A Commentary on the General System of Classification of Flowering Plants

Arthur Cronquist

(deceased)

Introduction

For the past century or more, major schemes of classification of flowering plants have attempted to portray the ensemble of similarities and differences among plants in an evolutionary context. Authors have implicitly or explicitly considered their schemes to be phylogenetic, and indeed C. E. Bessey (1915) labeled his epochal treatment The Phylogenetic Taxonomy of Flowering Plants. Phylogenetic in this traditional sense means simply that the scheme is compatible with presumed evolutionary relationships.

The much narrower and more rigid concept of "phylogenetic" recently promoted by cladistic theorists requires that any proper taxonomic group must include all the descendants of the nearest common ancestor of the group. A. Cronquist (1987) and others have discussed some of the problems of applying this concept. No general system of classification of flowering plants has yet been produced on cladistic principles.

The well-known and widely used system of Adolf Engler (H. G. A. Engler and K. Prantl 1887--1915) was considered by its author(s) to be phylogenetic at least in a general sense. That view, although defensible at the time, is no longer tenable. The system and its concepts do not meet the test of providing for all the evidence. The great weakness of the Englerian system, from a current point of view, is that it does not distinguish adequately between primitive simplicity and simplicity by reduction. Inasmuch as most students of the subject now agree that floral reduction has been a pervasive (though not exclusive) trend within the angiosperms, the system must be extensively recast.

The search for general agreement and a set of principles that will permit everything to fall into place has led taxonomists to revive, modify, and expand the concepts of floral evolution first presented in embryonic form by A. P. de Candolle in 1813. The system of G. Bentham and J. D. Hooker in their Genera Plantarum(1862--1883) was a lineal descendant of that of de Candolle. Although post-Darwinian in publication, it was pre-Darwinian in concept, and its authors never claimed anything else. It was, however, an important historical link in the progression from de Candolle's "natural" system to the avowedly phylogenetic system of C. E. Bessey. The oft-noted insertion of the gymnosperms between the monocotyledons and dicotyledons in the Bentham and Hooker system is a curious anachronism that does not seriously affect their treatment of the angiosperms.

The now widely accepted strobilar hypothesis of floral evolution dates from C. E. Bessey (1897), who saw "the reproductive strobilus in the form of a flower, in which the sterile leaves are well set off from those which bear the spores." Bessey's interpretation was a particular application and modification of F. O. Bower's (1894) theory of the strobilus in archegoniate plants. This interpretation is perfectly compatible in principle with A.P. de Candolle's pre-Darwinian views, with the addition that the relationships among the taxa are considered to be evolutionary rather than purely conceptual.

Authors since C. E. Bessey's time, including among others J. Hutchinson (1926--1934 et seq.), K. R. Sporne (1949 et seq.), A. L. Takhtajan (1959 et seq.), R. F. Thorne (1963 et seq.), and A. Cronquist (1968 et seq.), have built on his concepts of what we would now call polarity in the morphology of angiosperms. (Takhtajan's publications in this field actually began in 1942, but they did not begin to have a major impact on botanical thinking until the 1959 book here cited). Some differences of opinion about polarity remain, but all of the widely respected systems of classification proposed in the last several decades fit into the de Candolle-Bentham and Hooker-Bessey tradition (Takhtajan told me that he takes his inspiration more nearly from H. Hallier [1901 et seq.] than from Bessey, but his views on polarity are in the Besseyan mainstream, which he himself has helped to channel.) Considerable blocks from the Englerian system have remained virtually intact in the newer systems, but they are rearranged inter se to reflect more recent concepts of polarity.

The general system of classification of angiosperms used in the Flora of North America is the "integrated system" of A. Cronquist (1981, slightly modified in 1988). Among other modern systems, the integrated system is most similar to those of A. L. Takhtajan, as presented in progressively modified versions over several decades, most recently in 1986 and 1987. Although these two sets of systems are conceptually similar, Takhtajan recognizes a significantly larger number of families, orders, and subclasses than Cronquist. Furthermore, some families, such as the Euphorbiaceae and Urticaceae, appear in very different places in the works of these two authors. Differences in the rank at which taxa are received (and thus in the number recognized at a given level) are inherently unresolvable except by fiat. Differences in the position of various taxa in different systems often reflect the difficulty of distinguishing convergence or parallelism from synapomorphy. Molecular techniques may eventually help to resolve some of these problems, but the time (1992) is not quite yet.

Dicots and Monocots

It has been recognized for more than a century that the angiosperms form a natural group that consists of two subgroups. These two subgroups have usually been called dicotyledons and monocotyledons (or equivalent names with Latinized endings), from the most nearly constant of the several differences between them. As formal taxa they are here considered to be classes, called Magnoliopsida and Liliopsida. Less formally they may be called dicots and monocots.

Both groups occur in a wide variety of habitats, but the dicotyledons are the more diverse in habit. About half of all the species of dicots are more or less woody-stemmed, and many of them are definitely trees, usually with a deliquescently branched trunk. The monocots, in contrast, are predominantly herbaceous. Fewer than 10% of all monocots are woody, and most of these belong to the single large family Arecaceae (Palmae), which has only a few species in the area of our flora. Woody monocots usually have an unbranched (or sparingly branched) stem with a terminal crown of leaves, a habit that is rare among dicots. The difference in habit is partly a reflection of the complete absence of typical cambium in monocots, in contrast to its usual presence in dicots.

