Original Articles
Tail length estimation from sacrocaudal skeletal morphology in catarrhines
2013 Volume 121 Issue 1 Pages 13-24
Details
2013 Volume 121 Issue 1 Pages 13-24
Tail morphology in primates is important for interpreting functional adaptation and phylogeny. Tail length is probably the most remarkable trait. Establishing usable methods to predict the tail length of extinct primates is part of the basis for the reconstruction of primate evolution, particularly of hominoids. Previous studies revealed that sacral morphology often predicts tail length. However, most previous studies have only attempted this by using categories (short, long, etc.). A problem with those studies is that the range of short tail length is wide. Accordingly, this study aimed to quantitatively estimate tail length in catarrhines with intermediate tail lengths. Sacra and proximal caudal vertebrae (first to third) of 89 hybrid individuals between Japanese macaque (Macaca fuscata) and Formosan rock macaque (Macaca cyclopis) were measured. These hybrid macaques were phylogenetically controlled but varied greatly in their relative tail length (tail length/head and body length = 18.8–88.8%), and thus were an excellent sample to obtain general regression formulae to estimate catarrhine tail length. A total of 15 predicting models were devised and five formulae performed well. The utility of these formulae were confirmed by application to 15 species/subspecies of catarrhine taxa.
Tail length in primates varies greatly between species or even among local conspecific populations (Napier and Napier, 1967; Fooden, 1975, 1988, 2006; Napier, 1981; Hamada et al., 2005, 2008a; see also Table 1). Thus, tail length is an important trait for understanding primate adaptation and phylogeny, for example thermoregulation (Wilson, 1972; Fooden and Albrecht, 1999), balance in arboreal activity (Wilson, 1972; Larson and Stern, 2006) in Old World monkeys, and tail suspension or prehensility in New World monkeys (German, 1982; Rosenberger, 1983; Lemelin, 1995; Youlatos, 1998, 2002; Cant et al., 2001, 2003; Schmitt et al., 2005; Organ, 2007, 2010). In primates, the tail is markedly reduced in several lineages (e.g. apes, some macaques, lorises, some subfossil lemurs), but the evolutionary process and adaptive significance of tail reduction has not been thoroughly established. It is, therefore, important to obtain a reliable estimate of tail length from partial skeletal elements for the fossils of extinct primates.
| No.a | Taxon | Sex | N | Mean RTL (%) | RTL (%) |
|---|---|---|---|---|---|
|
| |||||
| Hybrid macaque | M | 30 | — | 18.3–88.8b | |
| F | 58 | — | 18.9–87.2b | ||
| NA | 1 | — | (38.9b) | ||
| 1 | Cercopithecus mitis kolbi | M | 1 | 113.2b | |
| F | 3 | 124.9b | |||
| 2 | Cercopithecus mitis stuhlmanni | M | 3 | 145.6b | |
| F | 2 | 140.0b | |||
| 3 | Cercopithecus neglectus | M | 2 | 127.8b | |
| F | 1 | 122.2b | |||
| 4 | Cercopithecus aethiops | M | 2 | 131.0b | |
| F | 2 | 143.4b | |||
| 5 | Erythrocebus patas | M | 2 | 84.2c | |
| F | 3 | 100.0c | |||
| 6 | Cercocebus torquatus | M | 1 | 111.5d | |
| F | 1 | 108.5d | |||
| 7 | Papio anubis | M | 1 | 68.0b | |
| 8 | Papio hamadryas | M | 10 | 73.3c | |
| F | 7 | 67.2c | |||
| 9 | Macaca arctoides | M | 1 | 10.3d | |
| 10 | Macaca fascicularis | M | 10 | 117.6e | |
| F | 14 | 116.3e | |||
| 11 | Macaca mulatta | M | 11 | 43.3e | |
| F | 10 | 44.0e | |||
| 12 | Macaca radiata | M | 10 | 107.0f | |
| F | 7 | 110.0f | |||
| 13 | Macaca assamensis | M | 5 | 35.0f | |
| F | 4 | 40.0f | |||
| 14 | Macaca nemestrina | M | 6 | 36.5g | |
| F | 10 | 36.5g | |||
| 15 | Colobus guereza | M | 2 | 102.2b | |
| F | 2 | 106.7b | |||
Previous attempts have addressed this issue using categorical methods. Ankel (1965, 1972) estimated tail length from the sacrum. She calculated the sacral index and explored the relationship between this index and tail length variation based on different numbers of caudal vertebrae. The sacral index devised in her study evaluated the relative development of neural tissue at the caudal sacral canal opening compared to the cranial canal opening. Twenty-one primate species (three species of prosimians, five species of platyrrhines, and 13 species of catarrhines including hominoids) were classified into several groups (e.g. “totally reduced tail,” “reduced tail,” and “prehensiled tail” (in her 1972 study, “tailless forms,” “long-tailed primates,” and “prehensile-tailed forms” were used)). The sacral index successfully distinguished these groups, which had different tail lengths and different numbers of caudal vertebrae. The index value in “reduced tail” became smaller than longer tailed species. Using this index, she also suggested that Pliopithecus (Epipliopithecus), a middle Miocene catarrhine from Europe, had a pendant “long-tail” (the index values were above those for “reduced tail” and below those for “prehensiled tail”) that consisted of 10–15 caudal vertebrae (Ankel, 1965, 1972), though the interpretation that Pliopithecus was tailless was dominant (Zapfe, 1958). Ward et al. (1991) devised a tapering index of the sacrum and applied it to the sacrum of Proconsul heseloni, an early Miocene hominoid from western Kenya, and concluded that it had lost its tail. Their tapering index was calculated from the cranial, caudal width and the vertebral body length of the last sacral vertebra; and tailless primate had a larger tapering index because the caudal end articulated with vestigial coccygeal vertebra rather than a robust Ca1. Nakatsukasa et al. (2003, 2004) also investigated taillessness in Miocene hominoids by using different criteria, such as the shape of the transverse process or trabecular mesh, and concluded that Nacholapithecus kerioi (Nakatsukasa et al., 2003) and Proconsul heseloni (Nakatsukasa et al., 2004) also lacked tails.
