Molecules, Morphology, Fossils, and the Relationship
of Angiosperms and Gnetales
Section of Evolution and Ecology, University of California, Davis, California 95616
Received May 15, 1997; revised December16, 1997
Morphological analyses of seed plant phylogeny agree
that Gnetales are the closest living relatives of angio-
sperms, but some studies indicate that both groups are
monophyletic, while others indicate that angiosperms
are nested within Gnetales. Molecular analyses of
several genes agree that both groups are monophy-
letic, but differ on whether they are related. Conflicts
among morphological trees depend on the interpreta-
tion of certain characters; when these are analyzed
critically, both groups are found to be monophyletic.
Conflicts among molecular trees may reflect the rapid
Paleozoic radiation of seed plant lines, aggravated by
the long branches leading to extant taxa. Trees in
which angiosperms are not related to Gnetales conflict
more with the stratigraphic record. E ven if molecular
data resolve the relationships among living seed plant
groups, understanding of the origin of angiosperm
organs will require integration of fossil taxa, necessar-
ily using morphology.
?1998 Academic Press
Oneofthemost intractableproblems in plant system-
atics and evolution is the origin of angiosperms, or
flowering plants, the dominant plant group on earth
today. One aspect of this question is which of the other
seed plants, or ‘‘gymnosperms,’’aremost closely related
to angiosperms—their closest outgroup (sister group)
or outgroups. Toa paleobotanist, this involves a consid-
eration of the many extinct Paleozoic and Mesozoic
seed plants, but to a molecular systematist, who can
study only extant organisms, the question is which
living seed plants are closest to the angiosperms. The
only living group that has attracted significant atten-
tion is the Gnetales (Martens, 1971; Friedman, 1996),
which include three divergent genera. Gnetum is a
tropical rainforest tree or vine, with leaves virtually
identical to those of a dicotyledonous angiosperm.
Ephedra is a desert shrub with small, pointed leaves,
but as in Gnetum these are opposite. Welwitschia, in
the deserts of southwest Africa, produces only one pair
of strap-shaped leaves. All three genera resemble most
angiosperms in having special water-conducting ves-
sels in the wood, and they all have reproductive struc-
tures organized into compound strobili and seeds with
a micropylar tube. Recently, Friedman (1990, 1992) and
Carmichael and Friedman (1996) confirmed earlier
reports that Gnetales have part of the angiosperm
process of double fertilization: both of the two sperm
produced by the male gametophyte (in the pollen) fuse
with nuclei in the female gametophyte (in the ovule),
but thesecond fusion produces an additional embryoor
embryos rather than triploid endosperm tissue, as in
angiosperms. The other living seed plants (cycads,
Ginkgo, and conifers) have rarely been associated with
Molecular data may also help clarify the ‘‘rooting’’ of
the angiosperms, or determining which angiosperm
groups are ‘‘basal’’ (i.e., derived from the first splits in
angiosperm phylogeny). This may help in reconstruct-
ing states in the common ancestor of living angio-
sperms and thus in connecting them with other groups.
However, in this article I will consider only relation-
ships of angiosperms with other seed plants.
At first sight, morphological and molecular data on
the relationship of angiosperms and Gnetales may
appear toconflict, but closer examination suggests that
they are actually complementary. On one question,
whether angiosperms and Gnetales are both monophy-
letic, molecular data arein agreement, whereasmorpho-
logical analyses have given inconsistent results. How-
ever, on the question of whether the two groups are
related at all, it is the molecular data that are inconsis-
tent and morphological data that are unanimous. I will
argue that the disagreements among morphological
analyses are due to questionable interpretations of
characters, and when these are corrected one obtains
results that agree with molecular data. The disagree-
ments among molecular analyses may reflect inherent
limitations of molecular data in resolving ancient radia-
tions. I will also argue that consideration of the great
diversity of fossil plants is necessary in order tounder-
stand the origin of angiosperm organs, since structures
in living seed plants are too highly modified, and that
fossils may provide independent tests of hypotheses
MOLECULAR PHYLOGENETICS AND EVOLUTION
Vol. 9, No. 3, J une, pp. 448–462, 1998
ARTICLE NO. FY980506
Copyright?1998 by Academic Press
All rights of reproduction in any form reserved.
based on molecular data, both as additional taxa and as
sources of stratigraphic evidence that bears on tree
topology. Since fossils can be integrated into phylog-
enies only through their morphological characters,
thesearearguments for continuedattention tomorphol-
ogy, even if molecules become the primary basis for
reconstructing relationships among living groups.
HYPOTHESES ON RELATIONSHIPS
One of the earliest hypotheses, proposed by Wett-
stein (1907) but often called the ‘‘englerian’’ theory,
postulated that angiosperms were derived from Gne-
tales, which were in turn related to conifers. This was
based on comparisons between thecompound strobili of
Gnetales, which consist of bracts with simple, uni-
sexual ‘‘flowers’’ in their axils, with one orthotropous
(erect) ovule in the female flowers, and the inflores-
cences of the wind-pollinated ‘‘Amentiferae,’’ which are
also made up of simple, unisexual flowers, with one
orthotropous ovule. This view fell out of favor with
increasing evidence that Amentiferae are advanced
within angiosperms, compared to ‘‘Magnoliidae,’’ such
as Magnolia, with showy, bisexual flowers, and this
conclusion has been amply confirmed by both fossil and
molecular data (Doyle, 1978; Chase et al., 1993; Crane
et al., 1995). However, englerian ideas are far from
dead, since one magnoliid group, Chloranthaceae, is
also like Gnetales in having opposite leaves, inflores-
cences of simpleflowers, and oneorthotropous ovule.
A competing theory was proposed by Arber and
Parkin (1907), inspired by the large, bisexual ‘‘flowers’’
of Mesozoic Bennettitales, which had cycad-like leaves.
In Bennettitales, the ovules were borne on simple
stalks, rather than anything like a carpel, the closed
ovule-bearing organ of angiosperms. Arber and Parkin
therefore proposed that the two groups were derived
froma common ancestor with pinnatesporophylls, with
the megasporophylls folded to enclose the ovules in
angiosperms, but reduced in Bennettitales. Arber and
Parkin (1908) thought that Gnetales were also related
toangiosperms, but they interpreted their simple flow-
ers as reduced rather than primitive.