Differences exist in the underground as well as the aerial parts of dicots and monocots. In monocots the primary root soon aborts, and the mature root system is wholly adventitious. Many dicots likewise have an adventitious root system, but a primary root system, derived from the radicle, is more common. The adventitious, fibrous root system of monocots is a consequence of the absence of cambium. Having no adequate means of secondary thickening, individual roots cannot persist, enlarge, and ramify. The roots of some monocots do manage to penetrate deeply into the soil, but the largest single family, the Orchidaceae, is shallow-rooted and mycorrhizal, and the next largest family, the Poaceae, tends to exploit mainly the upper part of the soil, often forming a dense turf. Another large family, the Liliaceae, often has contractile roots that pull the bulb progressively deeper into the soil with the passing years. Creeping rhizomes, which may penetrate to any depth, serve as rootstocks for many of the Poaceae, Cyperaceae, Liliaceae, and other monocots.

All differences between dicots and monocots are subject to overlap or exception. The most nearly constant difference is the number of cotyledons, but some dicots have only one cotyledon, and some members of both groups have an undifferentiated embryo without cotyledons. The several differences between dicots and monocots are summarized in table 14.1.

---

TABLE 14.1. Main Differences between Dicots and Monocots

---

Dicots	Monocots
Cotyledons 2 (seldom, 1, 3 or 4, or the embryo seldom undifferentiated)	Cotyledon 1 (or the embryo sometimes undifferentiated)
Leaves mostly net-veined	Leaves mostly parallel-veined
Intrafascicular cambium usually present	Intrafascicular cambium lacking; usually no cambium or any sort
Vascular bundles of the stem usually borne in a ring that encloses a pith	Vascular bundles of the stem generally scattered, or in 2 or more rings
Floral parts, when of definite number, typically borne in sets of 5, less often 4, seldom 3 (carpels often fewer)	Floral parts, when of definite number, typically borne in sets of 3, seldom 4, almost never 5 (carpels often fewer)
Pollen typically 3-aperturate, or of 3-aperturate-derived type, except in a few of the more archaic families	Pollen 1-aperturate or of 1-aperturate-derived type
Mature root system either primary or adventitious, or both	Mature root system wholly adventitious

---

It is widely agreed that the monocots are derived from primitive dicots, and that therefore the monocots must follow rather than precede the dicots in any proper linear sequence. The solitary cotyledon, parallel-veined leaves, absence of a cambium, dissected stele, and adventitious root system of monocots are all regarded as apomorphic characters within the angiosperms, and any plant that was plesiomorphic (i.e., more primitive than the monocots) in these several respects would certainly be a dicot. Monocots are more primitive than the bulk of the dicots in having mostly 1-aperturate pollen, but several of the more archaic families of dicots also have 1-aperturate pollen.

Despite the individual failure of all the characters used to separate monocots from dicots, there is seldom any problem in assigning a particular family or order to the one group or the other. Only the Nymphaeales (here, as customarily, assigned to the dicots) excite any continuing controversy in this regard (R. W. Haines and K. A. Lye 1975). Both Takhtajan and I regard Nymphaeales as the probable sister group of the monocots.

The Subclasses of Dicots

The dicots are here considered to consist of six subclasses, Magnoliidae, Hamamelidae, Caryophyllidae, Dilleniidae, Rosidae, and Asteridae (fig. 14.1). These are groups that cohere on the basis of all our information but that cannot precisely be defined phenetically. They can be characterized only in generalities and in terms of "critical tendencies" (a term suggested by H. F. Wernham 1912). They are polythetic groups, established by the accretion of apparently related members. Our minds demand some sort of organization of the many orders of dicots into a smaller number of affinity groups, even though the resulting groups are poorly characterized.

Magnoliidae

In general, it may be said that the Magnoliidae are the basal complex from which all other angiosperms have been derived. The Magnoliidae consist of nearly all of what has been informally called the Ranalian complex, plus a few other families. In cladistic terms the subclass is a prime example of a paraphyletic group, that is, a group that does not contain all the descendants from the immediate common ancestor.

The Magnoliales may be briefly characterized as those families of Magnoliidae that have ethereal oil cells, 1-aperturate (or sometimes inaperturate or 2-aperturate) pollen, and usually hypogynous flowers with a well-developed perianth of usually separate tepals. They are all woody plants. All of the putatively primitive features of angiosperms occur in one or another family of Magnoliales, but some of these features occur in other orders as well.

Based on the comparative morphology of modern species and on early Cretaceous fossil pollen and leaves, the Magnoliales must be considered the most archaic order of flowering plants (fig. 14.2). The fossil record clearly shows that the earliest angiosperm pollen grains were monosulcate and that simple, entire leaves with relatively poorly organized pinnate venation are the ancestral archetype. Among modern angiosperms, only the Magnoliales can accommodate this combination of characteristics.

No one family of Magnoliales can be considered ancestral to the rest of the order. Each family presents its own combination of primitive and advanced characters.

The Nymphaeales may be briefly characterized as those families of the Magnoliidae that are aquatic herbs without ethereal oil cells and without vessels in the shoot. Most of them have 1-aperturate or inaperturate pollen, but the distinctive Nelumbonaceae have 3-aperturate pollen. Most of the Nymphaeales have long-petiolate leaves with broad, cordate to hastate or peltate, floating blades, but in Ceratophyllum the leaves are all slender and submersed, and in Nelumbo many of the leaves are emergent.

