American Journal of Botany 90(9): 1263–1270. 2003.

INVITED SPECIAL PAPER

INCONGRUENCE BETWEEN CLADISTIC AND

TAXONOMIC SYSTEMS1

VERNE GRANT
Section of Integrative Biology, University of Texas, Austin, Texas 78712 USA

Cladistic and taxonomic treatments of the same plant group usually exhibit a mixture of congruences and incongruences. The
question arises in the case of the incongruences as to which version is right and which is wrong. Many cladists believe that cladistics
is a superior approach and gives the best results. There are several conceptual and methodological differences between cladistics and
taxonomy that cause incongruence. One important conceptual difference is the use of different criteria for grouping: order of branching
vs. similarity and difference (clades vs. taxa). Another is the policy regarding paraphyletic groups: to ban them in cladistics but ignore
the ban in taxonomy. These two differences automatically lead to some incongruences. One approach is not right and the other wrong;
each is operating by its own standards. However, when cladists apply the paraphyly rule to a taxonomic system and conclude that it
needs revision to eliminate paraphyly, as cladists often do, they are judging the taxonomic system by a wrong standard. Several
differences between the two schools in the use and handling of characters can also cause incongruence. First consider phenetic
characters. Taxonomy uses a very wide range of these, whereas phenetic cladistics sets restrictions on the selection of characters,
which deprive it of potentially useful evidence. Taxonomic systems generally rest on a broader empirical foundation than phenetic
cladistic systems. Next, consider molecular cladistics, which is the leader in the use of DNA evidence. Two sources of incongruence
between molecular cladistics and taxonomic systems can come into play here. First, the molecular evidence used in cladistics comes
mainly from cytoplasmic organelles, whereas taxonomic systems are based on characters that are determined mainly by the chromo-
somal genome. More generally, the database in a molecular cladogram is, in itself, too narrow to serve as a foundation for an organismic
classification. In cases of incongruence, the molecular evidence can be a reliable indicator of taxonomic relationships sometimes,
misleading other times, and may afford no clear basis for a systematic decision. In this situation, it is helpful, indeed necessary, to
integrate the molecular evidence with the phenetic evidence and bring more characters to bear on the question.

Key words: cladistics; molecular systematics; paraphyly; phylogenetics; systematics; taxonomy.

Research in plant systematics is currently being carried out system needs to be revised to conform to the cladograms. Ex-
by workers belonging to two main rival schools—cladistics amples are numerous in the current journal literature.
(phylogenetics) and taxonomy—which have different concep- It will be shown that conclusions of this sort are unwar-
tual frameworks and use different organizing principles, cri- ranted. Incongruences between cladograms and taxonomic sys-
teria, terminology, and types of evidence. Cladistic and taxo- tems of the same group are to be expected, especially in
nomic studies generally lead to partially different or incongru- groups of large size. This is a result of the differences between
ent results. Not surprisingly, cladists and taxonomists hold cladistics and taxonomy in goals and methods. Cladistics pro-
very different views regarding the value and role of their own duces cladifications, not taxonomic classifications, and the two
field and that of the alternative school. The various differences differ in the way natural variation is grouped (Mayr, 1997;
between the schools will be discussed in this paper. Mayr and Bock, 2002).
Systematic texts written by cladists present taxonomy as an A division between cladistics and taxonomy is inevitable
old-fashioned field that has made valuable contributions in the given the mutually exclusive criteria and methodology of the
past, but has a large subjective component, and is being re- two fields. The division is reinforced by the promotional ef-
placed by the new cladistics with its set of objective formal forts of some strongly committed cladistics. Widespread mis-
procedures (e.g., Wiley et al., 1991; Moritz and Hillis, 1996; understanding of taxonomy among cladists is another contrib-
Judd et al., 1999, 2002). Some cladists even suggest that we uting factor. However, although division cannot be eliminated
abandon the Linnaean hierarchy (de Queiroz, 1997). In other entirely, it can be moderated. There is good reason for workers
words, cladistics is superior to taxonomy. We will examine to combine the best results of the two approaches. Some cla-
dists combine morphology-based taxonomy with cladistic re-
this viewpoint.
sults in the interpretation stage of the investigation; and some
A standard research protocol in plant cladistics is to pick a
taxonomists, myself included, find molecular cladograms to be
group with its existing classification system(s), subject it to a very interesting and valuable, and use them in their work. We
cladistic analysis, often a molecular cladistic one, plot the re- will return to this aspect of the relationships between the fields
sults on cladograms, compare these with the preexisting tax- in the discussion.
onomic system, note the incongruences, and conclude that the Attention is called here to some previous critiques of cla-
distics by taxonomists: Cronquist (1987); Stuessy (1990,
1
Manuscript received 21 November 2002; accepted 11 April 2003.
1997); Mayr and Ashlock (1991); Mayr (1995, 1997); Mayr
The manuscript was read by Ernst Mayr, George Diggs, Billie L. Turner, and Bock (2002); Brummitt (1997, 2002); Turner (1998);
Tom Wendt, Karen A. Grant, and two anonymous reviewers. I gratefully ac- Grant (1998, 2001a); Diggs et al. (1999) (Appendix 6); Diggs
knowledge their helpful suggestions. and Lipscomb (2002).
1263
1264 AMERICAN JOURNAL OF BOTANY [Vol. 90

