Mammalian anatomists and paleontologists working primarily with osteological data have long been intrigued with the problem of eutherian diversification, including both interordinal relationships and the timing of the placental mammal radiation (McKenna, 1975; Novacek, 1992). Cladistic analyses of morphological characters for placental mammal orders have suggested a variety of superordinal hypotheses (Table 1). Molecular trees based on single gene segments were often in conflict with each other and with morphology, but larger nuclear gene data sets that include longer and/or multiple gene segments have converged on a well-supported superordinal tree topology that divides placental orders into four major groups: Afrotheria, Xenarthra, Laurasiatheria, and Euarchontoglires (Madsen et al., 2001; Murphy et al., 2001a, 2001b; Scally et al., 2001; Amrine-Madsen et al., 2003). Analyses of independent molecular and genomic data sets, specifically whole mitochondrial genomes and rare genomic changes (RGCs), are congruent with this four-clade classification (Hudelot et al., 2003; Waddell and Shelley, 2003; Murphy et al., 2004; Reyes et al., 2004; Springer et al., 2004, 2005; Gibson et al., 2005; Kriegs et al., 2006; Nishihara et al., 2006; Kjer and Honeycutt, 2007). For example, Nishihara et al. (2006) found six, nine, and nine L1 retroposon insertions supporting the monophyly of Afrotheria, Euarchontoglires, and Laurasiatheria, respectively.

Table 1

Superordinal groups based on morphological data.

Superordinal group Taxa 
Altungulata Paenungulata, Perissodactyla 
Anagalida Glires, Macroscelidea 
Archonta Dermoptera, Chiroptera, Primates, Scandentia 
Edentata Xenarthra, Pholidota 
Glires Rodentia, Lagomorpha 
Lipotyphla Afrosoricida, Eulipotyphla 
Paenungulata Hyracoidea, Proboscidea, Sirenia 
Tethytheria Proboscidea, Sirenia 
Ungulata Altungulata, Cetartiodactyla, Tubulidentata 
Volitantia Dermoptera, Chiroptera 
Superordinal group Taxa 
Altungulata Paenungulata, Perissodactyla 
Anagalida Glires, Macroscelidea 
Archonta Dermoptera, Chiroptera, Primates, Scandentia 
Edentata Xenarthra, Pholidota 
Glires Rodentia, Lagomorpha 
Lipotyphla Afrosoricida, Eulipotyphla 
Paenungulata Hyracoidea, Proboscidea, Sirenia 
Tethytheria Proboscidea, Sirenia 
Ungulata Altungulata, Cetartiodactyla, Tubulidentata 
Volitantia Dermoptera, Chiroptera 

Whereas nuclear, mitochondrial, and RGC data corroborate each other in supporting the four major clades, only Xenarthra had been previously hypothesized based on morphology. In addition, the morphology-based clades Altungulata, Anagalida, Archonta, Edentata, Lipotyphla, Ungulata, and Volitantia (Table 1) are all incompatible with the four-clade classification of mammals derived from molecular and genomic data. The molecular/genomic tree also suggests that morphology-based groups of placental orders (Novacek, 2001) are often homoplastic constellations of taxa that have evolved independently in separate geographic venues rather than monophyletic assemblages that descended from a common ancestor.

Scotland et al. (2003) argued for a more limited role for morphology-based phylogenetic analyses in reconstructing the tree of life. Other authors (Jenner, 2004; Wiens, 2004; Smith and Turner, 2005) have offered detailed critiques of Scotland et al.'s (2003) main thesis and maintain the view that morphological data remain “crucial in reconstructing the phylogeny of the earth's biota” (Smith and Turner, 2005:171). In particular, Wiens (2004), Jenner (2004), and Smith and Turner (2005) all note the importance of morphological data for reconstructing relationships of fossil taxa. We also recognize the primacy of morphological data for reconstructing relationships of extinct taxa. However, the failure of morphological data alone to recover a tree compatible with the four-clade interordinal partitioning of placental mammals raises serious doubts about the ability of current morphological cladistic studies to accurately reconstruct relationships for extinct forms.

Correlated character evolution related to diet and/or locomotion in independent lineages is perhaps the most dangerous pitfall of morphological cladistics, where instead the implicit assumption is that morphological characters evolve independently of each other in separate lineages. This assumption may be warranted for some characters, but there are also studies of the genetics of skeletal variation in mammals that provide evidence for positively and negatively correlated traits in skeletal size and shape. Chase et al. (2002) examined variation in the canid skeleton and found that individuals with relatively small pelvic girdles and lumbar vertebrae also tend to have large attachment sites for jaw and neck muscles, relatively large posterior faces, and small anterior faces. More generally, this finding implies that the size and strength of the pelvic and head-neck musculoskeletal systems are inversely correlated (Chase et al., 2002). Chase et al. (2002) also found that metrics of skull and limb length are inversely correlated with metrics of skull width and height. Carrier et al. (2005) found a negative correlation between the size of the pelvis and dimensions of distal limb bones. This negative correlation represents a functional trade-off between high-speed, energy-efficient running versus limb strength. Negative correlations between pelvis size and metrics of limb robustness are also seen in the transition from Australopithecus (large pelvis and less robust limb bones) to Homo (smaller pelvis and more robust limb bones; Wolpoff, 1999).

Kangas et al. (2004) discovered evidence for correlated character evolution in their genetic study of ectodysplasin expression in mouse and found that increased expression of this single protein resulted in an increase in the number of teeth, an increase in the number of cusps on teeth, changes in the shapes and positions of cusps, and the formation of longitudinal crests. Numerous characteristics related to these variables are routinely scored in matrices of dental characters (Luo et al., 2001; Meng et al., 2003). Kangas et al. (2004:211) concluded that “most aspects of tooth shape have the developmental potential for correlated changes during evolution which may, if not taken into account, obscure phylogenetic history.” Kangas et al. (2004) did not deny the potential for dental characters to change independently, but their study (p. 214) demands that “developmental nonindependence should not be excluded from the hypotheses considered in evolutionary taxonomy.”

Perhaps the most spectacular example of correlated character evolution is Volitantia (Chiroptera + Dermoptera), which is supported by as many as 17 putative synapomorphies (Simmons and Geisler, 1998). However, molecular and genomic studies provide congruent support for the inclusion of chiropterans in the superordinal clade Laurasiatheria, whereas dermopterans are members of the superordinal clade Euarchontoglires (Waddell et al., 1999, 2001; Teeling et al., 2000; Murphy et al., 2001b; Amrine-Madsen et al., 2003; Reyes et al., 2004; Nishihara et al., 2006; Kjer and Honeycutt, 2007). It is now manifest that putative synapomorphies of Volitantia are homoplastic features shared in common by Chiroptera and Dermoptera and that numerous derived characters related to gliding (Dermoptera) and powered flight (Chiroptera) have evolved independently in a highly correlated fashion in these two taxa. Gunnell and Simmons (2005) provide a cogent ecological explanation: that morphological similarities shared by these taxa were derived independently due to the demands of an arboreal environment.

The emerging molecular and genomic consensus of four major groups of placental mammals has implications for early placental biogeography as Afrotheria and Xenarthra are of putative Gondwanan origin based on the fossil record of constituent orders whereas Laurasiatheria and Euarchontoglires are of putative Laurasian origin (Eizirik et al., 2001; Madsen et al., 2001). Molecular dating analyses with relaxed clock methods suggest that the four superordinal groups all diverged from each other in the Cretaceous (Hasegawa et al., 2003, Springer et al., 2003; van Rheede et al., 2006; Murphy et al., 2007). By contrast, Asher et al. (2003, 2005) and Meng et al. (2003) concluded that molecular studies that place interordinal divergences in the Cretaceous (Hasegawa et al., 2003; Springer et al., 2003) are contradicted by phylogenetic analyses that place Cretaceous eutherians outside of crown-group Placentalia. Asher et al. (2003), Zack et al. (2005), and Tabuce et al. (2007) have all suggested that Afrotheria is Holarctic in origin based on the placement of extinct northern hemisphere condylarths within and/or at the base of Afrotheria in phylogenetic analyses with combined molecular-morphological data (molecular data coded as missing for fossil taxa; Asher et al., 2003) or morphological data alone (Zack et al., 2005; Tabuce et al., 2007). For example, Zack et al. (2005, p. 498) stated “identification of macroscelidean relatives in the North American Palaeocene argues against an African origin for Afrotheria, weakening support for linking placental diversification to the break-up of Gondwana.” One difficulty with Zack et al.'s (2005) conclusion is that estimating the place of origin based on the oldest afrotherian fossils, which in this case are North American aspheliscines, does not follow logically. Other variables, most importantly the topology itself, are of fundamental importance in reconstructing the place of origin. If we accept the highly nested position that Zack et al. (2005) discovered for North American aspheliscines (see Zack et al., 2005: fig. 3c), it is clear that Afrotheria must be substantially older than the aspheliscine fossils and that the origin of Afrotheria cannot be inferred from the geographic distribution of aspheliscines. A second difficulty with these studies is the assumption that morphological data, and methods for analyzing these data, give reliable phylogenetic solutions for the placement of key fossil taxa, even though morphological cladistics has so far failed to recover three of the fundamental clades of living placentals (Afrotheria, Euarchontoglires, Laurasiatheria) that are supported by multiple lines of independent molecular and genomic evidence. These same studies that question the African origin of Afrotheria based on the placement of extinct condylarths also fail to recover the monophyly of extant afrotherians in analyses based on morphology alone, and only recover a clade similar to Afrotheria by concatenating morphological characters with molecular data (Asher et al., 2003) or reducing taxon sampling to a small number of placental orders (Zack et al., 2005; Tabuce et al., 2007).

