OUP user menu

When teeth and bones disagree: body mass estimation of a giant extinct rodent

Virginie Millien, Helene Bovy
DOI: http://dx.doi.org/10.1644/08-MAMM-A-347R1.1 11-18 First published online: 17 February 2010

Abstract

Body size is correlated with virtually every morphological, physiological, and life-history trait in mammal species. As a consequence, estimates of body size of fossil species are often used for paleoecological reconstructions. Characters used as proxies for body mass in extinct species include teeth, skull, and skeletal measurements. We show that the body-mass estimates of extinct species from living taxa can be misleading and depend largely on the morphological variable selected as a proxy for body mass. We also discuss statistical tools that are available to assess the accuracy of body-mass estimates in extinct species. Here, we focus on the revision of the mass estimate of the giant Miocene fossil rodent Phoberomys pattersoni (Venezuela), the 2nd largest rodent ever reported, with an estimated body mass between 436 and 741 kg. This is far beyond the range of average body masses in living rodents, which vary from several grams to 40 kg. We conclude that body mass of Phoberomys was most likely overestimated. The species P. pattersoni likely weighed between 220 kg and 280 kg, the mass of a horse or a large antelope.

Key words
  • allometry
  • body size
  • body-size estimation
  • Miocene
  • Phoberomys
  • Rodentia

Body size is probably the single most obvious and important character of an organism. Most physiological and life-history traits scale with size, including life span, metabolic rate, fasting endurance, rate of development, fecundity, interspecific relations, and numerous other morphological, physiological, behavioral, and ecological factors (Calder 1984; Peters 1983; Schimdt-Nielsen 1984).

For these reasons a long-standing tradition has persisted among vertebrate paleontologists of estimating body mass in fossil mammal species (Damuth and MacFadden 1990). Body size in fossil species is, in principle, relatively easy to characterize, and this has led to comparisons across a wide range of taxa (Alroy 1998; Damuth and MacFadden 1990). Compared to bones or other tissues, teeth are favored material for paleontologists because they are typically well preserved. For instance, body mass of many fossil mammals has been estimated from the length or the surface of the 1st lower molar (Conroy 1987; Creighton 1980; Gingerich et al. 1982; Legendre 1989; Martin 1990). Other dimensions of the upper and lower 2nd molars (Schwartz et al. 1995) and upper and lower incisors (Millien-Parra 2000; Parra and Jaeger 1998) also have been used to estimate body size in fossil rodents. Because they are related to locomotion and body support, measurements of postcranial elements such as the length of the humerus (Gingerich 1990), the diameter of long bones (Biknevicius et al. 1993), diameters of metapodials (Alberdi et al. 1995), or the area of the astragalus (Martinez and Sudre 1995) are also strongly related to body mass in mammals and are used as a proxy for body mass in extinct species.

Rodents represent nearly one-half of all mammalian species (Wilson and Reeder 2005). Since their origin, rodents have dominated communities of mammalian herbivores in both number of species and biomass. This diversity, both taxonomic and ecological, means that rodent species are found in a wide range of environments (Landry 1970), which makes this group an excellent system for both paleontological and ecological studies.

Body mass in extant rodent species covers more than 4 orders of magnitude, from as little as 3.7 g on average in the African pygmy mouse (Mus minutoides) to 40 kg on average in the largest rodent, the South American capybara (Hydro-choeris hydrochaerisSilva and Downing 1995). Some individuals of H. hydrochaeris can weigh as much as 81 kg in the wild (Ferraz et al. 2005), but the largest rodent species ever described are now extinct. Gigantism has been documented in numerous island rodents (review in Millien et al. 2006). For instance, the large extinct rodent genus Amblyrhiza was described from Quaternary deposits in the West Indies (Cope 1883). Cope (1883) estimated the largest Amblyrhiza to be “larger than a male Virginian deer.” This was confirmed by Biknevicius et al. (1993), who proposed an estimated body mass for A. inundata of 50 to >200 kg. Extremely large forms of fossil rodents also are known from the Pleistocene of North America and the Pliocene of South America. Giant beavers of the Quaternary of North America represent a classic example of gigantism. Although the 2 living species, Castor canadensis and C. fiber, weigh on average between 9 and 42 kg (Silva and Downing 1995), the body mass of their extinct relative Castoroides ohiensis was estimated between 46 and 202 kg (Reynolds 2002).

