Printer Friendly

Numerical and comparative analyses of the modern systems of classification of the flowering plants.

II. Introduction

Often we are given no idea what evidence is used in erecting the classification. - M. T. Ghiselin (1984)

Plant taxonomists can refer to four major and modern classification systems: Cronquist (1988), Dahlgren et al. (1985; see also G. Dahlgren, 1989a, 1989b), Takhtajan (1987), and Thorne (1992b). These systems are important because they serve as a basis for phylogenetic studies, whether the latter use morphological or molecular data. For example, Chase et al. (1993) wrote that their analysis stems from the works of Cronquist (1981) and of Dahlgren et al. (1985). Also, in the discussion of their results, they compare intuitively their cladogram, obtained from rbcL data, to Cronquist's, Dahlgren's, Takhtajan's, and Thorne's classifications. Since studies in molecular systematics increase in number and because they use the modern classifications as a starting point, it is becoming important to understand the structure of these different classifications and to define their underlying principles and concepts. Yet modern classifications are little studied; we do not know, among other things, if they differ. A more profound knowledge of these recent classifications matters not only because we use them as a basis for our taxonomic analyses but also because we are comparing them to our results, judging qualitatively whether they are congruent or not. Indeed, such comparisons are precarious, for it is difficult to compare what is known to what is hardly known.

Apart from a few studies that used the size distribution of taxa (Holman, 1985, 1992; Cronk, 1989, 1990), historical studies of classification schemes have remained qualitative (Lu, 1981; Barabe & Brouillet, 1982; Barabe, 1984, 1993; Stevens, 1984b, 1986, 1994). Sivarajan (1991) mentioned that knowing the history of classification systems is the only way to foster in taxonomists a true understanding of the modern systems. For Stevens (1994: 269), "systematics can free itself of its undue reverence for tradition if systematists come to understand more about the development of that tradition in its historical context."

Nonetheless, the history of plant systematics awaits further development. Apart from works on Linnaeus or Adanson, few studies have been done on other plant taxonomists (Turrill, 1963; Kiger, 1971; Barabe & Brouillet, 1982; Mabberley, 1985; Williams, 1988; Barabe & Vieth, 1990; Stevens & Cullen, 1990; Endress, 1993; Stevens, 1984a, 1984b, 1991, 1994; Cain, 1994; Cuerrier et al., 1996). Among these, few have insisted on recent aspects of taxonomy.

Benson (1957), Lu (1981), and Woodland (1991) have graphically represented a theoretical genealogy of classification systems. Figure 1 shows that, for these authors, all modern systems originate directly or indirectly from the Bessey system; one should recognize, though, that Hallier's system is not the German counterpart of Bessey's system (see Cuerrier, in press), contrary to what Core (1955), Benson (1957), Stace (1989), and Woodland (1991) have written. Cronquist (1965: 285) stipulated: "Nearly all modern systems of angiosperms fall into the de Candolle-Bentham and Hooker-Bessey tradition." Cuerrier et al. (1992) have otherwise shown that if Engler's classification topology differs from Bessey's, their ordinal circumscriptions (i.e., family affinities) stay statistically similar.

Therefore, in this article, we wish to extend our knowledge of the modern and past systems of classification of the Angiosperms by comparing classifications of Cronquist (1988), Dahlgren et al. (1985), G. Dahlgren (1989a), Takhtaj an (1987), and Thorne (1992b) with one another, using statistical tools. A comparative study of the principles underlying these systems will parallel the numerical comparisons and they will serve to interpret the statistical results. Stevens (1986, 1994) mentioned that modern classifications were derived from those of the past. Hence, we will also compare the modern classifications to those published at the beginning of this century by Bessey (1915), Engler (1909), and Hallier (1912). Takhtajan's (1987) classification will be compared to Gobi's (1916) since Takhtajan himself asserted that he was influenced by Gobi. Our study will answer specifically the following questions:

* Are modern classifications statistically similar?

* Are they significantly different from classifications published at the end of the 19th and beginning of the 20th centuries?

* Are they more similar to Bessey's classification than to other past classifications as underlined by many authors (see above)?

* What are the relationships among the seven classifications studied in this article? Are these relationships the same as the ones depicted by Benson, Lu, and Woodland?

Answering these questions will help to verify the thesis defended, among others, by Stevens (1994), who stipulated that modern classifications have changed little over the last two centuries. Is this absence of changes visible solely in classification topologies or is it equally visible in the classificatory approach and in the principles used by taxonomists?

III. Material and Methods

For numerical analyses, only the most recent classification published by modern authors has been used: Cronquist (1988), Dahlgren et al. (1985; for Dicotyledons, G. Dahlgren, 1989a), Takhtajan (1987), and Thorne (1992b). [Takhtajan's (1997) new book, which includes a classification scheme, was only available after all comparisons were completed and the article written.] As for the past classifications, Bessey's (1915), Engler's (1909), Gobi's (1916), and Hallier's (1912) were chosen. Analyses have been done independently for Monocotyledons and Dicotyledons, because these two group circumscriptions are identical among all authors. Additionally this was done to avoid an overestimation of similarity among classifications caused by identical partitions at the most inclusive rank (Lapointe & Legendre, 1990). Moreover, monocotyledon classifications are usually formed independently of those of Dicotyledons.

Two types of statistical analyses were first performed: Mantel and consensus tests. The former is a local test and the latter is a global one. That is, the Mantel test compares terminal taxa only, while the consensus test takes into account taxonomic ranks and classification topologies (or structures) as well. All three formal properties of nested classification are thus analyzed (Lapointe & Legendre, 1990). One must realize, though, that these tests do not evaluate another component of evolutionary classifications: the linear sequence of taxa (from ancestor to descendant). This information is more or less clearly marked in classifications and thus remains elusive. For example, for Thorne, each taxon in his classification is analogous to a ladder at the bottom of which lie primitive groups and at the top, more advanced groups. It is impossible to statistically analyze this information since it is not always available and, moreover, because Linnean classification expresses it only with great difficulty (Cuerrier et al., 1992, 1996; Rieppel, 1991). Even though authors establish ancestor-descendant sequences, they acknowledge at the same time that taxa are not direct ancestors but solely the nearest groups.

For both Mantel and consensus tests, matrix dimensions have to be identical for objects to be comparable. This constraint requires that only families recognized by authors in their respective classifications must be used for comparative purpose (Cuerrier et al., 1992), even if families of one author are considered as subfamilies by another, which happens frequently. In the latter case, we have ignored the information. This constraint is a lesser evil because it has more to do with similarities than with differences at distinctive ranks; and similarity among classifications is already strong (Table I). Between modern and past classifications an important disparity in family number can be seen: Modern classifications have more families. Therefore, comparisons involving these two groups of classifications have been performed with those subfamilies (mostly those of past classifications) taken into account when recognized at the family level by the other author (mostly modern).

When simultaneous statistical tests are carried out, p-values (i.e., level of significance) need readjustment (Wright, 1992). To do this we have chosen the Bonferroni correction, which divides the level of significance by the number of tests performed. This correction limits the probability of a type I error (erroneous rejection of the null hypothesis) and renders more rigorous the series of tests used.
Table I

Results of the consensus (NISI) and Mantel (r) tests between modern
classifications. Mantel tests were performed at the ordinal level,
and only between the monocotyledon classifications, due to the
algorithm limits. For [Alpha] = 0.01, the associated probability of
the Normalized Intermediate Similarity Index, P(NISI), and of the
standardized Mantel statistic, P(r), is p = 0.00167, after
correcting the level with the Bonferroni method (level/number of
pair-wise comparisons). The number of permutations has been
established at 4000 for the Monocotyledons and at 10000 for the

Classifications NISI P(NISI) r P(r)


Monocotyledons (106 OTUs) 0.85021 0.00025 0.59443 0.00049
Dicotyledons (306 OTUs) 0.74547 0.00009 ? ?


Monocotyledons (107 OTUs) 0.73486 0.00025 0.36922 0.00049
Dicotyledons (313 OTUs) 0.88514 0.00009 ? ?


Monocotyledons (91 OTUs) 0.78652 0.00025 0.57833 0.00049
Dicotyledons (290 OTUs) 0.74141 0.00009 ? ?


Monocotyledons (111 OTUs) 0.80338 0.00025 0.55035 0.00049
Dicotyledons (361 OTUs) 0.78661 0.00009 ? ?


Monocotyledons (103 OTUs) 0.91234 0.00025 0.89108 0.00049
Dicotyledons (318 OTUs) 0.87091 0.00009 ? ?


Monocotyledons (103 OTUs) 0.64725 0.00025 0.26337 0.00049
Dicotyledons (331 OTUs) 0.75969 0.00009 ? ?