The phylogenetic placement of most mammal fossils cannot be tested directly with molecular data owing to the degradation of DNA. Nevertheless, the emergence of four superordinal clades (Afrotheria, Xenarthra, Laurasiatheria, Euarchontoglires) that are supported by analyses of nuclear gene sequences, mitochondrial genome sequences, and genome-wide screens for rare genomic changes provides an opportunity to conduct tree congruence tests with morphological data. Here, we use a novel method that combines molecular and morphological data and treats living orders as pseudoextinct to assess whether morphological data alone are sufficient to provide congruent phylogenetic results in the absence of molecular data. Unlike Scotland et al. (2003), who criticized morphological cladistics largely on the basis of the presumed behavior of morphological data (see Scotland et al., 2002: fig. 1), our pseudoextinction approach provides a test case of the behavior of real morphological data in phylogenetic analyses. Additional analyses examine whether morphological data are more incongruent with molecular data than are equivalently sized molecular data partitions with each other. Our results call for improved methods for collecting and analyzing morphological data to resolve deep-level phylogenetic relationships among extinct and living mammalian taxa.

Figure 1

Phylogenetic tree with posterior probabilities derived from MrBayes 3.1.1 (Ronquist and Huelsenbeck, 2003) and a 14,326-bp data set (TreeBASE S1792, M3274) that included segments from 20 different nuclear genes (ADORA3, ADRA2B, ADRB2, APOB, APP, ATP7A, BDNF, BMI, BRCA1, CNR1, CREM, EDG1, GHR, IRBP, PLCB4, RAG1, RAG2, TYR, VWF, ZFX). Accession numbers for previously published molecular data are given in online Supplemental Material (http://www.systematicbiology.org). New molecular sequences were obtained as previously described (Madsen et al., 2001; Murphy et al., 2001a, 2001b; Amrine-Madsen et al., 2003). Accession numbers for new sequences are EF104981 to EF104995 (see online Supplemental Information at http://www.systematicbiology.org). Posterior probabilities are averages based on two independent runs. The average standard deviation of split frequencies for the two runs was 0.004. Asterisks denote 44 taxa that were included in the pseudoextinction analyses. Abbreviations as follows: M = Marsupialia; X = Xenarthra; L = Laurasiatheria; E = Euarchontoglires; A = Afrotheria; B = Boreoeutheria.

Figure 1

Phylogenetic tree with posterior probabilities derived from MrBayes 3.1.1 (Ronquist and Huelsenbeck, 2003) and a 14,326-bp data set (TreeBASE S1792, M3274) that included segments from 20 different nuclear genes (ADORA3, ADRA2B, ADRB2, APOB, APP, ATP7A, BDNF, BMI, BRCA1, CNR1, CREM, EDG1, GHR, IRBP, PLCB4, RAG1, RAG2, TYR, VWF, ZFX). Accession numbers for previously published molecular data are given in online Supplemental Material (http://www.systematicbiology.org). New molecular sequences were obtained as previously described (Madsen et al., 2001; Murphy et al., 2001a, 2001b; Amrine-Madsen et al., 2003). Accession numbers for new sequences are EF104981 to EF104995 (see online Supplemental Information at http://www.systematicbiology.org). Posterior probabilities are averages based on two independent runs. The average standard deviation of split frequencies for the two runs was 0.004. Asterisks denote 44 taxa that were included in the pseudoextinction analyses. Abbreviations as follows: M = Marsupialia; X = Xenarthra; L = Laurasiatheria; E = Euarchontoglires; A = Afrotheria; B = Boreoeutheria.

Expanded Molecular Sampling

We performed Bayesian analyses with an expanded molecular data set that builds on the nuclear data of Murphy et al. (2001b) and Amrine-Madsen et al. (2003) (Fig. 1) and includes four marsupial outgroups, 53 placental taxa, and 20 different nuclear gene segments. Sequences for each gene segment were aligned using SOAP (Löytynoja and Milinkovitch, 2001) with 25 different combinations of gap opening and gap extension penalties following Westerman et al. (2002). A total of 14,326 aligned sites were conserved across all 25 alignments and were retained for phylogenetic analyses. The concatenated DNA alignment is available in online Supplemental Material (http://www.systematicbiology.org). ModelTest (Posada and Crandall, 1998) analyses were performed with each gene segment to determine the most appropriate model of sequence evolution as indicated by the Akaike Information Criterion. The 57-taxon molecular data set was analyzed with MrBayes 3.1.1 (Ronquist and Huelsenbeck, 2003; Ronquist et al., 2005), which simultaneously performs two Metropolis-coupled MCMC runs and checks for convergence on the fly after discarding the first 25% of tree samples as burn-in. Analyses were run with three heated chains and one cold chain for 5,000,000 generations. Chains were sampled every 1000 generations. Chains started with random trees and the following priors: all trees equally probable; Dirichlet probability density for base frequencies (1,1,1,1); Dirichlet probability density for rate matrix (all substitution types set at 1.0); uniform distribution for the shape (α) of the gamma distribution (0.1, 50.0); uniform distribution for proportion of invariant sites (0.0, 1.0); exponential (10.0) distribution for branch lengths. Bayesian analyses provided strong support for the four-clade classification (Afrotheria, Xenarthra, Laurasiatheria, Euarchontoglires) as well as for an association of Laurasiatheria and Euarchontoglires in the superordinal group Boreoeutheria (Fig. 1).

Pseudoextinction Analyses

Forty-four of the 57 taxa included in our molecular data set are also represented in the morphological data set of Asher et al. (2003; data set available at http://people.pwf.cam.ac.uk/rja58/; also see online Supplemental Material at http://www.systematicbiology.org), which is the largest currently available morphological data set to include representatives of all extant orders of placental mammals. The 44 taxa represented in our molecular data set that overlap with Asher et al.'s (2003) morphological data set are denoted in Figure 1 with asterisks. Asher et al.'s (2003) data set includes 185 osteological characters and 11 soft-tissue characters. Molecular and morphological data from Asher et al. (2003) were concatenated into a mixed data set for these 44 taxa. Our combined data set included two marsupial outgroups and representatives of all 18 placental orders.

DNA molecules and soft-tissue morphological characters are not routinely preserved for fossil mammals. In phylogenetic analyses with the 44-taxon data set, taxa representing each placental order (one order at a time) were treated as if the order was extinct by coding both molecular and 11 soft-tissue morphological characters as missing. The remaining 185 morphological characters are osteological and were retained for phylogenetic analyses with the pseudoextinct order. We assessed whether each pseudoextinct order was recovered as monophyletic on the most parsimonious tree(s) and in Bayesian analyses. We also assessed whether each pseudoextinct order remained in the same superordinal group (Afrotheria, Xenarthra, Laurasiatheria, Euarchontoglires) as in Figure 1 or moved elsewhere on the tree. In cases where individual pseudoextinct orders moved elsewhere on the tree we performed additional pseudoextinction analyses using taxonomic subsets of these orders. The superordinal groups Afrotheria, Euarchontoglires, and Laurasiatheria were also treated as pseudoextinct. Bayesian analyses with the 44-taxon mixed data set were performed with MrBayes 3.1.1 (Ronquist and Huelsenbeck, 2003; Ronquist et al., 2005). Analyses were run for a minimum of three million generations with four chains. Longer runs with up to eight chains were performed if necessary to achieve convergence. We used a threshold of 0.05 for the average standard deviation of split frequencies as a diagnostic for convergence. A lower threshold value (i.e., 0.01) is preferable, but achieving this level of convergence proved impractical for some of our analyses with pseudoextinct taxa. We note, however, that more than 75% of our analyses exceeded the desired 0.01 threshold. Burnin was set to include the first 25% of tree samples. We used the Lewis (2001) model for morphological characters. Settings for molecular partitions were as described in Figure 1 legend. Parsimony analyses with the mixed data set were performed with PAUP 4.0b10 (Swofford, 2003).

Analyses with the 44-taxon data set that treated entire molecular or morphological superordinal groups as pseudoextinct and examined the resulting deployment of taxa over the remaining tree are summarized in Table 2. When all afrotherian orders were pseudoextinct, they dispersed to Euarchontoglires (Macroscelidea), Laurasiatheria (Afrosoricida, Proboscidea, Sirenia, Hyracoidea), and Xenarthra (Tubulidentata). With the exception of pangolins, which joined Xenarthra, all of the pseudoextinct laurasiatherian orders moved to Afrotheria. Among the five orders in Euarchontoglires, rodents and lagomorphs joined elephant shrews in Afrotheria, whereas tree shrews, primates, and flying lemurs joined bats to reconstitute the morphological clade Archonta (parsimony) or moved to the base of the tree where they were a monophyletic sister-group to all other placentals (Bayesian).

