Journal of Mammalian Evolution

, Volume 13, Issue 2, pp 89–123

Spiny Norman in the Garden of Eden? Dispersal and early biogeography of Placentalia

Authors

    • Museum of Biological DiversityOhio State University
    • Department of Evolution, Ecology, and Organismal BiologyOhio State University at Newark
  • Christine M. Janis
    • Department of Ecology and Evolutionary BiologyBrown University
Original Paper

DOI: 10.1007/s10914-006-9006-6

Cite this article as:
Hunter, J.P. & Janis, C.M. J Mammal Evol (2006) 13: 89. doi:10.1007/s10914-006-9006-6

Abstract

The persistent finding of clades endemic to the southern continents (Afrotheria and Xenarthra) near the base of the placental mammal tree has led molecular phylogeneticists to suggest an origin of Placentalia, the crown group of Eutheria, somewhere in the southern continents. Basal splits within the Placentalia have then been associated with vicariance due to the breakup of Gondwana. Southern-origin scenarios suffer from several problems. First, the place of origin of Placentalia cannot be reconstructed using phylogenetic reasoning without reference to outgroups. When available outgroups are considered, a Laurasian origin is most parsimonious. Second, a model of pure vicariance would require that basal placental splits occurred not with the breakup of Gondwana, but of Pangea in the Late Triassic—Early Jurassic. This event long preceded even the oldest molecular divergence estimates for the Placentalia and was coeval only with the earliest mammals in the fossil record. Third, a problem with the number of dispersal events that would be required emerges under different southern-origin scenarios. In considering the geographic distribution of the major placental clades at their first appearance (mostly Early Cenozoic), it becomes clear that a Laurasian center of origin would require fewer dispersal events. Southern-origin models would require at least twice the number of dispersal events in comparison with a model of Laurasian origins. This number of required dispersal events increases if extinct groups of placental mammals are also considered. Results are similar assuming a morphology-based phylogeny. These facts, along with earlier findings speaking against a major placental radiation deep in the Cretaceous without leaving fossil evidence, suggest an origin of Placentalia somewhere in Laurasia with few supraordinal splits occurring before the last 5–10 million years of the Cretaceous.

Keywords

PlacentaliaEutheriaBiogeographyPhylogenyDispersalVicariance

Introduction

The time and place in which the Placentalia—the crown group of the extant eutherian mammals (Novacek et al., 1997)—originated and diversified are sources of debate among paleontologists and neontologists. Temporally, nearly all of the extant orders of placental mammals appeared in the fossil record within a short span of time near the beginning of the Cenozoic ∼65–55 million years ago (Ma). However, molecular clock estimates of divergence time suggest earlier splits among ordinal lineages over a longer span of time in the Cretaceous 105–65 Ma (Eizirik et al., 2001; Murphy et al., 2001b; Springer et al., 2003), with claims of divergences as deep as 129 Ma (Kumar and Hedges, 1998). Geographically, the diversity of stem eutherians in the Late Cretaceous of Asia tens of millions of years before their undoubted occurrence in North America (Cifelli, 2000), and numerous placental mammal dispersals out of Asia in the Early Cenozoic (Beard, 1998, 2002), suggest the importance of Asia in early stem eutherian and later crown placental mammal evolution.

Early molecular divergence estimates, however, along with negative evidence (paucity of Late Cretaceous Gondwanan mammalian fossils), have led molecular phylogeneticists to suggest a hidden radiation of placental mammals in the Cretaceous somewhere among the southern continents (Hedges et al., 1996). This suggestion has been further fueled by the consistent support for an endemic basal, or near basal, African clade, Afrotheria, in molecular phylogenetic trees (Springer et al., 1997; Stanhope et al., 1998; Eizirik et al., 2001; Madsen et al., 2001; Murphy et al., 2001a, 2001b; van Dijk et al., 2001). Paleontologists have emphasized dinosaur extinction and ecological release as important factors in the evolutionary radiation of (including within) the placental mammal orders (Archibald and Deutschman, 2001), while molecular workers have invoked continental drift and vicariance in isolating the ordinal lineages from one another (Hedges et al., 1996; Eizirik et al., 2001).

The notion that basally-branching placental mammal clades might represent some vicariance-related relicts of ancestral stocks is at first glance very appealing. This notion fits in both with the breakup of Gondwana, and also with molecular evidence that would place the divergence of major placental clades much earlier than their appearance in the fossil record. The early divergence estimates and the reasoning employed in constructing a scenario of Gondwanan placental radiation concern us. Foote et al.(1999) pointed out that there was little evidence for any modern placental groups prior to the start of the Cenozoic (see also Novacek et al., 1997, 2000; Novacek, 1999; Wible et al., 2004). They also noted, based on quantitative estimates of preservation using data from other, nonplacental mammals, that if modern placentals had radiated deep in the Cretaceous without leaving fossils, the preservational record would have to be at least an order of magnitude worse than actually observed. This argument holds whether one supposes that the record of modern placental mammals begins ∼65 Ma with the first appearance of undoubted members of the extant orders (Table 1); ∼88–85 Ma with the occurrence of possible members of a supraordinal group of modern placentals, the “zhelestids” (Archibald, 1996; Nessov et al., 1998); or even accepting only the more modest, and better justified, claim of ∼105 Ma for the first placental mammal divergences (see reply by Foote and colleagues in Archibald et al., 1999).
Table 1

Distribution of Placental Mammals at Their Appearance in the Fossil Recorda

Afrotheria

Africa (Early Eocene), possibly Eurasia (Late Paleocene)b

Xenarthra

South America (Late Paleocene)c

Euarchontoglires

 

 Euarchontad

North America and Eurasia (Early Paleocene)e

 Gliresf

Asia (Early Paleocene, possibly Late Cretaceousg)

Laurasiatheria

 

 Eulipotyphlah

Laurasia (Early Paleocene)

 Chiroptera

North America and Australia (Latest Paleocene or Early Eocene)

 Carnivorai

North America (Early Paleocene)

 Pholidotai

Europe (Middle Eocene), possibly North America (Early Paleocene)j

 Perissodactylai

Laurasia (Early Eocene), possibly earlier in Asia (Late Paleocene)k

 Cetartiodactylal

Laurasia (Early Eocene)

Extinct boreoeutherian orders

 

 “Proteutheria”m

Laurasia (Early Paleocene, possibly Late Cretaceous)

 “Amblypoda”m

Laurasia and South America (Early Paleocene)

 Creodontam

Laurasia (Late Paleocene) and Africa (Early Eocene)

 “Meridiungulata”m

South America (Late, possibly Early Paleocene)

aFrom McKenna and Bell (1997) and Murphy et al.(2001b), supplementary material, with additions. See text for full explanation.

bAfrotheria comprises the orders Proboscidea (elephants), Sirenia (sea cows) (these two grouped together as the Tethytheria), Hyracoidea (hyraxes), Tubulidentata (aardvarks), Macroscelidea (elephant shrews), and Afrosoricida (tenrecs and chrysochlorids, or golden moles). In the morphological phylogeny, the first four groups are classified as “subungulates” or “paenungulates” within the grandorder Ungulata; elephant shrews are grouped with Glires as the Anagalida; and the afrosoricids are grouped with the Lipotyphla (=Insectivora). Possible proboscideans, the anthracobunids, occur in Asia in the Early Eocene, soon after the appearance of Proboscidea in northern Africa. Embrithopoda, an extinct Afrotherian order, occur in Eurasia and Africa (Early Eocene), but possible embrithopods, the phenacolophids, occur earlier (Late Paleocene) in Asia. See text for discussion of the problems of the biogeography of the Tubulidentata.

cThe Xenarthra is comprised of the sloths, anteaters, and armadillos, plus several extinct groups. Several Early Cenozoic Laurasian “edentate” mammals (Palaeanodonta, Ernandon in the Late Paleocene of Asia, and Eurotamandua in the Middle Eocene of Europe) have been allied with the Xenarthra in the past. Palaeanodonts and Ernandon instead may be members of, or allied with the Pholidota (Rose and Lucas, 2000). Although it is unlikely that Eurotamandua is a mymecophagid (i.e., anteater) as originally described, its phylogenetic placement remains unresolved. Eurotamandua could be a sister-taxon to the Pilosa (i.e., sloths and anteaters), a stem xenarthran, or a pholidotan (Gaudin and Branham, 1998). Eurotamandua, Ernanodon, and palaeanodonts, however, all lack truly xenarthrous vertebrae (Gaudin, 1999), suggesting that their resemblances to Xenarthra may be due to convergence.

dPrimates, Scandentia (tree shrews), and Dermoptera (flying lemurs). The “Archonta” of the morphological tree also includes the Chiroptera (bats).

ePlesiadapiformes, no record of others (Euprimates) until the Early Eocene.

fRodentia and Lagomorpha (rabbits and hares). In morphological trees, the Glires are often grouped with the Macroscelidea (elephant shrews).

gLate Cretaceous if Asian zalambdalestids are allied with Glires; Novacek et al.(1997) and Wible et al.(2004) reported that zalambdalestids and other Cretaceous Mongolian eutherians retain primitive postcranial and cranial features that place them outside of crown Placentalia, whereas work by Archibald et al.(2001) on dental characters supports a link with Glires.

hEulipotyphla includes Erinacidae (hedgehogs), Soricidae (shrews), and Talpidae (moles). The Lipotyphla of the morphological tree also includes Tenrecidae (tenrecs) and Chrysochloridae (golden moles).

iCarnivora (carnivores), Pholidota (pangolins), and Perissodactyla (odd-toed ungulates) are grouped together as the Ferungulata in the molecular tree.

jPalaeanodonts and epoicotheres, possible pholidotans, occur in North America from the Early Paleocene (Rose and Lucas, 2000).

kMeng et al.(1998) and Bowen et al.(2002).

lCetartiodactyla, a grouping also found in the morphological tree, consists of Cetacea (whales) and Artiodactyla (even-toed ungulates).

mThe composition and affinities of these extinct groups are discussed in the text.

Less extreme hypotheses involving shorter periods of nonpreservation (∼5–10 million years) are more plausible, for example, if most interordinal divergences occurred between ∼75 or 70 and 65 Ma, or if fewer placental mammals lineages extended deep into the Cretaceous (e.g., <7). Although recent molecular results now support a more clustered timing of most divergence events suggestive of rapid radiations (Eizirik et al., 2001; Springer et al., 2003) rather than the widely spaced estimates originally proposed (Hedges et al., 1996; Kumar and Hedges, 1998), the earliest divergence estimates are still ∼105 Ma. Other paleontologists have also cast doubt on the earliest molecular divergence time estimates on other evidence including the nonrandom temporal clustering of ordinal appearance events in the Early Paleogene (Archibald and Deutschman, 2001), “ghost lineages” implied in a morphology-based phylogeny (Novacek, 1999), and the use of “control” groups with similar preservation potential as placental mammals (Benton, 1999; Novacek, 1999).

