Pteridosperms are the backbone of seed-plant phylogeny1

Pteridosperms are the backbone of seed-plant phylogeny1

Hilton, Jason

HILTON J. (School of Geography, Earth and Environmental Sciences, University of Birmingham, Edgbaston, Birmingham, B15 2TT, UK) AND R. M. BATEMAN (Natural History Museum, Cromwell Road, London, SW7 5BD, UK). Pteridosperms are the backbone of seed-plant phylogeny. J. Torrey Bot. Soc. 133: 119-168. 2006.-Using Doyle (1996) as a starting point, we compiled a morphological cladistic matrix of 54 coded taxa (31 wholly extinct, and 23 at least partly extant) and 102 informative characters in order to explore relationships among gymnosperms in general and pteridosperms in particular. Our core analysis omitted six supplementary fossil taxa and yielded 21 most-parsimonious trees that generated two polytomies in the strict consensus tree, both among pteridosperms; the first affected several hydraspermans, and the second affected the three peltasperm/ corystosperm taxa analyzed. The resulting topology broadly resembled topologies generated during previous morphological cladistic analyses that combined substantial numbers of extant and extinct higher taxa. Each of the five groups that include extant taxa was relatively well resolved as monophyletic and yielded the familiar Anthophyte topology (cycads (Ginkgo (conifers (Gnetales, angiosperms)))), strongly contradicting most recent DNA-based studies that placed Gnetales as sister to, or within, conifers. These five extant groups were embedded in the derived half of a morphologically diverse spectrum of extinct taxa that strongly influenced tree topology and elucidated patterns of acquisition of morphological character-states, demonstrating that pteridosperms and other more derived “stem-group” gymnosperms are critical for understanding seed-plant relationships. Collapses in strict consensus trees usually reflected either combinations of data-poor taxa or “wildcard” taxa that combine character states indicating strongly contradictory placements within the broader topology. Including three progymnosperms in the analysis and identifying the aneurophyte progymnosperm as outgroup proved crucial to topological stability. An alternative progymnosperm rooting allowed angiosperms to diverge below cycads as the basalmost of the extant groups, a morphologically unintuitive position but one that angiosperms have occupied in several recent molecular studies. We therefore believe that such topologies reflect inadequate rooting, which is inevitable in analyses of seed plants that use only extant taxa where the outgroups can only be drawn from ferns and/or lycopsids, groups that are separated from extant seed-plants by a vast phylogenetic discontinuity that is bridged only by wholly fossil groups. Given the rooting problem, and the poverty of the hypotheses of relationship that can be addressed using only extant taxa, morphology-based trees should be treated as the initial phylogenetic framework, to subsequently be tested using molecular tools and employing not only molecular systematics but also evolutionary-developmental genetics to test ambiguous homologies. Among several possible circumscriptions of pteridosperms, we prefer those that imply paraphyly rather than polyphyly and exclude only one monophyletic group, providing one cogent argument for the inclusion of extant cycads in pteridosperms. Although pteridosperms cannot realistically be delimited as a monophyletic group, they remain a valuable informal category for the plexus of gymnosperms from which arose several more readily defined monophyletic groups of seed-plants. The ideal solution of recognizing several monophyletic groups, each of which combines a “crown-group” with one or more pteridosperms, is not yet feasible, due to uncertainties of relationship and difficulties to satisfactorily delimiting the resulting groups using reliable apomorphies. Exploration of the matrix demonstrated that coding all of the organs of a plant (extinct or extant) and dividing significantly polymorphic coded taxa are highly desirable, thereby justifying the substantial investment of time required to reconstruct individual conceptual whole plants from disarticulated fossil organs.

Key words: character analysis, cladistics, morphology, paleobotany, phylogeny, pteridosperms, seed ferns, seed plants

Seed plants are first recognized in the fossil record in the Fammenian stage of the Late Devonian (Rothwell and Scheckler 1988), and today are generally regarded as being represented by five extant groups (cycads, Ginkgo, conifers, Gnetales, and angiosperms). It is hardly surprising that these ancient and phylogenetically disparate groups vary greatly in both species diversity and ecological importance, also contrasting in their breadths of geographical distribution and their respective ranges of environmental tolerance (DiMichele et al., this volume). Despite the considerable attention given to seed plant phylogeny, their higher-level relationships remain poorly resolved and controversial (cf. Nixon et al. 1994, Rothwell and Serbet 1994, Doyle 1996, 1998; Bowe et al. 2000, Chaw et al. 2000, Magallon and Sanderson 2002, Soltis et al. 2002, Burleigh and Mathews 2004).

Over recent years a growing body of molecular research has been focused explicitly on this portion of the green-plant tree of life, primarily with the intention of more conclusively determining the phylogenetic relationships among major groups. However, no overriding consensus has emerged, and many of the individual analyses present results that are not only technically incongruent but often are also strongly contradictory (see recent summary by Burleigh and Mathews 2004). It is especially apparent that molecular trees strongly disagree with those determined through various morphological investigations. Moreover, different molecular studies disagree significantly with one another. Compounding this problem is the fact that seed plants have experienced widespread extinction though their estimated 363 million year history, so that the extant members represent only a modest (and arguably somewhat unrepresentative) proportion of the major seed-plant lineages that have existed though geological time. This causes a fundamental paradox for the current trend of prioritizing molecular over morphological data in phylogeny reconstruction, as the molecularly recalcitrant extinct groups are critical for understanding deep divergence patterns and major character transitions among seed plants (e.g., Doyle 1998, Crane et al. 2004).

One of the most significant differences between seed-plant phylogenies produced from morphological and molecular data, and among contrasting molecular analyses, is the position of Gnetales (including three extant genera: Gnetum, Welwitschia, and Ephedrd) with respect to other seed plants. Morphological evidence characteristically groups Gnetales with flowering plants, resulting in the much-discussed ‘Anthophyte hypothesis’ of relationships. However, because the taxa selected and characters defined differ significantly among morphological studies, several contrasting arrangements of extant seed-plant lineages have been proposed (Fig. 1) and no consensus Anthophyte arrangement has emerged. In contrast, the more comprehensive of the recent multigene analyses (e.g., Burleigh and Mathews 2004) consistently group Gnetales with extant conifers. Gnetales are either placed within conifers as sister to Pinaceae, termed the ‘Gnepine’ hypothesis, or as sister to all extant conifers, termed the ‘Gnetifer’ hypothesis (Fig. 2). However, yet other molecular analyses place Gnetales as sister to all other extant seed plants, generating the ‘Gnetales-sister’ hypothesis. Although most recent molecular phylogenies favor Gnepines (e.g., Bowe et al. 2000, Chaw et al. 2000, Magallon and Sanderson 2002, Burleigh and Mathews 2004), contrasting results can be generated relatively easily by varying specific details of the analysis conducted, notably the selection of taxa scored, genie regions sequenced, or tree-building algorithms employed (e.g., Rydin and Kallersjo 2002, Rydin et al. 2002). Sequence similarities of several key genes influencing reproductive development suggest a close relationship between the gnetalean Gnetum and pines/spruces (papers by Frohlich, Theissen et al., Schneider et al. in Cronk, Bateman, and Hawkins 2002). Nonetheless, the great molecular disparities that are the inevitable consequence of deep divergences leave molecular trees vulnerable to analytical artefacts such as long-branch attraction (Doyle 1998); they are especially difficult to untangle by molecular means when the divergences are closely spaced, and are even more challenging when those closely spaced divergences occurred deep in time (Bateman 1999).

While molecular investigations proceed apace, relatively little attention is being paid to using well-established morphological data matrices to test (largely independently) the molecularly-derived Gnetales-sister, Gnepine, and Gnetifer hypotheses. We therefore face a conceptual dichotomy in the evolutionary analysis of seed plants: Do we simply accept what is effectively a “majority rule consensus” of extantonly molecular topologies, which groups Gnetales with conifers, or do we continue to explore the far better sampling of higher taxa that is inherent in morphology-only analyses, which are able to combine extant and extinct seed plants independent of their stratigraphie ages?

Although analyses combining morphological and molecular matrices may in theory present a viable way forward, previous studies have shown that the sheer volume of molecular characters generally outweighs the numerically limited morphological equivalents (e.g., Doyle 1998). Certainly, it becomes impractical to simultaneously analyze extant and extinct taxa; rather, extinct taxa are inserted a posteriori into a phylogenetic framework established using only the extant taxa, thereby effectively undermining the true value of the unique morphological character combinations observed in many fossil taxa.

Although this paper represents a continuation of earlier morphology-based cladistic analyses of seed plants, we have chosen to focus primarily on what we perceive as the crucial role of pteridosperms within seed-plant phylogeny. An important motivation for developing this study has been the relative hiatus over the last decade in morphological analyses of seed plants, following a veritable avalanche of phylogenetic studies in the golden years of the mid-1990s (e.g., Rothwell and Serbet 1994, Nixon et al. 1994, Doyle 1996). Admittedly, during the subsequent, disappointingly quiescent decade, relatively few fossil pteridosperms have been reconstructed as conceptual whole plants suitable for inclusion in such analyses. Nonetheless, rather than simply reanalyzing previously existing (and arguably already over-analyzed) taxa, we have incorporated new data on several fossil seed plants subject to varying degrees of reconstruction in order to better populate the experiments that we designed to explore the significance of pteridosperms in seed-plant phylogeny. Moreover, during the same time period, several other extinct seed plants have been reconstructed, including certain fossil cordaitean coniferophytes and bona fide conifers, that together are likely to necessitate modification of previously determined phylogenetic relationships among seed plants.

We therefore assess phylogeny on an expanded morphological data set that spans not just seed ferns but representatives of all major seedplant groups (though we have deliberately not expanded the angiosperm content of the matrix, in deference to the related, angiosperm-dominated analysis of Doyle, this volume, and Bateman, Hilton, and Rudall, in press). In addition to inferring seed-plant relationships per se, we also explore the pros and cons of including both fully and partially reconstructed fossil taxa in phylogenetic analyses, attempting to discern the minimum degree of conceptual reconstruction of a pteridosperm that is necessary to yield an acceptably reliable placement and a potentially stable topology (cf. Bateman 1992, Bateman and Simpson 1998).

Methods. The matrix for the present analysis was developed from the published matrix of Doyle (1996), which in our view offered the best combination of coded taxa and characters on which to build our hypotheses. We believe that subsequent changes to the taxa analyzed (Appendix 1) and characters used (Appendix 2) were sufficient to distinguish this matrix (Appendix 3) and this analysis from their predecessors. These changes are only summarized below, as more detailed information on specific taxon sampling and character states is presented in the appendices.

SUMMARY AND RATIONALE FOR TAXON CHANGES. Core Taxa. The most obvious modification was expanding the spectrum of coded taxa to include progymnosperms, a putatively paraphyletic grade group of plants that morphological evidence convincingly demonstrates to include the ancestor of seed plants (Beck and Wight 1988) and that is widely recognized as basally divergent within lignophytes (progymnosperms plus seed plants: cf. Crane 1985, Doyle and Donoghue 1986, 1992, Nixon et al. 1994, Rothwell and Serbet 1994). Progymnosperms are valuable in giving credible polarities to characters within basal seed ferns. In addition, they test the controversial Beck hypothesis of a diphyletic origin of seed plants (Beck 1957, 1970, 1971, 1976, 1981), which contrasts most starkly with the Rothwell hypothesis of seedplant monophyly (Rothwell 1982, 1986, Rothwell and Scheckler 1988, Rothwell and Serbet 1994).

Within progymnosperms, Tetraxylopteris is the most comprehensively delimited aneurophytalean progymnosperm known from both morphology and anatomy (Beck and Wight 1988). Aneurophytalean progymnosperms occupied the basal-most position within lignophytes in most previous analyses (Doyle and Donoghue 1992, Nixon et al. 1994, Rothwell and Serbet 1994). Archaeopteris spp. represents a composite of the stems of Callixylon and foliage of Archaeopteris and is based on vegetative and fertile morphology and anatomy (Doyle and Donoghue 1992, Rothwell and Serbet 1994). Archaeopteris is the most comprehensively known archaeopteridalean progymnosperm, and shares with seed plants characteristics such as possession of heterospory and similarities of growth architecture, including leaf form (Beck and Wight 1988). Cecropsis luculentus is the youngest known progymnosperm but it can be scored only for the anatomy and organization of its fertile shoot system. In the only cladistic treatment in which it has been included to date, Cecropsis was placed as sister to Archaeopteris, this pairing being sister to the monophyletic seed-plant clade (Rothwell and Serbet 1994).

As the focus of this paper is pteridosperms, we have added several extinct seed ferns to the matrix in order to investigate their phylogenetic positions and their respective impacts on previously determined phylogenetic relationships. The Upper Devonian seed fern Laceya hibemica (as reconstructed by Matten 1992) has been included to incorporate known variation in the morphology of basal-most hydrasperman seed ferns early in their radiation. We have also added Bilignea cf. solida (as provisionally reconstructed by Bateman 1988, Bateman and Rothwell 1990) from the Lower Carboniferous of Scotland. Bilignea has reproductive features that demonstrate it to be a hydrasperman-type seed fern (setting aside a remarkable teratological cupule-pair that that bears both ovules and pollen organs: Long 1977a; Bateman and DiMichele 2002), but it differs from other hydrasperman seed ferns in wood structure and growth architecture. It is tentatively interpreted as one of the smaller members of the presumed arborescent seed ferns that diversified in the Early Carboniferous (Galtier, this volume). The Lower Carboniferous seed fern Lyrasperma (sensu Long 1960a, 1964) is markedly different from all other currently recognized hydrasperman taxa in bearing ovules that are not radially symmetrical but are winged. Unlike most other hydrasperman ovules, those of Lyrasperma possess two vascular bundles that pass though the integument in the major plane, a feature more typical of cardiocarpalean-type ovules from higher seed ferns and coniferophytes (Hilton et al. 2003).

Moving on to the higher seed ferns, we segregated Medullosans sensu Doyle (1996) into Medullosa and Quaestora, following Rothwell and Serbet (1994), in order to permit the possible emergence of relationships with the expanded range of basal seed ferns and to better represent the abundance of forms encompassed by the medullosans.

Since the last major morphological analyses of seed plants were conducted, several wholeplant Paleozoic coniferophytes have been reconstructed. Although the Cathaysian cordaitean conferophyte Shanxioxylon (as reconstructed by Wang et al. 2003) is undoubtedly cordaitean, it shares with conifers helical fertile shoot systems and small needle-like leaves, thus providing potentially important information regarding the relationships between cordaiteans, conifers, and pteridosperms. To accommodate the addition of Shanxioxylon and other recently described Paleozoic conifers, we segregated Cordaites sensu Doyle (1996) into Cordaixylon and Mesoxylon, again largely following Rothwell and Serbet (1994).

Within the conifers we included the Paleozoic conifer Thucydia (as recently reconstructed by Hernandez-Castillo et al. 2001a) in order to augment characters of the walchian conifers, as its morphology is more typical of modern conifers than is the previously analyzed extinct genus Emporia. The composite Mesozoic conifer family Cheirolepidiaceae (Watson 1988, Rothwell et al. 2006) has also been included in the matrix in an attempt to bridge the considerable morphological gap that separates Paleozoic conifer families from their extant relatives. For example, Cheirolepidiaceae has previously been proposed as a likely candidate for a ghost lineage leading to Pinaceae (e.g., Smith and Stockey 2002), and at present this is the only Mesozoic conifer family that is sufficiently well known for inclusion into this analysis. In addition, Cheirolepidiaceae possess striate pollen grains similar to those of fossil and living Gnetales, providing potential morphological evidence of the link between extant conifers and Gnetales advocated in numerous molecular studies. However, their unique combination of a proximal tetrad attachment scar with structures indicating distal germination offers possible links to other extinct seed-plant groups. Also within conifers, we have changed the name for the composite Taxodiaceae sensu Doyle (1996) to Cupressaceae. Doyle’s coding included both Taxodiaceae s.l. plus Cupressaceae, and subsequent molecular investigations have shown that Taxodiaceae represents a paraphyletic assemblage relative to a monophyletic Cupressaceae s.s., necessitating expanded delimitation to a new Cupressaceae s.l. (e.g., Burleigh and Mathews 2004).

We divided Cycadales sensu Doyle (1996) into Cycadaceae and Zamiaceae (largely following Nixon et al. 1994, but also consistent with the molecular groupings obtained more recently by Hill et al. 2003, Rai et al. 2003) in order to eliminate several polymorphic cells from the data matrix and to provide a crude test of the monophyly of cycads.

Our decision not to expand the range of extant angiosperme in our matrix beyond that of Doyle (1996) was taken primarily for pragmatic reasons, namely to avoid pre-empting the morphological cladistic reanalysis of angiosperm relationships presented by Doyle in this volume. We recognize that this decision leaves us open to potential criticism for omitting the current preferred molecular candidate for the most basally divergent extant angiosperm, Amborella. However, we have also failed to include in our analysis the taxon that had previously occupied that role for several years but now appears in increasingly derived positions in molecular phylogenies, namely the specialized aquatic Ceratophyllum. We suspect that Amborella, clearly a mainstream member of the morphologically diverse grade of basal extant angiosperms, does not deviate sufficiently strongly in morphology from other extant angiosperms included in our matrix to radically alter the resulting topology (cf. Buzgo et al. 2004, Doyle, this volume). Also within angiosperms we have modified the names of several taxa to incorporate systematic advances made since the publication of Doyle’s (1996) matrix (as detailed in Appendix 1); these changes include adoption of Piperaceae, Aristolochioidaea, and Magnoliaceae. Lastly, we have entirely omitted the Lower Jurassic Piroconites, which was interpreted by Doyle (1996) as sister to the extant Gnetales, as we discerned critical ambiguities in its reconstruction.

