the paleogene mammalian fauna of santa rosa, amazonian peru

The Paleogene MammalianFauna of Santa Rosa,

Amazonian PeruEdited By Kenneth E. Campbell, Jr.

Natural History Museum of Los Angeles County

Science Series 40 December 27, 2004

TH

E PA

LE

OG

EN

E M

AM

MA

LIA

N FA

UN

A O

F SAN

TA

RO

SA, A

MA

ZO

NIA

N PE

RU

/ CA

MPB

EL

L

900 Exposition BoulevardLos Angeles, California 90007

THE PALEOGENE MAMMALIAN FAUNA OFSANTA ROSA, AMAZONIAN PERU

Cover Illustration and Plate I (above): SEM stereopairs of representative molar teeth of new genera andspecies from Santa Rosa local fauna. a–e, Rodents: a–c, family Agoutidae, three gen. et spp. nov.; d, e(reversed), family Echimyidae, two gen. et spp. nov. Upper left molars; anterior to right, labial at top. f–j,Marsupials: f–g, Didelphimorphia, ?Herpetotheriidae gen. et sp. nov., last lower left molar; f, occlusal view;g, labial view; h, Polydolopimorphia, family indet., gen. et sp. nov., upper left molar, occlusal view; i,Paucituberculata, Caenolestidae gen. et sp. nov., upper left molar, occlusal view; j, Polydolopimorphia,family Prepidolopidae, gen. et sp. nov., right upper molar. Scale bars 5 1 mm.

THE PALEOGENE MAMMALIANFAUNA OF SANTA ROSA,

AMAZONIAN PERU

Edited ByKenneth E. Campbell, Jr.

NO. 40SCIENCE SERIES

NATURAL HISTORY MUSEUM

OF LOS ANGELES COUNTY

SCIENTIFIC PUBLICATIONS COMMITTEE

John Heyning, Deputy Director for Research and Collections

John M. Harris, Committee Chairman

Brian V. Brown

Gordon Hendler

Joel W. Martin

Xiaoming Wang

K. Victoria Brown, Managing Editor

Natural History Museum of Los Angeles CountyLos Angeles, California 90007

ISSN 1-891276-27-1Published on 27 December 2004

Printed in the United States of America

PREFACE

This volume describes most of the known mam-malian fauna of the Santa Rosa local fauna of east-ern Peru, the first Paleogene vertebrate fauna to bereported from the forested, tropical lowlands of theAmazon Basin. It represents the work of many in-vestigators, some of whom had a much more dif-ficult task than others because of the unequal rep-resentation of the different mammalian groups pre-sent in the paleofauna. The articles are heavilyweighted toward taxonomic descriptions of speciesnew to science, but there is also much discussion ofthe relationships of the new forms and what theymight indicate about the evolution of the SouthAmerican mammalian fauna. As the first descrip-tion of a Paleogene fauna from Amazonia, this pub-lication presents critical baseline studies that will beof value to all students of South American verte-brate paleontology. I consider it a privilege to haveserved as both an editor and a contributor for thispublication, and I hope its contents will prove asexciting to the reader as they have been to me.

This work is the culmination of the first phase inefforts to describe the mammalian paleofauna fromSanta Rosa. I say first phase because, although theSanta Rosa local fauna has been sampled, the sam-ple has not been thorough. These papers provide adetailed look at an abundant and diverse mam-malian paleofauna, but they are based on a rela-tively small sample. Future excavations at the SantaRosa site, including more rapid processing ofscreen-washed matrix using heavy liquid tech-niques, will undoubtedly yield both a vast collec-tion of vertebrate specimens and many more speciesnew to science.

Any work as extensive as this cannot includemany of the more recent advances that have oc-curred in South American vertebrate paleontology.For example, work continues in Argentina on refin-ing the dates of Paleogene South American LandMammal Ages (SALMAs); a second rodent, onemuch more derived than any from Santa Rosa, hasbeen reported from the Tinguirirican faunas ofChile; formal recognition of a TinguiriricanSALMA has been proposed; and preliminary stud-ies of armored edentates from Santa Rosa appearto support an Eocene age for the Santa Rosa paleo-fauna. None of these developments alter the taxo-nomic conclusions presented herein, but some ofthem might be important for a future age assign-ment for the Santa Rosa local fauna.

Field research in vertebrate paleonotology in thevast tropical lowlands of the Amazon Basin is chal-lenging, but the obstacles are probably no greaterthan comparable research taking place in otherparts of the world, such as Antarctica or the Sa-hara. What is really difficult is overcoming the

deeply entrenched view that paleontological re-search in Amazonia is bound to fail because thereis nothing there to discover. Our research dispelsthis notion. Moreover, the specimens describedherein are significant not just because they repre-sent species new to science. As the first Paleogenevertebrate fauna for half of the South Americancontinent, the Santa Rosa local fauna has providedwonderful new perspectivies on early South Amer-ican mammals and opened the door to entirely newconsiderations of the evolution of major lineages ofSouth American mammals. Equally important, ifnot more so, is the impetus given to the search fornew Paleogene sites in Amazonia. The outcrops arethere. As in paleontological fieldwork anywhere, itis only necessary to find the right ones.

I wish to take this opportunity to thank all thecontributors for their enthusiasm and commitmentto this multifaceted project, as well as acknowledgetheir patience during the long process of bringingthe results of their research to fruition. I thank allof the many reviewers for their gracious contribu-tions of time and effort, which contributed to theoverall improvement of this volume.

I also wish to thank those who have contributedin so many other ways to this work. John G. Wig-more has been an especially enthusiastic supporterof our Amazonian field work. He participated inthe Santa Rosa discovery expedition and in follow-ing years provided funding that supported addi-tional expeditions. Additional funding for field-work came from the National Geographic Society,Winifred Rhodes, and Richard Seaver. Encourage-ment for our field efforts in western Amazoniacame from the Instituto Geologico, Minero, y Me-talurgico (INGEMMET) of Peru, and especiallyIng. Hugo Rivera Mantilla, Director Tecnica ofINGEMMET, and Ing. Oscar Palacios, DirectorGeneral de Geologıa de INGEMMET. Withouttheir desire to seek basic information on the geo-logical resources of Peru, coming from whateverdiscipline, we would never have succeeded in ourefforts to find such a site as Santa Rosa. Criticallogistical support for airlifting field equipment andmany heavy bags of matrix from the isolated out-post of Breu to Pucallpa was provided by the Pe-ruvian army. Last, but certainly not least, I thankmy field companions of many years of sloggingthrough the mud of Amazonia, Carl David Fraileyand Lidia Romero-Pittman. Ultimately, it must berecognized that it was their indefatigable effortsand good cheer, as well as their sharp eyes for tinyfossils, that made this work possible.

Kenneth E. Campbell, Jr.August 2004Los Angeles, California

Science Series, Number 40, pp. 1–163Natural History Museum of Los Angeles County, 2004

CONTENTS

The Paleogene Santa Rosa Local Fauna: Introduction...................................................... 1Kenneth E. Campbell, Jr.

The Paleogene Santa Rosa Local Fauna of Amazonian Peru: Geographic and GeologicSetting ................................................................................................................ 3

Kenneth E. Campbell, Jr., Carl David Frailey, and Lidia Romero-Pittman

New Paleogene Marsupials from the Amazon Basin of Eastern Peru ............................... 15Francisco J. Goin and and Adriana M. Candela

Paleogene Notoungulates from the Amazon Basin of Peru ............................................. 61Bruce J. Shockey, Ralph Hitz, and Mariano Bond

Paleogene Rodents from Amazonian Peru: The Santa Rosa Local Fauna.......................... 71Carl David Frailey and Kenneth E. Campbell, Jr.

Incisor Enamel Microstructure of South America’s Earliest Rodents: Implications forCaviomorph Origin and Diversification ................................................................ 131

Thomas Martin

A Possible Bat (Mammalia: Chiroptera) from the ?Eocene of Amazonian Peru ................ 141Nicholas J. Czaplewski and Kenneth E. Campbell, Jr.

Enigmatic Mammal from the Paleogene of Peru .......................................................... 145Francisco J. Goin, Emma Carolina Vieytes, Marıa Guiomar Vucetich, Alfredo A. Carlini,

and Mariano Bond

The Santa Rosa Local Fauna: A Summary .................................................................. 155Kenneth E. Campbell, Jr.


The Paleogene Santa Rosa Local Fauna: Introduction

Kenneth E. Campbell, Jr.1

The physiographic region known as lowland Ama-zonia comprises approximately 40 percent of theSouth American continent. Yet, prior to the discov-ery of the Santa Rosa local fauna of eastern Peru,absolutely nothing was known of the fauna thatinhabited this portion of the continent in the Paleo-gene. Although the eastern Amazonian lowlands ofPeru are known to have numerous paleontologicalsites containing vertebrates, some of which haveproduced some major paleontological surprises(e.g., Willard, 1966; Frailey, 1986; Kay and Frailey,1993; Campbell et al., 2000), these sites all datefrom the Neogene. Remarkably, because of the di-versity of the mammalian taxa represented in thecollection from Santa Rosa and the abundance ofspecimens of so many of the taxa in the paleofauna,an exceptionally good first picture of mammalianlife during the slice of time represented by the SantaRosa local fauna has emerged. Although still in-complete, this picture provides us with a wealth ofdata about certain groups of mammals. In time, aswe secure more specimens, the nonmammalian ver-tebrates are studied, and we better understand thedata available, the Santa Rosa local fauna will pro-vide much crucial information for paleoecologicalreconstructions of the Paleogene of Amazonia.

To many, perhaps the most surprising aspect ofAmazonian vertebrate paleontology is that thereare any fossils to be found in these vast tropicallowlands. This attitude may well be derived fromthe false assumption that fossils are not to be foundin tropical areas, which is based in large part, per-haps, on the erroneous assumption that fossils arenot preserved under tropical conditions. This lineof reasoning ignores the fact that many of westernNorth America’s fossiliferous lower Tertiary siteswere formed in tropical settings. The only differ-ence is that today the Amazon Basin is covered withtropical vegetation, whereas western North Amer-ica now encompasses many arid and semiarid re-gions where extensive rock exposures are common,long stratigraphic sections are exposed, and fossilscan be found with relative ease. In lowland Ama-zonia, outcrops are found only along river cutbanksand good exposures are few. Nonetheless, fossilvertebrates are fairly common in certain formationsunderlying the thick cover of tropical vegetation ofmodern Amazonia.

In the tropical lowlands of western Amazonia,two factors make paleontological investigationspossible. The first is that most of western Amazonia

1 Vertebrate Zoology, Natural History Museum of LosAngeles County, 900 Exposition Boulevard, Los Angeles,California 90007. Email: [email protected]

has undergone slight, broadscale uplift as a conse-quence of the latter phases of the Andean Orogeny,and a few limited areas have seen significant localfolding and faulting as well. The overall uplift hasnot been great, averaging perhaps only a few tensto a hundred meters, but it has facilitated exposureof Neogene deposits throughout the region. Olderdeposits are exposed in more limited areas, wherelocal uplift has exceeded several hundred meters. Insome instances, the entire Tertiary sequence maycrop out along rivers crossing anticlinal structuresformed during late Tertiary/Quaternary uplift inwestern Amazonia (Kummel, 1948). This is becauseof the fairly shallow depth to underlying basementrock and the relatively unconsolidated nature of theTertiary sediments. The largest of these uplifted ar-eas is the Sierra de Divisor, a narrow, folded andfaulted zone that extends over a linear distance ofabout 240 km along the border region betweenPeru and Brazil. The Santa Rosa locality was foundwithin this structural zone.

The second factor is that, because of the estab-lishment of the modern Amazonian drainage sys-tem in the late Neogene, the rivers in the regionflow in entrenched valleys. This means that whenthe rivers come into contact with the sides of theirvalley, they usually form cutbanks where one canobserve the stratigraphic column and where fossil-iferous deposits are exposed. The height of the cut-banks in the valley walls varies considerably, but itseldom exceeds 80 m. Cutbanks are difficult towork because they are often nearly vertical, but inmany areas, broad shelves occur near the waterlineduring the low water periods of the dry season.

The common image of the Amazon Basin as avast, waterlogged swampland is highly overrated,and mostly erroneous, because the areas subjectedto flooding are almost all within the entrenched riv-er valleys of the basin and approximately 80 per-cent of the basin is never flooded. The fact thatthere are large areas subject to flooding is a resultof the large size of many of the Amazonian riversand because these rivers have formed large flood-plains in broad valleys carved into largely uncon-solidated sediments. In Amazonia, larger rivers in-evitably have broader floodplains, which results infewer cutbanks into older rock formations on thesides of the valley. More exposures of valley wallsper river kilometer are to be found in the smallerrivers and headwater streams that have very smallor no floodplains because their channels have yetto reach base level and they are constantly cuttinginto the rock formations that form the valley sides.Unfortunately, these rivers are logistically the most

2 m Science Series 40 Campbell: Santa Rosa Introduction

difficult to work because they shallow rapidly dur-ing the dry season and are subject to blockage bytree falls, both factors that can make them impass-able.

The Santa Rosa local fauna was discovered inJuly 1995 in a cutbank of only a few meters heightalong the Rıo Yurua, Peru (Rıo Jurua in Brazil)(Campbell et al., 1996). About 400 kg of sedimentswere screen-washed as a test that year, using ordi-nary fiberglass window screen. The concentratedmatrix produced hundreds of microvertebrate spec-imens. A return visit to the Santa Rosa locality in1998 resulted in a much larger sample. Only asmall portion of the matrix from that expeditionhas been picked for microfossils, but that portionhas proven to be as productive as the 1995 sample.The remainder of that matrix will undoubtedlyyield thousands of specimens. However, the smallparticle size of this matrix, all of which is less than4 mm diameter, precludes finding additional mam-malian specimens bearing more than a single tooth,except for possibly the very smallest of the taxarepresented. There will undoubtedly be additionalreports of new mammalian taxa from the SantaRosa local fauna in the future, but it is probablyfair to say that they will represent the rarer, al-though potentially more interesting, components ofthe paleofauna.

The Santa Rosa local fauna contains a few largevertebrates, represented mostly by bone fragments,but its significance lies in its rich and diverse mi-crofauna. Although mammals are the thrust of thecollection of papers presented in this volume, fish,amphibians, reptiles, and birds are also present inthe microfauna. The majority of the taxa presentare represented only by isolated teeth, but theynonetheless give us a very informed first glimpse ofan early Tertiary, low latitude, tropical vertebratefauna of South America.

The most abundant, and probably diverse, of allthe vertebrates are the fish. At least five families arepresent (J. Lundberg, personal communication),represented by thousands of small to tiny teeth. Thefish fauna is being described separately and will un-doubtedly provide important information for pa-leoecological reconstruction at the site. Reptiles arerepresented by crocodilians, turtles, and possiblylizards. There are many hundreds of small, isolatedteeth of crocodilians of various types, and they rep-resent the second most common group of verte-brates in the paleofauna, after the fish. Unfortu-nately, these numerous teeth cannot be identified tospecies, so they do not provide more than generalinformation on environmental conditions. Isolatedturtle osteoscutes are not uncommon in the paleo-fauna, but they are also insufficient for identifica-tion. A few fossil specimens are considered to bepossible lacertilian jaws, but as yet, they have notbeen studied. The rarest vertebrates in the paleo-fauna are amphibians and birds. Both of these lattergroups are to date represented by fewer than a doz-

en specimens, and, unfortunately, these are toofragmentary to provide taxonomic characters thatwould assist in identification. That both groups arerepresented, however, does indicate the breadth ofthe fauna preserved at Santa Rosa and its potentialfor preserving additional components of the mam-malian paleofauna not yet identified.

The only invertebrate fossils found at Santa Rosato date are crab claws, probably of taxa belongingto the family Pseudothelphusiidae (J. Martin, per-sonal communication). These fossils are fairlyabundant and display a wide size distribution.Charophytes are also reported to occur in the de-posits at Santa Rosa (M. Aldana, personal com-munication; personal observation), but they havenot been described in the literature.

The papers in this volume describe the mamma-lian portion of the Santa Rosa local fauna; its im-portance for understanding early Tertiary mam-malian faunas of low latitude, tropical South Amer-ica; and, on a larger scale, its significance to ourunderstanding of mammalian evolution in the earlyTertiary in South America. This research would nothave been possible without the enthusiastic coop-eration and participation of our Peruvian col-leagues at the Instituto Geologico, Minero, y Me-talurgico (INGEMMET). On behalf of all the au-thors in this volume, I thank them. To facilitate theresearch on the Santa Rosa local fauna, specimenshave been catalogued into the collections of theVertebrate Paleontology Section of the Natural His-tory Museum of Los Angeles County. Followingcompletion of the research on the Santa Rosa localfauna, all holotypes and representative series of alltaxa will be returned to Peru.

LITERATURE CITEDCampbell, K.E., Jr., C.D. Frailey, and L. Romero-Pittman.

1996. The first Paleogene fauna from the AmazonBasin and its bearing on the first occurrence of ro-dents in South America. Journal of Vertebrate Pale-ontology 16(3)(Suppl.):25A.

. 2000. The late Miocene gomphothere Amahu-acatherium peruvium (Proboscidea: Gomphotheri-idae) from Amazonian Peru: Implications for theGreat American Faunal Interchange. Instituto Geo-logico Minero y Metalurgico, Serie D: Estudios Re-gionales, Boletın 23:1–152.

Frailey, C.D. 1986. Late Miocene and Holocene mam-mals, exclusive of the Notoungulata, of the Rıo Acreregion, western Amazonia. Contributions in Science374:1–46.

Kay, R.F., and C.D. Frailey. 1993. Large fossil platyrrhinesfrom the Rio Acre local fauna, late Miocene, westernAmazonia. Journal of Human Evolution 25:319–327.

Kummel, B. 1948. Geological reconnaissance of the Con-tamana Region, Peru. Geological Society of Ameri-ca, Bulletin 59:1217–1265.

Willard, B. 1966. The Harvey Bassler collection of Peru-vian fossils. Bethlehem, Pennsylvania: Lehigh Uni-versity, 255 pp.

Received 2 July 2003; accepted 29 December 2003.


The Paleogene Santa Rosa Local Fauna of Amazonian Peru:Geographic and Geologic Setting

Kenneth E. Campbell, Jr.,1 Carl David Frailey,2 andLidia Romero-Pittman3

ABSTRACT. The Santa Rosa local fauna of eastern Peru is the first low-latitude, low-elevation, early Tertiary vertebrate paleofauna known for the tropical regions of SouthAmerica. The paleofauna is derived from coarse-grained fluvial deposits within the mas-sive series of clay-dominated continental red beds of eastern Peru. Here, we summarizethe geologic history and rock formations of the western Amazon Basin and describe thegeologic features of the Rıo Yurua and the locality where the Santa Rosa local faunawas found. The age of the Santa Rosa local fauna is estimated as late Eocene on thebasis of the stage of evolution of the marsupials and rodents in the fauna. This age isconsistent with that of the Paleocene–Eocene Yahuarango Formation, one of the oldestof the continental red beds of eastern Peru.

RESUMEN. La fauna local de Santa Rosa, del Peru oriental, es la primera paleofaunamamıfera del Terciario temprano de baja latitud y baja elevacion conocida para lasregiones tropicales de Sudamerica. Aquı la historia geologica y las formaciones rocosasdel oeste de la cuenca Amazonica son resumidas y las caracterısticas geologicas del RıoYurua y de la localidad de Santa Rosa son descritos. La edad estimada de la fauna localde Santa Rosa es eoceno tardıo basado en el estado de evolucion y la combinacion detaxa presentes en la fauna. Una edad eoceno conforme a la edad paleoceno–eoceno dela Formacion Yahuarango, una de las mas antiguos de las capas rojas continentales delPeru oriente.

INTRODUCTION

The Santa Rosa local fauna of eastern Peru is thefirst Paleogene vertebrate fauna known from thetropical lowlands of the Amazon Basin (Campbellet al., 1996). Unfortunately, a significant factor lim-its our placement of the Santa Rosa paleofauna intoan overall scheme for the evolutionary developmentof the South American vertebrate fauna: we cannotplace the paleofauna within a stratigraphic contextwith any great degree of certainty. The requisitegeologic data that would facilitate dating of thefauna are just not currently available for the remoteregion in which the Santa Rosa locality lies. Itmight even be argued that the historic lack of Pa-leogene vertebrate faunas from the lowland tropicalregions of South America has contributed to the


2 Johnson County Community College, Overland Park,Kansas 66210; Museum of Natural History, University ofKansas, Lawrence, Kansas 66045; Research Associate,Vertebrate Paleontology, Natural History Museum of LosAngeles County, 900 Exposition Boulevard, Los Angeles,California 90007. Email: [email protected]

3 Instituto Geologico, Minero, y Metalurgico (INGEM-MET), Av. Canada 1470, San Borja, Apartado 889, Lima41, Peru. Email: [email protected]

overall lack of detailed knowledge about the geo-logic evolution of the Amazonia. We include herea brief summary of the geologic history of easternPeru because the region may be unfamiliar to many.This also allows us to place the discussion of thegeology of the Rıo Yurua into a broader regionalcontext. The stage of evolution of the most abun-dant mammalian taxa provides an approximate agefor the fauna, and from that point, we discuss thepossible stratigraphic context of the Santa Rosa lo-cal fauna.

GEOGRAPHIC SETTING

The Santa Rosa locality occurs on the left bank ofthe Rıo Yurua, at 98299390 S, 728459480 W (Fig. 1),Provincia Atalaya, Departamento de Ucayali, Peru.The site is approximately 7.0 km (by air) north ofthe small community and army post of Breu (alsoknown under the older name of Tipishca), approx-imately 7.5 km (by air) due south of the Peru–Brazilborder, and 230 km (by air) east-southeast fromPucallpa (on the Rıo Ucayali), the closest Peruviancity of any size. The elevation of the airstrip at Breuis approximately 235 m above mean sea level,whereas the elevation at the Santa Rosa locality isestimated to be approximately 215 m above meansea level on the basis of Sheet 2352 [Santa Rosa]of the Peru 1:100,000 topographic series. The fossil

4 m Science Series 40 Campbell et al.: Geology and Geography

Figure 1 Map showing the location of the Santa Rosa fossil locality in eastern Peru. Inset. Upper left: map showing theSanta Rosa locality on a detailed outline of the Rıo Yurua river channel as it appears on the Peru 1:250,000 Rıo Torollucmap, 1984, based on LANDSAT imagery. Inset. Upper right: map of Peru, showing location of lower map.

locality takes its name from the small village ofSanta Rosa, which is located on the right bank ofthe river 2.0 km due south of the locality and be-tween the locality and the community of Breu far-ther upriver (Fig. 1). It should be noted that onSheet 2352 [Santa Rosa] (first edition; undated) andon the corresponding geologic map, the village ofSanta Rosa is misplaced. It is mapped on the curveof the river exposing the Santa Rosa locality, on thebank opposite the locality, but it should be mapped2.0 km due south. Another anomaly is that on boththe 1995 and 1998 expeditions, our GPS unitplaced the locality at 98299390 S, but calculating thelatitude from Sheet 2352 [Santa Rosa] places the

locality at 98289390 S. We cannot explain the oneminute difference, and we do not know which isthe correct figure.

LOCALITY DESCRIPTION

The fossiliferous deposits occur in a long (.100 m)cut bank. About 5 m of red clay (thickness depen-dent on river level) are unconformably overlain byunconsolidated, buff sands (Figs. 2, 3). Above thesands, which appear to be of fairly recent originand probably represent terrace deposits, the forest-ed riverbank slopes steeply away from the river toa higher terrace level. The fossiliferous horizons are

Science Series 40 Campbell et al.: Geology and Geography m 5

Figure 2 View of the Santa Rosa fossil locality from the east. A coarse-grained fossiliferous lens can be seen at the feetof C. Frailey (left) and J. Wigmore (opening sample bag). The primary site of the locality is located near the west end ofthe outcrop, approximately over the head of Wigmore. The top of the Yahuarango Formation can be seen at the shoulderlevel of L. Romero-Pittman (walking on right).

a number of small fluvial deposits, which we inter-pret as bed load deposits (Figs. 4, 5), composed ofcoarse sand of mixed composition, calcareous nod-ules, and small clay pebbles. The largest fossilspresent are turtle shell fragments and fragments oflarger vertebrate limb bones. The coarse fluvial de-posits lie in horizons, or lenses, that are all less than30 cm thick, are of variable lengths, and, in theoutcrop, usually feather out laterally. These lenses,and particularly the longer, more pebbly horizonsthat can be meters in length, slope from left to rightacross the face of the outcrop (Figs. 3, 4). Whetherthis slope reflects the original slope of deposition oris the result of uplift is unknown. The lenses maybe closely stacked vertically, or they may occur asquite isolated lenses in the mass of red clay. On theleft side of the outcrop, a number of the lenses endabruptly, as if at a fault, or at the edge of a large,subsequent paleochannel incision (Fig. 3). Thesmall area exposed and the lack of detailed geologicdata for the region make interpretation of structur-al geology at the site difficult.

The fossils from Santa Rosa are in situ and showno signs of being reworked. It is unlikely that themany small, fragile dentigerous mandibles andmaxillae could have withstood reworking, andmany of the delicately edged, isolated microfossilsshow no wear. In addition, many of the microclastsand microfossils show signs of having undergone

extreme shear stresses that resulted in many visiblefractures, especially in the microclasts (Fig. 6) andthe mammal teeth, as can be seen in photographsof specimens in the papers that follow. Many of theclasts (e.g., Figs. 6A–B) have intersecting planes offractures that can be interpreted to delineate thegreatest principal stress axis. As a consequence ofthese fractures, and because there was no subse-quent natural cementation at the fracture sites, inthe process of recovery many of the fossils breakinto fragments following fracture lines, just as theywould have broken had they been reworked intransport from deposits in higher terrain. The frac-turing of the clasts and fossils indicate that the fos-siliferous beds must have been subjected to extremepressure, most probably resulting from the multiplephases of tectonism that gave rise to the Andes (see‘‘Geologic Setting’’). The presence of similar micro-fractures has not been observed in any of the Mio-cene microfaunas we have obtained from Amazon-ia.

GEOLOGIC SETTING

GEOLOGY OF EASTERN PERU

The modern Amazon Basin formed as a conse-quence of the Andean Orogeny, a series of orogenicepisodes that began in the Cretaceous (Megard,1984, 1987; Sebrier and Soler, 1991). These epi-


Figure 3 The most thoroughly sampled portion of the Santa Rosa locality as seen from across the Rıo Yurua. The slightlysloping conglomeratic lenses that bear the fossils stand out in relief against the surrounding clays in the center and towardthe right of the photograph. The lenses terminate abruptly on the left, either at a fault or at the edge of a large paleo-channel. The top of the Yahuarango Formation is near the top of the photograph (in shadows).

Figure 4 A view of the primary fossiliferous portion of the Santa Rosa locality, including the cleft seen in the right ofFigure 3. The top of the Yahuarango Formation is at the top of this photograph. Note the decrease in the quantity ofthe conglomeratic lenses to the right. The water level in this view is about 1 m higher than in Figure 3, covering the shelfseen in the latter figure.

sodes were a result of the subduction of the NazcaPlate under the western continental margin ofSouth America as the continent moved westward,away from Africa. As the Western Cordillera of theAndes began to emerge in the Cretaceous, a flood

of clastic sediments began accumulating in thebroad area between the shield rocks of the Brazilianand Guianian highlands and the rising Andes. Theestimated thickness of the accumulated Tertiary de-posits varies from 1700 to 2000 m (Jenks, 1956;


Figure 5 A close-up of the excavated face of the outcrop reveals that, in this area, the conglomeratic lenses are dominantover the pure clay horizons, one of which is seen directly above the scale running across the width of the photograph.This view is of a small area just to the left of the cleft in the right of Figure 3, where the conglomeratic lenses appear tobe quite thick.

Figure 6 A high-pressure environment resulted in the fracturing of many microclasts and mammal teeth found in thefossiliferous horizons of the Yahuarango Formation. A, Well-rounded bone fragment; B, fish spine; C, possible seed; D,elongated pebble. Scale 5 5 mm.

Ruegg, 1956) to 10,000 m (Campbell and Burgl,1965). Kummel (1948) gives a thickness of about3000 m for the post-Cretaceous Contamana Groupin north central Peru, which includes all but thevery youngest of the Tertiary formations. The Ter-tiary deposits thin eastward as they overlap thebasement rocks of the Brazilian shield. The Peru-vian Andes appear to have experienced six relative-ly short phases of compression as the subducting

plate squeezed the continental crust against theshield rocks (Megard, 1984, 1987; Sebrier and So-ler, 1991). Each new phase saw the zone of com-pression move eastward as the continental crustwas shortened, with the last major compressionalphase, which formed the Sub-Andean Thrust andFold Belt, occurring during the late Miocene (;7Ma; the Quechua III phase of Megard, 1984; Par-do, 1982).


Figure 7 Geologic map of a portion of the western Amazon Basin and the Eastern Cordillera of Peru showing locationof the Santa Rosa locality (1) on the Rıo Yurua (Rıo Jurua in Brazil). The Cushabatay (2) Mountains form the spur ofthe Eastern Cordillera leading into the Contamana (3) and Contaya (4) Mountains and the Sierra de Divisor. The anticlinalstructure of the long, folded and faulted belt running east-southeast from the Eastern Cordillera is clear, although themap is not particularly detailed because of its scale. The city of Pucallpa (5) is on the Rıo Ucayali. Geologic units identifiedon this map: J2–3 5 Jurassic, K1 and K2 5 Cretaceous, N 5 Neogene, Q1 5 Pleistocene, Q2 5 Holocene, P 5 Permian,PG 5 Paleogene, S-D 5 Silurian/Devonian. Adapted from Choubert and Faure-Muret (1976).

The last major marine incursion into AmazonianPeru from the Pacific occurred in the Oligocene andresulted in the deposition of the Pozo Formation.The lowland connection between the Amazon Ba-sin and the Pacific Ocean through which this ma-rine incursion passed, and which probably servedfor the later westward drainage of the basin, is gen-erally assumed to have been closed in the Miocene,although there are no data that bear on the actualtime of closure. The most likely time for this eventwas during uplift of the Eastern Cordillera duringthe latter part of the Quechua I phase of the An-dean orogeny [bracketed between 20 and 12.5 Maby Megard (1984, 1987); placed at 25–17 Ma byNoble et al. (1990)] or, less likely, early in the Que-chua II phase [placed at 12–8 Ma in northern Peruby Noble et al. (1990) and at 9.5–8.5 Ma by Me-gard (1984, 1987)]. The closure of the western con-nection may have created a freshwater ‘‘interiorsea’’ within the basin (Putzer, 1984), with drainagenorthward into the Caribbean until the eastwarddrainage pattern of the Amazonian river systemwas established (Hoorn et al., 1995). During the

late Tertiary the Amazon Basin was never far abovesea level, and transgressions of marine/brackish wa-ter from the Caribbean region may have reachedcentral Amazonia in the Miocene, and possiblyeven in the Pliocene (Ihering, 1927; Sheppard andBate, 1980; Nuttall, 1990; Hoorn, 1994; Rasanenet al., 1995). Sedimentation within lowland Ama-zonia is currently very limited (Dumont et al.,1990; personal observation), with most modern riv-ers flowing in entrenched valleys.

In addition to the Sub-Andean Thrust and FoldBelt, several anticlinal and domal structures in theAndean Foreland Basin of eastern central Peru wereuplifted during the late Miocene Quechua III phaseof the Andean orogeny (Pardo, 1982; Megard,1984), and most of these uplifted areas occur in afairly narrow, curving structural zone extendingeast-southeast from the Eastern Cordillera (Fig. 7).The northernmost structural feature that forms partof this structural zone is a spur of the Eastern Cor-dillera, the Cushabatay Mountains. The Conta-mana and Contaya Mountains, separated from theCushabatay Mountains by the Rıo Ucayali, contin-


ue the structural zone southeastward and form theconnecting link to the Sierra de Divisor. The Sierrade Divisor is a long group of narrower anticlinalstructures that run south-southeastward and thatform the basis of the border between Peru and Bra-zil for a distance of almost 240 km. These struc-tures are fault bounded, in part, on their easternflanks, and transverse faulting has also been re-ported. Portions of these uplifted areas remain to-day as topographic highs in the Amazonian low-lands. Cretaceous strata are reported to crop outwhere the tops of some of these structures havebeen eroded away (e.g., Kummel, 1948; Valenzuelaand Zavala, 1998), and rare instances of even olderrocks are recorded (Fig. 7). On the geologic mapof Peru (INGEMMET, 1975), the Santa Rosa lo-cality is located about 50 km south of the mappedlimit of the series of uplifted anticlinal structuresthat comprise the Sierra de Divisor. However,ONERN (1980) and Valenzuela and Zavala (1998)describe Paleogene strata closer than 50 km fromthe Santa Rosa area.

Almost all of the Tertiary deposits of westernAmazonia are referred to as the ‘‘Capas Rojas,’’ or‘‘red beds,’’ and four formations of red beds arerecognized in Peru. Of these, three [the Huchpa-yacu Formation (Cretaceous–?Paleocene), the Ya-huarango Formation (Paleocene–Eocene), and theChambira Formation (Miocene) (Fig. 8)] are allsimilar in that they are lithologically monotonous,consisting of alternating beds of sands, silts, mud-stones, unconsolidated clays, and occasional evap-orites. Stratification is irregular, with abundantcross-bedding, channeling, and lenticulation(Ruegg, 1947, 1956; Kummel, 1948). There is littlesize sorting of the clasts, which are mineralogicallyheterogeneous and unaltered, suggesting both anAndean origin and rapid deposition. This interpre-tation is reinforced by an overall decrease in clastsize to the east. Despite the gross similarities amongthese formations, Ruegg (1947, 1956) states that,with experience, the lower Tertiary (Paleogene), orpre-Pozo Formation, red beds can be distinguishedin the field from the upper Tertiary (Neogene), orpost-Pozo Formation, red beds. We have also ar-rived at this conclusion.

A fourth ‘‘red bed’’ formation is the Miocene Ipu-ruro Formation, the lithologic characteristics ofwhich set it off from the three older red bed for-mations (Kummel, 1948). There is a possibility thatthe lithologic change between the red beds of theChambira Formation and those of the more vari-able, darker sediments of the younger Ipururo For-mation is a result of the blockage of the westwardflow of the Amazonian drainage and its reorgani-zation into a northward, and, later, west- to east-flowing drainage system, but this remains to beconfirmed. This suggestion assumes that the drain-age reorganization would have dramatically affect-ed environments of deposition within the basin.

Two key lower Tertiary formations (Pardo andZuniga, 1976) occur within the red bed sequence

and serve as markers to separate the red bed for-mations in many areas in northern and easternPeru. The first of these marker beds is the CasaBlanca Formation (Kummel, 1948), which consistsof about 60 m of massive, white, quartzose sand-stones at its type locality. This formation is of uni-form lithology, has a wide areal distribution in east-ern Peru, and separates the upper Cretaceous–?Pa-leocene Huchpayacu Formation and the Paleocene–Eocene Yahuarango Formation. The Casa BlancaFormation signifies a period when the source areafor sediments being deposited in western Amazoniashifted from the emerging Andes to the eastern Bra-zilian shield. This formation is considered essen-tially isochronous because it is rather thin, but areliable age for this unfossiliferous formation hasnot been established.

The second key formation is the ?Eocene–Oli-gocene Pozo Formation (Williams, 1949), a 400-m-thick marine sequence at the type locality, with fos-sils of foraminiferans, ostracodes, and marine mol-lusks. This formation separates the Paleocene–Eo-cene Yahuarango Formation from the MioceneChambira Formation. It is well known from Ec-uador south to Yurimaguas, Peru (Pardo and Zu-niga, 1976), and it has been identified in well coresin the Rıo Ucayali valley (Martınez, 1975). How-ever, it is less well known farther to the southeast,and it was not reported in the Santa Rosa area byONERN (1980) or Valenzuela and Zavala (1998).As pointed out by Flynn and Swisher (1995), thereare few instances in the Cenozoic of South Americawhere fossiliferous marine strata intertongue withfossiliferous terrestrial sequences; thus, this impor-tant means of dating strata and correlating amongdisparate regions is seldom available.

Although the sequence of Tertiary formations ineastern Peru is fairly well known, the lack of nu-merical age dates for most of these formations re-mains a problem for interpreting the ages of theircontained faunas. This problem becomes clearwhen comparing the stratigraphic charts of Guiza-do (1975) (Fig. 8), Pardo and Zuniga (1976) (Fig.8), and Valenzuela and Zavala (1998) (chart notincluded here), and when comparing these chartswith the age assignments of Kummel (1948) for theformations of eastern Peru. There are datable se-quences within these older strata (Kummel, 1946;Ruegg, 1947, 1956), but to our knowledge therehave been no efforts to date them by modern ra-diometric methods. Securing numerical age datesfor stratigraphic horizons near the key marker for-mations is necessary for anchoring the stratigraphiccolumn of eastern Peru and for resolving conflictinginterpretations related to the chronologic occur-rence of these beds. The only numerical age datesfor strata of lowland Amazonia are those of Camp-bell et al. (2001), and these pertain only to theyoungest of the Neogene strata.

On the basis of the stage of evolution of the mar-supials and rodents, the age of the Santa Rosa localfauna is placed at ?Eocene [Mustersan South Amer-


Figure 8 These two charts illustrate the recognized Cenozoic stratigraphic column in eastern Peru. The two key markerbeds—the white sandstone Casa Blanca Formation and the marine Pozo Formation—are seen to bracket the Paleocene–Eocene Yahuarango Formation, although this is clearly the case only in the northern portion of the Ucayali Valley. Theage of the Santa Rosa local fauna, on the basis of taxa present and their stage of evolution, is estimated to be Eocene,which would place the fossiliferous deposit within the Yahuarango Formation. Above, adapted from Guizado (1975).Below, adapted from Pardo and Zuniga (1976). (Both translated from Spanish).

ican Land Mammal Age (SALMA) (see Campbell,2004)]. If this age estimation is correct, it can behypothesized that the fauna must come from theYahuarango Formation, which lies between the twokey marker formations cited above. Although Par-do and Zuniga (1976) illustrate the YahuarangoFormation as passing from the mid-Eocene into theOligocene, Guizado (1975) suggested that deposi-tion of the overlying Pozo Formation was initiatedin the late Eocene, thereby limiting the YahuarangoFormation to the Eocene. Cruz et al. (1997) also

dated the Pozo Formation to the Eocene on the ba-sis of its abundant microfossils. Cruz et al. (1997)assigned a Paleocene age to the Yahuarango For-mation on the basis of the presence of age-indica-tive charophytes, whereas Valenzuela and Zavala(1998) suggested an age of upper Maestrichtian tomid-Eocene. Thus, primarily on the basis of micro-fossil data unknown to earlier workers, more recentauthors have tended to assign an older age to theYahuarango Formation.

Termination of deposition of the Pozo Formation


and the beginning of deposition of the ChambiraFormation is placed at about the Oligocene–Mio-cene boundary (Martinez, 1975; Pardo and Zuniga,1976; Cruz et al., 1997). If this age assignment iscorrect, the Chambira Formation is far too youngto host the Santa Rosa local fauna.

One difficulty in referring the Santa Rosa localityto the Paleocene–Eocene Yahuarango Formationwith complete confidence is that if the Pozo For-mation does not occur as far southeast as the SantaRosa locality, then the fossiliferous deposits theremight have been deposited contemporaneouslywith the Pozo Formation as a continuation of theYahuarango Formation into the Oligocene, or as anolder portion of the Chambira Formation. If thiswere the case, then it might be more difficult toseparate the Yahuarango Formation from the over-lying Chambira Formation because of lithologicsimilarities, and it might also be more difficult todetermine the time of initiation of deposition of theChambira Formation in southeastern Peru. Al-though the Pozo Formation has been reported insubsurface logs of wells drilled in the Rıo Ucayalivalley to the west of the Santa Rosa locality at ap-proximately the same latitude (Martinez, 1975),and Cruz et al. (1997) reported it in the Sierra deDivisor about 200 km north-northwest of SantaRosa, we know of no data that document its dis-tribution close to the Santa Rosa locality.

GEOLOGY OF THE RIO YURUA

The only published studies of the geology of theRıo Yurua are those of ONERN (1980) and Val-enzuela and Zavala (1998). ONERN’s (1980) geo-logical work concentrated on geomorphology andstratigraphy, whereas that of Valenzuela and Za-vala (1998) concentrated on stratigraphy with sum-maries of geomorphology and historical geology.We disagree with several aspects of these interpre-tations, as detailed below and as summarized in ourgeneralized stratigraphic column for the region(Fig. 9).

The red beds comprising the stratigraphic se-quence along the Rıo Yurua were identified byONERN (1980) as the lower Tertiary HuayabambaFormation (also known as the lower red beds, low-er Capas Rojas, lower Puca, Huchpayacu Forma-tion, and Yahuarango Formation), the upper Ter-tiary Chambira Formation (also known as the up-per red beds, upper Capas Rojas, and upper Puca),and the Ipururo Formation. We refer to the Ya-huarango Formation those strata cropping outalong the Rıo Yurua that ONERN (1980) referredto the lower Tertiary Huayabamba Formation andthe upper Tertiary Chambira Formation.

Valenzuela and Zavala (1998) interpreted the de-posits cropping out along the Rıo Yurua to com-prise, in their terms, the Miocene Chambira For-mation, the Mio-Pliocene Ipururo Formation, andthe capping Plio-Pleistocene Ucayali Formation.The Chambira Formation is described as the basal

unit exposed along the river, and they include thelocality of the Santa Rosa local fauna in this for-mation, even providing a photograph of the SantaRosa locality (Valenzuela and Zavala, 1998; foto16). Thus, what they refer to as the Chambira For-mation, we interpret as the Yahuarango Formation.

Along the Rıo Yurua, the Yahuarango Formationconsists of moderately indurated massive clays andmudstones, with occasional bedding visible. Thinhorizons a few centimeters thick of unconsolidated,variously colored fine sand of mixed compositionare frequently encountered. Also frequently en-countered are calcareous and hematitic conglom-eratic lenses, the latter appearing to replace the for-mer. The calcareous conglomerates appear to cor-respond to channel or bed load deposits that varyin thickness from 2 to 30 cm, separated by beds ofclay. The channel and bed load deposits compriseclasts that range from very fine-grained to 12 cmin longest dimension, with fine and medium-sizedgrains predominating. Most of the larger clasts cor-respond to calcareous concretions, clay balls, frag-ments of bone, and plant matter, with much claymatrix and calcareous cement. Small pebbles arealso found. The calcareous fluvial deposits are thefossiliferous horizons at the Santa Rosa locality.

At the top of the Yahuarango Formation alongthe Rıo Yurua is a clear unconformity that sepa-rates it from the overlying formation (Fig. 9). Thisunconformity is found throughout western Ama-zonia, and it was called the Ucayali Peneplane byKummel (1948). The Ucayali Peneplane apparentlyformed as a consequence of regionwide erosion fol-lowing the uplift associated with the early late Mio-cene Quechua II Andean compressional event (seereview of this topic in Campbell et al., 2000). Kum-mel (1948) referred to the strata overlying the un-conformity as the Ucayali Formation, but for rea-sons of priority, we prefer the designation of Madrede Dios Formation (Campbell et al., 2000). BothONERN (1980) and Valenzuela and Zavala (1998)call the formation overlying the unconformity theIpururo Formation, the next youngest formation af-ter the Chambira Formation in the recognizedstratigraphic sequence of eastern Peru. However,from their descriptions of the lithology of the for-mation, it is clear that it is the Mio-Pliocene Madrede Dios Formation. Unfortunately, Valenzuela andZavala (1998) also applied the name Ucayali For-mation to unnamed Plio-Pleistocene terrace gravelsunrelated to the name-bearing formation of Kum-mel (1948).

The Madre de Dios Formation comprises theyoungest part of the stratigraphic column along theupper Rıo Yurua. This is a very complex formation,with numerous gradual and abrupt facies changes.The deposits include soft beds of clay that varyfrom narrowly laminar to massive in form; clayeysilts and silty sands; and cross-bedded, medium- tocoarse-grained, well-sorted, unconsolidated sand-stones. Paleochannels are often clearly visible. Thebasal horizons, which appear to be prograding del-


Figure 9 The general lithostratigraphic column for the Rıo Yurua. The only Quaternary deposits seen along the RıoYurua are those covering the river terraces and the Recent alluvium of the modern floodplains.

taic deposits (i.e., foreset beds) are primarily con-glomeratic, grading rapidly upward into sands andsilts. The basal conglomeratic horizons are oftenvery fossiliferous, producing both micro- and ma-cromammal fossils. On the basis of an 40A/39A ashdate of 9.01 6 0.28 Ma (Campbell et al., 2001)and the fossil taxa found in the basal horizons ofthe Madre de Dios Formation (Frailey, 1986), theage of the older portion of the formation is inter-preted as upper Miocene (Huayquerian SALMA;9–6 Ma). The uppermost horizon of this formationwas dated to 3.12 6 0.02 Ma (Campbell et al.,2001), but this horizon did not crop out along thatportion of the Rıo Yurua studied. For a more de-tailed review of the Madre de Dios Formation, seeCampbell et al. (2000).

SUMMARY

We recognize two formations cropping out alongthe Rıo Yurua at the southern end of the Sierra deDivisor structural zone, as opposed to the three for-mations previously described as being present. Thetwo formations we recognize, which are separatedby the Ucayali Unconformity, are the Paleocene–Eocene Yahuarango Formation and the Mio-Plio-cene Madre de Dios Formation. Recognition of theYahuarango Formation as the fossiliferous stratumis tentative, pending more detailed stratigraphicdata. The fossiliferous horizons producing the San-ta Rosa local fauna are coarse-grained to conglom-eratic bed load or channel deposits that occur with-in the massive red clays of what we interpret to be


the Yahuarango Formation. On the basis of theirstage of evolution, the marsupials and rodents ap-pear to date to the late Eocene, or MustersenSALMA, which is consistent with the age of por-tions of the Yahuarango Formation as presently in-terpreted. The Madre de Dios Formation is identi-fied as cropping out along the Rıo Yurua on thebasis of its characteristic lithostratigraphy.

ACKNOWLEDGMENTS

We thank Ing. Hugo Rivera Mantilla, Director Tecnica ofINGEMMET, and Ing. Oscar Palacios, Director Generalde Geologıa de INGEMMET, who have been very sup-portive of our geological and paleontological studies inPeru. We thank Javier Guerro, Fritz Hertel, Carina Hoorn,and Michael Woodburne for helpful reviews of our man-uscript. The National Geographic Society and John G.Wigmore provided financial support for the field workleading to the discovery of the Santa Rosa local fauna.Winifred Rhodes, Richard Seaver, and John Wigmore pro-vided funding for the 1998 expedition to Santa Rosa.

LITERATURE CITED

Campbell, C.J., and H. Burgl. 1965. Section through theEastern Cordillera of Colombia, South America.Geological Society of America, Bulletin 76:567–590.

Campbell, K.E., Jr. 2004. The Santa Rosa local fauna: Asummary. In The Paleogene mammalian fauna ofSanta Rosa, Amazonian Peru, ed. K.E. Campbell, Jr.Natural History Museum of Los Angeles County,vol. 40, 155–163.

Campbell, K.E., Jr., C.D. Frailey, and L. Romero-Pittman.1996. The first Paleogene fauna from the AmazonBasin and its bearing on the first occurrence of ro-dents in South America. Journal of Vertebrate Pale-ontology 16(3)(Suppl.):25A.

. 2000. The late Miocene gomphothere Amahu-acatherium peruvium (Proboscidea: Gomphotheri-idae) from Amazonian Peru: Implications for theGreat American Faunal Interchange. Instituto Geo-logico Minero y Metalurgico, Serie D: Estudios Re-gionales, Boletın 23:1–152.

Campbell, K.E., Jr., M. Heizler, C.D. Frailey, L. Romero-Pittman, and D.R. Prothero. 2001. Upper Cenozoicchronostratigraphy of the southwestern Amazon Ba-sin. Geology 29(7):595–598.

Choubert, G., and A. Faure-Muret. 1976. Atlas geolo-gique monde. Geological World Atlas. Paris:UNESCO.

Cruz, N. de la, J. Zedano-C., and A. Zapata-M. 1997.Geologıa de los cuadrangulos de San Roque, RıoCallerıa, San Lucas, Pucallpa, Nuevo Utiquinia,Cantagallo y Divisor Yurua Ucayali. Instituto Geo-logico Minero y Metalurgico, Serie A: Carta Geolo-gica Nacional, Boletın 102:1–155.

Dumont, J.F., S. Lamotte, and F. Kahn. 1990. Wetland andupland forest ecosystems in Peruvian Amazonia:Plant species diversity in the light of some geologicaland botanical evidence. Forest Ecology and Man-agement 33/34:125–139.

Flynn, J.J., and C.C. Swisher III. 1995. Cenozoic SouthAmerican Land Mammal Ages: Correlation to globalgeochronologies. In Geochronology time scales andglobal stratigraphic correlation, ed. A. Berggren,D.V. Kent, and J. Hardenbol. Society of Economic

Paleontologists and Mineralogists (SEPM) SpecialPublication, 54:317–333.


Guizado, J. 1975. Las molasas del pliocenico–cuaternariosdel Oriente Peruano. Boletın de la Sociedad Geolo-gica del Peru 45:25–44.

Hoorn, C. 1994. Fluvial palaeoenvironments in the intra-cratonic Amazonas Basin (Early Miocene–early Mid-dle Miocene, Colombia). Paleogeography, Palaeocli-matology, Palaeoecology 109:1–54.

Hoorn, C., J. Guerrero, G.A. Sarmiento, and M.A. Lor-ente. 1995. Andean tectonics as a cause for changingdrainage pattern in Miocene northern South Amer-ica. Geology 23:237–240.

Ihering, H. von. 1927. Die Geschichte des AtlantischenOzeans. Jena, Germany: Fischer, 237 pp.

(INGEMMET) Instituto Geologico, Minero, y Metalur-gico. 1975. Mapa Geologico del Peru. Lima, Peru:INGEMMET.

Jenks, W.F. 1956. Peru. Geological Society of America,Memoir 65:213–247.

Kummel, B. 1946. Estratigrafıa de la region de Santa Cla-ra. Estudio preliminar. Boletın de la Sociedad Geo-logica del Peru 19:133–152.

. 1948. Geological reconnaissance of the Conta-mana Region, Peru. Geological Society of America,Bulletin 59:1217–1265.

Martınez, M. 1975. Tectonica del area Ucayali Central.Boletın de la Sociedad Geologica del Peru 45:61–82.

Megard, F. 1984. The Andean orogenic period and its ma-jor structures in central and northern Peru. Journalof the Geological Society of London 141:893–900.

. 1987. Structure and evolution of the Peruvian An-des. In Anatomy of mountain ranges, ed. J.-P. Schaerand J. Rodgers, 179–210. Princeton: Princeton Uni-versity Press.

Noble, D.C., E.H. McKee, T. Mourier, and F. Megard.1990. Cenozoic stratigraphy, magmatic activity,compressive deformation, and uplift in northernPeru. Geological Society of America Bulletin 102:1105–1113.

Nuttall, C.P. 1990. A review of the Tertiary non-marinemolluscan faunas of the Pebasian and other inlandbasins of north-western South America. Bulletin,British Museum of Natural History (Geology) 45:165–371.

(ONERN) Oficina Nacional de Evaluacion de RecursosNaturales. 1980. Inventario, evaluacion e integra-cion de los recursos naturales de la zona de los RiosAlto Yurua—Breu. Lima, Peru: ONERN.

Pardo, A.A. 1982. Caracterısticas estructurales de la fajasubandina del norte del Peru. In Simposium ‘‘Ex-ploracion Petrolera en las cuencas subandinas de Ve-nezuela, Colombia, Ecuador y Peru,’’ 1–8. Bogota,Colombia: Asociacion Colombiana de Geologos yGeofisicos de Petrol.

Pardo, A.A., and F. Zuniga. 1976. Estratigrafıa y evolu-cion tectonica de la region de la selva del Peru. ParteII. Mesozoico y Cenozoico. Segundo Congreso La-tinoamericano de Geologıa, Memorias. Boletın Geo-logica, Caracas, Publicacion Especial 7:588–608.

Putzer, H. 1984. The geological evolution of the AmazonBasin and its mineral resources. In The Amazon, ed.H. Sioli, 15–46. Dordrecht, The Netherlands: Dr. W.Junk.


Rasanen, M.E., A.M. Linna, J.C.R. Santos, and F.R. Ne-gri. 1995. Late Miocene tidal deposits in the Ama-zonian Foreland Basin. Science 269(5222):386–390.

Ruegg, W. 1947. Estratigrafıa Comparada del OrientePeruano. Boletın de la Sociedad Geologica del Peru20:57–100.

. 1956. Geologıa y Petroleo en la Faja SubandinaPeruana. Simposio Sobre Yacimientos de Petroleo yGas, XX Congreso Geologico International IV,Mexico, 89–139.

Sebrier, M., and P. Soler. 1991. Tectonics and magmatismin the Peruvian Andes from late Oligocene time tothe Present. Geological Society of America, SpecialPaper 265:259–278.

Sheppard, L.M., and R.H. Bate. 1980. Plio-Pleistocene os-tracods from the upper Amazon of Colombia andPeru. Paleontology 23(1):97–124.

Valenzuela-O., G., and B. Zavala-C. 1998. Geologıa delos cuadrangulos de Puerto Putaya 18-p, Rıo Sha-huinto 18-q, Jacaya 19-p, Santa Rosa 19-q, Rıo Breu19-r, Parantari 20-p, Breu 20-q, y Rıo Piquiyacu 20-r. Instituto Geologico Minero y Metalurgico, SerieA: Carta Geologica Nacional, Boletın 109:1–137.

Williams, M.D. 1949. Depositos terciarios continentalesdel valle del Alto Amazonas. Boletın de la SociedadGeologica del Peru, Volumen Jubilar 50(Parte II):1–13.



New Paleogene Marsupials from the Amazon Basin of Eastern Peru

Francisco J. Goin1,2 and Adriana M. Candela1

ABSTRACT. An analysis of 79 fossil specimens, most comprising tiny, isolated upperand lower molars recovered from Paleogene levels near Santa Rosa in the PeruvianAmazon, led to the recognition of eight new genera and eleven new species of extinctmarsupials: Rumiodon inti gen. et sp. nov. (Didelphimorphia, cf. Herpetotheriidae), Pa-tene campbelli sp. nov. (Sparassodonta, Hathliacynidae), Incadolops ucayali gen. et sp.nov. (Polydolopimorphia, Prepidolopidae), Wamradolops tsullodon gen. et sp. nov. (Po-lydolopimorphia, family indeterminate), Hondonadia pittmanae sp. nov. (Polydolopi-morphia, family indeterminate), Perulestes cardichi and P. fraileyi gen. et spp. nov. (Pau-cituberculata, Caenolestidae), Sasawatsu mahaynaq gen. et sp. nov. (Paucituberculata,cf. Palaeothentidae), Kirutherium paititiensis gen. et sp. nov. (Microbiotheria, Micro-biotheriidae), Wirunodon chanku gen. et sp. nov. (order and family indeterminate), andKiruwamaq chisu gen. et sp. nov. (order and family indeterminate). The marsupial thatbest resembles W. chanku is Kasserinotherium tunisiense, from the lower Eocene ofAfrica. Perulestes and Sasawatsu represent early stages in the evolution of the paucitu-berculatan quadrangular upper molar; their combined features confirm that paucituber-culatans and polydolopimorphians do not belong to a natural group. Wamradolops,Incadolops, and Hondonadia add significant information on the evolution of severalmajor polydolopimorphian lineages. A formal suprageneric taxonomy for the order Po-lydolopimorphia, included in a major (unnamed) taxon together with microbiotherians(and diprotodontians?) is proposed. Frugivorous or frugivorous–insectivorous adaptivetypes are dominant among the Santa Rosa marsupials, with W. tsullodon the most abun-dant, comprising almost half of all marsupial specimens. A comparative analysis of theSanta Rosa marsupials with those of other Paleogene South American faunas failed tocorrelate it with any known South American Land Mammal Age. We conclude that theage of this fauna is, most probably, middle to late Eocene, although the possibility ofan early Oligocene age should not yet be discarded.

RESUMEN. El analisis de 79 especımenes fosiles consistentes en su mayor parte enmolares superiores e inferiores de muy pequeno tamano, procedentes de niveles paleo-genos cercanos a la localidad de Santa Rosa, en la Amazonia peruana, permitio el re-conocimiento de once nuevas especies de marsupiales extinguidos: Rumiodon inti gen.et sp. nov. (Didelphimorphia,? Herpetotheriidae), Patene campbelli sp. nov. (Sparasso-donta, Hathliacynidae), Incadolops ucayali gen. et sp. nov. (Polydolopimorphia, Prepi-dolopidae), Wamradolops tsullodon gen. et sp. nov. (Polydolopimorphia, familia indet.),Hondonadia pittmanae sp. nov. (Polydolopimorphia, familia indet.), Perulestes cardichiy P. fraileyi gen. et spp. nov. (Paucituberculata, Caenolestidae), Sasawatsu mahaynaqgen. et sp. nov. (Paucituberculata, cf. Palaeothentidae), Kirutherium paititiensis gen. etsp. nov. (Microbiotheria, Microbiotheriidae), Wirunodon chanku gen. et sp. nov. (ordeny familia indet.) y Kiruwamaq chisu gen. et sp. nov. (orden y familia indet.). El unicomarsupial conocido que recuerda a W. chanku es Kasserinotherium tunisiense, delEoceno temprano de Africa. Perulestes y Sasawatsu representan estadıos tempranos enla evolucion del patron molar de los Paucituberculata; su combinacion de caracteresconfirma que los Paucituberculata y los Polydolopimorphia no constituyen un gruponatural. Los polidolopimorfios Wamradolops, Incadolops y Hondonadia agregan infor-macion significativa sobre la evolucion de varios linajes de este orden. Se propone for-malmente una nueva taxonomıa supragenerica del Orden Polydolopimorphia, el cual seincluye en un grupo mayor (innominado) junto con los Microbiotheria (¿y los Dipro-todontia?). Los tipos frugıvoros o frugıvoro–insectıvoros son dominantes en esta fauna,siendo W. tsullodon el mas abundantemente representado, con casi la mitad de los es-

1 Departamento Paleontologıa Vertebrados, Museo de La Plata, Paseo del Bosque s/n, 1900 La Plata, Argentina. Email:[email protected] and [email protected]

2 Correspondence

16 m Science Series 40 Goin and Candela: Paleogene Marsupials

pecımenes. Un analisis comparativo de los marsupiales de Santa Rosa con aquellos deotras faunas paleogenas sudamericanas no permite correlacionarla con ninguna de lasedades-mamıfero reconocidas para el Paleogeno sudamericano. Sin embargo, se concluyeque la edad de esta fauna es, mas probablemente, Eoceno medio a tardıo, aunque tam-poco deberıa descartarse una edad Oligoceno temprano.

INTRODUCTION

There is a strong bias in our knowledge of SouthAmerican Paleogene land mammals resulting fromtwo main factors: most early Tertiary fossil sites arelocated in the southernmost part of the continent(e.g., Patagonia, Argentina; see Fig. 1), and a largeproportion of the known mammals are of moderateto large size. This bias partially reflects both thehistory of South American paleomammalogy, be-ginning with the Ameghinos’ pioneering studies incentral and southern Argentina, and the collectingtechniques traditionally employed (naked-eyesearch for fossils). This bias is especially importantin regard to South American marsupials for tworeasons. First, as far as living marsupials are con-cerned, the majority of taxa is presently distributedin tropical, warm, and wet regions of South Amer-ica. This suggests that, even though environmentalconditions were probably different from those ofthe present during Paleogene times, an importantdiversity of extinct taxa that is not presently re-corded should be expected in tropical South Amer-ica. Second, most South American marsupials are,and most probably have always been, very small tomedium-sized forms. This is not reflected in theknown Paleogene fossil record of marsupials. Aquick review of the literature on Paleogene mam-mals from Argentina is illustrative. From four Pa-tagonian localities of the Riochican (late Paleocene)South American Land Mammal Age (SALMA; Bajode la Palangana, Canadon Hondo, Cerro Redondo,and Pan de Azucar; see Fig. 1), Marshall et al.(1983) listed several medium-sized sparassodontsand a few polydolopines, which are the largest po-lydolopimorphians. Among Casamayoran SALMA(conventionally, early Eocene, but see below),mammals from Patagonia, which form one of thebest known Tertiary faunas of South America, arefive sparassodont genera, three polydolopines, asquirrel-sized caroloameghiniid, and one didelphi-morphian opossum (Coona pattersoni Simpson,1938, similar in size to the largest of living mar-mosines), and only one, very small, microbiotheriidopossum (Eomicrobiotherium gaudryi Simpson,1964). Mustersan SALMA (conventionally, early tomiddle Eocene, but see below) marsupials from Pa-tagonia are, in turn, very poorly represented, withonly two sparassodonts and one polydolopineknown. Divisaderan SALMA (?early Oligocene)mammals are known mostly from the type locality,Divisadero Largo, in western Argentina. Marsupialremains from this locality are restricted to a fewspecimens referable to the extremely specialized,rat-sized taxa of the family Groeberiidae. Finally,

Deseadan SALMA (middle to late Oligocene) mar-supials from Argentina include three medium-sized(Notogale Loomis, 1914) to very large (Pharso-phorus Ameghino, 1897; Proborhyaena Ameghino,1897) sparassodonts and three rat-sized paucitu-berculatans (see Marshall et al., 1983; Pascual etal., 1996).

In recent years, the use of screen-washing tech-niques applied to fossil prospecting and collectinghas changed our perception of the diversity ofSouth American fossil mammals. For example, inone Patagonian locality, the middle Paleocene LasFlores Formation, the application of these tech-niques helped R. Pascual, M. McKenna, C. Tam-bussi, and A. Carlini recover an impressive varietyof mammal remains, mostly isolated molars, rang-ing from tiny to moderately large-sized specimens.Among the marsupials, there seem to be no fewerthan 30 species representative of all major lineagesof South American metatherians (Goin et al.,1997). The diversity of this fauna suggests that avery significant part of South American marsupialhistory, especially that of the Paleogene, is still tobe discovered.

The second aspect that precludes a better under-standing of mammalian diversity and evolutionduring the Tertiary is the already mentioned lackof tropical, mammal-bearing, stratigraphic levels.In southern South America during the Paleogene,there was a shift in environmental conditions, be-ginning in the Eocene. Commenting on their ‘‘Pre-patagonian Faunistic Cycle’’ (Casamayoran to Div-isaderan SALMAs), Pascual et al. (1996:285–286,italics ours) mention that, ‘‘By its own, the Pata-gonian mammal record testifies to the climatic andrelated environmental shifting trend that occurredthroughout this Cycle: the ratio of hypsodont tobrachydont ‘ungulates’ increased from the early tothe later part. . . . This trend is compatible with theprogressive shift from forested warm habitats to ex-tensive temperate grassland plains. . . . Those ver-tebrates indicative of warm environments disap-peared from middle Patagonian latitudes alreadyfrom the middle part (Mustersan SALMA) of thisCycle. . . . Even most of those ‘ungulate’ speciesrepresenting the end of this cycle (DivisaderanSALMA) recorded in latitudes (338 S) not far fromnorthern Patagonia . . . curiously show primitivestates with respect to their Patagonian counter-parts.’’ Moreover, even the Divisadero Largo mam-mals may represent a relatively marginal fauna inthe whole South American context for that age. Indiscussing why the extremely specialized groeberiidmarsupials were, by that time, solely recorded atDivisadero Largo, without known ancestors or de-

Science Series 40 Goin and Candela: Paleogene Marsupials m 17

Figure 1 Map showing the fossil site near Santa Rosa, ineastern Peru, and other (mostly) Paleogene localities men-tioned in the text (see ‘‘Introduction’’ and ‘‘Discussion’’).1, Laguna Umayo (Peru; latest Cretaceous or earliest Pa-leocene); 2, Tiupampa (Bolivia; Tiupampian SALMA, ear-ly Paleocene); 3, Punta Peligro (Argentina; Peligran SAL-MA, early Paleocene); 4, Itaboraı (Brazil; Itaboraian SAL-MA, middle Paleocene); 5, Yacimiento Las Flores (Argen-tina; Itaboraian SALMA, middle Paleocene); 6, GranBarranca or Barranca Sur, Lago Colhue Huapi (Argentina;late Paleocene to Miocene); 7, Bajo de la Palangana (Ar-gentina; Riochican SALMA, late Paleocene); 8, CanadonHondo (Argentina; Riochican SALMA, late Paleocene); 9,Cerro Redondo (Argentina; Riochican SALMA, late Pa-leocene); 10, Pan de Azucar (Argentina; Riochican SAL-MA, late Paleocene); 11, Lumbrera Formation (Argentina;Casamayoran SALMA, conventionally early Eocene, butsee ‘‘Introduction’’); 12, Antofagasta de la Sierra (Argen-tina; Mustersan SALMA, conventionally middle Eocene,but see ‘‘Introduction’’); 13, Gran Hondonada (Argentina;Mustersan SALMA); 14, Divisadero Largo (Argentina;Divisaderan SALMA, ?late Eocene–early Oligocene); 15,Termas del Flaco (Chile; ‘‘Tinguirirican’’ age, early Oli-gocene); 16, La Venta (Colombia; Laventan SALMA, mid-dle Miocene).

scendants, Simpson suggested that they ‘‘. . .evolved in what are now (and quite likely werethen) the tropics and [are] picked up in our recordonly when [they] spread rather briefly to what wasfor [them] a marginal area’’ (Simpson, 1970:17; cit-ed also in Pascual et al., 1994:257). The recognitionof different climatic–environmental conditions be-tween southernmost South America and northernregions of the continent beginning in the middleEocene implies that future prospecting in the latterwill lead to the recovery of quite different faunas.

These new faunas will probably include taxa rep-resenting generalized forms with respect to contem-porary, related taxa recorded in Patagonia.

In this work, we analyze a new marsupial faunarecovered from Paleogene levels in a fossil localitynear Santa Rosa at a very low latitude in the east-ern Amazonian lowlands of Peru. The fauna pre-sented here consists of 79 specimens, most of themvery small, isolated molars, representing ten genera,eight of which are new, and eleven new species re-ferable to all major lineages (orders) of SouthAmerican marsupials. Some of the new taxa, how-ever, cannot be referred to any known lineage. Thedescription of the new taxa, the analysis of theiraffinities, and the discussion of the probable age ofthe fossil association constitute the main goals ofthis study. As stated below, some of these taxa rep-resent ‘‘missing links’’ in the evolution of SouthAmerican marsupials. Other taxa comprise ex-tremely derived, even aberrant marsupials that maynever have been recorded before and possibly neverwill be again, whereas a few can be positively re-lated to well-known taxonomic counterparts insouthern South America. In this sense, this contri-bution adds knowledge about the regional differ-entiation of South American marsupials during akey period (middle to late Paleogene; see ‘‘Discus-sion’’) of their history.

Relevant to the discussion of the age of SantaRosa mammals (see ‘‘Discussion’’) is the recent con-tribution by Kay et al. (1999), who reported thefirst radiometric age determinations and associatedmagnetic polarity stratigraphy for the Casamayor-an SALMA. Their data indicate that the Barrancansubage of the Casamayoran is much younger thanpreviously thought; that is, 36 Ma instead of 55–50 Ma. Thus, the Casamayoran SALMA would belate Eocene in age, whereas the Mustersan SALMAwould be practically at the Eocene–Oligoceneboundary. This means that no SALMAs are recog-nized for the early and middle, or most, of the Eo-cene. It could be that, whereas the Barrancan Sub-age is late Eocene, the Vacan Subage of the Casa-mayoran SALMA actually represents a fraction ofthe middle Eocene. This, however, has not beenpostulated on the basis of radiometric or paleo-magnetic evidence.

The Santa Rosa locality (Provincia Atalaya, De-partamento de Ucayali, Peru) is located approxi-mately 2 km north of the village of Santa Rosa onthe northwest bank of the Rıo Yurua near the con-fluence between it and the Rıo Beu, at 98299390 S,and 728459480 W (Fig. 1). Campbell et al. (2004)suggest that the fossil-producing deposits at the sitebelong to the Paleocene–Eocene Yahuarango For-mation (also known as the Huayabamba Forma-tion, lower red beds, lower Capas Rojas, lowerPuca, or Huchpayacu Formation). The fossiliferousbeds at Santa Rosa correspond to calcareous, clay-ey channel deposits consisting of mostly fine-grained sediments with horizons containing largeconcentrations of calcareous concretions, bone,


teeth, shell, and plant material. Most of the speci-mens studied in this work came from screen-wash-ing of about 400 kg of matrix from these beds atSanta Rosa in 1995. Several additional specimensresulting from a return trip to the site in 1998 arealso included. Available unsorted concentrate andadditional field work have the potential to producemany dozens of additional specimens. Field work,unusually difficult compared with more ‘‘normal’’paleontological work in higher latitudes, was car-ried out by Kenneth E. Campbell, Carl D. Frailey,Lidia Romero-Pittman, and Manuel Aldana.

ABBREVIATIONS AND TERMINOLOGY

LACM Natural History Museum of Los AngelesCounty, California, USA

MLP Departamento Paleontologıa Vertebra-dos, Museo de La Plata, La Plata, Argen-tina

MNRJ Museu Nacional de Rio de Janeiro, BrazilSALMA South American Land Mammal Age

Upper incisors I1, I2, I3, I4, I5

Upper canine C1

Upper premolars P1, P2, P3

Upper molars M1, M2, M3, M4

Lower incisors I1, I2, I3, I4

Lower canine C1

Lower premolars P1, P2, P3

Lower molars M1, M2, M3, M4

Upper/lower molar(locus unknown) Mx/Mx

Deciduous thirdupper/lower premolar dp3/dp3

Stylar cusps A, B, C, D StA, StB, StC, StD,respectively

Nomenclature for molar cusps and cristae usedin this study is shown in Figure 2. All measure-ments are in millimeters; L 5 length, W 5 width;an asterisk (*) after some measurements indicatesthat they are approximate. When commenting onthe inferred feeding habits of marsupials, we use theterm faunivorous to refer to animals that eat bothinvertebrates and small vertebrates.

SYSTEMATIC PALEONTOLOGY

Supercohort Marsupialia Illiger, 1811

Order Didelphimorphia Gill, 1872

cf. Family Herpertotheriidae Trouessart,1879

Rumiodon gen. nov.

TYPE SPECIES. Rumiodon inti sp. nov.DIAGNOSIS. As for the type, and only known,

species.ETYMOLOGY. From the Quechua rumi,

‘‘stone,’’ and the Greek odontos, ‘‘tooth.’’

Rumiodon inti sp. nov.Figures 3A–B

HOLOTYPE. LACM 144515, isolated right M1.HYPODIGM. The holotype; LACM 140661,

isolated right M?2; and LACM 140687, isolated leftlower premolar.

DIAGNOSIS. Rumiodon inti differs from otherdidelphimorphians in the following combination ofcharacters: upper molars much longer than wide,with reduced protocones and well-developed post-metacristae; centrocrista with labial vertex endingvery close to labial face of the tooth; paracone andmetacone well separated; paraconule vestigial orabsent; metaconule reduced; stylar shelf reduced;StA, StB, StC, and StD, at least in M1, anteropos-teriorly aligned and labiolingually compressed; StBsubequal to StD and larger than StC, which is, inturn, larger than StA.