Angiosperms have also been associated with the
Mesozoic genus Caytonia, which had palmately com-
poundleaves with simplereticulatevenation andmega-
sporophylls bearing two rows of anatropous (reflexed)
‘‘cupules,’’ each containing several ovules (Gaussen,
1946; Stebbins, 1974; Doyle, 1978). This would explain
the fact that the ovules of angiosperms are usually
anatropous and bitegmic (with twointeguments). If the
number of ovules in a Caytonia cupule was reduced to
1, the cupule would resemble an angiosperm ovule,
with the outer integument corresponding tothe cupule
wall. The carpel might be derived by widening and
folding of thesporophyll rachis.
Others have linked angiosperms to the Permian
glossopterids, which had simple leaves with simple
reticulatevenation andleaflikeovulatestructures bear-
ing one or several cupules or ‘‘sporophylls’’ (Stebbins,
1974; Retallack and Dilcher, 1981). The cupules could
be transformed into bitegmic ovules as proposed for
Caytonia, while the carpel could be derived from the
Phylogenetic Analyses Based on Morphology
Of the several cladistic analyses of seed plants based
on morphological data, all agree that Gnetales are the
closest living relatives of angiosperms, but some analy-
ses indicate that both groups are monophyletic, while
others indicate that angiosperms are nested within
Gnetales. They also differ in how fossils are interpo-
lated among living taxa—some link angiosperms and
Gnetales with Caytonia and glossopterids, whileothers
link them with coniferopsids (conifers, ginkgos, Paleo-
zoic cordaites). For this reason, they vary greatly in
implications for character evolution, such as origin of
the carpel. Some of these scenarios might never be
envisioned if only living taxa were considered, as is the
case with molecular analyses (cf. Doyle and Donoghue,
In the analysis of Crane (1985), angiosperms and
Gnetales were sister groups. The two were in turn
related to Bennettitales and the Mesozoic genus Pen-
toxylon, and the whole clade was associated with
glossopterids, Caytonia, and Triassic corystosperms.
Since Bennettitales, Gnetales, and angiosperms all
haveflowerlikestructures, this would support anArber
and Parkin (1907) scenario for floral evolution, with
angiosperm carpels and bennettitalian stalked ovules
both derived from pinnate sporophylls like those of
To test these results, Doyle and Donoghue (1986,
1992) added several characters that might support
alternative relationships, but this had little effect
(Fig. 1). Most importantly, angiosperms wereseparated
from Gnetales, as the sister group of Bennettitales,
Pentoxylon, and Gnetales. These four groups were
called ‘‘anthophytes’’ because their common ancestor
wouldhavehadflowerlikestructures. Caytonia was the
sister group of the anthophytes, supporting the homol-
ogy of thebitegmic ovulewith theCaytonia cupule. The
flowers of Gnetales would represent the culmination of
a trend for reduction and simplification. Ovules of
Gnetales have two integuments; the outer of these
would correspond to the perianth of Bennettitales.
These inferences illustrate the importance of fossils:
without fossils interpolated among extant taxa, there
would be noCaytonia prototype for the bitegmic angio-
sperm ovule, and the outer integument of angiosperms
might be derived from the outer integument of Gne-
RELATIONSHIP OF ANGIOSPERMS AND GNETALES
Unfortunately, these results were not robust: Doyle
and Donoghue (1987, 1992) found very different trees
that were only one or two steps longer. These include
‘‘neo-englerian’’ trees, with Gnetales basal in antho-
phytes and anthophytes linked with conifers (Fig. 2).
This would imply that the simple flowers of Gnetales
are primitive and homologous with the fertile short
shoots of Paleozoic conifers. The angiosperm outer
integument might be derived from the outer integu-
ment of Gnetales, itself derived from sterile append-
ages on the fertile short shoot; the carpel might corre-
Anthophytes werealsolinked with conifers in Rothwell
and Serbet (1994), while angiosperms and Gnetales
were sister groups within anthophytes. Angiosperms
and Gnetales were linked with conifers in analyses of
living plants only by Loconte and Stevenson (1990) and
DoyleandDonoghue(1992). However, this result would
not necessarily imply a neo-englerian scenariofor floral
evolution, since it would also be consistent with trees
with glossopterids and Caytonia interpolated between
conifers and anthophytes.
to the subtendingbract (Doyle,1994).
F IG. 1.
scenariofor floral evolution, with shading indicating homologous parts (modifiedfrom Doyle, 1994). Names of extant taxa aregiven in boldface
Representative most parsimonious seed plant tree of Doyle and Donoghue (1992), showing evolution of leaf morphology and a
J AMES A. DOYLE
Another weakness of these studies is that they
treated angiosperms as a single taxon, which required
debatable assumptions on basic states. For example,
Doyle and Donoghue (1986, 1992) scored angiosperms
as having spiral leaves and anatropous ovules. How-
ever, opposite leaves and orthotropous ovules occur in
Chloranthaceae, which are among the oldest angio-
sperms in the Cretaceous fossil record (Crane et al.,
1995) and have been considered a link between angio-
sperms and Gnetales (Meeuse, 1972; Hickey and Tay-
In an attempt to correct this problem, Doyle et al.
(1994) includedninepotentially basal angiospermtaxa,
to which Doyle (1996) added two more and a J urassic
relative of Gnetales, Piroconites (Kirchner, 1992; van
Konijnenburg-van Cittert, 1992), which had glossop-
terid-like reproductive structures. The two analyses
gavesimilar results. In Doyle(1996), Gnetales werethe
closest living group to angiosperms, but Caytonia was
located on the angiosperm line, rather than below the
anthophytes, and in many trees glossopterids were the
sister group of anthophytes (Fig. 3). This implies that
the common ancestor of angiosperms and Gnetales had
glossopterid-like ovulate structures, retained up to
F IG. 2.
leaf morphology and a scenariofor floral evolution (modified from Doyle, 1994).
Representative most parsimonious neo-englerian tree based on the data set of Doyle and Donoghue (1992), showing evolution of
RELATIONSHIP OF ANGIOSPERMS AND GNETALES
F IG. 3.
evolution of theovulatestructures. GNET, Gnetales. Dechellyia is a LateTriassic fossil tooincompletely known tobeincluded in theanalysis.
Representative most parsimonious seed plant tree of Doyle (1996), showing evolution of leaf morphology and a scenario for
Piroconites on the line to Gnetales, and anatropous
cupules arose on the Caytonia–angiosperm line, rather
than being ancestral in anthophytes.
