The prehensile tail, capable of suspending the entire body weight of an animal (Emmons and Gentry, 1983), is thought to have evolved twice (in parallel) in New World monkeys (Platyrrhini): once in the monophyletic Atelinae (Alouatta, Ateles, Brachyteles, Lagothrix), and once in Cebus1 (Napier, 1976; Rosenberger, 1983) (Fig. 1). However, placed in a larger mammalian context, the parallel evolution of prehensile tails in primates is not unique. In fact, prehensile tails may have evolved independently as many as 14 times among 50 extant mammalian genera (Emmons and Gentry, 1983; Emmons, 1990; Nowak, 1991; Bergeson, 1996). Clearly, these expensive anatomical investments, which represent a larger proportion of total body mass than nonprehensile tails (Grand, 1977, 1983), are effective adaptations for life in the forest canopy. However, despite their abundance among arboreal mammalian taxa, and despite their importance for balance, feeding behavior, and, certainly, locomotion, little is known about the mechanical structure of the prehensile tail and how it differs from the nonprehensile tail. The goal of this study is to assess the differences in tail structure among prehensile- and nonprehensile-tailed platyrrhines, by examining the external and internal geometries of caudal vertebrae in this group of primates.
Differences in positional behavior and tail-use among prehensile- and nonprehensile-tailed platyrrhines have been correlated with different tail morphologies. For example, prehensile-tailed monkeys are thought to have relatively expanded sacroiliac joints, relatively long proximal tail regions, and distal caudal vertebrae that are relatively short and wide when compared with nonprehensile-tailed taxa (Ankel, 1962, 1965, 1972; Grand, 1977; German, 1982; Rosenberger, 1983). However, while these previous studies have successfully identified general morphology patterns associated with different tail-use behaviors, all have focused on external geometric bone shape and/or structure and with limited quantitative data. Thus, the purpose of this study is twofold: (1) to quantify and statistically compare external bone dimensions of caudal vertebrae in prehensile- and nonprehensile-tailed platyrrhines and (2) to examine the internal geometry of caudal vertebrae in prehensile- and nonprehensile-tailed platyrrhines. In this manner, more nuanced insights into the mechanical differences of tail-use in these different tail types may be gained.
The primate tail (both prehensile and nonprehensile) can be divided into two regions based on distinctive external morphologies of the caudal vertebrae: proximal and distal (Schmidt, 1886; Ankel, 1962, 1965, 1972) (Figs. 2, 3a–c). Proximal caudal vertebrae bear ventral and neural arches and a pair of transverse processes (Fig. 3a), and they articulate with one another by way of zygapophyseal joints similar to those in the lumbar spine. The presence of zygapophyseal articulations in the proximal tail suggests a certain degree of sagittal plane flexibility in this region but limited movement in other planes (Shapiro, 1993, 1995; Ward, 1993). Distal caudal vertebrae are distinct from proximal vertebrae in that the former are longer and more rounded in cross-section, have two pairs of (relatively reduced) transverse processes, and articulate with one another solely through intervertebral discs (Fig. 3c). The distinct morphologies of distal caudal vertebrae, then, allow absolutely greater ranges of intervertebral joint motion and in all directions. Situated at the junction between proximal and distal tail regions is a transition vertebra (defined conventionally as the last element of the proximal tail), which is characterized by zygapophyseal articulations cranially and an intervertebral disc articulation caudally (Figs. 2 and 3b).
Within the proximal region of the tail and proceeding distally to the longest caudal vertebra (always in the distal region), vertebral body length steadily increases, while neural arch development, transverse process development, and ventral arch development all decrease (German, 1982; Organ, 2007). This characteristic increase in relative vertebral length has led some recent researchers to define a third region of the tail: the transitional region. This region extends from the first element distal to the transition vertebra up to and including the longest caudal vertebra (German, 1982; Lemelin, 1995; Schmitt et al., 2005; Organ, 2007; Organ et al., 2009) (Fig. 2).
Although both prehensile and nonprehensile primate tails always have these regional morphologic differences, there are some anatomical disparities between the two. For example, the proximal region is said to be relatively longer in prehensile-tailed taxa (Dor, 1937; Ankel, 1962, 1965, 1972). However, the robusticity (i.e., width-to-length ratio at the proximal end of the vertebral body) of proximal caudal vertebrae does not differ between tail types (German, 1982). This is not surprising, given that the movements of the proximal tail region (i.e., sagittal flexion and extension, limited by zygapophyseal articulations) are fairly consistent motions in both prehensile- and nonprehensile-tailed taxa. The vertebral robusticity of the distal tail, however, differs notably among prehensile- and nonprehensile-tailed animals. In this case, distal caudal vertebrae are consistently more robust (i.e., wider for a given length) in prehensile-tailed monkeys, with this robusticity steadily increasing proximodistally within the vertebral sequence. In contrast, the degree of caudal vertebral robusticity in nonprehensile-tailed animals tends to be constant along the entire distal region (German, 1982). In both groups, length of the distal vertebrae decreases along the sequence. However, although width of the vertebrae decreases proportionately in the nonprehensile-tailed group, width is more consistent among vertebrae in the prehensile-tailed group. Thus, greater relative robusticity of the distal caudal vertebrae of prehensile-tailed monkeys is a product of relatively shorter vertebrae, and indicates these vertebrae may be structured to resist higher bending (and torsional) moments associated with tail prehension and suspension (German, 1982). Yet, despite these differences in gross external vertebral morphology, how such differences affect the mechanics of tail-use is unclear. Although research on muscle architecture indicates that prehensile tail lateral flexor musculature is capable of producing higher contraction forces in prehensile tails than in nonprehensile tails (Organ et al., 2009), whether external geometric considerations of vertebral shape reflect such differences in muscle force is unknown. Moreover, it is likely that these assessments of “robusticity” reflect only relative muscle attachment size (i.e., transverse process breadth) and may not reflect higher resistance to bending and/or torsion.
Although external measurements can often provide mechanically relevant information about the principal axes of stress and strain within an element, more reliable and accurate information about mechanical structure and loading history can be obtained by examining the true cross-sectional properties of the element (i.e., second moment of area, section modulus) (Ruff, 1987, 2002; Selker and Carter, 1989; Daegling, 1989, 2002; Daegling and Grine, 1991; Biknevicius, 1993; Jungers et al., 1998). Perhaps most importantly, external breadth measurements do not always reflect similar patterns of variation in mechanical loading when compared to cross-sectional properties (Daegling, 1989, 2002; Daegling and Grine, 1991; Jungers et al., 1998; Ruff, 2002), further underscoring the importance of evaluating the internal architecture of tail vertebrae. By examining external and internal geometry together, this study attempts to further evaluate the mechanical structure of New World monkey tails.
