Higher-level systematics of the monocotyledons: an assessment of current knowledge and a new classification

Full text

Monocots:
Systematits
and
Euolution.
(2000).
Eds K.L.
\(ilson and
D.A. Morrison.
(CSIRO:
Melbourne)
Hlcuen-levEl
sysrEMATtcs
oF
THE MoNocorYLEDoNs:
AN
ASSESSMENT
OF CURRENT
KNOWLEDGE
AND A
NEW CLASSIFICATION
Mark W.
ChaseA'F,
Douglas
E. Sottif
,
Pamela S.
SottiP,
PaulaJ.
Rudat/,
Michaet
F, Faf
,
William
H. Hahnc,
Stuart Sulliua#,
JffieyJosep#,
Mia
MolurayA'E,
PaulJ.
KoresA'E,
ThomasJ.
GiunishD,
KennethJ.
Sytsma,D
andJ.
Chris
PireP
ARoyal
Botanic Gardens,
Kew,
Richmond, Surrey
TV9
3DS,
llK-
BD.p"rt-.rt
of
Botany,'Washington
State
University,
Pullnan,
\Washington'
USA.
CCERC,
Columbia
Universiry,
1200
Amsterdam
Avenue, New
York, NY
10027, USA.
DD.p".t^.rt
of Botany,
University
of
'Wisconsin,
Madison,
\(isconsin
53706-1381,
USA
oCurr.rt
address:
Department
of Botany
and Microbiology,
Oklahoma
University,
Norman,
Oklahoma
73019,
USA.
FCorresponding
author.
Abstroct
Since
the t me ofthe
last
major conlerence
on
rnonocoty edons,
great progress tn understanding
oftheir higher
evel
relationshipshastal<enplace;
muchofths
sdueanenormousincrease
ntheamountof
DNAsequence
data
co lected.
We
present
here
the resuts
of a comb
ned analysis of
three
genes, two
plastid and one
nuclear,
for 126
monocots,
as we
I
as
22 magnol
id outgroups.
These trees
are highly
congruent
with
previous higher
evel ana
yses
of
plastid rbcL sequences,
but with
one s
gn f
cant
d fference:
much of the
rnonocot tree
is
we
I
supporled.
Based on
the resu
ts of a similar
three-gene
study encompass
ng a
I
seed
plants, the monocots
are
weakly
supported
as a
member of a
clade composed
of CaneL
aceae
Winteraceae,
Ch oranthaceae,
Laurales,
lYagnoliales
and
Piperaes,
although
none of
their inter
reationships
receives more
than 5O%
bootstrap
sup-
port.
Al shorlest
trees
place
the
monocots
as sister
to the rest of
the taxa.
Wlthin the
strongy supported
monocots,Acoraceaearewel
supporledasthesister-grouptoal
therest.Aismataes(stronglysuppodedand
encornpass
ng a
I a ismatid
fam lies
plus Araceae and
Tofie diaceae)
are supporled
as the
next diverging
order,
fol owed by
Pandanales,
Dioscoreaes,
Li ales,
and finally
Asparagales
commeiinoids.
Al of
these arge cades
have at
least some
bootstrap
support, but
their inter
relat onships
all receive
ess than 50%
bootstrap
suppoft
Withnthecommelinoids,al
shortesttreesplacethepamsassistertotherest,
lolowedbyDasypogonaceae'
Zingtberales
Commel
naes and
Poales.
Al of these cades
receive
at least
weal< bootstrap
suppod, but
therr
inter
relatonships
receive
less than 50%
bootstrap
supporl.
The reason
for th s
Lack of support
rs the shorl
nternal
branches
nthisportionofthetree,
lYoredata,bothmorphoogcal
andmolecuar,arerequredtobet-
ter
est
mate these suprafami
ial
patterns,
but
the overal
congruence
of the three
gene patterns
p
us the
hlgh
leves of
nternaL support
for
nearly all ordinal
groupings
gives
us
confidence
that a new
classfcation
ofthe
monocots
based
on these
results is desirab
e, This
new monocot
classiicatlon,
published elsewhere
as
parl
of
an
overall ordinal
classificatlon
of the
fbmi ies of
flowering
pants,
s reproduced
here
with some
minor updating
based
on
results with
strong suppoft
lrom
the analyses
presented here.
Key words:
monocots,
phylogeny, bootstrap,
Acoraceae,
Trurjdaceae,
Dioscoreales,
Pandanales,
Commel
na es, Poa
es
lHtRooucrtohl
\Within
the
last decade,
major
strides
in the understanding
of
higher
level
phylogenetic
patterns
of
the seed plants
have
occurred,
largely
due to
the enormous
amount
of DNA sequence
data
that has
become
available
(Chase
er
al. 1993; Sokis
et al.
1997a,
1998;
Kallersjo et al.
1998; Savolainen
et
a/. in ptess).
The volume
ofsequence
data produced
has been staggering
(for
example,
2538 rbcL sequences
were
analysed
in the
Killersjo ez
a/. 1999 study),
whereas
three genes,
totaling
4549 base-pairs,

Mark W. Chase et al.
have
been collected
for
each of
567
seed plants in Soltis et al.
(560
angiosperms plus seven
gymnosperm
outgroups; submit-
ted). No other major
group
of organisms has been the focus
of
such
large-scale
and
intensively
sampled analyses. From the tra-
ditional
perspective, such vast amounts of data are not realisti-
cally analysed in a robust rranner
(Patterson
et al.
1993'
Hlllis
l995;
Graur et al. 1996), but many recent
simulations
(Graybeal
1998; Hillis 1998; Poe 1998) as well
as empirical studies have
demonstrated that larger, better sampled matrices
are not only
less problematic than previously envisaged
(Chase
and Cox
I998; Soltis et al. 1997b,1998), they are even
superior to smaller
thotigh
more rigorously
analysed data sets
(Halanych
1998). Bet-
ter
sampling
(i.e.
rnore
thorough taxonomically) is a more effec-
tive way to enhance accuracy
than an
increase
in the amount of
data in the analysis,
particularly
if it
breaks up long branches
(Graybeal
1998). Furthermore,
consistency of the relationships
recovered from the three individual
gene
trees for the
angiosperms is remarkable
(Chase
and Cox l99B), a fulther indi-
cation ofthe likely accuracy ofthese results.
Dahlgren and Rasmussen
(1983)
and Dahlgren
et a/.
(1985)
began the cladistic analysis of higher-level relationships
across
monocots
based on morphology and anatomy, leading to
rhe
rec-
ognition of new, well-supported
groups
such
as
Asparagales,
and
of overlooked affinities between other
groups
(e.g.,
Trilliaceae
ar.rd Stemonaceae). Chase et a/.
(1993)
and Duvall
er al.
(1993)
revolutionised monocot
systematics by deriving a ur-ritary
phylog-
eny based on analysing an extensive
data set of plastid rlcl DNA
sequences. Several key
papers,
covering relationships
in each of
the major
groups
as well as the monocots
as a whole, appeared in
Rudall et al.
(1995);
notable among these
were an analysis of the
difficult commelinoid
group
by Kellogg
and
Linder
(l995),
a re-
examination of relationships across monocots
based on morphol-
ogy by Stevenson and Loconte
(1995),
and
a combined analysis
of
molecular
and
rnorphological
data by Chase et a/.
(1995).
Finally, Civnish et al.
(1999)
analysed relationships among com-
melinoids
usir-rg
a codon-weighted
analysis of rbcL sequences,
including representatives
of several previously unstudied
groups.
Despite
this success
in
sorting out many of the major
patterns,
it
must at the
same time be acknowledged that not everything is
well supported and
clear, and there
is
still need for more study.
The rr-rajor improvements of the last few
years have had the effect
of
focusing
attention on the taxa and
patterns
for
which addi-
tional
resources
and study are needed. The
purpose ofthis paper
is
to summarise with respect to the mor-rocots
and their closest
relatives
the patterns of relatior-rship that
seem clear, as well as to
identi$'
the topics that require further
study. \7e therefore
present
here
the
results
ofan analysis ofthree
genes, plastid r/cl,
plastid
atpB,
and
nuclear I
BS ribosomal DNA
(IDNA),
for
over
100 monocot species. These
data are largely those collected for a
seed plant analysis
(Sokis
et a/.
submitted) that will be
published
elsewhere, but we have extended
the monocot sampling of that
analysis for this report.
(Due
to
the
length
limitations imposed
for this volume, we will
publish a
more
extensive and mole prop-
erly documented study elsewhere.) Finally,
we present here an
amended classification of the monocots
based on these results
and the formal reclassification
of the flowering
plants published
by the Angiosperm Phylogeny
Group
(APG
1998). For consist-
4
ency, we follow this classification throughout this
paper
with the
exceptions noted with an asterisk: these were taxa not
placed by
the
APG
because too little suppolt or information was available
then, but
with
additional
sampling and the support now availa-
ble we have moved these families into
orders already
recognised
by the APG classification. Taxa that are
differently
classified
from the APG
classification are marked with
t;
these are cases in
which the APG classification recognised a family that we con-
sider here
to be
included
within another family. For example,
because
Orectanthe
(Abolbodaceae)
falls as s.ister to the rest of
Xyridaceae,
we
think
that
it is
better to
include
it in the latter.
MnreRral AND METHoDS
DNA samples for the
great
majoriry
of these ta-ra
have
been
extracted from either fresh
or silica-gel-dried tissue samples by a
standard 2X CTAB
protocol
modified from
that of Doyle and
Doyle
(1987).
Vouchers and full citations for most of these
DNA samples are listed in Chase et al.
(1995a)
and Rudall ar a/.
(.1997);
we attempted
to use the same DNA samples as in these
previous
published
studies. From these total DNA templates,
each of the three
genes
was amplified and sequenced
directly
from
double-stranded PCR templates using standard techniques
with
the PCR and sequencing primers described in Ll.ed6 et a/.
(1998),
Hoot
e/ al.
(1995),
and Soltis et a/.
(1997a),
for rbcL,
atpB, and 1BS rDNA, respectively.
Alignment of sequences for these three
genes was
easily
accom-
plished by eye because no insertions/deletions occur for rbcL and
are rare
and short
for
atpB and 1BS
rDNA.
As described in Soltis
et a/.
(1997a),
a
few
ambiguously aligned
regions
of 18S rDNA
were excluded from
this analysis.
Because
the
issue
ofcongruence
and combinability
of
these
three
matrices has
been dealt with
elsewhere
(Chase
and
Cox
1998;
Soltis
rr al. 19976, 1998),
we
present here
(Fig.
1) results of the bootstrap for the combined
matrix in
which all three
loci
were present for all taxa. In addi-
tion,
we show one of
the individual
trees
from
this analysis
(Fig.
24, B).
\7e
have adopted the following categories for
describing
bootstrap support: weak,
50-74Vo;
moderate,
75-84o/o;
strong,
>85%0.
Groups
not
present
in at least
50%
ofthe
bootstrap
repli-
cates are not reported because these are often not
present
in the
strict consensus tree ofthe shortest trees and are liable to change
with
the addition of
more
taxa or data.
\7e performed a second analysis in which we included placehold-
ers for families from which only or-re or two of the three
genes
were sequenced. For example, one species, Sciaphila sp.
(Triurid-
aceae), has only an lBS rDNA sequence; no plastid loci would
amplify for this achlorophyllous family, and so it appears that we
will
have to
base our estimates on patterns on
just
this
nuclear
gene until
additional loci can
be sequenced. One ofthese
trees is
illustrated in Fig.
3
A, B. Because large amounts of missilrg data
weal<en estimates of internal support, we do not consider this
analysis to be as
important
in establishing patterns as the first one
in
which all ta-xa
have
all three genes
included.
Other studies
in
which a certain amount of data was absent
for
some taxa
have
shown that the effect on
general patterns
of the optimal tree
found with all loci for all taxa is minimal; this result is obvious
here as well. The missing data do result in lower bootstrap per-
centages
for
some groups, although the effect
is not huge
since