In contrast, in the analysis of Nixon et al. (1994),
which included 18 angiosperms, angiosperms were
nested within Gnetales, which werethereforeparaphy-
letic rather than monophyletic (Fig. 4). Angiosperms
were linked with Gnetum and Welwitschia, whereas
Ephedra was one or two nodes lower. In most trees,
anthophytes were linked with conifers, never with
glossopterids and Caytonia. These results imply that
angiosperms were derived from a gnetalian proto-
type—a neo-englerian scenario (cf. Fig. 2). Consistent
with this, Chloranthus or Casuarina was basal in
angiosperms, rather than Magnoliales or Nymphaeales
(Donoghue and Doyle, 1989; Doyle et al., 1994; Doyle,
1996). Angiosperms were also nested within Gnetales
in a smaller analysis by Hickey and Taylor (1996),
linked directly with Gnetum.
Phylogenetic Analyses of Molecular Data
In contrast, all major molecular analyses strongly
support angiosperms and Gnetales as monophyletic
groups, thus contradicting Nixon et al. (1994) and
Hickey andTaylor (1996), but they disagreeon whether
the two groups are related at all. Simplified molecular
trees are presented in Fig. 5, together with arrange-
ments of extant taxa based on morphology (Figs. 5a–
5d). I will pass over two early studies because of the
short length of sequences and small number of taxa
analyzed: Hori et al. (1985), based on 5S rRNA, which
did not include Gnetales; and Troitsky et al. (1991),
based on 5S, 5.8S, and short 18S sequences, which
separated Gnetum and Ephedra, possibly because Wel-
witschia was not included.
In Hamby and Zimmer (1992), based on most of the
18S rRNA gene and smaller portions of 26S, angio-
sperms andGnetales wereboth monophyletic, but their
relationships were unstable. In a neighbor-joining
analysis of 60 taxa, with seed plants rooted by Psilotum
and Equisetum, Gnetales were linked with angio-
sperms, while cycads, Ginkgo, and conifers formed a
cladeat thebaseof seedplants (Fig. 5e). In a parsimony
analysis the two angiosperm outgroups were reversed,
sothat Gnetales werebasal in seed plants (Fig. 5f), but
this was favored over the other arrangement by only
one step. When 72 taxa were analyzed with parsimony
(Doyleet al., 1994), therelationshipof angiosperms and
Gnetales was favored. Both trees correspond to the
same unrooted tree of seed plants; the difference is
whether outgroups attach tothe line tocycads, Ginkgo,
and conifers or the line to Gnetales. There is reason to
question any rooting based on living outgroups, which
branched from seed plants beforetheMiddleDevonian.
In morphological analyses including fossils (Figs. 1–4),
the crown-group (all derivatives of the most recent
common ancestor of extant seed plants) was rooted
F IG. 4. Representativeseed plant treeof Nixon et al. (1994).
RELATIONSHIP OF ANGIOSPERMS AND GNETALES
with Late Devonian–Carboniferous progymnosperms
and seed ferns, which are closer than living outgroups,
and the root was attached among cycads, Ginkgo, and
Doyle et al. (1994) performed bootstrap and decay
analyses to evaluate the strength of the rRNA results.
Because they questioned the rooting based on Psilotum
and Equisetum, this was an unrooted analysis. Angio-
sperms andGnetales werestrongly supportedas clades,
at bootstrap percentages of 100 and 99%, respectively,
and the twowere linked at the 88% level. This does not
directly evaluate whether angiosperms and Gnetales
form a clade, but it does measure support for an
unrooted tree in which the two taxa could be a clade
Trees with Gnetales basal in seed plants and angio-
sperms linked with cycads, Ginkgo, and conifers (Fig.
5f) were also found in analyses of the chloroplast gene
rbcL by Albert et al. (1994) and Doyle and Sanderson
(1997). In thelatter analysis, bootstrap support for this
rooting (i.e., for seed plants other than Gnetales) was
83%, but Gnetales and angiosperms weremonophyletic
at the96 and99% levels, respectively. However, Hasebe
et al. (1992) found a different tree when they converted
rbcL data into amino acid sequences and analyzed
them by neighbor joining, with angiosperms the sister
group of other seed plants (Fig. 5g). Gymnosperms
therefore formed a clade, but its bootstrap support was
only 63%. Gnetales were basal in gymnosperms, so
angiosperms and Gnetales could be linked if seed
plants were rooted differently. Stefanovic (1996) found
a similar tree based on 28S rDNA; again, bootstrap
support for thegymnosperms was low, 66%.
In contrast, when Hasebe et al. (1992) analyzed their
rbcL data with maximum likelihood, angiosperms were
the sister group of gymnosperms but Gnetales were
nested within the latter, linked with cycads (Fig. 5h).
This tree cannot be converted into one with angio-
sperms and Gnetales linked by rerooting seed plants.
Similar trees, but with Gnetales linked with conifers
(Fig. 5i), were found by Goremykin et al. (1996), based
on chloroplast rDNA ITS sequences, by Chaw et al.
(1997), based on 18S rDNA, and by Bowe and DePam-
philis (1997), based on the mitochondrial gene coxI.
However, in Goremykin et al., bootstrap support was
only 53–58% for the gymnosperms and 50–56% for
Gnetales plus conifers, in contrast to90–92% for angio-
sperms and 99–100% for Gnetales. In Chaw et al.
(1997), bootstrap support was 100% for both angio-
sperms and Gnetales, but only 56–75% for gymno-
sperms. Support for Gnetales plus conifers was higher,
84%, but it should be noted that this is less than the
88% value for angiosperms plus Gnetales in the rRNA
analysis of Doyle et al. (1994). Chaw et al. (1997) cited
absences of angiosperm indels as evidence for the
monophyly of gymnosperms, but such absences are
F IG. 5.Relationships of living groups found in morphological and molecular analyses of seed plants.
J AMES A. DOYLE
presumably symplesiomorphies that occurred in the
common ancestor of all seed plants and are therefore
irrelevant tothestatus of gymnosperms.
