Reconstructing the Locomotor Repertoire of Protopithecus brasiliensis. I. Body Size


  • Lauren B. Halenar

    Corresponding author
    1. The Graduate Center, Department of Anthropology, City University of New York, New York Consortium in Evolutionary Primatology (NYCEP), New York, New York
    • The Graduate Center, Department of Anthropology, City University of New York, 365 Fifth Avenue, New York, NY 10016
    Search for more papers by this author
    • Tel: (201) 321-3986


An accurate body size estimate is essential for reconstructing and interpreting many aspects of the paleobiology of an extinct taxon. With this in mind, the purpose of this study is two-fold: first, to create statistically robust predictive regression equations for body mass, total body length, and head and body length from postcranial elements using a platyrrhine reference sample, data that do not exist elsewhere in the literature; and, second, to apply those regression equations to the “giant” subfossil platyrrhine Protopithecus brasiliensis, a little-studied taxon represented by a nearly complete skeleton. Building on results of previous work with other primate groups, different skeletal elements, subgroups of the reference sample, and regression models lead to different body size estimates with different standard errors and prediction errors. However, relatively tight clusters of estimates around 20 kg, total length of 1,675 mm, and head and body length of 710 mm are obtained, placing the fossil in the size range of a large male baboon. While not quite as large as the original 25 kg body mass estimate for the fossil, this new estimate is still approximately 150% larger than the largest living New World monkey. Confirmation of its place in a large-bodied size class of platyrrhines has a profound effect on reconstructing the locomotor repertoire of Protopithecus and the evolutionary trajectory of the alouattin lineage. Anat Rec, , 2011. © 2011 Wiley Periodicals, Inc.

Remains of Protopithecusbrasiliensis were first discovered in 1836 by the Danish naturalist Peter Wilhelm Lund. A left proximal femur and right distal humerus were found in the Lagoa Santa cave system in Minas Gerais, Brazil, and despite their large size and robusticity, Lund recognized them as belonging to a New World monkey (Lund,1838). These fossils were the first to be correctly recognized as a primate and were included in Darwin's (1859) discussion of South American fossil monkeys in On the Origin of Species. Together with a nearly complete skeleton found 150 years later in the separate Toca da Boa Vista cave system in the neighboring state of Bahia, Lund's discovery makes up one-half of the known Pleistocene primate “megafauna” of Brazil. The other half is the subfossil skeleton of Caipora bambuiorum, also nearly complete. These two specimens include largely undescribed crania; mandibles; mostly complete upper and lower dentitions; cervical, thoracic, and caudal vertebrae; scapular and innominate fragments; complete upper and lower limb bones; carpals, tarsals, metacarpals, metatarsals, and phalanges. The reconstructed feet are the only complete cheiridia in the platyrrhine fossil record (Cartelle and Hartwig,1996; Hartwig and Cartelle,1996). The original publications describing the fossil material suggested that Caipora was morphologically comparable to a large Ateles (Cartelle and Hartwig,1996). Protopithecus, on the other hand, showed a more complicated mosaic pattern with frugivorous teeth in an otherwise Alouatta-like skull on top of a postcranial skeleton adapted for spider monkey-like suspensory locomotion (Hartwig and Cartelle,1996). Only limited additional analysis has been done on either of these fossils (Heymann,1998; Jones,2008; Halenar,2009,2010; Rosenberger et al., in review).

One of the most intriguing differences between Protopithecus and its living ateline relatives is the large size and robusticity of its long bones. Using a regression equation for estimating body mass based on femoral head volume in catarrhine primates (Ruff,1990), the original individual discovered by Lund in the 1800s was estimated to have a body weight of 22.7–24.1 kg (Hartwig,1995). At that time, Hartwig recognized that using catarrhines as a regression model was a fallback approach since comparable data and regression equations for living atelines did not exist. He also noted that Protopithecus is so much larger than modern atelines that it would fall outside of the range of any regression plot generated by a more appropriate platyrrhine reference sample (Hartwig,1995). Despite these caveats, Hartwig and Cartelle (1996) used Ruff's (1990) femoral head equation when the more complete Protopithecus skeleton was discovered; the new material was estimated at 25 kg, a value approximately twice that of the largest living platyrrhines. Prediction error, measures of estimate consistency and accuracy, and standard error were not provided in either publication.

A vast literature exists on the topic of body mass estimation in fossil human and non-human primates. The problems encountered by Hartwig (1995) in choosing an appropriate reference sample and dealing with fossils of large body size have been discussed at length and solved to varying degrees of satisfaction in more recent publications involving other taxa. For example, in order to estimate the body size of the truly giant subfossil lemurs, extrapolation into body size ranges outside of anything seen in living primates was required (Godfrey et al.,1995; Jungers et al.,2002,2008). The best estimates of body size were produced using humeral and femoral midshaft circumferences against a primate reference sample as well as a mammalian reference sample which represented a wide range of body sizes and positional behaviors (Godfrey et al.,1995). Similarly, Delson et al. (2000) discussed the problem of investigating very large fossils and concluded that it is the estimates for these taxa that come with the most error, regardless of the alternative materials and methods chosen. Other studies estimating the body weights of fossil primates (e.g., Dagosto and Terranova,1992; Aiello and Wood,1994) have shown that different regression models and reference samples produce not only different estimates of body mass but also different ranges and confidence intervals.

From a phylogenetic perspective, the appropriate reference sample for estimating body size of Protopithecus would include the four genera of ateline primates (Alouatta, Lagothrix, Ateles, and Brachyteles) suggested to be its closest living relatives (Hartwig and Cartelle,1996). Platyrrhine-only craniodental regression equations are now available from two studies (Meldrum and Kay,1997; Sears et al.,2008), although their value in this context remains somewhat limited. The lower molars of Protopithecus are not preserved and canine size and temporal line strength suggest the cranium belongs to a male (Hartwig and Cartelle,1996), so the regression equations of Meldrum and Kay (1997) based on female platyrrhine molar dimensions are not directly useful here. The equations of Sears et al. (2008) come with poor R2 values and relatively high error sum of squares (ESS) values; of the 80 craniodental variables investigated, the highest R2 value is 0.636 and the majority are below 0.5. It could be that there are so many variable cranial morphologies in their reference sample (i.e., callitrichine skulls are very different from Alouatta skulls), that those differences are making the scatter around the regression line too diffuse for accurate predictions. There are several other reasons for caution when applying the Sears et al. (2008) equations to other fossil taxa. For example, the equations produced in that study are not tested for accuracy on individuals of known body mass so there is no evidence for certain variables over- or under-estimating body mass as the authors discuss.