Recently, Russo and Shapiro (2011) clarified the relationships between sacral morphology and relative tail length (RTL = tail length/head and body length × 100%) in catarrhines. They classified 22 catarrhine taxa into four groups: “long” (100.0% ≤ RTL), “short” (37.0% ≤ RTL < 100.0%), “very short” (0.0% < RTL < 15.0%), and “absent” (RTL = 0%; hominoids). The study tested the utility of sacral measurements to distinguish these four groups. Three parameters, concerning the shape of the sacral caudal articular surface, the sacrocaudal articulation angle (Schmitt et al., 2005) and the lateral expansion of the transverse processes, varied between these categories. Among these features, the sacrocaudal articulation angle (Schmitt et al., 2005) distinguished three groups, “long,” “short,” and “very short” groups, separately. The other features were useful to separate “long” and “tailless” groups from the “short”/“very short” groups although the study did not distinguish “short” and “very short” groups. Russo and Shapiro (2011) provided useful results for interpreting primate tail length. However, the “short”-tailed group in their study still covers a broad range of tail length (37.0% ≤ RTL < 100.0%) and correlations between these parameters and RTL are low within each group. This study built on these previous results with the intention of estimating primate tail length with greater precision.
Specifically, this study aimed to devise reliable formulae for predicting the RTL quantitatively from the sacrum and proximal caudal vertebrae by using a phylogenetically narrow catarrhine sample (hybrid individuals of Japanese macaque and Formosan rock macaque). The study then tested the broader utility of the predictive formulae by applying the functions to 15 catarrhine species/subspecies other than Japanese and Formosan rock macaques (Table 1).
This study collected predictive measurements from the skeletal specimens of hybrid Japanese macaques (Macaca fuscata) and Formosan rock macaques (Macaca cyclopis). The skeletal specimens are housed at the Primate Research Institute of Kyoto University. The two species express close phylogenetic relatedness but have significantly different tail lengths (Fooden, 2006). Tail lengths and numbers of caudal vertebrae in these hybrids vary greatly depending on the degree of hybridization (Hamada et al., 2008b). Additional components of morphological variability such as inter-temporal distance (Mouri et al., 2008) and the size of incisors and molars (Kunimatsu and Yamamoto, 2008) also vary between these two species. Although M. fuscata and M. cyclopis are allopatrically distributed, the artificial introduction of M. cyclopis into Japan has caused sporadic hybridization in a certain region. About 50 years ago, a troop of M. cyclopis escaped from an abandoned private zoo park in Wakayama prefecture. The domesticated macaques then merged with wild local troops of M. fuscata and began to produce fertile offspring (Kawamoto et al., 1999, 2001; Ohsawa et al., 2005). As the population of hybrid macaques increased, the local government eradicated the groups to prevent further anthropogenic genetic disturbance in the wild M. fuscata populations following the Guidelines of Wild Life Management set by the Ministry of Environment. The captured macaques were killed.
Eighty-nine of these adult specimens were selected. The term adult references full emergence of the third molars. Their RTLs (tail length/head and body length (sitting height) × 100) were calculated using somatometric tail length and sitting height from the cadavers (Hamada et al., 2008b). Generally, RTL is calculated from tail length and head and body length; however, head and body length were not measured in the hybrid samples. Concerning this matter, Dr Hamada noted that the difference between head and body length and sitting height is lesser than 5% in the hybrids and the intraspecies RTL difference was much bigger (personal communication, 2011). Therefore in this study, I decided to use sitting height instead of head and body length. RTLs of the adult hybrids varied from 18.3% to 88.8% (from 101 to 470 mm in tail length).
Twenty linear measurements were collected from the sacrum and the first to third caudal vertebrae (Ca1–Ca3) using digital sliding callipers with an accuracy of 0.1 mm (Mitsutoyo Co., Ltd.) (Figure 1). Previous studies have revealed that the caudal end of the sacrum which was measured relative to the cranial end tends to reflect tail length strongly (Ankel, 1965, 1972; Ward et al., 1991; Russo and Shapiro, 2011). Six measurements from the caudal side of the sacrum including the measurements used in the previous studies (measurements No. 5 and 6 from Ankel (1965, 1972); measurements No. 7, 8, and 11 from Russo and Shapiro (2011); measurement No. 12 is original to this study; see Figure 1) were therefore collected. In addition, 14 measurements on the cranial and caudal sides of the sacrum were also collected. For example, measurements No. 1, 2, 5, 6 are used to calculate Ankel’s sacral index (Ankel, 1965, 1972). Sacral breadth was measured at three different levels and the total sacral length was measured. Generally, the measurements taken from the caudal side of the sacrum decreased when tail length became shorter. The Ca1–Ca3 are readily identified morphologically even from isolated fossilized materials and were used in this study. More distal caudal vertebrae are identifiable only in cases where a substantial number of caudal vertebrae are associated.