most taxa are at least two of the tree
genes
included
(compare
Fig. 2
with
Fig.
3).
\X/e
included this anlysis so that family reple-
sentation
was increased.
Both
sets of searches
were conducted with PAUP* 4.0d64
(D.
Swofford
1999) and consisted of 1000 replicates of random
taxon-entry orders with sub-tree pruning and
regrafting
(SPR)
swapping, but holding only l0 trees per step to
reduce
the time
spent in swapping on trees at suboptimal
lengths.
Shortest trees
collected
in
these
replicates
were
then
subsequently
swapped to
completion. Successive approximations weighting
(S'!7;
Farris
1969) was used to check for
patterns
that might be created by
sites that changc more frequently, and the trees produced were
similal to those found with
Fitch
parsimony
(equal
weights,
unordered
characters; Fitch 1971). \7e illustrate, with their Fitch
branch
lengths
(autapomorphies
included), the
single
tree found
with
successive weightir-rg
in
both searches, but
all
gror-rps
that are
not found in the strict consensus tree of the shortest Fitch trees
are marked with solid arrowheads
(Figs
2,
3).
Bootstrap analyses
(Felsenstein
I
985)
were conducted with
250 replicates, each
using SPR swapping
and holding
only
25 trees
per
step to reduce
time spent in swapping or-r large numbers of trees; this
procedure
can result in only an underestimate of support if any effect is to
occur
(less
thorough swapping with SPR finds larger
groups
less
effectively).
Both matrices are available from M'ffC
(m.chase@rbgkew.org.uk)
upon request; the new sequcnces will be deposited
in
GenBank after
submission of the
larger
three-gene
analysis
(Sokis
et a/. in
prep.).
Resulrs
The bootstrap consensus tree
(Fig.
l) is
highly resolved, although
there are rwo regions that collapse
into
polytomies as a
result
of
having
less
than
50%
support.
The monocot
portion of
the
angiosperm bootstrap tree
presented
here includes as well the sis-
ter
groups
of the monocots
(Fig.
1), which collectively includes
Canellaceae \(/interaceae, Laurales, Magnoliales and Piperales,
although
inter-relationships within
this clade
receive less than
50%o
bootstrap support. Bootstrap support
for
the
inclusion
of
the
monocots in
this eurnagnoliid
clade is weak in the complete
seed plant
analysis of all three
genes
(56%;
Soltis rr a/. inprep),
but in all ofthe large analyses
published
so far
(Chase
et al. \993;
Soltis er al. 1997a; Savolainen et a/.
in
press) there
is
a similar
pattern.
(Unless
noted
otherwise, the
term
'support'
is
used
here
only
in reference
to the bootstrap
and
does
not refer to character
support
for clades.)
The monocots are strongly supported as
monophyletic
(100%;
Figs 1,2A). \Tithin the monocots, there is strong support for
Acorus
(Acoraceae;
Acorales) as the sistel to the rest of the
llronocots
(95%).
The next-branching order, Alismatales
(including
the alismatid
families
plus
Araceae
and
Tofieldiaceae),
are both
moderately
supported
(75o/o)
and excluded
(78%)
from
the
remair-rder
of the
rnonocots. Relationships withir-r the sister
group of
Alismatales are
one of
the two major
points of uncer-
tainty within the monocots. Although moderately supported
(7\o/o),
the bootstrap consensus tree leaves this clade as a large
polytomy composed of the lollowing supported
groups
(Fig.
1):
/apono
/irion-Petrosauia
(99olo;
Petrosaviaceae,
which
were listed
SupnnrnvtLlnL cLAsstFtcATtoN oF THE tyoNocors
Bootstrap
Consensus
./
...
"t-C
(85-1oo)
,** l=""""
**
moderate
(75-84)
,/ t**
{
Arecates
*weak(50'74)
/
.(Zinsiberares
{
(.ott",,nu,",
./
\ ljasypogonaceae
-{
eumagnolirds
Fig. l. Bootstrap consensus tree
(250 replicates) for taxa for which all
three
genes
were collected. Groups
occurring in less than 50% of these
replicates are indicated as polytomies. Orders and family circumscrip-
tions are as in APG
(1998).
as
separate
farnilies in the APG
(1998)
classification; such
changes in circumscription are denoted by a
f
in the text and in
the revised classification in the Appendix),
the commelinoid
families and orders
(77o/o:
see below),
Dioscoreales
(667o,
including here Nartheciaceae,
which were unplaced in the APG
1998 classification; such changes
are
denoted
with an asterisk in
the text and
revised classification in the Appendix), Pandanales
(99olo),
Liliales
(97o/o)
and Asparagales
(56%).
A.rnong the families of Liliales included
in Fig. 24,
three
clades
are supported: Melanthiaceae
(100o/o;
including Trilliaceae),
Alstroemeriaceae Colchicaceae
(1
00o/o),
and
Smilacaceae-fupog-
onaceae-Liliaceae
(99%;
the last
including
Clintonia, Scoliopus
and
Triqtrth,
supported at
9B%).
\X/ithin
Pandanales, Stemon-
aceae-Velloziaceae
(73o/o)
are sister to Cyclanthaceae-Pandan-
aceae
(91%);
within Dioscoreales, Burmanniaceae are sister
to
Dioscoreaceae-Taccaceae
(100%),
and
Nartheciaceae* are sister
to these three
(66%).
Among the families of Asparagales
included in Fig. 28, there is
only weak support
(560/o)
for the
inclusion
of
Asteliaceae, Bland-
fordiaceae, Boryaceae,
Hypoxidaceae and Orchidaceae. Mr-rch
stronger support
for
their sister gloup
exists
(96%;
the
rest
of
Asparagales
sensu Chase
et al. 1995a, b). Within
Asparagales,
several other relationships are strongly supported:
Agavaceae-

44
Mark W. Chase et a/.
Sco/iopus
Tricyrtis
Calochortus
Clintonia
Lilium
Smilax
Ripogonum
Trillium
Xercphyllum
Chamaelirium
Bomarea
Colchicum Colchicaceae
-t
Liliaceae I
I
Liliales
Jmr
racaceae
A
227
Anthericaceae-Behniaceae
(100%),
Amaryllidaceae
(:99o/o),
Asphodelaceae-Xanthorrhoeaceae
(77
o/o),
Convallariaceae s.r.
(100%,
including
Nolinaceae and Ruscaceae),
Hyacinthaceae
(9
5o/o),
Hyacinthaceae-Themidaceae
(7
4o/o), kidaceae
(
1 00%),
Orchidaceae
(100%),
Tecophilaeaceae
(100%),
and Tecophilae-
aceae-Ixioliriaceae
(550/o).
Much of the spine within Asparagales
is also moderately to strongly supported.
Although the commelinoid clade is supported, in the bootstrap
consensus tree there is a tetrachotomy at the base composed of
the following clades: Arecaceae
(100%),
Dasypogonaceae
(
1 00%), Commelinales-Zingiberales
(7
lo/o) and P oales
(97
o/o).
Among
the
families
of Commelinales
included,
Commel-
inaceae-Haemodoraceae-Pontederiaceae
(81%)
are sister to
6
Asparag
a
les/com me I i noid s
cYclanthus.
cvcranthaceae I
5pnaeretclen,a
Freycinetia Pandanaceae Pandanales
Barbacenia Velloziaceae
I
Stemona Stemonaceae
,l
Jaoonolirion
-l
petrosavia
Petrosaviaceae
Dioscorea Dioscoreaceae
I
Tacca Taccaceae
Dioscoreales
Burmannia Burmanniaceae
__1
Aletris Nartheciaceae
Vallisneria Hydrocharitaceael
Zostera Zosteraceae
I
Aponogeton Aponogetonaceael
3tr:,X",;;Xf,X!;
Araceae
irismatares
PJT".
.. Tofierdiaceae I
lottelclra
J
74
Acorus
Annona
Asimina
Eupomatia
Degeneria
Ripogonaceae
-'lMetanthiaceae
l
Acoraceae
lAnnon"""""
l
Acorales
eumagnoliid
outgroups
113
72
72
77
63
10r
64
65
114
51
70
45
100
47
100
29
19
129
43
Calycanthus Calycanthaceae
Aristolochia
-
Lactois
Asarum
Aristolochiaceae
Saruma
Peoeromia
--l
Piper
-
PiPeraceae
Saururus Saururaceae
Chloranthus
--l
Sarcandra Chloranthaceae
Hedyosmum
l
Fig. 2. The single shortest tree found with successive weighting from the first analysis
(as
in
Fig. I
).
Numbers above
the
branches are the numbers of
estimated substitutions (ACCTRAN optimisation; autapomorphies included). Groups that are notfound in all of the nine shonest Fitch trees are indi-
cated by an arrowhead;
groups
that received less than 50% bootstrap are noted with a
'-'.
A. The outgroup and basal-most branches of this
single tree.
Philydraceae
(100%),
and, although the families of Zingiberales
are supported,
their inter-relationships all have less than
507o
support.
Among the families of Poales* included, Bromeiiaceae*
are sister
(excluded:
55o/o)
to all the rest, followed by Sparga-
niaceae, which are sister
(86%)
to nvo clades: i)
Xyridaceae
sister
(760/o)
to
Juncaceae-Cyperaceae
(100o/o)
and ii) Flagellariaceae
sister
(7
8
%)
to Restionacea e-P oaceae
(620/o)
.
The nine shortest Fitch trees obtained
in
this analysis
had I0,872
steps with a consistency index
(CI,
including autapomorphies) of
0.30 and a retention
index
(RI)
of 0.54.
The
single
tree found
with successive weighting
had 1816.78413 steps with a CI
=
0.67
and an Ri
=
0.71; the Fitch tree length of the S\7 tree was
10,889
(17
steps longer than the shortest Fitch tree) with a
Degeneriaceae
Magnoliaceae

