Systematic Botany (2000), 25(4): pp. 668–691
q Copyright 2000 by the American Society of Plant Taxonomists
Phylogenetic Relationships in the Commelinaceae: I. A Cladistic
Analysis of Morphological Data
TIMOTHY M. EVANS1
Department of Botany, University of Wisconsin, 430 Lincoln Drive, Madison, Wisconsin 53706
Present address, author for correspondence: Department of Biology, Hope College, 35 East 12th Street,
Holland, Michigan 49423-9000
1
ROBERT B. FADEN
Department of Botany, NHB 166, National Museum of Natural History, Smithsonian Institution,
Washington, DC 20560
MICHAEL G. SIMPSON
Department of Biology, San Diego State University, San Diego, California 92182
KENNETH J. SYTSMA
Department of Botany, University of Wisconsin, 430 Lincoln Drive, Madison, Wisconsin 53706
Communicating Editor: Richard Jensen
ABSTRACT. The plant family Commelinaceae displays a wide range of variation in vegetative, floral, and
inflorescence morphology. This high degree of variation, particularly among characters operating under
strong and similar selective pressures (i.e., flowers), has made the assessment of homology among morphological characters difficult, and has resulted in several discordant classification schemes for the family. Phylogenetic relationships among 40 of the 41 genera in the family were evaluated using cladistic analyses of
morphological data. The resulting phylogeny shows some similarity to the most recent classification, but with
some notable differences. Cartonema (subfamily Cartonematoideae) was placed basal to the rest of the family.
Triceratella (subfamily Cartonematoideae), however, was placed among genera within tribe Tradescantieae of
subfamily Commelinoideae. Likewise, the circumscriptions of tribes Commelineae and Tradescantieae were
in disagreement with the most recent classification. The discordance between the phylogeny and the most
recent classification is attributed to a high degree of convergence in various morphological characters, particularly those relating to the androecium and the inflorescence. Anatomical characters (i.e., stomatal structure), on the other hand, show promise for resolving phylogenetic relationships within the Commelinaceae,
based upon their agreement with the most recent classification.
The Commelinaceae, a well defined family of 41
genera and about 650 species, is characterized by
several features including a distinct closed leaf
sheath, a succulent leaf blade, and three-merous
flowers with distinct petals and sepals (Cronquist
1981; Faden 1985; Faden and Hunt 1991). The genera are largely tropical and subtropical, but several
extend into temperate regions. The greatest diversity is in Africa, where, along with Madagascar,
nearly half the genera and about 40% of the species
are found (Faden 1983a).
There is a natural division of taxa between the
Old and the New World. The seven subtribes of
tribe Tradescantieae (sensu Faden and Hunt 1991)
are each wholly confined to either the Eastern or
Western hemisphere, whereas the tribe Commeli-
neae is found mainly in Africa and Asia. Only six
genera (Aneilema, Buforrestia, Commelina, Floscopa,
Murdannia, and Pollia) have indigenous species in
both hemispheres.
Older classifications of the Commelinaceae relied
heavily on floral features (see Faden and Hunt 1991,
for a review). Woodson (1942) first used inflorescence characters for higher-level classification, but
he was largely unfamiliar with the Old World genera. Pichon (1946) employed anatomical characters
to separate the genus Cartonema as the family Cartonemataceae, but he returned to more traditional
characters to define his ten tribes of Commelinaceae. Brenan (1966) used a variety of morphological
characters to define 15 informal ‘‘groups’’ of genera. These have served as the basis of surveys of
668
2000]
669
EVANS ET AL.: COMMELINACEAE
anatomical (Tomlinson 1966, 1969) and cytological
(Jones and Jopling 1972) characters in the family,
but they have little taxonomic significance.
The most recent classification, which will be used
in this study, was put forward by Faden and Hunt
(1991). In their classification, the Commelinaceae is
divided into two unequal subfamilies (Table 1).
Subfamily Cartonematoideae contains the tribes
Cartonemateae and Triceratelleae, each comprising
a single genus. Subfamily Commelinoideae contains the remaining 38 genera that are arranged in
two tribes, Tradescantieae and Commelineae, the
former of which is divided into seven subtribes.
Evans (1995) conducted a cladistic analysis of the
Commelinaceae using both morphological and molecular characters. This paper represents a modification and expansion of the morphological components of that study.
Relationships of Commelinaceae to other Monocot Families. The family Commelinaceae is the
namesake for the order Commelinales. Cronquist
(1981) defined the order by the presence of perfect
flowers that are adapted for general insect pollination, having showy petals that are differentiated
from the sepals. He included three other families in
this order: Xyridaceae, Rapateaceae, and Mayacaceae. He separated the Commelinaceae from the
other three families by their well-defined and
closed leaf sheath, and the succulent blade.
Dahlgren et al. (1985) included the same families
in Commelinales as Cronquist, but with the addition of the Eriocaulaceae. Eriocaulaceae was placed
in its own order by Cronquist, who maintained that
it was most likely derived within Xyridaceae. Dahlgren et al. (1985) distinguished the Commelinaceae
from the other commelinoid families by the presence of raphides, an amoeboid tapetum, and the
leaf characters used by Cronquist (closed leaf
sheath and succulent blade).
Although cladistic analyses of morphological
data support a monophyletic Commelinales sensu
Dahlgren et al. (1985; Stevenson and Loconte 1995),
recently published molecular data suggest otherwise. Nucleotide sequence data from the chloroplast
gene rbcL (Chase et al. 1993; Clark et al. 1993; Duvall et al. 1993) place the Commelinaceae near the
Pontederiaceae, Philydraceae, and Haemodoraceae,
all of which are part of a larger clade containing
the families of the order Zingiberales. The Rapateaceae (the only other representative of Commelinales
included in the previous molecular studies) is
placed in a different clade that contains, among
other families, Poaceae and Cyperaceae. Combined
molecular and morphological studies are similar to
the molecular phylogenies (e.g., Chase et al. 1995;
Linder and Kellogg 1995). Recent work by Givnish,
Evans, Pires, and Sytsma (Givnish et al. 1995, 1999)
generally supports these findings. It also places the
Hanguanaceae as the sister family of the Commelinaceae and the Xyridaceae and Eriocaulaceae in
the grass/sedge clade.
Certain morphological characters also lend support to the separation of the Commelinaceae from
other families of the original Commelinales (sensu
Cronquist 1981, or Dahlgren et al. 1985). As discussed above, Dahlgren et al. (1985) separated the
Commelinaceae from the other commelinoid families in part by the presence of raphides, which are
absent in the Mayacaceae, Rapateaceae, Xyridaceae,
and Eriocaulaceae (Dahlgren and Clifford 1982).
Raphides are widespread, however, in the Pontederiaceae, Philydraceae, and Haemodoraceae (Dahlgren et al. 1985). Additionally, whereas the Commelinaceae were separated from other families in
the Commelinales by the presence of an amoeboid
tapetum, the genus Pontederia and all investigated
members of the Haemodoraceae also have an amoeboid tapetum (Dahlgren and Clifford 1982; Simpson 1988, 1990). Finally, flowers in the Commelinaceae show a strong tendency toward zygomorphy, whereas they are generally actinomorphic in
the other members of the Commelinales sensu
Cronquist (1981) or Dahlgren et al. (1985). Zygomorphic flowers are common, however, in the Pontederiaceae, Philydraceae, and Haemodoraceae.
Although numerous taxonomic treatments have
been produced for the Commelinaceae, there has
been little consensus as to which characters should
be used to define relationships among the genera.
The earlier systems relied heavily upon features of
the highly variable androecium, but they ignored
characters of the inflorescence. The classification
produced by Faden and Hunt (1991), which will be
used throughout this study, incorporated a broad
range of information, such as characters from morphology, anatomy, and palynology. None of these
classifications, however, is rooted in an evolutionary
framework. The purpose of this study was to examine the intergeneric relationships in the Commelinaceae using morphological and anatomical
characters in a cladistic analysis.
MATERIALS
AND
METHODS
Operational Taxonomic Units (OTU’s). A critical step in any cladistic analysis is the selection of
670
SYSTEMATIC BOTANY
[Volume 25
TABLE 1. Classification of the Commelinaceae (as presented by Faden and Hunt 1991). *indicates Old World subtribe,
**indicates New World subtribe.
Family: Commelinaceae R. Br.
Subfamily: Cartonematoideae (Pichon) Faden & D. Hunt
Tribe: Cartonemateae (Pichon) Faden & D. Hunt
Cartonema R. Brown
Tribe: Triceratelleae Faden & D. Hunt
Triceratella Brenan
Subfamily: Commelinoideae (Brückner) Faden & D. Hunt
Tribe: Tradescantieae (Meisner) Faden & D. Hunt
Subtribe: *Palisotinae Faden & D. Hunt
Palisota Reichb.
Subtribe: *Streptoliriinae Faden & D. Hunt
Streptolirion Edgew.
Spatholirion Ridley
Aetheolirion Forman
Subtribe: *Cyanotinae (Pichon) Faden & D. Hunt
Cyanotis D. Don (Including
Amischophacelus R. Rao & Kamm.)
Belosynapsis Hassk.
Subtribe: *Coleotrypinae Faden & D. Hunt
Coleotrype C. B. Clarke
Porandra D. H. Hong
Amischotolype Hassk.
Subtribe: **Dichorisandrinae (Pichon) Faden & D. Hunt
Dichorisandra Mikan
Siderasis Raf.
Geogenanthus Ule
Cochliostema Lem.
Undescribed genus
Subtribe: **Thyrsantheminae Faden & D. Hunt
Thyrsanthemum Pichon
Gibasoides D. Hunt
Tinantia Scheidw.
Elasis D. Hunt
Matudanthus D. Hunt
Weldenia Schult. f.
Subtribe: **Tradescantiinae Rohw.
Gibasis Raf.
Tradescantia L. (Including Setcreasea Schumann & Sydow, Separotheca Waterf., Cymbispatha Pichon, Campelia Rich., Rhoeo Hance, Zebrina Schnizl.)
Callisia Loefl. (Including Hadrodemas H. Moore, Cuthbertia Small,
Aploleia Raf., Leiandra Raf., Phyodina Raf., Leptorhoeo C. B. Clarke ex Hemsley)
Tripogandra Raf.
Sauvallea Wright
Tribe: Commelineae (Meisner) Faden & D. Hunt
Stanfieldiella Brenan
Floscopa Lour.
Buforrestia C. B. Clarke
Murdannia Royle
Anthericopsis Engl.
Tricarpelema J. K. Morton
Pseudoparis H. Perrier
Polyspatha Benth.
Dictyospermum Wight
Pollia Thunb.
Aneilema R. Brown
Rhopalephora Hassk.
