INTRODUCTION

Marsupials, although depauperate in South America today (represented by only a small number of didelphid, caenolestid, and microbiotheriid species) were vastly more diverse during most, if not all, of the Cenozoic. South America's highly endemic marsupial faunas during this time reflect the continent's long-term geographic isolation, as do its contemporaneous and equally peculiar placental mammal radiations. Marsupials have long been recorded from the classic deposits of Patagonia and surrounding Argentina, augmented by newer discoveries of important assemblages in lower latitudes and the Andean Cordillera. These include some of the earliest (Tiupampan South American Land Mammal “Age” [SALMA], early Paleocene) and best-preserved marsupicarnivorans, a tropical-latitude Deseadan SALMA assemblage (Salla) from Bolivia, the Colombian middle Miocene tropical fauna of La Venta (Laventan SALMA), and the unusual suite of didelphimorphians from the Paleocene (Itaboraian SALMA, early Paleocene) of Brazil (see Flynn and Wyss, 1998 for summary). Since 1986, we and our colleagues have uncovered early–middle Cenozoic mammal faunas in several regions of the Chilean southern Andes (Wyss et al., 1987, 1990, 1992, 1993a, 1993b, 1994, 1996; Flynn et al., 1988, 1991, 1995, 2002a, 2002b, 2003; Charrier et al., 1990, 1996, 1997, 2002; Dodson, 1994; Flynn and Swisher, 1995; Hitz, 1997; Flynn and Wyss, 1998, 1999; Hitz et al., 2000; Reguero et al., 2003; Croft, 2000, 2001; Croft et al., 2003a, 2003b). These faunas expand the geographical, temporal, paleoenvironmental, and taxonomic sampling of South American marsupials and other mammals, while providing important information on the geologic history of the Andes.

Recent efforts include extensive paleontological reconnaissance within the south-central Andean Main Range, roughly between 33.5° and 38°S latitude. This work has led to the recovery of several Cenozoic mammal faunas from the Abanico (= Coya-Machalí) Formation, the first known for the region, including the original discovery of the Tinguiririca Fauna. Better illuminating the Eocene-Oligocene transition than perhaps any other land mammal fauna in South America, the Tinguiririca Fauna includes polydolopid, groeberiid, and didelphimorph marsupials (Flynn and Wyss, 1999). Since that initial find, the Abanico Formation (and its lateral equivalents) of the central Chile Andean Main Range has yielded seven or more Cenozoic faunas, ranging in age at least from the Eocene (Casamayoran) through the early Miocene. Preparation of the hard and brittle volcaniclastic matrix, in which these fossils are typically found, is extremely difficult and labor intensive, with description of key material proceeding as it becomes available. We provide here the first report of a polydolopine skull. Polydolopids were among the earliest-described South American fossil marsupials; although they have a long temporal range and are moderately common elements of most early Cenozoic faunas, this specimen represents only the second polydolopid cranium known. This skull is the first fossil to be described from a locality just north of the Río Cachapoal, discovered in 1996 (Charrier et al., 1997). Analysis of this new specimen from another new fossil-bearing locality within the Abanico Formation in the Chilean Andes seems especially fitting for this festschrift volume, as it provides new systematic and geologic data from strata and a region in which we collaborated early on with our Ph.D. advisor and friend, Malcolm C. McKenna.

Simpson (1928, 1948) and Paula Couto (1952) presented important early analyses of the Polydolopidae. Marshall (1982) provided the most recent comprehensive review of the group, and Marshall et al. (1990) proposed a higher-level phylogeny for fossil and modern marsupials. A selection of more recent studies of polydolopids and potential close relatives includes Pascual (1980, 1981), Pascual and Bond (1981), Woodburne and Zinsmeister (1984), Case et al. (1988), and Goin and Candela (1995, 1996). McKenna and Bell (1997) summarized the taxonomy and temporal occurrences for polydolopid taxa then known. Flynn and Wyss (1999) described the youngest known polydolopid, Polydolops abanicoi, discussed its systematic and temporal implications, and provided the first phylogenetic definitions for various taxonomic names previously applied to components of this clade (e.g., Marsupialia, Polydolopimorphia, Polydolopoidea, Polydolopiformes, Polydolopidae, and Polydolopinae).