The largest estimated body masses for fossil rodents were assigned to the South American rodents Phoberomys pattersoni (Sanchez-Villagra et al. 2003) and Josephoartigasia monesi (Rinderknecht and Blanco 2008). A genus in the family Neoepiblemidae, Phoberomys (Bondesio and Bocquentin Villanueva 1988) encompasses 7 South American species (Horovitz et al. 2006; Negri and Ferigolo 1999), of which P. pattersoni is the 2nd largest. A nearly complete skeleton of P. pattersoni was found in the Urumaco formation in northwestern Venezuela (Horovitz et al. 2006; Mones 1980), and its body mass was estimated at >700 kg based on femur diameter (Sanchez-Villagra et al. 2003). Most recently, however, 2 studies based on tooth size revised body-mass estimates of P. pattersoni (Hopkins 2008) down to 200–300 kg and that of J. monesi (Millien 2008) to as low as 350 kg, lower than the estimate of 1,000 kg proposed by Rinderknecht and Blanco (2008). These conflicting results highlight the controversy associated with estimating body mass of fossil species that have no Recent equivalent and estimates that depend on the methods that are used. For this study we used many different measurements of the fossil giant rodent Phoberomys and assessed the level of confidence in the body-mass estimates derived from each one.

The most commonly used method of estimating body mass in extinct species begins by calculating the allometric relation between a given skeletal character and body mass in a number of extant species. This relation is then applied to extinct species to estimate body mass. This procedure is very straightforward but requires a number of assumptions and therefore has limitations (Damuth and MacFadden 1990; Egi 2001; Reynolds 2002; Schmidt-Nielsen 1984; Smith 1980, 1996). These limitations derive from the reference data set, the choice of the skeletal character, and the statistical analysis and presentation of the data (Reynolds 2002).

When estimating body mass in extinct taxa, researchers often assume that the relation between a skeletal character and body mass is the same for recent and extinct species (i.e., the allometric coefficient is constant). Because this is impossible to validate empirically, the choice of the reference data set used to reconstruct body mass in extinct species is critical. First, some researchers suggest that the taxa chosen to calculate the allometric relation should be close relatives of the extinct species, but others argue that a wider taxonomic sample offsets other potential bias (Schwartz et al. 1995). Second, because one is looking for the allometric relation at the level of the order (or maybe family), the data set should be composed only of interspecific data and the intraspecific variability is of no interest for estimating body mass in extinct species. Third, allometric relations should be applied to extinct species that fall within the range of size and body proportions observed in the reference data set (Damuth and MacFadden 1990; Reynolds 2002; Schmidt-Nielsen 1984). Last, as in any statistical procedure, sample size should be large enough to allow as much confidence in the results as possible.

A 2nd set of assumptions centers on the choice of the skeletal character used to estimate body mass. Limb bones are widely accepted as a good indicator of body mass because of their role in skeletal support (Damuth and MacFadden 1990; Egi 2001). However, dental and other cranial features also provide accurate estimations of body mass (e.g., Damuth and MacFadden 1990). Because of different functional constraints on teeth and the appendicular skeleton, body-mass estimation from tooth measurements (i.e., feeding attributes) can lead to values consistently different from those estimated from long-bone measurements (i.e., locomotor attributes).

To assess the validity of the regression equation, some crucial descriptive statistics, such as the standard error of the estimate and prediction errors, are often overlooked (Smith 1980, 1996). These statistics provide an indication of the accuracy of the body-mass estimate and should always accompany it. We examine the limitations described above of the allometric method for estimating body mass in extinct species and focus on the revision of the mass estimate of the giant Miocene fossil rodent P. pattersoni from Venezuela.

Materials and Methods

Data set.—Data were collected for 36 species belonging to 32 genera and 18 families of rodents (Appendix I). A sizeable majority (28 species) of the species were hystricomorph rodents, the suborder of Rodentia to which P. pattersoni also belongs. Because P. pattersoni is one of the largest rodents ever described, we also included representatives of large species of myomorph, castorimorph, and sciuromorph rodents. Taxonomy followed Wilson and Reeder (2005). Body-mass-data were obtained from specimen tags and from average values published in the literature (Silva and Downing 1995). The data set covered a large range of average body masses, from 124 g in the Val's gundi (Ctenodactylus vali), to 40 kg in the capybara (H. hydrochaeris). Measurements for the extinct giant rodent P. pattersoni and other Phoberomys species were obtained from the literature (Horovitz et al. 2006; Sanchez-Villagra et al. 2003) and by us from undescribed material from Universidad Nacional Experimental Francisco de Miranda Venezuela (UNEFM).