Supnnrnvtrrnr
cLASsrFtcATroN oF
THE MoNocors
Dimerocostus I
--l
--l
Alonocostus I I I
Tapeinochilos
lCostaceae I I
Cosfus
--.1
I I
frilAl?:
lMarantaceae I I
Hedvchium I I I
Tiniihpr I
I I
Gi;bb;
lzrngrberaceae
Zingiberales
Rredelia
--l I I
ff3XZuo"r",ru.
cannaceae
mm
gl
inoids
Stretitzia
'
]str"titriu"&
Ravenala
I I
Orchidantha Lowiaceae
I I
Heliconia
Heliconiaceae
I I
Ensete
I I I
Yri,l;'n
_.]Musaceae
I I
Pontederia
pontederiaceae
---]
I
Tradescantia
Commelina
Anigazanthos
-
nu".ooo,u????"
Commelinales
Hellnholtzia I I I
philvdrela
Phitydraceae
I I
(l,ier,,<
I I
Htivhlii"poru
cYPeraceae
I
I
Juhcus
'
--
Juncaceae Poales
I
Xvis Xvridaceae
I I
Xvis Xvridaceae I
l)ru7a _ I
zda
Poaceae
I
Restlo
-
Restionaceae
I
Flagellaria
Flagellariaceae
I
Sparganium Sparganiaceae
I
ti3l,!ffr,"",.,u
Bromeliaceae
--J
Calectasta
AiSA?f&r*,.a
Bromlriaceae
-J
Calectasta
Dasvnooon
Dasypogonaceae
Phoenix I
----1
P,odococcus
I
'6;;;"
Arecaceae
Arecales
Chdmaedorea I
Calamus
I
Behnia
Chlorophytum
Agave
Bowiea
Scilla
Muilla
Asparagus
Ruscus
Smilacina
Lhiope
Convallaria
Nolina
Allium
Clivia
Hippeastrum
Bulbine
Xanthorrhoea
Xeronema
Aristea
Gladiolus
Cyanella
Tecophilaea
lxiolirion
Borva
Empodtum
Rhodohypoxis
Milligania
Blandfordia
Cypripedium
Oncidium
Aposfasia
Behniaceae
Anihericaceae
Aoavaceae
-l"
I
Hyacinthaceae
Themidaceae
Asparagaceae
Convallariaceae
Alliaceae
I Amarvllidaceae
I
-Asphodelacea"
AsParagales
Xanthorrhoeaceae
Xeronemataceae
I lrid"""""
l
I
Tecophilaeaceae
lxioliriaceae
Boryaceae
-
Hypoxidaceae
-
A"t"liu""u"
Blandfordiaceae
(
)rch rd acea e
l
Fig. 28. The Asparagales and commelinoid portion
of the same tree. Orders and family circumscriptions are as in APG
(1998)
with only slight modi-
fication
(see
Appendix l).
CI
=
0.30 and an RI
=
0.54.The only major
differences between
the
S\X/
and Fitch
trees were 1
)
the position of Allium, which was
sister in the latter ro Asparagus and
sister
to Amaryllidaceae in
the
former;
2)
Melanthiaceae sister to Colchicaceae in the latter and
sister to
Liliaceae-Ripogonaceae-Smilacaceae
in the former;
3)
Boryaceae sister to the
Orchidaceae clade
in
the latter and sister
to the rest of Asparagales in the former; and 4) Dasypogonaceae
sister to Arecaceae in
the
latter
and sister to rest of the commeli-
noid families in the former. None of these alternative
positions
receive
bootstrap support greater than
50%o.
In
the second analysis
(with
some ta-xa
missing
one or two of
the
genes)
the same
general
patterns as described above were observed.
\(e illustrate here the single successive
(SVf
weighting tree with
its Fitch
branch
lengths
displayed
(Fig.
3A,B).
There were a total
of
384
most parsimonious Fitch trees
(groups
not present in the
strict consensus tree of these trees are marked with arrowheads).
The Fitch trees had 12,632
steps
with a
CI
=
0.28
(including
autapomorphies) and an zu
=
0.53. The single StW tree
had
IB95
.607 69 steps with a CI
=
0.65 and an RI
=
0.70;
the Fitch
sta-
tistics for this tree were 72,648 steps with a CI
-
0.28 and

Marh W. Chase et a/.
zu
-
0.53. In other experinents with addition of taxa for which
112 to
Il3
of the data are missing, placements are close to those
four.rd later when all data are
pi.esent
(Chase
er a/. 1995b; R.
G.
Olmstead, pers. comm.). The overall patterns are highly similar to
the ar-ralysis with ta-ra for which all three
genes
were included.
\We
will mention here
only the placements of the additional ta-ra.
Although we estimated bootstrap support and indicate this on Fig.
3,
we included this only to
demonstrate
the
generally
lower
per-
8
Lilium I
Tutipa
I
Clintonia I
Lloydia
I
Liliaceae
Calochortus I
Sco/lopus
I
Tricyrtis
l
Ripogonum
Ripogonaceae
Lapageria Philesiaceae
Smilax
Smilacaceae
Tillium
tr:^t'i!{i',i,f^
cor ch icaceae
Veratrum
9;',!i!J\iH
0,,-
Me ranth iaceae
Bomarea Alstroemeriaceae
les
Butomus
Sagittaria
Vallisneria
Zostera
Butomaceae
Alismataceae
Hydroc ha ri taceae
Zosteraceae
eumagnoliid
outgroups
I
I
I
l
Fig. 3. The single shortest tree found with successive weighting from the second analysis in which several taxa were missing one or two of the
genes
(see
text for details). Numbers above the branches are the numbers of estimated substitutions
(ACCTRAN
optimisation; autapomorphies
included).
Groups that are
not found in
all of the 384 shortest
Fitch
trees are
indicated
by an arrowhead;
groups that received less than 50% bootstrap are noted
with a'
'.
A. The outgroup and basal-most branches of this single SW tree.
l
Lilia
Cyclanthus
Cvclanthaceae
-l
Sphaeradenia
'
I
F_re.vci\gtia
Pand.anaceae
pandanales
Sciaphila Triuridaceae
'"'i""""'--
Barbacenia Velloziaceae
I
Stemona
Stemonaceae
_l
As
para ga
les/co m m
e
I i no id
s
Dioscorea Dioscoreaceae
^-l-
Tacca
Taccaceae
UlOSCOfealeS
Burmannia Burmanniaceae
-J
Aletrts Nartheciaceae
iftf:::,{i:,
]
P"t,o."ui"""""
Aponogeton Aponogetonaceae
Alismatales
7t-9". .. Tofierdiaceae I
tottetctta
I
Gymnostachys
Araceae I
Spathiphyllum_
I
Acorus Acoraceae Acorales
Annona
_l
-
Asimina
An nonaceae
Eupomatia
-Eupomatiaceae
Degeneria Degeneriaceae
Magnolia Magnoliaceae
Canella
Ci nnamode nron
Uanellaceae
THj,,," I
*,n,",u""u"
.-J
ulnntmomum I
sassafras
I
Lauraceae
Calycanthus Galycanthaceae
Aisblochia
-
Lactois
I
Asarum I
Aristoloch
iaceae
Saruma
I
Peoeromia |
-.
Pi$er
I
PrPeraceae
Saururus Saururaceae
Chtoranthus
_l
Sarcandra
I
Chloranthaceae
Hedvosmum I
centages and will not discuss this further here, other than to
note
that
in
some cases groups with
high
support drop
to
or-rly
moderate
support with the addition of taxa with missing data. For example,
the addition of Sciaphila
(Triuridaceae)
for
which only
1BS rDNA
is
present
resr-rlts in the bootstrap percentage for Pandanales falling
from99o/o
(Fig.
1, 2,A) to75o/o
(Fig.
3A).
In a separate 1BS rDNA
analysis, Sciaphila also fell within Pandanales, but without boot-
strap
percentage greater
than
50olo
(results
not shown).