In this paper, we mainly refer to the newer phylogeny of mammals supported by molecular data (see Springer et al., 1997; Stanhope et al., 1998; Eizirik et al., 2001; Madsen et al., 2001; Murphy et al., 2001a, 2001b; van Dijk et al., 2001; Springer et al., 2003), because this phylogeny has been used in the proposals of southern placental origins. However, we also consider how our proposals fit in with the more traditional phylogeny obtained from morphological data (see McKenna, 1975; Novacek, 1986; Novacek and Wyss, 1986; Novacek, 1992; McKenna and Bell, 1997; Novacek, 1999). The differences between these phylogenies are illustrated in Figs. 1 and 2, which are discussed later in the context of the biogeographic scenarios. Table 1 also notes some of the differences between the two phylogenies.
https://static-content.springer.com/image/art%3A10.1007%2Fs10914-006-9006-6/MediaObjects/10914_2006_9006_Fig1_HTML.gif
Fig. 1

Biogeographic and temporal occurrences of the major clades of placental mammals, phylogeny derived from molecular methodologies (Murphy et al., 2001b). Key: Solid dark bars = Afrotheria; Scallop-shaded bars = Xenarthra; Hatched bars = Boreoeutheria (left-to-right hatching = Euarchontoglires, right-to-left hatching = Laurasiatheria); Horizontally-striped bars = extinct taxa

https://static-content.springer.com/image/art%3A10.1007%2Fs10914-006-9006-6/MediaObjects/10914_2006_9006_Fig2_HTML.gif
Fig. 2

Biogeographic and temporal occurrences of the major clades of placental mammals, phylogeny derived from morphological data (Novacek, 1992)

Alternative interpretations to Foote et al.(1999)

Since the publication of Foote et al.(1999), other quantitative studies incorporating data on fossil occurrences have been used to argue for early-origins hypotheses. Indeed, the approach of Foote et al.(1999) itself would be compatible with any early-origins hypothesis in which the amount of missing evolutionary history assumed by the hypothesis fall below certain thresholds. The amount of missing evolutionary history, which increases with the number of lineages and the length of time thought to be missing from the fossil record, cannot be greater than allowed given empirical estimates of the preservation rate of Late Cretaceous placental mammals, which Foote and colleagues calculated in multiple ways from the known occurrences of all known Late Cretaceous mammals (Foote et al., 1999) or just Late Cretaceous eutherian mammals (see reply by Foote and colleagues in Archibald et al., 1999). In this way, the approach by Foote et al.(1999) established explicit criteria which early-origins hypotheses must meet in order to be plausible.

In one example already mentioned earlier, Archibald and Deutschman (2001) demonstrated using Monte Carlo methods that a nonrandom clustering of placental ordinal appearance events occurred in the Early Cenozoic and that this clustering could not be attributed to better sampling in the Early Cenozoic as compared to the Late Cretaceous. These authors used this result to reject the “Short Fuse Model” hypothesis of placental ordinal origins, in which the modern orders of placental mammals diversified deep in the Cretaceous, a hypothesis considered by these authors to be at odds with a literal reading of the fossil record of placental appearances. Archibald and Deutschman (2001) interpreted their result that the majority of placental ordinal originations must have occurred in the Early Cenozoic, to be compatible with either the “Explosive Model” or the “Long Fuse Model” of placental ordinal origins. According to the scheme developed by Archibald and Deutschman (2001), in the “Explosive Model” splits between ordinal lineages occurred after the K/T boundary, whereas in the “Long Fuse Model” stem lineages of the placental orders (but not the crown members of the orders themselves) extend into the Late Cretaceous. As the authors themselves argue, their analysis does not so much support the “Long Fuse Model” explicitly as it simply does not deal with the issue of the duration of these stem lineages at all because their approach addresses ordinal appearances only. In other work, however, Archibald has argued for the extension of placental supraordinal lineages into the Late Cretaceous on the basis of phylogenetic interpretation of “zhelestid” (see Archibald, 1996) and zalambdalestid (see Archibald et al., 2001) fossils, indirectly in support of a “Long Fuse Model”.

Archibald and Deutschman (2001) explicitly recognized that these three models of ordinal origination and diversification form a continuum, with each model differing from the other two in the assumed time since the origin of the placental orders and supraordinal lineages. These models also differ in their implied sum of species durations of placental mammals that would have escaped preservation. In this way, these models, or rather the continuum from which they derive, can be related directly to the origins hypotheses tested by Foote et al. (1999 and in Archibald et al., 1999) by altering three parameters: length of missing time, number of lineages, and diversification model. Although Archibald and Deutschman’s (2001) approach does not address the duration of ordinal stem lineages or supraordinal lineages, the approach of Foote et al.(1999) and in Archibald et al., 1999) explored a range of parameter values and diversification models along this continuum and concluded that an early-origins hypothesis, such as a “Long Fuse Model”, could be plausible so long as the combination of length of missing time and number of lineages remains below a certain threshold (for example, less than 5– 10 million years and fewer than seven lineages). More extreme versions of the “Long Fuse Model” would likely result in larger sum of species durations than can be accepted.

In another alternative approach, Tavare et al.(2002) recently proposed an age estimate of 81.5 Ma for the last common ancestor of extant primates, ∼27 million years (Myr) before the appearance of modern primates in the fossil record. Tavare et al.(2002) calculated this age-estimate by extrapolating a logistic diversification model, fit to the known Eocene-Recent primate fossil record, into the Cretaceous and Paleocene, where no modern primate fossils occur. The new age-estimate depends on the extension of Cenozoic-like primate diversification into the Cretaceous, which may or may not be appropriate. For this reason and for phylogenetic reasons, both of which are discussed later, we recommend caution in interpreting the generality of the results of Tavare et al.(2002) for primates to all placental mammals.

The model used by Tavare et al.(2002) assumes logistic diversification of modern primates from their origin to their Early Eocene appearance (54.8 Ma), a time span that includes the Cretaceous–Tertiary (K/T) boundary (65 Ma). The authors assume the same average species duration, 2–3 Myr implying an extinction rate of 0.5–0.3 per lineage per Myr, throughout diversification including pre-Eocene times. In this way, their model assumes that pre-Eocene and Eocene-Recent diversification dynamics were similar and that modern primate diversification was unperturbed at the K/T boundary. This assumption may be unrealistic. For North American mammals, the highest extinction rate observed in the fossil record is associated with the K/T boundary, followed by the highest observed origination rate (Alroy, 1999). The K/T boundary also marks a transition in diversification dynamics between low Cretaceous rates to high rates characteristic of the Paleocene evolutionary radiation (Van Valen, 1985; Alroy, 1999). Given a Paleocene burst in origination, as observed in other mammals, it may be unnecessary to hypothesize such a lengthy time of primate diversification.

Tavare et al.(2002) also compared the preservation rate implied by their logistic diversification model and the 81.5 Ma divergence estimate, to preservation rates (r) calculated by Foote et al.(1999) from actual first and last occurrences of Late Cretaceous mammalian species. Because the pre-Eocene primate preservation rate calculated by Tavare et al.(2002) is an order of magnitude lower than Foote et al.’s (1999) rate for Late Cretaceous mammals as a group, Tavare et al.(2002) concluded that Foote et al.’s (1999) preservation rate overestimates that of pre-Eocene primates. However, this comparison may not be valid. Tavare et al.’s (2002) primate preservation rate is logically equivalent to rmax, which is a maximum preservation rate given certain hypotheses about evolutionary history, and not r, which is an empirical estimate of preservation rate (Foote et al., 1999). That is, the primate preservation rate estimated by Tavare et al.(2002) is a hypothesis of what the preservation rate of this group must have been in order for their diversification model and estimated time of origin to be compatible with complete nonpreservation. In a similar way, Foote et al.(1999) calculated rmax, a hypothetical maximum allowable preservation rate, under a variety of diversification models and parameter values t (time of clade origin) and n (number of lineages) for crown placental mammals. Foote et al.(1999) then tested these hypothetical rmax values against empirical estimates of preservation rate. In order to maintain comparability between the two approaches, Tavare et al.’s (2002) primate rate should be tested against an appropriate empirically derived estimate of preservation rate. If the preservation rate calculated by Foote et al.(1999) for Late Cretaceous mammals overestimates that of pre-Eocene primates (as per Tavare et al., 2002), then no appropriate test of the maximum allowable preservation rate derived by Tavare et al.(2002) exists. In this way, Tavare et al.(2002) preservation rate hypothesis for pre-Eocene primates remains untested.

Nevertheless, modern primates did not evolve in a phylogenetic vacuum. Regardless of whether one accepts a molecular or morphological phylogeny, the origin of modern primates at 81.5 Ma would imply the contemporary occurrence of other lineages of modern placental mammals. One must also take into account the nonpreservation of these lineages, some of which are thought to have originated or diversified in the northern hemisphere where the Late Cretaceous and especially Paleocene mammalian fossil records are reasonably well sampled. Considering primates alone, as if they were the only group of modern placental mammals, the approach of Foote et al.(1999) would offer little objection to their nonpreservation for ∼27 Myr (Table 2). However, the probability of complete nonpreservation diminishes if one takes into account the relatives of primates. Modern primates at 81.5 Ma would imply the origin of at least five other lineages (Afrotheria, Xenarthra, Scandentia+Dermoptera, Glires, and Laurasiatheria) assuming a molecular phylogeny (Murphy et al., 2001b) or four lineages (Edentata, Scandentia, Volitantia, and a hypothetical lineage ancestral to all other placental mammals) assuming a morphological phylogeny (Novacek, 1992). Even reducing T, the time gap, to 17 Myr to recognize an appearance of primates that coincides with the first undoubted occurrence of modern placental mammal orders ∼65 Ma, the phylogenetic implications of 81.5 Myr old primates would remain considerable (Table 2). For this reason, we consider it probable that modern primates began to diversify much more recently than 81.5 Ma.
Table 2

Analysis of Early-Origin Hypotheses Implied by Results of Tavare et al.(2002)

 

N

T

S

rmax

P

Primates only

1

27

27

0.026

0.44

Strepsirrhines + Haplorhines

2

27

39

0.018

0.31

Crown placental mammals (molecular tree)

6

17

47

0.015

0.24

Crown placental mammals (morphological tree)

5

17

42

0.016

0.28

N, number of lineages present at first fossil appearance; T, time from postulated origin to first fossil appearance; S, implied sum of species durations under exponential diversification; rmax, maximum preservation rate compatible with non-preservation; P, probability of complete non-preservation given r=0.03 per lineage per Myr, calculated as per Foote et al.(1999).