Unlike most orders of angiosperms, the Nymphaeales have an obvious niche. They inhabit still waters, and many of their characteristics reflect adaptation to this habitat. They do not occupy this habitat to the exclusion of other groups, however. Nymphoides (Menyanthaceae) is very nymphaeaceous in aspect and habitat, and Myriophyllum (Haloragaceae) likewise recalls Ceratophyllum.

Based on leaf fossils, it appears that the Nymphaeales have a long history, going all the way back to the Albian stage of the Lower Cretaceous. The fairly numerous and varied Albian fossils of this sort are coming to be called nymphaeaphylls. The monosulcate pollen of many of the modern Nymphaeaceae and Cabombaceae bespeaks an ancestry among the more primitive dicotyledons, but it does not specify the time of origin. Nymphaealean pollen is in fact not very distinctive, and it can be traced with some certainty only to the uppermost Cretaceous.

It is here considered that the modern Nymphaeales all descend from a group of primitive dicotyledons that took to an aquatic habitat and became herbaceous very early in the history of the angiosperms. A subsequent early dichotomy resulted in a line leading to the modern Nelumbonaceae, and a line leading eventually to the other four families. The modern families of the order thus represent a series of isolated endlines, comparable on a smaller scale to the series of isolated endlines that comprise the families of the modern orders Magnoliales and Laurales.

The Ranunculales and Papaverales form a pair of closely related orders that stand somewhat apart from the remainder of the Magnoliidae. The mostly apocarpous flowers of the Ranunculales and the isoquinoline alkaloids of both orders tie the pair to the Magnoliidae, but the mostly 3-aperturate pollen separates them from other Magnoliidae except Illiciales and Nelumbo, and the absence of ethereal oil cells separates them from other orders except the Nymphaeales. The scanty fossil record does not carry the pair back beyond the Tertiary. Conceivably these two orders are of relatively recent origin, despite their retention of some archaic features.

Hamamelidae

The Hamamelidae are a group of mostly wind-pollinated families with reduced, usually apetalous flowers that are often borne in catkins. This subclass consists chiefly of the core of the traditional Amentiferae, after some unrelated families such as the Salicaceae have been excluded. Plants with highly reduced flowers also occur in each of the other subclasses of dicots, but the Hamamelidae are an ancient major group with reduced flowers.

The Hamamelidales are morphologically central to their subclass. Except for the highly archaic Trochodendrales (two monotypic families), all the other orders appear to tie back directly or indirectly to the Hamamelidales (fig.14.3).

The Hamamelidae can be traced back through the platanoid line (Hamamelidales) to near the middle of the Albian (final) stage of the Lower Cretaceous period. Members of the orders Juglandales, Myricales, Fagales, and Casuarinales have a distinctive sort of pollen that appears to take its evolutionary origin in the Normapolles complex of Middle Cenomanian (early Upper Cretaceous) time. The Urticales also have a long fossil record, with pollen dating from the Turonian, some 90 M.Y.B.P. (the next stage above the Cenomanian).

The Urticales have traditionally been associated with some of the other orders here referred to subclass Hamamelidae. The reduced flowers provide the obvious initial basis for such a treatment, but vegetative anatomy (O. Tippo 1938; E. M. Sweitzer 1971) and leaf venation (J.A. Wolfe 1973) have also been adduced to support the association. Even the pollen is said to resemble that of the Normapolles group of early Upper Cretaceous fossils. If the virtually apocarpous archaic genus Barbeya is, as by most authors, associated in some way with Urticales, it further strengthens the connection of the latter to Hamamelidae rather than to any other subclass.

A longstanding school of thought, exemplified most recently by R. F. Thorne (1973), C. C. Berg (1977), and A.L. Takhtajan (1987), holds that the Urticales are allied to the Malvales rather than to the Hamamelidales. The stratified phloem and presence of mucilaginous cells and ducts in both orders have played a large role in this assessment of relationship. Thorne has tried to mitigate the problem of Barbeya by excluding it from Urticales and considering it to be incertae sedis.

Caryophyllidae

The Caryophyllidae consist of the large order Caryophyllales plus two smaller orders (Polygonales and Plumbaginales) that are customarily associated with it (fig. 14.4). Each of the three orders is well marked, but no one distinctive feature marks the subclass. The simplest way to characterize the group is to say that it consists of those dicotyledons that have bitegmic, crassinucellar ovules and either have betalains instead of anthocyanins or have free central or basal placentation in a compound ovary. Furthermore, most species are herbaceous, and woody species usually have anomalous secondary growth or otherwise anomalous stem structure. The stamens, when numerous, originate in centrifugal sequence, and the pollen grains are usually 3-nucleate. The food storage tissue of the seed is typically starchy, and very often has clustered starch grains.

The fossil record as presently interpreted carries the Caryophyllidae back only to the Maastrichtian epoch of the latest Upper Cretaceous, some 70 M.Y.B.P. (pollen of Amaranthaceae or Chenopodiaceae). The relatively short fossil history, as contrasted to the Magnoliidae, Hamamelidae, and Rosidae, is consonant with the primitively herbaceous habit of Caryophyllidae. Aside from the ancestors of the Nymphaeales, dicotyledonous herbs apparently played only a negligible role in the vegetation of the Cretaceous.