Supplementary Taxa. We believe that all of the fossil taxa listed thus far have been reconstructed with sufficient rigor to represent bona fide whole-plant species (Laceya, Bilignea, Lyrasperma, Shanxioxylon, Thucydia, Cecropsis) or composite higher taxa at either the generic (Tetraxylopteris, Archaeopteris) or family (Cheirolepidiaceae) level. However, we are less confident about the completeness, accuracy, and/or reliability of inferred organ correlations of the data-sets for the following fossil taxa. We therefore opted to treat them as supplementary taxa, until further details of their morphology and/or anatomy strengthen their respective reconstructions into whole-plant species. Final decisions regarding whether to treat each coded taxon as core or supplementary were taken after the initial rounds of tree-building, so that we were also able to determine whether particular taxa behaved as “wildcards”, destabilizing the overall topology as a result of their relatively ambiguous placement.

Our confidence in the data-set for the hydrasperman pteridosperm Pitus, as reconstructed from the Lower Carboniferous of Scotland by Long (1963, 1979), was constrained by knowledge that alternative combinations of organs (and thus of characters) have been proposed for this reconstructed plant (e.g., Retallack and Dilcher 1988), and that more than one biological species of Pitus is likely represented by the organs thus far documented in the genus (Galtier, this volume).

The more derived pteridosperm Nystroemia pectiniformis, reconstructed by Hilton and Li (2003) from the Permian of China, is important as it has non-radially symmetrical ovules that were interpreted as hydrasperman in organization, even though the plant bore large flattened leaves on shoots distinct from those generating the fertile organs (Hilton and Li 2003, Wang et al. 2003). However, other key features of the plant, including its stelar organization and its cuticle, are unknown. Also, the nature of its branching organization is equivocal; Wang et al. (2003) argued that it bore bracts with axillary shoots and more closely resembled a conifer.

The Paleozoic coniferophyte Barthelia (Rothwell and Mapes 2001) and the putative vojnovskyalean coniferophyte Sergeia (Rothwell et al. 1996) are both treated as supplementary taxa. The former remains relatively enigmatic and the latter is known only known from its ovulate shoot systems. Fortunately, Paleozoic and Mesozoic coniferophytes are currently the subject of ongoing phylogenetic investigations (Hernandez-Castillo et al. in progress).

We also treated as supplementary taxa the controversial fossil angiosperms Archaefructus (Sun et al. 1998, 2001, 2003), as reinterpreted by Friis et al. (2003), and Sinocarpus (Leng and Friis 2003). Both taxa have extensive missing data and are still under active investigation; Archaefructus having been subjected to strongly contrasting interpretations. Otherwise, we did not significantly modify the angiosperm taxa included in the analysis of Doyle (1996), as (a) relationships among angiosperms were not the primary focus of this study, and (b) we were aware that a companion treatment was being produced for this volume by Doyle (this volume).

SUMMARY AND RATIONALE FOR CHARACTER CHANGES. Expanding the matrix of Doyle (1996) to include progymnosperms required several changes to the delimitation of characters and character states, most notably introducing new characters to distinguish progymnosperms from seed plants. Specifically, these involved an additional character contrasting the homosporous and heterosporous life histories (character 1), and expanding the branching characters to contrast apical branching in progymnosperms with axillary branching in seed plants (character 4). Other changes associated with the addition of Tetraxylopteris spp., Archaeopteris spp., and Cecropsis included characters distinguishing them from seed plants based on the presence/ absence of ovules/seeds (character 60) and an integument (character 63). Inclusion of progymnosperms also necessitated reordering of a few characters so that the likely plesiomorphic states within the data matrix (and in particular in the outgroup taxon Tetraxylopteris) scored 0; these included secretory structures (character 32) and structures surrounding the megasporangium/ ovule (character 38). The final set of changes necessitated by the addition of progymnosperms affected the terminology applied to the units bearing the reproductive organs. We have expanded ovule and ovule-bearing characters (characters 34-36, 52-53, 59, 71) to include megasporangia and ovules, and have expanded pollen organ and fertile shoot characters (characters 44-48, 78-81) to include microsporangia and pollen organs. (These changes are further elaborated in the descriptions of individual characters in Appendix 2.)

The addition of new basal seed-fern taxa (Laceya, Bilignea, Lyrasperma) required several changes to the characters used by Doyle (1996), in order to accommodate the greater degree of variation observed in these taxa. Following Galtier (1988), a new state was introduced for the stele (character 18) in progymnosperms, segregating the former state 0 (protostele or arcuate primary xylem segments) so that a revised state 0 (lobed protostele or actinostele) is contrasted with state 1 (parenchymatized protostele). We also recognized as character 30 the presence of a sparganum cortex as a distinguishing feature of basal seed ferns and some cordaitean coniferophytes (it is present in Mesoxylon and Cordaixylon but absent from Shanxioxylon). Segregating the medullosans of Doyle (1996) into Quaestora and Medullosa required modification of the nodal anatomy character (20) to include new state (4): leaf traces from two or more protoxylem strands or bundles over a length of stem, reflecting the taxonomic treatment and character coding of Rothwell and Serbet (1994).

We accommodated additional variation recognized within ovules and seeds by adding a new state to character 61 relating to ovule symmetry. This followed inclusion of Shanxioxlon, in which the ovules, like those of Emporta, are bilaterally symmetrical but do not exhibit 180° rotational symmetry (Rothwell 1986). To better incorporate this condition we segregated the platyspermic condition of Doyle (1996) into two states, one for 180° rotational symmetry and the other for bilateral symmetry; this treatment followed Rothwell and Serbet (1994) and Rothwell (1986).

Doyle (1996) viewed hydrasperman reproduction as a presence/absence character, whereas in the current treatment we disassembled this syndrome into several non-identical presence/absence characters in order to analyze these characters independently, as their distribution across basal seed plants is not uniform. These included presence/absence of a membranous pollen chamber floor (character 66) and of a central column (character 68), both of which are essential parts of the hydrasperman reproduction apparatus but are also present in some (but not all) medullosan ovules. We also added a new, at least partially independent, character for patterns of sealing of the integumentary apex post-pollination (character 65), accommodating information presented by Serbet and Rothwell (1995). While revising these character codings we also reworded the definitions of the nucellar apex, adopting the term ‘salpinx’ to replace the previously ambiguously defined term ‘lagenostome’ (see below and Fig. 3). Lastly, we treated the distinction between multiple integumentary bundles and two bundles positioned in the major plane (character 70) as fundamental within seed plants.

Other changes to the matrix included the addition of a new character (27) for the recognition of xylem pit-tori. As elaborated by Doyle (2004, this volume), this feature potentially unites Gnetales with conifers. Character 46, describing the style of dehiscence in seed plants, was receded from Doyle (1996), following recognition that this character was originally mis-scored by Nixon et al. (1994). We also modified the coding of proembryo tiering (character 97) in Gnetales, following the recommendation of Doyle (pers. comm. 2004; Doyle, this volume).

Ovule position (character 35) was rescored for cycads, following Doyle (pers. comm. 2004). A new character relating to the presence/absence of girdling leaf traces (character 14) was added to the matrix, as this condition became a synapomorphy of cycads once they had been segregated into Cycadaceae and Zamiaceae.

Finally, our detailed analyses of a wide range of anatomically preserved hydrasperman ovules led us to revise the terminology used to describe the nucellar apex in order to eliminate several long-standing ambiguities. The resulting terminology (compared with several previous accounts in Fig. 3) accords with, but elaborates, the conclusions of Rothwell (1986).

CLADISTIC METHODOLOGY. In keeping with relative time of appearance in the fossil record, and widely held views on character polarities, the primitive progymnosperm Tetraxylopteris spp. was specified as outgroup in all tree-building analyses. In total, 47 ingroup taxa were analyzed (Appendix 2). Of these, 24 were wholly extinct, whereas a further 23 (11 of them angiosperme) contained at least one extant species. Together, these coded taxa represented all previously recognized major groups of lignophytes, including progymnosperms (Archaeopteris spp., Cecropsis), hydrasperman seed ferns (Elkinsia, Laceya, Bilignea, Lyrasperma, Heterangium, Lyginopteris), medullosan seed ferns (Medullosa, Quaestora), higher seed ferns (sensu Rothwell and Serbet 1994: Callistophyton, Corystosperms, Autunia, Peltaspermum, Glossopteris, Caytonia), cycads (Cycadaceae, Zamiaceae), Ginkgo, cordaitean coniferophytes (Cordaixylon, Mesoxylon, Shanxioxylon), extinct (Emporta, Thucydia, Cheirolepidiaceae) and extant (Pinaceae, Podocarpaceae, Araucariaceae, Cupressaceae, Cephalotoxus, Taxaceae) conifers, Gnetales (Gnetum, Ephedra, Welwitchia), Bennettitales, and extant angiosperms (Magnoliaceae, Eupomatia, Austrobaileya, Chloranthaceae, Laurales, Winteraceae, Eudicots, Aristolochioidaea, Piperaceae, Nymphaeales, Monocots).

The six supplementary taxa consisted of the hydrasperman seed fern Pitus, the presumed derived seed fern Nystroemia, the enigmatic coniferophytes Barthelia and Sergeia, and the extinct angiosperms Archaefructus and Sinocarpus (Appendix 1).

In total, 102 characters were employed (Appendix 2), spanning the range of observations known for individual plant taxa and categorized as follows: life cycle and growth architecture (characters 1-8), morphology and anatomy of vegetative units (characters 5-12), stem anatomy and organization (characters 13-33), organization of fertile shoot systems (characters 34-58), megasporangia, ovules, and megaspores (characters 59-77), microspore and pollen morphology and ultrastructure (characters 78-87), and gametophytes, fertilization, embryology, and germination (characters 88-102).

Data were compiled on a PC using Nexus Data Editor (NDE) version 0.5.0 (Page 2001) and analyzed using PAUP 4.Ob 12 (Swofford 2002). The resulting topologies and characterstate distributions were further explored using MacClade 4 (Maddison and Maddison 2000). Within each taxon, characters that could not be scored (either because the information was not available or the character was inapplicable) were scored as ‘?’ and treated as missing data in the analysis. Characters scored for more than one state in a taxon (a relatively high probability when, as here, some coded taxa are composite higher taxa) were treated as polymorphic. Heuristic searches were undertaken on uniformly unordered characters using TBR branch swapping and with 1000 random addition replicates and MULPARS in effect. Bootstrap analyses were undertaken using the same heuristic search strategy and 100 random additions. Decay indices (= Bremer support indices: Bremer 1988) were calculated using TreeRot v2 (Sorenson 1999).

In order to explore the relative strengths of the phylogenetic signals contributed by different organs and combinations of organs, the initial matrix was partitioned according to the seven categories outlined in Table 1. Each character category was then subjected to both Approaches 1 and 2 of Bateman and Simpson (1998): respectively, the generation of semistrict consensus trees (sensu Bremer 1990) from the submatrices, and summary of homoplasy indices based on the relative performance of character categories across the original, whole-matrix analysis.

Results. MATRIX COMPLETENESS. The full matrix of 54 coded taxa and 102 parsimony-informative characters contained 31.5% ambiguous cells (30.7% coded missing or gaps, plus 0.8% coded as polymorphic), though this figure decreased by 3.5% in the core matrix following exclusion of the six wildcard taxa. The least complete of all the coded taxa, the fossil angiosperm Sinocarpus, contained 80% ambiguous cells, and even the least complete core taxon, the pteridosperm Aurunia, contained 67% ambiguous cells. By contrast, the most complete core taxa, the extant angiosperme Winteraceae and Austrobaileya, contained only 5% ambiguous cells. With regard to characters, only five characters (C1, 35, 54, 59, 60) were unambiguously coded for all 54 taxa, while ambiguous cell frequencies approximated 80% in several characters and reached 87% in Character 57.

WHOLE-PLANT ANALYSES. Strict Consensus Topology Obtained From the Core Analysis. The analysis using all 102 characters and all 48 core taxa generated a single island (cf. Maddison 1991) of 21 most-parsimonious trees, each 356 steps in length with a CI of 0.457 and RI of 0.801. Homoplasy values for the other experiments (see below) are very similar to that of the core analysis. We will outline the topology by working from the base to the apex of the “ladderized” tree.

The strict consensus (Fig. 4) places Archaeopteris and Cecropsis as sisters adjacent to the outgroup taxon Tetraxylopterls, reaffirming the widely held view of progymnosperms as a basal within the lignophyte clade and demonstrating them to be paraphyletic to the robustly monophyletic seed-plants.

Immediately above the progymnosperms is a polytomy that consists of all six analyzed taxa of hydrasperman pteridosperms (Elkinsia, Laceya, Bilignea, Lyrasperma, Heterangium, Lyginopteris: their relationships are explored in greater detail in the next subsection). Next to diverge are the medullosan pteridosperms Medullosa and Quaestora, which form a weakly supported clade, followed by Callistophyton representing the callistophytalean pteridosperms. This is in turn followed by the most primitive extant lignophytes, the cycads; the two chosen families are well-supported as a monophyletic group.

The next node is a polytomy subtending the three peltaspermalean pteridosperm taxa analyzed in the peltasperm/corystosperm groups. However, the supposed peltasperm Autunia consistently grouped with corystosperms rather than with Peltaspermum, implying that corystosperms should be accommodated within an expanded peltasperm group (these relationships are explored in greater detail in the next subsection).

The next node subtends Ginkgo, marking the basalmost position for a plant traditionally recognized as a coniferophyte. Immediately above this node, two major clades diverge: the remaining coniferophytes, and the gnetalean plus glossophyte clade (Fig. 4).

The clade of coniferophytes s.s. (i.e., excluding Ginkgo) diverges basally into cordaite and conifer clades. Cordaites resolve Shanxioxylon as sister to Cordaixylon plus Mesoxylon that is sister group to an extinct/extant conifer group including the Palaeozoic, Mesozoic, and modern conifers. The two basal nodes in the conifers subtend two fossil taxa: first Thucydia and then Emporta. Above Emporta is a clade that comprises the extant families Pinaceae and Podocarpaceae, followed by the extinct conifer family Cheirolepidiaceae, then Araucariaceae, then Cupressaceae, and finally Cephalotaxus plus Taxaceae.

The sister clade to coniferophytes s.s. dichotomizes basally into glossophytes and a well-supported monophyletic Gnetales clade, within which Ephedra is similarly well-supported as sister to Gnetum plus Welwitschia.

The glossophyte clade consists of a grade of four extinct taxa that we have collectively termed the “higher seed ferns”; subtending a strongly monophyletic angiosperm clade. Among the higher seed ferns, glossopterids are tentatively placed as sister to Pentoxylon; this pairing is succeeded by Bennettitales and then by the much-discussed Mesozoic pteridosperm Caytonia, which occupies the coveted position of sister-group to the angiosperms.

Relationships inferred within angiosperms are somewhat unconventional, and none gained bootstrap support. Basalmost are Nymphaeales, followed by Piperaceae and then monocots plus Aristolochioidaea. Eudicots are shown as sister to a clade of primitive dicots containing Winteraceae, Laurales plus Chloranthaceae, Austrobaileya, and Eupomatia plus Magnoliaceae.

In summary, several widely recognized higher taxa are shown to be monophyletic (e.g., medullosan seed ferns, cycads, cordaites, conifers, Gnetales, and angiosperms), while coniferophytes s.1. would be rendered monophyletic simply by omitting Ginkgo to generate coniferophytes s.s. In contrast, several other widely recognized higher taxa are unequivocally paraphyletic. Progymnosperms are paraphyletic to seed plants; hydrasperman seed ferns are probably paraphyletic to the remaining seed plants, and gymnosperms are paraphyletic to angiosperms. More pertinent to this paper, all five “crown groups” that are defined by nodes immediately subtending extant seed plants are embedded within the highly morphologically divergent pteridosperms, and each “crown group” has as its sister group one or more taxa traditionally viewed as pteridosperms.

Variation Among the Most-Parsimonious Trees. Incongruences among the 21 most-parsimonious trees generated two polytomies reflecting local instability in the strict consensus tree (Fig. 4): the first is the basal node within seed plants (here termed the hydrasperman polytomy), whereas the second occurs higher in the tree, immediately above cycads, and affects the trio of Peltaspermum, Autunia, and corystosperms (here termed the peltasperm polytomy).

The hydrasperman polytomy affects all six coded taxa of hydrasperman pteridosperms (Elkinsia, Laceya, Bilignea, Lyrasperma, Heterangium, Lyginopteris). It reflects eight alternative most-parsimonious arrangements (Fig. 5), ranging from a fully ladderized paraphyletic grade below all other seed plants that is consistent with their respective first appearances in the fossil record (Fig. 5a) to a monophyletic hydrasperman clade (Fig. 5h) that strongly contradicts first appearances. Each of the seven paraphyletic arrangements (Fig. 5a-g) has at its base successive divergences of the two oldest taxa, Elkinsia followed by Laceya, whereas they form a sister pair in the monophyletic hydrasperman arrangement (Fig. 5h). Bilignea is sister to Lyrasperma in three of the seven topologies (Fig. 5b-d) while Bilignea and Lyrasperma form a polytomy within the hydrasperman pteridosperms in two of the topologies (Fig. 5f-g). The greatest topological instability is caused by Lyginopteris and Heterangium. They are never paired but always closely juxtaposed, and in five of the seven topologies they form a grade immediately below the remaining seed plants (albeit together with Bilignea plus Lyrasperma in Fig. 5e). It is unclear whether Heterangium (Fig. 5a, d, g) or Lyginopteris (Fig. 5c, f) is sister to the non-hydrasperman seed plants.