TAXA VS. CLADES

Taxonomy throughout its long history has grouped natural

variation or biodiversity according to similarity and difference.
As Whewell put it (1859, book 16, p. 357): botanical classi-
fication depends on ‘‘the ideas of Likeness and Differences.’’
The basic units, taxa, are similarity groups that differ from
other similarity groups; and taxonomic classifications are
groupings of lower-level taxa within higher-level taxa. Before
Darwin, the similarity within low-level taxa was attributed to
an Act of Creation and the hierarchy of taxa to the Plan of
Creation. Since Darwin, of course, this pattern has been un-
derstood to be a result of branching phylogeny.
Darwinism brought about modifications of the similarity cri-
terion. The goal is to block out monophyletic similarity
groups. Similarities resulting from convergence are to be seg-
regated into different taxa. It is recognized that microscopic
Fig. 1. Phylogenetic tree of three species groups, A, B, and C. Degree of
or biochemical similarities may sometimes be more reliable phenetic difference between them is symbolized by circles and triangles. A
guides to monophyletic relationships than gross morphological and B are moderately different; C is very different from A and B. A, B, and
characters. C will be grouped in different ways by cladistics and taxonomy. Further ex-
The systematic units of cladists are putative ancestor-de- planation in text.
scendant lineages or clades; and these are grouped into a hi-
erarchical pattern of branching according to the order of
branching. Thus, cladistics groups a different aspect of natural
variation than the one used in taxonomy. MONOPHYLY AND PARAPHYLY
In some cases, there may be little or no difference between
a cladification and a classification. Incongruences between the Monophyly is defined differently in evolutionary biology
two modes of grouping are likely to appear in systems of units and cladistics. In evolutionary biology, the term monophyletic
at middle or higher rank. Consider a midsize or large group refers to a group of organisms that is descended from its most
containing an extant ancestral genus (A) and several different recent known or inferred common ancestor (Haeckel, 1866;
derived genera (D1, D2 . . . ). Cladists will group these all in Mayr and Ashlock, 1991). A monophyletic group in the tra-
the same major clade, whereas a taxonomist will consider the ditional sense of the word may include all or only a part of
character differnces between the A and D genera and put them the descendants of the common ancestor, and the ancestor may
in different groups. be a taxon of various ranks. In cladistics, by contrast, a mono-
phyletic group is a group consisting of all the inferred de-
Brown (1938) found a nectary character that, with other
scendants of an ancestral species (Hennig, 1950, 1966; Mayr
characters, distinguishes an extensive line of evolution in the
and Ashlock, 1991; Judd et al., 1999, 2002).
hypogynous dicots. In some members of the primitive family
In cladistics, a group that includes only some of the de-
Theaceae (Eurya) and the more advanced family Clethraceae scendants of the ancestral species is not monophyletic but
(Clethra), the basal part of the superior ovary is glandular and paraphyletic. The objective of cladistics is to block out mono-
secretes nectar, but there is no disk-shaped nectary. In the hy- phyletic clades. Therefore the elements of a paraphyletic group
pogynous members of the Ericaceae, the basal nectary is a must be rearranged so that they do form one or several clades.
defined disk, and such a disk is present in the more advanced Degree of phenotypic differentiation is not a criterion in the
families Primulaceae and Polemoniaceae. These families range grouping of the organisms into clades.
through a series of grades in other floral characters: free petals Visualize a simple phylogenetic tree consisting of two pri-
and numerous stamens (Theaceae); free petals and 10 stamens mary branches, A and B, derived from a common stem species
(Clethraceae); sympetalous with 10 stamens (Ericaceae); sym- (Fig. 1). A and B are similar but well-differentiated species
petalous with five stamens (Polemoniaceae, Primulaceae) (see groups. A third branch C consisting of species that are phe-
Cronquist [1981] for details). notypically very different from A and B arises from the base
Cladists have restudied the group, using new and old evi- of branch B. Taxonomists, applying the criterion of degree of
dence, and have confirmed the phyletic relationships of the difference, and unconcerned about paraphyly, may treat A and
core group, while making transfers of some peripheral families B as sections of one polymorphous but monophyletic genus
(Angiosperm Phylogeny Group, 1998; Judd et al., 1999, 2002; and C as a distinct monophyletic genus. Cladists, on the other
Anderberg et al., 2002). To emphasize phylogeny, they include hand, may group A, B, and C into one all-inclusive clade or
the entire phylad in a single large order Ericales. split the group into some smaller clades.
Accuracy or inaccuracy is not the issue in this case or in Thus a taxonomic system can be putatively monophyletic
scores of other cases like it. The difference in treatment stems for taxonomists but paraphyletic and in need of revision for
from the criteria used. If monophyly is the primary criterion, cladists. The terminological source of misunderstanding could
we get the cladistic order Ericales, which is very heteroge- be prevented if we used the qualifiers, monophyletic sensu
neous. For taxonomists, the objective is to group the Ericaceae taxonomy and monophyletic sensu cladistics, but we don’t do
and other similar families (Clethraceae, Monotropaceae, Epa- this. Instead we have numerous instances of cladists declaring
cridaceae, etc.) together in a relatively homogeneous, and thus an existing taxonomic treatment to be nonmonophyletic when
more useful, order Ericales. it is quite monophyletic by taxonomic standards.
September 2003] GRANT—INCONGRUENCE BETWEEN CLADISTIC AND TAXONOMIC SYSTEMS 1265