As summarized in Fig. 5, all morphological analyses
(Figs. 5a–5d) agree that angiosperms and Gnetales are
related, but some indicate that both groups are mono-
phyletic (Figs. 5a–5c), while others indicate that Gne-
tales are paraphyletic and angiosperms are nested
within them (Fig. 5d). The relationships of other seed
plants are poorly resolved, but they always form a
paraphyletic basal series. Molecular studies (Figs. 5e–
5i) all strongly support the monophyly of both angio-
sperms and Gnetales. Some indicate that the two
groups are related (Fig. 5e), but most imply that they
are not. In some analyses (Figs. 5f and 5g), angio-
sperms and Gnetales could be related if seed plants
were rooted differently, but in others (Figs. 5h and 5i),
they cannot be related under any rooting. However,
support for thesearrangements is weak.
In the following sections, I will adduce evidence that
the molecular results are more credible in cases where
their internal support is strong—that angiosperms and
Gnetales are both monophyletic—and the morphologi-
cal results where these are strongly supported—that
the two groups are related. It should be noted that the
disagreements among analyses of the two kinds are
somewhat different. With the molecular studies, many
of the conflicts are between analyses of different genes
(although some involve method of analysis or taxon
sampling), whereas with the morphological analyses
the conflicts are between data sets made up of overlap-
ping characters, which, however, havebeen interpreted
differently. The fact that molecular systematics in-
volves fewer problems of character analysis (except
sometimes in sequence alignment) is one of its great
strengths. The conflicts among morphological analyses
might thus seem to confirm the view that morphologi-
cal data are hopelessly subjective and can be reinter-
preted to support any hypothesis one likes. Part of the
following discussion is therefore a defense of morphol-
ogy against this charge: it is often possible to decide
objectively between interpretations of characters. In
addition, I will argue that fossils may provide indepen-
dent criteria for choosing among molecular hypotheses.
Conflicts among Morphological Analyses
With morphological data, the main disagreement on
relations of living taxa concerns whether Gnetales are
monophyletic or paraphyletic, as in Nixon et al. (1994)
and Hickey and Taylor (1996). One could simply claim
that the molecular data refute these hypotheses, and
this might be a valid position. However, the same
conclusion can be reached by closer examination of the
relevant morphological characters. I will concentrate
on theanalysis of Nixon et al.; that of Hickey andTaylor
was based on toofew taxa and character definitions too
heavily linked with oneevolutionary scenario.
Even without considering individual characters, there
is evidence that support for the paraphyly of Gnetales
in the Nixon et al. (1994) data set is weak. In an
analysis of this data set, Donoghue, Doyle, and Fried-
man (cited in Doyle, 1996) found that bootstrap support
for the angiosperm–Gnetum–Welwitschia clade was
only 54%. Furthermore, when Albert et al. (1994)
combined the Nixon et al. data set with rbcL data for
extant taxa, they obtained trees in which Gnetales
weremonophyletic.An even moreintriguing result was
that anthophytes were linked with Caytonia and glos-
sopterids rather than with conifers. This may seem
paradoxical, since there are no molecular data on
Caytonia. However, it can be explained as a result of
strong molecular support for Gnetales, weak morpho-
logical support for theangiosperm–Gnetum–Welwitschia
clade, and the fact that angiosperms share morphologi-
cal features with Caytonia, even in theNixon et al. data
set. When angiosperms arenested within Gnetales, the
latter features have no effect on the position of antho-
phytes as a whole. But when the rbcL data pull
Gnetales together as a clade, angiosperms go to the
base of anthophytes, and their features draw the whole
group together with Caytonia.
The key characters responsible for the result of
Nixon et al. (1994) are the five synapomorphies that
link angiosperms with Gnetum and Welwitschia, tothe
exclusion of Ephedra. Consideration of these charac-
ters (documented in Doyle, 1996) may illustrate some
general principles about pitfalls in the use of morpho-
logical characters and ways to avoid them. These
examples emphasize the necessity of considering struc-
tures in their positional and developmental contexts.
Onecharacter is cellularization of thefemalegameto-
phyte (in the ovule). The basic condition in seed plants
is alveolar: gametophyte ontogeny begins with a phase
of free-nuclear divisions and then cell walls form in a
honeycomb-likepattern aroundindividual nuclei. How-
ever, Gnetum and Welwitschia show a nonalveolar
pattern, in which cell walls form irregularly around
several nuclei. In angiosperms, the female gameto-
phyte is reduced toan egg cell and twosynergids at the
micropylar end, three antipodal cells at the other end,
and two free polar nuclei in the middle. Nixon et al.
(1994) scored angiosperms like Gnetum and Wel-
witschia, presumably becausethemiddleofthegameto-
phyte remains free-nuclear. However, the egg, syner-
gids, and antipodals of angiosperms areall uninucleate
cells, and in this sense they are more like other seed
plants. One solution would be to score angiosperms as
unknown. However, it is also possible to redefine the
character in terms of a related but less ambiguous
distinction: cellularization resulting in uninucleate vs
multinucleatecells. Under this definition, angiosperms
RELATIONSHIP OF ANGIOSPERMS AND GNETALES
can be unambiguously scored as having the basic state
and Gnetum and Welwitschia the derived state. This
suggests a general principle: when choosing among
alternative distinctions between character states, one
should define states such that critical taxa can be
A related character is the presence or absence of
archegonia, the original egg-containing organ in land
plants. Ephedra has archegonia, with an egg and neck
cells, but they are absent in angiosperms, Gnetum, and
Welwitschia. This may seem like a clear, objective
distinction, but closer examination shows that the
conditions in angiosperms and the two gnetalian gen-
era arevery different. In Gnetum, themicropylar end of
the gametophyte remains free-nuclear, and free nuclei
function as eggs (Martens, 1971; Carmichael andFried-
man, 1996); in Welwitschia, the whole gametophyte
becomes cellularized, but individual nuclei of multi-
nucleate cells act as eggs (Carmichael and Friedman,
personal communication). In contrast, although angio-
sperms have no neck cells and thus no archegonia, the
egg is a normal, uninucleate cell. I see no basis for
equating these two states or for considering one a
modification of the other; in fact, one might argue that
they resulted from different modifications of the basic
gametophyte ontogeny. In the face of such divergent
scenarios for character evolution, bias in favor of one
scenario over another can be avoided by treating the
alternativeconditions (threein this case) as states of an
Similar problems concern embryogeny, which passes
through an initial free-nuclear phasein cycads, Ginkgo,
and conifers but proceeds entirely by cellular divisions
in angiosperms, Gnetum, and Welwitschia. The critical
taxon is Ephedra, which Nixon et al. (1994) and previ-
ous workers (Crane, 1985; Doyle and Donoghue, 1986,
1992) interpreted as free-nuclear, because the two
zygotes divide toproduce eight free nuclei. However, as
emphasized by Friedman (1994), the following stages
are more like those in angiosperms, Gnetum, and
Welwitschia, in that each of the eight free nuclei (now
surrounded by cell walls) produces a separate embryo
by cellular divisions, whereas in cycads, Ginkgo, and
conifers many free nuclei contribute to one embryo.
