Resolving the overall pattern of marattialean fern phylogeny

Structure of the analyses

Several types of analyses to resolve and test relationships among marattialean ferns were performed. The type 1 analysis employed morphological characters to resolve relationships among living and extinct marattialean taxa to provide the first cladistic hypothesis for the overall pattern of relationships within the clade. Type 2 analyses (i.e., “total evidence analyses”) employed the combination of morphological and nucleotide sequence characters described earlier to analyze the overall pattern of marattialean fern phylogeny. These analyses were performed using all of the taxa included in the type 1 analysis, and also after pruning 19 wildcard taxa from among fossils that are least completely characterized (i.e., have large numbers of character states coded as unknown) and/or introduce large numbers of character conflicts among species (i.e., species of Marattiopsis Schimper, Scolecopteris Zenker, and similar psaroniaceans).

Type 3 analyses explored relationships among living marattialean species using different taxon and character sets to test the effects of those variations on the resulting systematic topologies. Variations of type 3 analysis were (a) an 18-taxon set abridged from the matrix of Murdock (2008a) using the nucleotide sequences developed by Murdock (2008a), (b) the same 18-taxon set using morphological characters developed for the current study, or (c) the same 18-taxon set using nucleotide sequence characters. Four additional variations of type 3 analyses were conducted of the 18-taxon set with the addition of different outgroup and ingroup taxa, and analyzed with the same combination of nucleotide sequence and morphological characters used in analysis 2. Variations in the additional taxa included were (d) Psilotum nudum to root the tree, (e) Botrychium virginianum to root the tree, (f) the outgroup taxa Botrychium virginianum, Psilotum nudum, and Equisetum spp. with Botrychium virginianum to root the tree, and (g) the same outgroup taxa and rooting as analysis 3f plus the ingroup species Marattia excavata, Angiopteris tonkinensis, and A. henryi. Taxa included in each type 3 analysis are recorded on the trees included in Appendix S1 (see the Supplemental Data with this article).

Analysis 1

The analysis of living and extinct species using morphological characters was conducted using 94 taxa, including 10 fossil outgroup taxa, three living outgroup species, 21 living ingroup species, and 60 extinct ingroup species. One outgroup taxon (i.e., Equisetum spp.) was coded using common characters of living and extinct species of that genus, and the tree was rooted with the Lower Devonian basal polysporangiophyte Aglaophyton major (Kidston & Lang) D.Edwards. The matrix included 124 characters of which 94 are activated (i.e., used in the analyses). The other 30 characters were deactivated in WinClada because they are either invariant or systematically uninformative, but are retained for possible use by future workers. Deactivated characters are highlighted as such in Appendix S2 and include nos. 11, 16, 19, 20, 25, 29, 39, 43, 45, 51, 57, 60, 61, 65, 71, 72, 73, 75, 77, 78, 80, 81, 82, 91, 104, 105, 106, 107, 108, and 123 when the matrix is numbered from 1.