Despite these issues, the Sears et al. (2008) equations were applied to Protopithecus, and not surprisingly, the results beg further investigation. The fossil mandible is incomplete so data are not available for implementing mandibular length, their best predictor variable (R2 = 0.636, ESS = 0.117). Ranking second and third, bizygomatic width (R2 = 0.626, ESS = 0.166) and skull length (R2 = 0.537, ESS = 0.196) of Protopithecus give a mass estimate of approximately 12 kg, or about the size of a large male Alouatta pigra. This is much lower than either of the original estimates (Hartwig,1995; Hartwig and Cartelle,1996) and seems an unlikely value both because of the low R2 values of the equations and the fact that the fossil bones themselves are much larger than the bones of an extant 12 kg platyrrhine. For example, total skull length of Brachyteles (usually cited as the largest ateline at around 12 kg) is 115 mm while for Protopithecus it is 150 mm; in the postcranial skeleton, femoral length of Brachyteles is 202 mm and in Protopithecus is 237 mm (Hartwig and Cartelle,1996). Attempting to use these newly published equations again serves to highlight the fact that the choice of reference group and predictor variable will affect the final estimate. The development of more accurate platyrrhine-based regression models for application to the fossil record is a necessary step forward.

Body size of an individual or species is related to many other aspects of its biology and behavior (for recent review see Bernstein (2010)). For fossil taxa, reconstructing diet and/or locomotion are most often the focus of studies attempting to estimate body size (e.g., Fleagle,1978,1999; Jungers,1984; Kay,1985). “Kay's Threshold” of 500 g has long been used to separate small-bodied insectivores from larger-bodied folivores (Kay,1975; Gingerich,1981). Similarly, a limit of 10–12 kg has been suggested for the metabolically efficient use of ricochetal brachiation (Preuschoft and Demes,1985; see below). This article provides a new set of regression equations for predicting body weight and linear dimensions of body size based on a large series of New World monkey species and specimens. Focus is placed on the postcranial skeleton as a source of size information, especially joint surfaces that have been shown to be valuable in other studies that estimate fossil body mass in various primate groups (e.g., Jungers,1990; Godfrey et al.,1995; Delson et al.,2000; Ruff,2003). These new estimates are then used as a basis for preliminary discussion of possible locomotor behaviors used by Protopithecus and how the combination of body size reduction, hyoid size increase, and decrease in acrobatic locomotion influenced the evolution of extant Alouatta.


A comparative sample across primates was used consisting of adult non-zoo, non-pathological individuals from collections at the American Museum of Natural History (AMNH) in New York, the National Museum of Natural History (NMNH) in Washington, DC, and the Museu Nacional do Rio de Janeiro, Brazil (MN) (Table 1). In choosing the sample, several things were considered. Conroy (1987) strongly advises using a restricted taxonomic sample to create predictive equations and also strongly advises against extrapolating outside the size range of that comparative sample. Bracketing the fossil of interest between larger and smaller relatives would seem to adhere to these suggestions; however, Delson et al. (2000) state that using a comparative sample that includes one very large and one very small species leads to a regression line anchored by those points and less influenced by the species in the middle of the range. (The problem of extrapolation will be discussed further below with regard to choosing a specific regression model.) Larger-bodied Old World monkeys and apes were included in the sample as one way to account for a fossil that is comparatively larger than its closest platyrrhine relatives. Callitrichines were not included as they are much smaller and exhibit subfamily-specific postcranial allometries that differ from the patterns seen in the rest of the platyrrhines, potentially due to a dwarfing event in their evolutionary history (Ford,1980; Ford and Corrucini,1985) and/or their claw-like nails which change their biomechanical relationship to arboreal substrates during locomotion (Jungers,1985).

Table 1. Taxa involved in calculating regression equations
    TOTL (mm)TrL (mm)Body weight (g)
  • a

    N = number of individuals with associated total length (TOTL) and head and body length (TrL) measurements in the AMNH and NMNH catalogs.

  • b

    Average body weights from Smith and Jungers (1997), Delson et al. (2000), and Di Fiore and Campbell (2007); see those sources for sample composition and ranges.

BrachytelesarachnoidesM     11,139
F     10,111
PitheciamonachusM     2,610
F     2,110
F     2,770
CebusalbifronsM1880 400 3,180
F1865 385 2,520
F11,390 575 9,100
F     5,700
brunnescensM1565 530  
fascicularisM11,092 500 5,360
F1925 390 3,590
thibetanaM     15,200
F     9,500
tonkeanaM     14,900
F     9,000
mulattaM     7,710
F     5,370
sylvanusM     14,530
F     10,140
F     9,820
PresbytispileatusM11,340 580 12,000
F11,240 480 9,860
cristatusM11,248 546 6,610
F11,222 532 5,760
johniM11,523 711 12,000
F11,530 616 11,200
melalophosM11,136 590 6,590
F11,297 587 6,470
obscurusM11,094 389 7,900
F11,230 480 6,260
F   480 6,137
frontatusM11,120 490480–5005,560
F31,1831,160–1200  5,670
potenzianiM     6,170
F     6,400
SemnopithecusentellusM     14,533
F     10,533
PantroglodytesM3840825–925  49,567
F2680387–830  40,367
SymphalangussyndactylusM1612   11,900
F2564548–580  10,700
PropithecusverreauxiM     3,475
F     3,615
IndriindriM     5,830
F     6,840

A Microscribe 3DX digitizer was used to collect a set of x, y, z coordinate points that define various aspects of the long bones, including: the distal humerus; proximal ulna; complete, proximal, and distal femur; complete, proximal, and distal tibia; and the talus (Fig. 1; Table 2). These elements are all well-preserved in Protopithecus on at least one side of the body. The independent variable used here in estimating body size by linear regression is the centroid size (the square root of the sum of squared distances of a set of landmarks from their centroid; Bookstein,1991) of these skeletal elements. Centroid size for each element in the sample was calculated using the morphologika2 v.2.5 software package (O'Higgins and Jones,2006) (Table 3). Centroid size will give a better estimate of overall joint size than simple areas calculated from 2D length and breadth measurements. Overall joint size is important to capture accurately as the joint surfaces are the weight-bearing part of the limb bones that are more closely related to “body size” than full bone lengths which vary more widely under multiple selective factors in mammals (Ruff,1987,1990,2002; Scott,1990; Anyonge,1993). Postcranial remains, in general, have been shown to be useful in similar previous studies, yielding more statistically robust equations than teeth and cranial variables (e.g., Ruff et al,1989; Jungers,1990; Rafferty et al.,1995; Ruff,2002).

Figure 1.

3D landmarks used to calculate the centroid size of all postcranial elements. Photographs are of Protopithecus. A = Anterior (left) and posterior (right) distal humerus, B = proximal ulna, C = anterior (left) and posterior (right) femur, D = proximal (left) and distal (right) left tibia, E = dorsal (top left), plantar (top right), medial (bottom left), and lateral (bottom right) talus. Scale bars = 1 cm. See Table 2 for landmark definitions.