Polydolopids, like most marsupials, are rare in comparison to placentals within most Cenozoic South American assemblages and had been reported only from Paleocene and Casamayoran assemblages until the 1970s (see Marshall, 1982: 9). They do, however, occur with some frequency in many pre-Deseadan assemblages, with well-documented occurrences now ranging from the Paleocene (Itaboraian and Riochican SALMAs; Paula Couto, 1952; Marshall, 1982; Goin and Candela, 1995; and possibly the older Tiupampan SALMA, “Epidolops sp.”, Gayet et al., 1991) to the earliest Oligocene Tinguiririca Fauna (representing a new pre-Deseadan, post-Mustersan SALMA, the Tinguirirican, see Flynn and Swisher, 1995, Flynn and Wyss, 1999, Flynn et al., 2003), a span of some 30 million years. Polydolops itself is quite long-ranging, with species known from that entire span, excepting the earliest (Tiupampan and Itaboraian SALMAs) part during which only Epidolops, the polydolopine nearest outgroup, and Amphidolops (= Seumadia) yapa, an advanced polydolopine, are known from Patagonian Argentina and Brazil (Paula Couto, 1952; Marshall, 1982; Goin and Candela, 1995; Bond et al., 1995; see Gayet et al., 1991, for possible older, Tiupampan SALMA occurrence of Epidolops in Bolivia). In addition, the polydolopids reported from Antarctica (Antarctodolops, Eurydolops) have been considered to be closely related to particular Polydolops species (Woodburne and Zinsmeister, 1984; Case et al., 1988), either rendering Polydolops paraphyletic or the Antarctic taxa assignable to Polydolops (see discussion in Flynn and Wyss, 1999).

SYSTEMATICS

DISCUSSION

The skull of Polydolops mckennai is the first recovered for a polydolopine and only the second for any polydolopid, providing evidence bearing on questions of polydolopid monophyly, basal diversification, and the features typifying the group and subclades ancestrally. A full revision of polydolopoid phylogenetics would be worthwhile, especially in light of advances in systematic methods, description of many new taxa since Marshall's (1982) review of polydolopids, and proposals that some taxa currently recognized may not be monophyletic or that the important Antarctic taxa (Antarctodolops and Eurydolops) may be nested within various species assigned to Polydolops. Although outside the scope of this brief initial report, such an analysis will yield better understanding of patterns of morphological evolution within the clade and of early Cenozoic biogeographic relationships among “Gondwanan” continents. For example, Eurydolops is a taxon known only from a single tooth (considered P3 by Case et al., 1988). Although this taxon has reasonably been considered to be a polydolopine, the shape and morphology of the available tooth has many resemblances to the P2 of Epidolops; if Eurydolops is in fact an epidolopine rather than polydolopine, this would have important biogeographic and temporal ramifications.

Polydolops mckennai is readily distinguishable from the other taxon known from a skull (Epidolops ameghinoi), as well as other non-Polydolops polydolopines, in a variety of features (see diagnosis). In overall size, P. mckennai is smaller than Epidolops and Eudolops, but larger than most other polydolopines.