Measurements.—Eleven measurements were taken from the skull, teeth, and appendicular skeleton of each specimen: the condylobasal length of the skull, the length of the upper cheek toothrow, the anteroposterior diameter of the lower and upper incisors, the transverse diameter of the lower and upper incisors, the maximum length and width of the 1st lower molar, the length of the humerus and femur, the anteroposterior diameter at 65% from the distal articular surface of the femur, and the anteroposterior diameter at 35% from the distal articular surface of the humerus. Incisor measurements were defined in Millien-Parra (2000), and long-bone measurements were as in Biknevicius et al. (1993). We used the mean value in the analyses when several specimens were available from a single species. Only adult specimens for which we could record both dental and skeletal measurements were considered. Measurements were taken with Mitutoyo digital calipers (Mitutoyo Corporation, Kanogawa, Japan) to the nearest 0.02 mm.

Statistical analyses.—All measurements were log transformed before further analyses. The log-transformed values were tested for normality by 1-sample Kolmogorov-Smirnov tests (Zar 1999). Pairwise relationships between body mass and skeletal or dental measurements were fitted to an allometric model that is expressed by a power function: Y = aXb, where a is a constant, b the allometric coefficient, Y is the dependent variable (body mass, in g), and X is the independent variable (dental or long-bone measurements, in mm). The log transformation of this function results in a linear relationship, log Y = log a + b log X, which was fitted to the data by a least-squares criterion (Smith 1984). Three additional parameters were computed: the coefficient of determination (r2); the standard error of the estimate (SEE), which provides a comparative index of the accuracy of the different models computed (Smith 1984); and the average absolute value of individual percent prediction errors (%PE). Individual percent prediction errors were calculated as %pe = [(true mass — estimated mass)/estimated mass] × 100 after the data had been converted back to linear scale (Smith 1980). A stepwise multiple regression model (backward, automatic, P to enter = 0.15, P to remove = 0.15) was then applied to the data. All statistical analyses were performed using SYSTAT software (SYSTAT Software Inc. 2004).

Skeletal analysis.—Because Phoberomys specimens fall far outside the range of sizes known in Recent rodents, we investigated whether extant rodents can be used as a reference sample to estimate body mass in the extinct taxon. Our approach was to compare bone and teeth proportions in living species and to assess whether these proportions are conserved in the giant fossil species. Bivariate plots were used to compare bone and teeth proportions in living species with those observed in Phoberomys. These visual examinations of the data helped us to detect the models that, although statistically significant, may not be the most appropriate to estimate body mass in Phoberomys. We also assessed the relevance of the allometric models by calculating ratios between various measurements of the fossil taxon and comparing them to values obtained from reference specimens of Recent species.

Results

Models calculation.—A Kolmogorov-Smirnov statistic adjusted for multiple tests showed that all variables were normally distributed. All regressions using log body mass as the dependent variable were significant (all P < 0.0001). For 3 of the allometric relations calculated, some individual data points had high leverage values (leverage > 0.3—Fox 2008), and the models were recalculated without these individual points (Table 1). The capybara had relatively long ml and upper toothrow compared to other rodents, which resulted in high leverage values (0.29 in both cases). Although coefficients of determination were high for all allometric relations calculated (r2-values ranging between 0.82 and 0.89; Table 1), SEEs also were relatively high, ranging from 0.21 to 0.27 (Table 1). Estimation of body mass from dental or skeletal measurements thus can vary within an interval of ±21−27% around its real value.

View this table:
Table 1

Parameters of the equations used to estimate body mass (g) in Phoberomys. Skull: condylobasal length of the skull; UTRL: upper toothrow length; Lower AP and Upper AP: lower and upper incisor anteroposterior diameters (mm); Lower T and Upper T: lower and upper incisor transverse diameters (mm); ml length and ml width: 1st lower molar maximum length and width (mm); n: sample size; r2: coefficient of determination; SEE: standard error of the estimate; %PE: average absolute percent prediction error, n = 35.

