ailurids and frugivorous and insectivorous members of the Procyonidae and Viverridae. Thus we are testing whether the dietary soft tissue correlates that we have observed in felids (this article) and primates (Perry et al., 2011) can be generalized to all Carnivora. We have further plans to explore these anatomies in phalangerid and dasyurid marsupials to test our hypotheses that these architecture/diet relationships are similar across Mammalia.

Additionally, with our improved understanding of the scaling of soft tissue anatomy in modern felids, we are closer to being able to more accurately reconstruct the BFs of fossil species. Several studies have done this, but until now data on the fiber architecture have been missing—resulting in discussions of BFs in relative terms rather than in performance estimates (Christiansen, 2011). For many years, students of sabertooths have estimated weak BFs for these enigmatic fossil taxa (Matthew, 1910; Kurtén, 1954), though recent studies (e.g., Christiansen, 2011) have questioned these assumptions. Likewise, the gape capabilities of sabertooths have also received much attention (Matthew, 1910; Merriam and Stock, 1932; Kurtén, 1954; Emerson and Radinsky, 1980; Turner and Antón, 1997; Christiansen, 2006; Hartstone-Rose et al., 2007).

We plan to gather information on the origin and insertion areas for the masticatory muscles. This, coupled with a dataset on muscle architecture in the same individuals will allow us to gauge how well the osteological measurements predict the soft-tissue ones. We plan to use correlations between the two datasets to predict soft-tissue dimensions in fossil carnivorans based on similar osteological predictors. Although several assumption underlie this method (e.g., that all fibers originate from the bony origin area and insert on the bony insertion area). Nevertheless, good correlations between muscle architecture and bony measurements would represent a novel and useful contribution to paleobiology.

With a further understanding of the relationship between the soft-tissue anatomy (described in the current study) and its osteological correlates (Hartstone-Rose and Perry, 2007b), we believe that we can advance the discourse on the masticatory abilities of some fossil felids, for example, the sabertooths; the osteological correlates of PCSA and FL will ultimately allow us to reconstruct estimates of BF and gape in these enigmatic species.


The authors thank Aaron Honeycutt of Foxfire Taxidermy, Carolina Tiger Rescue and its staff (especially Kathryn Bertok and Pam Fulk), the Smithsonian National Zoological Park and its staff, Laura J. Mitchell for help with data collection and two anonymous reviewers for valuable and substantive feedback.