The second polytomy in the strict consensus also reflects uncertain relationships among pteridosperms and is confined to what we term peltaspermalean pteridosperms. The affected node, which is located immediately above the node from which cycads arise, subtends Peltaspermum, Autunia, and corystosperms (Fig. 6). The composite taxon corystosperms is consistently sister to the peltasperm genus Autunia, but this pairing occurs below the node at which Peltaspermum arises in seven trees (Fig. 6a), above the node from which Peltaspermum arises in eight trees (Fig. 6b), and as sister to Peltaspermum in six trees (Fig. 6c). This topology therefore suggests that Peltaspermales s.s. are not monophyletic.

EXPERIMENTS WITH SUCCESSIVE ADDITIONS OF SUPPLEMENTARY TAXA. We then used the core analysis as a framework for investigating the topological effect of successively adding to the matrix single supplementary fossil taxa, in order to explore both their likely phylogenetic positions and, in the case of those acting as “wildcard” taxa, to infer the underlying cause of the topological instability introduced into the tree. We reiterate that these taxa were treated as supplementary not because of the effect that they induced when added to the analysis but in recognition that they are less well characterized than the core taxa. The six supplementary fossil taxa can be viewed as three pairs: two pteridosperms, two coniferophytes, and two angiosperms. Note that, in this and the following two subsections, sets of most-parsimonious trees represent a single island unless otherwise stated.

Addition of the hydrasperman pteridosperm Pitus to the core analysis had little effect, generating 24 most-parsimonious trees 358 steps long. Thus, tree length increased by two steps, and Pitus simply added a seventh coded taxon to the hydrasperman polytomy.

Addition of the more derived pteridosperm Nystroemia generated 21 most-parsimonious trees 358 steps long. Inclusion of Nystroemia added only two steps to the tree and left the core topology intact, simply inserting Nystroemia onto the backbone of the strict consensus between the node subtending the medullosan seed fern clade and that subtending Callistophyton. This topology confirmed the assignment of Nystroemia to pteridosperms proposed by Hilton and Li (2003).

Inclusion of the enigmatic coniferophyte Sergeia similarly added only two steps, but it behaved as a wildcard taxon; the number of mostparsimonious trees rocketed to 1898. It induced in the strict consensus tree a major polytomy at the base of the coniferophytes plus higher taxa and caused the medullosan clade to collapse. It also shifted the node subtending Gnetales from immediately below glossopterids plus Pentoxylon (Fig. 4) to immediately above, so that Gnetales became sister to Bennettitales plus Caytonia plus angiosperms.

Inclusion of Barthelia also added two steps, and had similar topological effects to adding Sergeia, though it generated only 98 most-parsimonious trees of 358 steps.

Moving on to consider enigmatic extinct angiosperms, inclusion of Sinocarpus resulted in 21 most-parsimonious trees, two steps longer. The topology of the core analysis was unaltered, and Sinocarpus was placed within the basal extant angiosperms as sister to Piperaceae.

Inclusion of Archaefructus in the core analysis added four steps and generated 88 most-parsimonious 360 step trees. It induced in the strict consensus a major polytomy at the base of angiosperms, collapsing the relationships previously inferred among Nymphaeales, Piperaceae, Aristolochioidaea plus monocots, and eudicots.

Thus, of our six supplementary fossil taxa, three had a radical effect on the nature and/or resolution of the resulting consensus topology, whereas the remaining three did not. The effect of adding combinations of supplementary taxa is, to some extent, predictable from their behaviour when added individually. In general, addition of two systematically distant taxa had little effect, whereas addition of two systematically closer taxa caused serious instability. For example, including both Pitus and Nystroemia generated most-parsimonious trees in which both taxa occupied the same positions as when they were inserted singly, producing 24 mostparsimonious trees of 360 steps, four steps longer than the core analysis. Similarly, adding Nystreoemia plus Sinocarpus generated most-parsimonious trees in which both taxa occupied the same positions as when they were inserted singly, producing 21 most-parsimonious 360 step trees.

In contrast, adding both Archaefructus and Sinocarpus again resulted in a polytomous relationship among several groups of basal angiosperms, generating 399 most-parsimonious trees of 365 steps. Likewise, adding the both enigmatic coniferophytes Sergeia and Barthelia induced a polytomy at the base of the coniferophytes, yielding 231 most-parsimonious 359 step trees. As in their individual effects, Gnetales were shifted one node up the backbone, but in addition Cheirolepidiaceae was placed as sister to all extant conifers, forming a clade that was in turn visualized as sister-group to all other extinct conifers plus cordaites.

EXPERIMENTS INVOLVING CHANGES IN SPECIFIED OUTGROUPS. Our core analysis included three progymnosperms, among which we specified Tetraxylopteris, widely viewed as the most primitive of the three, as outgroup. However, experimentation showed that choice of progymnosperms to be included in the analysis, and choice of the progymnosperm(s) to be identified a priori as outgroup(s), had profound effects on the most-parsimonious topologies generated from our matrix (Fig. 7).

For example, simply omitting Cecropsis from the analysis generated three most-parsimonious trees 354 steps long. Hydraspermans form a monophyletic group with Lyginopteris at its base (Fig. 7a), a result consistent with some of the most-parsimonious trees generated from the core analysis (Fig. 5h) but highly discordant with first appearances in the fossil record. Omitting Archaeopteris from the analysis but retaining Tetraxylopteris as sole outgroup generated 21 most-parsimonious trees of 353 steps. This simply generated a polytomy among the hydraspermans (Fig. 7b), as in the core analysis (Fig. 4). Omitting both of the heterosporous progymnosperms, Archaeopteris and Cecropsis, generated 18 most-parsimonious trees of 346 steps. These resolved Elkinsia as basal, followed by Laceya, though the remaining hydrasperman taxa formed a polytomy (Fig. 7c). More significant changes resulted from the omission of Tetraxylopteris and the ensuing selection of Archaeopteris as outgroup. This rooting resulted in a basal pteridosperm clade that consisted of all callistophyte/ medullosan and hydrasperman taxa (18 trees of 350 steps: Fig. 7b), but with their relative placements reversed relative to the core analysis; Callistophyton occupies the basalmost position within this clade, followed by medullosan and hydrasperman taxa. Above this basal pteridosperm clade occurs a monophyletic cordaite/ conifer clade. Other significant differences from the core analysis (Fig. 4) include the position of Ginkgoales and cycads (now represented as sisters above the cordaite/ conifer clade) and Gnetales. In contrast, the arrangements within both the conifer and angiosperm clades (Fig. 7d) are retained from the core analysis.

The final rearrangement omitted both Tetraxylopteris and Archaeopteris, thus of necessity using Cecropsis (Stubblefield and Rothwell 1989) as outgroup. This experiment generated 18 most-parsimonious trees of 348 steps and yielded a topology radically different from any of the other strict consensus trees generated in this analysis (Fig. 7e). Effectively, the overall polarity of evolution in seed plants is reversed. The clade comprising Bennettites as sister to Caytonia plus angiosperms that terminates the core analysis (Fig. 4) is transferred to the node immediately above the Cecropsis outgroup. Moreover, the remaining clades diverge in reverse order, from glossopterids through Gnetales, coniferophytes, Ginkgoales, peltasperms and cycads, and culminating in the hydrasperman pteridosperms (Fig. 7e). This remarkable topology overturns not only the fossil record but also every conventional wisdom on patterns of morphological evolution in seed plants, and amply demonstrates the risk of selecting as sole outgroup a coded taxon that is relatively rich in both missing values and derived character states.

EXPERIMENTS WITH TOPOLOGICAL CONSTRAINTS: Is MONOPHYLY OF EXTANT GYMNOSPERMS CREDIBLE? Rapidly accruing molecular data strongly support not only monophyly of extant angiosperms, which was recovered from our morphological analysis, but also monophyly of extant gymnosperms, which was not (for recent summary see Burleigh and Mathews 2004). We therefore analyzed our matrix using the enforced topological constraints option of PAUP. This approach retained only those trees that comply with monophyly of both extant gymnosperms, and extant angiosperms.

The constrained search used the same strategy as the core analysis and generated 1964 mostparsimonious trees of 360 steps, four steps longer than the unconstrained result. However, variation among the most-parsimonious trees is sufficient to generate a strict consensus tree that breaks the enforced constraints of the individual most-parsimonious trees. The resulting polytomy includes extant gymnosperms plus angiosperms plus their extinct sister-taxa (Fig. 8). The medullosan and corystosperm plus Autunia clades collapse, as do many of the nodes within the coniferophytes. Relationships are largely preserved within the glossophyte clade but, even more critically, Ginkgoales and Gnetales are drawn into the large (and decidedly unhelpful) polytomy that has appeared at the base of the coniferophytes.

The selected most-parsimonious constrained tree (Fig. 9) is well resolved. It contains only one polytomy and that is of minor importance, failing to resolve the relationship between the hydraspermans Bilignea and Lyrasperma. The hydrasperman taxa diverge from the main axis in a credible sequence (cf. Fig. 5g), and the relationships among the more primitive pteridosperm groups are consistent with those in the unconstrained analysis. Relationships within the glossophyte clade and the coniferophyte clade are also unchanged.

However, enforcing monophyly on the extant gymnosperms pulls the cycads up the tree into a predictable position as basal to the other extant seed plants, immediately below the peltasperm grade. Equally significant is the translocation of Gnetales from being sister to glossophytes (Fig. 4) to being sister to coniferophytes (Fig. 9). Collapsing this topology to extant taxa only yields a Gnetifer result (cf. Fig. 2b), wherein Gnetales are sister-group to conifers (Chaw et al. 1997, 2000, Bowe et al. 2000, Rydin et al. 2002, Soltis et al. 2002). This relationship implies an expanded concept of gnetifers that also includes extinct conifers and cordaitean coniferophytes. This result accords with recent studies of Gnetalean morphogenesis (Mundry and Stiitzel 2004), and reinforces interpretations of cordaitean coniferophytes as sister-group to conifers s.l. (e.g., Rothwell 1988).

Discussion. COMPARISON WITH PREVIOUS TOPOLOGIES. Overview. When our 21 mostparsimonious trees (e.g., Fig. 10) are converted to a strict consensus tree (Fig. 4), most of the nodes that collapse subtend the basal seed-fern grade. The only exception is the node subtending corystosperms plus Autunia plus Peltaspermum, which also collapses. Thus, with the exception of the basal pteridosperms, we can use our strict consensus tree as a basis for topological comparison with the results of previous phylogenetic studies.

The earliest morphological cladistic of analysis combining extant and extinct coded taxa of seed plants, by Hill and Crane (1982), had limited taxon sampling and generated trees non-algorithmically. Thus, the most appropriate initial yardstick for comparison is the trees generated from the rather similar matrices of Crane (1985) and Doyle and Donoghue (1986), both of which subsequently underwent a series of generally relatively minor modifications (Crane 1985, 1988, Doyle and Donoghue 1992). Substantially different matrices were developed by Loconte and Stevenson (1990), which reached its maximum development as Nixon et al. (1994), and by Rothwell and Serbet (1994), who presaged the present study in paying particular attention to progymnosperms and early pteridosperms.

More recently, emphasis has switched from morphological to molecular cladistic analyses, which are of necessity confined to extant taxa. When reviewed historically through the last decade it becomes evident that these studies have shown an overall trend of increase in number of coded taxa and/or the number of genomic regions that are sequenced and subsequently combined in the analysis (Hamby and Zimmer 1992, Hasebe et al. 1992, Albert et al. 1994, Goremykin et al. 1996, Chaw et al. 1997, 2000, Qiu et al. 1999, Sanmigullin et al. 1999, Winter et al. 1999, Bowe et al. 2000, Frohlich and Parker 2000, Nickrent et al. 2000, Gugerli et al. 2001, Mathews and Donoghue 2002, Rydin et al. 2002. Soltis et al. 2002, Becker and Theissen 2003. Rai et al. 2003, Burleigh and Mathews 2004).

The following topological comparisons focus less on the many similarities than on substantial deviations between previous topologies and our own.

Crane plus Doyle and Donoghue. Crane (1985) analyzed 20 coded taxa, all of which are represented in our study by the same or very similar taxonomic entities. The comparison here is with the strict consensus of five most-parsimonious trees (Crane’s fig. 20). Much of the ensuing discussion focused on the Anthophyte clade, which contained the three extant gnetaleans as sister to angiosperms, with this entire group sister to bennettites plus Pentoxylon. None of these relationships emerged from our study, which placed Caytonia as sister to angiosperms, and glossopterids as sister to Pentoxylon. In the study of Crane (1985), these two coded taxa were placed in a large clade of seedplants that lacked flowers, as sister to corystosperms. Gingko was placed as sister to conifers plus conifer relatives, while medullosans were tentatively placed as sister to cycads, this clade being inserted at the base of the non-flowerbearing seed plants. Lyginopteris was placed below the other pteridosperms, immediately above the single progymnosperm outgroup, in a topology consistent with our own.

Doyle and Donoghue (1986, 1987) preferred to focus their discussion on one of 36 most-parsimonious trees from their 20-taxon matrix (their fig. 4). Although the range of taxa and characters broadly resembled those of Crane (1985), the topology differed in several places, and more closely resembled our own. Bennettites plus Pentoxylon plus Gnetales were sister to angiosperms, and this version of the Anthophyte clade was in turn sister to first Caytonia and then glossopterids. Immediately beneath this clade, cycads appeared in a relatively derived position, as sister to Peltaspermum. As in Crane (1985), ginkgos were sister to conifers plus conifer relatives. Basal lignophytes were more intensively sampled by Doyle and Donoghue, branching off from the base in a sequence broadly congruent with our own preferred most-parsimonious tree: aneurophytes > archaeopterids > lower lyginopterids > higher lyginopterids > medullosans. The subsequent modifications of Doyle and Donoghue (1992, fig. 2) served only to slightly reposition corystosperms.

The reanalysis of Doyle (1996) increased the number of coded taxa to 36, mainly by splitting the previously unified angiosperm category into 11 coded taxa (but at the same time omitting progymnosperms from the opposite end of the tree). The resulting strict consensus of 123 mostparsimonious trees (Doyle 1996, fig. 5) contained many more unresolved nodes clustered in three polytomies: (1) ginkgos versus cordaites versus conifers, (2) Callistophyton versus corystosperms versus Autunia versus peltasperms versus cycads versus confers plus conifer relatives, and (3) glossopterids versus Pentoxylon versus Gnetales versus bennettites versus Caytonia plus angiosperms. Thus, inevitably, much of Doyle’s topology is congruent (or at least not incongruent) with our own, differing primarily in treating ginkgos as close relatives of conifers. The basal portion of the cladogram is less wellsampled than that of Doyle and Donoghue (1986), and remains consistent with the present topology: Elkinsia > Lyginopteris > medullosans. As the present study was derived by modification of the Doyle (1996) matrix, topological similarities were likely.

The analyses of Doyle (1996) and ourselves show that merging Peltaspermum and Autunia into a composite peltasperm group is unwarranted (cf. Rothwell and Serbet 1994). However, this conclusion in turn challenges our own treatment of corystosperms as a single composite coded taxon, thus precluding effective testing of the monophyly of Corystospermales s.s. (cf. Taylor and Taylor 1993, Klavins et al. 2002, Taylor et al., this volume). The relatively large proportion of missing values in the relevant data-sets (approximately one third) also raises concerns, and suggests that this group should be a high priority for further paleobotanical investigation. For the present, we recommend treating these taxa as a single informal group, the peltaspermalean pteridosperms (Fig. 4).

Rothwell and Serbet. Rothwell and Serbet (1994, fig. 4b) presented a strict consensus of 12 most-parsimonious trees for their 27 coded taxa, which broadly resembled the range used in previous studies, differing primarily in the addition of a few fossil progymnosperms, pteridosperms, and conifers. Their study once again recovered the Anthophyte clade (though showing bennettites as polytomous with angiosperme and Gnetales). Conifers and conifer relatives formed a large polytomy that encompassed ginkgos and acted as de facto sister to the anthophytes. Sister to both was a polytomy of Callistophyton versus peltasperms versus corystosperms plus glossopterids plus Caytonia, a topology radically different from that presented here. The basal portion of their study encompassed a similar range of coded taxa to our own, and generated a similar but better resolved sequence of divergences: aneurophytes > archaeopterids plus Cecropsis > Elkinsia > Heterangium > Lyginopteris > Medullosa plus Quaestora.

Loconte and Stevenson, and Nixon et al. The last morphological comparison is with the matrix of Loconte and Stevenson (1990), which was subsequently assimilated into the more detailed analysis of Nixon et al. (1994, fig. 3). Nixon et al. analyzed 46 coded taxa, including most of those analyzed by previous workers but also dividing cycads into three families, bennettites into three genera, and angiosperms into 18 entities. Their strict consensus of 834 most-parsimonious trees yielded a polytomy among the majority of the angiosperm taxa, another among the three bennettites and, most critically, a polytomy between corystosperms versus Lepidopteris versus Tatarina (both peltasperms) versus glossopterids versus Caytonia versus Pentoxylon versus Ginkgo versus cordaites versus conifers versus a restricted version of the Anthophyte clade that contained bennettites plus Gnetales plus angiosperme.

Other notable features of their topology that contrast with our own include their unique placement of Ephedra below the remaining gnetaleans, and the relatively early divergence of a cycad clade immediately above Medullosaceae and immediately below Callistophyton, thus inverting the respective positions of these two groups in our tree. In contrast, Medullosaceae, Lyginopteris, and the two progymnosperms analyzed by Nixon et al. (1994) occupied their usual respective phylogenetic placements, consistent with our own results.