Analysis of the 94 living and extinct taxa using 94 morphological characters yields 2645 trees of 543 steps (CI = 23, RI = 72), in which 30 nodes collapse in the strict consensus tree (Fig. 1). That tree shows a monophyletic Marattiales with a basal polytomy consisting of the Pennsylvanian age fossil Radstockia kidstonii Taylor (Fig. 1, at red arrow), a clade that consists of 29 extinct and living species that conform to our previous circumscription of the crown group family Marattiaceae (Rothwell et al., 2018a), and a clade of 51 extinct species that conform to our previous characterization of the stem group family Psaroniaceae (Fig. 1). The outgroup is completely resolved, with the living homosporous pteridophyte, Psilotum nudum (L.) P.Beauv., forming the sister group to Marattiales (Fig. 1). The Marattiales clade has a basal polytomy consisting of the Pennsylvanian age fossil Radstockia kidstonii Taylor, a clade of Marattiaceae, and a clade of Psaroniaceae. The Marattiaceae clade has a basal polytomy that consists of Marattia asiatica (Kawasaki) Harris, Marattiopsis patagonica Escapa, Bomfleur, Cuneo & Scasso, M. crenulatus Lundblad, and a well-resolved tree that consists of the 21 species of living plus six extinct species of the Marattiaceae. Living species group together to form clades that attach to the stem of the Marattiaceae tree in an arrangement that generally reflects their accepted generic identities (sensu Murdock, 2008b; Fig. 1). The Permian Qasimia schyfsmae + the Jurassic Marattia anglica are attached to the stem at the node immediately distal to the basal marattiaceous polytomy. Attached at sequentially more distal nodes on the stem are the Cretaceous Marattiopsis vodrazkae, a clade consisting of Christensenia aesculifolia (Blume) Maxon + (Danaea geniculata Raddi + (D. wendlandii Rchb. f. + D. c.f. grandifolia Underw.)), Angiopteris tonkinensis (Hayata) J.M.Camus, a clade consisting of the Triassic Danaeopsis fecunda Halle + (Angiopteris henryi + (A. chingii J.M.Camus + A. itoi (W.C.Shieh) J.M.Camus) + (A. evecta (Forst.) Hoffm. + (A. angustifolia C.Presl. + A. smithii Racib.))), the Jurassic Marattia aganzhenensis S.Yang, and a clade consisting of (Ptisana fraxinea (Sm.) Murdock + (P. salicina (Sm.) Murdock + (P. melanesica (Kuhn) Murdock + P. sylvatica (Blume) Murdock))) + (Marattia laxa Kunze + Marattia excavata Underw.) + (Marattia douglasii (C. Presl.) Baker + (M. alata Sw. + (E. kaulfussii (J.Sm.) J.Sm. + E. pittieri (Maxon) Christenh.))).

The other large clade of marattialeans, the stem group Psaroniaceae, is far less completely resolved than the rest of the marattialean phylogeny. This is particularly evident at the apex of the tree, where there are 31 Scolecopteris and Scolecopteris-like species that attach at a large polytomy (Fig. 1). Most of those attach as single species, but there also are three clades of two species, one clade of three species and one clade of four species (Fig. 1). The lack of resolution in this area of the tree reflects high levels of character conflict among relatively similar extinct species that are incompletely characterized, and many of which are represented by only fertile frond fragments and spores.

Because the family Psaroniaceae is poorly resolved in our strict consensus tree (Fig. 1), we calculated a majority rule consensus tree (i.e., from the “majority fools” option of WinClada) for the phylogeny of the Psaroniaceae (Fig. 2). This tree, while far less reliable than the strict consensus topology, does provide an opportunity to test the most widely recognized pre-cladistic systematic hypothesis of relationships among species of the psaroniaceous genus Scolecopteris (Fig. 2; Millay, 1976; Lesnikowska, 1982). The majority rule consensus tree recovers the Latifolia group (green letter L). By contrast, species of the Oliveri group (orange letter O), species of the Minor group (red letter M), and species of the Alta group (blue letter A) each resolve as nonmonophyletic groups (Fig. 2). Although, with the exception of S. nigra, which occurs elsewhere on the tree, other species of the Alta group do form a clade of four species (Fig. 2).

Analysis 2

The living and extinct species from analysis 1 were analyzed with a combination of morphological characters from analysis 1 and the nucleotide sequence characters developed by Murdock (2008a) yields 11,597 trees of 1246 steps (CI = 54, RI = 80) in which 66 nodes collapse in the strict consensus tree (Appendix S3a). In the results of the “total evidence” analysis, the strict consensus tree is poorly resolved, the backbone of the tree collapsing to a large polytomy from which almost all of the extinct species diverge as single species, two, three, or four species clades (Appendix S3a). Only one clade of fossils with as many as seven species diverges from the polytomy, and that clade is incompletely resolved. While living species of Danaea, Ptisana, Eupodium, and Angiopteris each resolve a clade, no other commonly recognized generic groupings are recovered (Appendix S3a). In the majority rule consensus tree of these results (Appendix S3b), enough of the backbone of the tree is restored to resolve the Marattiaceae and Psaroniaceae clades, but more specific relationships among species are probably highly unreliable.