MATERIALS AND METHODS
The comparative osteologic sample used in this study consists of 61 individuals representing eight platyrrhine genera (Table 1). These taxa were specifically chosen for study because they represent the full range of tail-use behaviors observed in New World monkeys, including both instances of the evolution of prehensile tails in this primate group—the Atelines and Cebus—which are known to exhibit different patterns of prehensile tail use (Table 2). Additionally, the sample includes at least one genus from each of the five platyrrhine subfamilies (Schneider and Rosenberger, 1996), thereby distributing the taxa roughly equally across the higher level phylogeny.
All individuals in a genus comprise a single species except Pithecia: P. pithecia (N = 5), P. monachus (N = 3).
All body masses taken from Ford and Davis (1992) except those for Alouatta palliata, which were taken from recorded unpublished data courtesy of Dr. Kenneth Glander, Duke University.
Calculated as the sum total of craniocaudal length (CCL) for all caudal vertebrae (in mm).
Average positions of the TV, LV, and MDV (see text for descriptions) for a genus can be deduced: the TV and LV are the last elements of the proximal and transitional regions, respectively; the MDV is located at the midpoint of the distal region. In Ateles for example, the TV is the eighth vertebra in the caudal sequence (on average). The LV is the twelfth vertebra (8 proximal + 4 transitional vertebrae), and the MDV is the twenty-second vertebra [proximal + transitional + ½ distal vertebrae; i.e., 8 + 4 + (20/2)].
All specimens studied here are adults, as determined by complete epiphyseal fusion of long bones, innominates, and scapulae, with no visible trace of the physeal plate. Whenever possible, skeletal ages were additionally corroborated by complete adult dental eruption sequences. A number of specimens exhibited incomplete fusion of the vertebral bodies but were still deemed fully adult based on the above criteria because age at which growth plates fuse within caudal vertebrae is unknown. In such cases, individuals with visible physeal lines but otherwise complete adhesion of the subarticular bone of the vertebral centrum were included.
The comparative sample includes only wild-caught animals as animals kept in captivity may alter loading regimes on their tails, which may affect the morphologies associated with normal behavior. For similar reasons, this sample was also limited to use of only nonpathologic specimens free of obvious signs of osteoarthritis or bone fracture.
Each taxon is represented by a minimum of six individuals and a maximum of nine. These small sample sizes are the direct result of a paucity of caudal vertebral material in most museum collections. Therefore, it was not possible to equally distribute the sample across the sexes. Similarly, no effort was made to evaluate potential differences between sexes in any of the taxa for any of the variables examined here, as the number of individuals per sex was too small for even nonparametric statistical testing. The pooling of sexes in this analysis is justified, however, because there are no known sex differences in tail-use behaviors among the taxa examined (Bergeson, 1996; Wheeler and Ungar, 2001). Also because of limited specimen availability, the Pithecia sample included two species, P. pithecia (N = 5) and P. monachus (N = 3); whereas, all other taxon samples included only one species. Ultimately, the osteologic sample included all usable specimens in a given taxon (see Appendix for more detail) at the Smithsonian National Museum of Natural History (Washington, DC); the American Museum of Natural History (New York, NY); and the under-publicized collection of mantled howling monkeys (Alouatta palliata) from La Pacifica, Guanacaste, Costa Rica, under the curatorial care of Dr. Kenneth Glander, Department of Evolutionary Anthropology, Duke University (Durham, NC).
Measurements of Mechanical Structure
General tail morphology.
The morphology of each individual tail examined in this study was assessed by recording the total number of vertebrae, number of vertebrae in each tail region, and craniocaudal length (CCL) (see later) of each vertebra. These data were used to describe relative number of vertebrae and relative length of each tail region, and were compared across taxa to quantitatively analyze the general trends in different tail regions between prehensile- and nonprehensile-tailed platyrrhines described qualitatively by Dor (1937) and Ankel (1962, 1965, 1972).
In addition to these general assessments of morphology, additional external linear measurements (see later) and internal (cross-sectional) geometric measurements were recorded from three vertebrae of each tail: the transition vertebra (TV), the longest caudal vertebra (LV), and the vertebra that falls at the midpoint (in terms of vertebral number) of the distal region of the tail [mid-distal vertebra (MDV)] (Fig. 2). Measurements were taken from both articulated and disarticulated skeletons (see Appendix), although the vast majority were from disarticulated skeletons, as fully articulated caudal sequences are rare in museum collections. These particular vertebrae were chosen because they are easy to identify based on external morphology and they have provided diagnostic utility in previous studies (German, 1982; Youlatos, 2003; Organ, 2007; Organ et al., 2009).
The TV can be identified by the presence of zygapophyseal articulations with the vertebra immediately craniad, and an intervertebral disc joint articulation with the vertebra immediately caudad; the LV is simply the vertebra with the longest CCL; and the MDV is identified numerically within the distal tail sequence (as described earlier). The physical locations of these vertebrae within the caudal sequence make them ideal for examining structural properties in different regions of the tail because they allow comparisons between all three regions of the tail, and represent vertebral segments where changes in mechanical loading should be high. Specifically, the TV represents the vertebral segment where articulations between vertebrae change from zygapophyseal to intervertebral disc joints; the LV is the longest vertebra in the sequence, thus bending (and torsion) should be highest in this vertebra among all caudal vertebrae; the MDV falls roughly in the area where tail-wrapping around a substrate occurs in prehensile-tailed animals (Organ, personal observation).
External measurements of caudal vertebrae.
Linear measurements of all caudal vertebrae in each tail were collected with Mitutuyo 15-cm sliding digital calipers and were entered into a Microsoft Excel 2003 customized spreadsheet with a USB input tool connected directly to a Dell Latitude D610 laptop PC. Each measurement was recorded three times and the trials averaged to minimize the potential for significant measurement error. The linear metrics collected in this study were chosen because they have known or suspected biomechanical/functional significance (i.e., biomechanical bone lengths, muscle moment arms, etc.), or because they have been used in previous studies (German, 1982; Youlatos, 2003; Schmitt et al., 2005) (Fig. 4):
aCraniocaudal length (CCL): maximum craniocaudal length of the vertebra, measured in lateral view
bProximal transverse process breadth (PTPB): maximum breadth of the proximal transverse process, measured in superior view
cHemal process height (HPH): maximum height of the hemal process, measured from proximal transverse process to hemal process, in proximal view
Measurement errors of linear metrics were evaluated for all vertebrae at the genus level and across the total sample. Percent measurement errors for each measure were computed by comparing the average deviation of each of the three trials (per measurement per vertebra) against the mean of the three trials, multiplied by 100 (Auerbach and Ruff, 2006; similar to White, 2000). Errors for the total sample ranged from just under 1% to ∼2.5%. At the genus level, measurement error ranged from less than 0.5% to about 4.5%, although the vast majority was less than ∼2.5%. The highest measurement errors, as expected, occurred with the smallest taxa (e.g., Saguinus and Aotus), as small deviations in measurements were more significant by comparison because of the diminutive size of the vertebrae in these taxa. Nevertheless, because between-taxon variation was greater than within-taxon variation for these measurements (Organ, 2007), measurement error should not affect interpretation of results.
