The classification of angiosperms at the higher level remains a most controversial field. It is true that their formal subdivision in dicotyledons and monocotyledons is undisputed, and disagreement about the recognition of families is largely due to the fact that a growing number of authors tend to elevate all subfamilies to the rank of family. I cannot see any benefit in this practice if the segregate families remain the closest relatives of the family in which they were formerly included; it is a principle of this present work not to follow this trend. However, above the level of family one enters a really murky area, in which many conflicting classifications exist. Between the levels of family and class the categories order, superorder, and subclass are in usage, but the respective taxaof different authors are often not comparable. This is to be expected, since methodology and objectives of different workers, though sometimes not explicitly stated, seem to diverge widely. Thus Dahlgren and Thorne use only the ranks of order and superorder, while Cronquist makes use of orders and subclasses; Takhtajan uses all three categories. Also the number of superorders recognized varies drastically; the dicotyledons are subdivided into 19 superorders by Thorne (1983), 25 superorders by Dahlgren (see G. Dahlgren 1989 a, b), and 38 superorders by Takhtajan (1987). Certainly all of these taxa have a degree of "naturalness", but their relatively large number in conjunction with the disagreement among taxonom-ists about their circumscription and nomenclature will not help them to be accepted in broader circles.
This seems to be quite different with the concept of subclass, which (at least in the dicotyledons) is the only category above the level of order that has come into general use. Therefore subclasses are used as a yardstick for the allocation of the more than 300 dicotyledon families to different volumes of this work. For example, the Polygonaceae and Plumbaginaceae, conventionally included in the same subclass as the
1 Doyle and Donoghue (1986) identified at least nine apomor-phies that unite the angiosperms: sieve tubes and companion cells derived from the same initials; stamens with two lateral pairs of pollen sacs; a closed carpel with stigmatic pollen germination; a hypodermal endothecium in the anther; lack of a laminated endexine; a megaspore wall without sporopollenin; a three-nuclear male gametophyte with neither prothallial cells nor a sterile cell; a megagametophyte with only eight nuclei (or various related conditions); and double fertilization with associated endosperm formation.
Caryophyllales, are treated in this volume, although I cannot see that the three groups are closely related. Admittedly, this seems to be a kind of "warehouse systematics"; however, the treatment of two families in the same volume clearly does not imply a connotation of relationship.
The concept of subclass was introduced by Takhtajan (1964) and subsequently remodeled both by him (latest version 1987) and by Cronquist (1968, 1981,
1988). In the dicotyledons these two authors largely agree in subclass delimitation (although minor differences exist), whereas in the monocotyledons the differences between their classifications are considerable. When the subclasses were designed, they were largely based on conventional alliances, such as the Polycarpicae, Apetalae, Sympetalae, Enantioblastae, etc., which had been shaped in the 19 th and early 20 th century. However, the characters on which they are based are few and/or not universally applicable. Above all this applies to the subclasses Rosidae and Dilleniidae, but also to the Hamamelididae, all of which can scarcely be characterized consistently, and which seem to be more or less confluent when considering the affinities of their primitive members. The purported difference between Rosidae and Dilleniidae in stamen development (centripetal vs. centrifugal) can only be observed in polyandrous groups, but many members of these subclasses have oligan-drous flowers. Moreover, the direction of stamen development is not. always a reliable character: in the Hamamelidaceae, for instance, both kinds of stamen development were shown to be present side by side (Endress 1976).
Nevertheless, all these subclass concepts are of some utility for handling angiosperm diversity. The Dilleniidae, for example, could be reduced to a coherent core if the Dilleniales, Theales, and iridoid-con-taining orders were removed and the Urticales and Santalales added (Huber 1990). Such a re-shaped subclass would then encompass the Capparidales, Viólales, Málvales, Euphorbiales, and Guttiferales with their respective satellite groups, but would intergrade without any clear demarcation (through Stachyurus, the Buxaceae and their allies) with the Hamamelidae.
Due to its multiple relations, the subclass Rosidae cannot be sharply delimited. The mutual relations between the Cunoniales, Hamamelidales, the numerous pinnately leaved orders ("Pinnatae"), and the Cornales blur the definition of this subclass. This is also reflected by the multiple relationships of mid-Cretaceous fossils such as the Debaya-Dewalquea complex and Sapindopsis, which suggest links to the plata-noids, Fagaceae, and pinnately leaved rosids (Crane
The subclass Hamamelididae was long known to be utterly heterogeneous, and various attempts were made to weed out the inconsistent elements. Even if the Urticales are removed from it2, whose affinity with the Malvales seems well settled (now also in terms of leaf architecture, see Wolfe 1989), there remain startling multiple relations of the higher to the lower Hamamelididae on the one hand and to the Cunoniales and other rosid elements on the other. This situation is well expressed in the classification of Dahlgren (see Dahlgren 1989 a,b).
Phytochemical evidence, which has led to the recognition of fundamentally different lines within the subclass Asteridae, was the main reason to split off the new subclass Lamiidae (Takhtajan 1987), resulting in two much more homogeneous groupings.
In the subclass Caryophyllidae, two extraneous elements, the Polygonales and Plumbaginales, are attached to a clearly monophyletic core, the Caryophyllales. This alignment is sanctioned mainly by tradition and the lack of a better alternative. The Magnoliidae s. 1. (in the sense of Cronquist, i.e., incl. Ranunculidae in the sense of Takhtajan) are certainly monophyletic in the traditional sense in spite of their relictual nature (and are not a mere "grade taxon" as sometimes suggested).3 In my opinion, the recognition of the Ranunculidae at the same rank as the Magnoliidae would make the latter paraphyletic. The Magnoliidae s.l. are considered paraphyletic by many authors who interpret the rest of the dicotyledons (and the monocotyledons) as derived from them; to my mind, this view is disputable (see. p. 10).
It goes without saying that finer primary subdivisions of the monocotyledons and dicotyledons into superorders must result in more homogeneous groups. Here the problem arises - as often in classification - that close relationships can be recognized between some of these superorders, allowing to unite them into taxa of higher rank, whereas other superorders appear isolated. Again one realizes that these taxa are hardly equivalent as evolutionary entities.
2 Because most users would expect the Urticales in the Hamamelididae, I have included them in this Volume, being aware that they are strange bedfellows here.
3 If the Nepenthaceae (to my mind related to Droseraceae), Paeonia and Glaucidium, included by Takhtajan (1987), and the Sabiaceae and Coriariaceae, hesitatingly included by Cronquist (1981), are excluded.
Was this article helpful?