Commelina L. (Including Phaeosphaerion Hassk., Commelinopsis Pichon)
2000]
EVANS ET AL.: COMMELINACEAE
evolutionary units, or operational taxonomic units
(OTU’s). It is assumed that the units being examined have all descended from a most recent common ancestor, and represent distinct evolutionary
entities. If the OTU’s are polyphyletic, then inaccurate (‘‘misleading’’) reconstructions of their evolutionary history may be produced (Nixon and Davis 1991; Donoghue 1994).
Of the 41 genera in the Commelinaceae, ten are
monotypic (Aetheolirion, Anthericopsis, Elasis, Gibasoides, Matudanthus, Sauvallea, Streptolirion, Tapheocarpa Conran, Triceratella, and Weldenia), and there is
no difficulty in treating each as an OTU. For each
of the remaining 31 genera, however, claims of
monophyly might be misleading. The genus Callisia, for example, as circumscribed by Faden and
Hunt (1991), contains 20 species in seven sections
(Hunt 1986b). There is no single morphological
character that defines the genus, and its questionably monophyletic status (Hunt 1986b) might produce misleading relationships in a cladistic analysis.
Likewise, the genus Tradescantia sensu Faden and
Hunt (1991) is large (containing about 70 species)
and morphologically diverse. Some members of this
genus until recently have been placed into as many
as six other genera (Setcreasea, Separotheca, Cymbispatha, Campelia, Rhoeo, and Zebrina). Hunt (1975,
1980, 1986a) has gradually lumped all of these genera into Tradescantia, but the monophyly of the genus is yet to be demonstrated.
Although it is clear that large genera such as Callisia and Tradescantia must be treated cautiously in
a cladistic analysis, the same difficulties may also
arise for smaller, more ‘‘clearly defined’’ genera.
Rhopalephora, for example, is distinguished from
Aneilema by a combination of characters: inflorescence axis very short, stamen filaments fused basally, ovary and capsule densely covered by hook
hairs, and dorsal capsule valve deciduous (Faden
1977). All of these characters occur within Aneilema,
but not in this combination, and no single morphological character is unique to Rhopalephora. The implied hierarchical relationships between the two
genera suggests one or the other of them may be
paraphyletic.
To avoid the problems discussed above, each
OTU should be treated in one of three ways: 1) it
should be clearly demonstrated to be monophyletic;
2) if polymorphic for any character, it should be
split into smaller groups that are monomorphic
(Nixon and Davis 1991); or 3) it should be represented in the analysis by exemplar species (Simp-
671
son 1990; see also Bininda-Emonds et al. 1998).
Some of the difficulties of addressing the first option (monophyly) are discussed above. The theoretical and practical difficulties in dividing all polymorphic OTU’s into monomorphic units is discussed in some detail in Donoghue (1994).
Therefore, in this analysis, a single species was
scored for each genus where possible (Table 2).
Data for individual species of four genera (Porandra,
Pseudoparis, Rhopalephora, and an undescribed genus) were unavailable, so they were left as polymorphic to reflect variation within each genus. An
attempt was made, whenever possible, to select the
same species being used in a complementary molecular cladistic analysis of the Commelinaceae
(Evans et al., in prep.) so that the results from the
two data sets would be more directly comparable.
On the basis of the recent molecular studies on
monocots (Chase et al. 1993; Clark et al. 1993; Duvall et al. 1993; Givnish et al. 1999), representatives
of the families Pontederiaceae (Heteranthera) and
Haemodoraceae (Haemodorum) were used as outgroups. These two taxa were selected due to their
putatively basal position in their respective families
(Simpson 1990; Graham et al. 1998) and availability
of morphological data. Forty ingroup and two outgroup genera or exemplars were included in this
study (Appendix 1, 2). The recently described genus Tapheocarpa Conran (1994) was not included
due to insufficient data.
Character Selection. Characters were scored
from both living and pressed plants housed at the
US National Herbarium (US). Observations were
made on living plant material growing in the
Smithsonian Institution Botany Research Greenhouses and in the field. Where living or pressed
plants were not available for examination, information was taken from published literature (Table
2). Forty seven characters were scored.
Phylogenetic Analysis. Data were entered into
a matrix using the computer program MacClade
version 3.05 (Maddison and Maddison 1992), and
the phylogenetic analyses were performed with
PAUP* version 4.0b2a (Swofford 1999). A total of
3.7 percent of the cells of the matrix were scored as
‘‘missing,’’ either because the data were unavailable
or because particular characters were not applicable
to certain taxa.
In an initial analysis, all characters were treated
as unordered. A multiple-islands approach was
used to find the most parsimonious trees (modified
from Olmstead et al. 1993; see also Maddison 1991;
Olmstead and Palmer 1994). To evaluate the sup-
672
[Volume 25
SYSTEMATIC BOTANY
TABLE 2. Primary literature sources for information at the species level used in a cladistic analysis of morphological
characters in the Commelinaceae.
Species
Aetheolirion stenolobium Forman
Amischotolype hispida (Lesson & A. Rich.) D. Y. Hong
Anthericopsis sepalosa (C. B. Clarke) Engl.
Aneilema calceolus Brenan
Belosynapsis kewensis Hassk.
Buforrestia mannii C. B. Clarke
Callisia elegans Alexander ex. H. E. Moore
Cartonema philydroides F. Muell.
Cochliostema odoratissimum Lemaire
Coleotrype natalensis C. B. Clarke
Commelina benghalensis L.
Cyanotis barbatus D. Don
Dichorisandra hexandra (Aubl.) Standl.
Dictyospermum montanum Wight
Elasis hirsuta Hunt
Floscopa africana (P. Beauv.) C. B. Clarke
Geogenanthus poeppigii (Miq.) Faden
Gibasis geniculata (Jacq.) Rohweder
Gibasoides laxiflora (C. B. Clarke) Hunt
Matudanthus nanus (Martens & Gal.) Hunt
Murdannia edulis (Stokes) Faden
Palisota ambigua (P. Beauv.) C. B. Clarke
Pollia mannii C. B. Clarke
Polyspatha paniculata Benth.
Porandra
Pseudoparis
Rhopalephora
Sauvallea
Siderasis fuscata (Lodd.) H. E. Moore
Spatholirion longifolium (Gagnep.) Dunn
Stanfieldiella imperforata (C. B. Clarke) Brenan
Streptolirion volubile Edgew.
Thyrsanthemum floribundum (Martens & Galeotti) Pichon
Tinantia erecta (Jacq.) Schltdl.
Tradescantia virginiana L.
Tricarpelema gigantea (Hassk.) Hara
Triceratella drummondii Brenan
Tripogandra amplexicaulis (Klotzsch) Woodson
Weldenia candida Schultes s.
Undescribed Genus (Faden, in prep.)
port for each node, 100 bootstrap replicates were
conducted. Due to the high number of trees produced, a maximum of 5000 trees per replicate was
saved.
As a measure of which characters are more phylogenetically informative and which are more homoplasious (with respect to the topologies of the
trees produced), as well as to arrive at a more stable
topology, the characters were re-weighted a posteriori (using the ‘‘REWEIGHT CHARACTERS’’ command
Source
Forman 1962
Clarke 1881
Brenan 1966; Faden, unpubl. data
Faden 1991
Clarke 1881; Faden, unpubl. data
Brenan 1960; Faden, unpubl. data
Moore 1958
Brenan 1966; Faden, unpubl. data
Read 1965; Faden, unpubl. data
Obermeyer and Faden 1985
Obermeyer and Faden 1985
Brenan 1966
Matuda 1955
Morton 1966; Faden, unpubl. data
Hunt 1978; Faden, unpubl. data
Berhaut 1988
Moore 1954
Hunt 1985
Hunt 1978; Faden, unpubl. data
Hunt 1978; Faden, unpubl. data
Faden 1980
Clarke 1881
Brenan 1966
Clarke 1881; Faden, unpubl. data
Hong 1974; Faden, unpubl. data
Faden, unpubl. data
Faden 1975, 1977
Faden, unpubl. data
Clarke 1881 (as Pyrrhema); Faden, unpubl. data
Faden 1985
Brenan 1960
Forman 1962
Hunt 1993
Hunt 1993
Anderson and Woodson 1935
Morton 1966; Faden, unpubl. data
Brenan 1961
Hunt 1993
Hunt 1993; Faden, unpubl. data
Faden, unpubl. data
in PAUP) based on their fit to the trees produced
in the unordered analysis. The multi-step analysis
described above was then repeated using the reweighted characters. This process was repeated until the same set of most parsimonious trees was
produced in two consecutive runs. PAUP allows the
character to be re-weighted based on consistency
index (CI), retention index (RI), or rescaled consistency index (RC). The maximal value of each of
these indices was used in three separate analyses.
2000]
673
EVANS ET AL.: COMMELINACEAE
In addition to the unordered analysis, an analysis
was performed in which three character transformations were ordered (fruit locule number [Character 35] and seed number per locule [Characters
37 and 38]). The same procedure described above
was used in this analysis to find multiple islands
of most parsimonious trees.
For purposes of character state mapping, one of
the most parsimonious trees was arbitrarily selected for illustration. However, due to the lack of resolution of several clades in the strict consensus tree,
distributions were examined on all of the most parsimonious trees.
RESULTS
Unordered Analysis. In the unordered analysis,
154 equally most parsimonious trees were produced, with a length (excluding autapomorphies) of
239 steps and a consistency index (CI) of 0.43
(length 5 272 and CI 5 0.44 when autapomorphies
are included). Figure 1 illustrates the tree chosen
for character state reconstructions.
Cartonema is sister to the rest of Commelinaceae.
The remaining genera form four lineages (Fig. 1):
Callisia, which is sister to the remaining genera;
Weldenia; and two largely unresolved clades. Within
the clade that is sister to Anthericopsis, 15 genera
(Buforrestia, Floscopa, Pseudoparis, Tricarpelema, Polyspatha, Aneilema, Commelina, Rhopalephora, Dictyospermum, Palisota, Tinantia, Cochliostema, Aetheolirion,
Geogenanthus, and an undescribed genus) form a
clade that is supported by two characters, antesepalous staminal filament lengths (Character 26),
and the presence of a zygomorphic androecium
(Character 31; note: Floscopa produces an asymmetric androecium). For ease of reference, this clade
will be referred to as the ‘‘zygomorphic clade.’’ The
remaining genera (Murdannia, Gibasis, Elasis, Thyrsanthemum, Gibasoides, Matudanthus, Coleotrype, Porandra, Amischotolype, Cyanotis, Belosynapsis, Triceratella, Streptolirion, Spatholirion, Sauvallea, Tradescantia,
and Tripogandra) form a large clade that is supported by the presence of bearded antesepalous
staminal filaments (Characters 22, 23). For ease of
reference, this clade will be referred to as ‘‘Clade
1.’’