The method of choice by previous researchers investigating the external geometry of caudal vertebrae has been to calculate indices that take into account the relative expansion of muscle attachment sites, primarily the transverse process (German, 1982; Youlatos, 2003; Schmitt et al., 2005). The transverse processes of caudal vertebrae are the major attachment sites for mm. intertransversarii caudae, the prime lateral flexors and rotators of the tail (Dor, 1937; Lemelin, 1995; Organ, 2007; Organ et al., 2009). In addition, the transverse processes of the proximal-region vertebrae serve as the fleshy origins of mm. flexor caudae longus et brevis, the primary flexors of the tail. Thus, given the relative importance of transverse processes for lateral and dorsoventral flexor muscle attachments, it is not surprising that previous research has focused on describing potential differences in transverse process expansion between prehensile and nonprehensile tails.
Relative transverse process expansion (TPE) of the TV, LV, and MDV was calculated (see later) and compared against a novel index, relative hemal process expansion (HPE). The HPE index is a measurement of the relative expansion of the distal (tendinous) attachment site for mm. flexor caudae longus et brevis. The HPE index does not consider muscle belly size (as the TPE does). Instead, the HPE index reflects only the dorsoventral bending moment arm of these muscles, and represents the mechanical advantage that these muscles possess during tail flexion (assuming equivalent muscle architecture across taxa).
Relative process expansion indices of the TV, LV, and MDV were calculated as follows: a. TPE = PTPB/CCL × 100 b. HPE = HPH/CCL × 100.
Internal measurements of caudal vertebrae.
To evaluate the internal geometry of caudal vertebrae, this study utilizes general engineering beam theory (Huiskes, 1982) to model the strength and rigidity of long bones. Caudal vertebrae are not long bones in the traditional sense; however, they are always at least twice as long (craniocaudally) as they are wide (mediolaterally) and, therefore, can still be modeled as beams (i.e., short beams).
Cross-sectional images of the TV, LV, and MDV were obtained at 50% of CCL with a peripheral quantitative computed tomography (pQCT) scanner (Stratec XCT SA+, Norland Medical Systems) and customized internal pQCT software (Norland/Stratec XCT 5.40). The use of pQCT was preferred over standard computed tomography (CT) for two reasons. First, pQCT technology was specifically designed to evaluate the cross-sectional properties of small bones (e.g., human wrist bones or small rodent limb bones) (Rüegsegger et al., 1976), thus providing adequate resolution for analyzing small caudal vertebrae. Second, pQCT data have been shown to accurately predict bone mineral density (Augat et al., 1998), long bone bending rigidity (Martin et al., 2004), and long bone bending strength (Ferretti et al., 1996; Moisio et al., 2003), which are all important components for analyzing bone mechanical structure.
Prior to scanning, the scanner was calibrated daily against a phantom cone with known densitometric properties. Specific scanning protocols for caudal vertebrae utilized pixel sizes of 0.1 mm2, slice thicknesses of 1.00 mm, and scanning speeds of 3.00 mm/sec for three blocks, which yielded 540 projections. For the purposes of this study, cortical bone was of more interest than cancellous bone, because (1) cancellous bone accounts for less than 15% of total bone in these vertebrae (see Burr and Piotrowski, 1982), (2) it does not appear to be directed along the principal axes of stress and strain, and (3) its contribution to total mechanical strength of long bones (at the midshaft or midcentrum) can generally be considered negligible in such circumstances (Ruff, 1983). To differentiate between bone types in the cross-sectional images obtained from the pQCT, cortical bone was isolated from cancellous bone using a density threshold of 500 mg/cm3 combined with internal software algorithms2 designed to segment bone types.
Caudal vertebrae were positioned in the scanner with the ventral surface facing the scanning platform using plasticine putty placed under the vertebral body (Fig. 4). The coronal (x) axis was defined by the mediolateral axis running through the right and left proximal transverse processes3 and was aligned parallel to the scanning platform. The sagittal (y) axis ran perpendicular to the coronal axis at the mediolateral midpoint of the proximal and distal articular surfaces. Finally, the longitudinal (z) axis was defined as the intersection of the coronal and sagittal axes and was aligned with the long axis of the scanning platform. The above standardization protocol was adapted from protocols published previously by Ruff (2002).
Bending and torsional stresses are known to predominate in most long bone diaphyses during active locomotion (e.g., Wainwright et al., 1976; Ruff, 2002; Biewener, 2003), and modeling caudal vertebrae as short beams inherently assumes similar predominant loading regimes. The present analysis of internal geometric structure explicitly focuses on bending and torsion of the caudal vertebrae (to the exclusion of axial tension and compression, which may better characterize the traction forces experienced by the tail during full suspension) to compare with the external linear dimensional analyses presented here and elsewhere (German, 1982). Therefore, the cross-sectional geometric parameters of primary interest are the ones that reflect bending and torsion: second moments of area and section moduli. Second moments of area and section moduli are calculated about either the neutral axis of the caudal vertebrae (for bending) or the centroid of the cross-section (for torsion). The (torsional) polar moment of area (J) is calculated as the sum of two orthogonal (bending) second moments of area (I). The section modulus is equivalent to the second moment of area (I for bending, J for torsion) divided by the perpendicular (or radial) distance from the axis (or centroid) to the outermost fiber of the section (e.g., Timoshenko and Gere, 1972; Ruff, 2000, 2002; Organ et al., 2006). Thus, because the polar section modulus (Zp)—a direct measure of torsional strength—is calculated from the polar moment of area (J), Zp can also be considered a proximate measure of (twice) average bending strength (Ruff, 2002; Organ, 2007; Organ and Lemelin, 2009), and the most appropriate singular measure of mechanical structure in platyrrhine caudal vertebrae. [For a more thorough discussion of mechanical design in bones, see Wainwright et al. (1976)]. After pQCT scanning was complete, the torsional section modulus (Zp) for each caudal vertebra was calculated using an internal loop algorithm in the pQCT scanning software, equivalent to algorithms traditionally used for this purpose (e.g., Nagurka and Hayes, 1980; Ruff, 2002).