Table 2. Anatomical landmark definitions
Distal humerusAnterior
A1Most lateral point
A2Most medial point
A3Most lateral point on capitulum
A4Most medial point on capitulum
A5Most superiolateral point on trochlea (excluding capitulum)
A6Most inferiomedial point on trochlea (excluding capitulum)
A7Most superior point on medial epicondyle
A8Most inferior point on medial epicondyle
A9Most superior point on lateral epicondyle
A10Most inferior point on lateral epicondyle
P1Most superior point of olecranon fossa
P2Most inferiomedial point of olecranon fossa
P3Most inferiolateral point of olecranon fossa
P4Deepest point of olecrannon fossa
P5Most medial point on trochlea
P6Most lateral point on trochlea
P7Tip of the medial epicondyle
Proximal ulnaU1Most proximal point on the olecranon
U2Most medial point on the maximum constriction of the olecranon
U3Most lateral point on the maximum constriction of the olecranon
U4Most posterior point on the olecranon
U5Most anterio-medial point of the olecranon
U6Most anterio-lateral point of the olecranon
U7Most medial point on the wing of the proximal articular facet
U8Most lateral point on the wing of the proximal articular facet
U9Most anterior point on the proximal border of the proximal articular facet
U10Most disto-medial point of the proximal articular facet
U11Most disto-lateral point of the proximal articular facet
U12Deepest point in the midline of the trochlear notch
U13Most posterio-lateral point of the distal articular facet
U14Most anterior point of the distal articular facet
U15Most anterior point of the radial facet
U16Most posterior point of the radial facet
U17Most proximal point of the radial facet
U18Most distal point of the radial facet
U19Deepest point in the radial facet
F1Middle of fovea capitus
F2Most proximal point on the femoral head
F3Most proximal point on the facet margin around the femoral head
F4Most distal point of the facet margin around the femoral head
F5Most anterior point of the facet margin around the femoral head
F6Most posterior point of the facet margin around the femoral head
F7Maximum point of constriction on ridge running from lesser trochanter to the femoral head
F8Deepest point of the proximal neck
F9Middle of the trochenteric fossa
F10Tip of greater trochanter
F11Most lateral point of greater trochanter
F12Most proximoanterior point of the greater tubercle
F13Tip of lesser trochanter
F14Origin of pectineal line
F15Most medial point
F16Most lateral point
F17Most proximomedial point of the patellar articular surface
F18Most proximolateral point of the patellar articular surface
F19Most proximal point of the patellar articular surface
F20Most distal point of the patellar articular surface
F21Most distal point of the medial condyle
F22Most distal point of the lateral condyle
F23Most posterior point of the medial condyle
F24Most posterior point of the lateral condyle
F25Most proximomedial point of the posterior aspect of the medial condyle
F26Most proximolateral point of the posterior aspect of the lateral condyle
F27Most proximolateral point of the posterior aspect of the medial condyle
F28Most proximomedial point of the posterior aspect of the lateral condyle
F29Most posterio-medial point of the groove (notch)
F30Most posterior point of the groove (notch)
Tib1Most anterior point on the medial tibial condyle on outer edge of articular surface
Tib2Most medial point on the medial tidial condyle
Tib3Most posterior point on the medial tibial condyle
Tib4Most lateral point on the medial tibial condyle
Tib5Most anterior point on the lateral tibial condyle
Tib6Most medial point on the lateral tidial condyle
Tib7Most posterior point on the lateral tibial condyle
Tib8Most lateral point on the lateral tidial condyle
Tib9Most anterior point on the tibial tuberosity
Tib10Most anterior point on the talar facet
Tib11Most medial point on the talar facet
Tib12Most posterior point on the talar facet
Tib13Midpoint of the lateral edge of the talar facet
Tib14Midpoint of the groove between the talar facet and the medial malleolus
Tib15Most anterior point of the groove between the talar facet and the medial malleolus
Tib16Most posterior point of the groove between the talar facet and the medial malleolus
Tib17Most medial point of the medial malleolus
Tib18Distal tip of the medial malleolus
Tib19Most lateral point on the fibular facet
TalusDorsal surface
Tal1Most distal point of the trochlear groove
Tal2Most distal point of contact between the medial malleolar facet and the trochlear surface
Tal3Most dorsal point on the medial facet margin
Tal4Most proximal point of contact between the medial malleolar facet and the trochlear surface
Tal5Most proximal point of the trochlear groove
Tal6Most dorsal point on the lateral facet margin
Tal7Most dorsal point on the trochlear groove
Tal8Most distal point on medial malleolar facet
Tal9Most distal point on lateral malleolar facet
Tal10Most plantar point on lateral malleolar facet
Tal11Deepest (most medial) point on lateral malleolar facet
Tal12Most dorsal point of the navicular facet
Tal13Most medial point of the navicular facet
Tal14Most lateral point of the navicular facet
Tal15Most proximal point of contact between the lateral malleolar facet and the trochlear surface
Tal16Most distal point of contact between the lateral malleolar facet and the trochlear surface
Tal17Most plantar point on medial malleolar facet
Tal18Most distal point of the navicular facet
Plantar surface
Tal19Most disto-lateral point of the posterior calcaneal facet
Tal20Lateral end of the narrowest distance across the posterior calcaneal facet
Tal21Most proximo-lateral point of the posterior calcaneal facet
Tal22Most proximal point of the posterior calcaneal facet
Tal23Most proximo-medial point of the posterior calcaneal facet
Tal24Medial end of the narrowest distance across the posterior calcaneal facet
Tal25Most disto-medial point of the posterior calcaneal facet
Tal26Most distal point of the posterior calcaneal facet
Tal27Deepest (most dorsal) point of the posterior calcaneal facet
Tal28Most proximal point of the anterior calcaneal facet
Tal29Most medial point of the anterior calcaneal facet
Tal30Most distal point of the anterior calcaneal facet
Tal31Most lateral point of the anterior calcaneal facet
Tal32Most plantar point on the navicular facet
Tal33Most lateral point of contact between the navicular facet and the distal calcaneal facet
Table 3. Species mean centroid sizes
   Distal humerusProximal ulnaFull femurProximal femurDistal femur
F139.6 139.8 1400 142.2 143.2 
guaribaM0  0  0  0  0  
pigraM248.247.3–49.2150.2 1422 152.2 148.2 
F142.1 0  0  0  0  
AtelesbelzebuthM148.8 141.7 1489 157.4 158.6 
F144.5 137.1 1496 150.1 152.4 
geoffroyiM0  0  0  0  0  
paniscusM149.3 139.7 0  0  0  
F0  0  0  0  0  
F0  0  0  0  0  
F140.3 138.6 1454 147.4 147.8 
F142.2 237.033.0–41.02403388–418242.839.2–46.4242.739.9–45.5
brunnescensM147.1 146.0 1437 153.4 157.9 
nigerM0  0  0  0  0  
F135.3 135.9 1362 142.7 143.4 
F0  0  0  0  0  
tonkeanaM151.2 148.3 1431 159.6 155.2 
F0  0  0  0  0  
F0  0  0  0  0  
sylvanusM152.4 149.9 1491 156.7 156.9 
F142.8 141.6 0  0  0  
F145.8 141.5 1531 153.7 153.0 
PresbytispileatusM143.9 143.9 1484 151.3 151.1 
F142.5 140.7 0  0  0  
johniM145.2 144.2 1565 154.6 156.2 
F136.4 134.9 1502 143.1 145.4 
F137.1 135.6 1420 142.7 143.8 
potenzianiM138.7 134.2 1458 142.4 144.3 
F0  0  0  0  0  
SemnopithecusentellusM144.1 146.4 1535 152.7 154.7 
F0  0  0  0  0  
SymphalangussyndactylusM155.6 143.8 1545 160.3 158.9 
F252.551.9–53.0239.939.2–40.81524 153.5 151.6 
IndriindriM138.1 130.9 1585 148.8 150.8 
F141.9 133.8 1629 151.0 153.2 
Total n  328  308  320  320  320  
   Full tibiaProximal tibiaDistal tibiaTalus
F1294 130.1 121.8 148.6 
 guaribaM   0  0  0  
 F   226.025.5–26.4217.917.7–18.1240.240.1–40.4
 pigraM1325 134.6 122.3 151.2 
 F2254253–2540  0  0  
AtelesbelzebuthM1398 140.1 129.7 0  
F1393 136.6 127.2 159.4 
geoffroyiM0  0  0  0  
paniscusM0  0  0  0  
F0  0  0  0  
F0  0  0  0  
F1338 138.7 122.1 152.9 
F1364 132.4 119.8 247.946.0–49.7
F0  0  0  148.5 
brunnescensM1344 140.9 124.8 0  
F1245 123.1 114.3 136.1 
nigerM0  0  0  0  
F1282 131.3 120.5 0  
thibetanaM1322 142.6 124.7 458.855.6–60.3
F0  0  0  0  
tonkeanaM1345 139.5 123.8 154.7 
F0  0  0  0  
mulattaM0  0  0  153.5 
F0  0  0  0  
sylvanusM0  0  0  159.1 
F0  0  0  152.6 
F1401 138.1 124.8 162.4 
F1398 146.0 128.9 365.463.7–67.5
PresbytispileatusM1363 141.8 124.8 0  
F1326 136.7 120.6 0  
F1348 134.1 119.4 246.144.7–47.6
johniM1432 143.6 125.7 157.6 
melalophosM1375 134.5 121.1 250.949.0–52.7
F1377 133.4 120.0 148.3 
obscususM1310 131.8 118.5 250.546.2–54.8
F1321 134.0 118.7 0  
frontatusM0  0  0  248.947.3–50.5
F0  0  0  249.347.6–51.0
potenzianiM0  0  0  146.4 
F0  0  0  0  
SemnopithecusentellusM0  0  0  157.1 
F0  0  0  0  
F1369 140.0 122.6 150.3 
IndriIndriM1420 137.8 123.3 0  
F1457 138.1 122.6 155.3 
Total n  245  245  245  209  