Measurements taken: (A) cranial articular surface of the sacrum (1, sagittal diameter of the sacral canal cranial opening; 2, transverse diameter of the sacral canal cranial opening; 3, transverse diameter of the cranial articular surface; 4, sagittal diameter of the cranial articular surface); (B) caudal articular surface (5, sagittal diameter of the sacral canal caudal opening; 6, transverse diameter of the sacral canal caudal opening; 7, transverse diameter of the caudal articular surface; 8, sagittal diameter of the caudal articular surface); (C) dorsal view of the sacrum (9, cranial breadth of the sacrum; 10, inter-prezygapophyseal breadth; 11, breadth at the transverse process of the last sacral vertebra; 12, inter-postzygapophyseal breadth); (D) ventral view of the sacrum (13, ventral length; 14, middle breadth); (E) dorsal view of the proximal caudal vertebrae (15, breadth at the transverse process (1st caudal vertebra); 17, breadth at the transverse process (2nd caudal vertebra); 19, breadth at the transverse process (3rd caudal vertebra)); (F) lateral view of the proximal caudal vertebrae (16, length of the vertebral body (1st caudal vertebra); 18, length of the vertebral body (2nd caudal vertebra); 20, length of the vertebral body (3rd caudal vertebra))
In order to devise a set of RTL prediction formulae, these measurements were standardized in two different ways and selected for performance of the multiple regression analyses. The following rule was used for the strength of correlations in this study: 0.5 ≤ r < 0.7 was defined as a moderate positive correlation and r ≥ 0.7 was defined as a strong correlation (Sokal and Rohlf, 1995).
Standardization of measurementsTo predict RTL, body size effects must be removed from the measurements. I standardized the measurements by the geometric mean of the cranial articular surface of the sacrum, which was calculated as (GMscra = (measurements No. 3 × No. 4)0.5) (Figure 1). GMscra was correlated strongly with the head and body length (r = 0.74), but poorly with tail length (r = −0.43). Measurements were divided by GM-scra.
Selection of variables from the standardized measurementsThe correlation matrix of the standardized measurements (except measurements No. 3 and 4, which were used for size standardization) and RTL was calculated (Table 2). Measurements that show a low correlation (r < 0.5) with RTL were eliminated from further steps of the analyses. Then, in order to avoid multicollinearity in the multiple regression analysis, pairs of measurements that show a high correlation with each other (r > 0.7) (Table 2) were selected and the term that expressed the weaker correlation with RTL was dropped. The remaining standard measurements were used for multiple regression analysis.
| Index | Abbreviation | Definitiona |
|---|---|---|
| Transverse diameter of the sacral canal (cranial sacral vertebra vs. caudal sacral vertebra) | TDSC | No. 6/No. 2 |
| Sagittal diameter of the sacral canal (cranial sacral vertebra vs. caudal sacral vertebra) | SDSC | No. 5/No. 1 |
| Transverse diameter of the sacral articular surface (cranial sacral vertebra vs. caudal sacral vertebra) | TDSAS | No. 7/No. 3 |
| Sagittal diameter of the sacral articular surface (cranial sacral vertebra vs. caudal sacral vertebra) | SDSAS | No. 8/No. 4 |
| Mediolateral and dorsoventral breadth of the articular surface of the last sacral vertebra | MLDV | No. 7/No. 8 |
| Inter-zygagophyseal breadth of the sacrum (cranial sacral vertebra vs. caudal sacral vertebra) | SZL | No. 12/No. 10 |
| Sacral breadth 1 (cranial vs. middle level) | SB1 | No. 14/No. 9 |
| Sacral breadth 2 (cranial vs. caudal level) | SB2 | No. 11/No. 9 |
| Sacral breadth 3 (middle vs. caudal level) | SB3 | No. 11/No. 14 |
| Breadth at the transverse process (last sacral vertebra vs. 1st caudal vertebra) | LSTP1 | No. 15/No. 11 |
| Breadth at the transverse process (last sacral vertebra vs. 2nd caudal vertebra) | LSTP2 | No. 17/No. 11 |
| Breadth at the transverse process (last sacral vertebra vs. 3rd caudal vertebra) | LSTP3 | No. 19/No. 11 |
| Breadth at the transverse process (1st caudal vertebra vs. 2nd caudal vertebra) | TP12 | No. 17/No. 15 |
| Breadth at the transverse process (1st caudal vertebra vs. 3rd caudal vertebra) | TP13 | No. 19/No. 15 |
| Breadth at the transverse process (2nd caudal vertebra vs. 3rd caudal vertebra) | TP23 | No. 19/No. 17 |
| Length of the vertebral body (1st caudal vertebra vs. 2nd caudal vertebra) | VB12 | No. 18/No. 16 |
| Length of the vertebral body (1st caudal vertebra vs. 3rd caudal vertebra) | VB13 | No. 20/No. 16 |
| Length of the vertebral body (2nd caudal vertebra vs. 3rd caudal vertebra) | VB23 | No. 20/No. 18 |
Multiple regression analysis was performed using all remaining variables. Akaike’s information criterion (Akaike, 1973) was used to select the best multiple regression model by a stepwise method (backward/forward). A stepwise method makes backward/forward steps (increasing/decreasing variables) from a full model (a model using all variables the performer chooses) and chooses the variables that maximize the F-value or P-value.
Reducing variables from the devised regression formulae and obtaining new modelsOther models were calculated by eliminating some measurements of some bony elements or bony parts from the regression formulae devised by the stepwise method. This operation was conducted based on the assumption that various different elements or parts were preserved in fossil specimens. For example, no fossil specimens contain intact sacra and proximal caudal vertebrae, and even though sacra or caudal vertebrae are preserved, the neural arch and transverse processes are often broken.