When 19 taxa with particularly incomplete coding of morphological characters and/or that produce high levels of homoplasy in the results (i.e., wild card taxa) are omitted from the matrix, analysis of the remaining 75 taxa yields 668 trees of 1153 steps (CI = 59, RI = 82) in which only 13 nodes collapse in the strict consensus tree (Fig. 3). These results produce a fairly well-resolved tree in which there is a monophyletic Marattiales with two subclades that diverge at the basal node (Fig. 3). One of the resulting clades includes all of the living Marattiaceae and is roughly equivalent to the traditional concept of the crown group Marattiaceae (Fig. 3). The other clade is roughly equivalent to the traditional concept of stem group Psaroniaceae (Fig. 3).

The marattiaceous clade has seven subclades attached to the stem. The Pennsylvanian Radstockia kidstonii, which forms a polytomy with species of the Psaroniaceae and the Marattiaceae in results of analysis 1 (red arrow in Fig. 1), is attached at the basal node of the Marattiaceae clade in the results of this total evidence analysis (Fig. 3). In ascending order of divergence from the stem distal to Radstockia kidstonii are a clade of Eupodium species, an incompletely resolved clade of Ptisana species and extinct species of Marattia and Marattiopsis, a clade of Christensenia aesculifolia + three species of Danaea, a clade of living and extinct Marattia species, and an incompletely resolved clade of living and fossil species of Angiopteris and the extinct Danaeopsis fecunda (Fig. 3). With the exception of two polytomies in the mid-region of the tree, relationships among the psaroniaceous species also are highly resolved (Fig. 3). The Triassic Scolecopteris antarctica Delevoryas, Taylor & Taylor and the Cretaceous Escapia christensenioides are the only species of the Psaroniaceae clade from Mesozoic strata. The remaining species are from the Paleozoic, emphasizing that the Psaroniaceae was most diverse during the Pennsylvanian and basal Permian and that it declined in prominence at the end of the Paleozoic.

Analyses 3

Analysis 3a used the same 18 species from the Murdock (2008a) study as in analysis 1, analyzed with the same nucleotide sequence character set developed by Murdock (2008a) and used in analysis 1. Analysis of the resulting matrix of 18 taxa by 532 informative characters yielded one most parsimonious tree of 688 steps (CI = 80, RI = 90). This fully resolved tree (Appendix S1a) is concordant with the tree topology obtained in the analysis of Murdock (2008a). All of the generic-group clades resolved by Murdock (2008a) are retained in the same relationships to each other (Fig. 4a; i.e., Danaea spp. + (Eupodium spp. + (Ptisana spp. + (Christensenia aesculifolia + (Marattia spp. + Angiopteris spp.)))).

Analysis 3b used the same 18 taxa as analysis 3a, which were analyzed with the same morphological character set used in analysis 1. Analysis of this matrix of 18 taxa and 94 characters yielded three trees of 129 steps (CI = 65, RI = 78). Five nodes collapse in the strict consensus tree. Species of Danaea occur at the base of the tree, Christensenia aesculifolia is attached at the next node, and species of Marattia, Eupodium, Ptisana, and Angiopteris form a terminal polytomy (Appendix S1b).

Analysis 3c uses the same 18 taxa as analyses 3a and 3b, which use a combination of the same nucleotide sequence characters as in analysis 3a and the same morphological characters in analysis 3b. Analysis of this matrix of 18 taxa and 626 characters yields one most parsimonious tree of 826 characters (CI = 78, RI = 88). The topology of this tree differs from that of analysis 3a (Appendix S1c) by attaching Christensenia aesculifolia to the stem immediately distal to Danaea spp. and having a clade consisting of (Eupodium spp. + Ptisana spp.) + (Marattia spp. + Angiopteris spp.) diverge at the next more distal node (Appendix S1c).

Analysis 3d used the same 18 living marattialean species as analysis 3c and Botrychium virginianum to root the tree. Results of that analysis yield six most parsimonious trees of 851 steps (CI = 76, RI = 87). Seven nodes collapse in the strict consensus tree. In that tree, the backbone collapses to a polytomy at the generic level (Appendix S1d).