This suggests that the occurrence of two free-nuclear
divisions in Ephedra is a secondary advance. Whether
or not this hypothesis is correct, it casts doubt on the
scoring of Ephedra. Again, this character can be re-
placed with a related one, whether each embryo is
derived from several free nuclei or from a single cell by
cellular divisions, for which Ephedra can beunambigu-
ously scored as having thelatter state.
A final case involves redundancy of two characters
used by Nixon et al. (1994): number of nuclei in the
male gametophyte (in the pollen) and presence or
absence of a stalk cell. In the basic male gametophyte
ontogeny in seed plants, successive divisions produce
three or more sterile nuclei, the last called the stalk
cell, andtwosperm, but in Gnetum andWelwitschia the
number of nuclei is reduced tofour and in angiosperms
to three. However, reduction from five to four nuclei in
Gnetum and Welwitschia occurs by loss of thestalk cell,
so the changes in the two characters are equivalent.
The stalk cell would alsobe lost in angiosperms during
reduction to three nuclei, whether this occurred by the
same or another route. Treating the stalk cell as a
separatecharacter thereforecounts thesameevolution-
ary event twice; on the general principle that such
weighting should be avoided, this character should be
When these characters are redefined in a more
neutral manner (Doyle, 1996) and the rest of the Nixon
et al. (1994) data set is left unchanged, one obtains
trees in which Gnetales and angiosperms are both
monophyletic. The bootstrap support for Gnetales is
only moderate (75%), but this is as high as for many
groups whose monophyly has rarely been questioned
Conflicts among Molecular Analyses
In discussing the results of molecular analyses, I
noted reasons to doubt the rootings obtained for seed
plants as a whole. Seed plants are a good example of a
radiation that occurred in a relatively short period a
long time ago, a situation in which molecular data are
least likely to be reliable, because of the small number
of molecular synapomorphies between nodes, thepossi-
bility that thesewereerased by later changes (multiple
hits), and the related problem of long branch attraction
(Felsenstein, 1978; Donoghue et al., 1989; Donoghue
and Sanderson, 1992).
This point is illustrated by Fig. 6, the tree in Fig. 3
(Doyle, 1996) with conifers and angiosperms reduced to
single taxa, plotted against geologic time. Black lines
show the known ranges of each group, white lines
‘‘ghost lineages’’ (Norell, 1992), where the cladogram
predicts that the line existed because its sister group is
already known in the fossil record, but the line itself is
not. Ghost lineages have been extended downward the
least possible distance (minimum implied gap: Benton
and Storrs, 1994), resulting in several apparent poly-
chotomies. Four living seed plant lines separated in the
Carboniferous; plants on their stem-lineages occur in
the Late Carboniferous or Permian, and most if not all
of the many seed plants known from the Early Carbon-
iferous are more primitive than their reconstructed
common ancestor (e.g., lacking such crown-group apo-
morphies as endarch primary xylem, pollen with honey-
comb-type exine structure and air sacs, and bilateral
ovule symmetry). Although the first records of the two
anthophyte lines are Late Triassic, it is possible that
they diverged in the Permian. If glossopterids are
paraphyletic (they have only two questionable autapo-
morphies, saccate and striate pollen), the ghost lin-
J AMES A. DOYLE
eages leading to both lines may be represented in the
Permian by plants known as glossopterids (Doyle,
Other insights may be gained by considering how
unparsimonious trees found with molecular data are in
terms of a morphological data set (Doyle, 1996). Figure
7 shows one of two trees found after forcing angio-
sperms tobethesister group of other living seed plants
(by adding 10 characters, with angiosperms and Elkin-
sia scored as 0, other living seed plants as 1, and fossils
as unknown, then subtracting 10 steps from the result-
ing trees). Contrary to Troitsky et al. (1991), this
arrangement does not mean that angiosperms and
gymnosperms were independently derived from Devo-
nian progymnosperms. Becauseof thestem, pollen, and
ovule advances shared by living seed plants, the angio-
F IG. 6.
10-million-year intervals, with known ranges, ghost lineages implied by the tree, and measures of consistency with the stratigraphic record
(seetext for discussion). Ranges based on Stewart and Rothwell (1993) and other general references.
Most parsimonious seed plant tree of Doyle (1996) (Fig. 3) plotted against geologic time (Palmer, 1983), approximated to
RELATIONSHIP OF ANGIOSPERMS AND GNETALES
sperm line splits off above medullosans, still linked
with Bennettitales and Caytonia. The next branch
leads toGnetales, Pentoxylon, glossopterids, andPiroco-
nites; the last to cycads, Ginkgo, conifers, and various
fossils. In effect, this tree reroots the seed plant crown-
group with angiosperms and their fossil relatives at the
Based on the present data set, this tree is 4 steps
longer than the most parsimonious trees, but it may
actually be even less parsimonious, because angio-
sperms and Gnetales have five other character states,
listed in Fig. 7, that are reconstructed as ancestral but
are more likely derived. Because the basal taxa are
fossils in which these characters are unknown, it is
most parsimonious to assume that they had the state
found in angiosperms and Gnetales. However, more
distant outgroups or other data suggest that the earli-
est seed plants actually had the alternative state. Each
character may thereforeinvolvea hidden extra step.
One of these characters is plotted in Fig. 7, presence
of a tunica in the apical meristem. This is an outer cell
layer that produces only epidermal tissue, whereas in
other groups the surface cells produce both epidermal
and internal tissues. The latter condition is more like
that in lower vascular plants, which have one apical
cell that produces both outer and inner tissues. Simi-
larly, the Ma ¨ule reaction (or the related abundance of
syringaldehyde as a lignin oxidation product) is absent
in most lower vascular plants (except Selaginella,
Equisetum, and Dennstaedtia; Gibbs, 1957; Logan and
Thomas, 1985). Logan and Thomas (1985) linked abun-
danceof syringaldehydetopresenceof fibers, which are
lacking in the wood of early seed plants. Thickness of
the nucellar cuticle is poorly documented in Paleozoic
seeds, but the thickened condition of Caytonia and
most anthophytes was first noted by Harris (1954) in
Mesozoic seeds. It is unlikely that the first seed plants
haddoublefertilization, which is associatedwith persis-
tence of a nucleus next to the egg (Donoghue and
Scheiner, 1992); in lower vascular plants, the adjacent
neck and ventral canal cells degeneratebeforefertiliza-
tion, as in cycads, Ginkgo, and conifers. Lower vascular
F IG. 7.