Table 2

Results of parsimony and Bayesian analyses when extant placental superordinal groups are treated as pseudoextinct. Parsimony results are based on the most parsimonious tree(s) for each pseudoextinction analysis (1000 randomized taxon input orders; TBR branch-swapping); parsimony bootstrap percentages are based on 500 pseudoreplications with 10 randomized taxon input orders per pseudoreplicate. Posterior probabilities for Bayesian analyses are average values based on two independent runs. Bayesian analyses were run until the average standard deviation of split frequencies was ≤ 0.05. Molecular and soft morphological characters were both scored as missing for pseudoextinct taxa.

 Parsimony results Bayesian results 
 
 

 
Superordinal group Monophyletic? Location on shortest tree(s) when pseudoextinct + bootstrap percentages Monophyletic? Location on tree when pseudoextinct 
Afrotheria No Tubulidentata joins Xenarthra (71%); paenungulates and Afrosoricida join Laurasiatheria (35%) where paenungulates are sister to perissodactyls (25%), tenrecid is sister to soricid (29%), and chrysochlorid is sister to talpid (64%); macroscelideans move to Euarchontoglires and are sister to Gliresa (31%) No Tubulidentata moves to Xenarthra (0.67) and is the sister to Dasypodidae (0.53); all other afrotherians join Laurasiatheria (0.80) where macroscelideans are sister to Erinaceidae (0.87), Afrosoricida joins Talpidae (0.73), and paenungulates form a clade with perissodactyls (0.88) 
Euarchontoglires No Glires moves to Afrotheria (16%) and is sister to Macroscelidea (2 trees; 24%) or Afrosoricida + Macroscelidea (1 tree; 12%); Dermoptera, Scandentia, and Primates join Laurasiatheria (14%) with Dermoptera sister to Chiroptera (43%) and reconstitution of Archonta (7%) No Glires moves to Afrotheria (0.83) and joins Macroscelidea (0.82); Euarchonta remains monophyletic (0.84) and moves to the base of Placentalia where there is a split between Euarchonta and all other placentals (0.60) 
Laurasiatheria No Pholidota joins Xenarthra (48%) as sister to Myrmecophagidae (74%); all other taxa move to Afrotheria (10%) where perissodactyls + artiodactyls join paenungulates (75%); Erinaceidae sister to Macroscelidea (63%); talpid joins chrysochlorid (59%); soricid sister to talpid + chrysochlorid + tenrecid (40%); chiropterans + cetaceans + carnivores join eulipotyphlans + afrosoricidans + macroscelideans (5%)b No Pholidota joins Xenarthra (0.95) as sister to Myrmecophagidae (0.97); Chiroptera joins Euarchontoglires (0.75) as sister to Dermoptera (0.65); eulipotyphlans, artiodactyls, and perissodactyls join Afrotheria where Erinaceidae is sister to Macroscelidea (0.93), soricid joins tenrecid (0.50), talpid joins chrysochlorid + soricid + tenrecid (0.93), and artiodactyls + perissodactyls form a monophyletic clade (0.95) that joins Hyracoidea (0.99) inside of a paraphyletic Paenungulata; carnivores and cetaceans are excluded from a clade that contains all other placental mammals (0.78) 
 Parsimony results Bayesian results 
 
 

 
Superordinal group Monophyletic? Location on shortest tree(s) when pseudoextinct + bootstrap percentages Monophyletic? Location on tree when pseudoextinct 
Afrotheria No Tubulidentata joins Xenarthra (71%); paenungulates and Afrosoricida join Laurasiatheria (35%) where paenungulates are sister to perissodactyls (25%), tenrecid is sister to soricid (29%), and chrysochlorid is sister to talpid (64%); macroscelideans move to Euarchontoglires and are sister to Gliresa (31%) No Tubulidentata moves to Xenarthra (0.67) and is the sister to Dasypodidae (0.53); all other afrotherians join Laurasiatheria (0.80) where macroscelideans are sister to Erinaceidae (0.87), Afrosoricida joins Talpidae (0.73), and paenungulates form a clade with perissodactyls (0.88) 
Euarchontoglires No Glires moves to Afrotheria (16%) and is sister to Macroscelidea (2 trees; 24%) or Afrosoricida + Macroscelidea (1 tree; 12%); Dermoptera, Scandentia, and Primates join Laurasiatheria (14%) with Dermoptera sister to Chiroptera (43%) and reconstitution of Archonta (7%) No Glires moves to Afrotheria (0.83) and joins Macroscelidea (0.82); Euarchonta remains monophyletic (0.84) and moves to the base of Placentalia where there is a split between Euarchonta and all other placentals (0.60) 
Laurasiatheria No Pholidota joins Xenarthra (48%) as sister to Myrmecophagidae (74%); all other taxa move to Afrotheria (10%) where perissodactyls + artiodactyls join paenungulates (75%); Erinaceidae sister to Macroscelidea (63%); talpid joins chrysochlorid (59%); soricid sister to talpid + chrysochlorid + tenrecid (40%); chiropterans + cetaceans + carnivores join eulipotyphlans + afrosoricidans + macroscelideans (5%)b No Pholidota joins Xenarthra (0.95) as sister to Myrmecophagidae (0.97); Chiroptera joins Euarchontoglires (0.75) as sister to Dermoptera (0.65); eulipotyphlans, artiodactyls, and perissodactyls join Afrotheria where Erinaceidae is sister to Macroscelidea (0.93), soricid joins tenrecid (0.50), talpid joins chrysochlorid + soricid + tenrecid (0.93), and artiodactyls + perissodactyls form a monophyletic clade (0.95) that joins Hyracoidea (0.99) inside of a paraphyletic Paenungulata; carnivores and cetaceans are excluded from a clade that contains all other placental mammals (0.78) 
a

On bootstrap trees the most common position for macroscelideans is sister to lipotyphlans (45%).

b

On bootstrap trees the most common position for chiropterans is sister to Dermoptera (42%).

The results of analyses that treated individual orders as pseudoextinct are summarized in Table 3. In both parsimony and Bayesian analyses, 4 of 18 orders (Afrosoricida, Eulipotyphla, Cetartiodactyla, Xenarthra) were not recovered as monophyletic when they were coded as pseudoextinct. Rodents were paraphyletic or monophyletic in parsimony analyses and monophyletic in Bayesian analyses (Table 3). However, only 5 of 18 orders (rodents, lagomorphs, scandentians, sirenians, proboscideans) consistently remained in the molecularly defined superordinal group, whereas the other 13 orders were deployed to one or more of the other superordinal groups on some or all of the most parsimonious trees and/or in the Bayesian analyses (Table 3). Among afrotherian orders, Afrosoricida, Hyracoidea, and Macroscelidea moved to Laurasiatheria and aardvark nested within Xenarthra as the sister-taxon to the armadillo. Among the orders that moved to Laurasiatheria, afrosoricidans joined eulipotyphlan insectivores (shrews, moles, hedgehogs), macroscelideans joined Erinaceidae (hedgehogs), and hyracoids became the sister taxon to Perissodactyla. When laurasiatherian orders were treated as pseudoextinct, pholidotans joined Xenarthra; carnivores, perissodactyls, and eulipotyphlans moved to Afrotheria; chiropterans either stayed in Laurasiatheria or moved to Euarchontoglires; and cetartiodactyls split into a cetacean group that either joined Xenarthra (parsimony) or remained in Laurasiatheria (Bayesian), and an artiodactyl assemblage (paraphyletic in parsimony analyses and monophyletic in Bayesian analyses) that remained within Laurasiatheria. When xenarthrans were pseudoextinct, the myrmecophagid (anteater) moved to Laurasiatheria (sister-taxon to Pholidota), whereas the dasypodid (armadillo) moved to Afrotheria (sister-taxon to Tubulidentata). Placental orders in Euarchontoglires were the most stable when treated as pseudoextinct and the only movements to other superordinal groups were Dermoptera joining Pholidota in Laurasiatheria and primates moving to Afrotheria (Table 3).

Table 3

Results of parsimony and Bayesian analyses when placental orders are pseudoextinct. Parsimony results are based on the most parsimonious tree(s) for each pseudoextinction analysis (1000 randomized taxon input orders; TBR branch-swapping); parsimony bootstrap percentages are based on 500 pseudoreplications with 10 randomized taxon input orders per pseudoreplicate. Posterior probabilities for Bayesian analyses are average values based on two independent runs. Bayesian analyses were run until the average standard deviation of split frequencies was ≤ 0.05. Molecular and soft morphological characters were both scored as missing for pseudoextinct taxa.