The ancestry of Caryophyllidae may lie in or near Ranunculaceae. In the absence of known fossil connections, it may be supposed that the common ancestor of the Caryophyllidae was an herb with hypogynous flowers and separate carpels, without petals. The number of potentially ancestral groups is thus immediately limited. The possibility that members of the Ranunculaceae may be at least collateral ancestors is bolstered by the fact that some of them have pollen very much like that of many Caryophyllales. The floral trimery of the Polygonaceae also has ample precedent in the Ranunculaceae. The evolutionary significance of the centrifugal androecium in Caryophyllales can scarcely be evaluated until a satisfactory general interpretation of the origin of centrifugality is achieved. Further speculation about the ancestry of Caryophyllidae is hampered by the uncertainty about the affinity of Polygonales and Plumbaginales to Caryophyllales.

Chemical and ultrastructural studies of the past two or three decades have led to a consensus as to the contents and characterization of the Caryophyllales, now one of the best defined major orders of angiosperms. All investigated members of the order have a characteristic type of sieve-tube plastid that is unknown in other angiosperms (H.-D. Behnke 1976). The plastid contains a set of bundles of proteinaceous filaments that collectively form a subperipheral ring. Often there is a larger central protein crystalloid, which may be either globular or polyhedral. Sieve-tube plastids with proteinaceous inclusions occur in some other dicotyledons (notably some members of the Magnoliidae and Fabaceae), but in those groups the inclusions do not form a ring of filaments.

Ten of the twelve families of Caryophyllales consistently produce betalains but lack anthocyanins. Betalains are otherwise unknown in angiosperms, although they have been found in some Basidiomycetes. The absence of betalains and presence of anthocyanins in the Caryophyllaceae and Molluginaceae led some botanists a few years ago to propose a fragmentation of the order, but this difference is now generally acknowledged to be of no more than subordinal importance.

Dilleniidae

The weakest distinction among the subclasses of dicotyledons is that between the Dilleniidae and Rosidae. Both groups are more advanced than the Magnoliidae in one or another respect, but less advanced than the Asteridae. The two taxa are kept apart as subclasses because each seems to constitute a natural group separately derived from the ancestral Magnoliidae rather than because of any definitive distinguishing characteristics. The same sorts of evolutionary advances have occurred in both groups, but with different frequencies. Despite the lack of solid distinguishing criteria, it is conceptually more useful to hold the two as separate subclasses than to combine them into one or to abandon any attempt at the organization of the Magnoliopsida into subclasses.

The Dilleniidae cannot be fully characterized morphologically. Except for the rather small (400 species) order Dilleniales, the vast majority of the Dilleniidae are sharply set off from characteristic members of Magnoliidae by being syncarpous. With few exceptions, the species of Dilleniidae with numerous stamens have the stamens initiated in centrifugal sequence. In this respect, they differ from Rosidae, in which the species with numerous stamens usually have a centripetal sequence of development.

More than a third of the species of Dilleniidae have parietal placentation, in contrast to the relative rarity of this type in the Rosidae. (Application of the term "parietal" is here restricted to compound ovaries.) Another third of the species (not the same third) are sympetalous, but only a very few of these (e.g., Diapensia) have isomerous, epipetalous stamens alternate with the corolla lobes and also unitegmic, tenuinucellar ovules as in Asteridae. Sympetaly is rare in the Rosidae. Ovules in the Dillenidae as a whole are bitegmic or less often unitegmic, with various transitional types, and they range from crassinucellar to tenuinucellar. Often they are bitegmic and tenuinucellar, a combination rare outside this group.

Compound leaves with distinct, articulated leaflets are much more common in the Rosidae than in the Dilleniidae. Uniovulate or biovulate locules are much less common in the Dilleniidae than in the Rosidae, but they are well represented in the Malvales, whose position in the Dilleniidae is well established. Not many of the Dilleniidae have a typical nectary disk of the sort so common in the Rosidae, but other types of nectaries are common.

It seems clear that the Dilleniidae take their origin in the Magnoliidae. The apocarpous order Dilleniales, especially the family Dilleniaceae, forms a connecting link between the two subclasses. If the rest of the Dilleniidae did not exist, the Dilleniales could easily be accommodated as a peripheral order of Magnoliidae. On the other hand, the Dilleniales do not closely resemble any one family of Magnoliidae, and they are sharply set off by a series of chemical features.

Unlike the vast majority of the Magnoliidae, the Dilleniaceae have ellagic acid, proanthocyanins, and raphides. The ethereal oil cells so characteristic of the woody Magnoliidae are absent from the Dilleniaceae. The Dilleniaceae are almost without alkaloids, and the few that have been found in some species have nothing to do with the characteristic benzyl-isoquinoline alkaloids of Magnoliidae.

Furthermore, there is a still unresolved controversy about the ancestry of the multistaminate, centrifugal androecium such as that in the Dilleniidae. If, as some authors maintain, this kind of androecium is necessarily derived from an ordinary, cyclic, oligostemonous androecium by secondary increase in number of stamens, then the gap between the Dilleniaceae and Magnoliidae is further widened.

The centrifugal androecium and frequently campylotropous or amphitropous ovules of the Dilleniidae recall the Caryophyllidae, but the latter have their own set of specialized features not found in the Dilleniaceae and other Dilleniidae. Any evolutionary relationship between the Dilleniaceae and Caryophyllidae must be rather remote in time.