A posteriori Re-weighting. Each of the reweighted analyses (CI-, RI-, and RC-based; Table 3)
produced topologies that were similar to the unordered tree. The RI-based analysis produced 45
equally most parsimonious trees (not shown) that
were a subset of the 154 trees found in the unor-
dered analysis. The CI- and RC-based analyses produced identical sets of 15 equally most parsimonious trees (not shown) that differed from the unordered topology in the lineages sister to the zygomorphic clade. In the unordered tree (Fig. 1), the
zygomorphic clade has a sister group consisting of
Stanfieldiella, Pollia, Siderasis, and Dichorisandra. The
CI- and RI-based analyses produced a set of nested
relationships from the more basal Siderasis to, sequentially, Pollia, Stanfieldiella, Anthericopsis, Dischorisandra, and the zygomorphic clade.
Ordered Character Analysis. When assumptions about the sequence of evolutionary steps in
characters 35, 37, and 38 were incorporated, 115
equally most parsimonious trees were produced
(not shown). This topology is nearly identical to the
unordered topology, except that it showed greater
resolution in the zygomorphic clade. The undescribed genus and Cochliostema were united in one
clade and Polyspatha, Aneilema, and Commelina
formed an unresolved clade.
DISCUSSION
The discordant classifications that have been proposed for the Commelinaceae reflect the high degree of uncertainty in homology among morphological characters in the family. Likewise, the low
CI values for many characters in this study (Table
3) demonstrate a high amount of homoplasy in the
family. The cladistic analysis presented here, although supporting particular elements of some previous classifications, does not agree closely with
any one of them. The zygomorphic clade (Fig. 1)
includes all but three genera (Pollia, Stanfieldiella,
and Murdannia) of Faden and Hunt’s (1991) tribe
Commelineae. However, several members of Tradescantieae are also included in this clade. Likewise, few elements of Brenan’s (1966) classification
are supported. Of the 15 informal ‘‘groups’’ put forward by Brenan, one, ‘‘Group VI’’ (5 subtribe Cyanotinae), is monophyletic here. Relatively narrow
circumscriptions of several of Brenan’s groups (i.e.,
the inclusion of only one or two genera per group)
make a direct comparison of relationships among
those genera difficult. Pichon’s (1946) classification
included ten tribes, five of which contain only a
single genus. Of the remaining five tribes, not one
is supported by this analysis. The large number of
taxonomic groups with only a single genus in Pichon’s and Brenan’s systems, as well as lack of statistical support for most clades in this analysis, suggest that there is a high degree of uncertainty in
674
SYSTEMATIC BOTANY
[Volume 25
FIG. 1. A representative of the 154 equally most parsimonious trees found in the unordered analysis of 47 morphological characters in Commelinaceae using Haemodorum and Heteranthera as outgroups. Dashed lines represent branches
that collapse in the strict consensus of most parsimonious trees. Bootstrap values greater than 50% are shown above
branches. Subtribal and tribal affinities are shown to right of generic names. ‘‘Clade 1’’ and the ‘‘zygomorphic clade’’
discussed in the text are indicated.
assessing homology of structures among genera of
Commelinaceae.
Basal Lineages in the Commelinaceae. Each
analysis divided the Commelinaceae into several
main lineages, with Cartonema sister to the rest of
the family and Callisia sister to everything except
Cartonema. Callisia is a member of subtribe Tradescantiinae, a well defined group based on features of the inflorescence (cincinni contracted,
fused in bifacial pairs; Faden and Hunt 1991). The
New World distribution of this entire subtribe, as
well as its putatively recent origin (Faden and
2000]
675
EVANS ET AL.: COMMELINACEAE
TABLE 3. Character statistics from each analysis. CI 5 consistency index; RI 5 retention index; RC 5 rescaled
consistency index; WCI 5 character weight after re-scaling according to the CI; WRI 5 character weight after rescaling
according to the RI; WRC 5 character weight after re-scaling according to the RC.
Weight
Character
States
Steps
CI
RI
RC
WCI
WRI
WRC
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
01
01234
01
012
01
012
01
01
012
01
01
01
01
01
01
01
01
01
01
01
01
01
01
01
01
012
012
012
012
01
012
012
01
012
012
01
0123
0123
0123
01
01
01
0123
01
01
012
01
3
13
9
6
2
7
5
4
7
3
4
5
3
2
2
3
1
1
2
3
3
3
3
1
3
2
2
7
9
2
5
3
7
3
4
6
7
8
11
6
2
5
9
3
2
2
1
.333
.308
.111
.333
.500
.286
.200
.250
.286
.333
.250
.200
.333
.500
.500
.333
1.00
1.00
.500
.333
.333
.333
.333
1.00
.333
1.00
1.00
.286
.222
.500
.400
.667
.143
.333
.500
.167
.286
.250
.273
.167
.500
.200
.333
.333
.500
1.00
1.00
.500
.308
.429
.556
0.00
.444
.636
.400
.375
.600
0.00
.429
.333
.500
.500
.500
—
1.00
.500
.333
.778
.882
.882
1.00
.889
1.00
1.00
.167
.222
0.00
.812
0.00
0.00
.667
.333
.444
.737
.667
.500
.500
.500
.750
.667
.600
.667
1.00
1.00
.167
.095
.048
.185
0.00
.127
.127
.100
.107
.200
0.00
.086
.111
.250
.250
.167
—
1.00
.250
.111
.259
.294
.294
1.00
.296
1.00
1.00
.048
.049
0.00
.325
0.00
0.00
.222
.167
.074
.211
.167
.136
.083
.250
.150
.222
.200
.333
1.00
1.00
333
307
100
286
500
286
200
250
286
333
250
200
333
500
500
333
1000
500
1000
333
250
500
500
1000
250
1000
1000
333
222
500
500
667
143
333
500
125
286
250
273
200
500
167
375
333
500
1000
1000
286
357
400
500
0
500
583
286
375
800
0
375
667
333
333
500
1000
1000
1000
333
778
889
889
1000
842
1000
1000
333
333
0
813
0
0
667
333
444
609
545
421
455
500
750
722
800
667
1000
1000
167
94
36
127
0
127
127
100
107
200
0
86
111
250
250
167
1000
417
1000
111
167
471
471
1000
208
1000
1000
111
49
0
438
0
0
222
167
28
211
167
136
120
250
115
271
200
333
1000
1000
676
SYSTEMATIC BOTANY
Hunt 1991) shed doubt on the basal position of
Callisia.
The position of Cartonema sister to the rest of the
family is consistent with other data, particularly
rbcL sequences (Evans 1995; Evans et al., in prep.).
On the basis of various morphological characters
(e.g., actinomorphic flowers with six fertile stamens,
the absence of raphide canals), Cartonema has been
postulated to be one of the more primitive members of the family (Pichon 1946; Hutchinson 1959;
Brenan 1966; Faden and Hunt 1991).
Relationships in the Tradescantieae. With the
exceptions of Murdannia (tribe Commelineae) and
Triceratella (subfamily Cartonemateae), Clade 1 consists exclusively of members of Tradescantieae (Fig.
1). Several members of the tribe, however, are
placed in other parts of the tree. As discussed
above, Callisia, for example, is basal to all other genera except Cartonema. Likewise, eight genera are
placed either sister to the zygomorphic clade (Dichorisandra and Siderasis) or within it (Palisota, Tinantia, undescribed genus, Cochliostema, Aetheolirion,
and Geogenanthus). The position of Dichorisandra
and Siderasis as the sister group of Pollia (tribe Commelineae) is supported by the presence of biseriate
seeds (Character 36). This character, however, is
highly variable within the family. Within the zygomorphic clade and its sister clade, biseriate seed
arrangement has arisen a minimum of three times
(Fig. 2). This character is even more variable in other most parsimonious trees, some of which force
Aetheolirion and Geogenanthus to gain biseriate seeds
independently. Likewise, relative filament length
within the antepetalous staminal whorl (Character
27) is the single character that supports the clade
containing Aetheolirion, Cochliostema, Geogenanthus,
Tinantia, and the undescribed genus as the sister
group of the zygomorphic clade. This character too
is variable within the zygomorphic clade (not
shown).
The taxonomic placement of the genus Triceratella
has been problematic. When Triceratella was originally described, a loose affinity with Stanfieldiella
was suggested (Brenan 1961). Tomlinson (1964)
[Volume 25
used anatomical characters to suggest that the genus may form a link between Cartonema, which he
placed in a separate family, and the Commelinaceae. He indicated that it might be treated as a separate subfamily of Commelinaceae. Faden and Hunt
(1991) placed Triceratella and Cartonema in a subfamily within the Commelinaceae, with each genus
in a separate tribe.
The morphological data presented here place Triceratella as a sister group of subtribe Cyanotinae
(Cyanotis and Belosynapsis), nested well within the
Tradescantieae. The clade containing Belosynapsis,
Cyanotis, Sauvallea, Spatholirion, Streptolirion, and Triceratella is supported by a single character (sessile
flowers). The presence of sessile flowers is labile
within Commelinaceae, having been lost once and
gained at least three times (Fig. 2). Although a relationship between Triceratella and the Streptoliriinae has some support, it needs further study. Until
less homoplasious characters are found to support
such a relationship, it will remain suspect.
Of the seven subtribes circumscribed by Faden
and Hunt (1991), only two, Coleotrypinae and Cyanotinae, are supported as monophyletic in this
study. The remaining subtribes are polyphyletic
and largely unresolved. Two genera in subtribe
Thyrsantheminae (Thyrsanthemum and Elasis) contribute to a polytomy in Clade 1. Thyrsantheminae,
as defined by Faden and Hunt (1991), is based
largely upon the absence of characters or the presence of putatively primitive characters, and may be
an unnatural assemblage (R. Faden, unpubl. data).
Recent molecular data also suggest that this subtribe is polyphyletic (Evans et al., in prep.), and the
lack of resolution and weak support in this analysis
illustrate the difficulty in finding taxonomic characters that will unambiguously unite these taxa
with other genera.
Relationships in the Commelineae. Ten of the
thirteen genera in tribe Commelineae were placed
within the zygomorphic clade (Fig. 1). Of the three
remaining Commelineae genera, two (Pollia and
Stanfieldiella) are part of the sister group of the zygomorphic clade; Murdannia is nested within Clade
→
FIG. 2. Distribution of characters on the representative tree (see Fig. 1) from the unordered analysis of 47 morphological characters in Commelinaceae. Left: Seed arrangement. Biseriate seed arrangement has arisen at least three times
within the zygomorphic clade and its sister clade. One additional step is required in some other equally most parsimonious trees due to shifts in position of Aetheolirion and Geogenanthus. Right: Pedicel. The presence of pedicellate flowers
is homoplasious within Commelinaceae. The distribution of this character is not altered in any other most parsimonious
trees.