 Parsimony results Bayesian results 
 
 

 
Order Monophyletic? Location on shortest tree(s) when pseudoextinct + bootstrap percentages Monophyletic? Location on tree when pseudoextinct 
Afrosoricida No Moves to Laurasiatheria (96%) and joins Eulipotyphla (92%); chrysochlorid sister to talpid (69%); tenrec sister to soricid (1 tree; 28%) or talpid + chrysochlorid (1 tree; 32%) No Moves to Laurasiatheria (1.00); chrysochlorid joins talpid (0.95) and tenrec joins chrysochlorid + talpid (0.80) 
Carnivora Yes (86%) Moves to Afrotheria (42%) where carnivores are sister to paenungulates (31%) Yes (1.00) Moves to base of Placentalia [basal split between carnivores (1.00) and all other placentals (0.99)] 
Cetartiodactyla No Cetaceans move to Xenarthra (16%); artiodactyls remain in Laurasiatheria (39%) and form clade with perissodactyls (72%) No Remains in Laurasiatheria (0.67); artiodactyls are monophyletic (0.96) and join perissodactyls (0.93); cetaceans are a monophyletic (1.00) sister to chiropterans + artiodactyls + perissodactyls + pholidotans + carnivores (0.52 for cetaceans + sister taxon) 
Chiroptera Yes (100%) Moves to Euarchontoglires (2 trees; 39%) where chiropterans join Dermoptera (33%) or remain in Laurasiatheria (1 tree; 37%) where they are sister to Erinaceidae (17%) Yes (1.00) Chiropterans join Euarchontoglires (0.47) as a monophyletic sister taxon to this superordinal group 
Dermoptera NA Moves to Laurasiatheria (51%) where Dermoptera is sister to Pholidotaa (16%) NA Remains in Euarchontoglires (0.83) and is the sister to Scandentia (0.82) 
Eulipotyphla No Moves to Afrotheria (78%) where Erinaceidae joins Macroscelidea (68%), talpid joins chrysochlorid (61%), and soricid is sister to talpid + chrysochlorid + tenrecid (47%) No Moves to Afrotheria (1.00) with Erinaceidae as sister to Macroscelidea (1.00), talpid as sister to chrysochlorid (0.63), and soricid as sister to tenrecid (0.78) 
Hyracoidea NA Moves to Laurasiatheria (95%) and is the sister to Perissodactyla (62%) NA Moves to Laurasiatheria (1.00) and is the sister to Perissodactyla (1.00) 
Lagomorpha Yes (92%) Remains in Euarchontoglires (68%) and Glires (73%) Yes (1.00) Remains in Euarchontoglires (1.00) but joins murid rodents (0.61) 
Macroscelidea Yes (100%) Moves to Laurasiatheria (60%) and is sister to Erinaceidae (51%) Yes (1.00) Moves to Laurasiatheria (0.99) and is sister to Erinaceidae (0.88) 
Perissodactyla Yes (60%) Moves to Afrotheria (50%) where perissodactyls are sister to Proboscidea + Sirenia (22%) Yes (1.00) Moves to Afrotheria (0.99) as sister to Hyracoidea (0.99) 
Pholidota NA Moves to Xenarthra (79%) and is sister to Myrmecophagidae (78%) NA Moves to Xenarthra (1.00) and is sister to Myrmecophagidae (1.00) 
Primates Yes (42%) Moves to Afrotheria (2 trees) or remains in Euarchontoglires (5 trees) Yes (1.00) Remains in Euarchontoglires (1.00) as sister to Scandentia (1.00) 
Proboscidea NA Remains in Afrotheria (71%) but is sister to Sirenia (79%) NA Remains in Afrotheria (0.98) but is sister to Sirenia (0.98) 
Rodentia Yes (24%) on 2 of 4 minimum length trees Remains in Euarchontoglires (80%) where Glires is monophyletic 92%) Yes (1.00) Remains in Euarchontoglires (1.00) as sister to Lagomorpha (1.00) 
Scandentia NA Remains in Euarchontoglires (64%) and is sister to primates (2 trees; 31%), Glires (1 tree; 12%), or Glires + Dermoptera (1 tree; 15%) NA Remains in Euarchontoglires (0.97) as sister to Dermoptera (0.59) 
Sirenia NA Remains in Afrotheria (58%) but is sister to Proboscidea (65%) NA Remains in Afrotheria (1.00) but is sister to Proboscidea (1.00) 
Tubulidentata NA Moves to Xenarthra (60%) where Tubulidentata is sister to Dasypodidae (1 tree; 31%) or is sister to Xenarthra (1 tree; 29%) NA Moves to Xenarthra (0.81) and joins Dasypodidae (0.53) 
Xenarthra No Dasypodid moves to Afrotheria (64%) and is sister to Tubulidentata (43%); myrmecophagid moves to Laurasiatheria (84%) and is sister to Pholidota (91%) No Dasypodid moves to Afrotheria (1.00) and is sister to Tubulidentata (1.00); myrmecophagid moves to Laurasiatheria (1.00) and is sister to Pholitoda (1.00) 
 Parsimony results Bayesian results 
 
 

 
Order Monophyletic? Location on shortest tree(s) when pseudoextinct + bootstrap percentages Monophyletic? Location on tree when pseudoextinct 
Afrosoricida No Moves to Laurasiatheria (96%) and joins Eulipotyphla (92%); chrysochlorid sister to talpid (69%); tenrec sister to soricid (1 tree; 28%) or talpid + chrysochlorid (1 tree; 32%) No Moves to Laurasiatheria (1.00); chrysochlorid joins talpid (0.95) and tenrec joins chrysochlorid + talpid (0.80) 
Carnivora Yes (86%) Moves to Afrotheria (42%) where carnivores are sister to paenungulates (31%) Yes (1.00) Moves to base of Placentalia [basal split between carnivores (1.00) and all other placentals (0.99)] 
Cetartiodactyla No Cetaceans move to Xenarthra (16%); artiodactyls remain in Laurasiatheria (39%) and form clade with perissodactyls (72%) No Remains in Laurasiatheria (0.67); artiodactyls are monophyletic (0.96) and join perissodactyls (0.93); cetaceans are a monophyletic (1.00) sister to chiropterans + artiodactyls + perissodactyls + pholidotans + carnivores (0.52 for cetaceans + sister taxon) 
Chiroptera Yes (100%) Moves to Euarchontoglires (2 trees; 39%) where chiropterans join Dermoptera (33%) or remain in Laurasiatheria (1 tree; 37%) where they are sister to Erinaceidae (17%) Yes (1.00) Chiropterans join Euarchontoglires (0.47) as a monophyletic sister taxon to this superordinal group 
Dermoptera NA Moves to Laurasiatheria (51%) where Dermoptera is sister to Pholidotaa (16%) NA Remains in Euarchontoglires (0.83) and is the sister to Scandentia (0.82) 
Eulipotyphla No Moves to Afrotheria (78%) where Erinaceidae joins Macroscelidea (68%), talpid joins chrysochlorid (61%), and soricid is sister to talpid + chrysochlorid + tenrecid (47%) No Moves to Afrotheria (1.00) with Erinaceidae as sister to Macroscelidea (1.00), talpid as sister to chrysochlorid (0.63), and soricid as sister to tenrecid (0.78) 
Hyracoidea NA Moves to Laurasiatheria (95%) and is the sister to Perissodactyla (62%) NA Moves to Laurasiatheria (1.00) and is the sister to Perissodactyla (1.00) 
Lagomorpha Yes (92%) Remains in Euarchontoglires (68%) and Glires (73%) Yes (1.00) Remains in Euarchontoglires (1.00) but joins murid rodents (0.61) 
Macroscelidea Yes (100%) Moves to Laurasiatheria (60%) and is sister to Erinaceidae (51%) Yes (1.00) Moves to Laurasiatheria (0.99) and is sister to Erinaceidae (0.88) 
Perissodactyla Yes (60%) Moves to Afrotheria (50%) where perissodactyls are sister to Proboscidea + Sirenia (22%) Yes (1.00) Moves to Afrotheria (0.99) as sister to Hyracoidea (0.99) 
Pholidota NA Moves to Xenarthra (79%) and is sister to Myrmecophagidae (78%) NA Moves to Xenarthra (1.00) and is sister to Myrmecophagidae (1.00) 
Primates Yes (42%) Moves to Afrotheria (2 trees) or remains in Euarchontoglires (5 trees) Yes (1.00) Remains in Euarchontoglires (1.00) as sister to Scandentia (1.00) 
Proboscidea NA Remains in Afrotheria (71%) but is sister to Sirenia (79%) NA Remains in Afrotheria (0.98) but is sister to Sirenia (0.98) 
Rodentia Yes (24%) on 2 of 4 minimum length trees Remains in Euarchontoglires (80%) where Glires is monophyletic 92%) Yes (1.00) Remains in Euarchontoglires (1.00) as sister to Lagomorpha (1.00) 
Scandentia NA Remains in Euarchontoglires (64%) and is sister to primates (2 trees; 31%), Glires (1 tree; 12%), or Glires + Dermoptera (1 tree; 15%) NA Remains in Euarchontoglires (0.97) as sister to Dermoptera (0.59) 
Sirenia NA Remains in Afrotheria (58%) but is sister to Proboscidea (65%) NA Remains in Afrotheria (1.00) but is sister to Proboscidea (1.00) 
Tubulidentata NA Moves to Xenarthra (60%) where Tubulidentata is sister to Dasypodidae (1 tree; 31%) or is sister to Xenarthra (1 tree; 29%) NA Moves to Xenarthra (0.81) and joins Dasypodidae (0.53) 
Xenarthra No Dasypodid moves to Afrotheria (64%) and is sister to Tubulidentata (43%); myrmecophagid moves to Laurasiatheria (84%) and is sister to Pholidota (91%) No Dasypodid moves to Afrotheria (1.00) and is sister to Tubulidentata (1.00); myrmecophagid moves to Laurasiatheria (1.00) and is sister to Pholitoda (1.00) 
a

Dermoptera is sister to Laurasiatheria rather than Pholitoda on the bootstrap tree.