In the second analysis, indices that are empirically expected to reflect tail development were calculated. Several indices or angles were calculated from sacral measurements that were previously used to evaluate tail length (Ankel, 1965, 1972; Ward et al., 1991, Russo and Shapiro, 2011). Ankel (1965, 1972) demonstrated that the size of the caudal opening of the sacral canal is larger relative to the cranial opening in long-tailed primates. The sacral index she used was defined as: (No. 5 × No. 6)/(No. 1 × No. 2) (Figure 1). Ward et al. (1991) also reached a similar conclusion based on the analysis of the sacral tapering angle to show that Proconsul did not have a tail. In other words, Ankel (1965, 1972) and Ward et al. (1991) showed that the development of the morphology of sacral caudal end compared to the cranial end accurately reflects tail length. Russo and Shapiro (2011) calculated ML/DV ratios (mediolateral breadth of the last sacral vertebra’s articular surface divided by its dorsoventral breadth (measurements No. 7/No. 8 in Figure 1)). Likewise, this study calculated 18 indices from paired measurements of the corresponding cranial and caudal elements (Table 2). The study focused on six measurements on the caudal side of the sacrum (measurements No. 5, 6, 7, 8, 11, and 12), which likely reflect tail development, and divided them by their corresponding cranial measurements. These indices included ones suggested in the previous studies: TDSC (transverse diameter of the sacral canal) and SDSC (sagittal diameter of the sacral canal) were from Ankel (1965, 1972); MLDV (mediolateral and dorsoventral breadth of the articular surface of the last sacral vertebra) was from Russo and Shapiro (2011). The indices were used for multiple regression analyses following the same procedures as noted above (Table 4).
| TDSC | SDSC | TDSAS | SDSAS | MLDV | SZL | SB1 | SB2 | SB3 | LSTP1 | LSTP2 | LSTP3 | TP12 | TP13 | TP23 | VB12 | VB13 | VB23 | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| RTL | −0.30 | 0.06 | 0.62 | 0.72* | −0.58* | 0.67* | 0.20 | 0.73* | 0.67 | −0.47 | −0.47 | −0.19 | −0.26 | 0.14 | 0.40 | 0.46 | 0.51* | 0.41 |
| TDSC | 1.00 | −0.24 | −0.10 | −0.21 | −0.31 | −0.26 | −0.01 | −0.09 | −0.09 | −0.03 | 0.01 | 0.02 | 0.05 | 0.04 | 0.00 | −0.28 | −0.26 | −0.14 |
| SDSC | 1.00 | 0.07 | 0.06 | −0.63 | 0.05 | 0.17 | 0.25 | 0.20 | −0.09 | −0.18 | −0.09 | −0.22 | −0.03 | 0.13 | 0.26 | 0.28 | 0.17 | |
| TDSAS | 1.00 | 0.73 | 0.36 | 0.38 | 0.04 | 0.67 | 0.67 | −0.40 | −0.54 | −0.19 | −0.48 | 0.10 | 0.49 | 0.56 | 0.57 | 0.37 | ||
| SDSAS | 1.00 | −0.32 | 0.54 | 0.03 | 0.69 | 0.69 | −0.41 | −0.48 | −0.16 | −0.36 | 0.14 | 0.46 | 0.48 | 0.55 | 0.44 | |||
| MLDV | 1.00 | −0.53 | −0.07 | −0.51 | −0.49 | 0.36 | 0.37 | 0.19 | 0.23 | −0.05 | −0.26 | −0.38 | −0.54 | −0.56 | ||||
| SZL | 1.00 | 0.06 | 0.57 | 0.56 | −0.39 | −0.45 | −0.32 | −0.29 | −0.03 | 0.19 | 0.43 | 0.47 | 0.37 | |||||
| SB1 | 1.00 | 0.23 | −0.10 | 0.09 | 0.02 | −0.13 | −0.09 | −0.25 | −0.19 | 0.14 | 0.23 | 0.28 | ||||||
| SB2 | 1.00 | 0.94 | −0.77 | −0.82 | −0.53 | −0.51 | 0.01 | 0.40 | 0.55 | 0.62 | 0.49 | |||||||
| SB3 | 1.00 | −0.81 | −0.85 | −0.50 | −0.49 | 0.09 | 0.47 | 0.52 | 0.56 | 0.41 | ||||||||
| LSTP1 | 1.00 | 0.85 | 0.50 | 0.22 | −0.25 | −0.44 | −0.30 | −0.36 | −0.29 | |||||||||
| LSTP2 | 1.00 | 0.70 | 0.70 | 0.09 | −0.40 | −0.59 | −0.61 | −0.41 | ||||||||||
| LSTP3 | 1.00 | 0.60 | 0.71 | 0.38 | −0.50 | −0.65 | −0.62 | |||||||||||
| TP12 | 1.00 | 0.50 | −0.14 | −0.70 | −0.66 | −0.38 | ||||||||||||
| TP13 | 1.00 | 0.78 | −0.32 | −0.45 | −0.47 | |||||||||||||
| TP23 | 1.00 | 0.15 | 0.00 | −0.23 | ||||||||||||||
| VB12 | 1.00 | 0.88 | 0.42 | |||||||||||||||
| VB13 | 1.00 | 0.80 | ||||||||||||||||
| VB23 | 1.00 |
Bold type indicates the correlation coefficient of candidates of variables;
In order to determine the relative contribution of each variable to the prediction models, standard partial regression coefficients (SPRCs) were calculated. SPRCs were calculated by multiplying non-standardized coefficients by the ratio between the standard deviation of each independent variable and the standard deviation of the dependent variable.