Three aspects of “body size” were predicted for the fossil: body weight (kg); total length (TOTL; mm) which includes the length of the tail (TAILL); and trunk length, head, and body (TrL; mm) which includes the length of the skull and trunk (TOTL = TAILL + TrL; Ford and Corrucini,1985). The two length measurements are important in this case because they can be used to calculate a tail to trunk length ratio, which can in turn be used to begin to answer questions about whether Protopithecus had a prehensile tail like its extant ateline relatives. TOTL and TrL measurements were taken from the AMNH and NMNH catalogs for all specimens in the comparative sample (N = 345; Table 1). Only about a third of those specimens had associated body weights (N = 102; Table 4), so male and female species mean body weights were instead taken from the literature to create the body weight predictive equations (Table 1). Using body weight, TOTL, and TrL will mitigate difficulties inherent in using weight alone as proxy for body size, as individual body weight can be expected to vary seasonally due to food availability (e.g., Terborgh,1983), based on female reproductive condition, and with regard to recording techniques in the field or in the lab. TOTL and TrL are not without problems, as differing allometries of the tail, skull, and trunk are at work in different taxonomic subgroups of the sample (see below). All three values are reported to give the most comprehensive estimate possible for the overall size of the extinct animal.

Table 4. Individuals with associated body weights
GenusSpeciesSpecimen no.Body weight (g)GenusSpeciesSpecimen no.Body weight (g)
  518226F5,000  33718M790
  518227F5,500  503709M840
 caraya211605M8,300  337316M660
  518232M6,500  337317F670
  518236M7,300  337315F690
  518241M7,200 apella211578M3,900
  518230M5,000  452616M10,600
  211498M9,300 guereza452619M11,400
  211504M8,300  452621M7,300
  518233F3,500  452620M6,300
  518237F3,400  452625M9,900
  518238F4,800  452629M10,200
  518240F4,600  452628M9,200
  518243F5,400  452635M11,400
  211497F5200  452643M9,000
  211501F4,900  452622M10,600
  518231F3,500  452641F9,000
 guariba518244M7,200  452624F7,400
  518247M5,300  452636F7,500
  518248M4,600  452642F6,400
  518254M6,200  452634F9,600
  518251M6,200  452502M8,500
  518246F4,600  452498F6,900
  518252F5,000 mulatta173813M12,700
  518253F3,700 sylvanus476786F9,000
  338106M8,600  384229M25,000
  338107M7,700  452509M20,000
  338104F6,600  384228F18,000
 seniculus169608M6,500  384227F17,000
  169638M6,900 rubicundus151826M6,100
  169609M7,000 frontatus151825M5,600
  169619M6,800  151823F5,900
  169634M6,900  151820F5,700
  169639M5,000  154362F5,300
  169777M7,000 potenziani121673M7,100

Separate data points for male and female species mean values were used to develop the predictive regression equations. For those predicting body weight, three different subsets of the full comparative sample were used: atelines only, platyrrhines only, and the full sample. For the length measurements, only the platyrrhine sample was used in the final analysis because it became evident in preliminary studies that different tail proportions are exhibited by platyrrhines, arboreal cercopithecoids, terrestrial cercopithecoids, and hominoids with no tails. These four groups can be distinguished on the basis of differing slopes when the regression is shown graphically (Fig. 2).

Figure 2.

Ln centroid size of the proximal ulna plotted against ln TOTL of all individuals in the sample with this associated measure as an exploratory analysis of allometric patterns within the sample. Three distinct groups are bound by convex hulls: platyrrhines (blue), cercopithecoids (green), and hominoids (red). The cercopithecoid group is further divided into arboreal (top) and terrestrial clusters (bottom). This grouping pattern reflects different tail proportions and was seen in every element when plotted against TOTL. For this reason, only the platyrrhine reference sample was used when predicting TOTL and TrL of the fossils.