Pollen that appears to represent the Dilleniidae dates from about the beginning of the Upper Cretaceous, but this early pollen is not clearly referable to an order. Otherwise the fossil record as presently understood gives no clear indication of the origin of the group. If my hypothesis of chemical evolution (A. Cronquist 1977b, elaborated in 1988) is correct, the Dilleniidae probably originated about the same time as the Hamamelidae and the Rosidae. The Hamamelidae and the more archaic members of the Dilleniidae and Rosidae characteristically rely heavily on hydrolyzable tannins as defensive weapons. The more advanced members of the latter two groups have largely discarded tannins in favor of more recently evolved defenses.

In contrast to their evident separation from the Magnoliidae and Caryophyllidae, the Dilleniaceae are obviously allied to such syncarpous families as the Actinidiaceae and Theaceae. For purposes of conceptual organization, however, it is useful to put the largely apocarpous family Dilleniaceae in a separate order from the largely syncarpous Theales. The Dilleniaceae appear to be the modern remnants of the group that gave rise not only to Theales but also to nearly all the rest of the subclass Dilleniidae. The Paeoniaceae, the only other family of the Dilleniales, stand somewhat apart and do not appear to be in the main line of evolution of the subclass.

Theales are the central group of Dilleniidae, from which all the other orders except the Dilleniales appear to have evolved. The Malvales, Lecythidales, Violales, Capparales, and Nepenthales all appear to have arisen from a common complex in the Theales. The Salicales are an amentiferous offshoot from the Violales, and the Batales appear to be related to the Capparales. The remaining four orders take their origin in a different part (or parts) of the Theales. The Ericales are evidently allied to the Actinidiaceae (Theales), and the Diapensiales appear to be allied to the Ericales. The Ebenales and Primulales are somewhat more remote, but they may be allied to each other and to a lesser extent to Ericales. Only the Theales provide a reasonably likely origin for these groups.

The concepts of relationships within Dilleniidae that are here expounded may be conveniently expressed in the phylogenetic diagram (fig. 14.5).

Rosidae

The 18 orders that make up the Rosidae evidently cohere as a natural group. Only the Euphorbiales and Rafflesiales are obviously debatable, the latter because their morphological reduction in association with parasitism makes their affinities hard to establish. The position of the Euphorbiaceae has long been uncertain. Some authors have associated the Euphorbiaceae with Malvales, others with orders here assigned to Rosidae, and some have used the apparently dual affinity of the Euphorbiaceae to cast doubt on the rosid-dilleniid distinction.

The Buxaceae and Simmondsiaceae, often (as here) associated with the Euphorbiaceae, may not properly belong there. Takhtajan may well be right in putting these two small families into the Hamamelidae in the more recent versions of his system. For the Buxaceae alone, that would present no great problem. Unfortunately the foliaceous-accrescent sepals of Simmondsia are quite out of harmony with the Hamamelidae, yet the affinity of Simmondsia (even as a separate family Simmondsiaceae) to the Buxaceae is generally acknowledged. The position of the several families of Euphorbiales as here constituted would be a suitable subject for molecular techniques, especially because of the wide diversity of recent opinions based on phenetic data.

The Rosales as here constituted form an exceedingly diverse order, standing at the evolutionary base of their subclass. In effect, they are what is left after all the more advanced, specialized orders of Rosidae have been delimited (fig. 14.6). Aside from internal phloem and a parasitic or highly modified aquatic habit, most of the features that mark the more advanced orders of Rosidae (and indeed even some of the features of Asteridae) can be found individually within the order Rosales, but in the Rosales these features do not occur in the combinations that mark the more advanced groups.

Two characteristics that are very common in the Rosales are much less common among other orders of the subclass. These are a polymerous androecium and an apocarpous gynoecium (although the gynoecium is monocarpous in the large order Fabales). Furthermore, a great many of the Rosales have more or less numerous ovules per carpel, and relatively few have only one or two. In contrast, many of the other orders of Rosidae have only one or two ovules per carpel, and most of those that have numerous ovules are well differentiated from the Rosales in other respects.

The order Fabales, as here treated, consists of three closely allied families, the Fabaceae, Caesalpiniaceae, and Mimosaceae. The existence of three major groups (here called families) that collectively constitute a larger group (here called an order) is widely admitted. The rank of the groups is in dispute, however. Many authors have preferred to recognize a single family Leguminosae (or Fabaceae sensu lato), with three subfamilies. The broadly defined family is then often included in Rosales. There is no objective right or wrong here. I prefer the treatment here presented as being more in harmony with the customary definitions of families of angiosperms. The legumes, whether treated as one family or three, form such a coherent group with such an abundance of genera and species that they would appear to dominate any other order to which they might be assigned. At some 18,000 species, they are second only to Asterales among dicotyledonous orders. Morphologically, the Fabales form one of the better-defined orders of dicotyledons. The combination of compound, stipulate leaves and flowers with a single carpel will distinguish the vast majority of Fabales from virtually all other groups.

The closest linkage of the Fabales to Rosales lies in the fairly small (300--400 species), mainly tropical family Connaraceae, which is not represented in our flora. The Connaraceae lie in the nebulous area where the Rosales, Fabales, and Sapindales join at their evolutionary base.