SupnnrnvtLtnL clAsstFtcATtoN oF
THE
t'4oNocors
Dimerocostus
--l
Monocostus
Tapeinochilos
Uc6taceae
Cc fil-s
Calathea i..
Merente
MaranECeae
Globba i
Zindiber
Heilvchium
Zln$OeraCeae
Rielletia
Canna
Cannaceae
PhenakosDermum
Stretitzia'
Streliziaceae
Ravenala
Orchidantha
-Lowiaceae
Heliconia
Heliconiaceae
E sefe
Musella
Musaceae
Musa l
Pontederia
Ponbderiaceae
Tradescantia
Commelinaceae
Hanguana
Hanquanaceae
Anigozanthos
Haeh,'todoraceae
Fil,n ?!i:,:
Phirydraceae
9:.f"
Poaceae
Anafthria
Anarfrriaceae
Resfio
Restjonaceae
Flagellaria
F laqellariaceae
Iachnocaulon
xyis
Xwidaceae
Orechnthe
cYPerus
cweraceae
Knyncospora
)
''
lriarEa
Podococcus
Phoenix
Calamus
Calectasia
Dasypogon
Convallaria
Liriope
Rus cus
Smilacina
Nolina
Calibanus
Comospermum
Peliosanthes
Eriospermum
Asparagus
Sowerbaea
Behnia
Chlorophytum
Chlorogalum
Agave
Anemarrhena
Scilla
Whiteheadia
Bowiea
Muilla
Alliam
lpheion
Clivia
Hiweastum
Caesra
Hemerocallis
Phormium
Bulbine
Xanthorrhoea
Xeronen a
Aristea
Gladiolus
Cyanella
Tecophilaea
lxiolirion
Empodium
Rhodohypoxis
Millidania
BIan-dfordia
Cypripedium
Oncidium
Apostasia
Borya
Aspa ragaceae
Laxmantaceae
Behniaceae
Andredcaceae
]Agauaceae
I
Hvacintraceae
Themidaceae
Asohodelaceae
X ahhorr*roea ce ae
Xeronemabceae
:,::"-'*'"
Asparagales
Alliaceae
l'
Amaryllidaceae
I
Hemerocallidaceae I
II
.l
I
Zingiberi
I
l
Commeli
l
I
Poales
l"o*
.l
lcomm
I
-
Arecales
__l
elinoids
I
I
l
ales
linal
Arecaceae
Dasypogonaceae
C onvallariaceae
Anemarrhenaceae
lT":"
lecoohrlaeaceae
-
lxioliiiaceae
Hypoxidaceae
-
Asteliaceae
Blandfordiaceae
O rchidaceae
-
Boryaceae
Fig, 3B. The Asparagales and commelinoid
portion
of the same
tree. Note that this terminal pair of clades is not present in all shortest
trees, but
it is
favoured by successive weighting of these data. Orders and family circumscriptions are as in APG
(1998)
with minor modifications
(see
Appendix l).
67
^-
64
ri;J,2's,
lg
,o
'tcjote
19
)J
17
--'-"'{E

Mark W. Chase
et al.
As mentioned above, one
newly
placed
family is Triuridaceae*,
which falls deeply
within Pandanales with Freycinetia
(Pandan-
aceae). The
mitochondrial
genes
that several research groups are
now sequencing should
be able to confirm this placement
(aryA,
Davis
and
Stevenson;
matR, Chase and
Qiu).
Several
additional
families are represented
in Asparagales. These include: Sotaerbaea
(Laxmaniaceae
atpB and rbcL) which fails with Asparagaceae
(but
with less than
50olo
support)
;
Caesia,
Hemerocal/is and
Phormium, all of
Hemerocallidaceae
(atpB
and rbcL;
I00o/o),
which fall with
Asphodelaceae and Xanthorrhoeaceae
(this
clade
has
760/o
support);
and Anemarrhena, Anemanhenaceae
(atpB
and
rbcL), which falls
(84%)
with
Behniaceae, Agavaceae, and
Anthericaceae s.s. Convallariaceae
(100o/o)
contains Nolina and
Calibdnus
(fomerly
of
Nolinaceae; atpB and rbcL). The addition
of lpheion
(Alliaceae;
atpB and rlcl) makes A.lliaceae fall
consist-
ently with Amaryllidaceae
(68%),
whereas in the all-data
matrix,
Alliumwas
highly unstable in its
position,
lalling
only
in the
S'X/
tree with
Amaryllidaceae
(see
above).
\l/ithin
the commelinoids, Arecaceae are sister
to Dasypogonaceae,
but
with support less than
50olo.
Additional family placements
include Hanguanaceae*
(.aeB
and rbcL) sister to Commelinales,
Rapateaceae*
(rbcL
only) sister
to all of Poales, Mayacaceae*
(1BS
rDNA and r/rl) sister
to the sedge clade
(Cyperaceae,
Juncaceae
and Thurniaceae;
for the last of which both Prionium and
Thurnia
have rbcL only).
Arrarthriaceae, Anarthria
(atpB
and
1BS IDNA),
falls
with
Restionaceae Poaceae, but
with less than
50olo
support.
Orectanthe
(rlcl;
Abolbodaceae) is sister to
-!rzi
(Xyridaceae),
making recognition of
Abolbodaceae unnecessaryt.
Typha
(Typhaceae;
1BS rDNA and rbcL) also falls into
Poales
but
not
with Sparganiaceae;
these two are successive sister groups
to all of
Poales except
Rapateaceae. Poa.les here receives scant
5170
boot-
straP suPport.
DrscussroN
The bootstrap
consensus tree
(Fig.
1)
does
not identi$r a sister
group
for the monocots, but the
first analysis
(for
which a-ll criti-
cal
taxa have all three genes present)
is consistent with all previous
analyses
(Chase
et a/.
1993;
Soltis
er dl. 7997a Savolainen et al.
in
press)
in placing the
monocots as the sister of the eumagnoliids:
Canellaceae-\Tinteraceae,
Chloranthaceae, Laurales,
Magno-
liales, and
Piperales. Amborellaceae, Austrobaileyaceae,
Cerato-
phyllaceae,
Illiciaceae-Schisandraceae,
Nymphaeaceae and the
eudicots
have always been consistently outside
of eumagnoliid
clade. \7e cannot explicitly
address the issue of sister-group
rela-
tionships within the eumagnoliids
in this analysis due to the
absence ofyet more distantly
related taxa, but it should be
noted
that the network produced
here is congruent with previously pub-
lished
molecular analyses that showed the
monocots not to have a
single
family as their sister
group.
Analyses of non-molecular data
by Stevenson
and Loconte
(1995)
did
not include any woody
members of this
group,
so their
result
of
Aristolochiaceae alone as
sister to the monocots
cannot be considered a complete evalua-
tion of monocot relationships.
The non-molecular trees of Nandi
et al.
(1998)
also placed
Aristolochiaceae and
related families
(Piperaceae,
etc.) as sister to the monocots
(with
a bootstrap of
83%) and
this pair together as sister to the other
members of the
eumagnoliid
clade
(termed
there
as magnoliid I), so there
is
at
t0
least a degree of similariry benveen
the DNA trees
illustrated here
and other
rypes
of data; a sister-group
relationship
for monocots
and Aristolochiaceae
(and
perhaps
its piperalean relatives)
cannot
yet be considered strongly
refuted.
The eumagnoliid group
(here
defined to
include the monocots)
shares a
considerable number of traits
(Savolainen
et al. in
press),
so
nearly all
data
collected thus
far can be considered
largely in
agreement,
and it is simply the details
that need further clarifica-
tion. A relationship of monocots
to either Chloranthaceae
as sug-
gested
by some authors
(Carlquist
1992), Piperaceae
(Burger
1977)
or Nymphaeaceae
(see
review by Les and Schneider
1
995)
appears
unlikely, although
the weak bootstrap support
for the eumagnoliid
group demonstrated
here
(56%)
cannot yet
be considered the
final
word. If the monocots are sister
to the most of the
eumagnoliid
clade, then this would
make the situation much
more complex
than if they have a single sister
family. No matter which group(s)
the monocots ultimately
have as their sister group,
they will always
be quite
isolated and diver-gent from them.
The
question
of the root within the
monocots has long been
con-
troversial, and that of Acorus
as sister to all the
rest as first
revealed by Duvall et
al.
(1993)
has not been demonstrated
by
anything other than sequence
data. Stevenson
and Loconte
(1995)
used only
herbaceous magnoliids as outgroups,
and their
rooting
within the former dioscoreoid
families
(Dioscoreaceae,
Smilacaceae, Stemonaceae,
Taccaceae,
Trilliaceae, etc.)
was
almost certainly due to convergence
among taxa adapted
to for-
est edges and understory
(Givnish
1979; Chase
et a/. 1995b); and
such a rooting
is
strongly
refuted by all sequence
data so far pro-
duced
(rbcL,
atpB, IBS IDNA, plastid
rys4,
Nadot et al.
1995,
and mitochondrial
aqpA, Davis al al.
1998) as well as evidence
from
intron splicing patterns in
mitochondrial
genes
(Qiu
and
Palmer 1997). Furthermore, the dioscoreoid
lamilies are
hetero-
geneous
in all characters except
those associated with
leaf archi-
tecture.
The
vanilloid
orchids have similar
net-veined
leaves
(Cameron
and Dickison 1998), but
these are clearly orchids
(Chase
rr al. 1994: Cameron et a/.
1999), again demonstrating
that unrelated
monocots can
convergently have evolved
such
dicotlike
leaves.
A rooting among Acoraceae,
alismatid families,
Araceae and
former melanthioid lilies such
as Petrosaviaceae,
Nartheciaceae
and Tofieldiaceae provides
sLrpport for a concept
of plesiomor-
phic characters
for the monocots
(much
like those of
tValker
1986). Obtaining
such a rooting
is
unlikely
if using only
mor-
phological data
because several of
these traits ate unknown out-
side
the monocoryledons. These
would include partially
fused or
follicular fruits, hollow styles, septal
nectaties, P2 rype sieve-ele-
ment plastids
(Behnke
1995) and ensiform
leaves
(Rudall
and
Furness 1997).
To achieve such modifications
relative to their
magnoliid
relatives, the ancestral
monocots must have under-
gone
a drastic reorganisation of their
vegetative organs, and
this
too has been hypothesised by several
previous workers
(Stebbins
1974, amongothers).
In many
groups
of angiosperms, such
reor-
ganisations and shifts of
life history strategies are associated
with
higher rates of sequence
divergence, and this
is certainly true of
the monocots,
which are marked by
long branches relative to
those of
nearby clades, except for Chloranthaceae
and Piperales
(Fig.
2,A) which may have
also
undergone
similar
(but
parallel)