MEASUREMENTS. LACM 144515: L 5 1.50,W 5 0.90; LACM 140661: L 5 1.59, W 5 1.19;mm.

TYPE LOCALITY. Santa Rosa, Peru. LACM lo-cality 6289.

TYPE HORIZON AND AGE. ?YahuarangoFormation. ?Eocene.

ETYMOLOGY. From the Quechua Inti, ‘‘sun,’’supreme god of the Inca people.

DESCRIPTION. The holotype (LACM 144515)is complete and unworn. It is clearly longer thanwide, with the protocone reduced and the postme-tacrista well developed. The metacone is much larg-er and higher than the paracone; both cusps areplaced well apart. The stylar shelf is high and re-duced in its labiolingual width. The centrocrista ishigh and labially invasive. The preprotocrista con-nects labially with the anterobasal cingulum,whereas the postprotocrista extends basally pastthe metacone’s posterolingual face. All stylar cusps,but especially the anterior ones, are labiolinguallycompressed. StA is small, but clearly distinct andwell projected anteriorly. StB is slightly larger thanStD in size and height; StC is smaller and equidis-tant to both. The paraconule is vestigial or absent;the metaconule is very reduced. StB and the para-cone are very close to each other, and the postme-tacrista is nearly parallel to the dental axis. Theselast two features suggest that the holotype is an M1.

Several features present in LACM 140661, in-cluding the proximity of the paracone and StB, theorientation of the postmetacrista, and the very an-terior placement of the protocone, suggest that thistooth is also an M1. However, its large size andmore robust protocone and metacone lead us toregard it as a possible M2. LACM 140661 hasstrongly worn cusps and crests, especially the post-metacrista. The protocone is quite reduced and lowrelative to the stylar shelf. Consequently, the trigonbasin is reduced, but deep. The protocone and thepre- and postprotocristae are strongly worn,though a small metaconule can be observed on thedistal end of the latter. The metaconule is labiolin-


Figure 2 Nomenclature for molar teeth used in this study, exemplified by a didelphimorphian opossum (A, B) and bythe generalized polydolopimorphian Rosendolops primigenium (from Goin and Candela, 1996) (C). A, C, Upper rightmolars; B, lower right molar. Abbreviations for upper molars: A, B, C, D, stylar cusps A, B, C, and D, respectively; Ac,anterior cingulum; Me, metacone; Mec, metaconule (in C, Ml is metaconule); Pa, paracone; Pac, paraconule; pome,postmetacrista (postparacrista 1 premetacrista 5 centrocrista); popa, postparacrista; Postpc, postprotocrista; Pr, proto-cone; Prepc, preprotocrista; prme, premetacrista; prpa, preparacrista. Abbreviations for lower molars: ac, anterior cin-gulum; co, cristid obliqua; ec, entocristid; en, entoconid; hld, hypoconulid; hy, hypoconid; me, metaconid; pa, paraconid;pc, postcingulum; popc, postprotocristid; pr, protoconid; prpc, preprotocristid. The preprotocristid and the postparacristidform the paracristid; the postprotocristid and the postmetacristid form the metacristid. Not to scale.


Figure 3 Rumiodon inti gen. et sp. nov. LACM 144515 (holotype), isolated right M1: A, occlusal view; B, occlusal-lingual view. Rumiodon sp. LACM 140583: C, left M3; D, M4. Scale 5 1 mm.

gually compressed and develops a posterior crestthat does not end in the base of the metacone but,instead, runs a short distance at the base of thiscusp, parallel to the postmetacrista. The paraconeis clearly smaller and lower than the metacone, andit is very close to and paired with StB. However,the preparacrista does not seem to have connectedwith this stylar cusp at its lingual slope, but insteadat its frontal face. Both the postparacrista and the

premetacrista penetrate deeply to the stylar shelf sothat, in occlusal view, the vertex of the centrocristaends very close to StC. The postmetacrista is quitedeveloped, and it is almost as long as the anteriorwidth of the tooth. The metacone is strong andhigh, about twice the size of the paracone, althoughboth cusps are very worn. StC is small and round-ed. StD is larger and higher than StB, and it devel-ops an anterior and a posterior crest. The antero-


labial end of the tooth is broken, so the presenceor absence of a StA cannot be determined. A nar-row anterior cingulum is present.

LACM 140687 is an isolated left lower premolar,broken in its anterior portion. Apparently it wasunicuspid, with a sharp central cusp. The wholepremolar is strongly compressed labiolingually. Theposterior crest is sharp edged, ending at the base ina narrow posterior cingulum

COMMENTS. Rumiodon inti is a didelphimor-phian opossum not larger than medium-sized livingmarmosines. The strong development of the cuttingcrests of the upper molars and the reduction of theprotocone of M1 probably reflect faunivorous oreven carnivorous feeding habits.

Despite general morphological similarities be-tween R. inti and some Neogene didelphids, severalfeatures suggest closer affinities with Paleogenerather than with Neogene opossum groups (seealso, ‘‘Discussion’’). Rumiodon inti differs fromNeogene marmosines and didelphines in the fol-lowing combination of characters.

1. In both marmosines and didelphines the para-and metaconules, if present, are extremely smallor vestigial; in Rumiodon, the paraconule is ves-tigial, but the metaconule is distinct, althoughreduced, in at least one molar (LACM 140661).

2. The stylar shelf is very narrow in Rumiodon,unlike in most Neogene taxa of Didelphimor-phia. When it is very narrow among the latter,as in species of the carnivorous Sparassocynidaeor in didelphines that have specialized toward acarnivorous diet (e.g., Hyperdidelphys Ameghi-no, 1904), it occurs together with the reductionor loss of the stylar cusps. In Rumiodon, as inseveral Paleogene opossums, the stylar shelf isnarrow, but there is no reduction of the stylarcusps.

3. The centrocrista penetrates the stylar shelf deep-ly labially in Rumiodon, as in the Paleogene De-rorhynchidae and Herpetotheriidae, but not inNeogene opossums.

4. Although an StC is present in R. inti, as in sev-eral Neogene opossums, this cusp is clearly equi-distant from StB and StD, an infrequent featurein Neogene didelphines (but not in some mar-mosines in which StC is present).

5. The extent of the reduction of the protocone andthe great development of the postmetacrista andits orientation relatively parallel to the dentalaxis are infrequently seen in Neogene opossums.

Most of the features mentioned above relate Ru-miodon with several Paleogene species of Didel-phimorphia of both South America and the Hol-arctic region. Particularly interesting are the follow-ing: (a) Several features are shared by Rumiodonand some European Herpetotheriidae of the generaAmphiperatherium Filhol, 1879, and PeratheriumAymard, 1850 (see ‘‘Discussion’’). In the M1 of Am-phiperatherium minutum (Aymard, 1846), for ex-ample, stylar cusps B, C, and D are clearly set apart

from each other, the protocone is reduced, the post-metacrista is well developed, and the metaconule ismore developed than the paraconule, which is ves-tigial or absent (see Crochet, 1980). These speciesalso show several similarities between their lowermolars and those of Rumiodon (e.g., unreduced hy-poconulids, anterobasal cingula well developed, lit-tle reduced paraconids, and entoconids large andvery compressed laterally). Most of these charactersare also shared by R. inti and two European speciesof the genus Peratherium: P. matronense Crochet,1979, and P. monspeliense Crochet, 1979. (b) Thedeep centrocrista, high stylar shelf, distinct StC thatis aligned with StB and StD, and reduction of theprotocone are all features suggesting affinities withDerorhynchus Paula Couto, 1952a, and allied taxafrom the middle Paleocene of South America (Oliv-eira, 1998). In fact, many of these features areshared between South American derorhynchids andEuropean herpetotheriids. A review of the Itabor-aian ‘‘opossum-like’’ marsupials from Brazil ledOliveira (1998) to assign several taxa to the Her-petotheriidae. Marshall (1987) originally regardedderorhynchines as a subfamily of Didelphidae, theDerorhynchinae, which Goin et al. (1999) raised tofamily rank. Affinities between derorhynchids andherpetotheriids have been mentioned previously byone of us (Goin, 1991, 1995). Goin (1991) sug-gested that both lineages, together with several ad-ditional taxa from the Paleogene of South America,might constitute a natural group, having a sistergroup relationship to Neogene and Recent SouthAmerican didelphimorphians.

We find more affinities among R. inti and Paleo-gene opossums not belonging to the Didelphidae(Marmosinae 1 Didelphinae), such as Europeanherpetotheriids and South American derorhynchids,than with any known group of South AmericanNeogene didelphimorphians. As already stated, weare inclined to consider that a relationship with theHerpetotheriidae may be a plausible systematic hy-pothesis. Pending a review of these and other Pa-leogene didelphimorphians, we provisionally re-gard R. inti as a probable herpetotheriid.

Rumiodon sp.

REFERRED SPECIMENS. LACM 140583, leftmandibular fragment with almost complete M3–4

(Figs. 3C–D), and LACM 144518, isolated left tri-gonid.

MEASUREMENTS. LACM 140583: LM3 52.4, WM3 5 1.12, LM4 5 2.12, WM4 5 1.08; mm.

DESCRIPTION. LACM 140583 is a left man-dibular fragment with the two last molars, shownseparately in Figs. 3C–D. The horizontal ramus istypically didelphimorphian, with the origin of theascending ramus behind the M4. Both molars havestrong, well-developed preprotocristids and wideanterobasal cingula that continue posteriad almostto the base of the hypoconids. Paraconids are welldeveloped and have a lingual position. Protoconids


Figure 4 Patene campbelli sp. nov. LACM 140628 (ho-lotype), isolated left M4: A, occlusal view; B, posteriorview. Scale 5 1 mm.

are very high and robust. Metaconids are interme-diate in height between para- and protoconids. Tal-onids are shorter than trigonids. Hypoconulids arewell developed in both molars, and the entoconidsare higher than the hypoconids and very com-pressed labiolingually, developing sharp entocris-tids. Hypoconulids are placed almost immediatelybehind the entoconids. The talonid basin is shallowin both molars, and the hypoconids are not veryprominent. The talonid of M4 is somewhat morereduced than that of M3. The postprotocristids areeven, high, and perpendicular to the talonid plane.At the base of this crest, lingually and behind themetaconid, is a small crest projecting posteriad,having a deep notch at its contact with the ento-cristid. The entoconid of M4 (Fig. 3D) is displacedsomewhat posteriorly with respect to that of M3

(Fig. 3C), in which it forms a distinct wall lingualto the talonid.

LACM 144518, an isolated left trigonid, is evenlarger than the trigonids of LACM 140583. Theprotoconid is broken, but its relative size at thebase indicates that it was larger than the metaconidand the paraconid. The paracristid is well devel-oped, as is the hypoconulid notch. The anterobasalcingulum is moderately developed. At the metacris-tid, it is evident that the cristid obliqua ended an-teriorly at a point below the tip of the protoconid.Behind the metaconid is a small crest, possibly con-nected to the pre-entocristid.

COMMENTS. Although comparable featuressuggest rather strongly that the holotype of Rumio-don inti, LACM 140661, and LACM 140583 be-long to the same taxon, it must be pointed out thatthe upper molars of R. inti are somewhat smallerthan what might be expected from the size of thepreserved M3–4 referred to Rumiodon sp. It could

be argued that the holotype of R. inti is, in fact, adeciduous molar (dp3). However, the paracone ofthat specimen is not fused to StB, a common featurein dp3s of species of the Didelphimorphia, and theproportions of the holotypical specimen are similarto those of the M1s of many other didelphimor-phians.

Although the lower molars in the mandibularfragment assigned here to Rumiodon sp. are largerthan the upper counterparts of R. inti, the cusp andcrest development of both lower molars seem toagree with the lower molar morphology one mightexpect on the basis of the upper molars. The ob-served differences in size may simply reflect intra-specific variability. However, the size differencesseem significant to us. In the absence of more com-plete materials of upper and lower molars, we pro-visionally refer the mandibular fragment with M3–4

to Rumiodon sp.

Order Sparassodonta Ameghino, 1894

Family Hathliacynidae Ameghino, 1894

Patene Simpson, 1935

Patene campbelli sp. nov.Figure 4

HOLOTYPE. LACM 140628, isolated left M4.HYPODIGM. The holotype only.DIAGNOSIS. Differs from Patene simpsoni Pau-

la Couto, 1952b, in its smaller size and by the pres-ence, in M4, of a more reduced metacone and lessdeveloped anterobasal cingulum. It differs from Pa-tene colhuapiensis Simpson, 1935, by its smallersize and by having M4 with persistent anterobasalcingulum, paraconule and metacone more devel-oped, and trigon basin wider.

MEASUREMENTS. L 5 4.92, W 5 2.0; mm.TYPE LOCALITY. Santa Rosa, Peru. LACM lo-

cality 6289.TYPE HORIZON AND AGE. ?Yahuarango

Formation. ?Eocene.ETYMOLOGY. Honoring Kenneth E. Campbell,

codiscoverer of the Santa Rosa local fauna of east-ern Peru, in recognition of his paleontological re-search in tropical South America.

DESCRIPTION. The holotype is an almost com-plete, somewhat worn, isolated left M4. The para-cone is, by far, the highest cusp of the tooth (Fig.4B). It is followed in size by the metacone, whichis very reduced and fused to the paracone at thebase, and then by the paraconule, which is slightlyhigher than the protocone. The metacone is morelingually placed than the paracone, and it developsa narrow labial cingulum that extends from almostthe fusion point with the paracone to the parastylarcorner of the tooth (Figs. 4A–B). The preparacristais straight and sharp. It does not reach the paras-tylar corner, but only to a more lingual point withinthe stylar shelf, at the base of which the anteriorcingulum of the tooth originates. This cingulum is


Figure 5 Incadolops ucayali sp. nov. LACM 144513 (ho-lotype), left upper molar: A, occlusal view; B, lingual view.Scale 5 1 mm.

very narrow and disappears before contacting theanterior crest of the paraconule (Fig. 4A). The par-aconule is moderately developed and, at the pointwhere it fuses with the paracone, it develops a smallcutting crest. The trigon basin is relatively wide.

COMMENTS. This new species of Amazonianmarsupial is referable to the genus Patene, and itappears most closely related to its most generalizedspecies, P. simpsoni from the middle Paleocene–lower Eocene of Argentina and Brazil. Patenecampbelli is a very generalized hathliacynid, asshown by the relatively wide trigon basin, the per-sistence of the paraconule, the presence of a vesti-gial metacone basally fused with the paracone, andthe persistence, although reduced, of an anterobasalcingulum. All of these features are well developedin P. simpsoni, whereas they are almost absent inP. colhuapiensis from the Riochican (upper Paleo-cene) of central Patagonia. Consequently, the newspecies P. campbelli is morphologically intermedi-ate between P. simpsoni and P. colhuapiensis, butcloser to the former. The size of a domestic cat, P.campbelli is the smallest species of this genus, andit is almost the smallest of the known sparasso-donts, with only the oldest representatives of theorder, Allqokirus australis Marshall and Muizon,1988, and Mayulestes ferox Muizon, 1994, bothfrom the lower Paleocene of Tiupampa, Bolivia, be-ing smaller.

Order Polydolopimorphia Ameghino, 1897

Family Prepidolopidae Pascual, 1980

Incadolops gen. nov.TYPE SPECIES. Incadolops ucayali sp. nov.DIAGNOSIS. As for the type, and only known,

species.ETYMOLOGY. Inca, after the Inca civilization

and people, and dolops, a Greek term meaning‘‘ambusher,’’ and a common ending of several po-lydolopimorphian generic names.

Incadolops ucayali sp. nov.Figure 5

HOLOTYPE. LACM 144513, isolated left Mx.HYPODIGM. The holotype only.DIAGNOSIS. Differs from other prepidolopids

(Prepidolops Pascual, 1980a, and Punadolops Goinet al., 1998) in the following features: upper molarswith less difference in height between the protoconeand the paracone–metacone; centrocrista less open,with premetacrista and postparacrista longer andcloser to each other at their labial ends; metaconuleclearly distinct and bearing small postcingulum atits base; anterobasal cingulum present (vestigial orabsent in other prepidolopids), paracone and StBnot as close to each other (especially relative totheir positions in Prepidolops didelphoides Pascual,1980a), and trigon basin relatively larger (especial-ly relative to that of Prepidolops didelphoides).

MEASUREMENTS. L 5 1.50; W 5 1.79; mm.



ETYMOLOGY. Ucayali, for the Departamentode Ucayali, in eastern Peru, wherein is found thetype locality.

DESCRIPTION. LACM 144513 is subtrapezo-idal in shape, wider than long, and with bunoidcusps. The protocone is strong and high, with itsanterior face flat, supporting a large trigon basin.The metaconule is well developed, and it is onlyslightly displaced posterolingually. The premeta-conular crest does not meet the base of the meta-


cone; instead, it points anteriad, toward the trigonbasin. The metacone is larger and higher than theparacone. Stylar cusps B and D are large and con-nected to each other by sharp cristae. The posteriorcrest of StB and the anterior crest of StD are longand sharp edged. At the anterolabial corner of thetooth is a small, anteroposteriorly compressed StAthat is connected with a short and wide anteriorcingulum. The centrocrista is open (Fig. 5A): thepostparacrista ends labially near the posterior crestof StB, whereas the premetacrista ends near the an-terior crest of StD. The preparacrista ends at thelingual base of StA. The postmetacrista is short andpoorly developed.

COMMENTS. The following features that arepresent in Incadolops ucayali are diagnostic of thePolydolopimorphia: pairing of the paracone andmetacone with the StB and StD, respectively, andonly relative enlargement of the StB and StD (seeGoin and Candela, 1996; and below). In turn, theabsence of a ‘‘hypocone’’ (enlarged, displaced meta-conule) and the presence of a large trigon basin aredistinctive features of the Prepidolopidae.

Incadolops is more generalized than other pre-pidolopids (Prepidolops and Punadolops) in thefollowing features: a lower paracone and metaconerelative to the protocone, persistence of stylar cuspsA and E, better developed metaconule, persistenceof the anterior cingulum, less open centrocrista;that is, the labial ends of the postparacrista and thepremetacrista are set closer to each other.

Order Polydolopimorphia

Family Indeterminate

Wamradolops gen. nov.TYPE SPECIES. Wamradolops tsullodon sp. nov.DIAGNOSIS. As for the type, and only known,

species.ETYMOLOGY. From the Quechua wamra,

‘‘child,’’ and the Greek dolops, ‘‘ambusher,’’ a com-mon ending of several polydolopimorphian genericnames; in reference to the small size of the typespecies in comparison to other polydolopimorphianmarsupials.

Wamradolops tsullodon sp. nov.Figures 6–9

HOLOTYPE. LACM 140590, a right maxillafragment with complete P3–M2 (Fig. 6).

HYPODIGM. The holotype; LACM 140584,right horizontal ramus with M1–3 and alveoli of M4

(Fig. 8); LACM 140640, isolated left P3; LACM140601, isolated left ?dp3; LACM 140654, isolatedright M1 (Figs. 9A–C); LACM 140645, isolatedright M1; LACM 140657, isolated left M1; LACM140660, isolated left M1; LACM 140656, isolatedright M1 (Figs. 9D–F); LACM 140672, isolatedfragmentary right M2; LACM 140655, isolatedright M2; LACM 140658, isolated right M2 (Figs.9G–I); LACM 140609, isolated right M2; LACM

140659, isolated left M2; LACM 140648, isolatedM2; LACM 140588, isolated left M3; LACM140626, isolated left M3 (Figs. 7B–C); LACM140587, isolated left M3; LACM 140586, isolatedright M3; LACM 140589, isolated left M3; LACM140608, isolated left M4; LACM 140594, isolatedright P2; LACM 140595, isolated left P2; LACM140591, partial left maxilla with P3–M1; LACM140597, isolated right M1, very worn; LACM140616, fragment of an isolated right M1, withoutthe anterior portion; LACM 144510, isolated leftM1; LACM 140593, isolated right M1 (Fig. 7A);LACM 140585, isolated right M2 (Figs. 7D–E);LACM 140596, isolated right M3; LACM 140592,isolated left M3 (Figs. 7F–G); LACM 140625, iso-lated left M4; LACM 144512, isolated left M4.

DIAGNOSIS. Upper molars strongly heterodont,with large, posterolingually displaced metaconule;M1, and to a lesser extent M2, with very reducedtrigon basin as a consequence of the more labialand posterior position of protocone; protoconeslightly lower than labial cusps; St?C, intermediate,but far labial to StB and StD, present only in M1;and (preprotoconal?) cingulum strong, anterolin-gual to protocone, especially in M1.



MEASUREMENTS. See Table 1.ETYMOLOGY. From the Quechua tsulla, ‘‘un-

equal,’’ and the Greek odontos, ‘‘tooth,’’ in refer-ence to the strong heterodonty of the molar teeth.

DESCRIPTION. Specimens LACM 140594 andLACM 140595 are P2s, a right and left, respective-ly. The anterior root of P2 is more compressed thanthe posterior root, although it supports the largerand taller cusp of the tooth, which has a long an-terior edge. The posterior root supports a verymuch shorter lobe that also forms a cusp, which isconnected with the anterior cusp by means of asharp crest. The anterior and posterior lobes of P2

are separated by vertical ridges on both the lingualand labial sides, with the lingual ridge being higher.The posterior face is flat, and the enamel appearsthicker on the labial face than on the lingual face.The P2 is smaller than the P3.

The holotype, LACM 140590, is a right maxillawith complete P3–M2 (Fig. 6). In the P3, the anteriorand posterior lobes are similar in size, and the lin-gual groove that separates them is wide and deep.Where the cusps merge, they form a continuous an-teroposterior ridge, very compressed labiolingually.

The M1 is much wider than the M2, and it showsthe greatest differences with respect to the gener-alized tribosphenic pattern. It is suboval in occlusalview, and the protocone is very close to the para-and metacone, being oddly placed between and lin-gual to them. Consequently, it does not take partin the anterolingual corner of the tooth. The trigonbasin is almost absent. A strong ?preprotoconalcingulum develops in this corner and continues be-


Figure 6 Wamradolops tsullodon gen. et sp. nov. LACM 140590 (holotype), partial right maxilla (bone not shown) withcomplete P3–M2: A, lingual view; B, occlusal view. Abbreviations: me, metaconid; ml, metaconule; pa, paraconid; pc,?preprotoconal cingulum; pr, protoconid; stb, stylar cusp B; st?c, possible stylar cusp C; std, stylar cusp D. Scale 5 1 mm.

low and lingual to the protocone. There are notraces of StA or StE. St?C (see below) is present andfound on the labial side of the tooth. On the otherhand, StB and StD are much more lingually placedon the stylar shelf, being opposite and paired to theparacone and metacone, respectively. A pair of veryshort crests connect the paracone with StB (the pre-paracrista?), and StB with St?C. The postparacristais well developed. It does not connect immediatelywith the labial end of the premetacrista, but doesso instead by means of a short ridge. In turn, StDdevelops anterior and posterior crests, the formerconnecting with the premetacrista, whereas the lat-ter connects with the postmetacrista at the posteriorside of the tooth. The metaconule is placed in theposterolingual corner of the tooth; it is smaller andlower than the proto- and metacone, and it devel-ops a short anterior crest toward the protocone. Insome of the preserved molars, the preprotoconalcingulum reaches the base of the metaconule pos-teriorly. The whole molar is bunoid, with a shortcrown and very low cusps, subequal in height. Inshort, the odd shape of the M1 is mainly a result ofthe unequal location of the stylar cusps (quite labial

St?C, whereas StD and StB are lingually displaced,almost paired to the metacone and paracone, re-spectively) and to the very unusual location of theprotocone (Figs. 6, 7A).

The M2 is clearly smaller than M1. In the M2, theSt?C is absent; StB and StD have a more labial po-sition within the stylar shelf than in M1, and themetaconule is proportionally larger than in M1. Thepreprotocrista connects with a long cingulum that,running along the anterolabial half of the tooth,continues on the labial side almost to the base ofStD. The preprotoconal cingulum is somewhat lessdeveloped than in M1. The metaconule is not unitedto the protocone by a crest; instead, a lingual flexseparates the metaconule and protocone. A vesti-gial postprotoconal cingulum links the metaconuleand protocone at their bases. Posteriorly and labi-ally, the cingulum develops a crest that reaches thebase of the metacone. The paracone is somewhatsmaller than StB, and it is firmly attached to thatcusp. The preparacrista is absent. The metacone isslightly larger and taller than StD, and it is locatedslightly posterior to StD. The centrocrista is open.The postmetacrista is very short. The premetacrista


Figure 7 Wamradolops tsullodon gen. et sp. nov. LACM 140593, right M1: A, occlusal view. LACM 140626, right M3:B, occlusal view; C, labial view. LACM 140585, right M2: D, occlusal view; E, anterior view. LACM 140592, left M3:F, occlusal view; G, anterior view. Scale 5 1 mm.


Figure 8 Wamradolops tsullodon gen. et sp. nov. LACM 140584, right mandible with M1–3: A, labial view; B, occlusalview. Scale 5 1 mm.

is longer and oriented toward the apex of StD. Asthe position of the protocone in this molar is more‘‘normal,’’ that is, more lingually placed, the trigonbasin is wider than that of M1.

The M3 is somewhat smaller than the M2, al-though essentially similar to the latter. It differsfrom the M2 in the larger size of the protocone; inhaving the anterior face of the tooth more planar,whereas the posterior side is more rounded; by theweaker development of the metaconule; and by thepresence of a continuous and broad lingual cingu-lum between the protocone and the metaconule.There is no labial cingulum, and the protocone iseven more ‘‘normally’’ placed than in the M2.

The M4 is smaller than the M3 and subtriangularin outline because of the absence of a metaconule.The StB is completely fused to the paracone. It ismuch taller and robust than StD, which is appar-ently fused to the metacone. The protocone is a lowcusp, and the trigon basin is proportionally wide.

The only preserved lower premolar is LACM140640, a left P3. It is a large and robust tooth,much larger than the M1. The anterior half of theP3 is much more compressed than the posteriorhalf, and its anterior edge is not very sharp. The

posterior half develops a cutting edge at its apex.The posterior face of this tooth is almost flat.

The horizontal ramus is thick and robust, with avery deep masseteric fossa and the ascending ramusbeginning just behind the M3. As the ramus de-creases in height from back to front, molars areincreasingly higher from M4 to M1; thus, the top ofthe tooth row appears horizontal, as is also the casein the Bonapartheriidae.

The M1 is a relatively high-crowned tooth, withthe trigonid very compressed laterally and the tal-onid short and subequal in width to the trigonid.Its labial side is more than twice as high as thelingual side. This feature, known as ‘‘unilateralhypsodonty’’ in several rodents, is similar to, butmuch more pronounced than, that seen in prepi-dolopid or bonapartheriid polydolopimorphians.The anterior and posterior faces of M1 are flat. Theprotoconid is very high and pointed, and its ante-rior and posterior cristae are aligned. The metaco-nid is much smaller and fused at its base to theprotoconid. The paraconid is very small, not con-nected to the straight preprotocristid, and in a lin-gual position, almost on the anterior side of thetooth. In several unworn first molars, a small crest,


Figure 9 Wamradolops tsullodon gen. et sp. nov. LACM 140654, right, unworn M1: A, labial view; B, occlusal view; C,posterior view. LACM 140656, right, worn M1: D, labial view; E, occlusal view; F, posterior view. LACM 140658, rightM2: G, labial view; H, occlusal view; I, posterior view. Scale 5 1 mm.


Table 1 Measurements (mm) of upper and lower molars of Wamradolops tsullodon gen. et sp. nov. L 5 length, W 5width.

Measurement (mm)

Specimen

M1

L W

M2

L W

M3

L W

M4

L W

Upper molarsLACM 140590LACM 140591LACM 140616LACM 140597LACM 140593

1.201.12

1.081.12

1.681.481.521.441.44

0.84 1.20

LACM 140585LACM 140596LACM 140592

0.96 1.200.840.92

1.281.16

Lower molarsLACM 140584LACM 140660LACM 140654LACM 140656LACM 140645

1.201.161.08

0.760.800.881.040.76

1.00 0.72 0.84 0.68

LACM 140657LACM 140658LACM 140655LACM 140659LACM 140609

1.16 0.881.121.040.961.12

0.640.800.680.68

LACM 140589LACM 140588

0.840.96

0.880.80

LACM 140587LACM 140626

0.96 0.800.80 0.64

resulting from the fusion of the para- and metaco-nid slopes, can be observed on the lingual face ofthe protoconid. The postprotocristid is atypical. In-stead of connecting directly with the metaconid,thus forming, in occlusal view, an angle with thepreprotocristid, it extends along the posterior sideof the protoconid and then turns abruptly towardthe metaconid. The entoconid is conical and higherthan the hypoconid. The talonid basin is very nar-row. The hypoconulid is well developed and placedposterolingually to the entoconid. A narrow post-cingulum extends along the posterior side of thetooth. There is a vestigial anterobasal cingulum.

The M2 is smaller than M1, and the difference inheight between the higher labial and the lower lin-gual side is less than in M1. In the trigonid, theparaconid is very compressed anteroposteriorly asa consequence of the shortening of the trigonid onits lingual face, and it is closer to the metaconidthan in M1. The anterior side of the tooth is notflat but, instead, shows a concavity known as thehypoconulid fossa. The preprotocristid is shorterand directed downward, and the protoconid andthe metaconid are subequal in height and more sep-arate from each other than they are on the M1. Thetalonid basin is somewhat broader than on M1, andthe entoconid and hypoconid are subequal inheight. The hypoconulid is centered on the poste-

rior side of the tooth, and it is more compresseddorsoventrally. The postcingulum is also broaderthan that of the M1. There is a slightly differenti-ated anterobasal cingulum.

The M3 is subquadrangular in outline and limitedto four cusps: the metaconid, which is apparentlyfused with the paraconid, the protoconid, the en-toconid, and the hypoconid. The protoconid is low-er than the metaconid, and the hypoconid is largerand taller than the entoconid. The trigonid is no-tably anteroposteriorly compressed, forming, infact, a crest transverse to the dental axis (the post-protocristid). There is a short anterobasal cingulumlabial to the protoconid that continues posteriadbasal to it. The talonid basin is proportionally larg-er than in the anterior molars. The hypoconulid isvery reduced.

The M4 is probably represented by LACM140608. It is a small tooth, clearly longer thanwide, both at the trigonid and at the talonid. Themetaconid is taller than the protoconid, and, at itsanterior face, it is partially fused to the paraconid.The entoconid is somewhat taller and stronger thanthe hypoconid. The hypoconulid is very small andlow, and it occupies a central position on the pos-terior face of the talonid. The talonid basin is pro-portionally very wide.

LACM 140601 seems to represent an incomplete


left dp3. Its morphology is closely similar to that ofM1, except for the greater proximity of the ento-conid to the trigonid and the vestigial size of themetaconid. Consequently, the trigonid is shaped al-most exclusively by the protoconid and a strongpreprotocristid.

COMMENTS. Among all the polydolopimor-phians compared with Wamradolops tsullodon, thenew taxon shows affinities with several groups thathitherto were known exclusively from the SouthAmerican Paleogene: Prepidolopidae, Bonaparth-eriidae, Groeberiidae, Polydolopidae (but restrictedto Polydolpinae s.s.; see below), and, belonging toan undetermined family, Roberthoffstetteria Mar-shall, Muizon, and Sige, 1983. Affinities with thefirst two families seem particularly close. In somefeatures, Wamradolops seems to be an intermediateform between Prepidolops and BonapartheriumPascual, 1980a. In the bonapartheriids, as in themost specialized prepidolopids, the upper and low-er molars are strongly heterodont. In the uppers,the metaconule forms a well-developed hypocone-like structure, and the para- and metacone are ba-sally fused to StB and StD, respectively. In turn, theanterior lower molars, M1–2, are notably com-pressed labiolingually to the point where there ispractically no talonid basin, and they are relativelyhigh crowned (Fig. 9). Finally, the mandibular ra-mus is very high (i.e., dorsoventrally deep), and thetwo last molars, M3–4, are placed on the ascendingsurface of the ramus together with the origin of themasseteric crest (Fig. 8A). Upper and lower pre-molars are deeply modified as cutting blades, eventhough they develop a moderate talon at their lin-gual sides. In the Prepidolopidae the molars aresomewhat less heterodont and more low crowned,and the mandibular ramus is robust and thick, butdorsoventrally shallow. The upper molars haveonly a weakly developed, and sometimes absent,metaconule without forming a ‘‘hypocone.’’ Thelower molars are shorter, less compressed, and pre-serve a slightly reduced talonid basin. Finally, thepremolars, although hypertrophied and sharpedged, are somewhat more generalized than thoseof Bonapartherium. In Wamradolops, there is amixture of characters of both families, as well assome other features, both primitive and derived,that are exclusive to this genus. Among the primi-tive characters, the persistence of postparacrista,pre- and postmetacrista, and a stylar cusp (St?C) inM1 is noteworthy. Among the derived characters isthe peculiar position of the protocone, especially inthe M1, in that it is very close to the para- and themetacone and almost leaves no space for the trigonbasin (see above). In addition, a peculiar lingualcingulum extends across the base of the protocone.

In other features, Wamradolops also seems to beintermediate between Prepidolops and Bonaparth-erium. For example, a well-differentiated metaco-nule is in the position of a ‘‘hypocone,’’ but it is lessdeveloped than in Bonapartherium; the trigonid ofthe M1 is very compressed labiolingually, much

more than in Prepidolops, although not reachingthe degree of compression seen in Bonapartheriumwhere the trigonid cusps are indistinguishable. Themandibular ramus is robust, broad, and short, asin Prepidolops, although the last lower molars(M3–4) are placed in an ascending manner followingthe origin of the masseteric crest, as in Bonaparth-erium. The upper third premolar is very com-pressed, as in Bonapartherium, but without the lin-gual ‘‘talon’’ characteristic of the P3 of the latter.

In short, although it is difficult to decide to whichfamily Wamradolops belongs, it seems that the newAmazonian taxon is closer to prepidolopids andbonapartheriids, especially to the latter, than to anyother known polydolopimorphian. However, sev-eral features show the resemblance of Wamradol-ops to other South American Paleogene marsupialsas well. These are as follows.

1. One of the most unusual derived features pres-ent in the upper molars, and especially in thefirst upper molar, of W. tsullodon is the some-what posterolabial ‘‘shifting’’ of the protocone(Figs. 6, 7A). As a consequence of this, the tri-gon basin is reduced relative to the size of themolar. This same feature is also present, thoughless extreme, in all upper molars of Roberthoff-stetteria nationalgeographica Marshall, Muizon,and Sige, 1983, a peculiar marsupial from thelower Paleocene (Tiupampian SALMA) of Boliv-ia, originally regarded by Marshall and Muizon(1988) as a caroloameghiniid. Roberthoffstetter-ia also shows a somewhat lingual position of StBand StD with respect to StC. The former are, asa consequence, relatively close to the paraconeand metacone, respectively (Figs. 6B, 7A), butnot as close as in Wamradolops.

2. The ‘‘shifting’’ of the protocone as mentionedabove is also present in the first two upper mo-lars of Groeberia minoprioi Patterson, 1952,from upper Eocene (Divisaderan SALMA) levelsin western Argentina. As in Wamradolops,Groeberia has its protocone clearly displaced la-bially in M1, less displaced in M2, and almost ina ‘‘normal’’ position in M3. As a consequence,in the first two upper molars of Wamradolopsand Groeberia, the most lingual cusp is the ‘‘hy-pocone,’’ or displaced metaconule. Finally, bothWamradolops and groeberiids have teeth with avery thick enamel layer (see Figs. 9D–F). Thisfeature is also present, but to a lesser extent, inbonapartheriids. Several derived features incommon between Wamradolops and the recent-ly described, basal argyrolagoid Klohniacharrieri Flynn and Wyss, 1999, from lowerOligocene levels of central Chile are alsorelevant at this point, as discussed below (see‘‘Discussion’’). These and other considerationsargue in favor of the polydolopimorphianaffinities of the groeberiids (and of argyrolagidsand patagoniids as well, see below).

3. Several of the above-mentioned derived features


Figure 10 Hondonadia pittmanae gen. et sp. nov. LACM 140615 (holotype), isolated upper left molar: A, occlusal view;B, anterior view. Abbreviations: me, metaconid; pa, paraconid; stb, stylar cusp B. Scale 5 1 mm.

are also present in polydolopinepolydolopimorphians. These features include theshort, dorsoventrally shallow and thickhorizontal ramus, strong heterodonty, bunoidcrowns with very low cusps, the presence of astrong labial cingulum in the most anteriorupper molars and, in the anterior upper molars,an unusually sloped labial face as a consequenceof the lingual displacement of StB and StD. Infact, several of the most notable derived featuresof Wamradolops may constitute theplesiomorphic condition for the aberrantlyspecialized molar pattern of polydolopines. Forinstance, the ‘‘shifted’’ protocone, together withthe strongly developed preprotoconal cingulum,could anticipate the trilobed lingual pattern ofthe upper molars of polydolopines. And themost labial stylar cusp present in the first uppermolar of Wamradolops may not be homologousto StC of other marsupials; instead, it might bean accessory cusp that has emerged from thelabial cingulum. The latter is the case with thelabial accessory cusps and cuspules present inpolydolopines.

Leaving aside the question as to what lineage W.tsullodon belongs, this taxon is important for ourunderstanding of the Polydolopimorphia as awhole. The assemblage of derived features presentin the dentition of Wamradolops strongly suggeststhat it, plus numerous other taxa such asRoberthoffstetteria, Gashternia Simpson, 1935;prepidolopids; bonapartheriids; epidolopines;polydolopines; groeberiids; patagoniids; andargyrolagids belong to a monophyletic group (see‘‘Discussion’’).

Hondonadia Goin and Candela, 1998Hondonadia pittmanae sp. nov.

Figure 10HOLOTYPE. LACM 140615, an isolated, bro-

ken left Mx.

HYPODIGM. The holotype and LACM 144516,a partial ?left Mx.

DIAGNOSIS. Differs from Hondonadia feruglioiGoin and Candela, 1998, by having paracone un-reduced, but rather high and styliform, subequal toStB; metacone larger than paracone and positionedeven more labially.

MEASUREMENTS. LACM 140615: L 5 1.52*;mm.



ETYMOLOGY. In honor of our colleague LidiaRomero-Pittman, of INGEMMET, Peru, codiscov-erer of the Santa Rosa local fauna and collector ofthe first marsupial specimen from Santa Rosa.

DESCRIPTION. The holotype is highly frag-mented, as it lacks almost the entire lingual halfand part of the stylar shelf behind StB and labialto the metacone. In occlusal view, both the anteriorand posterior edges of the tooth converge lingually,suggesting that if there had been any metaconule,it was not developed enough to change the sub-triangular pattern of the tooth. Only the anteriorhalf of the stylar shelf is preserved, where a smallbut conspicuous StA and a tall, pointed, and sub-oval in section StB can be observed (Figs. 10A–B).On the posterior side of this cusp is a small cuttingcrest, with a minute cusp on its apex, almost fusedto the StB. The paracone is in front of the StB; it iswell developed, subequal in height, and very closeto this cusp. The preparacrista connects with theStA. The postparacrista seems to meet the preme-tacrista in a very labial point, immediately behindthe StB. The metacone is larger and taller than theparacone, and it is much more labially placed thanthe latter. Both the paracone and the metacone arestrongly asymmetric in their labial and lingualsides. Both have convex lingual sides descending


smoothly toward the protocone (Fig. 10B), but onthe other hand, the labial side of both cusps is pla-nar, almost concave. Labial to StB is a relativelyundeveloped cingulum. The anterior cingulum ofthe tooth is well developed, extending linguallyalong the base of the paracone.

COMMENTS. The twinning of the paraconeand StB and, almost surely, of the metacone andStD, constitutes a polydolopimorphian synapomor-phy (Goin and Candela, 1996). Among represen-tatives of this order, the few assessable features ofthe Santa Rosa species agree well with those ofHondonadia, whose type species, H. feruglioi,comes from Mustersan SALMA (now recognized aslate Eocene/early Oligocene; Kay et al., 1999) levelsat Gran Hondonada, in central Patagonia (Goinand Candela, 1998). Derived features shared by H.feruglioi and the Amazonian species include: uppermolars with trapezoidal outline in occlusal view,preparacrista connecting with StA, open centrocris-ta, metacone more labially placed than the para-cone, and paracone somewhat reduced. In turn, al-most all derived features of both species of Hon-donadia suggest close affinities with the generalizedpolydolopimorphian Rosendolops primigeniumGoin and Candela, 1996. That is, the labial sidesof the para- and metacone are flat, almost concave;in both genera, the anterior cingulum of the uppermolars is well developed, and StB has a posterior,but not an anterior, crest; the preparacrista con-nects with StA, and the metacone is more labialthan the paracone. On the other hand, the generadiffer in that Hondonadia has its paracone less re-duced; the centrocrista is, apparently, not as openas in Rosendolops; and StA is more developed. Thelabial displacement of the metacone (Fig. 10A) withrespect to the paracone is a unique, derived feature,incipiently present in Rosendolops and more obvi-ously developed in Hondonadia, especially in H.pittmanae.

In our original description of the holotypical, anduntil now only known, species of the genus, H. fer-uglioi, we regarded it as a ‘‘?Polydolopimorphia’’(Goin and Candela, 1998:81). Even though we not-ed its similarities with Rosendolops, the large sizeof StD (reduced or subequal to StB in other poly-dolopimorphians, larger than StB in many pauci-tuberculatans) raised doubts about its ordinal as-signation. The recognition of new taxa of general-ized paucituberculatans in the Santa Rosa faunasheds much light on the early phases in the evolu-tion of representatives of both orders (see below),and confirms the assignment of Hondonadia to thePolydolopimorphia. Additionally, several upperand lower molars referable to a new, still undescrib-ed species of Hondonadia have been recovered re-cently from ?post-Mustersan–?pre-Deseadan levelsof the southern cliffs of Lake Colhue Huapi, in cen-tral Patagonia (Goin and Candela, 1997). Thesebetter preserved specimens have a set of derivedfeatures that add decisive evidence in favor of the

polydolopimorphian affinities of this genus (seealso Goin et al., 1998b).

Rosendolops primigenium was originally, andtentatively, referred to the Prepidolopidae by Goinand Candela (1996). Its affinities with Hondonadia,as well as the new information provided by the gen-eralized prepidolopid Incadolops ucayali from San-ta Rosa, raises doubts regarding its familial assign-ment. In our present state of knowledge, we preferto regard both Rosendolops and Hondonadia asbelonging to an as yet undetermined family of Bon-apartherioidea (see Table 5).

Order Paucituberculata Ameghino, 1894

Family Caenolestidae Trouessart, 1898

Perulestes gen. nov.TYPE SPECIES. Perulestes cardichi sp. nov.INCLUDED SPECIES. Perulestes cardichi sp.

nov. and P. fraileyi sp. nov.DIAGNOSIS. Perulestes gen. nov. differs from all

other genera of caenolestids in the following fea-tures: lower molars with shorter and narrower an-terobasal cingula, M1 with protoconid much higherthan metaconid and paraconid more reduced, M2–4

with trigonids having metaconids less anteriorlydisplaced and with very small paraconids, hypo-conulids reduced. Upper molars with anterior cin-gula and paracone present, though very reducedand basally fused to the StB in M1–2 and absent inM4; StB clearly larger than StD in M1, though sub-equal in height in M2–4; metacone paired with StDbut clearly visible; postmetacrista and StA persis-tent.

ETYMOLOGY. Peru, for Peru, the country oforigin of the two known species of this genus, andthe Greek lestes, meaning ‘‘robber’’ or ‘‘pirate’’ and,by extension, ‘‘carnivorous,’’ a common ending forgeneric names of paucituberculatan marsupials.

COMMENTS. See comments under Perulestescardichi sp. nov.

Perulestes cardichi sp. nov.Figures 11, 12

HOLOTYPE. LACM 140649, a complete, iso-lated right M1 (Figs. 11A–B).

HYPODIGM. The holotype; LACM 140606, al-most complete, isolated left M2 (Fig. 12); LACM140604, almost complete, isolated right M2;LACM 140652, isolated right M2 with talonid bro-ken; LACM 149364, isolated left M2; LACM140603, isolated right trigonid of M3; LACM140605, isolated left M3; LACM 140599, isolatedright talonid of M4; LACM 140610, isolated leftM?3, broken in its lingual half; LACM 140624,fragment of left M1 with only its parastylar corner;LACM 140650, isolated right M1, complete, butvery worn; LACM 140686, fragment of left M2

with only parastylar corner; LACM 149363, iso-lated left M2; LACM 140703, isolated, very wornleft M2; LACM 140613, complete, isolated right


Figure 11 Perulestes cardichi gen. et sp. nov. LACM 140649 (holotype), right M1: A, occlusal view; B, lingual view.LACM 140613, right M4: C, occlusal view; D, lingual view. Abbreviations: me, metaconid; ml, metaconule; pa, paraconid;pr, protoconid. Scale 5 1 mm.

M4 (Figs. 11C–D); LACM 140618, partial right M4

with only posterior half; and LACM 140614, par-tial left M4 with only posterior half.

DIAGNOSIS. Perulestes cardichi sp. nov. differsfrom the other species of the genus, P. fraileyi sp.nov., in its smaller size and by having lower molarswith hypoconulid less developed and less promi-nent posteriorly and metacone less displaced pos-teriorly with respect to StD of upper molars.

MEASUREMENTS. See Table 2.



ETYMOLOGY. Honoring Ing. Augusto Car-dich, Peruvian Professor of Andean Archaeology atthe Museo de La Plata, for his contributions to theknowledge of the South American past.

DESCRIPTION. Perulestes cardichi is similar insize to Pichipilus centinelus Marshall and Pascual,


Figure 12 Perulestes cardichi gen. et sp. nov. LACM140606, left M2: A, occlusal view; B, labial view. Scale 51 mm.

1977, from the middle Miocene of Patagonia. Onthe basis of the preserved remains, the reduction inthe size of the tooth row from front to back (i.e.,larger to smaller, respectively) is less marked thanin other caenolestids. No incisors, canines, premo-lars, or first molars of this species are known. Inthe M2, the trigonid is much shorter and somewhatnarrower than the talonid. The anterobasal cingu-lum is short and narrow. The protoconid is large,occupying almost the whole labial half of the tri-gonid, but much lower than the metaconid. Theprotoconid and the metaconid are aligned almosttransversely to the dental axis, though the meta-conid apex is slightly anterior to that of the pro-toconid. The paraconid is a very small cusp. It iseven lower than the protoconid, and it is positionedsomewhat labially to the base of the anterior slopeof the metaconid, but it is not fused to this cusp.The anterior end of the preprotocristid does notpoint toward the paraconid; instead, it descendsgradually, without forming a notch, until it mergeswith the most anterior portion of the anterobasalcingulum. In one of the specimens (LACM 140606;Fig. 12), a minute crest (?postparacristid) can beseen extending labially almost to the preprotocris-tid. Between the paraconid and the anterior edge ofthe anterobasal cingulum is a small, shallow con-

cavity: the hypoconulid fossa. Because of the pe-culiar position of the paraconid, the anterior faceof the trigonid is more planar than pointed. Behindand lingual to the metaconid is a small, sharp crestconnected to the entocristid. Likewise, behind andlabial to the protoconid, another crest is present,though less developed. The metacristid is flat andtransverse to the talonid basin; its notch is wide andV-shaped in posterior view.

The talonid is broad and elongated, being ap-proximately two-thirds of the total length of thetooth. Its basin is low relative to the trigonid, andlabially it is limited by the cristid obliqua, whichextends past the labial edge of the talonid to endanteriorly at the posterolabial corner of the proto-conid. The hypoconid is broad and low. The ento-conid is notably developed, being very high and lat-erally compressed; consequently, it forms a lingualwall for the entire talonid. Immediately behind andsomewhat labial to the entoconid is a minute hy-poconulid. The posterior face of the talonid has nopostcingulum.

The M3 is subequal in size to M2, and it showsseveral differences with the latter: the trigonid hasits labial side somewhat longer than the lingualside; the protoconid and the metaconid are sub-equal in height; the paraconid is even more reducedthan in M2, and it is positioned more lingually; theanterobasal cingulum is more vertical; the notch atthe metacristid is more open, and the crest is notso perpendicular to the talonid plane but, instead,forms a gentle slope at its union with the talonid;the crest behind the metaconid is proportionallymore developed; the talonid is proportionally short-er than in M2, and the difference in height betweenits basin and that of the talonid is less; the ento-conid and metaconid are more anteriorly placed,with the latter being positioned more medially thanthe protoconid; and the posterior face of the toothis not flat but, instead, is more convex than in M2.

The M4 is clearly smaller than M2–3. The talonidhas the same proportions as in M3, with the follow-ing differences: the cristid obliqua is not parallel tothe dental axis but ends anteriorly at the middlepoint of the base of the protoconid, the entoconidis more styliform in outline, and the hypoconulid isrelatively larger.

The upper molars, which decrease in size fromM1 to M4, are subtrapezoidal in outline, with thelabial side longer than the lingual side and with theanterior face straighter than the posterior, which ismore rounded. Relative to other Paucituberculata,the most distinctive feature of the upper molars isthe persistence, at least in M1–2, of a vestigial para-cone (Figs. 11A–B), which is very low and attachedto the lingual slope of StB. The paracone, thoughvery small, is clearly present in M1, whereas in M2,it seems to be even more fused to the base of StB.The paracone is present in M3, but it is small andpartially fused to StB, almost to the tip. Finally, inthe M4, the paracone is either absent or completelyfused to StB (Figs. 11C–D). In turn, the metacone


Table 2 Measurements of upper and lower molars of Perulestes cardichi and P. fraileyi gen. et spp. nov. L 5 length,W 5 width.

Measurement (mm)

M1

L W

M2

L W

M3

L W

M4

L W

Perulestes cardichi

Upper molarsLACM 140649LACM 140650LACM 149363LACM 140703LACM 140613LACM 140618

2.362.20

2.202.00

1.851.70

2.002.10

1.52 1.881.68

Lower molarsLACM 140606LACM 140604LACM 149364LACM 140652LACM 140605LACM 140603

1.802.201.60

1.121.601.20

1.961.96

1.321.201.20 0.92

Perulestes fraileyi

Upper molarsLACM 140651 2.92

Lower molarsLACM 140653LACM 140600LACM 140622

1.361.281.36

is present in every molar and ends posteriorly in asmall, though visible, postmetacrista directed to-ward the metastylar corner. The metacone is sub-equal in height to the StD and faces this cusp in theM1; in M2–3 it is placed slightly behind it, whereasin M4 it is positioned even more posteriorly.

The M1 shows a short, wide anterior cingulumthat ends labially in a small StA. Labial to this cusp,on the labial side of the tooth is an enamel thick-ening directed back and downward, surroundingStB labially. StB is larger and taller than StD; sharpcrests join both cusps. Below and labial to thesecrests there is a short labial cingulum. There is noStC. StD develops a short posterior crest that joinsthe postmetacrista at the metastylar corner of thetooth. The metacone is slightly lower and subequalin size to StD. On the anterolingual side of thiscusp, a rounded edge runs lingually toward themetaconule. The protocone is low, but wide, andlimits lingually the small trigon fossa. Anteriorlyand labially, it develops a preprotocrista connectedto the paracone. Posteriorly, the postprotocristaconnects with a hypertrophied metaconule, the ‘‘hy-pocone,’’ that forms a major part of the posterolin-gual corner of the tooth. The metaconule is some-what smaller, but taller, than the protocone. It de-velops a postmetaconular crest that extends alongthe posterior side of the tooth and ends basally nearthe metastylar corner. LACM 140650 is an almost

complete, although worn, tooth recognized as aright M1.

LACM 149363, LACM 140686, and LACM140703 are identified as probable M2s. The secondspecimen is only a fragment of the parastylar cornerof a left molar, and the last is worn down to thevery base of its cusps. Major differences of LACM149363 with the first molar are its smaller size, alarger anterobasal cingulum, and a less developedparacone. The M4 has the anterior face much lon-ger than the posterior face, a consequence of themore labial position of the metaconule. The meta-cone is subequal and very close to StD; it is posi-tioned more posteriorly than in M3. StB is largerthan StD. There is no paracone, or, if present, it iscompletely fused to StB. The anterior cingulum ofthe tooth is straight and does not end labially as adifferentiated StA. The metaconule and the proto-cone are subequal in height. Below and lingual tothe metaconule is an enamel thickening that shapesa narrow cingulum. The metacone and the meta-conule are linked by the respective facing edges ofeach cusp. StD has no posterior crest; consequently,only the postmetacrista reaches the metastylar cor-ner of the tooth.

COMMENTS. Of the two species of Perulestesrecognized here, P. cardichi is better represented,including remains referable to almost the entire up-per and lower molar series. Only specimens of M1


Figure 13 Perulestes fraileyi gen. et sp. nov. LACM140651 (holotype), incomplete right M4, occlusal view.Abbreviations: me, metaconid; pr, protoconid; std, stylarcusp D. Scale 5 1 mm.

are missing, although they are represented in P.fraileyi sp. nov., (see below). Several derived fea-tures present in Perulestes warrant its placement inthe Paucituberculata. In the upper molars, StB andStD are very large and labiolingually compressed.The paracone is reduced and fused basally to StB,thus indicating an early stage of the trend seen inother paucituberculatans where the paracone iscompletely fused to StB. In the lower molars, theparaconid is reduced, the paracristid lacks a notch,and the entoconid is laterally compressed (Fig. 12).Compared with the various lineages of paucituber-culatans, it is evident that Perulestes lacks severalderived features diagnostic of the Palaeothentidaeand Abderitidae, including an extreme reduction orabsence of metacone, hypertrophy of M1, and ab-sence of paraconid in M2–4. Perulestes also lacksseveral derived features diagnostic of Caenolestinaecaenolestids, including upper molars with lowmetaconule, or ‘‘hypocone,’’ which more resemblesa posterolingual cingulum than a cusp, and lowermolars with a dorsoventral compression of the hy-poconulid and with hypoconid high and salient.

Perulestes differs from caenolestids of the sub-family Pichipilinae in having reduced paraconids,the preprotocristid parallel to the dental axis andnot connected to the paraconid, and the anteriorface of the trigonid flat and transverse to the dentalaxis. Interestingly, all of these features are sugges-tive of those already present in the Palaeothentidaeand Abderitidae. Another generalized feature ofPerulestes, compared with the genera of Pichipili-nae, is the placement of the metaconid with respectto the protoconid: both are aligned transverse tothe dental axis, although the metaconid is slightlyanterior. On the other hand, in Pichipilus Ameghi-no, 1890, and Phonocdromus Ameghino, 1894, theanterior location of the metaconid is much moremarked. Finally, the paraconid of Perulestes is notpaired with the metaconid but is notably reducedin size and somewhat medially placed in the trigo-nid. In Pichipilus and Phonocdromus, as in Qui-rogalestes almagaucha Goin and Candela, 1998, aprimitive species of Pichipilinae from the lower Eo-cene of Patagonia, the paraconid is clearly pairedwith the metaconid, so that in the worn molars itis difficult to recognize both cusps independently.On the other hand, as occurs in Pichipilus andPhonocdromus, in Perulestes fraileyi, the paraconidand the metaconid of the M1 are normally separat-ed from each other, so that the trigonid shows anormal prefossid (i.e., trigonid basin). Also, theprotocristid of this molar shows a deep notch, asin the pichipilines.

In short, it is difficult to precisely assign Perules-tes to one of the known subfamilies, or even fam-ilies, of Paucituberculata, mainly because most fea-tures present in Perulestes, especially in the uppermolars, seem to be generalized for the whole order,and in several features, especially in the lower mo-lars, Perulestes is intermediate or transitional be-tween Caenolestinae 1 Pichipilinae caenolestids on

the one hand and Palaeothentidae 1 Abderitidaeon the other (see ‘‘Discussion’’). In overall features,however, Perulestes seems to constitute a general-ized, but peculiar, genus of the family Caenolesti-dae, probably closer to pichipilines than to caeno-lestines.

Perulestes fraileyi sp. nov.Figures 13, 14

HOLOTYPE. LACM 140651, an incompleteright M4, without its anterior half and part of thestylar shelf (Fig. 13).

HYPODIGM. The holotype; LACM 140622,right M1 without talonid (Fig. 14A); LACM140600, right M1 without the talonid; LACM140653, right M?1 without the trigonid (Fig. 14B);LACM 144511, isolated left M3; and LACM144514, isolated right M1.

DIAGNOSIS. Perulestes fraileyi sp. nov. differsfrom the type species of the genus by its larger size,and in having lower molars with hypoconulid bet-ter developed and more prominent posteriorly, and


Figure 14 Perulestes fraileyi gen. et sp. nov. LACM140622, right M1 lacking the talonid: A, occlusal view.LACM 140653, right M?1 lacking the trigonid: B, occlusalview. Scale 5 1 mm.

upper molars with metacone positioned more pos-teriorly with respect to StD.

MEASUREMENTS. See Table 2.TYPE LOCALITY. Santa Rosa, Peru. LACM lo-


Formation. ?Eocene.ETYMOLOGY. Honoring Carl D. Frailey, co-

discoverer of the Santa Rosa local fauna of easternPeru; in recognition of his paleontological work intropical South America.

DESCRIPTION. Only two trigonids are knownfor the M1: LACM 140622 and LACM 140600.

Both trigonids are identical in size and morphology,although LACM 140622 (Fig. 14A) is less wornand forms the basis for the following description.The protoconid is the largest and tallest cusp of thetooth. Together with the paracristid, it representstwo-thirds of the total width of the trigonid. Theparacristid is well developed, lacks a notch and, inocclusal view, is not straight but describes a slightlydescending curve toward its contact with the para-conid. The paraconid is very small and linguallyoriented. The trigonid fossa is open lingually, andthe slopes of the three cusps merge on it. The meta-conid is intermediate in size and height betweenthose of the paraconid (lower) and the protoconid(taller). The metacristid notch is deep. Behind boththe metaconid and the protoconid is a small sharpcrest; the one behind the protoconid is more devel-oped. The anterobasal cingulum is vestigial.

The only preserved talonid assignable to this spe-cies, LACM 140653, also seems to be an M1 on thebasis of its size and proportions (Fig. 14B). In com-parison with the M2–4 of the type species of thegenus, this specimen has less of a height differencebetween the entoconid and the hypoconid; the hy-poconulid is stronger, more projected backward,and positioned more labially. In turn, the entoconiddoes not seem to have developed a sharp entocris-tid. As in the type species of the genus, there is nopostcingulum on the posterior side of the talonidof this specimen.

The isolated left M3 differs from the M3 of Per-ulestes cardichi in that the posterior face of the tri-gonid is more vertical, with the protocristid andparaconid comparatively larger. The paraconid ismore ‘‘normally’’ positioned on the lingual side ofthe trigonid. The paracristid is not straight but, in-stead, forms an almost right angle at the meetingof the preprotocristid and the postparacristid

The only upper molar assignable to this speciesis the holotype (LACM 140651; Fig. 13), which,judging from the preserved portion, has the sameoverall proportions and structure as the M4 of P.cardichi. Aside from its much larger size, the great-est differences with the M4 of the holotypical spe-cies of the genus are the following: the metacone ismore posteriorly placed; StD seems to have a pos-terior crest moderately developed; there is a narrowcingulum lingual to the protocone; the metacone isless twinned to StD; the metaconule is lower thanthe protocone; the premetacrista is still present; andthe tooth has a thicker enamel cover, forming su-perficial ‘‘wrinkles’’ in some places, such as theunion between the protocone and the metaconuleand the space between StD and the metacone. Asin P. cardichi, a small lingual crest, the premeta-conular crest, connects the metacone with the meta-conule.

COMMENTS. Perulestes fraileyi is twice the sizeof the holotypical species of the genus, P. cardichi.Unfortunately, the only element preserved for bothspecies is the M4. The differences, other than size,


Figure 15 Sasawatsu mahaynaq gen. et sp. nov. LACM140668 (holotype), isolated right M3: A, lingual view.LACM 140647, upper molar (M?4): B, occlusal view. Ab-breviations: me, metaconid; pa, paraconid; pr, protoconid;stb, stylar cusp B; std, stylar cusp D. Scale 5 1 mm.

Figure 16 Sasawatsu mahaynaq gen. et sp. nov. LACM140611, isolated left M2: A, occlusal view; B, lingual view.Abbreviations: en, entoconid; me, metaconid; hld, hypo-conulid; hy, hypoconid; pr, protoconid. Scale 5 1 mm.

between the M4 of both species are relatively minorand strengthen their assignment to a single genus.

cf. Family Palaeothentidae Sinclair, 1906

Sasawatsu gen. nov.

TYPE SPECIES. Sasawatsu mahaynaq sp. nov.DIAGNOSIS. As for the type, and only known,

species.ETYMOLOGY. From the Quechua sasa, ‘‘diffi-

cult,’’ and watsu, ‘‘root’’; in reference to its prob-lematic, inferred ancestry.

Sasawatsu mahaynaq sp. nov.Figures 15, 16

HOLOTYPE. LACM 140668, a complete, iso-lated right M3 (Fig. 15A).

HYPODIGM. The holotype; LACM 149366,isolated right M3; LACM 140695, distal portion oflower right incisor (I2); LACM 140607, fragmentof horizontal ramus with intra-alveolar portion ofhypertrophied incisor, M2 complete, and alveoli ofP2–M1 and M3–4; LACM 140611, complete, isolat-

ed left M2 (Fig. 16); LACM 140646, partial rightM?2 without posterior half of talonid; LACM144509, isolated right M4; LACM 140647, upper?right molar (M?4), lingual half only (Fig. 15B).

TENTATIVELY REFERRED SPECIMEN.LACM 149365, right M?2 (see ‘‘Comments’’).

DIAGNOSIS. Sasawatsu mahaynaq differs fromthe remaining species of Paucituberculata in the fol-lowing combination of features: molars large andlow-crowned; M1 only moderately larger than M2;lower molars with distinct, although very reduced,paraconid, which is lingual and close to, but notfused with, metaconid; upper molars with paraconeand metacone very close to StB and StD, respec-tively, and much smaller than the latter; stylar cuspsnot labiolingually compressed.

MEASUREMENTS. See Table 3.TYPE LOCALITY. Santa Rosa, Peru. LACM lo-


Formation. ?Eocene.ETYMOLOGY. From the Quechua mahaynaq,

‘‘unique, unmatched.’’DESCRIPTION. LACM 140607 preserves most

of the horizontal mandibular ramus. This bone isrobust and fusiform, reaching its greatest height be-low the M3. In lateral view, the alveolar plane isstraight until about the anterior root of the M3. Theposterior root of the M3 and both roots of the M4


Table 3 Measurements of upper and lower molars of Sasawatsu mahaynaq gen. et sp. nov. L 5 length, W 5 width.

Measurement (mm)

M2

L W

M3

L W

M4

L W

Upper molarsLACM 140668LACM 140647LACM 149366

3.28

3.10

4.20

4.052.92

Lower molarsLACM 140661LACM 140607LACM 140646LACM 149365

3.803.88

4.10

2.402.602.402.90

are placed on an ascending plane that follows theorigin of the masseteric crest. This crest begins be-hind M4, as in most Palaeothentidae and unlike inthe Abderitidae, in which the ascending ramus aris-es labial to M4. In occlusal view, the alveolar rowis not straight but curved, being outwardly convex.The horizontal ramus has a flat or slightly concavelingual side and a convex labial side. There arethree mental foramina in the labial side. The largestis below the P3, and two other very small foramina,one below the M1 and the other below the anteriorroot of the M2, are found more posteriorly. On thelingual side of the horizontal ramus the symphysisextends posteriad to a point below the posterioredge of P3.

Only two teeth are preserved in this ramus, aproximal portion of the hypertrophied incisor (I2)and the M2. Anterior to the M2 are the alveoli ofthe M1 and, anterior to those, another four alveolithat decrease in size anteriad. Although the man-dible is broken anteriorly, it can be seen that ante-rior to the latter alveoli is a bony space that, as adiastema, separated them from the hypertrophiedincisor. Finally, between the diastema and this in-cisor, part of the alveolus of a small, single-rootedand somewhat procumbent tooth can be seen. Thisalveolus probably belonged to the second, whichwas probably the last, incisor of the series. It isdifficult to specify the homologies of the teeth an-terior to the M2 because of the poor preservationof their alveoli. The two alveoli immediately ante-rior to M2 correspond to those of the M1, whereasthe most anterior alveolus after the diastema maybelong to the canine and the next to three single-rooted premolars or, more probably, to one single-rooted premolar and one two-rooted premolar.Consequently, we think that the most probablelower dental formula was: I/2, C/1, P/2, M/4.

The root of the incisor, LACM 140695, is sub-oval in cross section. In its distal portion, the inci-sor is relatively short, robust, and somewhat spat-ulate, with its larger diameter in the vertical planeand with the enamel coat covering the labial and

ventral sides. In LACM 140695, there is a smallhorizontal wear facet on the tip.

No lower premolars or M1s are known. Judgingfrom the alveoli of the M1, it was moderately largerthan M2.

The M2 is robust, with the labial side consider-ably higher than the lingual side, and the trigonid,which is anteroposteriorly short, being longer onits labial side than on its lingual side. The proto-conid is robust, although lower than the metaconid.The paraconid is very small and positioned at thebase of the anterolingual slope of the metaconid, towhich it is much closer than to the protoconid. Theparacristid forms a right angle in its distal portion,such that the distal third of it, which connects withthe paraconid, is perpendicular to the dental axis.The metaconid develops a crest posteriorly thatconnects with the entocristid. The talonid is as wideas the trigonid, though clearly longer. The entoco-nid is tall, is subequal in height to the metaconid,and expands anteriorly as a crest that forms a lin-gual border to the talonid. Two vertical grooves,one anterior and the other posterior to the ento-conid cusp, can be observed on its internal face.The hypoconulid is short and occupies a more lin-gual position than in most Palaeothentidae and Ab-deritidae.

The M3, in comparison with the M2 (Fig. 16), isslightly smaller. It has a difference in height be-tween the labial and lingual sides of the tooth; ahypoconulid that is less distinct and more centrallypositioned; the paracristid shorter; the paraconidmore mesial and closer to the main axis of the par-acristid, being equidistant from the protoconid andthe metaconid; the vertical grooves on the internalface of the entoconid deeper; and the anterobasalcingulum better developed.

The M4 is much smaller than the M3, but similarin general shape. The paracristid is not straight but,instead, forms in occlusal view an almost right an-gle at the union between the postparacristid and thepreprotocristid. The protoconid is positionedslightly anteriorly compared with the metaconid,


and both the entoconid and the hypoconid havenoticeable grooves on the slopes that face the tal-onid basin. The talonid basin of the M4 is compar-atively smaller than that of the M3, and on its pos-terior face, the hypoconulid is positioned more cen-trally and ventrally.

Only three upper molars have been identified: theholotype (LACM 140668; Fig. 15A); LACM149366, which is essentially identical to the holo-type; and LACM 140647 (Fig. 15B), a very frag-mented specimen, probably an M4. The holotype isan M3, with its anterior side much longer than theposterior side and with a large central basin. Theprotocone is robust and occupies almost all the lin-gual side of the tooth. StB is also very large, al-though low. There is no StC or StA. StD is smallerthan StB, and anteriorly, it develops a short crest.The paracone and the metacone are very reducedand tightly paired to StB and StD, respectively. Themetaconule is poorly differentiated. There are nosharp cristae.

LACM 140647 (M?4) is smaller than the othertwo upper molars, and it preserves only its lingualhalf. Only a robust protocone and a poorly devel-oped metaconule bordering lingually a large centralbasin can be distinguished. No vestiges of the para-cone or the metacone can be observed, althoughthis may be because the tooth is broken labially.

COMMENTS. Even though generalized in sev-eral features of upper molar morphology, includingthe persistence of a vestigial paracone and poorlydeveloped metaconules, Sasawatsu mahaynaq is astrongly derived species of Paucituberculata. This isa result of its adaptations toward frugivorous feed-ing habits. All cheek teeth are brachydont, withvery low and bunoid cusps, wide basins, and acomplete absence of cutting crests. As in Perulestes,at least the M3 preserves a vestige of the paracone,which is almost completely fused to the base of StB.Unlike Perulestes, the metacone is notably reducedand attached to the lingual side of StD. In the lowermolars, the paraconid occupies a relatively mesialposition. In some molars, such as the M3, it is al-most equidistant from the protoconid and themetaconid. The lower dental formula of this pecu-liar paucituberculatan suggests that it belongs tothe Palaeothentidae or Abderitidae. However, judg-ing from the M1 alveoli, this molar was not muchlarger than the M2; in turn, in members of thesefamilies, the M2 is much larger than the M3. On theother hand, the assemblage of features of upper andlower teeth resembles more those of the Palaeo-thentidae than those of the Caenolestidae or theAbderitidae.

Sasawatsu is one of the largest (squirrel-sized)and most specialized marsupials of the Santa Rosalocal fauna. As with Perulestes, persistence of theparacone of the upper molars suggests an earlystage in the development of the quadrangular molarof the Paucituberculata.

LACM 149365, an isolated right lower molar, ishere regarded as a probable M2 assignable to this

species. It is similar to other M2s of S. mahaynaq,differing only by having a slightly larger size, ahigher labial face, the paraconid closer to the meta-conid, and a less distinct hypoconulid that is placedhigher than in the other M2s. It could be the casethat this molar actually represents a first lower mo-lar. However, its size does not match the size of thealveoli for the first molar in LACM 140607, whichindicates a larger tooth. Second, its morphologydoes not fit with that of any known paucitubercu-latan first lower molar. Among the Caenolestidae,the trigonid is very open lingually; in the Palaeo-thentidae, the trigonid is extremely open, and thecristid obliqua meets the metaconid at its anteriorend; and in the Abderitidae, the trigonid of the M1

is comparatively very large and bladelike. None ofthese features occur in LACM 149365, and givenall other sources of evidence, the possibility thatthis molar might belong to a new, distinct clade ofpaucituberculatans not recorded up to now seemsremote. A more plausible hypothesis is that intra-specific variability (e.g., sexual dimorphism) is re-sponsible for the observed differences betweenLACM 149365 and other specimens here assignedto M2s.

Distinct striations occur on the enamel surface ofmost specimens here assigned to S. mahaynaq.They are obvious on the labial surfaces of lowermolars and on the lingual surfaces of the upper mo-lars. They seem to be perikamata, which have beendescribed for other mammals (e.g., Moss-Salentijnet al., 1997). Perikamata consist of alternating ridg-es and grooves on the enamel surface, shaping abanded pattern that is approximately parallel to thedental axis. Perikamata are considered to be theresult of differential growth patterns during the ear-ly phase of amelogenesis. They are common in pri-mates, including man. To our knowledge, this fea-ture is not present in any other South Americanfossil or living marsupials.

Order Microbiotheria Ameghino, 1889

Family Microbiotheriidae Ameghino, 1889

Kirutherium gen. nov.TYPE SPECIES. Kirutherium paititiensis sp. nov.DIAGNOSIS. As for the type, and only known,

species.ETYMOLOGY. From the Quechua kiru, ‘‘tooth’’

and the Greek therion, ‘‘beast,’’ a common endingof generic names for microbiotherian marsupials.