2000]
EVANS ET AL.: COMMELINACEAE
677
678
SYSTEMATIC BOTANY
1. The only synapomorphy that supports the zygomorphic clade is the presence of a zygomorphic
androecium (Character 31), a character that is variable within the family as a whole. However, all
members of this clade, except Palisota, share a common stomatal type consisting of six cells, with
small terminal subsidiary cells (Figs. 3 and 4). Although this stomatal type is restricted to genera belonging to the tribe Commelineae, it is not a synapomorphy for the zygomorphic group, as it is also
found in Pollia, Anthericopsis, and Murdannia. This
anatomical character is in agreement with a molecular phylogeny for Commelinaceae (Evans 1995;
Evans et al. 2000), and the distribution of stomatal
types here raises questions regarding the placement
of Murdannia within Clade 1, and calls for a reevaluation of the homology of the filament bearding
between Murdannia and the other genera in that
clade.
Although the unordered analysis produced a
highly unresolved zygomorphic clade, the reweighting analyses each yielded a smaller clade
within the zygomorphic group containing Polyspatha, Aneilema, Rhopalephora, Commelina, Dictyospermum, and Tricarpelema. This clade is supported by
a reduction in seed number in each locule (Tricarpelema is polymorphic for this character). As relationships in the tribe Commelineae have been difficult to evaluate due to a lack of informative characters (Faden and Hunt 1991), the formation of distinct clades based on fruit characteristics may
provide some insight as to where such characters
may be found.
A Comparison of Different Methods of Parsimony
Analysis. Several different methods of character
weighting and ordering were explored with this
data set, and the question arises as to which one(s)
are the best approximation of the true history of
the family. The unordered analysis contains the
fewest a priori assumptions about character evolution, as it incorporates few assumptions about character weight or direction of transformation. The absence of additional assumptions, however, imposes
a penalty in terms of the potential loss of phylogenetic information. For example, the number of
locules in the fruit (Character 35) may represent a
series, in which fruits with three unequal locules
(e.g., one locule is reduced) are intermediate between fruits with three locules and fruits with two
locules. By coding such characters as ordered, hypotheses about evolutionary trends may be incorporated in estimates of phylogenetic relationships.
In addition to ordering, characters were also re-
[Volume 25
weighted on the basis of three different measures
of their internal consistency (CI, RI, and RC). Each
of these re-weighting schemes starts with the unordered topology, and then re-evaluates the fit of
the characters to that topology. The CI, a measure
of the fit of the characters to the tree, is derived as
the ratio of the number of state changes for a character in the data to the number of state changes for
the character on the hypothesized tree (Kluge and
Farris 1969). A potential problem with the consistency index, however, is that it only measures how
well a character fits onto a particular topology; it
does not evaluate how well that character actually
supports the topology (i.e., if it is synapomorphic for
any clade). Farris (1989) developed the retention index (RI) and rescaled consistency index (RC) in an
attempt to incorporate the internal consistency of
each character with respect to the worst and best
possible fit to a tree. The RI and RC both incorporate the minimum number of steps a character must
undergo on any tree, as well as the actual amount
of synapomorphies on the hypothesized tree.
By using these values as bases for re-weighting
characters, slightly different properties of the characters are being selected. When CI is used, for example, each character is re-weighted according to
the number of steps it undergoes when mapped
onto the tree. However, this scheme does not take
into account whether the character supports any of
the clades in the tree. Incorporating RI and RC,
however, does take this into account. A good example of the differences in these measures can be
seen with Character #5 (number of flowers per cincinnus). This character has a CI of 0.5 (Table 3).
However, it does not unite any clades (Fig. 4), so it
lends no support to the topology. Because it does
not support a clade, it has an RI and RC value of
zero (Table 3). Therefore, the CI re-weighting
scheme assigned a relatively high weight to that
character whereas the RI and RC decreased its
weight.
While each of the re-weighting schemes produced phylogenies similar to the unordered phylogeny (RC- and CI-based analyses produced a subset of the unordered trees), they all showed greater
resolution than the unordered analysis. Whereas
the unweighted analysis produced a nearly completely unresolved zygomorphic clade, each of the
re-weighting analyses produced a highly resolved
zygomorphic clade. The re-weighting analyses differed from each other mainly in the relationships
of the sister taxa to the zygomorphic clade. These
differences reflect biases toward different charac-
2000]
EVANS ET AL.: COMMELINACEAE
679
FIG. 3. Diagrams depicting developmental pathways for the four stomatal types found in the Commelinaceae: stomata with two subsidiary cells, four subsidiary cells, and two types with six subsidiary cells. Abbreviations: t.m.c.,
terminal subsidiary cell mother cell; g.m.c., guard cell mother cell; l.m.c., lateral subsidiary cell mother cell; t.s.c., terminal
subsidiary cell; l.s.c. (1), innermost lateral subsidiary cell; l.s.c. (2), outermost lateral subsidiary cell (from Tomlinson
1966).
680
SYSTEMATIC BOTANY
[Volume 25
2000]
681
EVANS ET AL.: COMMELINACEAE
ters, depending on how well they support the initial (unordered) phylogeny (e.g., Character 5, number of flowers per cincinnus, as discussed above).
It is unclear which of these values (CI, RI, or RC)
will provide a more accurate reflection of the suitability of each character in reconstructing phylogeny, but they do provide information about which
characters are more homoplasious and/or more informative with respect to a particular data set and
topology. It is evident that the number of flowers
per cincinnus, for example, does not contribute to
the phylogenetic interpretation provided by the rest
of the data. Furthermore, if the data set as a whole
is misleading, then re-weighting the characters a
posteriori may reinforce the incorrect parts of the
phylogeny and reduce the support for some more
accurate areas.
The Commelinaceae are quite diverse morphologically, and, as has been demonstrated in this
study, many of the morphological characters in the
family are highly homoplasious (e.g., characters relating to pollination). Many classifications of the
family have been based in large part on these homoplasious characters, thereby creating unnatural assemblages (an obvious example of this is Brenan’s
[1966] ‘‘Group I’’, a diverse and artificial group
united mainly by inflorescence characters and a dehiscent fruit). Although relatively few anatomical
characters have been examined in great enough detail for a comprehensive anatomy-based cladistic
analysis, anatomical characters (e.g., stomatal characters; Fig. 3) appear to hold great promise for evaluating relationships at the tribal and, possibly, subtribal levels.
ACKNOWLEDGEMENTS. This paper represents a portion
of T. Evans’ Doctoral thesis in the University of Wisconsin
Department of Botany. We wish to thank committee members Tom Givnish, Bob Kowal, David Spooner, and Ted
Garland for reading an earlier draft of the manuscript.
Additionally, we thank Kandis Elliot for preparation of the
figures. The United States National Herbarium and Smithsonian Institution Botany Research Greenhouses provided
plant material. This work was supported in part by a
Graduate Student Research Grant from the American Society of Plant Taxonomists, and by a Short Term Visitor
Grant and Graduate Student Fellowship from the Smithsonian Institution.
LITERATURE CITED
ANDERSON, E. and R. E. WOODSON. 1935. The species of
Tradescantia indigenous to the United States. Contributions from the Arnold Arboretum of Harvard University 9: 1–132.
BININDA-EMONDS, O. R. P., H. N. BRYANT, and A. P. RUSSELL. 1998. Supraspecific taxa as terminals in cladistic
analysis; implicit assumptions of monophyly and a
comparison of methods. Biological Journal of the Linnean Society 64: 101–133.
BERHAUT, J. 1988. Pp. 130–131 in Flore Illustrée du Sénégal:
Monocotyledones et Pteridophytes. Dakar: Clairafrique.
BRENAN, J. P. M. 1960. Notes on African Commelinaceae:
II: The genus Buforrestia C. B. Cl. and a new related
genus, Stanfieldiella Brenan. Kew Bulletin 14: 280–286.
———. 1961. Triceratella, a new genus of Commelinaceae
from southern Rhodesia. Kirkia 1: 14–19.
———. 1966. The classification of Commelinaceae. Botanical Journal of the Linnean Society 59: 349–370.
CHASE, M. W., D. E. SOLTIS, R. G. OLMSTEAD, D. MORGAN,
D. H. LES, B. D. MISHLER, M. R. DUVALL, R. A. PRICE,
H. G. HILLS, Y-L. QIU, K. A. KRON, J. H. RETTIG, E.
CONTI, J. D. PALMER, J. R. MANHART, K. J. SYTSMA, H.
J. MICHAELS, W. J. KRESS, K. G. KAROL, W. D. CLARK,
M. HEDREN, B. S. GAUT, R. K. JANSEN, K-J. KIM, C. F.
WIMPEE, J. F. SMITH, G. R. FURNIER, S. H. STRAUSS, QY. XIANG, G. M. PLUNKETT, P. S. SOLTIS, S. M. SWENSEN, S. E. WILLIAMS, P. A. GADEK, C. J. QUINN, L. E.
EGUIARTE, E. GOLENBERG, G. H. LEARN, S. W. GRAHAM, S. C. H. BARRETT, S. DAYANANDAN, and V. A.
ALBERT. 1993. Phylogenetics of seed plants: an analysis of nucleotide sequences from the plastid gene
rbcL. Annals of the Missouri Botanical Garden 80:
528–580.
———, D. W. STEVENSON, P. WILKIN, and P. J. RUDALL.
1995. Monocot systematics: a combined analysis. Pp.
685–730 in Monocotyledons: systematics and evolution,
eds. P. J. Rudall, P. J. Cribb, D. F. Cutler, and C. J.
Humphries. Kew: Royal Botanic Garden.
←
FIG. 4. Distribution of characters on representative tree (see Fig. 1) from the unordered analysis of 47 morphological
characters in Commelinaceae. Left: Stomatal structure. This anatomical character may be useful for higher level relationships within the family. Although it is homoplasious here, its distribution is congruent with recent molecular data
(Evans 1995, in prep.). Note that the four-celled stomatal type is one step shorter in equally most parsimonious trees
that unite Palisota with the clade containing Cochliostema, Tinantia, and the undescribed genus. The number of steps for
the other stomatal types is not altered in other equally most parsimonious trees. Right: Number of flowers per cincinnus.
Although this character possesses a relatively high consistency index (CI 5 0.75), it does not unite any clades and thus
has a retention index of zero. The distribution of this character is not altered in any other most parsimonious trees.
682
SYSTEMATIC BOTANY
CLARK, W. D., B. S. GAUT, M. R. DUVALL, and M. T. CLEGG.
1993. Phylogenetic relationships of the BromelifloraeCommeliniflorae-Zingiberiflorae complex of monocots based on rbcL sequence comparisons. Annals of
the Missouri Botanical Garden 80: 987–998.