reorganisarions of habit
(Carlquist
1992; Carlquist et al.
1995).
\(ithin
the monocots, several
groups
illustrate
this phenomenon
of innovation and sequence divergence:
Arecaceae,
Dasypogo-
naceae, Dioscoreales, Orchidaceae,
Zir-rgiberales and the
grami-
noid
clades of
Poales. Such episodes of drastic innovation and
sequence
divergence occur wirhin several clades of eudicots as
well and should
make fascinating comparative studies.
There are, however, few
non-monocot
angiosperms that
approach the
habits
of these putatively primitive
monocots,
which should now be the
focus
of better
informed research into
plant
habit.
One of
the most interesting aspects is how some
monocots have reinvented dicot-like vegetative traits or devel-
oped
achlorophyllous forms. Many of these modifications are
also correlated:
net-veined leaves
and
achlorophylly in the Dio-
scoreales
(Burmannia
and Thismia for the latter and Dioscorea for
the former),
Pandanales
(Sciaphila
and
Stemona)
and Vanilloi-
deae of Orchidaceae
(Pseudouanil/a
and Epistephiurz; Cameron
and Dickison
1998); achlorophylly and anomalous secondary
growth in Iridaceae
(Geosiris
for the former and Nivenioideae the
latter); and achlorophylly
and arborescent
primary growth
in
Pandanales
(Sciaphila
and Velloziaceae-Pandanaceae). Furness
and Rudall
(this
volume) suggested that there is also a correlation
between
simultaneous microsporogenesis and either achloro-
phylly
(which
they telmed saprophytism) or wind pollination,
both of which are also correlated with
higher levels
of sequence
divergence
(see
branch
lengths in the
graminoid
clade, Fig. 28).
Understanding
the cause/effect of such phenomena is important
to our
understanding of how diversification and evolutionary
innovation take place.
Patterns of relationship that now are
quite
clear and previously
unpredicted
include the association of Pandanaceae Cyclanth-
aceae with Stemonaceae and Velloziaceae
. This
surprising group,
first observed with
just
rbcL data
(Chase
ar al.
1995a),
agin
raises questions
about vegetative modification within lineages
and how reliably
gross
morphology can be
interpreted.
A close
relationship for the first pair of families was
long
suspected, but
only Dahlgren et al.
(7985)
had
ever suggested that
Arecaceae
and Pandanaceae-Cyclanthaceae were
not reasonably
closely
related, and still they only
hypothesised
that
the relationship
was
most
likely
somewhat
more
distant
and not to other taxa. Now
the task
is to find the evidence supporting such a relationship,
and some of this work is reported in Caddick et a/.
(this
volume).
A possible further addition to
Pandanales is Triuridaceae
(here
represented L>y Sciaphila), with which
they
share successive
microsporogenesis
(however,
this is
putatively plesiomorphic
in
monocots, as opposed
to
simultaneous
microsporogenesis, which
is
the general
rule in Dioscoreales except for Burmanniaceae;
Furness and Rudall, this volume).
Unlike
the situation wirh Sciaphila, Burmanniaceae and Thismi-
aceae appear to contain highly sequence-divergent but
intact
cop-
ies of rbcL and atpB. The association of Dioscoreaceae*Taccaceae
with Thismiaceae and Burmanniaceae
represents
an unlikely
grouping
from the standpoint oftraditional
ideas
ofrelationships
and their
highly and
differently specialised
habits. Nonetheless,
they
have cor-rsistently appeared with Dioscoreaceae-Taccaceae
and in the combined analysis do so with high levels of internal
support
(see
also Caddick
et al.
this
volume, for a further
treat-
SupRnrnvrrrnr cLAssrFrcATroN
oF THE MoNocors
ment of these families). Most of this clade have simultaneous
microsporogenesis, except for Nartheciaceae, which
are
only
weakly associated here
(680/o;
Fig.
2A.) with
Dioscoreales.
Reported first by Givnish et a/.
(lL)99),
we
confirm as well here
that Thurnia and
Prionium
are
closely related. This is another
unpredicted
relationship, but this pair of
genera
are so closely
related that maintenance of separate families seems unnecessary.
Prionium has always been anomalous in
Juncaceae
(e.g.
Munro
and Lir-rder 1997), so its removal seems prudent.
Both
genera
are
fibrous water plants occupying relictual sites
in
the Guyana
ShieldiAmazon basin of South
America and in
southern
Africa,
respectively. In most respects, these two plants are
quite
similar,
although the
former is a mostly submersed aquatic and the
latter
a large erect emergent
(Givnish
et al. 1L)99).
Because
of
their
close relationship and the need to
remove Prionium from
Jun-
caceae, we place these
genera
together
in
Thurniaceae+.
There are
yet
two major issues to be sorted out in monocot sys-
tematics. The first is the exact relationships of the
lilioid orders,
Asparagales,
Dioscoreales,
Liliales, and Pandanales. Although the
orders
themselves now seem well established, with perhaps only
the basal
portions
of Asparagales requiring some
further
atten-
tion, their inter-relationships are unclear and
receive less than
50%o
bootstrap support. The branches
in
this portion ofthe
tree
are all short
(Fig.
2A),
and
it
appears that
what is required is
more data.
The result
of Chase
er a/.
(19951:),
in which these
formed a monophyletic
group
supported by the synapomorphy
of an inferior ovary, has never been reproduced solely
with
molecular data.
\With
the
latter,
the
rypical
pattern
has
been
like
the one seen
here in
the second
analysis, with the orders forming
successive sister groups to the
final Asparagales-commelinoid
pair, but
none
of these
relationships has more than
50%
boot-
strap support.
There are few other obvious characters that could
operate
as
synapomorphies
for any of these
groups.
The phylo-
genetic
structure within these orders is also
largely
established
(see
Caddick et al.,
Fay et al. andRudall et a/., all this volume, for
reviews
of Dioscoreales,
Asparagales and Liliales, respectively),
with
few
genera overall
still requiring attention
(not
all are
sequenced, but
little
doubt
about their relationships exists, e.g..
Ornithoglossum
and
Schelhammera
of Colchicaceae). The three
remaining major
question
marks are in Asparagales:
Herreriopsis
(endemic
to Madagascar), Hesperocallis
(native
to the deserts of
western
North America),
and
Hagenbachia
(occurrins
in
Central
America and the Amazon basin).
The
second
region
of
major
uncertainry
is the inter-relationship
of orders within the commelinoids.
Zingiberales and
Commeli-
nales
appear
now
to be well supported
as sister
orders
(found
as
well in codon-weighted rbcL analyses by Givnish et al. 1999),
although the exact position of Hanguanaceae temains unclear.
Morphological data
(Rudall
et al. 1999) place
Hanguana
as sister
to Zingiberales, whereas the shortest
molecular trees
place
it
nearby
as sister to Commelinales
(which
has moderate
suppoft
with atpB and rbcL;1BS
rDNA is
still
missing for Hanguana).
Bromeliaceae are clearly members
of
Poales
(and
are so
placed
here). Although
Givnish
et al.
(1999)
found that with codon-
weighted rlcl sequence data Mayacaceae and Rapateaceae asso-
ciate with Bromeliaceae, this pattern
is highly
unstable and
receives less
than
50olo
bootstrap support. \fith more data and
il