All means were log-transformed using the natural logarithm. Because of the bias introduced by de-transforming the solution to each equation to get the predicted body size estimates, a correction factor needs to be applied (Smith,1993). Previous publications have chosen to use the Quasi-Maximum Likelihood Estimator (QMLE = exp [MSE/2] where MSE is the residual mean square error of the regression equation; Delson et al.,2000; Jungers et al.,2008) and that procedure is followed here as well. It should be noted that the QMLE has been found to overcompensate for bias (Ruff,2003) and is only one relatively small factor to consider in evaluating regression equations (Smith,1993). Because it is directly related to the standard error of the equation, the magnitude of the correction factor will depend on the strength of the relationship between the two variables used to produce the equation; the better the fit, the smaller the correction (Dagosto and Terranova,1992). The QMLE and other de-transformation bias correction factors were developed with reference to ordinary least squares regression so they were only applied to the estimates from those equations (see below).

Univariate regression is used with each element alone (see Ruff (2003) for discussion of why it is not necessarily helpful to use multivariate step-wise regressions with several elements combined, especially when the locomotor pattern of the fossil taxon of interest is an unknown). As the literature on the topic of body size estimation is equivocal, both ordinary least squares (OLS, Model I) and reduced major axis (RMA, Model II) regression equations were produced. OLS regression should be used in body size prediction because there is a clear dependent variable (body size) that is conditional upon an independent variable (skeletal estimator) (McArdle,2003; Smith,2009 and references within). However, OLS regression does not deal well with cases of extrapolation (Ricker,1973; Conroy,1987). RMA regression has been used previously in cases where the fossil under investigation is outside the range of body size reflected in the comparative sample used to create the model (Jungers,1988; Aiello,1992; Ruff,1998) and reporting estimates based on multiple models has been suggested as a necessary step in cases of extrapolation (Delson et al.,2000). Here, both OLS and RMA were used for each element against the smaller-bodied ateline and platyrrhine samples, and only OLS was used for the full sample which includes the larger-bodied Old World monkeys and apes.

Comparing equations derived from different reference samples and regression models to determine the “best” equation was done by calculating the standard error of the estimate (SEE) and mean prediction error (MPE) for each equation (Dagosto and Terranova,1992; Delson et al.,2000). These are important values to consider in ranking prediction equations; just because an equation has a high R2 does not mean the equation is actually a good predictor. This issue was discussed by Conroy (1987) who had very high R2 values for all equations (0.91–0.99), but when the regression lines were shown graphically, many points fell outside of their 95% confidence intervals. The SEE is a test of consistency or dispersion around the regression line while the MPE is a test of how accurate the equation is at predicting the size of individuals with known dimensions. The equation for calculating %SEE for log-transformed data used by Ruff (2003) is used here as well:

equation image

The equation for calculating prediction error is:

equation image

The MPE is then calculated for each variable as the mean of the absolute values of the prediction errors summed over all taxa (Delson et al.,2000). The MPE should give the same ranking of predictor variables as the SEE (Ruff,2003). All individuals in the sample had associated TOTL and TrL measurements taken from the museum catalogs with which to calculate prediction errors. The subset of individuals that also had associated weights recorded in the museum catalogs were used to calculate MPE for the body weight estimation equations (Table 4).


A relatively wide range of body size estimates were recovered for Protopithecus: 12–35 kg; 1,479–1,887 mm TOTL; and 613–831 mm TrL. These ranges reflect the use of different skeletal elements, reference samples, and regression models (Tables 5 and 6). The low values of these ranges are all larger than the average body weight, TOTL, and TrL for any of the modern platyrrhine taxa in the reference sample (Figs. 3–5). The low TOTL value of 1,479 mm is outside the range of all living platyrrhine species (Fig. 4), but the low TrL value of 613 mm is inside the range of values for some of the larger species of Alouatta (Fig. 5). The extreme highs and lows should most likely be disregarded as they come from skeletal elements that give very low or very high estimates in every configuration of the data. Choosing the “best” equation and “best” estimated value from among the results is a moot point, given the range of values reported here. However, the equation with the combined highest R2 (= 0.98), lowest %SEE (= 11.0), MPE (= 14.7), and QMLE (= 1.005) is that using the distal humerus with a platyrrhine-only reference sample; this gives an estimate of 28 kg for the Cartelle specimen and 24 kg for the Lund specimen. Condensing all of the values reported in Tables 5 and 6 into an average, disregarding the obvious extreme outliers in estimate and confidence statistics, gives a body weight of approximately 23 kg, 1,675 mm TOTL, and 710 mm TrL. As an alternative to compiling an average value, a histogram of all body weight predictions shows 19, 21, and 25 kg as the most frequent estimates with a reasonable range from 17 to 29 kg (Fig. 6). These modal estimates are predicted by equations in all three permutations of the reference sample (Table 5). Out of all of the species in the comparative sample, the predicted values for Protopithecus are most comparable to those of a large male Papio or Nasalis (Figs. 3–5).

Table 5. Body weight estimates for Protopithecus brasiliensis
  Body weighta (kg)SlopeInterceptR2%SEEMPEQMLE
  • a

    For all paired estimates, the first is for the Cartelle specimen and the second is for the Lund specimen.

Ateline reference sample
Distal humerusOLS25, 222.6641−1.20070.8511.813.21.006
RMA28, 252.8962−2.0768  14.7 
Proximal ulnaOLS131.9111.81120.347.516.11.028
RMA213.2599−3.167  17.9 
Full femurOLS121.8492−2.31290.6519.415.91.016
RMA142.2921−4.9879  19.8 
Proximal femurOLS19, 191.93611.4740.8014.312.91.009
RMA21, 212.16280.60993  14.0 
Distal femurOLS151.85521.7610.6319.921.91.016
RMA182.3307−0.05746  24.5 
Full tibiaOLS111.2791.50160.519.718.61.019
RMA141.7864−1.4247  19.5 
Proximal tibiaOLS172.0281.78820.6818.619.31.015
RMA202.46670.25931  20.0 
Distal tibiaOLS121.47754.28570.5422.519.11.021
RMA152.01222.6322  20.5 
RMA212.6052−1.3268  22.3 
Platyrrhine reference sample
Distal humerusOLS28, 242.8741−2.00230.9811.014.71.005
RMA29, 252.9099−2.1322  15.1 
Proximal ulnaOLS182.8626−1.72730.9023.616.71.023
RMA193.013−2.2609  17.1 
Full femurOLS173.1683−10.3590.8728.428.91.032
RMA193.4051−11.761  31.0 
Proximal femurOLS32, 312.9936−2.59290.9221.922.71.020
RMA35, 333.1282−3.0876  22.6 
Distal femurOLS212.8278−1.98310.9122.027.31.020
RMA232.956−2.4544  26.5 
Full tibiaOLS193.0895−9.08290.7441.739.01.063
RMA233.5882−11.909  41.8 
Proximal tibiaOLS273.1941−2.31230.9220.421.91.017
RMA293.3201−2.7355  21.6 
Distal tibiaOLS182.54620.94490.9024.325.41.024
RMA202.68640.53375  23.8 
RMA293.4107−4.4983  24.2 
Full reference sample
Distal humerusOLS28, 252.6351−0.98480.8925.116.81.025
Proximal ulnaOLS182.6011−0.66230.8727.919.91.031
Full femurOLS132.6665−7.4130.7343.927.41.068
Proximal femurOLS26, 252.6714−1.39360.9122.717.71.021
Distal femurOLS192.6081−1.17230.9222.219.21.020
Full tibiaOLS162.8064−7.48770.7147.642.21.075
Proximal tibiaOLS192.6463−0.5120.9124.128.41.023
Distal tibiaOLS212.64840.75350.8926.521.21.027
Table 6. Total length (TOTL) and trunk length (TrL) estimates for Protopithecus brasiliensis
  TOTLa (mm)SlopeInterceptR2%SEEMPEQMLETrL (mm)SlopeInterceptR2%SEEMPEQMLE
  • a