CLARKE, C. B. 1881. Commelinaceae. Pp. 113–324 in Monographiae phanerogamarum, 3, eds. A. De Candolle and
C. De Candolle. Paris: G. Masson.
CONRAN, J. G. 1994. Tapheocarpa (Commelinaceae), a new
Australian genus with hypogeous fruits. Australian
Systematic Botany 7: 585–589.
CRONQUIST, A. 1981. An integrated system of classification
of flowering plants. New York: Columbia University
Press.
DAHLGREN, R., and H. T. CLIFFORD. 1982. The monocotyledons: a comparative study. New York: Academic Press.
———, F. N. RASMUSSEN, and P. F. YEO. 1985. The families
of the monocotyledons: structure, function, and taxonomy. Berlin: Springer Verlag.
DONOGHUE, M. J. 1994. Progress and prospects in reconstructing plant phylogeny. Annals of the Missouri Botanical Garden 81: 405–418.
DUVALL, M. R., M. T. CLEGG, M. W. CHASE, W. D. CLARK,
W. J. KRESS, H. G. HILLS, L. E. EGUIARTE, J. F. SMITH,
B. S. GAUT, E. A. ZIMMER, and G. H. LEARN. 1993.
Phylogenetic hypotheses for the monocotyledons constructed from rbcL sequence data. Annals of the Missouri Botanical Garden 80: 607–619.
EVANS, T. M. 1995. A phylogenetic analysis of the Commelinaceae based on morphological and molecular
data. Ph.D. thesis, University of Wisconsin, Madison.
———, R. B. FADEN, and K. J. SYTSMA. 2000. Homoplasy
in the Commelinaceae: a comparison of different classes of morphological characters. Pp. 557–566 in Proceedings of the Second International Conference on the
Comparative Biology of the Monocots, Sydney, September
1998, eds. K. L. Wilson and D. A. Morrison. Melbourne: CSIRO.
FADEN, R. B. 1975. A biosystematic study of the genus
Aneilema R. Br. (Commelinaceae). Ph.D. thesis, Washington University, St. Louis.
———. 1977. The genus Rhopalephora Haask. (Commelinaceae). Phytologia 37: 479–481.
———. 1980. The taxonomy and nomenclature of some
Asiatic species of Murdannia (Commelinaceae): the
identity of Commelina medica Lour. and Commelina tuberosa Lour. Taxon 29: 71–83.
———. 1983a. Phytogeography of African Commelinaceae. Bothalia 14: 553–557.
———. 1983b. Floral dimorphism, pollen dimorphism and
floral biology in a West African Commelinaceae, Palisota hirsuta (Thunb.) K. Schum. American Journal of
Botany 70: 113.
———. 1985. Commelinaceae. Pp. 381–387 in The families
of the monocotyledons, eds. R. M. T. Dahlgren, H. T.
Clifford, and P. F. Yeo. Berlin: Springer.
———. 1988. Reproductive biology of Palisota Reichb., an
African endemic genus of Commelinaceae. Abstracts,
[Volume 25
A.E.T.F.A.T., Twelfth Plenary Meeting, 4–10 September, 1988. Hamburg. P. 58.
———. 1991. The morphology and taxonomy of Aneilema
R. Brown (Commelinaceae). Smithsonian Contributions to Botany 76: 1–166.
——— and D. R. HUNT. 1991. The classification of the
Commelinaceae. Taxon 40: 19–31.
——— and K. E. INMAN. 1996. Leaf anatomy of the African genera of Commelinaceae: Anthericopsis and Murdannia. Pp. 464–471 in The biodiversity of African plants,
eds. L. J. G. van der Maesen, X. M. van der Burgt,
and J. M. van Madenbach de Rooy. Procedings of the
XIVth A.E.T.F.A.T. Congress, 22–27 August, 1994,
Wageningen, The Netherlands. London: Kluwer Academic Publishers.
FARRIS, J. S. 1989. The retention index and rescaled consistency index. Cladistics 5: 417–419.
FORMAN, L. L. 1962. Aëtheolirion, a new genus of Commelinaceae from Thailand, with notes on allied genera. Kew Bulletin 16: 209–221.
GIVNISH, T. J., T. M. EVANS, and K. J. SYTSMA. 1995. The
Commelinales: an extreme example of convergent
evolution in monocots. American Journal of Botany
82(6, Supplement): 132.
———, ———, J. C. PIRES and K. J. SYTSMA. 1999. Polyphyly and convergent morphological evolution in
Commelinales and Commelinidae: evidence from
rbcL sequence data. Molecular Phylogenetics and Evolution 12: 360–385.
HONG, D. 1974. Revisio Commelinacearum Sinacarum.
Acta Phytotaxonomica Sinica 12: 459–488.
GRAHAM, S. W., J. R. KOHN, B. R. MORTON, J. E. ECKENWALDER, and S. C. H. BARRETT. 1998. Phylogenetic
congruence and discordance among one morphological and three molecular data sets from Pontederiaceae. Systematic Biology 47: 545–567.
HUNT, D. R. 1975. The reunion of Setcreasea and Separotheca
with Tradescantia. American Commelinaceae: I. Kew
Bulletin 30: 443–458.
———. 1978. Three new genera in Commelinaceae. Kew
Bulletin 33: 331–334.
———. 1980. Sections and series in Tradescantia. American
Commelinaceae: IX. Kew Bulletin 35: 437–442.
———. 1985. A revision of Gibasis Rafin.: American Commelinaceae XII. Kew Bulletin 41: 107–129.
———. 1986a. Campelia, Rheo, and Zebrina united with
Tradescantia. American Commelinaceae: XIII. Kew
Bulletin 41: 401–405.
———. 1986b. Amplification of Callisia Loefl. American
Commelinaceae: IX. Kew Bulletin 41: 407–412.
———. 1993. Commelinaceae. Pp. 130–201 in Flora NovoGaliciana, Vol. 13, Limnocharitaceae to Typhaceae, ed.
R. McVaugh. Ann Arbor: University of Michigan Herbarium.
HUTCHINSON, J. 1959. Commelinaceae, Cartonemataceae,
in Families of flowering plants, Second Edition. Volume
2, pp. 561–566. Oxford: Oxford University Press.
JONES, K. and C. JOPLINH. 1972. Chromosomes and the
2000]
683
EVANS ET AL.: COMMELINACEAE
classification of the Commelinaceae. Botanical Journal
of the Linnean Society 65: 129–162.
KLUGE, A. G. and J. S. FARRIS. 1969. Quantitative phyletics
and the evolution of anurans. Systematic Zoology 18:
1–32.
LEE, R. E. 1961. Pollen dimorphism in Tripogandra grandiflora. Baileya 9: 53–56.
LINDER, H. P. and E. A. KELLOGG. 1995. Phylogenetic patterns in the Commelinid clade. Pp. 473–496 in Monocotyledons: systematics and evolution, eds. P. J. Rudall, P.
J. Cribb, D. F. Cutler, and C. J. Humphries. Kew: Royal
Botanic Garden.
MADDISON, D. R. 1991. The discovery and importance of
multiple islands of most-parsimonious trees. Systematic Zoology 40: 315–348.
MADDISON, W. P. and D. R. MADDISON. 1992. MacClade:
Analysis of phylogeny and evolution. Version 3.0. Sunderland, Massachusetts: Sinauer Associates.
MATUDA, E. 1955. Las Commelinaceae Mexicanas. Sobretiro de los Anales del Instituto de Biologı́a 26: 303–
432.
MOORE, H. E. 1954. The Seersucker plant—Geogenanthus
undatus. Baileya 2: 41–45.
———. 1958. Callisia elegans, a new species, with notes on
the genus. Baileya 6: 135–147.
MORTON, J. K. 1966. A revision of the genus Aneilema R.
Brown (Commelinaceae) with a cytotaxonomic account of the West African species. Botanical Journal
of the Linnean Society 59: 431–479.
NIXON, K. C. and J. I. DAVIS. 1991. Polymorphic taxa, missing values and cladistic analysis. Cladistics 7: 233–
241.
OBERMEYER, A. and R. FADEN. 1985. Commelinaceae. Pp.
23–60 in Flora of southern Africa, Vol. 4, Pt. 2, ed. O. A.
Leistner. Cape Town: Government Printer.
OLMSTEAD, R. G, B. BREMER, K. M. SCOTT, and J. D. PALMER. 1993. A molecular systematic analysis of the Asteridae sensu lato based on rbcL sequences. Annals of
the Missouri Botanical Garden 80: 700–722.
——— and J. D. PALMER. 1994. Chloroplast DNA systematics: a review of methods and data analysis. American Journal of Botany 81: 1205–1224.
PICHON, M. 1946. Sur les Commélinacées. Notulae systematicae. Museum National d’histoire Naturelle. Phanérogamie. Paris 12: 217–242.
READ, R. W. 1965. Cochliostema velutinum (Commelinaceae):
a new species from Colombia. Baileya 13: 8–15.
SIMPSON, M. G. 1988. Embryological development of Lachnanthes caroliniana (Haemodoraceae). American Journal of Botany 75: 1394–1408.
———. 1990. Phylogeny and classification of the Haemodoraceae. Annals of the Missouri Botanical Garden
77: 722–784.
STEVENSON, D. W. and H. LOCONTE. 1995. Cladistic analysis of monocot families. Pp. 543–578 in Monocotyledons: systematics and evolution, eds. P. J. Rudall, P. J.
Cribb, D. F. Cutler, and C. J. Humphries. Kew: Royal
Botanic Garden.
SWOFFORD, D. L. 1999. Phylogenetic analysis using parsimony
(* and other methods) Version 4.0. Sunderland, Massachusetts: Sinauer Associates.
TOMLINSON, P. B. 1964. Notes on the anatomy of Triceratella
(Commelinaceae). Kirkia 4: 207–212.
———. 1966. Anatomical data in the classification of Commelinaceae. Botanical Journal of the Linnean Society
59: 371–395.
———. 1969. Commelinaceae. Pp. 12–63 in Anatomy of the
monocotyledons, 3. Ed. C. R. Metcalfe, Oxford: Clarendon Press.
WOODSON, R. E. 1942. Commentary on the North American genera of Commelinaceae. Annals of the Missouri
Botanical Garden 29: 141–154.
APPENDIX 1
Description and rationale for characters and coding used in this analysis. A brief justification for
the rationale behind the ordering scheme for characters 35 (fruit locule number), 37 (seed number per
dorsal locule), and 38 (seed number per ventral locule) is included. Although the characters were not
polarized a priori, they were scored so that those
states that have traditionally been believed to be
primitive in the family (based on observed trends
in monocots as a whole) were given a state of ‘‘0’’,
and those states that have been considered more
derived were scored as ‘‘1’’ or higher.