Mark W. Chase et a/.
dilferent ta-ron sampling,
Mayacaceae in our analyses
(Fig.
28)
are sister
to the sedge clade
(Cyperaceae,
Juncaceae
and Thur-
niaceae),
whereas Rapateaceae fall either as sister to the whole of
Poales or to Bromeliaceae. Like Bromeliaceae,
Typhaceae are
clearly members of Poales, and
Eriocaulaceae-Xyridaceae are sis-
ter to the graminoid
clade
(Anarthriaceae,
Ecdeiocoleaceae, Flag-
ellariaceae,
Poaceae and Restionaceae). The
remaining
ordinal
questions
concern the positions ofArecales and Dasypogonaceae,
and again
the blanches here are short, which means that
more
data
are needed to resolve these patterns clearly.
A Nrw ClassrrrcarroN
FoR MoruocorvlEDoNs
\il/e
now lack information for only rwo monocot families, Cor-
siaceae
and Hydatellaceae
(several
families
are
not represented in
this analysis but have been included
in
other published studies).
The latter has been considered a
member
of
the commelinoid
clade, but this position
is
based on scanty
evider-rce and thelefore
may prove
incorrect. The former may
yet prove
to be part of
Burmanniaceae.
Considering
dre extent of unknown
relation-
ships
at the last monocot symposium
in 1993
(Royal
Botanic
Gardens,
Kew), we
can
now
say
much more with some certainry.
and the
few relationships that we still do not know are compara-
tively insignificant. Such great
advances make feasible and desir-
able a
new
classificatior-r
of the monocot families.
The
process
of producing a classification
is no longer the work of
a single or
few individuals
(i.e.,
experts), but rather with the
application of
cladistic nomenclature the researchers themselves
can produce
a classification directly from their
results.
\7e
have
applied these principles to the
monocotyledons, and such a col-
laborative classification should be considered essentially
author-
less. The era of the
'expert'
taxonomist
who
sifts
the available
information base and
intuitively weights some of these data to
produce a classification
is now over. The new system
is
appropri-
ately synthetic
and therefore reflects all the available data.
Some have argued
that classifications with Linnaean categories
are
now
unnecessary,
that we can substitute the cladogram
with
named nodes for the use of categories
(De
Queiroz
1997; De
Queiroz
and
Gauthier
1994), a.nd we agree that
lor many
studies
a classification does not contain enough detail
to be useful and so
trees from specific studies
must
be
consulted. Lack of stabiliry,
specifically discovering
that clearly monophyletic
groups
were
nested in other taxa of the same rank,
has
been
argued as a prime
reason why hierarchical classification should be
abandoned.
Eliminating the hierarchical
nature
of
classification does not
solve the problem of which groups should
be named, and in this
respect the APG classification
followed the recommendations of
Backlund and
Bremer
(1998)
that list, after monophyly, those
criteria
that should be considered if there are alternative
circum-
scriptions, all of which are
monophyletic.
Systematists
have
spent the last nvo centuries
educating
generations
of scientists in
all fields about
how
to
consult and
use
classifications, and ifwe
now must
teach
them instead the intricacies of how to
interpret
and use our
cladograms, then we not only lose the
results
of
all
our predecessors'
efforts to teach classification but
also have to
introduce a large body of
jargon-filled
and
sometimes
obtuse the-
ory into common use;
this
seems
entirely unnecessary when we
have a simple
hierarchical system that can no\.v be made more
predictive
and
stable
than ever before. It is not the use ofhierar-
t7
chical classification
that was fundamentally
flawed, but rather
specific
classifications that were based on
intuition rather than
ob.jective criteria
for evaluating data. Cladograms
themselves are
not inherently better than
hierarchical classifications
(the
admo-
nition'garbage
in,
garbage
out'comes to mind), but
rather it is
the qualiry of
information that is paramount. \(/e
now have the
possibility
of having predictability and
stability with both clado-
grams
and classification and,
ifthe
classification
does not contain
enough
detail
for the research being
considered, then there are
many ways for the discerning
researcher to find the detail
required
(websites,
such
as TreeBase, etc.). The simplicity of
a
classification serves
an extremely useful and
accommodating
point of
entry into the increasingly
complex and wonderful
world of phylogenetics. \(e
admonish our colleagues to
forego
the discussion of which
we should have, and
focus instead on
how to improve both cladograms
and hierarchical
Linnaean clas-
sification.
They
both serve
important and
reconcilable uses.
There may be some people who are still uneasy
with the thought
ofa system ofclassification
based so extensively on
the results of
DNA sequence studies.
After all, we
have
only
three
genes
sam-
pled
from the more than
10,000 that most plants contain, so
why should this system be so
'definitive'?
This was also the
feel-
ing
of
many of the authors of
earlier
papers
based entirely on
studies of rbcL
gene
sequences
(Chase
er al. 1993 Duvall
et a/.
1993; Chase et al.
I995a). As the evidence
from
other genes
began to appear,
it
became
increasingly clear that
the
patterns
detected were
highly congruent
with those
produced
wirh rbcL.
It is also clear that there was simultaneously
a high degree of sim-
ilarity in the monocot gene trees,
the system of Dahlgren
et al.
(1985)
and the unrooted
network of Stevenson and
Loconte's
(1995)
morphological study
(for
which the rooting
was
different
but the nerwork
highly
similar;
Chase et al.
1995b). Thus, as the
body of congruent
evidence accumulated,
the reasons for not
classif ing became more trivial.
In 1996,
collaborations berween
molecular systematists and
anatomists produced
the first papers
to address some of the dis-
crepancies
in the Dahlgren et dl. system
for the monocoryledons
(Chase
rr a/. 1996; Fay and Chase
1996; Rudall and Chase
1996). The logic behind such
changes was
irrefutable: it is
impossible for there to be
such congruence between unrelated
classes of
information ur-rless both are detecting
evidence of the
same patterns
(i.e.
evolutionary
history). Thus formal revision of
classification became not only possible
but also desirable.
The
new classification presented
in Appendix 1 is the
logical
outcome
of this process.
This
new classification is also unlike any
that have preceded
it in
several
additional ways. First, as stated
above, it is a collaborative
and synthetic system and
includes results produced by
many
researchers.
It could be argued that
it is highly skewed toward
DNA data because
thele are so many
more variable sites in
genes
than
there are morphological characters,
but this is a trivial
objection
since it is also clear that
there is no majol discrepancy
between
patterns found with all
classes ofdata.
Second,
the results upon which the system
are based are repeat-
able
(others
can analyse the matrices which
are available to a1l
researchers) and reproducible
(other
researchers can collect
addi-
tional data and determine
if the same patterns are
found). Thus,

if these
phylogenetic
patterns are refutable, then a classification
built
upon these patterns can also be indirectly
refuted. The
only
exception to this is when there are several alternative classifica-
tions that all recognise
monophyletic
groups;
and even this
pre-
dicament can be
addressed by following a set of reasonable
guidelines
(Backiund
and Bremer 1998) that include considera-
tion of
previous
widespread use,
ma-rimising
support
and evi-
dence
from other data, in addition to size
(e.g.
we
should strive
to eliminate monogeneric
families
simply because
they are
often
based on autapomorphies perceived
to
'isolate'
them from their
more homogenous sister group).
Implementation
of
monophyly
is
paramount,
and the use of these other considerations will
nearly always result in a clear superioriry of one alternarive over
others; in particular, patterns ofsupport
(the
bootstrap,
etc.) and
availability of diagnostic
morphological characters
provide pow-
erlul di.criminatory
ancillary crireria.
Lastly, because
of the large amount of data and high levels of
internal
support,
the new classification should also be predictive
to an unprecedented extent.
This makes
the
new
classification
a
unique and powerful tool
for all
botanists.
It could be argued
that
all
previous
classifications claimed to be predictive, bur it
was always clear to those who wished to use classification as a
tool
that the existence of so many competing systems precluded
this
possibiliry
(i.e.
because they did not agree, they could
not
a1l
be simultaneously correct). The new classification rises above this
problem.
If predictions are made, and the
results
are
found to
refute the
patterns upon
which the classification is based, then
these resuits can be incorporated and the classification thus made
more accurate. There is no need for a competing classification
ever again because this one becomes more valuable with time,
and availabiliry on rhe internet will mal<e the
most recent
changes accessible to everyone
(it
can now be
found at
http:/iuu,'w.systbot. uu.se/classifi cation/classifi
cation9B.html and
http ://www.rbgkew. org. uk/whatson/summary. html).
For those who still harbour doubts about the timeliness of the
production of a new classification, we pose the
following two
questions:
(i)
what
is
desirable about the
alternatives, and
(ii)
what more can you ask of a
classification? The alternatives are all
cleally
inferior,
and
the new classification has everything that
could be
considered desirable in a scientific context: the patterns
upon
which it is based are repeatable, reproducible, and
refutable,
thus making classification
indirectly
able
to
be
evaluated ir-r the
same way that other
hypotheses are evaluated for accuracy. It is
true that we do
not know if reliance upon measures of internal
support
(i.e.
confidence estimates) such as the bootstrap,
jack-
knife and Bremer support can be misleading, but these are the
only measures of internal support currently available to us.
More
important than internal support
is
the argument based on
con-
gruence:
it is most
unlikely
that
such
similar
patterns
could be
produced
with either faulty data or improper methods of analysis.
In reality, the massive increase in
gene
sequence data should de-
escalate one of the more divisive arguments
in
modern systemat-
ics, that over methodological issues
(e.g.
which method of analy-
sis is the most reliable and consistent). Debates over
methodology are still significant,
but these can now assume a
position of secondary
importance
(i.e.
when overall patterns are
not clear) and be better assessed by comparing performance on
subsets of the larger matrix to simulate situations in which pat-
SupRnrlvrrrnr
clAssrFlcATroN oF THE MoNocors
terns are weak
(as
in Siddall 1998). Both parsimony and
ma-ri-
mum likelihood ar-e thought to be inconsistent over some
range
of branchJength
inequalities
(Felsenstein
1978; Siddall
1998),
so we must never consider that any
method is completely reliable
and always pursue additional
studies of other data; lack of pre-
dictability
thus
becomes
the clue that
phylogenetic
patterns are
inaccurate and our method of analysis
inconsistent. \(/e have
faith that the patterns obtained in the angiosperms are
reasona-
bly accurate because they are well corroborated by
many types
of
data, and this means that we can proceed
with a wide range
of
exciting and
groundbreaking
studies on plant
evolutior-r
guided
by the possession of
a tool more predictive than anyone
previ-
ously
thought
possible.
\fle fully expect that the coming years
will be among the most exciting ever, a
new
golden
age of bot-
any, in which major advances will
appear in many fields as
researchers focus on new questions
in
studies
enhanced by an
evolutionary perspective.
AcruowleocEMENTs
'ife
would like to thank the
many people who plovided tissue
samples used
in this
study.
In
particular,
we wish to acknowledge
Dr Hideki Takahashi of the Botanic Garden of Hokkaido Uni-
versity for
srpplyingJaponolirion,
Dr Kenneth Cameron
lor
col-
lecing Petrosauia in Borneo, Drs Maria do Carmo
Amaral
and
Volker Betrich for Mayaca and Dr
John
Dransfield
for
Sciaphila.
Collection of the 1BS rDNA data was supported by
a National
Science
Foundation
Grant
(U.S.A.;DEB-9707868)
to D. and P.
Soltis.
Repeneruces
Angiosperm Phyogeny Group
[APG;
Eds K
Bremer, lY. W.Chase and
P, Stevensl,
(
I 998) An ordinal classiflcation
fbr the fami ies of flow-
ering
p
ants.
Annals of the Mtssouri Botantcol Gcrden 85, 53 553.
Bac<lund,A., and Bremer, K
( 998) To
beornotto
be Prncrples of
class
fcat
on and
monotypic
plant
fam I es. Icxon
47,39 I 404.
Behnl<e, H.-D.
(
1995). P-type s eve-element
plast
ds and the systemat
ics of the Arales
(sensu
Cronquist
l9BB)
wth
S type
plastrds
rn
Ptstio.
Plont
Systemdtics and
Evolutton 195,87 19.
Burger,
W.
C
(
977) The P
perales
and the monocots: alternate
hypotheses fbr the origin of monocotyledonous
flowers. Botanicol
Review 43, 345-393,
Caddrc<,
L. R., Rudal, P.
1.,
Wl<in, P., and Chase, lY. W.
(this
volume).
Yams and the r a ies: systematrcs of
D
oscorea es.
?p. 475487.
Cameron, K, Yl,, and
Dicl<son, W C
(1998)
Fol ar architecture of
van
loid
orchids:
insights nto the evoJution of ret cu ate eaf vena
tion
in monocotyledons. Botanical
lournal
of the
Linneon
Society
128,
15 10.
Cameron, K, lY,, Chase,
lY. W., Whitten, W. lY., Kores, P.
J., Jarre
I D,
C.,
A
bert,
V A., Yukawa, T., Hils, H. G., and Goldman
D. H.
(.999)
A
phylogenetc
anayss of the Orchidaceae: evdence
from
rbcl nucleotide
sequences.
Amerrcan
lournol
ofBotony 86,2A8 ))4.
Carlqurst,
S
(
1992) Wood anatomy and stem of Chloronthus:
summary
of wood anatomy ol Chloranthaceae, with comments
on relation-
shrps, vessel essness, and the orgin
of monocotyledons. /AWA Bu/ie-
rin
13,
3 6.
Carqurst,
S., Dauer, K., and N shimura, S Y
(1995)
Wood
and
stem
anatomy of Saururaceae wth
reference to ecology,
phylogeny,
and
orgin of the
monocotyedons. IAWA Bulletin 16, 133 50.
Chase,
lY. W., Sotis, D. F., Olmstead, R, G,,
lYorgan, D., les, D. H.,
lYish er, B., Duval
,
lY, R,,
Price, R. A., Hils, H. G.,
Qiu,
Y-1., Kron, K.
A,, Rettg,
I.
H., Conti,
E., PaLmer,.1. D., lYanharl,
I.
R., S;'tsma, K.
1.,
t3