The Myrtales, with some 14 families and more than 9000 species, are one of the better defined large orders of dicotyledons. In addition to the mostly strongly perigynous to epigynous flowers and the overlapping similarities among the constituent families, the order is marked by two otherwise uncommon anatomical features: internal phloem and vestured pits in the vessel segments.

The most discordant family of the Myrtales, and the only one that still provokes controversy about its possible inclusion in the order, is the Thymelaeaceae. The Thymelaeaceae are marked by their usually pseudomonomerous ovary, often unusual pollen, and a distinctive set of secondary metabolites, commonly including the simple coumarin daphnin (or allied compounds). Furthermore, some few members of the family are unusual in Myrtales in having essentially hypogynous flowers. More ordinary kinds of pollen and gynoecia, with transitional types, also occur in the family, however. Thus it is unnecessary to seek placement of the Thymelaeaceae in another order. Indeed the internal phloem, vestured pits, and strongly perigynous, polypetalous to apetalous flowers of characteristic members of the family would be out of harmony with any other order that might be suggested as a haven for it.

Furthermore, the characteristic obturator of the Thymelaeaceae, though not identical in detail, might be compared with the obturator of the Combretaceae, and the glandular-punctate leaves of some Thymealeaceae recall those of the Myrtaceae. It would, of course, be possible to recognize an order Thymealeales to provide for this one family, as some authors have done, but the segregate order would still stand alongside Myrtales. We should also note that P. G. Martin and J. M. Dowd (1986) linked the Thymelaeaceae to the Myrtales on the basis of the sequence of amino acids in the terminal 40 residues of the smaller subunit of ribulose biphosphate carboxylase.

The Sapindales, with 15 families and about 5400 species, form a well-characterized natural group. Only two families are really peripheral. The Staphyleaceae connect the Sapindales to the ancestral Rosales, in the vicinity of the Cunoniaceae, and the Zygophyllaceae are suggestive of the Geraniales. These two families also differ from the bulk of the order in often having more than two ovules per locule.

The features common to most members of the Sapindales, which make it useful to distinguish them as a group from the Rosales, are the compound or cleft leaves, haplostemonous or diplostemonous androecium, well-developed nectary disk, and syncarpous ovary with a limited number of ovules (usually only one or two) in each locule. All of these features can be found individually in the Rosales, but mostly not in combination.

Asteridae

The Asteridae are the best characterized subclass of dicotyledons, marked by their sympetalous flowers, in which the stamens are isomerous and alternate with the corolla lobes, or fewer than the corolla lobes. Much less than 1% of the species of Asteridae fail this test, and probably no more than 1% of the species that do meet the test do not belong to Asteridae. The tenuinucellar ovule with a massive single integument is a further marker of the group, but there are more exceptions to this feature, both within and without the Asteridae. The Callitrichales are the most aberrant order in the subclass. Except for the vestigial calyx in Hippuris, the Callitrichales lack a perianth entirely.

Chemically, the Asteridae are noteworthy for the frequent occurrence of iridoid compounds, the usual absence of ellagic acid and proanthocyanins, and the apparently complete absence of betalains, benzyl-isoquinoline alkaloids, and mustard oils. The absence of betalains and benzyl-isoquinoline alkaloids tends to set Asteridae off from most Caryophyllidae and Magnoliidae, respectively, and the usual absence of ellagic acid and proanthocyanins tends to set them off from Rosidae, Dilleniidae, and Hamamelidae. The distinction is far from absolute, however, because these substances are missing from many members of the subclasses they are supposed to characterize.

Ancestry of the Asteridae very probably lies in the order Rosales sensu latissimo. Iridoid compounds, a sympetalous corolla, stamens isomerous and alternate with the petals, a compound pistil with numerous ovules on axile placentas, unitegmic ovules, and tenuinucellar ovules occur in this order, but not in combination. The nectary disk of many Asteridae also finds a ready precedent in Rosales, as do the simple, stipulate, opposite leaves of the more archaic members of the group. The combination of these features into a functional whole marks the transition from the ancestral Rosales to the first members of the Asteridae. All the other orders of Rosidae are already too advanced to serve as plausible ancestors of the Asteridae.

The Asteridae make a relatively late entry into the fossil record. There is no reason to suppose that the group originated before the beginning of the Tertiary period, and its members began to play a prominent role in the flora of the world only during the Oligocene.

The Asteridae are the most advanced subclass of dicotyledons, and possibly the most recently evolved; only the Caryophyllidae may be more recent. More than any other subclass, they exploit specialized pollinators and specialized means of presenting the pollen. It seems likely that the rise of the Asteridae is correlated with the evolution of insects capable of recognizing complex floral patterns.

The Asteridae form such a coherent group that delimitation of the orders becomes problematical. The Gentianales appear to stand near the base of the subclass (fig. 14.7). The Rubiales link to both the Gentianales and the Dipsacales. The Menyanthaceae and Buddlejaceae, often included in Gentianales, are here referred to Solanales and Scrophulariales, respectively.

The largest order of Asteridae, in terms of number of families, is Scrophulariales. The Scrophulariaceae are the largest family of the order and are also central to it. Four of the other families (Acanthaceae, Bignoniaceae, Globulariaceae, and Orobanchaceae) are connected to the Scrophulariaceae by genera or groups of genera that have by different authors been referred to the central or the peripheral family. Although the seven remaining families of the order are more sharply delimited, most of them may logically be considered to be specialized derivatives of the Scrophulariaceae.