Figure 2 pictures the creation of the matrices used by both tests and shows the steps leading to meta-analyses (clustering analyses and ordinations of the seven classifications, i.e., a classification of classifications).


Based on the calculation of a consensus index, this test is meant to measure the similarity between two classifications. As was pointed out previously, Lapointe and Legendre's (1990) algorithm was used because it takes advantage of the three properties of dendrogram (taxonomic ranks, family position, and topology). Their algorithm proceeds through a double permutation of the ultrametric matrices formed from the above-mentioned classifications. Cuerrier et al. (1992) applied this consensus test in their comparison of Bessey's and Engler's classifications. The null hypothesis works as follows: The two classifications compared are no more similar than classifications randomly generated. If the null hypothesis is rejected, then the two classifications compared are considered to be statistically similar. Furthermore, the consensus indices obtained (which are normalized intermediate similarity indices, or NISI) can serve to form an input matrix that can be used for clustering analyses and ordinations. Contrary to matrix correlations (r) obtained from Mantel tests, NISI are not affected by the number of objects or matrix dimensions (F.-J. Lapointe, pers. comm.).


This permutation test was first conceived by Mantel (1967) to statistically compare two matrices of the same dimension and produce a matrix correlation. Cuerrier et al. (1990, 1992) give a detailed description of the Mantel test as it is applied in this paper. The null hypothesis can be stated as follows: Matrix A does not statistically resemble matrix B more than does a matrix randomly generated. So the null hypothesis is accepted if the content of taxa differs between the classifications being compared, whereas it is rejected if contents are similar. Results of Mantel tests among dicotyledon classifications of modern authors have not been retained since the operational taxonomic units (OTUs; i.e., number of families) exceeded the algorithm limit. However, results of Mantel tests between matrices at the rank of subclass and superorder are discussed; they were all significant, and could possibly be correctly interpreted.

Matrix correlations of the Mantel tests have further served to produce input matrices for clustering analyses and ordinations of the seven classifications. However, matrix correlations are affected by matrix dimensions (number of objects; i.e., OTUs). Podani (pers. comm.) mentioned that interpretations of these analyses using matrix correlations based on different OTUs must be made with extreme precaution (see below).


As stated above, the results of both the Mantel and consensus tests can serve to generate classifications (dendrograms and ordinations) of the classifications compared, so that these can be called metataxonomic analyses or recta-analyses, in which OTUs themselves happen to be the seven classifications.

Two different clustering strategies were used: single linkage and complete linkage. When dendrograms from both types of analysis had identical topologies, only the one produced by single linkage was reproduced in this paper. One should know that single linkage merges objects and groups together while contracting the space of reference; thus a chaining of groups with only one object can follow (Sneath & Sokal, 1973; Abbott et al., 1985). On the contrary, the dilation of space occurring with complete linkage favors the distance between clusters and creates, in general, a smaller number of clusters. Thus, complete linkage can reduce the similaxity between objects or between clusters which can lead to artificial clusters (Abbott et al., 1985). Program NCLAS of the SYN-TAX IV software (Podani, 1990) was used to perform all clustering analyses. Once dendrograms of monocotyledon and of dicotyledon classifications were independently obtained, repeated analyses were done using an input matrix of combined results (i.e., NISI of Monocotyledons + NISI of Dicotyledons/2). This procedure is justified because the index used is normalized.


Again, using the consensus test results (NISI) as an input matrix, principal coordinate analyses (PRINCOOR) were carried out. Since the Mantel test results in matrix correlations (r), and not in similarity indices, principal component analyses (PRINCOMP, option "PCA using correlation") were chosen instead. As was done with the clustering analyses, consensus test results obtained by comparing monocotyledon classifications with one another and dicotyledon classifications with one another were combined. The new input matrix served to produce an ordination of angiosperm (Monocotyledons and Dicotyledons) classifications.

IV. Results


Results taken from the multiple consensus and Mantel tests between modern classifications of Monocotyledons and of Dicotyledons are compiled in Table I.

Results from both tests are in agreement: The null hypothesis is always rejected. Indeed, monocotyledon classifications of Cronquist, Dahlgren et al., Takhtajan, and Thorne are statistically similar, as much for their overall structure as for their content of orders. Modem authors share similar monocotyledon and dicotyledon classifications at the family level and above. Minor differences seen among modern classifications are not important enough for statistical results to be affected.


Table II summarizes the results of the consensus and Mantel tests between modern and past classifications. In general, the classificatory structure does not differ significantly among authors. However, the null hypothesis is accepted for some comparisons: between the classifications of Cronquist and Bessey (Dicotyledons), between those of Dahlgren and Engler (Dicotyledons), and between those of Takhtajan and Bessey (Monocotyledons). These classifications differ statistically. Even though the null hypothesis is also accepted when Cronquist's and Bessey's classifications (Monocotyledons) are compared - or Cronquist's and Engler's (Dicotyledons), or Takhtajan's and Bessey's, or Takhtajan's and Engler's (Dicotyledons), or Thorne's and Engler's (Dicotyledons) - the probabilities obtained are near the level of significance. Indeed, without the Bonferroni correction, results would have emphasized the similarity among classifications. This correction is used as a conservative measure that accentuates the rigor when simultaneous tests have been repeated. The dicotyledon and monocotyledon classifications produced by Hallier are both similar to modern classifications. The monocotyledon classification of Engler is also similar to those of the modern authors. Therefore, differences are observed between modern and past classifications only when Bessey's (Monocotyledons and Dicotyledons) or Engler's (Dicotyledons) classifications are involved.

Comparisons between modern authors and the three main authors, who published their systems at the end of the 19th and beginning of the 20th centuries, do not statistically reveal significant differences in the content of orders (Table II). Thus, the ordinal circumscription did not change significantly over the last century.

In addition to the three past authors, Gobi's (1916) classification was compared to Takhtajan's (Table III), since the latter recognized the influence of Gobi's system as well as those of Bessey and more especially Hallier. Consensus tests show that Takhtajan's classification is more similar to Gobi's for the Monocotyledons (r = 0.67003) and to Hallier's for the Dicotyledons (r = 0.66705). Likewise, results of the Mantel test indicate that Takhtajan's content of orders resembles more that of Gobi and less those of Hallier and Engler (Monocotyledons). However, differences among these results are shallow, especially the ones stemming from the Mantel test. One must realize that 1) the dicotyledon classification of Engler differs significantly from that of Takhtajan (p [greater than] 0.0025) and 2) more importantly, Bessey's classification differs statistically from Takhtajan's (consensus test; Table III). Also, results (i.e., correlations) of the Mantel test between Bessey and Takhtajan are the weakest (Table III).

Tables IV and V give the results of the Mantel tests at the superorder and subclass ranks. These comparisons examine the familial content of both taxonomic ranks and assess their similarity among authors who use them in their classification. Results originating from the dicotyledon classifications of the modern authors have been transcribed even though they remain uncertain, this uncertainty being caused by the constraint imposed by algorithms upon the number of objects that can be treated at once. The superorder content among taxonomists is statistically similar (Table IV).

At the subclass rank, dissimilarity is found between the classifications of Bessey and of Takhtajan. Contrary to this dissimilarity and the one shown by their topology (consensus test; Table III), the content of orders is similar between these authors (Mantel test; Table III). The [TABULAR DATA FOR TABLE II OMITTED] [TABULAR DATA FOR TABLE III OMITTED] asymmetry observed at the subclass rank (Liliidae of Takhtajan vs. Alternifoliae-Cotyloideae of Bessey) then vanishes at the ordinal rank. The content of subclasses also differs between Engler and Bessey or Hallier, the difference being so high that correlations are negative (Table V). No other comparisons of the content of subclasses recorded in Table V display any significant differences; therefore, the content of subclasses is statistically similar.
Table IV

Results of the Mantel tests between classifications using
super-ordinal level. P(r) = associated probability of standardized
Mantel statistic (r). The level of significance has been adjusted
through the Bonferroni method (0.01/number of pair-wise
comparisons); it is then equal to 0.00333 (Monocots) and 0.00167
(Dicots). The number of permutations has been established at 4000
for the Monocotyledons and at 10000 for the Dicotyledons. Results
shown in italics must be carefully interpreted since matrices used
for the test were near the algorithm limits.