Vicariance, dispersal, and the “Garden of Eden”

Foote et al.(1999) proposed several alternatives that might explain the conflict between the molecular estimates of divergence time and their proposal that such estimates are unlikely (see Benton, 1999, for a similar list). Although the favored alternative was temporal variation in rate of sequence divergence associated with evolutionary radiation, Foote et al.(1999) conceded that modern placental mammals could have radiated somewhere on the southern continents, where the Late Mesozoic fossil record was considerably poorer. They postulated (at the time with intentional irony) the notion of some southern “Garden of Eden” in which placental mammals could have diversified undetected. They suggested that future paleontological discoveries could test the “Garden of Eden” hypothesis. Perhaps unfortunately, this notion of a mammalian “Garden of Eden” has since been taken up and championed by other authors (Archibald et al., 1999; Murphy et al., 2001b; Rich et al., 2001b) in a somewhat different sense from that originally intended. Arguably, the “Garden of Eden” seems to have come to mean any Gondwanan root of modern placental mammals rather than its original sense of a large-scale evolutionary radiation of placentals in Gondwana deep in the Cretaceous.

However, any scenario that claims for an ancient (i.e., prior to ∼75 or 70 Ma) origin for the placental groups on the southern continents needs to deal with several thorny issues. First, there is a taxon-sampling problem inherent in basing inferences of the distribution of only the extant clades, that is, the crown group. Claims for a southern origin for Placentalia are supposed to have gained support by the placement of Afrotheria and Xenarthra, clades endemic to the southern continents Africa and South America, respectively, in basal positions in the phylogenetic tree of living placental mammals (Eizirik et al., 2001; Madsen et al., 2001; Murphy et al., 2001a, 2001b). In actuality, inferring the place of origin of placental mammals in this manner (i.e., from the root of the tree), without reference to appropriate outgroups, produces equivocal results (Fig. 3(A)). This reasoning also ignores the extinct outgroups to Placentalia, which are useful in framing biogeographic scenarios. For example, if the known Cretaceous eutherians in Laurasia are nonplacentals, that is outside of the crown group (as per Novacek et al., 1997, 2000; Novacek, 1999; Ekdale et al., 2004; Wible et al., 2004), then the last common ancestor of modern placental mammals is reconstructed as most likely Laurasian and the geographic distribution of the Afrotheria and Xenarthra becomes best explained by dispersal out of Laurasia (Fig. 3(B)). One might argue that stem eutherians or even basal placental mammals of the right age merely await discovery on some southern continent that has hitherto been poorly sampled paleontologically. Doing so, however, would require hypothetical dispersal events of either an ancestral placental lineage from Laurasia to Gondwana or of multiple stem eutherian lineages from Gondwana to Laurasia.
https://static-content.springer.com/image/art%3A10.1007%2Fs10914-006-9006-6/MediaObjects/10914_2006_9006_Fig3_HTML.gif
Fig. 3

Phylogenetically derived biogeographic reconstructions of Placentalia. Topology from Murphy et al. (2001) without an outgroup (A) and with Asioryctotheria as an outgroup to Placentalia (B). Reconstruction B would be the same even if the Zalambdalestidae is removed from the Asioryctotheria as suggested by Archibald et al.(2001), if Metatheria (including marsupials), an initially Laurasian group, is instead used as an outgroup, or if the last common ancestor of marsupials and placentals, a Laurasian taxon according to a growing body of fossil evidence (Luo et al., 2001, 2002, 2003; Ji et al., 2002), is used as an outgroup. designates an entirely extinct taxon

Even assuming a southern origin of Placentalia, a second issue of concern is the origin of the northern groups (i.e., Boreoeutheria of Murphy et al., 2001b, which includes all extant orders exclusive of Afrotheria and Xenarthra). Because many of these groups (or at least basal members of these clades) are already present in Laurasia (North America and/or Eurasia) at the start of the Cenozoic, it is unlikely that they could represent a single immigration event of an ancestral stock at 65 Ma. On the other hand, it is also unlikely that the ancestral stock of the Boreoeutheria migrated deep in the Cretaceous to the northern latitudes and subsequently diversified there because the Laurasian Cretaceous fossil record is sufficiently well sampled to be confident that at least some of these groups would have left a record (as per Foote et al., 1999). That is, the hypothesis of southern origins/ancient splits of major placental clades would necessitate that the diversification of the Boreoeutheria occurred in some southern “Garden of Eden”, where the major groups remained undetected until dispersal to the northern hemisphere near the start of the Cenozoic. There is also the problem that these boreoeutherian clades then left little or no evidence of their original presence on any southern continent at the beginning of the Cenozoic, but that is another issue.

Given that the southern-origins hypothesis of Placentalia is in need of further testing, we present here an evaluation of this hypothesis based on the notion of parsimony of dispersal events. We construct different biogeographic scenarios, each involving the origination of placentals on different continental land masses (Laurasia, Africa, South America, and Australia). We then evaluate the “biogeographic parsimony” of each scenario in terms of the number of dispersal events that would be required to account for the observed distribution of major placental clades at the start of the Cenozoic. Given that intercontinental dispersal events are generally considered to be rare and unlikely events (McKenna, 1973), the scenario that requires the fewest number of dispersals (i.e., the most biogeographically parsimonious) should be favored, all other things being equal. Application of the principle of parsimony is not particular to biogeographic events, but rather is a general operational principle of scientific model building and selection among competing theories.

As detailed later, a southern origin of placental mammals (on any continent), with the branching pattern of phylogeny representing initial vicariance events followed by the dispersal of the boreoeutherians, results in the need to account for a large number of dispersal events. In contrast, a northern (Laurasian) origin for all placental mammals, with the dispersal of the more basal clades into the southern continents (unrelated to the pattern of continental breakup, and most likely at a much later date) results in many fewer dispersal events to account for. Of course, it is not required that the actual evolutionary history of placental mammals unfolded in the most parsimonious manner in terms of numbers of dispersal events. Nevertheless, biogeographic scenarios should be judged on their own merits in accounting for the known facts in addition to their congruence to phylogenetic (e.g., Fig. 3) and evolutionary hypotheses. To our knowledge, this paper is the first attempt to explore fully the consequences of competing biogeographic scenarios of placental mammal origins for the dispersal history of this clade (see Stewart and Disotell, 1998, for an analogous application to hominoids).

The problem of extinct placental groups

The picture of placental mammal diversification is somewhat complicated by the consideration of certain extinct groups that, at least on morphological grounds, are most likely within crown Placentalia (Table 1; Figs. 1 and 2). A consideration of the extant placental mammal orders alone would ignore these groups, but some of them are important to biogeographic scenarios, especially those whose geographic and temporal distributions differ from their nearest presumed relatives. Although our study yielded similar results with and without these extinct groups, they should be considered, even if only for the sake of completeness. We will consider some major ones in this paper. We have chosen, for the purposes of this paper, to consider these various groups as representing clades requiring no more than a single dispersal event each. We fully acknowledge the probable paraphyly or even polyphyly of some or all of these groups, and some of these problems are discussed later. However, as the phylogeny of some of these groups is poorly resolved, constructing alternate scenarios based on alternate phylogenies, each requiring multiple dispersals would have been extremely cumbersome. Note that in considering these groups as single clades we are biasing our argument against our main thesis; that is, we are probably underestimating the number of dispersals.

The first of these groups are the endemic South American ungulates or “Meridiungulata” including the orders Litopterna, Notoungulata, Pyrotheria, Astrapotheria, and Xenungulata. These orders of hoofed mammals radiated during the Cenozoic in South America, when this continent was isolated from the other continents, converging on a range of body forms seen in extant orders of ungulates and rodents. Because the enigmatic Laurasian order Arctostylopida is no longer considered to be related to the notoungulates (Cifelli et al., 1989), the “meridiungulates” can be considered a completely South American group (except for a few species that migrated north during the Great American Interchange in the Plio-Pleistocene, ∼2 Ma).

It is debatable whether these orders form a monophyletic group, the Meridiungulata as per McKenna (1975) and McKenna and Bell (1997), or whether one or more of these orders shares a more recent ancestry with eutherians of other continents. For example, litopterns and the didolodont and mioclaenid archaic ungulates of the Paleocene of South America may be derived from a North American mioclaenid ancestor; that is, from the otherwise Laurasian ungulates (Cifelli, 1983, 1993; Muizon and Cifelli, 2000). On the other hand, pyrotheres and xenungulates may share an ancestry with the uintatheres, a group with a Laurasian distribution, from a Pseudictops-like “anagalidan” (i.e., Glires-related) ancestor (Lucas, 1993). Muizon and Cifelli (2000) suggested, assuming a northern continent origin, that a minimum of two migrations into South America would be needed to account for the presence there of the “Meridiungulata”: one that gave rise to the South American mioclaenids, didolodonts, litopterns, and notoungulates on the one hand, and a second that gave rise to pyrotheres, astrapotheres, and xenungulates on the other.

Grafting these extinct groups into the molecular phylogenetic tree is difficult (Archibald, 1999a,b). For the purposes of this paper, we consider the meridiungulates as a single clade, as the sister taxon to the Ferungulata plus Cetartiodactyla (see Fig. 1 and Table 1 for an explanation of the mammalian orders included in these superordinal groupings). In the morphological tree they would be the sister taxon of the Perissodactyla plus Cetartiodactyla (Fig. 2). If a diphyletic origin is assumed, considering current ideas on the ancestry of each group, one would most probably root litopterns and notoungulates within the Ferungulata+Cetartiodactyla, and pyrotheres, xenungulates, and astrapotheres with the Glires. Regardless of their exact placement, the important point is that both groups would be nested well within the Boreoeutheria. As a result, any biogeographic scenario for placental mammals must account either for the migration of the “meridiungulates” into South America, or the migration of the rest of the Boreoeutheria out of South America leaving the “meridiungulates” behind.

The second group is the “Proteutheria”, a paraphyletic or polyphyletic assemblage of primitive insectivorous mammals including leptictids, palaeoryctids, pantolestids, and apatemyids. “Proteutherians” include Laurasian, mainly early Cenozoic (it is uncertain whether any Cretaceous forms belong to the well-established Cenozoic “proteutherian” groups) small-bodied mammals that have diverged little from the ancestral eutherian morphotype. Nevertheless, the “proteutherians” are probably not all basal eutherians, as a convincing case has been made that at least the Cenozoic leptictids may be closely related to the extant lipotyphlan insectivores (Novacek, 1986; MacPhee and Novacek, 1993), that is, the Eulipotyphla of Murphy et al.(2001b). (See Table 1 for an explanation of the difference between Lipotyphla and Eulipotyphla.) Furthermore, none of the “proteutherian” groups has been linked with either the Xenarthra or any of the groups considered within the Afrotheria, and thus the whole assemblage is best considered boreoeutherian.