Kirutherium paititiensis sp. nov.Figure 17

HOLOTYPE. LACM 140620, an isolated leftM3, broken at its labial third.

HYPODIGM. The holotype only.DIAGNOSIS. Kirutherium paititiensis differs

from other microbiotheriids in the following fea-tures: upper molars with vestigial para- and meta-conules that are elongate and partially fused with


Figure 17 Kirutherium paititiensis gen. et sp. nov. LACM140620 (holotype), isolated left M3 broken labially: A,occlusal view; B, lingual view. Abbreviations: me, meta-conid; pa, paraconid; pr, protoconid. Scale 5 1 mm.

pre- and postprotocrista, respectively; trigon basinvery deep, having its lingual face divided by twolabiolingually oriented grooves; postmetacristapoorly developed.

MEASUREMENTS. LACM 140620: L 5 1.23*,W 5 1.42; mm.



ETYMOLOGY. Paititiensis, from Paititi, a lostcity of the Incas that was probably located in thetropical forests east of the Andean Cordillera.

DESCRIPTION. The holotype, an isolated upperleft molar (M3), is very similar in size to that of theonly living microbiotheriid, Dromiciops gliroidesThomas, 1894. The labial third is broken in such away that it lacks almost all of the stylar shelf (Fig.17A). However, it is evident, judging from the pre-served posterolabial corner of the tooth, that thestylar shelf was short, as in Dromiciops Thomas,1894, or Microbiotherium Ameghino, 1887. Thepara- and metacone are subequal in size and height,and they are well separated from each other. Bothcusps have their labial sides descending smoothlytoward the labial face of the crown. The centro-crista is straight and forms a wide valley at the con-tact between the postparacrista and the premeta-crista. The postmetacrista is only partially pre-served, although, judging from its overall propor-tions, it was relatively short. Neither thepreparacrista nor stylar cusps were preserved. Theposterior third of the stylar shelf, which is pre-served in the holotype, shows no stylar cusps, thussuggesting that, as in other species of the familyMicrobiotheriidae, stylar cusps were extremely re-duced or absent. The protocone is wide, and thetrigon basin very deep. On the lingual face of thetrigon basin are a couple of peculiar, shallowgrooves that border the protocone at its lingualbase (Fig. 17A). The protocone is slightly eccentricin that it is oriented anteriorly. Both the para- andmetaconule are vestigial, elongate, and partiallyfused with the labial ends of the pre- and postpro-tocrista, respectively (Fig. 17B). There are no an-terior or posterior cingula in the preserved portionof the tooth.

COMMENTS. Most of the features of Kiruther-ium paititiensis constitute a combination of primi-tive and derived characters unequivocally diagnos-tic of the Microbiotheriidae: large protocone, wideand deep trigon basin, straight centrocrista, verynarrow stylar shelf with vestigial or absent stylarcusps, relatively short postmetacrista, para- andmetacone subequal in size and height, and vestigialpara- and metaconule. The overall proportions ofthe tooth, including the relatively short postmeta-crista, agree with those of the M3 of other micro-biotheriids with known upper dentitions. Kiruth-erium seems to be close to Microbiotherium andDromiciops in having a wide protocone and deeptrigon basin. Khasia cordillerensis Marshall andMuizon, 1988, from lower Paleocene (TiupampianSALMA) levels at Tiupampa, Bolivia, has its pro-tocone more anteroposteriorly compressed and avery reduced stylar shelf without cusps or crests.

Order and Family IndeterminateWirunodon gen. nov.

TYPE SPECIES. Wirunodon chanku sp. nov.DIAGNOSIS. As for the type, and only known,

species.


Figure 18 Wirunodon chanku gen. et sp. nov. LACM140621 (holotype), isolated upper left molar (M?3): A, oc-clusal view; B, lingual view. Scale 5 1 mm.

ETYMOLOGY. From the Quechua wiruna,‘‘small mallet,’’ and the Greek odontos, ‘‘tooth,’’ inreference to the relatively small protocone and tri-gon basin of the holotype of the type species.

Wirunodon chanku sp. nov.Figures 18, 19

HOLOTYPE. LACM 140621, an isolated upperleft M?3 (Fig. 18).

HYPODIGM. The holotype and LACM 140677,an isolated right trigonid (Fig. 19).

DIAGNOSIS. Upper molars with almost straightcentrocrista; both protocone and trigon basin re-duced; stylar shelf wide; postmetacrista well devel-oped; preparacrista connects with stylar cusp A;stylar cusps B and D reduced and labiolinguallycompressed; paracone very reduced relative tometacone, suboval in section, and with its long axispeculiarly oriented with an anterolabial–posterolin-gual axis; para- and metaconules absent.

MEASUREMENTS. LACM 140621: L 5 1.27,

W 5 1.12; LACM 140677: L (trigonid) 5 0.74, W(trigonid) 5 0.78; mm.



ETYMOLOGY. From the Quechua chanku,‘‘opossum.’’

DESCRIPTION. The holotype is very small,comparable in size to an M3 of the smallest livingmarmosines. The protocone is eccentric, almostaligned in an axis with StB and the paracone. Theparacone is much lower and reduced with respectto the metacone, and the centrocrista is almoststraight. The anterior cingulum is wide and extendslingually in the preprotocrista. There is no para- ormetaconule. The stylar shelf is wide, and both StBand StD are positioned very labially. StD is morelabiolingually compressed than StB. Both StC andStE are absent. The postmetacrista is well devel-oped; the preparacrista apparently connects with asmall StA. The paracone and metacone have flatlabial surfaces (Fig. 18A), thus both cusps are sub-oval in section, rather than conical, with their lon-ger axes parallel to the dental row. The paracone,however, is slightly twisted on an anterolabial–pos-terolingual axis. For this reason the postparacristais not completely straight, but forms instead, in oc-clusal view, a minute notch at its contact with thepremetacrista (Fig. 18A). The relative width of thestylar shelf and the development of the postmeta-crista agree with the pattern seen in the M3 of other‘‘opossum-like’’ marsupials.

LACM 140677 is an isolated right trigonid of alower molar. The anterobasal cingulum is narrow,and it extends basally almost to the paracristid.This crest is well developed as a consequence of theconsiderable height of the protoconid. The para-conid is well developed, although it is slightly lowerthan the metaconid (Fig. 19B). The paracristidnotch is deep. Even though the talonid is brokenaway, it can be observed that its basin was veryshallow. The anterior end of the cristid obliquameets the metacristid at a point below the notch,thus suggesting that this specimen is an M4.

COMMENTS. This enigmatic marsupial is dif-ficult to assign to any known marsupial order. Onthe one hand, the persistence of an almost straightcentrocrista suggests that it belongs to the Peradec-tia or the Microbiotheria. However, a relationshipto the latter can be rejected on the basis of almostall the remaining available characters: the trigonbasin is reduced (very wide in Microbiotheriidae);the protocone is not wide, but rather narrow (widein Microbiotheriidae); the stylar shelf is wide (veryreduced in Microbiotheriidae); and the metacone ismuch larger and taller than the paracone (paraconeand metacone subequal in Microbiotheriidae). Thislast feature also differentiates Wirunodon from thePeradectia, as do the smaller protocone, StB, andStD; the great development of the postmetacrista;the absence of para- and metaconules; and the pre-


Figure 19 Wirunodon chanku gen. et sp. nov. LACM 140677, right trigonid: A, labial view; B, lingual view. Scale 51 mm.

paracrista connected with StA. In fact, most derivedfeatures of Wirunodon are consistent with its as-signment to the Didelphimorphia, among which areforms with a well-developed postmetacrista, a cleardifference in height between para- and metacone,StB and StD relatively reduced and labiolinguallycompressed, and para- and metaconule reduced orabsent. However, in all known Didelphimorphiathe centrocrista is V-shaped, although with varia-tions (Goin, 1997). Finally, sparassodont affinitiesof Wirunodon can be ruled out on the basis of itswide stylar shelf, shallow ectoflexus, and moder-ately separated, rather than close, paracone andmetacone.

The existence of South American peradectiansstrongly specialized toward frugivorous habits,such as the Caroloameghiniidae (Goin et al.,1998a), suggests that on this continent other pera-dectian lineages may have differentiated towardfaunivorous or even carnivorous habits. On theother hand, it is also reasonable to suppose that thealmost straight centrocrista of Wirunodon is sec-ondarily derived from a more typical didelphimor-phian condition. Regarding the lower molars ofWirunodon, the trigonid structure in specimenLACM 140677 is highly generalized, although itshows a long and deep paracristid, along with awell-developed postmetacrista of the holotype. Theconnection of the cristid obliqua to a point belowthe notch of the paracristid anteriorly is a primitivefeature that is present in all peradectids and in nu-merous Paleogene didelphimorphians.

The only marsupial known to us that resemblesW. chanku, with its unique combination of ple-siomorphic, apomorphic, and autapomorphic fea-tures, is Kasserinotherium tunisiense Crochet,

1986, from the lower Eocene of Tunisia (Africa),which is regarded as a peradectid. In particular, oneof the specimens assigned to this species (EY 12;see Crochet, 1986:924, figs. 2a–c), referred to as anM1 or M2, closely resembles the holotype of W.chanku in having a very small size, straight centro-crista, conules absent, reduced protocone and tri-gon fossa, reduced and labiolingually compressedstylar cusps, wide stylar shelf, and well-developedpostmetacrista. The fact that the preparacrista endslabially at the anterolabial corner of the tooth(StA?) may be another feature in common betweenWirunodon and Kasserinotherium. Even thoughthe preparacrista of the M3 of K. tunisiense seemsto connect with StB, this may not be the case inmore anterior molars (see Crochet, 1986:fig. 2a). Infact, Crochet (1986:924) notes that, in specimenEY 12, ‘‘L’angle forme par la centrocrete et la par-acrete est plus ouvert qu’il en l’est habituellementpour une dent de ce rang chez les marsupiaux prim-itifs.’’

Kiruwamaq gen. nov.

TYPE SPECIES. Kiruwamaq chisu sp. nov.DIAGNOSIS. As for the type, and only known,

species.ETYMOLOGY. From the Quechua kiru,

‘‘tooth,’’ and wamaq, ‘‘strange, odd.’’

Kiruwamaq chisu sp. nov.Figure 20

HOLOTYPE. LACM 140619, an almost com-plete, isolated left M?3.

HYPODIGM. The holotype only.DIAGNOSIS. Kiruwamaq chisu differs from all


Figure 20 Kiruwamaq chisu gen. et sp. nov. LACM 140619 (holotype), isolated upper left molar (M?3): A, occlusal view;B, lingual view. Scale 5 1 mm.

other known South American ‘‘opossum-like’’ mar-supials in the following combination of characters:very small size; reduced protocone; paracone muchsmaller, more labially oriented, and very close tothe metacone; straight centrocrista; postmetacristawell developed; stylar shelf very wide only in itsanterior half; StD very small, labially oriented andlabiolingually compressed; preprotocrista connectscontinuously with anterior cingulum of tooth.

MEASUREMENTS. L 5 0.85, W 5 1.19; mm.TYPE LOCALITY. Santa Rosa, Peru. LACM lo-


Formation. ?Eocene.ETYMOLOGY. From the Quechua chisu, ‘‘very

small.’’DESCRIPTION. The holotype is broken at its

anterolabial end, including the area of StB and theanterolabial end of the preparacrista (Fig. 20A). Itis a very wide and short tooth, with all its crestssharp and with the crown much higher at the post-metacrista than at the preparacrista. The protoconeis notably reduced and low; consequently, the tri-gon basin is vestigial. The preprotocrista does notend at the paracone base; instead, it runs anteriorto it and continues to the anterior cingulum. Thepostprotocrista is, in contrast, very short and endsat the base of the metacone. These features, togeth-er with the labial retraction of the paracone, give apeculiar, asymmetric shape to the whole tooth. Theparacone is much smaller and lower than the meta-

cone. It, with the metacone, does not form an axisparallel to the dental axis; instead, it is positionedmore labially and very close to the metacone. Thecentrocrista is straight. The preparacrista and thepostmetacrista are sharp. The anterior faces of theparacone and the metacone are, as in Wirunodonchanku, extremely flat and perpendicular to theplane of the stylar shelf. The preparacrista is bro-ken at its labial end; thus, its eventual connectionwith any anterior stylar cusp cannot be observed.However, it can be seen that the preparacrista waslong, although shorter than the postmetacrista. Theectoflexus was apparently deep. The postmetacristais notably developed, and, in occlusal view, it isperpendicular to the dental axis. The stylar shelf isvery wide for its anterior half. Both the prepara-crista and the postmetacrista form gradual slopeson the anterior and posterior limits of the stylarshelf; consequently, the stylar shelf is basin-shapedinstead of flat. The great relative development ofthe postmetacrista suggests that this tooth corre-sponds to an M3.

COMMENTS. Kiruwamaq chisu is an unusualSouth American mammal, the affinities of whichare a mystery. The possession of a straight centro-crista is a generalized feature present in the Pera-dectia and Microbiotheria and in some Sparasso-donta. However, most remaining derived featuresof K. chisu exclude it from these groups. This spe-cies could also be thought to belong to the Didel-phimorphia. In fact, some lineages of this order,


Table 4 List of marsupial taxa from the Santa Rosa localfauna.

Order Didelphimorphia Gill, 1872cf. Family Herpetotheriidae Gray, 1821

Rumiodon inti gen. et sp. nov.Order Sparassodonta Ameghino, 1894

Family Hathliacynidae Ameghino, 1894Patene campbelli sp. nov.

Order Microbiotheria Ameghino, 1889Family Microbiotheriidae Ameghino, 1889

Kirutherium paititiensis gen. et sp. nov.Order Polydolopimorphia Ameghino, 1897

Family Prepidolopidae Pascual, 1980Incadolops ucayali sp. nov.

Family indeterminateWamradolops tsullodon gen. et sp. nov.Hondonadia pittmanae sp. nov.

Order Paucituberculata Ameghino, 1894Family Caenolestidae Trouessart, 1898

Perulestes cardichi gen. et sp. nov.Perulestes fraileyi gen. et sp. nov.

cf. Family Palaeothentidae Sinclair, 1906Sasawatsu mahaynaq gen. et sp. nov.

Order and Family indeterminateWirunodon chanku gen. et sp. nov.Kiruwamaq chisu gen. et sp. nov.

such as the Sparassocynidae, have also developedlarge postmetacristae, a strongly reduced protoconeand paracone, and a paracone close to the meta-cone. However, the peculiar shape of the protoconein this species, just as with the width and form ofthe stylar shelf, is not seen in any known Didel-phimorphia.

The strong reduction of the protocone and theparacone, the proximity of the latter to the meta-cone, and the large development of the postmeta-crista are derived characters diagnostic of the Spar-assodonta. But, the width of the stylar shelf makesit unlikely that it belongs to this order. Some fea-tures of K. chisu, however, suggest possible affini-ties with the most primitive sparassodont, Allqok-irus australis, from the lower Paleocene (Tiupam-pian SALMA) of Bolivia. This species also has itsparacone somewhat labially oriented relative to themetacone, a preprotocrista that continues labiallyin the anterior cingulum of the tooth, a preproto-crista and postmetacrista that are well developed,and a labially compressed StD with an elongatedanterior crest. Kiruwamaq chisu differs from A.australis primarily by lacking para- and metaco-nules and in having a larger reduction of the pro-tocone and a larger relative development of thepostmetacrista. All of these features suggest a morederived condition for K. chisu. Thus, it could beargued, though the evidence is still scarce, that, onthe basis of its small size and the relatively largedevelopment of all shearing crests, the Amazonianspecies belongs to a lineage derived from minute-sized hathliacynid sparassodonts of insectivorous–faunivorous habits.

Finally, some other features of Kiruwamaq mightsuggest different affinities. For example, the re-markable width of the holotype, relative to the totallength of the tooth, makes it typically triangular,with the labial face much shorter than the othertwo. A similar outline, even more prominent thanin Kiruwamaq, can be seen in the enigmatic ?mar-supials of the family Necrolestidae. Known onlyfrom Miocene levels (Santacruzian SALMA) of Pa-tagonia, these were probably fossorial forms whoseaffinities with marsupials are still doubtful. The ex-treme simplification of the molar pattern in necro-lestids hampers more precise comparisons.

DISCUSSION

A NOVEL PALEOGENE MARSUPIAL FAUNAFROM TROPICAL SOUTH AMERICA

The new taxa described herein (Table 4) constituteone of the most outstanding marsupial assemblagesdiscovered in recent years. They add significantknowledge about the Paleogene marsupial diversityin tropical South America, and, most notably, pro-vide evidence regarding the early phases in the evo-lution of ‘‘pseudodiprotodont’’ (sensu Ride, 1962)marsupial lineages.

Size

Almost all specimens recovered at this new localityare small to very small. The largest species repre-sented is the carnivore Patene campbelli, which wasabout the size of a domestic cat, followed by therat- or squirrel-sized paucituberculatans Perulestesfraileyi and Sasawatsu mahaynaq. All of the othersare of the size of small mice, probably not exceed-ing the weight of the living microbiotheriid Drom-iciops gliroides (20–50 g; Hershkovitz, 1999:33).

Feeding Habits

The large proportion of small frugivorous or fru-givorous–insectivorous taxa among the Santa Rosamarsupials is striking. Their size and inferred dietare suggestive of arboreal locomotion; however, nocranial or postcranial remains that could test thishypothesis have been recovered so far.

The very fragmentary preserved holotype ofHondonadia pittmanae has a morphological pat-tern seen in other generalized polydolopimorphianswith inferred insectivorous–frugivorous diets (e.g.,Rosendolops). The paucituberculatans Perulestescardichi and P. fraileyi, though different in size, alsoshow features suggesting frugivorous feeding hab-its, such as poorly developed cristae and wide pro-tocones and talonids. Sasawatsu mahaynaq seemsmore strictly adapted to a frugivorous diet, withvery low crowned molars; a complete absence ofshearing cristae; and wide trigon basins, proto-cones, and talonids. The microbiotheriid Kiruther-


ium paititiensis is reminiscent, in upper molar mor-phology, of the insectivorous–frugivorous Dromi-ciops gliroides, the only living microbiotheriid mar-supial. Wirunodon chanku is so small that itsprobable feeding habits are difficult to interpret.This taxon has reduced protocones, well-developedpostmetacristae and paracristids, and reduced sty-lar cusps, all features that characterize many fossiland living opossums of, at least partially, carnivo-rous feeding habits. But, because of the minute sizeof W. chanku, it seems more likely that its diet con-sisted of small insects or worms, rather than ofsmall vertebrates. Another tiny ?opossum recoveredfrom Santa Rosa, K. chisu, seems to have beenadapted to feed on small insects. The small proto-cone, and large postmetacrista and paracristid ofRumiodon inti suggest partially faunivorous feed-ing habits.

Wamradolops tsullodon is, by far, the most abun-dant marsupial of the Santa Rosa local fauna. Al-most half of all marsupial specimens from the SantaRosa fauna are assignable to this species. Several ofthe most characteristic features of this taxon areprobably adaptations related to the breaking downof relatively hard items, such as seeds, small nut-like structures, or bark, or the shearing of fleshyfood into smaller pieces. These features includedeep and strong mandibles; very large upper andlower premolars with shearing blades; molars withlow cusps and, especially in the anterior lower mo-lars, unilateral hypsodonty; a thick enamel layer onall teeth; and a reduction of shearing cristae in up-per and lower molars. All of these features, plus itsvery small size, suggest mainly frugivorous feedinghabits. Among South American marsupials, the?early Oligocene (Divisaderan SALMA) Groeberi-idae constitute the group most extremely special-ized toward these inferred feeding habits. To Simp-son (1970), the diet of groeberiids possibly includedbark, wood-burrowing insects, or fruits with hardshells. Several extant mammals, such as the primateDaubentonia E. Geoffroy, 1795, from Madagascar,or the marsupial Dactylopsila Gray, 1858, alsoshare these feeding habits from New Guinea andAustralia (Pascual et al., 1994). Groeberiids showa unique set of cranial and dental adaptations thatinclude short, very high horizontal rami that areventrally fused; a very deep palate, which, togetherwith the curved edge of the fused mandibles, sug-gests the existence of a strong, large (parrot-like?)tongue (Pascual et al., 1994); upper and lower in-cisors very large and strong; teeth markedly heter-odont, that is, comprising different morphofunc-tional units that have a thick enamel layer; a skullwith a relatively enormous, tardigrade-like ventralprojection of the masseteric process of the zygoma,which suggests a strong horizontal component ofthe masseteric muscles, etc. Even though no cranialremains of Wamradolops are known, several of itsmost derived dental features are also present ingroeberiids: for example, heterodont cheek-teethwith a thick enamel layer; upper molars with a pos-

terolabially ‘‘displaced’’ protocone, resulting in theenlarged metaconule being the most lingual cusp ofthese molars; and lower molars with an anteriorlydisplaced entoconid.

Negative Evidence

An additional aspect of the Santa Rosa marsupialassociation comes from negative evidence, that is,the complete absence of polydolopines (Polydolo-pimorphia, Polydolopidae). Polydolopines areamong the best represented taxa in Paleogene levelsof Patagonia, but they have never been recorded atnorthern latitudes. Even though we are cautiouswith this negative evidence, it is consistent withwhat has been found from Paleogene levels northof Patagonia; for example, the Lumbrera Forma-tion (?lower Eocene) and Antofagasta de la Sierrafauna (late Eocene), both from northwestern Ar-gentina, and the Itaboraı Formation (ItaboraianSALMA) of eastern Brazil all lack polydolopines.Regarding the middle upper Paleocene ItaboraıFormation, it is interesting to note that, althoughno polydolopines are among the extremely richmarsupial fauna recovered from it (see, e.g., Mar-shall, 1987), the epidolopine Epidolops ameghinoiPaula Couto, 1952b, is extremely well represented,comprising about 40 percent of all marsupial spec-imens from this locality. On the other hand, half ofthe several hundred marsupial specimens collectedat a contemporary (Itaboraian SALMA) locality inthe Las Flores Formation in central Patagonia areassignable to epidolopines and polydolopines (thelatter forms including half a dozen new, still un-named, species; Goin et al., 1997). Polydolopinesare the most abundant mammal remains known todate from middle Eocene levels of the La MesetaFormation in the Antarctic Peninsula. The youngestpolydolopine known, Polydolops abanicoi Flynnand Wyss, 1999, also comes from southern SouthAmerica: the Termas del Flaco region in centralChile, from 348599 S. Its age has been estimated asearly Oligocene (‘‘Tinguirirican’’ age; Flynn andWyss, 1999, and references cited therein; Kay et al.,1999). There seems to be a relationship, probablytrophic, between the multituberculate-like polydol-opines and southern South American and Antarcticforests where Nothofagus sp. (the Southern Beech)is dominant and podocarps and araucarian conifersare common (Case, 1988). In turn, NothofagusBlume, 1850, is considered an indicator of cool,humid climatic conditions. Case (1988) correlatesone lower Eocene locality in Seymour Island (RV-8425, ‘‘TELM 2’’), which is dominated by a large-leafed species of Nothofagus, with cooling condi-tions.

SIGNIFICANCE OF THE SANTA ROSAMARSUPIALS

‘‘Opossum-like’’ Marsupials

Rumiodon inti may be, if its precise affinities areconfirmed (see ‘‘Systematic Paleontology’’), one of


the first records of a South American didelphimor-phian of the family Herpetotheriidae (see also Oliv-eira, 1998). Interestingly, this would not be the onlyherpetotheriid record for this continent. Work inprogress by Goin and Forasiepi on the middle latePaleocene marsupials of the Las Flores Formationof central Patagonia (Goin et al., 1997), suggeststhat several of the recognized taxa from there be-long to this group of didelphimorphians. A reviewof several taxa from the Itaboraı Formation leadsto similar conclusions (Goin, 1991; Oliveira, 1998;Oliveira and Goin, in press). All of these formsshare a combination of generalized and derived fea-tures that are not present in the Neogene opossumsof the family Didelphidae, including well-developedparaconules, metaconules, or both; a narrow stylarshelf; centrocrista ending very labially on the stylarshelf; a well-developed StC that is clearly separatefrom StB and StD; and unreduced hypoconulidsand entoconids.

Reig (1981) recognized the subfamily Herpeto-theriinae, including the genera HerpetotheriumCope, 1873; Peratherium; and Amphiperatheriumthat were known from lower Tertiary levels ofNorth America, Europe, and northern Africa. Reig(1981:56) regarded herpetotheriids as closely relat-ed to peradectids (his ‘‘Peradectinae’’) rather thanthe didelphids, a conclusion also shared by Reig etal. (1987; but see the proposed classification of ‘‘di-delphimorphs’’ by these authors in the same work,pp. 81–82, and Marshall, 1987). Previously, Cro-chet (e.g., 1980) reviewed the European genera Per-atherium and Amphiperatherium, including themin his tribe Didelphini. It should be noted that Cro-chet’s (1980) concept of Didelphidae, Didelphinae,and even of Didelphini was extremely wide. Tohim, didelphids comprise five subfamilies: Didel-phinae, Sparassocyninae, Microbiotheriinae, Glas-biinae, and Caroloameghiniinae, some of which(i.e., his Glasbiinae, Microbiotheriinae, and Caro-loameghiniinae) are presently placed in differentfamilies or orders (see, e.g., Marshall et al., 1990).In turn, the Didelphinae included the tribes ‘‘Di-delphini’’ (Didelphimorphia) and ‘‘Peradectini’’(currently referred to the order Peradectia). Finally,his ‘‘Didelphini’’ included an extremely wide varietyof opossums, several of which are presently allo-cated in different families and orders (e.g., Miran-datherium Paula Couto, 1952c). Perhaps because ofhis broad approach to the group, his diagnosis ofthe tribe ‘‘Didelphini’’ stressed features currently re-garded as diagnostic of all didelphimorphians (Cro-chet, 1980:58; Goin, 1991). However, Crochet hadalready realized the different nature of the SouthAmerican early Tertiary didelphimorphian radia-tion with respect to that of the Neogene, findinginconclusive affinities between several South Amer-ican Paleogene opossums and European herpetoth-eriids. Regarding the possible origin of his ‘‘Didel-phini,’’ he noted: ‘‘La presence a Sao Jose de Ita-borai et dans l’Eocene europeen de Didelphini pre-sentant une dilambdodontie talpidienne peut

egalement etayer la notion de liens entre les faunnesdidelphidiennes neotropicale du Paleocene et hol-arctique de l’Eocene. Mais ce phenomene peut aussietre la consequence d’une convergence heterochro-ne. L’absence de ce type morphologique de molairesuperieure dans la documentation nord-americainedisponible n’est pas un obstacle a l’une ou l’autrede ces hypotheses . . . les genres actuels d’origineessentiellement sud-americaine s’individualisent aucours d’une seconde radiation adaptative, dontl’age ne peut encore etre precise par manque dedonnes concertant les periodes Eocene et Oligo-cene.’’ (Crochet, 1980:225; italics ours). A reviewof South American Neogene opossums by one ofus (Goin, 1991, 1994, 1995) led to the same con-clusions: Neogene didelphids are derived in differ-ent ways than Paleogene didelphids. To the set ofmolar features mentioned above that separate her-petotheriids from didelphids, all of which are veri-fiable in the Peruvian R. inti, are several other char-acters that deserve further consideration.

1. The horizontal ramus in didelphines is anteriorlyshortened with respect to that of herpetother-iids, with the symphysis reaching posteriad to apoint below the posterior edge of P2. In the her-petotheriid specimens, the mandible is moreelongated anteriorly, and the symphysis endsposteriorly just below the P2. In herpetotheriidswith known rostral, or anterior, portions of themandibles, the entire rostrum seems to havebeen more elongate, as shown in Amphiperath-erium lamandini (Filhol, 1876); A. fontense Cro-chet, 1979; A. ambiguum Filhol, 1877; A. exileGervais, 1848–1852; Peratherium elegans (Ay-mard, 1846); and P. caluxi Filhol, 1877 (seeCrochet, 1980:figs. 102a–c, 123, 130, 140, 207,235a–c). In all of these taxa the horizontal ra-mus is very elongate, with the symphysis endingposteriorly below the P2 and the mental foram-ina much larger than that of didelphines.

2. In didelphids, the P2 is subequal or larger thanthe P3; in herpetotheriids, the P3 is usually largerthan the P2.

3. Lower incisors are, in most didelphids, subequalin size. In herpetotheriids with preserved ante-rior dentitions, the second lower incisor (ho-mologous to I3 of Hershkovitz, 1982) is largerthan the rest. This derived character is preservedin several specimens belonging to different spe-cies reviewed by Crochet (1980), including A.minutum, A. ambiguum, and Peratherium per-rierense Crochet, 1979 (see Crochet, 1980:figs.87a–b, 129, 217, 232). Curiously, differences inthe relative sizes of the lower incisors amongSouth American didelphids have only been re-corded in Paleogene opossums such as the Ita-boraian Derorhynchus (Paula Couto, 1962;Marshall, 1987) and Eobrasilia Simpson, 1947(Marshall, 1984). Herpetotheriid affinities ofDerorhynchus and several other opossums werediscussed by Goin (1991).


4. Where known, upper incisors of herpetotheriidsdiffer in relative sizes when compared with thoseof South American Neogene opossums: I2 ismuch smaller than I1 or I3 (see Crochet, 1980:fig. 119). Among South American Neogeneopossums, only the first (I1) or last (I5) upperincisors have differences in relative sizes (Taka-hashi, 1974). All of these features, plus thosementioned previously for the molar teeth, con-stitute a heterobathmic (sensu Hennig, 1968)distribution of generalized and derived featuresfor each clade (i.e., Didelphidae and Herpeto-theriidae), thus suggesting that they are sistergroups. Until a thorough review of all PaleogeneSouth American opossum taxa is completed, itis not possible to be more precise as to theirprobable relationships. Goin (1991, 1995) sug-gested that the concept of Didelphoidea shouldbe restricted exclusively to Neogene and RecentSouth American opossums. In his systematicproposal, these include the Didelphidae (s.s.),Caluromyidae, and Sparassocynidae. On theother hand, herpetotheriines (excluded from theDidelphidae) and several other taxa known ex-clusively from South American Paleogene levels,such as the Derorhynchidae, Eobrasiliinae, andProtodidelphidae, may be the sister group of theDidelphoidea.

The above considerations have important con-sequences in paleobiogeographic hypotheses re-garding vicariance, dispersal, or both, as events inthe early history of didelphimorphian opossums.Particularly debated have been the biogeographicrelationships of didelphimorphian and peradectianopossums between South America, North America,Europe, and Africa (see, e.g., Marshall, 1987; Reiget al., 1987; Marshall and Muizon, 1988; Marshallet al., 1990; Marshall and Sempere, 1993; Petersand Storch, 1993). In discussing the origin of theMessel (middle Eocene of Germany) birds andmammals, Peters and Storch (1993) suggested thatEuropean Amphiperatherium and Peradectes Mat-thew and Granger, 1921, may have used quite dif-ferent routes and time periods to reach Europe.‘‘North or South America provided the earliest ev-idence of Peradectes (earliest Paleocene), while Eu-rope is third (Early Eocene). . . . This suggests a dis-persal of Peradectes from North America to Europevia northern North Atlantic bridges during the Ear-ly Eocene. Didelphines, on the other hand, ap-peared about contemporaneously in the early Eo-cene on the European and North American conti-nent. . . . The fossil records of marsupials fromSouth America, on the other side, includes the ear-liest known didelphines, coming from the latestCretaceous or early Paleocene of Bolivia and Peru. . . , and Crochet and Sige (1983) correspondinglyhypothesized didelphine dispersal from SouthAmerica into Europe via Africa’’ (Peters and Storch,1993:264). The record of South American ?herpe-totheriid opossums, as we suggest here for R. inti

and as is argued elsewhere (Oliveira and Goin, inpress), makes the hypothesis of Crochet and Sige(1983) even more plausible. Regarding the allegedNorth American origins of European herpetother-iids, it should be mentioned that South Americanand European members of this family are closer toeach other than either of them is to Herpetotheriumand allies (sensu Korth, 1994), and the molar pat-tern of the former is plesiomorphic with respect tothat of the latter.

The other ‘‘opossum-like’’ marsupial in the SantaRosa local fauna is the enigmatic Wirunodon chan-ku, the familial or ordinal assignation of which isvery speculative on the basis of available material.Its almost straight centrocrista suggests affinitieswith the Peradectia, but nearly all the remainingcharacters are derived for both the Peradectia andthe Microbiotheria. Many of these characters arepresent in several lineages of South American di-delphimorphians. To us, the only marsupial com-parable to W. chanku is Kasserinotherium tuni-siense, a ?peradectian from the lower Eocene of Tu-nisia (Africa). If their affinities are confirmed, thiswould be another piece of evidence supportingSouth American and African biogeographic rela-tionships during the early Paleogene (see above).However, on the basis of available material, it isvery difficult to precisely assign this peculiar Ama-zonian marsupial to any known family or order.Both hypotheses here presented, that is, either per-adectian or didelphimorphian affinities, are admis-sible. Regarding the first, it is interesting to noteCrochet’s (1986:926) suggestion that peradectids,in their present concept, may constitute a paraphy-letic taxon.

‘‘Pseudodiprotodont’’ Marsupials

Three unquestionable species of polydolopimor-phian marsupials occur in the Santa Rosa fauna:Incadolops ucayali, Hondonadia pittmanae, andWirunodon tsullodon. All three taxa bear signifi-cant information concerning the evolutionary his-tory of the order, on which we comment below.

Incadolops ucayali is the most generalized pre-pidolopid known to date. Its upper molar structuresheds light on how the pairing of the paracone andmetacone with StB and StD was achieved. The cen-trocrista is still generalized for polydolopimorphi-ans, although already open. The stylar shelf is no-tably reduced, and a large trigon basin, typical ofother prepidolopids, is already present. Its tendencytoward bunodonty accounts for many secondarilyderived features present in representatives of thisorder. The generalized features of Incadolops, com-pared with other prepidolopids, create doubts as tothe inclusion of Rosendolops primigenium in thisfamily (see ‘‘Systematic Paleontology’’).

Hondonadia pittmanae is one of two marsupialsof the Santa Rosa local fauna related at the genericlevel with other marsupials from southern SouthAmerica (the other being Patene campbelli). Com-


Table 5 Suprageneric classification of microbiotheriansand polydolopimorphians (in its current concept; see textfor explanations). The inclusion of these two lineages ina higher category (‘‘taxon unnamed’’) implies that both ofthem (plus the ?Diprotodontia; see Discussion section) be-long to a natural group.

Taxon unnamedOrder Microbiotheria

Family MicrobiotheriidaeOrder Polydolopimorphia

Plesion GlasbiidaeSuborder Bonapartheriiformes, new rank

Superfamily Bonapartherioidea, new rankFamily PrepidolopidaeFamily Bonapartheriidae

Subfamily Bonapartheriinae, new rankSubfamily Epidolopinae

Family GashterniidaeFamily indeterminate

Hondonadia, RosendolopsSuperfamily Argyrolagoidea

Family GroeberiidaeFamily PatagoniidaeFamily Argyrolagidae

Bonapartheriiformes indeterminateWamradolops

Suborder PolydolopiformesPlesion RoberthoffstetteriaFamily Polydolopidae

?Order Diprotodontia

menting on the type species of the genus, H. fer-uglioi, we noted that ‘‘. . . the acquisition of thepolydolopimorphian molar pattern had alreadyhappened by Paleocene times, for example, Epidol-ops, Gashternia (Candela et al., 1995), or, at themost, by the early Eocene, for example, Prepidol-opidae, Bonapartheriidae (Pascual, 1980a, 1980b,1983). Hondonadia represents the second knowncase of the persistence of this same basic pattern inits most generalized state well into the Paleogene.The first known case was that of Rosendolops pri-migenium [Goin and Candela, 1996]. . . . Hondon-adia is more primitive than Rosendolops in havingstylar cusps B and D not so closely twinned withthe paracone and metacone, respectively, in that thestylar cusps are less compressed labio-lingually andin the persistence of a preprotoconal cingulum nearthe anterolingual face of the crown’’ (Goin andCandela, 1998:83; Spanish in the original). How-ever, in at least one feature, Hondonadia is morederived than Rosendolops: in the even more labialposition of the metacone relative to the paracone.This last feature, probably unique among metathe-rians, reinforces our previous argument that Rosen-dolops and Hondonadia are closely related. Bothgenera are here referred to as belonging to an un-determined family of the new taxon Bonaparther-ioidea (see below and Table 5).

Wamradolops tsullodon is enigmatic in that it

shares a number of derived features with severaldifferent lineages of polydolopimorphians (see‘‘Systematic Paleontology’’ and below). Severalstriking molar similarities between Wamradolopsand Groeberia have already been noted. Other den-tal elements, such as the incisors, have not yet beenrecovered for the Peruvian marsupial. Some con-spicuous, derived features of Wamradolops, such asthe enlarged, plagiaulacoid-like upper and lowerlast premolars, are not present in Groeberia. Final-ly, though strong and relatively deep, the horizontalramus in Wamradolops is not as deep or extensivelyinflected as in Groeberia, resembling more the ra-mus of bonapartheriids. Wamradolops also resem-bles the early Paleocene Roberthoffstetteria nation-algeographica in the lingual ‘‘displacement’’ of sty-lar cusps B, ?C, and D in the M1–3. Taking intoaccount the overall similarities between Wamra-dolops, the prepidolopids, and, especially, the bon-apartheriids (see ‘‘Systematic Paleontology’’), itcould be argued that groeberiids may represent anextremely specialized, later version of this originalstock. An alternative hypothesis is that Wamradol-ops belongs to a different polydolopimorphian lin-eage than that including prepidolopids and bona-partheriids. This hypothetical lineage might be re-lated to groeberiids, for which it represents a ple-siomorphic sister taxon, but at the same time, itmight be convergent on them to some extent.

Paucituberculatan marsupials from Santa Rosaare also crucial for interpreting the early stages inthe evolution of the order. Perulestes and Sasawatsuare important taxa for the interpretation of the evo-lution of the paucituberculatan molar pattern andadaptations. They demonstrate that the acquisitionof the quadrangular upper molar pattern in bothmajor lineages of South American ‘‘pseudodipro-todont’’ marsupials, that is, the Paucituberculataand the Polydolopimorphia, followed quite differ-ent pathways.

In his review of South American paucitubercu-latans, Marshall (1980) named as the ‘‘intermediatecuspule’’ the vestigial cusp that can be seen on thelingual side of the enlarged StD in upper molars ofgeneralized caenolestids, such as StilotheriumAmeghino, 1887, and Caenolestes Thomas, 1895.In later works (e.g., Marshall, 1987) he recognizedthat this cusp is homologous to the metacone,which becomes minute and almost fuses with theStD in these caenolestids. The Peruvian paucituber-culatans described here, especially P. cardichi, showan early stage in the reduction of the metacone,thus confirming the homologies of this cusp. In theupper teeth of this species, the metacone of M1–3 isonly moderately reduced, although clearly pairedwith StD, showing for this character the most ple-siomorphic condition known for the order. Themost interesting feature in the upper molars of Per-ulestes is, undoubtedly, the persistence of the para-cone, at least in M1–2. Although this cusp is obvi-ously paired with StB in P. cardichi, this taxon re-veals an early stage in the gradual reduction and


movement of the paracone toward StB, and the fi-nal ‘‘absorption’’ of the paracone by StB can beseen.

The Santa Rosa marsupials are very useful forrevealing differences between the upper molar pat-tern of the Paucituberculata and that of the Poly-dolopimorphia. In the latter, there is an even pair-ing of the para- and metacone with StB and StD,respectively. In the Prepidolopidae, as in the Bon-apartheriidae and Gashterniidae, the resulting pair-ing did not result in the disappearance of the para-or the metacone but, instead, gave a fusion of bothcusps with the stylar cusps. All of these cusps aresubequal in size and height, and their pairing im-plied a notable reduction of the labial half of theupper molars, that is, of the stylar shelf. As in pau-cituberculatans, the metaconule is strongly devel-oped, forming a hypocone-like structure. But thiscusp is crestless and level with the protocone, whichis quite different from the paucituberculatan meta-conule. Both the paracone and metacone are, inmost polydolopimorphians, positioned very labial-ly. In short, the quadrangular upper molar of po-lydolopimorphians is formed by the ‘‘paracone 1StB’’ in the parastylar corner, the ‘‘metacone 1StD’’ in the metastylar corner, the metaconule in theposterolingual corner, and the protocone on the an-terolingual face.

The acquisition of a quadrangular molar patternin the Paucituberculata followed a different evolu-tionary pathway. Here, the paracone was notablyreduced and, finally, it fused with StB to form asingle cusp, whereas the metacone persisted as anindependent cusp, even though reduced to the pointof being vestigial in even the most generalizedgroups (e.g., Caenolestinae). StB and StD are notreduced or equal in height to the paracone andmetacone; instead, they are relatively enormous,occupying most of the width of an unreduced stylarshelf. Also, the metaconule, even though it evolvedas a ‘‘hypocone,’’ persisted as a cusp, including thepre- and postmetaconular cristae. The tip of themetaconule is invariably higher than that of theprotocone, and it is placed lingual to it. The mor-phology of the upper molars of P. cardichi showshow this process unfolded (see ‘‘Systematic Pale-ontology’’). Thus, Perulestes represents an earlyversion of an evolutionary process experienceduniquely by the Paucituberculata among SouthAmerican ‘‘pseudodiprotodont’’ marsupials. A sim-ilar case is represented by Sasawatsu mahaynaq, thesecond genus of paucituberculatan marsupials rec-ognized here. It is difficult, however, to allocate Sas-awatsu to one of the recognized families of the or-der (see ‘‘Systematic Paleontology’’), even thoughthe available evidence suggests that Sasawatsu isnot a caenolestid but a more advanced paucituber-culatan, probably a palaeothentid. If this is thecase, it may be concluded that the ‘‘absorption’’ ofthe paracone by StB and the reduction and pairingof the metacone and StD were acquired indepen-

dently via convergence in different lineages of thisorder.

THE SANTA ROSA FAUNA AND ITSSIGNIFICANCE FOR UNDERSTANDING THEPOLYDOLOPIMORPHIAN RADIATION

In this section, we briefly discuss and integrate re-cent research on the evolution of the Polydolopi-morphia (Goin and Candela, 1998; Goin et al.,1998a; Goin, 2003; Goin et al., 2003) with the newevidence offered by the Peruvian taxa. On this ba-sis, a formal classification of the Polydolopimor-phia is proposed (Table 5).

1. It has already been argued that ‘‘pseudodipro-todont’’ marsupials, that is, the Pseudiprotodon-tia of Kirsch et al. (1997) that includes the Pau-cituberculata and the Polydolopimorphia, donot form a natural group. Instead, the earlyphases of their respective evolutionary histories,as revealed by the paucituberculatans Perulestesand Sasawatsu, on the one hand, and by Wam-radolops plus other early polydolopimorphianson the other, suggest the different origins, ho-mologies, and evolutionary processes experi-enced by each lineage. Upper and lower molarsynapomorphies of the Polydolopimorphia arethe following: upper molars show a great reduc-tion of the stylar shelf; the paracone and themetacone are paired with StB and StD, respec-tively; all these cusps are subequal in size andheight; the metaconule is large, ‘‘hypocone-like,’’equal in height to the protocone, and, in mostlineages, positioned on the lingual face of thetooth; lower molars have short talonids; the hy-poconid is not labially salient; and the cristidobliqua is parallel to the dental axis. It is clearto us that both the tempo and mode of evolutionof polydolopimorphian molar patterns, as wellas of the gliriform incisors and the plagiaula-coid-like cheek teeth (see, e.g., Marshall, 1980,1982), are indicative of separate origins for thelineages. As discussed below and elsewhere(Goin and Candela, 1998; Goin et al., 1998a;Goin, 2003), polydolopimorphian origins are re-lated to the microbiotherian radiation, whereasthe paucituberculatan origins are related tothose of the didelphimorphians.

2. Several groups regarded as polydolopimor-phians since Marshall’s (1987) review of the‘‘opossum-like marsupials’’ from the middle Pa-leocene Itaborai fauna of Brazil can be excludedfrom this order. Neither the Caroloameghiniidaenor the Protodidelphidae show the set of derivedfeatures that characterize the Polydolopimor-phia in its present concept. Goin et al. (1998b,1998c) have already argued for the peradectianaffinities of the Caroloameghiniidae and for theplacement of the Protodidelphidae within theDidelphimorphia. In our opinion, the Protodi-delphidae include only the following genera:Carolocoutoia Goin et al., 1998c; Robertbutler-


ia Marshall, 1987; Protodidelphis Paula Couto,1952a; and, probably, Guggenheimia PaulaCouto, 1952a.

3. Palangania brandmayri Goin et al., 1998a, fromthe upper Paleocene of Patagonia, sheds consid-erable light on the origins of the Polydolopi-morphia. Palangania shares several derived fea-tures with the Microbiotheriidae, including lowtrigonid cusps, posterolingually displaced meta-conid relative to the protoconid, and paraconidclearly smaller than the metaconid (althoughsubequal in height in Palangania), anterobasalcingulum very reduced or absent, reduced hy-poconulid, and large protocone. On the otherhand, it also displays derived features character-istic of the most primitive polydolopimorphians.It shows an incipient pairing of StB and StDwith the paracone and metacone, respectively;there is some reduction in width of the stylarshelf; and the preparacrista points to StA, thepostmetacrista is moderately developed, the cen-trocrista is vestigial, and the metaconule is welldeveloped (Goin et. al., 1998a). Most notably,Palangania shows, in most of these synapomor-phies, a molar pattern quite similar to that ofGlasbius Clemens, 1966, from the Upper Cre-taceous of North America, usually regarded asa peradectian. Goin et al. (1998a) suggested thatthe difficulties in recognizing whether P. brand-mayri is a Microbiotheria or a Polydolopimor-phia are a result of both lineages belonging to anatural group. In this paper, we formally pro-pose these previous suggestions and considerthat: (1) microbiotherians and polydolopimor-phians (and diprotodonts?, see below) belong toa natural group and (2) glasbiids (including Pa-langania, and probably Chulpasia Crochet andSige, 1993, from the lower Paleocene of Peru)constitute the plesiomorphic sister group of allother polydolopimorphians (Plesion Glasbiidae;see Table 5).

Upper and lower molar morphology of thebetter known Neogene microbiotherians seemsto be too derived to be considered ancestral forknown polydolopimorphians. Most microbio-theriid specimens are known from Miocene lev-els in Patagonia (e.g., Microbiotherium), as wellas from the living Dromiciops gliroides. TheSanta Rosa microbiotheriid Kirutherium paiti-tiensis does not shed light on this topic, as itshares closer affinities with Microbiotheriumand Dromiciops than with any other member ofthe order. If Khasia cordillerensis, from the low-er Paleocene of Bolivia, is a microbiothere (andnot a pediomyid peradectian, as also could beargued), it also is too derived to be regarded asa model of the generalized microbiothere molarpattern. Nonetheless, Goin et al. (1998a) re-ported the discovery of upper molars referableto the microbiotheriid Mirandatherium from themiddle upper Paleocene Las Flores Formation incentral Patagonia that have a morphology quite

different from the upper molars referred to thesame genus by Marshall (1987). These uppermolars seem to represent the plesiomorphic pat-tern for the Microbiotheria. Interestingly, theupper molars of Mirandatherium show derivedfeatures that, although incipiently developed,are suggestive of those observable in polydolo-pimorphians (e.g., reduced stylar shelves; dis-tinct stylar cusps StB and StD, which are incip-iently paired with the paracone and metacone,respectively; and an enlarged metaconule, butnot paraconule). Mirandatherium seems to beonly one step behind glasbiids, the latter regard-ed here as generalized polydolopimorphians.

4. A basic dichotomy can be recognized within thePolydolopimorphia. On the one hand is the sub-order Polydolopiformes (see Table 5), which in-cludes Roberthoffstetteria nationalgeographicaplus the polydolopines s.s (i.e., PolydolopsAmeghino, 1897, and allies). On the other hand,the remaining members of the order are placedin the suborder Bonapartheriiformes (new rank).Goin et al. (2003) mention several derived fea-tures exclusive of the former: the alignment ofthe paraconule, protocone, and metaconule in alingual row; well-expanded anterior and poste-rior cingula, which are level with the trigon ba-sin; variable occurrence of accessory cuspules onthe labial face of the upper molars; and labiallyplaced paraconid with respect to the metaconidin the lower molars. In turn, derived features ofthe Bonapartheriiformes include a larger meta-conule, the absence of StC and StE, and reduc-tion of the paraconid.

The long-accepted sister group relationshipbetween epidolopines and polydolopines, whichcomposed the Polydolopidae of traditional lit-erature, was challenged by Goin and Candela(1995). Epidolops shares with prepidolopids,bonapartheriids, Gashternia, and Wamradolopsthe basic set of derived features leading to theacquisition of the quadrangular molar pattern,while lacking the more advanced features thatrelate the groeberiids, argyrolagids, and pata-goniids in a group of their own (see below). Theaffinities between Epidolops and Gashterniawere suggested by Candela et al. (1995). In ourpresent state of knowledge, we think that theinclusion of Epidolops in the Bonapartheriidae,in its own subfamily, the Epidolopinae Marshall1982 (see Table 5), is the best systematic hy-pothesis at hand.

5. Three groups usually regarded as paucitubercu-latans (i.e., the groeberiids, argyrolagids, andpatagoniids) should be recognized as part of thePolydolopimorphia (suborder Bonapartheriifor-mes) in a group of their own: the Argyrolago-idea (Table 5). Klohnia charrieri, from lower Ol-igocene levels of central Chile, was consideredby Flynn and Wyss (1999) to be the closestknown relative of Patagonia Pascual and Carli-ni, 1987 (Patagoniidae), and they regarded both


taxa as the nearest outgroup to the Groeberi-idae. Although originally described as a pauci-tuberculatan, we find that Klohnia is, on thecontrary, a very interesting polydolopimorphian.Klohnia is derived in many features, includingthe possession of an enlarged, gliriform lowerincisor, loss of one molar (M4/M4?), a significantdiastema between I?1 and P3, and upper and low-er cheek teeth with a thick enamel layer. The lossof the last molar is uncommon among marsu-pials. This feature is present only in a few po-lydolopimorphians, such as the polydolopinesand the prepidolopid Punadolops alonsoi Goinet al., 1998b, even though the reduction of theM4/M4 is already noticeable in several othermembers of this order, including Bonaparther-ium, Prepidolops, Epidolops, and Wamradol-ops. The enlarged, anteriorly shifted entoconidof Klohnia (the ‘‘isolated lingual ‘talonid’ cusp’’of Flynn and Wyss, 1999:358) is also character-istic of Groeberia and Proargyrolagus Wolff,1984. This feature is already, although incipi-ently, developed in Wamradolops. In the uppermolars, the polydolopimorphian pairing of theparacone and metacone with StB and StD, to-gether with the enlargement of the metaconuleforming the posterolingual corner of the teeth,is also present in Klohnia (see Flynn and Wyss,1999:fig. 2G). Finally, a thick enamel layer ispresent in the cheek teeth of Bonapartherium,Prepidolops, Wamradolops, Gashternia, andEpidolops. Klohnia is even closer to groeberiids,patagoniids, and argyrolagids in its great en-largement of the first upper and lower incisors,the latter having an intra-alveolar extension thatis lingual to the cheek teeth series, and in thatthe last upper and lower premolars are not hy-pertrophied, but of normal size relative to themolars. On the other hand, the labial ‘‘shifting’’of the protocones in the upper molars (seeabove) relates Wamradolops more closely togroeberiids (and argyrolagids). In short, andpending a phylogenetic analysis of these taxa,we find that the molar morphologies of bothWamradolops and Klohnia represent an incipi-ent stage in the evolution of the more derivedGroeberiidae, Argyrolagidae, and Patagoniidae.Within the Argyrolagoidea, Koenigswald andGoin (2000) agree with Flynn and Wyss (1999),in that groeberiids and patagoniids are moreclosely related to each other than either is to theArgyrolagidae.

6. Within the Bonapartheriiformes we recognize abasic dichotomy. The Bonapartherioidea (newrank) include the Prepidolopidae, Bonaparther-iidae, Gashterniidae, Rosendolops, and Hon-donadia; and the Argyrolagoidea include theGroeberiidae, Patagoniidae, Argyrolagidae, andKlohnia. Derived features of the Bonaparther-ioidea include very large, plagiaulacoid-like lastpremolars. Derived features of the Argyrolago-idea include the great enlargement of the first

lower incisor, reduction of the P3/P3, metaconulelarger than the protocone, enlarged parastylararea, and lower molars with the paraconid an-teriorly displaced. Upper and lower molars arehypsodont in some lineages (e.g., the Patagoni-idae and most members of the Argyrolagidae).

7. The placement of microbiotherians and poly-dolopimorphians within a natural group maygive a new insight on the evolution of at leastsome of the Australian marsupial lineages. Asstated by Archer et al. (1999:7): ‘‘Although thereis general consensus, following studies of tarsalanatomy by Szalay (1982), that microbiotheri-ans share a special relationship with Australianmarsupials, there are arguments about the pre-cise nature of their placement within Australi-delphia, some concluding a special relationshipwith dasyuromorphians and peramelemorphians. . . , others a sister taxon relationship with das-yuromorphians . . . , dasyuromorphians and no-toryctids . . . , diprotodontians . . . , and stillothers with the whole of the Australian marsu-pial radiation . . . , or the Australian radiationplus the early South American genus Andino-delphys. . .’’

Unfortunately, there is no information avail-able for polydolopimorphians on molecular, cra-nial, or postcranial evidence in order to testthese earlier hypotheses because all members ofthis order are extinct and most are representedby fragmentary jaws, teeth, or the association ofboth. A few cranial remains are known for Ep-idolops and Bonapartherium (Paula Couto,1952b; Pascual, 1980a), but they are quite frag-mentary and, in the case of Bonapartherium,very badly preserved.

Ride (1993) argued for the derivation of themacropodid molar pattern from a microbioth-erian morphotype (see also Kirsch et al., 1997,and references cited in pp. 252–254). Ride’s ar-guments were based on the molar morphologyof Dromiciops. As is the case for polydolopi-morphians, the pattern seen in both Dromiciopsand Microbiotherium is already too derived toregard it as ancestral for the Diprotodontia. Ear-lier microbiotheriids, to the contrary, show anupper and lower molar pattern much more suit-able for inferring diprotodont origins. Paleocenemicrobiotherians, such as Mirandatherium andother, still undescribed taxa from the middle up-per Paleocene Las Flores Formation, show acombination of generalized and derived featuresquite suggestive of a pattern ancestral to dipro-todonts. Even some generalized polydolopimor-phians, such as the prepidolopid Incadolops,bear derived features that could be interpretedas diprotodontian synapomorphies. The char-acteristic W-shaped diprotodontian ‘‘ectoloph’’(preparacrista 1 postparacrista 1 premetacrista1 postmetacrista), for instance, is already pre-sent in Incadolops. In turn, the molar pattern ofboth prepidolopids and Mirandatherium sp.


would seem to exclude any possible derivationfrom them of dasyuromorphians, peramelemor-phians, or allied taxa.

COMMENTS ON THE AGE OF THE SANTAROSA MARSUPIALS

Tropical faunas, both extant and extinct, frequentlyinclude taxa that have plesiomorphic features rel-ative to contemporary, closely related taxa fromhigher latitudes (Goin, 1997). This is true, for ex-ample, in living didelphids. Lestodelphys halli(Thomas, 1921), by far the most specialized taxonof the clade composed of Thylamys Gray, 1843,Gracilinanus Gardner and Creighton, 1989, andLestodelphys Tate, 1934, is presently distributed al-most exclusively in Patagonia, ranging from south-ern Mendoza Province in western Argentina toeastern Santa Cruz Province in southernmost Ar-gentina. Gracilinanus, in turn, has most of its spe-cies distributed in middle latitudes, whereas themore generalized genera Marmosa Gray, 1821, andMicoureus Lesson, 1842, have most of their speciesdistributed in low, intertropical latitudes. Goin(1997) suggested a similar case for middle Miocene(Laventan SALMA) marsupials of La Venta, Co-lombia. Among the La Venta marsupials, the per-sistence of Pachybiotherium Ameghino, 1902, a ge-nus not present in the Patagonian fossil record afterColhuehuapian (early Miocene) time is noteworthy.Also, two thylacosmilid sparassodonts from LaVenta (Anachlysictis gracilis Goin, 1997, and aspecimen belonging to a still unnamed genus andspecies) represent successive ‘‘stages of evolution,’’or morphotypes, leading to the extremely special-ized craniodental pattern seen in late Miocene thy-lacosmilids from southern South America. The San-ta Rosa marsupials might represent a similar case.It appears that at least some marsupial taxa rep-resent the persistence of generalized types in strati-graphic levels younger than those deposits of south-ern South America that include closely related taxa.

It is also difficult to constrain the age of the sed-iments bearing this Amazonian marsupial fauna onthe basis of the recognized taxa because all are newspecies, and the overwhelming majority are new athigher taxonomic levels. Of the eleven species rec-ognized in this study, eight are assignable to newgenera, and only two (Patene campbelli and Hon-donadia pittmanae) can be assigned to previouslyknown genera. In addition, two of the Amazonianspecies, Wirunodon chanku and Kiruwamaq chisu,cannot be assigned even at ordinal level. Below,each taxon is discussed in detail.

The genus Patene is known from three species:P. colhuapiensis, from upper Eocene (CasamayoranSALMA) levels of Patagonia; P. simpsoni, whosebiochron ranges from the middle upper Paleocene(Itaboraian SALMA) of Patagonia and Brazil to theCasamayoran SALMA of northwestern Argentina(Goin et al., 1986; Bond et al., 1995); and the newspecies from Santa Rosa, P. campbelli. The Ama-

zonian species shares more similarities with P.simpsoni, which is, in turn, the most generalized ofthe genus.

The available material of K. chisu (an M?3) doesnot permit a secure assignment to any of the fam-ilies or orders of Metatheria currently recognized.A better knowledge of this species, however, mayestablish its relationship to the generalized sparas-sodonts of the family Hathliacynidae (see ‘‘System-atic Paleontology’’). But, Kiruwamaq is so uniquelyderived that its affinities, beyond its possible as-signment to this family, cannot be established. Ofall the hathliacynids compared with Kiruwamaq,the species that appeared to be the closest was theoldest and most generalized of the family: Allqok-irus australis Marshall and Muizon, 1988, from thelower Paleocene (Tiupampian SALMA) of Bolivia.

Alternative hypotheses on the peradectian or di-delphimorphian affinities of W. chanku are men-tioned above. One hypothesis is that Wirunodon isan early didelphimorphian, derived in some char-acters but not in others, such as the persistence ofa predilambdodont centrocrista. The dilambdo-dont, or V-shaped, centrocrista appears as early asthe early Paleocene (Tiupampian SALMA) in sev-eral South American opossums (Marshall et al.,1995, and references cited therein). However, indental terms, dilambdodonty is the main derivedfeature diagnostic of early didelphimorphians. Asecond hypothesis is that Wirunodon is, in fact, avery derived peradectian specialized toward fauni-vorous feeding habits. The known biochron of per-adectians in South America ranges from the lowerPaleocene of Laguna Umayo and Tiupampa, wherethe most generalized forms are recorded (Sige,1972; Muizon, 1991), to the last record of the spe-cialized Caroloameghiniidae in the late Eocene(Casamayoran SALMA). We suggested above thatWirunodon might be related to Kasserinotheriumtunisiense, which is known only from the lower Eo-cene of Chambi, in northern Africa (Crochet,1986).

The microbiotheriid Kirutherium paititiensisshares closer affinities with the Oligocene–MioceneMicrobiotherium and the living Dromiciops thanwith the oldest microbiotheriids hitherto recog-nized [i.e., Khasia cordillerensis from the lower Pa-leocene of Bolivia, or Mirandatherium, from themiddle upper Paleocene (Itaboraian SALMA) ofArgentina and Brazil]. However, the upper denti-tions of several Paleogene and Neogene micro-biotheriids, which could be even closer to Kiruth-erium than the genera mentioned above, are un-known. These include the Casamayoran SALMAEomicrobiotherium gaudryi and Pachybiotheriumacclinum Ameghino, 1902, from the lower Mio-cene (Colhuehuapian SALMA) of Patagonia.

Rumiodon inti is a ?herpetotheriid didelphimor-phian with several features in common with Paleo-gene species of marsupials of Europe. Among theEuropean forms, R. inti is comparable to Amphi-peratheium minutum (Aymard, 1846), whose


biochron ranges from the late Eocene to the middleOligocene, and with two Eocene species of Pera-therium: P. monspeliense Crochet, 1979, and P. ma-tronense Crochet, 1979 (Crochet, 1980). Likewise,R. inti has several features in common with stillunnamed species of herpetotheriines from the mid-dle upper Paleocene (Itaboraian SALMA) of Pata-gonia and Brazil (Oliveira, 1998; Goin and Fora-siepi, unpublished data).

Hondonadia pittmanae is recognized on the basisof a very fragmentary upper molar. The type speciesof the genus comes from Gran Hondonada (upper-most Eocene, Mustersan SALMA; Goin and Can-dela, 1998) in central Patagonia. Several specimenscoming from ?post-Mustersan to ?pre-‘‘Tinguiriri-can’’ levels at the classic southern cliffs of Lake Col-hue Huapi may also be referred to this genus (Goin,unpublished data).

Incadolops ucayali is slightly more generalizedthan other known prepidolopids (e.g., species ofPrepidolops and Punadolops) known from Casa-mayoran and Mustersan levels of northwestern Ar-gentina.

Wamradolops tsullodon is intermediate in sev-eral, but not all, features between prepidolopidsand bonapartheriids. Both groups are simulta-neously recorded in different, upper Eocene (Cas-amayoran and Mustersan SALMAs) sites in north-western Argentina. Some derived features alsosuggest that Wamradolops may represent an ances-tral lineage of the ?early Oligocene (DivisaderanSALMA) groeberiids from western Argentina, aswell as of a basal argyrolagoid, Klohnia charrieri,from the lowermost Oligocene (‘‘Tinguirirican’’age) of central Chile.

Both Perulestes and Sasawatsu are, in their gen-eralized features, comparable only to a new, stillundescribed caenolestid recovered from ‘‘Tinguirir-ican’’ age (lower Oligocene) levels at the southerncliffs of Lake Colhue Huapi in central Patagonia(Goin, unpublished data). The new Patagoniancaenolestid is represented by no less than a dozentiny, upper and lower molars. Interestingly, the up-per molars have their paracone even less reducedthan that of Perulestes and Sasawatsu, their ‘‘hy-pocone’’ is well developed in M1 but vestigial inM3, and there is some posterior displacement of themetacone relative to StD. In these last features, thePatagonian taxon is very similar to P. cardichi.

In summary, two of the Peruvian taxa show af-finities with genera of the Casamayoran SALMA.A third taxon related at the generic level with otherSouth American marsupials is Patene campbelli,which seems more closely related to a middle latePaleocene species (P. simpsoni). Other taxa (e.g.,Wirunodon and Rumiodon) seem to be pre-Casa-mayoran, whereas the remaining ones provide noconfident age-related information. Although it isevident that this fauna is middle Paleogene in age,it is impossible to assign it to any of the recognizedSALMAs of the South American biostratigraphicscheme. Taking into account all these consider-

ations, the most likely hypothesis is that the SantaRosa mammal-bearing levels are middle to upperEocene in age, although a lower Oligocene age,while improbable, cannot be discarded.

SUMARIO EN ESPANOL

Existe un marcado sesgo en nuestro conocimientode los mamıferos sudamericanos paleogenos debidoa dos factores principales: en primer lugar, la mayorparte de las localidades fosilıferas se localizan en elextremo austral de este continente (esto es, en laPatagonia; vease la Fig. 1); en segundo lugar, buenaparte de los mamıferos exhumados hasta ahora esde moderado a gran tamano. En parte, esto reflejatanto la historia de la paleomastozoologıa sud-americana, comenzando con los trabajos pionerosde Ameghino en la Argentina central y del sur,como tambien las tecnicas tradicionales de recolec-cion (busqueda de fosiles a ojo desnudo). En lo querespecta a los marsupiales sudamericanos, este ses-go es particularmente importante dado que la may-or parte de los marsupiales sudamericanos vivienteses de distribucion intertropical y que estos mamı-feros son, y muy probablemente lo han sido tam-bien en el pasado, formas de muy pequeno a me-diano tamano. Esto no esta siendo reflejado ac-tualmente en el registro paleogeno de los marsupi-ales. La utilizacion, en anos recientes, de tecnicasde lavado y tamizado sedimentario para la pros-peccion fosilıfera ha cambiado nuestra percepcionsobre la diversidad de estos mamıferos.

El segundo aspecto que oscurece un mejor con-ocimiento sobre la radiacion temprana de los mar-supiales sudamericanos es la ausencia de nivelesmamalıferos paleogenos de ubicacion intertropical.Al respecto, en este trabajo analizamos una nuevafauna de marsupiales procedente de niveles paleo-genos de la localidad de Santa Rosa, en la Ama-zonia peruana. La asociacion estudiada incluye 79especımenes, en su mayor parte molares superiorese inferiores aislados, muy pequenos, representativosde once nuevas especies referibles a todos los linajesmayores (ordenes) de marsupiales sudamericanos.La localidad fosilıfera de Santa Rosa (latitud98299390 S, longitud 728459480 W) se ubica 2 kmal norte del pueblo de Santa Rosa (Provincia Ata-laya, Departamento de Ucayali, Peru), sobre el mar-gen oeste del Rıo Yurua, cerca de su confluenciacon el Rıo Beu. Los fosiles proceden de depositosde canal expuestos en dicho margen. Campbell etal. (2004) sugieren que estos depositos pertenecena la Fm. Yahuarango, tambien conocida como Fm.Huayabamba, Capas Rojas inferiores, Puca inferi-or, o Fm. Huchpayacu, de posible edad Paleoceno-Eoceno. Los niveles fosilıferos corresponden a de-positos de canal de tipo calcareo, mayormente rel-lenos con concreciones calcareas, huesos, dientes,conchillas y material vegetal.

Los nuevos taxones aquı descriptos (Tabla 4; Re-sumen) constituyen una de las mas notables asocia-ciones de marsupiales descubiertas en los ultimos


anos. Su estudio agrega informacion significativasobre la diversidad paleogena de estos mamıferosy, especialmente, proporciona nuevas evidenciassobre las fases tempranas de la evolucion de losmarsupiales ‘‘pseudodiprotodontes’’ (sensu Ride,1962), esto es, los Polydolopimorphia y los Pauci-tuberculata. Rumiodon inti gen. et sp. nov. (Didel-phimorphia, ?Herpetotheriidae) presenta sus mo-lares superiores mucho mas largos que anchos, conlos protoconos reducidos y postmetacristas bien de-sarrolladas. Patene campbelli sp. nov. (Sparasso-donta, Hathliacynidae) difiere de las otras especiesdel genero por su menor tamano y por el desarrollointermedio del metacono y del cıngulo anterobasalen el M4. Incadolops ucayali gen. et sp. nov. (Po-lydolopimorphia, Prepidolopidae) muestra rasgosgeneralizados en relacion con otros prepidolopidos(metaconulo mas desarrollado, menor diferencia dealtura entre el protocono, por un lado, y el para-cono y el metacono por el otro). Wamradolops tsul-lodon gen. et sp. nov. (Polydolopimorphia, fam. in-det.) es intermedio en varios rasgos (pero no todos)entre los prepidolopidos y los bonaparteridos. Losmolares superiores de Hondonadia pittmanae sp.nov. (Polydolopimorphia) tienen el metacono may-or que el paracono y mucho mas labialmente ubi-cado. Perulestes cardichi y P. fraileyi gen. et spp.nov. (Paucituberculata, Caenolestidae) son gener-alizados para todo el Orden Paucituberculata alpresentar el protoconido del M1 mucho mas altoque el metaconido y, en los molares superiores, elparacono aun presente, si bien ya muy reducido yfusionado basalmente a la cuspide estilar B en losM1–2. Sasawatsu mahaynaq gen. et sp. nov. (Pau-cituberculata, cf. Palaeothentidae) presenta unacombinacion de rasgos primitivos y derivados: mo-lares bunoides, sin crestas, con la persistencia delparacono y metacono como cuspides individuales.Kirutherium paititiensis gen. et sp. nov. (Micro-biotheria, Microbiotheriidae) difiere de otros mi-crobioteridos en que en los molares superiores per-sisten paraconulos y metaconulos vestigiales y elon-gados, parcialmente fusionados con la pre- y lapostprotocrista; la cuenca del trıgono es muy pro-funda y la postmetacrista esta aparentemente pocodesarrollada. Wirunodon chanku gen. et sp. nov.(orden y familia indet.) es una minuscula, enigma-tica ‘‘comadreja’’ que muestra caracteres tantoprimitivos (e.g., molares superiores con la centro-crista casi recta) como derivados (e.g., protocono ycuenca del trıgono reducidos, amplia plataformaestilar, postmetacrista bien desarrollada, prepara-crista conectando con la cuspide estilar A). Kiru-wamaq chisu gen. et sp. nov. (orden y familia in-det.) difiere del resto de los marsupiales sudameri-canos en que los molares superiores tienen el pro-tocono muy reducido, el paracono cercano almetacono, mucho mas pequeno que este ultimo ymas labialmente orientado, la centrocrista recta, lapostmetacrista bien desarrollada y la plataforma es-tilar muy amplia.

Casi todos los ejemplares recuperados en Santa

Rosa son de tamano pequeno a muy pequeno. Pa-tene campbelli, del tamano de un gato, es el masgrande hasta ahora registrado. Le siguen los Pau-cituberculata Perulestes fraileyi y S. mahaynaq, deltamano de una ardilla. El resto es una serie de for-mas similares a pequenos ratones, las que proba-blemente no excedıan en peso a aquel del marsupialviviente Dromiciops gliroides (entre 20 y 50 gra-mos). En relacion con sus habitos alimentarios, esllamativa la gran proporcion de formas cuya mor-fologıa dentaria implica dietas frugıvoras o frugı-voro–insectıvoras. Su tamano y habitos inferidossugieren que buena parte de ellas pudo haber ten-ido un estilo de vida arborıcola. Un aspecto adi-cional de la asociacion de Santa Rosa procede dela evidencia negativa: no se registra un solo mar-supial polidolopino (Polydolopimorphia, Polydol-opidae). Estos ultimos, cuya dentadura es similar ala de los extintos multituberculados, son muy abun-dantes en niveles paleogenos de la Patagonia, perono aparecen nunca en latitudes mas bajas. Pareceexistir cierta relacion, probablemente trofica, entrelos polidolopinos y las floras australes subantarticasdominadas por especies del genero Nothofagus,junto con podocarpaceas y araucarias.