Character (all in mm)SlopeInterceptr2SEE%PE
Skull3.488−3.3320.940.15730.65
UTRL2.698−0.0380.890.21543.36
Upper AP3.2371.3150.820.27457.17
Upper Ta2.4452.0800.830.26254.98
ml lengthb2.9021.4340.890.21844.62
ml width3.1771.3510.870.22645.76
Lower AP2.9261.6010.820.27346.03
Lower Ta2.2742.2570.820.26956.74
Femur length2.825−1.9640.860.24048.59
Femur diameter2.6350.9030.840.25747.45
Humerus length2.827−1.6780.850.24950.06
Humerus diameter2.5261.4570.880.21739.42
  • a Cryptomys ochraceocinereus removed from the data set, n = 34.

  • b Ondatra zibethicus removed from the data set, n = 34.

Multiple regressions included only the following independent characters: Upper toothrow length, Lower anteroposterior diameter of the incisor, Femur diameter, Humerus length, and Humerus diameter (model 1; Table 2). This last model, with an SEE of 0.12, is more accurate in estimating body mass than any of the pairwise models. Because not all variables were available for individuals of Phoberomys, a 2nd multivariate model was calculated that included the following independent variables: Upper toothrow length, Femur diameter, Humerus length, and Humerus diameter (model 2; Table 2). A 3rd model, which did not include long-bone diameters, also was calculated (model 3, variables included: Upper toothrow length, Femur length, and Humerus length).

View this table:
Table 2

Parameters of the multivariate predicting equation (model 2) used to estimate body mass (g) in Phoberomys; R2: square multiple coefficient of determination; SEE: standard error of the estimate; PE: average absolute prediction error; n = 35. Model 1: F5,29 = 188.66, P < 0.001, SEE = 0.119, R2 = 0.97, PE = 0.208; model 2: F4,30 = 193.61, P < 0.001, SEE = 0.131, R2 = 0.96, PE = 0.206; model 3: F 3,31 = 129.50, P < 0.001, SEE = 0.181, R2 = 0.93, PE = 0.351. UTRL: upper toothrow length; Lower AP: lower incisor anteroposterior diameter (mm).

EffectCoefficienttP
Model 1
Constant−1.060−3.1180.004
UTRL1.5195.9760.000
Lower AP1.0662.6940.012
Femur diameter1.3035.1100.001
Humerus length1.0654.0260.001
Humerus diameter−1.616−2.9430.006
Model 2
Constant−0.915−2.4790.019
UTRL1.2884.8970.000
Femur diameter1.2924.6100.000
Humerus length1.0593.6390.001
Humerus diameter−0.553−1.3180.198
Model 3
Constant−1.019−3.1100.004
UTRL1.6134.9840.000
Humerus length0.9420.4510.655
Femur length0.3601.2620.216

Skeletal analysis.—Scatter plots of humerus diameter against its length (Fig. la) and femur diameter against its length (Fig. lb) show that the giant fossil rodent Phoberomys has relatively large bone diameters. The linear relation between log-transformed diameter (independent variable) and log-transformed length (dependent variable) were used to calculate percent prediction errors (%pes) of estimated diameters from the length of long bones. The specimen P. pattersoni UNEFM-VF-020 has a humerus diameter 47% and a femur diameter 82% larger than in a “normal” rodent. On the contrary, Phoberomys spp. specimens CIAAP-1438 and UNEFM-VF-010 seem to better fit a rodent design with %pe values of −3% and −4%, respectively, for humerus diameter. Yet the %pe for the femur diameter in specimen CIAAP-1438 is 163%. Both of these specimens each represent a single individual and may belong to Phoberomys species other than pattersoni. The large size in Phoberomys thus seems to be accompanied by an increase in the robustness of the long bones, and body-mass estimates from long-bone diameters are likely to be overestimated.

Fig. 1

Bivariate plots of a) humerus diameter against humerus length and b) femur diameter against femur length. Regression equations used to calculate the estimated value of diameters from length were as follow: diameter = 1.049 × length − 1.121 (humerus, log-transformed) and diameter = 0.099 × length − 0.977 (femur, log-transformed). Filled square: reference sample for 35 rodent species; open circles: Phoberomys pattersoni and Phoberomys spp.

When compared with recent rodents, P. pattersoni possessed an upper toothrow of usual length relative to the lengths of its humerus and femur (Fig. 2). The capybara has a relatively long toothrow compared to its femur length and humerus length. The upper toothrow length and the lengths of the humerus and femur are thus recommended for the estimation of body mass in the extinct rodent Phoberomys spp.