Analysis 3e used the 18 taxon and character set from analysis 3c, except that Psilotum nudum replaced Botrychium virginianum to root the tree. That analysis yields one most parsimonious tree of 845 steps (CI = 76, RI = 87), with a topology of generic groupings that is the same as that in the results of analysis 3c (Appendix S1e).

Analysis 3f used the same 18 marattiaceous taxa as analysis 3c, the same three living outgroup taxa used in analysis 1 (i.e., Botrychium virginianum, Equisetum spp., and Psilotum nudum), and was rooted with B. virginianum. That analysis yielded one most parsimonious tree of 877 steps (CI = 75, RI = 86), with a topology that is equivalent to the results of analyses 3c and 3e (Appendix S1f).

Analysis 3g used the same 18 marattiaceous ingroup taxa, 3 outgroup taxa, and rooting as analysis 3f, plus three additional ingroup marattiaceous species (i.e., Angiopteris tonkinensis, A. henryi, and Marattia excavata) that were not included in the Murdock (2008a) analysis and that have been coded only for the morphological characters. The last three species were added to the matrix to test the stability of the results when additional living taxa are included in the analysis. Results of analysis 3g yielded two most parsimonious trees of 898 steps (CI = 74, RI = 86). One node collapses in the strict consensus tree, producing a polytomy within the Angiopteris clade. Other than the addition of two additional species of Angiopteris and one additional species of Marattia, that polytomy is the only substantive difference between the results of analysis 3f and 3g (Appendix S1g).

Relationships among stem group (Psaroniaceae) species

Results of our analysis with morphological characters also produced a well-resolved basal region of the Psaroniaceae clade, but there is a large polytomy of 31 Scolecopteris and Scolecopteris-like species distally (Fig. 1). Species with unresolved relationships are primarily from the time segment (i.e., Pennsylvanian-Early Permian) when the clade was most species rich (Rothwell et al., 2018a). A lack of resolution in that part of the tree is probably the product of several factors including (1) incomplete preservation leading to differential character codings for the included species, (2) character codings for only some parts of the sporophyte (i.e., primarily fertile pinnules and spores), and (3) an incompletely sampled fossil record.

We tested our results using comparisons to pre-cladistic groupings of the most species-rich genus Scolecopteris (Millay, 1979) and found differing degrees of concordance in results of the contrasting methodologies. Because there is inadequate resolution of scolecopterid species in the strict consensus tree of our analysis using morphological characters (Fig. 1), we have computed a majority rule consensus tree for the results (Fig. 2). In the more highly resolved topology represented by the majority rule consensus tree of the psaroniaceous clade (Fig. 2), one of the pre-cladistic groups appears to be robust, while others are not. Whereas, all species of the Latifolia group form a clade at the apex of the tree, species of the Alta group, the Minor group, and the Oliveri group all form nonmonophyletic assemblages in our results (Fig. 2).

Incomplete resolution of the Psaroniaceae tree, and the contrasting topologies of relationships inferred from the results of the phylogenetic and pre-cladistic methodologies, suggest that the sets of morphological characters used in the pre-cladistic treatments of Scolecopteris have the potential to capture at least some of the phylogenetic signal needed to resolve relationships among species of Scolecopteris. However, these results also reveal that both better characterizations of extinct species and additional refinement of characters will be required to clarify patterns of diversity during the segment of time when marattialean ferns were undergoing their most explosive evolution and had their greatest species richness (i.e., Middle Pennsylvanian-Early Permian; Rothwell et al., 2018a).

Earlier phylogenetic analyses

Because most previous phylogenetic analyses of Marattiales focus either on living Marattiaceae (Hill and Camus, 1986a; Murdock, 2008a) or extinct taxa of Psaroniaceae (Lesnikowska, 1982), and because most resolve relationships using either nucleotide sequence or morphological characters (Hill and Camus, 1986a, 1986b; Liu et al. 2000), the results presented herein are the first to be generated using a total evidence approach (Kluge, 1989). Our analyses also are the first to employ subsets of the taxa and/or characters in the total evidence analysis as reciprocal tests of the resulting hypotheses of relationships (Figs. 1-4).