It is more difficult to distinguish P. mckennai from other species of Polydolops, as many of them are known from less complete or noncomplementary material (e.g., P2–3 known only for P. thomasi; upper molars known only for P. thomasi and P. serra), but it is larger than almost all of them (see diagnosis) and differs in cheek-tooth proportions. The largest species of Polydolops, P. mckennai and P. mayoi, are represented by material that is not directly comparable (i.e., skull and upper dentition versus lower dentition only), so the approximate sizes can only be inferred through relative sizes of upper and lower teeth in closely related taxa for which both are known. Upper cheek teeth of P. mckennai are ∼10–20% larger than those of P. thomasi, and the lower cheek teeth in the holotype of P. mayoi are 2–26% longer and 20–35% wider than the average for P. thomasi. Thus, among Polydolops species, P. mckennai may most closely approximate P. mayoi in size (and possibly in having elongate cheek teeth), although they likely differ in shape and proportions, particularly in P. mckennai having proportionally larger and more elongate P2–3. The most distinctive aut\[chapomorphy of P. mckennai, distinguishing it from all other species of Polydolops, is the extreme elongation of all the cheek teeth (table 6.2). In P. mckennai all cheek teeth are at least 15% longer than they are wide, and P2 is almost twice as long as it is wide. In other species of Polydolops the upper cheek teeth are squared (approximately equal in length and width) or wider than long, except P2, which is much less elongate in those other species. The extreme elongation of both P2 and P3 appears to be a diagnostic aut\[chapomorphy for P. mckennai, distinguishing the species from all other polydolopids. Although Eudolops tetragonus also has an elongate P2 and slightly elongate P3, they are not nearly so elongate as in P. mckennai (e.g., P2: 60% versus 90% longer than broad) and although both teeth are sectorial in both taxa, their morphologies are quite distinct. The primitive condition for molar shape in polydolopids appears to be either squared (as in most polydolopines) or broader than long (as most marked in the polydolopine proximal outgroup Epidolops). Although the elongate molars in Eudolops hernandezi make inferences about the ancestral molar shape in polydolopids and polydolopines a bit problematic, clear distinction of this taxon from Polydolops and the roughly squared or relatively broad molars in all other polydolopids (including the other species of Eudolops) support the proposed reconstruction. Thus, the very elongate molars in P. mckennai appear to be diagnostic autapomorphies. Peculiarly, P. mckennai appears to have many fewer buccal and stylar cusps on the upper molars than do other species of Polydolops. This distinction, as well as the very elongate cheek teeth, might be used to argue for exclusion of mckennai from a Polydolops clade. However, currently lacking complementary material for many of the other polydolopine taxa (e.g. skulls for all, anterior dentitions for most, and any upper teeth in some), and given the apparent synapomorphies in P2–3, cheek-tooth row orientation, and other molar morphology, we conservatively choose to ally the new species with other species of Polydolops.

The anterior teeth (and consequently dental formulas) are poorly known in polydolopids, owing to the dearth of specimens with a preserved anterior mandibular ramus or rostrum; indeed, until recently this region of the dentition was reasonably well known in just a single taxon, Epidolops ameghinoi. Preservation of almost the entire rostrum and mandibular ramus in taxa from the Abanico Formation, Polydolops mckennai (this paper) and P. abanicoi (Flynn and Wyss, 1999), providing the first evidence of the anterior upper dentition in any polydolopid other than Epidolops, strengthens homology inferences for the anterior teeth in polydolopids (particularly identification of P1/p1 and the enlarged tooth) and supports the conclusion that these are synapomorphic in all polydolopids (see below).

P. mckennai documents the occurrence of upper incisors (at least one, and possibly two or more) in Polydolops; Epidolops had three upper incisors (Paula Couto, 1952; Marshall, 1982). As in Epidolops, the incisor(s) were not located on the anterior rim of the snout, however, but rather were set more posteriorly on the buccal rim of the palate, closely appressed to several other teeth. The lateral position and development of a battery of anterior upper teeth [incisor(s), canine, P1] seem congruent with the great enlargement and procumbency of the lower canine (located near the midline and symphysis of the mandibular rami), and associated great reduction or loss of all lower incisors. In addition, this specimen establishes the expected, but previously undocumented, presence of a large, vertically oriented upper canine. The clear presence of P1 in P. mckennai contradicts the earlier inference that P1 and p1 were both absent in Polydolops (Marshall, 1982), and is consistent with the evidence from P. abanicoi of a small, anteriorly positioned p1 in the clade (Flynn and Wyss, 1999). Thus, the presence of P1 in P. mckennai and p1 in P. abanicoi (also likely in P. clavulus, as noted by Ameghino, 1903; see Flynn and Wyss, 1999), and similar P1/p1 in the polydolopine outgroup Epidolops, suggests that a small P1/1, located far anteriorly on the ramus or maxilla (close to the canine and separated from P2/2 by a long diastema), was the ancestral condition for polydolopids and polydolopines. The homology of this suite of modifications (very large upper and lower canines; procumbency of the lower anterior teeth, particularly the canines; great reduction or loss of upper and lower incisors; reduction of P1/p1; and long P1/p1–P2/p2 diastema separating the anterior dental battery from the cheek-tooth row) between Epidolops and polydolopines strengthens the hypothesis of polydolopid monophyly supported by these synapomorphies (in addition to those in the cheek-tooth battery).