Examination of the applicability of modelsThe utility of the predicting formulae was tested by applying the results to 15 catarrhine species/subspecies other than Japanese and Formosan rock macaques (Table 1). Apes were not included in the samples because the number of sacral vertebrae and caudal vertebrae are greatly different between extant apes and the monkeys. The prediction formulae were calculated from the hybrid macaques possessing three sacral vertebrae. Therefore, these formulae are designed for ‘monkey-like’ specimens with three or four sacral vertebrae (as per Zapfe, 1958). This purported shortcoming does not limit the utility of this study in reconstructing the tail reduction process in hominoid evolution. In fact, Rose et al. (1996) inferred from the sacral morphology that the tailless Nacholapithecus kerioi had fewer sacral vertebra like monkeys. This is also supported by a greater number of lumbar vertebrae. Among the comparative cercopithecid samples, skeletons associated with individual RTL records were limited to six species/subspecies: Cercopithecus mitis kolbi, C. mitis stuhlmanni, C. neglectus, C. aethiops, Papio anubis, Colobus guereza were associated with the field record; RTLs of M. arctoides and Cercocebus torquatus were directly measured from the associated skin. For seven other species, the RTL was obtained from the literature: Erythrocebus patas, P. hamadryas, Napier (1981); M. fascisularis, M. mulatta, Fooden (2006); M. radiata, M. assamensis, Fooden (1988); M. nemestrina, Fooden (1975). These specimens are housed at the Natural History Museum, London and Primate Research Institute of Kyoto University.
The precision of each model was evaluated by the adjusted R2 and %SEE (standard error of estimates). The adjusted R2 is calculated from the SEE and the sample standard deviation as follows. The smaller SEE becomes, the closer to 1 if the adjusted R2 becomes more precise. The model follows from:
Assuming, measured value: x, estimated value: y, and sample size: n.
Then, SEE = sqrt[∑(x − y)2/(n − 2)].
The sample standard deviation of y (σy2) = ∑(y − ymean)/(n − 1)
Adjusted R2 = (σy2 − SEE2)/σy2
The %SEE (SEE/average of estimated values × 100%) of each model was also calculated and the results are shown in Table 5. The estimated values of these 15 species were plotted (Y) on the real values (X) in scatter plots (Figure 2). The precision of the estimated values was evaluated by using the 95% confidence interval (CI). In Figure 2, the reference line (X–Y) was drawn with the 95% CI (x±1.96 SEE). If the values are plotted over the maximum limit of the CI or under the minimum limit, then the values can be considered over-or underestimated, respectively.
| Model | Used bony element(s) | Adjusted R2 | %SEE | Modelsa,b,c |
|---|---|---|---|---|
| 1 | Sacrum and 2nd caudal vertebra | 0.64 | 11.67 | RTL = 8.233 × No. 8 + 6.981 × No. 12 + 3.313 × No. 17 − 50.643 |
| 2 | Sacrum | 0.65 | 12.24 | RTL = 9.828 × No. 8 + 8.168 × No. 12 − 50.557 |
| 3 | Sacrum | 0.50 | 11.63 | RTL = 13.494 × No. 8 − 50.117 |
| 4 | Sacrum | 0.42 | 13.39 | RTL = 12.308 × No. 12 − 51.126 |
| 5 | 1st caudal vertebra | 0.56 | 16.62 | RTL = 9.921 × No. 15 + 6.269 × No. 16 − 50.082 |
| 6 | 2nd caudal vertebra | 0.51 | 16.62 | RTL = 7.963 × No. 17 + 7.320 × No. 18 − 50.226 |
| 7 | 3rd caudal vertebra | 0.49 | 11.32 | RTL = 8.350 × No. 19 +6.457 × No. 20 − 51.419 |
|
| ||||
| 8 | Sacrum | 0.69 | 8.23 | RTL = 7.7591 × SZL + 6.2601 × SB2 + 4.4097 × SDSAS − 0.9741 × MLDV + 49.4459 |
| 9 | Sacrum | 0.70 | 15.05 | RTL = 7.885 × SZL + 4.833 × SDSAS + 6.358 × SB2 − 49.437 |
| 10 | Sacrum | 0.68 | 8.98 | RTL = 8.661 × SB2 + 8.112 × SZL − 2.563 × MLDV + 49.476 |
| 11 | Sacrum | 0.46 | 13.13 | RTL = 12.983 × SDSAS − 50.111 |
| 12 | Sacrum | 0.46 | 13.47 | RTL = 12.972 × SZL − 51.161 |
| 13 | Sacrum | 0.54 | 10.40 | RTL = 14.648 × SB2 − 49.584 |
| 14 | Sacrum | 0.28 | 18.57 | RTL = −10.173 × MLDV + 49.976 |
| 15 | 1st and 3rd caudal vertebrae | 0.21 | 17.60 | RTL = 8.484 × VB13 − 52.049 |
The results of model 1 (A), 2 (B), 8 (C), 9 (D), and 10 (E) (Table 5) applied to catarrhine samples. The solid line is the X–Y reference line and the two dashed lines enclosing the reference line show the 95% confidence limits. The plotted numbers are the same as in Table 1. The results for males and females were plotted separately (male: diamonds; female: circles). The two vertical lines show the RTL variation of the hybrid macaques and the vertical dashed lines show the ‘short’ category in Russo and Shapiro (2011).