Table 4 shows the results of pseudoextinction analyses for taxonomic subsets of topologically unstable placental orders (i.e., orders that moved to different superordinal groups in Table 3). Most taxa remained in the same superordinal group that they reside in on the molecular tree shown in Figure 1. Macroscelideans, which moved to Laurasiatheria as the sister to Erinaceidae when both species were coded as pseudoextinct, remained in Afrotheria when only one species was pseudoextinct. Similarly, carnivores remained in Laurasiatheria when only one of two carnivore taxa was pseudoextinct. However, there were six taxa (Myrmecophagidae, Dasypodidae, Chrysochloridae, Tenrecidae, Talpidae, Erinaceidae) that moved to a different superordinal group when they alone were pseudoextinct in both parsimony and Bayesian analyses (Table 4).

Table 4

Results of parsimony and Bayesian analyses when subsets of placental orders are pseudoextinct. Parsimony results are based on the most parsimonious tree(s) for each pseudoextinction analysis (1000 randomized taxon input orders; TBR branch-swapping); parsimony bootstrap percentages are based on 500 pseudoreplications with 10 randomized taxon input orders per pseudoreplicate. Posterior probabilities for Bayesian analyses are average values based on two independent runs. Bayesian analyses were run until the average standard deviation of split frequencies was ≤ 0.05. Molecular and soft morphological characters were both scored as missing for pseudoextinct taxa.

 Parsimony results Bayesian results 
 
 

 
Taxon Monophyletic? Location on shortest tree(s) when pseudoextinct + bootstrap percentages Monophyletic? Location on tree when pseudoextinct 
Myrmecophagid N/A Moves to Laurasiatheria (92%) and is sister to Pholidota (90%) N/A Moves to Laurasiatheria (1.00) and is sister to Pholidota (1.00) 
Dasypodid N/A Moves to Afrotheria (75%) and is sister to Tubulidentata (46%) N/A Moves to Afrotheria (0.78) and is sister to Tubulidentata (0.77) 
Chrysochlorid N/A Moves to Laurasiatheria (84%) and is sister to talpid (62%) N/A Moves to Laurasiatheria (84%) and is sister to talpid (62%) 
Tenrecid N/A Moves to Laurasiatheria (96%) and is sister to talpid (2 trees; 33%), erinaceid (1 tree; 20%), or soricid (1 tree; 20%) N/A Moves to Laurasiatheria (1.00) and is sister to soricid (0.83) 
Macroscelides Yes (100%) Remains in Afrotheria (86%) and is sister to Elephantulus (100%) Yes (1.00) Remains in Afrotheria (1.00) and is sister to Elephantulus (1.00) 
Elephantulus Yes (100%) Remains in Afrotheria (72%) and is sister to Macroscelides (100%) Yes (1.00) Remains in Afrotheria (1.00) and is sister to Macroscelides (1.00) 
Soricida No Remains in Laurasiatheria (54%) and is sister to talpid (40%) N/A Moves to Afrotheria (0.71) and is sister to tenrecid (0.70)* 
Talpidae No Moves to Afrotheria (91%) and is sister to Afrosoricida (76%) N/A Moves to Afrotheria (0.98) and is sister to chrysochlorid (0.84) 
Erinaceidae No Moves to Afrotheria (75%) and is sister to Macroscelidea (57%) N/A Moves to Afrotheria (0.85) and is sister to Macroscelidea (0.85) 
Yangochiroptera Yes (83%) Remains in Laurasiatheria (100%) and is sister to Pteropodidae (100%) Yes (1.00) Remains in Laurasiatheria (1.00) and is sister to Rousettus (0.80) 
Pteropodidae Yes (87%) Remains in Laurasiatheria (100%) and is sister to Yangochiroptera (100%) Yes (1.00) Remains in Laurasiatheria (100%) and is sister to Phyllostomidae (100%) 
Feliformia N/A Remains in Laurasiatheria (99%) and is sister to Caniformia (81%) N/A Remains in Laurasiatheria (1.00) and is sister to Caniformia (1.00) 
Caniformia N/A Remains in Laurasiatheria (99%) and is sister to Feliformia (78%) N/A Remains in Laurasiatheria (1.00) and is sister to Feliformia (1.00) 
Ceratomorpha No Remains in Laurasiatheria (84%); Ceratomorpha is paraphyletic to Hippomorpha with tapirid as sister to Hippomorpha (2 trees; 44%) or rhinocerotid as sister to Hippomorpha (1 tree; 33%) Yes (0.96) Remains in Laurasiatheria (0.96) and is sister to Hippomorpha (0.96) 
Hippomorpha N/A Remains in Laurasiatheria (99%) and is sister to tapirid (2 trees; 45%) or rhinocerotid (1 tree; 32%) N/A Remains in Laurasiatheria (0.99) and is sister to tapirid (0.93) 
Cetacea Yes (100%) Moves to Xenarthra (22%) and is sister to myrmecophagid (2 trees; 22%) or remains in Laurasiatheria (62%) and is sister to Chiroptera (1 tree; 20%) Yes (1.00) Remains in Laurasiatheria (0.87) in a clade with carnivores and Pholidota (0.63) 
Lama N/A Remains in Laurasiatheria (100%) and is sister to Ruminantia (66%) N/A Remains in Laurasiatheria (1.00) and is sister to Ruminantia (1.00) 
Sus N/A Remains in Laurasiatheria (95%) and is sister to a clade that contains other cetartiodactyls + perissodactyls (40%) N/A Remains in Laurasiatheria (0.93) and is sister to Lama (0.80) 
Ruminantia N/A Remains in Laurasiatheria (99%) and is sister to Lama (73%) N/A Remains in Laurasiatheria (1.00) and is sister to Lama (1.00) 
Hippopotamus N/A Remains in Laurasiatheria (92%) and is sister to Ruminantia (24%) N/A Remains in Laurasiatheria (0.92) and is sister to Lama (0.69) 
Homo N/A Remains in Euarchontoglires (73%) and is sister to Tarsius (65%) N/A Remains in Euarchontoglires (1.00) and is sister to Tarsius (0.96) 
Strepsirrhine N/A Remains in Euarchontoglires (75%) and is sister to Tarsius + Homo (67%) N/A Remains in Euarchontoglires (1.00) and is sister to Tarsius (0.91) 
Tarsius N/A Remains in Euarchontoglires (78%) and is sister to Homo (68%) N/A Remains in Euarchontoglires (1.00) and is sister to Homo (0.59) 
 Parsimony results Bayesian results 
 
 