El analisis de los marsupiales Polydolopimorphiay Paucituberculata exhumados en Santa Rosa escrucial para interpretar la evolucion temprana delos representantes de ambos grupos sudamericanosde ‘‘pseudodiprotodontes,’’ los que, a la luz de estosresultados, no conforman un grupo natural. El pa-tron cuadrangular de los molares superiores en am-bos grupos es similar, aunque el camino evolutivoque siguio cada uno de ellos es diferente. En losPolydolopimorphia se aprecia el apareamiento delparacono y el metacono con las cuspides estilaresB y D, respectivamente. En las formas basales deeste grupo, dicho apareamiento no produjo la de-saparicion del para- ni del metacono, sino en cam-bio la fusion en la base de estas cuspides con lasestilares. Todas estas cuspides son subiguales en ta-mano y altura, y su apareamiento implico una no-table reduccion de la region labial (i.e., la platafor-ma estilar) de los molares superiores. Al igual queen los Paucituberculata, el metaconulo esta muybien desarrollado, formando una estructura similara un hipocono. Sin embargo, a diferencia de losPaucituberculata, su metaconulo no tiene crestas yesta nivelado con el protocono; ambas cuspides es-tan ubicadas al mismo nivel sobre la cara lingualdel diente. En suma, el patron cuadrangular delmolar superior de los polidolopimorfios esta com-puesto por el paracono y la StB en la esquina par-astilar, el metacono y la StD en la metastilar, el me-taconulo agrandado en la esquina posterolingual yel protocono en la anterolingual. La adquisicion deun molar superior cuadrangular en los Paucituber-culata siguio una distinta vıa evolutiva. En estos, elparacono se redujo notablemente, hasta fusionarsecon la StB en una unica cuspide, mientras que elmetacono persistio como cuspide independiente, sibien se redujo al punto de quedar vestigial, incluso


en los grupos mas generalizados (i.e., Caenolesti-dae) sobre la ladera lingual de la StD. Por su parte,las cuspides estilares B y D no se redujeron o ni-velaron en altura con el paracono y el metacono,sino que son de tamano muy grande y ocupan lamayor parte de la (no reducida) plataforma estilar.El metaconulo, que tambien evoluciono a modo de‘‘hipocono,’’ persistio como cuspide, incluyendo suscrestas pre- y postmetaconulares. Su cuspide es in-variablemente mas alta que aquella del protoconoy se ubica mas lingualmente que este ultimo.

El estudio de los marsupiales polidolopimorfiosde Santa Rosa permite incrementar notablementenuestro conocimiento sobre las relaciones interge-nericas e incluso interfamiliares de este grupo (Ta-bla 5). Nuestras conclusiones son las siguientes.

1. Los Microbiotheria, Polydolopimorphia y, po-siblemente, los Diprotodontia de Australasiaconforman un grupo natural (el ‘‘Taxon innom-inado’’ de la Tabla 5).

2. Los Glasbiidae (representados en el Cretacicotardıo norteamericano por el genero Glasbius,en el Paleoceno de la Patagonia por Palanganiay, probablemente, en el Paleoceno temprano dePeru por Chulpasia) constituyen el grupo her-mano plesiomorfo de todos los restantes poli-dolopimorfios.

3. Los representantes de este Orden muestran lasiguiente serie de sinapomorfıas: molares super-iores con la plataforma estilar reducida, para-cono y metacono apareados con las cuspides es-tilares B y D, respectivamente, todas estas cus-pides son subiguales en tamano y altura; meta-conulo agrandado y a la misma altura que elprotocono y, en casi todos los linajes, ocupandotoda la esquina posterolingual del diente; mo-lares inferiores con los talonidos cortos, el hi-poconido poco saliente labialmente y con la crıs-tida obliqua paralela al eje dentario.

4. Varios grupos que habıan sido incluidos en esteOrden desde la revision de Marshall (1987) de-berıan ser excluidos del mismo. Por ejemplo, nilos Caroloameghiniidae ni los Protodidelphidaemuestran la serie de rasgos derivados que car-acteriza a los Polydolopimorphia en su concep-cion actual.

5. Se reconoce una dicotomıa basica entre los re-presentantes de este Orden: por un lado, Rob-erthoffstetteria nationalgeographica mas los pol-idolopinos s.s. (i.e., Polydolops y formas afines),los que se agrupan aquı en el Suborden Poly-dolopiformes; sinapomorfıas de este grupo son,en los molares superiores, el alineamiento delparaconulo, protocono y metaconulo en una hil-era lingual, el desarrollo de amplios cıngulos an-terior y posterior, nivelados con la cuenca deltrıgono y una variable existencia de cuspulas es-tilares accesorias; en los molares inferiores, laubicacion labial del paraconido con respecto almetaconido. En segundo lugar, los restantesmiembros del Orden componen el nuevo Subor-

den Bonapartheriiformes, cuyos rasgos deriva-dos diagnosticos incluyen un metaconulo muyagrandado, ausencia de las cuspides estilares Cy E y la reduccion del paraconido.

6. Tres familias generalmente asignadas a los Pau-cituberculata son aquı incluidas entre los Bona-partheriiformes: Groeberiidae, Argyrolagidae yPatagoniidae. Estas familias comprenden la Su-perfamilia Argyrolagoidea, cuyos rasgos deri-vados incluyen el agrandamiento del primer in-cisivo inferior, la reduccion de los ultimos pre-molares superiores e inferiores, la presencia deun metaconulo mayor que el protocono, un areaparastilar agrandada y el desplazamiento ante-rior del paraconido en los molares inferiores. Elresto de los Bonapartheriiformes (los Prepidol-opidae, Bonapartheriidae, Gashterniidae, Ro-sendolops y Hondonadia) se incluyen en la Su-perfamilia Bonapartherioidea, cuyos rasgos de-rivados incluyen la posesion de ultimos premo-lares muy grandes, de tipo ‘‘plagiaulacoideo.’’

La estimacion de la edad de los sedimentos fos-ilıferos de Santa Rosa es problematica. Por un lado,es evidente que la fauna de Santa Rosa representaun Paleogeno medial; por el otro, sin embargo, noes posible referirla con certeza a ninguna de las eda-des-mamıfero sudamericanas (SALMAs) reconoci-das para el Paleogeno de este continente. Dos delos taxones de Santa Rosa muestran afinidades anivel generico con formas de la SALMA Casama-yorense. Un tercer taxon, la nueva especie Patenecampbelli, remite a formas afines del Paleoceno me-dial (P. simpsoni). Otros taxones (e.g., Wirunodon,Rumiodon) parecen ser pre-Casamayorenses, mien-tras que los restantes no proveen informacion sig-nificativa en torno a la posible edad de esta fauna.Teniendo en cuenta estos datos, como ası tambienresultados recientes sobre la datacion absoluta dela Subedad Barranquense de la SALMA Casama-yorense, nuestra hipotesis mas plausible es que lossedimentos de Santa Rosa fueron depositados dur-ante el Eoceno medio o tardıo aunque una edadOligoceno temprano no deberıa descartarse porcompleto.

ACKNOWLEDGMENTS

We thank Kenneth E. Campbell and Carl D. Frailey formaking available to us the marsupial specimens from San-ta Rosa for study and for many helpful comments. PatriciaSarmiento and Rafael Urrejola helped us make the SEMphotos. Augusto Cardich kindly helped us with Quechuavocabulary, which we used in the naming of the new taxa.Discussions with Marcelo Reguero and Mariano Bond,from the Museo de La Plata, regarding the age of SantaRosa mammal fauna, were enlightening. Finally, we thankChristian de Muizon and an anonymous reviewer for theiruseful comments and criticisms on the manuscript.

LITERATURE CITED

Ameghino, F. 1887. Enumeracion sistematica de las es-pecies de mamıferos fosiles coleccionados por CarlosAmeghino en los terrenos eocenos de Patagonia aus-


tral y depositados en el Museo de La Plata. Boletındel Museo de La Plata 1:1–26.

. 1890. Los plagiaulacıdeos argentinos y sus rela-ciones zoologicas, geologicas y geograficas. Boletındel Instituto Geologico Argentino 11:143–201.

. 1894. Enumeration synoptique des especies demammiferes fosiles des formations eocenes de Pata-gonie. Boletın de la Academia de Ciencias de Cor-doba 13:259–452.

. 1897. Mammiferes cretaces de l’Argentine (Deux-ieme contribution a la connaissance de la faunemammalogique des Couches a Pirotherium). Boletındel Instituto Geologico Argentino 18:406–521.

. 1902. Premiere contribution a la connaissance dela faune mammalogique des couches a Colpodon.Boletın de la Academia de Ciencias de Cordoba 17:71–141.

. 1904. Nuevas especies de mamıferos cretacicos yterciarios de la Republica Argentina. Anales de laSociedad Cientıfica Argentina 56:193–208 (1903);57:162–175, 327–341 (1904); 58:35–71, 182–192,225–291 (1904).

Archer, M., R. Arena, M. Bassarova, K. Black, J. Bram-mall, B. Cooke, P. Creaser, K. Crosby, A. Gillespie,H. Godhelp, M. Gott, S. Hand, B. Kear, A. Krik-mann, B. Mackness, J. Muirhead, A. Musser, T. My-ers, N. Pledge, Y. Wang, and S. Wroe. 1999. Theevolutionary history and diversity of Australianmammals. Australian Mammalogy 21:1–45.

Aymard, A. 1846. Essai monographique sur un genre nou-veau de mammifere fossile trouve dans la Haute-Loire et nomme Entelodon. Annales des SocieteAgricole du Puy 12:227–267.

. 1850. Compte rendu de la seance du 13 avril1849, reponse a M. Robert. Annales des SocieteAgricole du Puy 14:80–86.

Blume, K.L. 1850. Lugduno-Batavum sive stirpium exo-ticarum, novarum vel minus cognitarum ex vivis autsiccis brevis expositio et descriptio. Museum Botan-icum Leiden 1(15):225–240.

Bond, M., A.A. Carlini, F.J. Goin, L. Legarreta, E. OrtizJaureguizar, R. Pascual, and M.A. Uliana. 1995. Ep-isodes in South American land mammal evolutionand sedimentation: testing their apparent concomi-tance in a Paleocene succession from Central Pata-gonia. VI Congreso Argentino de Paleontologıa yBioestratigrafıa, Trelew, Actas: 47–58.

Campbell, K.E., Jr., C.D. Frailey, and L. Romero-Pittman.2004. The Paleogene Santa Rosa local fauna of Am-azonian Peru: Geographic and geologic setting. InThe Paleogene mammalian fauna of Santa Rosa,Amazonian Peru, ed. K.E. Campbell, Jr. Natural His-tory Museum of Los Angeles County, Science Series40, 3–14.

Candela, A.M., F.J. Goin, and R. Pascual. 1995. Los Po-lydolopimorphia (Mammalia, Marsupialia) de laFormacion Las Flores (Paleoceno medio, Patagoniacentral, Argentina). Acta Geologica Lilloana 18(1):149.

Case, J.A. 1988. Paleogene floras from Seymour Island,Antarctic Peninsula. Geological Society of America,Memoir 169:505–521.

Clemens, W.A. 1966. Fossil mammals of the Type LanceFormation, Wyoming. Part II: Marsupialia. Univer-sity of California Publications in Geological Sciences62:1–122.

Cope, E.D. 1873. Third notice of extinct vertebrata from

the Tertiary of the plains. Paleontological Bulletin16:1–8.

Crochet, J.-Y. 1979. Les marsupiaux fossiles d’Europe. 78Reunion Annuelle des Sciences de la Terre, Lyon, p.137.

. 1980. Les Marsupiaux du Tertiaire d’Europe. Par-is: Editions de la Fondation Singer-Polignac, 279 pp.

. 1986. Kasserinotherium tunisiense nov. gen., nov.sp., troisieme marsupial decouvert en Afrique (Eo-cene inferieur de Tunisie). Comptes Rendus del’Academie des Sciences, Paris 302(14):923–926.

Crochet, J.-Y., and B. Sige. 1983. Les mammiferes Mon-tiens de Hainin (Paleocene moyen de Belgique) PartIII: Marsupiaux. Palaeovertebrata 13(3):51–64.

. 1993. Les mammiferes de Chulpas (formationUmayo, transition Cretace-Tertiaire, Perou): donneespreliminaires. Documents du Laboratoire de Geo-logie 125:97–107.

Filhol, H. 1876. Mammiferes fossiles nouveaux proven-nant des depots de phosphate de Chaux du Quercy.Comptes Rendus de l’Academie des Sciences, Paris82:289.

. 1877. Recherche sur les Phosphorites du Quercy.Annales de la Societe Geologique 7:1–220; 8(1):1–340.

. 1879. Etude des mammiferes fossiles de St-Ger-and-le-Puy. Biblioteque Ecole des Hautes Etudes 10:1–253.

Flynn, J.J., and A.R. Wyss. 1999. New Marsupials fromthe Eocene–Oligocene transition of the AndeanMain Range, Chile. Journal of Vertebrate Paleontol-ogy 19(3):533–549.

Gardner, A.L., and G.K. Creighton. 1989. A new genericname for Tate’s (1933) microtarsus group of SouthAmerican mouse opossums (Marsupialia, Didelphi-dae). Proceedings of the Biological Society of Wash-ington 102(1):3–7.

Geoffroy Saint-Hilaire, E. 1795. Untitled. La Decade Phi-losophique, Litteraire et Politique: Par une Societedes Republicains 4(28):195.

Gervais, P. 1848–1852. Zoologie e Paleontologie Fran-caises (animaux vertebres) ou nouvelles recherchessur les animaux vivantes et fossiles de la France, vol.1, 1st ed. Paris: Arthus Bertrand, 271 pp.

Goin, F.J. 1991. Los Didelphoidea (Mammalia, Marsupi-alia, Didelphimorphia) del Cenozoico tardıo de laRegion Pampeana. Unpublished thesis, UniversidadNacional de La Plata, 327 pp.

. 1994. Living South American opossums are notliving fossils. VI International Theriological Con-gress (Univ. Of South Wales, Kensington), Sydney,Abstracts, p. 112.

. 1995. Los Marsupiales. In Evolucion biologica yclimatica de la Region Pampeana durante los ultimoscinco millones de anos. Un ensayo de correlacioncon el Mediterraneo occidental, ed. M.A. Alberdi, G.Leone, and E.P. Tonni, 165–179. Madrid: MuseoNacional de Ciencias Naturales/CSIC.

. 1997. New clues for understanding Neogene mar-supial radiations. In A history of the Neotropicalfauna. Vertebrate paleobiology of the Miocene inColombia, ed. R.F. Kay, R.H. Madden, R.L. Cifelli,and J. Flynn, 185–204. Washington, D.C.: Smithson-ian Institution Press.

. 2003. Early marsupial radiations in South Amer-ica. In Predators with pouches: The biology of car-nivorous marsupials, ed. M. Jones, C. Dickman, and


M. Archer, 30–42. Colingwood: CSIRO (Common-wealth Scientific and Industrial Research Organiza-tion).

Goin, F.J., and A.M. Candela. 1995. Una nueva especiede Epidolops Paula Couto, 1952 (Marsupialia, Po-lydolopimorphia, Polydolopidae). Consideracionessobre el patron molar inferior de los Epidolopidae.VI Congreso Argentino de Paleontologıa y Bioestra-tigrafıa, Trelew, Actas, pp. 143–148.

. 1996. A new Early Eocene polydolopimorphian(Mammalia, Marsupialia) from Patagonia. Journalof Vertebrate Paleontology 16(2):292–296.

. 1997. New Patagonian marsupials from Amegh-ino’s ‘‘Astraponoteen Plus Superieur’’ (post-Muster-san–pre-Deseadan age). Journal of Vertebrate Pale-ontology 17(3)(Suppl.):2.

. 1998. Dos nuevos marsupiales ‘‘Pseudodiproto-dontes’’ del Paleogeno de Patagonia. In Paleogeno deAmerica del Sur y de la Penınsula Antartica. Asocia-cion Paleontologica Argentina, Buenos Aires, Publi-cacion Epecial, 5:79–84.

Goin, F.J., A.M. Candela, M. Bond, R. Pascual, and V.Escribano. 1998a. Una nueva ‘‘comadreja’’ (Mam-malia, Marsupialia, ?Polydolopimorphia) del Paleo-ceno de Patagonia, Argentina. In Paleogeno deAmerica del Sur y de la Penınsula Antartica. Asocia-cion Paleontologica Argentina, Buenos Aires, Publi-cacion Especial, 5:71–78.

Goin, F.J., A.M. Candela, and C. de Muizon. 2003. Theaffinities of Roberthoffstetteria nationalgeographica(Marsupialia) and the origin of the polydolopine mo-lar pattern. Journal of Vertebrate Paleontology23(4):869–876.

Goin, F.J., A.M. Candela, and A. Forasiepi. 1997. New,middle Paleocene marsupials from central Patagonia.Journal of Vertebrate Paleontology 17(3 Suppl.):49A.

Goin, F.J., A.M. Candela, and G. Lopez. 1998b. MiddleMiocene marsupials from Antofagasta de la Sierra,northwestern Argentina. Geobios 31(1):75–85.

Goin, F.J., E.V. Oliveira, and A.M. Candela. 1998c. Car-olocoutoia ferigoloi n. gen. et sp. (Protodidelphidae),a new Paleocene ‘‘opossum-like’’ marsupial fromBrazil. Palaeovertebrata 27(3–4):145–154.

Goin, F.J., R.M. Palma, R. Pascual, and J. Powell. 1986.Persistencia de un primitivo Borhyaenidae (Mam-malia, Marsupialia) en el Eoceno temprano de Salta.Aspectos geologicos y paleoambientales relaciona-dos. Ameghiniana 23(1–2):47–56.

Goin, F.J., M. Woodburne, J. Case, S.F. Vizcaıno, and M.Reguero. 1999. New discoveries of ‘‘opossum-like’’marsupials from Antarctica (Seymour Island, MiddleEocene). Journal of Mammalian Evolution 6(4):335–365.

Gray, J.E. 1821. On the natural arrangement of vertebroseanimals. London Medical Repository 15(1):296–310.

. 1843. List of the specimens of Mammalia in thecollection of the British Museum. London: BritishMuseum of Natural History, 266 pp.

. 1858. List of species of Mammalia sent from theAru Islands by Mr. A.R. Wallace to the British Mu-seum. Proceedings of the Zoological Society of Lon-don 1858:106.

Hennig, W. 1968. Elementos de una sistematica filogene-tica. Buenos Aires: Eudeba, 353 pp.

Hershkovitz, P. 1982. The staggered marsupial lower thirdincisor (I3). Geobios, memoire special 6:191–200.

. 1999. Dromiciops gliroides Thomas, 1894, last ofthe Microbiotheria (Marsupialia), with a review ofthe Family Microbiotheriidae. Fieldiana Zoology(New Series) 93:1–60.

Kay, R.F., R.H. Madden, M.G. Vucetich, A.A. Carlini,M.M. Mazzoni, G.H. Re, M. Heizler, and H. San-deman. 1999. Revised age of the CasamayoranSouth American Land Mammal ‘‘Age’’—Climaticand biotic implications. Proceedings of the NationalAcademy of Sciences, Washington 96(23):13235–13240.

Kirsch, J.A.W., F.-J. Lapointe, and M. Springer. 1997.DNA-hybridization studies of marsupials and theirimplications for metatherian classification. Austra-lian Journal of Zoology 45:211–280.

Koenigswald, W.v., and F.J. Goin. 2000. Enamel differ-entiation in South American marsupials and a com-parison of placental and marsupial enamel. Palaeon-tographica A255:129–168.

Korth, W.W. 1994. Middle Tertiary marsupials (Mam-malia) from North America. Journal of Paleontology68:376–396.

Lesson, R.P. 1842. Nouveau tableau du regne animal.Mammiferes. Paris: Arthus Bertrand, 204 pp.

Loomis, F.B. 1914. The Deseadan Formation of Patagon-ia. Concord, New Haven: The Rumford Press, 232pp.

Marshall, L.G. 1980. Systematics of the South Americanmarsupial family Caenolestidae. Fieldiana Geology(New Series) 5:1–145.

. 1982. Systematics of the extinct South Americanmarsupial family Polydolopidae. Fieldiana Geology(New Series) 12:1–109.

. 1984. The lower jaw of Eobrasilia coutoi Simp-son, 1947, a unique Didelphoid (not Borhyaenoid)marsupial from the Paleocene of Brasil. Journal ofPaleontology 58:173–177.

. 1987. Systematics of Itaboraian (middle Paleo-cene) age ‘‘opossum-like’’ marsupials from the lime-stone quarry at Sao Jose de Itaborai, Brasil. In Pos-sums and opossums: Studies in evolution, vol. 1., ed.M. Archer, 91–160. Sydney: Surrey Beatty & SonsPty Limited and The Royal Zoological Society ofNew South Wales.

Marshall, L.G., J.A. Case, and M.O. Woodburne. 1990.Phylogenetic relationships of the families of marsu-pials. In Current mammalogy, vol. 2, ed. H. Geno-ways, 433–505. New York: Plenum Press.

Marshall, L.G., and C. de Muizon. 1988. The dawn ofthe age of mammals in South America. NationalGeographic Research 4(1):23–55.

Marshall, L.G., C. de Muizon, and B. Sige. 1983. LateCretaceous mammals (Marsupialia) from Bolivia.Geobios 16:739–745.

Marshall, L.G., C. de Muizon, and D. Sigogneau-Russell.1995. Pucadelphys andinus (Marsupialia, Mamma-lia), from the early Paleocene of Bolivia. Memoiresdu Museum National d’Histoire Naturelle 165:1–164.

Marshall, L.G., R. Hoffstetter, and R. Pascual. 1983.Mammals and stratigraphy: Geochronology of thecontinental mammal-bearing Tertiary of SouthAmerica. Palaeovertebrata, Memoire Extraordinaire1983:1–93.

Marshall, L.G., and R. Pascual. 1977. Nuevos marsupi-


ales Caenolestidae del ‘‘Piso Notohippidense’’ (SWde Santa Cruz, Patagonia) de Ameghino. Sus apor-taciones a la cronologıa y evolucion de las comuni-dades de mamıferos sudamericanos. Publicacion delMuseo Municipal de Ciencias Naturales de Mar delPlata (Lorenzo Scaglia) 2(4):91–122.

Marshall, L.G., and T. Sempere. 1993. Evolution of theNeotropical Cenozoic land mammal fauna in its geo-chronologic, stratigraphic, and tectonic context. InBiological relationships between Africa and SouthAmerica, ed. P. Goldblatt, 328–392. New Haven,Connecticut, and London: Yale University Press.

Matthew, W.D., and W. Granger. 1921. New genera ofPaleocene mammals. American Museum Novitates13:1–7.

Moss-Salentijn, L., M.L. Moss-Salentijn, and M. Sheng-Tien Yuan. 1997. The ontogeny of mammalianenamel. In Tooth enamel microstructure, ed. W.v.Koenigswald and P.M. Sanders, 5–30. Rotterdam:A.A. Balkema.

Muizon, C. de. 1991. La fauna de mamıferos de Tiupam-pa (Paleoceno Inferior, Formacion Santa Lucia), Bo-livia. In Fosiles y Facies de Bolivia. Vol. 1. Vertebra-dos, ed. R. Suarez-Soruco. Revista Tecnica de Yaci-mientos Petrolıferos Fiscales Bolivianos, 12(3–4):575–624.

. 1994. A new carnivorous marsupial from the Pa-leocene of Bolivia and the problem of marsupialmonophyly. Nature 370:208–211.

Oliveira, E.V. 1998. Taxonomia, Filogenia e Paleobiogeo-grafia de marsupiais ‘‘Poliprotodontes’’ do Mesopa-leoceno da Bacia de Itaborai, Rio de Janeiro, Brasil.Unpublished thesis, Universidade Federal do RioGrande do Sul, Porto Alegre, 327 pp.

Oliveira, E.V., and F.J. Goin. In press. Marsupiais fosseisdo Paleoceno de Itaboraı: Origem, Irradiacao e His-toria Biogeografica. In Marsupiais do Brasil, ed.N.C. Caceres. Rio de Janeiro: Universidade Federaldo Rio de Janeiro.

Pascual, R. 1980a. Nuevos y singulares tipos ecologicosde marsupiales extinguidos de America del Sur (Pa-leoceno Tardıo o Eoceno Temprano) del NoroesteArgentino. II Congreso Argentino de Paleontologıay Bioestratigrafıa, Buenos Aires, Actas 2:151–173.

. 1980b. Prepidolopidae, Nueva Familia de Mar-supialia Didelphoidea del Eoceno Sudamericano.Ameghiniana 17(3):216–242.

. 1983. Novedosos marsupiales paleogenos de laFormacion Pozuelos (Grupo Pastos Grandes) de laPuna, Salta, Argentina. Ameghiniana 20(3–4):265–280.

Pascual, R., and A. A. Carlini. 1987. A new superfamilyin the extensive radiation of South American Paleo-gene marsupials. Studies in Neotropical Mammalo-gy: Essays in Honor of Philip Hershkovitz. FieldianaZoology 39:99–110.

Pascual, R., F.J. Goin, and A.A. Carlini. 1994. New dataon the Groeberiidae: Unique late Eocene–early Oli-gocene South American marsupials. Journal of Ver-tebrate Paleontology 14(2):247–259.

Pascual, R., E. Ortiz Jaureguizar, and J.L. Prado. 1996.Land mammals: Paradigm for Cenozoic SouthAmerican geobiotic evolution. Munchner Geowis-senschaftlich Abhandlungen (A) 30:265–319.

Patterson, B. 1952. Un nuevo y extraordinario marsupialdeseadiano. Revista del Museo Municipal de Cien-

cias Naturales y Tradicional de Mar del Plata 1:39–44.

Paula Couto, C. de. 1952a. Fossil mammals from the be-ginning of the Cenozoic in Brazil. Marsupialia: Di-delphidae. American Museum Novitates 1567:1–26.

. 1952b. Fossil mammals from the beginning of theCenozoic in Brazil. Marsupialia: Polydolopidae andBorhyaenidae. American Museum Novitates 1559:1–27.

. 1952c. A new name for Mirandaia riberoi PaulaCouto 1952. Journal of Mammalogy 33(4):503.

. 1962. Didelfıdeos fosiles del Paleoceno de Brasil.Revista del Museo Argentino de Ciencias Naturales‘‘Bernardino Rivadavia,’’ Zoologıa 8(12):137–166.

Peters, D.S., and G. Storch. 1993. South American rela-tionships of Messel birds and mammals. Kaupia 3:263–269.

Reig, O.A. 1981. Teorıa del origen y desarrollo de la fau-na de mamıferos de America del Sur. MonographiaeNaturae 1:1–162.

Reig, O.A., J.A.W. Kirsch, and L.G. Marshall. 1987. Sys-tematic relationships of the living and NeocenozoicAmerican ‘‘opossum-like’’ marsupials (Suborder Di-delphimorphia), with comments on the classificationof these and of the Cretaceous and Paleogene NewWorld and European metatherians. In Possums andopossums: Studies in evolution, vol. 1, ed. M. Ar-cher, 1–89. Sydney: Surrey Beatty & Sons and theRoyal Zoological Society of New South Wales.

Ride, W.D.L. 1962. On the evolution of Australian mar-supials. In The evolution of living organisms, ed.G.W. Leeper, 281–386. Melbourne: Melbourne Uni-versity Press.

. 1993. Jackmahoneya gen. nov. and the genesis ofthe macropodiform molar. Association of Australa-sian Palaeontologists Memoir 15:441–459.

Sige, B. 1972. La faunule de Mammiferes du Cretace Su-perieur de Laguna Umayo (Andes Peruviennes). Bul-letin du Museum National d’Histoire Naturelle, 3ºser., 99, Sciences du Terre 19:375–409.

Simpson, G.G. 1935. Descriptions of the oldest knownSouth American mammals, from the Rıo Chico For-mation. American Museum Novitates 793:1–25.

. 1938. A new marsupial from the Eocene of Pa-tagonia. American Museum Novitates 989:1–5.

. 1947. A new Eocene marsupial from Brazil.American Museum Novitates 1357:1–7.

. 1964. Los mamıferos Casamayorenses de la Co-leccion Tournouer. Revista del Museo Argentino deCiencias Naturales, Paleontologıa 1:1–21.

. 1970. Addition to the knowledge of Groeberia(Mammalia, Marsupialia) from the mid-Cenozoic ofArgentina. Breviora 362:1–17.

Szalay, F.S. 1982. A new appraisal of marsupial phylogenyand classification. In Carnivorous marsupials, vol. 2,ed. M. Archer, 621–640. Mosman, New SouthWales: Royal Zoological Society of New SouthWales.

Takahashi, F. 1974. Variacao morfologica de incisives emDidelfideos (Marsupialia, Didelphinae). Anais da Ac-ademia Brasileira de Ciencias 46(3–4):413–416.

Tate, G.H.H. 1934. New generic names for two SouthAmerican marsupials. Journal of Mammalogy 15:154.

Thomas, O. 1894. On Micoureus griseus, Desm., with thedescription of a new genus and species of Didelphi-


dae. Annals and Magazine of Natural History (Series6) 14:184–188.

. 1895. Descriptions of four small mammals fromSouth America. Including one belonging to the pe-culiar marsupial genus ‘‘Hyracodon’’ Thomes. An-nals and Magazine of Natural History (Series 6) 16:367–370.

. 1921. A new genus of opossum from southernPatagonia. Annals and Magazine of Natural History(9a Series 8) 1921:136–139.

Wolff, R.G. 1984. New early Oligocene Argyrolagidae(Mammalia, Marsupialia) from Salla, Bolivia. Jour-nal of Vertebrate Paleontology 1:108–113.



Paleogene Notoungulates from the Amazon Basin of Peru

Bruce J. Shockey,1 Ralph Hitz,2 and Mariano Bond3

ABSTRACT. Paleogene-grade notoungulates (order Notoungulata) from the SantaRosa local fauna of eastern Peru are documented and described. The specimens includeat least two species of toxodont notoungulates (suborder Toxodontia) and the first re-cord of an interatheriid (suborder Typotheria) from the Amazon Basin. The toxodontspecimens include isolated teeth from a sheep-sized animal and a tooth from perhapsthe smallest known toxodont. The larger toxodont specimens include an upper molarhaving distinctive, open, multiple cristae; a leontiniid-like upper premolar; and a lowermolar with the general morphology of toxodontids, leontiniids, and Deseadan or youn-ger notohippids. The interatheriid specimen is a lower jaw with molars morphologicallysimilar to the ‘‘Tinguirirican’’ Eopachyrucos plicifera and an unnamed, small interath-eriid from Salla, Bolivia (Deseadan South American Land Mammal Age). The toxodontspecimens, as well as the interatheriid, are comparable to known notoungulates rangingin age from ‘‘Tinguirirican’’ to Deseadan (31.5–24 Ma). These specimens offer tantaliz-ing hints regarding the notoungulate fauna of the Amazon Basin and suggest a regionto explore for clues regarding the rapid radiation of advanced notoungulates in the latePaleogene and early Neogene.

RESUMEN. Se describen y documentan notoungulados (Orden Notoungulata) de gra-do paleogeno procedentes de la fauna local de Santa Rosa, en el este de Peru. Losespecımenes incluyen al menos dos especies de notoungulados toxodontidos (SubordenToxodontia) y el primer registro de un interaterido (Suborden Typotheria) de la CuencaAmazonica. Los materiales de toxodontidos incluyen dientes aislados de un animal deltamano de una oveja, ası como tambien un diente de, tal vez, el mas pequeno toxodon-tido conocido. Los especımenes asignables al toxodontido de mayor tamano incluyenun molar superior con crestas distintivas, abiertas y multiples, un premolar superior detipo leontınido y un molar inferior cuya morfologıa coincide con la de los toxodontidos,leontınidos y notohıpidos deseadenses o mas tardıos. El ejemplar de interaterido consisteen una mandıbula inferior cuyos molares son parecidos a los de Eopachyrucos plicifera,del ‘‘Tinguiririquense’’, y a los de un pequeno y aun innominado interaterido de Salla,Bolivia (SALMA Deseadense). Los especımenes de toxodontidos, ası como el de inter-aterido, son comparables a los notoungulados que se registran entre el ‘‘Tinguiriri-quense’’ al Deseadense (31.5–24 Ma). Estos materiales ofrecen pistas muy interesantesen relacion con la fauna de notoungulados de la Cuenca Amazonica, y sugieren unaregion a explorar a los efectos de encontrar las huellas de la rapida radiacion de no-toungulados avanzados ocurrida a fines del Paleogeno y comienzos del Neogeno.

INTRODUCTION

Despite the relatively good mammalian record forthe Cenozoic of South America, there remain sig-nificant gaps in our knowledge of the evolutionaryhistory of notoungulates. These gaps are both spa-

1 Department of Zoology, University of Florida, 223Bartram, Gainesville, Florida 32611. Email: [email protected]

2 Tacoma Community College, 6501 South 19th Street,Tacoma, Washington 98466. Email: [email protected]

3 Departamento Paleontologıa Vertebrados, Museo deLa Plata, Paseo del Bosque s/n, 1900 La Plata, Argentina.Email: [email protected]

tial and temporal; that is, a large geographic region(Amazonia) is underrepresented in the fossil recordand depositional hiatuses are present, even in thewell-explored higher latitudes. Thus, the discoveryof Paleogene-grade notoungulates in the AmazonBasin of Peru is noteworthy, despite their limitednature.

The notoungulate remains from Santa Rosa in-clude several isolated teeth referable to the subor-der Toxodontia and a single mandibular fragmentwith molars referable to the family Interatheriidae,suborder Typotheria. The toxodont teeth are of twodistinct size classes. Most of the specimens are froma roughly sheep-sized animal, but one specimen ismuch smaller. The interathere specimen is also

62 m Science Series 40 Shockey et al.: Paleogene Notoungulates

quite small. Although none of the notoungulatespecimens from Santa Rosa can be confidently re-ferred to any known genus, they are similar to taxathat lived in the time interval after the MustersanSouth American Land Mammal Age (SALMA) andinto the Deseadan SALMA.

Paleogene notoungulates have been virtually un-known from the Amazon Basin. Even early Neo-gene, that is, Miocene, notoungulates are poorlyknown from this region. A single upper molar, theholotypical and only known specimen of Purperiacribatidens (Paula Couto, 1981), was referred tothe Notohippidae (Purperia Paula Couto, 1982a, isa replacement for the preoccupied generic nameMegahippus Paula Couto, 1981). Restudy of thisspecimen by Ribeiro and Bond (1999) indicatesthat it is not referable to this family. It may be aleontiniid, but it can only be confidently referred tothe subordinal level (Toxodontia). A single, lowermolar tooth fragment noted by Paula Couto(1982b) apparently is the only other evidence ofnotohippids in the Amazon Basin. However, a va-riety of toxodontids are known from Amazonia.Paula Couto (1981) reported the middle Miocenenesodontine toxodontid Nesodon sp. from thewestern Amazon. He also reported (Paula Couto,1982b) a dozen other genera of toxodontids thatcame from western Amazonia, all of which werereferred to the more derived and geologically youn-ger subfamilies Toxodontinae and Haplodontheri-inae of the family Toxodontidae. The presence ofinteratheriids in the Amazon Basin had been unre-corded, but they are known from the mid-Mioceneof La Venta, Colombia (Stirton, 1953) farther tothe north.

To facilitate the discussion of the possible age ofSanta Rosa, we will informally refer to the intervalafter the Mustersan SALMA and before the Desea-dan SALMA as ‘‘Tinguirirican’’ (Wyss et al., 1994;Flynn and Swisher, 1995), an age roughly equiva-lent to Ameghino’s ‘‘Astraponoteen plus superieur’’(see Bond et al., 1996; Kay et al. 1999; and Hitz etal., 2000). With respect to the geochronology of theTinguirirican faunal interval, Wyss et al. (1994) re-ported an isotopic age of 31.5 Ma for the fossil-bearing horizons at the principal locality of Tin-guiririca, Chile. They provided an additional iso-topic date from below the fossiliferous sequencethat suggested that the duration of the Tinguiriricaninterval might have extended to 37.5 Ma. Recentisotopic dates reported by Kay et al. (1999) fromthe Sarmiento Formation at the Gran Barranca, Ar-gentina, however, appear to better constrain the du-ration of the Tinguirirican faunal interval. Kay etal. (1999) provided isotopic age data from the Bar-rancan substage of the Casamayoran SALMA, theupper age limit of which is about 35.3 Ma. Thislimit dictates that the Mustersan must fit between31.5 and about 35 Ma, implying that the durationof the Tinguirirican faunal interval cannot extendas far back in time as previously conceived. Becauseof the present flux in our understanding of the du-

ration of the Tinguirirican interval, we will simplyrefer to the isotopic age of the Tinguirirican faunaas 31.5 Ma, with the understanding that it extendsan unknown, but probably short, amount of time,into the earlier Oligocene.

The toxodont specimens are not referable to anyknown genera, but they are too incomplete to serveas useful holotypes (see Pascual, 1965). The referralor naming of the interatheriid specimen is also de-ferred until more complete material becomes avail-able.

All specimens discussed herein came from Nat-ural History Museum of Los Angeles County(LACM) locality 6289, about 2 km north of thevillage of Santa Rosa, Peru, on the west bank ofthe Rıo Yurua near its confluence with the Rıo Beu,in Provincia Atalaya, Departamento Ucayali, east-ern Peru (98299390 S, 728459480 W; see Campbellet al., 2004).

Abbreviations of the museums from which com-parative notoungulates were examined are as fol-lows.

LACM Natural History Museum of Los AngelesCounty, Los Angeles, California, USA

UF Vertebrate Paleontology Collection, Flor-ida Museum of Natural History, Univer-sity of Florida, Gainesville, Florida

FMNH Field Museum of Natural History, Chi-cago, Illinois


Order Notoungulata Roth, 1902

Suborder Toxodontia Owen, 1853

Toxodontia is one of the three currently recognizedsuborders of the Notoungulata (Cifelli, 1993; Mc-Kenna and Bell, 1997). It includes five families: theIsotemnidae, Homalodotheriidae, Leontiniidae,Notohippidae, and Toxodontidae. The phylogenet-ic analysis of Cifelli (1993) suggests that the Iso-temnidae is basal to the ‘‘advanced Toxodontia’’and that the Homalodotheriidae were derived froma specialized lineage of isotemnids. The ‘‘advancedToxodontia’’ include the leontiniids, notohippids(sensu Simpson, 1945), and toxodontids. All of thetoxodont specimens noted in this work are referredto the ‘‘advanced Toxodontia’’ (sensu Cifelli, 1993).

None of the toxodont specimens from SantaRosa were found in an association that would sug-gest that they represent a single individual. Mostare of a similar size, such that they could have comefrom the same species. At this time it is more pru-dent to describe the larger toxodont specimens asindividual teeth rather than distinct taxa.

One toxodont specimen (LACM 140570), how-ever, is so small that it is very distinct, not just fromthe other toxodont specimens from Santa Rosa, butfrom all other known taxa of this suborder. It isdescribed and discussed separately.

Science Series 40 Shockey et al.: Paleogene Notoungulates m 63

Figure 1 Occlusal views of notoungulate teeth from Santa Rosa, Peru. A, LACM 140569, left M1 or M2; B, LACM144304, left P4; C, LACM 144305, right M?1, illustrating the 7/9 morphology discussed in the text; D, LACM 140570,left lower molar of small ?notohippid; E, LACM 140571, interatheriid molars of partial mandible, M1–trigonid of M3.

LARGER SPECIMENS OF TOXODONTIA

The larger notoungulate teeth recovered from SantaRosa are from a medium-sized (more or less sheep-sized) toxodont. They are well preserved and in-structive; thus, they are described and discussed be-low.

MATERIAL. LACM 140569, left M1 or M2;LACM 144304, left upper premolar; LACM144305, right M1 or M2; LACM 144307, left M1

or M2 (missing the hypoconid) is comparable toLACM 144305. Collected by K.E. Campbell, C.D.Frailey, L. Romero-Pitman, and M. Aldana in1995, 1998.


Genus and Species IndeterminateFigure 1A

MATERIAL. LACM 140569.DESCRIPTION. This upper left molar (M1 or

M2) is distinctive in having multiple and open cris-tae. It is mesodont and has an oblique protolophthat is significantly longer than the transversely ori-ented metaloph. The anteroposterior dimension ofthe tooth is 18.7 mm and the transverse diameteris 18.4 mm. The crown height at the paracone ridgeis 17.2 mm. Enamel is missing from the posteriorportion of the ectoloph, but the dentine border iscomplete enough so that its form is unambiguous;that is, it is not strongly undulated. The paraconeridge is weak, and the shape of labial dentine sug-gests only a slight bulge at the metacone. Three ma-jor cristae project from the anterior half of the ec-toloph toward the medial sagittal plane of the

tooth. Two smaller cristae lie posterior to the third;none unite with the crochet. These five cristae givethe labial side of the infolding of enamel a complexbranching appearance. This infolding has not beenisolated as a fossette, suggesting that the tooth isminimally worn. Another millimeter or so of wearwould have isolated this infolding as the major fos-sette. The short posterior cingulum forms an oval-shaped fossette that is partially bisected by a pos-terior enamel projection, giving the fossette some-what of a figure eight appearance. There is no ex-ternal cingulum, but a distinct anterolingualcingulum is present. The lingual portion of this cin-gulum forms a cuspule. The tooth lacks any traceof cementum.

COMMENTS. The most distinctive feature ofLACM 140569 is the supernumerary cristae, ofwhich none unite with the crochet, or with eachother medially. Although the tendency to developcuspules adjacent to the ectoloph is pronounced innotoungulates (Patterson, 1934), such cuspules aretypically minute and disappear or unite rapidlywith wear. Rarely do they form more than two dis-tinct cristae. The ‘‘advanced Toxodontia’’ typicallyhave only two cristae in their upper molars, withthe posterior one uniting with the crochet. How-ever, multiple cristae are present in the molars ofmany leontiniids [e.g., Huilatherium pluriplicatumVillarroel and Guerrero, 1985; Colpodon Burmeis-ter, 1885 (personal observation of FMNH 13310);Taubatherium major Soria and Alvarenga, 1989;and T. paulacoutoi Soria and Alvarenga, 1989], butthey unite and form multiple labial fossettes (or‘‘enamel-lined pits’’ of some authors, e.g., Villarroeland Colwell Danis, 1997).


The cristae of LACM 140569 are not united (donot form fossettes); thus, they are more similar inappearance to those of Plexotemnus complicatissi-mus Ameghino, 1904, and closely related taxa thatSimpson (1967) united under a single genus, Acoe-lohyrax Ameghino, 1902 (this synonymy and useof the generic name Acoelohyrax has been disput-ed; see Bond and Lopez, 1993). These Plexotem-nus-like taxa have well-formed multiple cristaethat, like LACM 140569, give the medial marginof the ectoloph an undulating appearance. Also, thecristae tend to remain distinct from the crochet.Distinctive undulated medial margins of the ecto-loph are present in Plexotemnus (CasamayoranSALMA) and Puelia coarctatus (Ameghino, 1901)(Mustersan SALMA). However, LACM 140569differs from these taxa by lacking the medial bulgeof the posterior cingulum and by being a little larg-er and higher crowned. LACM 140569 differs frommolars of known leontiniids by its smaller size, byits relatively smaller transverse dimension, and bynot having cristae and crochet united.

Genus and Species IndeterminateFigure 1B

MATERIAL. LACM 144304.DESCRIPTION. The size of this upper premolar

suggests that it came from an animal a little largerthan the animal from which came the upper molardiscussed above (LACM 140569).

The occlusal view of this premolar (Fig. 1B) hasa roughly square appearance, with a longer trans-verse than anteroposterior dimension (length 516.2 mm; width 5 18.5 mm). There is a sharpgroove in the ectoloph between the parastyle andthe paracone ridge and a slight bulge at the meta-cone. Otherwise, the ectoloph is fairly straight. Theparastyle is distinctly shorter than the paracone andmetacone, which are about the same height. An ex-ternal cingulum covers the base of the ectoloph,running from the posterior margin of the parastyleto the posterior border of the tooth. The paraconeridge, including the external cingulum, is 17.2 mmtall.

The occlusal surfaces of the paracone and meta-cone show some wear, which has left subtle, butunmistakable oblique (posterolabial to anterolin-gual) wear lines. The metaloph and the medial bor-der of the posterior cingulum reach the protocone.Tiny cuspules cover the enamel between the pos-terior cingulum and the metacone ridge. A deepcentral valley separates the protocone from the ec-toloph. There is no protoloph.

The protocone is robust and, though it is nearlyas high as the paracone and metacone, it is verylittle worn, remaining completely covered in enam-el. There is a vertical groove of the protocone sim-ilar to, but weaker than, that of Leontinia Amegh-ino, 1895 (see Patterson, 1934). An anterior cin-gulum rings the anterior base of the protocone and

extends to the parastyle. The center of the lingualsurface of the protocone lacks a cingulum.

COMMENTS. The lack of a protoloph, com-bined with the reasonably well-developed antero-lingual cingulum, is suggestive of the ‘‘leontiniid de-pression’’ (sensu Colwell, 1965; see also Villarroeland Colwell Danis, 1997) of leontiniid premolars.LACM 144304 is much smaller than premolars ofLeontinia gaudryi Ameghino, 1895, and Scarrittiacanquelnesis Simpson, 1934, but its anteroposteriordimension is within the size range of P3s and in thelower end of the size range of P4s of Ancylocoelusfrequens Ameghino, 1895. It is almost as large asthe P4 of a small, unnamed leontiniid from Salla,Bolivia (Shockey, 1997b:fig. 3.1). Not only does theform of the tooth suggest a leontiniid referral, butalso something of the function is demonstratedwith the oblique macrowear marks on the tooth.This is a conspicuous feature seen in virtually allspecimens examined of Leontinia, Scarrittia Simp-son, 1934, and a currently unnamed leontiniid ge-nus from Salla (Shockey, 1997b). We provisionallyrefer this tooth to the Leontiniidae.

Genus and Species IndeterminateFigure 1C

MATERIAL. LACM 144305.DESCRIPTION. LACM 144305, a right M?1,

has a length of 16.8 mm and a width of 6.8 mm.In occlusal view, it appears as a seven-over-nine(7/9; Fig. 1C), the trigonid having the shape of thenumber seven and the talonid forming a nine (theentolophid fossettid being the hole in the circle partof the nine). This 7/9 morphology (with an ento-lophid fossettid) is typical of right lower molars ofDeseadan notohippids (except Morphippus Amegh-ino, 1897, and Eurygenium Ameghino, 1895), De-seadan-to-Santacrucian toxodontids, and allknown leontiniids (left lower molars of these ani-mals appear as the mirror image of 7/9). This 7/9morphology is a putative synapomorphy for the‘‘advanced Toxodontia’’ (Cifelli, 1993). The molarat hand is smaller than those of most leontiniids,and it is most comparable in size to those of ne-sodontine toxodontids and Deseadan notohippids.

Like the previous specimens described, LACM144305 is little worn. The trigonid is less wornthan the talonid, having its occlusal area on a plane3 mm higher than that of the talonid. The lingualinfolding of enamel is deep, but it would have be-come very shallow with 1–2 mm of wear and dis-appear with just 3–4 mm of wear. The most dis-tinctive feature of the trigonid is the sharp, verticalridge along the protocone just anterior to the in-folding of labial enamel at the trigonid–talonid bor-der.

The talonid is more deeply worn, and little re-mains of the lingual infolding of enamel that sep-arates it from the trigonid. With about another 1mm of wear, the trigonid and talonid would haveunited, leaving just a trigonid–talonid fossettid. The


entolophid is robust and has a crescent-shaped en-tolophid fossettid. The hypolophid is distinct fromthe entolophid, but these would have merged withanother 1 mm of wear.

LACM 144305 is very similar in size, crownheight, and morphology to the M1 from a nearlycomplete mandible of a nesodontine toxodontid,Proadinotherium sp. of Salla (UF 149222; seeShockey, 1997b:fig. 3.12). Most of the few differ-ences in the two may be attributed to the greaterwear of the trigonid of the Salla specimen. How-ever, the labial surface of the trigonids of the molarsof the Salla specimen are rounded, lacking the dis-tinctive, sharp, vertical ridge seen on LACM144305.

COMMENTS. Although LACM 144305 is near-ly identical to the Proadinotherium sp. specimenfrom Salla, the 7/9 morphology of the lower molarsof the ‘‘advanced Toxodontia’’ is seen in Deseadan-grade notohippids and leontiniids, as well as neso-dontine toxodontids, so it is of little value in deter-mining phylogenetic position more precisely thanas unspecialized ‘‘advanced Toxodontia.’’ The la-bial vertical ridge of the trigonid, however, is fre-quently seen in specimens of leontiniids.

Discussion of Larger Specimens of Toxodontia

The toxodont specimens described above are ofsimilar size, suggesting the possibility that they arefrom the same species, but there was no close as-sociation of any of these teeth upon collection thatwould suggest that they were from a single individ-ual. LACM 144305 and the other lower molar(LACM 144307; not figured) are near-perfect mir-ror images of one another, and they certainly camefrom the same species, if not the same individual.

However, it is difficult to interpret the upper mo-lar (LACM 140569) and the upper premolar(LACM 144304) as coming from the same species.The upper premolar is most likely that of a leon-tiniid, and it is within the size range of the P3s ofthe smallest leontiniids and within the lower end ofthe size range of the P4s (e.g., Ancylocoelus Amegh-ino, 1895, and a small species of an unnamed genusat Salla; Shockey, 1997b). The upper molar (LACM140569), however, is smaller than the M1s of thesmallest leontiniids. The undulated median surfaceof the ectoloph of the M1 is exactly the morphologythat one would expect to have led to the multiple,enamel-lined pits seen in a variety of leontiniid mo-lars. Thus, it might represent a primitive leontiniidor an animal closely related to leontiniids. Like-wise, the generalized toxodont morphology of thelower molars (LACM 144305 and LACM 144307)suggest that they are from an animal near the baseof the ‘‘advanced Toxodontia’’ radiation and might,too, be from a primitive leontiniid or closely relatedtaxon.

The relative lack of wear of these teeth may re-present a sample bias or a bias in preservation; thatis, young individuals died and were preserved. It

may, however indicate environmental conditions inwhich nonabrasive forage was available to andconsumed by these herbivores.

SMALL SPECIMEN OF TOXODONTIA

cf. Notohippidae Ameghino, 1895

Genus and Species IndeterminateFigure 1D

MATERIAL. LACM 140570.DESCRIPTION. This M1 or M2 is similar to, but

distinctly smaller than, those of the ‘‘dwarf’’ noto-hippid Rhynchippus pumilus (Ameghino, 1897). Itis also lower crowned than R. pumilus, or anyknown Deseadan or younger notohippid. The en-tolophid is distinct in having a somewhat undulatedanterolingual border. The entolophid is fairly ro-bust and contains a subtriangular fossettid. No oth-er fossettids are present.

The anteroposterior dimension of the talonid is5.9 mm at the occlusal level, and the estimated totalocclusal anteroposterior length is about 7.5 mm.The greatest height of the crown is 7.3 mm.

COMMENTS. This specimen is much too smallto belong to the same taxon as any of the toxo-dontid specimens described above. It has bothprimitive and derived characters, in that it isbrachydont to mesodont, but it shows the derivedentolophid fossettid. This entolophid fossettid isseen in most Deseadan notohippids, but not in pre-Deseadan genera such as Eomorphippus Roth,1902 (see Simpson, 1967), or the Plexotemnus-liketaxa. An entolophid fossettid appears in early tox-odontids (and is lost in late Miocene and youngertoxodontids), all known leontiniids, and most De-seadan or younger notohippids. It may represent asynapomorphy for leontiniids, toxodontids, andadvanced notohippids. A phylogenetic analysis hasshown, however, that it is just as likely for this traitto have evolved independently in leontiniids(Shockey, 1997a). If, indeed, the entolophid fosset-tid were a synapomorphy for these diverse Desea-dan taxa, then it would have evolved at some timeprior to the Deseadan.

SUBORDER TYPOTHERIA ZITTEL, 1893

Family Interatheriidae Ameghino, 1887

Subfamily Interatheriinae Simpson, 1945

Genus and Species IndeterminateFigure 1E

MATERIAL. LACM 140571, partial left man-dible with M1–3 (only the trigonid of M3 remains).

DESCRIPTION. LACM 140571 is the only in-terathere specimen known from the Amazon Basin.Persistent labial and lingual sulci on the molars ren-der them bilobed and identify this specimen as aninterathere, although some basal interatheres haveweakly developed sulci (Hitz, 1997). Additionally,the extremely hypsodont, or possibly hypselodont,


molars of LACM 140571 identify this specimenmore narrowly as belonging to the subfamily Inter-atheriinae.

Compared with known interatheriines, LACM140571 is very small. The length and width of theM1 is 4.3 by 2.4 mm and of the M2 is 4.6 by 2.4mm. The metaconids on M1 and M2 are pro-nounced and project lingually. On the lingual sideof the talonid of both M1 and M2 there is a slightlingual extension, positioned just posterior to themetaconid. This feature is more pronounced on M1.The trigonid of the M3 is smaller than that of M1

or M2, and the metaconid is less pronounced thanthat of the preceding molars. The labial side of themandible is smooth with no visible foramina. Thelack of foramina labially supports the interpreta-tion that these teeth are the M1–3. Most interatheresobserved by us have a mental foramen on the labialside of the mandible below the P4. The lack of suchon LACM 140571 suggests that the specimen rep-resents a more posterior portion of the jaw thanthe premolar area. The lingual side of the mandiblehas a shallow excavation on the posterior end, in-terpreted here as the mandibular fossa.

By virtue of its size and the lingually projectingmetaconids of the molars, LACM 140571 resem-bles Eopachyrucos plicifera Ameghino, 1901, andan unnamed species from Salla, Bolivia (Hitz,1997). Given its fragmentary condition, however,referral of this specimen to either taxon is prema-ture.

COMMENTS. The upper Oligocene Salla bedsof Bolivia preserve several interathere taxa, thesmallest of which ranges widely in size (MacFaddenet al., 1985; Hitz, 1997). The dentition of LACM140571 resembles a few of the more diminutivespecimens (UF 91302, UF 91644, UF 91643, UF49266) from this smaller taxon, both in size andgeneral morphology. LACM 140571 also resemblesa specimen of Eopachyrucos plicifera (MLP 12-1529), a taxon known from several localities fromsouthern Argentina (Gran Barranca, Chubut; Can-adon Blanco, Chubut; Rocas Bayas, Rıo Negro).Eopachyrucos plicifera was first allocated to theHegetotheriidae by Ameghino (1901), but a recent,emended diagnosis places it within the Interatheri-idae (Hitz et al., 2000).

The upper Oligocene Salla beds have long beenrecognized as belonging to the Deseadan SALMA(MacFadden et al., 1985). The age of the depositsbearing E. plicifera is more problematic. Recent ev-idence (Bond et al., 1996), however, suggests thatthey fall within a hiatus in the SALMA sequence,between the Mustersan and Deseadan SALMAs.Bond et al. (1996) identified a stratigraphic level atthe Gran Barranca, Chubut, that appears distinctfrom Deseadan and Mustersan deposits and termedthis interval the ‘‘Astraponoteen plus superieur’’level. Several localities in Argentina have been ten-tatively correlated to the Astraponoteen plus super-ieur level at the Gran Barranca, including CanadonBlanco and Rocas Bayas (Hitz et al., 2000). Fur-

thermore, the Astraponoteen plus superieur levelhas been tentatively correlated to the Tinguiriricafauna of central Chile on the basis of faunal simi-larities [e.g., interatheres (see Hitz et al., 2000) andmarsupials (Goin and Candela, personal commu-nication)].

DISCUSSION

It is unclear whether the larger toxodont specimensrepresent one, two, or three species. Nothing re-garding their collection or morphology clearly re-lates them to one another, but the similar size sug-gests the possibility.

The upper molar (LACM 140569) is distinctivein having ununited multiple cristae. This is similarto that seen in the Casamayoran to MustersanPlexotemnus-like toxodontids, but LACM 140569is larger, is higher crowned, and lacks the lingualbulge of the posterior cingulum seen in the Plexo-temnus-like taxa. Alternatively, the multiple cristaemay suggest a relationship with the leontiniids,which tend to have multiple cristae, although theyare expressed as labial fossettes or enamel pits.LACM 140569 also differs from leontiniids by hav-ing linear dimensions that are only 90 percent ofthose of the smallest known leontiniids. The upperpremolar (LACM 144304) has a morphology (e.g.,‘‘leontiniid basin’’) comparable to leontiniids, al-though it is smaller than all but the smallest mem-bers of the family. Should LACM 140569 andLACM 144304 belong to a single species, it wouldlikely be a primitive leontiniid. The lower molar(LACM 144305) has the general morphology of the‘‘advanced Toxodontia’’ (sensu Cifelli, 1993). Al-though it is almost identical to the M1 of Proadi-notherium sp. (a nesodontine toxodontid) of Salla,the single significant difference (the sharp verticalridge of the labial trigonid) is a character suggestiveof leontiniids. Thus, all three of the larger speci-mens may have come from a very small and prim-itive leontiniid.

The diminutive ?notohippid, LACM 140570, iscomparable to Deseadan taxa, but it is lowercrowned. The interatheriid, LACM 140571, is mostcomparable to the early Oligocene ‘‘Tinguirirican’’Eopachyrucos plicifera and an unnamed interatherefrom the upper Oligocene (Deseadan) of Salla, Bo-livia.

The toxodont specimens discussed above arehigher crowned than those of Mustersan species ofToxodontia, but generally lower crowned thanthose of the Deseadan (see Shockey, 1997a). Al-though there appears to be a strong relationshipbetween crown height and the geological age of atleast notohippids (Shockey, 1997a), some cautionshould be exercised before extrapolating this prin-ciple to tropical regions. For example, Cifelli andGuerrero (1997) noted that proterotheriid litop-terns from the low-latitude, middle Miocene LaVenta fauna of Colombia actually had significantlylower crowned cheek teeth than geologically older


proterotheriids from the Santacrucian faunas ofsouthern Patagonia. This suggests that regional en-vironmental differences influence the evolution ofhypsodonty.

The problems associated with the use of high-latitude fossils to estimate the age of tropical faunaswere reviewed by Madden et al. (1997). These in-vestigators found that correlation by faunal resem-blance or by ‘‘guide’’ fossils tended to overestimatethe age of the La Venta fauna. For example, Kay etal. (1987) correlated the La Venta fauna with theSantacrucian SALMA by using the Simpson coef-ficient of faunal similarity (Simpson, 1960). Also,some ‘‘guide’’ fossils (e.g., Prothoatherium, Pachy-biotherium) suggested a Colhuehuapian ‘‘age’’ forthe considerably younger La Venta fauna.

Although much of the La Venta fauna appearsarchaic for its age, it should be noted that somerelatively ‘‘advanced’’ lineages are also found there.For example, the first records of Dasypodini andTolypeutinae armadillos occur at La Venta (Carliniet al., 1997). The general lack of similarity betweenLa Venta and Patagonian faunas of similar age sug-gests that there was little faunal exchange betweenthe northern and southern extremes of the conti-nent, at least during the middle Miocene. We op-timistically note, however, that Stirton’s (1953)original estimate of a ‘‘Friasian’’ age for the La Ven-ta fauna (see Madden et al., 1997, for discussion)was within a couple million years of its actual age(13.5–11.6 Ma; Flynn et al., 1997).

Given the caveats noted above, we offer an im-precise, first-order estimate of a ‘‘Tinguirirican’’ toDeseadan SALMA age for the notoungulates ofSanta Rosa on the basis of their comparative mor-phology with notoungulates of known age and ontheir degree of hypsodonty. Other, better represent-ed faunal members of the Santa Rosa local fauna[e.g., marsupials (Goin and Candela, 2004) and ro-dents (Frailey and Campbell, 2004)] appear to con-strain this age further.

CONCLUSIONS

This work records the occurrence of Paleogene-grade notoungulates from the Amazon region ofeastern Peru. They include toxodonts of two dis-tinct size classes and the first record of an intera-there from the Amazon Basin. These notoungulates,on the basis of their degree of hypsodonty and sim-ilarity to known taxa, appear to date from the earlyOligocene (‘‘Tinguirirican’’ age) to the late Oligo-cene (Deseadan SALMA), or, in absolute terms,from about 31.5 to 24 Ma. The larger toxodontspecimens share features with the early ‘‘advancedToxodontia,’’ suggesting that they represent taxanear the base of the radiation of leontiniids, Desea-dan notohippids, and toxodontids. LACM 140569presents characters suggestive of an advanced Plex-otemnus-like toxodont or a primitive leontiniid, or,curiously, both. These limited specimens cannotnarrowly constrain the age of the Santa Rosa fau-

na, but the older end of the age estimate is close tothe age estimates derived from the numerous mar-supial and rodent specimens from Santa Rosa.

Although the notoungulates described in thiswork are fragmentary, they offer tantalizing hintsof taxa unknown in the rest of the Neotropics. Giv-en the tremendous gap in our knowledge of thetaxa that gave rise to the explosive radiation of no-toungulates in the Deseadan, the discovery of thesePaleogene-grade notoungulates at Santa Rosa com-pels further exploration in tropical South America.Such endeavors have the potential to fill both thegeographic and temporal gaps in our knowledge ofthe history of life in South America.

ACKNOWLEDGMENTS

We thank K.E. Campbell for inviting us to examine theseinteresting specimens. We also appreciate the thoughtful,critical, and thus helpful, reviews of earlier versions of thiswork by D. Croft and R. Madden. An International Re-search Fellowship from the National Science Foundationprovided support for BJS during the initial phases of thisstudy.

LITERATURE CITED

Ameghino, F. 1895. Premiere contribution la connaissancede la fauna mammalogique de couches a Pyrother-ium. Boletın Instituto Geografico Argentino 15:603–660.

. 1897. Mammiferes cretaces de l’Argentine. Deux-ieme contribution a la connaissance de la faunamammalogique des couches a Pyrotherium. BoletınInstituto Geografico Argentino 18:406–521.

. 1901. Notices preliminaires sur des ongules nou-veaux des terrains cretaces de Patagonie. Boletın dela Academia Nacional de Ciencias de Cordoba 16:349–429.

. 1902. Notices preliminaire sur des mammiferesnouveaux des terrain cretaces de Patagonie. BoletınAcademia Nacional de ciencias Cordoba 17:5–70.

. 1904. Nuevas especies do mamıferos, cretaceos yterciarios de la Republica Argentina. Anales de So-ciedad de Ciencias de Argentina, Buenos Aires 56:193–208.

Bond, M., and G. Lopez. 1993. El primer Notohippidae(Mammalia, Notoungulata) de la Formacion Lum-brera (Grupo Salta) del Noroeste Argentino. Consi-deraciones sobre la sistematica de la Familia Noto-hippidae. Ameghiniana 30:59–68.

Bond, M., G. Lopez, and M. Reguero. 1996. ‘‘Astrapon-oteen plus superieur’’ of Ameghino: Another intervalin the Paleogene record of South America. Journalof Vertebrate Paleontology 16(3)(Suppl.):23A.

Burmeister, G. 1885. Examen crıtico de los mamıferos yreptiles fosiles denominados por D. Augusto Bra-vard. Annales Museo Nacional Buenos Aires 3:95–174.

Campbell, K.E., Jr., C.D. Frailey, and L. Romero-Pitman.2004. The Paleogene Santa Rosa local fauna of Am-azonian Peru: Geographic and geologic setting. InThe Paleogene mammalian fauna of Santa Rosa,Amazonian Peru, ed. K.E. Campbell, Jr. Natural His-tory Museum of Los Angeles County, Science Series40, 3–14.

Carlini, A.A., S.R. Vizcaıno, and G.J. Scillato-Yane. 1997.Armored xenarthrans: A unique taxonomic and eco-


logic assemblage. In Vertebrate paleontology of theNeotropics: The Miocene fauna of La Venta, Colom-bia, ed. R.F. Kay, R.H. Madden, J.J. Flynn, and R.L.Cifelli, 213–226. Washington, D.C.: Smithsonian In-stitution Press.

Cifelli, R.L. 1993. The phylogeny of the native SouthAmerican ungulates. In Mammal phylogeny: Placen-tals, vol. 2, ed. F.S. Szalay, M.J. Novacek, and M.C.McKenna, 195–216. New York: Springer-Verlag.

Cifelli, R., and J. Guerrero. 1997. Litopterns. In Verte-brate paleontology of the Neotropics: The Miocenefauna of La Venta, Colombia, ed. R.F. Kay, R.H.Madden, J.J. Flynn, and R.L. Cifelli, 289–302.Washington, D.C.: Smithsonian Institution Press.

Colwell, J. 1965. A new notoungulate of the family Leon-tiniidae from the Miocene of Colombia. Unpub-lished Master’s thesis, University of California,Berkeley.

Flynn, J.J., J. Guerrero, and C.C. Swisher III. 1997. Geo-chronology of the Honda Group. In Vertebrate pa-leontology of the Neotropics: The Miocene fauna ofLa Venta, Colombia, ed. R.F. Kay, R.H. Madden, J.J.Flynn, and R.L. Cifelli, 44–59. Washington, D.C.:Smithsonian Institution Press.

Flynn, J.J., and C.C. Swisher III. 1995. Cenozoic SouthAmerican Land Mammal Ages: Correlation to globalgeochronologies. In Geochronology time scales andglobal stratigraphic correlation, ed. A. Berggren,D.V. Kent, and J. Hardenbol. Society of EconomicPaleontologists and Mineralogists (SEPM) SpecialPublication, 54:317–333.

Frailey, C.D., and K.E. Campbell, Jr. 2004. Paleogene ro-dents from Amazonian Peru: The Santa Rosa localfauna. In The Paleogene mammalian fauna of SantaRosa, Amazonian Peru, ed. K.E. Campbell, Jr. Nat-ural History Museum of Los Angeles County, Sci-ence Series 40, 71–130.

Goin, F.J., and A. Candela. 2004. New Paleogene mar-supials from the Amazon Basin of eastern Peru. InThe Paleogene mammalian fauna of Santa Rosa,Amazonian Peru, ed. K.E. Campbell, Jr. Natural His-tory Museum of Los Angeles County, Science Series40, 15–60.

Hitz, R. 1997. Contributions to South American mam-malian paleontology: New Interatheres (Notoungu-lata) from Chile and Bolivia, Typothere (Notoun-gulata) phylogeny, and paleosols from the Late Ol-igocene Salla beds. Unpublished Ph.D. dissertation,University of California, Santa Barbara.

Hitz, R., M. Reguero, A. Wyss, and J.J. Flynn. 2000. Newinteratheriines (Interatheriidae, Notoungulata) fromthe Paleogene of central Chile and southern Argen-tina. Fieldiana, New Series 42:1–26.

Kay, R.F., R.H. Madden, J.M. Plavcan, R.L. Cifelli, andJ. Guerrero Dıaz. 1987. Stirtonia victoriae, a newspecies of Miocene Colombian primate. Journal ofHuman Evolution 16:173–196.

Kay, R.F., R.H. Madden, M.G. Vucetich, A.A. Carlini,M.M. Mazzoni, G.H. Re, M. Heizler, and H. San-deman. 1999. Revised geochronology of the Casa-mayoran South American Land Mammal Age: Cli-matic and biotic implications. Proceedings of theNational Academy of Science 96(23):13235–13240.

MacFadden, B.J., K.E. Campbell, R.L. Cifelli, O. Siles,N.M. Johnson, C.W. Maeser, and P.K. Zeitler. 1985.Magnetic polarity stratigraphy and mammalian fau-na of the Deseadan (Late Oligocene–Early Miocene)

Salla beds of northern Bolivia. Journal of Geology93(3):223–250.

Madden, R.H., J. Guerrero, R.F. Kay, J.J. Flynn, C.C.Swisher, and A.H. Walton. 1997. The LaventanStage and Age. In Vertebrate paleontology in theNeotropics: The Miocene fauna of La Venta, Colom-bia, ed. R.F. Kay, R.H. Madden, J.J. Flynn, and R.L.Cifelli, 499–519. Washington, D.C.: Smithsonian In-stitution Press.

McKenna, M., and S. Bell. 1997. Classification of mam-mals above the species level. New York: ColumbiaUniversity Press, 631 pp.

Pascual R. 1965. Los Toxodontidae de la Formacion Ar-royo Chasico de la Provincia de Buenos Aires.Ameghiniana 19:19–35.

Patterson, B. 1934. Upper premolar–molar structure inthe notoungulata with notes on taxonomy. Geolog-ical Series of Field Museum of Natural History 6:91–111.

Paula Couto, C. de. 1981. Fossil mammals from the Ce-nozoic of Acre Brazil. IV—Notoungulata Notohip-pidae and Toxodontidae Nesodontinae. Anais do IICongresso Latinoamericano de Paleontologia, PortoAlegre 2:461–477.

. 1982a. Purperia, a new name for MegahippusPaula Couto, 1981. Iheringia 7:69–70.

. 1982b. Fossil mammals from the Cenozoic ofAcre Brazil. V—Notoungulata Nesodontinae (II),Toxodontinae and Haplodontheriinae, and Litopter-na, Pyrotheria and Astrapotheria (II). Iheringia 7:5–43.

Ribeiro, A.M., and M. Bond. 1999. Consideraciones so-bre Purperia cribatidens (Paula Couto, 1981) (No-toungulata, Toxodontia) del Cenozoico de Acre,Brasil. Ameghiniana 36(1):107.

Roth, S. 1902. Notas sobre algunos neuvos mamıferosfosiles. Anals de Museo La Plata 10:251–256.

Shockey, B.J. 1997a. Two new notoungulates (Family No-tohippidae) from the Salla beds of Bolivia (Desea-dan: Late Oligocene): Systematics and functionalmorphology. Journal of Vertebrate Paleontology17(3):584–599.

. 1997b. Toxodontia of Salla, Bolivia (Late Oli-gocene): Taxonomy, systematics, and functionalmorphology. Unpublished Ph.D. dissertation, Uni-versity of Florida, Gainesville, Florida.

Simpson, G.G. 1934. A new notoungulate from the earlyTertiary of Patagonia. American Museum Novitates735:1–3.

. 1945. The principles of classification and a clas-sification of mammals. Bulletin of the American Mu-seum of Natural History 85:1–350.

. 1960. Notes on the measurement of faunal resem-blance. American Journal of Science 258A:300–311.

. 1967. The beginning of the age of mammals inSouth America, part 2. Bulletin of the American Mu-seum of Natural History 137:1–259.

Soria, M.F., and H. Alvarenga. 1989. Nuevos restos demamıferos de la Cuenca de Taubate, Estado de SaoPaulo, Brasil. Anais Academia Brasileira de Ciencias61:157–175.

Stirton, R.A. 1953. A new genus of interatheres from theMiocene of Colombia. University of California Pub-lications in Geological Sciences 29:265–348.

Villarroel, C., and J. Colwell Danis. 1997. A new leonti-niid notoungulate. In Vertebrate paleontology of theNeotropics: The Miocene fauna of La Venta, Colom-bia, ed. R.F. Kay, R.H. Madden, J.J. Flynn, and R.L.


Cifelli, 303–318. Washington, D.C.: Smithsonian In-stitution Press.

Villarroel, C., and J. Guerrero. 1985. Un nuevo y singularrepresentante de la familia Leontiniidae? (Notoun-gulata, Mammalia) en el Mioceno de La Venta, Co-lombia. Geologıa Norandina, Bogota 9:35–40.

Wyss, A.R., J.J. Flynn, M.A. Norell, C.C. Swisher III, M.J.

Novacek, M.C. McKenna, and R. Charrier. 1994.Paleogene mammals from the Andes of central Chile:A preliminary taxonomic, biostratigraphic, and geo-chronologic assessment. American Museum Novita-tes 3098:1–31.