CHARACTER 1. INFLORESCENCE PEDUNCLE
DEVELOPMENT. The inflorescence may be pedunculate or sessile to subsessile. The inflorescences of Haemodorum and Heteranthera are pedunculate.
This character is divided into two states: 0 5 inflorescence peduncle well developed; 1 5 inflorescence peduncle absent or nearly so.
CHARACTER 2. CINCINNUS BRACT DEVELOPMENT. The bract of the cincinnus may be
large, small or lacking. If present, it may be persistent or cauducous. In Commelina, the cincinnus bract
encloses the inflorescence and is spathe-like. In all
three genera of the subtribe Streptoliriinae sensu
Faden and Hunt (Streptolirion, Spatholirion, and Aetheolirion), bracts may be large and foliaceous,
small, or lacking. In Stanfieldiella and Floscopa, a
clear gradation from large foliaceous bracts at the
lower region of the inflorescence, to small bracts at
the distal regions often can be found within a single
inflorescence.
The development of the cincinnus bract is variable in the outgroups. The bracts in Haemodorum are
small and cauducous or absent, and in Heteranthera
they are large and foliaceous.
This character is divided into four different char-
684
SYSTEMATIC BOTANY
acter states: 0 5 cincinnus bracts small and persistent; 1 5 cincinnus bracts small and cauducous or
absent; 2 5 cincinnus bracts all large and foliaceous
or spathaceous; 3 5 some cincinnus bracts large
and foliaceous and some small or lacking; 4 5 cincinnus bracts grading from large and foliaceous at
the proximal region of the inflorescence to small or
lacking at the distal regions.
As it is difficult to make any reliable assumptions
regarding the order of transformation of this character, it is left unordered in the second analysis.
CHARACTER 3. CINCINNUS PEDUNCLE DEVELOPMENT. The individual cincinni may be either pedunculate or sessile (or nearly so). In some
genera (see Character 7) the cincinnus is reduced
to a single flower, making it problematic to distinguish between the cincinnus peduncle and the
flower pedicel (Character 7). In these cases, it is
possible to distinguish between the two structures
by their positional relationship to the floral bracteole (see Character 4). When a floral bracteole is present, any stalk that is proximal to the bracteole (i.e.,
between the bracteole and the main axis) is considered to be the cincinnus peduncle, while any stalk
that is distal to the bracteole is interpreted as the
flower pedicel.
The cincinni of Haemodorum and Heteranthera are
distinctly pedunculate.
This character is divided into two states: 0 5 cincinnus peduncle well developed; 1 5 cincinnus peduncle absent or nearly so.
CHARACTER 4. BRACTEOLE DEVELOPMENT.
Each flower in the cincinnus may be opposed by a
small bracteole. Most commonly, the bracteole is
persistent and not perfoliate, but in two genera,
Aneilema and Rhopalephora, as well as in some Murdannia species, it is usually perfoliate. Several genera (Palisota, Streptolirion, Spatholirion, Siderasis, Cochliostema, Sauvallea, Murdannia, Tricarpelema, Dictyospermum, and Commelina) contain species in which
the bracteole is cauducous or absent. The bracteole
is absent in Heteranthera and persistent, not perfoliate in Haemodorum.
This character is divided into three states: 0 5
bracteole persistent, not perfoliate; 1 5 bracteole
persistent, perfoliate; 2 5 bracteole cauducous or
absent.
No clearly ordered series is apparent for this
character, so it is left unordered in the second analysis.
CHARACTER 5. NUMBER OF FLOWERS PER
CINCINNUS. The basic unit of the inflorescence
in the Commelinaceae is the helicoid cyme, or cin-
[Volume 25
cinnus, implying the presence of more than one
flower per primary inflorescence. Belosynapsis, Undescribed Genus, Sauvallea, Murdannia, and Aneilema, as well as Cartonema, each contain species interpreted here as having cincinni that, through reduction, consist of a single flower.
Several flowers per cincinnus are found in the inflorescences of Haemodorum and Heteranthera.
This character is divided into two states: 0 5
sometimes more than one flower per cincinnus; 1
5 always only one flower per cincinnus.
CHARACTER 6. FLOWER TYPES PRESENT.
While Commelinaceae typically have perfect flowers, two different types of andromonoecy have arisen within the family. In seven genera (Palisota, Tinantia, Tricarpelema, Polyspatha, Aneilema, Rhopalephora,
and some species of Dichorisandra), perfect and staminate flowers may be found in random positions
on the same plant. Members of the subtribe Streptoliriinae, as well as Pseudoparis and Commelina, possess both perfect and staminate flowers that are
largely separate from one another. Haemodorum and
Heteranthera produce only perfect flowers.
This character is divided into three states: 0 5
flowers all perfect; 1 5 flowers perfect and staminate, distribution random; 2 5 flowers perfect and
staminate, types largely separate.
It may be a safe assumption that the perfect flower type is primitive in the family, but there is no
way to discern which type of andromonoecy is
more derived than the other, or whether they both
arose independently from the perfect flower state.
Therefore this character is left unordered in the second analysis.
CHARACTER 7. PEDICEL PRESENCE/ABSENCE. Flowers in the Commelinaceae may be either pedicillate or sessile to subsessile (see Character 3 for a discussion about the identification of
the flower pedicel and the cincinnus peduncle in
one-flowered cincinni). All of the genera in the tribe
Commelineae possess pedicellate flowers, whereas
approximately half of the genera in the Tradescantieae, as well as the genus Triceratella, have sessile
or subsessile flowers. The flowers in Haemodorum
are pedicellate and in Heteranthera are sessile.
This character is divided into two states: 0 5
flowers pedicellate; 1 5 flowers sessile or nearly so.
CHARACTER 8. SEPAL COLOR. In most genera, members of the outer perianth whorl are distinctly sepal-like, although they may be slightly colored. The sepals of Palisota and the three genera of
the subtribe Streptoliriinae are petal-like and colored the same as the petals. This state is also found
2000]
EVANS ET AL.: COMMELINACEAE
in some species of Porandra and Amischotolype, and
has arisen (almost certainly independently) in Pollia
as well. Both outgroup genera produce tepals, so
the ‘‘sepals’’ were scored as petal-like.
This character is divided into two states: 0 5 outer perianth whorl distinctly sepal-like (although
sometimes colored); 1 5 outer perianth whorl similar to the petals.
CHARACTER 9. SEPAL (OUTER PERIANTH
WHORL) WIDTH. While sepal width has not
been used as a taxonomic character in much of the
family, it has been useful in defining the subtribe
Streptoliriinae. In most genera in the family, the sepals are narrower than the inner (or lateral) petals.
In the Streptoliriinae, as well as in some species of
Callisia and Floscopa, the sepals are noticeably wider
than the inner petals. Additionally, several genera
possess sepals that are nearly equal in width to the
inner petals. Both outgroups produce tepals, so the
‘‘petals’’ and ‘‘sepals’’ were scored as equal in
width.
This character is divided into three states: 0 5
sepals narrower than the inner petals; 1 5 sepals
equal in width to the inner petals; 2 5 sepals wider
than the inner petals.
CHARACTER 10. PETAL SIZE AND FORM.
The corolla in the Commelinaceae is composed of
three petals and may be either actinomorphic or
zygomorphic. The zygomorphic state is expressed
through one or both of the following features: 1)
enlargement or reduction in size of the outer petal
(the terms ‘‘inner’’ and ‘‘outer’’ are meant to represent the positions of the petals in the bud, where
the outer petal is the single petal that slightly envelops the two lateral, or inner petals [Fig. 5]; they
do not necessarily correspond to the position of the
petals relative to the main axis of the inflorescence,
as the flower may sometimes be rotated 60 degrees;
Faden 1975); or 2) shape of the outer petal (the outer
petal may be saccate and/or strongly reflexed in
the zygomorphic corollas). A zygomorphic corolla
is found primarily in some genera of the tribe Commelineae (Floscopa, Buforrestia, Tricarpelema [in some
species], Polyspatha, Aneilema, Rhopalephora, and
Commelina). Strongly zygomorphic flowers are rare
in tribe Tradescantieae but occasionally occur (e.g.,
Tinantia anomala [Torrey] C. B. Clarke). It should be
noted, however, that several genera in the tribe Tradescantieae (e.g., Cochliostema) express a slight tendency toward zygomorphic corollas (the outer petal
is subequal to the inner petals), particularly in
flowers that are oriented horizontally instead of
vertically. Because zygomorphy in these genera is
685
FIG. 5. Floral diagram illustrating the relative position
of each floral unit in the Commelinaceae. OS 5 outer sepal; IS 5 inner sepal; OP 5 outer petal; IP 5 inner petal;
OAS 5 onter antesepalous stamen; IAS 5 inner antesepalous stamen; OAP 5 outer antepetalous stamen; IAP 5
inner antepetalous stamen. The ‘‘outer’’ and ‘‘inner’’ positions refer to the relative location of each structure in the
developing bud.
slight (particularly when compared to the strongly
zygomorphic corollas found in the Commelineae),
or is due primarily to the androecium, these genera
are treated here as having actinomorphic corollas.
The tepals in Haemodorum and Heteranthera are all
equal.
This character is divided into two states: 0 5 petals all equal or subequal (corolla mostly actinomorphic); 1 5 one petal strongly differentiated
from the other two (corolla zygomorphic).
CHARACTER 11. PETAL FUSION. The petals
may be either free or fused, as in Cyanotis, Coleotrype, Weldenia, and some species of Tradescantia. The
degree of fusion is variable, ranging from the relatively short corolla tubes in Cyanotis to the long
corolla tube in Weldenia.
The tepals in Haemodorum are distinct and in Heteranthera are connate. Because of the uncertainty of
homology between petal connation and tepal connation, this character was scored as ‘‘unknown’’ in
Heteranthera. This character is divided into two
states: 0 5 petals distinct; 1 5 petals at least basally
connate.
686
SYSTEMATIC BOTANY
CHARACTER 12. PETAL FORM. The shape of
the petals is variable in that some species have distinctly clawed petals and others do not. Because the
outer petal may evolve independently of the two
inner petals, it is possible to have a situation in
which two of the petals are clawed and the third is
not (Faden 1991). The petals in Haemodorum and
Heteranthera are not clawed.
This character is divided into two states: 0 5 petals not clawed; 1 5 at least one petal clawed.
CHARACTER 13. PETAL FRINGING. Although the petal margin is usually not fringed in
the Commelinaceae, fringed margins are found in
three genera within the subtribe Dichorisandrinae
(Geogenanthus, Undescribed Genus, and Cochliostema). The petals are not fringed in either outgroup
genus.
This character is divided into two states: 0 5 petal margins entire; 1 5 petal margins fringed.