 
Taxon Monophyletic? Location on shortest tree(s) when pseudoextinct + bootstrap percentages Monophyletic? Location on tree when pseudoextinct 
Myrmecophagid N/A Moves to Laurasiatheria (92%) and is sister to Pholidota (90%) N/A Moves to Laurasiatheria (1.00) and is sister to Pholidota (1.00) 
Dasypodid N/A Moves to Afrotheria (75%) and is sister to Tubulidentata (46%) N/A Moves to Afrotheria (0.78) and is sister to Tubulidentata (0.77) 
Chrysochlorid N/A Moves to Laurasiatheria (84%) and is sister to talpid (62%) N/A Moves to Laurasiatheria (84%) and is sister to talpid (62%) 
Tenrecid N/A Moves to Laurasiatheria (96%) and is sister to talpid (2 trees; 33%), erinaceid (1 tree; 20%), or soricid (1 tree; 20%) N/A Moves to Laurasiatheria (1.00) and is sister to soricid (0.83) 
Macroscelides Yes (100%) Remains in Afrotheria (86%) and is sister to Elephantulus (100%) Yes (1.00) Remains in Afrotheria (1.00) and is sister to Elephantulus (1.00) 
Elephantulus Yes (100%) Remains in Afrotheria (72%) and is sister to Macroscelides (100%) Yes (1.00) Remains in Afrotheria (1.00) and is sister to Macroscelides (1.00) 
Soricida No Remains in Laurasiatheria (54%) and is sister to talpid (40%) N/A Moves to Afrotheria (0.71) and is sister to tenrecid (0.70)* 
Talpidae No Moves to Afrotheria (91%) and is sister to Afrosoricida (76%) N/A Moves to Afrotheria (0.98) and is sister to chrysochlorid (0.84) 
Erinaceidae No Moves to Afrotheria (75%) and is sister to Macroscelidea (57%) N/A Moves to Afrotheria (0.85) and is sister to Macroscelidea (0.85) 
Yangochiroptera Yes (83%) Remains in Laurasiatheria (100%) and is sister to Pteropodidae (100%) Yes (1.00) Remains in Laurasiatheria (1.00) and is sister to Rousettus (0.80) 
Pteropodidae Yes (87%) Remains in Laurasiatheria (100%) and is sister to Yangochiroptera (100%) Yes (1.00) Remains in Laurasiatheria (100%) and is sister to Phyllostomidae (100%) 
Feliformia N/A Remains in Laurasiatheria (99%) and is sister to Caniformia (81%) N/A Remains in Laurasiatheria (1.00) and is sister to Caniformia (1.00) 
Caniformia N/A Remains in Laurasiatheria (99%) and is sister to Feliformia (78%) N/A Remains in Laurasiatheria (1.00) and is sister to Feliformia (1.00) 
Ceratomorpha No Remains in Laurasiatheria (84%); Ceratomorpha is paraphyletic to Hippomorpha with tapirid as sister to Hippomorpha (2 trees; 44%) or rhinocerotid as sister to Hippomorpha (1 tree; 33%) Yes (0.96) Remains in Laurasiatheria (0.96) and is sister to Hippomorpha (0.96) 
Hippomorpha N/A Remains in Laurasiatheria (99%) and is sister to tapirid (2 trees; 45%) or rhinocerotid (1 tree; 32%) N/A Remains in Laurasiatheria (0.99) and is sister to tapirid (0.93) 
Cetacea Yes (100%) Moves to Xenarthra (22%) and is sister to myrmecophagid (2 trees; 22%) or remains in Laurasiatheria (62%) and is sister to Chiroptera (1 tree; 20%) Yes (1.00) Remains in Laurasiatheria (0.87) in a clade with carnivores and Pholidota (0.63) 
Lama N/A Remains in Laurasiatheria (100%) and is sister to Ruminantia (66%) N/A Remains in Laurasiatheria (1.00) and is sister to Ruminantia (1.00) 
Sus N/A Remains in Laurasiatheria (95%) and is sister to a clade that contains other cetartiodactyls + perissodactyls (40%) N/A Remains in Laurasiatheria (0.93) and is sister to Lama (0.80) 
Ruminantia N/A Remains in Laurasiatheria (99%) and is sister to Lama (73%) N/A Remains in Laurasiatheria (1.00) and is sister to Lama (1.00) 
Hippopotamus N/A Remains in Laurasiatheria (92%) and is sister to Ruminantia (24%) N/A Remains in Laurasiatheria (0.92) and is sister to Lama (0.69) 
Homo N/A Remains in Euarchontoglires (73%) and is sister to Tarsius (65%) N/A Remains in Euarchontoglires (1.00) and is sister to Tarsius (0.96) 
Strepsirrhine N/A Remains in Euarchontoglires (75%) and is sister to Tarsius + Homo (67%) N/A Remains in Euarchontoglires (1.00) and is sister to Tarsius (0.91) 
Tarsius N/A Remains in Euarchontoglires (78%) and is sister to Homo (68%) N/A Remains in Euarchontoglires (1.00) and is sister to Homo (0.59) 
a

The average standard deviation of split frequencies remained at 0.10 after Bayesian analyses were run for 80 million generations with eight chains (one cold, seven hot) for each analysis.

Data Congruence

Current morphological data sets are small relative to molecular data sets and it remains possible that larger morphological data sets will overcome problems that beset current data sets if these problems are statistical in nature and are tied to small sample size. Alternatively, there may be fundamental differences between molecular and morphological data. We examined congruence within the molecular data set and between the molecular and morphological data after partitioning the molecular data into 21 partitions that contained the same number of phylogenetically informative characters as the morphological data set. First, we identified 12 gene segments (ADORA3, ADRA2B, ADRB2, APOB, APP, ATP7A, BDNF, BRCA1, CREM, EDG1, PLCB4, VWF) with taxonomic overlap that comprised all placental orders as well as five lineages of afrosoricidan and eulipotyphlan insectivores. Taxonomic overlap for these 12 segments included 32 placental taxa. This resulted in an alignment that included 9179 nucleotide positions, of which 3818 were phylogenetically informative. Next, the 3818 phylogenetically informative characters were used to generate 21 different molecular partitions, each of which contained 175 informative characters and matched the size of the morphological data set (175 informative characters) of Asher et al. (2003) for the same 32 placental taxa. Data for each partition were then mapped on to the most parsimonious tree(s) for each of the other data partitions and the percentage increase in the number of steps was calculated. Incongruence was indexed as the percentage increase in the number of steps for each partition relative to the number of steps on the most parsimonious tree(s) for each data partition.

The morphological data consistently emerged as the most incongruent data partition, even though molecular data were segregated into partitions that contained the same number of informative characters (175) as the morphological data set (Fig. 2). The increase in tree length associated with mapping molecular data for one partition on to the best tree(s) for another molecular partition ranged from 2.9% to 14.9% (mean = 6.7%, gray diamonds in Fig. 2) for 420 comparisons; mapping molecular data partitions on to the morphological tree (21 comparisons) resulted in increases in tree length that ranged from 9.5% to 18.1% (mean = 14.2%, open circles in Fig. 2); mapping morphological data on to trees derived from molecular data partitions (21 comparisons) resulted in increases in tree length that ranged from 15.7% to 25.5% (mean = 19.9%, black triangles in Fig. 2).

Figure 2

Plot of minimum tree length for 22 different data partitions (21 molecular, one morphological; x-axis) versus the increase in tree length when each data partition is mapped on to the best tree(s) for each of the other data partitions (y-axis). In cases where a partition was mapped on to more than one equally most parsimonious tree for another data partition, we plotted the midpoint value for the percentage increase in tree length. The 21 molecular partitions (P) arbitrarily followed the sequential gene order in our concatenated molecular data set, irrespective of gene boundaries, and included characters from the following gene segments: P1 (ADRA2B); P2 (ADRA2B); P3 (ADRA2B, ADORA3, ADRB2); P4 (ADRB2); P5 (ADRB2, APOB); P6 (APOB, APP); P7 (APP, ATP7A); P8 (ATP7A); P9 (BDNF, BRCA1); P10 (BRCA1); P11 (BRCA1); P12 (BRCA1); P13 (BRCA1); P14 (BRCA1); P15 (BRCA1); P16 (BRCA1); P17 (BRCA1, CREM, EDG1); P18 (EDG1); P19 (EDG, PLCB4, VWF); P20 (VWF); P21 (VWF). PAUP 4.0b11 (Swofford, 2003) was used to find the most parsimonious tree(s) for each data partition. We employed heuristic searches with tree-bisection and reconnection branch swapping and 1000 randomized taxon input orders. The 32 taxa included in partition congruence analyses were tamandua, armadillo, hedgehog, mole, shrew, tenrec, golden mole, sirenian, hyrax, elephant, elephant shrew, aardvark, mouse, hystricid, North American porcupine, pika, flying lemur, tree shrew, strepsirrhine, human, flying fox, rousette fruit bat, whale, dolphin, hippo, ruminant, pig, horse, ceratomorph, cat, caniform, pangolin.

Figure 2

Plot of minimum tree length for 22 different data partitions (21 molecular, one morphological; x-axis) versus the increase in tree length when each data partition is mapped on to the best tree(s) for each of the other data partitions (y-axis). In cases where a partition was mapped on to more than one equally most parsimonious tree for another data partition, we plotted the midpoint value for the percentage increase in tree length. The 21 molecular partitions (P) arbitrarily followed the sequential gene order in our concatenated molecular data set, irrespective of gene boundaries, and included characters from the following gene segments: P1 (ADRA2B); P2 (ADRA2B); P3 (ADRA2B, ADORA3, ADRB2); P4 (ADRB2); P5 (ADRB2, APOB); P6 (APOB, APP); P7 (APP, ATP7A); P8 (ATP7A); P9 (BDNF, BRCA1); P10 (BRCA1); P11 (BRCA1); P12 (BRCA1); P13 (BRCA1); P14 (BRCA1); P15 (BRCA1); P16 (BRCA1); P17 (BRCA1, CREM, EDG1); P18 (EDG1); P19 (EDG, PLCB4, VWF); P20 (VWF); P21 (VWF). PAUP 4.0b11 (Swofford, 2003) was used to find the most parsimonious tree(s) for each data partition. We employed heuristic searches with tree-bisection and reconnection branch swapping and 1000 randomized taxon input orders. The 32 taxa included in partition congruence analyses were tamandua, armadillo, hedgehog, mole, shrew, tenrec, golden mole, sirenian, hyrax, elephant, elephant shrew, aardvark, mouse, hystricid, North American porcupine, pika, flying lemur, tree shrew, strepsirrhine, human, flying fox, rousette fruit bat, whale, dolphin, hippo, ruminant, pig, horse, ceratomorph, cat, caniform, pangolin.

With one exception, these results show that for every molecular data partition, the alternate tree from morphology was worse (longer) than any tree from another molecular data partition (containing the same number of parsimony informative characters). In addition, all of the mappings of morphological data on to molecular trees were worse than any molecular tree with different molecular data. Thus, there is more conflict between the morphological data and any molecular data partition than between all possible pairs of molecular data partitions. These findings also suggest that current molecular and morphological data for placental mammal orders are not readily miscible and that the disagreement between morphological and molecular data is not simply a sample size (i.e., number of characters) problem.

Osteology remains the only source of data for most fossils, but methods for coding and analyzing such data are ineffective at discriminating between homology and homoplasy at the level of placental interordinal relationships. The previously mentioned problem of correlated character evolution (Chase et al., 2002; Kangas et al., 2004; Carrier et al., 2005) may contribute to incongruence between molecular and morphological data. Given that character matrices used by Asher et al. (2003), Zack et al. (2005), and Tabuce et al. (2007) include precisely the kinds of characters that might be expected to be convergent in mammals with similar diet and/or locomotion, the potential problem of correlated character evolution must be addressed.