Goin and Candela (1995) suggested that Polydolopidae might not be monophyletic, given the highly modified lower first molar of polydolopines and the difficulty in deriving them from a hypothesized epidolopine (= Epidolops) ancestry. Epidolops represents a lineage that diverged from the remaining polydolopids (a clade termed the Polydolopinae) at least as early as the Itaboraian SALMA. Although Epidolops (E. ameghinoi) represents the earliest known occurrence of a polydolopid (polydolopines are first recorded in the somewhat younger [Flynn and Swisher, 1995] Riochican SALMA), and does have some quite different aspects of molar morphology, it is the proximal outgroup of the polydolopines (with its own autapomorphies) rather than an ancestor. Therefore, one should not necessarily expect to be able to derive every feature seen in polydolopines from the condition in Epidolops, but rather from a more generalized condition that marked the common ancestor of both lineages.

Epidolops is distinguished from polydolopines by marked differences in P2/p2 and some aspects of molar morphology, loss of M4/m4 (in polydolopines), and significant rotation of the greatly expanded P3/p3 blades in Epidolops. With discovery of the first polydolopine skull (Polydolops mckennai), comparison to the other available polydolopid cranium (Epidolops ameghinoi, Paula Couto, 1952; Marshall, 1982) reveals differences in several distinctive features between the two lineages, although caution must be exercised in interpreting these differences (due to dorsoventral crushing of both specimens), and understanding their distribution in other polydolopids and phylogenetic implications awaits discovery of crania for other taxa. The cheek-tooth row is straight in P. mckennai, rather than V-shaped or arcuate (P3 is angled sharply outward and the tooth rows converge posteriorly in E. ameghinoi), proportionally much longer, and with more similarly sized cheek teeth (contra the dominance of a huge P3 in E. ameghinoi). The skull of Epidolops ameghinoi appears to be markedly more robust than P. mckennai, with a very broad rostrum and massive zygomatic arches. The anterior root of the latter is much broader (forming a flattened shelf) and more anteriorly placed relative to the cheek-tooth row (spanning the anterior edge of P3 to the posterior end of M2) in E. ameghinoi. In contrast, the anterior zygomatic root in P. mckennai spans only the anterior edge of M1 to the middle of M2. The snout is narrower and less flattened anteriorly in P. mckennai, and the canine alveolus is less robust and more laterally compressed. The basicranial portion of the skull and the braincase appear to be much smaller, proportionally, in E. ameghinoi than in P. mckennai.

All polydolopids share a large suite of unusual, and apparently homologous, dental and gnathic features, including inflexion of the portion of the mandibular ramus bearing the anterior dentition relative to that bearing the cheek teeth, great enlargement of upper and lower canines, procumbency of the enlarged lower canine, reduction of P1/p1, elongate diastema between the canine and cheek teeth, greatly enlarged and bladelike sectorial P3/p3, and quadrate and highly cuspate molars. The two available skulls suggest polydolopids are characterized by elongate snouts, broad and relatively flat (dorsoventrally) skulls, and robust zygomatic arches, although crushing of both of the known skulls makes interpretation of these as synapomorphies of the clade tentative. Although similar conditions for some of these appear sporadically among other marsupial taxa, some of these resemblances can be clearly shown to be nonhomologous (e.g., see discussions in Ride, 1962; Marshall, 1982). For example “diprotodont” procumbent anterior teeth, or “plagiaulacoid” shearing blades, occur in other marsupials, but these generally are in different tooth loci than in polydolopids. Similarly, quadrate cheek teeth (sometimes accompanied by a diastema from the anterior tooth battery) occur in other taxa, but anterior and cheek teeth typically are aligned linearly along the ramus. Given their close correspondence in form and apparent homology, we consider this suite of features as synapomorphies indicative of the unique common ancestry of all polydolopids, and thus we conservatively regard Epidolops and polydolopines as constituting a monophyletic group (see also Flynn and Wyss 1999).