Single-factor analyses of variance (ANOVA) with post-hoc Tukey's HSD tests were used to evaluate differences among taxa in external and internal geometric properties. All linear and cross-sectional variables were natural-log transformed prior to statistical testing; percentage data were subjected to the arcsine transformation. Also prior to analysis, all variables were tested for normality of distribution and homogeneity of variance using the Komogorov-Smirnov (with Lilliefors correcting factors) and Levene's tests, respectively. All variables were normally distributed; however, not all variables exhibited homogeneity of variance across all taxa. In these cases (indicated where appropriate), the Games-Howell post-hoc test was used in lieu of the Tukey's HSD test (Zar, 2010). Alpha levels for all statistical tests were set at α = 0.05.
The analyses of general tail morphology and external geometry are size-independent, as they represent proportions and ratios that are calculated in a consistent manner across taxa. However, analyses of internal geometry (cross-sectional properties) are inherently size-dependent and require some form of size-standardization to compare across taxa with differing body masses. Traditionally, cross-sectional properties have been standardized by body mass × bone length, a proxy for “body size.” Unfortunately, body masses were not available for the majority of museum specimens used for this study, preventing use of the traditional size-standardization. Instead, all cross-sectional properties were standardized by CCL of the vertebra (i.e., bone length without any consideration of body mass). This approach is appropriate because species-mean caudal vertebral length data in platyrrhines scale isometrically with body mass data collected from the literature (Organ, 2007; also see German, 1982). It is important to emphasize that the prehensile-tailed platyrrhines are also the largest-bodied platyrrhines (Table 1).
Notched box plots (McGill et al., 1978), a variation on more traditional box plots, are used here to graphically represent genus data distributions for external and internal geometric properties. In notched box plots, the middle horizontal “waist” is the median, the outer horizontal limits of the boxes are the hinges encompassing the interquartile range (25%–75%), and the “whiskers” represent all other data points except outliers more than 1.5 times the interquartile range. The “notches” in the sides of the boxes (i.e., extending out from the median) represent ∼95% confidence intervals around the median, such that if the notches about two medians do not overlap, the medians for those two taxa are considered significantly different. In some cases, the vertical extent of the notch exceeds the interquartile range (a result of an arbitrary constant used in calculating these notches, McGill et al., 1978; Ruff, 2002) and, thus, should only be used as a rough guide to differences among groups. Instead, as noted earlier, differences in size-adjusted internal geometric properties were evaluated with ANOVA and post-hoc tests.
General Tail Morphology
Average tail lengths and number of caudal vertebrae for each taxon are presented in Table 1. The total number of caudal vertebrae range from 22 in Pithecia to 33 in Saguinus, with all other taxa falling between. In terms of absolute tail length, the longest tails belong to prehensile-tailed Ateles (740 mm) and the shortest to nonprehensile-tailed Saimiri (348 mm). There appears to be a relationship between overall tail length (in mm) and body size (body mass), as the largest platyrrhines tend to have the longest tails. However, even though the largest-bodied taxa are those with prehensile tails, there does not appear to be any relationship between total number of caudal vertebrae and prehensility of the tail, as tails with the least (Pithecia) and greatest (Saguinus) number of vertebrae both belong to nonprehensile-tailed monkeys.
Results of the ANOVA testing potential differences among tail regions of the various taxa indicate significant mean differences among taxa with regard to the relative number of proximal [F(7,53) = 18.00, P < 0.001] and distal vertebrae [F(7,53) = 14.31, P < 0.001], as well as differences in relative lengths of the proximal [F(7,53) = 14.45, P < 0.001] and distal regions [F(7,53) = 7.92, P < 0.001] of the tail (Table 3 and Fig. 5). No differences were found among taxa in the transitional region of the tail either in terms of relative number of vertebrae [F(7,53) = 1.98, P = 0.075] in the region or relative length [F(7,53) = 1.58, P = 0.162]. Post-hoc Tukey's tests suggest that taxa with the longest proximal regions tend to be the prehensile-tailed taxa (i.e., Ateles, Alouatta, and Cebus), and these taxa also tend to have the greatest number of caudal vertebrae in the proximal region (Table 3 and Fig. 5). However, proximal tail region length in Cebus and Lagothrix is not statistically different from any of the nonprehensile-tailed platyrrhines other than Pithecia,4 which consistently has the shortest proximal region of any platyrrhine studied here. In addition, nonprehensile-tailed Saimiri has an equivalent percentage of proximal-region vertebrae compared to prehensile-tailed Cebus and Lagothrix (Table 3 and Fig. 5).
Table 3. Summary statistics for tail regions in platyrrhines
Prehensile-tailed; Mean, mean percentage for genus; SE, standard error of the mean.
-eResults of Tukey's HSD post-hoc multiple comparisons for the percentage of caudal vertebrae that are located in the proximal region: possession of same letter = nonsignificant difference (i.e., P > 0.05), with “b” representing the highest group of indistinguishable values, “c” next highest, etc.
-iResults of Tukey's HSD post-hoc multiple comparisons for percentage of total tail length made up by the proximal tail region: possession of same letter = nonsignificant difference (i.e., P > 0.05), with “f” representing the highest group of indistinguishable values, “g” next highest, etc.
-lResults of Tukey's HSD post-hoc multiple comparisons for the percentage of caudal vertebrae that are located in the distal region: possession of same letter = nonsignificant difference (i.e., P > 0.05), with “j” representing the highest group of indistinguishable values, “k” next highest, etc.
-pResults of Tukey's HSD post-hoc multiple comparisons for percentage of total tail length made up by the distal tail region: possession of same letter = nonsignificant difference (i.e., P > 0.05), with “m” representing the highest group of indistinguishable values, “n” next highest, etc.
In contrast to the pattern seen for the proximal region of the tail, the platyrrhines with the longest distal regions are those with the shortest proximal regions (i.e., the nonprehensile-tailed taxa such as Saguinus, Aotus, and Pithecia) (Table 3 and Fig. 5), and this pattern generally holds for percentage of distal region vertebrae as well (Table 3 and Fig. 5). As in the proximal region, the relative number of vertebrae and relative length of the distal tail region in Lagothrix and Saimiri falls right at the intersection between the prehensile-tailed taxa and the nonprehensile-tailed taxa (Table 3), where they cannot be statistically differentiated from either group.