Doyle(1996), showing evolution of theapical meristem. MP, most parsimonious; below arelisted fivecharacters not documented in fossils that
the tree implies are ancestral in seed plants but are more likely to be derived based on more distant outgroups or other evidence, implying
additional steps (seetext for discussion).
Seed plant tree with angiosperms forced below other living seed plants, as in some molecular analyses, based on the data set of
J AMES A. DOYLE
plants are like angiosperms and Gnetales in having
cellular embryogeny, but early seed plants had large
egg cells (Stewart and Rothwell, 1993), which are
correlated with free-nuclear embryogeny, presumably
becauseit is difficult topartition a largecell.
Gnetales arenot associated with conifers in Fig. 7, as
they were in cpITS, 18S, and coxI trees (Goremykin et
al., 1996; Chaw et al., 1997; Bowe and DePamphilis,
1977; Fig. 5i). Apparently, even when angiosperms are
assumed to be basal in seed plants, there is more
morphological support for an arrangement in which
angiosperms and Gnetales form adjacent lines than
there is for a link between Gnetales and conifers. Trees
with angiosperms basal and Gnetales (plus Piroconites
and Bennettitales) linked with conifers are 10 steps
longer than the shortest trees. Trees with Gnetales
basal in living seed plants, as in rbcL analyses (Albert
et al., 1994; Doyle and Sanderson, 1977; Fig. 5f), are 5
F IG. 8.
ranges, ghost lineages implied by thetree, and measures of consistency with thestratigraphic record (seetext for discussion).
Seedplant treewith angiosperms forcedbelow other living groups of seedplants (Fig. 7), plottedagainst geologictime, with known
RELATIONSHIP OF ANGIOSPERMS AND GNETALES
steps longer than the shortest trees, not counting the
same five characters that probably involve additional
Molecular trees that disagree with morphology can
alsobequestioned becausethey conflict with thestrati-
graphic record. This is illustrated by two trees plotted
against geologic time, with known ranges and ghost
lineages: Fig. 6, a morphological tree; and Fig. 8, the
tree in Fig. 7 with angiosperms basal, as in molecular
trees of the type in Fig. 5g. The percentage of the total
tree length consisting of ghost lineages is an inverse
measure of consistency with the stratigraphic record
(Smith and Littlewood, 1994). A direct measure is the
relative completeness index (RCI), or 1 minus the ratio
of ghost lineages to observed ranges, expressed as a
percentage (Benton and Storrs, 1994). In the morpho-
logical tree (Fig. 6), the taxa that split off first are the
oldest Late Devonian and Carboniferous seed ferns,
followed by Permian and then Mesozoic groups. This
implies 470 My of ghost lineages and 1960 My of
observed ranges, or 19% ghost lineages and RCI ? 76%.
However, when angiosperms are basal (Fig. 8), the
sequence is from the oldest to the youngest groups of
seed plants and then back to taxa of intermediate age.
As a result, this tree implies 710 My of ghost lineages,
or 27% ghost lineages andRCI ? 64%. Thenumbers are
identical when Gnetales are basal among living seed
plants (Fig. 5f). These contrasts are all the more
significant because most of the increase in ghost lin-
eages involves extension of the lines leading to angio-
sperms and Gnetales down into the Carboniferous,
which probably has the best plant record of any period,
whereas theLatePermian andEarly Triassicgaps seen
in both trees fall in themost poorly known interval.
The reasons that molecular data may give incorrect
relationships of angiosperms and Gnetales are sug-
gestedby a study on rbcL andtheageof angiosperms by
Doyle and Sanderson (1997), which indicates that
evolution of this gene is far from clocklike. Assuming
that hepatics split from other land plants in the Late
Ordovician, the shortest tree, with Gnetales basal in
seed plants (Fig. 5f), implies that conifers and cycads
diverged in theMiddleJ urassic, 130 million years after
the first records of both lines (Middle Pennsylvanian
and Early Permian, respectively). Figure 9 shows an
rbcL tree found after imposing five constraints: Lycopo-
dium basal in vascular plants, as inferred from fossil
and molecular data (Raubeson and J ansen, 1992; Ken-
F IG. 9.
other evidence(seetext). Cycads, Ginkgo, and conifers areshort branches relativetoangiosperms and Gnetales.
Phylogram of seed plants based on rbcL (Doyleand Sanderson, 1997), with fivetopological constraints for greater congruencewith
J AMES A. DOYLE
rick and Crane, 1997); cycads, Ginkgo, and conifers as a
paraphyletic grade and Gnetales linked with angio-
sperms, as in morphological analyses; and monocots
and eudicots as clades, as indicated by larger rbcL
analyses (Chaseet al., 1993; otherwiseangiosperms are
rooted next to Oryza or Pisum). Figure 9 is a phylo-
gram, so branch lengths correspond to the amount of
molecular evolution. Conifers and especially cycads
and Ginkgo are very short branches relative to both
lower groups and angiosperms and Gnetales, while
within angiosperms some herbaceous lines (Oryza,
Pisum, Nicotiana) are especially long (as noted by
Bousquet et al., 1992). Theseresults can beexplained if
rbcL evolution slowed going from early vascular plants
to early seed plants, but then accelerated again in
angiosperms and Gnetales and still morein herbaceous
angiosperms.As a result, angiosperms or Gnetales may
be drawn to the base of seed plants by long branch
effects, specifically reversals to outgroup states on the
long lines leading tothesetaxa.
These exercises support the view that morphological
and molecular data may complement each other in
different parts of a phylogeny: morphology resolves
conflicts among molecular analyses by indicating that
angiosperms andGnetales arerelated, whereas molecu-
lar data resolve morphological conflicts by implying
that both groups are monophyletic. It is possible that
the relationship of angiosperms and Gnetales will
eventually be resolved using molecular data alone:
combining data from many genes, discovery of struc-
tural changes, or improvements in methods of analysis
that overcome the problem of long branch attraction.