Characters 14 through 33 represent the androecial characters. Because these characters comprise
such a large proportion of the those used in this
study, and because of the complex evolutionary
trends found in the androecium, it is necessary to
first discuss the manner in which these characters
were distinguished from each other before describing each one in detail.
The simplest type of androecium in the Commelinaceae is composed of two whorls of three stamens each, all equal and fertile (Fig. 5). Stamens of
the outer whorl are antesepalous, and those of the
inner whorl are antepetalous. Within each whorl,
one stamen may evolve independently of the other
two. In order to accurately score the stamens, the
androecium was divided into four different components. Each whorl was divided into two units, a
single outer stamen, and two inner stamens. The
terms ‘‘outer’’ and ‘‘inner’’ here do not refer to the
location of the stamens relative to the axis of the
inflorescence, but rather to the sepal or petal in
front of which they are attached. For example, the
outer antesepalous stamen (the outer stamen of the
outer whorl) is the stamen that is attached in front
of the single sepal that slightly encloses the other
two sepals in bud. Likewise, the inner antesepalous
stamens arise in front of the two inner sepals in the
bud. The same convention is used for the antepetalous stamens. Although it may appear unnecessarily complex, this system for naming the members of the androecium accomplishes two purposes:
1) it divides the androecium into individually
evolving units (as opposed to artificially treating
each whorl, or the entire androecium, as a single
[Volume 25
character); and 2) it allows for the consistent recognition of homologous stamens, regardless of the
orientation of the flower to the main axis of the inflorescence (in some genera, the entire flower is rotated 60 degrees).
CHARACTERS 14–17. STAMEN PRESENCE/
ABSENCE. As was noted above, the basic (and
putatively primitive) form of the androecium in
monocots is the presence of two whorls of three
stamens each. One, two, or all three stamens may
be absent or vestigial in either whorl. For example,
the outer antesepalous stamen (OAS) is absent in
Dictyospermum, as well as some species of Palisota,
Dichorisandra, Floscopa, Murdannia, Pollia, Aneilema,
and Rhopalephora. Both outgroup genera lack the antesepalous whorl of stamens.
Each evolutionary component in the androecium
is scored for its presence or absence: Character 14—
outer antesepalous stamen; Character 15—inner antesepalous stamens; Character 16—outer antepetalous stamen; Character 17—inner antepetalous stamens. Each of these characters is divided into two
states: 0 5 stamen(s) present; 1 5 stamen(s) absent.
CHARACTERS 18–21. STAMEN FERTILITY.
The stamens may be either fertile or sterile. While
most of the sterile stamens produce no pollen, a few
species possess antherodes (staminode anthers)
that produce sterile pollen (e.g., the antepetalous
stamen in Tripogandra grandiflora (Donnell-Smith)
Woodson [Lee 1961], the outer antepetalous stamen
in Aneilema hockii DeWild. [Faden 1991], and the inner antepetalous stamens in Palisota [Faden 1983b,
1988]). It should be noted that although the sterile
anthers that produce no pollen (or in which no anther is produced) and those that do produce sterile
pollen were both scored the same (i.e., ‘‘sterile’’),
the homology of these different types of sterility is
suspect. The presence of anthers that produce sterile pollen is rare, however, and is almost certainly
autapomorphic for those taxa that express it. If a
particular stamen or whorl is absent or vestigial in
a genus (see Characters 14–17), then stamen fertility is not applicable, and it was scored as ’’?’’. The
antesepalous whorl of stamens is absent, and the
antepetalous whorl is fertile in Haemodorum and
Heteranthera.
Each unit of the androecium, then, is scored for
staminal fertility: Character 18—outer antesepalous
stamen; Character 19—inner antesepalous stamens;
Character 20—outer antepetalous stamen; and
Character 21—inner antepetalous stamens. These
characters are divided into two states: 0 5 stamen(s) fertile; 1 5 stamen(s) sterile.
2000]
EVANS ET AL.: COMMELINACEAE
CHARACTERS 22–25. STAMINAL FILAMENT
BEARDING. The filaments may be either glabrous
or bearded, often with distinctly moniliform hairs.
Although several different types and degrees of
bearding may be found (even when comparing different filaments from the same flower), information
is lacking regarding homology of the different hair
types. Therefore, the plants are scored on the simple basis of the presence or absence of any type of
hairs.
The stamen filaments of Haemodorum and Heteranthera are all glabrous.
Each unit of the androecium is scored for the
presence/absence of staminal hairs: Character 22—
outer antesepalous filament; Character 23—inner
antesepalous filaments; Character 24—outer antepetalous filament; and Character 25—inner antepetalous filaments. These characters are divided
into two states: 0 5 staminal filament(s) glabrous;
1 5 staminal filament(s) bearded.
CHARACTERS 26–27. RELATIVE FILAMENT
LENGTH WITHIN A WHORL. A considerable
amount of variation in staminal length is found in
the Commelinaceae, particularly within the tribe
Commelineae. While absolute filament lengths are
not informative due to variation in overall flower
size within the family, relative length among staminal filaments is directly comparable, even among
flowers of varying sizes. The length of each filament
relative to the others was determined (both within
a whorl and between whorls; see also Characters
28–29) by first comparing the length of the outer
filament with the inner filaments in the same whorl,
and then comparing each outer filament with the
two inner filaments of the other whorl (Characters
28–29). Thus, all of the length variation may be included without any unnecessary redundancy in the
character coding.
The antepetalous stamens are equal in Haemodorum and Heteranthera.
The filament length comparisons within a whorl
are scored as follows: Character 26—relative filament length in the antesepalous whorl; and Character 27—relative filament length in the antepetalous whorl. These characters are divided into three
states: 0 5 filaments equal; 1 5 outer filament longer, inner filaments shorter; 2 5 outer filament
shorter, inner filaments longer.
There is no obvious evolutionary sequence
among these character states, so they are left unordered in the second analysis.
CHARACTERS 28–29. RELATIVE FILAMENT
LENGTH BETWEEN WHORLS. In comparing the
687
filaments between whorls, it is useful to distinguish
between a ‘‘top’’ (sometimes called ‘‘posterior’’ or
‘‘posticous’’ in horizontal flowers) and a ‘‘bottom’’
(or ‘‘anterior’’ or ‘‘anticous’’) of the flower (see Fig.
5). The ‘‘top’’ was earlier defined as the half of the
flower containing the outer antesepalous and inner
antepetalous stamens, while the ‘‘bottom’’ of the
flower contains the remaining three. Because the
outgroup genera lack the antesepalous whorl, these
characters were not applicible, and were scored as
’’?’’.
Relative filament length between whorls, then, is
scored as two characters: Character 28—relative
length of outer antesepalous (OAS) and inner antepetalous (IAP) filaments (‘‘top’’); and Character
29—relative length of inner antesepalous (IAS) and
outer antepetalous (OAP) filaments (‘‘bottom’’).
They are divided into three states: 0 5 filaments
equal; 1 5 antesepalous filament(s) longer, antepetalous filament(s) shorter; 2 5 antesepalous filament(s) shorter, antepetalous filament(s) longer.
There is no obvious evolutionary sequence
among these character states, so they are left unordered in the second analysis.
CHARACTER 30. FILAMENT ADNATION TO
THE COROLLA. Epipetalous filaments are found
in Coleotrype and Weldenia, as well as some species
of Tradescantia. The filaments of Haemodorum are
epipetalous and those of Heteranthera are free. This
character is divided into two states: 0 5 filaments
free; 1 5 filaments epipetalous.
CHARACTER 31. ANDROECIAL SYMMETRY.
While the androecial characters above provide a
fairly complete description of the variation found
either within a single stamen (or pair of stamens
when considering the lateral ones) or between different individual components of the androecium,
they do not completely describe the arrangement of
the androecium as a whole. As was mentioned earlier, the putatively primitive state for the androecium in the Commelinaceae is a radially symmetrical
arrangement of six fertile stamens. Almost half of
the genera, however, possess species with a zygomorphic (bilaterally symmetrical) androecium. Additionally, the androecium in Cochliostema and
many species of Murdannia is neither actinomorphic
nor perfectly zygomorphic, but asymmetric because
one or more of the stamens curve to one side of the
flower. In many Murdannia species, the stamens
show a mirror-image symmetry between flowers,
similar to enantiostyly, but due to the curvature of
the stamens instead of the style (Faden, unpubl.
data). Although the homology of the asymmetrical
688
SYSTEMATIC BOTANY
androecia in these genera is suspect, it cannot be
assumed that they arose independently, particularly
for a family in which such heavy emphasis has been
placed on androecial characters. The androecium is
actinomorphic in both outgroup genera. This character is divided into three states: 0 5 androecium
actinomorphic at anthesis; 1 5 androecium zygomorphic (bilaterally symmetrical) at anthesis; 2 5
androecium asymmetrical at anthesis.
CHARACTER 32. ANTHER DEHISCENCE.
Anther dehiscence in the Commelinaceae is almost
always through longitudinal slits. Porandra, Dichorisandra, and some species of Coleotrype exhibit poricidal anther dehiscence through terminal pores or
slits. In Cyanotis anthers, dehiscence occurs basally.
Both outgroup genera exhibit terminal longitudinal
anther dehiscence. This character is divided into
three states: 0 5 anther dehiscence longitudinal; 1
5 anther dehiscence poricidal, by terminal pores or
slits; 2 5 anther dehiscence poricidal basally.
There is no obvious evolutionary sequence
among these character states, so this character is
left unordered in the second analysis.
CHARACTER 33. ANTHER CONNECTIVE
WIDTH. The width of the connective relative to
the anther sacs is variable in the Commelinaceae.
In most genera, the connective is narrower than the
anther sacs. Nine genera, however, possess species
with broad connectives. Both outgroup genera have
narrow connectives. This character is divided into
two states: 0 5 connective narrower than the anther
sacs; 1 5 connective broader than the anther sacs.
CHARACTER 34. STYLE CURVATURE. The
single style in a flower may be central in the flower,
or it may have a distinct curve to one side. The
three genera in subtribe Coleotrypinae (Coleotrype,
Porandra, and Amischotolype), as well as Gibasoides,
Tinantia, and some species of Aneilema and Commelina have curved styles. Additionally, some species of Murdannia exhibit enantiostyly, in which the
direction of curvature of the style is predictable (i.e.,
there are ‘‘left’’ and ‘‘right handed’’ flowers; Faden,
unpubl. data). The style in Haemodorum is regularly
curved (enantiostylous) and in Heteranthera is
straight. This character is divided into three states:
0 5 style always central; 1 5 style irregularly
curved; 2 5 style regularly curved (enantiostylous).
CHARACTER 35. FRUIT LOCULE NUMBER.