The order Asterales consists of the single worldwide family Asteraceae (Compositae), with perhaps as many as 20,000 species. The Asteraceae are one of the more successful families of flowering plants, represented by numerous genera, species, and individuals. Not many of them are forest trees, and only a few are aquatic, but otherwise they exploit most of the obvious kinds of ecological opportunities open to angiosperms.

Affinities of the Asteraceae have been vigorously but inconclusively debated. Traditionally they have been thought to be allied to the Campanulaceae, and often even included in the same order. I have argued (1977, 1981) that patterns of relationship within the family suggest that the tribe Heliantheae is the basal group, and that their immediate ancestors must have been woody plants with opposite leaves and a cymose inflorescence, as in the Rubiaceae. Ongoing studies of chloroplast DNA in the Asteraceae and some other families by R.K. Jansen and J.D. Palmer (e.g., 1988) and their associates suggest that Barnadesia and its immediate allies in the tribe Mutisieae may be the sister group of all the other Asteraceae. The consequences of this view to the taxonomic and phylogenetic interpretation of the family are only beginning to be explored.

Monocots

The differences between monocots and dicots and the probable origin of the monocots from dicots have already been discussed. Some further considerations are in order here.

The fossil pollen record suggests that the origin of monocots from primitive dicots in Aptian-Albian times was the first significant dichotomy in the evolutionary diversification of the angiosperms. The wide variety of monocotyledonous leaves found in the late Lower Cretaceous and through the Upper Cretaceous attests to the continuing diversification of the group during this time, but most of these leaves cannot be referred with any certainty to modern groups.

The first modern family of monocots to be clearly represented in the fossil record is the Arecaceae (subclass Arecidae), near the base of the Santonian (or perhaps the Conacian) epoch, but palms are surely not primitive monocots. Their large, distinctive, readily fossilized leaves merely make the group easy to recognize from its inception. Pollen that probably represents Cyperales or Restionales (subclass Commelinidae) appears in the late Upper Cretaceous, probably before the Maastrichtian, and the distinctive leaves of Zingiberales show up in the Maastrichtian. Pollen thought to represent Pandanus occurs in Maastrichtian deposits in North America.

The Alismatidae and Liliidae are not certainly recognizable before the Tertiary, but some of the miscellaneous Upper Cretaceous fossil monocot leaves might well belong to one of these groups. We can be reasonably sure that the palms and the Zingiberales did not arise long before the first appearance of their characteristic leaves in the fossil record, but we cannot be so confident that other large groups did not long antedate their first identifiable fossils. Thus the fossil record, as presently understood, is compatible with any of several different views about the Cretaceous diversification of monocots. The principal constraint is the recognition of a very early dichotomy between monocots and dicots.

The dicots that gave rise to the monocots may have had apocarpous flowers with a fairly ordinary (not highly specialized) perianth, and with 1-aperturate pollen. They must have been herbs without a very active cambium, and they presumably had laminar placentation. The only modern group of dicots that meets these specifications is the Nymphaeales.

It is not here suggested that the Nymphaeales are directly ancestral to the monocots as a whole, but rather that the premonocotyledonous dicots were probably something like the modern Nymphaleales. As noted, an aquatic group of angiosperms with leaves much like those of the modern Nymphaeales was already proliferating in the Albian epoch of the Lower Cretaceous. The modern Nymphaeales are aquatic, mostly lack vessels, and show tendencies toward the fusion of two cotyledons into one.

An interesting difference between monocots and dicots is that whereas in dicots the vessels appear first, phyletically, in the secondary wood of the stem and spread to other tissues and organs, in monocots they appear first in the roots. This fact led V. I. Cheadle (1953) and others to suppose that vessels originated independently in the two classes. They therefore consider that the evolutionary divergence of the two classes preceded the origin of vessels. As I have argued elsewhere (1988), it is at least equally plausible that in the ancestral premonocots, as in their probable relatives the Nymphaeales, vessels were phyletically lost in association with the aquatic habitat. Loss of the cambium eliminated at one stroke all vessels that had not worked their way, phyletically, into the primary tissues. Subsequent evolution of vessels in the monocots had to begin essentially de novo, in association with the return to a terrestrial habitat.

The typical parallel-veined leaf of monocots is here considered to be a modified, bladeless petiole. This morphological interpretation for the leaves of Sagittaria was proposed a century and a half ago by A. P. de Candolle (1827, vol. 1, p. 286). It was further elaborated in evolutionary terms and applied to monocots as a whole by A.Arber in 1925. It is the only hypothesis known to me that permits all the information about monocots to fall into place and make sense. Even the ontogeny of the typical monocot leaf is highly compatible with the petiolar hypothesis. The blade typically develops from a portion of the leaf primordium somewhat behind the tip and matures basipetally; the primordial tip is inactive or produces only a terminal point or small appendage on the blade.

Terrestrial monocots with a well-defined, net-veined leaf blade are here considered to be derived from ancestors with narrow, parallel-veined leaves lacking a well-defined blade. All transitional stages can be seen in several families. An attempt to read the system the other way (R.M.T. Dahlgren et al. 1985) means that we must start with broad, more or less net-veined leaves in diverse groups of monocots having little to do with each other, and have all these converge in both floral and vegetative characters into a hopelessly polyphyletic core of typical monocots.