Because both the time of appearance and the geographic distribution of “proteutherians” coincide approximately with that of eulipotyphlans, biogeographic scenarios must account for the simultaneous appearance of the two “insectivore” groups: “Proteutheria” and Eulipotyphla. Although possible “proteutherians” are also known from the Late Cretaceous Deccan intertrappean beds of India (Prasad and Sahni, 1988; Prasad et al., 1995) and the early Paleocene of South America, it is unknown exactly how these fossils pertain to the well-circumscribed “proteutherian” groups of the Early Cenozoic of North America and Eurasia. Rana and Wilson (2003) recently compared the teeth of Deccan eutherians with the teeth of a host of other problematic eutherians from the Late Cretaceous of Laurasia, and found similarities that they interpreted to be due to a common, perhaps boreoeutherian ancestry (i.e., neither demonstrably xenarthran nor afrotherian), an observation still most consistent with a migration into India from Eurasia following collision.

The third group is the Creodonta, a group of archaic carnivorous mammals known mainly from Laurasia (Gunnell, 1998), although a radiation of derived hyaenodontid creodonts is also known in the Eocene through Miocene of Africa (Gheerbrant, 1995; Holroyd et al., 1996; Morales et al., 1998a, 1998b; Holroyd, 1999; Gheerbrant et al., 2002). Creodonta may be monophyletic or diphyletic with the two distinctive families, Oxyaenidae and Hyaenodontidae, arising separately (Gunnell, 1998). Creodonta and Carnivora are usually grouped together as the Ferae in morphological phylogenies (see Fig. 2). Creodonta may be most closely related to the Carnivora among extant mammals with both arising from the palaeoryctid “proteutherians” (Gunnell, 1998). Alternatively the two orders may instead have different origins, with carnivorans arising from ancestors that were dentally more primitive than the palaeoryctids (Fox and Youzwyshyn, 1994). Nevertheless, creodonts have not been linked with either the Xenarthra or the Afrotheria and, given their possible sister-taxon relation with Carnivora and likely ancestry from palaeoryctid “proteutherians”, are best considered boreoeutherians, at least provisionally.

The early history of the Creodonta and Carnivora is complicated, with carnivorans appearing first in the Early Paleocene in North America, followed by oxyaenid creodonts in North America and hyaenodontid creodonts in Asia in the Late Paleocene. Finally, an invasion of Asian-derived hyaenodontids into North America occurred near the Paleocene/Eocene boundary, along with many other mammalian taxa. Even if creodonts are sister taxa to carnivorans, their complicated early history suggests at least several separate dispersal events.

The fourth group is the “Amblypoda”, a probably polyphyletic assemblage of large ungulate-like mammals including the orders Pantodonta, Dinocerata (the uintatheres), Taeniodonta, and Tillodonta. “Amblypods” had a Laurasian distribution with the exception of a single rather derived pantodont from the Early Paleocene of South America. Muizon and Cifelli (2000) thus hypothesized three dispersal events to account for all the eutherians in South America at the start of the Cenozoic (one for pantodonts and two for the “meridiungulates”). Also, as mentioned above, the uintatheres may be related to the South American pyrotheres and xengulates, united in a Uintatheriamorpha and sharing an ancestry among basal Glires. Pantodonts and tillodonts may be more closely related to each other than to other groups, and both groups seem to share an Asian ancestry from a “proteutherian” ancestor. Taeniodonts have been considered by many to be derived from a Didelphodus-like “proteutherian”, and transitional forms have been described (Eberle, 1999; but see Fox and Naylor, 2003, for an alternative interpretation). It is thus likely that all of the “amblypods” are boreoeutherian.

An additional important order, which on all morphological evidence belongs with the Afrotheria, is the Embrithopoda (including the rhinoceros-like arsinoetheres). The possible sister-taxon relationship of embrithopods to Proboscidea, the geographic distribution of the embrithopods around the Tethys seaway (i.e., both Asia and northern Africa), and the early appearance of possible proboscideans, the anthracobunids, in the Late Paleocene of Asia, all point to ambiguity in the place of origin of the Proboscidea (Beard, 1998) and other afrotherians. Moreover, the oldest known proboscideans in Africa, previously thought to be of Late Paleocene (Gheerbrant et al., 1996), are now considered to be of Early Eocene (Gheerbrant et al., 2002), that is, younger than comparable fossils in Asia. At the very least, even if an African origin of Proboscidea and the Afrotheria is assumed, biogeographic scenarios must still account for the presence in Eurasia of embrithopods and anthracobunids. Further complicating the issue of Afrotherian origins is postcranial evidence rooting another afrotherian group, the elephant shrews (order Macroscelidea), among Laurasian hyopsodontid condylarths (Zack et al., 2005), something suspected from previously reported dental similarities (Simons et al., 1991; Butler, 1995). Thus, there is growing evidence from the fossil record for the origin of Afrotheria, or afrotherian subgroups, in Laurasia.

One final group of eutherian mammals that we do not consider further here, but which may eventually prove to be important for biogeographic scenarios, are the archaic ungulates of the order Condylarthra. Condylarths are a paraphyletic or possibly polyphyletic group of eutherian mammals thought to be ancestral to, or the sister taxon of the living ungulate mammals (Archibald, 1998). Condylarths appeared in the fossil record near the K/T boundary, and subsequently underwent a radiation in the Paleogene primarily in North America (Hunter, 1997, 1999) and Eurasia (Gingerich et al., 1997, 1999), with a separate radiation in South America of didolodont and mioclaenid condylarths (Muizon and Cifelli, 2000). Dental resemblances between the Late Cretaceous zhelestids and later (Paleogene) condylarths suggest that the Condylarthra may have roots in the Late Cretaceous (Archibald, 1996; Nessov et al., 1998). However, given that this resemblance is largely restricted to the teeth and jaws, the inference that condylarths and zhelestids are closely related should be drawn only with caution (Novacek, 1999). Moreover, cranial material attributed to zhelestids displays primitive features otherwise found outside of crown placentals (Ekdale et al., 2004). It also remains unclear whether any of the condylarths (e.g., Protungulatum) is closely related to any of the living ungulate orders (Ji et al., 2002). Furthermore, extant ungulates only form a monophyletic group in morphological (Novacek, 1992), but not molecular (Murphy et al., 2001a, 2001b) phylogenies, further obscuring the relationships of the Condylarthra. We decided not to include the Condylarthra in counts of dispersal events, assuming either that condylarths are unrelated to living ungulates, or that their distribution does not require that they dispersed separately from the extant ungulates.

Biogeographic scenarios

Most commonly, area is optimized on a phylogeny, and biogeographic history reconstructed as if area were a character that can evolve (Beard, 1998; Stewart and Disotell, 1998; Strait and Wood, 1999; Stevens and Heesy, 2000; Murray, 2001). In this section, we instead adopt a different approach to evaluate paleobiogeographic scenarios keeping in mind that the best justified of them should optimize several criteria. First, the amount of missing evolutionary history required by a model should be minimized, in terms of number of missing lineages and length of missing time, to plausibly acceptable levels (Foote et al., 1999). Second, a model should be compatible with the constraints implied by current paleogeographic reconstructions and phylogenetic branching patterns. These constraints, however, should not be considered as hard and fast because they are based on estimates for the configuration of the continental land-masses and the relationships of placental mammal orders, respectively. Third, a model should account for the extinct orders to minimize taxon-sampling problems. Fourth, and most important for this study, a model should minimize the number of hypothetical dispersal events that they require following our maxim of “biogeographical parsimony”. Although other recent treatments (e.g., Archibald, 2003) have discussed biogeographic aspects of placental origins in light of current phylogenetic hypotheses and paleontological knowledge, our approach is the first to explore fully the consequences of alternative scenarios for the dispersal history of placental mammals.

Here we consider biogeographic models in the following order: (1) a model invoking pure vicariance; (2) a set of models assuming a molecular phylogeny invoking a mixture of vicariance and dispersal assuming either a southern-continent (Africa, South America, and Australia) or a northern-continent origin of Placentalia, with brief comments on corresponding models invoking pure dispersal; and (3) models assuming a morphology-based phylogeny.

Pure vicariance

A biogeographic model invoking only vicariance would claim that all of the splitting events among placental ordinal lineages occurred as a result of continental breakup. Although early claims that the southern continents were an important evolutionary center for placental mammals were somewhat vague about mechanism (Hedges et al., 1996), the most extreme version of these hypotheses would be one of pure vicariance. Given that the boreoeutherian placental orders seem endemic to the northern continents, vicariance cannot be the sole cause of placental distribution without accounting for some dispersal to the northern continents (mixed models, see later). The only other possible model of pure vicariance would be to relate placental diversification to the breakup not of Gondwana, but of Pangea, as implied by Hedges et al.(1996). This hypothesis would require that the major splits among placental lineages began with the breakup of Pangea starting with the separation of Laurasia and Gondwana in the Late Triassic and Early Jurassic (210–180 Ma) or with the further isolation of these supercontinents with the rise of the Tethys Sea by the Middle Jurassic (180–160 Ma). These events predate even the oldest molecular estimates for the start of the placental radiation by at least 30 million years, the appearance of stem eutherians by at least 60 million years, and the appearance of crown placentals by at least 75 million years. Indeed, they are congruent with the earliest fossil record appearance of any mammals. Thus, it is exceedingly unlikely that placental lineages were even in existence during the breakup of Pangea, and vicariance alone cannot explain modern placental diversification.

Mammals appeared before the breakup of Pangea, which undoubtedly profoundly affected the early evolutionary history of the group. The fossil record provides evidence for the importance of the breakup of Pangea for the major mammalian lineages known or inferred to have been present at that time. For example, the Gondwanatheria in South America and other Gondwanan landmasses, India and Madgascar (Krause et al., 1997), and the Multituberculata in North America and Eurasia, may represent vicariant mammalian lineages separated since the breakup of Pangea (Pascual, 1996). Similarly, the geographic distribution of Early Mesozoic mammals, Laurasian versus Gondwanan, is largely concordant with one recent phylogeny that calls for an early split between a Laurasian clade, Boreosphenida, that ultimately gave rise to metatherian and eutherian mammals, and a Gondwanan clade, Australosphenida, that gave rise to monotremes (Luo et al., 2001, 2002; Weil, 2001; but see Woodburne et al., 2003, for an alternative view). In either case, the mammalian lineages involved in this vicariance event are far removed from the extant orders of placental mammals.