    For all paired estimates, the first is for the Cartelle specimen and the second is for the Lund specimen.

Distal humerusOLS1725, 16560.83793.88450.896.97.41.002709, 6850.71553.51630.7011.48.11.006
RMA1781, 17050.887993.7028  7.7 775, 7430.857752.9998  8.5 
Proximal ulnaOLS14790.82123.9980.877.76.81.0036450.78913.29680.896.56.01.002
RMA15220.882553.7799  7.4 6590.835943.1302  6.3 
Full femurOLS15090.93731.37410.838.57.61.0036230.75451.64820.5713.810.21.008
RMA15691.02590.85124  7.9 6941.00280.18177  9.9 
Proximal femurOLS1809, 17890.89483.6270.877.56.31.003721, 7140.72043.46140.5913.49.61.008
RMA1887, 18650.958883.3918  6.5 831, 8220.937352.6653  9.4 
Distal femurOLS15950.83683.83930.896.66.41.0026500.67143.64050.6013.110.11.008
RMA16370.883623.6676  6.4 7240.863782.9348  9.7 
Full tibiaOLS15410.88911.89470.6612.78.81.0076130.64222.48450.3517.813.01.014
RMA17001.0960.726  8.6 7571.0872−0.029076  11.0 
Proximal tibiaOLS16890.91743.83660.877.67.61.0036820.74033.62410.5913.49.81.008
RMA17560.985323.609  7.9 7750.963192.8776  9.7 
Distal tibiaOLS15230.7534.70690.877.46.61.0026240.59664.35870.5713.610.51.008
RMA15680.805154.5543  6.7 6940.787073.8016  9.9 
RMA17571.03013.0117  7.8 7981.04362.1639  10.0 
Figure 3.

Body weight estimates for Protopithecus compared with the entire reference sample. Bars for extant species are the male average and female average taken from the literature. Bars for the fossil are the low estimate (full tibia OLS ateline sample), low average (OLS ateline sample), high average (RMA platyrrhine sample), and high estimate (Cartelle proximal femur RMA platyrrhine sample). The low estimate is larger than any extant platyrrhine species to its left but all estimates are in the range of the catarrhine species to its right.

Figure 4.

TOTL estimates for Protopithecus compared with the entire reference sample. Where all values were available in the museum catalogs, bars for the extant sample represent low value, female average, male average, and high value for each species. Bars for the fossil represent low estimate, average, and high estimate. As for body size, the low estimate is longer than any living platyrrhine to its left but in the range of the catarrhines to its right. However, the average length estimated for the fossil is longer than any extant primate species in the sample, perhaps reflecting its relatively long tail.

Figure 5.

Trunk length (TrL) estimates for Protopithecus compared to the entire reference sample. Where all values were available in the museum catalogs, bars for the extant sample represent low value, female average, male average, and high value for each species. Bars for the fossil represent low estimate, average, and high estimate. The low and average estimates for the fossil are within the range of all extant primates in the sample, due to the negative allometry of trunk length and body weight and/or the shorter trunks of species that use suspensory locomotion.

Figure 6.

Histogram showing the frequency of each body weight estimate predicted by the equations created here. Plotting the frequency in this way makes it easier to throw out extremely high and low estimates that were predicted a single time, or not at all. A range of 17–29 kg seems reasonable with 19, 21, and 25 kg being the most frequent predictions for Protopithecus. These three estimates were given by equations based on several skeletal elements in all three subsets of the reference sample.

The majority of the R2 values are above 0.8. Most equations have %SEE values around 15%–25%, with the total body length equations much lower at 6%–13%; Ruff (2003) concludes that the best one can do is probably 10%–15%. Similarly, low MPE values are comparable to those of the equations considered to be among the best by Delson et al. (2000). These low values were also used to justify including both male and female means in the comparative sample and not creating separate equations for each sex. Only three of the QMLE correction factors are above the 6% value suggested by Smith (1993) to be the maximum cutoff for usable equations.

The average TOTL of about 1,675 mm for Protopithecus is much longer than the average TOTL of any of the male Old World monkey species in the comparative sample, but the fossil's average TrL of about 710 mm is within the range of the larger males (Figs. 4 and 5). As Schultz (1938) and many others subsequently noted in catarrhines, trunk length [defined both as “head and body length” (Ford and Corrucini,1985) and “precaudal vertebral column length” (Majoral et al.,1997)] also scales negatively allometric with body mass across platyrrhine primates. This allometric scaling effect could partially explain the relatively long TOTL but relatively short TrL of the larger-bodied Protopithecus. The relative proportions of two different anatomical areas, the skull and the tail (see below), could also be affecting these length measurements. Because the “head and body length” measurement used here is longer than the “skeletal trunk length” measurement discussed in various allometric studies (e.g., Jungers,1984), including species in the comparative sample which vary greatly in relative skull length will affect comparisons with the fossil predictions. This would be especially problematic if the end point of the length was taken by some observers from glabella and by others from prosthion as the longer snout of Alouatta, Protopithecus, and the papionins would be skewing the results in an unequal fashion. These details are not recorded in the museum catalogs.

The centroid size of the full femur and tibia, values related to overall bone length, were the worst predictors of body size of those used here; this agrees with previous work where long bone lengths were less successful than various measures of joint sizes (e.g., Ruff,2003). Correlations related to joint surfaces at the proximal and distal end of the tibia are also not very strong; it has been suggested that the tibia is not as good as some of the other postcranial elements for body size estimation because of the variable role of the fibula in weight support (Ruff,2003). The distal humerus of Protopithecus, for which there are two specimens, presents a different problem. The specimen that is part of the more recently discovered complete skeleton from Bahia seems to overestimate body size, perhaps because it exhibits a very large brachioradialis flange which increases the measured width and height of the joint surface (Fig. 7). The other distal humerus specimen, which was part of the material discovered by Lund in Minas Gerais, is smaller and gives estimates closer to the average values for the three aspects of size (Fig. 7). This, along with several specific morphological differences in the two bones (Halenar, this volume), could potentially mean that the two humeri represent a male and female of a dimorphic species.