Matrices r P(r)

Dicotyledons (234 OTUs) 0.15225 0.00009
Dicotyledons (231 OTUs) 0.15872 0.00009
Dicotyledons (222 OTUs) 0.16315 0.00009
Monocotyledons (111 OTUs) 0.89550 0.00049
Dicotyledons (361 OTUs) 0.04308 0.00009
Monocotyledons (103 OTUs) 0.86566 0.00049
Dicotyledons (318 OTUs) 0.03460 0.00009
Monocotyledons (103 OTUs) 0.79567 0.00049
Dicotyledons (331 OTUs) 0.03596 0.00009
Table V

Results of the Mantel tests between classifications at the subclass
level. P(r) = associated probability of the standardized Mantel
statistic (r). The level of significance has been adjusted through
the Bonferroni method (0.01/number of pair-wise comparisons); it is
then equal to 0.00333 (Monocots) and 0.00010 (Dicots). The number of
permutations has been established at 4000 for the Monocotyledons and
at 10,000 for the Dicotyledons. The result shown in italics must be
carefully interpreted since matrices used for the test were near the
algorithm limits.

Matrices r P(r)

Monocotyledons (51 OTUs) 0.24186 0.00049
Dicotyledons (233 OTUs) 0.10838 0.00009
Dicotyledons (234 OTUs) -0.02089 0.15321
Dicotyledons (256 OTUs) 0.02534 0.00009
Monocotyledons (53 OTUs) -0.05682 0.09323
Dicotyledons (231 OTUs) 0.14034 0.00009
Dicotyledons (229 OTUs) 0.23302 0.00009
Dicotyledons (175 OTUs) 0.17941 0.00009
Monocotyledons (107 OTUs) 0.52233 0.00049
Dicotyledons (313 OTUs) 0.12216 0.00009
Dicotyledons (240 OTUs) -0.04124 0.00019
Dicotyledons (230 OTUs) 0.16894 0.00009
Dicotyledons (175 OTUs) 0.20300 0.00009


Results of the above statistical tests were used to produce input matrices which in turn were helpful to create a classification of classifications. Input matrices can be obtained from the primary author.

1. Monocotyledons

The dendrogram and ordination [ILLUSTRATION FOR FIGURES 3 & 4 OMITTED] obtained from the results of the consensus tests involving monocotyledon classifications have produced similar relationships among the seven classifications. The dendrogram [ILLUSTRATION FOR FIGURE 3 OMITTED] reveals two groups of classifications: past (Bessey, Engler, and Hallier) and modern (Cronquist, Dahlgren, Thorne, and Takhtajan). Engler's and Hallier's classifications form a group to which Bessey's classification is grafted. The first two axes of the ordination show a structure similar to the hierarchical clustering. Modem and past classifications are separated into two groups along axis 1 and Engler-Hallier and Dahlgren-Thorne are combined in two groups whereas Bessey and Takhtajan are isolated along axis 2. Classifications of Dahlgren and Thorne always constitute a cluster in all dendrograms; this cluster is also found in ordinations.

Results of the Mantel tests using monocotyledon classifications generate two different dendrograms depending upon the method [ILLUSTRATION FOR FIGURES 5 & 6 OMITTED]. The complete linkage maximizes the grouping of Bessey's and Hallier's classifications and minimizes the similarity between Takhtajan's and the other modern classifications [ILLUSTRATION FOR FIGURE 5 OMITTED]. Nonetheless, on these dendrograms, the same cluster appears composed of the classifications of Engler, Cronquist, Dahlgren, and Thorne. The groupings in modern and past classifications are absent; both the single and complete linkages join together the classifications of Engler and Cronquist, due to the high correlation (r = 0.61962). This cluster then joins that of Dahlgren and Thorne. Complete linkage [ILLUSTRATION FOR FIGURE 5 OMITTED] groups the classifications of Bessey and of Hallier and links the resulting group to the cluster comprising Cronquist, Engler, Dahlgren, and Thorne; whereas the single linkage [ILLUSTRATION FOR FIGURE 6 OMITTED] adds successively to the same cluster the classifications of Takhtajan, of Hallier, and finally of Bessey.

The ordination obtained from the principal component analysis [ILLUSTRATION FOR FIGURE 7 OMITTED] illustrates a cluster comprising the classifications of Engler, Cronquist, Thorne, and Dahlgren; whereas Bessey's, Hallier's, and especially Takhtajan's classifications are isolated, as in the clustering analyses. Whatever the axes used, Engler and Cronquist as well as Thorne and Dahlgren are always linked together, and Bessey, Hallier, and Takhtajan undergo slight changes of position, imitating in that regard the results of clustering analyses.

2. Dicotyledons

The dendrogram produced by single linkage using the results of the consensus tests pictures the relationships among all seven dicotyledon classifications [ILLUSTRATION FOR FIGURE 8 OMITTED]. Dahlgren and Thorne again are combined in a cluster to which is joined those of Cronquist and Takhtajan. Thus, modern classifications are clustered together as in Figures 3 and 4 (produced from the monocotyledon classifications). Thereafter, Hallier, Engler, and Bessey are successively linked to the "modern" cluster. The same clusters resulted from the ordination [ILLUSTRATION FOR FIGURE 9 OMITTED]: Modern classifications form a group, whereas Hallier's joins first, followed by Engler's and Bessey's classifications.

3. Angiosperms

Results of the consensus tests done separately on the Monocotyledons and the Dicotyledons can be combined for further clustering and ordination analyses. The single linkage dendrogram [ILLUSTRATION FOR FIGURE 10 OMITTED] shows a topology different from that of the complete linkage dendrogram [ILLUSTRATION FOR FIGURE 11 OMITTED]. These two dendrograms present, however, some similarity in the grouping of the modern classifications. Not surprisingly, this grouping was also obtained when the Monocotyledons and the Dicotyledons were analyzed separately [ILLUSTRATION FOR FIGURES 3 & 8 OMITTED]. In Figure 11 the complete linkage joins the three past classifications as in Figure 3. As for the single linkage dendrogram, the classification of Bessey is isolated whereas those of Hallier and Engler constitute a cluster which then joins the modern classification cluster [ILLUSTRATION FOR FIGURE 10 OMITTED]. The ordination originating from the same input matrix separates the modern from the past classifications [ILLUSTRATION FOR FIGURE 12 OMITTED]. Again, Bessey is isolated, whereas Hallier and Engler form one group and Thorne and Dahlgren form another. The Cronquist-Takhtajan cluster, shown on the dendrograms of Figures 10 and 11, stands out again when axes 1 and 3 (not shown) are used in the same analysis that generates Figure 12.

V. Discussion

Using the results of the consensus and Mantel tests, and of the meta-analyses, the four questions raised in the introduction can be tentatively answered.


The classificatory structure and content of orders (and of superorders or subclasses) do not differ statistically among the modern classifications. The classifications of Cronquist, Dahlgren, Takhtajan, and Thorne show more global resemblance than local differences. Indeed, Figures 3 and 4 (dealing with the monocotyledon classifications), Figure 8 (dealing with the dicotyledon classifications), and especially Figures 10, 11, and 12 (dealing with the angiosperm classifications) indicate that the four modern classifications join together in a cluster on the basis of the similarities obtained from the two types of tests used.

This similarity of classifications contradicts the authors of the systems. For instance, Takhtajan (1964:160; 1980) wrote: "In the systems of classification of the higher plants proposed during the last decades there reign extraordinary differences of opinion on the content and size of the higher taxa." Cronquist (1976: 2) echoed the notion: "The difficulties, and the great differences of opinion, come instead in how to organize the families of monocots and dicots into orders and superorders or subclasses." Thorne (1977) also emphasized the enormous difference between his system and those of the other modern authors. These opinions cannot be entirely accepted. Cronquist (1969b, 1974), however, stated that his system was similar to that of Takhtajan. Since 1957, these two taxonomists frequently consulted one another about the modifications they brought to their classifications (Cronquist, 1969b, 1974). Obviously, when an author proposes a new system, he underlines the differences and originality of his own system, whereas the same author neglects these differences and, instead, emphasizes the resemblances when he or she wants to show that all systems converge due to the advancement of knowledge in systematics (Stevens, pers. comm.). Many authors have also stipulated that modern classifications resemble one another (among others, Stevens, 1986). But does this resemblance reflect a shared taxonomic approach? Tables VI-VIII allow us to examine the numerical results in greater depth and formulate an answer to that question.