Relative Muscle Attachment Development
Results of the ANOVA on TPE index are presented in Table 4 and Figure 6, and indicate that the taxa differ in the relative expansions of their transverse processes [TV: F(7,52) = 6.21, P < 0.001; LV: F(7,52) = 20.53, P < 0.001; MDV: F(7,51) = 186.70, P < 0.001]. There is clearly a gradient from most-expanded to least-expanded transverse processes that roughly corresponds to the distribution of tail prehensibility in the platyrrhine sample (Fig. 6). However, TPE indices do not systematically distinguish between types, except in the distal tail region (MDV), where transverse processes are relatively more expanded in prehensile-tailed taxa. For the LV, it is clear that the prehensile-tailed atelines have more expanded transverse processes than the nonprehensile-tailed platyrrhines. However, the one taxon that is statistically indistinguishable from both the atelines and the nonprehensile-tailed taxa is Cebus.
Table 4. Summary statistics for muscle attachment indices in platyrrhines
Prehensile-tailed; TPE, transverse process expansion; HPE, hemal process expansion; Mean, mean percentage for genus; SE, standard error of the mean.
-dResults of Games-Howell post-hoc multiple comparisons for TV TPE indices: possession of same letter = nonsignificant difference (i.e., P > 0.05), with “b” representing the highest group of indistinguishable values, “c” next highest, etc.
,fResults of Games-Howell post-hoc multiple comparisons for TV HPE indices: possession of same letter = nonsignificant difference (i.e., P > 0.05), with “e” representing the highest group of indistinguishable values, “f” next highest.
,hResults of Games-Howell post-hoc multiple comparisons for LV TPE indices: possession of same letter = nonsignificant difference (i.e., P > 0.05), with “g” representing the highest group of indistinguishable values, “h” next highest.
,jResults of Games-Howell post-hoc multiple comparisons for LV HPE indices: possession of same letter = nonsignificant difference (i.e., P > 0.05), with “i” representing the highest group of indistinguishable values, “j” next highest.
-oResults of Games-Howell post-hoc multiple comparisons for MDV TPE indices: possession of same letter = nonsignificant difference (i.e., P > 0.05), with “k” representing the highest group of indistinguishable values, “l” next highest, etc.
-rResults of Games-Howell post-hoc multiple comparisons for MDV HPE indices: possession of same letter = nonsignificant difference (i.e., P > 0.05), with “p” representing the highest group of indistinguishable values, “q” next highest, etc.
HPE index results, found in Table 4 and Figure 7, also differ among the taxa [TV: F(7,52) = 13.02, P < 0.001; LV: F(7,52) = 43.06, P < 0.001; MDV: F(7,51) = 63.95, P < 0.001]. However, unlike the TPE indices above, the HPE indices perform better in distinguishing tail types, especially in the transitional (LV) and distal (MDV) regions of the tail (Table 4 and Fig. 7). In the proximal region, prehensile-tailed taxa have the highest HPE indices, which are significantly higher than all nonprehensile-tailed taxa except Pithecia (Table 4 and Fig. 7). The range of values for Pithecia extends well into the ranges of prehensile-tailed taxa, and therefore, the genus possesses a HPE index that is statistically indistinguishable from prehensile and nonprehensile tails (Fig. 7).
Caudal Vertebral Structural Properties
Results for the size-adjusted polar section modulus are presented in Table 5 and Figure 8. Size-adjusted polar section moduli differ among the taxa for each of the three vertebral locations [TV: F(7,52) = 57.71, P < 0.001; LV: F(7,52) = 77.77, P < 0.001; MDV: F(7,50) = 36.86, P < 0.001]. All prehensile-tailed taxa (both atelines and Cebus) have caudal vertebrae that are relatively stronger in torsion (and bending) than those of nonprehensile-tailed taxa regardless of which vertebra is examined: TV, LV, or MDV (Table 5 and Fig. 8). The one exception to this generalization is the size-adjusted torsional strength of the TV of Pithecia, which is statistically indistinguishable from prehensile-tailed Cebus and all nonprehensile-tailed taxa.
Table 5. Summary statistics for size-adjusted torsional strength (polar section modulus) of platyrrhine caudal vertebrae
Ln (TV Zp/CCL)
Ln (LV Zp/CCL)
Ln (MDV Zp/CCL)
Prehensile-tailed; TV, transition vertebra; LV, longest caudal vertebra; MDV, mid-distal vertebra; Zp, polar section modulus; CCL, craniocaudal length; Mean, mean log ratio for genus; SE, standard error of the mean.
-eLetters indicate the results of Games-Howell post-hoc tests for TV: possession of same letter = nonsignificant difference between taxa (P > 0.05). b, highest mean: c, next highest; etc.
-iLetters indicate the results of Games-Howell post-hoc tests for LV: possession of same letter = nonsignificant difference between taxa (P > 0.05). f, highest mean; g, next highest; etc.
-lLetters indicate the results of Games-Howell post-hoc tests for MDV: possession of same letter = nonsignificant difference between taxa (P > 0.05). g, highest mean; k, next highest; etc.
The objectives of this study are to quantify and statistically compare the external and internal geometries of prehensile- and nonprehensile-tailed platyrrhine caudal vertebrae. The data presented here strongly support the predicted relationships between generalized tail-use behaviors (i.e., prehensile vs. nonprehensile) and general tail structure and caudal vertebral structure in platyrrhine primates.
General Tail Morphology
Based on previous work by Dor (1937) and Ankel (1962, 1965, 1972), prehensile-tailed taxa were expected to possess longer proximal tail regions than their nonprehensile-tailed counterparts, both in terms of relative tail region length and relative number of vertebrae. The theoretical proposal behind this hypothesis is purely functional: longer proximal tail regions are essential for increasing proximal vertebral column flexibility (especially in vertebral sequences with zygapophyseal articulations) because the range of motion at zygapophyseal joints is severely restricted when compared with other vertebral joints (e.g., Ward, 1991, 1993; Shapiro, 1993, 1995). Thus, adding more zygapophyseal joints in series allows for greater absolute flexibility of the proximal tail, even though flexibility at each joint remains limited. In addition to this increased absolute flexibility, however, longer proximal regions also provide some level of increased columnar stiffness, because movements between zygapophyseal joints other than dorsoventral flexion/extension (i.e., lateral flexion, rotation) are still restricted compared to similar movements between the intervertebral disc joints in the transitional and distal regions of the tail. Moreover, while longer proximal regions, in effect, tend to increase the number of zygapophyses (and thus, ultimate range of motion), they also extend the relative stiffness of the vertebral column more distally within the tail by adding more zygapophyses in series. A proximal tail morphology such as this would certainly benefit prehensile-tailed animals, providing both adequate flexibility and stiffness.