Hileman et al. (2001) carried out a molecular cladistic anal- Johnson et al. (1996) and Porter (1997) made molecular
ysis of Arbutus and related genera in the tribe Arbuteae. Ar- cladistic analyses of the Polemoniaceae, using a chloroplast
butus is a well-defined genus, well differentiated from related gene in the 1996 study and a ribosomal DNA segment in the
genera such as Comarostaphylis, Arctostaphylos, and several 1997 study. The species of Gilia do not form a single clade
small genera (Wells, 2000). In the molecular cladogram, how- in the molecular cladograms, but are dispersed among other
ever, the western North American species of Arbutus are wide- temperate genera of the family. Johnson et al. (1996) conclud-
ly separated from the Mediterranean species of Arbutus. Com- ed that the Gilia of Grant (1959) and Day (1993a, b) is ex-
arostaphylis and Arctostaphylos lie between the two clades of tremely polyphyletic. They use the term ‘‘extreme polyphyly’’
Arbutus. Hileman et al. (2001) conclude that Arbutus is not as a conclusion based on their molecular cladograms. How-
monophyletic. ever, the incongruent part in the cladogram is a mixture of
An alternative and more plausible interpretation of the clad- polyphyly and paraphyly (see Grant [1998, 2001a]).
ogram of Hileman et al. (2001) should be considered. The Porter and Johnson (2000) expanded the molecular evidence
genus Arbutus, consisting of relatively mesic trees, is basal in into a cladification of the family. In this they segregate the
the tribe Arbuteae and has diverged into two branches, a west- Gilia leptalea group and G. tenerrima out of Gilia, as Grant
ern North American branch and a Mediterranean branch, dif- and Day had also done (Grant, 1999; Grant and Day, 1999).
fering in inflorescence and leaf characters. The xerophytic In their system they go on to split the core genus Gilia (of
shrubby genera Comarostaphylis, Arctostaphylos, and related Grant [1999] and Grant and Day [1999]) into six genera: a
small genera (see Wells [2000] for details) are derivatives of shrunken Gilia sensu stricto, three former sections elevated in
Arbutus in the western United States and Mexico. This plau- rank (Aliciella, Giliastrum, Saltugilia), and two new small
sible phylogeny would account for the nested position of the genera (Dayia, Bryantiella) (Porter and Johnson, 2000).
shrubby genera in the cladogram. Some incongruences between the taxonomic and cladistic
New genera originate from preexisting genera. The ancestral treatments are the result of making a new cladistic study of
genus has often radiated into two or more sections in its par- an old system and finding flaws in it, which have also been
ticular adaptive zone. A new divergent species may arise from found and corrected by the taxonomists. These incongruences
a species in a section of the ancestral genus, occupy a new will be set aside for this discussion.
adaptive zone, and, in time, radiate in its zone and become a There remains the enormous difference between the taxo-
new genus. When a cladogram is plotted for a group like this, nomic and cladistic treatments of the core genus Gilia: as a
it will show the pattern just described: primary clades (A and single multisection genus or as six smaller genera. This part
B) for sections of the basal genus and a nested clade (C) for of the overall incongruence is due to the difference between
the derived genus (Fig. 1). taxonomic and molecular cladistic concepts and methods. One
The ban on paraphyletic groups is a source of much unnec- treatment cannot be said to be right or wrong. However, there
essary confusion in systematics. The paraphyly concept be- is an open question concerning the relative merits of the tax-
longs to cladistics, not taxonomy, and when it is applied to onomic and cladistic approaches. Furthermore, it is inappro-
taxonomic classifications, as it commonly is, it conflicts with priate to apply the term ‘‘polyphyletic’’ in its cladistic sense
the similarity/difference criterion. The solution is to simply to a competent taxonomic treatment that differs from a mo-
ignore or reject the paraphyly rule in taxonomy. lecular cladistic treatment.
The paraphyly concept has been criticized by numerous au- The core genus Gilia (with usually x 5 9 but also x 5 8)
thors: Cronquist (1987); Mayr and Ashlock (1991); Mayr is basal within the tribe Gilieae and is putatively ancestral to
(1995, 1997); Mayr and Bock (2002); Rieseberg and Brouillet a group of genera with x 5 7 (Eriastrum, Ipomopsis, and Lan-
(1994); Sosef (1997); Brummitt (1997, 2002); Grant (2001a); gloisia) (Grant, 1959, 2001a). In the molecular cladogram of
Diggs et al. (1999) (Appendix 6); Diggs and Lipscomb (2002). Johnson et al. (1996) the latter three genera are nested between
I have not seen any response to these criticisms. the two subgenera of the core genus Gilia. For Johnson et al.
(1996), this pattern means that Gilia sensu lato is ‘‘polyphy-
PARAPHYLY VS. MONOPHYLY IN GILIA letic,’’ or more correctly nonmonophyletic or paraphyletic. As
I see it, the observed pattern is in agreement with the phylo-
Gilia (Polemoniaceae) is a large, multi-section, and taxo- genetic hypothesis outlined earlier of a polymorphous ances-
nomically difficult genus. It was an artificial catchall genus in tral genus Gilia and a derived group of genera with x 5 7
the nineteenth century but became progressively more natural (Grant, 2001a).
in the twentieth century as disparate blocks of species were
segregated out of it (see Grant [1999]). By the time of the PHENETIC CHARACTERS
Grant (1959) treatment, the more obvious disparate elements
were gone and the genus was thought to be natural. But later External morphological features are always, of necessity, the
work revealed some previously undetected disparate species. first type of character to be used in the macrotaxonomic ex-
The first of these was the Gilia leptalea group, which was ploration of a plant group above the species level. Some of
segregated into a special section for holding purposes (Day, these morphological characters are discrete, but others are
1993a, b) and later transferred to Allophyllum (Grant and Day, quantitative and hard to define. Micromorphological, cytolog-
1999). Several other disparate species were also segregated out ical, physiological, and chemical characters have all found a
of Gilia at this time (Grant, 1999; Grant and Day, 1999). The place in the pantheon.
core genus Gilia as presently constituted (in Grant and Day A broad database consisting of information about diverse
[1999]; Grant [1999, 2001a]) remains large and polymor- characters and types of characters is essential for taxonomy.
phous, with two subgenera and six sections, but has enough One line of evidence frequently conflicts with another line in
common characteristics to justify the conclusion that it is a broad database, and then the question as to how to handle
monophyletic. the discordant variation arises. Picking one prominent char-
1266 AMERICAN JOURNAL OF BOTANY [Vol. 90