KEY CHARACTERS SEPARATING THE MAJOR GROUPS OF PTERIDOSPERMS. This paper deliberately focuses more on topologies than the underlying character-state transitions. Nonetheless, in order to explore which character states are most important for distinguishing among, and determining the relative phylogenetic placements of, pteridosperms, we used MacClade to explore character-state distributions across the selected most-parsimonious tree under Acctran optimization (Fig. 10).

Two characters that link basal seed ferns to progymnosperms (effectively symplesiomorphies in the context of this matrix) are summarized in Figure 11. Radial symmetry of the microspore/ pre-pollen/ pollen (character 90: Fig. lia) characterizes progymnosperms and hydrasperman seed ferns but it is absent from medullosan seed ferns. Radial seed symmetry (character 61: Fig. 11b) is observed in all hydrasperman seed ferns included in this analysis except Lyrasperma (bilateral symmetry also characterizes other, as yet unreconstructed hydrasperman seed ferns: e.g., Hilton et al. 2003), and it is also present in Zamiaceae and Gnetum. Also, the reliability of bilateral seed symmetry as a phylogenetic character has been challenged, as it may in part reflect developmental fine-tuning to the architecture of the surrounding organs (Rothwell 1986, Doyle 1996, Hilton et al. 2003). Seed symmetry is among the nine unambiguous character-state transitions on the branch underpinning the clade of callistophytalean seed ferns (Fig. 12). This morphological long branch emphasizes the distinctness of the callistophytalean seed ferns from the hydrasperman and medullosan groups, and identifies an obvious phylogenetic location for inferring missing taxa or saltational macroevolution (Bateman and DiMichele 1994a).

The distributions of states of other characters are more complex. Patterns of cauline protoxylem organization (character 17: Fig. 13a) show that a single central strand unites Tetraxylopteris with the majority (but not all) of the hydrasperman taxa, so that derived sympodial architectures are present not only in all other seed plants but also in the heterosporous progymnosperms Archaeopteris and Cecropsis, suggesting two independent origins of the derived state. Likewise, a sparganum/dictyoxylon cortex (character 30: Fig. 13b) is present in Tetraxylopteris, hydrasperman, medullosan, and callistophytalean seed ferns, but also the cordaitean sister-genera Cordaixylon and Mesoxylon. In this topology the sparganum/dictyoxylon cortex has been lost once in progymnosperms, on the branch leading to Archaeopteris plus Cecropsis, and a second time on the branch immediately above the node subtending Callistophyton. Both cauline protoxylem (Fig. 13a) and cortex type (Fig. 13b) are important, as these vegetative characters link the primitive progymnosperm Tetraxylopteris to primitive hydrasperman pteridosperms such as Elklnsia. This topology effectively by-passes the vegetatively specialized progymnosperm clade of Archaeopteris plus Cecropsis, permitting the aneurophyte origin of seed plants postulated by Rothwell and Erwin (1987). In this context, the placement of the heterosporous progymnosperms Archaeopteris and Cecropsis as a clade rather than a grade is critical to our interpretation of character-state transitions. Nonetheless, the transition from a homosporous to a heterosporous life history (e.g., Bateman and DiMichele 1994b) contributes seven unambiguous character-state changes (Fig. 12) that help to place the heterosporous progymnosperms as sister to the seed plants, a position supported by some other characters such as leaf architecture (e.g., Rothwell and Serbet 1994).

Other characters that are important for separating basal seed ferns are pollination biology and features of the nucellar apex that include integumentary sealing of the nucellar apex following pollination (character 65), pollen chamber type (character 67: Fig. 13c), and mode of pollen chamber sealing (character 69: Fig. 13d). A membranous pollen chamber floor and the absence of pollen chamber sealing unify the six hydrasperman taxa but provide a link with one or other rather than both of the medullosans: Medullosa lacks a membranous pollen-chamber floor, while Quaestora indulges in pollen-chamber sealing; both states are characteristic of more derived pteridosperms, including Callistophyton, and demonstrate an important role for medullosans as potential phylogenetic intermediates between early hydraspermans and more derived seed plants.

Detailed analysis of individual characters and their evolutionary significance across the entire seed-plant tree will be presented elsewhere (work in progress).

PARTITIONING THE MATRIX. Homoplasy indices in subsets of the full matrix. Bateman and Simpson (1998, figs 1, 19) divided 129 published morphological cladistic matrices into mutually exclusive subsets of vegetative and reproductive characters in order to compare, via regression against homoplasy values for the complete set, the relative strength of signals as reflected in ensemble values for the consistency index (CI) and retention index (RI) (their Approach 2). They observed on average only slightly less homoplasy when regressing the subsets of reproductive characters against the whole matrices. As expected, that pattern is also evident in the present matrix; relative to the whole matrix, reproductive characters have an ensemble CI 0.04 greater and an RI just 0.02 greater (Table 1). However, when the reproductive subset is compared directly with the vegetative subset, the differences in levels of homoplasy appear more profound; CI is 0.11 greater and RI 0.07 greater.

These character subsets are further partitioned in Table 1, which reveals a more complex pattern of distribution of homoplasy among organs, but also demands a brief explanation of the relationship between CI and RI. RI values are typically greater than CI values for that same subset of morphological characters (e.g., Givnish and Sytsma 1997), for a variety of reasons detailed by Bateman and Simpson (1998). Not surprisingly, that general pattern is repeated in this study, where the two measures of homoplasy yield a correlation coefficient of 0.67, indicating significant divergence. Approximately two-fold differences are evident between the least and most homoplastic of the seven subsets, but the identity of these subsets differs between the two measures: gametophytes/ embryogeny versus microsporangia/ microspores for CI, but gametophytes/ embryogeny versus overall architecture for RI. Values of CI and RI are most similar in architecture (low value for RI, albeit based on a sample of only six characters) and gametophytes/ embryogeny (high values for RI and CI, obtained from a more character-rich subset).

Similar contrasts in homoplasy levels between organ-based subsets were noted by Bateman and Simpson (1998, figs 14, 17) in their detailed analysis of the fossil rhizomorphic lycopsids: the obvious group to compare with seed plants, given that they independently achieved a similar level of vegetative and reproductive sophistication (Bateman 1994). Sadly, the unusually low level of homoplasy observed in gametophyte/ embryogeny characters for seed plants cannot realistically be compared with the very few such characters that fell into this category in the Iycopsid study. It is tempting to speculatively interpret the low homoplasy in these small-scale and, in many cases, internal characters as reflecting the likelihood that they are under less direct influence from the environment. However, in contrast, the lowest level of homoplasy observed in a category of characters common to both studies was observed in fertile shoots, offering modest encouragement to those systematists who are inclined to emphasize such characters.

Both studies revealed relatively high levels of homoplasy for overall architecture, which can change rapidly and radically during evolution and exerts a strong influence over habitat preference. High homoplasy levels in microspores were also common to the two studies, and are most obviously attributable to the ease with which a single-celled organ can be evolutionariIy modified and/or the direct relationship between its morphology and highly specialized function.

The relatively high homoplasy in stem characters was similarly common to the two studies, but it also serves to illustrate the limitations of causally interpreting homoplasy indices in isolation of other considerations. In the rhizomorphic lycopsids, as in the seed plants, RI and CI were only moderate, but nonetheless they contributed the key synapomorphies delimiting many of the more important (and more biologically credible) clades. In particular, they operated in synergy with cone (i.e., reproductive shoot) characters to dictate the main structures evident in the cladograms (Bateman and Simpson 1998), and the likelihood exists that such synergies between characters are also of crucial importance in the present study. Exploring this area requires greater emphasis on the nodebased comparisons that underpin Bateman and Simpson’s “Approach 1”.

Tree-building from subsets of the full matrix. Approach 1 of Bateman and Simpson (1998) requires the generation of a semi-strict consensus tree from the full matrix and another from each subset of characters for the same range of coded taxa, so that the analyst can determine the proportion of congruent nodes between the subset consensus and the full consensus (effectively acting as a relative “true tree”). Here, just one character subset was subjected to Approach 1. The group chosen was the 19 megaspore/ovule characters (C59-77: Table 1) that are, by definition, of particular interest in any study of seed-plant origins.

The semistrict consensus for all characters and the 48 core taxa resolved 41 of a potential maximum of 47 nodes (87%). By contrast, the semistrict consensus tree for the megapore/ovule characters alone (summarizing 100,000 stored trees 52 steps in length) resolved only six nodes (13% of the potential maximum). Three of these nodes were contradicted by relationships evident in the full analysis; each simply linked pairs of coded taxa via single putative synapomorphies (Elkinsia plus Laceya; Bennettitales plus WeIwitschia; Winteraceae plus Nymphaeales). Although the remaining three resolved nodes were congruent with the full analysis, none was especially profound. Two simply united the pair of heterosporous progymnosperms (Archaeopteris plus Cecropsis) with all seed plants, while the third grouped the three magnolialean angiosperms analyzed (Magnolia plus Eupomatia plus Austroballeya). It is noteworthy that the poor and substantially inaccurate resolution obtained for seed-plant ovules parallels that obtained for the corresponding megaspore character category in the rhizomorphic lycopsid study of Bateman and Simpson (1998, fig. 10), where only one accurate node and one inaccurate node were resolved out of a potential maximum of 14.

Thus, at least within the context of the present matrix, ovule characters alone generate a very poorly resolved phylogeny, and half of the few relationships that they do imply are presumed to be inaccurate due to their incongruity with trees generated from the all-organ matrix. Clearly, matrices coding all of the organs of a plant are far more powerful in taxonomically broad cladistic analyses, justifying the substantial investment of time required to reconstruct individual conceptual whole plants (cf. Chaloner 1986, Bateman and Rothwell 1990, Bateman 1992, Hilton et al. 2003, Wang et al. 2003, Hilton and Bateman 2005).

DEFINING AND DELIMITING PTERIDOSPERMS. Taxonomic Inflation. Pteridosperms have traditionally been defined as seed plants that bore seeds directly on fern-like leaves (e.g., Oliver and Scott 1904, Seward 1917, Scott 1923; for recent synthesis see Rothwell, this volume). Not surprisingly, this definition accords well with those hydrasperman and medullosan taxa that prompted early recognition of pteridosperms as an important, previously unrecognized group of plants that could readily be diagnosed by a remarkable and novel combination of morphological characters (e.g., Halle 1927, 1929). With the benefit of hindsight, colored by a century of subsequent research, we can see that the range of morphology accepted in the group known as pteridosperms has since been expanded greatly, both toward the base of the seed-plant phylogeny to encompass lyginopterids and other hydrasperman taxa other than Lyginopteris as delimited by Oliver and Scott (1904), and upward toward the five extant groups of seed plants, encompassing several other groups of putative seed ferns (Taylor et al., this volume). This constitutes a classic case of taxonomic inflation.

Expanding the concept towards the base of the phylogeny in order to include basalmost hydrasperman seed ferns (sensu Rothwell and Serbet 1994) incorporates taxa that have fertile shoot systems that are clearly separated from the pinnate leaves. This architecture is evident in the oldest well-known seed fern, Elkinsia (Serbet and Rothwell 1992), as well as other primitive Devonian-Carboniferous taxa such as Laceya. However, it is also evident in more derived hydrasperman taxa such as Nystroemia (sensu HiIton and Li 2003). Both Elkinsia and Laceya occupied basal positions, consistent with their earliest appearances in the fossil record, in the majority of our most-parsimonious trees obtained by us (Fig. 10). This demonstrates that the traditional view of pteridosperms as bearing seeds (or single seeds enveloped by tightly fitting cupules, as in the archetypal pteridosperm, Lyginopteris oldhamia) directly on the leaves is a derived character.

Passing up the tree from medullosans (Fig. 4), the traditional concept of pteridosperms accords well with the characters of callistophytaleans (Rothwell 1980, 1981, Galtier and Béthoux 2002), as they bore seeds directly on fern-like leaves. However, it is less applicable to the grade of peltaspermalean seed ferns that includes corystosperms. Ovulate organs of peltasperms comprise a central axis bearing alternate branches that terminate in cupules or cupular disks that bear ovules on their adaxial surface (Harris 1937, Townrow 1960, Taylor et al. this volume). The cupular disks have been described as megasporophylls (Taylor et al., this volume) but are distinct from foliage of the same plants and are not fern-like. Corystosperms possess fertile shoots separated from vegetative leaves that comprise helically inserted laterals, each terminating in recurved cupules bearing ovules (Klavins et al. 2002, Taylor et al., this volume). This organization is more reminiscent of the separation between fertile and vegetative organs in Elkinsia and certain other hydrasperman taxa, and demonstrates evolutionary loss in Mesozoic seed plants of seeds bom on fern-like leaves.

Moving further up the tree, neither Pentoxylon nor Bennettittales have previously been regarded as bona fide seed ferns. Both of these taxa have complex fertile shoot systems rather than seeds attached directly to fern-like leaves. Bennettitales have long been widely viewed as a distinct group of seed plants. Cladistic treatments have consistently treated Pentoxylon as closely related to Bennettitales, Gnetales, and angiosperms plus Gnetales, together united as the Anthophyte clade (Nixon et al. 1994, Rothwell and Serbet 1994, Doyle 1996). However, the current analysis placed Pentoxylon as sister to the glossopterid seed ferns (Figs 4, 10), which despite possessing many derived character states do still bear seeds directly on foliar organs (Doyle, this volume, Gould and Delevoryas 1977, Taylor et al., this volume) that could in theory be homologized with highly modified leaves. This observation suggests that Pentoxylon may be better viewed as a similarly highly derived seed fern. Most importantly, Rothwell and Serbet’s (1994) now widely accepted revision of the fertile shoot system of Pentoxylon reinterpreted its supposed cone axis as a sporophyll bearing at its apex a crowded aggregation of ovules (Rothwell and Serbet 1994, p. 479). This interpretation implies that Pentoxylon bore ovules on a leaf-like structure, albeit one that had been highly reduced, strengthening the phylogenetic link between Pentoxylon and glossopterids, and thus suggesting that a seed-fern affinity for Pentoxylon may be justified.

Early investigations (e.g., Chamberlain 1913) interpreted Bennettittales as being closely related to cycads, based on what are now widely perceived as convergent morphological similarities of growth form and leaf architecture (e.g., Crane 1985, 1988). More recent treatments have inferred close relationships of Bennettitales with Gnetales plus angiosperms; thus, Bennettitales played an important role in the development of the Anthophyte hypothesis (e.g., Doyle and Donoghue 1986, Nixon et al. 1994, Rothwell and Serbet 1994, Doyle 1996). However, in the present analysis (Fig. 4), Bennettitales diverge immediately above glossopterids plus Pentoxylon and immediately below Caytonia, another taxon that has long been viewed as a highly derived pteridosperm but that bears its ovules on a specialized axial structure.

In Caytoniales, ovules are born in fertile shoots that comprise a slender axis with suboppositely arranged cupules that bear numerous ovules attached directly to the midrib (Harris 1940, 1964). The midrib suggests that the cupule in Caytonia may be a modified leaf, thereby establishing that its ovules were borne on a leaf or leaf homologue. However, this arrangement is fundamentally different from that exhibited in other seed ferns. Moreover, the ovules were enclosed in a cupule that was sufficiently modified from the plesiomorphic form to suggest analogy with the carpel, and thus encouraged pre-cladistic hypotheses that Caytonia was a possible ancestor of the angiosperms (e.g., Thomas 1925, Long 1977a). Not surprisingly, the phylogenetic position and taxonomic affinity of Caytonia remain problematic.

If we accept the inferred pteriodspermalean affinities of glossopterids, Pentoxylon, and Caytonia, Bennettitales are sandwiched between two groups that, rightly or wrongly, are widely regarded as being seed ferns. However, the ovulate receptacles of Bennettitales are unlike the ovulebearing structures of bona fide seed ferns (and also of Pentoxylon: Rothwell and Serbet 1994), reinforcing their interpretation as a distinct group of seed plants. It is highly desirable to expand future cladistic analyses of seed plants to include more than one reconstructed bennettite, in order not only to test their monophyly but also to determine the sequence in which the character states that distinguish Bennettitales were acquired. Given the present topology (and the uncertainties that surround it), we provisionally treat Caytonia, Bennettitales, and Pentoxylon plus glossopterids as collectively constituting a paraphyletic grade of ‘higher seed ferns’ that immediately subtends angiosperms. However, we recognize that this interpretation could easily change with more comprehensive taxon sampling within corystosperms, glossopterids, and Bennettittales. In the present analysis, each of these diverse groups is represented only by a single placeholder, so their monophyly effectively remains untested.

Determining the Best Methodology for Delimiting a Phylogenetic Group. It has become increasingly popular in cladistic circles (e.g., Smith 1994, Kitching et al. 1999) to distinguish between a “crown group” of extant species plus any extinct species nested within it (e.g., species A-H, including the extinct species D and G, in Fig. 14) and a typically paraphyletic “stemgroup” of related but extinct species (species I and J in Fig. 14). In morphological phylogenetic analyses the “stem-group” often provides one or more taxa selected by the analyst as outgroups (species J in Fig. 14), since extinct species are judged on average more likely than extant species to exhibit suites of plesiomorphic character states. Bateman and DiMichele (2003) argued that this conventional distinction between crown group and stem group is meaningless, serving primarily to imply that fossil species are less important than extant species for phylogeny reconstruction. Prioritizing crown groups places undue emphasis on the most deeply divergent of the extant species under consideration (species H in Fig. 14), emphasis that merely reflects the happenstance of that species’ survival to the present day. It also arbitrarily differentiates between fossils nested within the crown group and those diverging below the node that delimits the crown group (e.g., species D and G versus I and I in Fig. 14). Moreover, both the crown group and the stem group are relative concepts. For example, reducing the phylogenetic spectrum encompassed by the fictitious analysis to species A-G would allow us to view species G as stem group relative to crown group A-F, and further reducing the spectrum to species A-D would allow us to view species D as stem group relative to crown group A-C.