Paleogene Rodents from Amazonian Peru:The Santa Rosa Local Fauna

Carl David Frailey1 and Kenneth E. Campbell, Jr.2

ABSTRACT. The first large and diverse sample of ?Eocene rodents from South Amer-ica consists of forms that are more primitive than those of the Oligocene genera towhich they are surely allied. The teeth of these rodents are brachydont, are quadrangularwith rounded corners in occlusal outline, and have low crests and four distinct cusps.The clean, scissorslike cutting surface of younger caviomorphs appears to be in its for-mative stages. All of these early rodent taxa share a basic tetralophodont enamel patternthat illustrates the beginning of patterns seen in the modern rodent families Erethizon-tidae, Echimyidae, and Agoutidae, and one genus shows consistent and characteristicsigns of pentalophodonty. The pentalophodont condition appears to have evolved fol-lowing the merging of minor cuspules into a distinct crest within the metaflexus, which,with wear, produced the fifth loph, a neoloph. All available mandibular rami are hys-tricognathous, indicating an early appearance of this character.

New genera include Eopululo (Erethizontidae, one new species), Eoincamys (Agou-tidae, two new species), Eobranisamys (Agoutidae, two new species), Eopicure (Agou-tidae, one new species), Eosallamys (Echimyidae, two new species), Eoespina (Echimyi-dae, one new species), Eosachacui (Echimyidae, one new species), and Eodelphomys(Echimyidae, one new species). Fragmentary or excessively worn specimens of severalother indeterminate taxa indicate a greater diversity in the Santa Rosa local fauna thanthat suggested by the named taxa.

The Santa Rosa local fauna is dated to the ?Eocene (early Mustersan South AmericanLand Mammal Age, SALMA) on the basis of the stage of evolution of the includedrodent taxa and that of other mammalian taxa in the fauna. The presence of such ahighly diverse rodent fauna at this early date in the South American fossil record requiresa reexamination of hypotheses of hystricognath origin and dispersal. Heretofore, hy-potheses regarding the origins of the South American hystricognaths have placed theirroots in North America or Africa, with dispersal to South America occurring sometimebefore the Deseadan SALMA or, more recently, the post-Mustersan/pre-Deseadan ‘‘Tin-guirirican’’ time. The diverse ?Eocene Santa Rosan rodent fauna increases the possibilitythat the South American hystricognaths might have originated after vicariant isolationfrom the ancestral rodent lineage during the breakup of Gondwana. This hypothesisrequires a much older date of origination for South American hystricognath rodentsthan previously suggested by the fossil record.

RESUMEN. La primera muestra de roedores del ?Eoceno de America del Sur consisteen formas mas primitivas que aquellas de los generos oligocenos con las cuales estanseguramente relacionadas. Los dientes de estos roedores son braquidontes, cuadrangu-lares, con esquinas redondeadas en su perfil oclusal y tienen crestas bajas y cuatro cus-pides diferenciadas. Parecen encontrarse en una etapa formativa las superficies netas ycortantes de los caviomorfos de menor edad. Todos estos taxones de roedores tempranoscomparten un patron basico tetralofodonte, ilustrativo de los comienzos de los patronesobservables en las familias modernas de roedores Eretizontidae, Echimyidae y Agouti-dae; uno de los generos muestra signos consistentes y caracterısticos de pentalofodoncia.La condicion pentalofodonte parece haber evolucionado a partir de la fusion de cuspulasmenores en una cresta diferenciada dentro del metaflexo; con el desgaste se produce elquinto lofo, un neolofo. Todas las ramas mandibulares estudiadas son hystricognatas,indicando la temprana aparicion de este caracter.

1 Johnson County Community College, Overland Park, Kansas 66210; Museum of Natural History, University ofKansas, Lawrence, Kansas 66045; Research Associate, Natural History Museum of Los Angeles County, 900 ExpositionBoulevard, Los Angeles, California 90007. Email: [email protected]

2 Vertebrate Zoology, Natural History Museum of Los Angeles County, 900 Exposition Boulevard, Los Angeles, Cal-ifornia 90007. Email: [email protected]

72 m Science Series 40 Frailey and Campbell: Paleogene Rodents

Los nuevos generos incluyen a Eocoendu (Erethizontidae, una especie nueva), Eoin-camys (Agoutidae, dos nuevas especies), Eobranisamys (Agoutidae, dos nuevas especies),Eopicure (Agoutidae, una nueva especie), Eosallamys (Echimyidae, dos nuevas especies),Eoespina (Echimyidae, una nueva especie), Eosachacui (Echimyidae, una nueva especie)y Eodelphomys (Echimyidae, una nueva especie). Aparecen tambien especımenes frag-mentarios o excesivamente desgastados pertenecientes a varios otros taxones indeter-minados, lo que revela una diversidad aun mayor en la fauna local de Santa Rosa.

La edad de la fauna local de Santa Rosa se postula como ?Eoceno (SALMA Mustersensetemprano) sobre la base de el grado evolutivo de los roedores estudiados, ası como tambiende otros taxones de mamıferos presentes en la misma. La presencia de una asociacion tandiversa de roedores en esta etapa temprana del registro fosil sudamericano requiere unarevision de las hipotesis sobre el origen y distribucion de los histricognatos. Hasta ahora,dichas hipotesis han postulado el origen norteamericano o africano de los histricognatos,con una dispersion hacia America del Sur ocurrida algo antes de la SALMA Deseadenseo, mas recientemente, hacia los tiempos ‘‘Tinguiririquenses’’ (post-Mustersense/pre-De-seadense). La diversa fauna de roedores de Santa Rosa sugere la posibilidad de que loshistricognatos sudamericanos pudieron haberse originado a partir del aislamiento vicar-iante del linaje ancestral de roedores durante la particion de Gondwana. Esta hipotesisrequiere un tiempo de origen de los histricognatos sudamericanos muy anterior al sugeridopor el registro fosil.

INTRODUCTION

The presence of rodents in the Eocene record ofSouth America has been hypothesized for manyyears (for summary discussions, see George, 1993;Lavocat, 1993; Wood, 1993; Vucetich et al., 1999).The great diversity of uniquely South American ro-dents in the Oligocene (and their total absence inolder faunas) suggested an undiscovered Eocene ap-pearance followed by rapid diversification. Rapiddiversification, indeed, for among rodents of theDeseadan South American Land Mammal Age(SALMA) are 19 genera within six families of theorder Rodentia (following the taxonomic arrange-ment of McKenna and Bell, 1997). Radiometricdates indicate that the Deseadan SALMA extendsfrom 29.4 Ma to younger than 25.65 Ma in Bolivia(Kay et al., 1998), or from 29 to 27 Ma in Pata-gonia (Flynn and Swisher, 1995). The sudden, mid-Tertiary appearance of a diverse suite of rodents ina continent that had been isolated from all othercontinents for the first half of the Tertiary has ledto considerable debate as to how and when rodentsarrived in South America.

The first look at the rodent fauna of the supposedperiod of arrival and diversification was the discov-ery of a rodent in the pre-Deseadan Tinguiririca lo-cal fauna of Chile (Wyss et al., 1993), which wasdated to 31.5 Ma (early Oligocene) and originallythought to have possibly extended back as far as37.5 Ma (late Eocene; Wyss et al., 1993, 1994;Flynn and Wyss, 1999). Wyss et al. (1993) left thetaxonomic placement of their single specimen as‘‘Dasyproctidae?, unnamed new genus and species’’because it was too badly worn to reveal many tax-onomic characters. Nonetheless, the mandibular ra-mus is clearly hystricognathous and dentally alliedwith later caviid rodents. From this specimen itcould be inferred that by ‘‘Tinguirirican’’ time, tax-

onomic diversification of South American rodentswas well underway.

But it is the Santa Rosa local fauna that finallyreveals the extent of rodent diversification in theinterval surrounding the Eocene–Oligocene transi-tion and that answers many questions about thenature of these earliest caviids. From a small sampleof concentrated matrix taken from the Santa Rosalocality (Campbell, 2004) in 1995, about 200 com-plete, isolated rodent teeth, one intact upper toothrow and partial maxilla, and numerous isolated in-cisors and tooth fragments resulted from the screen-washing of approximately 400 kg of sediment. Thissample was enlarged by a return expedition in1998. Much of the concentrated matrix collectedthen, all of which is less than 4 mm diameter, hasnot yet been sorted, but several partial dentitionsrecovered in the field and several dozens of addi-tional isolated teeth sorted from a sample of thematrix were a significant addition to the sample.Three families, eight genera, and eleven species arerepresented by sufficiently diagnostic material topermit identification and description herein (Table1). The majority of these taxa (the erethizontids be-ing the principal exceptions) appear to be closelyallied to taxa from the Deseadan SALMA depositsof Salla, Bolivia, yet consistently demonstrate moreprimitive dental features. Six additional species,represented by material considered inadequate fordescription, are also present. These are undoubt-edly new to science and could be named if the spec-imens were more numerous or more complete.

MATERIALS AND METHODS

The taxonomic placement of genera within familiesof the suborder Hystricognatha in this paper fol-lows McKenna and Bell (1997). The parvorder Ca-viida Bryant and McKenna, 1995, corresponds tothe restricted concept of Caviomorpha, that is,

Science Series 40 Frailey and Campbell: Paleogene Rodents m 73

Table 1 Classification of the Santa Rosan rodents de-scribed in this paper.

Order Rodentia Bowdich, 1821Suborder Hystricognatha Woods, 1976

Infraorder Hystricognathi Tullberg, 1899Family Erethizontidae Bonaparte, 1845

Eopululo new genusEopululo wigmorei new species

Parvorder Caviida Bryant and McKenna, 1995Superfamily Cavioidae Fischer de Waldheim, 1817

Family Agoutidae Gray, 1821Subfamily Dasyproctinae Gray, 1825

Eoincamys new genusEoincamys pascuali new speciesEoincamys ameghinoi new species

Eobranisamys new genusEobranisamys romeropittmanae new

speciesEobranisamys riverai new species

Eopicure new genusEopicure kraglievichi new species

Genus and species indeterminate AGenus and species indeterminate BGenus and species indeterminate C

Superfamily Octodontoidea Waterhouse, 1839Family Echimyidae Gray, 1825

Subfamily Heteropsomyinae Anthony, 1917Eosallamys new genus

Eosallamys paulacoutoi new speciesEosallamys simpsoni new species

Eoespina new genusEoespina woodi new species

Eosachacui new genusEosachacui lavocati new species

Subfamily Adelphomyinae Patterson andPascual, 1968Eodelphomys new genus

Eodelphomys almeidacomposi new speciesSubfamily indeterminate

Genus and species indeterminate AGenus and species indeterminate BGenus and species indeterminate C

South American hystricognaths exclusive of the Er-ethizontidae (Bryant and McKenna, 1995). Dental-ly, the erethizontids are similar to the other generaof South American rodents, especially so in the San-ta Rosa local fauna, but this family is excludedfrom the Caviida because of other characteristics.Much rich discussion of the evolution, and there-fore of the systematic relationships, of these rodentsand their appropriate taxonomy exists in the sci-entific literature. The Santa Rosan rodents bringnew evidence to these discussions, as will be pre-sented. However, a discussion of family level rodentsystematics, and a detailed defense of the choice offamilial designations herein for certain genera, isnot the purpose of this paper, and indeed, such is

not possible without parts of the skeleton beyondthe teeth currently available to us. All comparativeanalyses are at the generic level and will alwaysmention the particular genus or genera under con-sideration. Subsequent familial placement of theSanta Rosa genera is with the closest comparablegenera of a younger age. Higher level taxonomicreassortment of genera, should it be necessary inthe future, should not affect the validity of the ge-neric comparisons presented in this paper.

The choice of names for genera in this paper re-flects the similarities to Oligocene and later generaof South American rodents and the probable Eo-cene age (see ‘‘Discussion’’) for these new rodents.The choice of species names honors people whohave contributed to vertebrate paleontology inSouth America.

A complete classification of the taxa discussed inthis paper is presented in Table 1. Measurementsof all teeth discussed in the text are given in Ap-pendix 1. Appendix 2 is a photo gallery of the ro-dent specimens from the Santa Rosa local faunadiscussed herein.

All specimens are currently housed in the collec-tions of the Section of Vertebrate Paleontology,Natural History Museum of Los Angeles County(LACM). Unless otherwise specified, specimennumbers refer to this collection. Other acronymsused in this paper: PU, Princeton University; YPM,Yale Peabody Museum.

GENERAL DESCRIPTIVE ODONTOLOGY

In general, the teeth of these early rodents are whatwould be expected of primitive forms, and they areconsistent with the theoretical description for earlycaviid rodents given by Wood (1981:83). They aresquat, or brachydont, and quadrangular withrounded corners when viewed in occlusal outline.Upper teeth display only slight unilateral hypso-donty. In general, lophs and lophids are broadbased and taper to narrow crests. Distinct cusps arepresent in early wear stages, particularly in thosetaxa that appear to be near Branisamys Hoffstetterand Lavocat, 1970, from Salla, Bolivia. Wear stagesare seen in which the lophs or lophids have notfused, and in which the wear surface is sloping andmay extend onto the side of the tooth. All of theteeth are fundamentally similar, with taxonomicdifferences found in the size and placement of thelophs and lophids. The basic patterns of the SantaRosan genera are diagramed in Figure 1. In mostinstances, and with a little practice, these diagramsare sufficient to make an identification to genus ofan unknown tooth.

The species of rodents documented by these teethrepresent the earliest known South American mem-bers of the Hystricognathi. The presence of recog-nizable cusps indicates that these teeth are not farremoved from bunodonty. The degree of brachy-donty is such that minor wrinkles and bumps in thecrown surface quickly, with wear, become part of


Figure 1 Dental terminology and comparative sketches of the occlusal patterns of the Santa Rosan rodent teeth. Arrowsindicate anterior (to left) and labial directions. Dental nomenclature used in this paper is indicated. With wear, a flexusmay become a fossette and a flexid may become a fossettid. Lophs shown terminating without connecting with otherlophs or cusps indicate lophs that are unconnected initially, but which connect with wear. Upper teeth. Cusps: HY,hypocone; PA, paracone; PR, protocone; ME, metacone. Lophs: ATL, anteroloph; ML, metaloph; NEL, neoloph; PL,posteroloph; PRL, protoloph. Flexuses and fossettes: HF, hypoflexus; MSF, mesoflexus; MTF, metaflexus; NF, neoflexus;PF, paraflexus. Lower teeth. Cuspids: END, entoconid; HYD, hypoconulid; MED, metaconid; PRD, protoconid. Lophids:ALD, anterolophid; HLD, hypolophid; MLD, metalophid, PLD, posterolophid. Flexids and fossettids: AFD, anteroflexid;HFD, hypoflexid; MSFD, mesoflexid; MTFD, metaflexid.

the occlusal surface. The simple patterns, often sub-tle distinctions that separate taxa, and the innu-merable variations, some exasperatingly interme-diate between taxa as defined here, made for greatdifficulty in the separation and identification ofsome genera and species.

Deseadan SALMA representatives of the rodentfamilies documented at Santa Rosa have higher

crowned teeth (although still low crowned whencompared with later genera), nontapering and moredistinct lophs and lophids, fusion of lophs and lo-phids during early wear, and relatively indistinctcusps and cuspids. The occlusal view of wear pat-terns seen in the Santa Rosan rodents and theiryounger Deseadan relatives are so similar in somecases that one can mentally see the derivative pat-


Figure 2 Differing interpretations of dental features of early rodents as displayed in an upper left molar (anterior to left).A, From Patterson and Wood (1982) and followed in this paper; B, from Lavocat (1976); C, from Flynn et al. (1986).HY, hypocone; M, mure; ME, metacone; ML, metaloph; MSL, mesoloph; MSLL, mesolophule; MST, mesostylid; MTC,metaconule; NEL, neoloph; PA, paracone; PR, protocone; PRL, protoloph.

tern created by no more than a slightly increasedcrown height and the loss of separate identities forthe individual cusps, the suppression of a crest ormure, or the stabilization of what, in some SantaRosan forms, appear to be simple irregularities inthe crown surface. Such changes as these are notan insignificant matter, however, and they mightonly have been achieved over great lengths of time.Currently, the lack of age data prevents us fromobtaining, or offering, insights into rates of evolu-tion of early caviid rodents. Nonetheless, it is fairto suggest that the Santa Rosan rodents were mov-ing along the evolutionary path that led to some ofthe known Oligocene genera of South American ro-dents.

In that the dental patterns of the rodent teeth inthe Santa Rosa local fauna are very similar to thosedescribed by Patterson and Wood (1982) for theDeseadan SALMA rodents of Bolivia, for the sakeof direct comparison, and consistent with our in-terpretation of dental homologies, we use theircharacter designations with only slight modifica-tion. This terminology is derived from Wood andWilson (1936) and illustrated in Figures 1 and 2.On upper molars, the protoloph is designated as thecrest that connects the paracone and hypocone [theterms loph(s) and crest(s) are used interchangeablyin this paper]. This crest is invariably the strongerand the first of the interior crests to form as thecrown wears. Such usage is somewhat at odds withstandard caviid nomenclature in which the connec-tion between the protoloph and metaloph is iden-tified simply as a mure and the segment of the crestbetween the hypocone and the junction of the pro-toloph and metaloph is not further identified otherthan as part of the hypocone (e.g., Patterson andWood, 1982). As defined here, the continuous pro-toloph standardizes the dental nomenclature for allof the Santa Rosan genera, in particular includingEoincamys new genus, in which no other crestunites with the protoloph.

The metaloph, as here defined, is a weak crestthat decreases in height toward the interior of the

crown surface. Its union with the protoloph, whichoccurs in all genera except Eoincamys new genus,is near the midpoint of the protoloph and markedby a deflection in that crest. Among the upper teethof these genera, the protoloph is frequently, but notconsistently, wide at this point and short crests mayextend toward both the metaloph and the proto-cone. These crests then become connecting crests ascrown wear progresses. This might be the expres-sion of a mesocone (sensu Lavocat, 1981; Vucetichand Verzi, 1990) or metaconule (sensu Flynn et al.,1986). If so, this argues against identification of thethird loph as a mesoloph or metaconule in the San-ta Rosan rodent teeth in that in these genera thethird loph does not progress with wear toward thelabial margin. Rather, it proceeds internally from alabial cusp, which we identify as the metacone.

Patterson and Wood (1982, and references there-in) identify the fourth crest in the upper teeth ofcaviids as a neoloph and refer to the third crestfrom the anterior as the metaloph. We find the ter-minology of Patterson and Wood (1982) more inaccord with the crown patterns of the Santa Rosanrodents and in contradiction to the terminologyand hypotheses of pattern origin expressed by Lav-ocat (1976, 1981), Flynn et al. (1986), Vucetichand Verzi (1990), and Candela (1999) (see ‘‘Dis-cussion’’).

On the lower teeth, the patterns are simpler. Thepreeminent crest that corresponds to the protolophof the uppers is the hypolophid. In our usage, thehypolophid connects the entoconid and protoconid.Patterson and Wood (1982) use the term hypolo-phid, but it is not included in their figure 1, and itdoes not correspond exactly to our usage. The la-bial, oblique portion of what we call the hypolo-phid is termed the ectolophid by Patterson andWood (1982). This crest is not identifiable in Eoin-camys new genus. In describing the crests of cha-pattimyid lower teeth, Flynn et al. (1986) identifythe second of four crests as a metalophulid II. Aswith their nomenclature for the upper teeth of theserodents, this designation is probably correct for the


rodents that they studied, but unlikely for the SantaRosan rodents in which the labial terminus of thesecond lophid is a large cuspid, here termed themetaconid, and the lophid therefore identified as ametalophid. What Flynn et al. (1986) call the me-talophid is our anterolophid.


Order Rodentia Bowdich, 1821

Suborder Hystricognatha Woods, 1976

Family Erethizontidae Bonaparte, 1845

Eopululo new genusTYPE SPECIES. Eopululo wigmorei new species.INCLUDED SPECIES. Monotypic.DIAGNOSIS. Teeth brachydont with major,

identifiable cusps. Upper molars square with round-ed corners in occlusal outline. Protoloph and me-taloph parallel to each other and to transverse axisof tooth; approximately equal in length and width.Parafossette, mesofossette, and metafossette large,nearly equal in size, and oval (parafossette) to egg-shaped (metafossette) in outline.

ETYMOLOGY. Eos, dawn, Greek; pululo, localPeruvian name for a porcupine.

Eopululo wigmorei new speciesFigure 12A; Appendix 2

HOLOTYPE. LACM 143269, right Mx.REFERRED SPECIMENS. Mx: left, 149489;

right, 143272. Mx: left, 143410, 149465; right,143263 (partial), 143268 (partial). All cataloguenumbers pertain to LACM collections.

DIAGNOSIS. As for the genus.MEASUREMENTS. See Appendix 1.TYPE LOCALITY. Santa Rosa, Peru. LACM lo-


Formation. ?Eocene.ETYMOLOGY. Named in honor of John G.

Wigmore, a distinguished former member of theBoard of Governors of the Natural History Muse-um of Los Angeles County, who financially sup-ported the Rio Yurua expeditions and participatedin the Santa Rosa discovery expedition of 1995.

DISCUSSION. This taxon is one of the largestspecies recovered from Santa Rosa. The three upperteeth assigned to it are fragmental or in an ad-vanced stage of wear. In one, LACM 143272, thedental pattern is almost gone. LACM 143269 is thebest of the three specimens, and the following de-scription is taken primarily from that specimen.Complete teeth are rounded squares in occlusalview. As is typical of erethizontid teeth, the twocentral crests are thin, parallel, and at nearly 90degrees to the anteroposterior axis of the tooth.They are approximately equal in length and prom-inence. The protocone is connected to the proto-loph by a weak mure. All three labial folds of thistooth have been closed into fossettes, although it is

clear that the mesofossette was the last to be encir-cled and that this was only achieved at the ad-vanced state of wear seen in this particular tooth.All three fossettes are large and subequal in size,with the metafossette only slightly larger than theothers. The three labial fossettes are large, which isconsistent with the rounded outline of the toothand the central placement of the protoloph and me-taloph. In outline, these three labial fossettes are acurved oval (parafossette, curving around the pro-toloph), rectangular in form (mesofossette), orovate (metafossette). The shape of the metafossetteis created by the greater width of the posterolophat its labial end. At this position in teeth of Pro-tosteiromys Wood and Patterson, 1959, the poster-oloph may be crenulated or even have a neoloph,but no such irregularities are present on the holo-type of Eopululo wigmorei. The hypoflexus is nar-row, but it gives no indication of closure into a fos-sette. On the heavily worn and damaged specimen,LACM 143272, all fossettes are nearly gone, butbecause they were flat bottomed, rather than steep-ly sloped toward their centers, their former extentand the outlines of the crests are still evident. Aswell as can be determined, the surviving, remnantpattern of that tooth matches the description of thespecimen that is better preserved.

LACM 143410 is a nearly complete lower tooth,but LACM 143268 is only a small fragment ofwhat was probably a lower tooth. Both show alarge, deep, and roughly rectangular mesoflexid.On the more complete specimen, LACM 143410,it can be seen that only slight wear has caused thelabial ends of the anterolophid and metalophid tojoin; that is, the anterofossettid has formed, and theposterior two crests have nearly encircled the me-taflexid.

Another lower tooth, LACM 143263, is missingmost of the anterolophid and the labial one-thirdof the tooth. What survives is very representativeof lower teeth found among species of the Erethi-zontidae. The anterofossettid is on the verge offorming, although wear on the tooth is slight. Withadditional wear, the metafossettid and then the me-sofossettid would form, in that order, with the me-soflexid remaining free until advanced wear. Thetwo central crests (i.e., the metalophid and hypo-lophid) are nearly parallel to each other and thetransverse axis of the tooth. The metalophid is onlyslightly thinner than the hypolophid. The large andoval anteroflexid and metaflexid are highly char-acteristic of the elongate lower teeth seen in speciesof the Erethizontidae.

Parvorder Caviida Bryant andMcKenna, 1995

Superfamily CavioidaeFischer de Waldheim, 1817

Family Agoutidae Gray, 1821

Subfamily Dasyproctinae Gray, 1825


Eoincamys new genusTYPE SPECIES. Eoincamys pascuali new spe-

cies.INCLUDED SPECIES. Eoincamys ameghinoi

new species.DIAGNOSIS. Enamel pattern similar to that of

Incamys Hoffstetter and Lavocat, 1970, but differsby having low, rounded lophs and recognizablecusps. Teeth brachydont with four broad upperlophs; major cusps evident in lophs. Protoloph runsdiagonally across the tooth. Connections betweenprotoloph and posteroloph incomplete until latewear stage. Metaloph with weak or no connectionto posteroloph.

ETYMOLOGY. Eos, dawn, Greek; Incamys, agenus of Deseadan rodent, indicating similarity tothat genus.

DISCUSSION. This is one of the most easilyidentifiable dental patterns among the generallysimilar rodent teeth from Santa Rosa. In the upperteeth, the protoloph dominates the occlusal surface,and it is usually unconnected with either the pos-teroloph or metaloph. The metaloph tends to par-allel the curvature of the protoloph and therebyunite with the posteroloph rather than the proto-loph. The protoloph never unites with the antero-loph; that is, the paraflexus and hypoflexus are con-fluent as a single flexus. This is very similar to thecondition in the Oligocene genus Incamys (Fig. 14),although because of a higher crown height the par-afossette of Incamys more often opens onto the la-bial margin.

A comparable loph arrangement exists in lowerteeth assigned to this genus, but here the hypolo-phid is the dominant crest. Like the metaloph, themetalophid is comparatively small and subject tomore variation. The metalophid may unite with thehypolophid, may be discontinuous at its middle, or,through miscellaneous enamel spurs, may unitewith both the hypolophid and anterolophid.

A small metaloph/metalophid is present in theinitial wear stages of upper and lower molars ofIncamys, and the metaflexus and anteroflexid arediscernible. There is an indication in both upperand lower teeth of Eoincamys that a three-cresteddental pattern typical of Incamys is developingthrough a process whereby the metaloph/metalo-phid is being eliminated in upper and lower teeth,respectively, by a reduction and union with the pos-terior crest (upper teeth) or anterior crest (lowerteeth).

Eoincamys pascuali new speciesFigures 3, 12B, 13A, 14; Appendix 2

HOLOTYPE. LACM 143318, left Mx.REFERRED SPECIMENS. P4: left, 143321;

right, 143313, 143322. M3: right, 143324. Mx: left,143314, 143316, 143319, 149463; right, 143312,143317, 143323, 149478. dp4: left, 143335,143456; right, 143328, 143331. P4: 143301,143333, 143334, 143336. M1: left, 143303. M3:

right, 149445, 149449. Mx: left, 143304, 143306,143307, 143308, 143309, 149446; right, 143299,143300, 143305, 143310, 143427, 143434,143504, 149452, 149453. All catalogue numberspertain to LACM collections.

DIAGNOSIS. The larger of two species referredto Eoincamys. Crown with four crests. Anteroloph,protoloph, and metaloph parallel. Metaloph andposteroloph of same length.

MEASUREMENTS. See Appendix 1.TYPE LOCALITY. Santa Rosa, Peru. LACM lo-


Formation. ?EoceneETYMOLOGY. Pascuali, in honor of Rosendo

Pascual, a friend, colleague, and preeminent mam-malian paleontologist of Argentina.

DESCRIPTION. Upper Teeth. The protoloph,hypoloph, and metaloph are approximately paral-lel, smoothly curved, and run diagonally across theocclusal surface of the tooth. The metaloph andposteroloph are much reduced in length and con-fined to the posterolateral corner of the tooth.These two crests are almost completely united inthe unworn condition at both their lingual and la-bial margins. The uniting of these crests causes anoval metafossette, rather than a metaflexus, to be aprominent feature of these teeth. In two specimens,LACM 143317 and 143319, a short enamel spur,which we interpret to be spatially equivalent to aneoloph, extends into the small metafossette. Thisspur is here considered to be a variation in the an-terior enamel wall of the posteroloph rather than aneoloph. The labial portion of the anteroloph isclosely associated with the paracone, and it is sep-arated from the rest of the anteroloph by a con-striction in the crest. The length of the anterolophthat is lingual to the constriction is about the sameas that of the metaloph, but much wider.

Lower Teeth. In occlusal outline, the teeth havea flattened anterior margin and a curved posteriormargin. The hypolophid is comparatively straight,with only slight deflections of small enamel spursat the location of potential (i.e., with wear) con-nections with the metalophid and posterolophid.The hypofossettid and metafossettid are combinedthroughout most of the wear stages, and they areopen at both labial and lingual margins. The me-talophid commonly unites with the lingual end ofthe anterolophid to create a U-shaped curve, open-ing laterally, that persisted on many specimens untillate wear was achieved.

The lower molars of Eoincamys pascuali are sim-ilar to those of the Deseadan taxon Incamys boli-vianus Hoffstetter and Lavocat, 1970, in that thedental pattern is dominated by two oblique lophids,the hypolophid and posterolophid. The protoconidis the labial terminus of the hypolophid. The anter-olophid and metalophid unite on the lingual border.With moderate wear, the metalophid joins the hy-polophid, but a connection between the anterolo-


Figure 3 Eoincamys pascuali. Variations in dental and wear patterns in upper and lower teeth. A reversed drawing isindicated by ‘‘r.’’ P4: A, LACM 143322. Mx: B, LACM 143318 (holotype of genus); C, LACM 143319; D, LACM143323r; E, LACM 143316; F, LACM 143324r. P4: G, LACM 143336r. Mx: H, LACM 143308; I, LACM 143299r; J,LACM 143306; K, LACM 143305r; L, LACM 143434r. Scale 5 1 mm.

phid and hypolophid does not occur until late wear.The size and position of the anterolophid and me-talophid of Eoincamys and Incamys are very simi-lar, except that in Incamys the metalophid fuseswith the anterolophid to create a single crest ante-rior to the hypolophid. There is no complete murebetween the hypolophid and posterolophid; that is,the hypoflexid and metafossettid are fully conflu-ent, although remnants of a partial mure are pres-ent in most specimens of E. pascuali, and the limitsof the hypoflexid and metafossettid are discernible.A complete mure is more likely to form, albeitweakly, in the sister species, E. ameghinoi new spe-cies.

Eoincamys ameghinoi new speciesFigures 4, 12C; Appendix 2

HOLOTYPE. LACM 143339, left Mx.REFERRED SPECIMENS. Maxilla, partial, with

P4–M2, 149495. P4: left, 149473; right, 149430.

M3: left, 143315; right, 143340. Mx: left, 143320,143338, 143396 (partial), 149470; right, 143279,143437, 143468, 149450. P4: left, 143332; right,143445. Mx: left, 143280, 143435, 149435,149442; right, 143507, 149444. All cataloguenumbers pertain to LACM collections.

DIAGNOSIS. The smaller of two species as-signed to this genus. Metaloph short, small, and notconnected to posteroloph until advanced wear.Paracone and metacone large and conical.



Formation. ?Eocene.ETYMOLOGY. Ameghinoi, in honor of Carlos

and Florentino Ameghino, pioneering vertebratepaleontologists of Argentina.

DISCUSSION. The teeth of Eoincamys ameghi-noi appear to be even more brachydont than those


Figure 4 Eoincamys ameghinoi. Variations in dental and wear patterns in upper and lower teeth. A reversed drawing isindicated by ‘‘r.’’ Mx: A, LACM 143339 (holotype); B, LACM 143338; C, LACM 143340r; D, LACM 143279r. P4: E,LACM 143445r. Mx: F, LACM 143507r; G, LACM 143435; H, LACM 143280. Scale 5 1 mm.

of E. pascuali, although the difference is small andquestionable because of variations resulting fromwear. The greater brachydonty may explain themore variable patterns seen in teeth of E. ameghi-noi in that the occlusal surface is more likely to beaffected by minor irregularities in the crown sur-face.

Upper Teeth. This small species is most easilyseparated from E. pascuali by its proportionallysmaller metaloph. The effective result is to have asingle valley anterior to the protoloph that consistsof the paraflexus and hypoflexus, with a V-shaped,combined mesofossette and metafossette posteriorto the protoloph. The latter are joined medially.The combined posterior fossette, rather than aflexus, is created by the encircling of the posteriortwo-thirds of the occlusal surface by a nearly con-tinuous crest composed of the protoloph, postero-loph, and labial margin of the tooth. Only shallownotches exist on the labial margin to indicate theentrances to the three labial valleys.

Lower Teeth. As in the uppers, and as is char-acteristic of this genus, an incipient three-crestedpattern is indicated by the reduction of one crestfrom the occlusal surface. In this case, the reducedcrest is the metalophid. The union of this lophidwith the anterolophid at their lingual ends is not sostrong as in E. pascuali, and the curved, U-shapedmetalophid/anterolophid is not nearly so stronglyformed. That union is approached in several spec-

imens (LACM 143332 and 143507), but in thesespecimens there appears to be a typically tetralo-phodont pattern. However, in these specimens, themetalophid is incomplete at its center. In one spec-imen (LACM 143445, a P4), at a wear stage equiv-alent to the aforementioned specimens, the meta-lophid has disappeared into a broadly confluentanterolophid/metalophid. It seems that the meta-lophid is a highly variable part of the dental patternof this genus, and more so in E. ameghinoi than inE. pascuali.

Eobranisamys new genus

TYPE SPECIES. Eobranisamys romeropittmanaenew species.

INCLUDED SPECIES. Eobranisamys riverainew species.

DIAGNOSIS. Enamel pattern similar to that ofBranisamys Hoffstetter and Lavocat, 1970, but dif-fers in having low, rounded crests and identifiablecusps.

ETYMOLOGY. Eos, dawn, Greek; Branisamys,a genus of Deseadan agoutid, indicating affinitywith that genus.

DISCUSSION. Upper Teeth. The specimens re-ferred to this genus are similar to teeth of themonotypic Deseadan genus Branisamys in that theupper molars are pentalophodont with the antero-loph completely separate until at least moderate


wear is attained. The fifth loph appears to haveformed from the alignment of enamel irregularitieson the surface of the metaflexus, or from crenula-tions on the anterior edge of the posteroloph, orboth, supporting the conclusion of Patterson andWood (1982) that the tetralophodont enamel pat-tern is primitive for the parvorder Caviida. Theterm ‘‘neoloph’’ is therefore used in this paper forthe short loph between the metaloph and the pos-teroloph. The anteroloph and protoloph, includingthe hypocone, are strongly curved, producing aquarter-moon shape to the combined paraflexusand hypoflexus. In several specimens, even aftermoderate wear, the metaloph does not connect withthe protoloph. The neoloph and posteroloph arealways fused in Branisamys, but may or may notbe fused in specimens of Eobranisamys.

Lower Teeth. The lower teeth of Eobranisamyshave four strong crests. The four lophids are ap-proximately equal in thickness, with the metalo-phid being the narrowest. The posterolophid is con-vexly curved, but the anterolophid is straight. Theanterolophid roughly parallels the metalophid andthe medial half of the hypolophid, with slight var-iations that are important in distinguishing species.More so than in any other Santa Rosan genus, afterfusion of the metalophid with the hypolophid it isthe metalophid that appears to be a continuous,gently curved crest from the lingual surface to theprotoconid, to which the lingual half of the hypo-lophid has joined. The hypolophid turns at the me-dial extent of the hypoflexid and curves to meet themetalophid. With slight wear, the hypolophid isconnected to the posterolophid by a small mure.The posterolophid is continuous with the hypocon-ulid. The mesoflexid lies on the central labiolingualplane.

The dp4 of Eobranisamys has a modification ofthe enamel pattern of the lower molars, with theposterior half of the tooth the least altered (Figs.13B–D). A complete and independent posterolo-phid, which terminates labially in the hypoconid,occupies the posterior border of the tooth. The hy-polophid reaches from the lingual margin, begin-ning at the entoconid, to the anteroposterior mid-line of the tooth, as in the molars. Unlike with themolars, the hypolophid of the dp4 has a weak con-nection to the protoconid and stands as an abbre-viated lophid. As a result, the metalophid, whichconnects the metaconid and protoconid, extendstransversely across the occlusal surface as an iso-lated lophid. It connects to the hypolophid at theanteroposterior midline only during late wear stag-es of the tooth. In its unworn state, the anterolo-phid is composed of two or more unnamed cuspids.With a little wear of the crown, these cuspids fuseto create a lophid that eventually unites with themetalophid at the anteroposterior midline and, an-teriorly, with a small cingulate crest. The small, an-terior cingulate cuspid and the short anterolophidcreate the forward elongation and narrowing of thedp4.

This pattern differs from that of the dp4 of Bran-isamys, in which the protoconid is equally con-nected to the hypolophid and to the metaconid.Stated differently, the metalophid of Branisamys ismore molariform. The anterior cingulate cuspid isvariable in size and position in Eobranisamys, butin Branisamys it appears to be larger and less lo-phate.

Eobranisamys romeropittmanae new speciesFigures 5, 12D, 13B, 14; Appendix 2

HOLOTYPE. LACM 144297, partial maxillawith right M1–2.

REFERRED SPECIMENS. Maxilla, partial,right, with M2–M3, 144296; M3: left, 143354,143381, 149490; right, 143364, 143365, 149493.Mx: left, 149482; right, 143355, 143356, 143357,143358, 143359, 143360, 143362. Mandible, par-tial, right, with dp4–M2, 144293; left, with dam-aged dp4–M2, 144292. dp4: left, 143352, 143423(partial); 143505 (partial); right, 143419. M1: left,149487. M3: left, 143347. Mx: left, 143341,143342, 143343, 143346, 149462; right, 143266,143344, 143345, 149483. All catalogue numberspertain to LACM collections.

DIAGNOSIS. Larger of two species assigned toEobranisamys. Metaloph short and not fully con-nected to protoloph, which connects to hypocone;neoloph well established.



Formation. ?Eocene.ETYMOLOGY. Romeropittmanae, in honor of

Lidia Romero-Pittmann, a friend, colleague, andcodiscoverer of the Santa Rosa local fauna.

DISCUSSION. Upper Teeth. Of the two speciesreferred to this genus, the upper teeth of Eobrani-samys romeropittmanae display the greatest sepa-ration of the anteroloph from the protoloph. In al-most all specimens there is an enamel spur, or adeflection in the curve of the protoloph, to indicatea potential connection with the anteroloph. How-ever, a connection is apparently never achieved,even in advanced wear. In two teeth (LACM143365 and 143381), the stage of wear is such thatthe labial ends of the anteroloph and protoloph areunited. Even here, there is no mure that connectsthese lophs. Also, the anteroloph is more stronglycurved than in the other species, and it often hascompletely turned toward the posterior as it reach-es the protocone. The protoloph and metaloph arenot parallel. Instead, the protoloph is angled to-ward the anterolabial margin of the tooth and themetaloph approximates the labiolingual plane. Themetaloph and neoloph appear to be the last to jointhe other crests as wear progresses. The neofossetteis large, circular to oval in outline, and closes earlywith wear.

Lower Teeth. The lower teeth are quadrangular


Figure 5 Eobranisamys romeropittmanae. Variations in dental and wear patterns in upper and lower teeth. A reverseddrawing is indicated by ‘‘r.’’ Mx: A, LACM 143356r; B, LACM 143362r; C, LACM 143359r; D, LACM 143358r. M3:E, LACM 143364r; F, LACM 143365r. Mx: G, LACM 143345r; H, LACM 143343; I, LACM 143346; J, LACM143266r. M3: K, LACM 143347. Scale 5 1 mm.

in outline. Of the two species referred to this genus,the lower teeth of E. romeropittmanae are the morelophate. The metaconid and entoconid are smalland obscure. The mesoflexid is nearly as wide asthe adjacent flexids, and it curves sharply anteriadat its lateral terminus. The metalophid and hypo-lophid are oblique to the anteroposterior axis of thetooth, and they approximate the parallel curvatureof the posterolophid.

The lower deciduous premolar has a metalophidand an anterolophid that are curved and that restobliquely on the occlusal surface; they parallel thehypolophid and posterolophid. The anterior cin-gulate cusp, rather than being centrally located asin Branisamys and the other species of Eobranisa-mys, is displaced to the anterolingual corner (Fig.13B).

Two partial dentitions in fragmentary mandiblesare referred to this species (LACM 144292 and144293; Appendix 2). Both specimens are brokenat the mandibular symphysis and the ascendingrami are missing. LACM 144293 preserves an ex-

cellent dentition from dp4 to M2. The teeth are bad-ly damaged on LACM 144292, but from the po-sition of the roots and length of the first tooth(crown broken away), a dp4 was present. The den-tal patterns of M1–2 cannot be determined, and onlythe roots of the M3 are preserved. The allocationof LACM 144292 to E. romeropittmanae is basedon the resemblance of this mandibular ramus tothat of LACM 144293.

Both mandibular rami are short, robust, and ful-ly hystricognathous. The lower incisors extend be-neath the tooth rows. The proximal, or posterior,end of the lower incisor is exposed by breakage inLACM 144292. The incisor can be seen posteriorto the M3 only a few millimeters below the poste-rior alveolar margin of M3. The masseteric fossa isdeep, smoothly concave, and terminates anteriorlyas a blunt point that reaches to the posterior mar-gin of dp4. The diastema is short (approximately1.5 times the length of dp4) and shallow. A circularmental foramen is located beneath the posteriorthird of the diastema and about halfway between


Figure 6 Eobranisamys riverai. Variations in dental and wear patterns in upper and lower teeth. A reversed drawing isindicated by ‘‘r.’’ M1: A, LACM 143503r (holotype); B, LACM 143372; C, LACM 143377. Mx: D, LACM 143366; E,LACM 143367. Scale 5 1 mm.

the lateral alveolar ridge and the lower margin ofthe mandible at the diastema. When viewed withthe anterior part of the incisor placed vertically, thetooth row does not tilt toward the midline to theextent seen in modern, or even Deseadan, taxa ofthe Caviida. Instead of an approximate 30 degreetilt as mentioned by Patterson and Wood (1982) forthe rodents from Salla, Bolivia, the tilt is no morethan 10 degrees. Similarly, the lateral torsion of theocclusal plane, in which P4 or dp4 is more verticalthan the inwardly tilting posterior tooth row, is lessevident in Eobranisamys than in Branisamys (Di-nomyidae; Patterson and Wood, 1982:438, fig.25F) or in Incamys (Agoutidae, personal observa-tion).

Although the dental patterns of LACM 144292and 144293 are obscured by wear and breakage,one aspect of these dentitions is especially interest-ing and pertinent to the evolution of the Caviida.In LACM 144293, all the molars were fully eruptedand worn, and yet the dp4 was present with no re-placement tooth in the ramus (see X-ray, Appendix2). In LACM 144292, the first tooth is brokenaway, but the size and position of its roots indicatea long, anteriorly tapered dp4 instead of a quadran-gular P4. Again, there is no evidence of a permanentP4 in the mandible.

The possibility that dp4s were retained and notreplaced by P4s resurrects an old discussion aboutrelationship between Old and New World hystri-cognaths. The retention of the dp4 was cited as acharacter in common between Caviida (‘‘Caviom-orpha’’) and the Old World hystricognaths(‘‘Phiomorpha’’) by Hoffstetter (1975:520, cited byPatterson and Wood, 1982:498). Phiomorpha istreated as a junior synonym of the infraorder Hys-tricognathi Tullberg, 1899, by McKenna and Bell(1997). In thryonomyids (Old World), deciduouspremolars are a permanent part of the tooth row

by the early Oligocene (Patterson and Wood, 1982:498). Normal replacement of the deciduous pre-molars in the caviids was used by Wood (1981:80,88) to exclude the African hystricognaths from aposition of direct ancestry to the caviids because theAfrican forms retained the dp4 throughout life,whereas all known Oligocene genera of caviids re-placed the dp4 (Wood and Patterson, 1959:305).This statement was adjusted by Patterson andWood (1982:391) to say that the Echimyidae re-tained dp4 from the Colhuehuapian SALMA (earlyMiocene) on, but that in the Deseadan SALMA(late Oligocene) replacement was typical. The fos-sils from Santa Rosa create some questions in thisregard. For example, Branisamys (Deseadan SAL-MA; Agoutidae) replaced the dp4 (Patterson andWood, 1982). But no replacement, permanentfourth premolars of any species of Eobranisamys(Agoutidae) are identified in the sample from SantaRosa, and the two mandibular fragments attributedto E. romeropittmanae retained the dp4 as a per-manent part of the tooth row with no indication ofa P4 underneath. Permanent lower premolars ofother species are represented in the collection. Now,with the apparent development of this trait (i.e.,retention of the dp4) in at least one genus of theSanta Rosan rodents, the proposed relationship be-tween South American caviids and the African hys-tricognaths might be strengthened, but because thisfeature is not universal among the Santa Rosan ro-dents, it might represent only a common propensitywithin species of the Hystricognatha.

Eobranisamys riverai new speciesFigures 6, 12E, 13C–D; Appendix 2

HOLOTYPE. LACM 143503, right M1.REFERRED SPECIMENS. M2: right, 143378.

M3: left, 143302; right, 143428. Mx: left, 143372,


Figure 7 Eopicure kraglievichi. Variations in dental andwear patterns in upper teeth. A, LACM 143264 (holotypeof genus); B, LACM 149455; C, LACM 143353; D,LACM 149479. Scale 5 1 mm.

143373, 143377, 149474; right, 149464. dp4: left,143368 (partial), 143369, 143506; right, 143270(partial), 143371. Mx: left, 143366, 143367,149456, 149461, 149475; right, 149476, 149485.All catalogue numbers pertain to LACM collec-tions.

DIAGNOSIS. Smaller of two species assigned tothis genus. Metaloph connected to protoloph; neo-loph variable in shape and position, with weak orno connection to the posteroloph medially.



Formation. ?Eocene.ETYMOLOGY. Riverai, in honor of Hugo Ri-

vera Mantilla, Director Tecnico, Instituto Geolo-gico Minero y Metalurgico (INGEMMET), whohas long encouraged and facilitated our researchinto the prehistory of Amazonian Peru, and with-out whose support the Santa Rosa local faunawould remain unknown.

DISCUSSION. Upper Teeth. The anteroloph isless curved than in Eobranisamys romeropittman-ae, and it does not extend posteriad to the labiolin-gual midline of the tooth. The union between an-teroloph and protoloph appears to occur late, andin one specimen (LACM 143373) it might neverhave occurred as the crown became worn. The pro-toloph and metaloph diverge, but less so than in E.romeropittmanae, and as a result the mesoflexus issmaller. There is a strong deflection in the curvatureof the protoloph at the point where it meets themetaloph. When these two crests unite, the proto-loph and metaloph appear to be attached to op-posite ends of a mure from which the metalophcontinues to the hypocone. The neofossette is com-paratively smaller than that of E. romeropittmanae.

Lower Teeth. Lower teeth tend to be quadran-gular, but rectangular in outline rather than squareas in E. romeropittmanae. The metaconid and en-toconid are large cusps, equal to the hypoconid insize, and present as knobs at the medial terminusof their respective lophids. The entoconid exceedsthe metaconid in size. An enlarged anteromedialportion of the anterolophid, possibly the antero-conid, is equal to the hypoconid in size and isplaced diagonally to the hypoconid, its posterolat-eral opposite. In one specimen, LACM 143366, theanterofossettid is invaded by enamel projectionsfrom both the anterolophid and metalophid, reflect-ing what is probably individual variation. The me-soflexid is constricted at its medial opening.

Eopicure new genus

TYPE SPECIES. Eopicure kraglievichi new spe-cies.

INCLUDED SPECIES. Monotypic.DIAGNOSIS. Upper molars strongly rectangular

in occlusal outline. Protocone conical and partiallyisolated from the other major cusps. Metaflexus oc-

cupied either by a small, isolated cusp or enamelspurs from adjoining crests.

ETYMOLOGY. Eos, dawn, Greek; picure, localPeruvian name for an agoutid (Agoutidae), indicat-ing affinity with that group.

DISCUSSION. This genus is placed in the Agou-tidae, subfamily Dasyproctinae, on the basis of sim-ilarities to Eobranisamys in the arrangement of thecrests and in the multiple enamel extensions intothe metaflexus that are suggestive of the neoflexusof Eobranisamys.

Eopicure kraglievichi new speciesFigures 7, 12F; Appendix 2

HOLOTYPE. LACM 143264, left Mx.REFERRED SPECIMENS. Mx: left, 143353,

149455; right, 149454, 149466, 149479. All cat-alogue numbers pertain to LACM collections.

DIAGNOSIS. As for the genus.ETYMOLOGY. Kraglievichi, in honor of Lucas

Kraglievich, a prominent vertebrate paleontologistof Argentina in the first one-third of the twentiethcentury, among whose publications are several per-taining to fossil rodents.

DESCRIPTION. In occlusal outline, these upperteeth have the shape one expects in lower molars,in that the length exceeds the width by an averageof 25 percent (range, 17–40 percent). All six teethhave slight unilateral hypsodonty, which is char-acteristic of upper teeth. Two of the six (LACM143264 and 149455) are unworn and rootless andmay never have fully erupted. One other tooth,LACM 143353, has had the roots broken in sucha way as to partially obscure the single medial root


expected in an upper tooth, but it appears to bepresent. Three teeth (LACM 149454, 149466, and149479) have enough of the roots present to safelysay that they are upper teeth, and, by inference,that the others are as well.

Other aspects of the occlusal surfaces of theseteeth are highly definitive. Of all the rodent speci-mens from Santa Rosa, these six teeth display theweakest expression of lophodonty. The poorlyformed crests exist more as a series of connectedcusps, some elongated to form a portion of a crestand others less so and more suggestive of individualcones. The most obvious expression of this is theisolation of the protocone, which exists more as aseparate conical cusp than as the enlarged end ofthe anteroloph in five of the six specimens. Only inLACM 149455 is the protocone definitely contin-uous with the anteroloph in a totally unworn con-dition, and for that reason this specimen is the leastconfidently assigned to this species. The protolophand metaloph are close and approximately parallel.The metaloph follows a smooth curve and contin-ues as the medial half of the protoloph to the hy-pocone. The lateral half of the protoloph curvessharply near the middle of the occlusal surface andmeets the metaloph. This part of the enamel patternis very similar to that of Eobranisamys. Innumer-able accessory cusps are present on the crowns ofthese teeth. Most frequently these are found in themetaflexus (all specimens), but variations in thecrests are also present in two specimens (LACM143353 and 149454) in which the paraflexus is in-vaded by a spur from the small cuspid that marksthe lateral terminus of the anteroloph.

In three specimens (LACM 143264, 143353, and149479), a small isolated cusp is in the center ofthe metaflexus, whereas in the others, and mostdramatically so in the unworn crown of LACM149455, one to three small enamel spurs enter themetaflexus from the metaloph and posteroloph.

An intriguing feature is the presence of an iso-lated cusp in the hypoflexus of three specimens(LACM 143353, 149454, and 149479). This cuspis especially large in LACM 149479.

On the most worn of the six specimens, LACM149479, an unusual pattern of wear has completelyremoved the protocone. The site is occupied by acircular pit that extends well into the dentine. Evenat this advanced and strange degree of wear, thecircular margin of the protocone is recognizable.The protocone was separated from nearby crestsfor most of its circumference.

DISCUSSION. This taxon is represented in thefauna by only six teeth, all part of the upper den-tition. Lower teeth of this species may lie unrec-ognized among unreferred specimens or specimensreferred to other species in this fauna, but the un-usual features of the upper teeth suggest that com-parably diagnostic features should be present in thelower teeth as well, but no such teeth have beenfound. The more elongate upper molars have anocclusal outline that one would associate with low-

er molars, but the arrangement of roots and uni-lateral hypsodonty are typical of upper teeth.Should any of these six teeth eventually prove to belower teeth, they are sufficiently different from allother lower teeth in the collection to ensure the re-tention of this taxon.

Indeterminate Specimens

After long consideration, the following four teethhave been isolated as unreferable material. They areplaced within the family Agoutidae, subfamily Da-syproctinae because of their general similarities toEoincamys and Eobranisamys. Although there areonly one or two specimens per unnamed taxon,they are sufficiently distinctive to suggest that theyrepresent real, but very rare, taxa in the fauna. Wethink this treatment reflects, probably more accu-rately, the species diversity of the Santa Rosa localfauna. The possibility remains, however, that someof these specimens are no more than curious dentalvariations of named Santa Rosan species.

Agoutidae, Genus and SpeciesIndeterminate A

Appendix 2

REFERRED SPECIMEN. LACM 144299, rightMx.

DESCRIPTION AND DISCUSSION. A singlelower molar represents an unknown taxon that hasa dental pattern similar to that of Eobranisamys.This tooth has four lophids and a dental pattern inwhich the metalophid and hypolophid are close to-gether and angle to meet each other before con-tinuing to the protoconid as the hypolophid. Theunion of the metalophid and hypolophid occurs atthe center of the occlusal surface.

The tooth of this species differs from those attrib-uted to species of Eobranisamys in that it is muchlarger and the metalophid is centrally placed. Iden-tification below the family level is not justified forthis tooth because of the relatively simple and sim-ilar patterns seen in almost all rodent lower teethin this local fauna. The slight differences from low-er teeth of Eobranisamys that have been noted mayor may not be significant.

Agoutidae, Genus and SpeciesIndeterminate B

Appendix 2

REFERRED SPECIMEN. LACM 143375, rightMx.

DESCRIPTION AND DISCUSSION. This spec-imen is very similar to teeth that have been assignedto Eobranisamys riverai, but it differs in several as-pects, one of which might prove to be highly di-agnostic. This molar is slightly larger than the up-per molars of E. riverai, and the union of the me-taloph with the protoloph occurs closer to the hy-pocone. More significantly, however, the neolophdoes not curve posteriorly to meet with the poster-


oloph. Instead, it extends mediad as a long, straightcrest that does not meet another crest, and it wouldnot have connected with another crest until veryextensive wear of the tooth had occurred. This isso unusual that we cannot place this specimen with-in the genus Eobranisamys at this time. If addition-al specimens appear in which this condition can beidentified as a natural variation on the pattern ofEobranisamys, then LACM 143375 might be as-signed to E. riverai. If the position of the neolophremains consistent in other specimens found in thefuture, then this specimen would represent anothergenus closely related to Eobranisamys.

Agoutidae, Genus and SpeciesIndeterminate C

Appendix 2

REFERRED SPECIMENS. LACM 143349 and143351, left Mx.

DISCUSSION. These specimens are much likethose assigned to Eobranisamys romeropittmanae.They would, however, be the largest lower molarsof that species in the collection. These teeth are typ-ical for this group, in that they have four crests inwhich the anterior three converge on the protoco-nid, and in which the connection between the hy-polophid and posterolophid is weak. The most un-usual element of these teeth is unique among spec-imens in the fauna. Near the middle of the metal-ophid are two opposing, but offset, enamel spurs.One faces anteriad and the other posteriad. Thereis considerable variation in surface dental featuresamong the teeth in this collection that frequentlyblur taxonomic distinctions, and in most cases,these variations are taken here as evidence of aclose relationship among taxa because of the earlystage of caviid diversification represented at SantaRosa. On these specimens, however, the enamelspurs are not just odd variations. They are crisplydefined aspects of the grinding surface that extendfrom the floors of the bordering fossettids. This isso unlike any other lower teeth in the sample thatwe think it best to recognize them as unique in thepresent discussion of the local fauna.

Family Echimyidae Gray, 1825

Subfamily Heteropsomyinae Anthony, 1917

Eosallamys new genus

TYPE SPECIES. Eosallamys paulacoutoi newspecies.

INCLUDED SPECIES. Eosallamys simpsoni newspecies.

DIAGNOSIS. Molar teeth brachydont. Enamelpattern similar to that of Sallamys Hoffstetter andLavocat, 1970, but retains four low and roundedlophs with expansive cusps, as opposed to highercrowned pattern of three lophs after only moderatewear and cusps that are less separable from lophsin Sallamys.

ETYMOLOGY. Eos, dawn, Greek; Sallamys, agenus of Deseadan echimyid of South America.

DISCUSSION. On both upper and lower teethof the two species referred to this genus, the posi-tions of major cusps are indicated by terminalswellings in the crests, which is consistent withbrachydonty. These cusps continue to be evidentthrough moderate crown wear and disappear onlyas the enamel crests become broadly confluent.These species are placed near the Deseadan genusSallamys on the basis of the reduced metaloph (up-per teeth) and metalophid (lower teeth) that withwear disappear into union with the posterolophand anterolophid. In so doing, a three-crested den-tal pattern very similar to the higher crowned teethof Sallamys is achieved from what was originally atetralophodont pattern.

The upper teeth of this genus have four principallophs, of which the diagonal protoloph dominatesthe occlusal surface. The central two lophs are in-clined slightly in the direction of the hypocone. Themetaloph is much smaller than the protoloph andvariable in its size and point of contact with theprotoloph. Frequently, it is doubly curved in a sin-uous, flat S-shape that turns anteriad at its contactwith the protoloph. In the terminology chosen forthis paper, the protoloph is considered to be thecrest that extends between the paracone and thehypocone, rather than the shorter portion betweenthe paracone and the site of union between the pro-toloph and the metaloph. This decision was madebecause in most of the Santa Rosan rodents, in-cluding Eosallamys, the metaloph does not connectto the hypocone as early as does the labial half ofthe protoloph, and a strong crest runs directly fromthe paracone to the hypocone, with only a slightdeflection where it is joined, or is to be joined, bythe metaloph. However, in Eosallamys, it is thesmall and initially incomplete metaloph and not theprotoloph that extends as a smooth curve from themetacone toward the hypocone. When the crestsare fully united, the appearance is that of a metal-oph that extends between the metacone and hypo-cone and to which the protoloph joins near the cen-ter of the occlusal surface.

The hypoflexus is deep in Eosallamys, beingplaced at an angle greater than 45 degrees with re-spect to the labial margin of the tooth, and it ap-pears to terminate at the base of the labial half ofthe protoloph. In the other genera of Santa Rosanrodents, the hypoflexus appears to be a continua-tion of the parafossette, with the two fossettes sep-arated only by the connection between the proto-loph and anteroloph. Frequently, but not invari-ably, the metafossette is invaded by spurs or cren-ulations from the posteroloph and with wear maybe divided into labial and lingual halves. Two spec-imens identified as upper third molars of E. pau-lacoutoi new species (LACM 143363) and E. simp-soni new species (LACM 143380) have strangelyshaped crests that connect the posteroloph with themetaloph and disrupt the metafossette. This ap-


pears to be a variable aspect of the teeth in thisgenus, and particularly so with regard to M3s.

The following description is derived mostly fromthe lower teeth of E. simpsoni, for which bettermaterial is available. In the lower teeth of Eosal-lamys, four crests are present with the hypolophidas the strong, centrally positioned loph. As in theuppers, the two central lophids incline slightly to-ward the protoconid, which is in the opposite di-rection of that seen in the uppers. The metalophid,like the metaloph, is relatively smaller and sinuous(S-shaped) in outline. The belly of the S-shapemarks the position of a mure that, with wear, canseparate the anterofossettid into labial and lingualsubfossettids (as happens with the metafossette inthe upper teeth). It is the anterior fossette in thelower teeth, versus the posterior fossette in the up-per teeth, in which irregularities in the borderingenamel form partial lophids and spurs that enterthe fossettid. The metalophid is not complete in allspecimens, and it may curve to meet the anterolo-phid rather than connecting with the hypolophid(LACM 143311 and 143412), further obscuringthe anterofossettid. This condition anticipates theloss of this lophid by its incorporation within abroad anterolophid, as seen in Sallamys. The hy-poflexid is deep and positioned at an angle greaterthan 45 degrees with respect to the lingual marginof the tooth. It appears to terminate at the base ofthe lingual half of the hypolophid rather than existas an extension of the metalophid.

The four-lophed pattern seen in the lower teethreferred to this genus is in direct contrast to thethree-lophed pattern seen in lower cheek teeth ofSallamys. The three-lophed pattern in Sallamys isformed from the reduction of the labial portion ofthe metalophid and the fusion of the lingual portionwith the anterolophid (Patterson and Wood, 1982:figs. 5C,E). Some lower teeth of E. simpsoni newspecies show the beginning of this process (LACM143311, 143412, and 143413).

The octodontid affinities of this genus are indi-cated by the deep hypoflexid that effectively dividesthe tooth into anterior and posterior halves. Thelabiolingual transverse midplane of the tooth ismarked by the anterior edge of the hypolophid.

In the two species assigned to this genus, the en-toconid is present as an enlarged, medial tip of thehypolophid, but the metaconid is small and vari-able in size, and in some specimens of E. simpsoninew species, it cannot be distinguished from the re-mainder of the metalophid. The anteroflexid is thelargest entrant, with the mesoflexid and metaflexidof approximately equal size.

The anterolophid and posterolophid curve lin-gually to meet the two central lophids. For the twonamed species, with moderate wear, the anterofos-settid forms before the posterofossettid. The me-sofossettid is the last of the flexids to close.

Eosallamys paulacoutoi new speciesFigures 8, 12G, 13E–F; Appendix 2

HOLOTYPE. LACM 143422, maxilla, partial,right, with P4–M3.

REFERRED SPECIMENS. P4: left, 149460; right143415, 149480. M3: right, 143363. Mx: left,149494; right, 143267, 144302. dp4: left, 143418(partial), 143451, 149458 (partial); right, 143370,143420, 143450. P4: 149492. M3: right, 143262.Mx: right, 143265, 143361 (partial), 144524. Allcatalogue numbers pertain to LACM collections.

DIAGNOSIS. The larger of two species assignedto this genus. Posteroloph thick; metafossette heavi-ly invaded by enamel spurs. Hypoflexus narrow,forming a hypofossette with moderate to heavywear.



Formation. ?Eocene.ETYMOLOGY. Paulacoutoi, in honor of Carlos

de Paula Couto, a prominent Brazilian vertebratepaleontologist who published many papers on fossilvertebrates of Amazonia.

DISCUSSION. Specimens of this species are un-common in the Santa Rosa local fauna, but an ex-cellent, complete upper tooth row is included in thecollection, and it has been designated as the holo-type. It is moderately worn, but the major elementsof the dental patterns of the various teeth are evi-dent.

Upper Teeth. In Eosallamys paulacoutoi, the me-tafossette is shallow and partially filled with irreg-ular crenulations of the posteroloph. With wear, themetafossette is the first fossette to close. It then dis-appears into several small enamel lakes as the me-taloph fuses with the posteroloph in a pattern thatanticipates the condition of the Oligocene genusSallamys (Patterson and Wood, 1982:fig. 4A). Thehypoflexus appears to be unusually narrow. In theholotypical specimen, the hypoflexus has closedinto a hypofossettid in the first two teeth of thetooth row. In LACM 143363, an M3, the postero-loph is highly crenulated in a manner that is ap-parently typical for this species. In addition, a singleauxiliary crest from the metaloph to the postero-loph divides the metafossette into two subfossettes.That condition is seen again in an M3 of E. simp-soni new species (LACM 143380).

Lower Teeth. The enamel pattern in lower teethis unusual in that a small, enamel spur may extendin a curve from the lingual apex of the anterolophidto attach to the metalophid near its center. Thiscreates a small fossettid from the lingual corner ofthe anterofossettid. This unusual element is consis-tent with the substantial variation seen in the cor-responding valley of the upper teeth, the metafos-sette, and may indicate commensurate variation inthis part of the lower teeth. The metalophid andhypolophid are slightly oblique in the posterolin-


Figure 8 Eosallamys paulacoutoi. Variations in dental and wear patterns in upper and lower teeth. A reversed drawingis indicated by ‘‘r.’’ P4–M3: A, LACM 143422r (holotype of genus). Mx: B, LACM 143267r; C, LACM 143363r. M3: D,LACM 143262r. Mx: E, LACM 143265r. Scale 5 1 mm.

gual to anterolabial direction. The metaconid isclearly present, even in a tooth that shows moder-ate wear. Protoconid and hypoconid are of approx-imately equal size.

Eosallamys simpsoni new speciesFigures 9, 12H, 13G, 14; Appendix 2

HOLOTYPE. LACM 143424, right Mx.REFERRED SPECIMENS. P4: left, 149451. M3:

left, 143380; right, 149448. Mx: left, 143374,143376, 143398, 149457, 149459, 149481,149488; right, 143411, 143429, 143430, 143441(partial), 149472, 149477. dp4: left, 143452; right,143406 (partial), 143407 (partial), 143443 (par-tial), 149467 (partial). P4: left, 143311. M3: left,143284. Mx: left, 143276, 143416 (partial); right,143273, 143275, 143412, 143413, 143414,149491. All catalogue numbers pertain to LACMcollections.

DIAGNOSIS. The smaller of two species referredto this genus. Paracone and metacone large and ap-proximately equal in size. Metafossette not heavily

invaded by enamel spurs. Hypoflexus broadlyopen.



Formation. ?Eocene.ETYMOLOGY. Simpsoni, in honor of George

Gaylord Simpson, who was so instrumental to ourunderstanding of the Age of Mammals in SouthAmerica.

DESCRIPTION. Upper Teeth. In many ways,these teeth are a smaller version of those referredto Eosallamys paulacoutoi, and were it not for thesignificant size difference it would be difficult toseparate some of them. The smaller teeth of E.simpsoni also have a more primitive appearance,with a bulbous paracone and metacone, but sub-stantial variation in this character is found amongthe teeth referred to this species. Enamel intrusionsinto the metafossette may be small (LACM 143429and 149459) or large enough to be called a neoloph(LACM 143374 and 143398). In the latter two


Figure 9 Eosallamys simpsoni. Variations in dental and wear patterns in upper and lower teeth. A reversed drawing isindicated by ‘‘r.’’ Mx: A, LACM 143429r; B, LACM 143430r; C, LACM 143424r (holotype). M3: D, LACM 143380.Mx: E, LACM 143411r. P4: F, LACM 143311. Mx: G, LACM 143273r; H, LACM 143412r; I, LACM 143275r; J, LACM143276; K, LACM 143413r. M3: L, LACM 143284. Scale 5 1 mm.

specimens, the neoloph is formed by two enamelprojections from the posteroloph that meet, but donot completely unite, and therefore do not dupli-cate the strong and highly characteristic neoloph ofEobranisamys.

Lower Teeth. The lower teeth have four lophids,unlike the characteristic three lophids of later echi-myids such as Sallamys. However, the most fre-quent site of variation is within the metalophid, andit is probably this lophid that is suppressed in thelower teeth of later echimyids.

An echimyid feature of these lower teeth thatseems to be less evident in E. paulacoutoi is therelative positions of the hypoflexid and the lingualhalf of the hypolophid. The hypoflexid terminatesnear the center of the occlusal surface, where thehypolophid turns to meet the metalophid beforecontinuing to the protoconid. The lingual half ofthe hypolophid appears to rest upon the lingual ter-

minus of the hypoflexid. The effect is to separatethe curve of the hypoflexid from the curve of themetafossettid and to associate the hypoflexid withthe hypolophid instead.

Eoespina new genus

TYPE SPECIES. Eoespina woodi new species.INCLUDED SPECIES. Monotypic.DIAGNOSIS. Upper teeth tetralophodont,

rounded crowns with thick, tapered marginalcrests, thin internal crests, and large bulbous cusps.

ETYMOLOGY. Eos, dawn, Greek; espina, Span-ish, spine or thorn.

DISCUSSION. Numerous specimens representthe smallest rodent in the Santa Rosa local fauna.They are placed in family Echimyidae, subfamilyHeteropsomyinae on the basis of the emended di-agnosis of that subfamily by Patterson and Wood


(1982:282–283) and because of their basic similar-ity to the teeth of Eosallamys paulacoutoi.

Upper Teeth. The upper teeth of this genus aresimilar to those of Eosallamys in that they have atetralophodont pattern with three labial folds, ofwhich the middle fold, the mesoflexus, is the short-est and most narrow. The protoloph and metalophdiverge only slightly toward the labial margin ofthe tooth. This genus differs from Eosallamys andEosachacui new genus (see below) in that the pos-teroloph lacks the extensive crenulations that par-tially fill the metafossette of those genera.

Other than the unusual and possibly aberrantteeth of Eopicure, in which crests are poorlyformed, the upper cheek teeth of Eoespina appearto be the most primitive among the Santa Rosantaxa. These teeth are rounded in both occlusal andlateral views. The major cusps are large, cone-shaped, and stand well above the level of the crestsin an unworn tooth, and they retain this height dis-tinction until the crown is well worn. All crests arebroadly based and taper toward the occlusal sur-face. The thick anterior and posterior crests com-bine with the four bulbous cusps to nearly encirclethe occlusal surface. The metacone is the smallestprincipal cusp, although this size difference is lessthan that in teeth of other Santa Rosan taxa. Twointernal crests, the metaloph and labial half of theprotoloph, are thin, low, and nearly parallel. Thesetwo crests are separate or weakly connected to thelingual half of the protoloph in an unworn tooth.Thus, until moderate wear unites them, there ap-pear to be three short, thin crests within the interiorof the occlusal surface (i.e., the two aforementionedlabial crests and the oblique, lingual half of the pro-toloph).

The tetralophodont P4 is about half the size ofeither of the first two upper molars. There is con-siderable variation in the size and placement of theprotoloph and metaloph. These crests may appearas a small version of what is seen in a typical uppermolar, or both may exist as projections from thelabial margin of the tooth. In occlusal outline, theP4 is round with a nearly continuous marginal crest.

Lower Teeth. All permanent lower teeth are te-tralophodont, with little evidence of irregularitiesin the anterolophid or anteroflexid that typify sev-eral other taxa in this fauna. The description of theupper teeth as having large, conical cusps and thickmarginal crests applies to the lower teeth as well,although the size distinction between the marginaland interior crests is not as strong. The metaconidis the smallest identifiable cuspid; the metalophid isthe smallest crest and in some examples, probablyP4s, it is incomplete. The reduction to near loss ofthe metalophid is part of the subfamilial diagnosisof Patterson and Wood (1982).

Eoespina woodi new speciesFigures 10, 12I, 13H–I; Appendix 2

HOLOTYPE. LACM 143286, left Mx.REFERRED SPECIMENS. P4: left, 144301;

right, 143432, 143459, 149468. M2: left, 149441.M3: left, 143288, 143298; right, 143297. Mx: left,143291, 143293, 143397, 144300, 149431,149432, 149471, 149484; right, 143289, 143290,143295, 143296, 143384, 143455, 143457,149427, 149433, 149436, 149437. Mandibularfragment, right, with P4–M2, 144294. dp4: left,149440; right, 143391, 143403, 143404 (partial),143405, 143448 (partial). P4: left, 143283. M1: left,149425, 149469; right, 149434. M3: left, 143386,143438; right, 143282, 143440, 143461. Mx: left,143389, 143436, 143446; right, 143281, 143327,143330, 143385, 143390, 143439, 143447,149429, 149486. All catalogue numbers pertain toLACM collections.



Formation. ?Eocene.ETYMOLOGY. Woodi, in honor of Albert

Wood, who contributed so much to our under-standing of rodent evolution.

Eosachacui new genus

TYPE SPECIES. Eosachacui lavocati new spe-cies.

INCLUDED SPECIES. Monotypic.DIAGNOSIS. Union of the protoloph and me-

taloph occurs near center of occlusal surface. Me-soflexus widens laterally. Protoloph large, slightlycurved, with only a slight flexure at its union withthe metaloph. Posteroloph highly variable, withmultiple enamel projections into metaflexus, whichmay extend to metaloph.

ETYMOLOGY. Eos, dawn, Greek; sacha cui, lo-cal Peruvian name for a spiny rat (Echimyidae).