In general, the data presented here support this hypothesis, as prehensile-tailed platyrrhines tend to have longer proximal regions than nonprehensile-tailed platyrrhines. There are, however, exceptions to this rule (e.g., Lagothrix and Saimiri). These results support qualitative assessments made by Dor (1937) and Ankel (1962, 1965, 1972), where the prehensile-tailed atelines (i.e., Ateles, Lagothrix, Brachyteles, Alouatta) tended to have the most proximal-region vertebrae, the nonprehensile-tailed platyrrhines tended to have the fewest proximal-region vertebrae, and prehensile-tailed Cebus was intermediate in number of proximal-region vertebrae. Although German (1982, p 455) suggested that “… in the New World monkeys, there are equal proportions of proximal vertebrae irrespective of use of tail in locomotion,” the results presented here do not support this conclusion.
The comparatively low percentage of proximal-region vertebrae demonstrated here for prehensile-tailed Lagothrix is consistent with previous work by Schmitt et al. (2005), who reported similar results. These authors argued that the shorter proximal tail region of Lagothrix would be capable of far less extension than in the other atelines, especially when combined with a comparatively obtuse angle of the sacro-caudal articulation (Schmitt et al., 2005). Yet, despite clear differences in the kinematics of tail-use between Lagothrix and Ateles (Schmitt et al., 2005), there is no obvious reason why Lagothrix would require a shorter (and therefore less flexible) proximal tail region, even if they use their tails (1) to help mediate lateral sway with every hand-hold during tail-assisted brachiation and (2) to firmly grasp supports as the so-called fail-safe mechanism proposed by Stern et al. (1980) and Jungers and Stern (1981).
Nonprehensile-tailed Saimiri, with its comparatively long proximal tail region, is particularly interesting, given the anecdotal reports of prehensile tail-use in juveniles (e.g., Rosenblum, 1968; Thorington, 1968), where the ability to use the tail in a prehensile manner is lost during postnatal ontogeny. If long proximal tail regions are essential for increasing both flexibility and stiffness in prehensile tails, the presence of such morphology in Saimiri may be related to tail-use behavior in the early postnatal period of life. Structural analysis of an ontogenetic series of Saimiri may provide the clues to the functional significance of adult morphology of this taxon.
It is also interesting to note some general patterns in the distal tail regions of the taxa studied here. Prehensile-tailed individuals tend to have significantly shorter distal tail regions (in terms of relative tail length) than nonprehensile-tailed individuals. However, despite the fact that the distal tail region is shorter in prehensile-tailed animals than in nonprehensile-tailed ones, the median number of distal region vertebrae does not appear to differ drastically among taxa (Table 1). Although not directly examined in this study, this general pattern would seem to indicate that prehensile-tailed animals have relatively shorter distal region vertebrae than their nonprehensile-tailed counterparts, as they have (on average) the same number of distal-region vertebrae comprising a shorter total length. If this were confirmed empirically, it would suggest that the distal tail regions of nonprehensile-tailed taxa are more flexible than those of prehensile-tailed taxa because, in the case of the latter, displacement of one end of the caudal vertebral column relative to the other decreases when individual segments are decreased in length (Ward, 1993). However, a corollary of such a reduction in length of the individual segments is a concomitant reduction in bending moments about the intervertebral discs between each of the segments (Ward, 1993), which may be particularly important in prevention of irreparable damage to intervertebral cartilages during prehensile tail-use behaviors.
Relative Muscle Attachment Development
Previous work by German (1982), Jones (2004), and Schmitt et al. (2005) suggested differences in muscle attachment site morphology between prehensile and nonprehensile caudal vertebrae, however with some differences in the details among studies. Results of the present study were expected to confirm that muscle attachment site expansion is useful for distinguishing among prehensile and nonprehensile tail vertebrae, especially distally within the sequence. In addition, a novel muscle attachment site expansion index (HPE) was evaluated against the traditional TPE index, in hopes that it would provide a means to distinguish proximal caudal vertebrae of prehensile and nonprehensile tails (unlike TPE).
As expected, results presented here indicate differences in muscle attachment site morphology broadly related to tail function: prehensile-tailed taxa tend to have more expanded muscle attachment sites than nonprehensile-tailed taxa. These differences become increasingly drastic (and statistically significant) further distally within the tail vertebral sequence in both groups, confirming the results of previous analyses (German, 1982; Youlatos, 2003; Schmitt et al., 2005). That said, results of the muscle attachment site analyses are not straightforward.
TPE can distinguish the vertebrae of prehensile- and nonprehensile-tailed platyrrhines only in the distal region of the tail; in the other two regions, there are no consistent patterns of variation among prehensile and nonprehensile tail vertebrae. HPE performs much better in distinguishing between tail types, although only in the transitional (LV) and distal (MDV) tail regions. In the proximal region (TV), prehensile-tailed taxa and nonprehensile-tailed taxa cluster at the high and low ends of the total sample range, respectively, while Pithecia falls in between (and is statistically indistinguishable from either group) (Fig. 7).
On closer examination however, it is evident that the large range of Pithecia for TV HPE is attributable to the composition of the generic sample itself. Recall that the Pithecia sample is the only skeletal sample in this study consisting of more than one species: P. pithecia (N = 5) and P. monachus (N = 3). There are no differences between Pithecia species for LV or MDV HPE indices (Fig. 9). Interestingly, HPE values for the TV in these two taxa cluster at the two limits of the total generic range (Fig. 9). However, the two species are not statistically different from one another (non-parametric Mann-Whitney U = 13.00, df = 6, P = 0.101) probably due to the very small sample sizes; but this likely accounts for the normal distribution of the data for the genus as a whole.
Researchers have only recently recognized differences in postcranial anatomy of P. monachus relative to its congeners, and the predominant theme among these reports is that P. monachus may not specialize in vertical clinging and leaping as much as simple quadrupedal arboreal locomotion (e.g., Meldrum and Lemelin, 1991; Davis et al., 2006; Walker and Davis, 2007). Furthermore, some of this research suggests that the postcranial morphology of P. monachus may more closely resemble the morphologies observed for Chiropotes, another nonprehensile-tailed pitheciine platyrrhine. As Fleagle and Meldrum (1988) predicted, based on its hindlimb morphology, Chiropotes may spend a significant proportion of time in hindlimb suspension, where the middle section of the tail is draped over the support (e.g., Walker, 1993, 1996; Meldrum, 1998). It may be that P. monachus exhibits similar tail-bracing behavior, but with the proximal region braced instead. Future studies of positional behavior in this taxon should help in understanding why this species exhibits high HPE values on par with prehensile-tailed platyrrhines.