First, the same conceptual constraint plays a fundamental role in all four systems: the unique origin of the Angiosperms and of most taxa (Table VIII). This constraint is not accepted by all botanists (Meeuse, 1987, 1992; Krassilov, 1991; Hughes, 1994). Second, all authors agree in viewing one of the families of the Magnolianae as the most primitive taxa in the Angiosperms and the Nymphaeales as the probable ancestor of the Monocotyledons. These similarities vanish, however, in the choice of the probable ancestor of the Angiosperms and of the most primitive family of the Monocotyledons (Table VI). Thus, for Thorne (1992a) and Dahlgren et al. (1985), the Melanthiaceae and Trichopodaceae are the primitive families of the Monocotyledons, respectively, whereas for Cronquist (1988) and Takhtajan (1987), the Butomaceae are still viewed, in a more traditional fashion, as one of the most primitive members of the Monocotyledons. Therefore, there are both differences and resemblances among the four modern authors about ancestors and primitive taxa (Table VI). Figures 8 and 10-12 support the dichotomy Dahlgren+Thorne vs. Cronquist+Takhtajan, although, as noted before, all four classifications are very similar to one another. Thus, classifications can be similar and yet be in disagreement about the origin of some taxa. Among past authors, Hallier and Engler have produced similar monocotyledon classifications [ILLUSTRATION FOR FIGURES 3 & 4 OMITTED], even if they did propose different origins of that taxon. For example, Engler saw in the Pandanales, among others, a primitive order; Hallier held the idea that the Liliaceae were the most primitive family of the Monocotyledons, which were derived from a Pro-Berberideae very similar to the Lardizabaleae (Cuerrier, in press). Here one must keep in mind that the consensus test does not evaluate sequences of primitive and derived taxa (ancestors-descendants). This information does not constitute one of the formal properties of dendrograms; a Linnean scheme can hardly represent it. For an evolutionary interpretation of modern as well as past classifications, one must have recourse to the text or to those diagrams that explicitly illustrate the genealogy of plant taxa. For example, Takhtajan's (1987) phylogenetic diagrams (genealogical trees) cannot be extirpated from his classification. Unless one adopts a convention which is not inherent to the Linnean scheme, the phylogenetic information will not be recorded in the classification. Indeed, modern classifications insufficiently render the angiosperm phylogeny. This explains, at least in part, why modern classifications resemble one another in spite of divergent viewpoints concerning the phylogeny of the flowering plants.

Results of statistical analyses seem to underline a common taxonomic approach by the four authors. Table VIII shows the conceptual concordance among systems. All authors adhere to the classical primitive flower concept from which they formulate their evolutionary tendencies. They believe in the same tendencies (Table VIII), notably those of Bessey (dicta). Cronquist (1957, 1983), for instance, recognized the paternity of those principles and evolutionary tendencies published by Bessey (1915) and consolidated by Thorne (1958, 1963). Some of these general tendencies, however, are not accepted by Dahlgren et al. (1985). Barabe (1984, 1993: tab. 2) has shown how few changes there were among the tendencies used by Bessey, Engler, Hutchinson, and Takhtajan.

Also, in rank determination, a similar approach can be observed (Table VIII) where authors take into consideration the taxonomic experience, the need for clarity (didactic value), and the size of discontinuities (i.e., gaps). For them, ranks have a practical dimension, even though they refer to the size of discontinuities. Therefore, the rank determination is based on conflicting criteria: on one hand, a need to establish conventions, and on the other [TABULAR DATA FOR TABLE VI OMITTED] [TABULAR DATA FOR TABLE VII OMITTED] hand, a need to represent phylogenetic data. Stevens (1996) mentioned that such a problem also arose in the work of Bentham and Hooker. But, then, on which principle do authors establish the size of discontinuities? Such a principle still needs to be found or made explicit.

This subjectivity in regard to the determination of taxonomic ranks has produced classifications that differ in the number of ranks used (Table VII). Similar attitudes, then, do not guarantee similar results, and here, more than anywhere else, the idiosyncrasy of the four modern authors is particularly strong. Cronquist (1988), Dahlgren et al. (1985), and G. Dahlgren (1989a) use two ranks higher than that of family, but one of them differs for these two authors [TABULAR DATA FOR TABLE VIII OMITTED] without, however, affecting the global similarity of their classifications. Thorne (1992b) and Takhtajan (1987) had recourse to three and four ranks, respectively. Again, this difference is less important than taxon circumscription itself, since the consensus tests have shown the global similarity among the four classifications. Thus, as it will become clear below, the modification of the rank of a taxon, whose circumscription remains similar, is but a minor fact.

The taxonomic approach of the modern authors also includes the use of intermediates for establishing the affinity of taxa. Equally important is the value of new characters that authors judge in regard to their own classifications. The similarity in these notions and in the ones discussed above explains, at least in part, the resemblance among modern classifications [ILLUSTRATION FOR FIGURES 10-12 OMITTED].

These authors diverge, however, in their choice of characters; this is particularly true of Dahlgren, who emphasizes the use of phytochemical and embryological characters. In Table VIII, the other notions showing dissimilarity among authors either have no consequence for classifications or are simply deceptive. For example, the concept of heterochrony, used mostly by Takhtajan (1943, 1954, 1972, 1976, 1983), applies more at the level of explanation than at taxon delimitations and the classificatory approach per se.

The similarity among modern classifications is imputable not only to common principles and to a reciprocal influence but also, and more especially, to the use of past classifications as a starting point. Here, statistical tests performed between modern and past classifications were of particular interest.


Modern classifications present a high resemblance to past classifications, to which they were compared (Table II). One must keep in mind, however, that classifications published at the end of the 19th and beginning of the 20th centuries (e.g., van Tieghem [1918] and Camel [1889]) did differ from those that we have chosen to compare (i.e., Bessey, Engler, Hallier). These classifications seem to have had a minor influence, at the most, on modern ones. These classification schemes never gained any importance among botanists as in the case of Bessey, Hallier, and especially Engler, and modern authors do not refer to them (except in the case of Takhtajan, who mentioned the influence of van Tieghem on his system [D. Stevenson, pers. comm.]). A sociohistorical study would be needed to understand why some classifications became popular on the international scene while others did not.

Even though the number of taxa differs among the different categories used by past (Bessey, Engler, Hallier) and modern authors (Cronquist, Dahlgren, Takhtajan, Thorne), the global similarity of classifications was not significantly affected. Results of the Mantel tests indicate that taxa have similar delimitations. Table VII clearly shows the increase in the number of families among modern authors in comparison to past ones. This modification is a minor one since newly recognized families by modern authors were taxa relegated to a lower rank by past taxonomists. During the 19th and 20th centuries, many taxa had their rank changed (especially from subfamily or genus to family) without changes in their circumscription (Watson, 1964; see also Stevens, 1986, 1994; Thorne, 1992b: 244; Cuerrier et al., 1996). Watson (1964: 280) stated that"too many botanists seem to accept traditional systems uncritically and, unwittingly, to base their conclusions on nineteenth-century taxonomic philosophy." Watson's explanation is plausible, as the similarity among modern and past classifications tends to show (Table II).

Cronquist (1979) drew attention to the correspondence between the pre-Darwinian concept of natural classification and that of evolutionary classification. For him, an evolutionary approach to classification is compatible with that which prevailed before Darwin. This compatibility or correspondence appears to be connected to the concept of classification itself, a concept that does not allow an easy importation of phylogenetic data. Table VIII clearly shows among modern authors the practical and didactic goal of classifications. This constraint minimizes the impact of the Darwinian theory on the structure of classifications. For instance, this constraint can already be seen in the work of Bessey, Engler, and Baillon (Cuerrier et al., 1996); while Stevens (1984b) discussed it at some length. Therefore, a classification is concerned with two main features: its stability, which serves, among other criteria, as an argument between taxonomists to value their classifications (Barabe & Brouillet, 1982); and its memory device, which helps the human mind to grasp more easily the large amount of information secured in a classification.

In spite of the great similarity between modern and past classifications, some clustering analyses have underlined a temporal separation between these two groups of classifications [ILLUSTRATION FOR FIGURES 3, 4, 11 & 12 OMITTED]. This separation could be the consequence of the reciprocal influence among modern authors and of recent developments in different botanical disciplines, leading to an increasing number of available characters. In other respects, the clustering analysis, using the results of the Mantel tests on the monocotyledon classifications, shows that Engler shares more relationships with the modern authors than with the past ones [ILLUSTRATION FOR FIGURES 5-7 OMITTED]. This result will allow us to answer the third question raised in the introduction.


Many taxonomists insist that Bessey's classification (1915) is the major influence on the modern systems [ILLUSTRATION FOR FIGURE 1 OMITTED]. Analyses conducted in this study indicate that modern classifications are, on the contrary, less similar to that of Bessey than to that of Engler or Hallier, in the structure as well as in the content of orders. It should be noted that the classifications of Cronquist and Bessey differ, even though the first author firmly recognized the influence of the latter: "We are all - or nearly all - Besseyans" (Cronquist, 1988: 162). He also admits that "large blocks, and groups of blocks, of the Engler system remain in all present systems, merely rearranged with respect to each other" (Cronquist, 1988: 162). This statement appears to be true: Orders circumscribed by modern authors resemble less those of Bessey and Hallier than those of Engler (Table II; [ILLUSTRATION FOR FIGURES 5-7 OMITTED]), notwithstanding the supposed absurdities of the Engler system emphasized by Thorne (1973). Metataxonomic analyses show that Bessey's classification is either isolated or linked to the other past classifications; it is never directly connected to the modern classifications [ILLUSTRATION FOR FIGURES 3-12 OMITTED]. It seems that Bessey's position as an outlier can be explained by his use of an analytic or dichotomic approach based on few characters. Cuerrier et al. (1992) underlined this difference in approach between Bessey and Engler, resulting in fewer taxa at the level of the most inclusive ranks in Bessey's classification. Indeed, in Table VII one can see that Bessey uses two taxa at the subclass rank, whereas Cronquist uses five or six. At the rank of superorder, Bessey's classification has five taxa whereas Thorne's, Dahlgren's, and Takhtajan's classifications have 19, 25, and 39 taxa, respectively. The large divisions in Bessey do not statistically differ, however, from those of the modern authors (Tables IV and V).