acter and using it consistently will not do; long experience logenetic relationships in the Crassulaceae inferred from chlo-
shows that this leads to artificial classifications. The opposite roplast DNA restriction-site variation’’ (van Ham and Hart
extreme is to use all available characters, pool them, and treat [1998], example taken at random).
them by some morphometrical or biometrical method, so as to The DNA that is currently being used in plant molecular
derive a quantitative measure of difference. This treatment has systematic studies comes mainly from chloroplasts, mitochon-
its place, but it also has a big disadvantage. Some characters dria, and ribosomes. These are all organelles, and the first two
are really more valuable markers than others, and the valuable are part of the cytoplasmic genome. Ribosomes are nuclear in
characters can be swamped out by the trivial ones in comput- origin, but are atypical in that category, being encoded in a
ing a multicharacter index value. special site on the nucleolus-organizing chromosome pair. The
A third way, which many taxonomists have adopted, some- chromosomal genome, with the exception of this special site,
times intuitively, is to look for one or a few diagnostic char- is not represented in the sample of DNA being used in plant
acters that are supported by various other characters and that systematics and phylogenetics.
hold up throughout a group. We never know beforehand where The chromosomal genome (exclusive of the nucleolus-or-
in the exo- or endophenotype we will find the diagnostic char- ganizing region) contains the Mendelian genes that determine
acter; it can be anywhere. And there is no formal recipe for the vast majority of phenotypic character differences that are
finding it. The favorable conditions for discovery are a wide used in plant taxonomy and morphological cladistics, as shown
array of characters, an exploratory state of mind, and a trial- by numerous studies (see Grant [1975] for review). Molecular
and-error approach. cladistics and morphology-based systematics are dealing with
There is a subjective element in the third approach. Two different parts of the overall genome.
taxonomists working on the same group can emphasize dif- Furthermore, a molecular cladistic study generally utilizes
ferent primary characters and obtain different results. The one or only a few genes, whereas the phenotypic characters
problem usually becomes resolved eventually with the search used in morphological systematics are expressions of scores
for and finding of additional characters. New characters often or hundreds of genes. The molecular database is good as far
strengthen the case for one of the older classifications. as it goes but is very narrow.
The third approach has a basis in evolution. Organismic Molecular cladograms are being compared with phenotype-
groups are products of adaptive radiation. The ancestral spe- based classifications in one plant group after another in the
cies in an adaptive radiation has some pivotal character that current literature. In any given group, one frequently finds
enables it to occupy and diversify in a new adaptive zone. congruence between cladogram and classification for some
Taxonomists can use the pivotal character or some character characters and incongruence for others. This is as expected on
associated with it as a key character. The key character may the considerations presented previously.
have an obvious adaptive value, as for instance seed pods and Reed and Frankham (2001) assembled 71 cases in plant and
beans with stored nutrients in the Leguminosae, or its adaptive animal groups in which data are available for both molecular
value may be unknown to us, or it may be simply a marker variation and quantitative phenotypic variation. They comput-
of the unseen true pivotal character. In any case, such a key ed the correlation coefficients between the molecular and phe-
character defines a natural group. notypic variation in the 71 datasets. The correlation is gener-
A cladistic analysis usually starts with an existing taxonom- ally weak, with a wide range (r 5 20.88 to 10.90) and a
ic system and uses the taxonomic characters employed in it. mean value of r 5 0.217. These results show that molecular
However, cladistics uses a selected subsample of the taxonom- characters usually do not reflect accurately the quantitative
ic characters. First, it uses the derived characters, but not the phenotypic variation in a group (Reed and Frankham, 2001).
putatively ancestral ones. Second, it uses phenotypic charac- Chloroplast genes have been used extensively in plant cla-
ters that segregate into two or three discrete character states distic studies. Chloroplasts are semi-independent organelles
and rejects quantitative characters that exhibit continuous var- that can become disassociated from their normal nuclear ge-
iation. Cladistics then processes the characters according to set nome (Rieseberg and Soltis, 1991). Consequently, a particular
formal protocol. Finally, cladistics analyzes the phenotypic chloroplast gene may or may not be a reliable indicator of the
variation so as to portray clades, not taxa. overall genome of the species in which it is found, depending
The differences between cladistics and taxonomy in meth- on its history.
odology can be expected to produce different results, and they Cronn et al. (2002) analyzed molecular characters in diploid
do. I mentioned earlier the differences between the two schools species of Gossypium belonging to the various chromosomal
in the way they circumscribe the order Ericales. These differ- genome groups (A, B, D, etc.). They used three types of DNA
ences extend to the other orders of angiosperms as treated and one or more genes in each type: ribosomal DNA (one
taxonomically by Cronquist (1981) and cladistically by the gene), other chromosome segments (11 loci), and cpDNA
Angiosperm Phylogeny Group (1998) and Judd et al. (1999, (four genes). The cladograms for the 12 loci in the two types
2002). of nuclear DNA are in agreement with the other evidence con-
cerning phylogeny of Gossypium. But the cladograms for the
MOLECULAR CHARACTERS four chloroplast genes are incongruent with this evidence and
give incorrect phylogenetic indications.
DNA sequence data have often resolved systematic ques- Chloroplast gene trees differ from nuclear ribosomal DNA
tions where phenotypic evidence was inconclusive and have trees, not only in Gossypium as just noted, but also in Triticeae
revealed relationships that were not suspected on the basis of (Gramineae) (Kellogg et al., 1996), Ceanothus (Rhamnaceae)
traditional characters. DNA evidence is intrinsically powerful. (Hardig et al., 2000), Osmorhiza (Umbelliferae) (Yoo et al.,
Many molecular cladists appear by their actions to regard it 2002), and Phlox (Polemoniaceae) (Ferguson and Jansen,
as a panacea for deciding systematic questions. Papers bearing 2002).
titles like the following are common in the literature: ‘‘Phy- The species of Eucalyptus subgen. Monocalyptus in south-
September 2003] GRANT—INCONGRUENCE BETWEEN CLADISTIC AND TAXONOMIC SYSTEMS 1267