The question remains, however, why TPE indices do not fare as well as HPE indices in distinguishing between tail types. TPE indices not only assess the relative mechanical advantage incurred by mm. intertransversarii caudae (the primary rotators and lateral flexors of the tail), but they also reflect the relative size of the proximal attachments for mm. flexor caudae longus et brevis (the primary dorsoventral tail flexors) (German, 1982; Lemelin, 1995; Jones, 2004; Schmitt et al., 2005; Organ, 2007). Hemal processes and their associated hemal arches (i.e., chevron bones) are the distal attachment sites for the long external tendons of mm. flexor caudae longus et brevis. Therefore, HPEs also reflect mechanical advantage that these muscles have during contraction. However, because mm. flexor caudae longus et brevis do not have fleshy origins on hemal processes (only tendinous attachments), muscle belly size does not directly factor into bony morphology. Even though it has long been thought that areas of muscle belly attachment sites (i.e., TPE) vary in predictable ways with muscle size (and by extension, muscle force) (e.g., Hildebrand, 1994), we now understand that the relationship between muscle scars and muscle size is much more complex (e.g., Antón, 1996; Zumwalt, 2005, 2006). Making no assumptions about the internal architecture of caudal musculature between tail types and/or taxonomic groups, examining HPEs (instead of TPEs) effectively takes muscle size (i.e., force) out of the equation, and instead only measures mechanical advantage (i.e., distance). Therefore, if lateral and dorsoventral caudal flexor musculature is capable of producing equal forces in both tail types, and if those forces are equitably applied to the hemal arches in both tail types, the larger expansions of prehensile tail hemal processes from the center of the joint will create a proportionally larger bending moment during tail flexion than the smaller expansions of nonprehensile tail hemal processes.
Caudal Vertebral Structural Properties
Internal geometry of caudal vertebrae was expected to differ among prehensile- and nonprehensile-tailed taxa, contributing to higher torsional and bending strength and rigidity in the prehensile tail vertebrae. Furthermore, these predicted differences were expected to increase further distally within the tail, following previous analysis of external geometry (German, 1982).
As predicted, there are clear differences in the relative torsional (and bending) strengths of caudal vertebrae across tail types: prehensile tail caudal vertebrae are structured to resist relatively higher magnitudes of loading than nonprehensile tail caudal vertebrae, and thus are stronger in bending and torsion than their functional analogs in nonprehensile-tailed animals. Moreover, the differences between tail types become more drastic further distally within the tail vertebral sequence. These results accord well with cross-sectional properties reflecting resistance to pure axial tension and compression (not examined here), where the relative vertebral cortical area is highest in prehensile-tailed platyrrhines, and the disparity between tail types becomes more drastic in the transitional and distal regions of the tail (Organ, 2007).
The results of the present study, however, do not completely distinguish between tail types in the proximal tail region, as the TV of Pithecia exhibits torsional strength properties statistically equivalent to both prehensile-tailed Cebus and the rest of the nonprehensile-tailed taxa. Therefore, on the surface these results would appear to mirror the results of previous studies of external geometry (German, 1982). German (1982) explained the lack of differences between the proximal regions of prehensile- and nonprehensile-tailed platyrrhines by concluding that proximal tail-use by these animals should be so similar that differences in structure likely would not exist. Specifically, German (1982) suggested that monkeys, irrespective of tail type, abduct, adduct, flex, and extend their tails and that these motions should be very similar across taxa. Therefore, because these movements are so similar, no differences in morphology should be observable between tails.
Aside from the Pithecia data, the results of the present analysis indicate that the internal structure of tail vertebrae differ between prehensile and nonprehensile tails, even in the proximal tail. The data presented here indicate that size-standardized internal structural data of caudal vertebrae provide much more reliable approximations of tail-use behavior, suggesting once again that internal data may detect differences not always reflected by external considerations alone (e.g., Ruff, 1987, 2002; Selker and Carter, 1989; Daegling, 1989, 2002; Daegling and Grine, 1991; Biknevicius, 1993; Jungers et al., 1998; Organ et al., 2006).
Functional Convergence of Tail Structure in Atelines and Cebus
Prehensile tails are thought to have evolved twice (in parallel) in New World monkeys (Napier, 1976; Rosenberger, 1983). The prehensile tail found among the monophyletic Atelinae (Alouatta, Ateles, Brachyteles, and Lagothrix) represents one of these evolutionary events, while the second event occurred somewhere in the evolutionary history of the genus Cebus. Differences in tail-use behaviors exist between the atelines and Cebus (Table 2), perhaps resulting from different evolutionary histories of their prehensile tails. Atelines often employ full or hindlimb-assisted tail suspension during locomotor and feeding bouts, whereas Cebus regularly wraps its tail around a substrate to form a tripodal posture with its hindlimbs during feeding bouts, but rarely suspends its entire body weight from its tail (Bergeson, 1996; also see Table 2).
Given these differences in tail-use behaviors, it is not surprising that some details of the functional anatomy of prehensile tails also differ between these taxa. The ateline prehensile tail is categorized as having a hairless volar pad on the ventral surface of the distal tail, allowing it maintain sufficient friction against the substrate during prehension and suspension (Lemelin, 1995; Organ, 2007). The prehensile tail of Cebus, however, does not have a volar pad, maintaining a completely furred distal tail. Furthermore, while the tendons of mm. flexor caudae longus (FCL, the primary dorsoventral flexor of the tail) and extensor caudae lateralis (ECL, a major tail extensor) cross the same number of vertebral joint segments in Atelines and Cebus—FCL crosses 4–7 segments in prehensile tails and 15–17 segments in nonprehensile tails; ECL crosses 7–9 segments in prehensile tails and ∼12 segments in nonprehensile tails (Lemelin, 1995)—the morphology of the ECL muscle bellies differs remarkably. In the Atelines, the ECL muscle bellies are diminutive and dorsoventrally flattened, while in Cebus (and nonprehensile-tailed platyrrhines) ECL is bulky and well-developed (Lemelin, 1995). Furthermore, the mm. intertransversarii caudae (the primary lateral flexors and rotators of the tail) are capable of producing similar maximum muscle forces (tetanic tension) in Cebus and atelines, higher than possible in nonprehensile tails (Organ et al., 2009). All of these morphologies taken together suggest that the prehensile tail of Cebus is intermediate in form compared to the atelines on the one hand, and the nonprehensile-tailed platyrrhines on the other.