However, morphological analysis will still be needed in
order to integrate fossils into the tree of living taxa,
which is essential for understanding how the living
groups and their structures originated.
I thank Simon Tillier for his kind invitation to participate in this
symposium, and Geeta Bharathan, Michael Donoghue, and Michael
Sanderson for valuablediscussions of theoretical issues.
Albert, V. A., Backlund, A., Bremer, K., Chase, M. W., Manhart, J . R.,
Mishler, B. D., and Nixon, K. C. (1994). Functional constraints and
rbcL evidence for land plant phylogeny. Ann. Mo. Bot. Gard. 81:
Arber, E.A. N., andParkin, J . (1907). On theorigin of angiosperms. J .
Linn. Soc. London Bot. 38: 29–80.
Arber, E. A. N., and Parkin, J . (1908). Studies on the evolution of the
angiosperms: The relationship of the angiosperms tothe Gnetales.
Ann. Bot. London 22: 489–515.
Benton, M. J ., and Storrs, G. W. (1994). Testing the quality of the
fossil record: Paleontological knowledge is improving. Geology 22:
Bousquet, J ., Strauss, S. H., Doerksen, A. H., and Price, R. A. (1992).
Extensive variation in evolutionary rate of rbcL gene sequences
among seed plants. Proc. Natl. Acad. Sci. USA 89: 7844–7848.
Bowe, L. M., and DePamphilis, C. W. (1997). Conflict and congruence
among three plant genetic compartments: Phylogenetic analysis of
seed plants from gene sequences of coxI, rbcL and 18S rDNA. Am.
J . Bot. 84(6, Suppl.): 178–179.
Carmichael, J . S., and Friedman, W. E. (1996). Doublefertilization in
Gnetum gnemon (Gnetaceae): Its bearing on theevolution of sexual
reproduction within theGnetales and theanthophyteclade. Am. J .
Bot. 83: 767–780.
Chase, M. W., and 41 others (1993). Phylogenetics of seed plants: An
analysis of nucleotide sequences from the plastid gene rbcL. Ann.
Mo. Bot. Gard. 80: 526–580.
Chaw, S.-M., Zharkikh, A., Sung, H.-M., Lau, T.-C., and Li, W.-H.
(1997). Molecular phylogeny of extant gymnosperms and seed
plant evolution: Analysis of nuclear 18S rRNA sequences. Mol.
Biol. Evol. 14: 56–68.
Crane, P. R. (1985). Phylogenetic analysis of seed plants and the
origin of angiosperms. Ann. Mo. Bot. Gard. 72: 716–793.
Crane, P. R., Friis, E. M., and Pedersen, K. R. (1995). The origin and
early diversification of angiosperms. Nature374: 27–33.
Donoghue, M. J ., and Doyle, J . A. (1989). Phylogenetic analysis of
angiosperms and therelationships of Hamamelidae. In ‘‘Evolution,
Systematics, and Fossil History of the Hamamelidae’’ (P. R. Crane
and S. Blackmore, Eds.), Vol. 1, pp. 17–45. Clarendon, Oxford.
Donoghue, M. J ., Doyle, J .A., Gauthier, J ., Kluge, A. G., and Rowe, T.
(1989). The importance of fossils in phylogeny reconstruction.
Annu. Rev. Ecol. Syst. 20: 431–460.
Donoghue, M. J ., and Sanderson, M. J . (1992). The suitability of
molecular and morphological evidencein reconstructing plant phylog-
eny. In ‘‘Molecular Systematics of Plants’’ (P. S. Soltis, D. E. Soltis,
and J . J . Doyle, Eds.), pp. 340–368. Chapman and Hall, New York.
Donoghue, M. J ., and Scheiner, S. M. (1992). The evolution of
endosperm: A phylogenetic account. In ‘‘Ecology and Evolution of
Plant Reproduction’’(R. E. Wyatt, Ed.), pp. 356–389. Chapman and
Hall, New York.
Doyle, J . A. (1978). Origin of angiosperms. Annu. Rev. Ecol. Syst. 9:
Doyle, J . A. (1994). Origin of the angiosperm flower: A phylogenetic
perspective. Plant Syst. Evol. Suppl. 8: 7–29.
Doyle, J . A. (1996). Seed plant phylogeny and the relationships of
Gnetales. Int. J . Plant Sci. 157(6, Suppl.): S3–S39.
Doyle, J . A., and Donoghue, M. J . (1986). Seed plant phylogeny and
theorigin of angiosperms:An experimental cladisticapproach. Bot.
Rev. 52: 321–431.
Doyle, J . A., and Donoghue, M. J . (1987). The importance of fossils in
elucidating seed plant phylogeny and macroevolution. Rev. Palaeo-
bot. Palynol. 50: 63–95.
Doyle, J . A., and Donoghue, M. J . (1992). Fossils and seed plant
phylogeny reanalyzed. Brittonia 44: 89–106.
Doyle, J .A., Donoghue, M. J ., andZimmer, E.A. (1994). Integration of
morphological and ribosomal RNA data on the origin of angio-
sperms. Ann. Mo. Bot. Gard. 81: 419–450.
Doyle, J . A., and Sanderson, M. J . (1997). Fossils, molecular clocks,
and theageof angiosperms. Am. J . Bot. 84(6, Suppl.): 132.
Felsenstein, J . (1978). Cases in which parsimony or compatibility
methods will bepositively misleading. Syst. Zool. 27: 401–410.
Friedman, W. E. (1990). Sexual reproduction in Ephedra nevadensis
(Ephedraceae):Further evidenceofdoublefertilization in a nonflow-
ering seed plant. Am. J . Bot. 77: 1582–1598.
RELATIONSHIP OF ANGIOSPERMS AND GNETALES
Friedman, W. E. (1992). Evidence of a pre-angiosperm origin of Download full-text
endosperm: Implications for the evolution of flowering plants.
Friedman, W. E. (1994). The evolution of embryogeny in seed plants
and the developmental origin and early history of endosperm. Am.
J . Bot. 81: 1468–1486.
Friedman, W. E. (Ed.) (1996). Biology and evolution of the Gnetales.
Int. J . Plant Sci. 157(6, Suppl.): S1–S125.
Gaussen, H. (1946). Les Gymnospermes, actuelles et fossiles. Trav.