Results from the early phylogenetic analysis of Lesnikowska (1982) recognized similar (if not entirely consistent) levels of partial concordance between relationships inferred from the cladistic and pre-cladistic approaches (Millay, 1976, 1979; Lesnikowska, 1982) to those found in our analysis (Fig. 2). However, the overall pattern of marattialean phylogeny was not addressed in the Lesnikowska (1982) analysis.

Results of the Hill and Camus (1986a) analysis recognize monophylesis for each of the living genera Marattia s.l., Danaea, Christensenia, and Angiopteris (i.e., including species sometimes assigned to Archangiopteris Christ & Giesenh., Macroglossum Copel., and Protomarattia Hayata in studies that predate that of Murdock [2008a]), and in this respect are concordant with those of our analyses (Figs. 1, 3, 4). Our results also resolve a polytomy among the genera Angiopteris, Marattia, and Danaea in the strict consensus tree of Hill and Camus (1986a) that anticipates the divergent topologies for relationships among these same genera that occur in the results of our analyses (Fig. 4).

A slightly revised analysis by Hill and Camus (1986b) was rooted with Ophioglossum and added the fossils Radstockia kidstonii, Qasimia schyfsmae, and Angiopteris blackii Hill to the Hill and Camus (1986a) matrix. In a representative most parsimonious tree of those results (fig. 1 of Hill and Camus, 1986b), the integrity of clades that conform to the living genera is largely maintained, but Marattia spp. form a paraphyletic assemblage. Taxa of that tree are generally arranged as Christensenia + (Psaronius-Scolecopteris + (Marattia laevis [= Eupodium pittieri] + (Radstockia + Danaea spp. + (Marattia spp. + (Qasimia schyfsmae +Marattia costulisora Alston + (Marattia fraxinea + Angiopteris blackii + Angiopteris spp. sensu Murdock)))))). The exact topology of the strict consensus tree for the most parsimonious trees from those results was not figured. However, Hill and Camus (1986b) do report that there are some rearrangements within smaller clades and one additional polytomy in the strict consensus tree.

The results of Liu et al. (2000) agree with those of our analyses by resolving Marattiales as a monophyletic group in which the living genera diverge from the stem of the tree at apical nodes (Liu et al., 2000). However, by contrast to the distinct Psaroniaceae and Marattiaceae clades resolved in our results (Figs. 1, 3), the strict consensus tree of Liu et al. (2000) resolves Marattiaceae as a tree in which Psaroniaceae forms a paraphyletic assemblage with respect to a monophyletic Marattiaceae (fig. 1a of Liu et al., 2000). When the extinct genera are deleted from the tree of Liu et al. (2000), arrangement of the terminal taxa (sensu Murdock, 2008a) conforms to Marattia + (Angiopteris [sensu Murdock, 2008a] + (Danaea + Christensenia)). That topology does not occur among the results found in our analyses (Fig. 4).

Relationships among living species of Marattiaceae

In the most comprehensive recent systematic treatment of living Marattiaceae, the results (Murdock 2008a, b) support the inclusion of three small traditionally recognized genera (i.e., Archangiopteris, Macroglossum, and Protomarattia) within the genus Angiopteris (Murdock, 2008a), a result also found in the tree of Hill and Camus (1986b). The Murdock (2008a) analysis also placed the traditionally recognized species of Marattia s.l. into three distinct clades and allowed for the segregation of the included species to Marattia s.s., Ptisana, and Eupodium (Murdock, 2008b). Topologies that underlie those generic revisions are probably heavily influenced by the molecular component of the characters analyzed by Murdock (2008; i.e., Appendix S1a), but our inclusion/exclusion and alternative character set experiments (Appendix S1a–S1g) reveal that morphological characters also lend support for Murdock's generic revisions (Murdock, 2008b).