In fact, the key difference between extinct and extant species is practical rather than philosophical; specifically, the far greater testability that is inherent in the living relative to the dead. Examples of such tests include the ability to directly observe both ontogeny and “behaviour,” to distinguish ecophenotypic from genuinely heritable influences on phenotype, to suppress or enhance genes within individuals, and transfer genes among individuals, in order to better understand their function, and to explore crucial aspects of gene expression such as pleiotropy and epigenetics (Bateman and DiMichele 2002, 2003). In addition, the theoretical impossibility, and practical implausibility, of extracting DNA from fossils older than ca 100 ka (e.g., Pääbo et al. 2004) means that all major taxa of seedplants not represented in the extant flora are in practice impervious to DNA-based analyses.

A related philosophical debate in phylogenetics pertinent to this discussion surrounds the controversial “Phylocode” (cf. Benton 2000, Cantino and de Queiroz 2000, Nixon and Carpenter 2000, Forey 2002), which seeks to establish formal, node-based definitions of monophyletic groups as an alternative to the long-established Linnaean hierarchy. Although we have yet to be convinced of the overall merit of this classificatory scheme, its advent has usefully rekindled a more urgent debate regarding the best protocol for delimiting monophyletic groups in cladograms. The three alternatives, based on nodes, stems, or apomorphies, are summarized with regard to the status of fictitious group AD in Fig. 14. A node-based group arbitrarily excludes all seven character-state transitions occurring on the branch immediately subtending the most recent common ancestor of the group A-D, whereas a stem-based group equally indiscriminately includes all those transitions. In contrast, an apomorphy-based concept forces the systematist to define the derived group on the basis of a single apomorphy that is considered by the systematist to be most diagnostic. For example, one could use double fertilization leading to a functional triploid endosperm to delimit the angiosperme (e.g., selecting apomorphy 4 to delimit group A-D in Fig. 14).

Ideally, the character state regarded as diagnostic is readily recognized in the phenotype, is readily fossilized, originated only once (i.e., is not a convergence/ parallelism), and has been retained in the clade throughout all of the subsequent evolutionary events. The value of the character will be further strengthened if the genetic transition that caused its first appearance can also be determined. Sadly, in practice, few characters match this ideal. It often proves necessary to select more than one diagnostic apomorphy from the relevant branch, while recognizing that that adding further coded taxa to the analysis could separate the supposedly diagnostic characters. Nonetheless, this apomorphybased approach negates two long-standing controversies: whether to include in phylogenetic analyses fossil species, and whether to award a node immediately subtending a fossil species equal standing with a node immediately subtending an extant species, since in an apomorphy-based classification the answer to both questions must by definition be affirmative. In the hypothetical example of Fig. 14, fossil species D performs a vital role in demonstrating that apomorphies 1-5 arose in the lineage before apomorphies 6-7. Equally importantly, the apomorphy-based approach allows us to distinguish between the point of origin of a clade and any subsequent radiation within that clade, an especially crucial distinction for any biologist seeking ever-popular “key innovations”.

Strengths and Weaknesses of Applying an Apomorphy-Based Approach. If we return to the traditional definition of a pteridosperm, as a plant bearing seeds on fern-like foliage, we can immediately see that the definition is not strictly apomorphy-based. Rather, it combines one synapomorphy (or, more accurately, a suite of functionally linked synapomorphies) relating to the seed habit with another suite of leaf characters that is, in the context of seed-plants, symplesiomorphic. Moreover, as we have already seen, the range of taxa, and thus of character states, encompassed by pteridosperms has been inflated beyond the original derived hydrasperman and medullosans taxa in both phylogenetic directions. Consequently, the criterion of fern-like leaves, however broadly defined, has been superseded.

However, the major delimitational problem presented by the pteridosperms only becomes fully apparent when we consider the relative phylogenetic positions of groups that (a) have rarely if ever been ascribed to the pteridosperms and (b) are resolved, generally with above-average confidence, as monophyletic. Firstly, it is no concidence that the majority of these clades contain extant species: angiosperms, Gnetales, conifers, Ginkgo (monophyly of Ginkgoales has not yet been adequately tested), and cycads. To these examples of monophyletic non-pteridosperm groups we can tentatively add the sistergroup to conifers, the wholly extinct cordai tes; they collectively form the relatively ill-defined coniferophytes. Each of the remaining clades is held together by several character-state transitions that generally include at least one reliable synapomorphy.

However, removing these clades leaves three exclusively fossil groups, two of which are clearly paraphyletic and the third equivocally paraphyletic. The basal seed-plant group could be described as the core pteridosperms; the equivocally paraphyletic hydraspermans subtending potentially monophyletic groups of first medullosans and then callistophytaleans (Fig. 4). Only the uncertainly placed, but robustly monophyletic, cycads separate the core pteridosperms from the ambiguously placed and poorly resolved peltasperm-corystosperm group. These taxa have been viewed as pteridosperms by some authors (e.g., Taylor et al., this volume) but as independent orders within the paraphyletic class Gymnospermopsida by others (e.g., Stewart and Rothwell 1993, and they played a key role in the development of the mostly male theory of angiosperm origins; Frohlich, in press). They are separated by no less than four nodes from the final paraphyletic group of four taxa (glossopterids plus Pentoxylon, Bennettitales, Caytonia), each of which has been treated as a separate order of gymnosperms in some classifications (e.g., Stewart and Rothwell 1993) but at least some of which have alternatively been placed in a broader pteridosperm group (e.g., Taylor and Taylor 1993, Taylor et al., this volume).

Thus, no feasible circumscription would allow pteridosperms to be viewed as monophyletic. Four possible circumscriptions (Figs 4, 10) merit serious consideration:

(1) Pteridosperms sensu stricto is a single paraphyletic and wholly extinct group containing the hydrasperman grade plus the medullosan and callistophytalean clades, subtending a crown group defined by the divergence of the cycads.

(2) Adding the peltasperm-corystosperm group to pteridosperms s.l. generates another wholly fossil paraphyletic group in which are embedded two monophyletic groups that contain extant taxa: the cycads, and the larger coniferophyte-glossophyte clade.

(3) The number of clades embedded in the paraphyletic pteridosperms could easily be reduced from two back to one by treating the cycads as a monophyletic group within pteridosperms (in other words, pteridosperms would no longer be viewed as a wholly extinct group).

(4) It would be possible to recognize pteridosperms sensu lato as a group that incorporated not only the core pteridosperms and peltasperms plus corystosperms but also glossopterids plus Pentoxylon, Bennettitales, and Caytonia.

The main disadvantage of option 4 is that is that it delimits an unequivocally polyphyletic group. The main advantage is that it incorporates all of the ambiguously placed fossil seed-plants, excluding only the well-defined monophyletic groups that include extant taxa. Although options 1-3 exclude fewer clades from pteridosperms, the arbitrary branches where pteridosperms are perceived to have evolved into one or more non-pteridospermous groups are short and weak. Ironically, although option 4 requires pteridosperms to be arbitrarily terminated at no fewer than four morphological branches, each of those branches is relatively long and well supported.

Any of these delimitation options renders pteridosperms extremely difficult to define on either an autapomorphy-based or a node-based system (assuming that a certain degree of robustness is expected of the relevant nodes). This is typical of the practical contortions necessary when one attempts to accommodate a non-monophyletic group in a fundamentally phylogenetic classification. It will be evident to any reader that a far more preferable solution would be to fragment and redistribute the erstwhile pteridosperm taxa, adding one or more pteridosperms to each of the better supported crown groups to which they are sisters and/or close relatives. Unfortunately, our knowledge of the most critical fossil seed plants remains inadequate to generate morphological trees of sufficient robustness to be confident in the relationships depicted (in other words, to generate the reliable apomorphies required to delimit the relevant clades). This problem, which will not be easily solved, is the subject of ongoing research.

LIKELIHOOD OF A CLOSE RELATIONSHIP BETWEEN CONIFERS AND GNETALES, AND OF A CONCOMITANTLY DISTANT RELATIONSHIP BETWEEN GNETALES AND ANGIOSPERMS. Numerous recent molecular phylogenetic analyses have proposed close relationships between extant conifers and extant Gnetales, generating the competing Gnepine, Gnetales-sister, and Gnetifer hypotheses (Fig. 2). These studies have generated much controversy, most frequently among proponents of each of the three topological hypotheses but also, on occasion, with authors who suspect that the reliability of these topologies is commonly being exaggerated. Concerns expressed include (a) the small number of taxa analyzed in order to focus on generating large numbers of molecular characters (see below), (b) the frequent combination of several different genie regions that obscures the often highly contradictory phylogenetic signals obtained from the individual regions, and (c) one possible explanation for (b); recent evidence that lateral gene transfer (often among highly phylogenetically disparate lineages) is commonplace in such phylogenetically critical taxa as Gnetum (Won and Renner 2003) and Amborella (Bergthorsson et al. 2004).

Admittedly, the molecular phylogenies have received further support from gene duplication studies. Sister relationships have been inferred between several orthologous MADS-box genes found in the gnetalean Gnetum on the one hand and in either Pinus or Picea on the other. Certainly, the C-class genes COM3 and DAL2, the AGL6-group genes GGM9 and DALl, and the B-class genes COM2 and PRDGL (papers by Frohlich, Theissen et al., Schneider et al. in Cronk et al. 2002) have provided strong, if circumstantial, evidence that most strongly favors the Gnepine hypothesis.

Discussions favoring one of these hypotheses over the other two have demonstrated that, as expected, these incongruent and/or weak topologies reflect the effects of contrasts in overall taxon sampling (Rydin and Kallersjo 2002, Rydin et al. 2005), outgroup sampling to determine rooting (Rydin and Kallersjo 2002, Rydin et al. 2002, 2005), and in the genomic compartments analyzed, which routinely generate conflicting signals (Rydin et al. 2002, 2005, Burleigh and Mathews 2004). Also critical is transition bias at third positions in codons (Frohlich, in press). Thus, there is as yet no molecular consensus regarding which of these three hypotheses should be preferred. However, consensus has arisen among most molecular researchers that Gnetales are, among extant taxa, either sister to, or nested within, conifers.

In the morphological analysis presented here, Gnetales are not directly related to conifers. In the major divergence between coniferophytes s.s. (less Ginkgoales, but including all conifers) and glossophytes (including angiosperms), Gnetales consistently occur (albeit without bootstrap support) on the glossophyte side of the major divide, being placed as basal to the remaining glossophytes (Fig. 4). Only in the topologically constrained morphological analysis, in which monophyletic angiosperms and monophyletic extant gymnosperms were artificially imposed on the matrix, were some of the resulting mostparsimonious trees consistent with a Gnetalessister result (albeit with the extinct Cheirolepidiaceae reliably nested within the extant conifers: Figs 8, 9). We will therefore briefly review those morphological characters that might suggest a close relationship between Gnetales and conifers (Fig. 15).

Within seed plants, Gnetales share with the extinct cordaitean coniferophyte Mesoxylon branching that is axillary with multiple buds, whereas all other extinct and extant seed plants have axillary branching with single buds (character 4: Fig. 15a). Metaxylem lacking scalariform tracheids (character 22) occur in Gnetales, Ginkgoales, and all extant conifers, but they do not characterize the extinct conifer Emporia, nor cordaitean coniferophytes (character 22: Fig. 15b). The presence of tori in xylem pits also unites Gnetales with extant conifers and Ginkgoales, but in the intervening extinct conifers and cordaitean coniferophytes this character is either absent (Shanxioxylon, Cordaixylon) or unknown (Mesoxylon, Thucydia, Emporia) (character 27, Appendix 2: Fig. 15c). This feature has only recently emerged as a phylogenetic character, and so it is scored as unknown in most fossil taxa; it merits further investigation. The final character potentially uniting conifers with Gnetales is proembryo tiering (character 98: Fig. 14d). This also could not yet be scored for most of the fossil taxa, as it requires detailed information on the ontogeny of the embryo; it too clearly merits more detailed investigation.

One crucial problem evident in previous morphological cladistic analyses and clearly identified here is the length of the branch that subtends the extant Gnetales (Fig. 12). These nine unambiguous character changes could in theory indicate a saltational macroevolutionary event (Bateman and DiMichele 1994a), but they represent several contrasting structures in the plant that are highly unlikely to have been developmentally correlated and so have been vulnerable to saltation. It seems more likely that this long branch reflects missing taxa that, if reconstructed, would allow us to infer the sequence of acquisition of these crucial character states. Although fossil Gnetales are now known from the Cretaceous (Rydin et al. 2004, Yang et al. 2005), the present reconstructions lack sufficient characters; their completeness is comparable with that of our weakest supplementary taxon, the fossil angiosperm Sinocarpus. Thus, inclusion of these fossil gnetaleans into morphological phylogenetic analyses is not yet sufficiently justified.

The corollary to the molecular nesting of Gnetales within coniferophytes is the extraordinary basal position within the seed-plant “crown group” frequently occupied by angiosperms (e.g., Rydin and Källersjö 2002) and the consequent collapse of the previously dominant Anthophyte hypothesis (but see Rothwell and Stockey 2002). In effect, the polarity of previous morphologically based trees has been partially reversed, a phenomenon often associated with uncertain rooting. In this context, our experiments with including various combinations of the three coded progymnosperms, and with selecting particular progymnosperms as the outgroup, are highly relevant. Most notably, including only one progymnosperm, Cecropsis, and treating it as the sole outgroup, re-rooted the typical topology such that angiosperms were the basalmost of the extant groups (Fig. 5e). Thus, we are inclined to view topologies (based on any kind of data) that show an early divergence of angiosperms from seedplants, preceding that of cycads, as merely demonstrating the vital importance of including sufficient well-chosen taxa to root the trees successfully.

We do not believe that adequate rooting is possible in analyses that use only extant taxa (including those based on DNA characters), where the outgroups must of necessity be drawn from ferns and/or lycopsids, which are separated from extant seed-plants by a vast phylogenetic discontinuity. We are in practice attempting to untangle a deep and relatively rapid radiation among the seed plants, and one that has left only highly selective representatives (most relatively derived) in the extant flora. This is a worst-case scenario for reconstructing phylogeny using sequence data (Bateman 1999). As concluded by Rydin and Källersjö (2002, p. 496) in their review of molecular evidence of seed-plant phylogeny, “we could easily obtain seemingly well-supported, yet conflicting, phylogenies simply by substituting a few terminals . . . Taxon sampling effects have been an underestimated source of phylogenetic error [and] if deep divergences are analyzed using restricted data sets, . . . reproducibility should be explored by conducting additional analyses in which taxa are exchanged . . . Considering the possible rooting problems of isolated groups, .. . putative seed plant stem groups such as the Paleozoic ‘seedferns’ and extinct lineages within the crown group-Bennettitales, cordaites, Mesozoic seed ferns, and additional representatives of Ginkgoales-are likely to provide important, and possibly indispensable, information on the evolutionary history of seed plants.”

We too are confident that further progress in understanding seed-plant relationships will depend heavily on gaining further knowledge of carefully targeted anatomically preserved fossils, most likely developed in tandem with reappraisal of homologies among extant seedplants via evolutionary-developmental genetic studies (e.g., Frohlich and Parker 2000). However, at least some of the much-cited evidence of seed-plant relationships emanating from evolutionary-developmental genetic studies is open to alternative explanations, as it primarily echoes the molecular phylogenetic results by emphasizing the distinctness of angiosperms relative to all other extant seed-plant groups (for example, the assumed loss of the NEEDLY copy of LEAFY from angiosperms: Frohlich and Parker 2000). The frequently found phylogenetic placement of angiosperms between supposedly monophyletic extant ferns and supposedly monophyletic extant gymnosperms will carry more credence if a significant number of genie duplications and/or rearrangements can be demonstrated, some of which unite angiosperms with all ferns but have been lost from all gymnosperms while others unite angiosperms with all gymnosperms but are absent from all ferns.

In the meantime, Gnetales and conifers are obvious priorities for phylogenetic studies integrating extinct and extant taxa (e.g., HernandezCastillo et al. 2005, Rydin et al. 2005), though much more could also be learned from relatively early diverging extant lineages, notably Ginkgo and cycads. In this context, recent studies of male morphogenesis in the cycad genus Zamia suggest derivation from a medullosan pteridosperm (Mundry and Stützel 2003), implying a Carboniferous origin for a lineage that on molecular evidence diverged after the angiosperms.

Overall, we are confident that resolving specific issues such as the close relationship (or, arguably, the lack of a close relationship) between Gnetales and conifers cannot be solved simply by isolating the problem within the chosen portion of the seed-plant phylogeny. Rather, taxa believed to be more distantly related may in practice prove to have a critical effect on topologies and homology assessments. We would therefore be surprised if pteridosperms did not continue to have a highly significant bearing on this ongoing, critical debate.

Conclusions. ( 1 ) Our morphological cladistic analysis of 48 core taxa (25 wholly extinct, the majority pteridosperms, and 23 at least partly extant) and 102 informative characters yielded 21 most-parsimonious trees via heuristic searches. The alternative topologies together generated two polytomies in the strict consensus tree, both occurring among taxa widely viewed as pteridosperms; the first affected several hydraspermans, and the second affected the three peltasperm/ corystosperm taxa analyzed.