Lab. Forest. Toulouse, Tome II Etud. Dendrol., Sect. 1, Vol. 1, Fasc.
3, Chap. 5, pp. 1–26.
Gibbs, R. D. (1957). TheMa ¨ulereaction, lignin, and therelationships
between woody plants. In ‘‘The Physiology of Forest Trees’’ (K. V.
Thimann, Ed.), pp. 269–312. Ronald Press, New York.
Goremykin, V., Bobrova, V., Pahnke, J ., Troitsky, A., Antonov, A., and
Martin, W. (1996). Noncoding sequences from the slowly evolving
chloroplast inverted repeat in addition torbcL data donot support
gnetalean affinities of angiosperms. Mol. Biol. Evol. 13: 383–396.
Hamby, R. K., and Zimmer, E. A. (1992). Ribosomal RNA as a
phylogenetic tool in plant systematics. In ‘‘Molecular Systematics
of Plants’’(P. S. Soltis, D. E. Soltis, and J . J . Doyle, Eds.), pp. 50–91.
Chapman and Hall, New York.
Harris, T. M. (1954). Mesozoic seed cuticles. Sven. Bot. Tidskr. 48:
Hasebe, M., Kofuji, R., Ito, M., Kato, M., Iwatsuki, K., and Ueda, K.
(1992). Phylogeny of gymnosperms inferred from rbcL gene se-
quences. Bot. Mag. 105: 673–679.
Hickey, L. J ., and Taylor, D. W. (1996). Origin of the angiosperm
flower. In ‘‘Flowering Plant Origin, Evolution and Phylogeny’’ (D.
W. Taylor and L. J . Hickey, Eds.), pp. 176–231. Chapman and Hall,
Hori, H., Lim, B.-L., and Osawa, S. (1985). Evolution of green plants
as deduced from 5S rRNA sequences. Proc. Natl. Acad. Sci. USA
Kenrick, P., and Crane, P. R. (1997). ‘‘The Origin and Early Diversifi-
cation of Land Plants: A Cladistic Study,’’ Smithsonian Institution,
Kirchner, M. (1992). Untersuchungen an einigen Gymnospermen der
fra ¨nkischen Rha ¨t-Lias-Grenzschichten. Palaeontographica Abt. B
Loconte, H., and Stevenson, D. W. (1990). Cladistics of the Sper-
matophyta. Brittonia 42: 197–211.
Logan, K. J ., and Thomas, B. A. (1985). Distribution of lignin
derivatives in plants. New Phytol. 99: 571–585.
Martens, P.(1971).‘‘LesGne ´tophytes,’’Encyclopedia ofPlantAnatomy,
Vol. 12, No. 2, Borntraeger, Stuttgart.
Meeuse, A. D. J . (1972). Taxonomic affinities between Piperales and
Polycarpicaeandtheir implicationsin interpretativefloral morphol-
ogy. In ‘‘Advances in Plant Morphology’’ (Y. S. Murty, B. M. J ohri,
H. Y. Mohan Ram, and T. M. Varghese, Eds.), pp. 3–27. Sarita
Prakashan, Meerut, India.
Nixon, K. C., Crepet, W. L., Stevenson, D., and Friis, E. M. (1994). A
reevaluation of seed plant phylogeny. Ann. Mo. Bot. Gard. 81:
Norell, M.A. (1992). Taxic origin and temporal diversity: Theeffect of
phylogeny. In ‘‘Extinction and Phylogeny’’ (M. J . Novacek and Q. D.
Wheeler, Eds.), pp. 88–118. Columbia Univ. Press, New York.
Palmer, A. R. (1983). The Decade of North American Geology 1983
geologic timescale. Geology 11: 503–504.
Raubeson, L.A., and J ansen, R. K. (1992). Chloroplast DNA evidence
on the ancient evolutionary split in vascular land plants. Science
Retallack, G., and Dilcher, D. L. (1981). Arguments for a glossopterid
ancestry of angiosperms. Paleobiology 7: 54–67.
Rothwell, G. R., and Serbet, R. (1994). Lignophytephylogeny and the
evolution of spermatophytes: A numerical cladistic analysis. Syst.
Bot. 19: 443–482.
Smith,A. B., and Littlewood, D. T. J . (1994). Paleontological data and
molecular phylogenetic analysis. Paleobiology 20: 259–273.
Stebbins, G. L. (1974). ‘‘Flowering Plants: Evolution above the
Species Level,’’ Harvard Univ. Press, Cambridge, MA.
Soltis, D. E., Soltis, P. S., Nickrent, D. L., J ohnson, L.A., Hahn, W. J .,
Hoot, S. B., Sweere, J . A., Kuzoff, R. K., Kron, K. A., Chase, M. W.,
Swenson, S. M., Zimmer, E. A., Chaw, S.-M., Gillespie, L. J ., Kress,
W. J ., and Sytsma, K. J . (1997). Angiosperm phylogeny inferred
from 18S ribosomal DNA sequences. Ann. Mo. Bot. Gard. 84: 1–49.
Stefanovic, S. (1996). ‘‘Phyloge ´nie Mole ´culaire et Relations des
Familles de Conife ´rales Base ´e sur le Se ´quenc ¸age de l’ADN Ribo-
somique28S,’’Diplo ˆmed’E´tudesApprofondies, Universite ´deParis.
Stewart, W. N., and Rothwell, G. W. (1993). ‘‘Paleobotany and the
Evolution of Plants,’’ 2nd ed., Cambridge Univ. Press, Cambridge,
Taylor, E. L., and Taylor, T. N. (1992). Reproductive biology of the
Permian Glossopteridales and their suggested relationship to
flowering plants. Proc. Natl. Acad. Sci. USA 89: 11495–11497.
Troitsky,A. V., Melekhovets,Yu. F., Rakhimova, G. M., Bobrova, V. K.,
Valiejo-Roman, K. M., andAntonov,A. S. (1991).Angiosperm origin
and early stages of seed plant evolution deduced from rRNA
sequencecomparisons. J . Mol. Evol. 32: 253–261.
Van Konijnenburg-van Cittert, J . H. A. (1992). An enigmatic Liassic
microsporophyll, yielding Ephedripites pollen. Rev. Palaeobot.
Palynol. 71: 239–254.
Wettstein, R. R. von (1907). ‘‘Handbuch der Systematischen Botanik,’’
Vol. 2, Franz Deuticke, Leipzig/Vienna.
J AMES A. DOYLE