There is far less consistency in the relationships among living genera of Marattiaceae as revealed by our analyses of alternative taxon and character sets (Fig. 4). Whereas the Danaea + (Eupodium + (Ptisana + (Christensenia + (Marattia + Angiopteris)))) topology obtained by Murdock (2008a) also occurs in our analysis of the abridged taxon set using Murdock's nucleotide sequence characters (Fig. 4A; Appendix S1a), it does not reappear in the results of any other of our inclusion/exclusion experiments (Fig. 4; Appendix S1). When the same 18 living species are analyzed using only the morphological characters developed for this analysis, Christensenia is displaced to the node above Danaea along the stem of the tree, and Marattia, Eupodium, Ptisana, and Angiopteris form a polytomy (Fig. 4B) in which species of Marattia s.s. do not form a clade (Appendix S1b). When the same 18 species are analyzed using a combination of the nucleotide sequence characters of Murdock (2008a) and the morphological characters developed in this study, full resolution of the topology is restored (Appendix S1c), but Christensenia occurs at the node distal to species of Danaea, and the remaining genera form a clade consisting of (Eupodium + Ptisana) + (Marattia + Angiopteris) (Fig. 4C).

As has been recognized for many years (i.e., Nixon and Carpenter, 1993), topologies of phylogenetic results may vary tremendously depending upon the outgroups included in the analysis and the taxon used to root the tree. When the 18 living species from Murdock are analyzed using the same morphological and nucleotide sequence characters as in the previous analysis, but the tree is rooted with Botrychium virginianum (Appendix S1d), the backbone of the tree collapses to a polytomy of all marattiaceous genera (Fig. 2D). However, if the same ingroup taxa and characters are again analyzed, but with Psilotum nudum replacing Botrychium virginianum to root the tree (Appendix S1e), full resolution of the tree is again restored. Results of the last analysis have the same topology (Fig. 4C) as described earlier for analysis of the 18-taxon set using a combination of morphological and nucleotide sequence characters (cf., Appendix S1c). The addition of two more outgroups (Appendix S1f), or of two more outgroups plus three more ingroup species (Appendix S1g) also does not alter the topology of the generic relationships (i.e., Fig. 4C). However, one node collapses within the Angiopteris clade in the results of the last analysis (Appendix S1g).

Two additional divergent topologies for relationships among genera with living species occur among the results of our various analyses. In the results of our analysis of living and fossil species using morphological characters, the tree is rooted between Danaea + Christensenia and a clade consisting of Angiopteris + (Ptisana + (Eupodium + Marattia) (Figs. 1, 4F). However, in the results of the total evidence analysis, the tree consists of Eupodium + (Ptisana + (Danaea + Christensenia) + (Marattia + Angiopteris)) (Fig. 3). Taken together, these analyses (Figs. 1, 3; Appendix S1) reveal that the topology of the various results is highly unstable (Fig. 4), and that phylogenetic relationships among living genera of Marattiaceae are not nearly as well resolved as commonly believed, a situation that also occurs when comparing clades of fossil and extant seed plants (Mathews, 2009). The realization that hypotheses of generic affinities for marattiacean fern species are far more robust than are any of the hypotheses of relationships among marattiaceous genera is consistent with the Bayesian posterior probability and maximum parsimony bootstrap values obtained by Murdock (2008a). Lower values for nodes connecting generic groupings in the Murdock results also are concordant with the inconsistency of interrelationships among living genera that occur in the results of our analyses (Fig. 4).