According to Takhtajan (1980: 235), his own system derives from those of Hallier and Gobi. Nonetheless, the monocotyledon classifications of Takhtajan and Hallier are less similar to one another than that of Hallier is to the other modern authors (Table III). The difference in the number of ranks between Takhtajan (4) and Hallier (1) contributes to diminish the similarity between their monocotyledon classifications. Moreover, Hallier (1912) starts his monocotyledon classification with the Liliaceae, which originate from a plant similar to Berberis, whereas Takhtajan (1987), in accordance with Bessey, begins his with the Butomaceae, which he then derives from the Nymphaeales.

Results of the Mantel tests have also shown that the classifications of Takhtajan and Bessey differ at the level of subclasses. This difference appears to be due to the asymmetry caused by Takhtajan's subclass Liliidae, since this subclass contains the majority of families whereas the three other subclasses are fewer in taxa. The classification of Bessey shows more symmetry (Cuerrier et al., 1992). One should also note that Takhtajan used four subclasses and Bessey only two (Table VII). These results again indicate that the classification of Bessey is unrelated to the modern ones, like that of Takhtajan.


The genealogy of the systems shown in Figure 1 has not been corroborated by metataxonomic analyses. As for the modern systems, Figure 1 reflects less the classification than the authors' opinions about ancestors and primitive taxa. In fact, Engler has influenced recent classifications more than Bessey did [ILLUSTRATION FOR FIGURES 5-7 OMITTED]. As for the primitive monocotyledon taxa, the ideas held by Dahlgren and Thorne refer to those of Hallier, not to those of Bessey or Engler (Table VI). Classifications of Dahlgren and Thorne show high similarities both at the level of the topology and the content of superior categories [ILLUSTRATION FOR FIGURES 3-12 OMITTED]. Indeed, Thorne mentions the influence that Dahlgren's studies on the Monocotyledons had on his own classification. These two authors collaborated on an article dealing with the Myrtales, but, nevertheless, differences of opinion were clearly stated in the article (Dahlgren & Thorne, 1984). Also, Thorne and Dahlgren used dahlgrenograms to represent taxon relationships, whereas Cronquist and Takhtajan chose minimum spanning trees, which allow direct ancestor-descendant links to be expressed (Table VIII). Therefore, a bird's-eye view (or a transversal section) for the first two authors and a side view (or a longitudinal section) for the latter two authors can be observed. This difference might be considered shallow if the "bubbles" of the dahlgrenogram are linked with one another. One should consider, however, dahlgrenograms as a bird's-eye view of Steiner trees (i.e., with no direct ancestor-descendant links). Throughout Takhtajan's numerous papers, trees have remained versions, more or less divergent, of a first genealogical tree published in 1942 (104, [ILLUSTRATION FOR FIGURE 2 OMITTED]).

Some results [ILLUSTRATION FOR FIGURES 8-11 OMITTED] have grouped together the dicotyledon or angiosperm classifications of Cronquist and Takhtajan. This grouping probably reflects the Russian's influence upon the American botanist: Indeed, Cronquist (1969a) embraced the subclasses of Takhtajan's Magnoliatae. Moreover, Cronquist knew the Russian language and could read the works of Russian botanists.

Thus, a genealogical hierarchical tree representing the filiation of the angiosperm systems of classification cannot be accepted as the ideal model: A network seems to be a more comprehensive model [ILLUSTRATION FOR FIGURE 13 OMITTED]. Indeed, the analyses have shown that all four modern systems are similar not only to one another but also to that of Bessey and, more especially, to those of Engler, Gobi, and Hallier. Moreover, this network of reciprocal influences follows a similar taxonomic approach by the authors; indeed, Table VIII indicates the concordance between notions or criteria used by the four modern authors. The fundamental principles at the basis of modern systems are those of Bessey. Nevertheless, they resemble those of Hallier (1912) and Engler (1909). Engler's influence is seen in ordinal delimitations (Table II). This influence can hardly be distinguished from those of Bessey and Hallier, since their contents of orders also resemble those of the modern authors (Cuerrier et al., 1992). In fact, the contents of orders are similar in all seven classifications. But it is more than reasonable to assume that the importance of Engler's Pflanzenfamilien for floras and herbaria may well have increased the influence of Engler's classification on modern taxonomists. It then follows that the Englerian system has become more familiar. Similarities between past and modern authors are multiple: the type of tree used to represent evolution, the rank determination, the assessment of the value of characters and types of characters used, the purposes of classification, the use of intermediates, the constraint of angiosperm monophyly (except for Engler), and the use of a logical language. Stevens (1986) also emphasizes the similarity of many systematic principles between evolutionary and pre-Darwinian classifications.

We wrote that principles of the modern taxonomists were those of Bessey. But, then, on what evidence do modern authors accept those Besseyan principles, when we know that Bessey built his classification first and afterwards proposed his dicta (Cuerrier et al., 1996)? And on which criteria should we establish their use at the different levels of the Linnean hierarchy? One must keep in mind that these are very general principles and are meant to be used at the angiosperm level. Therefore, principles should be handled carefully in establishing the affinities within taxa. Leroy (1993) also mentions that Takhtajan used evolutionary tendencies that were too general when he established the affinity of some orders or families.

We can resume the taxonomic approach of the four modern systematists while quoting Cronquist (1988: 2): "We try to group together the things that are most alike in all respects, and to separate them progressively from things they are progressively less like." In Buffon (1749), Fries (in Lindley, 1826), and de Candolle (1859), from the 18th century to the 20th, taxonomists had the same preoccupation: to group what is similar and to separate what is not. Stevens (1986: 325) wrote that the "correspondence between evolutionary and pre-Darwinian systematic practice is extensive." The difficulties encountered by Bessey at the beginning of the 20th century in establishing the homologies of structures and the affinities of taxa still remain two daunting problems even today (Stevens, 1994; Cuerrier et al., 1996).

Classifications have changed little over centuries, even though new data have been discovered and used. This is in part because recent systems are still based on morphological data stemming mostly from the flower, and in part because they were influenced by past classifications. But other complementary explanations further illuminate the resemblance between modern and past classifications. We will discuss below three of these important explanations.


1. Typology and Conceptual Constraint

The four modern taxonomists claim explicitly or implicitly to posit their work in the classical concept of the plant and more especially of the flower. Therefore, data are always constrained by a theory that has not changed since Goethe and A.-P. de Candolle. In spite of the works published by Lam, Melville, Meeuse, and Croizat, the classical theory is still accepted without question. And yet, the ideas developed by Croizat (1964) - on strobili, flowers, and inflorescences in general; on the distinction between carpel and ovary; and on the definition of morphological structures in particular - illustrate that we should not take for granted the foundation of classification systems on a theory of morphology. Similarly, Meeuse (1987, 1992) warns us about such notions as monophyly and flower, while Sattler (1991, 1992) demonstrates the dynamic and continuum notion of the plant. These works, as Leroy (1993) noted, generate numerous ideas and fruitful reasoning. These studies force us to recognize, at least, the elements of constraint often hidden in postulates tacitly acknowledged to be at the very basis of classification systems. Meeuse (1987: 43) wrote: "The published systems of classification of the Flowering Plants purported to be phylogenetic by their originators are purely typological, even if only for the reason that they are exclusively based on certain features of extant plants which were molded into a preconceived pattern of concepts and interpretations." Indeed, taxonomists of the modern systems try to unravel the angiosperm origin through a comparison of extant plants (e.g., see Cronquist, 1969c; Takhtajan, 1958, 1980, 1991). The resulting classification is more horizontal than vertical. The latter would integrate paleobotanical data and the time aspect. If, as understood by modern systematists, evolution is mostly a vertical process, then a methodology that would lead to a horizontal classification improperly represents the phylogeny. Indeed, Linnean classifications cannot support and illustrate verticality. They hardly formulate the affinity of taxa, let alone their origin.