Mark \Y. Chase et al.
iYichaeis, H.
J.,
l(ress, W.
j.,
Karo1, l(. G., Clad<, W. D., Hedrdn, lY.,
Gaut, B. S.,
jansen,
R. K., Km, K-J., Wmpee, C. F., Smth,J, F-,, Furnier,
G, R,, Straus, S. H., Xrang
Q
Y,,
Plunl<ett,
G. Y,, Sotis,
P.
S., Swen
son,
S. Y., W I
ams,
S. E.,
Gadel<,
P. A.,
Qu
nn, C.
J.,
Eguiafte, L.,
Golenberg, E,, Learn, G, H., Graham, S, W,, Barrett, S, C, H,, Day-
anandan, S., and
Albert, V A
(
993) Phylogenetics
ofseed
plants:
an analysis of nucleotide sequences from the
pastd gene
rbcl.
Annals of the
Missouri Botantcctl
Carden 80,528 580.
Chase,
lY. W., Cameron, K. lY., H lls, H. G., and
larrell,
D
(
994)
lYo ecu ar systemat cs of the Orchidaceae and other ilord mono-
cots, In
'Proceedings
of the l4th
Word
Orchid Conference,'
(Ed,
A,
Prdgeon.)
pp.
6
73.
(HIYSO:
London.)
Chase, lY, W,, Duva
l,
lY, R,, HiLs, H, G,, Conran,
I,
G,, Cox, A, V,,
Egu
ar1e,
L. E., Harlwel
,
J.,
F-ay, f4. F.,
Cadd cl<,
L. R., Cameron, l(. lY.,
and Hoot, S.
(
995a). lYolecular
phylogenetrcs
of Li ranae. ln
'Yono-
cotyledons: Systematcs and Evoutron'(Eds
P.l. Rudal, P.J
Crbb,
D. F.
Cut er, and C.
J.
Humphries.)
pp.
109 I
37
(Royal
Botan
c Gar
dens: l(ew.)
Chase,
lY. W.,
Stevenson, D,
W,,
W
<in,
P. and Rudall, P,
I
(1995b)
llonocot
systematrcs: a combined analysis.
n
'lYonocotyledons:
Systematcs and Evo ution'
(Eds
P.
J.
Rudal, P.
J.
Crbb, D. F. Cutler,
and C.
I.
Humphries.)
pp.
685
730
(Royal
Botanic
Gardens:
Kew.)
Chase, lY. W., Rudal, P.
J.,
and Conran,'J G
(1996)
New crrcumsc
p-
trons and a
new lamiy
of asparagod
l res: genera formery
ncuded
n Anthercaceae. K.ew Bullettn
5l,
667
68A.
Chase, M. W., and Cox, A. V. 1998, Gene sequences, col aboration,
and analysis of arge data sets.
Australion
Systemdtic
Botony I l,
)t5 7)9.
Dahlgren, R,
and
Rasmussen, l-=. N.
(1983).
Ylonocotyledon
evo
ution:
characters and
phylogenetic
estimation. Evolutionory Brology 16,
255-395,
Dahlgren, R. ll. T., C fford, H. T., and Yeo, P. F-.
(1985).'Ihe
Fam lies of
the lYonocoty edons.'
(Springer-Verlag:
Bedrn).
Davis,J.
1.,
Simmons,
Yl. P.,
Stevenson,
D. W.
and
Wendel
I
F
(1998)
Data decisveness, data
quarty,
and incongruence n
phyogenetc
analysis: an examp e from the monocotyiedons us ng mitonchon-
drial ctpA sequences. Systemdtic
Btology 47,)87 3lA.
De
Queiroz,
K
(
997)
The
Lnnaean
herarchy
and the evolutoniza
tion of taxonomy, with an emphasis on the
problem
of
nomencla
ture. A/iso 15, 125 144.
De
Queiroz,
K,, and Gauth er,)
(
994)
Toward
a
phyogenetc
system
ol b ologrcal
nomenclature. Trends rn Ecology
and
Evolutron
9,
)7
3l
.
Doyle,
J. J.,
and Doye,
)
L
(
987) A rapd DNA
isoation procedure
lrom smal
quantties
of
fresh leaf
tssue.
Phytochemtcol Bullettn 19,
|l t5.
Duva
l, Y. R.,
C egg,
lY. T.,
Chase,
lY. W., Clar<, W. D.,
(ress,
W.
1.,
Hils,
H. G., Eguiate, L. F., Smth,
I.
F-., Gaut, B. S., Zimmer, E. A., and
Learn, G. H
Jr
(
993) Phylogenetc hypotheses forthe monocoty-
ledons constructed
from
rbcl sequence data.
Anncis
of the
Mtssourt
Botanical Carden BO,647 619
Farris,
I.
S.
(
1969) A successive approximations weighting approach to
character weighting. Systemdtlc Zoology
18,
374 385.
Fay, Y. F., and Chase, lY W
(
996). Resurrection of Themidaceae for
the
Brodicec
allance and
recircumscription
of
Al iaceae, Amaryl r-
daceae, and
Agapantho
deae.
Tcron 45,141 451.
Fay, 1Y. F., Ruda I, P.
J.,
Sul ivan, S., Stobarl, K, L,, de Bru
1n,
A.
Y,, Reeves,
C,,
Qamaruz-Zaman,
F,, Hong, W.
P.,
loseph, 1.,
Hahn, W.
J.,
Con
ran,
J,
G,, and Chase, Y W
(th
s vo ume).
Phy
ogenet
c studies of
Asparagales based on four
plastrd
DNA reg ons. Pp. 360-37 .
Felsenste n
I
(
978) Cases
rn
whrch
pars
mony or compatib lity meth-
ods
wll be
posit
ve
y
mrs ead ng. Systematic Zoology 27
,40
-
I 0,
Felsenste n
I
(1985)
Conldence
I mits on
phylogenetrcs:
an approach
us
ng the
bootstrap.
Evolution 39,783-9L
t4
Fitch, W. V
(197
).
Towards deln ng the course of evolution: m nr
mum change lor a specific tree topology. Sfstemdtlc
Zoology 20,
44646
Furness, C. A,, and Rudal P
'J
(thsvoume).The
systematc signfcance
of simultaneous
cytol< nesis dur ng m crosporogenes s n monocoty-
edons. Pp. 89-193.
Givnish, T.
).
979. an the adaptive s
gn f
cance of leaf
form,
In
'
Top cs
n Pant Popuation
Bology'(Eds
O.
T.
Sobrig, S.
jain
G. B
lohnson,
and
P
H. Raven)
pp.
375407.
(Coumbia
Unrversity Press: New
Yorl<.)
Givnish,
T.
1.,
Evans, T. Yl., Pires,
l.
C., and
Sytsma,
K.
)
(
999) Poyphyy
and convergent
evolut on in Commel na es and Commel n dae: ev
-
dence from rbcL sequence daIa, Molecular Phylogenettcs and
Evolu-
tion
12,360
385.
Graur,
D., Duret L.
and Gouy,
[1
(1996)
Phyogenetic
positon
of the
order lagomorpha
(rabbits,
hares and all es). Ncture 379, 333-334
Graybea,
A ( 998) Ls
t betterto add
taxa or characters to a diflcult
phyogenetic problem?
Systemctic Brolc,tgy 47,9 17.
Halanych, K f1
(1998)
lagomorphs misplaced by more characters and
fbwer taxa. Systemctlc Btology 47, l38 46.
Hi lis, D. l'1
(1995)
Approaches for assessing
phylogenetic
accurac).
Syslemctlc Biology 44,3 6.
Hi lis, D. V
(1996)
nlerrng compex
phylogenies.
Ncture 383,
r30
3
r,
Hi lis, D. iY
(998)
Taxonomc samping,
phylogenetc
accuracy, and
nvest
gator
bias Systemot/c
Brology 47,3
B.
Hoot,S. B,, Cuham, A,, and Crane, P R
(
995) The utlityofctpBgene
sequences
in resolving
phylogenetic
re
ationsh
ps:
compar son
wthin rbcl and lBs ribosoma DNA sequences in the Lardzaba-
aceae.
Annols
of the
Missoun Botantcctl
Carden 82,
194 )07.
Kiilerslo, X,, Farrs,
J,
S,, Chase, lY.
W., Bremer, B., Fay,
Y. F., and
Bremer K.
(1999).
Simutaneous
parsimony
jac<knfe
analyss of
2538 rbcl DNA sequences Teveas suppot for malor cades of
green p
ants, land
plants,
seed
plants,
and flower ng
p
ants, P/ont Sys-
temdtlcs and Evolutron 213,259 287,
Ke logg, E. A., and Linder, H P
(1995)
Phy ogeny of Poales, ln
'Yono
cotyledons: Systematcs and Evoluton
(Eds
P,
J
Rudail, P.
J.
Crbb,
D. F. Cutler, and C.
J.
Humphries.)
pp.
5 i I 5'12
(Royal
Botanic Gar-
dens, Kew,)
Les, D. H., and Schneder, E. L.
(1995).
Ihe Nymphaeales, Alismatdae,
and the theory of an aquatic monocotyledon origin, ln
'lYonocoty
edons: Systematics and
Evolution' (Eds. P.
I.
Rudal
,
P.
I.
Cribb,
D. F.
Cuter, and C.
I.
Humphres)
pp
23 42
(Roya
Botanc Gardens:
Kew.)
Led6, lY. D., Crespo, lY.8., Cameron, K. lY., Fay, lY. F., and Chase, lY.
W
(1998)
Systematics of Piumbagnaceae based upon cladrstic
analysis of rbcL sequence data. Systemctlc Botany 23, )l )9.
lYunro,
S.
1.,
and
linder, N P
(1997)
The
embryology and systematic
reationshps of Pdonium serratum
(Juncaceae:
juncaies).
American
Journal
of Botany 84, 850-860,
Nadot,
S.,
Bittar,
G.,
Carter, L., Lacrox, R.,
and
Leleune, B
(
995) A
phylogenet
c ana
ys
s of monocoty edons based on the ch oroplast
gene
rps4, using
parsimony
and a
new numerical phenetics method.
Moleculor Phylogenetlcs
dnd
Evolutton 4 )57 )B).
Nand
,
O., Chase,
Yl. W., and Endress, P K
(
l99B) A comb ned clad s
t c analys s of angiosperms us ng rbcl and non-mo ecular data sets,
Annals of the Mtssourt Botanrcal Carden 85, 37-2 2,
Patlerson, C. D., Wiliams, D. lY., and Humphries, C
'J
(I993)
Congru-
ence between moecular and morphologcal
phyogenes,
Annual
Revtew of Ecology and Systemattcs 24 53
|
BB.
Poe, S,
(
998), Sensitivity ol
phyogeny
estrmation to taxonomic sam
pl
ng.
Systemdtlc
Blology 47, B 3l.
Qiu,
Y. L.,
and
Palmer,
)
D
(1997)
lYitochondrial genome
evolution
and land
plant phylogeny,
American
Journol
of Botany 84, I 3- 14,
[Abstract]