Fig. 2

Bivariate plots of a) the ratio of upper toothrow length (UTRL) to humerus length against the ratio of UTRL to femur length, and b) the ratio of humerus length to femur length against the ratio of UTRL to femur length. Filled square: reference sample for 35 rodent species; open circle: Phoberomys pattersoni UNEFM-VF-020. The arrow points to the symbol for Hydrochoeris hydrochaeris.

Unfortunately, the dimensions of the 1st lower molar are known for only 1 specimen of P. pattersoni (UNEFM-VF-025) and 1 specimen of Phoberomys sp. (CIAAP-1449) for which no other measurement is available. The same problem existed for the diameters of the upper and lower incisors that are known from only 3 isolated incisors. These dimensions cannot be compared to other skeletal dimensions of Phoberomys. As a consequence, although they may be accurate, one has to be extremely cautious in the interpretation of body-mass estimates for Phoberomys based on measurements on incisors and on the 1st lower molar.

Body-mass estimates.—Estimated body-mass values for P. pattersoni ranged from 221 to 460 kg and from 565 to 774 kg in UNEFM-VF-020 and UNEFM-VF-025, respectively (Table 3). Estimates for UNEFM-VF-025 were based on the dimensions of the 1st lower molar only. Average prediction errors for each of the allometric relations indicate that P. pattersoni may have weighed as little as 111 kg but also could have weighed over a ton. This last figure was based on a body-mass estimate from the length of the 1st lower molar. Allometric models based on the length of the upper toothrow and the humerus and femur lengths yielded an estimated body mass of 221–281 kg. Estimated body mass for P. pattersoni UNEFM-VF-020 was 568 kg and 346 kg based on multivariate models 2 and 3, respectively. As opposed to model 2, model 3 did not include long-bone diameters. The other smaller species of Phoberomys had estimated body masses ranging from 53 to 912 kg. When considering values obtained from the upper toothrow and from the humerus and femur lengths only, estimated body mass for Phoberomys spp. ranged from 96 to 183 kg (Table 3).

View this table:
Table 3

Body-mass estimates for Phoberomys pattersoni and Phoberomys spp. and associated prediction errors (%PEs). UTRL: upper toothrow length; Lower AP and Upper AP: lower and upper incisor anteroposterior diameters, respectively (mm); Lower T and Upper T: lower and upper incisor transverse diameters, respectively (mm); ml length and ml width: 1st lower molar maximum length and width, respectively (mm). Model 2: multivariate model including UTRL, Femur length, and Humerus length as independent variables. An asterisk (*) indicates that part of M3 is broken. UNEFM = Universidad Nacional Experimental Francisco de Miranda, Venezuela. CIAAP = Centro de Investigaciones Antropológicas, Arqueológicas y Paleontológicas, Coro, Venezuela.

CharacterMeasurement (mm)Body mass (kg)%PEMass -PE (kg)Mass + PE (kg)
Phoberomys pattersoni
UNEFM-VF-020
UTRL*108.1281± 43.4159403
Femur length402247± 48.6127367
Femur diameter64460± 47.524168
Humerus length305221± 50.1111332.
Humerus diameter45430± 39.4260599
UNEFM-VF-025
ml length34.3774± 44.64291,119
ml width24.3565± 45.8306823
Phoberomys spp.
CIAAP 1438
Femur length355174± 48.689258
Femur diameter83912± 47.54791,344
Humerus length280174± 50.187261
Humerus diameter26.8116± 39.470162
CIAAP 441
Femur length287.596± 48.649142
Humerus length284.8183± 47.596269
UNEFM-VF-010
Humerus length270157± 50.178235
Humerus diameter26107± 39.465150
CIAAP 1449
ml length18.2123± 44.668178
ml width19259± 45.8141378
UNEFM “c”
Lower AP13.276± 46.041111
Lower T12.758± 56.72592
UNEFM “f ”
Lower AP14.396± 46.052140
Lower T12.860± 56.72693
UNEFM “e”
Upper AP14.45117± 57.250184
Upper T12.153± 55.02483

