The dinomyids appear in South America in the Oligocene, but it was in the Miocene and the Pliocene that these rodents show a spectacular radiation, with many different species recorded in Argentina, Brazil, Uruguay, Venezuela, Bolivia and Colombia (

Mones 1986;

White & Alberico 1992). However, not much is known about these mammals because most of the fossils recovered are isolated teeth and incomplete mandibular remains. These kinds of fossils do not give much information about the palaeobiology, anatomy and phylogenetic relationship of the species. Nevertheless, the importance of this group is strongly supported by the great diversity of dental shapes and sizes, suggesting that dinomyids occupied many different ecological niches during the last 30

Myr in South America (

Fields 1957).

The new complete skull offers important clues for palaeobiological inferences. It seems strange that with such a large and strong skull *J. monesi* nevertheless had very slender zygomatic arches in comparison with other hystricognath rodents. In addition, the grinding teeth are very small and the pterygoid fossa is slightly reduced implying a minor lateral component to chewing. These features suggest that *J. monesi* had weak masticatory muscles for grinding food and probably did not have the abrasive diet typical of other hystricognath rodents. Palaeobiologically this could imply a diet composed of soft vegetation and perhaps fruit. A reliance on aquatic plants would appear entirely congruent with the inferred palaeoenvironment and also the relatively small size of the teeth.

The surprisingly large body size of this species suggests a particular palaeobiology for a rodent ( and ). Body mass estimations based on predictive equations from data on living rodents are not very reliable due to the large body mass gap between the largest living rodent (capybara) and the much larger giant extinct rodent (; see

Biknevicius *et al*. 1993;

McFarlane *et al*. 1998). However, it is interesting to obtain some body mass estimation due to the ecological and life-history implications of size (

Peters 1983). In any case, the skull described here is clearly larger than the poorly preserved skull of

*Phoberomys* (family Neoepiblemidae) reported previously as the largest rodent that ever existed with a body mass estimated using postcranial measures (

Sánchez-Villagra *et al*. 2003). The best comparison of size would be obtained by applying the same methods in both cases; but this cannot be done at present since

*we* have no postcranial bones of

*Josephoartigasia*.

In

Reynolds (2002), the body mass of the giant extinct rodent

*Castoroides* was estimated based on skull length, obtaining a maximum value of 200

kg. If we apply the same method (with the same allometric relation) to

*J. monesi*, we obtain a mean body mass of 1400

kg with a standard deviation of 533

kg and extreme values of 716 and 2250

kg.

However, the sample of living rodents used to estimate *Castoroides* body mass is very broad, including many groups not closely related with dinomyids. We have obtained a new allometric relation between skull length and body mass using 13 specimens of eight genera of hystricognath rodents, the closest living relatives of *Josephoartigasia* (see ).

According to this allometric relation, the mass of

*Josephoartigasia* was 1008

kg. Because the skull is almost complete, it is possible to check this estimate using other cranial measures. We have obtained allometric relations between body mass and six other standard cranial measures (

Vassallo 2000) for the same sample of hystricognath rodents, and estimated the mass of

*Josephoartigasia* using each of these (). The equations listed in were obtained by us from log–log linear regression by minimum squares. All the exponents are between 3.9 and 3.1. We estimated the body mass for the extinct rodent averaging the estimations for each cranial measure. A standard error and a maximum range of variation were also obtained by this method.

| **Table 2**Parameters (*a* and *b*) of the allometrical equations obtained to estimate body mass. (The general mathematical form is *M*=*aX*^{b}, where *M* is the body mass (kg) and *X* is one of the anatomical measurements (cm); abbreviations are the same as given in figure legend (more ...) |

The average of all seven body mass estimates gives 1211

kg with a standard deviation of 753

kg, when each of the estimates is given equal weight. The largest and smallest estimates of the mass of

*J. monesi* obtained are 2586

kg (from the extremely large width of the incisors) and 468

kg (from the very small zygomatic arches), respectively. Our method gives results very close to the ones following

Reynolds (2002).

Our results are clearly consistent with a body mass greater than 400

kg (estimated from forelimb bones data) and 700

kg (estimated from hind limb bones data) of

*Phoberomys* (

Sánchez-Villagra *et al*. 2003). Moreover, the fragments available for the cranium of

*Phoberomys* (

Horovitz *et al*. 2006) indicate a skull of approximately 65% the size of that of

*Josephoartigasia*. We can thus conclude with a high degree of confidence that our specimen of

*J. monesi* had a body mass about twice that of

*Phoberomys*, making it the largest rodent known to have existed.

Our result reinforces the conclusions of a previous work (Sánchez-Villagra 2003), indicating that Rodentia displays the second largest range of sizes among mammalian orders, after diprotodontian marsupials.