eastern Australia and Tasmania are grouped into several tax- flowering plants. Linnaeus realized that the sexual system was
onomic series. The cpDNA haplotypes are incongruent with artificial and regarded it as a precursor to a natural system yet
the taxonomic subdivisions, but correlate with geographical to be constructed.
areas. The species are known to hybridize when they coexist. Modern revisions and treatments based on molecular clad-
In this group the cpDNA variation appears to reflect chloro- ograms for one or a few genes have the attributes of the old
plast sharing resulting from hybridization and not the primary artificial systems. A molecular character is selected in advance
divergence of the species (McKinnon et al., 1999). by the worker(s); its variant forms are recorded throughout a
A causal connection between hybridization and the spread- group; and the group is subdivided according to the molecular
ing of chloroplast genes has been reported or suggested by a character.
number of authors (see Rieseberg and Soltis [1991] for re- Johnson et al. (1996) surveyed the chloroplast gene matK
view). Examples in addition to Eucalyptus occur in Zea, Tri- throughout the family Polemoniaceae, drew the cladograms,
ticeae, Osmorhiza, and Phlox (Doebley, 1989; Kellogg et al., and blocked out a series of subgroups designated by informal
1996; Ferguson and Jansen, 2002; Yoo et al., 2002). names. Porter and Johnson (2000) then used the matK gene
Some authors have held that the DNA sites assayed in mo- and several other organelle genes as a database for a formal
lecular systematic studies are better indicators of phylogeny system of named taxa. The arrangement of the taxa parallels
than morphological characters because the former are more or the clades in the molecular cladograms. The primary clades
less isolated from environmental selection, whereas the latter become subfamilies, the secondary clades become tribes, and
are exposed to selection (e.g., Sytsma et al., 1990). There was the third-order clades are genera or small sets of genera. The
never much empirical evidence to support this assumption; in classification lacks phenotypic diagnosis of the taxa, but has
fact there is some recent evidence against it. Populations of good descriptions (see Grant [2001a] for analysis).
Triticum dicoccoides and Hordeum spontaneum living in eco- This cladification can be compared with broad-based taxo-
logically different sites in Israel have different forms of ribo- nomic systems of the family: the old one (Grant, 1959), which
somal DNA that are correlated with the environmental vari- was out of date in some respects when molecular studies were
ables. This finding suggests that at least some of the rDNA started in the 1990s, and more recent revisions, which took
variation is determined by selection (Gupta et al., 2002; Nevo the molecular evidence into consideration (Grant, 1998, 1999,
et al., 2002). 2001a; Grant and Day, 1999). The chloroplast gene matK is a
Assume, plausibly, that the organeller DNA sites used in pretty good marker, and there is much congruence between the
molecular systematics often have a selective value. Then taxonomic and molecular systems. But there are also many
consider this assumption in the context of mosaic evolution, incongruences. The molecular system has many unnecessary
the evolution of different characters or genes at different small genera. Some molecular groups are polyphyletic from
rates in the same group. The combination of the two factors failure to use characters of pollen morphology (see Grant
can explain discordance between molecular and phenetic in- [2001a]).
dicators of systematic relationships. Genes in cytoplasmic Molecular cladograms are very valuable. How useful they
organelles can change at a slow rate in a group while the are depends on how they are handled at the stage of interpre-
chromosomal genome determining many phenetic traits tation. A molecular cladogram that is translated directly into
evolves rapidly (or vice versa). Then the molecular evidence a cladification or classification is likely to yield a partly arti-
derived from organelles will indicate close relationships ficial system. The traditional taxonomic approach of integrat-
among lineages in the group, while phenetic characters in- ing new evidence with other lines of evidence and making
dicate more distant relationships between the same lineages informed judgement decisions will generally produce better
(or vice versa). results.

ARTIFICIAL SYSTEMS BACKGROUND KNOWLEDGE

The way in which characters are chosen determines the na- Taxonomists working on a revision or monograph routinely
ture of the resulting classification. One way is to select a char- study their plants in the field and garden, where feasible, as
acter in advance and use it to sort out the variant forms into well as in the herbarium. These contacts build up a body of
classes. This is a recipe for producing an artificial system of background knowledge and perceptions about the plants,
classification. Some of the classes may be natural or mono- which is hard to quantify but which contributes to sound tax-
phyletic, but it is very unlikely that all of them will be so. The onomic decisions.
approach that has worked for building natural classifications Some individual cladists do likewise, but some do not. In
is different. All available characters are considered, with no either case, familiarity with the plants in the field and garden
preconceived favorites; and sets of characters that define gen- is not a requirement in the protocol for a cladistic analysis.
eral similarities and differences are used. Often one or two of Furthermore, there is no place in the cladogram to put subtle
these are useful as key characters, but these are discovered, but significant impressions about plant affinities. Cladifications
not selected beforehand. have to make do with a narrower information base than that
A classical example of an artificial classification is the sex- in taxonomic classifications.
ual system of Linnaeus (1753, 1759). Linnaeus used stamen This difference in approach has a bearing on the problem
number, other stamen characters, and carpel number to group of identification of plant materials used in a revisional
angiosperms into 23 ‘‘classes.’’ The system puts unrelated study. A taxonomist or cladist who has studied a group in
genera in the same class (e.g., Pentandria), puts related genera the field and herbarium can make correct identifications. But
in different classes in some cases, and has some natural or what is the source of the identifications in a molecular cla-
nearly natural classes (e.g., Tetradynamia, equivalent to the distic study by a strictly laboratory worker? Is it the name
Cruciferae). A 24th class, Cryptogamia, was set up for non- on herbarium sheets that provide tissue samples? This could
1268 AMERICAN JOURNAL OF BOTANY [Vol. 90