2. Parallelism and Convergence

The importance of new data is often muted by using the concepts of parallelism and convergence. They are frequently used to devalue the importance of a character, without recurring to phylogenetic analysis. Thus, these concepts remain superficial. How can one discuss convergence without an organogenetic and ecological study within a phylogenetic frame? For example, Cronquist (1980) and Thorne (1981) disagree with Dahlgren's groupings based on phytochemical data. For them, iridoids would have evolved many times during the angiosperm history; therefore a case of parallelism without any analysis. Contrary to Thorne, Dahlgren (1983) reduced to nothing the value of stamen centrifugal development in some Caryophyllales, since it is merely a case of convergence. Even though Dahlgren's statement seems correct a priori, a phylogenetic study using flower ontogenetic data should have been done first. For many cases of supposed convergence, Thorne (1958) used his own personal experience in determining the value of characters. Unless this personal experience is based on a character analysis, such a taxonomic judgment constitutes an act of authority, not an accurate research of taxa affinities.

3. Cognitive Constraint and Historical Inertia

Systematists sometimes use their own classifications to determine the value of new characters. For Cronquist (1980: 4), "our general scheme is now good enough to warrant the greatest of caution in accepting radical changes." This assertion suffices for Cronquist and for Takhtajan (1980) to neglect data that do not agree with their classification schemes. Such an approach imparts a static element to classifications. Results of the tests we performed in the present study show that classifications have changed little over a century; therefore, they underline the static element of classifications whether or not such statism is imputable to the approach mentioned above. Stevens (1984b, 1986) pointed out that taxonomists try to confirm the existing taxonomic structure rather than to evaluate the one suggested by their data. Taxa formed over two centuries ago encapsulate a historical inertia linked to the concept of stability. The usefulness of classifications, and the pressure implicitly exerted to make them remain so, restricts the importance of taxonomic changes that may come from new data. After centuries of utilization, names of taxa are burdened with a historical weight which renders all modifications difficult to apply (Stevens, 1986, 1994). This inertia also explains the similarity of past and modern classifications. For Stevens (1986: 329), "taxa initially recognized seem immune from criticism and provide a pattern to which other taxa must conform." The worldwide use of the Englerian classification might have influenced, in an unconscious manner, taxon delimitations up to now.


Due to the constraints discussed above, it will not suffice in taxonomy to only use new data, as is often the case in morphological and molecular studies, to create the necessary opening for a complete revision of classifications. One should reexamine the meaning of "classification," and try to avoid evaluating data with existing classifications as a standard. The choice of taxa entering the input matrix for cladistic analyses should be based on more than one classification. For example, the analyses of Dicotyledons and Monocotyledons by Chase et al. (1993) and Qiu et al. (1993) rely on Cronquist (1981) and Dahlgren et al. (1985), respectively. For Donoghue and Sanderson (1992:359), "it is also ironic that even those who are most wary of morphological data nevertheless lean on it heavily in designing their own research, namely in choosing which groups to work on, which subgroups to sample, and so on." The use of traditional classifications in cladistic analyses remains a type of constraint, based on taxon definitions and on a scheme of homology that is hardly explicit (see also Meeuse, 1982). Since the four modern classifications are similar in their structure and the content of their taxa, as well as in their foundation and their taxonomic approach, it would be preferable to take into account opinions and phylogenetic hypotheses formulated by unorthodox authors (Hughes, Krassilov, Lam, Meeuse, Croizat). Molecular systematics offers new data; the latter are welcome, but other types of data (morphological, serological, phytogeographical, etc.) should not be excluded. Donoghue and Sanderson (1992; 342) rightly wrote: "Our argument is not against the use of molecular data; rather, it is against ignoring relevant morphological evidence."

Perhaps the importance of molecular studies in comparison to the morphological ones can be more readily appreciated at the level of preconceived ideas that Meeuse (1987) has so often denounced. For him, the angiosperm phylogeny is based "on the repetition of ingrained tenets or the use of the phrase: established ideas" (Meeuse, 1987: x). Molecular phylogenies have a chance to bypass a priori ideas from which morphological structures suffer. They can be used to examine the evolution of morphological structures (see Soltis & Soltis, 1995).

Molecular studies have already shown that the series of cladograms obtained from rbcL sequences not only resemble the classifications of Thorne (1992b) and Dahlgren et al. (1985) but, moreover, differ from the classification of Cronquist (1988). One should keep in mind, however, that Thorne's (1992a, 1992b) work follows the development of molecular systematics, which might explain its similarity with the study of Chase et al. (1993). Also, it seems that the divergence between the cladogram based on molecular data and Cronquist's classification is caused in part by the simplicity and the clarity that spring from the latter classification. Indeed, Cronquist's classification responds to a need for convention (didactic); it is thus a logical tool. Therefore, it is hardly surprising that it differs from a cladogram based on different conventions. It follows that Cronquist's classification and Chase et al.'s cladogram are two different objects. For example, the Ranunculales and Magnoliales of Qiu et al. (1993) are not conceptually those of the modern taxonomists. A quantitative comparison between modern classifications and a classification taken from the cladogram of Chase et al. (1993) would help to establish whether the resemblances mentioned above are statistically significant and, therefore, whether Thorne's classification is really similar to the cladogram (and that Cronquist's classification is really dissimilar).

VI. Conclusion

The results of the statistical tests used in this study are in agreement with the opinion stating that the modern classifications do not differ from those published at the end of the 19th and beginning of the 20th centuries. In spite of what authors have written, not only do modern classifications resemble one another, but also their content of orders has remained broadly that of the past taxonomists (Bessey, Engler, Gobi, Hallier). Modern classification systems of the flowering plants are Besseyan only in their adoption of an angiosperm ancestor near the Bennettitales; they are not Besseyan because their classifications have followed the classificatory scheme and taxon circumscription of Bessey. Thus, the genealogies of systems drawn by Lu (1981) and Woodland (1991) are of limited value only. They can be viewed as a complementary viewpoint to ours, which renders the affinity among classifications as a network [ILLUSTRATION FOR FIGURE 13 OMITTED]. There is complementarity in the sense that the network has been based on comparisons involving the classificatory structure and/or the content of orders, whereas Lu's and Woodland's genealogical trees have been founded on the origin of the Angiosperms expressed by taxonomists.

This study reveals a historical inertia (classification relying on morphological structures) as well as the idiosyncrasies of all modern authors of classification systems. This means that modern taxonomists started their classification from existing classifications, modifying it as works and studies by morphologists, anatomists, organogeneticists, phytochemists, and others became available. At times, their own classification is used as a reference for establishing the value of "new" characters. It hardly comes as a surprise, then, that results of statistical analyses do not show an important amount of difference among the classifications compared, since circularity is therein inevitable. It is difficult to separate what creates the "stable" in the stability of classifications: either taxa described a century ago are still viewed as good taxa in general (they were already well-circumscribed) or taxa are maintained because the conceptual approach to taxonomy has remained identical over years. One explanation may be better than the other or both may have explanatory powers, yet they both imply that past classifications have had an undeniable influence on those of today.

The use of intermediates and of evolutionary tendencies and the importance of constructing taxon sequences by modern taxonomists are reminiscent of the approach followed more than a century ago by past taxonomists. The idea of tendency in itself is teleological (Bernier, 1984); it refers back to ideas like progression and progress, orthogenesis, creative force (i.e., Bildungstrieb). Many of these notions have more than a century of existence. Also, in the writings of modern taxonomists, one can still find a whiff of scala naturae. Indeed, notions like progress and perfection are to be found in the works of Cronquist (1969a), Takhtajan (1958, 1959, 1973), and Thorne (1992a, 1992b). For instance, these authors apply the metaphor of evolutionary scala when discussing the phylogenetic part of their work. In short, the taxonomic language has changed very little since the 19th century (Stevens, 1994).

Recent advances in molecular biology have proven important for taxonomic works, but a classification system is more than just a heap of data - it is also principles, methods, and a theory. The revision of the modern classifications which should spring from molecular works (as well as from any taxonomic studies), must run the risk of questioning the actual foundation of classifications. Also, molecular studies must keep clear of the constraints imposed by recent systems. The point of view still in use remains that of the classical or traditional taxonomy from Goethe, A.-L. de Jussieu, and A.-P. de Candolle to Cronquist, Dahlgren, Takhtajan, and Thorne. To paraphrase a Bachelardian idea (Bachelard, 1972): One could write that, in taxonomy, prudence has became an epistemological obstacle. With molecular data, taxonomy is on the threshold of renewal, because these data are independent of ideas and notions molded by a long history which has burdened the terminology used in morphology (notions like what is a flower, what is primitive, etc.). Therefore, results of molecular analyses could help to reinterpret the evolution of morphological characters and to reorient fruitful organogenetic studies. To rethink plant classifications from their basis will obviate the epistemological obstacle laid down by history more than a century ago.