Conclusions

We tested the ability of the most comprehensive morphological data set available to recapitulate the accepted phylogeny for living placental mammals, in the absence of molecular data, as a way to evaluate the accuracy of cladistic studies of fossil specimens that rely exclusively on morphology. We show that as many as 72% of the living placental orders move to a different superordinal group when molecular data are missing and that four of the 18 orders are rendered poly- or paraphyletic. Superordinal groups are never recovered as monophyletic when treated as pseudoextinct. We also show that there is fundamental incongruence between molecular and morphological data at the level of placental interordinal relationships. These results suggest that morphological studies of eutherian interordinal relationships have failed to separate homology and homoplasy and have consistently been misled by the latter. Cases in point include the recovery of Insectivora when either Eulipotyphla or Afrosoricida are pseudoextinct, Volitantia when chiropterans are pseudoextinct, Altungulata when perissodactyls are pseudoextinct, and Edentata when pholidotans are pseudoextinct. Further, these homoplastic groups that are recovered based on morphology often mix placental orders that originated on Gondwana versus Laurasia. The inadequacies of even the most extensive published morphological data for reconstructing higher level placental mammal phylogeny become even more apparent when entire superordinal groups are treated as extinct. Each of the four major clades disappears when it is pseudoextinct and is transmogrified into a polyphyletic assemblage that is integrated elsewhere across the tree. Notably, the apposition of southern versus northern hemisphere clades is all but lost when molecular data are missing for Afrotheria, Euarchontoglires, or Laurasiatheria.

Can we trust morphological cladistic analyses that place extinct aspheliscines within or at the base of Afrotheria? This question is comparable to asking whether or not pseudoextinct taxa remain in their original superordinal group or move to a different superordinal group. Our results show numerous instances of clades composed of one or two species (e.g., Afrosoricida, Perissodactyla, Macroscelidea, Carnivora, Pholidota, Xenarthra, Dasypodidae, Myrmecophagidae, Chrysochloridae, Tenrecidae, Talpidae, Erinaceidae) that move to a different superordinal group when they are pseudoextinct. These results suggest that Zack et al.'s (2005) and Tabuce et al.'s (2007) placement of aspheliscines could be a similar artifact and underscore the difficulty of reconstructing mammalian phylogenetic history for ancient, extinct lineages that are only known for osteological characters. Another problem associated with real fossils, and one that we did not address in our pseudoextinction analyses, is that fossil skeletons are rarely complete and some or many osteological characters must be coded as missing.

We agree with Jenner (2004), Wiens (2004), and Smith and Turner (2005) that systematists should continue to collect data for morphology-based phylogenetic analyses. However, our findings suggest that new methods for coding and analyzing morphological characters should be explored (also see Wiens, 2004), at least for analyzing difficult phylogenetic problems such as placental interordinal relationships. Lewis (2001) suggested scoring all variable morphological characters (i.e., informative and autapomorphic) rather than just informative characters and developed a model for analyzing appropriately scored data that will diminish certain long-branch attraction problems. Another strategy is to score morphological characters for stem or early crown representatives of placental orders. The rationale is that these taxa will more closely approximate ancestral states for living orders and reveal instances of homoplasy in more derived taxa. In one of their morphological analyses, Zack et al. (2005) scored an early Eocene perissodactyl (Hyracotherium) rather than a living representative of this order. However, perissodactyls still clustered as the sister-taxon to Hyracoidea. Tabuce et al. (2007) included stem representatives for four afrotherian orders (Macroscelidea, Proboscidea, Sirenia, Hyracoidea) and recovered a clade that grouped these taxa together to the exclusion of an early perissodactyl (i.e., Hyracotherium) and an early artiodactyl (i.e., Diacodexis). However, Hyracotherium grouped closer to paenungulates than to Diacodexis in some analyses (fig. 3b) and only joined Diacodexis when the taxonomic matrix was manipulated to exclude the Eocene fossil genus Anthracobune. These results highlight the challenges that face morphologists studying early placental diversification. Finally, studies that integrate genetics, developmental biology, and morphology and that unravel the causal links between genetic variation and morphological features (Kangas et al., 2004; Kassai et al., 2005) remain critical for developing a proper foundation for the phylogenetic analysis of morphological characters.

Problems associated with reconstructing relationships among extant placental orders with morphology alone are largely overcome by combining morphological data with molecular data because the phylogenetic signal associated with the relatively large molecular data set overwhelms the conflicting signal from the smaller morphological data set. However, combined data only mask the problem for extinct taxa where molecular data are still missing. Recovering a morphological tree that approximates the molecular/genomic tree of living placental mammals with its four superordinal clades is an important stepping-stone on the way to inferring phylogenetic relationships of extinct eutherian orders. Robust solutions for the phylogenetic placement of fossil taxa are a prerequisite for testing hypotheses concerning the biogeography and timing of the early diversification of mammals. In this context, one of the greatest challenges ahead for mammalian systematists is to tease apart homology and homoplasy in morphological characters.

acknowledgment

We thank Ron DeBry, Michael Woodburne, Michael Westerman, Carey Krajewski, John Gatesy, and two anonymous reviewers for reviewing earlier versions of the manuscript. This work was partially supported by the National Science Foundation (M.S.S. and W.J.M.).

References

Amrine-Madsen
H.
Koepfli
K.-P.
Wayne
R. K.
Springer
M. S.
A new phylogenetic marker, apolipoprotein B, provides compelling evidence for eutherian relationships
Mol. Phylogenet. Evol.
 , 
2003
, vol. 
28
 (pg. 
225
-
240
)
Asher
R. J.
Meng
J.
Wible
J. R.
McKenna
M. C.
Rougier
G. W.
Dashzeveg
D.
Novacek
M. J.
Stem Lagomorpha and the antiquity of Glires
Science
 , 
2005
, vol. 
307
 (pg. 
1091
-
1094
)
Asher
R. J.
Novacek
M. J.
Geisler
J. H.
Relationships of endemic African mammals and their fossil relatives based on morphological and molecular evidence
J. Mammal. Evol.
 , 
2003
, vol. 
10
 (pg. 
131
-
194
)
Carrier
D. R.
Chase
K.
Lark
K. G.
Genetics of canid skeletal variation: Size and shape of the pelvis
Genome Res.
 , 
2005
, vol. 
15
 (pg. 
1825
-
1830
)
Chase
K.
Carrier
D. R.
Adler
F. R.
Jarvik
T.
Ostrander
E. A.
Lorentzen
T. D.
Lark
K. G.
Genetic basis for systems of skeletal quantitative traits: Principal component analysis of the canid skeleton
Proc. Natl. Acad. Sci. USA
 , 
2002
, vol. 
99
 (pg. 
9930
-
9935
)
Eizirik
E.
Murphy
W. J.
O'Brien
S. J.
Molecular dating and biogeography of the early placental mammals
J. Hered.
 , 
2001
, vol. 
92
 (pg. 
212
-
219
)
Gibson
A.
Gowri-Shankar
V.
Higgs
P. G.
Rattray
M.
A comprehensive analysis of mammalian mitochondrial genome base composition and improved phylogenetic methods
Mol. Biol. Evol.
 , 
2005
, vol. 
22
 (pg. 
251
-
264
)
Gunnell
G. F.
Simmons
N. B.
Fossil evidence and the origin of bats
J. Mammal. Evol.
 , 
2005
, vol. 
12
 (pg. 
209
-
246
)
Hasegawa
M.
Thorne
J. L.
Kishino
H.
Time scale of eutherian evolution estimated without assuming a constant rate of molecular evolution
Genes Genet. Syst.
 , 
2003
, vol. 
78
 (pg. 
267
-
283
)
Hudelot
C.
Gowri-Shankar
V.
Jow
H.
Rattray
M.
Higgs
P. G.
RNA-based phylogenetic methods: Application to mammalian mitochondrial RNA sequences
Mol. Phylogenet. Evol.
 , 
2003
, vol. 
28
 (pg. 
241
-
252
)
Jenner
R. A.
Accepting partnership by submission? Morphological phylogenetics in a molecular millennium
Syst. Biol.
 , 
2004
, vol. 
53
 (pg. 
333
-
342
)
Kangas
A. T.
Evans
A. R.
Thesleff
I.
Jernvall
J.
Nonindependence of mammalian dental characters
Nature
 , 
2004
, vol. 
432
 (pg. 
211
-
214
)
Kassai
Y.
Munne
P.
Hotta
Y.
Penttila
E.
Kavanagh
K.
Ohbayashi
N.
Takada
S.
Thesleff
I.
Jernvall
J.
Itoh
N.
Regulation of mammalian tooth cusp patterning by ectodin
Science
 , 
2005
, vol. 
309
 (pg. 
2067
-
2070
)
Kjer
K. M.
Honeycutt
R. L.
Site specific rates of mitochondrial genomes and the phylogeny of eutheria
BMC Evol. Biol.
 , 
2007
, vol. 
7
 pg. 
8
 