DESCRIPTION. Upper Teeth. The paracone andmetacone are seen as enlarged cones on the labialends of the protoloph and metaloph. The proto-cone and hypocone are more compressed and in-corporated within the anteroloph and posteroloph.The occlusal surface is crossed by four crests, ofwhich the metaloph is the smallest. The two centralcrests diverge slightly toward the labial surface,with the metaloph placed transversely and the pro-toloph oblique to the transverse axis. The obliquityof the protoloph is created by the very anteriorplacement of the paracone, which lies even with, orslightly anterior to, the protocone. The protolophmaintains a constant width from paracone to hy-pocone. There is a deflection in the protoloph atthe point at which the metaloph unites with theprotoloph after moderate wear of the tooth surface,but this change of direction is directed toward theprotocone rather than toward the anteromedial di-rection, with the result that this change of directionis less abrupt than is frequently seen in taxa of thislocal fauna. For example, in Eobranisamys andEoespina the great deflection in the path of the


Figure 10 Eoespina woodi. Variations in dental and wear patterns in upper and lower teeth. A reversed drawing isindicated by ‘‘r.’’ P4: A, LACM 143459r. Mx: B, LACM 143286 (holotype of genus); C, LACM 143291; D, LACM143384r. P4: E, LACM 143283. Mx: F, LACM 143390; G, LACM 143446; H, LACM 143281r. M3: I, LACM 143440r.Scale 5 1 mm.

transverse protoloph creates a more rectangularbase of the mesofossette. At an early stage of union,the lingual end of the metaloph appears to turnback toward the deflection of the protoloph andthen to fuse with it. When fully united, the meta-loph appears to be the principal, most smoothlycontinuous crest running from the metacone to thehypocone. The union of these two crests has theappearance of a Y, as is seen in Eosallamys fromthis fauna. The large number of irregularities of theposteroloph, in which enamel spurs of various size,number, and position may occupy much of the me-tafossette, is another similarity to Eosallamys.These enamel spurs are particularly strong in theM3, and they may subdivide the metafossette intotwo or three smaller fossettes.

Three labial flexuses are present, which, withonly moderate wear, create three fossettes. The or-der of closing is, first the parafossette, then the me-tafossette, and finally, and only with extreme wear,the mesofossette. The resulting fossettes are roughlyoval in outline. The largest is the metafossette. Themesofossette and parafossette are approximatelyequal in size. The hypoflexus is deep and nearlyreaches the center of the occlusal surface. Its curvedaxis is continuous with that of the parafossette andthe two form a single curved valley in the unwornor slightly worn tooth. After considerable wear hasoccurred, these two flexuses are separated by onlya small mure.

Lower Teeth. The occlusal surfaces of lower teethof Eosachacui have four crests and three labialflexids. The anterolophid, hypolophid, and poster-olophid are approximately the same size. The me-talophid is small, but usually uniform in width andlength. In three specimens (LACM 143325,143329, and 149426; probably M1s, but possiblyP4s), the metalophid is incomplete at its middle sec-tion or turns anteriad at this point to join the an-terolophid rather than the hypolophid. When themetalophid unites with the hypolophid, as is typi-cal, it does so midway between the protoconid andthe lingual deflection of the hypolophid. In so do-ing, the angle of union is less than 45 degrees anda Y configuration of the crests occurs, as in theuppers. The lingual half of the hypolophid is trans-verse to slightly oblique toward the posterior. Thelingual half of the hypolophid is equal to or smallerthan the lingual half of the anterolophid.

Of the three flexids, the mesoflexid and meta-flexid are of nearly equal size and both are smallerthan the anteroflexid. On most specimens, the an-teroflexid is disrupted by either an irregular marginto the anterolophid or enamel spurs from the an-terolophid or metalophid.

One specimen, LACM 143394, which appears tobe a P4 but may be an M1, is unusual in that thereseem to be five crests present rather than four. Thisis surely an anomalous pattern because there ap-pear to be two protoconids. The more posterior of


the two is smaller and more of a buttressing cuspuleto the main protoconid. Two crests extend fromthese two protoconids toward the lingual side ofthe tooth. The posterior of the two anomalouscrests is recognizable as the hypolophid at its unionwith the metalophid. The anterior of the two crestsis composed of two principal cuspules, one ofwhich, the more lingual, is connected to an enamelspur that extends posteriad from the anterolophid.As one of the characters of Eosachacui is the pres-ence of cuspules and enamel spurs in the antero-fossettid, this is apparently an extreme example ofsuch variation.

The hypoflexid is deep, extending to near thecenter of the occlusal surface and terminating at thelingual deflection of the hypolophid. As such, thehypoflexid is continuous with the curve of the lin-gual half of the hypolophid rather than seeming tobe the continuation of the curve of the metaflexid.

DISCUSSION. These teeth represent anothersmall heteropsomyine rodent in the fauna that isonly slightly larger than Eoespina and with whichit shares many similarities. In the upper molars, themost significant difference is the presence of enamelvariations in the form of spurs or cuspules in theparaflexus versus a comparatively simple paraflexusin Eoespina. Also, the deflections of the protolophin Eosachacui are less severe than those seen inspecimens referred to Eoespina, and the medial ex-tent of the mesoflexus narrows in contrast to acomparatively blunt, square medial margin of thisflexus in Eoespina.

In the lower teeth, both genera have greatly re-duced the metalophid as a precursor condition tothe trilophid pattern that is characteristic of youn-ger echimyids. In Eosachacui, this crest generallymaintains a constant width and presence in the oc-clusal surface, but it may be broken, or constricted,or it may turn into the anteroflexid rather than jointhe hypolophid. The anteroflexid of Eoespina sel-dom has additional cuspules or crests. In Eoespina,the metalophid appears not to be uniform in eitherheight or thickness. The greatest height and thick-ness are found at its union with the hypolophid,and toward the metaconid the metalophid maynearly disappear and not be included as part of theocclusal surface. The close positions of the proto-lophid and metalophid in Eosachacui, and a resul-tant narrow mesoflexid versus a more open meso-flexid in Eoespina, might be constant features, butthe difference is slight. More variable is the orien-tation of the lingual half of the hypolophid. It istransverse or slightly oblique toward the posteriorin Eosachacui, and, in Eoespina, it appears to beeither transverse or slightly oblique toward the an-terior, producing a more rounded mesofossettid.

Eosachacui lavocati new speciesFigures 11, 12J; Appendix 2

HOLOTYPE. LACM 143294, left Mx.REFERRED SPECIMENS. P4: left, 143433,

143458, 143466. M3: left, 143401. Mx: left,143292, 143337, 143383, 143399,143444,143453; right, 143278, 143285, 143387, 143388,143454, 149438, 149439. dp4: left, 143326,143400, 143449 (partial); right, 143379, 143402.P4: left, 143394. M1: left, 143329, 149426; right,143274. M3: left, 143395. Mx: left, 143287,143393, 143408, 143417, 143431, 143442 (par-tial), 143443 (partial), 149447; right, 143325,143382, 143421, 149428. All catalogue numberspertain to LACM collections.



Formation. ?Eocene.ETYMOLOGY. Lavocati, in honor of Rene Lav-

ocat, for his contributions to our knowledge of ro-dent evolution.

Subfamily Adelphomyinae Patterson andPascual, 1968

Eodelphomys new genusTYPE SPECIES. Eodelphomys almeidacomposi

new species.INCLUDED SPECIES. Monotypic.DIAGNOSIS. Tooth with large entoconid occu-

pying posterolabial corner. Occlusal surface of low-er teeth dominated by an extended, straight, diag-onal hypolophid that connects entoconid with pro-toconid. Hypolophid and posterolophid weaklyconnected by a mure. Metalophid and anterolophidunconnected to hypolophid, but join to create alens-shaped anterofossettid.

ETYMOLOGY. Eos, dawn, Greek; delphomys,for sharing similarities with Paradelphomys Patter-son and Pascual, 1968, and Adelphomys Ameghi-no, 1887, fossil rodents from the Miocene of Ar-gentina.

DESCRIPTION. This single specimen is so dis-tinctive that we have departed from our practice ofusing only upper teeth as holotypes and of givingnames to only those taxa for which numerous spec-imens are available. The dental pattern of this spe-cies is typified by a long hypolophid that runs di-agonally across the occlusal surface from a largeentoconid to the protoconid, with only a minorcurve as it passes the hypoflexid. Smaller lophidsparallel the long hypolophid anteriorly and poste-riorly without uniting (i.e., without showing a con-fluence of dentine) until a late wear stage.

DISCUSSION. This genus is referred to the fam-ily Echimyidae, subfamily Adelphomyinae becauseof a dental pattern that is close to that of the Oli-gocene genus Paradelphomys, of that family andsubfamily. The occlusal surfaces of teeth from gen-era in this group are dominated by a strong, diag-onal lophid (the hypolophid) that is paralleled by ashorter lophid to the anterior (likely a combinedanterolophid and metalophid) and one to the pos-


Figure 11 Eosachacui lavocati. Variations in dental and wear patterns in upper and lower teeth. A reversed drawing isindicated by ‘‘r.’’ P4: A, LACM 143458. Mx: B, LACM 143294 (holotype of genus); C, LACM 143383; D, LACM143285r. P4: E, 143394. Mx: F, 143325r; G, LACM 143382r; H, LACM 143393; I, LACM 143329; J, LACM 143287;K, LACM 143274r. Scale 5 1 mm.

terior (the posterolophid). Both an anterolophidand a metalophid are present in Eodelphomys, butin close proximity to each other. This M3 differsfrom the molar of Paradelphomys in that it is ex-tremely brachydont, tetralophodont instead of tri-lophodont (the anterior two crests have fused inParadelphomys), and it has a larger entoconid. Theanterolophid and metalophid of Eodelphomys areseparate crests united at their two ends to create anisolated fossettid, the anterofossettid. These crestsmay have been separate in Paradelphomys at a veryearly wear stage, but they are fused in both teethof the specimen of Paradelphomys described byPatterson and Pascual (1968), a partial mandiblewith dp4–M1, and these crests are smaller thanthose of Eodelphomys.

The holotypical specimen and the specimen de-scribed by Patterson and Pascual (1968) are at sim-ilar stages of wear. A small mure between the pos-terolophid and hypolophid is present in Eodelpho-mys, but a potential connection is only suggested

in Paradelphomys, and it would not have beenachieved until late wear.

The combined anterolophid/metalophid of Par-adelphomys is united with the hypolophid labiallyin the M1 and at both ends in the dp4. This is anearlier union of these crests than is seen in the toothof Eodelphomys, and this earlier union is probablya result of the higher crown in Paradelphomys.

The dental pattern of a strong, central lophidmatched fore and aft by shorter lophids is alsocharacteristic of Eoincamys, Incamys, and somelater dasyproctids. However, there are numerousdifferences that offset the fundamental similarity.The weak connecting mure between the hypolophidand the posterolophid is probably a shared primi-tive character between Eodelphomys and these ear-ly dasyproctids in that this mure is part of the in-terconnected crest pattern in higher crowned mo-lars. The posterolophid of Eodelphomys differsfrom that of teeth in species of Eoincamys and In-camys in that its labial terminus does not reach the


Figure 12 Summary illustration of dental characteristics and size comparison of rodent taxa included in the Santa Rosalocal fauna. All teeth are drawn to the same scale. Both upper and lower teeth, if known, are presented for each taxon.All are drawn as left teeth with the anterior to the left and the protocone or protoconid at the lower left of each sketch.Catalogue numbers list the upper tooth and then the lower; the letter ‘‘r’’ indicates a reversed drawing. A, Eopululowigmorei, LACM 143269, LACM 143410; B, Eoincamys pascuali, LACM 143319, LACM 143306; C, Eoincamysameghinoi, LACM 143339, LACM 143435; D, Eobranisamys romeropittmanae, LACM 143359r, LACM 143341; E,Eobranisamys riverai, LACM 143373, LACM 143366; F, Eopicure kraglievichi, LACM 143353; G, Eosallamys paula-coutoi, LACM 143422r, LACM 143262r; H, Eosallamys simpsoni, LACM 143430r, LACM 143412r; I, Eoespina woodi,LACM 143286, LACM 143390; J, Eosachacui lavocati, LACM 143294, LACM 143382; K, Eodelphomys almeidacom-posi, LACM 144298r.


labial margin of the tooth. In Eodelphomys, thismay be a result of the reduced nature of the pos-terior half of the M3, but the same condition is seenin more anterior teeth in Paradelphomys. The me-talophid is reduced in Eoincamys and Incamys,whereas it is nearly as large as the anterolophid inEodelphomys.

Eodelphomys almeidacomposi new speciesFigure 12K; Appendix 2

HOLOTYPE. LACM 144298, mandibular frag-ment, right, with M3.


cality 6289.TYPE AND HORIZON AND AGE. ?Yahuar-

ango Formation. ?Eocene.ETYMOLOGY. Almeidacomposi, in honor of

Diogenes de Almeida Compos, a friend, colleague,and leading vertebrate paleontologist of Brazil.

Family Echimyidae, Genus and SpeciesIndeterminate A

Appendix 2

REFERRED SPECIMENS. LACM 143409, par-tial right Mx.

DISCUSSION. A large echimyid in the fauna isindicated by a single specimen. The upper toothfragment displays a tetralophodont pattern inwhich the hypoflexus lies in the same axis as thelabial portion of the protoloph, and the metalophjoins the protoloph away from the center of theocclusal surface and toward the hypocone. Theseare characters of Eosallamys. Differences thatmight be diagnostic are the narrow mesofossetteand the proximity of the tips of the metaloph andposteroloph. In this tooth, the metaloph and pos-teroloph would join at an early wear stage, longbefore the union of the anteroloph and protoloph.

Echimyidae, Genus and SpeciesIndeterminate B

REFERRED SPECIMEN. LACM 143348, leftMx.

DISCUSSION. This single specimen is very sim-ilar to those assigned to Eosallamys. Very charac-teristically, the anterofossettid has been divided intolabial and lingual subfossettids by a short crest thatextends anterolaterally from the metalophid to theanterolophid. Additionally, the metafossettid is sim-ilarly divided, although incompletely so, even atthis late wear stage. The largest species assigned toEosallamys is E. paulacoutoi, to which this speci-men could possibly be referred, but it is about 30percent larger than the lower teeth assigned to thatspecies. The subjective decision at this time is toseparate it from E. paulacoutoi as perhaps indica-tive of another echimyid in the fauna.

Echimyidae, Genus and SpeciesIndeterminate C

REFERRED SPECIMEN. LACM 143277, leftMx.

DISCUSSION. This single tooth is referred to thefamily Echimyidae on the basis of its similarity toDeseadomys Wood and Patterson, 1959. As in De-seadomys, the metafossette formed well before theparafossette or mesofossette, and, with moderatewear, it is approaching obliteration. The mesofos-sette is more rectangular than that of Deseadomys,although, like Deseadomys, the metaloph does turntoward the posteroloph at its labial end in the for-mation of the metafossette. This specimen exhibitsapproximately the same amount of wear as that ofan upper dentition referred to Deseadomys aram-bourgi (Wood and Patterson, 1959:fig. 4), but at asimilar wear stage the hypoflexus of the Santa Rosaspecimen has been closed into a hypofossette thatmatches the insubstantial metafossette. No suchcorrespondence is seen in the referred upper denti-tion of Deseadomys, which displays what is for thiswear stage in all rodents a more typically open hy-poflexus.

This specimen is within the size range of Desea-domys arambourgi (Wood and Patterson, 1959:ta-ble 1).

Suborder Hystricognatha Woods, 1976


SPECIMENS. LACM 143271, 143350, 143392,143425, 143426, 143460, 143462, 143463,143464, 143465, 143467, and many unnumbered,fragmentary specimens of upper and lower teeth.

DISCUSSION. A number of teeth, approximate-ly 400, are included in the sample from the SantaRosa local fauna that cannot be identified, mosteven to family. Well over half of these teeth areincisors, mostly fragments. The remainder are smallfragments that consist of as little as one crest, andseveral specimens are so highly abraded that thedental patterns are totally obscured. The latterspecimens may be important in paleoecological re-construction in that they were found mixed togeth-er with completely intact and unabraded speci-mens. Many of the fragmentary teeth are a resultof fracturing in situ, followed by separation of piec-es during the recovery of the fossils (see Campbellet al., 2004).

DISCUSSION

FAUNAL CHARACTERISTICS

The rodents of the Santa Rosa local fauna appearto represent an early stage in the adaptive radiationof a vertebrate lineage, wherein the members of thenewly formed lineages, possibly excepting the Er-ethizontidae, are not distantly removed from theircommon origin. The impression derived from thedetailed study of these taxa is that the differences


Figure 13 Comparison sketches of deciduous premolar (dp4) patterns of Santa Rosan rodents. A reversed drawing isindicated by ‘‘r.’’ A, Eoincamys pascuali, LACM 143335; B, Eobranisamys romeropittmanae, LACM 143352; C,Eobranisamys riverai, LACM 143371r; D, Eobranisamys riverai, LACM 143369; E, Eosallamys paulacoutoi, LACM143420r; F, Eosallamys paulacoutoi, LACM 143451; G, Eosallamys simpsoni, LACM 143452; H, Eoespina woodi,LACM 143391r; I, Eoespina woodi, LACM 143405r. Not to scale.

in tooth morphology among the Santa Rosan ro-dents are slight variations on a theme that in an-other context might serve to differentiate genera ofone family of rodents. Overall, the distinctionsamong the dental patterns of the several genera arerather subtle. Despite this subtlety of the observeddifferences, however, they are consistent, and trendslinking some Santa Rosan rodents with later taxacan be identified with confidence.

The rodents of Santa Rosa all have a commonpattern comprising four principal cusps that areconnected by low, broad-based, and tapering crestson both the upper and lower teeth. The cusp pat-terns of the upper and lower teeth (Fig. 1) appearto be arranged around cusps homologous to thoseseen in the teeth of taxa within the Sciuromorphaor Myomorpha, in which the paracone/paraconidof the tritubercular pattern has been reduced in allteeth. If one were to mentally reverse the trend to-ward lophodonty observed in these and youngerSouth American rodents, a version of a hypotheti-cal ancestor would appear in which the dental pat-tern was bunodont, but possibly not basined, withfour primary cusps in which the upper labial andlower lingual cusps were positioned closer to eachother than were the opposite pair. The dental con-figurations and size diversity of all the Santa Rosanrodents described in this paper are compared in Fig-ure 12.

The basic pattern of P4s in the Santa Rosa sam-

ple, excluding those of the readily diagnostic Eoin-camys, consists of two anterior crests and two pos-terior crests that converge, respectively, on the pro-tocone and hypocone. Only a small connection, aweak mure, between the protoloph and metalophdeveloped as wear altered the occlusal pattern. Thehypoflexus is nearly encompassed by the protoconeand hypocone (the latter being the smaller of thetwo cusps), and with only slight wear this flexus isentirely enclosed as a small, deep, medial enamellake. These teeth are so symmetrical that at times,it is difficult to determine which is the anterior sur-face. The pattern seen in these P4s is here inter-preted as a modification of a molariform patternthat has been compressed onto a nearly circular oc-clusal surface. Species assignments of the P4s werebased on perceived similarities to upper molar teeththat are of the appropriate size.

The deciduous lower premolars of rodents in theSanta Rosa local fauna are also basically similar,and they are important to the larger considerationsof caviomorph relationships. Patterson and Wood(1982:508) stated, ‘‘The best opportunity for com-paring thryonomyoid and caviomorph antemolarteeth lies in the deciduous dentition, dm .’’ The4

4

dm of these authors are the retained, or perma-44

nent, dp of others, and the latter terminology is so44

used here. In this study, dp4s are identified for Eoin-camys, Eobranisamys, Eosallamys, and Eoespina(Fig. 13). Within those genera, referral to species


was based on size. The posterior two-thirds of thetooth can be described with the terminology of thelower molars, but the pattern of the anterior one-third of the tooth is highly variable and often ob-scured by wear. The anterior extension of the enam-el pattern is created by two crests that replace theanterolophid of permanent teeth. These two crestsapparently originate as two parallel plates that,with only a little wear, might join in a variety ofways. The primary distinctions among the dp4s ofthese four genera lie in the positions and inclina-tions of the hypolophid and metalophid, whichmatch the corresponding elements in the permanentlower teeth of the respective genera.

In Eoincamys (Agoutidae), this means that thehypolophid is a diagonal crest that extends acrossthe occlusal surface, the metalophid is reduced, andthere is no mure connecting the hypolophid to theposterolophid. For Eobranisamys (Agoutidae), bythe terminology of this paper, the metalophid ap-pears to be a continuous, transverse crest that isconnected to the lingual half of the hypolophid. Inthe deciduous lower premolars allocated to the twogenera of Echimyidae (i.e., Eosallamys and Eoes-pina), the hypolophid is slanted in the anterolabialdirection for its labial half. The metalophid is some-what reduced, although larger than that of Eoin-camys. It is smaller than the metalophid of Eob-ranisamys, and it tends to be inclined slightly to-ward the anterolingual margin.

Indeed, because of the basic, simple structureseen in the teeth of all of the rodents from SantaRosa, had South American rodents become extinctafter deposition of the Santa Rosa local fauna, it ispossible that all of these early rodent genera couldhave been grouped closely together taxonomically,perhaps within a single family. With younger taxafor comparison, the separation of the Santa Rosanrodents into several families is justified and betterreflects the radiation from an ancestral morphotypepossibly not much older than Santa Rosan time.The erethizontids are generally considered to be alone branch that separated early and has continuedto the present. Eopululo, as our single, most prob-able erethizontid, is, therefore, more separated tax-onomically in this paper than might be justified ondental patterns alone, which is all that is availablenow.

EARLY DIVERGENCE IN CAVIID RODENTS

Wood (1949) and Wood and Patterson (1959:294–295) gave considerable importance to Platypitta-mys (Deseadan SALMA; Patagonia, Argentina) asa structural ancestor, at least, of the caviid rodents.Platypittamys was placed in the Acaremyidae (Oc-todontoidea) by Wood (1949). The postulated im-portance of this genus in caviid evolution was notchanged in Wood and Patterson (1959:295), but itstaxonomic placement was described more cautious-ly as ‘‘primitive Octodontidae.’’ Currently (McKen-na and Bell, 1997), Platypittamys is placed within

the subfamily Acaremyinae, family Octodontidae,although Vucetich and Kramarz (2003), in a cla-distic analysis, place Platypittamys as a sister taxonto a group consisting of Acaremyidae (as used bythem and restricted to three genera) and anothergroup made up of a mixture of ‘‘acaremyids,’’ oc-todontids, and echimyids.

However, the supposed primitive dental patternof Platypittamys and the retention of primitive fea-tures in living octodontids (Wood and Patterson,1959:295) have given, in our estimation, undue im-portance to the Octodontidae as a basal group inthe radiation of the caviids. It is our view that thedental characters of Platypittamys are more spe-cialized than those seen in the Santa Rosan rodentsin features that herald the highly characteristicmodern octodontid dental pattern. These charac-ters would include the large hypoflexus(-id), which,with the mesoflexus(-id), forms a distinct waist be-tween the anterior and posterior halves of a tooth,and the shallow anterior and posterior fossettes andfossettids. In this paper, no Santa Rosan rodents areplaced in the Acaremyinae, and we suggest that theOctodontidae, like the Erethizontidae, represents alineage of South American rodents separate fromthat which gave rise to all other taxa of caviids. Infact, the overall resemblance of the dental patternseen in Santa Rosan rodents to the primitive ereth-izontid pattern, that is, a short hypoflexus(-id) andthree opposing fossette(-id)s of nearly equal lengthand depth, leads us to conclude that the erethizon-tid dental pattern is closer to the ancestral caviidcondition than that seen in the Acaremyinae andthat the roots of the caviid radiation go much deep-er than the earliest appearance of the acaremyines,including Platypittamys. Because the basic SantaRosan rodent dental pattern appears to be that ofan echimyid, we agree with the statement of Vuc-etich and Kramarz (2003:443), ‘‘The Octodontidaeprobably originated from an ancestor closer to theEchimyidae than the Acaremyidae . . . .’’

TETRALOPHODONTY VERSUSPENTALOPHODONTY

All but one of the Santa Rosan rodent genera havetetralophodont upper and lower teeth, the hypo-thetical ancestral number of lophs proposed byWood (1949) for erethizontids and caviids. The sin-gle exception, Eobranisamys, has the basic tetral-ophodont pattern, but it also has a small, accessoryloph that projects from the posteroloph into themetaflexus on its upper teeth, which creates a pen-talophodont pattern. This loph is variable in sizeand location. It is clearly not a primary loph, butrather an addition to the tetralophodont conditioncreated by the alignment of enamel irregularities onthe surface of the metaflexus, or from crenulationson the anterior surface of the posteroloph, or both.Similar irregularities cause a noticeably crenulatedanterior margin of the posteroloph in two otherSanta Rosan genera, Eopululo and Eosallamys. It


might yet prove to be that the original number oflophs in caviids was five, the hypothetical penta-lophodont ancestral condition argued by many(Lavocat, 1981; Vucetich and Verzi, 1990; Candela,1999, 2002), but it appears to us that the variationin what we refer to as the neoloph, and in its coun-terpart the anterolophid, is greater than whatwould be expected if it were an additional, or fifth,loph in the process of being lost. In most of theSanta Rosan rodent upper molars, there is an abun-dance of minor cuspules in the posterior part of themetaflexus, of which only some are connected tothe posteroloph and produce a crenulated marginwith wear. In our view, this variation is consistentwith the establishment of dental patterns proceed-ing from the union of minor cuspules that, by SantaRosan time, had perfected a fifth loph in only onegenus, Eobranisamys.

The basic pentalophodont pattern of upper mo-lars seen in many Old World hystricognaths is, wethink, correctly identified with an additional crestplaced third in order from the anterior. This is ei-ther the mesoloph (Lavocat, 1981, and referencestherein) or a mesolophule (Flynn et al., 1986; Fig.2), and it might, in fact, be either in different gen-era. In this instance (i.e., Old World hystricog-naths), it is the fourth crest that extends toward theinterior of the crown from a large cusp, the meta-cone, and which should therefore be called a me-taloph. In contrast, the third labial cusp in Eobran-isamys, at the labial union of the posteroloph anda neoloph, when such a union occurs, appears tous to be an enlargement of the most labial of aseries of cuspids that compose the unworn poster-oloph. The proposal that this third cusp might bethe metacone and the central, or second labial, cuspthe mesostyle [as in the theoretical pentalophodontprimitive condition of these rodents proposed byLavocat (1981) and Candela (1999)] is not accept-ed in this paper because of the overall similarity incusp placement and prominence among all of theSanta Rosan rodent teeth, including the pentalo-phodont Eobranisamys, and because of the variablenature of the posteroloph in which the minor cus-pules comprising the labial enlargement of this lophwear into a highly variable, crenulated anteriorborder.

Butler (1985) argued that it would be uniqueamong rodents for the paracone and metacone ofthe caviids to be closer together than the protoco-nid and hypoconid, rather than alternate withthem, as must be true if the dental homologies ofPatterson and Wood (1982) are correct. Nonethe-less, as we identify the cusps in the Santa Rosanrodents, that is exactly the situation we find, andthis organization of the cusps might be a definitivefeature of the Caviida. We consider the nonalter-nating placement of these cusps as a primitive fea-ture similar to the placement of these cusps in someearly North American (including Mexico) rodentssuch as Protoptychus Scott, 1895; JaywilsonomysFerrusquia-Villafranca and Wood, 1969; Mysops

Leidy, 1871; or Marfilomys Ferrusquia-Villafranca,1989. The most obvious difference between thesegenera and the Santa Rosan rodents is that the pro-toloph, as used in this paper, touches the postero-labial margin of the protocone. Further separationof the protoloph from the protocone, either untilfull separation is achieved or until the protoloph isconnected to the protocone only by a short mure,is a short step to the basic dental pattern for upperteeth of Santa Rosan rodents. Although we consid-er the structural plan of the dental pattern in theseNorth American rodents to be similar to what wemight expect to see in pre–Santa Rosan SouthAmerican rodents, the fact that all of the above arecontemporaneous or younger than the structurallymore derived Santa Rosan rodents removes themfrom consideration as possible ancestral lineagesfor the South American rodents.

To argue for pentalophodonty as the most prim-itive dental pattern among the Santa Rosan rodentswould require one of two hypothetical evolutionarypaths for the teeth of all erethizontid and caviidgenera, except Eobranisamys. Sensu Lavocat(1981), the metacone and metaloph were lost anda mesostyle enlarged in all of the tetralophodontforms except Eoincamys, in which the mesolophwas lost. Alternatively, sensu Flynn et al. (1986),the mesolophule was lost and the metacone movedanteriorly, or the metacone and the labial portionof the metaloph was lost and the labial terminus ofthe mesolophule was enlarged into a cusp that re-placed the metacone. Coincident with these chang-es suggested by Flynn et al. (1986), the orientationof the lingual portion of the metaloph moved an-teriorly to unite with the protoloph, except in Eoin-camys, in which it moved posteriorly to unite withthe posteroloph. Parsimony alone would argueagainst these possible routes to tetralophodonty.We take the position that the great variation seenon the anterior border of the posteroloph is not ametaloph or metalophule in various stages of re-duction, but rather a feature that is common amongthe teeth of the Santa Rosan rodents. In Eobrani-samys, alone among the genera of Santa Rosan ro-dents, this feature has become clarified, prominent,and a characteristic structure.

Additional cusps and crests are common on teethof those North American and African early Tertiaryrodents that have been proposed as morphologicalprototypes of the caviomorphs. No such additionsare evident on the relatively clean cusp morphologyof the Santa Rosan rodents. Most notably, there isno clear indication of a mesocone on any rodentmolar from Santa Rosa. In unworn teeth from thissample, the crests appear as a series of minor cus-pules. It might yet prove to be that one of theseminor cuspules is a mesocone, but, if so, it is notvery much different from nearby cuspules. With nomesocone there can be no mesoloph (see Wood andWilson, 1936, for cusp terminology) on any rodenttooth from Santa Rosa. The crest that might beidentified as such by those who advocate a penta-


lophodont ancestor for the caviids originates notfrom the center of the crown, in the position of themesocone, but from a labial cone, from which itextends inward with decreasing height. We inter-pret that labial cone as the metacone; therefore, theprimitive pattern would be tetralophodont. For therodents of Santa Rosa, either simplification of thedental pattern from a pentalophodont conditionhad occurred, or an abundance of accessory cuspstructures in these taxa that would have led to pen-talophodonty had not yet developed. On the basisof their extreme brachydonty, in which the dentalpatterns are not much elevated from a bunodontcrown, the overall similarity of the rodent teethfrom Santa Rosa to each other, and a predominancein the fauna of tetralophodont upper molars, weprefer the latter argument.

CREST OBLIQUITY AND CHEWING

In rodents, a scissorslike cutting surface is formedby passing oblique crests of upper and lower teethagainst each other. As a function of brachydontyand the large primary cusps in all rodent molarsfrom Santa Rosa, this shear is poorly formed, andthe flattened, scissorslike cutting surface of Neo-gene and Recent caviids was not yet developed.Rather, wear appears to have proceeded in a dis-tinct pattern. First, the lingual side of the uppermolars and the labial side of the lower molarswould wear (see various wear stages of teeth in Ap-pendix 2). Then a sloping surface would extendacross the loph(-id)s to the tips of the labial (upper)and lingual (lower) cusps. Further wear flattenedthe lingual (upper) and labial (lower) portion of theocclusal surface, leading to a steeply sloping wearpattern on the opposite side that extended to thetips of the cusps (paracone/metacone, uppers; meta-conid/entoconid, lowers). The tips of these cuspswould then wear into a keyhole pattern. This pro-duced a roughly S-shaped, or terraced, wear sur-face. On upper molars, the wear surface extendedfrom the lingual side of the tooth, across abouttwo-thirds to three-fourths of the tooth, then slopedsteeply to the tips of the paracone/metacone.

This terraced wear pattern is different from thatobserved by Vucetich and Verzi (1996) in that thereis no wear on the anterior or posterior sides of thecusps. In fact, there is no indication of a labiolin-gual stroke to the chewing motion at any wearstage. This suggests that the Santa Rosan rodentshad already achieved, or nearly achieved, a form ofpropalinal chewing. This could be considered gradeD of Butler (1985), except that the lateral (upper)and lingual (lower) cusps remain elevated above thewear surface until a late wear stage is reached. Inthe later wear stages, the S-shaped occlusal surfacebecame flattened and took on an appearance simi-lar to later caviids, except that by this stage, theinitial brachydont condition resulted in very shal-low fossette(-id)s, as opposed to the deeper fosset-te(-id)s in younger caviids. This type of chewing

might not have been universal among the Santa Ro-san rodents, and it still needs to be confirmed byanalysis of wear striae on the occlusal surfaces, butit appears to have been the most important modeof chewing in these rodents.

In Eopululo, as in all erethizontid rodents, theobliquity of the crests is less severe than in otherSouth American rodents. The occlusal surface is di-visible into two equal parts that are separated by ashallow hypoflexus(-id) and a thin, straight meso-flexus(-id). Anterior and posterior halves of the oc-clusal surface are clearly separated on the basis ofpositioning and union of the crests. The posteriorhalf is primarily a connection between the meta-cone and hypocone. The protoloph remains an el-ement of the anterior portion of the tooth that isstrongly connected with wear to both the proto-cone and the metaloph. The combination of twostraight central crests and two curved marginalcrests on both upper and lower teeth in this genus,combined with the offset of the upper and lowerteeth, always positions a straight crest against acurved crest during the chewing motion.

In the upper teeth of the other genera of rodentsfrom Santa Rosa, there is a sweep of crests towardthe posterior. This is most dramatic in Eoincamys,in which the metaloph sweeps posteriad to unitewith the posteroloph rather than with the hypo-cone. Eobranisamys, Eosallamys, and Eoespinahave similar, but not as extreme, patterns in thatthe metaloph is the straighter of the two centrallophs and the protoloph curves to meet it. In Eos-allamys, the inclination of the two central crestsand the accessory spurs that extend into the meta-flexus are diagnostic. Eoespina is distinguished bya weak connection of the metaloph to the proto-loph, whereas the distinct neoloph of Eobranisamysdistinguishes that genus.

The lower teeth display generalities similar to theuppers, but with an opposing obliquity. The lowerteeth of Eopululo have the most equally balanceddental patterns, which consist of four lophids, ofwhich the central two are of approximately equallength, and where the primary connection of theprotoconid is with the hypolophid. The hypolophidis connected to the metalophid and the hypoconid(Fig. 12). The posterior lophids of the other rodentgenera display a sweep anteriad in which the hy-polophid, uniting the entoconid with the protoco-nid, dominates the pattern. The metalophid be-comes the most variable element in the patterns. InEoincamys, the metalophid is short and much sep-arated from the hypolophid. In Eosallamys, Eob-ranisamys, and Eoespina, the metalophid appearsto be the primary crest, with the hypolophid joiningit at its midpoint between the metaconid and pro-toconid (or, said another way, the hypolophidmakes a strong deflection at its point of union withthe metalophid). This condition holds in Eoespina,even though the metalophid makes a weak connec-tion with the hypolophid.


AGE OF THE SANTA ROSAN RODENTS

Before the discovery of the Santa Rosa local fauna,the oldest extensive samples of South American ro-dents were from the combined Deseadan SALMASalla and Lacayani local faunas of Bolivia and theCabeza Blanca local fauna of Argentina (Vucetich,1991). The diversity of these faunas indicated thatrodents had been present in South America forsome time prior to Deseadan time. A single speci-men, a partial mandibular ramus from the pre-De-seadan Tinguiririca local fauna of Chile (Wyss etal., 1993), provided support for the view that ro-dents were in South America well before the De-seadan SALMA. This ramus is hystricognathous, asare all of the partial mandibles found at SantaRosa. The hystricognathous mandible is one inwhich the angular process of the mandible is notpositioned in the same vertical plane as the toothrow, but instead is positioned lateral to that plane.

The teeth of the Tinguirirican specimen are in anadvanced wear stage, but some observations arenonetheless relevant to the discussion of Santa Ro-san rodents. In the original description of the spec-imen, Wyss et al. (1993) suggested that the closestcomparison among Deseadan genera was withBranisamys of Bolivia. On that basis, the specimenwas questionably referred to the Dasyproctidae(Agoutidae, Dasyproctinae, in this paper). With thebenefit of the Santa Rosan sample, both in terms oftaxonomic diversity and abundant examples ofwear stages, we suggest that the Tinguirirican spec-imen is better placed within the Echimyidae. Thisspecimen shares echimyid features with Eosallamys(?Eocene, this paper) and Sallamys (Oligocene) inthat all lower teeth of these taxa possess a deephypoflexid, which is equal in length to the opposingflexids and which terminates at the base of the hy-polophid, and four slightly oblique lophids, ofwhich the anterior two are the first to fuse and forma single lophid. This is illustrated in Sallamys,where four lophids are found only in the earliestdental wear stage, whereas later wear stages showonly the characteristic three lophids.

Of some note is that the crown height of thecheek teeth of the Tinguirirican rodent far exceedsthat of any Santa Rosan species. Even highly worn,as they are, the cheek teeth of this rodent have asmuch crown remaining as an unworn tooth of anySanta Rosan rodent. In its degree of hypsodonty,the Tinguirirican specimen more closely matchesteeth of Sallamys, or other moderately hypsodontOligocene caviomorphs, than it does the brachy-dont teeth of the Santa Rosan rodents. As a generaltrend, hypsodonty is an apomorphic characterwithin the caviids. With only one rodent specimenfrom the Tinguiririca local fauna, conclusions aretenuous, but on the basis of the degree of fusion ofthe anterior two lophids of the lower molars andthe lesser degree of hypsodonty, we consider theSanta Rosan rodents to be ‘‘pre-Tinguirirican’’ inage.

Wyss et al. (1994:25), in discussing single-crystal40Ar/39Ar dates relevant to the age of the Tinguirir-ican fauna, suggested that it is ‘‘. . . at least asyoung as about 31.5 Ma . . . ,’’ but that it ‘‘. . . maybe as old as late Eocene (circa 37.5 Ma), and mighttherefore span the Eocene–Oligocene boundary.’’The former age was determined for deposits im-mediately above and within the upper part of thefossiliferous horizons, whereas the latter was ob-tained from immediately below the fossiliferous ho-rizons. Flynn and Wyss (1999) also placed the agebetween these two dates, indicating a pre-Desea-dan, post-Mustersan age on the basis of the occur-rence of several higher level taxa.

However, Kay et al. (1999), in recent efforts todate the Casamayoran age fauna at Gran Barrancain Chubut, Argentina, have demonstrated that theBarrancan ‘‘subage’’ of the Casamayoran SALMAis middle late Eocene, rather than late early Eoceneas previously thought. Kay et al. (1999) place theage of the Barrancan ‘‘subage’’ at 36.6–35.34 Ma,or 37.60–35.69 Ma, depending on to which sub-chron of the geomagnetic polarity time scale theplagioclase 40Ar/39Ar date is referred. This meansthat the age of the overlying Mustersan faunas atthe Gran Barranca must also be revised upward,and Kay et al. (1999) suggest that the MustersanSALMA must now straddle the Eocene–Oligoceneboundary, or range from approximately 35 to 32Ma. This period of time was derived by taking intoconsideration the ‘‘Astraponoteen plus Superieur’’local fauna of the Gran Barranca, which overliesthe Mustersan SALMA faunas and appears to becontemporaneous with the Tinguiririca local faunaof Chile (Bond et al., 1996; Flynn and Wyss, 1999),and the fossiliferous beds producing the latter,which are dated to 31.5 Ma. Thus, there wouldseem to be room for only about a 0.5 Ma time gapbetween the dated ‘‘Tinguirirican’’ age faunas ofChile and the Mustersan faunas.

The Deseadan SALMA local fauna of Salla, Bo-livia, is currently dated to 29.4–25.5 Ma (Kay etal., 1998), whereas the age of the Deseadan SAL-MA local fauna of Scarritt Pocket in Argentina iscurrently placed at 29–27 Ma (Flynn and Swisher,1995). Thus, the post-Mustersan, pre-DeseadanTinguiririca local fauna would seem to be restrictedto a possible temporal range of 32–29.4 Ma, andit is confidently tied only to the date of 31.5 Ma,or early Oligocene. The reported approximately37.5 Ma age of the horizon immediately below thefossiliferous horizon at the Tinguiririca locality(Wyss et al., 1994) appears to be unrelated to theage of the Tinguiririca local fauna.

The age of the Santa Rosa local fauna is unre-solved (Campbell, 2004). This is a consequence oftwo major factors. First, no currently availablestratigraphic data are useful in dating the fossilif-erous horizon producing the Santa Rosa local fauna(see Campbell et al., 2004). Second, most of themammalian taxa are new at the generic level, andmany are new at higher taxonomic levels (see other


papers in this volume), so it is impossible to cor-relate the Santa Rosa local fauna with other faunasof known age. We have interpreted the age of theSanta Rosan rodents to be pre-‘‘Tinguirirican’’ onthe basis of the degree of hypsodonty and the de-gree of fusion of the anterior two lophids of thelower molars, relative to the single rodent from theTinguiririca local fauna. The lack of a sizable timegap, if any at all, between ‘‘Tinguirirican’’ time andthe Mustersan SALMA leads us to the conclusionthat the Santa Rosan rodents must be at least asold as the Mustersan, and probably early Muster-san.

With their revision of the age of the Casamayor-an SALMA, Kay et al. (1999) were able to relatethe climatic changes observed in the marine sedi-mentary record of the South Atlantic with those ofcontinental paleofloras and paleomammals. Theysuggested that the mammalian faunas experienceda dramatic change in response to climatic change,and one of the ways this was expressed was via amarked increase in the number of hypsodont taxain South America after 36 Ma and before 32 Ma.They also provided data demonstrating a signifi-cant correlation between climatic variables and de-gree of hypsodonty in sigmodontine rodents. If, andwe recognize it is a big if, the early caviomorphs ofSouth America responded to climatic change in thelate Eocene in the same way as the mammalian her-bivores discussed by Kay et al. (1999), then it ispossible that the brachydont rodents of Santa Rosawould fall somewhere in the time span between 36Ma and ‘‘Tinguirirican’’ time, or before, prior to theonset of hypsodonty.

Still, we know nothing about the Paleogene florasof western Amazonia, and local conditions mayoverride general faunal tendencies. An example ofthe latter that might be pertinent is the greater hyp-sodonty of the Deseadan SALMA rodents of La-cayani, Bolivia, relative to those of the DeseadanSALMA rodents of Salla, Bolivia. Thus, we canonly offer this suggestion as a possibility; that is,that there might be a correlation between hypso-donty in early caviids and large-scale climaticchange and that the brachydont Santa Rosan ro-dents predate this period of climatic change. None-theless, we are confident that the Santa Rosan ro-dents are at least Mustersan SALMA in age andthat they are probably early Mustersan SALMA inage. They might, quite possibly, be even older. Thisconclusion is consistent with the age suggested bythe marsupials from the Santa Rosa local fauna(Goin and Candela, 2004). For the reasons given,we refer to the age of the Santa Rosa local faunaas ?Eocene.

RELATIONSHIPS OF THE SANTA ROSANRODENTS

In this paper, we describe eight genera and elevenspecies, with several additional genera and speciesindicated by fragmentary material. This diversity is

equivalent to the most diverse Deseadan localitiesknown (Vucetich, 1991). Even given the similarityobserved among the various rodent taxa from San-ta Rosa, this diversity indicates that rodents werein South America for some time before depositionof the Santa Rosa local fauna. Although an ?Eoceneage assignment does have an element of uncertaintybecause numerical age dates are lacking, we thinkit is supported by the stage of evolution of the ro-dents, in comparison with other early rodents ofSouth America. An ?Eocene age places the rodentfauna of Santa Rosa contemporaneous with or pre-dating taxa that might otherwise have been consid-ered among the ancestral groups from which camethe caviids. Furthermore, the replacement of the in-distinct time frame ‘‘pre-Deseadan’’ with the equal-ly indistinct time frame ‘‘pre-Mustersan,’’ duringwhich the development of typical caviid features ofSouth American rodents would have taken place,has a significant effect on discussions relating to theorigin of South American rodents. When one con-siders that the diversity of the known Deseadan ro-dent faunas led previous authors to propose a lateEocene radiation of rodents in South America, fol-lowing an even earlier arrival in South America(Patterson and Wood, 1982; Marshall and Sem-pere, 1993; Vucetich et al., 1999), it is reasonableto conclude that the highly diverse and significantlyolder Santa Rosan rodents push the origin of ca-viids even farther back and into the early Cenozoic.

The complex history of rodent classification re-flects the presence in rodents of a mosaic of char-acters that one might expect in lineages descendentfrom a common ancestor, as well as differing opin-ions of character preeminence by those who studythese animals. In studies of rodent relationships, theconsensus seems to be that the caviids represent anatural group (George, 1993, and references there-in; Hartenberger, 1998), but the relationship of ca-viids to rodents of other continents is less certain.This question is complicated by the anatomical sim-ilarity that is so evident among rodents of Europe,Africa, and South America. The origin of these sim-ilarities, however, whether phylogenetic or throughconvergence, is unknown.

The question of the origin of the South Americanrodents has been extensively discussed, with pro-ponents principally aligned on one of two sides.The advocates of a North American origin were ledby Wood (1981, 1985, 1993). Those who were per-suaded that Africa had been the original home ofSouth American rodents were led by Lavocat(1974, 1976, 1993). The latter viewpoint is alsoadvanced by Martin (1992, 1994, 2004), whonotes the presence of the same three subtypes ofmultiserial enamel within the Santa Rosan rodentsthat are seen in all South American caviids and theEocene/Oligocene thryonomyoids of Africa.

However, the dental features that seem to linkcaviids with African rodents are seen in the lateEocene/Oligocene rodents (Fig. 14). For example,Eoincamys and Incamys (Agoutidae) from South


Figure 14 Comparative upper molar dental patterns of some Santa Rosan genera (middle column) with Oligocene generafrom South America (left column) and late Eocene/Oligocene genera from Africa (right column). A reversed drawing isindicated by ‘‘r.’’ Sources: Incamys bolivianus (M1–2, PU 21728 (in part), Bolivia; Patterson and Wood, 1982:fig. 17B,Agoutidae). Eoincamys pascuali (LACM 143319, this paper; Agoutidae). Gaudeamus aegyptius (M1–2, YPM 18044r (inpart), Egypt, Simons and Wood, 1968:fig. 14A; Phiomyidae). Sallamys pascuali (M1, PU 20982, Bolivia, Patterson andWood, 1982:fig. 5A; Echimyidae). Eosallamys simpsoni, LACM 143424r; this paper; Echimyidae). Metaphiomys schaubi(M2, YPM 21320 (in part), Egypt, Patterson and Wood, 1982:fig. 30F; Phiomyidae). Branisamys luribayensis (M1, PU21955r (in part), Bolivia, Patterson and Wood, 1982:fig. 24B; Agoutidae). Eobranisamys romeropittmanae, LACM143358r; this paper; Agoutidae). Phiomys andrewsi (M2, YPM 18035r, Egypt, Simons and Wood, 1968:fig. 2D; Phio-myidae). Not to scale.

America resemble Gaudeamus Wood, 1968(Thryonomyidae), from the late Eocene Fayum fau-nas of Africa. In Gaudeamus, the trilophid patternresembles the pattern of upper molars in Incamysafter moderate wear has united the metaloph withthe posteroloph. Compared with Eoincamys andIncamys, the dental pattern of Gaudeamus mightbe considered the next step in character develop-ment.

Dental patterns of the Eocene/Oligocene Africangenera Metaphiomys Osborn, 1908 (Diamanto-myidae Schaub, 1958), and Phiomys Osborn, 1908(Phiomyidae), also bear a general resemblance to

those of Eobranisamys and Branisamys (Agouti-dae). However, as discussed in justification of thechoice of dental nomenclature used in this paper(see ‘‘General Descriptive Odontology’’ section), weidentify the five crests in the upper teeth of Eob-ranisamys as the anteroloph, protoloph, metaloph,neoloph, and posteroloph. The five crests of the‘‘phiomorphs’’ are the anteroloph, protoloph, me-soloph or mesolophule, metaloph, and posteroloph.

When comparing Eosallamys and Sallamys(Echimyidae) with Metaphiomys or Phiomys, it ini-tially appears that similar dental patterns wereachieved by reduction of the metaloph and enlarge-


ment of the neoloph. However, on closer inspec-tion, it seems more likely that the pattern of Sal-lamys was achieved by a reduction and redirectionof the metaloph to meet the posteroloph at its mid-point. These conditions, and a dental spur at themidpoint of the posteroloph (in the position of aneoloph), occur in specimens of Eosallamys.

In each example cited above, the closer compar-ison is between the ?Eocene and Oligocene SouthAmerican rodents and a lineal relationship is rea-sonable. Similarities seen between molar patterns ofearly caviids and contemporaneous African hystri-cognathi seem to be superficial rather than basedon homologies. At best, if one were attempting tostrengthen the argument for a relationship betweenSouth American and African rodents, it could bepostulated that the similar patterns of caviids andsome African/Asian rodents might have been inde-pendently derived from a more primitive, probablybasined, molar pattern.

In Patterson and Wood (1982), Wood states thatreal or incipient hystricognathy is present in someNorth American rodents by the late Paleocene inthe western United States or the Eocene in Mexico.His use of ‘‘incipient hystricognathy’’ as a characterwas challenged by Dawson (1977) and Korth(1984); nevertheless, hystricognathous rodentswere apparently present in North America by theEocene (Wood, 1972, 1973). Prolapsus Wood,1973, the North American genus in question, wasused to support the hypothesis of a North Ameri-can origin for the ‘‘Caviomorpha’’ (Caviida). But,hystricognathy appears to be an early feature of thecaviids, as indicated by the fact that it is a featureof the partial mandibles in the collection from San-ta Rosa, and the presence of this feature in a NorthAmerican rodent that is approximately contempo-raneous to the Santa Rosan rodents cannot be usedto support a North American origin of the caviom-orphs. It is just as possible that hystricognathy oc-curred independently or owes its presence in NorthAmerica to dispersal northward.

Leaving aside the question of hystricognathy, wenote that there are no dental patterns that wouldlink early North American rodent genera withthose of the Santa Rosa local fauna. For example,although unworn cylindrodont teeth (Cylindrodon-tidae) have a pattern that is similar to the generalstructure of Santa Rosan teeth (i.e., both have sim-ple patterns built around four crests), the symmetryof the crests in the cylindrodonts and their char-acteristic hypsodonty is unlike that of the Santa Ro-san rodents. If the caviids had a North Americanorigin, then that source would have had to haveentered South America at a very early date, leavingno record of taxa with similar dental characters inNorth America, or at least no record that has beendiscovered to date. The characteristic caviid pat-terns seen in Santa Rosan rodents would then havedeveloped in place in South America from this un-known source.

The Santa Rosan rodents do not provide enough

evidence to tip the scales toward either a NorthAmerican or African origin of the South Americanrodents. However, the timing of the dispersal event,if that were the source of South American rodents,must have occurred earlier than previously thought.We note that the teeth of the Santa Rosan rodentsare readily referable to extant caviid families on thebasis of their similarities to Oligocene representa-tives of these families. However, if, as we suspect,the Santa Rosan rodents date from the early Mus-tersan or before, then these advanced dental andmandibular features appear at a time when, for themost part, Eocene rodents elsewhere in the worldretained evidence of tritubercularity and only incip-ient hystricognathy. A much earlier placement ofthe time of origination of the South American ro-dents may reinvigorate arguments heretofore leftdormant for lack of intervening fossils such as arefound at Santa Rosa.

For example, in a discussion of South Americanzoogeography, Croizat (1971, 1979) suggested thatduring the Cretaceous certain ‘‘proto-rodents’’ inthe unified fauna of the Gondwanan continents hadprogressed toward some hystricognath characters[e.g., perhaps the subtypes of multiserial enameldiscussed by Martin (2004)]. These characters wereto become fully formed in African and South Amer-ican rodents only after the separation of those con-tinents and their faunas. In making this suggestion,Croizat placed the ‘‘caviomorphs’’ among the au-tochthonous fauna of South America, or FaunalStratum I of Simpson (1950), which includes theXenarthra, Pyrotheria, Litopterna, Notoungulata,and Marsupialia, a position strengthened some-what by the discovery of the Santa Rosa local fau-na.

An interesting variation on this theme was putforth by Landry (1957:90), but his hypothesis haslanguished for lack of fossil evidence. Landry sug-gested that the order Rodentia originated in the lateCretaceous or earliest Paleocene. Furthermore, hetentatively suggested that the hystricomorphs werethe result of the first adaptive radiation of rodents;a radiation that was later replaced throughout theworld by the sciurognaths, or Paramys and its rel-atives, during the early Eocene. In Landry’s hypoth-esis, New World and Old World hystricomorphsare relics of that first radiation.

Phylogenetic reconstruction based on DNA anal-ysis has placed the separation of the caviids as anearly event in the development of the Rodentia, soearly that they might very well be treated as taxo-nomic equivalents to the order Rodentia (Graur etal., 1991; D’Erchia et al., 1996). Although DNAevidence no longer points to a separation of thecaviids from the rest of Rodentia as initially pro-posed (Frye and Hedges, 1995), isolation of thisgroup seems certain to have occurred long beforethe time of the Santa Rosa local fauna.

Although highly controversial when proposed,the hypotheses of Landry (1957), Croizat (1971,1979), and Graur et al. (1991) present an alterna-


tive scenario for the origin of South American ca-viids into which the Santa Rosan rodents might beplaced. It is widely recognized that the isolation ofSouth America as an island continent during theTertiary strongly influenced the evolutionary devel-opment of the fauna. There are a number of majortaxonomic groups that are primarily, or wholly, re-stricted to a South American distribution duringthis period of isolation (Pascual et al., 1985), butthere is also substantial evidence for faunal ex-change between South America and North Americaand between South America and Africa in the lateMesozoic and early Cenozoic. Furthermore, in eachcase, this movement was apparently in both direc-tions, as South America contributed taxa to bothNorth America (Estes and Baez, 1985; Gingerich,1985) and Africa (George and Lavocat, 1993;Goldblatt, 1993). Viewed through the prism of theSanta Rosa local fauna, the autochthonous Tertiaryfauna of South America, including rodents, is per-haps better seen as the natural development of anisolated portion of the Gondwanaland fauna ratherthan comprising just the archaic remnants of morewidely distributed, and more rapidly evolving, Oldand New World faunas.

The stage of evolution of the rodents from theSanta Rosa local fauna indicates that they are olderthan the rodents from the Deseadan SALMA of Sal-la, Bolivia, and the single specimen from Tinguirir-ica, Chile. Inasmuch as the caviids are generallyconsidered to be conservative in the rate of evolu-tionary change of their tooth morphology, thesmall, but significant, differences seen between taxaof the Salla local fauna and Santa Rosa local faunasuggest that the Santa Rosa taxa may be consider-ably older than those from Salla, in spite of firstimpressions based on similar features. The diversityat Santa Rosa is not as great as that seen amongall known Deseadan rodent faunas in South Amer-ica taken together, but it is greater than that of thenext youngest rodent faunas. We still have only asmall sample from Santa Rosa, and intensivescreen-washing could produce many thousands ofspecimens and probably many additional taxa.Nonetheless, the present known diversity at SantaRosa is sufficient to indicate that rodents were inSouth America for some time before the accumu-lation of those deposits. On the basis of the evi-dence presented above, we are inclined to acceptthe hypothesis that the caviomorph rodents ofSouth America represent an autochthonous evolu-tionary lineage derived from an ancestral groupthat formed part of a greater Gondwanaland fauna.

SUMMARY

The rodents of the Santa Rosa local fauna describedin this paper constitute the oldest known rodentsfrom South America. They are dated to the ?Eocene(Mustersan SALMA) on the basis of their moreprimitive form compared with the post-Mustersan,pre-Deseadan rodent from the Tinguiririca local

fauna of Chile. This age is also consistent with thestage of evolution of other mammals in the SantaRosa local fauna, but there are as yet no other con-firming age data. There is great taxonomic diversityin this fauna, with at least eight genera and perhapsas many as seventeen species present, eleven ofwhich are described in this paper. These species areplaced within three well-known families of SouthAmerican rodents—the Erethizontidae, Agoutidae,and Echimyidae—because of identifiable evolution-ary trends already underway by Santa Rosan time.Nonetheless, a horizontal classification on the basisonly of the features of the teeth of the Santa Rosanrodents could easily justify inclusion of the SantaRosan rodents within one family. Most of the spec-imens consist of isolated teeth that range in sizefrom approximately 1 mm to just under 4 mm inwidth.

The teeth of all of the Santa Rosan rodents arefundamentally similar. They are quadrangular inoutline, with rounded corners, and brachydont, andthey have low crests among only four recognizablecusps. With the exception of Eobranisamys, theteeth, both uppers and lowers, are primarily tetra-lophodont. Still, a developing fifth loph, a neoloph,is present as a variable feature on many teeth in theSanta Rosa collection. Usually it consists of minorcrenulations in the anterior surface of the postero-loph or a series of aligned enamel irregularitieswithin the metaflexus. It might also form from asubstantial spur that extends into the metaflexus,or two such spurs that approach each other to forma complete neoloph. Such crenulations and enamelirregularities are present in variable form in severalSanta Rosan taxa (e.g., Eoincamys, Agoutidae; andEosallamys and Eoespina, Echimyidae). The fifthloph on the upper teeth of Eobranisamys (Agouti-dae) is consistently present, but variable, and it isthe best formed fifth loph of those found in anygenus within the local fauna. In the Oligocene er-ethizontid Protosteiromys (P. medianus in Woodand Patterson, 1959:fig. 30), a complete neoloph(our usage) is present, evidently marking the sta-bilization in the erethizontids, as well as in someagoutids, of what is a variable feature in most ro-dents of Santa Rosan age. The overall similaritiesin crown height and shape and the arrangement ofthe lophs suggest that these rodents were not farremoved from their common ancestry, even thoughthey were highly advanced for Eocene rodents.

Two partial mandibles referred to Eobranisamysshow evidence of advanced caviomorph features.The rami are fully hystricognathous, and the lowerdeciduous premolars were still in place and in useas part of a complete lower tooth row that includedM3. Hystricognathy has been found in an early Ol-igocene rodent of Chile (Wyss et al., 1993) and pos-sibly in Eocene rodents of North America (Wood,1972), and this trait seems to have been an earlycharacteristic among South American rodents. Hys-tricognathy might be significant in understandingthe history of rodents in North America, but it can-


not be used as evidence that South American ro-dents are derived from North America because thegroups on the two continents were roughly contem-poraneous.

Furthermore, the two partial mandibles of Eo-branisamys have dp4s in place, with moderate towell worn molars. There is no indication of un-erupted P4s beneath the dp4s, and only deciduousfourth premolars are referred to the genus Eobran-isamys. The retention of deciduous premolars wascited as a character in common between Caviidaand the Old World ‘‘Phiomorpha’’ by Hoffstetter(1975:520, in Patterson and Wood, 1982:498).That view is strengthened by the evidence fromthese two rami.

The Santa Rosan rodents, primitive as they arewhen compared with later caviids, are nonethelesshighly advanced for Eocene rodents, and a clearprogression from the taxa of Santa Rosa toward theDeseadan SALMA taxa of Salla, Bolivia, and Pa-tagonia, Argentina, can be identified. The occlusalpatterns of both upper and lower teeth show nolingering evidence of tritubercularity in upper teeth,or basinal expansion of the talonid at the expenseof the trigonid and marginal placement of the cus-pids in lower teeth. The fundamental dental patternof rodents from Santa Rosa is one in which thereare four large cusps in both upper and lower teethand in which the labial cusps (in the uppers) orlingual cuspids (in the lowers) are more closelyspaced to the labiolingual axis of the tooth. Eachmolar has four low crests, with a principal crestplaced diagonally on the occlusal surface and be-tween the paracone and hypocone (upper teeth), orthe entoconid and protoconid (lower teeth). BySanta Rosan time, the basic caviid patterns were setfor the families Agoutidae, Echimyidae, and per-haps Erethizontidae. If an Eocene age for the SantaRosa local fauna is borne out, then the rodentswould appear to be much farther along in their evo-lutionary direction than Eocene rodents in otherparts of the world, and it would seem fruitless toseek their ancestry in either North America or Af-rica, as was the practice when no rodents olderthan the Oligocene were known from South Amer-ica.

The early age of these rodents, and their well-defined and characteristic caviid features, is suffi-cient cause to resurrect hypotheses of caviid diver-gence that place this group in isolation in the earlyCenozoic. This might have occurred as the result ofvicariance brought about by the breakup of Gond-wana. The obvious and often remarked upon sim-ilarities between South American and African hys-tricognaths would therefore be the manifestationsof a once common gene pool. However, that thecaviids are related to North American paramyidscannot be ruled out entirely. Certainly, instances offaunal and floral exchange between South America,North America, and Africa in the early Tertiary aregrowing more evident, and the precursors of thecaviids could have come from any direction. None-

theless, the Santa Rosan rodents suggest that theywere diverging from a common center, had not pro-ceeded far in their evolutionary development, andwere diverging in ways characteristic of later ca-viids without displaying similarities to Eocene gen-era in North America or Africa. From this, we con-clude that the caviids constitute a long and separatelineage within Rodentia and that they should berecognized as a component of the autochthonousfauna of South America.

ACKNOWLEDGMENTS

The authors gratefully acknowledge the field companion-ship and financial support of John G. Wigmore, codiscov-erer of the Santa Rosa local fauna. Additional funding forfield efforts was provided by the National Geographic So-ciety. Tim Barnhart and Bret Gustafson, Johnson CountyCommunity College, assisted with the preparation of theillustrations. Larry Martin, Morton Green, and Robert W.Wilson, University of Kansas Museum of Natural History,read drafts of the manuscript and provided notes andcomments. Spirited reviews of the manuscript were of-fered by Malcolm McKenna and Guiomar Vucetich, forwhich we are grateful. We thank Francisco Goin for thetranslation of the abstract into Spanish. The X-ray pho-tographs were made by John Chorn, University of KansasMuseum of Natural History.

LITERATURE CITED

Ameghino, F. 1887. Enumeracion sistematica de las es-pecies de mamıferos fosiles coleccionados por CarlosAmeghino en los terrenos eocenos de Patagonia aus-tral y depositados en el Museo La Plata. Boletın delMuseo La Plata (Buenos Aires) 1:1–26 (del aparta-do).

Bond, M., G. Lopez, and M. Reguero. 1996. ‘‘Astrapon-oteen plus superieur’’ of Ameghino: Another intervalin the Paleogene record of South America. Journalof Vertebrate Paleontology 16(3)(Suppl.):23A.

Bryant, J.D., and M.C. McKenna. 1995. Cranial anatomyand phylogenetic position of Tsanganomys altaicus(Mammalia: Rodentia) from the Hsanda Gol For-mation (Oligocene), Mongolia. American Museumof Natural History Novitates 3156:1–42.

Butler, P. 1985. Homologies of molar cusps and crests, andtheir bearing on assessments of rodent phylogeny. InEvolutionary relationships among rodents. A multi-disciplinary analysis, ed. W.P. Luckett and J.-L. Har-tenberger, 381–401. New York: Plenum Press.

Campbell, K.E., Jr. 2004. The Santa Rosa local fauna: Asummary. In The Paleogene mammalian fauna ofSanta Rosa, Amazonian Peru, ed. K.E. Campbell, Jr.Natural History Museum of Los Angeles County,Science Series 40, 155–163.


Candela, A. 1999. The evolution of the molar pattern ofthe Erethizontidae (Rodentia, Hystricognathi) andthe validity of Parasteiromys Ameghino, 1904. Pa-laeovertebrata 28(1):53–73.

. 2002. Lower deciduous tooth homologies in Er-


ethizontidae (Rodentia, Hystricognathi): Evolution-ary significance. Acta Palaeontologia Polonica 47(4):717–723.

Croizat, L. 1971. De la ‘‘pseudovicariance’’ et de la ‘‘dis-jonction illusoire.’’ Anuario da Sociedade Broteriana37:115–141.

. 1979. Review of Biogeographie: Fauna und Florader Erde und ihre Geschichtliche Entwicklung by P.Banarescu and N. Boscaiu. Systematic Zoology28(2):250–252.

Dawson, M.R. 1977. Late Eocene rodent radiations:North America, Europe, and Asia. Geobios, SpecialMemoir 1:195–209.

D’Erchia, A.M., C. Gissi, G. Pesole, C. Saccone, and V.Arnason. 1996. The guinea-pig is not a rodent. Na-ture 381:597–599.

Estes, R., and A. Baez. 1985. Herpetofaunas of North andSouth America during the Late Cretaceous and Ce-nozoic: Evidence for interchange? In The greatAmerican biotic interchange, ed. F.G. Stehli and S.D.Webb, 140–197. New York: Plenum.

Ferrusquia-Villafranca, I. 1989. A new rodent genus fromcentral Mexico and its bearing on the origin of theCaviomorpha. In Papers on fossil rodents in honorof Albert Elmer Wood, ed. C.C. Black and M.R.Dawson. Natural History Museum of Los AngelesCounty, Science Series 33, 91–117.

Ferrusquia-Villafranca, I., and A.E. Wood. 1969. Newfossil rodents from the early Oligocene Rancho Gai-tan local fauna, northeastern Chihuahua, Mexico.The Pearce-Sellards Series, Texas Memorial Museum16:1–13.

Flynn, J.J., and C.C. Swisher. 1995. Cenozoic SouthAmerican Land Mammal Ages: Correlation to globalgeochronologies. Society of Economic Petrologistsand Mineralogists, Special Publication 54:317–333.

Flynn, J.J., and A.R. Wyss. 1999. New marsupials fromthe Eocene–Oligocene transition of the Andean mainrange, Chile. Journal of Vertebrate Paleontology19(3):533–549.

Flynn, L., L.L. Jacobs, and I.U. Cheema. 1986. Baluchi-myinae, a new ctenodactyloid rodent subfamily fromthe Miocene of Baluchistan. American MuseumNovitates 2841:1–58.

Frye, M.S., and S.B. Hedges. 1995. Monophyly of the Or-der Rodentia inferred from mitochondrial DNA se-quences of the genes for 12S rRNA, 16s rRNA, andtRNA-valine. Molecular Biology and Evolution12(1):168–176.

George, W. 1993. The strange rodents of Africa and SouthAmerica. In The Africa–South America connection,ed. W. George and R. Lavocat. Oxford Monographson Biogeography, 7:119–141. New York: OxfordUniversity Press.

George, W., and R. Lavocat (eds.). 1993. The Africa–South America connection. Oxford Monographs onBiogeography, no. 7. New York: Oxford UniversityPress, 166 pp.

Gingerich, P.D. 1985. South American Mammals in thePaleocene of North America. In The Great AmericanBiotic Interchange, ed. F.G. Stehli and S.D. Webb,123–137. New York: Plenum Press.

Goin, F., and A. Candela. 2004. New Paleogene marsu-pials from the Amazon Basin of eastern Peru. In ThePaleogene mammalian fauna of Santa Rosa, Ama-zonian Peru, ed. K.E. Campbell, Jr. Natural HistoryMuseum of Los Angeles County, Science Series 40,15–60.

Goldblatt, P. (ed.). 1993. Biological relationships betweenAfrica and South America. New Haven: Yale Uni-versity Press, 630 pp.

Graur, D., W.A. Hide, and W.H. Li. 1991. Is the guinea-pig a rodent? Nature 351:649–652.

Hartenberger, J.-L. 1998. Description of the radiation ofthe Rodentia (Mammalia) from the late Paleocene tothe Miocene—Phylogenetic consequences. ComptesRendus de l’Academie des Sciences Serie II, (A)326(6):439–444.

Hoffstetter, R. 1975. El origen de los Caviomorpha y elproblema de los Hystricognathi (Rodentia). I Con-greso Argentino de paleontologia y bioestratigrafia,Actas 1(2):505–528.

Hoffstetter, R., and R. Lavocat. 1970. Decouverte dans leDeseadien de Bolivie de genres pentalophodontes ap-puyant les affinites africaines des rongeurs caviom-orphes. Comptes Rendus de l’Academie des Sciences,(D) 271:172–175.

Kay, R.F., B.J. MacFadden, R.H. Madden, H. Sandeman,and F. Anaya. 1998. Revised age of the Salla beds,Bolivia, and its bearing on the age of the DeseadanSouth American Land Mammal ‘‘Age.’’ Journal ofVertebrate Paleontology 18(1):189–199.


Korth, W.W. 1984. Earliest Tertiary evolution and radia-tion of rodents in North America. Bulletin of theCarnegie Museum of Natural History 24:1–71.

Landry, S.O. 1957. The interrelationships of the New andOld World hystricomorph rodents. University ofCalifornia Publications in Zoology 56:1–118.

Lavocat, R. 1974. The interrelationships between the Af-rican and South American rodents and their bearingon the problem of the origin of South Americanmonkeys. Journal of Human Evolution 3:523–526.

. 1976. Rongeurs caviomorphes de l’Oligocene deBolivie. II. Rongeurs du bassin Deseadien de Salla-Luribay. Palaeovertebrata 7(3):15–90.

. 1981. The implications of rodent paleontologyand biogeography to the geographical sources andorigin of platyrrhine primates. In Evolutionary bi-ology of the New World monkeys and continentaldrift, ed. R.L. Ciochon and A.B. Chiarelli, 93–102.New York: Plenum Press.

. 1993. Conclusions. In The Africa–South Americaconnection, ed. W. George and R. Lavocat. OxfordMonographs on Biogeography, 7:142–150. NewYork: Oxford University Press.

Leidy, J. 1871. Notice of some extinct rodents. Proceed-ings of the Academy of Natural Science, Philadelphia1871:230–232.

Marshall, L., and T. Sempere. 1993. Evolution of the Neo-tropical Cenozoic land mammal fauna in its geo-chronologic, stratigraphic, and tectonic context. InBiological relationships between Africa and SouthAmerica, ed. P. Goldblatt, 328–392. New Haven,Connecticut, and London: Yale University Press.

Martin, T. 1992. Schmelzmikrostruktur in den Inzisivenalt- und neuweltlicher hystricognather Nagetiere. Pa-laeovertebrata, Memoire extraordinaire, pp. 1–168.

. 1994. African origin of caviomorph rodents is in-


dicated by incisor enamel microstructure. Paleobi-ology 20:5–13.

. 2004. Incisor enamel microstructure of SouthAmerica’s earliest rodents: Implications for caviom-orph origin and diversification. In The Paleogenemammalian fauna of Santa Rosa, Amazonian Peru,ed. K.E. Campbell, Jr. Natural History Museum ofLos Angeles County, Science Series 40, 131–140.

McKenna, M.C., and S.K. Bell. 1997. Classification ofmammals above the species level. New York: Colum-bia University Press, 629 pp.

Osborn, H.F. 1908. New fossil mammals from the FayumOligocene, Egypt. American Museum of NaturalHistory, Bulletin 24:265–272.

Pascual, R., M.G. Vucetich, G.J. Scillato-Yane, and M.Bond. 1985. Main pathways of mammalian diver-sification in South America. In The great Americanbiotic interchange, ed. F.G. Stehli and S.D. Webb,219–247. New York: Plenum.

Patterson, B., and R. Pascual. 1968. New echimyid ro-dents from the Oligocene of Patagonia, and a syn-opsis of the family. Museum of Comparative Zool-ogy, Harvard University, Breviora 301:1–14.

Patterson, B., and A.E. Wood. 1982. Rodents from theDeseadan Oligocene of Bolivia and the relationshipsof the Caviomorpha. Bulletin of the Museum ofComparative Zoology 149(7):371–543.