Discussion

Despite a relatively large sample size of reference data, the regression statistics were not sufficient to assign complete confidence to any single morphological character. The most notable results from the calculation of the allometric models were the consistently high correlation coefficients. This is typical of interspecific correlations and also of large sample size (Schwartz et al. 1995); however, the correlation coefficient is not the most convincing index value (Smith 1980). To assess the validity of the regression %PE is a much more useful tool because it determines how accurately the regression model predicts the estimate (Schwartz et al. 1995; Smith 1980). The %PE in this study is high for all equations (%PE > 30 for all characters considered). However, this does indicate that the values found for any of these equations are rough estimates because %PE and SEE also can be misleading (Roth 1990). Logged values when untransformed are biased (Smith 1993; Sprugel 1983). Because estimates are used in the calculation of both %PE and SEE, this bias is inevitably reflected in these 2 statistics. Furthermore, equations that underestimate body size lead to higher %PE values because of the smaller estimate in the denominator (Roth 1990). Other factors that could explain the high %PE values are high measurement errors or the heterogeneity of the sample (Van Valkenburgh 1990).

The choice of variables obviously has a large impact on body-mass estimates, and this is confirmed by the large variation in the present study. Estimates based on skull bones or teeth are often dismissed (Egi 2001; Smith 1996), arguably due to the lack of stress imposed on the cranium by body mass. Instead, limb-bone measurements are commonly accepted (Damuth and MacFadden 1990; Egi 2001), and it has been assumed that the size of the femur, which bears much of an animal's body weight, should be a reliable proxy for its body mass (Sanchez-Villagra et al. 2003). However, although cross-sectional areas of long limb bones are recognized as good indicators of body mass, in certain groups (e.g., Hyaenodontidid) femoral cross-section equations result in overestimated body mass (Egi 2001). We showed that the range of body-mass estimates for Phoberomys, and in particular for P. pattersoni, is large and depends considerably on the character used.

Sanchez-Villagra et al. (2003) ed that the ratio of femur diameter to humerus diameter of P. pattersoni was 1.42, a value beyond the range of extant caviomorph rodents that were used as a reference sample in their study. Furthermore, although structural variation in limb bones of rodents is rather limited, limbs of the capybara exhibit features, such as a thick humerus and long olecranon, that most likely are related to its semiaquatic lifestyle and large body mass (Elissamburu and Vizcaino 2004). Our skeletal analysis revealed that Phoberomys shows some morphological peculiarities, with unusually robust long bones (i.e., large bone diameters) when compared to extant rodent species. We thus conclude that body-mass estimates of Phoberomys derived from diameters of long bones will tend to be overestimates. This corroborates the results presented by Sanchez-Villagra et al. (2003) in which the largest body-mass estimates were obtained from the femur and humerus diameters.

We could not establish the accuracy of the estimates based on the 1st lower molar because of incompleteness of the fossil remains of Phoberomys. Instead, we recommend the use of the length of the upper toothrow, length of the femur, length of the humerus, or a combination of all 3 to estimate body mass of Phoberomys. Using these 3 characters, P. pattersoni would have weighed between 220 and 340 kg. This estimate is 55–70% lower than the original figure of 740 kg proposed for P. pattersoni (Sanchez-Villagra et al. 2003). However, these values are still 5–8 times greater than the average body mass of the largest living rodent, the capybara, and greater than the mass of other fossil giants described in the past (e.g., the giant beaver [C. ohiensis] or the giant hutia [A. inundata]), except for the recently described J. monesi.

Although P. pattersoni is certainly one of the largest rodents ever described, because of the lack of phylogenetically similar taxa with comparable bone and tooth proportions, body-mass estimates for P. pattersoni should be regarded as conditional. Estimation of body mass of extinct taxa that fall outside the range of size of the reference sample has been regarded as one of the largest causes for overestimated values (Damuth and MacFadden 1990; Reynolds 2002; Schmidt-Nielsen 1984).

Factors that led Phoberomys and other extinct taxa to reach such gigantic sizes are unknown. Island species provide the best recorded and some of the most spectacular examples of evolution in body size, including gigantism in small mammals. However, none of the known island rodent species have reached a size comparable to those of Phoberomys or the even larger Josephoartigasia. Several hypotheses have been formulated to explain gigantism among island rodents (review in Millien and Jaeger 2001). Most of them rely on changes in interspecific interactions between species on islands, such as a reduction or loss of predation or competition, or both. Another set of hypotheses suggests the adaptation of isolated species to drastic changes in their physical environment. The sediments in the Urumaco formation indicate a coastal wetland with some.lagoons separated by sandy barriers, suggesting that P. pattersoni was semiaquatic, similar to H. hydrochaeris (Horovitz et al. 2006). Other fossil rodents of unusually large size have been found in the same formation, including the genus Telicomys, which may have been 70% the size of P. pattersoni (Horovitz et al. 2006). The gigantism of Phoberomys may represent an adaptation to locomotion and foraging in a coastal wetland environment, a hypothesis corroborated by the morphology of its massive and heavy limbs.