Rudal,
P,
J,,
and Chase, lY. W
(1996)
Systematcs
of
Xanthor
rhoeaceae
sensu /dtor evidence lor
polyphyy.
Telopea 6, 629 647.
Rudal
,
P.J., Furness,
C.
A.,
Fay, Y, F., and Chase, lY, W.
(1997).
lYicro-
sporogenesis and
pol
en suLcus
type
in Asparagales ([i ianae),
Ccnc-
dton
lournol
of
Botony
75, 408-430.
Rudall, P.
J.,
and Furness,
C
A (1997)
Systemat cs of Acorus: ovule and
anther. /nternctlonal
lournal
of Plant Sctence I 58, 540
65
L
Rudal
,
P.
J.,
Stevenson, D. W.,
and Llnder, H, ?
(
999) Struccure and
systematcs of Hanguana,
a
monocotyledon
of uncerlain affnty.
Austrolion Systemctlc Botany 12,
3
I
330.
Ruda l, P,
J,,
Stobart, K, L,, Hong, W.-P., Conran,
j.
G., Furness,
C.
A., Kte,
G., and Chase,
l'4.
W.
(this
volume). Cons der the il es: systematics
of Li iales. Pp.347 359.
Savolanen, V., Chase, X, W., Hoot, S. B., lYorton, C. lY.,
Soltis,
D. E.,
Bayer, C., Fay, M. F.,
de
Bruiln, A. Y.,
Sul ivan, S. and
Q
u, Y,-1.
(2000),
Phylogenetcs offlowerng
plants
based
upon a combined
analysis of
p
astid ctpB and rbcl
gene
sequences.
Systemottc
Btology:
ln
press.
Siddal
,
lY. F
(1998)
Success of
parsimony
in
the lourtaxon case: long-
branch repuisron by lil<el hood n
the
F-arris Zoae.
Aodistics 14,
)09 ))4.
Soltrs, D. E,, Soltrs, P. S., Nicl<rent, D. L.,
lohnson,
L A., Hahn, W.
J,,
Hoot,
S.
8.,
Sweere,
j,
A,, Kuzofl R. l(, i(ron, K. A., Chase, M. W.,
Swensen, S. lY., Z mmer, E. A.,
Chaw, S.
lY.,
Gilespie, L,
J,,
Kress, W,
J,,
and Sy4sma, K
)
(1997a)
Angrosperm
phylogeny
nlerred
lrom
BS ribosomal DNA sequences. Annals of the Mtssouri Botanical
Gorden 84,
149.
Soltis, D. F., Hibsch-letter,
C., Solts,
P.
S., Chase, lY, W,, and Farris,J, S,
(1997b)
Molecular
phylogenetic
relatonships
among angiosperms:
an over^view based on
rbcL
and
IBS
rDNA sequences. ln
'Evolutron
and Diversrflcation of Land Plants'
(Eds
K. watsul<
and
P.
H.
Raven)
pp,
57- 1 78
(Spr
nger-Vedag: Tokyo).
Sotis,
D. E.,
Solts,
P.
S.,
Mor1,
lY,, Chase, X, W,, Savolainen, V., Hoot, S.
B.,
and
lYodon,
C
lY (1998) lnlerring
compLex
phylogenies
using
parsimony:
an empircai
approach using three arge
DNA
data sets
lor ang osperms, Systemctlc Biology 47,32 42.
Stebbins, G
L (1974)
'Fowerng
Pants, Evoluton above the Species
Level.
(Be
knap Press: Cambridge, Massachusetts.)
Stevenson,
D. W.,
and
Loconte,
H
(1995)
Cladistic analysis of mono-
cot
familes. n
'iYonocotyledons:
Systematrcs and Evoluton'
(Eds
P.
J.
Rudal, P.J. Cdbb, D. F. Cutler,
and C.J.
Humphres.) pp.543
578
(Roya
Botan c Gardens: Kew,)
Wall<er,
I.
W (1986)
Classfication and evolution of the monocotyle-
dons. Amertcan
]ournal
of
Botony 73,
7
46.
SupnnrnvrLtar clAsstFrcATroN
oF
THE
MoNocors
Appenorx l.
A phylogenetic classification
of the families of monocoryledons
(families
placed
since publication of the Angiosperm Phylogeny
Group classification, Annak of the Missouri Botanical
Garden,
7998, are marked with an asterisk; orders that are
differently
composed
are
marked
with
S;
families recircumscribed relative to
the APG classification
are
marked
with a
t).
UNPLACED:
Corsiaceae
Petrosaviaceaet
(inciuding
Japonolirionaceae)
Hydatellaceae*
(moved
from Poales)
Aconnles
Acoraceae
Alrsmarnles
Alismataceae
Aponogetonaceae
Araceae
Butomaceae
Cymodoceaceae
Hydrocharitaceae
Juncaginaceae
Limnocharitaceae
Posidoniaceae
Potamogetonace ae
Ruppiaceae
Scheuchzeriaceae
Tofieldiaceae
Zostetaceae
LtLlotDs
Aspanacales
Agapanthaceae
Agavaceae
A]liaceae
Amaryllidaceae
Anemarrhenaceae
Anthericaceae
Aphyllanthaceae
Asparagaceae
Asphodelaceae
Asteliaceae
Behniaceae
Blandfordiaceae
Boryaceae
Convallariaceae
Doryanthaceae
Hemerocallidaceae
Herreriaceae
H
esperocal
I idaceae
Hyacinthaceae
Hypoxidaceae
Iridaceae
Ixioliriaceae
Lanariaceae
Laxmaniaceae
Orchidaceae
t5

Mark W Chase et al.
Tecophilaeaceae
Themidaceae
Xanthorrhoeaceae
Xeronemataceae
Drosconeales$
Burmanniaceae
Dioscoreaceae
Nartheciaceae*
Taccaceae
Thismiaceae
Trichopodaceae
Lrlrnles
Alstroemeriaceae
Campynemataceae
Colchicaceae
Liliaceae
Luzuriagaceae
Melanthiaceae
Philesiaceae
Ripogonaceae
Smilacaceae
Panoarales$
Cyclanthaceae
Pandanaceae
Stemonaceae
Triuridaceae*
Velloziaceae
COMMELINOIDS
UNPLACED:
Dasypogonaceae
Anecares
Arecaceae
Comurlrules$
Commelinaceae
Haemodoraceae
Hanguanaceae*
Philydraceae
Pontederiaceae
Poales$
Anarthriaceae
Bromeliaceae*
Centrolepidaceae
Cyperaceae
F,cdeiocoleaceae
Eriocaulaceae
Flagellariaceae
Joinvilleaceae
Juncaceae
Mayacaceae*
Poaceae
Rapateaceae*
Restionaceae
Sparganiaceae
Thurniaceaet
(including
Prioniaceae)
Typhaceae
Xyridaceaet
(including
Abolbodaceae)
ZtHctgenaLes
Cannaceae
Costaceae
Heliconiaceae
Lowiaceae
Marantaceae
Musaceae
Strelitziaceae
Zingtberaceae
t6