| No. 1 | No. 2 | No. 5 | No. 6 | No. 7 | No. 8 | No. 9 | No. 10 | No. 11 | No. 12 | No. 13 | No. 14 | No. 15 | No. 16 | No. 17 | No. 18 | No. 19 | No. 20 | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| RTL | 0.34 | 0.43 | 0.20 | 0.41 | 0.59 | 0.76* | 0.15 | −0.20 | 0.73 | 0.61* | 0.31 | 0.29 | 0.65 | 0.56 | 0.63* | 0.69* | 0.74 | 0.69 |
| No. 1 | 1.00 | 0.47 | 0.22 | 0.69 | 0.38 | 0.36 | 0.38 | 0.18 | 0.18 | 0.43 | 0.51 | 0.26 | 0.31 | 0.50 | 0.50 | 0.41 | 0.31 | 0.36 |
| No. 2 | 1.00 | 0.57 | 0.64 | 0.49 | 0.55 | 0.29 | −0.14 | 0.46 | 0.20 | 0.47 | 0.44 | 0.59 | 0.43 | 0.61 | 0.54 | 0.50 | 0.58 | |
| No. 5 | 1.00 | 0.22 | 0.43 | 0.38 | 0.07 | −0.37 | 0.34 | −0.08 | 0.31 | 0.26 | 0.36 | 0.23 | 0.34 | 0.25 | 0.35 | 0.32 | ||
| No. 6 | 1.00 | 0.40 | 0.52 | 0.29 | −0.11 | 0.38 | 0.39 | 0.40 | 0.32 | 0.53 | 0.43 | 0.60 | 0.49 | 0.46 | 0.48 | |||
| No. 7 | 1.00 | 0.81 | 0.31 | −0.15 | 0.7 | 0.32 | 0.48 | 0.29 | 0.72 | 0.59 | 0.54 | 0.74 | 0.72 | 0.70 | ||||
| No. 8 | 1.00 | 0.30 | −0.30 | 0.81 | 0.47 | 0.51 | 0.32 | 0.80 | 0.62 | 0.65 | 0.78 | 0.82 | 0.82 | |||||
| No. 9 | 1.00 | 0.25 | 0.28 | 0.30 | 0.38 | 0.58 | 0.40 | 0.25 | 0.27 | 0.35 | 0.10 | 0.39 | ||||||
| No. 10 | 1.00 | −0.27 | 0.05 | −0.08 | −0.09 | −0.24 | −0.04 | −0.13 | −0.17 | −0.30 | −0.28 | |||||||
| No. 11 | 1.00 | 0.53 | 0.39 | 0.36 | 0.80 | 0.50 | 0.52 | 0.74 | 0.64 | 0.81 | ||||||||
| No. 12 | 1.00 | 0.24 | 0.21 | 0.46 | 0.42 | 0.44 | 0.52 | 0.35 | 0.51 | |||||||||
| No. 13 | 1.00 | 0.46 | 0.57 | 0.66 | 0.57 | 0.65 | 0.38 | 0.67 | ||||||||||
| No. 14 | 1.00 | 0.56 | 0.23 | 0.46 | 0.39 | 0.14 | 0.48 | |||||||||||
| No. 15 | 1.00 | 0.49 | 0.71 | 0.76 | 0.61 | 0.81 | ||||||||||||
| No. 16 | 1.00 | 0.64 | 0.82 | 0.71 | 0.72 | |||||||||||||
| No. 17 | 1.00 | 0.60 | 0.68 | 0.61 | ||||||||||||||
| No. 18 | 1.00 | 0.70 | 0.94 | |||||||||||||||
| No. 19 | 1.00 | 0.61 | ||||||||||||||||
| No. 20 | 1.00 |
Bold type indicates the correlation coefficient of candidates of variables;
Statistical analyses were performed using R software version 2.9 (R Development Core Team, 2009).
Out of 10 measurements which had positive moderate correlation (r > 0.5) with RTL, four were selected for multiple regression analyses. Two variables were from the sacrum (measurements No. 8, No. 12) and two from Ca2 (measurements No. 17, No. 18) (Table 1). Using these measurements, a total of seven predictive models were produced (Table 5).
The best-fit regression formula selected by the stepwise method was model 1 (adjusted R2 = 0.64) (Figure 2, Table 5). This model used measurement No. 8 (sagittal diameter of the caudal articular surface of the sacrum), No. 12 (inter-postzygapophyseal breadth of the sacrum), and No. 17 (breadth at the transverse process of Ca2) to predict tail length. The SPRCs were 8.233 for No. 12, 6.981 for No. 8, and 3.313 for No. 17.
Model 2 was obtained using the same sacral measurements reported in model 1 (adjusted R2 = 0.65) (Figure 2, Table 5). The SPRCs were 9.828 for No. 8 (sagittal diameter of the caudal articular surface of the sacrum) and 8.168 for No. 12 (inter-postzygapophyseal breadth of the sacrum).
Models 3 and 4 were univariate regression models using sacral measurements No. 8 (sagittal diameter of the caudal articular surface of the sacrum) and No. 12 (inter-postzygapophyseal breadth of the sacrum), respectively. The adjusted R2 values of model 3 and model 4 were 0.50 and 0.42.
In order to check the possibility of tail length estimation from caudal vertebrae, Models 5–7 were calculated from the breadth of the transverse process and length of the vertebral body for each of the proximal caudal vertebrae (Table 5). The adjusted R2s of models 5, 6, and 7 were moderately high. The highest (0.56) was from Ca1.
Predictive models using the indicesTable 4 shows the correlation matrix of the 18 indices and RTL. Six indices had moderate correlations (r > 0.5) with RTL. After eliminating SB3 (middle vs. caudal level sacral breadth) and TDSAS (cranial vs. caudal transverse diameter of the sacral articular surface), which were strongly correlated with SB2 and SDSAS, respectively, five indices (SZL, SDSAS, MLDV, SB2, and VB13) were used for multiple regression analysis. Using these indices, eight models were produced (Table 5).
By the stepwise method, model 8, which uses four indices, was selected. These indices include SZL (ratio of cranial vs. caudal inter-zygophyseal breadth of the sacrum), SB2 (cranial vs. caudal sacral breadth 2, see Table 2, Figure 1), SDSAS (cranial vs. caudal sagittal diameter of the sacral articular surface), and MLDV (mediolateral vs. dorsoventral breadth of the articular surface of the last sacral vertebra). The adjusted R2 for model 8 was 0.69 and the SPRC of the variables was highest in SZL and lowest in MLDV. This model uses two indices (SDSAS and MLDV) reflecting the articular surface morphology of the last sacral vertebra. Therefore, two derivative models from this original model were calculated by dropping one of the two indices (model 9 without MLDV and model 10 without SDSAS). The adjusted R2 (0.70) for model 9 is as high as that produced in model 8 (Table 5, Figure 1). This value represents the highest score for the predicting formulae obtained by this study (Table 4). The SPRCs for this model were 7.885 for SZL, 6.358 for SB2, and 4.833 for SDSAS (Table 5). The adjusted R2 of model 10 was 0.68, and slightly lower than that of models 8 and 9. The SPRC was lowest for MLDV as in model 8.