Relationships among extinct species assigned to Marattia and Marattiopsis

Fossil marattiaceous ferns that closely resemble living species of Marattia have been inconsistently assigned either to Marattia or Marattiopsis and are variously interspersed among living species in the results of our analyses (Figs. 1, 3). To a certain degree, this inconsistency results from incomplete resolution of relationships, but it also reflects incomplete preservation of many extinct species, and until recently, inconsistent circumscription of the concepts for Marattia and Marattiopsis (Murdock, 2008a; Escapa et al., 2015). That inconsistency has been complicated further by the segregation of living species previously assigned to Marattia s.l. into three genera with diagnostic, but relatively subtle morphological differences (Murdock, 2008b). Recognizing that the characters required to determine whether a fossil belongs to Marattia s.s., Ptisana, or Eupodium (sensu Murdock, 2008b) are not preserved in most fossils, Bomfleur et al. (2013) proposed that the generic name Marattiopsis Schimper be conserved and employed for extinct Marattia-like species that cannot be assigned with confidence to one of the segregate genera with living species. While it is beyond the scope of the current investigation to propose formal nomenclatural revisions for the extinct species of Marattia, we fully endorse the Bomfleur et al. (2013) proposal and consider all of the extinct species of Marattia in our study (i.e., dark red species of Marattia in Fig. 3) to conform to the Bomfleur et al. (2013) concept of Marattiopsis.

Resolving the overall pattern of marattialean phylogeny

Our analyses demonstrate that living species of Marattiales occupy only part of one of the two major clades within this order of eusporangiate euphyllophytes (i.e., Domain of living Marattiaceae in Fig. 3) and therefore do not accurately reflect the overall pattern of phylogeny for the clade. This conclusion is emphasized by the results of inclusion/exclusion experiments of a “total evidence” data set (i.e., that includes living and fossil taxa that are analyzed by all available categories of characters) and is readily obvious from examination of the phylogenetic distribution of living marattialean species in those results (Fig. 3).

Advantages of this total evidence approach are not always immediately apparent, because the results of such analyses often provide less resolution and lower support values for resolved nodes than do analyses of living species that are analyzed with nucleotide sequence characters. For Marattiales, results of our initial total evidence analysis emphasizes this point by yielding a large number of most parsimonious trees that produce a poorly resolved strict consensus tree (Appendix S3a), and an incompletely resolved majority rule consensus tree (Appendix S3b). However, neither of these consensus trees produces a topology for relationships among living genera that is concordant with the results of analyses that include only living species (i.e., Murdock, 2008a; Appendix S1; Fig. 4). Although the fossil record remains sparsely sampled as compared to that of living species, and relationships remain incompletely resolved in our results, this first total evidence analysis of Marattiales with taxon sampling that is roughly proportional to the species richness of the clade through time (i.e., a majority of taxa being extinct) provides a more complete picture for the overall pattern of phylogeny than has previously been achieved for the clade.

The role of total evidence in resolving overall patterns of phylogeny

Over the past few years, a rapidly growing percentage of studies that include only living plants have endeavored to resolve overall patterns of phylogeny for an ever increasing spectrum of clades, with an emphasis on deeper and deeper nodes in the phylogenies. Those analyses of systematic relationships among living species typically produce much more highly resolved trees in which individual nodes are supported by much higher support values than do the results of analyses that also include fossils.

Against this background, but contrary to the assumption that the fossil record is unlikely to overturn patterns of relationships that are resolved using only living species (Patterson, 1981), numerous classic studies of plants and animals resolved divergent patterns of relationships when fossils were either included or excluded from analyses (Rothwell et al., 2018b, introduction to this issue). When such an assumption is viewed as a hypothesis and tested by inclusion/exclusion experiments with fossil Marattiales, it is clear that relationships among living species are dramatically altered by the inclusion of extinct species (cf., Figs. 1, 3, and 4). Our results for Marattiales are concordant with those for other clades of vascular plants (e.g., Rothwell and Nixon, 2006; Stockey et al., 2016), particularly within ancient clades that have suffered a large amount of extinction (Rothwell and Nixon, 2006). Therefore, it is imperative that we employ total evidence analyses (i.e., those that include a full spectrum of representative living and fossil taxa that are analyzed with morphological, nucleotide sequence, and other characters) if we are to accurately reconstruct the pattern of phylogeny for such clades. Within this context, unless hypotheses developed in the absence of fossils are accompanied by inclusion/exclusion experiments with fossils, and other tests using data from the fossil record, there is no way to determine whether such hypotheses reflect the overall pattern of phylogeny for the clade under consideration, or merely systematic relationships among living species.