(2) The resulting topology broadly resembled topologies generated in previous morphological cladistic analyses that combined substantial numbers of extant and extinct higher taxa. The closest comparison is with the analysis of Doyle (1996), whose matrix fomed the basis of the present study.

(3) Groups that include extant taxa (angiosperms, Gnetales, cycads, and, to a lesser extent, conifers) are relatively well resolved as monophyletic, aided by the relative completeness of their individual datasets. As summarized in Fig. 16, relationships inferred among the five extant groups are (cycads (Ginkgo (conifers (Gnetales (angiosperms))))), strongly contradicting most recent DNA-based topologies.

(4) However, the five groups containing extant taxa are embedded within the derived half of a morphologically diverse spectrum of extinct taxa, some or all of which can be assigned to the pteridosperms (Fig. 16). Although inclusion of these fossils weakens our perception of the robustness of the “crown groups”, it strongly influences the topology of the tree and elucidates patterns of acquisition of morphological characterstates. Thus, pteridosperms, and other fossil taxa derived from pteridosperms, are critical for understanding seed-plant relationships.

(5) Experiments with relatively poorly understood supplementary taxa suggest that large numbers of missing values can introduce local topological weaknesses, but that larger scale collapses in the strict consensus tend to reflect combinations of such taxa, or “wildcard” taxa that contain combinations of strongly phylogenetically informative characters indicating contradictory placements within the broader topology. We suspect that the 28% of ambiguous cells in our core matrix approaches the maximum that can be sustained in a matrix if it is also to contain sufficient structure to generate modest numbers of adequately resolved most-parsimonious trees.

(6) Experiments with including various combinations of the three coded progymnosperms, and with selecting particular progymnosperms as the outgroup, demonstrated that extreme topological instability could result. Most notably, including only one progymnosperm, Cecropsis, and treating it as the sole outgroup re-rooted the typical topology such that angiosperms were the basalmost of the extant groups. Admittedly, angiosperms have occupied this morphologically unintuitive position in several recent molecular trees. However, the Cecropsis-rooted morphological tree also reversed the sequence of divergence of most of the gymnospermous groups, generating a topology that entails a highly improbable sequence of character-state acquisitions and strongly contradicts first appearances of taxa observed in the fossil record.

(7) We therefore prefer to view topologies of any kind that show an early divergence of angiosperms from seed-plants, preceding that of cycads, as merely demonstrating the vital importance of including sufficient well-chosen taxa to root the trees successfully. We do not believe that this is possible in analyses that use only extant taxa (including those based on DNA characters), where the outgroups must of necessity be drawn from ferns and/or lycopsids, which are separated from extant seed plants by a vast phylogenetic discontinuity.

(8) Although pteridosperms cannot realistically be delimited as a monophyletic group, they remain a valuable informal category for the plexus of gymnosperms from which arose several more readily defined monophyletic groups of seed plants. The ideal solution of recognizing several monophyletic groups that incorporate both crown-groups and pteridosperms is not yet feasible, due to uncertainties of relationship and difficulties to satisfactorily delimiting the resulting groups.

(9) Several contrasting delimitations of pteridosperms are possible. These vary from a narrowly defined and wholly extinct paraphyletic group of hydraspermans plus medullosans plus callistophytaleans to a broadly defined polyphyletic group that also encompasses peltasperms plus corystosperms and glossopterids, Pentoxylon, Bennettitales, and Caytonia (Fig. 16). A case can also be made for including both extinct and extant cycads within pteridosperms. Each of the possible circumscriptions of the group readily identifies the base of the group but arbitrarily delimits the branch(es) where it is perceived to have transmogrified into one or more derived, monophyletic groups of non-pteridosperms.

(10) Our results do not support any of the recent, DNA-based hypotheses that place Gnetales as sister to, or embedded within, conifers. Rather, we have recovered a modified version of the Anthophyte hypothesis; groups successively diverging below angiosperms are Caytonia, Bennettitales, glossopterids plus Pentoxylon, and then Gnetales. Only when monophyly is artificially imposed on the gymnosperm “crown group” do Gnetales move to a position as sister to conifers (extant and extinct) plus cordaites; even then, they remain four nodes distant from the closest extant conifer group, Pinaceae plus Podocarpaceae.

(11) Exploration of the matrix organ by organ suggested that homoplasy levels were slightly higher in vegetative than in reproductive characters, and in large- and small-scale organs than in those of intermediate scale. Phylogenies constructed from data for individual organs such as ovules were poorly resolved, and the majority of the weakly supported groups that were recovered contradicted those generated from the entire matrix.

(12) Evidently, coding all of the organs of a plant (extinct or extant) is highly desirable and far more effective in any morphological cladistic analysis that addresses a broad taxonomic spectrum. In addition, some of the aggregate taxa coded in our analysis (e.g., corystosperms, glossopterids, Bennettitales) encompass an undesirably broad spectrum of variation, and so would benefit from division into multiple coded taxa. Together, these conclusions amply justify the substantial investment of time required to reconstruct individual conceptual whole plants from disarticulated fossil organs.

(13) It has become fashionable to couch morphological phylogenetic studies in the context of topologies generated from DNA-based data. Given the poverty of the hypotheses of relationship that can be addressed using only extant taxa, it could prove appropriate to apply the converse rationale. We suggest that the morphologybased trees could be treated as the initial phylogenetic framework that is subsequently tested using molecular tools. Rather than simply generating increasing numbers of often contradictory sequences from a few exemplar taxa, molecular tools may offer greater practical assistance to broad-brush reconstructions of seed-plant phylogeny by using an evolutionary-developmental genetic approach to explore specific cases of uncertain homologies presented by the morphological trees. Even then, our inability to explore the causes of morphological differentiation in key fossil groups such as pteridosperms (other than the presently under-researched cycads) will undoubtedly prove frustrating.

1 This work has been supported at various stages by the UK Natural Environment Research Council (Research Fellowship GT5/98/5/ES to JH), the Nuffield Foundation (NAL/00883/G to JH), and the Royal Society (2004 Conference Support to JH), all of which are gratefully acknowledged.

Literature Cited

ALBERT, V. A., A. BACKLUND, K. BREMER, M. W. CHASE, J. R. MANHART, B. D. MISHLER AND K. C. NIXON. 1994. Functional constraints and rbcL evidence for land plant phylogeny. Ann. Mo. Bot. Gard. 81: 543-567.

ALVIN, K. L., C. J. FRASER, AND R. A. SPICER. 1981. Anatomy and palaeoecology of Pseudofrenelopsis and associated conifers in the English Wealden. Palaeontology 24: 759-778.

AXSMITH, B. J., F. M. ANDREWS, AND N. C. FRASER. 2004a. The structure and phylogenetic significance of the conifer Pseudohirmerella delawarensis nov. comb, from the Upper Triassic of North America. Rev. Palaeobot. Palynol. 129: 251-263.

AXSMITH, B. J., M. KRINGS, AND K. WASELKOV. 2004b. Conifer pollen cones from the Cretaceous of Arkansas: implications for diversity and reproduction in the Cheirolepidiaceae. J. Palaeont. 78: 402-409.

AXSMITH, B. J., E. L. TAYLOR, T. N. TAYLOR, AND N. R. CUNEO. 2000. New perspectives on the Mesozoic seed fern order Corystospermales based on attached organs from the Triassic of Antarctica. Am. J. Bot. 87: 757-768.

BECKER, A. AND G. THEISSEN. 2003. The major clades of MADS-box genes and their role in the development and evolution of flowering plants. Mol. Phylog. Evol. 29: 464-489.

BATEMAN, R. M. 1988. Palaeobotany and palaeoenvironments of Lower Carboniferous floras from two volcanigenic terrains in the Scottish Midland Valley. Doctoral thesis, University of London, UK.

BATEMAN, R. M. 1992. Multiple hierarchies and plurality of species concepts: necessary evils in phylogenetically informative palaeobotanical classifications, pp. 21-22. In The Elusive Fossil Species (symposium). Fourth International Organisation of Palaeobotany Conference (Paris) Abstracts (OFP Information Special Volume 16B).

BATEMAN, R. M. 1994. Evolutionary-developmental change in the growth architecture of fossil rhizomorphic lycopsids: scenarios constructed on cladistic foundations. Biol. Rev. 69: 527-597.

BATEMAN, R. M. 1999. Integrating molecular and morphological evidence for evolutionary radiations, pp. 432-471. In P. M. Hollingsworth, R. M. Bateman, and R. J. Gornall [eds.], Molecular systematics and plant evolution. Taylor & Francis, London, UK.

BATEMAN, R. M. AND W. A. DIMICHELE. 1994a. SaItational evolution of form in vascular plants: a neoGoldschmidtian synthesis, pp. 61-100. In D. S. Ingram and A. Hudson [eds.], Shape and form in plants and fungi. Linnean Society, London, UK.

BATEMAN, R. M. AND W. A. DIMICHELE. 1994b. Heterospory: the most iterative key innovation in the evolutionary history of the plant kingdom. Biol. Rev. 69: 345-417.

BATEMAN, R. M. AND W. A. DIMICHELE. 2002. Generating and filtering major phenotypic novelties: neoGoldschmidtian saltation revisited, pp. 109159. In Q. C. B. Cronk, R. M. Bateman, and J. A. Hawkins [eds.], Developmental genetics and plant evolution. Taylor & Francis, London, UK.

BATEMAN, R. M. AND W. A. DIMICHELE. 2003. Genesis of phenotypic and genotypic diversity in land plants: the present as the key to the past. Syst. Biodiv. 1: 13-28.

BATEMAN, R. M. AND G. W. ROTHWELL. 1990. A reappraisal of the Dinantian floras at Oxroad Bay, East Lothian, Scotland. 1. Floristics and the development of whole-plant concepts. Trans Roy. Soc. Edin. B 81: 127-159.

BATEMAN, R. M. AND N. J. SIMPSON. 1998. Comparing phylogenetic signals from reproductive and vegetative organs, pp. 231-253. In S. 3. Owens and P. J. Rudall [eds.], Reproductive biology. Royal Botanic Gardens, Kew, UK.

BAUCH, J., W. lieSE, AND R. SCHULTZE. 1972. The morphological variability of the bordered pit membranes in gymnosperms. Wood Sci. Technol. 6: 165-184.

BECK, C. B. 1957. Tetraxylopteris schmidtii gen. et. sp. nov., a probable pteridosperm precursor from the Devonian of New York. Am. J. Bot. 44: 350367.

BECK, C. B. 1970. The appearance of the gymnospermous structure. Biol. Rev. 45: 379-400.

BECK, C. B. 1971. On the anatomy and morphology of lateral branch systems of Archaeopteris. Am. J. Bot. 58: 758-784.

BECK, C. B. 1976. Current status of the Progymnospermopsida. Rev. Palaeobot. Palynol. 21: 5-23.

BECK, C. B. 1979. The primary vascular system of Callixylon. Rev. Palaeobot. Palynol. 28: 103-115.

BECK, C. B. 1981. Archaeopteris and its role in vascular plant evolution, pp. 193-230. In K. J. Niklas [ed.], Paleobotany, Paleoecology, and Evolution, vol. 1. Praeger, New York, USA.

BECK, C. B. AND D. C. WIGHT. 1988. Progymnosperms, pp. 1-84 In C. B. Beck [ed.], Origin and evolution of gymnosperms. Columbia University Press, New York, NY.

BENSON, M. 1908. On the contents of the pollen chamber of a specimen of Lagenosloma ovoides. Bot. Gaz. 45: 409-412.

BENTON, M. J. 2000. Stems, nodes, crown clades and rank-free lists: is Linnaeus dead? Biol. Rev. 75: 633-648.

BERGTHORSSON, U., A. O. RICHARDSON, G. J. YOUNG, L. R. GOERTZEN AND J. D. PALMER. 2004. Massive horizontal transfer of mitochondrial genes from diverse land plant donors to the basal angiosperm Amborella. Proc. Nat. Acad. Sci. 101: 17747-17752.

BERTRAND, C. E. 1907. Les caractéristiques du genre Taxospermum de Brongniart. Bull. Soc. Bot. France 7: 1-213.

BERTRAND, C. E. AND B. RENAULT. 1886. Remarques sur le Paroxyton stephanense. C. R. Acad. Sci. Paris 103: 765-767.

BONAMO, P. M. AND H. P. BANKS. 1967. Tetraxylopteris schmidtii: its fertile parts and its relationships within the Aneurophytales. Am. J. Bot. 54: 755-768.

BOWE, L. M., M. G. COAT, AND C. W. DE PAMPHILIS. 2000. Phylogeny of seed plants based on all three genomic compartments: extant gymnosperms are monophyletic and Gnetales’ closest relatives. Proc. Nat. Acad. Sci. 97: 4092-4097.

BREMER, K. 1988. The limits of amino acid sequence data in angiosperm phylogenetic reconstruction. Evolution 42: 759-803.

BREMER, K. 1990. Combinable component consensus. Cladistics 6: 369-372.

BURLEIGH, J. G. AND S. MATHEWS. 2004. Phylogenetic signal in nucleotide data from seed plants: implications for resolving the seed plant tree of life. Am. J. Bot. 91: 1599-1613.

BUZGO, M., P. S. SOLTIS, AND D. E. SOLTIS. 2004. Floral developmental morphology of Amborella trichopoda (Amborellaceae). Int. J. Plant Sci. 165: 925-947.

CALDER, M. G. 1938. On some undescribed species from the Lower Carboniferous flora of Berwickshire, together with a note on the genus Stenomyelon Kidston. Trans Roy. Soc. Edin. 59: 309-331.

CANTINO, P. D. AND K. DE QUEIROZ. 2000. PhyloCode. Retreived June 2005 from Ohio University.

CARLQUIST, S. 1996a. Wood, bark, and stem anatomy of Gnetales: a summary. Int. J. Plant Sci. 157(6 Suppl.): S58-S76.

CARLQUIST, S. 1996b. Wood anatomy of primitive angiosperms: new perspectives and syntheses, pp. 6890. In D. W. Taylor and L. J. Hickey [eds.], Flowering plant origin, evolution, and phylogeny. Chapman & Hall, New York, NY.

CHALONER W. G. 1986. Reassembling the whole plant, and naming it, pp. 67-78. In R. A. Spicer and B. A. Thomas [eds.], Systematic and taxonomic approaches in palaeobotany. Systematics Association Special Volume No. 31, Clarendon Press, Oxford, UK.

CHAMBERLAIN, C. J. 1913. Macrozamia moorei, a connecting link between living and fossil cycads. Bot. Gaz. 55: 141-154.

CHAW, S. M., C. L. PARKINSON, Y. CHENG, T. M. VINCENT, AND J. D. PALMER. 2000. seed plant phylogeny inferred from all three genomes: monophyly of extant gymnosperms and origin of Gnetales from conifers. Proc. Nat. Acad. Sci. 97: 4086-4091.

CHAW, S. M., A. ZHARKIKH, H. M. SUNG, T. C. LAU AND W. H. Li. 1997. Molecular phylogeny of extant gymnosperms and seed plant evolution: analysis of nuclear 18S rRNA sequences. MoI. Biol. Evol. 14: 56-68.

COSTANZA, S. H. 1985. Pennsylvanioxylon of Middle and Upper Pennsylvanian coals from the Illinois basin and its comparison with Mesoxylon. Palaeontographica B 197: 81-121.

CRANE, P. R. 1985. Phylogenetic analysis of seedplants and the origins of the angiosperms. Ann. Mo. Bot. Gard. 72: 716-793.

CRANE, P R. 1988. Major clades and relationships in the higher gymnosperms, pp. 218-272. In C. B. Beck [ed.]. Origin and evolution of gymnosperms. Columbia University Press, New York, NY.

CRANE, P. R., P. HERENDEEN, AND E. M. FRIIS. 2004. Fossils and plant phylogeny. Am. J. Bot. 91: 1683-1699.

CRONK, Q. C. B., R. M. BATEMAN, AND J. A. HAWKINS. 2002. Developmental genetics and plant evolution. Taylor & Francis, London. 543 pp.

DELEVORYAS, T. AND J. MORGAN. 1954. A new pteridosperm from Upper Pennsylvanian deposits of North America. Palaeontographica B 96: 12-23.

DIMICHELE, W. A., T. L. PHILLIPS, AND H. PFEFFERKORN. 2006. Paleoecology of Late Paleozoic pteridosperms from tropical Euramerica. J. Torrey Bot. Soc. 133: 83-118.

DOYLE, J. A. 1996. seed plant phylogeny and the relationships of Gnetales. Int. J. Plant Sci. 157(suppl.): S3-S39.

DOYLE, J. A. 1998. Molecules, morphology, fossils, and the relationships of angiosperms and Gnetales. MoI. Phylog. Evol. 9: 448-462.

DOYLE, J. A. 2004. seed ferns and the origin of angiosperms. Abstract 81, Botanical Society of America Botany 2004 conference. Snowbird, USA.

DOYLE, J. A. 2006. seed ferns and the origin of angiosperms. J. Torrey Bot. Soc. 133: 169-209.

DOYLE, J. A. AND M. J. DONOGHUE. 1986. seed plant phylogeny and the origin of angiosperms: an experimental cladistic approach. Bot. Rev. 52: 321-431.

DOYLE, J. A. AND M. J. DONOGHUE. 1987. The importance of fossils in elucidating seed plant phylogeny and macroevolution. Rev. Palaeobot. Palynol. 50: 63-95.

DOYLE, J. A. AND M. J. DONOGHUE. 1992. Fossils and seed plant phylogeny reanalyzed. Brittonia 44: 89-106.