It is debatable why none of the giant rodents such as Phoberomys or Josephoartigasia in South America, Casto-roides in North America, or Amblyrhiza in the Caribbean persisted. The capybara is the only survivor of a once much diversified group of rodents of unusually large size. Today, the ecological niches of giant fossil rodents are filled by ungulates. The success of this group over large rodents may be related to a highly adapted digestive system that allows ungulates to rely on an entirely herbaceous diet despite their large size.

Despite our conclusion that the mass of P. pattersoni was overestimated, this species, with a mass we estimated to be 220–340 kg, remains one of the largest rodents ever described. Our investigation suggests that efforts should be made to account for sources of error when estimating body mass of extinct species.

Acknowledgments

We thank C. Denys for access to the collections at the Museum National d'Histoire Naturelle, Paris, M. Sanchez-Villagra and D. Auguilera for arranging the loan of Phoberomys specimens from UNEFM, and A. Gonzalez and 2 anonymous reviewers for their comments on the manuscripts. We also thank C. A. Indica for translating the Spanish summary. This research was funded by a grant from the Fonds Québécois de la Recherche sur la Nature et les Technologies (FQRNT 118967).

Appendix I

List of species used for the reference data set. Specimens were examined at the Museum National d'Histoire Naturelle (MNHN), Paris, and at the Redpath Museum (RM), McGill University, Montreal. Specimen numbers for each species are given in parentheses.

Atherurus macrourus (MNHN2001-470), Castor canadensis (MNHN 1996–2168, MNHN 1958–203, MNHN 1996–520), C. fiber (MNHN1994–2961, MNHN1998–1957, MNHN1998–1941, MNHN1994–2960), Cavia aperea (MNHN1986–470, MNHN1986–469), C. porcellus (MNHN 1962–2092), Chinchilla lanigera (MNHN1974–332, MNHN1974–333, MNHN1974–331) Coendou prehensilis (MNHN1997–643, MNHN1995–3239), Cricetomys sp. (MNHN1993–212), C. gambianus (MNHN1995–1565), Cryptomys ochraceocinereus (MNHN1956–717), Ctenodactylus vali (MNHN1951–254, MNHN1951–256, MNHN1951–255), Cuniculus paca (MNHN2005–1026), Dasyprocta sp. (RM2460), D. leporina (MNHN1998–677, MNHN1998–2253, MNHN1917–64, MNHN2006–503), Dinomys branickii (MNHN1990–658), Dolichotis patagonum (MNHN1974–87, MNHN1961–1035, MNHN1974–86), Echimys chrysurus (MNHN1999–1082), Erethizon dorsata (RM2474), Galea spixii (MNHN1972–857, MNHN1975–465, MNHN1975–455), Hydrochoeris hydrochaeris (MNHN1988–197, MNHN1962–2221, MNHN2001–1972, RM6025), Hystrix cristata (MNHN1990–662, MNHN1991–615), Lagidium peruanum (MNHN1971–361), Makalata didelphoides (MNHN1983–360, MNHN1972–853), Marmota marmota (MNHN1958–729, MNHN1996–2438), Myocastor coypus (MNHN1959–146, MNHN1990–651, MNHN1958–204, MNHN1952–876, RM1161), Myoprocta acouchy (MNHN1962–1329), Octodon degus (MNHN1958–747), Ondatra zibethicus (MNHN1983–865, MNHN1983–863, RM2475, RM2725), Petaurista petaurista (MNHN1982–843), Proechimys cuvieri (MNHN1998–689, MNHN1998–696, MNHN1995–3218, MNHN1981–409), Rattus norvEgiens (VM0), Sciurus niger (MNHN2001–106), Spermophilus mexicanus (RM1178, RM1139), Sphiggurus insidiosus (MNHN1997–641), Thrichomys apereoides (MNHN1972–853), Thryonomys swinderianus (MNHN1964–204, MNHN2003–70).

Footnotes

  • Associate Editor was Elizabeth R. Dumont.

Literature Cited

View Abstract