What is most remarkable, given these differences in soft-tissue morphologies among ateline and Cebus tails, is that the morphologies of caudal vertebrae of these taxa functionally converge on a general prehensile tail condition. Atelines and Cebus generally have more expanded proximal tail regions than nonprehensile-tailed platyrrhines. All prehensile-tailed platyrrhines examined here display more expanded muscle attachment sites in the distal tail region, suggesting (not surprisingly) that mechanical loading in the distal region of the tail is significantly greater in prehensile tails than in nonprehensile tails. However, more proximally within the tail, the results are less distinctive in that the proximal (TV) tail region of Pithecia monachus complicates the simple comparisons of HPE indices in prehensile and nonprehensile-tailed platyrrhines. In general, prehensile-tailed platyrrhines have caudal vertebrae with more expanded hemal processes and more expanded transverse processes than do their nonprehensile-tailed relatives. Finally, platyrrhine prehensile tail caudal vertebrae are stronger in torsion than functionally homologous nonprehensile caudal vertebrae, suggesting increased resistance to higher magnitudes of mechanical loading.
In sum, the general structure of tails in New World monkeys positively correlates with habitual tail-use behavior, in the sense that prehensile tails are mechanically designed to withstand higher bending and torsional stresses than nonprehensile tails. Similar tail morphology related to prehensility is found among procyonid carnivorans (and likely other mammalian groups as well), occupying similar (and often overlapping) Neotropical habitats (Organ, 2007). Thus, because platyrrhine evolution has produced two separate, but functionally convergent, solutions to the problem of navigating arboreal habitats, the next step should be to examine the platyrrhine fossil record for clues to these evolutionary events. For example, Cebupitheciasarmientoi (12.5 Ma, Miocene, La Venta, Colombia) (Stirton, 1951; Stirton and Savage, 1951; Meldrum and Lemelin, 1991), a pitheciine platyrrhine thought to occupy a phylogenetic position close to the base of the pitheciine radiation, has been reported to have a long, nonprehensile tail (Meldrum and Lemelin, 1991). These conclusions, however, are based on external measurements alone, and have included similar “robusticity” measures like TPE indices and width/length. Application of the novel HPE index to this fossil, in addition to examining the internal geometric structure of its caudal vertebrae, may provide further arguments for or against its presumed tail-use behavior (see Organ and Lemelin, 2009). This is especially pertinent given that Meldrum and Lemelin (1991) argued for many postcranial morphologic similarities between Cebupithecia and Pithecia monachus, which is shown here to possess unique proximal caudal vertebral morphology.
For access to specimens, the author wishes to thank R. Thorington, J. Mead, L. Gordon, J. Jacobs, and H. Kafka at the Smithsonian National Museum of Natural History; R. MacPhee, E. Westwig, and W. Harcourt-Smith at the American Museum of Natural History; and K. Glander at Duke University. The author is indebted to C. Ruff for his expert advice and use of his pQCT scanner for the duration of this project, and to M. Rose and K. Glander for sharing unpublished data. M. Teaford, C. Ruff, A. Taylor, P. Lemelin, and R. German provided valuable comments and discussions of earlier versions of this work. Finally, the author wishes to thank T. Smith, V. DeLeon, C. Steinkoenig, Q. Wang, J. Laitman, and two anonymous reviewers whose comments significantly improved this manuscript. A portion of this work was awarded the 2006 Mildred Trotter Prize by the American Association of Physical Anthropologists.
The tail of Cebus more specifically would be categorized by Emmons and Gentry (1983) as “semiprehensile” (Fig 2b), given the heavy reliance of hindlimb-assisted tail bracing (the so-called tripodal stance) this genus frequently exhibits. However, previous research suggests that the structure of caudal vertebrae and caudal musculature of Cebus is more than adequate to fully suspend the body weight of the animal (e.g., Lemelin, 1995; Organ, 2007; Organ et al., 2009), and these animals are sometimes observed engaging in full-body tail suspension (e.g., Bergeson, 1996). Therefore, the tail of Cebus is regarded as “prehensile” for this study, and is considered along with the fully prehensile tails of ateline platyrrhines.
Recently, Ashe et al. (2006) suggested that the specific algorithms included with the XCT software may underestimate bone strength data compared to more traditional methods like 3-point bending tests. However, these authors are careful to point out that the specimens they used (i.e., aged female human radii) were not ideal for such a comparative analysis, and their results may drastically differ from other similar analyses in the future. Such a finding, however, should not affect interpretations in the present analysis because all data were collected using the same method, and therefore results should be internally comparable.
Occasionally the TV has only one pair of transverse processes: the distal pair. In such cases, standardization of the vertebra prior to pQCT scanning involved using the dorsoventral midpoint of the proximal vertebral articular surface as a set point through which the coronal axis passed.
Lagothrix is also indistinguishable from Pithecia in percentage of total vertebrae in the proximal region.
APPENDIX: CAUDAL VERTEBRAL NUMBER, SERIATION, AND MISSING ELEMENTS
Caudal vertebrae were occasionally missing from museum skeletal collections, in part because they are difficult to preserve during skeletal maceration due to their size. In all instances in this study, the missing vertebrae belonged to the distal-most aspects of the vertebral sequence and not the proximal or middle regions. This was especially true for the smallest taxa, but it did occur with larger ones as well. Determining whether a caudal series was or was not complete proved particularly difficult when compounded with little comparative caudal vertebral number data in the literature. To determine the completeness of caudal vertebral sequences for this study, disarticulated tail skeletons were seriated proximodistally (see below) and compared when possible against articulated tail skeletons from the same taxa. In the event that no articulated specimens were available, the total number of disarticulated caudal vertebrae was compared against literature data (e.g., Dor, 1937; Ankel, 1962) and/or unpublished data kindly provided by Dr. Michael D. Rose (Department of Evolutionary Anthropology, Duke University).
Full articulation of the caudal vertebral sequence was not a requisite for inclusion of tail specimens in this study. Inclusion of partially and fully disarticulated tails was possible because the order of caudal vertebrae within the tail sequence can be relatively easily seriated based on well-documented changes in the relative external breadths and lengths of particular vertebral morphologies: relative transverse process breadth, zygapophyseal intrafacet distances, and hemal process expansion all decrease steadily throughout the tail (German, 1982; Organ, 2007), enabling specimens with disarticulated caudal vertebrae to be utilized if all vertebrae are present.
However, as also mentioned earlier, there were instances (19 of 88 individuals, 21.6%) where the distal-most vertebrae were missing from the sequence. In such instances, at most the two distal-most vertebrae were missing, which affected the data collected for this study in two ways. First, the overall length of the tail was underestimated in these individuals, although by relatively negligible amounts (at most 8–10 mm). The second way in which the data were affected is probably more significant in that the relative position of the mid-distal vertebra (MDV, Fig. 2) had to be approximated, and may have varied from its true position. The net effect of this estimation was an increase in the variance for MDV linear and structural data detailed in this article. However, in most cases the between-taxa variance was greater than the within-taxa variance for the measurements analyzed, suggesting that the error introduced by estimating the position of the MDV was negligible.