Mixed vicariance and dispersal: Africa as “Garden of Eden”

Biogeographic models invoking a mixture of vicariance and dispersal call for early splits among basal lineages due to vicariance followed by dispersal of other, higher branching lineages into new geographic regions. Because vicariance alone cannot explain all of the major splits among modern placental mammals (discussed earlier), we must turn our attention to mixed models. Recent molecule-based biogeographic models for placental origins (Eizirik et al., 2001; Murphy et al., 2001b) also have been of this mixed type, calling for early splits associated with the breakup of Gondwana giving rise to Afrotheria and Xenarthra, followed by dispersal into Laurasia. We consider in turn each major Gondwanan continent beginning with Africa, which has received much attention from molecular biologists and paleontologists as a potential source for modern placental mammals. No Late Cretaceous mammals are currently known from Africa with certainty (Krause et al., 2003, recently described a gondwanathere from Tanzania that may be Early or Late Cretaceous) leaving great opportunity for speculation about evolutionary events on this landmass during that time. We discuss this first example of Africa as a potential “Garden of Eden” in detail because many of the factors that we consider also apply to other southern continent “Garden of Eden” hypotheses.

Although current molecular phylogenies of modern placentals are compatible with African origins for the group as a whole, these phylogenies by themselves do not provide stronger support for an African origin than for, say, an origin in South America or one of the northern continents (Fig. 3(A)). The finding of an endemic African clade at the base of the placental tree in fact can be accounted for with equal likelihood by dispersal of the ancestral lineage of the group into Africa as by the group’s origin there in situ. One phylogenetic topology that would provide stronger support to the origin of Placentalia in Africa would be one in which subclades endemic to Africa were scattered throughout the family tree of placental mammals or formed a paraphyletic array at the base of Placentalia. Such is the case when inferring the likely place of origin of any clade within a phylogenetic context (see Fig. 4 for hypothetical examples).
https://static-content.springer.com/image/art%3A10.1007%2Fs10914-006-9006-6/MediaObjects/10914_2006_9006_Fig4_HTML.gif
Fig. 4

Hypothetical biogeographic reconstructions illustrating how combinations of topology and geographic distribution support biogeographic dispersal scenarios. Parts A and D show two simple topologies, using a three-taxon statement, which differ in how taxon b from region 2 is related to taxa a and c from region 1. Parts B and C follow from part A by the addition of an outgroup that in part B is endemic to region 2 and in part C is endemic to region 1. Similarly, parts E and F follow from part D by the addition of outgroups. The case illustrated by adding an appropriate outgroup to the topology of Murphy et al.(2001b) in Figs. 1 and 3 (above) is essentially identical to the situation illustrated in part F. Similarly, adding an appropriate outgroup to the topology of Novacek (1992) in Fig. 2 (earlier) and Fig. 13 (later) is the same as the situation illustrated in part C

Assuming Africa as a Mesozoic “Garden of Eden” for certain modern placental lineages, we derive the number of dispersal events required by this biogeographic model (Table 3). The distribution of modern placental mammal lineages can be accounted for initially without reference to their extinct relatives, which yields an absolute minimum estimate of the number of vicariance and dispersal events required, albeit an unrealistically low one (Fig. 5). For the African-origins mixed model, the isolation of the Xenarthra in South America from the Afrotheria and Boreoeutheria in Africa can be viewed as the sole vicariant event. It is curious that no xenarthran is known from the Late Cretaceous or Early Paleocene in South America. Despite many Early Paleocene localities known in both Patagonia and Bolivia, the Xenarthra did not appear in the South American record until the Late Paleocene.
Table 3

Vicariance and Dispersal Events Required Assuming Africa as a “Garden of Eden” for Placental Mammalsa

 

Extant groups only

Full record

Vicariance events

In South America

Xenarthra

Xenarthra

Dispersal events

To South America

 

“Meridiungulata”

  

“Amblypoda” (Pantodonta)

To Laurasia

 

Embrithopoda

 

Euarchonta

Euarchonta

 

Glires

Glires

 

Eulipotyphla

Eulipotyphla

  

“Proteutheria” (possibly)

 

Chiroptera

Chiroptera

 

Ferungulatab

Carnivora

  

Creodonta

  

Pholidota

  

Perissodactyla

 

Cetartiodactyla

Cetartiodactyla

  

“Amblypoda”

Dispersal events

6

13 (14 including “Proteutheria”)

Total events

7

14 (15 including “Proteutheria”)

aExtinct groups are marked with a.

bAssuming a northern origin of “Meridiungulata”.

https://static-content.springer.com/image/art%3A10.1007%2Fs10914-006-9006-6/MediaObjects/10914_2006_9006_Fig5_HTML.gif
Fig. 5

World map illustrating dispersal (single-headed arrows) and vicariance (double-headed arrow) events implied assuming Africa as a “Garden of Eden”. See text for explanation. Extant placental lineages only illustrated

The African origins model necessitates the dispersal of the Boreoeutheria into Laurasia. However, current fossil record and molecular evidence would constrain this dispersal to being a later one of several supraordinal lineages, rather than an earlier one of a single lineage. Current divergence time estimates place the split of Boreoeutheria from other placental mammals at ∼95 Ma and the split between the two major clades of Boreoeutheria, the Euarchontoglires and Laurasiatheria, at ∼80 Ma (Murphy et al., 2001b). In turn various splits between ordinal lineages within the Euarchontoglires and Laurasiatheria have been estimated at between ∼80 and 65 Ma (Eizirik et al., 2001), the last corresponding approximately to the Cretaceous-Tertiary boundary. Thus, the divergence time estimates require that the radiation of boreoeutherian orders occurred in the Late Cretaceous, leaving few if any fossils in evidence for it. It is highly unlikely that such a radiation, if it occurred in the Late Cretaceous, could have occurred in Laurasia without leaving any fossils. As discussed earlier, the estimated completeness of the known Late Cretaceous Laurasian mammalian record is too great by at least an order of magnitude for it to be likely that a major placental radiation could occur in Laurasia without leaving fossil evidence (Foote et al., 1999). Thus, a biogeographic scenario calling for dispersal of a single boreoeutherian ancestral lineage ∼95 Ma into Laurasia, or even two ancestral lineages for the two major boreoeutherian subclades ∼80 Ma soon after their origin, followed by subsequent diversification in Laurasia cannot be supported. The origin in Africa (or some other “Garden of Eden”) of several boreoeutherian supraordinal lineages and followed later by their dispersal into Laurasia in the last 5–10 million years of the Cretaceous is better compatible with the results of Foote et al.(1999). By this time, the boreoeutherian lineages listed in Table 3 (Euarchonta, Glires, Eulipotyphla, Chiroptera, Ferungulata, and Cetartiodactyla) would have been in existence if one accepts the divergence time estimates of Eizirik et al.(2001).

Adding extinct placental orders further complicates this scenario (Fig. 6). First, we consider the issue of the “Meridiungulata.” In order to assume that the Ferungulata had not yet split into its constituent lineages before dispersing to Laurasia, it is necessary to assume a northern origin of the “meridiungulates,” if they are ferungulate-derived, and their subsequent dispersal south into South America. Supposing instead a southern origin of these “meridiungulates” would require that at least some of the various ferungulate lineages had already split before dispersing to Laurasia, greatly increasing the number of required dispersal events. The “meridiungulates” must have made their way into South America by the Early Paleocene either as one or, more likely, two lineages (Muizon and Cifelli, 2000) either from the northern continents or from Africa, assuming Africa as the site of boreoeutherian diversification. Nevertheless, to be conservative and minimize the number of required dispersals, we assume a monophyletic “Meridiungulata” and a single dispersal into South America.
https://static-content.springer.com/image/art%3A10.1007%2Fs10914-006-9006-6/MediaObjects/10914_2006_9006_Fig6_HTML.gif
Fig. 6

World map illustrating dispersal (single-headed arrows) and vicariance (double-headed arrow) events implied assuming Africa as a “Garden of Eden”. See text for explanation. Both extant and extinct placental lineages illustrated

Secondly, afrotherian mammals in the form of the embrithopods and anthracobunids, must have made their way into Eurasia to be sampled there by the Late Paleocene and Early Eocene. Additionally, aardvarks (Tubulidentata) may represent another biogeographical problem. Although aardvarks are a primarily African group, their earliest occurrence is in the Oligocene of Eurasia. Although there is no African fossil record of mammals for most of the Oligocene, had aardvarks originally been present in Africa and then migrated to Eurasia, one would expect to find them in the Eocene African fossil record in the Fayum. However, aardvarks are rare in general as fossils through their entire record, and so this absence may not be significant. Finally, the Sirenia (sea cows), the sister taxon to the Proboscidea, first appear in the early Middle Eocene in both the western Atlantic (Domning, 2001a) and the northern Tethys (Zalmout et al., 2003), while the most primitive forms, documenting the transition to aquatic habits, occur in the Caribbean (Domning, 2001b). To be conservative, we count only one afrotherian dispersal into Eurasia, in the Paleocene.

Third, the near simultaneous appearance of leptictid “proteutherians” and eulipotyphlans in the Early Paleocene, and perhaps latest Cretaceous, of Laurasia, along with their possible sister-taxon relationship, are compatible with these appearances representing separate dispersal events. Nevertheless, they are also compatible with a single dispersal of the ancestral lineage of both groups into Laurasia. Because of this ambiguity, we treat as required by the model only one dispersal of “insectivores” (i.e., “Proteutheria” plus Eulipotyphla) into Laurasia acknowledging that this estimate is possibly low.

Fourth, the occurrence of the “amblypod” groups (pantodonts, taeniodonts, tillodonts, and uintatheres) in North America and Eurasia must be accounted for as, at the very least, a single dispersal, acknowledging that this estimate is almost assuredly too low.

Fifth, among the Amblypoda, pantodonts also must have made their way into South America by the Early Paleocene (Muizon and Cifelli, 2000), requiring dispersal from either the northern continents or, in this model, from Africa.

Finally, the complicated early dispersal history of the Creodonta and Carnivora, and its implications for the diversification of the Ferungulata and “Garden of Eden” scenarios, requires some consideration. The temporally and geographically disjunct appearances of carnivorans (Early Paleocene of North America), oxyaenid creodonts (Late Paleocene of North America), and hyaenodontid creodonts (Late Paleocene of Asia), imply separate origins and dispersals of at least carnivorans and creodonts, and possibly of the two families of creodonts as well. Other ferungulate lineages (Pholidota and Perissodactyla) appear thereafter in the Late Paleocene or Early Eocene of Eurasia and North America, but are unknown earlier on the northern continents despite a reasonably well-sampled Paleocene northern continent fossil record. The early appearance of Carnivora (Early Paleocene) and the nesting of this order deep within the Ferungulata in molecule-based phylogenies, imply that all of the ferungulate lineages, as well as Cetartiodactyla, the sister taxon of Ferungulata, must have been in existence in the Early Paleocene. The appearance of the ferungulate orders in a temporally and geographically widely discordant fashion suggests, assuming an African or other southern continent “Garden of Eden”, separate dispersal events for each lineage at different times and by different dispersal routes out of that “Garden of Eden” into the northern continents.