Univariate regression formulae (models 11, 12, 13, 14, 15) were calculated from SDSAS, SZL, SB2, MLDV, and VB13 (length of the vertebral body of third caudal vertebra compared to first caudal vertebra). The adjusted R2s of these models were 0.46, 0.46, 0.54, 0.28, and 0.21 respectively (Table 5). The adjusted R2 was the highest in model 13 among these five formulae, but the predictive power of these formulae was relatively low. By comparison of the adjusted R2 between model 11 and 14, it is apparent that MLDV’s explanation power of RTL is much lower than that of SDSAS. VB13 is the only index from the proximal caudal vertebrae among the selected indices. The adjusted R2 of the formula using VB13 is low (0.21).
ApplicationTo examine the utility of the obtained models, five formulae (adjusted R2 > 0.60) were selected (models 1, 2, 8, 9, 10) and applied to 15 catarrhine species/subspecies (Table 1, Figure 2). The average RTL values were different between sexes in the catarrhine species, and therefore the values for males and females were plotted separately in order to check the utility exactly (only males were available for Papio anubis and Macaca arctoides). The X–Y reference line and lines showing the 95% CI were drawn. The vertical dashed lines show the range of the ‘short’-tailed category in Russo and Shapiro (2011) and the vertical solid lines show the RTL range of the hybrid macaques used in the present study. Generally, the predicted values for taxa that had RTLs within the range 18.3–88.8% were plotted within 95% CI. Species whose RTL exceeded the range of the hybrid macaques were underestimated. All models seemed to predict everything as around 100 when the species whose RTL is more than 100% were applied. On the other hand, the RTL was overestimated in stump-tailed macaque (Macaca arctoides) with a diminutive tail.
Tail reduction has occurred in various primate taxa. The greatest example of tail reduction is, of course, observed in hominoids, where all extant members have no tail. Some functional and phylogenetical features which might affect tail reduction have been proposed, but the evolutionary process has not been clarified in any lineage since fossils documenting the tail reduction process have not been found. It is certain that future discoveries of fossils will reduce these gaps in the primate fossil record. However, given the rarity of caudal vertebrae preservation in the primate fossil record, it remains important to develop a method that predicts tail length from other aspects of skeletal morphology.
The predicting formulae were devised from hybrid macaques possessing three sacral vertebrae. Therefore, these formulae should be applicable for species with a ‘monkey-like’ sacrum (three or four sacral vertebrae). Deriving these models from a phylogenetically narrow sample does not, however, limit the utility of this study. In fact, Rose et al. (1996) inferred that the tailless Nacholapithecus kerioi had fewer sacral vertebrae, a trait suggesting greater similar to monkeys.
This study also differs from similar previous attempts in two respects. First, subjects were controlled with respect to phylogenetic factors by using Japanese–Formosan hybrid macaques and excluding potentially integrated traits as much as possible. Moreover, RTLs of the hybrids varied serially from 18.3 to 88.8%. Therefore, tail length associated morphological variation detected in this study retains a great functional quality, while minimizing phylogenetic influence, and is likely close to the general tendency in catarrhines. Second, tail length was estimated quantitatively among catarrhines with an intermediate tail length (15.0 < RTL < 100.0%). In previous studies, these catarrhines were lumped together as a single group (‘reduced tail’ or ‘short’) and intra-group differences were not analyzed (Ankel, 1965, 1972; Ward et al., 1991; Russo and Shapiro, 2011). Results from this study demonstrate that it is possible to estimate tail length with reasonable accuracy in catarrhines in this range (except diminutive-tailed species).
Skeletal morphological parameters that strongly reflect tail length were identified in the sacrum and proximal caudal vertebrae. It might be expected that formulae using both the sacrum and caudal vertebral measurements would perform well. While this is true, reliable formulae (adjusted R2 > 0.60) are also obtained from only the sacral measurements. This is a promising result for studies of fossils, since fossil sacral specimens are not frequently associated with proximal caudal vertebrae. Compared to the sacrum, the number of parameters defined in the proximal caudal vertebrae were fewer, so this is not to say that the proximal caudal vertebrae cannot serve as a good key to estimate tail length (e.g. see model 5), but instead to highlight the predictive value of sacral measurements. Future studies may succeed in finding key features of the proximal caudal vertebrae for tail length estimation (either quantitative or categorical) (see Nakatsukasa et al., 2004). However, it is noted that at least the axial length ratio between the proximal caudal vertebrae (or relative elongation of the Ca3 vertebral body) was a weak predictor of tail length if it is applied to catarrhines with intermediate length of tails (model 13).