Schaub, S. 1958. Simplicidentata. In: Traite de Paleonto-logie, vol. 6, no. 2, ed. J. Piveteau, 659–818. Paris:Masson et Cie.

Scott, W.B. 1895. Protoptychus hatcheri, a new rodentfrom the Uinta Eocene. Proceedings of the Academyof Natural Sciences, Philadelphia 1895:269–286.

Simons, E., and A.E. Wood. 1968. Early Cenozoic mam-malian faunas, Fayum Province, Egypt. PeabodyMuseum of Natural History, Yale University, Bulle-tin 28:1–105.

Simpson, G.G. 1950. History of the fauna of Latin Amer-ica. American Scientist 38(3):369–408.

Tullberg, T. 1899. Ueber das System der Nagethiere: einephylogenetische Studie. Nova acta regiae Societatisscientiarum upsaliensis 18(3):1–514.

Vucetich, M.G. 1991. Los roedores de Salla y Lacayani(Bolivia) y su correlacion con las otras faunas deedad Deseadense (Oligoceno). In Fosiles y Facies deBolivia, vol. 1, Vertebrados, ed. R. Suarez-Soruco.Santa Cruz, Bolivia: Revista Tecnica de YPFB, 12(3–4):625–629.

Vucetich, M.G., and A.G. Kramarz. 2003. New Miocenerodents from Patagonia (Argentina) and their bear-ing on the early radiation of the octodontoids (Hys-tricognathi). Journal of Vertebrate Paleontology23(2):435–444.

Vucetich, M.G., and D.H. Verzi. 1990. La pentalofodon-cia entre los roedores Caviomorpha: Simplesiomor-fıa o homoplasia? In Reunion conjunta de la Socie-dad Argentina para el Estudio de los Mamıferos

(SAREM) y la American Society of Mammalogists(ASM), Buenos Aires, 42.

. 1996. A peculiar octodontoid (Rodentia, Caviom-orpha) with terraced molars from the Lower Mio-cene of Patagonia (Argentina). Journal of VertebratePaleontology 16(2):297–302.

Vucetich, M.G., D.H. Verzi, and J.-L. Hartenberger. 1999.Review and analysis of the radiation of the SouthAmerican Hystricognathi (Mammalia, Rodentia).Comptes Rendu de l’ Academie des Sciences, Scienc-es de la terre et des planetes 329:763–769.

Wood, A.E. 1949. A new Oligocene rodent from Patagon-ia. American Museum of Natural History, Novitates1435:1–54.

. 1968. The African Oligocene Rodentia. Part II of:Early Cenozoic mammalian faunas, Fayum Province,Egypt. Peabody Museum of Natural History, YaleUniversity, Bulletin 28:23–105.

. 1972. An Eocene hystricognathous rodent fromTexas: Its significance in interpretations of continen-tal drift. Science 175:1250–1251.

. 1973. Eocene rodents, Pruett Formation, South-west Texas; their pertinence to the origin of theSouth American Caviomorpha. Texas Memorial Mu-seum, the Pierce-Sellards Series 20:1–40.

. 1981. The origin of the caviomorph rodents froma source in middle America: A clue to the area oforigin of the platyrrhine primates. In Evolutionarybiology of the New World monkeys and continentaldrift, ed. R.L. Ciochon and A.B. Chiarelli, 79–91.New York: Plenum Press.

. 1985. Northern waif primates and rodents. In TheGreat American Biotic Interchange, eds. F.G. Stehliand S.D. Webb, 267–282. New York: Plenum Press.

. 1993. The history of the problem. In The Africa–South America connection, no. 7, ed. W. George andR. Lavocat. Oxford Monographs on Biogeography,1–7. New York: Oxford University Press.

Wood, A.E., and B. Patterson. 1959. The rodents of theDeseadan Oligocene of Patagonia and the beginningsof South American rodent evolution. Bulletin of theMuseum of Comparative Zoology 120(3):279–428.

Wood, A.E., and R.W. Wilson. 1936. A suggested nomen-clature for the cusps of the cheek teeth of rodents.Journal of Paleontology 10(5):388–391.

Wyss, A.R., J.J. Flynn, M.A. Norell, C.C. Swisher III, R.Charrier, M.J. Novacek, and M.C. McKenna. 1993.South America’s earliest rodent and recognition of anew interval of mammalian evolution. Nature 365:463–437.

Wyss, A.R., J.J. Flynn, M.A. Norell, C.C. Swisher III, M.J.Novacek, M.C. McKenna, and R. Charrier. 1994.Paleogene mammals from the Andes of central Chile:A preliminary taxonomic, biostratigraphic, and geo-chronologic assessment. American Museum of Nat-ural History, Novitates 3098:1–31.



APPENDICES

Appendix 1 Crown measurements of Santa Rosan rodentteeth. Measurements of length by width are taken fromgreatest anteroposterior lengths and greatest labiolingualwidths as measured on the occlusal surface. The holotypeof a species is indicated by one asterisk (*); two asterisks(**) indicate the holotype of a genus. Measurements inparentheses are estimates made from damaged specimens.Specimens with pound sign (#) were too broken for mea-surements, but appear in Appendix 2.

Length 3 Width (mm)

Eopululo wigmorei

143269**143272149489

Mx

Mx

Mx

3.63 3 3.052.97 3 2.843.28 3 3.41

143263#

143268#

143410149465

Mx

Mx

Mx

Mx

(4.2) 3 (3.3)(3.7) 3 (3.0)

Eoincamys pascuali

143313143321

P4

P4

1.49 3 1.791.73 3 1.92

143322 P4 1.65 3 1.79143312143314143316143317143318**

Mx

Mx

Mx

Mx

Mx

2.06 3 2.441.78 3 1.941.82 3 2.172.31 3 2.492.18 3 2.11

143319143323143324149463149478

Mx

Mx

M3

Mx

Mx

1.81 3 2.091.77 3 2.171.53 3 1.701.90 3 2.221.94 3 2.23

143328143331143335143456

dp4

dp4

dp4

dp4

(1.6) 3 1.68— 3 1.61

1.85 3 1.58(1.6) 3 1.57

143301143333143334143336

P4

P4

P4

P4

1.84 3 1.661.72 3 1.411.73 3 1.701.73 3 1.60

143299143300143303143304143305

Mx

Mx

MI

Mx

Mx

1.99 3 1.782.10 3 1.781.90 3 1.502.06 3 1.721.81 3 1.61

143306143307143308143309143310

Mx

Mx

Mx

Mx

Mx

2.12 3 1.892.05 3 1.882.16 3 1.552.05 3 1.561.93 3 1.70

143427143434143504149445149446149449149452149453

Mx

Mx

Mx

M3

Mx

M3

Mx

Mx

— 3 (1.8)1.61 3 1.501.93 3 1.722.03 3 1.651.90 3 1.822.12 3 1.822.07 3 1.822.06 3 1.77

Appendix 1 Continued.

Length 3 Width (mm)

Eoincamys ameghinoi

149430149473

P4

P4

1.26 3 1.631.10 3 (1.2)

143279143315143320143338143339*

Mx

M3

Mx

Mx

Mx

1.72 3 1.801.43 3 1.481.47 3 1.531.73 3 1.691.65 3 1.67

143340143396143437143468149450149470

M3

Mx

Mx

Mx

Mx

Mx

1.61 3 1.76

1.40 3 (1.4)1.69 3 1.261.64 3 1.671.66 3 1.74

149495 P4

M1

M2

1.01 3 1.121.23 3 1.311.66 3 1.50

143332143445

P4

P4

1.55 3 1.451.32 3 1.38

143280143435143507149435

Mx

Mx

Mx

Mx

1.44 3 1.261.49 3 1.551.33 3 1.231.98 3 1.56

149442149444

Mx

Mx

1.87 3 1.451.51 3 1.25

Eobranisamys romeropittmanae

143354143355143356143357143358

M3

Mx

Mx

Mx

Mx

3.14 3 3.042.55 3 —3.05 3 2.973.01 3 (3.0)3.22 3 3.07

143359143360143362143364143365

Mx

Mx

Mx

M3

M3

3.15 3 3.153.29 3 (3.1)3.21 3 3.072.94 3 3.042.75 3 3.10

143381149482149490149493

M3

Mx

M3

M3

(3.0) 3 (3.1)2.82 3 2.783.16 3 3.353.38 3 2.91

144296 M2

M3

3.43 3 3.803.31 3 3.40

144297** M1 3.25 3 3.47M2 3.31 3 3.98

143352143419143423143505

dp4

dp4

dp4

dp4

3.67 3 2.413.29 3 2.19

— 3 2.35

144293 dp4

M1

M2

3.55 3 2.283.19 3 3.103.48 3 3.60

143266143341143342143343143344

Mx

Mx

Mx

Mx

Mx

3.33 3 2.953.11 3 3.033.70 3 3.393.23 3 3.153.15 3 2.92



Length 3 Width (mm)

143345143346143347144292149462

Mx

Mx

M3

M1

Mx

3.18 3 3.013.28 3 3.143.55 3 2.91(2.2) 3 2.063.28 3 —

149483149487

Mx

M1

3.23 3 3.203.21 3 3.07

Eobranisamys riverai

143302143372143373143377143378

M3

Mx

Mx

Mx

M2

2.61 3 2.642.49 3 3.042.15 3 2.532.62 3 2.802.65 3 2.94

143428143503*149464149474

M3

M1

Mx

Mx

2.04 3 2.272.62 3 2.632.68 3 (2.4)2.26 3 2.59

143368143369143270143371143506

dp4

dp4

dp4

dp4

dp4

— 3 (1.8)2.90 3 2.03

— 3 1.952.58 3 1.852.38 3 1.74

143366143367149456149461149475

Mx

Mx

Mx

Mx

Mx

2.84 3 2.542.62 3 2.423.01 3 2.393.39 3 (2.5)2.78 3 2.72

149476149485

Mx

Mx

2.98 3 2.682.81 3 2.49

Eopicure kraglievichi

143264**143353149454149455149466149479

Mx

Mx

Mx

Mx

Mx

Mx

2.75 3 2.302.78 3 2.292.49 3 2.132.96 3 2.122.63 3 2.012.54 3 2.48

Agoutidae, genus and species indeterminate A144299 Mx 4.24 3 4.08

Agoutidae, genus and species indeterminate B143375 Mx 2.96 3 2.82

Agoutidae, genus and species indeterminate C143349143351

Mx

Mx

3.84 3 2.923.64 3 2.99

Eosallamys paulacoutoi

143415 P4 2.10 3 2.25149460149480

P4

P4

2.16 3 2.342.35 3 2.58

143267143363144302149494

Mx

M3

Mx

Mx

2.66 3 2.732.86 3 3.052.50 3 2.432.56 3 2.65

143422** P4

M1

M2

2.43 3 2.882.70 3 2.993.12 3 3.47

M3 2.78 3 3.33


Length 3 Width (mm)

143370#

143418143420143450143451149458

dp4

dp4

dp4

dp4

dp4

dp4

— 3 2.00— 3 2.16

2.80 3 1.752.90 3 1.852.94 3 2.22(3.2) 3 (2.0)

149492143262

P4

M3

3.20 3 2.653.03 3 2.49

143265143361144524

Mx

Mx

Mx

3.08 3 —2.59 3 —2.70 3 2.52

Eosallamys simpsoni

149451 P4 1.91 3 1.67143374143376143380143398143411

Mx

Mx

M3

Mx

Mx

1.98 3 2.162.35 3 2.612.20 3 2.172.05 3 2.062.34 3 2.32

143424*143429143430143441149448149457

Mx

Mx

Mx

Mx

M3

M3

1.88 3 2.042.09 3 2.302.04 3 2.241.94 3 —1.89 3 1.762.06 3 1.95

149459149472149477149481149488

Mx

Mx

Mx

Mx

Mx

2.36 3 2.302.33 3 2.482.31 3 2.022.20 3 2.11(2.8) 3 (2.7)

143406143407#

143452149443#

149467#

dp4

dp4

dp4

dp4

dp4

— 3 1.48

2.20 3 1.47

143311 P4 2.32 3 1.94143273143275143276143284143412

Mx

Mx

Mx

M3

Mx

2.28 3 2.052.28 3 2.282.24 3 2.041.67 3 1.571.93 3 1.83

143413143414143416#

149491

Mx

Mx

Mx

Mx

1.86 3 1.772.26 3 2.38

2.76 3 2.48

Eoespina woodi

143432143459144301

P4

P4

P4

1.06 3 1.201.08 3 1.421.01 3 1.33

149468 P4 1.28 3 1.24143286**143288143289143290143291

Mx

M3

Mx

Mx

Mx

1.33 3 1.450.99 3 1.311.44 3 —1.42 3 1.551.36 3 1.56

143293143295

Mx

Mx

1.36 3 1.541.31 3 (1.2)



Length 3 Width (mm)

143296143297143298

Mx

M3

M3

1.41 3 1.501.12 3 1.301.30 3 1.37

143384143397143455143457144300

Mx

Mx

Mx

Mx

Mx

1.29 3 1.37— 3 1.50

1.32 3 1.431.36 3 1.731.39 3 1.67

149427149431149432149433149436149437

Mx

Mx

Mx

Mx

Mx

Mx

1.30 3 1.261.17 3 1.331.51 3 1.501.51 3 1.351.49 3 1.581.49 3 1.33

149441149471149484

M2

Mx

Mx

1.55 3 1.571.48 3 1.291.53 3 1.62

143391143403143404#

143405143448#

149440

dp4

dp4

dp4

dp4

dp4

dp4

1.48 3 0.931.33 3 0.92

1.32 3 0.96

1.50 3 1.02143283 P4 1.33 3 1.21144294 P4

M1

M2

1.28 3 1.301.43 3 1.421.48 3 1.51

143281143282143327143330143385143386

Mx

M3

Mx

Mx

Mx

M3

1.24 3 1.151.26 3 1.171.27 3 1.241.41 3 1.371.43 3 1.531.24 3 1.15

143389143390143436143438143439

Mx

Mx

Mx

M3

Mx

1.49 3 1.391.46 3 1.401.30 3 1.281.34 3 1.22(1.2) 3 (1.2)

143440143446143447143461149425

M3

Mx

Mx

M3

M1

1.12 3 1.151.53 3 1.421.04 3 1.121.29 3 1.15(1.5) 3 (1.4)

149429149434149447149469149486

Mx

M1

Mx

M1

Mx

1.55 3 1.391.24 3 1.401.62 3 2.141.41 3 1.381.21 3 1.30

Eosachacui lavocati

143433143458143466

P4

P4

P4

1.24 3 (1.3)1.27 3 1.461.03 3 (1.2)

143278143285

Mx

Mx

1.45 3 1.721.48 3 (1.6)

143292143294**

Mx

Mx

1.39 3 1.601.52 3 1.74


Length 3 Width (mm)

143337 Mx 1.36 3 1.54143383143387

Mx

Mx

1.55 3 1.60(1.6) 3 (1.7)

143388143399143401143444143453

Mx

Mx

M3

Mx

Mx

1.34 3 1.481.46 3 1.661.25 3 (1.3)1.37 3 1.411.42 3 1.62

143454149438149439

Mx

Mx

Mx

(1.6) 3 1.631.55 3 1.451.58 3 1.52

143326143379143400143402143449

dp4

dp4

dp4

dp4

dp4

— 3 1.381.60 3 1.031.53 3 1.08(1.5) 3 0.98

— 3 1.15143394143274

P4

M1

1.47 3 1.301.54 3 1.39

143287143325143329143382143393

Mx

Mx

M1

Mx

Mx

1.43 3 1.341.51 3 1.401.43 3 1.471.70 3 1.651.62 3 1.38

143395143408143417143421143431

M3

Mx

Mx

Mx

Mx

1.39 3 1.311.56 3 1.501.69 3 1.621.50 3 1.331.64 3 2.04

143442143443149426149428149447

Mx

Mx

M1

Mx

Mx

1.62 3 (1.6)(1.6) 3 —1.54 3 1.551.56 3 1.401.62 3 2.14

Eodelphomys almeidacomposi

144298** M3 3.81 3 3.38

Echimyidae, genus and species indeterminate A

143409 Mx 3.69 3 —

Echimyidae, genus and species indeterminate B

143348 Mx 3.98 3 3.00

Echimyidae, genus and species indeterminate C

143277 Mx 2.47 3 (2.4)

Hystricognatha, genus and species indeterminate

143271#

143350143392143425143426

Mx

Mx

Mx

1.69 3 1.99

143460143462143463143464143465143467


Appendix 2 Photographic catalogue of the Santa Rosan rodents. Upper teeth are followed by lower teeth.Order of species appearance follows that in text. For ease in identification of a tooth, see numerical sequenceunder taxon in Appendix 1.


Incisor Enamel Microstructure of South America’s Earliest Rodents:Implications for Caviomorph Origin and Diversification

Thomas Martin1

ABSTRACT. Six isolated rodent incisors from the Paleogene Santa Rosa locality ofeastern Peru exhibit a double-layered schmelzmuster with multiserial Hunter–Schregerbands (HSB) in the portio interna (PI), which is typical for caviomorph rodents. Allthree subtypes of multiserial HSB [parallel to acute angular interprismatic matrix (IPM);acute angular IPM; and rectangular, platelike IPM] observed in Oligocene and youngercaviomorphs are present. This variety of schmelzmuster reflects diversity in the ?EoceneSanta Rosa rodent fauna. Two incisors with rectangular, plate-like IPM in the PI indicatethe presence of members of Octodontoidea, a caviomorph superfamily characterized bythis autapomorphic character. The high diversity of incisor enamel microstructure canbe explained by postulating several waves of immigration from Africa, where the samesubtypes of HSB occur in Eocene–Oligocene Thryonomyoidea, or a longer evolutionaryhistory of South American caviomorphs.

RESUMEN. Seis incisivos aislados de roedores procedentes de la localidad paleogenade Santa Rosa, en el este del Peru, exhiben un schmelzmuster de dos capas, con bandasde Hunter–Schreger (HSB) multiseriales en la porcion interna (PI), un rasgo tıpico delos rodores caviomorfos. Se aprecian en ellos los tres subtipos de HSB multiserial ob-servados en los caviomorfos oligocenos y mas jovenes: matriz interprismatica (IPM)orientada desde paralela hasta en angulo agudo, IPM en angulo agudo y, finalmente,IPM rectangular, en forma de lamina o placa. Esta variedad de schmelzmuster reflejadiversidad en la fauna ¿eocena? de roedores de Santa Rosa. Dos incisivos con IPMrectangular en forma de lamina o placa indican la presencia de miembros de los Octo-dontoidea, una superfamilia de caviomorfos caracterizada por este rasgo autapomorfico.La alta diversidad en la microestructura del esmalte de estos incisivos puede ser expli-cada ya sea postulando varias oleadas inmigratorias a partir de Africa, en donde losmismos subtipos de HSB aparecen en los Thryonomyoidea del Eoceno–Oligoceno, obien a partir de una mas larga historia evolutiva de los caviomorfos sudamericanos.

INTRODUCTION

Incisor enamel microstructure is an important char-acter for phylogeny and systematics in Rodentia(Korvenkontio, 1934; Koenigswald, 1980; Martin,1993, 1997). The enlarged and ever-growing dI2/dI2 (Luckett, 1993) of rodents are highly modifiedteeth adapted for high stresses, which is reflectedby a very complex enamel microstructure (Fig. 1)that is unparalleled among mammals. Mammaliantooth enamel consists of needlelike hydroxyapatitecrystallites that are a few micrometers long. In mostnonholotherian Mesozoic mammals, the hydroxy-apatite crystallites are oriented perpendicular to theenamel–dentine junction (EDJ) and are arrangedmore or less parallel to each other (Sander, 1997).In Holotheria, first known in Late Jurassic Dry-olestidae (Martin, 1999b), the crystallites are bun-dled into fiberlike elements of 3–5 mm diameter

1 Institut fur Geologische Wissenschaften, FachrichtungPalaontologie, Freie Universitat Berlin, Malteserstraße 74-100, D-12249 Berlin, Germany. Email: [email protected]

known as enamel prisms (Lester and Koenigswald,1989; Wood and Stern, 1997). Enamel prisms orig-inate at the EDJ and run through the entire enamelthickness to the outer enamel surface (OES). Thecurrent model holds that each enamel prism isformed by a single cell (ameloblast) during its mi-gration from the basal membrane to the OES(Moss-Salentijn et al., 1997).

Enamel prisms are an important innovation inthe evolutionary history of mammalian teeth be-cause ameloblasts construct composite enamel tis-sue (Pfretzschner, 1988). In Mesozoic Holotheriaand many other basal mammals, the enamel prismsare embedded in a thick matrix of crystallites,which are oriented parallel to each other but arenot bundled. The crystallites form the so-called in-terprismatic matrix (IPM). In these early mammals,enamel prisms run parallel to each other from theEDJ to the OES, and the IPM crystallites run par-allel to the prism long axes. This enamel type iscalled radial enamel and characterizes, besides Me-sozoic Holotheria, many Paleogene mammals andsome Recent mammals as well [e.g., Lipotyphla and

132 m Science Series 40 Martin: Rodent Enamel Microstructure

Figure 1 Schematic drawing of rodent incisor enamel asseen in longitudinal section. The tip of the incisor is point-ing upward. EDJ, enamel–dentine junction; HSB, Hunter–Schreger band; OES, outer enamel surface; PE, portio ex-terna; PI, portio interna; PLEX, prismless external layer(after Martin, 1999a).

Chiroptera (Koenigswald, 1997)]. When eutherianteeth reach a certain size or are exposed to an in-creased stress pattern (e.g., crushing and grinding),enamel microstructure acquires a higher level ofcomplexity (Koenigswald et al., 1987). Decussatingenamel prisms strengthen the enamel and increaseits resistance against crack propagation (Pfretz-schner, 1988). Decussating prisms are arranged inlayers that generally are 10–15 prisms thick (Ko-enigswald and Pfretzschner, 1987). Within one lay-er, the prisms run parallel to each other; in adjacentlayers, they decussate at a high angle (about 90 de-grees). The prism layers are oriented parallel to theocclusal surface and surround the tooth crown. Inlongitudinal section, they appear as bands, whichare referred to as Hunter–Schreger bands (HSB).

In the highly specialized, ever-growing incisors ofrodents, enamel covers only the anterior surface ofthe tooth (Figs. 1, 2). Together with the softer, un-derlying dentine that wears at a faster rate, theenamel cover forms a self-sharpening cutting edge.Because of the high stresses incisor enamel experi-ences during gnawing, the enamel microstructurereaches a very high level of complexity. In general,rodent incisor HSB are thinner than those of othermammals. Rodent incisor enamel microstructureevolved independently from that of the cheek teeth(Koenigswald and Clemens, 1992), and it is partic-ularly useful for phylogenetic studies because inci-sor form and function are very similar in differentrodent groups in comparison to rodent cheek teeth.

On the basis of HSB thickness, Korvenkontio(1934) distinguished in rodent incisors three typesof HSB: multiserial, pauciserial, and uniserial. Mul-tiserial HSB are three to seven prisms wide, pauci-serial one to six (average three prisms), and uni-serial only one prism thick (Martin, 1993). Majorgroups of Rodentia can be characterized roughly byHSB type: for example, hystricognaths by multis-erial HSB, paramyids and ischyromyids by paucis-erial HSB, and myomorphs by uniserial HSB. Onthe basis of HSB thickness, Korvenkontio (1934)interpreted multiserial HSB as the most primitive,and pauciserial and uniserial as more derived enam-el types. Later, Martin (1992, 1993) demonstratedthat pauciserial HSB are the most primitive condi-tion and that structural characters such as orien-tation and thickness of the IPM are crucial for as-sessing primitive and derived stages. The HSBthickness alone is of little value.

A plesiomorphic condition is thick IPM sur-rounding each single enamel prism and runningparallel to the prism long axes. This IPM orienta-tion characterizes pauciserial HSB and is found inthe most basal rodents, such as early Paleogene is-chyromyids and ctenodactyloids, as well as in theoutgroup [e.g., Paleocene condylarths (Koenig-swald et al., 1987)]. Other characters of pauciserialHSB include the low inclination of HSB, missingtransition zones, and low enamel thickness [for de-tails, see Martin (1993)].

In uniserial HSB, band thickness is reduced to asingle prism. So far, this highly derived enamel typehas been detected only in Rodentia, where it char-acterizes sciurids, myomorphs, geomyoids, glirids,and certain theridomyids (Koenigswald, 1980;Martin, 1999a; Kalthoff, 2000).

In multiserial HSB, the number of prisms may beslightly increased compared with pauciserial HSB,but mostly overlaps the same range (three to fiveprisms per band). This demonstrates that HSBthickness is not a suitable character for distinctionbetween pauciserial and multiserial HSB. The mostimportant difference is the orientation of the IPM,which runs at an angle to prism long axes in mul-tiserial HSB. In rare cases, the IPM may run parallelto prism long axes, but in those instances, it ismuch thinner than in pauciserial HSB, and it doesnot surround each single prism. Another importantcharacter of multiserial HSB is the presence of tran-sition zones between HSB (Wahlert, 1984; Koenig-swald and Pfretzschner, 1987), where prisms switchfrom one HSB to the next higher on their way fromthe EDJ to the OES (Fig. 1). Multiserial HSB char-acterize hystricognath rodents, such as caviom-orphs, Thryonomyoidea, Pedetidae, and derivedCtenodactyloidea (Martin, 1992, 1993, 1995).Within caviomorphs, the Octodontoidea are char-acterized by a plate-like IPM, which runs at a rightangle to the prisms. This enamel type is consideredto be the most derived on the basis of biomechan-ical considerations and evolutionary occurrence.

Science Series 40 Martin: Rodent Enamel Microstructure m 133

Figure 2 Cross sections of rodent incisors from Santa Rosa. A, LACM 145444, upper; B, LACM 145447, upper; C,LACM 145446, lower; D, LACM 145445, lower; E, LACM 145442, lower; F, LACM 145443, lower. (Drawing by M.Bulang-Lorcher).

METHODS AND MATERIALS

The enamel preparation technique has been de-scribed in detail elsewhere (Martin, 1992, 1993).Small pieces of rodent incisors a few millimeterslong were embedded in polyester resin. Enamelsamples were cut in longitudinal and cross section,ground, polished, and subsequently etched with 2N hydrochloric acid for 2–4 seconds to make mor-phological details visible. After rinsing and drying,samples were sputter-coated with gold and studiedwith a scanning electron microscope (SEM) at mag-nifications between 3500 and 32000. The enamelsamples are housed in the author’s enamel collec-tion under the reference numbers given (MA plusnumber). For a glossary of terms see Koenigswaldand Sander (1997) and Martin (1999a).

Rodent incisors (Fig. 2) studied from the ?EoceneSanta Rosa locality, Natural History Museum ofLos Angeles County (LACM) locality 6789 (col-lected by K.E. Campbell, C.D. Frailey, and L. Rom-ero-Pitman; July 1995): LACM 145442, left lowerincisor, MA 259; LACM 145443, left lower incisor,MA 260; LACM 145444, right upper incisor, MA261; LACM 145445, right lower incisor, MA 262;LACM 145446, right lower incisor, MA 263;LACM 145447, right upper incisor, MA 264.

RESULTS

INCISOR MORPHOLOGY

Among the incisors, three size classes can be distin-guished (Fig. 2). Upper incisor LACM 145444 (Fig.

2A) (anterior–posterior diameter 2.9 mm) and low-er incisors LACM 145442 and 145443 (Figs. 2E–F; diameter of both about 2.8 mm) belong to thelargest group. They are very similar in cross-sec-tional shape and are characterized by the strongoverlap of the enamel cover on the lateral side. Up-per incisor LACM 145447 (Fig. 2B) is slightlysmaller (diameter 2.5 mm) and characterized by agroove on the lateral part of the enamel cover thatappears as an enamel indentation in cross section.Lower incisors LACM 145446 (Fig. 2C; diameter1.25 mm) and LACM 145445 (Fig. 2D; diameter1.1 mm) are much smaller, and their enamel coverdoes not extend very far on the lateral side.

ENAMEL MICROSTRUCTURE

Upper Incisors (Table 1)

LACM 145444 (MA 261). Incisor enamel is two-layered (Figs. 3A–B) with multiserial HSB in theportio interna (PI; Fig. 1) and radial enamel in theportio externa (PE; Fig. 1). HSB comprise three tofive prisms, and the thin IPM runs parallel to theprism long axes. Transition zones between HSB arewell developed. Prism cross sections are oval in thePI and lancet-shaped in the PE. HSB are inclined20 degrees apically. In the PE, inclination of prismsis 55 degrees. Enamel thickness is 235 mm, and PEcomposes 15–20 percent of the entire enamel thick-ness (42–48 mm). A prismless external layer (PLEX;Fig. 1) is missing.

LACM 145447 (MA 264). Incisor enamel is two-


Figure 3 SEM micrographs of incisor enamel longitudinal sections of upper rodent incisors from Santa Rosa. In allsections, EDJ (enamel–dentine junction) is to the left and the tip of the incisor is to the top. A, LACM 145444 (MA261), overview, scale bar 5 50 mm; B, LACM 145444 (MA 261), detail of PI (portio interna) with HSB (Hunter–Schregerbands); thin IPM (interprismatic matrix) runs parallel to the prism long axes, scale bar 5 10 mm; C, LACM 145447(MA 264), overview, scale bar 5 50 mm; D, LACM 145447 (MA 264), detail of PI with HSB; thin IPM runs at a smallangle or parallel to the prism long axes, scale bar 5 10 mm.

layered with multiserial HSB in the PI and radialenamel in the PE (Figs. 3C–D). HSB comprise threeto six prisms, and the thin IPM runs at a very acuteangle or parallel to the prism long axes. Transitionzones between the HSB are well developed. Prism

cross sections are oval in the PI and lancet-shapedin the PE. HSB inclination is 15–20 degrees. In thePE, prisms are inclined 55 degrees apically. Enamelthickness is 190 mm, and PE composes 15 percentof the entire enamel thickness (30 mm). A PLEX is


Tab

le1

Inci

sor

enam

elfe

atur

esof

the

spec

imen

sfr

omSa

nta

Ros

a.

Spec

imen

num

ber

Inci

sor

Schm

elzm

uste

rPr

ism

spe

rH

SB

Incl

inat

ion

ofH

SB(d

eg)

Enam

elth

ickn

ess

(mm

)Pe

rcen

tage

ofPE

thic

knes

s

PEpr

ism

incl

inat

ion

(deg

)IP

Min

PI

LA

CM

1454

44L

AC

M14

5447

Upp

erU

pper

Mul

tise

rial

Mul

tise

rial

3–5

3–6

2015

–20

235

190

15–2

015

55 55Pa

ralle

lPa

ralle

lto

acut

ean

gula

rL

AC

M14

5446

Low

erM

ulti

seri

al3–

440

135

2065

Para

llelt

oac

ute

angu

lar

LA

CM

1454

45

LA

CM

1454

42L

AC

M14

5443

Low

er

Low

erL

ower

Mul

tise

rial

Mul

tise

rial

Mul

tise

rial

3–4

3–4

3–4

35 35 25

130

230

240

15 1512

–15

65 70 70

Acu

tean

gula

rR

ecta

ngul

arR

ecta

ngul

ar

missing. The outer surface of the enamel layer isnot smooth, but bears a groove on the lateral sidethat is clearly visible in the overview of the incisorcross section (Fig. 2B). The groove has no influenceon the enamel microstructure.

Lower Incisors (Table 1)

LACM 145446 (MA 263). Incisor enamel is two-layered with multiserial HSB in the PI and radialenamel in the PE (Figs. 4A–B). HSB comprise threeto four prisms, and IPM runs at a very low angleto the prism long axes and anastomoses regularly.Transition zones between the HSB are well devel-oped. Prism cross sections are oval in the PI andlancet-shaped in the PE. HSB inclination is 40 de-grees, and, in the PE, prisms are inclined 65 degreesapically. Enamel thickness is 135 mm, and PE com-poses 20 percent of the entire enamel thickness.

LACM 145445 (MA 262). Incisor enamel is two-layered with multiserial HSB in the PI and radialenamel in the PE (Figs. 4C–D). HSB comprise threeto four prisms, and IPM runs at an acute angle tothe prism long axes and anastomoses regularly.Transition zones between adjacent HSB are well de-veloped. Prism cross sections are oval in the PI andlancet-shaped in the PE. HSB inclination is 35 de-grees, and, in the PE, prisms are inclined 65 degreesapically. Enamel thickness is 130 mm, and PE com-poses only 15 percent of the entire enamel thickness(20 mm). A PLEX is missing.

LACM 145442 (MA 259). Incisor enamel is two-layered with multiserial HSB in the PI and radialenamel in the PE (Figs. 5A–B). HSB comprise threeto four prisms, and the platelike IPM runs at a rightangle to the prism long axes and anastomoses onlyaccidentally. Transition zones between the HSB arewell developed. Prism cross sections are oval in thePI and lancet-shaped in the PE. HSB inclination is35 degrees, and, in the PE, prisms are inclined verysteeply (70 degrees) apically. Enamel thickness is230 mm, and the thin PE composes only 15 percentof the entire enamel thickness (35 mm). A PLEX ismissing.

LACM 145443 (MA 260). Incisor enamel is two-layered with multiserial HSB in the PI and radialenamel in the PE (Figs. 5C–D). HSB comprise threeto four prisms, and the plate-like IPM runs at aright angle to the prism long axes and anastomosesonly accidentally. Transition zones between theHSB are well developed. Prism cross sections areoval in the PI and lancet-shaped in the PE. HSBinclination is 25 degrees, and, in the PE, prisms areinclined very steeply apically (70 degrees). Enamelthickness is 240 mm, and the thin PE composes only12–15 percent of the entire enamel thickness (30mm). A thin PLEX is present.

DISCUSSION

The Santa Rosa rodents play a crucial role in ourunderstanding of the origin and diversification ofcaviomorphs because they likely represent the old-


Figure 4 SEM micrographs of incisor enamel longitudinal sections of lower rodent incisors from Santa Rosa. In allsections, EDJ (enamel–dentine junction) is to the left and the tip of the incisor is to the top. A, LACM 145446 (MA263), overview; scale bar 5 50 mm; B, LACM 145446 (MA 263), detail of PI (portio interna) with multiserial HSB(Hunter–Schreger bands); IPM (interprismatic matrix) runs at an acute angle to the prism long axes and anastomosesregularly, scale bar 5 10 mm; C, LACM 145445 (MA 262), overview, scale bar 5 50 mm; D, LACM 145445 (MA 262),detail of PI with multiserial HSB; IPM runs at an acute angle to the prism long axes and anastomoses regularly, scalebar 5 10 mm.

est rodents known from South America. There hasbeen a long-standing discussion on the origin of theSouth American caviomorphs (see Frailey andCampbell, 2004) because South America was an is-land continent during the late Mesozoic and most

of the Cenozoic eras. Historically, caviomorph ro-dents seemed to appear suddenly in the fossil recordtogether with platyrrhine primates during the De-seadan South American Land Mammal Age (SAL-MA; Simpson, 1980), which was for a long time


Figure 5 SEM micrographs of incisor enamel longitudinal sections of lower rodent incisors from Santa Rosa. In allsections, EDJ (enamel–dentine junction) is to the left and the tip of the incisor is to the top. A, LACM 145442 (MA259), overview, scale bar 5 50 mm; B, LACM 145442 (MA 259), detail of PI (portio interna) with multiserial HSB(Hunter–Schreger bands); plate-like IPM (interprismatic matrix) runs at a right angle to the prism long axes and doesnot anastomose, scale bar 5 10 mm; C, LACM 145443 (MA 260), overview, scale bar 5 50 mm; D, LACM 145443(MA 260), detail of PI with multiserial HSB; plate-like IPM runs at a right angle to the prism long axes and does notanastomose, scale bar 5 10 mm.


regarded as early Oligocene, but which more re-cently has been placed in the late Oligocene to earlyMiocene (MacFadden et al., 1985) or late early Ol-igocene to early Miocene (Marshall et al., 1986).The level of the Salla beds in Bolivia, from whichthe earliest platyrrhine primate Branisella comes, isnow dated between 27.02 and 25.82 Ma (Kay etal., 1998), and those authors place all of the fos-siliferous Salla beds in the Oligocene. Recently, amandible of a supposed dasyproctid from Tingui-ririca (Chile) has been described (Wyss et al.,1993), which predates the Deseadan level and hasbeen dated radiometrically at 31.5 Ma, possibly ex-tending back to ;37.5 Ma (early Oligocene or lateEocene). The Santa Rosa local fauna of easternPeru is the first known Paleogene mammalian pa-leofauna of tropical South America. On the basisof the evolutionary stage of the marsupials (Goinand Candela, 2004) and rodents (Frailey andCampbell, 2004), the age of the Santa Rosa localfauna is placed at late Eocene (Mustersan SALMA)or, less probably, early Oligocene (post-Mustersan,pre-‘‘Tinguirirican’’ age), following the change inthe age of the Mustersan SALMA by Kay et al.(1999). The Santa Rosa local fauna was collectedfrom fine- to coarse-grained bed load or channeldeposits within massive red bed clays that are ex-posed in the left bank of the Rio Yurua, approxi-mately 7 km north of Breu and 7.5 km south of theBrazilian border, Departamento de Ucayali, Peru.The source red beds are tentatively referred to thePaleocene–Eocene Yahuarango Formation (Camp-bell et al., 2004).

Two classical hypotheses have been put forwardfor the origin of caviomorphs. Either they are de-scendants of North American ischyromyids (Woodand Patterson, 1959; Wood, 1981, 1985; Pattersonand Wood, 1982) or African Thryonomyoidea(‘‘Phiomorpha’’)(Lavocat, 1969, 1971, 1974). Onthe basis of a study of incisor enamel microstruc-ture, Martin (1992, 1994) presented a clear case forthe origin of caviomorphs from Paleogene Thryon-omyoidea because both share the same derivedcharacters of multiserial schmelzmuster in the in-cisors. In this regard, the incisor enamel microstruc-ture of the Santa Rosan rodents plays a crucial rolebecause, as the oldest South American rodentsknown so far, they give valuable information on theevolutionary history of caviomorph schmelzmuster.All rodent incisors from Santa Rosa studied to dateexhibit a double-layered enamel with multiserialHSB, which is typical for caviomorph rodents(Martin, 1992). Among the six investigated inci-sors, all three subtypes of multiserial HSB are pre-sent, which are also observed in Oligocene andyounger caviomorphs. This indicates that the ?Eo-cene Santa Rosa rodent fauna was already diversi-fied. The observations at the enamel microstructurelevel are also supported by the macromorphologi-cal diversity of the investigated incisors; at leastthree groups can be distinguished (Fig. 2) by differ-ences in size and shape.

The incisor enamel schmelzmuster observed inthe Santa Rosan rodents are by no means moreprimitive than those of Deseadan SALMA or latercaviomorphs (Martin, 1992). None of the SantaRosan incisors exhibit pauciserial HSB or aschmelzmuster transitional from the pauciserial tothe multiserial condition. This supports the hypoth-esis that the caviomorph ancestors are descendantsof an African stock belonging to the stem lineageof Thryonomyoidea and caviomorphs (Martin,1992, 1994) and that the first African immigrantsinto South America introduced the multiserialschmelzmuster. Outside of South America, Eocene–Oligocene stem lineage representatives of AfricanThryonomyoidea, advanced Ctenodactyloidea, andPedetidae are the only rodents showing schmelz-muster with multiserial HSB in the incisors (Mar-tin, 1992, 1994). Northern Hemisphere Ischyro-myoidea are characterized by more primitiveschmelzmuster with pauciserial HSB, and in thisclade, a multiserial schmelzmuster never evolved(Martin, 1993).

Among caviomorph rodents, the superfamily Oc-todontoidea (comprising Octodontidae, Ctenomyi-dae, Abrocomidae, Echimyidae, Capromyidae, andMyocastoridae) is characterized by the presence ofrectangular, plate-like IPM in the HSB (Martin,1992). It is considered the most derived type ofHSB on the basis of biomechanical considerationsand evolutionary occurrence. This derived enamelsubtype was found in two of the Santa Rosan in-cisors (LACM 145442 and 145443; Fig. 5), whichalso are very similar to each other in size and grossmorphology (possibly belonging to the same spe-cies), indicating that members of the Octodonto-idea were already present. Other typical charactersof the octodontoid incisor schmelzmuster that wereobserved in the specimens from Santa Rosa are thehigh inclinations of HSB in the PI and of the radi-ally oriented prisms in the PE.

The diversity of the incisor schmelzmuster foundin the rodents from Santa Rosa makes it probablethat the immigration event took place a consider-able time span before that stratigraphic level. De-pending on the age of the Santa Rosa local fauna,an early or middle Eocene age for the arrival of theimmigrants could be projected. However, the oldestrecord of multiserial HSB in African ‘‘phiomorph’’incisors from Bir el Ater, Algeria (Martin, 1993), islate Eocene (Jaeger et al., 1985). In the Bir el Ater‘‘phiomorphs,’’ only multiserial HSB with acute an-gular IPM have been detected; the first multiserialHSB with rectangular, plate-like IPM appear in lateEocene/early Oligocene ‘‘phiomorphs’’ from the Fa-yum (Martin, 1992). Another possibility for the or-igin of caviomorphs in South America is the as-sumption of several waves of immigration of Afri-can ancestors shortly before the Santa Rosan level.In this case, it would have to be assumed that thecaviomorph stem-lineage immigrants were taxo-nomically diverse and that they already possesseddifferent subtypes of multiserial HSB.


A different hypothesis for the origin of cavio-morphs is presented by Frailey and Campbell(2004). From among the many, mainly isolated ro-dent teeth from Santa Rosa, they distinguish at leasteight genera with possibly 17 species in three fam-ilies (Erethizontidae, Agoutidae, and Echimyidae).According to the advanced molar morphology ofthe Santa Rosan caviomorphs, compared with con-temporaneous rodents from other parts of theworld (e.g., a four-crested molar pattern), an earlyseparation of the caviomorph lineage (late Creta-ceous or, at the latest, early Paleocene) is postulat-ed. The caviomorphs are regarded as an autochtho-nous South American group that had been separat-ed by the breakup of Gondwanaland from a com-mon Gondwanan stock. In this case, the similaritiesbetween South American and African hystricog-naths (including enamel microstructure characters)would be the result of a once common gene pool.

ACKNOWLEDGMENTS

I thank K.E. Campbell, Natural History Museum of LosAngeles County, for providing the rodent incisors fromSanta Rosa for study and for the invitation to contributeto this volume. He and C.D. Frailey permitted me to citefrom their chapters in this volume on the geology androdent fauna of Santa Rosa. I thank Francisco Goin forproviding a Spanish translation of the abstract. L. Flynn,W.v. Koenigswald, and J.H. Wahlert gave useful com-ments on the manuscript. This study was supported by aHeisenberg grant from the Deutsche Forschungsgemein-schaft (Ma 1643/5-1).

LITERATURE CITED

Campbell, K.E., C.D. Frailey, and L. Romero-Pittman.2004. The Paleogene Santa Rosa local fauna of Am-azonian Peru: Geographic and geologic setting. InThe Paleogene mammalian fauna of Santa Rosa,Amazonian Peru, ed. K.E. Campbell, Jr. Natural His-tory Museum of Los Angeles County, Science Series40, 3–14.

Frailey, C.D., and K.E. Campbell. 2004. Paleogene ro-dents from Amazonian Peru: The Santa Rosa localfauna. In The Paleogene mammalian fauna of SantaRosa, Amazonian Peru, ed. K.E. Campbell, Jr. Nat-ural History Museum of Los Angeles County, Sci-ence Series 40, 71–130.


Jaeger, J.-J., C. Denys, and B. Coiffait. 1985. New Phiom-orpha and Anomaluridae from the late Eocene ofnorth-west Africa: Phylogenetic implications. InEvolutionary relationships among rodents, ed. W.P.Luckett and J.-L. Hartenberger, 567–588. NewYork: Plenum Press.

Kalthoff, D.C. 2000. Die Schmelzmikrostruktur in den In-zisiven der hamsterartigen Nagetiere und andererMyomorpha (Rodentia, Mammalia). Palaeontogra-phica A 259:1–193.

Kay, R.F., R.H. Madden, M.G. Vucetich, A.A. Carlini,M.M. Mazzoni, G.H. Re, M. Heizler, and H. San-

deman. 1999. Revised age of the CasamayoranSouth American Land Mammal ‘‘Age’’—Climaticand biotic implications. Proceedings of the NationalAcademy of Sciences, Washington 96(23):13235–13240.

Kay, R.F., B.J. McFadden, R.H. Madden, H. Sandeman,and F. Anaya. 1998. Revised age of the Salla beds,Bolivia, and its bearing on the age of the DeseadanSouth American Land Mammal ‘‘Age.’’ Journal ofVertebrate Paleontology 18:189–199.

Koenigswald, W.v. 1980. Schmelzmuster und Morpholo-gie in den Molaren der Arvicolidae (Rodentia). Ab-handlungen der senckenbergischen naturforschendenGesellschaft 539:1–129.

. 1997. Brief survey of enamel diversity at theschmelzmuster level in Cenozoic placental mammals.In Tooth enamel microstructure, ed. W.v. Koenig-swald and P.M. Sander, 137–161. Rotterdam: Bal-kema.

Koenigswald, W.v., and W.A. Clemens. 1992. Levels ofcomplexity in the microstructure of mammalianenamel and their application in studies of systemat-ics. Scanning Microscopy 6:195–218.

Koenigswald, W.v., and H.U. Pfretzschner. 1987. Hunter–Schreger-Bander im Zahnschmelz von Saugetieren(Mammalia), Anordnung und Prismenverlauf.Zoomorphology 106:329–338.

Koenigswald, W.v., J.M. Rensberger, and H.U. Pfretz-schner. 1987. Changes in the tooth enamel of earlyPaleocene mammals allowing increased diet diversi-ty. Nature 328:150–152.

Koenigswald, W.v., and P.M. Sander. 1997. Glossary ofterms used for enamel microstructures. In Toothenamel microstructure, ed. W.v. Koenigswald andP.M. Sander, 267–280. Rotterdam: Balkema.

Korvenkontio, V.A. 1934. Mikroskopische Untersuchun-gen an Nagerincisiven, unter Hinweis auf dieSchmelzstruktur der Backenzahne. Annales Zoolo-gici Societatis Zoologicae-Botanicae Fennica Vana-mo 2:1–274.

Lavocat, R. 1969. La systematique des rongeurs hystri-comorphes et la derive des continents. Comptes Ren-dus de l’Academie Scientifique, Paris D 269:1496–1497.

. 1971. Affinites systematiques des caviomorphes etdes phiomorphes et origine africaine des caviomor-phes. Anais da Academia Brasiliera de Ciencias43(Suppl.):515–522.

. 1974. The interrelationships between the Africanand South American rodents and their bearing onthe problem of the origin of South American mon-keys. Journal of Human Evolution 3:323–326.

Lester, K.S., and W.v. Koenigswald. 1989. Crystallite ori-entation discontinuities and the evolution of mam-malian enamel or, when is a prism? Scanning Mi-croscopy 3:645–663.

Luckett, W.P. 1993. An ontogenetic assessment of dentalhomologies in therian mammals. In Mammal phy-logeny, Mesozoic differentiation, multituberculates,monotremes, early therians, and marsupials, ed. F.S.Szalay, M.J. Novacek, and M.C. McKenna, 182–204. Berlin, New York, Heidelberg: Springer-Verlag.

MacFadden, B., K.E. Campbell, R.L. Cifelli, O. Siles,N.M. Johnson, C.W. Naeser, and P.K. Zeitler. 1985.Magnetic polarity stratigraphy and mammalian fau-na of the Deseadan (late Oligocene–early Miocene)Salla beds of northern Bolivia. Journal of Geology93:223–250.


Marshall, L.G., R.L. Cifelli, R.E. Drake, and G.H. Curtis.1986. Vertebrate paleontology, geology, and geo-chronology of the Tapera de Lopez and ScarrittPocket, Chubut Province, Argentina. Journal of Pa-leontology 60:920–951.

Martin, T. 1992. Schmelzmikrostruktur in den Inzisivenalt- und neuweltlicher hystricognather Nagetiere. Pa-laeovertebrata, Memoire extraordinaire, 1–168.

. 1993. Early rodent incisor enamel evolution: Phy-logenetic implications. Journal of Mammalian Evo-lution 1:227–253.

. 1994. African origin of caviomorph rodents is in-dicated by incisor enamel microstructure. Paleobi-ology 20:5–13.

. 1995. Incisor enamel microstructure and phylo-genetic interrelationships of Pedetidae and Cteno-dactyloidea (Rodentia). Berliner geowissenschaftli-che Abhandlungen E 16:693–707.

. 1997. Incisor enamel microstructure and system-atics in rodents. In Tooth enamel microstructure, ed.W.v. Koenigswald and P.M. Sander, 163–175. Rot-terdam: Balkema.

. 1999a. Evolution of incisor enamel microstruc-ture in Theridomyidae (Rodentia). Journal of Ver-tebrate Paleontology 19:550–565.

. 1999b. Dryolestidae (Dryolestida, Mammalia)aus dem Oberen Jura von Portugal. Abhandlungender senckenbergischen naturforschenden Gesell-schaft 550:1–119.

Moss-Salentijn, L., M.L. Moss, and M.S. Yuan. 1997. Theontogeny of mammalian enamel. In Tooth enamelmicrostructure, ed. W.v. Koenigswald and P.M.Sander, 5–30. Rotterdam: Balkema.

Patterson, B., and A.E. Wood. 1982. Rodents from theDeseadan Oligocene of Bolivia and the relationshipsof the Caviomorpha. Bulletin of the Museum ofComparative Zoology 149:371–543.

Pfretzschner, H.U. 1988. Structural reinforcement andcrack propagation in enamel. In Teeth revisited: Pro-ceedings of the VII International Symposium on

Dental Morphology. Paris 1986, ed. D.E. Russell,J.P. Santoro, and D. Sigogneau-Russell. Memoirs duMuseum national d’Histoire naturelle C 53:133–143.

Sander, P.M. 1997. Non-mammalian synapsid enamel andthe origin of mammalian enamel prisms: The bot-tom-up perspective. In Tooth enamel microstructure,ed. W.v. Koenigswald and P.M. Sander, 41–62. Rot-terdam: Balkema.

Simpson, G.G. 1980. Splendid isolation: The curious his-tory of South American mammals. New Haven: YaleUniversity Press, 266 pp.

Wahlert, J.H. 1984. Hystricomorphs, the oldest branch ofthe Rodentia. Annals of the New York Academy ofSciences 435:356–357.

Wood, A.E. 1981. The origin of the caviomorph rodentsfrom a source in middle America: A clue to the areaof origin of the platyrrhine primates. In Evolutionarybiology of the New World monkeys and continentaldrift, ed. R.L. Ciochon and A.B. Chiarelli, 79–91.New York: Plenum Press.

. 1985. The relationships, origin, and dispersal ofthe hystricognathous rodents. In Evolutionary biol-ogy among rodents, ed. W.P. Luckett and J.-L. Har-tenberger, 475–513. New York: Plenum Press.

Wood, A.E., and B. Patterson. 1959. The rodents of theDeseadan Oligocene of Patagonia and the beginningsof South American rodent evolution. Bulletin of theMuseum of Comparative Zoology 120:282–428.

Wood, C.B., and D.N. Stern. 1997. The earliest prisms inmammalian and reptilian enamel. In Tooth enamelmicrostructure, ed. W.v. Koenigswald and P.M.Sander, 63–84. Rotterdam: Balkema.

Wyss, A.E., J.F. Flynn, M.A. Norell, C.C. Swisher III, R.Charrier, M.J. Novacek, and M.C. McKenna. 1993.South America’s earliest rodent and recognition of anew interval of mammalian evolution. Nature 365:434–437.



A Possible Bat (Mammalia: Chiroptera) from the ?Eocene ofAmazonian Peru

Nicholas J. Czaplewski1 and Kenneth E. Campbell, Jr.2

ABSTRACT. An isolated, tribosphenic lower molar was recovered by screen-washinglowland, continental deposits of probable Eocene age along the Rıo Yurua in easternPeru. The tooth appears to represent a member of the Chiroptera, and it shows simi-larities to the teeth of species of the family Noctilionidae, but it is incomplete andinsufficient for a familial identification. It forms part of the Santa Rosa local fauna, thefirst known Paleogene mammalian fauna from Amazonia. If correctly identified, thespecimen is the oldest known fossil bat in South America and only the third record ofa bat from the Paleogene of that continent.

RESUMEN. A partir del lavado de muestras sedimentarias de depositos continentalesde probable edad eocena, procedentes del Rıo Yurua en el este del Peru, ha sido recu-perado un molar tribosfenico aislado. El diente parece representar a un miembro de losChiroptera y muestra similitudes con aquellos de las especies de Noctilionidae, si bienel resto es incompleto e insuficiente como para asegurar su identificacion familiar. Formaparte de la fauna local de Santa Rosa, la primera fauna de mamıferos conocida para elPaleogeno de la Amazonia. Si el ejemplar ha sido correctamente identificado, constituyeel murcielago mas antiguo conocido para America del Sur y el tercer registro de unmurcielago paleogeno para este continente.

INTRODUCTION

Small mammals are relatively unknown as fossils inCenozoic deposits of lowland tropical South Amer-ica. Previously, only the middle Miocene beds at LaVenta, Colombia (Kay et al., 1997), and the upperMiocene Acre Conglomerate of southwestern Ama-zonia (Frailey, 1986; Kay and Frailey, 1993; Cza-plewski, 1996) have produced large numbers oftiny bones and teeth of small vertebrates (‘‘micro-vertebrates’’), including marsupials, primates, bats,and rodents. However, the technique of screen-washing to recover microvertebrates is just begin-ning to be applied on a wider scale to South Amer-ican sedimentary deposits. Even for large mam-mals, relatively few specimens are known from thetropical portion of the continent (see faunal list forwestern Amazonia in Campbell et al., 2000). Forthe Paleogene, previous records of microvertebratesfrom the lowland tropics are virtually nonexistent.

The remains of vertebrates were discovered atSanta Rosa in 1995 in the left bank of the Rıo Yu-rua near its confluence with the Rıo Beu (Campbellet al., 2004). The locality is at 98299390 S,728459480 W, approximately 7 km (by air) north of

1 Oklahoma Museum of Natural History, 2401 Chau-tauqua Avenue, and Department of Zoology, Universityof Oklahoma, Norman, Oklahoma 73072-7029. Email:[email protected]


Breu and 7.5 km (by air) south of the Peru–Brazilborder (see Campbell et al., 2004:fig. 1). The ele-vation is about 215 m above mean sea level. A pre-liminary sample of sediments from the locality wasprocessed by screen washing and yielded hundredsof rodent and marsupial teeth and jaws, promptinga return visit to the locality in 1998 to collect ad-ditional concentrated matrix (Campbell et al.,2004). The fossiliferous deposits occur as small flu-vial lenses (,30 cm thick) in a cutbank consistingof about 5 m of red clay unconformably overlainby buff sands. The fluvial lenses contain coarsesand, calcareous nodules, and clay pebbles. Thestage of evolution of certain of the other vertebratefossil groups in the Santa Rosa fauna, especially themarsupials and rodents, suggests a probable lateEocene age for the fauna (Frailey and Campbell,2004; Goin and Candela, 2004), and the beds inwhich the fossils occur at the Santa Rosa localityare interpreted as belonging to the Paleocene–Eo-cene Yahuarango Formation (Campbell et al.,2004).

Among the microvertebrate materials screen-washed from the Santa Rosa locality is an enig-matic tooth from a tiny, insectivorous mammal. Al-though the specimen is damaged, it is important forits geologic age, its occurrence in the Amazon Ba-sin, and the possibility that it might represent theoldest fossil bat from South America. Only two Pa-leogene bats have previously been reported fromSouth America. An incomplete skeleton of the mo-lossid bat Mormopterus faustoi Paula Couto, 1956,

142 m Science Series 40 Czaplewski and Campbell: ?Eocene Bat

Figure 1 Left lower molar (LACM 140623) attributed to Chiroptera from the Santa Rosa local fauna, Peru. A, lingualview; B, labial view; C, occlusal view (stereo pair).

was reported from the bituminous shales of the Up-per Oligocene Tremembe Formation in southeast-ern Brazil (Paula Couto, 1956, 1983; Paula Coutoand Mezzalira, 1971; Legendre, 1984, 1985). Asecond skeleton of a bat was also reported from theTremembe Formation (Leonardos, 1924; Mezzali-ra, 1964, 1966, 1989), but its familial identity isindeterminate (Czaplewski, personal observation).Although the second Tremembe skeleton is morenearly complete than that of M. faustoi, it iscrushed, and only small fragments of the teeth re-main. The Tremembe specimens are considered tobe of the Deseadan South American Land MammalAge (SALMA).


Infraclass Eutheria Huxley, 1880

Order Chiroptera Blumenbach, 1779

Family, Genus, and Species IndeterminateFigure 1

SPECIMEN. Natural History Museum of LosAngeles County (LACM) 140623, left lower molar.

LOCALITY AND AGE. Santa Rosa, Rıo Yurua,Departamento de Ucayali, Peru; ?Yahuarango For-mation. ?Eocene. LACM locality 6289.

DESCRIPTION. The molar is tribosphenic. Itbears an anteroposteriorly compressed trigonidwith protoconid, paraconid, and metaconid. Thetalonid portion includes a hypoconid, but the pos-terolingual corner is broken. Therefore, the hypo-conulid and entoconid are missing, so it is not pos-sible to know the morphology of these importantcusps. The remaining portion of the talonid isslightly less compressed than the trigonid. The pro-toconid is the tallest of the trigonid cusps, followedby the metaconid and paraconid. The bases of theparaconid and metaconid are at about the same lev-

el in lingual view. The hypoconid is about half theheight of the protoconid. The tooth measures ap-proximately 1.25 mm (estimated because of break-age) in anteroposterior length, 1.02 mm in trans-verse width of the trigonid, and approximately 1.02mm (estimated) in transverse width of the talonid.

The major cusps are slender. There are strong,continuous anterior, labial, and posterior cingulids(Fig. 1). A tiny, weak lingual cingulid is presentacross the foot of the trigonid valley. There is asmall swelling at the lingual extremity of the ante-rior cingulid. A tiny interdental contact wear facetoccurs on the anterior wall of the paraconid. Thelabial valley between protoconid and hypoconidextends across the tooth as a groove anterior to thecristid obliqua and up toward the apex of the meta-conid. The cristid obliqua extends from the hypo-conid toward, but does not reach, the apex of themetaconid. Instead, it descends from the hypoconiddown into the groove between talonid and trigonid,toward the base of the metaconid. In anterior view,the notch in the paracristid has an angle of about90 degrees. In posterior view, the notch in the pro-tocristid forms an angle of slightly less than 90 de-grees. Only half of the postcristid is preserved, butenough of it is present to indicate that, in occlusalview, the angle between it and the cristid obliqua isless acute than the angle formed by the paracristidand protocristid. The hypoconulid (assuming it waspresent) and the entoconid were probably close toeach other (i.e., the hypoconulid was not mediallyplaced). A distal metacristid with a small wear facetextends posterolabially from the metaconid. Theposterolabial trend of the metacristid is similar tothat in the posterior molars of some bats and dif-ferent from that of the anterior molars, suggestingthat the tooth could possibly be an M3.

DISCUSSION. The Santa Rosa specimen has atypical dilambdodont morphology that is generally

Science Series 40 Czaplewski and Campbell: ?Eocene Bat m 143

seen in Paleogene insectivorous mammals of manykinds, including certain metatherians, chiropterans,adapisoriculids, talpoids, nyctitheriids, soricids,plagiomenids, mixodectids, galeopithecids, and tu-paiids. Little attempt was made at comparisonswith this bewildering array of mammals in whichconvergence of characters of the lower molars isprobably rampant. Except for metatherians and theOligocene bats from the Tremembe Formation (seeabove), the taxa mentioned are unknown as fossilsin the Paleogene of South America. The Santa Rosaspecimen is referred with uncertainty to the Chi-roptera because it has a well-developed labial cin-gulid on the lower molars. A strong labial cingulidis probably a synapomorphy of early bats (Hand etal., 1994); it is lacking in metatherians.

The Santa Rosa tooth differs from lower molarsof most bats and insectivores in its cristid obliqua,which terminates on the base of the metaconid in-stead of on the middle of the posterior wall of thetrigonid below the notch in the protocristid. Onlyone New World family of bats, the Noctilionidae,has lower molars in which the cristid obliqua isshifted lingually onto the metaconid (Czaplewski,1996). In the Old World, this feature also occurs inNycteridae and Megadermatidae. However, inNoctilio Linneaus, 1766 (the only genus in the fam-ily Noctilionidae), the cristid obliqua is much moreprominent than in the Santa Rosa fossil.

Other differences between the Santa Rosa toothand those of Noctilio are that the talonid of molarsof Noctilio is much wider than the trigonid (ofequal width in LACM 140623), and the two moi-eties of the tooth are more transversely oriented(i.e., at right angles to the long axis of the toothrow) instead of slightly obliquely oriented. Alsolacking in the Santa Rosa tooth, but present in mo-lars of Noctilio is a wall-like distal metacristid–en-tocristid that closes off the talonid valley lingually.These characters are present in specimens of theextant N. albiventris Desmarest, 1818 from themiddle Miocene of Colombia, which is the earliestfossil record of a noctilionid (Czaplewski, 1997).At this time, it is impossible to say whether thecharacters of the Santa Rosa tooth might representa transitional grade in the development of thesecharacters of noctilionids. The differences seem topreclude assigning LACM 140623 to the Noctilion-idae.

The lower molars of a few Paleogene lipotyphlaninsectivores, such as adapisoriculids, nyctitheriids,talpids, and proscalopids, resemble LACM 140623in some of its details. Certain talpids and proscal-opids possess lower molars with a well-developedlabial cingulid and a labial valley that reaches mostof the way across the tooth to the metaconid. How-ever, in these taxa, the occlusal outline is usually inthe shape of a parallelogram with crests stronglyoblique to the long axis of the tooth row, more sothan in LACM 140623, and the cristid obliquareaches all the way to the apex of the metaconid,unlike the condition in the Santa Rosa tooth.

Many nyctitheriids and adapisoriculids have aprominent mesial cingulid like LACM 140623, butthey lack the labial cingulid on lower molars. Thelipotyphlans of the genus Wyonycteris Gingerich,1987, of the late Paleocene in North America andthe Paleocene–Eocene boundary of Europe (Smith,1995) are generally similar in morphology and sizeto LACM 140623. Wyonycteris chalix Gingerich,1987, was originally thought to be a bat, but Handet al. (1994) discussed its similarities with severalother groups of insectivorous placental mammalsand argued that it is not a bat. Smith (1995, 1997),who named an additional species, Wyonycteris ri-chardi Smith, 1995, felt that Wyonycteris is not atypical bat and thought that it probably representsa derived member of the Adapisoriculidae or Nyc-titheriidae. In W. chalix, the configuration of thecristid obliqua, the groove paralleling the cristidobliqua, and the lingual cingulid are virtually iden-tical to those in LACM 140623. Wyonycteris ri-chardi, on the other hand, lacks a lingual cingulidat the base of the trigonid valley, and it does notdevelop a small swelling at the lingual end of theanterior cingulid. But the molars of species ofWyonycteris differ from the Santa Rosa tooth inhaving a more oblique orientation of cristids,slightly less acute notches in the paracristid andprotocristid, a relatively larger paraconid, and incompletely lacking a labial cingulid.

LACM 140623 is tentatively considered to rep-resent a lower molar of a chiropteran. In the ab-sence of more diagnostic material, it is not possibleto identify the specimen accurately. The broken tal-onid cusps would be critical for determining wheth-er the tooth might represent an archaic, generalizedbat or a member of a modern clade of bats. Nev-ertheless, it is an important specimen in that it sug-gests the presence in the Santa Rosa local fauna ofan order of mammals barely known (or, if not Chi-roptera, unknown) in the Paleogene of South Amer-ica. If it truly belongs to a bat, there is a remotepossibility—on the basis of one dental character,the configuration of the cristid obliqua—that theSanta Rosa tooth represents a transitional form be-tween primitive bats and the endemic Neotropicalfamily Noctilionidae. It might also constitute theoldest record of Chiroptera in South America.

ACKNOWLEDGMENTSWe thank Tiffany Naeher for help in making the photo-graphs, Gregg Gunnell for making available a cast of W.chalix, and Francisco Goin for providing a Spanish trans-lation of the abstract. The assistance and cooperation ofIng. Hugo Rivera Mantilla, Director Tecnico of INGEM-MET, and Ing. Oscar Palacios, Director General de Geo-logıa of INGEMMET, are gratefully acknowledged. Wealso thank Suzanne Hand, Bernard Sige, and ThierrySmith for thoughtful reviews and helpful comments thatimproved the manuscript.

LITERATURE CITEDCampbell, K.E., Jr., C.D. Frailey, and L. Romero-Pittman.

2000. The late Miocene gomphothere Amahuacath-

144 m Science Series 40 Czaplewski and Campbell: ?Eocene Bat

erium peruvium (Proboscidea: Gomphotheriidae)from Amazonian Peru: Implications for the GreatAmerican Faunal Interchange. Instituto GeologicoMinero y Metalurgico, Serie D: Estudios Regionales,Boletın 23:1–152.

. 2004. The Paleogene Santa Rosa local fauna ofAmazonian Peru: Geographic and geologic setting.In The Paleogene mammalian fauna of Santa Rosa,Amazonian Peru, ed. K.E. Campbell, Jr. Natural His-tory Museum of Los Angeles County, Science Series40, 3–14.

Czaplewski, N.J. 1996. Opossums (Didelphidae) and bats(Noctilionidae and Molossidae) from the late Mio-cene of the Amazon Basin. Journal of Mammalogy77:84–94.

. 1997. Chiroptera. In Vertebrate paleontology inthe Neotropics: The Miocene fauna of La Venta, Co-lombia, ed. R.F. Kay, R.H. Madden, R.L. Cifelli, andJ.J. Flynn, 410–431. Washington, D.C.: SmithsonianInstitution Press.

Desmarest, A.G. 1818. Article ‘‘Noctilion ou bec de lie-vre.’’ Nouveau dictionnaire d’histoire naturelle, ap-pliquee aux arts, principalement a l’agriculture et al’economie rurale et domestique; par une societe denaturalistes, Nouveau Edition 23, 14–16. Deterville,Paris.

Frailey, C.D. 1986. Late Miocene and Holocene mam-mals, exclusive of the Notoungulata, of the Rio Acreregion, western Amazonia. Contributions in Science374:1–46.


Gingerich, P.D. 1987. Early Eocene bats (Mammalia, Chi-roptera) and other vertebrates in freshwater lime-stones of the Willwood Formation, Clark’s Fork Ba-sin, Wyoming. Contributions from the Museum ofPaleontology, The University of Michigan 27:275–320.


Hand, S., M. Novacek, H. Godthelp, and M. Archer.1994. First Eocene bat from Australia. Journal ofVertebrate Paleontology 14:375–381.

Kay, R.F., and C.D. Frailey. 1993. Large fossil platyrrhines

from the Rio Acre local fauna, late Miocene, westernAmazonia. Journal of Human Evolution 25:319–327.

Kay, R.F., R.H. Madden, R.L. Cifelli, and J.J. Flynn (eds.).1997. Vertebrate paleontology in the Neotropics:The Miocene fauna of La Venta, Colombia. Wash-ington, D.C.: Smithsonian Institution Press, 592 pp.

Legendre, S. 1984. Identification de deux sous-genres etcomprehension phylogenique du genre Mormopterus(Molossidae, Chiroptera). Comptes Rendus del’Academie de Sciences, Serie II, Paris 298:715–720.

. 1985. Molossides (Mammalia, Chiroptera) cen-ozoıques de l’Ancien et du Nouveau Monde; statutsystematique; integration phylogenique des donnees.Neues Jahrbuch fur Geologie und Palaontologie,Abhandlungen 170:205–227.

Leonardos, O.H. 1924. Os folhelhos petrolıferos do Valedo Paraıba. Engenharia, Rio de Janeiro 1(3):21–24.

Mezzalira, S. 1964. Novas ocorrencias de vegetais fosseiscenozoicos no Estado de Sao Paulo. IGG [InstitutoGeografico e Geologico], Sao Paulo 15:73–94(1961–1962).

. 1966. Os fosseis do Estado de Sao Paulo, SaoPaulo: Instituto Geografico e Geologico, Boletim, 45:1–132.

. 1989. Os fosseis do Estado de Sao Paulo, 2nd ed.revisao atual. Sao Paulo: Secretaria do Meio Am-biente, Instituto Geologico, Serie Pesquisa 1989, 142pp.

Paula Couto, C. de. 1956. Une chauve-souris fossile desargiles feuilletees Pleistocenes de Tremembe, Etat deSao Paulo (Bresil). Actes du IV Congres Internation-al du Quaternaire, Roma-Pise, aout-septembre 19531:343–347.

. 1983. Geochronology and paleontology of the ba-sin of Tremembe-Taubate, State of Sao Paulo. Iher-ingia, serie Geologia, Porto Alegre 8:5–31.

Paula Couto, C. de, and S. Mezzalira. 1971. Nova con-ceituacao geocronologica de Tremembe, Estado deSao Paulo, Brasil. Anais da Academia Brasileira deCiencias 43(Suppl.):473–488.

Smith, T. 1995. Presence du genre Wyonycteris (Mam-malia, Lipotyphla) a la limite Paleocene–Eocene enEurope. Comptes Rendus de l’Academie des Scienc-es, Paris 321:923–930.

. 1997. Les insectivores s.s. (Mammalia, Lipotyph-la) de la transition Paleocene–Eocene de Dormaal(MP 7, Belgique): Implications biochronologiques etpaleogeographiques. Memoires et Travaux del’E.P.H.E., Institut Montpellier 21:687–696.



Enigmatic Mammal from the Paleogene of Peru

Francisco J. Goin,1,3 Emma Carolina Vieytes,2

Marıa Guiomar Vucetich,1 Alfredo A. Carlini,1 and Mariano Bond1

ABSTRACT. We describe an isolated molariform tooth excavated from ?Eocene de-posits at Santa Rosa, Peru. The specimen most probably belongs to a mammal, and itmay correspond to a left upper last molar. Analysis of its enamel and dentine micro-structure leads to no definite conclusion as to its relationships, or even as to its enamel–dentine structural patterns. Taphonomic processes, for example, digestion of the speci-men by a predator, might be involved in some of its peculiar characteristics. Its uniquegross morphology prevents us from assigning it to any known mammal group withconfidence. Several features of its cusp and crest morphology resemble the general pat-tern of Late Cretaceous–Eocene gondwanatherian mammals, especially that of Ferug-liotherium windhauseni (Ferugliotheriidae).

RESUMEN. Se describe un molariforme aislado procedente de niveles paleogenos dela localidad de Santa Rosa, Peru. El ejemplar pertenece mas probablemente a un ma-mıfero, y podrıa corresponder a un ultimo molar superior izquierdo. El analisis de sumicroestructura no permite arribar a conclusiones definitivas sobre sus relaciones, osiquiera sobre el patron estructural del esmalte y la dentina. Se sugiere que procesostafonomicos, como la digestion del ejemplar por un depredador, podrıan estar relacio-nados con algunas de sus peculiares caracterısticas. Su morfologıa general, verdadera-mente unica, impide asignarlo a cualquiera de los grupos conocidos de mamıferos; variosrasgos de sus cuspides y crestas recuerdan mas a la estructura dentaria de los mamıferosGondwanatheria, del Cretacico Superior–Eoceno, especialmente a la de Ferugliotheriumwindhauseni (Ferugliotheriidae).