Although the other end-member hypothesis, that the ancestor of ferungulates dispersed from Africa into the northern continents then diversified in obscurity, cannot be discounted completely, it involves further complications. If this diversification occurred in North America, as might be implied by the first appearance of the group in North America (Carnivora in the Early Paleocene), then one must invoke dispersal from North America into Asia to account for the distribution of perissodactyls, pholidotans, and hyaenodontid creodonts at their appearance in the fossil record. If, on the other hand, this diversification occurred in Eurasia, as might be implied by the proximity of Eurasia to Africa, the hypothetical place of origin of the ancestral lineage, then one must invoke dispersal from Asia into North America to account for the distribution of first carnivorans, and later perissodactyls, and oxyaenid creodonts at their appearance in the fossil record. In both cases, one must also somehow account for the lack of fossils of the ferungulate orders on the northern continent where this diversification is alleged to have taken place in the Late Cretaceous.

In sum, a mixed vicariance and dispersal model of African origins requires somewhere between 6 (living lineages only) and 13 (14 including “Proteutheria”) dispersal events (living plus extinct lineages), or between 7 and 14 (15 including “Proteutheria”) total biogeographic events, in order to be reconcilable to the known Late Cretaceous through Early Cenozoic mammalian fossil record. Estimates of the number of dispersals of extinct lineages vary depending on the assumptions made regarding how many boreoeutherian lineages must have dispersed out of Africa into the northern continents, as well as how one considers the affinities of these extinct groups. However, considering extinct placental taxa requires more dispersals out of Africa than would be implied by considering only the extant placental orders. A model of pure dispersal, that is, one hypothesizing no vicariance events, invoking Africa as a “Garden of Eden” would differ from the earlier mixed model only by treating the occurrence of Xenarthra in South America as a dispersal event rather than a vicariance event. The total number of biogeographic events would remain the same.

Mixed vicariance and dispersal: South America as “Garden of Eden”

A mixed vicariance and dispersal model assuming that South America acted as a “Garden of Eden” for placental mammal orders is similar in several fundamental ways to the Africa as “Garden of Eden” hypothesis discussed earlier. Such a model also follows closely the biogeographic ideas proposed recently by (Murphy et al., 2001b), meriting some interest. We focus here on how a model of South American origins differs from a model of African origins.

First, the sole vicariance event caused by the separation of Africa from South America would have been the isolation of the basal lineage that gave rise to the Afrotheria in Africa, but there remains the need to invoke dispersal to account for presence of afrotherians (embrithopods and anthracobunids) in Asia (Table 4; Figs. 7 and 8). Second, no dispersal event is needed in this model of course to account for the presence of Xenarthra in South America. Nevertheless, the problem remains that no xenarthran is known on that continent until the Late Paleocene, despite a moderately good Early Paleocene record, including the presence of “meridiungulates”. Third, accounting for the presence in South America of “meridiungulates” by the Early Paleocene, as well as the presence of “amblypods” in both South America (pantodonts) and the northern continents by the same time, can be accomplished in a couple of different ways. Either these groups dispersed back into South America having arisen on the northern continents, which would add at least two, but more likely three dispersal events to the minimum number (Table 4, extant groups only; Fig. 7), or they were left behind when the remaining boreoeutherian lineages dispersed out of South America.
Table 4

Vicariance and Dispersal Events Required Assuming South America as a “Garden of Eden” for Placental Mammals, Following Initial Vicariance Model of Murphy et al.(2001b)

 

Extant groups only

Full record

Vicariance events

In Africa

Afrotheria

Afrotheria

Dispersal events

To Laurasia from Africa

 

Embrithopoda

To Laurasia from South America

  
 

Euarchonta

Euarchonta

 

Glires

Glires

 

Eulipotyphla

Eulipotyphla

  

“Proteutheria” (possibly)

 

Chiroptera

Chiroptera

 

Ferungulataa

Carnivora

  

Creodonta

  

Pholidota

  

Perissodactyla

 

Cetartiodactyla

Cetartiodactyla

  

“Amblypoda”

Dispersal events

6

11 (12 including “Proteutheria”)

Total events

7

12 (13 including “Proteutheria”)

aAssuming either a northern origin of “Meridiungulata”, which would require their dispersal back into South America by the Early Paleocene (not listed above), or a southern origin of “Meridiungulata” as the sister-taxon to Ferungulata.

https://static-content.springer.com/image/art%3A10.1007%2Fs10914-006-9006-6/MediaObjects/10914_2006_9006_Fig7_HTML.gif
Fig. 7

World map illustrating dispersal (single-headed arrows) and vicariance (double-headed arrow) events implied assuming South America as a “Garden of Eden”. See text for explanation. Extant placental lineages only illustrated

https://static-content.springer.com/image/art%3A10.1007%2Fs10914-006-9006-6/MediaObjects/10914_2006_9006_Fig8_HTML.gif
Fig. 8

World map illustrating dispersal (single-headed arrows) and vicariance (double-headed arrow) events implied assuming South America as a “Garden of Eden.” See text for explanation. Both extant and extinct placental lineages illustrated

In sum, the hypothesis of a South American “Garden of Eden” requires between 6 and 11 (12 including “Proteutheria”) dispersal events, or between 7 and 12 (13 including “Proteutheria”) total biogeographic events. As with the African-origins hypothesis, a pure dispersal South American-origins model would differ from this mixed model only in accounting for the presence of the Afrotheria in Africa as a dispersal event rather than a vicariance event.

Mixed vicariance and dispersal: Australia as “Garden of Eden”

The hypothesis of Australia as a “Garden of Eden” for placental mammals is one that has been argued recently on paleontological grounds (Rich et al., 2001b). Based on the occurrence of Ausktribosphenos and a related form, Bishops (see Rich et al., 2001a), mammals with teeth reminiscent of the tribosphenic teeth of advanced therian mammals, in the Early Cretaceous of Australia, and the resemblance of these teeth to those of erinaceid insectivores (hedgehogs and relatives), Rich et al. have argued that Australia in particular acted as a “Garden of Eden” for a hidden radiation of placental mammals deep in the Cretaceous. Because of the purported phylogenetic affinities of the fossil mammal concerned, we affectionately refer to the Australian origins hypothesis as the “Spiny Norman hypothesis”. (“Spiny Norman” was a fictional gigantic hedgehog, the size of a dirigible, which the mobster Dinsdale Piranha imagined to pursue him in a sketch from the British comedy television show Monty Python’s Flying Circus.)

The Spiny Norman hypothesis is plagued by problems quite apart from biogeographic concerns. First, the occurrence of a hedgehog in the Early Cretaceous would extend the temporal range of the Eulipotyphla back at least 20 million years before the current molecular estimate for the first split within Boreoeutheria, that between Euarchontoglires and Laurasiatheria, at ∼80 Ma (Murphy et al., 2001b). For this reason alone, Ausktribosphenos is probably too early to be directly relevant to the origins of any modern order of boreoeutherian placental mammal. Second, a recent phylogeny of Early Mesozoic mammals determined Ausktribosphenos to be part of a southern hemisphere clade of mammals, including monotremes, which evolved tribosphenic-like teeth convergently with true tribotheres (e.g., marsupials and placentals) of the northern hemisphere (Luo et al., 2001, 2002). According to this phylogenetic scheme, not only would the resemblance of Ausktribosphenos to northern hemisphere tribotheres be convergent in origin, but also so would be the special resemblance of this fossil mammal to erinaceids. Nevertheless, these studies by Luo et al. did not treat tooth morphology in great detail, leaving open the possibility that the tribosphenic appearance of Ausktribosphenos molars may be even more superficial than previously appreciated. Assuming a pretribosphenic ancestry (Archer, 1978; Archibald, 1999a,b, 2003; Pascual and Goin, 1999), an alternative interpretation of the molar morphology of Ausktribosphenos is possible that may better account for the molar cusps, crests, and occlusal wear of Ausktribosphenos than that originally proposed (Hunter, 2004, and unpublished). Counter arguments have been proposed to the particular phylogenetic scheme of Luo and colleagues (Sigogneau-Russell et al., 2001; Rich et al., 2002; Woodburne et al., 2003). The affinities of Ausktribosphenos continue to be debated, but within the context of a growing consensus that ausktribosphenids are not placentals (Sigogneau-Russell et al., 2001; Ji et al., 2002; Rauhut et al., 2002).

Nevertheless, giving Spiny Norman the benefit of the doubt, and treating this hypothesis as a mixed vicariance and dispersal model, one can derive the number of biogeographic events required by Spiny Norman in an Australian “Garden of Eden” (Table 5; Figs. 9 and 10). The initial vicariance events in this model would isolate Xenarthra in South America, Afrotheria in Africa, and Boreoeutheria in Australia approximately correlated to the breakup of Gondwana. The early isolation of Africa from the rest of Gondwana would make the occurrence of the Afrotheria in Africa just as reasonably explained by dispersal as vicariance. However, the hypothesis that South America was isolated from Australia, providing an opportunity for isolation and vicariance to occur, is problematic because the two land masses were likely connected via Antarctica through much of the Mesozoic and Early Cenozoic. This land connection almost certainly allowed faunal exchange between the southern landmasses, apart from Africa, during the Late Cretaceous and Early Cenozoic, including the dispersal of gondwanatherian mammals (Krause et al., 1997) and marsupials.
Table 5

Vicariance and Dispersal Events Required Assuming Australia as “Garden of Eden” for Placental Mammals, Incorporating Biogeographic Ideas of Rich et al.(2001b)

 

Extant groups only

Full record

Vicariance events

In Africa

Afrotheria

Afrotheria

In South America

Xenarthra

Xenarthra

Dispersal events

To South America

 

“Meridiungulata”

  

“Amblypoda” (pantodonts)

To Laurasia

Euarchonta

Euarchonta

 

Glires

Glires

 

“Insectivora”

Eulipotyphla

  

“Proteutheria” (possibly)

 

Chiroptera

Chiroptera

 

Ferungulataa

Carnivora

  

Creodonta

  

Pholidota

  

Perissodactyla

 

Cetartiodactyla

Cetartiodactyla

  

“Amblypoda”

  

Embrithopoda

Dispersal events

6

14

Total events

8

16

aAssuming either a northern origin of “Meridiungulata”, or an Australian origin as the sister-taxon of Ferungulata with dispersal to South America, most likely via Antarctica.

https://static-content.springer.com/image/art%3A10.1007%2Fs10914-006-9006-6/MediaObjects/10914_2006_9006_Fig9_HTML.gif
Fig. 9

World map illustrating dispersal (single-headed arrows) and vicariance (double-headed arrows) events implied assuming Australia as a “Garden of Eden.” See text for explanation. Extant placental lineages only illustrated

https://static-content.springer.com/image/art%3A10.1007%2Fs10914-006-9006-6/MediaObjects/10914_2006_9006_Fig10_HTML.gif
Fig. 10

World map illustrating dispersal (single-headed arrows) and vicariance (double-headed arrow) events implied assuming Australia as a “Garden of Eden”. See text for explanation. Both extant and extinct placental lineages illustrated

Although this connection of South America and Australia via Antarctica presents problems for considering vicariance as a factor in the split between Xenarthra and Boreoeutheria, it would seem nevertheless to provide a dispersal route for boreoeutherians (e.g., “meridiungulates” and pantodont “amblypods”) back into South America from Australia. If Eulipotyphla and the South American boreoeutherian lineages originated in Australia, other boreoeutherian groups would have to originate there as well, requiring their subsequent dispersal out of Australia into Laurasia (Table 5). Of course, these boreoeutherian taxa would have had to disperse and go extinct locally in Australia by the end of the Paleogene because none of these groups, with the exception of bats (Chiroptera, which probably represent Early Cenozoic volant dispersal from Asia), are known from the reasonably well-sampled fossil record of Australia during the Neogene–Recent.