Several measurements of the last sacral vertebra (inter-postzygapophyseal breadth (No. 12), sagittal diameter of the caudal articular surface (No. 8), and breadth at the transverse process (No. 11)) are functionally related to tail length (models 1, 2, 8, 9, 10). Inter-postzygapophyseal breadth likely reflects the rigidity for withstanding stresses induced by movements of a tail. This might be parallel to the relative narrowing of the caudal surface of the last sacral vertebra in shorter-tailed and tailless catarrhines (Ward et al., 1991). In macaques, no caudal muscles originate at the transverse process of the last sacral vertebra, but muscle bundles originating from the vertebral column and os coxa converge and pass on the ventral and dorsal sides of the transverse process (Tojima, 2010). The transverse process of the last sacral vertebra appears to serve an important function for maintaining the muscle bundles and tendons passing through on it (Tojima, 2010). Among Macaca species which differ markedly in tail length, the difference of the caudal musculature development with tail length is more marked in the extensors and abductors (MM extensor caudae lateralis, extensor caudae medialis, abductor caudae lateralis, abductor caudae medialis) compared to the flexors (MM flexor caudae longus, flexor caudae brevis, iliocaudalis, pubocaudalis, ischiocaudalis) (Tojima, 2010). In longer-tailed species, many researches clarified that tail extension seemed to have an important function for movement around the arboreal environment. Wilson (1972) and Larson and Stern (2006) described that primate tail played an important role in maintaining balance in an arboreal environment. Some behavioral studies have observed that arboreal primates hold their tails in an extended position during quadrupedal walking (Bernstein et al., 1978; Stevens et al., 2008). In addition, it is found that extension of the tail serves to reorient the body during leaping by counteracting the angular momentum produced by pelvic rotation during takeoff from horizontal substrates (Emerson, 1985; Günther et al., 1991; Essner, 2002). Dunbar and Badam (2000) reported that long tail of Macaca radiata have important prehensile and sensory functions in arboreal locomotion and posture. The transverse process of the last sacral vertebra seems to support these caudal muscles. In shorter-tailed species, some of the caudal muscles become tendinous before they pass by the last sacral vertebra. Russo and Shapiro (2011) also suggested that a lateral expansion of the transverse process may be related to an increase in the leverage of the proximal tail’s abductor musculature, and thus positively correlated with tail length.
The sagittal diameter of the caudal articular surface of the sacrum is probably related to a reduction in tail mobility (Russo and Shapiro, 2011). These authors noted that the shape of this articular surface becomes more circular in longer-tailed cercopithecoids compared to reduced-tailed catarrhines and suggested that a rounder morphology is related to a wide range of tail movements. As indicated, MLDV can predict RTL when it is used with the other indices in multiple regression formulae. However, the results of this study demonstrate that the sagittal diameter of the last sacral vertebra reflected RTL more strongly compared to the ratio of transverse vs. sagittal diameter. Sagittal diameter of the articular surface causes changes in its shape from elliptical to circular, meaning that the results obtained in this study are consistent with previous findings (Russo and Shapiro, 2011). This measurement might also reflect the difference of tail maintenance. In a troop of rhesus macaques, some individuals have been observed to walk with their tails erect (Altmann, 1962; Ojha, 1974). This behavior has been observed in other macaque species with short and long tails (wild Macaca fuscata, Itani, 1954; wild Macaca fascicularis, author’s personal observation, 2010). The frequency, total time, and the reason of this behavior are not fully understood. If this behavior is common in short-or long-tailed catarrhines, however, the last sacral vertebra will receive larger moments of force because of the tail weight and the movements against gravity in the sagittal direction in longer-tailed species. The relative diameter of sagittal direction is considered to become larger against the moment of force.
In order to test the utility of the formulae, this study calculated Ankel’s (1965) sacral index in the examined hybrid macaques. Among the 89 hybrid macaque individuals, there were 67 for which all four of the measurements (No. 1, 2, 5, 6 in Figure 1) could be obtained. Fifty-six out of 67 individuals fit into the ‘reduced tail’ category (roughly corresponding to 19.7–44.4% in RTL). Since the RTLs of the hybrid macaques ranged from 18.8 to 88.8%, this result supports the utility of the sacral index as a predictor of RTL. However, the correlation between the sacral index and RTLs in this sample was low (r = 0.32). Thus, while this index would be applicable for broad comparisons, alternative methods such as those devised in this study are necessary for finer-grained studies in ‘short’-tailed catarrhines.
Ward et al. (1991) devised a tapering index to estimate tail length. It is not possible to compare this method with the one reported here. The method reported by Ward et al. (1991) requires measurements of the cranial and caudal articular surfaces of the last sacral vertebra. The materials used by this study were all adults and the last two sacral vertebrae are completely fused. Technically, it is almost impossible to collect these measurements accurately from completely fused sacra because the border between the cranial end of the last sacral vertebra and the caudal end of the forward sacral vertebra is often unclear.
Prediction models obtained by this study performed well for various catarrhines whose RTLs were within the range of the reference hybrid macaques and result in the following important observations:
However, this study also revealed the difficulty in predicting very long or very short tail lengths using only the formulae for catarrhines with intermediate tail lengths. The formulae tend to underestimate RTLs in long-tailed catarrhines and to overestimate RTLs in diminutive-tailed catarrhines. Perhaps other morphological traits (or different combination of traits) that were not used in this study are applicable for tail length variation in those categories. In this study, standard osteometric techniques were employed. However, future studies might be encouraged to adopt finer-grained methods (e.g. 3-D digitizer or photogrammetry) or more detailed analyses of the relationships between the morphology of the proximal caudal vertebrae and tail length.
The author is grateful to Yuzuru Hamada of the Primate Research Institute, Kyoto University for his permission to study the hybrid macaques and his helpful suggestions and comments, and to Paula Jenkins and Louise Tomsett of the Natural History Museum, London for their help in studying the skeletal collections under their care. Masato Nakatsukasa of the Laboratory of Physical Anthropology, Kyoto University and Toshio Mouri of the Primate Research Institute gave me important advice and suggestions for the revision of this paper. Yoshi Kawamoto of the Primate Research Institute, Kyoto University gave me important advice and information about the genetic features of these hybrids. Wataru Yano gave me important advice and information, particularly on the statistical analyses. This study was supported by the Cooperation Research Program of the Primate Research Institute, Kyoto University, and by the Global COE Program A06 of Kyoto University.