Figure 7.

Protopithecus distal humerus showing the size difference between the specimen from the Toca da Boa Vista cave in Bahia, Brazil (left) and the specimen from Lagoa Santa in Minas Gerais, Brazil (right). Scale bar = 1 cm.

The performance of some of the other predictor variables investigated here is harder to explain. For example, it is unclear why the proximal femur gives such large estimates for the platyrrhine sample even though those equations have comparable confidence statistics to the rest of the “better” variables. Similarly, the ulna is producing very low estimates, despite having relatively good confidence statistics. The over- and under-estimating of various joint sizes could be investigated by producing bivariate plots of joint size and long bone length, a method used by Millien and Bovy (2010) to investigate allometric relationships in fossil rodents, where the sample included small-bodied extant and much larger fossil species. This would require long bone length measurements for each individual in the comparative sample, data which has yet to be collected. The proximal femur as defined here (Fig. 1C) includes more than just the femoral head, the element used to produce the original catarrhine-based body size estimates for Protopithecus (Hartwig,1995; Hartwig and Cartelle,1996). However, the results are comparable: a similarly large value of 25–26 kg is estimated here from the full sample equation. As mentioned above, it is much easier to pick out and discuss the less useful skeletal elements as opposed to determining the “best” ones to use. The distal humerus, proximal ulna, and proximal femur seem to produce more robust equations, but they still give variable estimates depending on the reference sample used.


Explaining the large body size of Protopithecus is a challenge crucial to illuminating various aspects of its paleobiology. Hartwig and Cartelle (1996) suggested that large size could be part of the general tendency of late Pleistocene mammals to increase in body mass (e.g., Martin and Klein,1984). However, while Protopithecus is much larger than any extant platyrrhine, it is not large for a primate and especially unremarkable when compared to other Pleistocene South American mammals such as the ground sloths and glyptodonts. This should be taken into consideration when the temptation arises to refer to Protopithecus as a “giant” subfossil platyrrhine, as has been done in the literature (e.g., Hartwig,1995). The relatively recent late Pleistocene date for the caves in which the Protopithecus remains were found (Auler et al.,2006) could be used to interpret Protopithecus as evidence for Cope's rule of phyletic size increase (Stanley,1973) at work in the ateline lineage. However, Protopithecus has been suggested to represent a late-surviving holdover from an earlier Miocene stage in alouattin evolution (Rosenberger et al.,2009) and a size decrease over evolutionary time makes more sense as an explanation for the evolution of various unique aspects of extant howler monkey morphology (see below). The size range of the entire platyrrhine radiation was broader in the past than it is today, and while it might not be immediately clear why the larger size class evolved and eventually went extinct, its relation to the living atelines can shed light on their evolutionary history.

An important part of the suspensory locomotion practiced by living ateline primates is the use of a prehensile tail which can support their entire body weight and which itself weighs as much as one of the hindlimbs (Grand,1984). The main skeletal distinctions between prehensile and non-prehensile tails are the overall length of the tail and the robusticity of the lumbar and caudal vertebrae (Schultz,1961; Ankel,1972; German,1982; Rosenberger,1983; Lemelin,1995; Meldrum,1998). While the Protopithecus caudal vertebrae were not included in the data analyzed here, inferences about tail length can still be made from the results of this study. Predicting the TOTL and TrL of the fossil allows the calculation of a tail to head and body length ratio, which indicates whether or not the tail is elongated relative to trunk length. The TAILL:TrL index for the fossil is 1.4 which is within the range of mean values for three of the extant prehensile-tailed ateline genera: Alouatta = 1.1, Lagothrix = 1.4, and Ateles = 1.7 (Table 7; data for Brachyteles was unavailable). However, this datum should not be used in isolation to “prove” the prehensility of the Protopithecus tail. Means of many non-prehensile-tailed species in the larger sample, such as Lophocebus (1.6) and Presbytis (1.4), are also high in the ateline range; just because a tail is elongated does not automatically mean it was prehensile. While detailed analyses of the fossil caudal vertebrae and dimensions of the sacral opening can contribute more to addressing this question (as in Youlatos,2003; Schmitt et al.,2005; Organ et al.,2009), Protopithecus is an ateline primate and since all of the other atelines have a prehensile tail, it is relatively safe to infer that the fossil also possesses this trait. The larger ecological picture also favors this interpretation. Various aspects of South American forest structure, such as the lack of lianas, have led to the parallel evolution of a prehensile tail used during locomotion in many mammalian lineages (Emmons and Gentry,1983; Lockwood,1999). However, it would still be informative to use principles of biomechanics to model the feasibility of a 23 kg monkey supporting its entire body weight with a fully prehensile tail.

Table 7. Tail length (TAILL) to trunk length (TrL) index
GenusSpecies TAILL:TrLGenusSpecies TAIL:TrL
Protopithecusbrasiliensis 1.4    
F1.2  F1.4
carayaM1.2 guerezaM1.1
F1.2  F1.1
F1.1  F1.6
F1.2 brunnecensM0.1
pigraM1.1  F0.1
F1.1 fascicularisM1.2
seniculusM1.1  F1.4
F1.5  F0.8
F1.6  F1.6
geoffroyiM1.7 cristatusM1.3
F1.6  F1.3
paniscusM1.9 johniM1.1
F1.8  F1.5
AotusazaraeM1.3 melalophosM0.9
F1.3  F1.2
ChiropotessatanusM1.0 obscurusM1.8
CebusalbifronsM1.2  F1.6
F1.1 rubicundusM1.4
apellaM1.1 frontatusM1.3
F1.2  F1.4

If Protopithecus was too large to hang by a prehensile tail, it is possible that it was also too large to be as acrobatic and suspensory as Hartwig and Cartelle (1996) originally suggested, in part based on its intermembral index of 104. This value is in the high end of the Ateles and Brachyteles range of 99–110 (Erikson,1963). However, the intermembral index increases with increasing body size (Jungers,1985) and when this is taken into account, Protopithecus is more similar in terms of body proportions to Pan than to Ateles (Heymann,1998). Evidence such as this led Heymann (1998) to propose that a terrestrial component to the locomotor repertoire would have been important for Protopithecus. His argument is based on the fossil's exceptionally large body size for a platyrrhine; throughout the short paper, he presents evidence for why their size would have made prehensile-tail assisted ateline-style “brachiation” an untenable locomotor behavior. For example, it has been suggested that 10–12 kg, the size of a large siamang, is the upper size limit for animal that can successfully sustain the metabolic costs of brachiation (Preuschoft and Demes,1985); if ateline-style “brachiation” is more energetically costly than that practiced by hylobatids (Swartz,1989), the 23 kg fossils are even more likely to be using an alternate form of locomotion. Even if Protopithecus was capable of practicing a suspensory mode of locomotion, Heymann (1998) refers to evidence suggesting that the South American forest structure cannot support a large-bodied suspensory taxon, even with the help of a prehensile tail (Emmons and Gentry,1983). He offers a rationale for a “high degree” of terrestriality being practiced by the fossils, but it is also reliant on circular logic. He argues, for example, if Protopithecus and Caipora were terrestrial, their large body size would be a good defense against predators such as jaguars and other large cats.