In sum, the hypothesis of Spiny Norman in an Australian “Garden of Eden” requires between 6 and 13 (14 including “Proteutheria”) dispersal events, or between 8 and 15 (16 including “Proteutheria”) total dispersal and vicariance events. In the case of Australia, a model invoking dispersal may have more merit than a mixed model because it is unlikely that South America was isolated completely from Australia during the Cretaceous. Nevertheless, the pure dispersal model requires two more dispersal events than the mixed model, resulting in the same total number of biogeographic events.

Pure dispersal: gondwanan origins

As discussed earlier, models invoking pure dispersal from a southern continent “Garden of Eden” require more dispersal events than the corresponding mixed model, for the same total number of biogeographic events. The same would also apply to hypotheses for the origin of Placentalia on other landmasses that were part of Gondwana or were in the southern hemisphere during the Cretaceous, such as India or Madgascar. Furthermore, evidence for persistent connections among all of the Gondwanan landmasses exclusive of Africa places limits on the role that vicariance could have played in causing isolation between South American mammals and mammals isolated on these eastern Gondwanan landmasses during the Cretaceous and Paleocene.

Pure dispersal: laurasian origins

The biogeographic hypothesis invoking a northern continent origin and radiation of Placentalia is considerably simpler than the southern origins models discussed earlier. Assuming northern origins and the molecular phylogeny of Murphy et al.(2001b), only between two and four dispersal events are required to account for the distribution of placental mammals at the base of the Cenozoic (Table 6; Figs. 11 and 12). Xenarthrans must have dispersed into South America by the Late Paleocene, as must have boreoeutherian mammals to account for the occurrence there of the “meridiungulates” and the pantodont “amblypods”. Finally, afrotherians must have migrated into Africa leaving behind embrithopods and anthracobunids in Eurasia.
Table 6

Dispersal Events Required with Laurasian Origins of Placental Orders

 

Extant groups only

Full record

To South America

Xenarthra

Xenarthra

  

“Meridiungulata”a

  

“Amblypoda” (pantodonts)

To Africa

Afrotheria

Afrotheriaa

Total

2

4

aSee text for alternative scenario.

https://static-content.springer.com/image/art%3A10.1007%2Fs10914-006-9006-6/MediaObjects/10914_2006_9006_Fig11_HTML.gif
Fig. 11

World map illustrating dispersal (single-headed arrows) and vicariance (double-headed arrow) events implied assuming Laurasian origins. See text for explanation. Extant placental lineages only illustrated

https://static-content.springer.com/image/art%3A10.1007%2Fs10914-006-9006-6/MediaObjects/10914_2006_9006_Fig12_HTML.gif
Fig. 12

World map illustrating dispersal (single-headed arrows) and vicariance (double-headed arrow) events implied assuming Laurasian origins. See text for explanation. Both extant and extinct placental lineages illustrated

Alternatively, one might instead hypothesize two dispersals: Afrotheria into Africa, followed by embrithopods and anthracobunids back into Eurasia. Even counting two rather than one “meridiungulate” dispersals into South America, and the alternative two-dispersal history of Afrotheria, the maximium number of dispersals is only six, as compared to between 11 and 13 (excluding “Proteutheria”) dispersal events under the different southern continent mixed models. While it is not possible to demonstrate definitively on these grounds that this northern origins model is correct, it is nevertheless more biogeographically parsimonious than southern origins models (Table 7; Figs. 11 and 12) and more consistent with current molecular phylogenies. For these reasons, we consider a scenario of Laurasian origins to be well supported.
Table 7

Dispersal Events Required Using Morphology-Based Phylogeny (Novacek, 1992)

African origins

 Extant groups: Xenarthra (South America); Pholidota, Glires, Lipotyphla, Archonta, Carnivora, Cetartiodactyla, and Perissodactyla (Laurasia)

 Extinct groups: “Meridiungulata” and Pantodonta (South America); Embrithopoda and Creodonta (Laurasia)

South American origins

 Extant groups: Pholidota, Glires, Lipotyphla, Archonta, Carnivora, Cetartiodactyla, and Perissodactyla (Laurasia); Macroscelidea, Tubulidentata, and Paenungulata (Africa)

 Extinct groups: “Amblypoda”, Creodonta, Embrithopoda (Laurasia)

Australian origins

 Extant groups: Xenarthra (South America); Pholidota, Glires, Lipotyphla, Archonta, Carnivora, Cetartiodactyla, and Perissodactyla (Laurasia); Macroscelidea, Tubulidentata, and Paenungulata (Africa)

 Extinct groups: “Meridiungulata” and Pantodonta (South America); Embrithopoda and Creodonta (Laurasia)

Laurasian origins (short fuse)

 Extant groups: Xenarthra (South America); Macroscelidea, Afrosoricida, Tubulidentata, and Paenungulata (Africa); see text for alternative scenario

 Extinct groups: “Meridiungulata” and Pantodonta (South America); Embrithopoda (Eurasia)

Furthermore, this northern origins model is most plausible if it also represents a kind of “explosive” model of placental origins (Archibald and Deutschman, 2001), though not one that would require interordinal splits to be restricted to the Cenozoic. According to such a model, all of the major interordinal splits among placental mammals would have occurred in Eurasia or North America during the Latest Cretaceous and extending into the Paleocene. The Laurasian fossil record is good enough for it to be unlikely, on probabilistic grounds, that a prolonged radiation (>5–10 million years) of many lineages (>7) could escape the notice of paleontologists (Foote et al., 1999). Therefore, we propose that the most likely model explaining crown placental biogeography is that of a northern continent origin and diversification involving few lineages until the last few millions years of the Cretaceous and Paleocene.

Biogeographic models assuming a morphological phylogeny

Although recent workers have used molecular phylogenies to infer a southern continent origin and radiation of modern placentals, ironically these phylogenies on closer inspection support a northern continent origin if they support any particular biogeographic hypothesis at all. Nevertheless, these molecular phylogenies are controversial, and not necessarily accepted by all morphological workers. Therefore, biogeographic models assuming a traditional morphological phylogeny (Novacek, 1992) should be considered as well in order to determine whether the main inference made here, that a northern continent origin is more biogeographically parsimonious, is robust to different phylogenetic scenarios.

With the inclusion of only the extant orders of placental mammals, either with or without an appropriate outgroup, biogeographic reconstructions based on the morphological tree of Novacek support an origin and radiation of Placentalia somewhere on the northern continents (Fig. 13). It is also important to consider the number of dispersal events implied by this branching topology (Table 7). A minimum of five dispersal events is required by phylogenetic concerns alone among the extant orders: (1) a dispersal of Xenarthra into South America; (2) dispersal of Macroscelidea into Africa; (3) dispersal of afrosoricids (conservatively only one, but conceivably separate dispersals of tenrecids and chyrsochlorids); and either (4a and 5a) separate dispersals of Tubulidentata and a lineage that gave rise to the Paenungulata (Hyracoidea, Proboscidea, and Sirenia) into Africa or (4b and 5b) dispersal of a lineage that gave rise to Tubulidentata, Perissodactyla, Hyracoidea, Proboscidea, and Sirenia into Africa followed by dispersal of Perissodactyla (and Tubulidentata) back into Laurasia. Adding the extinct groups increases this minimum number by three: (6 and 7) dispersal of “meridiungulates” and pantodonts into South America and (8) dispersal of embrithopods into Eurasia.
https://static-content.springer.com/image/art%3A10.1007%2Fs10914-006-9006-6/MediaObjects/10914_2006_9006_Fig13_HTML.gif
Fig. 13

Phylogenetically derived biogeographic reconstruction of Placentalia. Topology from Novacek (1992) without an outgroup

Assuming an origin and radiation of Placentalia on the southern continents increases this minimal number over that implied by the topology alone. For the African-origins model, a consideration of extant groups requires eight dispersal events (Table 7): (1) dispersal of Xenarthra into South America; (2) dispersal of Pholidota into Laurasia; (3) dispersal of Glires into Laurasia leaving Macroscelidea behind as a relictual African order; (4) dispersal of Lipotyphla into Laurasia leaving afrosoricids behind as African relicts (ignoring for simplicity the “Proteutheria”); (5 and 6) dispersal into Laurasia of the Archonta (Primates, Dermoptera, Scandentia, and Chiroptera) and Carnivora; and finally (7 and 8) dispersal of Cetartiodactyla and Perissodactyla into Laurasia leaving Tubulidentata and the remaining “subungulates” behind as relictual African orders. Adding extinct groups increases this number of required dispersal events by four: (9 and 10) dispersal of “meridiungulates” and pantodonts into South America, (11) dispersal of embrithopods into Eurasia, and (12) dispersal of creodonts into Laurasia. Other models of southern continent origins (South America and Australia) yield similar numbers of required dispersal events as the African-origins model (Table 7).

The biogeographic reconstruction of Placentalia based solely on the topology of the morphological phylogeny results in strong support for a Laurasian origin and radiation of the group (Fig. 13). The Laurasian-origin model under the morphological phylogeny is also more biogeographically parsimonious than any southern continent model, whether based on the molecular or morphological phylogeny (Table 8). Nevertheless, the minimum estimates of five (extant groups only) and eight (extant and extinct groups) dispersal events required by the Laurasian-origins model under a morphological phylogeny is not quite as biogeographically parsimonious as the Laurasian-origins model under a molecular phylogeny, which requires only between two and four dispersal events. However, this result should not be taken to indicate that the Laurasian-origins model under a molecular phylogeny is the “correct” one. Rather, the chief inference of a northern continent origin and radiation of Placentalia is robust to choice of a preferred phylogeny for the group.
Table 8

Summary of Number of Dispersal Events Required Under Competing Modelsa

 

Number of Dispersals (Mixed models)

 

Molecular

Morphological

Hypothetical “Garden of Eden”

  

 Africa

13

12

  Full record

  

  Extant groups only

  
 

6

8

 South America

11

13

  Full record

  

  Extant groups only

6

10

 Australia

13

15

  Full record

  

  Extant groups only

6

11

Short Fuse Model

  

 Laurasia

4

8

  Full record

  

  Extant groups only

2

5

aNot including “Proteutheria.”