P. mckennai is the first specimen described from new localities in the Abanico Formation of the Río Cachapoal drainage. It is not yet clear whether the material collected from the Cachapoal locality represents a single fauna/horizon, or several horizons, as our initial collection from this area (currently numbering several dozen specimens) was recovered both in situ and in blocks rolled from the steep valley slopes. The holotype of P. mckennai, as most of the other fossils obtained from this drainage, was found in a block derived from an uncertain stratigraphic horizon within the Abanico Formation. Preliminary assessment of the stratigraphic, structural, and topographic (from analysis of aerial photos) data, relative to more intensively studied sequences only a few kilometers to the south (the Río Las Leñas drainage; Flynn et al., 1995; Charrier et al., 1997, 2002; work in progress by R. Charrier and the authors), suggests that the very steeply dipping strata in this area likely represent only a relatively short interval of time, predating 16.1 Ma and 20.09 Ma levels (the latter containing a fossil platyrrhine, among other taxa; Flynn et al., 1995; Charrier et al., 2002) along the Río Las Leñas. Much or all of the stratigraphic interval at the Cachapoal locality appears to correlate with or underlie strata exposed at Río Las Leñas, which have been preliminarily dated at 29.3 ± 0.1 Ma (⁴⁰Ar/³⁹Ar, unpublished, C.C. Swisher III, personal commun.). We thus tentatively consider the material and localities in the immediate vicinity of the polydolopid skull find as representing a single faunal interval, herein termed the Cachapoal locality, although the presence of a mixed fauna from strata of several ages within the suite of material collected from rolled blocks must remain a viable alternative (Charrier et al., 1997). Continuing study of the fauna, geology, and geochronology by R. Charrier and the authors may provide stronger support for interpretation of a single fauna or further evidence for multiple horizons (and their ages).

The youngest previously known polydolopid, P. abanicoi, is from the pre-Deseadan, post-Mustersan Tinguiririca Fauna (Flynn and Wyss, 1999). Thus, the Cachapoal locality from which P. mckennai derives is no younger than earliest Oligocene (Tinguirirican SALMA, Flynn and Swisher, 1995; Flynn et al., 2003), or this taxon marks an age extension (Deseadan or younger) for the last appearance of polydolopids; the available evidence suggests that the former interpretation is more likely. The Cachapoal locality strata seem to lie within the core of a major anticline, with even more tightly folded anticlinal strata to the east (Charrier et al., 2002). Radioisotopic and new biochronologic data suggest that deposition in this area initiated by the earliest Oligocene (if not earlier, within the Eocene), and that compression and uplift postdated the early Miocene. Alternatively (but unlikely), if the Cachapoal locality strata (earliest Oligocene or older) are shown not to lie at the core of an anticline, but instead to overlie the 16–29 Ma sequence, it would document eastward thrusting of older strata after the middle Miocene. Such thrusting, although of uncertain timing relative to that in the Las Leñas/Cachapoal area, has been documented in the Abanico Formation along the Río Tinguiririca valley well to the south (Wyss et al., 1994, 1996; Charrier et al., 1994, 1996; see also Godoy and Lara, 1994; Sempere et al., 1994; Godoy et al., 1999). More robust information on the age and tectonostratigraphic relations of the Cachapoal locality strata will be essential to refining the model for basin development under an initial extensional tectonic regime and the timing of subsequent tectonic deformation, compression, and uplift (Charrier et al., 2002).