be very misleading. Identification is no small problem, but basis. Some molecular evidence may point to desirable
is rarely addressed in the materials and methods section of changes in the classification. Or the phenotypic evidence may
cladistic papers. The potential for error in published results be strong enough to override a molecular character in another
is very real. case and may warrant retention of the status quo pending
further study.
DISCUSSION The comparisons just made do not support the presumption
of a superiority of cladistics. The two schools are about equal-
Incongruences are ubiquitous in comparisons of cladograms ly effective, each in its own way, in two indicators (nos. 3 and
with taxonomic classifications. How do we handle them? Ad- 4). Cladistics is superior in one way (no. 6). Taxonomy is the
vocates of cladistics claim that it has a more repeatable and better approach in three indicators (nos. 1, 2, and 5). The rel-
objective method than taxonomy and is the preferred approach ative merits of the two schools are not as simple as the ad-
(e.g., Wiley et al., 1991; Moritz and Hillis, 1996; Judd et al., vocates of cladistics make them out to be.
1999, 2002). In cases of incongruence, the cladogram can be We can also compare the two schools with respect to the
presumed to be correct. Let us examine this assumption. We usefulness of their products for society. There is a long-stand-
will compare the cladistic and taxonomic approaches with re- ing social need for similarity/difference classifications. This is
spect to six indicators of effectiveness in reaching their re- why they have existed since prehistoric times (see Whewell,
spective goals. [1859], book 16, p. 357). First, there were the folk classifi-
(1) Cladistics is based on inferred phylogenies, which makes cations, then the herbals of medicinal plants in the European
for an uncertain foundation. Phylogenies of groups above the Middle Ages, then the pre-Darwinian taxonomists, then evo-
species level are, with rare exceptions, unverifiable hypothe- lutionary taxonomy. In modern societies, taxonomic classifi-
ses. Taxonomic systems are based on observable characters cations are used for identification and information retrieval by
and do not rest on phylogenetic hypotheses. the general public, by people in technological fields, and in
(2) Cladograms can be tested for accuracy in special cases biological research (including cladistics).
in which we have a known pedigree, as in cultivated plants Cladistics has no such general constituency. Phylogeny is
and long-term selection experiments. Such tests have been car- of only secondary interest to most people including most kinds
ried out in species of Avena, Hordeum, Gilia, and Helianthus. of biologists. Who needs a crocodile-bird clade when we al-
In each case the cladogram differed significantly from the ready have a bird taxon and a crocodile taxon?
known pedigree. Both reticulate and dichotomous pedigrees Some cladists see systematics going in the direction of cla-
were represented in the set of cases (Grant, 2001b). It is well distics in the future (e.g., Wiley et al., 1991; Moritz and Hillis,
known that cladograms do not pick up reticulate evolutionary 1996; de Quieroz, 1997; Judd et al., 1999, 2002). This pre-
patterns; we find that they do not always get dichotomous diction doesn’t take into account the special merits and general
branching patterns right either. usefulness to society of taxonomy. It is more likely that, for
(3) Taxonomic methods are designed to organize natural the present and immediate future, we will have a dual presence
variation into similarity groups (taxa) and hierarchies of such of both taxonomy and cladistics (Brummitt, 1997, 2002). It is
groups (classification systems). Cladistics does not do this; it also likely and desirable that the two schools will interact syn-
identifies clades and produces cladograms. Some incongruenc- ergistically with one another.
es between classifications and cladograms of the same group Sorting organismic variation into natural groups is not an
develop as a result of the differences in the units used, taxa exact science. The variation is multifaceted, the evolutionary
or clades. In such cases, both systems may be correct by their processes involved are multifarious, and the true phylogeny is
own standards. usually unknown. Taxonomy accepts these realities and rec-
(4) Cladistics places an arbitrary ban on paraphyletic ognizes the need for some subjective judgements in building
groups, whereas paraphyly is a nonissue in taxonomy. Appli- classifications. Cladistics was invented for the purpose of im-
cation or nonapplication of the paraphyly rule can make a proving on taxonomy (Hennig, 1950, 1966). Classification was
large difference in a system. The system can have one arrange- to be based on phylogeny, and phylogeny was to be recon-
ment of groups in a taxonomic classification and a different structed by an explicit objective protocol. How well does this
arrangement in its cladistic version because of the paraphyly plan work in practice? It works sometimes, but it is not good
rule alone. Where paraphyly is a source of incongruence, the enough. In cases in which cladifications can be compared with
incongruence should be regarded as an artifact of the cladistic up-to-date taxonomic classifications, the former are not con-
method and is not a sufficient reason to revise the taxonomic sistently better, and sometimes the classifications are better, as
system. we have seen.
(5) One of the big differences between taxonomic classifi- In this situation, the best results can be obtained by com-
cations and cladograms is the breadth of the information base. bining the strong points of the two approaches. For exam-
This base is very broad in taxonomy in which no class of ple: (1) The taxonomic order Ericales sensu stricto of older
characters is excluded from consideration. It is narrower in authors (Cronquist [1981] and others) is more generally use-
morphological cladistics because of the exclusion of certain ful than the cladistic order Ericales sensu lato (s.l.) of Judd
types of potentially valuable characters, and it is extremely et al. (1999, 2002). Preserve the Ericales sensu stricto for
narrow in molecular cladistics. Taxonomy has a more inclusive general use (and retain Ericales s.l. in cladistics.) (2) Let’s
empirical foundation than cladistics. not attempt to apply the paraphyly rule to taxonomic clas-
(6) One area in which cladistics is superior to traditional sifications where it does not fit and only causes distortions
taxonomy is the use of DNA evidence. Molecular cladograms in the system.
often show some incongruences with existing classifications. (3) The greatest strength of cladistics lies in its use of mo-
The conflict between molecular and traditional characters lecular evidence. The molecular cladograms can be and are
presents a problem and should be handled on a case-by-case translated directly into sets of molecular clades. However, it is
September 2003] GRANT—INCONGRUENCE BETWEEN CLADISTIC AND TAXONOMIC SYSTEMS 1269