INTRODUCTION

Among the thousands of vertebrate specimens fromthe Paleogene of Santa Rosa, Peru (see Campbell,2004), there is a single, multicuspate molariformtooth of uncertain allocation. The specimen is veryodd in that its external surface lacks the character-istic brightness of enamel; some of the cusps’ inter-nal slopes, as well as the external surface, are ir-regularly perforated by small holes and grooves,and its morphology is sufficiently strange to pre-clude any definite assignment to any known groupof mammals. In spite of its poor state of preserva-tion, its unusual morphology merits description.The purpose of this paper is to describe the newspecimen and to discuss its most relevant featuresand possible affinities.

1 Departamento Paleontologıa Vertebrados, Museo dela Plata, Paseo del Bosque s/n, 1900 La Plata, Argentina.Email: [email protected], [email protected], [email protected],[email protected]

2 Departamento Zoologıa Vertebrados, Museo de laPlata, Paseo del Bosque s/n, 1900 La Plata, Argentina.Email: [email protected]

3 Correspondence

SYSTEMATICS

Mammalia IndeterminateFigures 1–4

MATERIAL. LACM 149371. Isolated molari-form tooth lacking its roots. Natural History Mu-seum of Los Angeles County, Vertebrate Paleontol-ogy Collections. LACM 149371 was picked fromconcentrate obtained during the second, or 1998,field expedition to Santa Rosa. Fieldwork was car-ried out by Kenneth E. Campbell, Jr., Carl D. Frail-ey, and Manuel Aldana.

MEASUREMENTS. Maximum length 5 2.65mm; maximum width 5 2.20 mm; labial height 51.05 mm; lingual height 5 1.30 mm.

LOCALITY, GEOLOGY, AND AGE. The SantaRosa fossil locality is located about 2 km north ofthe village of Santa Rosa, on the west bank of theRıo Yurua near its confluence with the Rıo Beu, inProvincia Atalaya, Departamento Ucayali, easternPeru (98299390 S, 728459480 W; Campbell et al.,2004). LACM 149371 comes from fine- to coarse-grained bed load or channel deposits that probablybelong to the Yahuarango Formation [also knownas, or included within, the Huayabamba Forma-tion, lower red beds, lower Capas Rojas, lowerPuca, or Huchpayacu Formation (Campbell et al.,2004)]. The age of the Santa Rosa local fauna has

146 m Science Series 40 Goin et al.: Enigmatic Mammal

Figure 1 Stereopairs of LACM 149371. A, Occlusal view; B, occlusal-lingual view. (See Fig. 2 for postulated spatialorientation of the specimen).

been estimated as ?Eocene (Mustersan South Amer-ican Land Mammal Age) on the basis of the stageof evolution of the marsupials (Goin and Candela,2004) and rodents (Frailey and Campbell, 2004).

GROSS MORPHOLOGY. For descriptive pur-poses only, cusps have been designated with capitalletters, A to F, with the tips of the cusps in a dorsalposition. The nomenclature for these cusps, as wellas the presumed labiolingual and anteroposterioraxes of the tooth, are shown in Figure 2A. We can-not be positive that these axes are correct, but theywill be assumed to be correct for the purpose of thedescriptions that follow. When LACM 149371 isobserved under the stereomicroscope, the opaque,porous aspect of the whole crown is noticeable. Thecharacteristic brightness of an enamel surface is notobservable at all, except for small vitreous patchesat the tips of some cusps. Another unusual featureof the crown is the irregular presence of holes andgrooves on its surface. They are scattered along the

external faces of the crown, and several small, cir-cular holes are also present on the inner slopes ofthe cusps, as well as on the intercusp valleys (Fig.3F).

In occlusal view (Figs. 1A–B, 2A, 3A), the spec-imen has a triangular shape. It has six or seven pe-ripheral cusps, with contiguous cusps connected toeach other by low ridges. The cusps surround twodeep fossae, as well as a third, much smaller andshallower fossa. The fossae are centrally locatedand separated from each other by crests. The labialface of this tooth (Fig. 3B) is low, whereas the othertwo become progressively higher toward the lingualcorner (Fig. 3C). No cingulum of any kind is ob-servable in occlusal or any other view of the crown.

The crown pattern is complex. Most of the cuspslopes that face other cusps are smooth and ratherrounded, whereas those facing the center of thecrown are steep and flat. The most conspicuous fea-ture of the intercusp morphology is that deep val-

Science Series 40 Goin et al.: Enigmatic Mammal m 147

Figure 2 LACM 149371. A, Occlusal view; B, anterior view. Features labeled in A are the following: A–F, cusps mentionedin the text; af, anteriormost fossa; if, intermediate fossa; pf, posteriormost fossa; iv, intercusp valley. In B, the followingareas of the crown were examined for the microstructure analysis: 1, basal crack on the anterolingual edge of the tooth;2, tip of cusp E (see Figs. 3A–B); 3, area left of the crack on the anterolingual face of the crown (see Figs. 3C, E, F); 4,area above the crack noted as site 1 (see Fig. 3D); 5, area below the same crack (see Fig. 3G); 6, intercusp valley betweencusps E and F (see Fig. 3H).

leys on the external surface of the crown separatecontiguous cusps, but low ridges or crests internalto these valleys connect them.

Cusp A, the largest and highest cusp of the tooth,is triangular in occlusal view, with its posterior faceflat and sloping steeply ventrad (Fig. 2). Cusp B is

more rounded in section; its anterior face is slightlyconcave. A deep, external valley separates cusps Aand B. Lingual to this valley is a low ridge con-necting both cusps and surrounding the anteriorfossa at its anterolabial corner. Cusp B projects pos-teriad with a relatively long, moderately sloped


Figure 3 LACM 149371. A, Occlusal view; B, labial view; C, lingual view; D, anterolingual view; E, posterolingual view.Scale for A–E is the same. F, Detail of the anterolingual face of the crown shows the small holes, or dissolution points,that penetrate the tooth’s surface. The anterolingual face shown in D differs from that of Figure 1B, in that, in the latterphotograph, the base of the crown has been polished to the left of the crack.

crest that reaches the most posterior fossa, whichis the smallest of the fossae. Posterior to this fossais cusp C, the lowest of the tooth, which is subcir-cular in section. Its anterior slope, which faces thesmall fossa, is almost flat. Posterolingually, cusp C

is separated from cusp D by a deep valley and itslabial face is marked by a deep groove.

Cusp D is very odd, and may, in fact, representtwo subtriangular cusps fused at their bases. Thisstructure is crest shaped, and it surrounds the me-


dial fossa of the tooth lingually. Two crests origi-nating from cusp D separate the three fossae. Themost anterior crest, the longest and deepest one,almost reaches the base of cusp B. Cusp E is locatedat the lingual corner of the tooth; it is subtriangularin cross section. Its anterolabial face is almost flat,and it borders a deep valley separating it from cuspF. Cusp E is separated from cusps D and F bymeans of external valleys and connected to them byinternal crests. Cusp F is rounded in section, and itborders the largest fossa anteriorly.

The three aligned, successively smaller and shal-lower fossae are subcircular in section (Figs. 1A,2A). The largest is almost centrally located, and itis surrounded by four of the five cusps. The mostposterior fossa is extremely small, shallow, and bor-dered posteriorly by cusp C.

Ventrally, the crown and roots are broken. Pulpcavities for four roots can be distinguished. Thesecavities are large and subcircular in section. One ofthem, more suboval in profile than the others,seems to be filled with sediment. The two anteriorroots, as well as the two posterior roots, seem tobe partially confluent with each other. A muchsmaller cavity, located medially at the labial side ofthe tooth, suggests the probable existence of a fifthroot. Finally, there is a slight depression at the an-terolabial corner of the tooth; whether it representsan additional root is impossible to determine.

ENAMEL AND DENTINE MICROSTRUC-TURE. We analyzed the specimen to determine itsmicrostructure, but without embedding it in artifi-cial resin because of its unique condition. Treat-ment was restricted to three areas of the crown: (1)on a large crack located on the anterior face, (2) atthe tip of cusp E, and (3) the ventral face (see Fig.2B). The first and third areas were polished withgrit powder no. 1000 and then etched for 5–6 sec-onds with 10 percent HCl to make morphologicaldetails visible. Acid was applied to the tip of thecusp without polishing.

The results of the microstructure analysis wereinconclusive. Most examined areas of the tooth(anterior face, tip of cusp, and ventral face) have ahomogeneous internal structure. At intermediatemagnifications (3750–1000), we could distinguishflat, tubule-like structures that are parallel to eachother and rise toward the external surface at anangle of approximately 45 degrees (Fig. 4G). Ad-ditionally, either prisms or prism-like structures, to-gether with interprismatic structures, were ob-served (Figs. 4A–F). At higher magnifications(33500–10,000), however, neither the prisms northe matrix crystals could be distinguished. In trans-verse view, prism-like structures were observed; inthis view, they are suboval in section (Fig. 4A).Again, at higher magnifications, the crystallinestructure of these forms was not distinguishable(Fig. 4B). All structures seemed to be degraded tosome extent, making it impossible, at least on thepolished surfaces, to obtain more precise observa-tions. On an unpolished, but acid-treated surface of

the crown (equivalent to a tangential plane), theobserved pattern is reminiscent of Hunter–Schregerbands of the enamel of mammals (Fig. 4C). How-ever, this pattern is not uniform across the surfaceof the crown. In other areas, the observed patternwas that of the prism-like structures. In short,enamel structures are not unequivocally distin-guishable. In none of the examined areas of thetooth could we observe a dentine–enamel junction.

DISCUSSION

LACM 149371 is enigmatic in many ways. Re-garding its taxonomic assignment, the most plau-sible hypothesis for us is that it corresponds to amammal. Its crown is sufficiently complex to elim-inate the possibility that it may belong to anothervertebrate group, including several with complex,mammal-like postcanine dentitions (e.g., croco-diles; see Clark et al., 1989). The complexity of thecrown suggests that it is a molar. Its triangularshape, root number (four or five), and cusp dispo-sition suggest that most probably it is an upper mo-lar. The flat aspect and low crown of its longest facesuggest that it corresponds to the labial face (Fig.3B). In turn, the dramatic narrowing of the crowntoward the smallest of the cusps (C) suggests thatit may correspond to the posterior end of the tooth.In short, the tooth may be a last upper left molar.

Several features suggest some degree of chemicaldegradation of both the general morphology andmicrostructure of this specimen. The dull, porousaspect of the crown, as well as the absence of pris-matic structures at various points, suggests that theenamel is partially lacking or that it is highly mod-ified. The grooves and holes that penetrate severalareas of the crown surface are suggestive of disso-lution points. Taphonomic processes may be in-volved in the apparent degradation of these struc-tures. One possibility is that the specimen was de-posited under the influence of acidic undergroundwater (see, e.g., Fernandez Javo et al., 1998). Sev-eral patterns of molar digestion by avian predatorshave been proposed (Andrews, 1990). Some canidsand felids also digest small mammalian molars tothe same, or even greater, degree of destruction.Crocodilian digestion of mammalian teeth is alsoknown as a factor producing enamel denudation(Fisher, 1981). If this were the case for LACM149371, it would explain the unusual features not-ed above. Although we were unable to observe suchpreservation anomalies among available marsupialremains from the same matrix sample, other spec-imens of teeth from the Santa Rosa collection doshow evidence of acid etching and enamel denu-dation (K. Campbell, personal communication). Atleast some of the observed microstructure featuresof the tooth could correspond to dentine, not toenamel. It has been noted that in some mammals,such as gondwanatherians (Sigogneau Russell et al.,1991), the dentine tubules can be filled with calci-fied, dense, cylindrical structures.


Figure 4 Scanning electron micrographs showing different microstructure patterns in LACM 149371. A, Detail of cuspE showing ?enamel prisms emerging from the ?interprismatic matrix at the cusp surface; B, detail of A; C, unpolishedarea left of the crack on the anterolingual face, showing a pattern similar to that of Hunter–Schreger bands; D, ?enamelprisms, parallel to each other, above the crack on the anterolingual face of the crown; E, polished area to the left of the


←

crack on the anterolingual face, showing prism- and matrix-like structures set almost at right angles; F, detail of areadepicted in E, showing ?crystallites; G, ?filled dentine tubules below the crack on the anterolingual face; H, detail of the(unpolished) surface of the intercusp valley between cusps E and F.

The unusual shape of the valleys and ridges sep-arating and connecting the cusps, especially thethree anterior cusps (A, F, and E; Fig. 3D), remindus of the condition seen in several allotherians. Inthe multituberculate Essonodon browni Simpson,1927, for instance, the cusp arrangement of M2 isnot that of aligned cusps, but one of peripheralcusps around the crown; several of these cusps areanteroposteriorly compressed. Other multitubercu-lates, such as Meniscoessus robustus Cope, 1882,show a similar pattern in the intercusp relationship(external valleys, internal ridges; see, e.g., Clemens,1963; Archibald, 1982). None of these, however,show any indication of fossae between cusps.Moreover, molars of multituberculates are usuallyquadrangular, not triangular in shape, and theyusually show a longitudinal valley between twoparallel cusp series.

The same cusp features mentioned above forLACM 149371 remind us of the pattern observableamong gondwanatherian mammals. Originally re-garded as multituberculates (see, e.g., Krause et al.,1992; Kielan-Jaworowska and Bonaparte, 1996),gondwanatherians were considered more recentlyby Pascual et al. (1999) to be Mammalia incertaesedis. In particular, the cusp shape and intercuspmorphology of LACM 149371 resembles that ofFerugliotherium windhauseni Bonaparte, 1986a(Gondwanatheria, Ferugliotheriidae), from the up-per Cretaceous Los Alamitos Formation in north-ern Patagonia. No upper molars of these gondwan-atherians are known; thus, it is impossible to makeany direct comparison with LACM 149371. None-theless, it should be noted that several cusps in Fer-ugliotherium Bonaparte, 1986a, are anteroposteri-orly compressed and have low ridges or crests thatconnect with the center of the crown, as in the Pe-ruvian mammal. Ferugliotherium has only incipientfossae because it has quite brachydont molariformteeth. Other gondwanatherians, such as the Sud-americidae, include forms (Sudamerica ScillatoYane and Pascual 1985, Gondwanatherium Bona-parte, 1986b) with high-crowned molariform teeth.Interestingly, Sudamerica ameghinoi Scillato Yaneand Pascual, 1985, from the lower Paleocene of Pa-tagonia, has two to three fossae in its molariformteeth. Even though they are not very deep, they areclearly observable in lightly worn teeth (Koenig-swald et al., 1999).

Comparisons with early tribosphenic mammals(early Australosphenida and Boreosphenida sensuLuo et al., 2001, 2002) lead to no definite conclu-sion other than their sharing of triangular-shapedupper molars. Among early australosphenidans,upper molars of Ambondro Flynn et al., 1999, andSteropodon Archer et al., 1985, are unknown,

whereas those of Ausktribosphenos Rich et al.,1997, and more modern monotremes are definitelyaway from the pattern of the Peruvian mammal.The same can be said of its comparison with earlyboreosphenidans, which include Metatheria, Euthe-ria, and ‘‘tribotheres’’ (see Luo et al., 2002). All ofthese taxa show a more ‘‘typical’’ tribosphenic ar-rangement of cusps, crests, notches, basins, and cin-gula (see below).

No metatherian mammal known to us matchesthe morphology of LACM 149371. Even if we as-sume that the natural orientation of the tooth wasas suggested in Figure 2A, the intercusp morphol-ogy is not observable in fossil or living metatheri-ans. Neither are deep fossae limited by shallow cris-tae transverse to the main axis of the tooth seen inknown metatherians. Borhyaenoids, microbiother-ians, didelphoids, caenolestoids, and polydolopi-morphians of the families Bonapartheriidae, Prepi-dolopidae, and Gashterniidae have well-knownmolar patterns, none of which match that ofLACM 149371 (see, e.g., Marshall, 1978, 1980,1981, 1982a, 1982b; Pascual, 1980; Pascual andBond, 1981). Finally, none of the high-crownedmolars of taxa of the Patagoniidae and Argyrolag-idae have fossae of any kind (Simpson, 1970).

Assignment of this specimen to a still unknown,early edentate seems to us equally speculative onthe basis of the available evidence. Toothed eden-tates are either homodont and simple-crowned(e.g., armadillos, sloths), or they have complicatedcrowns without peripheral cusps surrounding cen-tral fossae (e.g., glyptodonts). There is no evidenceof multicuspate crowns among dasypodoids, whosemolars are suboval in section and have simplecrowns. Among glyptodontoids, even though thelobed aspect of their teeth could suggest a primitivemulticusp pattern, their size, hypsodonty, andcrown pattern are far from that of LACM 149371.Vermilinguans (Pilosa) lack teeth, either among ex-tinct forms or as early (fetal or abortive) dentitions.Unworn teeth of early tardigradans (Pilosa) are sin-gle-cusped, and their crowns are subelliptical, re-niform, or subrectangular in section (see Hoffstet-ter, 1958, 1982; Paula Couto, 1979).

Some early South American native ‘‘ungulates’’match the size of LACM 149371. Moreover, thistooth does remind us, in a very general sense, ofsome upper premolars of ‘‘condylarths’’ and basallitopterns, as well as notioprogonid and typotheriidnotoungulates (Cifelli, 1983; Muizon and Cifelli,2000; Soria, 2001). It could be the case that thisspecimen is actually an upper right molariformtooth of some unknown form of these ungulates,with an anterolabially salient parastyle. If this werethe case, however, cusp A (posterolabial, or ‘‘meta-


cone’’ in this interpretation) differs from that of allknown ungulates in being transversely orientedwith respect to the anterior ‘‘paracone.’’ Eventhough other features also remind us of some un-gulates, the presence of a ‘‘metacone’’ makes itmarkedly different from all of them.

Rodent affinities for this specimen are also highlyspeculative, even though some resemblances withearly forms can be noted. One of these is the earlyEocene Ivanantonia efremovi Shevyreva, 1989,from Mongolia, which is of uncertain affinitiesamong basal rodents. Lower molars of IvanantoniaShevyreva, 1989, show some anteroposterior align-ment of the cusps that, in its case, border a large,central groove. Hartenberger et al. (1997) suggest-ed affinities between Ivanantonia and the late Eo-cene, North American Nonomys Emry and Daw-son, 1972. Interestingly, Nonomys shows in its an-teriormost cheek tooth (M1) a parastylar area (‘‘an-terocone lobe’’ sensu Emry, 1981) that is wellprojected anteriorly. LACM 149371 also shows aprominent projection, which is also interpreted hereas external, but posterior instead of anterior. None-theless, Nonomys lacks the central fossae of the Pe-ruvian mammal, whereas the latter lacks any indi-cation of a cingulum. In none of the other knownearly Tertiary Glires (see, e.g., Dashzeveg and Rus-sell, 1988; Dashzeveg et al., 1998) could we findthe peculiar features observable in LACM 149371.

In short, of all mammals compared with the Pe-ruvian molariform tooth, we regard the Gondwan-atheria, and especially the Ferugliotheriidae, as ourbest estimate as to its possible allocation. If thisproves to be the case, LACM 149371 would rep-resent one of the latest known occurrences of agondwanatherian mammal. The youngest record ofsudamericid, but not of ferugliotheriid, gondwan-atherians was reported from the middle Eocene ofthe Antarctic Peninsula (Reguero et al., 2001).

ACKNOWLEDGMENTS

We thank K.E. Campbell for loaning us LACM 149371for study and for his help with the English version of thispaper. Patricia Sarmiento and Rafael Urrejola were veryhelpful in making the scanning electron micrographs. Thefollowing colleagues gave us useful comments on the stud-ied specimen: Rosendo Pascual, Adriana Candela, JavierGelfo, Claudia Montalvo, and Edgardo Ortiz Jaureguizar.Finally, we thank Zofia Kielan-Jaworowska, ThomasMartin, Malcolm McKenna, and an anonymous reviewerfor their useful criticisms and comments on the manu-script.

LITERATURE CITED

Andrews, P. 1990. Owls, caves and fossils. Chicago: TheUniversity of Chicago Press, 231 pp.

Archer, M., T.F. Flannery, A. Ritchie, and R.E. Molnar.1985. First Mesozoic mammal from Australia—AnEarly Cretaceous monotreme. Nature 318:363–366.

Archibald, J.D. 1982. A study of mammalia and geologyacross the Cretaceous–Tertiary boundary in GarfieldCounty, Montana. University of California Publica-tions in Geological Sciences 122:1–286.

Bonaparte, J.F. 1986a. Sobre Mesungulatum houssayi ynuevos mamıferos cretacicos de Patagonia, Argenti-na. Actas IV Congreso Argentino de Paleontolgıa yBioestratigrafıa (Mendoza) 2:48–61.

. 1986b. A new and unusual Late Cretaceous mam-mal from Patagonia. Journal of Vertebrate Paleon-tology 6:264–270.

Campbell, K.E., Jr. 2004. Introduction. In The Paleogenemammalian fauna of Santa Rosa, Amazonian Peru,ed. K.E. Campbell, Jr. Natural History Museum ofLos Angeles County, Science Series 40, 1–2.


Cifelli, R.L. 1983. The origin and affinities of the SouthAmerican Condylarthra and Early Tertiary Litopter-na (Mammalia). American Museum Novitates 2772:1–49.

Clark, J.M., L.L. Jacobs, and W. Downs. 1989. Mammal-like dentition in a Mesozoic crocodylian. Science244:1064–1066.

Clemens, W.A., Jr. 1963. Fossil mammals of the typeLance Formation, Wyoming. Part I. Introductionsand Multituberculata. University of California Pub-lications in Geological Science 48:1–105.

Cope, E.D. 1882. Mammalia in the Laramie Formation.American Naturalist 16:830–831.

Dashzeveg, D., J.-L. Hartenberger, T. Martin, and S. Le-gendre. 1998. A peculiar minute Glires (Mammalia)from the Early Eocene of Mongolia. In Dawn of theage of mammals in Asia, ed. C. Beard and M.R.Dawson. Bulletin of the Carnegie Museum of Nat-ural History 34:194–209.

Dashzeveg, D., and D.E. Russell. 1988. Paleocene and Eo-cene Mixodontia (Mammalia, Glires) of Mongoliaand China. Palaeontology 31:129–164.

Emry, R.J. 1981. New material of the Oligocene muroidrodent Nonomys, and its bearing on muroid origins.American Museum Novitates 2712:1–14.

Emry, R.J., and M.R. Dawson. 1972. A unique cricetid(Rodentia, Mammalia) from the early Oligocene ofNatrona County, Wyoming. American MuseumNovitates 2508:1–14.

Fernandez Javo, Y., C. Denys, P. Andrews, T. Williams, Y.Dauphin, and L. Humphrey. 1998. Taphonomy andpalaeoecology of Olduvai Bed—I (Pleistocene, Tan-zania). Journal of Human Evolution 34:137–172.

Fisher, D. 1981. Crocodilian scatology, microvertebrateconcentrations and enamel-less teeth. Paleobiology7:262–275.

Flynn, J.J., M. Parrish, B. Rakotosamimanana, W.F. Simp-son, and A.R. Wyss. 1999. A middle Jurassic mam-mal from Madagascar. Nature 401:57–60.


Goin, F.J., and A.M. Candela. 2004. New Paleogene mar-supials from the Amazon Basin of eastern Peru. InThe Paleogene mammalian fauna of Santa Rosa,Amazonian Peru, ed. K.E. Campbell, Jr. Natural His-


tory Museum of Los Angeles County, Science Series40, 15–60.

Hartenberger, J.-L., D. Dashzeveg, and T. Martin. 1997.What is Ivanantonia efremovi (Rodentia, Mamma-lia)? Palaontologische Zeitschrift 7:135–143.

Hoffstetter, R. 1958. Xenarthra. In Traite de Paleontolo-gie, ed. J. Piveteau, 6(2):535–636. Paris: Masson &Cie.

. 1982. Les Edentes Xenarthres, un groupe singu-lier de la faune Neotropicale. In Proceedings of thefirst meeting on paleontology, essential of historicalgeology (Venice, 1981), ed. E. Montanaro Gallitelli,385–443. Modena: S.T.E.M. Mucchi.

Kielan-Jaworowska, Z., and J.F. Bonaparte. 1996. Partialdentary of a multituberculate mammal from the LateCretaceous of Argentina and its taxonomic impli-cations. Revista del Museo Argentino de CienciasNaturales ‘‘Bernardino Rivadavia’’, Nueva Serie 145:1–9.

Koenigswald, W.v., F.J. Goin, and R. Pascual. 1999. Hyp-sodonty and enamel microstructure in the Paleocenegondwanatherian mammal Sudamerica ameghinoi.Acta Palaeontologica Polonica 44:263–300.

Krause, D.W., Z. Kielan-Jaworowska, and J.F. Bonaparte.1992. Ferugliotherium Bonaparte, the first knownmultituberculate from South America. Journal ofVertebrate Paleontology 12:351–376.

Luo, Z.-X., R.L. Cifelli, and Z. Kielan-Jaworowska. 2001.Dual origin of tribosphenic mammals. Nature 409:53–57.

Luo, Z.-X., Z. Kielan-Jaworowska, and R.L. Cifelli. 2002.In quest for a phylogeny of Mesozoic mammals.Acta Palaeontologica Polonica 47:1–78.

Marshall, L.G. 1978. Evolution of the Borhyaenidae, ex-tinct South American predaceous marsupials. Uni-versity of California Publications in Geological Sci-ences 117:1–89.

. 1980. Systematics of the South American marsu-pial family Caenolestidae. Fieldiana, Geology, NewSeries 5:1–145.

. 1981. Review of the Hathlyacyninae, an extinctsubfamily of South American ‘‘dog-like’’ marsupials.Fieldiana, Geology, New Series 7:1–120.

. 1982a. Systematics of the South American mar-supial family Microbiotheriidae. Fieldiana, Geology,New Series 10:1–75.

. 1982b. Systematics of the extinct South Americanmarsupial family Polydolopidae. Fieldiana, Geology,New Series 12:1–109.

Muizon, C. de, and R.L. Cifelli. 2000. The ‘‘condylarths’’(archaic Ungulata, Mammalia) from the early Paleo-

cene of Tiupampa (Bolivia): Implications on the or-igin of the South American ungulates. Geodiversitas22:47–150.

Pascual, R. 1980. Nuevos y singulares tipos ecologicos demarsupiales extinguidos de America del Sur (Paleo-ceno tardıo o Eoceno temprano) del Noroeste argen-tino. II Congreso Argentino de Paleontologıa yBioestratigrafıa y I Congreso Latinoamericano dePaleontologıa, Buenos Aires, Actas 2:151–173.

Pascual, R., and M. Bond. 1981. Epidolopinae, subfam.nov. de los Polydolopidae (Marsupialia, Polydolo-poidea). II Congresso Latino-Americano de Paleon-tologıa, Porto Alegre, Anais 1981:479–488.

Pascual, R., F.J. Goin, D.W. Krause, E. Ortiz Jaureguizar,and A.A. Carlini. 1999. The first gnathic remains ofSudamerica: Implications for gondwanathere rela-tionships. Journal of Vertebrate Paleontology 19:373–382.

Paula Couto, C. de. 1979. Tratado de Paleomastozoolo-gia. Rio de Janeiro: Academia Brasileira de Ciencias,590 pp.

Reguero, M.A., S.A. Marenssi, and S.N. Santillana. 2001.Antarctic Peninsula and South America (Patagonia)Paleogene terrestrial faunas and environments: Bio-geographic relationships. Ameghiniana 38(Suppl.):16R–17R.

Rich, T., P. Vickers-Rich, A. Constantine, T.F. Flannery, L.Kool, and N. van Klaveren. 1997. A tribosphenicmammal from the Mesozoic of Australia. Science278:1438–1442.

Scillato Yane, G.J., and R. Pascual. 1985. Un peculiar Xe-narthra del Paleoceno medio de Patagonia (Argen-tina). Su importancia en la sistematica de los Parath-eria. Ameghiniana 21:173–176.

Shevyreva, N.S. 1989. New rodents (Ctenodactyloidea,Rodentia, Mammalia) from the Lower Eocene ofMongolia. Paleontological Journal 23(3):57–68.

Sigogneau Russell, D., J.F. Bonaparte, R.M. Frank, and V.Escribano. 1991. Ultrastructure of dental hard tis-sues of Gondwanatherium and Sudamerica (Mam-malia, Gondwanatheria). Lethaia 24:27–38.

Simpson, G.G. 1927. Mammalian fauna of the Hell CreekFormation of Montana. American Museum Novita-tes 267:1–7.

. 1970. The Argyrolagidae, extinct South Americanmarsupials. Bulletin of the Museum of ComparativeZoology 139:1–86.

Soria, M.F. 2001. Los proterotheriidae (Litopterna, Mam-malia), sistematica, origen y filogenia. Monografıasdel Museo Argentino de Ciencias Naturales 1:1–167.



The Santa Rosa Local Fauna: A Summary

Kenneth E. Campbell, Jr.1

As noted by Patterson and Pascual (1972:249) overthree decades ago, ‘‘The [South American Tertiary]record has one great deficiency: it is very largelyrestricted to the southern parts of the continent.’’Although this deficiency has been reduced some-what for the Neogene of northern South Americaby progress in recording paleofaunas at La Venta,Colombia (Kay et al., 1997) and in western Ama-zonia (Frailey, 1986; Campbell et al., 2000), for thePaleogene record, it remains little changed. No low-latitude, tropical paleofauna had been recordedfrom the Paleogene prior to the Santa Rosa localfauna, and the geographically closest Paleogene lo-calities are those at Laguna Umayo, Peru (latestCretaceous or earliest Paleocene); Tiupampa, Boliv-ia [early Paleocene, Tiupampian South AmericanLand Mammal Age (SALMA)]; and Salla and La-cayani, Bolivia (late Oligocene, Deseadan SALMA;see Goin and Candela, 2004; fig. 1). Thus, beforethe discovery of the Santa Rosa locality, well overone-half of the South American continent was un-represented by Paleogene vertebrate faunas. For re-views of South American faunas and their ages, seePascual and Ortiz-Jaureguizar (1990), Marshalland Sempere (1993), and Flynn and Swisher(1995).

The Santa Rosa local fauna, with its abundantand diverse fossil assemblage, is a dramatic firstlook at Paleogene faunas of low-latitude, low-ele-vation, tropical South America. This fauna has pre-sented a wonderful opportunity for understandingtropical Paleogene faunas, but at the same time, be-cause there are no other benchmarks against whichto place the Santa Rosa local fauna, many limita-tions are placed on our interpretations of it. Per-haps the most important limiting factor is the lackof an accurate age for the paleofauna (see ‘‘Age ofthe Santa Rosa Local Fauna’’). Dating faunas bythe stage of evolution of their constituent taxa, aswas done for the mammals from Santa Rosa, canonly provide an approximate age, and it has a po-tential for a large margin of error, especially whenthere is only one Paleogene fauna for one-half acontinent. Without an accurate age for the paleo-fauna, it is not possible to place it within an overallscheme of evolutionary patterns for South Ameri-can paleofaunas with confidence.

Another important limiting factor is that becauseso many of the taxa are unique, it is difficult toknow where to place them in their respective line-ages. For the Santa Rosa local fauna, this difficulty


applies to taxa within well-known mammaliangroups, such as the marsupials (Goin and Candela,2004), and even to taxa within the larger groupMammalia (Goin et al., 2004). This factor compli-cates interpretation of the paleofauna as a wholebecause it is not known whether these unique taxaare a result of isolation, environmental stability inthe tropics, or some other reason.

I review here some of the most significant aspectsof the Santa Rosa local fauna as reported in thisvolume and present a perspective of its role in ourunderstanding of Paleogene tropical South Ameri-ca.

GEOLOGY

To anyone inexperienced in Amazonian geology,one single word descriptor would probably cometo mind: monotonous. Red clays, silts, and finesands abound, and few stratigraphic horizons standout as being significantly different from all others.To the practiced eye, however, subtle nuances doappear that allow differentiation of formations(Ruegg, 1947, 1956; personal observation). Furthercomplicating the interpretation of stratigraphy inlowland Amazonia is the fact that long stratigraph-ic sections are nonexistent, and only intermittentoutcrops along rivers provide a look at the stratig-raphy underlying the vast cover of Amazonian veg-etation.

The stratigraphy along the Rıo Yurua describedby Campbell et al. (2004) follows the typical pat-tern in lowland Amazonia. Below the upper Mio-cene–lower Pliocene Madre de Dios Formation,which is unusual among Amazonian formations inthat it does have a readily recognizable stratigraphy(Campbell et al., 2000), are exposed, moderatelyconsolidated, massive red clays and mudstoneswith thin horizons of less consolidated fine tocoarse sands. Thin conglomeratic lenses are alsocommon, consisting of calcareous concretions, clayballs, fragments of bone, silicified wood, and coarsesand with a matrix of clay. It was the latter type ofdeposit that contained the fossil vertebrates record-ed at Santa Rosa, Peru.

Identification of the Paleocene–Eocene Yahuar-ango Formation as the fossil-bearing unit at SantaRosa was based on the presumed Eocene age of thefossils, which in turn was based on the stage ofevolution of the most abundant mammals in thelocal fauna. As noted by Campbell et al. (2004),however, there are possible problems with this for-mational assignment, and additional stratigraphicdata are required before it can be confirmed. Moredetailed stratigraphic studies along the Rıo Yurua

156 m Science Series 40 Campbell: Summary

might provide these critical data, but it is possiblethat the formation producing the Santa Rosa localfauna will remain in question until a strong meansof correlation between the Santa Rosa local faunaand other faunas still to be found in areas wherethe stratigraphic sequence can be determined withconfidence.

MARSUPIALS

As a measure of the uniqueness of the marsupialpaleofauna from Santa Rosa, it should be notedthat Goin and Candela (2004) recognize eleven newspecies in ten genera (eight new), with all majorlineages of South American marsupials represented.In and of itself, finding so many new species in aregion previously unrepresented in the paleontolog-ical record is not surprising. In this case, however,only two of the new species are referable to previ-ously described genera, two new genera could notbe referred to known families, and an additionaltwo genera were only tentatively referred to knownfamilies. Furthermore, two new genera could notbe referred to any known family or order, which israther exceptional for South American marsupials.With no other tropical, low-latitude Paleogene fau-nas available, however, it is not possible at this timeto assign a level of significance to the uniquenessof this paleofauna. That is, there is a question as towhether the Santa Rosan marsupial paleofauna isrepresentative of Paleogene marsupial faunas ofAmazonia that were dramatically different fromcontemporary paleofaunas of higher latitudes, orwhether it represents only a temporal or geographicanomaly.

Goin and Candela (2004) describe two factorsthat have limited our knowledge about SouthAmerican marsupials, and Paleogene marsupials inparticular. One is the biased geographic distributionof Paleogene faunas of South America, as notedabove. The other is that most living marsupials areof small to very small size, an attribute that pre-sumably pertained to most extinct members of theMarsupialia in South America. Indeed, most of themarsupial specimens from Santa Rosa are very tinyspecimens. With the recent application of screen-washing techniques, such as those employed at San-ta Rosa to fossil-bearing sediments in South Amer-ica, an increasingly large number of new taxa ofmicrovertebrates are being recovered. Although thefossils recovered in this manner are often primarilyisolated teeth, these specimens still provide a wealthof new information about the species represented.

The Santa Rosan marsupials are quite informa-tive regarding the early phases of several marsupiallineages. For example, in their discussion of the‘‘opossum-like’’ marsupial Rumiodon inti, Goinand Candela (2004) identify and describe in detaila series of characters that separate the herpetoth-eriids (Herpetotheriidae) from didelphoids (Didel-phidae). Their analysis supports the concept of tem-porally separate radiations for the two groups (ear-

ly Tertiary for herpetotheriids; Neogene for didel-phoids).

Among the polydolopimorphians are what ap-pear to be very early representatives of that group.Incadolops ucayali is the most generalized prepi-dolopid (Polydolopimorphia, Prepidolopidae) cur-rently known. Hondonadia pittmanae is one of thetwo polydolopimorphian taxa that cannot beplaced to family, but it appears to be related at thegeneric level to marsupials of southern South Amer-ica. It is placed in a new taxon, the superfamilyBonapartherioidea, family indeterminate. Wamra-dolops tsullodon, also a polydolopimorphian, is apossible early representative of the lineage leadingto groeberiids (Groeberiidae), the origin of whichremains a mystery. It is enigmatic in sharing derivedfeatures with several lineages of polydolopimorphi-ans, and it is interpreted as suggesting monophy-letic group status for a variety of polydolopimor-phians. Goin and Candela (2004) describe how W.tsullodon, on the one hand, and the paucitubercu-latans Perulestes cardichi, P. fraileyi, and Sasawatsumahaynaq, on the other, reveal the details of ac-quisition of the upper molar pattern in the two ma-jor lineages of South American ‘‘pseudodiproto-dont’’ marsupials, the Polydolopimorphia and Pau-cituberculata, respectively. Perulestes cannot be as-signed with confidence to a known family ofPaucituberculata because most of its dental featuresare so generalized, although it is tentatively referredto the family Caenolestidae. Sasawatsu can only bereferred to the Palaeothentidae even more tenta-tively.

Incorporating the new evidence from the SantaRosan marsupials with that from new research onhigh-latitude marsupials, Goin and Candela (2004)propose and argue for a new classification of thePolydolopimorphia. This will surely stimulate newavenues of research into South American marsupi-als. It should also focus attention on the need foradditional Paleogene faunas from Amazonia tocomplement those of higher latitudes.

There are many highlights, in addition to the sys-tematic contributions, in Goin and Candela’s(2004) paper on the Santa Rosan marsupials. Theyfound, for example, that a strikingly large propor-tion of the marsupials were frugivorous or frugiv-orous–insectivorous. This interpretation was basedon structural features of the teeth, such as molarswith low cusps, heterodonty, thick enamel, and areduction of shearing crests, among other charac-ters. The size of these marsupials and their inferreddiet is suggestive of arboreal locomotion, althoughthis remains to be tested with currently unavailablecranial and postcranial material. Interestingly, onetaxon of unknown ordinal affiliation, Wirunodonchanku is so small its possible feeding habits aredifficult to interpret. Although the teeth of this spe-cies share features with opossums of carnivorousfeeding habits, its very small size must have limitedit to small insects or worms. The tiny opossum Ki-

Science Series 40 Campbell: Summary m 157

ruwamaq chisu might also have been limited tofeeding on small insects (Goin and Candela, 2004).

Wamradolops tsullodon was the most abundantmarsupial taxon, making up about 42 percent ofthe total marsupial specimens. Features of the teethof this species, plus its small size, suggested to Goinand Candela (2004) that it was probably mainlyfrugivorous. This species shares several of its mostderived characters with groeberiids, which havebeen interpreted as extremely specialized towardfrugivory, especially food items that are hard toprocess dentally (Simpson, 1970). This might wellbe true, but it might be useful to note that one ofthe most abundant types of insects in tropical for-ests are beetles, which have a characteristically hardcarapace. As pointed out by Simpson (1970), thegroeberiids are very limited in their temporal andspatial distribution, which suggested to him thattheir isolated occurrence, at the time of his writing,in the fossil record at Divisadero Largo, Argentina(Divisaderan SALMA, early Oligocene) resultedfrom a brief interval when their preferred(?) tropi-cal habitat expanded briefly into higher latitudes.

At the other end of the spectrum is the completeabsence of polydolopines (Polydolopimorphia, Po-lydolopidae), which are the best represented mar-supial taxa in Paleogene levels in Patagonia. This isconsidered significant because it appears that thedistribution of these multituberculate-like marsu-pials was also governed by the distribution of theirpreferred habitat. In this case, however, it was tem-perate, not tropical, habitats that were key to theirdistribution. The polydolopines appear to havebeen restricted to South American and Antarcticforests where beech trees, which prefer cool, humidclimates, were dominant and podocarps and arau-carian conifers were common.

The Santa Rosan marsupial paleofauna is alsovery intriguing for information pertaining to paleo-biogeographic reconstructions. For example, if con-firmed as a didelphimorphian, R. inti might be oneof the first South American records of the familyHerpetotheriidae, although work in progress byGoin and colleagues suggests that marsupials ofthis family were present in central Patagonia andeastern Brazil in the Paleocene (Goin and Candela,2004). Herpetotheriids were previously knownfrom lower Tertiary levels in North America, Eu-rope, and northern Africa. They are thought to berelated to North American and European peradec-tids.

Hypotheses of paleobiogeographic reconstruc-tions also must include Wirunodon chanku, orderand family indeterminate, although it might sharecharacters with both peradectians and early didel-phimorphians. This species appears closest to theearly Eocene Kasserionotherium tunisiense Cro-chet, 1986, a ?peradectian of Tunisia. Goin andCandela (2004) suggest that this relationship, ifconfirmed, could be evidence of a biogeographic re-lationship between South America and Africa in theearly Paleogene.

RODENTS

The rodents of the Santa Rosa local fauna representthe earliest known South American rodents. As not-ed by Frailey and Campbell (2004), they are fun-damentally similar to each other, sharing a commonpattern of four principal cusps connected by lowcrests. Available partial mandibles indicate that thespecies they represent were hystricognathous. Dif-ferences between the described genera are subtle,but consistent, and they allow the genera to beplaced into three families: Erethizontidae, Agouti-dae, and Echimyidae. In several cases, it is possibleto identify trends that relate genera from SantaRosa to younger South American taxa. Were it notfor these identifiable trends, it might have been dif-ficult to justify placing the Santa Rosan rodents intomore than a single family.

By coincidence, the number of new rodent taxais the same as that for marsupials: eleven new spe-cies in eight new genera. However, the diversity ofthe rodents is greater because there are at least sixpossible additional species of rodents that were notnamed because the material available was not con-sidered adequate for name-bearing types. Despitethe many similarities among the rodents of SantaRosa, this diversity is considered quite significant.

The Santa Rosan rodents make significant con-tributions to the debate as to whether the ancestralcaviid tooth condition was tetralophodont or pen-talophodont. In reference to this debate, Butler(1985:395) stated, ‘‘The problem [of the ancestralnumber of lophs] will be resolved only when theposition of the metacone is established by the dis-covery of a primitive form that still has traces ofcusps.’’ Were it so simple, the cuspate Santa Rosanrodents would seem to settle the question in favorof the tetralophodont condition because all of theSanta Rosan rodents have a basic tetralophodontpattern, and in only one genus (Eobranisamys) isthis basic pattern altered by the addition of an ac-cessory loph, a neoloph, to form a pentalophodontcondition. The neoloph appears to have formed bythe linear confluence and stabilization of minor cus-pules in the posterior part of the metaflexus, prob-ably together with the development and elongationof crenulations on the anterior surface of the pos-teroloph (Frailey and Campbell, 2004). Other taxaalso have similar cuspules in the metaflexus, butthey are highly variable and do not appear to havestabilized as a crest. Also pertinent to this questionis the fact that in the Santa Rosan rodents the lat-eral two cusps on the upper teeth (paracone andmetacone) and the medial two cusps on the lowerteeth (entoconid and metaconid are characteristi-cally closer to each other than are the cusps on theopposite side of each cheek tooth (protocone andhypocone in uppers; protoconid and hypoconid inlowers). This cusp arrangement, which occurs in allSanta Rosan rodents and is different from that ofother early rodents, was seemingly considered byButler (1985:395) to be a hypothetical arrange-


ment, and the lack of specimens showing such anarrangement was described as a ‘‘difficulty’’ in ar-guing for an ancestral tetralophodont condition incaviids. As an ancestral condition in South Ameri-can rodents, however, this cusp arrangement wouldappear to establish the basis for interpreting an im-portant point of evolutionary divergence in earlyrodents.

The flattened, scissorslike cutting surface of Neo-gene and Recent caviids was not yet developed inthe Santa Rosan rodents. Rather, the primary (i.e.,temporally the longest functional) wear surface ap-pears to have been terraced, but with no gross in-dication of labiolingual movement during chewing.Chewing motion appears to have been propalinal,as in modern caviids, but with elevated labial (up-per molars) and lingual (lower molars) cusps untila late wear stage. Although this may represent aninteresting type of early chewing pattern, detailedanalysis of wear striae is required before any defin-itive statement about chewing motion in Santa Ro-san rodents can be made (Frailey and Campbell,2004).

The similarity of all of the Santa Rosan caviidrodents to each other, as well as to the erethizon-tids, suggested to Frailey and Campbell (2004) thatthe hypothesis of Wood (1949) and Wood and Pat-terson (1959), wherein the Acaremyinae (familyOctodontidae) were considered to be the possiblestructural ancestors of caviid rodents, was doubt-ful. Frailey and Campbell (2004) placed no SantaRosan rodents in the Acaremyinae, suggesting thatthe latter were already more specialized than theformer. In this case, the Octodontidae, rather thanrepresenting the lineage that gave rise to the rest ofthe caviids, was a group that split off fairly earlyfrom the main group that gave rise to the rest ofthe caviids, as did the erethizontids. If so, then theerethizontid dental pattern would be closer to theancestral caviid pattern than that of the acaremyi-nes.

Although an ?Eocene age for the paleofauna issupported by available evidence (see ‘‘Age of theSanta Rosa Local Fauna’’), the lack of an accurateage assignment hinders discussion of the originsand relationships of the South American rodents.The Santa Rosan rodent fauna is as diverse as themost diverse Deseadan SALMA (late Oligocene) lo-calities known, which suggests that the origin andradiation of South American rodents must have oc-curred in the early Eocene, or before. Althoughprincipal previous hypotheses of South Americanrodent origins have focused on possible links toNorth American or African lineages, the ?EoceneSanta Rosan rodents would seem to raise objectionsto both of those possibilities (Frailey and Campbell,2004).

An alternative hypothesis, that is, that the SouthAmerican rodents are an autochthonous SouthAmerican group of mammals derived from an an-cestral Gondwanan lineage, has also been proposedby various authors, but it has not received much

attention, primarily because of the lack of earlyTertiary fossil evidence (see review of evidence inFrailey and Campbell, 2004). The Santa Rosan ro-dents seem to provide some of this missing evi-dence, and this alternative hypothesis is favored byFrailey and Campbell (2004). Clearly, however, ad-ditional, and older, fossil evidence is required be-fore there can be said to be strong support for thisalternative hypothesis.

Martin (2004) demonstrates that all three sub-types of multiserial Hunter–Schreger bands (HSB)present in Oligocene and younger ‘‘caviomorphs’’were already present in rodents from Santa Rosa,including what is considered to be the most derivedtype of HSB. He did not discover anything atypicalfor ‘‘caviomorph’’ rodents in the six isolated inci-sors he examined, or anything more primitive thanwhat is seen in Deseadan SALMA or later SouthAmerican rodents.

To account for the diversity of incisor schmelz-muster in the rodents from Santa Rosa, Martin(2004) suggests two possibilities. One is that a dis-persal event from African ancestors occurred in theearly or middle Eocene. A difficulty with this is thatthe oldest African records of rodents with multi-serial HSB are from the late Eocene (Jaeger et al.,1985; Martin, 1992), which would be contempo-raneous with the Santa Rosan rodents as currentlydated. The second possibility Martin (2004) sug-gests is that there were waves of dispersal of Afri-can ancestors carrying the different types of incisorschmelzmuster to South America shortly before thetime of Santa Rosa. In that event, however, onemight expect to find rodents in Africa that aremuch more similar in dental features to those ofSanta Rosa than are actually known from late Eo-cene sites in Africa.

NOTOUNGULATES

The several specimens of notoungulates, represent-ing at least three and possibly as many as six spe-cies, are the first Paleogene records of this groupfrom lowland Amazonia (Shockey et al., 2004). Al-though a variety of notoungulates have been re-corded from the Neogene of Amazonia, the inter-atheriid from Santa Rosa is the first record of anyage for that family in Amazonia. The notoungulatespecimens cannot be referred to any known generaor species, and placement even to family is uncer-tain in instances. The naming of new taxa was de-ferred until better material is available. With so fewspecimens and so few taxa, however, and no otherlowland tropical Paleogene fauna for comparison,the notoungulates can offer few hints regardingtheir overall diversity and the importance of thegroup to the Amazonian mammalian fauna.

OTHER MAMMALS

The only possible specimen of a bat from SantaRosa can only questionably be referred to the Chi-roptera because it is damaged (Czaplewski and


Campbell, 2004). This specimen is important, how-ever, because, if it is a bat, it represents only thethird Paleogene record of a bat from South Americaand, if its age were indeed Eocene, it would be theoldest bat known from South America. The speci-men would be equally important should additionalspecimens from the Santa Rosa locality show thatit is not a bat because, in that case, it would rep-resent an order of mammals previously unknownin the Paleogene of South America.

An unusual mammalian specimen from the SantaRosa local fauna described by Goin et al. (2004)defied all attempts to relate it to any known groupof mammals. This single, multicuspate molariformtooth seemed to lack enamel on its external surface,but it could not be determined whether this wasnatural or a result of taphonomic processes. Micro-structure analysis was suggestive, but not conclu-sive, for mammalian enamel structure. Goin et al.(2004) concluded that this specimen might be bestallocated to Gondwanatheria (family Ferugliother-iidae), but they admit that the tooth is quite baf-fling.

AGE OF THE SANTA ROSA LOCAL FAUNA

In their description of the geology of the SantaRosa locality, Campbell et al. (2004) pointed outthat there were no available stratigraphic data thatcould be of use in deriving the age of the SantaRosa local fauna. The outcrop at the locality is solimited in extent and unremarkable in its lithiccharacteristics that it offers few clues to its forma-tional context. Although Campbell et al. (2004)tentatively refer the fossiliferous deposit to the Pa-leocene–Eocene Yahuarango Formation, this isbased on the age of the fossils derived from thefaunal studies, not on stratigraphic data.

From their study of the marsupials, Goin andCandela (2004) concluded that the Santa Rosa lo-cal fauna was middle Paleogene in age but thatthere were insufficient data to assign it to any cur-rently recognized SALMA. They considered it mostlikely that the paleofauna was middle to late Eo-cene in age, but there remained a slight possibilitythat it was of early Oligocene age. The older age issuggested by the perceived similarities of the SantaRosan marsupials to early Tertiary South Americanmarsupials, and even to early Tertiary North Amer-ican, European, or African taxa. The majority ofthe eleven new taxa described appear to be relatedmost closely to Paleocene or Eocene taxa, a fewseem similar to early Oligocene taxa, and the othersare so unique they provide no age-relevant infor-mation. Precise assignment of an age to the mar-supials of the Santa Rosa paleofauna through cor-relation with paleofaunas of known ages is, there-fore, difficult.

The rodents from Santa Rosa are the oldestknown South American rodents (Frailey andCampbell, 2004), and, as such, they cannot be usedto derive an age for the Santa Rosa local fauna

through correlation with taxa elsewhere. On thebasis of the stage of evolution of their teeth, theSanta Rosan rodents are clearly more primitivethan the Deseadan SALMA (late Oligocene) ro-dents of Bolivia and Argentina, previously the old-est known rodent faunas of South America. Keycharacters include the brachydont condition of theSanta Rosan rodents, as opposed to the significant-ly more hypsodont state of younger rodents, andthe minimal degree of fusion of the anterior twolophids of the lower molars of the Santa Rosan ro-dents compared with the greater fusion in youngertaxa. The Deseadan SALMA rodents also had de-veloped a chewing surface much more similar tothat of younger caviids. The Santa Rosan rodentsare interpreted as being older than the single rodentfrom the Tinguiririca local fauna of Chile, which isdated to 31.5 Ma (early Oligocene; Wyss et al.,1994), for the first two reasons cited above. How-ever, with only a single, well-worn specimen fromthe Tinguiririca local fauna it is necessary to be abit cautious about carrying interpretations too far.Still, the abundant marsupial paleofauna from San-ta Rosa would appear to support the interpretationthat the Santa Rosan rodents are older than thatfrom the Tinguirirican fauna.

Information as to the age of the Santa Rosa localfauna on the basis of evidence from the notoun-gulates is mixed, but the lower end of the estimatedage range approaches that inferred from the mar-supials and rodents. According to Shockey et al.(2004), the toxodont specimens are higher crownedthan Mustersan SALMA species, but lowercrowned than Deseadan SALMA species of Toxo-dontia. In general, although the notoungulate spe-cies represented cannot be referred to known tax-onomic categories with confidence, they appearmost similar to known taxa that date from the earlyOligocene (‘‘Tinguirirican’’ age) to the late Oligo-cene (Deseadan SALMA). Until better material isavailable, however, how to interpret these hintsfrom the notoungulates regarding the age of thefauna remains problematic, as noted by Shockey etal. (2004).

The combined evidence from the marsupials androdents appears to support an age assignment oflate Eocene, although the age could be either older(if more weight were given to the marsupials) oryounger (if more weight were given to the notoun-gulates). Following the redating of Eocene SAL-MAs by Kay et al. (1999), the best compromiseappears to be to consider the Santa Rosa local fau-na as dating to the Mustersan SALMA until betterage-pertinent data are available.

ENVIRONMENT OF DEPOSITION ANDPALEOECOLOGY

As noted by Campbell et al. (2004), the fossils weredeposited in lenses of coarse-grained, mostly sand-sized sediment with a matrix of clay within massivebeds of clay. These lenses are not particularly thick,


ranging from just 1–2 cm to 30 cm in thickness,with abrupt transitions to fine-grained clay depositsabove and below. The presence of some large (i.e.,up to 8–10 cm) bone fragments indicates fairlyswift water movement, at least on occasion, butthere are no rock clasts above sand size. All of thelenses sampled at the Santa Rosa locality were fos-siliferous, although similar appearing coarse-grained lenses in other outcrops that were sampledwere not. That there were multiple depositionalevents, that is, episodes of fossiliferous lens for-mation, at the Santa Rosa site indicates a fairly sta-ble surrounding habitat from which the fossilscame, but shifting water currents probably depos-ited bed load in different areas at different times. Itis possible that analysis of the fish fossils from thesite will provide more information about the typeof aquatic environment in which the paleofaunawas deposited.

The reason for the rich accumulation of fossils atSanta Rosa is unclear. The largest mammals pre-served, the notoungulates, reached the size ofsheep, and there were a few large crocodilians thatmay have been able to seek them out as prey. Mostof the fossils, however, are microvertebrates, andthe vast majority of these are only a few millimetersin size. Both of the most abundant groups of mam-mals, the rodents and marsupials, are representedby older individuals, as indicated by extensive wearon the teeth, as well as extremely young individualswith no visible wear on the teeth. For the mostabundant taxa, a continuum of wear stages indi-cates individuals from a wide range of ages. Noneof the marsupials show any evidence of havingpassed through a predator, as might be indicated byacid etching of the enamel (Goin and Candela,2004), but there are numerous examples of etchedrodent teeth.

Goin and Candela (2004) suggested that manyof the marsupials might have been arboreal ani-mals, and although we know nothing of the habitsof the early rodents, it is probable that some ofthem were also arboreal, or at least scansorial. Awide sampling of arboreal taxa in the fossil recordis relatively rare, and it is indeed fortunate that wehave a good sample from the first Paleogene faunaof tropical Amazonia. The presence of a presumedbat and a few birds indicates that at least a smallportion of the flying guild was sampled as well.

In their discussion of individual taxon abundancein Deseadan SALMA rodent faunas, Patterson andWood (1982) pointed out that, within any givensample of rodents, whether fossil or Recent, onespecies seemed to be overwhelmingly dominant interms of numbers. In the Santa Rosa local fauna,the same is true for both the marsupials and ro-dents, although more so for the former than thelatter. For the marsupials, Wamradolops tsullodonrepresented almost 42 percent of the paleofauna,whereas for the rodents, Eoespina woodi repre-sented almost 25 percent of the fauna. Two speciesqualified as the next most common rodent, with

Eoincamys pascuali and Eobranisamys romeropitt-manae each representing about 13 percent of thepaleofauna.

If borne out, a Mustersan SALMA age for theSanta Rosa local fauna places it at a critical periodof biotic change in South America. Following Kayet al. (1999), the Mustersan SALMA is now con-sidered to range from approximately 35 to 32 Ma,or to straddle the Eocene–Oligocene boundary at33.7 Ma (Berggren et al., 1995). As Pascual andOrtiz-Jaureguizar (1990), Woodburne and Swisher(1995), and Kay et al. (1999) point out, this periodcoincides with major compositional changes inmammalian communities in North America (the‘‘Terminal Eocene Event’’), Europe (‘‘La GrandeCoupure’’), and Patagonia, South America. For Pa-tagonia, where most Paleogene mammalian faunasof South America are found, Kay et al. (1999) de-scribe the period from 34.5 to 28.5 Ma as one char-acterized by a transition from relatively warmer tocooler conditions. They describe this cooling ashaving the effect of decreasing mean annual precip-itation and temperature, increasing seasonal tem-perature amplitude, and decreasing the mean tem-perature of the coldest month. A dramatic increasein the number of hypsodont taxa in southern SouthAmerica was the most obvious result of these cli-matic effects. Hypsodonty appears to correlate witha more gritty diet, which can be brought on by anincrease in available siliceous grasses or dust (Kayet al., 1999). An increase in hypsodonty could eas-ily be tied to a decrease in tropical vegetation ascooling and a decrease in rainfall brought about areduction in the distribution of tropical ecozones.

On a global scale, it would be expected that cool-ing in the high latitudes would lead to a greaterheat gradient between the equatorial regions andthe poles. This would lead, in turn, to a greaterlevel of heat exchange between the tropics and thepolar regions, which undoubtedly would have af-fected (i.e., strengthened) global wind patterns.How this might have affected the climate, and es-pecially patterns of precipitation, in proto-Amazon-ia is unknown to me, but it would be unrealistic toexpect the equatorial regions of South America tobe unaffected by the dramatic changes in climateseen in the higher latitudes of that continent at theEocene–Oligocene transition.

Although details of the Eocene paleoclimate ofnorthern South America are not particularly wellknown, I think it is important to recognize the fol-lowing facts. First, although the Santa Rosa sitewas a few degrees farther south in the late Eocenethan it is today (possibly 2–5 degrees), it would stillhave been within existing climatic belts. Second, inthe mid- to late Eocene, the Andes were experienc-ing their first (Megard, 1984, 1987; Sebrier et al.,1988; Ellison et al., 1989; and Sebrier and Soler,1991) or second (Noble et al., 1990) of up to sixpost-Cretaceous phases of uplift. The elevation ofthe mountain ranges raised at this time, and that ofthose produced by the preceding Cretaceous uplift


events, is still unknown, but it can safely be as-sumed that these ranges did not begin to approachthe heights reached by today’s Andes. For example,an estimate by Jordan et al. (1997) was that Eoceneuplift accounted for 25–50 percent of the elevationof the Western Cordillera of the Central Andes.And after reviewing available data pertaining to theuplift history of the Central Andes, several degreesof latitude south of the Santa Rosa locality, Greg-ory-Wodzicki (2000) concluded that the WesternCordillera in that region was no more than half ofits modern elevation by 25–18 Ma (late Oligocene–early Miocene) and the Eastern Cordillera was at25–30 percent of its modern elevation by 20 Ma(early Miocene). Widespread uplift of the EasternCordillera apparently did not begin until the Oli-gocene. Extrapolation from these conclusions to theuplift history of the Andes much farther north, thatis, to the west of Santa Rosa, and to a period inthe late Eocene is not yet justifiable, but this infor-mation does provide a general sense of the scale ofthe Andes in the late Eocene.

One consequence of a low Andean range is thatorographic uplift, which is such a dominant featureof the climate of western Amazonia today andwhich produces so much of the precipitation inwestern Amazonia, probably would have been asmall fraction of what is seen today. It would alsohave occurred farther to the west. In the absence ofthe high Andes, much of the moisture now trappedand recirculated within the Amazon Basin wouldhave been carried westward over the Pacific coastof South America, producing a much drier proto-Amazonia.

Third, the Santa Rosa locality is today over 2400km from the northeast coast of South America, theclosest source for moisture-laden, tropical Atlanticwinds, and this distance was probably similar in theEocene. Depending on the seasonal strength of theprevailing trade winds and Eocene topography, pre-cipitation from these winds over northeastern andeastern South America might have limited theamount of atmospheric water available to distantwestern Amazonia and increased seasonality in thatregion.

Furthermore, edaphic conditions might have pro-moted wide expanses of savanna vegetation in pro-to-Amazonia, much as they do today, even thoughtoday the region is probably under a climatic re-gime with much greater precipitation. With lessrainfall, edaphic conditions might have contributedto even more widespread tropical savannas inSouth America.

Given the above facts, there is little reason toexpect that the present vast expanse of tropical for-est in Amazonia existed in the Paleogene. Indeed,it is perhaps more reasonable to speculate that theSanta Rosa local fauna represents a fauna from ariparian environment surrounded by savanna, acondition that would sustain brachydonty amongthe mammals of the forest (marsupials and rodents)and promote hypsodonty among the large mam-

mals of the savanna (the notoungulates). Hypo-thetically, these environmental conditions mightwell have existed for some time prior to the Eo-cene–Oligocene transition, and they might alsohave been even more dominate prior to the firstpost-Cretaceous uplift of the Andes. If this were thecase, using any type of ‘‘hypsodonty index’’ to es-timate the age of the Santa Rosa local fauna couldlead to significant error.

By itself, the Santa Rosa local fauna cannot beused to address questions of biotic change in thetropics at the Eocene–Oligocene transition, even ifsuch had been occurring at the time of deposition,because it represents only a thin slice of time.Nonetheless, I consider it doubtful that Amazoniadid not experience any affects from the climaticchanges then occurring in higher latitudes. As withso many other questions raised by the Santa Rosalocal fauna, that of how the Eocene–Oligocene cli-matic shift affected biotic communities in tropicalAmazonia can only be answered by additional Pa-leogene faunas from the region.

ACKNOWLEDGMENTS

I thank all of the authors who contributed to our under-standing of the Santa Rosa local fauna through their con-tributions to this volume. Their assistance and patience ismuch appreciated. I thank Nick Czaplewski, Carl Frailey,Francisco Goin, John Lundberg, and Thomas Martin fortheir comments that helped improve this summary.

LITERATURE CITED

Berggren, W.A., D.V. Kent, C.C. Swisher III, and M.-P.Aubry. 1995. A revised Cenozoic geochronology andchronostratigraphy. In Geochronology time scalesand global stratigraphic correlation, ed. A. Berggren,D.V. Kent, and J. Hardenbol. Society of EconomicPaleontologists and Mineralogists (SEPM) SpecialPublication, 54:129–212.

Butler, P. 1985. Homologies of molar cusps and crests, andtheir bearing on assessments of rodent phylogeny. InEvolutionary relationships among rodents. A multi-disciplinary analysis, ed. W.P. Luckett and J.-L. Har-tenberger, 381–401. New York: Plenum Press.

Campbell, K.E., Jr., C.D. Frailey, and L. Romero-Pittman.2000. The late Miocene gomphothere Amahuacath-erium peruvium (Proboscidea: Gomphotheriidae)from Amazonian Peru: Implications for the GreatAmerican Faunal Interchange. Instituto GeologicoMinero y Metalurgico, Serie D: Estudios Regionales,Boletın 23:1–152.

. 2004. The Paleogene Santa Rosa local fauna ofAmazonian Peru: Geographic and geologic setting.In The Paleogene mammalian fauna of Santa Rosa,Amazonian Peru, ed. K.E. Campbell, Jr. Natural His-tory Museum of Los Angeles County, Science Series40, 3–14.

Crochet, J.-Y. 1986. Kasserinotherium tunisiense nov.gen., nov. sp., troisieme marsupial decouvert en Af-rique (Eocene inferieur de Tunisie). Comptes Rendusde l’Academie des Sciences, Paris 302(14):923–926.

Czaplewski, N.J., and K.E. Campbell, Jr. 2004. A possiblebat (Mammalia: Chiroptera) from the ?Eocene ofAmazonian Peru. In The Paleogene mammalian fau-na of Santa Rosa, Amazonian Peru, ed. K.E. Camp-


bell, Jr. Natural History Museum of Los AngelesCounty, Science Series 40, 141–144.

Ellison, R.A., B.A. Klinck, and M.P. Hawkins. 1989. De-formation events in the Andean orogenic cycle in theAltiplano and Western Cordillera, southern Peru.Journal of South American Earth Sciences 2(3):263–276.

Flynn, J.J., and C.C. Swisher III. 1995. Cenozoic SouthAmerican Land Mammal Ages: Correlation to globalgeochronologies. In Geochronology time scales andglobal stratigraphic correlation, ed. A. Berggren,D.V. Kent, and J. Hardenbol, Society of EconomicPaleontologists and Mineralogists (SEPM) SpecialPublication, 54:317–333.



Goin, F., and A. Candela. 2004. New Paleogene marsu-pials from the Amazon Basin of eastern Peru. In ThePaleogene mammalian fauna of Santa Rosa, Ama-zonian Peru, ed. K.E. Campbell, Jr. Natural HistoryMuseum of Los Angeles County, Science Series 40,15–60.

Goin, F.J., E.C. Vieytes, M.G. Vucetich, A.A. Carlini, andM. Bond. 2004. Enigmatic mammal from the Paleo-gene of Peru. In The Paleogene mammalian fauna ofSanta Rosa, Amazonian Peru, ed. K.E. Campbell, Jr.Natural History Museum of Los Angeles County,Science Series 40, 145–153.

Gregory-Wodzicki, K. 2000. Uplift history of the Centraland Northern Andes: A review. Geological Societyof America Bulletin 112:1091–1105.

Jaeger, J.-J., C. Denys, and B. Coiffait. 1985. New Phiom-orpha and Anomaluridae from the late Eocene ofnorth-west Africa: Phylogenetic implications. InEvolutionary relationships among rodents, ed. W.P.Luckett and J.-L. Hartenberger, 567–588. NewYork: Plenum Press.

Jordan, T.E., J.H. Reynolds III, and J.P. Erikson. 1997.Variability in age of initial shortening and uplift inthe Central Andes, 16–338309 S. In Tectonic upliftand climate change, ed. W.F. Ruddiman, 41–61.New York: Plenum Press.

Kay, R.F., R.H. Madden, R.L. Cifelli, and J.J. Flynn (eds.).1997. Vertebrate paleontology in the Neotropics:The Miocene fauna of La Venta, Colombia. Wash-ington, D.C.: Smithsonian Institution Press, 592 pp.


Marshall, L., and T. Sempere. 1993. Evolution of the Neo-tropical Cenozoic land mammal fauna in its geo-chronologic, stratigraphic, and tectonic context. InBiological relationships between Africa and SouthAmerica, ed. P. Goldblatt, 328–392. New Haven andLondon: Yale University Press.

Martin, T. 1992. Schmelzmikrostruktur in den Inzisiven

alt- und neuweltlicher hystricognather Nagetiere. Pa-laeovertebrata, Memoire extraordinaire, 1–168.

. 2004. Incisor enamel microstructure of SouthAmerica’s earliest rodents: Implications for cavio-morph origin and diversification. In The Paleogenemammalian fauna of Santa Rosa, Amazonian Peru,ed. K.E. Campbell, Jr. Natural History Museum ofLos Angeles County, Science Series 40, 131–140.

Megard, F. 1984. The Andean orogenic period and its ma-jor structures in central and northern Peru. Journalof the Geological Society of London 141:893–900.

. 1987. Structure and evolution of the Peruvian An-des. In The anatomy of mountain ranges, ed. J.-P.Schaer and J. Rodgers, 179–210. Princeton, NewJersey: Princeton University Press.

Noble, D.C., E.H. McKee, T. Mourier, and F. Megard.1990. Cenozoic stratigraphy, magmatic activity,compressive deformation, and uplift in northernPeru. Geological Society of America Bulletin 102:1105–1113.

Pascual, R., and E. Ortiz-Jaureguizar. 1990. Evolving cli-mates and mammal faunas in Cenozoic South Amer-ica. Journal of Human Evolution 19:23–60.

Patterson, B., and R. Pascual. 1972. The fossil mammalfauna of South America. In Evolution, mammals,and southern continents, ed. A. Keast, F.C. Erk, andB. Glass, 247–309. Albany: State University of NewYork.

Patterson, B., and A.E. Wood. 1982. Rodents from theDeseadan Oligocene of Bolivia and the relationshipsof the Caviomorpha. Bulletin of the Museum ofComparative Zoology 149(7):371–543.

Ruegg, W. 1947. Estratigrafıa Comparada del OrientePeruano. Boletın de la Sociedad Geologica del Peru20:57–100.

. 1956. Geologıa y Petroleo en la Faja SubandinaPeruana. Simposio Sobre Yacimientos de Petroleo yGas, XX Congreso Geologico International IV,Mexico, 89–139.

Sebrier, M., A. Lavenu, M. Fornari, and J.-P. Soulas. 1988.Tectonics and uplift in Central Andes (Peru, Boliviaand northern Chile) from Eocene to Present. Geo-dynamique 3(1/2):85–106.

Sebrier, M., and P. Soler. 1991. Tectonics and magmatismin the Peruvian Andes from late Oligocene time tothe Present. Geological Society of America, SpecialPaper 265:259–278.

Shockey, B., R. Hitz, and M. Bond. 2004. Paleogene no-toungulates from the Amazon Basin of Peru. In ThePaleogene mammalian fauna of Santa Rosa, Ama-zonian Peru, ed. K.E. Campbell, Jr. Natural HistoryMuseum of Los Angeles County, Science Series 40,61–69.

Simpson, G.G. 1970. Addition to the knowledge of Groe-beria (Mammalia, Marsupialia) from the mid-Ce-nozoic of Argentina. Breviora 362:1–17.

Wood, A.E. 1949. A new Oligocene rodent from Patagon-ia. American Museum of Natural History, Novitates1435:1–54.

Wood, A.E., and B. Patterson. 1959. The rodents of theDeseadan Oligocene of Patagonia and the beginningsof South American rodent evolution. Bulletin of theMuseum of Comparative Zoology 120(3):279–428.

Woodburne, M., and C.C. Swisher III. 1995. Land mam-mal high-resolution geochronology, intercontinentaloverland dispersals, sea level, climate, and vicari-ance. In Geochronology time scales and global strati-graphic correlation, ed. A. Berggren, D.V. Kent, and


J. Hardenbol. Society of Economic Paleontologistsand Mineralogists (SEPM) Special Publication, vol.54:335–364.

Wyss, A.R., J.J. Flynn, M.A. Norell, C.C. Swisher III, M.J.Novacek, M.C. McKenna, and R. Charrier. 1994.Paleogene mammals from the Andes of central Chile:

A preliminary taxonomic, biostratigraphic, and geo-chronologic assessment. American Museum of Nat-ural History, Novitates 3098:1–31.


The Paleogene MammalianFauna of Santa Rosa,

Amazonian PeruEdited By Kenneth E. Campbell, Jr.

Natural History Museum of Los Angeles County

Science Series 40 December 27, 2004

TH

E PA

LE

OG

EN

E M

AM

MA

LIA

N FA

UN

A O

F SAN

TA

RO

SA, A

MA

ZO

NIA

N PE

RU

/ CA

MPB

EL

L

900 Exposition BoulevardLos Angeles, California 90007