DOYLE, J. A. AND P. K. ENDRESS. 2000. Morphological phylogenetic analysis of basal angiosperms: comparisons and combination with molecular data. Int. J. Plant Sci. 161(suppl.): S121-S153.

DUNN, M. D., G. W. ROTHWELL, AND G. MAPES. 2002. Additional observations on Rhynchosperma quinii (Medullosaceae): a permineralized ovule from the Chesterian (Upper Mississipian) Fayetteville Formation of Arkansas. Am. J. Bot. 89: 1799-1808.

EGGERT, D. A. AND T. DELEVORYAS. 1960. Callospermarion: a new seed genus from the Upper Pennsylvanian of Illinois. Phytomorphology 10: 131-138.

FAIRON-DEMARET, M. AND S. E. SCHECKLER. 1987. Typification and redescription of Moresnetia zalesskyi Stockmans, 1948, an early seed plant from the Upper Famennian of Belgium. Bull. L’Institut R. Sciences Naturelle de Belgique 57: 183-199.

FOREY, P. L. 2002. PhyloCode: no pain, no gain. Taxon 51: 43-54.

FRIIS, E. M., J. A. DOYLE, P. K. ENDRESS, AND Q. LENG. 2003. Archaefructus: angiosperm precursor or specialised early angiosperm. Trends Plant Sci. 8: 369-373.

FROLICH, M. E IN PRESS. Recent developments regarding the evolutionary origin of flowers. In D. E. Soltis, J. Leebens-Mack and P. S. Soltis [eds.], Developmental genetics of the flower. Elsevier, Amsterdam.

FROHLICH, M. F. AND D. S. PARKER. 2000. The mostly male theory of flower evolutionary origins. Syst. Bot. 25: 155-170.

GALTIER, J. 1988. Morphology and phylogenetic relationships of early pteridosperms, pp. 135-176. In C. B. Beck [ed.], Origin and evolution of gymnosperms. Columbia University Press, New York, NY.

GALTIER, J. 1991. The Late Carboniferous cupulate seed Gnetopsis eliptica Renault, pp. 351-357. In J. Kovar-Eder [ed.], Palaeovegetational development in Europe and regions relevant to its palaeofloristic evolution. Proceedings of the Pan-European Palaeobotanical Conference, Vienna, Austria.

GALTIER, J. 1992. On the earliest arborescent gymnosperms. Cour. Forsch.-Inst. Senckenberg 147: 119-125.

GALTIER, J. 2004. Arborescent seed ferns of the Pitus type. Abstracts, Botany 2004, Snowbird, Utah.

GALTIER, J. 2005. The diversification of early arborescent seed ferns. J. Torrey Bot. Soc. 133: 7-19.

GALTIER, J. AND BETHOUX, O. 2002. Morphology and growth habit of Dicksonites pluckenetii from the Upper Carboniferous of Graissessac (France). Geobios 35: 525-535.

GIVNISH, T. J. AND SYTSMA, K. J. 1997. Consistency, characters, and the likelihood of correct phylogenetic inference. MoI. Phylog. Evol. 7: 320-330.

GORDON, W. T. 1941. On Salpingostoma dasu: a new Carboniferous seed from East Lothian. Trans. Roy. Soc. Edin. 60: 427-464.

GOREMYKIN, V., V. BOBROVA, J. PAHNKE, A. TROITSKY, A. ANTONOV AND W. MARTIN. 1996. Noncoding sequences from the slowly evolving chloroplast inverted repeat in addition to rbcL data do not support Gnetalean affinities of angiosperms. MoI. Biol. Evol. 13: 383-396.

GOULD, R. E. AND T. DELEVORYAS. 1977. The biology of Glossopteris: evidence from petrified seed-bearing and pollen-bearing organs. Alcheringa 1: 387-399.

GUGERLI, E, C. SPERISEN, U. BÜCKLER, I. BRUNNER, S. BRODBECK, J. D. PALMER AND Y.-L. QIU. 2001. The evolutionary split of Pinaceae from other conifers: evidence from an intron loss and multigene phylogeny. Mol. Phylog. Evol. 21: 167-175.

HALLE, T. G. 1927. Palaeozoic plants from central Shansi. Palaeontologia Sinica Series A (Geological Survey of China, Peking) 2: 1-316.

HALLE, T. G. 1929. Some seed-bearing pteridosperms from the Permian of China. Kungl Svenska Vetenskapsakademiens Handlingar, Tredje Serien, 6: 3-24.

HASEBE, M., R. KOFUKI, M. ITO, M. KATO, K. IWATSUKI AND K. UEDA. 1992. Phylogeny of gymnosperms inferred from rbcL gene sequences. Bot. Mag., Tokyo, 105: 673-697.

HAMBY, R. K. AND E. A. ZIMMER. 1992. Ribosomal RNA as a phylogenetic tool in plant systematics, pp. 50-91. In P. S. Soltis, D. E. Soltis, and J. J. Doyle [eds.], Molecular systematics of plants. Chapman and Hall, New York, NY.

HAMMOND, S. E. 2004. Progymnosperms and the origin of the seed. Unpublished doctoral thesis, Cardiff University, UK.

HAMMOND, S. E. AND C. M. BERRY. 2005. A new species of Tetraxylopteris (Aneurophytales) from the Devonian of Venezuela. Bot. J. Linn. Soc. 148: 275-303.

HARRIS, T. M. 1937. The fossil flora of Scoresby Sound, east Greenland, Part 5. Stratigraphie relationships of the plant beds. Meddel. Gr0nland 113: 1-112.

HARRIS, T. M. 1964. The Yorkshire Jurassic flora. II. Caytoniales, Cycadales and Pteridosperms. British Museum (Natural History), London, UK.

HEMSLEY, A. R. 1993. A review of Palaeozoic seedmegaspores. Palaeontographica B 229: 135-166.

HERNANDEZ-CASTILLO, G. R., G. W. ROTHWELL, AND G. MAPES. 2001a. Thucydiaceae fam. nov., with a review and reevaluation of Paleozoic walchian conifers. Int. J. Plant Sci. 162: 1155-1158.

HERNANDEZ-CASTILLO, G. R., G. W. ROTHWELL, AND G. MAPES. 2001b. Compound pollen cones in Paleozoic conifers. Am. J. Bot. 88: 1139-1142.

HERNANDEZ-CASTILLO, G. R., G. W. ROTHWELL, R. A. STOCKEY, AND G. MAPES. 2003. Growth architecture of Thucydia mahoningensis, a model for primitive walchian conifer plants. Int. J. Plant Sci. 164: 443-452.

HERNANDEZ-CASTILLO, G. R., R. A. STOCKEY, G. W. ROTHWELL, AND G. MAPES. 2005. Systematics of the most ancient conifers. Abstracts 17th International Botanical Congress, Vienna. Abstract 4.9.4, p. 68.

HILL, C. R. AND P. R. CRANE. 1982. Evolutionary cladistics and the origin of angiosperme, pp. 269-361. In K. A. Joysey and A. E. Friday [eds.], Problems of phylogenetic reconstruction. Academic Press, New York, NY.

HILL, K. D., M. W. CHASE, D. W. STEVENSON, H. G. HILLS, AND B. SCHUTZMAN. 2003. The families and genera of cycads: a molecular phylogenetic analysis of Cycadophyta based on nuclear and plastid DNA sequences. Int. J. Plant Sci. 146: 933-948.

HILTON, J. AND R. M. BATEMAN. 2005. Reassessing seed ferns in seed plant systematics, evolution and phylogeny. XVII International Botanical Congress, Vienna. Abstract 4.9.3, p. 68.

HILTON J. AND D. EDWARDS. 1999. New data on Xenotheca devonica (Arber and Goode), an enigmatic early seed-plant cupule bearing preovules, pp. 75-90. In M. H. Kurmann and A. R. Hemsley [eds.], The evolution of plant architecture. Royal Botanic Gardens, Kew, UK.

HILTON, J. AND C. S. LI. 2003. Reinvestigation of Nyroestroemia pectinformis Halle, an enigmatic seed plant from the Late Permian of China. Palaeontology 46: 29-51.

HILTON, J., S. J. WANG, AND B. TIAN. 2003. Reinvestigation of Cardiocarpus minor (Wang) Li from the Early Permian Taiyuan Formation of northern China, and an evaluation of cardiocarpalean ovule taxonomy, systematics and phylogeny. Bot. J. Linn. Soc. 141: 151-175.

HILTON, J., S. J. WANG, W. Q. ZHU, B. TIAN, J. GALTIER, AND A. H. WEI. 2002. Callospermarion ovules from the Early Permian of northern China: palaeofloristic and palaeogeographic significance of callistophytalean seed-ferns in the Cathaysian flora. Rev. Palaeobot. Palynol. 120: 301-314.

JANSEN, S., B. CHOAT, S. VINCKIER, P. SCHOLS, AND E. SMETS. 2004a. Intervascular pit membranes with a torus in the wood of Ulmus (Ulmaceae) and related genera. New Phytol. 163: 51-59.

JANSEN, S., R. R. DUTE, P. GASSON, AND E. SMETS. 2004b. A preliminary survey of angiosperms with torus-bearing pit membranes. Abstract, Botany 2004, Snowbird, Utah.

JENNINGS, J. R. 1976. The morphology and relationships of Rhodea, Telangium, Telangiopsis, and Heterangium. Am. J. Bot. 63: 1119-1133.

JONGMANS, W. L. 1930. On the fructification of Sphenopteris hoeinghausii, and its relations with Lyginodendron olhamia and Crossotheca schalzlarensis. Jaarverslag: Geologish bureau voor net Nederlandsche mijnigebied, Heerlen, Netherlands.

KERP, J. H. A. 1988. Aspects of Permian palaeobotany and palynology. X. The West and Central European species of the genus Autunia Krasser emend. Kerp (Peltaspermaceae) and the form genus Rachiphyllum Kerp (callipterid foliage). Rev. Palaeobot. Palynol. 54: 249-360.

KIDSTON, R. 1887. On the fructifications of some ferns from the Carboniferous Formations. Trans. Roy. Soc. Edin. 33: 137-156.

KITCHING, I. J., P. L. FOREY, C. J. HUMPHRIES, D. J. SIEBERT, AND D. M. WILLIAMS. 1999. Cladistics: a practical course in systematics (2nd edn). Oxford University Press, Oxford, UK.

KLAVINS, S. D. AND L. C. MATTEN. 1996. Reconstruction of the frond of Laceya hibernica, a lyginopterid pteridosperm from the uppermost Devonian of Ireland. Rev. Palaeobot. Palynol. 93: 253-268.

KLAVINS, S. D., T. N. TAYLOR, AND E. L. TAYLOR. 2002. Anatomy of Umkomasia (Corystospermales) from the Triassic of Antarctica. Am. J. Bot. 89: 664-676.

LENG, Q. AND E. M. FRIIS. 2003. Sinocarpus decussatus gen. et sp. nov., a new angiosperm with basally syncarpous fruits from the Yixian Formation of northwest China. Plant Syst. Evol. 241: 77-88.

LOCONTE, H. AND D. W. STEVENSON. 1990. Cladistics of Spermatophyta, Brittonia 42: 197-211.

LONG, A. G. 1944. On the prothallus of Lagenstoma ovoides Will. Ann. Bot. 8: 105-120.

LONG, A. G. 196Oa. On the structure of Samaropsis scotica Calder (emended) and Eurystoma angulare gen. et sp. nov., petrified seeds from the Calciferous Sandstone Series of Berwickshire. Trans. Roy. Soc. Edin. 64: 261-284.

LONG, A. G. 1960b. On the structure of Calymmatotheca kidstoni (emended) and Genomosperma latens gen. et sp. nov. from the Calciferous Sandstone Series of Berwickshire. Trans. Roy. Soc. Edin. 64: 29-48.

LONG, A. G. 1960c. Stamnostoma huttonense gen. et sp. nov.: a pteridosperm seed and cupule from the Calciferous Sandstone Series of Berwickshire. Trans. Roy. Soc. Edin. 64: 201-217.

LONG, A. G. 1961. Some pteridosperm seeds from the Calciferous Sandstone Series of Berwickshire. Trans. Roy. Soc. Edin. 64: 401-419.

LONG, A. G. 1963. Some specimens of Lyginorachis papilio Kidston associated with stems of Pitys. Trans. Roy. Soc. Edin. 65: 211-224.

LONG, A. G. 1964. Some specimens of “Stenomyelon” and “Kalymma” from the Calciferous Sandstone Series of Berwickshire. Trans. Roy. Soc. Edin. 65: 435-446.

LONG, A. G. 1975. Further observations on some Lower Carboniferous seeds and cupules. Trans. Roy. Soc. Edin. 69: 278-293.

LONG, A. G. 1977a. Some Lower Carboniferous pteridosperm cupules bearing ovules and microsporangia. Trans. Roy. Soc. Edin. 70: 1-11.

LONG, A. G. 1977b. Observations on the Carboniferous seeds of Mitrospermum, Conostoma and Lagenostoma. Trans. Roy. Soc. Edin. 70: 37-61.

LONG, A. G. 1979. Observations on the Lower Carboniferous genus Pitus Witham. Trans. Roy. Soc. Edin. 70: 111-127.

MADDISON, D. 1991. The discovery and importance of multiple islands of most-parsimonious trees. Syst. Zool. 40: 315-328.

MADDISON, D. AND W. MADDISON. 2000. MacClade 4: analysis of phytogeny and character evolution. Sinauer Associates, Sunderland, MA.

MAGALLÓN, S. AND M. J. SANDERSON. 2002. Relationships among seed plants inferred from highly conservative genes: sorting conflicting phylogenetic signals among ancient lineages. Am. J. Bot. 89: 1991-2006.

MAPES, G. AND G. W. ROTHWELL. 1980. Quaestora amplecta gen. et sp. n., a structurally simple medullosan stem from the Upper Mississippian of Arkansas. Am. J. Bot. 65: 636-647.

MAPES, G. AND G. W. ROTHWELL. 1984. Permineralized ovulate cones of Lebachia from Late Palaeozoic Limestones of Kansas. Palaeontology 27: 69-94.

MAPES, G. AND G. W. ROTHWELL. 1988. Diversity amongst Hamilton conifers, pp. 225-244. In G. Mapes and R. H. Mapes [eds.], Regional geology and paleontology of the Upper Paleozoic Hamilton Quarry area in south-eastern Kansas. Guidebook no 6, Kansas Geological Survey, Lawrence, KS.

MAPES, G. AND G. W. ROTHWELL. 1991. Structural relationships of primitive conifers. Neues Jahrb. Geol. Paläont. Abh. 183: 269-287.

MASSELTER, T., T. SPECK, AND N. P. ROWE. 2005. Evolution of mechanical innovations and self repair in climbing plants. Abstracts 17th International Botanical Congress, Vienna, Austria. Abstract 5.4.3, p. 81.

MATHEWS, S. AND M. J. DONOGHUE. 2002. Analyses of phytochrome data from seed plants: exploration of conflicting results from parsimony and Bayesian approaches. «http://www.2002.botanyconference. org/section 12/abstracts/238.shtmt«

MATTEN, L. C. 1992. Reconstruction ofLaceya, an Upper Devonian seed plant, pp. 105. In Abstracts of the Fourth International Organisation of Palaeobotany Conference, Paris 1992.

MATTEN, L. C. AND T. FINE. 1984. Telangium schweitzeri sp. nov.: a gymnospermous pollen organ from the Upper Devonian of Ireland. Palaeontographica B 232: 15-33.

MATTEN, L. C., T. FINE, W. R. TANNER, AND W. S. LACEY. 1984. The megagametophyte of Hydrasperma tenuis Long from the Upper Devonian of Ireland. Am. J. Bot. 71: 1461-1464.

MATTEN, L. C., W. S. LACEY, AND D. EDWARDS. 1975. Discovery of one of the oldest gymnosperm floras containing cupulate seeds. Phytologia 32: 299-303.

MATTEN, L. C., W. S. LACEY AND R. C. LUCAS. 1980. Studies on the cupulate seed genus Hydrasperma Long from Berwickshire and East Lothian in Scotland and County Kerry in Ireland. Bot. J. Linn. Soc. 81: 249-273.

MAY, B. I. AND L. C. MATTEN. 1983. A probable pteridospemn from the uppermost Devonian near BaIlyheigue, Co. Kerry, Ireland. Bot. J. Linn. Soc. 86: 103-123.

MEYEN, S. V. 1988. Gymnosperms of the Angara flora, pp. 338-381. In C. B. Beck [ed.], Origins and evolution of the gymnosperms. Columbia University Press, New York, NY.

MEYER-BERTHAUD, B. AND W. E. STEIN. 1995. A reinvestigation of Stenomyelon from the Late Tournaisian of Scotland. Int. J. Plant Sci. 156: 863-895.

MEYER-BERTHAUD, B., T. N. TAYLOR, AND E. L. TAYLOR. 1993. Petrified stems bearing Dicroidium leaves from the Triassic of Antarctica. Palaeontology 36: 337-356.

MUNDRY, M. AND T. STÜTZEL. 2003. Morphogenesis of male sporangiophores of Zamia ambyphyllidia D. W. Stev. Plant Biol. 5: 297-310.

MUNDRY, M. AND T. STÜTZEL. 2004. Morphogenesis of the reproductive shoots of Welwitschia mirabilis and Ephedra distachya (Gnetales), and its evolutionary importance. Organ. Divers. Evol. 4: 91108.

NICKRENT, D. L., C. L. PARKINSON, J. D. PALMER, AND R. J. DUFF, 2000. Multigene phylogeny of land plants with special reference to bryophytes and the earliest land plants. MoI. Biol. Evol. 17: 18851895.