VII. Acknowledgments

Numerous articles needed for the completion of this manuscript were made available by S. Madrinan (Universidad de Los Andes) and P. F. Stevens (Harvard University). Dr. Stevens also read and commented on an earlier version of this manuscript. The English was kindly revised by J. St. John Winter. We acknowledge V. Perreault of the Inter-Library Loan Service (Universite de Montreal) for his ability to find so many articles from obscure periodicals. We thank F.-J. Lapointe for contributing his knowledge of tree comparisons, and J. Podani (L. Eotvos University [Budapest]) for kindly sending the MATTEST and MTCOM programs of his software SYN-TAX V; without his help, results would not have been so complete. Members of the IRBV Ecological Laboratory and H. Veronneau were kind enough to lend me their computers. J. Brisson has developed and made available his application permitting matrix conversions needed for all the different analyses. This paper has been written with the support of a FCAR grant (to AC) and a FCAR group grant (to LB and DB).

VIII. Literature Cited

Abbott, L. A., F. A. Bisby & D. J. Rogers. 1985. Taxonomic analysis in biology. Columbia University Press, New York.

Bachelard, G. 1972. L'engagement rationaliste. PUF, Paris.

Barabe, D. 1984. Les principes directeurs des systemes modernes de classification des angiospermes. Naturaliste Canad. 111: 21-30.

-----. 1993. Les systemes modernes de classification des angiospermes. Pp. 135-159 in O. Poncy (ed.), Systematique botanique: problemes actuels. Biosystema 10. Societe francaise de systematique, Paris.

----- & L. Brouillet. 1982. Commentaires sur le systeme de classification des Angiospermes de Takhtajan. Acta Biotheoretica 3: 127-141.

----- & J. Vieth. 1990. Les principes de systematique chez Engler. Taxon 39: 394-408.

Benson, L. 1957. Plant classification. D. C. Heath, Boston.

Bernier, R. 1984. The species as an individual: facing essentialism. Syst. Zool. 33: 460-469.

Bessey, C. E. 1915. The phylogenetic taxonomy of flowering plants. Ann. Missouri Bot. Gard. 2: 109-164.

Buffon, G. L. L., Comte de. 1749. Histoire naturelle, generale et particuliere, avec la description du Cabinet du Roy. Discours premier. Vol. 1. Imprimerie Royale, Paris.

Cain, A. J. 1994. Rank and sequence in Caspar Bauhin's Pinax. Bet. J. Linn. Sec. 114: 311-356.

Caruel, T. 1889. Conspectus familiarum phanerogamarum. Nuovo Giorn. Bot. It. 21: 132-137.

Chase, M. W., D. E. Soltis, R. G. Olmstead, D. Morgan, D. H. Les, B. D. Mishler, M. R. Duvall, R. A. Price, H. G. Hills, Y.-L. Qiu, K. A. Kron, J. H. Rettig, E. Conti, J. D. Palmer, J. R. Manhart, K. J. Sytsma, H. J. Michaels, W. J. Kress, K. G. Karol, W. D. Clark, M. Hedren, B. S. Gaut, R. K. Jansen, K.-J. Kim, C. F. Wimpee, J. F. Smith, G. R. Furnier, S. H. Strauss, Q.-Y. Xiang, G. M. Plunkett, P. S. Soltis, S. M. Swensen, S. E. Williams, P. A. Gadek, C. J. Quinn, L. E. Eguiarte, E. Golenberg, G. H. Learn Jr., S. W. Graham, S. C. H. Barrett, S. Dayanandan & V. A. Albert. 1993. Phylogenetics of seed plants: an analysis of nucleotide sequences from the plastid gene rbcL. Ann. Missouri Bet. Gard. 80: 528-580.

Core, E. E. 1955. Plant taxonomy. Prentice-Hall, Englewood Cliffs, NJ.

Croizat, L. 1964. Thoughts on high systematics, phylogeny and floral morphogeny, with a note on the origin of the Angiospermae. Candollea 19: 17-96.

Cronk, Q. C. B. 1989. Measurement of biological and historical influences on plant classification. Taxon 3: 357-370.

-----. 1990. The name of the pea: a quantitative history of legume classification. New Phytol. 116: 163-175.

Cronquist, A. 1957. Outline era new system of families and orders of Dicotyledons. Bull. Jard. Bet. Etat (Bruxelle) 27: 13-40.

-----. 1965. The status of the general system of classification of flowering plants. Ann. Missouri Bot. Gard. 52: 281-303.

-----. 1969a. Broad features of the system of Angiosperms. Taxon 18: 188-193.

-----. 1969b. The general system of classification of flowering plants. Ward's Bull. 8: 1-2, 6-7.

-----. 1969c. On the relationship between taxonomy and evolution. Taxon 18: 177-187.

-----. 1974. Current opinion on the evolutionary history and classification of flowering plants. Pp. 128-129 in Memoria del II Congresso Venezolano de Botanica. Merida.

-----. 1976. The taxonomic significance of the structure of plant proteins: a classical taxonomist's view. Brittonia 28: 1-27.

-----. 1979. Taxonomy. J. Coll. Sci. Teaching 9: 76-79.

-----. 1980. Chemistry in plant taxonomy: an assessment of where we stand. Pp. 1-27 in E A. Bisby et al. (eds.), Chemosystematics: principles and practice. Academic Press, London.

-----. 1981. An integrated system of classification of flowering plants. Columbia University Press, New York.

-----. 1983. Some realignments in the Dicotyledons. Nordic J. Bot. 3: 75-83.

-----. 1987. A botanical critique of cladism. Bot. Rev. (Lancaster) 53: 1-52.

-----. 1988. The evolution and classification of flowering plants. Ed. 2. The New York Botanical Garden, Bronx.

Cuerrier, A. In press. Hallier's system of plant classification: a numerical and historical analyses. Bot. Jahrb. 121. (Forthcoming, 1999.)

-----, L. Brouillet & D. Barabe. 1990. Numerical taxonomic study of the Begoniaceae using the Mantel test on leaf microcharacters. Taxon 39: 549-560.

-----, D. Barabe & L. Brouillet. 1992. Bessey and Engler: a numerical analysis of their classification of the flowering plants. Taxon 41: 667-684.

-----, R. W. Kiger & P. F. Stevens. 1996. Charles Bessey, evolution, classification and the New Botany. Huntin 9:179-213.

Dahlgren, G. 1989a. The last dahlgrenogram: system of classification of the Dicotyledons. Pp. 249-260 in K. Tan et al. (eds.), Plant taxonomy, phytogeography and related subjects. Edinburgh University Press, Edinburgh.

-----. 1989b. An updated angiosperm classification. Bot. J. Linn. Soc. 100: 197-203.

Dahlgren, R. 1980. A revised system of classification of the Angiosperms. Bot. J. Linn. Soc. 80:91-124.

-----. 1983. General aspects of angiosperm evolution and macrosystematics. Nordic J. Bot. 3:119-149.

-----, H. T. Clifford & P. F. Yeo. 1985. The families of the Monocotyledons: structure, evolution, and taxonomy. Springer-Verlag, Berlin.

----- & R. F. Thorne. 1984. The order Myrtales: circumscription, variation, and relationships. Ann. Missouri Bot. Gard. 71: 633-699.

De Candolle, A. 1859. Memoire sur la famille des Begoniacees. Ann. Sci. Nat., ser. 4, 11: 93-149.

Donoghue, M. J. & M. J. Sanderson. 1992. The suitability of molecular and morphological evidence in reconstructing plant phylogeny. Pp. 340-368 in P. S. Soltis et al. (eds.), Molecular systematics of plants. Chapman and Hill, New York.

Endress, P. K. 1993. Federico Delpino and early views on angiosperm origin and macroevolution. Diss. Bot. 196: 77-83.

Engler, A. 1897. Principien der systematischen Anordnung, insbesondere der Angiospermen: Nachtrage zu Teil II-IV. Pp. 5-14 in A. Engler & K. Prantl (eds.), Die naturlichen Pflanzenfamilien. Angiospermae. W. Engelmann, Leipzig.

-----. 1909. Syllabus der Pflanzenfamilien. Ed. 6. Gebruder Bomtraeger, Berlin.

-----. 1926. Die naturlichen Pflanzenfamilien. Angiospermae. Band 14a. W. Engelmann, Leipzig. Ghiselin, M. T. 1984. Narrow approaches to phylogeny: a review of nine books of cladism. Oxford Surv. Evol. Biol. 1: 209-222.

Gobi, C. 1916. A review of the system of plants. Petrograd.