Discussion

Several lines of evidence discussed earlier point to an origin of the crown group of eutherian mammals somewhere among the northern continents. Phylogenetic reconstruction of the geographic origin of crown placentals yields a likely northern continent origin, whether the phylogeny chosen is based on molecules or morphology (Figs. 3 and 13). This result is probably due to the northern distribution of all appropriate outgroups to crown placentals as well as most modern placental clades at their appearance in the fossil record (compare with the hypothetical examples in Fig. 4). Likewise, when one considers the implications of different biogeographic models of placental mammal origins, models invoking northern continent origins require fewer dispersal events than models invoking southern continent origins (Table 8). Intercontinental dispersal by mammals requires overcoming barriers to dispersal, especially between isolated landmasses. The fact that intercontinental dispersal is not unknown among land mammals over the course of the Cenozoic is largely due to the enormity of geologic time (that is, if one waits long enough even the most improbable dispersal events are likely to happen at least once) and to the transitory formation of land bridges. Nevertheless, all else being equal, a biogeographic hypothesis that implies fewer improbable dispersal events is more likely than another requiring more.

It is probable that the breakup of first Pangea and later Gondwana did profoundly affect the distribution of mammals and the course of mammalian evolution, though at a higher taxonomic level than the Placentalia. The separation of Laurasia from Gondwana in the Early Jurassic probably established isolated vicariant lineages that gave rise in Laurasia to multituberculates and advanced therians (i.e., metatherians and eutherians), and in Gondwana to gondwanatheres and monotremes among other lineages (Luo et al., 2001, 2002). In addition, certain lineages of nontherian mammals (triconodonts and symmetrodonts) and pretribosphenic mammals (eupantotheres) are known to have persisted much later and diversified on the southern continents, at least in South America, than on the northern continents (Bonaparte, 1990). In Patagonia, these lineages persisted into the Late Cretaceous and were only replaced by metatherians and eutherians sometime between the Campanian and the Early Paleocene, ∼71–65 Ma, when these groups appeared in both Patagonia and northern South America, most likely dispersing from North America (Pascual, 1996, 1998; Pascual et al., 2000). Floral differences between northern South America and Patagonia in the Late Cretaceous suggest that northern South America could have harbored a different Late Cretaceous mammalian fauna from Patagonia as well (Clemens, 2001a). At the present time, however, this hypothesis cannot be tested because no mammalian fossils have been reported from the Late Cretaceous of northern South America.

The effects of the breakup of Gondwana on the course of mammalian evolution are more difficult to assess because the Mesozoic fossil record of mammals on the southern continents is patchy in both space and time. Nevertheless, it is likely that the separation of Africa from South America, and at the same time from Antarctica and Australia as well, primarily would have affected the nontherian mammals (e.g., triconodonts, symmetrodonts, and eupatotheres) known to have inhabited Gondwana during the Jurassic and Early Cretaceous (Pascual, 1996, 1998; Vizcaino et al., 1998; Pascual et al., 2000). That is, one set of Gondwanan, nontherian mammals would have been isolated in Africa apart from related, nontherian mammals in the rest of Gondwana. It is likely that South America maintained a connection through Antarctica to Australia during the Late Cretaceous and Early Cenozoic. Such a connection would explain the occurrence of gondwanathere mammals in South America, India, and Madagascar (Krause et al., 1997), the dispersal of marsupials relatively late in the Late Cretaceous from North America through South America and Antarctica (Marshall, 1980; Eaton, 1993) to Madagascar (Krause, 2001; Averianov et al., 2003, presents a different interpretation), and the presence of typically South American mammals on western Antarctica in the Early Cenozoic (e.g., Woodburne and Case, 1996).

Several therian mammals with tribosphenic molars, the kind of molar teeth shared by derived (i.e., crown) therian mammals, metatherians and eutherians, have been reported from Gondwana, raising the issue that the southern continents might have been a center of evolution for crown therian mammals. These reports include the Jurassic (Flynn et al., 1999) and Late Cretaceous (Krause, 2001) of Madagascar, Jurassic of South America (Rauhut et al., 2002), the Early Cretaceous of northern Africa (Sigogneau-Russell, 1991a), the Early Cretaceous of Australia (Rich et al., 1997, 1999; Rich et al., 2001a), and the Late Cretaceous of India (Prasad and Sahni, 1988; Prasad and Khajuria, 1990; Prasad and Godinot, 1994; Prasad et al., 1994, 1995; Rana and Wilson, 2003; Khosla et al., 2004). The tribosphenidan from the Early Cretaceous of northern Africa, like the rest of the mammalian fauna from this locality (Sigogneau-Russell, 1991b), shares affinities with Laurasian taxa, suggesting that northern Africa was in connection with Laurasia at that time. Similarly, the eutherian mammals from the Late Cretaceous of India have apparent Laurasian affinities (Prasad et al., 1995; Rana and Wilson, 2003). The report from the Early Cretaceous of Australia is that of Ausktribosphenos and a related taxon, which has already been discussed. Most likely, the tribosphenic-like molars of Ausktribosphenos evolved convergently on those of Laurasian tribosphenidans (Archer et al., 1999; Luo et al., 2001, 2002), and Ausktribosphenos probably descended from pretribosphenic mammals living in Gondwana. The report from the Late Cretaceous of Madagascar is that of a marsupial, which the authors interpret, correctly in our opinions, as simply an earlier-than-expected arrival from North America via South America and Antarctica (Krause, 2001). The reported tribosphenidan Ambondro from the Jurassic of Madagascar has been interpreted either as the earliest occurrence of tribosphenic therians, that is earlier in Gondwana than in Laurasia (Flynn et al., 1999; Sigogneau-Russell et al., 2001), or, like Ausktribosphenos, the product of convergent evolution on Laurasian tribosphenidans (Luo et al., 2001, 2002). Ambondro has been joined by Asfaltomylos, a form of comparable morphology and similar antiquity in South America (Rauhut et al., 2002). If Ambondro is a true tribosphenidan closely related to Laurasian forms, then the earlier occurrence of tribosphenidans in Gondwana than in Laurasia has suggested to some workers a possible origin of Tribosphenida in Gondwana (Flynn et al., 1999; Sigogneau-Russell et al., 2001). Nevertheless, even if the origin of Tribosphenida began in Gondwana, subsequent diversification of the tribosphenidans, including the origin of Metatheria and Eutheria, appears to have been in Laurasia. This inference gains further support from the recent discovery of basal (i.e., stem) eutherians and metatherians (though well outside of crown placentals or marsupials) in the Early Cretaceous of China (Ji et al., 2002; Luo et al., 2003).

With the exception of the view advanced by Woodburne et al.(2003), which has not undergone scrutiny, a consensus has emerged among paleontologists that advanced therians with tribosphenic-like teeth in the Mesozoic of Gondwana have no direct bearing on the origin of crown placental mammals or even stem eutherian mammals. As a result, in order for there to have been an origin and diversification of crown placental mammals on a southern continent, it would be first necessary to hypothesize an initial dispersal of eutherians into Gondwana from Laurasia, followed by a later migration back to Laurasia. For this reason, under a southern-continent origin hypothesis, Gondwana is probably better viewed as a potential “Noah’s ark” (McKenna, 1973) rather than as a “Garden of Eden” for placental mammals. That is, eutherian mammals would have had to first embark on and later disembark from Gondwana, rather than originate in situ in Gondwana. Such is the case no matter which Gondwanan land mass is hypothesized to have harbored crown placental mammals for a time during the Cretaceous. Thus, a challenge for any southern-continent origins hypothesis is to explain not only how placental mammals dispersed from a Gondwanan landmass to Laurasia, but also how the ancestors of those placental mammals would have dispersed into Gondwana in the first place.

Why then have some recent workers argued for a southern continent origin and diversification of crown placental mammals, contrary to the traditional northern-origin view? It is possible that the phylogenetic and biogeographic implications of a southern origin have not been fully appreciated. Basing biogeographic interpretations solely on the distribution of extant placental taxa incurs a taxon-sampling problem. Such inferences cannot be drawn with confidence without reference to outgroups (Figs. 3 and 4). As a result, the occurrence of basal clades endemic to different and isolated southern continents (Xenarthra in South America and Afrotheria in Africa), taken in isolation, is as consistent with migration from Laurasia as it is with a southern origin for these clades. However, taken in the context of a probable origin of both the stem and the crown group of eutherian mammals in Laurasia, separate migrations out of Laurasia seems more likely.

It is also possible that it may be simply too easy to believe that newly appearing taxa in a geographic region, without obvious relatives or immediate ancestors in the same region, must have migrated into the region from somewhere else. Although such an inference is usually reasonable to some extent, two caveats must be kept in mind. First, the closer and less isolated the hypothetical source area, the more likely it is that taxa could actually disperse from there. For example, the hypothesis that southwestern North America could have served as a source for certain immigrants (e.g., the multituberculate Stygimys) into the North American western interior in the Paleocene (Clemens, 2001a, 2001b) is entirely plausible. Second, because of an under-appreciated aspect of the Noah’s ark problem, it is necessary to explain how the ancestors of immigrants arrived at the hypothetical source landmass in the first place. In the case discussed in this paper, supposing that crown placental mammals dispersed from a southern continent requires an explanation of how these mammals originally came to be on that southern continent. For these reasons, pointing out the incompleteness of the fossil record of Mesozoic mammals in the southern hemisphere and asserting that placental mammals could have originated and diversified there is not by itself a compelling argument.

Although the exact place of origin and diversification of crown placental mammals remains unknown, it seems more likely that this place was somewhere on the northern continents if one considers the broader phylogenetic, fossil, and biogeographic evidence discussed here. By itself, the neontological evidence from living placental mammals lacks the taxonomic scope and temporal depth to address this biogeographic problem adequately. Of course, we acknowledge that our view may need to be revised as subsequent discoveries come to light. However, we are less concerned with establishing the precise place of origin of placental mammals than simply formulating the best interpretation of the currently available data. There remain problems concerning the early biogeography of placental mammals, including the source for immigrants into the North American Western Interior in the Early Paleocene (Clemens, 2001a, 2001b) and later near the Paleocene/Eocene boundary (Krause and Maas, 1990; Beard, 1998, 2002; Clyde et al., 2001), as well as the time that xenarthrans arrived in South America (Pascual, 1996, 1998). However, these problems are fundamentally paleontological, rather than neontological in nature. It is hoped that in the future, these biogeographic problems will be addressed with the proper paleontological tools for their resolution.

Acknowledgements

Responding to comments by Mark Springer and three anonymous reviewers aided the growth and development of this manuscript.

Copyright information

© Springer Science+Business Media, Inc. 2006