NISHIDA, H., K. B. PIGG, AND J. F. RIOBY. 2003. Swimming sperm in an extinct Gondwanan plant. Nature 422: 396-397.

NIXON, K. C. AND J. M. CARPENTER. 2000. On the other ‘phylogenetic systematics.’ CIadistics 16: 298-318.

NIXON, K. C., W. L. CREPET, D. W. STEVENSON, AND E. M. FRIIS. 1994. A reevaluation of seed plant phylogeny. Ann. Mo. Bot. Gard. 81: 484-533.

OLIVER, F. W. 1903. The ovules of the older gymnosperms. Ann. Bot. 17: 451-476.

OLIVER, F. W. AND E. J. SALISBURY. 1911. On the structure and affinities of the Palaeozoic seeds of the Conostoma group. Ann. Bot. 25: 2-48.

OLIVER, F W. AND D. H. SCOTT. 1904. On the structure of the Palaeozoic seed Lagenostoma lomaxi, with a statement of the evidence upon which it is referred to Lyginodendron. Phil. Trans. R. Soc. Lond. 219: 193-247.

PÄÄBO, S., H. POINAR, D. SERRE, V. JAENICKE-DESPRES, J. HEBLER, N. ROHLAND, M. KUCH, J. KRAUSE, L. VIGILANT, AND M. HOFREITER. 2004. Genetic analyses from ancient DNA. Ann. Rev. Gen. 38: 645-679.

PAGE, R. D. M. 2001. Nexus Data Editor. «http://»

POORT, R. J. AND H. J. A. KERP. 1990. Aspects of Permian palaeobotany and palynology. XI. On the recognition of true Peltasperms in the Upper Permian of western and central Europe and a reclassification of the species formerly included in Peltaspermum Harris. Rev. Palaeobot. Palynol. 63: 197-225.

QIU Y. L., J. LEE, F. BERNASCONI-QUADRONI, D. E. SOLTIS, P. S. SOLTIS, M. ZANIS, E. A. ZIMMER, Z. CHEN, V. SAVOLAINEN AND M. W. CHASE. 1999. The earliest angiosperms: evidence from mitochondrial, plastid and nuclear genomes. Nature 402: 404-407.

RAI, H. S., H. E. O’BRiEN, P. A. REEVES, R. G. OLMSTEAD, AND S. G. GRAHAM. 2003. Inference of higher-order relationships in the cycads from a large chloroplast data set. MoI. Phylog. Evol. 29: 350-359.

RENAULT, B. 1885. Cours de Botanique fossile, 4: 1-232. Paris, France.

RETALLACK, G. J. AND D. L. DILCHER. 1988. Reconstructions of selected seed ferns. Ann. Mo. Bot. Gard. 75: 1010-1057.

ROTHWELL, G. W. 1970. Ontogeny of the Paleozoic ovule, Callospermarion pusillum. Am. J. Bot. 58: 706-715.

ROTHWELL, G. W. 1971. Additional observations on Conostoma anglo-germanicum and C. oblongum from the Lower Pennsylvanian of North America. Palaeontographica B 131: 167-178.

ROTHWELL, G. W. 1975. The Callistophytaceae (Pteridospermopsida). I. Vegetative structures. Palaeontographica B 151: 171-196.

ROTHWELL, G. W. 1980. The Callistophytaceae (Pteridospermopsida). II. Reproductive features. Palaeontographica B 173: 85-106.

ROTHWELL, G. W. 1981. The Callistophytales (Pteridospermopsida): reproductively sophisticated Paleozoic gymnosperms. Rev. Palaebot. Palynol. 32: 103-121.

ROTHWELL, G. W. 1982. New interpretations of the earliest conifers. Rev. Palaeobot. Palynol. 37: 7-28.

ROTHWELL, G. W. 1986. Classifying the earliest gymnosperms, pp. 137-162. In R. A. Spicer and B. A. Thomas [eds.], Systematic and taxonomic approaches in palaeobotany. Systematics Association Special Volume No. 31, Clarendon Press, Oxford, UK.

ROTHWELL, G. W. 1988. Cordiales, pp. 273-297 In C. B. Beck [ed.], Origin and evolution of Gymnosperms. Columbia University Press, New York, NY.

ROTHWELL G. W. 1993. Cordaixylon dumusum (Cordaitales). II. Reproductive biology, phenology, and growth ecology. Int. J. Plant Sci. 154: 572-586.

ROTHWELL, G. W. 2006. To “the upward outlook” of Wilson Nichols Stewart. J. Torrey Bot. Soc. 133: 1-3.

ROTHWELL, G. W. AND D. ERWIN. 1987. Origin of seed plants: an aneurophyte/seed fern link elaborated. Am. J. Bot. 74: 970-973.

ROTHWELL, G. W. AND G. MAPES. 2001. Barthelia furcata gen. et sp. nov., with a review of Paleozoic coniferophytes and a discussion of coniferophyte systematics. Int. J. Plant Sci. 162: 637-667.

ROTHWELL, G. W., G. MAPES, J. HILTON, AND N. T. HOLLINGWORTH. In press. Anatomy of cheirolepidiaceous pollen cones: Classostrobus crossii sp. nov. Int. J. Coal Geol.

ROTHWELL, G. W., G. MAPES, AND R. H. MAPES. 1996. Anatomically preserved vojnovskyalean seed plants in Upper Pennsylvanian (Stephanian) marine shales of North America. J. Paleont. 70: 1067-1079.

ROTHWELL, G. W. AND S. E. SCHECKLER. 1988. Biology of ancestral gymnosperms, pp. 85-134. In C. B. Beck [ed.], Origins and evolution of the gymnosperms. Columbia University Press, New York, NY.

ROTHWELL, G. W., S. E. SCHECKLER, AND W. H. GILLESPIE. 1989. Elkinsia gen. nov., a Late Devonian gymnosperm with cupulate ovules. Bot. Gaz. 150: 170-189.

ROTHWELL, G. W. AND R. SERBET. 1992. Pollination biology of Elkinsia polymorpha: implications for the origin of the gymnosperms. Cour. Forsch.-Inst. Senckenburg 147: 225-231.

ROTHWELL, G. W. AND R. SERBET. 1994. Lignophyte phylogeny and the evolution of spermatophytes: a numerical cladistic analysis. Syst. Bot. 19: 443-482.

ROTHWELL, G. W. AND STOCKEY, R. A. 2002. Anatomically preserved Cycadeoidea (Cycadeoidaceae), with a reevaluation of the systematic characters for the seed cones of Bennettitales. Am. J. Bot. 89: 1447-1458.

ROTHWELL, G. W. AND D. C. WIGHT. 1989. Pullaritheca longii gen. nov. and Kerryia mattenii gen. et sp. nov., Lower Carboniferous cupules with ovules of the Hydrasperma tenuis-iype. Rev. Palaeobot. Palynol. 60: 295-309.

RYDIN, C. AND M. KÄLLERSJÖ. 2002. Taxon sampling and seed plant phylogeny. Cladistics 18: 484-513.

RYDIN, C., M. KÄLLERSJÖ, AND E. M. FRIIS. 2002. seed plant relationships and the systematic position of Gnetales based on nuclear and chloroplast DNA: conflicting data, rooting problems, and the monophyly of conifers. Int. J. Plant Sci. 163: 197-214.

RYDIN, C., K. R. PEDERSEN, AND E. M. FRIIS. 2004. On the evolutionary history of Ephedra: Cretaceous fossils and extant molecules. Proc. Nat. Acad. Sci. 101: 16571-16576.

RYDIN, C., K. R. PEDERSEN, E. M. FRIIS, AND R R. CRANE. 2005. The Gnetales: fossils and phylogenies. Abstracts 17th International Botanical Congress, Vienna, Austria. Abstract 4.9.2. p. 68.

SAMIGULLIN, T. KH., W. F. MARTIN, A. V. TROITSKY, AND A. S. ANTONOV. 1999. Molecular data from the chloroplast rpoCI gene suggests a deep and distinct dichotomy of contemporary spermatophytes into two monophyla: gymnosperms (including Gnetales) and angiosperms. J. MoI. Evol. 49: 310-315.

SCHABILION, J. T. AND N. C. BROTZMAN. 1979. A tetrahedral megaspore arrangement in a seed fern ovule of Pennsylvanian age. Am. J. Bot. 66: 744-745.

SCHECKLER, S. E. AND H. P. BANKS. 1971. Anatomy and relationships of some Devonian progymnosperms from New York. Am. J. Bot. 58: 737-751.

SCHMID, R. 1967. Electron microscopy of wood of Callixylon and Cordaites. Am. J. Bot. 54: 720-729.

SCOTT, D. H. 1923. Studies in fossil botany. Black, London, UK.

SCOTT, D. H. 1924. Fossil plants of the Calamopitys type from the Carboniferous rocks of Scotland. Trans. Roy. Soc. Edin. 53: 569-596.

SERBET, R. AND G. W. ROTHWELL. 1992. Characterizing the most primitive seed ferns. 1. A reconstruction of Elkinsia polymorpha. Int. J. Plant Sci. 153: 602-621.

SERBET, R. AND G. W. ROTHWELL. 1995. Functional morphology and homologies of gymnospermous ovules: evidence from a new species of Stephanospermum (Medullosales). Can. J. Bot. 73: 650-661.

SEWARD, A. C. 1917. Fossil Plants, Vol. Ill, Pteridospermae, Cycadofilices, Cordaitales, Cycadophyta. Cambridge University Press, London and Cambridge, UK.

SMITH, A. B. 1994. Systematics and the fossil record: documenting evolutionary patterns. Blackwell, Oxford, UK

SMITH, S. Y. AND R. A. STOCKEY. 2002. Permineralized pine cones from the Cretaceous of Vancouver Island, British Columbia. Int. J. Plant Sci. 163: 185-196.

SOLTIS, D. E., P. S. SOLTIS, AND M. J. ZANIS. 2002. Phylogeny of seed plants based on evidence from eight genes. Am. J. Bot. 89: 1670-1681.

SORENSON, M. D. 1999. TreeRot, version 2. Boston University, Boston, MA.

SRIVASTAVA, S. C. AND J. BANERJI. 2001. Pentoxylon plant: a reconstruction and interpretation. Plant Cell Biol. Devel. (Szeged) 13: 11-18.

STEWART, W. N. AND ROTHWELL, G. W. 1993. Paleobotany and the evolution of plants. Cambridge University Press, Cambridge, UK.

STUBBLEFIELD, S. P. AND G. W. ROTHWELL. 1989. Cecropsis luculentum gen. et sp. nov., evidence for heterosporous progymnosperms in the Upper Pennsylvanian of North America. Am. J. Bot. 76: 1415-1428.

SUN, G., Q. J., D. L. DILCHER, S. ZHENO, K. C. NIXON, AND X. WANG. 2002. Archaefructaceae, a new basal angiosperm family. Science 296: 899-904.

SUN G., D. L. DILCHER, S. ZHENG, AND Z. ZHOU. 1998. In search of the first flower: a Jurassic angiosperm, Archaefructus, from Northeast China. Science 282: 1692-1695.

SUN, G., S. ZHENG, D. L. DILCHER, Y. WANG, AND S. MEI. 2001. Early angiosperms and their associated plants from western Liaoning, China. Shanghai Scientific and Technological Education Publishing House, China. 227 pp.

SWOFFORD, D. L. 2002. PAUP: Phylogenetic Analysis Using Parsimony, version 4.Ob. Illinois Natural History Survey, Champaign, IL.

TAYLOR, T. N. AND D. A. EGGERT. 1967. Petrified plants from the Upper Mississippian of North America. 1: The seed Rhynchosperma gen. nov. Am. J. Bot. 55: 306-313.

TAYLOR, T. N. AND S. E. SCHECKLER. 1996. Devonian spore ultrastructure: Rhabdosporites. Rev. Palaeobot. Palynol. 93: 147-158.

TAYLOR, T. N. AND E. L. TAYLOR. 1993. The biology and evolution of fossil plants. Prentice Hall, Englewood Cliffs, N J.

TAYLOR, E. L., T. N. TAYLOR, AND H. KERP. 2006. The Mesozoic seed ferns: Old paradigms, new discoveries. J. Torrey Bot. Soc. 133: 62-82.

THOMAS, H. H. 1925. The Caytoniales, a new group of angiospermous plants from the Jurassic rocks of Yorkshire. Phil. Trans. R. Soc. Lond. 213B: 299-363.

THOMAS, H. H. 1933. On some pteridospermous plants from the Mesozoic rocks of South Africa. Phil. Trans. R. Soc. London B 222: 193-265.

THOMAS, H. H. 1955. Mesozoic pteridosperms. Phytomorphology 5: 177-185.

TIAN, B. AND S. J. WANG. 1987. On cordaitalean stems in coal balls from Taiyuan Formation in Xishan, Taiyuan, Shanxi. Acta Palaeont. Sin. 26: 196-204 [in Chinese with English summary].

TIAN, B. AND S. J. WANG. 1988. Study on primary vascular systems of cordaitalean stems Shanxioxylon and Pennsylvanioxylon tianii in coal balls from Shanxi. Acta Palaeont. Sin. 27: 21-30 [in Chinese with English summary].

TOWNROW, J. A. 1960. The Peltaspermaceae, a pteridosperm family of Permian and Triassic age. Palaeontology 3: 333-361.

TRIVETT, M. L. 1992. Growth architecture, structure, and relationships of Cordaixylon iowense nov. comb. (Cordaitales). Int. J. Plant Sci. 154: 572-588.

TRIVETT, M. L. AND G. W. ROTHWELL. 1985. Morphology, systematics, and paleoecology of Paleozoic fossil plants: Mesoxylon priapi, sp. nov. (Cordaitales). Syst. Bot. 10: 205-223.

TRIVETT, M. L. AND G. W. ROTHWELL. 1988. Diversity among Paleozoic Cordaitales: the vascular architecture of Mesoxylon birame Baxter. Bot. Gaz. 149: 116-125.

WANG, J., H. W. PFEFFERKORN, B. SUN, AND L. LIU. 2003. Discovery of organic connection of Chiropteris Kurr and Nystroemia Halle from the Early Permian of Western Henan, China. Chinese Sci. Bull. 48: 2248-2252.

WANG, S. J., J. HILTON, B. TIAN, AND J. GALTIER. 2003. Cordaitalean seed plants from the Early Permian Taiyuan Formation of North China. I. Delimitation and reconstruction of the whole-plant Shanxioxylon sinense. Int. J. Plant Sci. 164: 89-112.

WANG, S. J. AND B. TIAN. 1991a. On male cordaitalean reproduction organs in coal balls from Taiyuan Formation, Xishan coal-field, Taiyuan, Shanxi. Acta Palaeontol. Sin. 30: 743-749 [in Chinese with English abstract].

WANG, S. J. AND B. TIAN. 199 Ib. A new species of petrified ovules of Late Palaeozoic age. Acta Bot. Sin. 33: 958-962 [in Chinese with English abstract].

WANG, S. J. AND B. TIAN. 1993a. On female reproductive cordaitalean organs in coal balls from Taiyuan Formation, Xishan coal-field, Taiyuan, Shanxi. Acta Palalaeontol. Sin. 32: 760-764 [in Chinese with English abstract].

WANG, S. J. AND B. TIAN. 1993b. A new cordaitalean root-species in Xishan coal-field, Taiyuan. Coal Geol. Explor. 21: 7-11 [in Chinese with English summary].

WATSON, J. 1988. The Cheirolepidiaceae, pp. 282-447. In C. B. Beck [ed.], Origin and evolution of gymnosperms. Columbia University Press, New York, NY.

WILLIAMSON, W. C. 1877. On the organisation of the fossil plants of the Coal Measures. Part 8. Ferns (continued) and gymnosperm stems and seeds. Phil. Trans R. Soc. Lond. 167: 213-270.

WILLIAMSON, W. C. AND D. H. SCOTT. 1895. Further observations on the organisation of the fossil plants of the Coal Measures. 3. Lyginodendron and Heterangium. Phil. Trans. R. Soc. 186: 703-779.

WINTER, K.-U., A. BECKER, T. MUNSTER, J. T. KIM, H. SAEDLER AND G. THEISSEN. 1999. MADS-box genes reveal that gnetophytes are more closely related to conifers than to flowering plants. Proc. Nat. Acad. Sei. 96: 7342-7347.

WON, H. AND S. S. RENNER. 2003. Horizontal gene transfer from flowering plants to Gnetum. Proc. Nat. Acad. Sei. 100: 10824-10829.

YANG, Y., B. Y. GENG, D. L. DILCHER, Z. D. CHEN, AND T. A. LOTT. 2005. Morphology and affinities of an early Cretaceous Ephedra (Ephedraceae) from China. Am. J. Bot. 92: 231-241.

Jason Hilton2,3

School of Geography, Earth and Environmental Sciences, University of Birmingham, Edgbaston, Birmingham, B15 2TT, UK

Richard M. Bateman

Natural History Museum, Cromwell Road, London, SW7 5BD, UK

2 We are grateful to Jim Doyle for providing the initial data matrix from which this analysis was founded and for much subsequent helpful discussion regarding the treatment of both taxa and characters. Brian Axsmith, Jerry Davis, Mike Frohlich, Steven Jansen, Paula Rudall, and Dennis Stevenson are thanked for insights into various characters and concepts. We also thank Mike Dunn and Gar Rothwell for the invitation to participate in this symposium volume, and for reviewing this manuscript.

3 Author for correspondence, E-mail: j.m.hilton@

Received for publication July 19, 2005, and in revised form September 20, 2005.

Copyright Torrey Botanical Society Jan-Mar 2006

Provided by ProQuest Information and Learning Company. All rights Reserved