Kriegs
J. O.
Churakov
G.
Keifmann
M.
Jordan
U.
Brosius
J.
Schmitz
J.
Retroposed elements as archives for the evolutionary history of placental mammals
PLoS Biol.
 , 
2006
, vol. 
4
 pg. 
e91
 
Lewis
P. O.
A likelihood approach to estimating phylogeny from discrete morphological character data
Syst. Biol.
 , 
2001
, vol. 
50
 (pg. 
913
-
925
)
Löytynoja
A.
Milinkovitch
M. C.
SOAP, cleaning multiple alignments from unstable blocks
Bioinformatics
 , 
2001
, vol. 
17
 (pg. 
573
-
574
)
Luo
Z.-X.
Cifelli
R. L.
Kielan-Jaworoska
S.
Dual origin of tribosphenic mammals
Nature
 , 
2001
, vol. 
409
 (pg. 
53
-
57
)
Madsen
O.
Scally
M.
Douady
C. J.
Kao
D. J.
DeBry
W.
Adkins
R.
Amrine
H.
Stanhope
M. J.
de Jong
W. W.
Springer
M. S.
Parallel adaptive radiations in two major clades of placental mammals
Nature
 , 
2001
, vol. 
409
 (pg. 
610
-
614
)
McKenna
M. C.
Luckett
W. P.
Szalay
F. S.
Toward a phylogenetic classification of the Mammalia
Phylogeny of the primates: A multidisciplinary approach
 , 
1975
New York and London
Plenum Press
(pg. 
21
-
46
)
Meng
J.
Hu
Y.
Li
C.
The osteology of Rhombomylus (Mammalia, Glires): Implications for phylogeny and evolution of Glires
Bull. Am. Mus. Nat. Hist.
 , 
2003
, vol. 
275
 (pg. 
1
-
247
)
Murphy
W. J.
Eizirik
E.
Johnson
W. E.
Zhang
Y. P.
Ryder
O. A.
O'Brien
S. J.
Molecular phylogenetics and the origins of placental mammals
Nature
 , 
2001
, vol. 
409
 (pg. 
614
-
618
)
Murphy
W. J.
Eizirik
E.
O'Brien
S. J.
Madsen
O.
Scally
M.
Douady
C. J.
Teeling
E.
Ryder
O. A.
Stanhope
M. J.
de Jong
W. W.
Springer
M. S.
Resolution of the early placental mammal radiation using Bayesian phylogenetics
Science
 , 
2001
, vol. 
294
 (pg. 
2348
-
2351
)
Murphy
W. J.
Pevzner
P. A.
O'Brien
S. J.
Mammalian phylogenomics comes of age
Trends Genet.
 , 
2004
, vol. 
20
 (pg. 
631
-
639
)
Murphy
W. J.
Pringle
T. H.
Crider
T. A.
Springer
M. S.
Miller
W.
Using genomic data to unravel the root of the placental mammal phylogeny
Genome Res.
 , 
2007
, vol. 
17
 (pg. 
413
-
421
)
Nishihara
H.
Hasegawa
M.
Okada
N.
Pegasoferae, an unexpected mammalian clade revealed by tracking ancient retroposon insertions
Proc. Natl. Acad. Sci. USA
 , 
2006
, vol. 
103
 (pg. 
9929
-
9934
)
Novacek
M. J.
Mammalian phylogeny: Shaking the tree
Nature
 , 
1992
, vol. 
356
 (pg. 
121
-
125
)
Novacek
M.
Mammalian phylogeny: Genes and supertrees
Curr. Biol.
 , 
2001
, vol. 
11
 (pg. 
R573
-
R575
)
Posada
D.
Crandall
K.
ModelTest: Testing the model of DNA substitution
Bioinformatics
 , 
1998
, vol. 
14
 (pg. 
817
-
818
)
Reyes
A.
Gissi
C.
Catzeflis
F.
Nevo
E.
Pesole
G.
Saccone
C.
Congruent mammalian trees from mitochondrial and nuclear genes using Bayesian methods
Mol. Biol. Evol.
 , 
2004
, vol. 
21
 (pg. 
397
-
403
)
Ronquist
F.
Huelsenbeck
J. P.
MrBayes 3: Bayesian phylogenetic inference under mixed models
Bioinformatics
 , 
2003
, vol. 
19
 (pg. 
1572
-
1574
)
Ronquist
F.
Huelsenbeck
J. P.
van der Mark
P.
MrBayes 3.1 manual
 , 
2005
Scally
M.
Madsen
O.
Douady
C. J.
de Jong
W. W.
Stanhope
M. J.
Springer
M. S.
Molecular evidence for the major clades of placental mammals
J. Mammal. Evol.
 , 
2001
, vol. 
8
 (pg. 
239
-
277
)
Scotland
R. W.
Olmstead
R. G.
Bennett
J. R.
Phylogeny reconstruction: The role of morphology
Syst. Biol.
 , 
2003
, vol. 
52
 (pg. 
539
-
548
)
Simmons
N. B.
Geisler
J. H.
Phylogenetic relationships of Icaronycteris Archaeonycteris Hassianycteris Palaeochiropteryx to extant bat lineages, with comments on the evolution of echolocation and foraging strategies in Microchiroptera
Bull. Am. Mus. Nat. Hist.
 , 
1998
, vol. 
235
 (pg. 
1
-
182
)
Smith
N. D.
Turner
A. H.
Morphology's role in phylogeny reconstruction: Perspectives from paleontology
Syst. Biol.
 , 
2005
, vol. 
54
 (pg. 
166
-
173
)
Springer
M. S.
Murphy
W. J.
Eizirik
E.
O'Brien
S. J.
Placental mammal diversification and the Cretaceous-Tertiary boundary
Proc. Natl. Acad. Sci. USA
 , 
2003
, vol. 
100
 (pg. 
1056
-
1061
)
Springer
M. S.
Murphy
W. J.
Eizirik
E.
O'Brien
S. J.
Rose
K. D.
Archibald
J. D.
Molecular evidence for major placental clades
The rise of placental mammals
 , 
2005
Baltimore
The Johns Hopkins University Press
(pg. 
37
-
49
)
Springer
M. S.
Stanhope
M. J.
Madsen
O.
de Jong
W. W.
Molecules consolidate the placental mammal tree
Trends Ecol. Evol.
 , 
2004
, vol. 
19
 (pg. 
430
-
438
)
Swofford
D. L.
PAUP*: Phylogenetic analysis using parsimony (*and other methods, Version 4.0b10
 , 
2003
Sunderland, Massachusetts
Sinauer Associates
Tabuce
R.
Marivaux
L.
Adaci
M.
Bensalah
M.
Hartenberger
J.-L.
Mahboudi
M.
Mebrouk
F.
Tafforeau
P.
Jaeger
J.-J.
Early Tertiary mammals from North Africa reinforce the molecular Afrotheria clade
Proc. Roy. Soc. B
 , 
2007
, vol. 
274
 (pg. 
1159
-
1166
)
Teeling
E. C.
Scally
M.
Kao
D. J.
Romagnoli
M. L.
Springer
M. S.
Stanhope
M. J.
Molecular evidence regarding the origin of echolocation and flight in bats
Nature
 , 
2000
, vol. 
403
 (pg. 
188
-
192
)
van Rheede
T.
Bastiaans
T.
Boone
D. N.
Hedges
S. B.
de Jong
W. W.
Madsen
O.
The platypus in its place: nuclear genes and indels confirm the sister group relation of monotremes and therians
Mol. Biol. Evol.
 , 
2006
, vol. 
23
 (pg. 
587
-
597
)
Waddell
P. J.
Kishino
H.
Ota
R.
A phylogenetic foundation for comparative mammalian genomics
Genome Informatics
 , 
2001
, vol. 
12
 (pg. 
141
-
154
)
Waddell
P. J.
Okada
N.
Hasegawa
M.
Towards resolving the interordinal relationships of placental mammals
Syst. Biol.
 , 
1999
, vol. 
48
 (pg. 
1
-
5
)
Waddell
P. J.
Shelley
S.
Evaluating placental inter-ordinal phylogenies with novel sequences including RAG1, γ -fibrinogen, ND6, and mt-tRNA, plus MCMC-driven nucleotide, amino acid, and codon models
Mol. Phylogenet. Evol.
 , 
2003
, vol. 
28
 (pg. 
197
-
224
)
Westerman
M.
Burk
A.
Amrine-Madsen
H. M.
Prideaux
G. J.
Case
J. A.
Springer
M. S.
Molecular evidence for the last survivor of an ancient kangaroo lineage
J. Mammal. Evol.
 , 
2002
, vol. 
9
 (pg. 
209
-
223
)
Wiens
J. J.
The role of morphological data in phylogeny reconstruction
Syst. Biol.
 , 
2004
, vol. 
53
 (pg. 
653
-
661
)
Wolpoff
M. H.
Paleoanthropology
 , 
1999
Boston, Massachusetts
McGraw-Hill
Zack
S. P.
Penkrot
T. A.
Bloch
J. I.
Rose
K. D.
Affinities of “hyopsodontids” to elephant shrews and a Holarctic origin of Afrotheria
Nature
 , 
2005
, vol. 
434
 (pg. 
497
-
501
)