Hallier, H. 1912. L'origine et le systeme phyletique des Angiospermes exposes a l'aide de leur arbre genealogique. Arch. Neerl. Sci. Exact Nat. Ser. 3, 1: 146-234.

Holman, E. W. 1985. Evolutionary and psychological effects in pre-evolutionary classifications. J. Classific. 2: 29-39.

-----. 1992. Statistical properties of large published classifications. J. Classific. 9: 187-210.

Hughes, N. F. 1994. The enigma of angiosperm origins. Cambridge University Press, Cambridge.

Kiger, R. W. 1971. Context and bases of the Besseyan dicta. M.A. thesis, University of Maryland, College Park.

Krassilov, V. A. 1991. The origin of Angiosperms: new and old problems. TREE 6:215-220.

Lapointe, F.-J. & P. Legendre. 1990. A statistical framework to test the consensus of two nested classifications. Syst. Zool. 39: 1-13.

Leroy, J.-F. 1993. Origine et evolution des Plantes a feurs: les Nympheas et le genie de la nature. Masson, Paris.

Lindley, J. 1826. Some account of the spherical and numerical system of nature of M. Elias Fries. Philos. Mag. J. 68: 81-91.

Lu, A.-M. 1981. A preliminary review of the modern classification systems of the flowering plants. Acta Phytotax. 8: 279-290.

Mabberley, D. J. 1985. Jupiter botanicus: Robert Brown of the British Museum. Cramer, Braunschweig.

Mantel, N. 1967. The detection of disease clustering and a generalized regression approach. Cancer Res. 27: 209-220.

Meeuse, A. D. J. 1982. Cladistics, wood anatomy and angiosperm phylogeny - a challenge. Acta Bot. Neerl. 31: 345-354.

-----. 1987. All about Angiosperms. Eburon, Delft.

-----. 1992. Angiosperm evolution: no abominable mystery. Eburon, Delft.

Podani, J. 1990. SYN-TAX IV: computer programs for data analysis in ecology and systematics on IBM-PC and Macintosh computers. ICEMST, Trieste.

Qiu, Y.-L., M. W. Chase, D. H. Les & C. R. Parks. 1993. Molecular phylogenetics of the Magnoliidae: cladistic analyses of nucleotide sequences of the plastid gene rbcL. Ann. Missouri Bot. Gard. 80: 587-606.

Rieppel, O. 1991. Things, taxa and relationships. Cladistics 7: 93-100.

Sattler, R. 1991. Plant morphology as applied to systematics. Pp. 433-437 in D. W. Woodland, Contemporary plant systematics. Prentice Hall, Englewood Cliffs, NJ.

-----. 1992. Process morphology: structural dynamics in development and evolution. Canad. J. Bot. 70: 708-714.

Sivarajah, V. V. 1991. Introduction to the principles of plant taxonomy. Ed. 2, N. K. B. Robson (ed.). Cambridge University Press, Cambridge.

Sneath, P. H. A. & R. R. Sokal. 1973. Numerical taxonomy: the principles and practice of numerical classification. W. H. Freeman, San Francisco.

Soltis, P. S. & D. E. Soltis. 1995. Plant molecular systematics: inferences of phylogeny and evolutionary processes. Evol. Biol. 28: 139-194.

Stace, C. A. 1989. Plant taxonomy and biosystematics. Edward Arnold, London.

Stevens, P. F. 1984a. Hauy and A.-P. Candolle: crystallography, botanical systematics, and comparative morphology, 1780-1840. J. Hist. Biol. 17: 49-82.

-----. 1984b. Metaphors and typology in the development of botanical systematics, 1690-1960, or the art of putting new wine in old bottles. Taxon 33: 169-211.

-----. 1986. Evolutionary classification in botany, 1960-1985. J. Arnold Arbor. 67: 313-339.

-----. 1991. George Bentham and the Kew rule. Pp. 157-168 in D. L. Hawksworth (ed.), Improving the stability of names: needs and options. Regnum Vegetabile 123. Koeltz Scientific, Konigstein.

-----. 1994. The development of biological systematics: Antoine-Laurent de Jussieu, nature, and the natural system. Columbia University Press, New York.

-----. 1996. How to interpret botanical classifications - suggestions from history. BioScience 47: 243-250.

----- & S. P. Cullen. 1990. Linnaeus, the cortex-medulla theory, and the key to his understanding of plant form and natural relationships. J. Arnold Arbor. 71: 179-220.

Takhtajan, A. 1942. The structural types of gynoecium and the placentation. Bull. Armenian Branch Acad. Sci. U.S.S.R. 3-4(17-18): 91-112.

-----. 1943. Correlations of ontogenesis and phylogenesis in higher plants. Trans. Erevan State Univ. 22: 71-176.

-----. 1954. Quelques problemes de la morphologic evolutive des Angiospermes. Essais Bot. 2: 779-793.

-----. 1958. Origin of angiospermous plants. Transl. by O. H. Gankin. Ed. by G. L. Stebbins. American Institute for the Biological Sciences, Washington, DC.

-----. 1959. Essays on the evolutionary morphology of plants. Transl. by O. H. Gankin. Ed. by G. L. Stebbins. American Institute for the Biological Sciences, Washington, DC.

-----. 1964. The taxa of the higher plants above the rank of order. Taxon 13: 160-164.

-----. 1966. Sistema i filogenija cvetkovyck rastenij (Systema et phylogenia Magnoliophytorum). Nauka, Moscow.

-----. 1972. Patterns of ontogenetic alternations in the evolution of higher plants. Phytomorphology 22: 164-171.

-----. 1973. Evolution und Ausbreitung der Blutenpflanzen. Gustav Fischer, Jena.

-----. 1976. Neoteny and the origin of flowering plants. Pp. 207-219 in C. B. Beck (ed.), Origin and early evolution of angiosperms. Columbia University Press, New York.

-----. 1980. Outline of the classification of flowering plants (Magnoliophyta). Bot. Rev. (Lancaster) 46: 225-359.

-----. 1983. Macroevolutionary processes in the history of the plant world. Bot. Zhurn. 68: 1593-1603.

-----. 1987. Systema Magnoliophytomm. Nauka, Leninopoli.

-----. 1991. Evolutionary trends in flowering plants. Columbia University Press, New York.

-----. 1997. Diversity and classification of flowering plants. Columbia University Press, New York.

Thorne, R. F. 1958. Some guiding principles of angiosperm phylogeny. Brittonia 10: 72-77.

-----. 1963. Some problems and guiding principles of angiosperm phylogeny. Amer. Naturalist 97: 287-305.

-----. 1968. Synopsis of a putatively phylogenetic classification of the flowering plants. Aliso 6: 57-66.

-----. 1973. Inclusion of the Apiaceae (Umbelliferae) in the Araliaceae. Notes Roy. Bot. Gard. Edinburgh 32: 161-165.

-----. 1976. A phylogenetic classification of the Angiospermae. Evol. Biol. 9: 35-106.

-----. 1977. Some realignments in the Angiospermae. PI. Syst. Evol., Suppl. 1: 299-319.

-----. 1981. Phytochemistry and angiosperm phylogeny: a summary statement. Pp. 233-295 in D. A. Young & D. S. Seigler (eds.), Phytochemistry and angiosperm phylogeny. Praeger, New York.

-----. 1992a. An updated phylogenetic classification of the flowering plants. Aliso 13: 365-389.

-----. 1992b. Classification and geography of the flowering plants. Bot. Rev. (Lancaster) 58: 225-350.

Turrill, W. B. 1963. Joseph Dalton Hooker: botanist, explorer and administrator. Thomas Nelson & Sons, London.

Van Tieghem, P. 1918. Elements de botanique. 2 vols. Ed. 5. Rev. & corrected by J. Costantin. Masson, Paris.

Watson, L. 1964. The taxonomic significance of certain anatomical observations on Ericaceae. New Phytol. 63: 274-280.

Williams, R. L. 1988. Gerard and Jaume: two neglected figures in the history of Jussiaean classification. Taxon 37: 2-34, 233-271.

Woodland, D. W. 1991. Contemporary plant systematics. Prentice Hall, Englewood Cliffs, NJ.

Wright, S. P. 1992. Adjusted P-values for simultaneous inference. Biometrics 48: 1005-1013.
COPYRIGHT 1998 New York Botanical Garden
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 1998 Gale, Cengage Learning. All rights reserved.

 Reader Opinion




Article Details
Printer friendly Cite/link Email Feedback
Author:Cuerrier, Alain; Brouillet, Luc; Barabe, Denis
Publication:The Botanical Review
Date:Oct 1, 1998
Previous Article:Taxaceae: the genera and cultivates species.
Next Article:The question of cotyledon homology in angiosperms.

Terms of use | Copyright © 2015 Farlex, Inc. | Feedback | For webmasters