Inferring terrestriality from large body size has been shown to be a tenuous connection at best; as Remis (1995) so eloquently observed, “Because positional behavior and substrate use vary with habitat, tree structure, and social context, we should avoid making predictions about modern or fossil taxa based on body size considerations alone” (p. 431). The same suggestion was made for several of the subfossil lemurs of Madagascar (which truly are “giants,” with body size estimates ranging from 50 to 200 kg) but detailed analyses of their postcranial morphology show none of the signature features of other living terrestrial primates (Jungers et al.,2002). Similarly, there are no indications in the joints of Protopithecus of adaptations such as the retroflexed medial epicondyle on the distal humerus or a posteriorly directed olecranon process on the proximal ulna that are found in terrestrial Old World monkeys; instead, the postcranial skeleton combines suspensory and climbing adaptations (Halenar,2010, this volume). However, the relatively short head and body length of Protopithecus predicted here, while potentially related to allometric effects (see above), could also be an adaptation to a more frequent use of terrestrial substrates (Anapol et al.,2005). A short overall vertebral column may be a feature of terrestrial primates, but a short lumbar region is also characteristic of suspensory primates like Ateles and the hylobatids (Erikson,1963; Jungers,1984; Jungers and Susman,1984; Majoral et al.,1997). The length of this specific part of the vertebral column is unknown for the incomplete fossil.

This topic remains relevant as a commitment to terrestrial locomotion as practiced by various Old World monkeys and strepsirhines of Madagascar is not seen in any extant New World monkey species. A related fossil from Cuba, Paralouatta varonai, has recently been suggested to be “predominantly semi-terrestrial” due to several unexpected similarities shared with the elbows and digits of cercopithecine primates, although the sparse fossil record of that taxon does not allow for firm conclusions (MacPhee and Meldrum,2006). Regardless of the broad locomotor category eventually assigned to the Brazilian fossils, the fact remains that Protopithecus was a very large monkey and there is no reason to suggest that they were not using the ground to some extent, as even the smaller Ateles and Brachyteles do today on an opportunistic basis to cross open areas in their habitat or utilize specific food sources (Campbell et al.,2005; Mourthe et al.,2007).

Protopithecus has been suggested to be a close relative of Alouatta (Hartwig and Cartelle,1996; Rosenberger et al., in review) and body size changes within the alouattin lineage fit in well with changes in other aspects of their morphology. Alouatta is a genus with unique adaptations in its hyoid bone, mandible, and cranial base for producing very loud long call vocalizations (Chapman,1929; Hershkovitz,1949; Kelemen and Sade,1960; Biegert,1963; Schön,1971; Watanabe,1982). The evolutionary history of these structures is not well known, but Protopithecus can provide some empirical evidence for examining Schön's suggestions pertaining to the anatomy of the Alouatta hyolaryngeal apparatus (Schön,1971,1976; Schön Ybarra,1984,1988,1995). He proposed that howler monkey ancestors were adapted for suspensory locomotion before the enlarged hyoid bone evolved; as the hyoid became larger and took up more space in the neck, acrobatic maneuvering became more challenging and locomotion changed to the more deliberate slow quadrupedalism that is associated with Alouatta today. The large body size of Protopithecus supports this scenario in several ways. The cranial base of Protopithecus is more flexed and not as elongated as that of Alouatta and is more similar overall to the generalized cranial base of Lagothrix (Halenar, unpublished data). It is possible that the fossil, at approximately 23 kg, had a large enough subbasal space to accommodate an expanded hyoid bone under an otherwise un-howler-like basicranium. If extant howler monkeys represent a size decrease over evolutionary time, characters like airorynchy, expanded gonial angles, and an unflexed cranial base would have developed to open up the subbasal space and maintain the howling ability that had become selectively important for aspects of their social behavior (see Baldwin and Baldwin,1976; Sekulic,1982; Sekulic and Chivers,1986; Whitehead,1987; Chiarello,1995; Whitehead,1995).

More work on the paleoenvironmental and biogeographic aspects of howler monkey evolution can shed light on the ecological variables involved in body size changes within the lineage (e.g., Rosenberger et al.,2009). Why Protopithecus and/or other howler fossil relatives evolved an enlarged hyoid and increased reliance on howling in the first place is yet to be determined, but the evolutionary history of alouattins is under investigation (Rosenberger et al., in review; Halenar, unpublished data).


After investigating the relationship between body size and a variety of skeletal dimensions in atelines, platyrrhines, and larger-bodied anthropoids, it can be concluded that despite reasonably strong statistical relationships between some of the elements, there is enough variation in any sample to make fossil body size prediction challenging. Even with all of the relevant numbers reported, it is very difficult to pick the “best” equation from those created in this study or the “best” skeletal element for use in future predictions of body size in fossil platyrrhines and primates more generally. Both of those decisions will change depending on the fossil or question of interest. Accordingly, it is desirable that multiple elements and equations be used in producing an estimate of any aspect of body size. Protopithecus is a special case where many elements of the same individual can be used at once; reporting as many estimates from as many elements as possible is always recommended. Additionally, since for statistical reasons there will never be a single value capable of summarizing the sample-based information on body size for an individual fossil or fossil taxon, it is desirable that mean estimates include a range of values and confidence statistics based on the particulars of the equation used. Having the opportunity to create so many equations at once suggests that there are some elements that should be avoided when making predictions, such as the tibia, which was always among the worst predictors in every configuration of the data presented here.

Despite these caveats, the original interpretation of Protopithecus as a member of a previously unknown large-bodied size class of the platyrrhine radiation can now be supported by an independent investigation and analysis. Inferences about its paleobiology and behavior, and what that means for the evolutionary history of the living ateline primates, can now be made more confidently with this in mind.


Thanks go to Dr. Castor Cartelle of the Museu de Ciências Naturais at the Pontificia Universidade Católica de Minas Gerais in Belo Horizonte, Brazil and Dr. Kim Aaris at the Universitets Zoologisk Museum in Copenhagen, Denmark for access to the fossil material as well as Eileen Westwig (AMNH), Linda Gordon (NMNH), and Dr. Leandro Salles (MN) for access to the extant primate collections. The author would like to thank Drs. Jeffrey Laitman, Alfred Rosenberger, and Walter Hartwig for sponsoring this special issue and allowing her to participate. Special thanks also go to Drs. Eric Delson, Melissa Tallman, Marian Dagosto, and Jeff Meldrum for their support and very helpful comments during the revisions of the manuscript.