highly desirable to incorporate the molecular evidence in tax- HAECKEL, E. 1866. Generelle Morphologie der Organismen. Georg Reiner,
onomic classifications. To do this effectively, the molecular Berlin, Germany.
HARDIG, T. M., P. S. SOLTIS, AND D. E. SOLTIS. 2000. Diversification of the
evidence should be handled in the taxonomic way; that is, North American shrub genus Ceanothus (Rhamnaceae): conflicting phy-
weigh molecular characters in the balance with other types of logenies from nuclear ribosomal DNA and chloroplast DNA. American
characters, and exercise taxonomic judgement to resolve con- Journal of Botany 87: 108–123.
flicts in the evidence. HENNIG, W. 1950. Grundzüge einer Theorie der phylogenetischen Systematik.
Deutscher Verlag, Berlin, Germany.
HENNIG, W. 1966. Phylogenetic systematics. University of Illinois Press, Ur-
LITERATURE CITED bana, Illinois, USA.
HILEMAN, L. C., M. C. VASEY, AND V. T. PARKER. 2001. Phylogeny and
ANDERBERG, A. A., C. RYDIN, AND M. KALLERSJO. 2002. Phylogenetic re- biogeography of the Arbutoideae (Ericaceae): implications for the Mad-
lationships in the order Ericales s.l.: analyses of molecular data from five rean-Tethyan hypothesis, Systematic Botany 26: 131–143.
genes from the plastid and mitochondria genomes. American Journal of JOHNSON, L. A., L. SCHULTZ, D. E. SOLTIS, AND P. S. SOLTIS. 1996. Mono-
Botany 89: 677–687. phyly and generic relationships of Polemoniaceae based on matK se-
ANGIOSPERM PHYLOGENY GROUP. 1998. An ordinal classification for the quences. American Journal of Botany 83: 1207–1224.
families of flowering plants. Annals of the Missouri Botanical Garden JUDD, W. S., C. S. CAMPBELL, E. A. KELLOGG, AND P. F. STEVENS. 1999.
85: 531–553. Plant systematics; a phylogenetic approach, 1st ed. Sinauer Associates,
BROWN, W. 1938. The bearing of nectaries on the phylogeny of flowering Sunderland, Massachusetts, USA.
plants. Proceedings of the American Philosophical Society 79: 549–595. JUDD, W. S., C. S. CAMPBELL, E. A. KELLOGG, P. F. STEVENS, AND M. J.
BRUMMITT, R. K. 1997. Taxonomy versus cladonomy, a fundamental contro- DONOGHUE, 2002. Plant systematics; a phylogenetic approach, 2nd ed.
versy in biological systematics. Taxon 46: 723–734. Sinauer Associates, Sunderland, Massachusetts, USA.
BRUMMITT, R. K. 2002. How to chop up a tree. Taxon 51: 31–41. KELLOGG, E. A., R. APPELS, AND R. J. MASON-GAMER. 1996. When genes
CRONN, R. C., R. L. SMALL, T. HASELKORN, AND J. F. WENDEL. 2002. Rapid tell different stories: the diploid genera of Triticeae (Gramineae). System-
diversification of the cotton genus (Gossypium, Malvaceae) revealed by atic Botany 21: 321–347.
analysis of sixteen nuclear and chloroplast genes. American Journal of LINNAEUS, C. 1753. Species plantarum, 1st ed. Stockholm, Sweden.
Botany 89: 707–725. LINNAEUS, C. 1759. Systema naturae, 10th ed. Stockholm, Sweden.
CRONQUIST, A. 1981. An integrated system of classification of flowering MAYR, E. 1995. Systems of ordering data. Biology and Philosophy 10: 419–
plants. Columbia University Press, New York, New York, USA. 434.
CRONQUIST, A. 1987. A botanical critique of cladism. Botanical Review 53: MAYR, E. 1997. This is biology; the science of the living world. Harvard
1–50. University Press, Cambridge, Massachusetts, USA.
DAY, A. G. 1993a. Gilia. In J. C. Hickman [ed.], The Jepson manual: higher MAYR, E., AND P. D. ASHLOCK. 1991. Principles of systematic zoology, 2nd
plants of California, 828–835. University of California Press, Berkeley, ed. McGraw-Hill, New York, New York, USA.
California, USA. MAYR, E., AND W. BOCK. 2002. Classifications and other ordering systems.
DAY, A. G. 1993b. New taxa and nomenclatural changes in Allophyllum, Zeitschrift für zoologische Systematik und Evolutionsforschung 40: 169–
Gilia, and Navarretia (Polemoniaceae). Novon 3: 331–340. 194.
DE QUEIROZ, K. 1997. The Linnean hierarchy and the evolutionization of MCKINNON, G. E., D. A. STEANE, B. M. POTTS, AND R. E. VAILANCOURT.
taxonomy, with emphasis on the problem of nomenclature. Aliso 15: 1999. Incongruence between chloroplast and species phylogenies in Eu-
125–144. calyptus subgenus Monocalyptus (Myrtaceae). American Journal of Bot-
DIGGS, G. M., AND B. L. LIPSCOMB. 2002. What is the writer of a flora to any 86: 1038–1046.