The fruit in the Commelinaceae is typically a loculicidal capsule (with the two notable exceptions of
Palisota, which produces a berry, and Pollia, which
produces an indehiscent, berry-like, crustaceous
fruit). Various degrees of zygomorphy are found in
[Volume 25
the fruits, which may have three equal locules (radially symmetrical), reduction in size of one locule,
forming three unequal locules (in Siderasis, Rhopalephora, and some species of Coleotrype, Aneilema,
and Commelina), or even the complete loss of one or
two locules (in Floscopa and Polyspatha, as well as
some species of Callisia, Aneilema, and Commelina).
Although it might be argued that there may be a
correlation between the degree of zygomorphy of
other parts of the flower (i.e., the corolla or the androecium) with the number of locules in the fruit
(thereby making this character redundant), it
should be noted that flowers of some genera are
radially symmetrical for one part of the flower and
zygomorphic for another. For example, both the corolla and the androecium in Siderasis are regular,
while the fruit is composed of two larger and one
smaller locules, and Palisota has a zygomorphic androecium and a berry composed of three equal locules. It is apparent that the evolution of the zygomorphic state may occur independently within different parts of the flower.
The fruits of both outgroup genera contain three
equal locules.
This fruit character is divided into three different
states: 0 5 fruit consisting of three equal locules; 1
5 fruit consisting of three unequal (to two) locules;
2 5 fruit consisting of two (or occasionally one)
locules. Because the states in this character form a
continuous series, they were ordered in the second
analysis as follows: 0 ↔ 1 ↔ 2.
CHARACTER 36. SEED ARRANGEMENT.
About one third of the genera contain species with
biseriate seed arrangement, while the remaining
genera have uniseriate seed arrangement.
Haemodorum and Heteranthera have biseriate seed
arrangement.
This character is divided into two states: 0 5
seed arrangement uniseriate; 1 5 seed arrangement
biseriate.
CHARACTERS 37–38. SEED NUMBER PER
LOCULE. The locule facing the outer sepal (dorsal locule) evolves independently of the locules facing the inner sepals (ventral locules). In Aneilema,
for example, the dorsal locule, when developed, always contains fewer seeds than each ventral locule.
The fruit of Dictyospermum, in contrast, always contains a single seed in each of the three locules. Additionally, some genera produce fruits with the putatively primitive state of more than two seeds in
each locule (e.g., Cochliostema and Weldenia). Finally,
some fruits contain all three locules, but one or two
of them do not produce seeds.
2000]
EVANS ET AL.: COMMELINACEAE
The fruits in Haemodorum consist of three locules,
each of which produces two seeds. The fruits of
Heteranthera consist of three locules, each of which
produces numerous seeds.
The number of seeds per locule, then, is treated
as two independent characters: Character 37—seed
number per dorsal locule; Character 38—seed number per ventral locule. These characters are divided
into four different states: 0 5 more than two seeds/
locule; 1 5 two seeds/locule; 2 5 one seed/locule;
3 5 zero seeds/locule (if the locule is entirely absent, then it is scored as ‘‘unknown’’).
Because the states in this character form a continuous series, they are ordered in the second analysis as follows: 0 ↔ 1 ↔ 2 ↔ 3 (note that these
numbers indicate the character state, not the actual
number of seeds/locule).
CHARACTER 39. EMBRYOTEGA POSITION.
All the seeds in the Commelinaceae possess a small
embryotega, or operculum, a cap-like structure
over the embryo. The embryotega may be dorsal,
semi-dorsal, semi-lateral, lateral, or terminal according to the genus. The outgroups do not produce an embryotega, so they are scored as unknown.
This character is divided into four states: 0 5 embryotega dorsal; 1 5 embryotega semi-dorsal to
semi-lateral; 2 5 embryotega lateral; 3 5 embryotega terminal. Although it is easy to envision a transformation series in which the position of the embryotega has moved around the seed (e.g., starting
from the dorsal position and moving through semidorsal and semi-lateral to the lateral), there are no
intermediates between the terminal position and
any of the others. Without further developmental
information to document the order of transformation, this character is left unordered in the second
analysis.
CHARACTER 40. HILUM SHAPE. The funicular scar is most commonly linear, but it is punctiform in Cartonema, Triceratella, Palisota, Belosynapsis, Weldenia, Callisia, and Pollia, as well as a few
scattered species of other genera. Haemodorum produces a punctiform hilum, and information was unavailable for Heteranthera, so it is treated as unknown. This character is divided into two states: 0
5 hilum punctiform to elliptic; 1 5 hilum oblong
to linear.
CHARACTER 41. ARIL. The genus Dichorisandra, as well as some species of Amischotolype and
Porandra produce arillate seeds. Depending on
which species are chosen to represent Amischotolype
and Porandra, this character could appear autapo-
689
morphic for Dichorisandra. Note, however, that
Amischotolype hispida (the species used in this analysis) does produce arillate seeds. Both outgroup
genera produce exarillate seeds. This character is
divided into two states: 0 5 seeds exarillate; 1 5
seeds arillate.
CHARACTER 42. PTYXIS. Ptyxis refers to the
folding of the leaves in bud. In the Commelinaceae,
the developing leaf is typically in one of two positions as it emerges from the bud: its margins may
be rolled into two distinct curls on the upper surface, scroll-like (involute) or they may be rolled into
a single curl (convolute; or ‘‘supervolute’’ in Dahlgren and Clifford 1982). Occasionally they are simply folded (conduplicate). Ptyxis in Haemodorum is
convolute, and information is unavailable for Heteranthera, so it is scored as unknown. This character
is divided into two states: 0 5 ptyxis convolute or
conduplicate; 1 5 ptyxis involute.
CHARACTER 43. STOMATAL STRUCTURE.
While most classifications for the Commelinaceae
have relied heavily upon traditional macro-morphological characters, attempts to incorporate anatomical characters have shown promise (e.g., Pichon
1946; Tomlinson 1966, 1969; Faden and Inman
1996). However, genus-specific data for most anatomical characters are not readily available. This
study incorporates four different anatomical characters (Characters 43–46).
Of the different types of anatomical data that
have been studied, the most useful is stomatal
structure. Four distinct types of stomata have been
described in the Commelinaceae (Fig. 3; Tomlinson
1966, 1969), the simplest of which have only two
subsidiary cells positioned laterally (found in Triceratella, some Cartonema species, and both outgroup genera). Most members of the tribe Tradescantieae have four-celled stomata, with two lateral
and two terminal subsidiary cells. The stomata of
Streptolirion, Spatholirion, and Aetheolirion (subtribe
Streptoliriinae), as well as Geogenanthus, are sixcelled, with four lateral and two terminal subsidiary cells. In these three genera, the terminal cells
are relatively large, extending to the outer edge of
the lateral cells. Finally, members of the tribe Commelineae possess stomata with four lateral and two
terminal subsidiary cells, but the terminal cells are
smaller, not extending to the outer edge of the outermost lateral cells. Stomata in Haemodorum and
Heteranthera are two-celled. This character is divided into four states: 0 5 stomata two-celled; 1 5
stomata four-celled; 2 5 stomata six-celled, termi-
690
SYSTEMATIC BOTANY
nal cells large; 3 5 stomata six-celled, terminal cells
small.
Although Tomlinson (1966) presents information
on the development of each of these stomatal types
(Fig. 3), he also mentions that the sequence of cell
divisions is variable. Until more about the development of the stomatal apparatus is understood, it
is difficult to make a priori hypotheses regarding
the order of character state transformations, so this
character is unordered in the second analysis.
CHARACTER 44. HOOK HAIRS. Tomlinson
(1966) described two main categories of hairs for
the Commelinaceae: ‘‘microhairs’’, which show limited variability in structure, and ‘‘macrohairs’’,
which are quite diverse. He recognized eight different types of macro-hairs, ranging from singlecelled papillae to multicellular uniseriate hairs. One
particular type of macro-hair which has been examined more closely by one of the authors (R. Faden, unpubl. data) is the ‘‘hook hair’’. These twocelled, hook shaped hairs have been observed on
the leaves of Polyspatha, Dictyospermum, Pollia, Rhopalephora, Aneilema, and some species of Commelina,
and are notably absent in the rest of the family
(Tomlinson 1966; R. Faden, unpubl. data). Hook
hairs are absent in both outgroup genera. This character is divided into two states: 0 5 hook hairs absent; 1 5 hook hairs present.
CHARACTER 45. GLANDULAR MICROHAIRS. Three-celled glandular microhairs (or
simply ‘‘microhairs’’ in Tomlinson 1966) are present on the leaves of every genus in the Commelinaceae except Cartonema and Triceratella. The distribution of these hairs on the leaf, stem, and flower
is quite variable, and Tomlinson has suggested that
their distribution might be useful for diagnosing
genera. As an example, he points out that few or
no microhairs are found on the adaxial leaf surface
[Volume 25
in Murdannia, but they are abundant in Aneilema.
Due to the difficulty of collecting data on hair distribution, as well as assessing homology on such
data, however, this character is simply scored for its
presence or absence.
Glandular hairs are present in some Haemodoraceae and some Pontederiaceae (Simpson 1990), but
they are largely restricted to the outer surfaces of
flowers and inflorescence branches. As the homology of these hairs with the vegetative glandular microhairs is suspect, both outgroup genera were
scored as lacking glandular microhairs.
This character is divided into two states: 0 5
glandular microhairs absent; 1 5 glandular microhairs present.
CHARACTER 46. RAPHIDE CANALS. Raphide canals are present in all genera of Commelinaceae except Cartonema. Indeed, the absence of raphide canals has been one of the most compelling arguments for the separation of Cartonema into its
own family (Pichon 1946; Hutchinson 1959; Tomlinson 1966). Of the remaining genera, only in Triceratella are the raphide canals located next to the
veins, instead of being evenly distributed throughout the leaf blade.
Raphide canals are absent in Haemodorum and
Heteranthera.
This character is divided into three states: 0 5
raphide canals absent; 1 5 raphide canals present
next to the veins; 2 5 raphide canals present, not
exclusively next to the veins.
CHARACTER 47. CLOSED LEAF SHEATH.
All members of the Commelinaceae possess distinct
closed sheathing leaf bases (Cronquist 1981; Dahlgren et al. 1985). Although both outgroup genera
possess a sheathing leaf base, it is open. This character was divided into two states: 0 5 closed
sheathing leaf bases absent; 1 5 closed sheathing
leaf bases present.
2000]
EVANS ET AL.: COMMELINACEAE
691
APPENDIX 2
Data matrix of morphological characters used in cladistic analysis of Commelinaceae. For alignment
purposes, characters that are polymorphic in a taxon are represented by a letter: A 5 0/1; B 5 0/2; C 5
1/2; D 5 0/1/2; E 5 2/3.