do? Evolutionary taxonomy or phylogenetic systematics? Sida 20: 647– MORITZ, C., AND D. M. HILLIS. 1996. The evolution of molecular system-
674. atics. In D. M. Hillis, C. Moritz, and B. K. Mable [eds.], Molecular
DIGGS, G. M., B. L. LIPSCOMB, AND R. J. O’KENNON. 1999. Illustrated flora systematics, 2nd ed., 1–13. Sinauer Associates, Sunderland, Massachu-
of north central Texas. Botanical Research Institute of Texas, Fort Worth, setts, USA.
Texas, USA. NEVO, E., A. B. KOROL, A. BEILES, AND T. FAHIMA. 2002. Evolution of wild
DOEBLEY, J. 1989. Molecular evidence for a missing wild relative of maize emmer and wheat improvement; population genetics, genetic resources,
and the introgression of its chloroplast genome into Zea perennis. Evo- and genome organization of wheat’s progenitor, Triticum dicoccoides.
lution 43: 1555–1559. Springer-Verlag, New York, New York, USA.
FERGUSON, C. J., AND R. K. JANSEN. 2002. A chloroplast DNA phylogeny PORTER, M. M. 1997. Phylogeny of Polemoniaceae based on ribosomal in-
of eastern Phlox (Polemoniaceae): implications of congruence and in- ternal transcribed DNA sequences. Aliso 15: 57–77.
congruence with the ITS phylogeny. American Journal of Botany 89: PORTER, M. M., AND L. A. JOHNSON. 2000. A phylogenetic classification of
1324–1335. Polemoniaceae. Aliso 19: 55–91.
GRANT, V. 1959. Natural history of the phlox family. Martinus Nijhoff, The REED, D. H., AND R. FRANKHAM. 2001. How closely correlated are molecular
Hague, Netherlands. and quantitative measures of genetic variation? A meta-analysis. Evolu-
GRANT, V. 1975. Genetics of flowering plants. Columbia University Press, tion 55: 1095–1103.
New York, New York, USA. RIESEBERG, L. H., AND L. BROUILLET. 1994. Are many plant species para-
GRANT, V. 1998. Primary classifications and phylogeny of the Polemoniaceae, phyletic? Taxon 43: 21–32.
with comments on molecular cladistics. American Journal of Botany 85: RIESEBERG, L. H., AND D. E. SOLTIS. 1991. Phylogenetic consequences of
741–752. cytoplasmic gene flow in plants. Evolutionary Trends in Plants 5: 65–
GRANT, V. 1999. Classification of the genus Gilia (Polemoniaceae). Phyto- 84.
logia 84: 69–86. SOSEF, M. S. M. 1997. Hierchical models, reticulate evolution, and the in-
GRANT, V. 2001a. A guide to understanding recent classifications of the fam- evitability of paraphyletic supraspecific taxa. Taxon 46: 75–85.
ily Polemoniaceae. Lundellia 4: 12–24. STUESSY, T. F. 1990. Plant taxonomy. Columbia University Press, New York,
GRANT, V. 2001b. Tests of the accuracy of cladograms in Gilia and some New York, USA.
other angiosperm genera. Plant Systematics and Evolution 230: 89– STUESSY, T. F. 1997. Classification: more than just branching patterns of evo-
96. lution. Aliso 15: 113–124.
GRANT, V., AND A. DAY. 1999. Transfer of some species from Gilia to Al- SYTSMA, K., J. F. SMITH, AND L. D. GOTTLIEB. 1990. Biogeography and
lophyllum and Tintinabulum, and the effects of the transfer on the generic evolution of morphology, breeding systems, flavonoids, and chloroplast
definition of Gilia (Polemoniaceae). Phytologia 84: 368–382. DNA in the four Old World species of Fuchsia (Onagraceae). Systematic
GUPTA, P. K., P. K. SHARMA, H. S. BALYAN, J. K. ROY, S. SHARMA, A. Botany 15: 280–295.
BEHARAV, AND E. NEVO. 2002. Polymorphism at rDNA loci in barley TURNER, B. L. 1998. Plant systematics: beginnings and endings. Aliso 17:
and its relation with climatic variables. Theoretical and Applied Genetics 189–200.
104: 473–481. VAN HAM, R. C., AND H. HART. 1998. Phylogenetic relationships in the
1270 AMERICAN JOURNAL OF BOTANY [Vol. 90

Crassulaceae inferred from chloroplast DNA restriction-site variation. WILEY, E. O., D. SIEGEL-CAUSEY, D. R. BROOKS, AND V. A. FUNK. 1991.
American Journal of Botany 85: 123–134. The compleat cladist; a primer of phylogenetic procedures. Museum of
WELLS, P. V. 2000. The manzanitas of California, also of Mexico and the world. Natural History, University of Kansas, Lawrence, Kansas, USA.
Published by the author, University of Kansas, Lawrence, Kansas, USA. YOO, K. O., P. P. LOWRY, AND J. WEN. 2002. Discordance of chloroplast and
WHEWELL, W. 1859. History of the inductive sciences, 3rd ed. Appleton, nuclear ribosomal DNA data in Osmorhiza (Apiaceae). American Journal
New York, New York, USA. of Botany 89: 966–971.