Three principal ways exist by which the typical monocot leaf can become broad and more or less net-veined. One way is to spread the main veins farther apart near the middle of the blade and amplify the cross-connections among them. Subsequently the main veins can fade out before reaching the leaf tip, so that a more or less palmate venation is established. Alisma, Sagittaria, Dioscorea, Smilax, Trillium, and many aroids exemplify this type of change. A second way is for each of the many closely set parallel veins to diverge in turn toward the margin, the outermost veins first, those nearest the midrib last. The result is a pinnately veined leaf with numerous closely parallel primary lateral veins. Members of Zingiberales reflect this sort of change. A third way, known only in the palms and Cyclanthales, differs from the second way in the intercalation of new tissue between the lateral veins during the early growth of the leaf. The plicate structure of palm and cyclanth leaves reflects this ontogeny.

If the interpretation here presented is correct, the aquatic ancestry of the monocots has had a profound effect on the subsequent evolutionary history of the group. As aquatic herbs, early monocots were preadapted to evolve terrestrial herbaceous forms, filling a niche (or set of niches) not then effectively occupied by dicots. Nevertheless, the evolution of an effective water-conducting system (with well-developed vessels), of expanded, net-veined leaves, of a branching, arborescent habit, and of a means of secondary thickening has not been easy for them. No monocot has evolved a coherent syndrome of these features that would permit a broad-scale evolutionary challenge to woody dicotyledons. Even among those monocots that have evolved a broad, more or less net-veined blade, traces of the ancestral parallel-veined pattern usually persist. We have noted that aroids, palms, and the Zingiberales have taken three essentially different routes in the evolutionary expansion of leaves, and the difference is reflected in the mature morphology.

Although the monocots are considered to have an aquatic ancestry, the situation is not simple. It appears that terrestrial monocots, derived from aquatic early monocots, have themselves repeatedly given rise to groups that have returned to the water. Among the modern Alismatidae, there appears to be a progressive adaptation to an aquatic and eventually marine habitat. In the subclass Arecidae, the mainly terrestrial family Araceae has some secondarily aquatic forms that point toward the thalloid, aquatic family Lemnaceae. In a third subclass, the Commelinidae, such aquatic families as the Mayacaceae, Sparganiaceae, and Typhaceae appear to be derived from terrestrial ancestors within the group. The aquatic habit of the Pontederiaceae (Liliidae) may likewise be secondary.

The nature of the single cotyledon in monocots has occasioned much study and controversy. Like the foliage leaves, the cotyledon often has a basal sheath surmounted on one side by a limb that may or may not be divided into blade and petiole. Typically, the sheath is closed and tubular, at least near the base. The vascular supply typically consists of two near-median bundles, as in the individual cotyledons of dicots. This implies that the monocot cotyledon is equivalent to a single leaf and is not a double structure as has sometimes been supposed. The sheathing base of the cotyledon is thus left unexplained, except that it is comparable to the sheathing base of a foliage leaf.

Drawing on evidence from living members of the Nymphaeales (dicots), I have suggested an alternative interpretation (1968, 1981, 1988): two ancestral cotyledons have become connate by their margins toward the base, forming a bilobed, basally tubular, compound cotyledon. One of the lobes has subsequently been reduced and lost, and its vascular supply suppressed, so that the embryo has, in effect, a single cotyledon with a sheathing, tubular base. I have further suggested that this modified cotyledonary structure has so firmly impressed itself on the growth pattern of the embryo that subsequent leaves are also built on the same plan. The sheathing base of monocot leaves is therefore a reflection of cotyledonary structure, rather than the reverse.

Regardless of the morphological nature of the single cotyledon, throughout the Liliopsida it is basically the same organ. Highly modified though the cotyledon may be in such plants as grasses, there is no need to assume that the monocotyledonous condition of the Liliopsida is of more than one origin. A more ample discussion of the nature and possible evolutionary history of the cotyledon in Liliopsida is provided by A. Arber (1925) and A. J. Eames (1961).

The unique septal nectary of many Liliopsida helps to unify the class and also to strengthen the concept that the Alismatidae are near basal. The structure is apparently unknown in the Magnoliopsida. According to W. H. Brown (1938), septal nectaries "occur in the septa between two carpels and represent places where the adjacent walls of the carpels have not fused. They discharge nectar to the outside by means of small openings. They are such complicated structures that they would seem to indicate a relationship between all plants having them."

Septal nectaries are characteristic of those Arecaceae that are nectariferous, and of the Liliales, Bromeliales, and Zingiberales. Not every genus in every family of these orders has septal nectaries, but they are common enough so that their absence is exceptional rather than typical. The Smilacaceae and the tribe Tulipeae of the Liliaceae are among the more notable exceptions. The complex, external nectaries of some of the Zingiberales are evidently derived from septal nectaries (V. S. Rao 1970). Some of the Orchidaceae also have modified septal nectaries. It will be noted that septal nectaries occur in three of the four subclasses of Liliopsida that are typically syncarpous. Most of the Commelinidae lack nectaries entirely.

The antecedents of septal nectaries probably lie in the mostly apocarpous subclass Alismatidae. As W. H. Brown (1938) has pointed out, Sagittaria and other Alismatidae have nectaries between the petals and staminodes, and between and around the staminodes and lower carpels. Alisma, with a single whorl of separate carpels, has a nectary at the base of the slit between any two adjacent carpels. The palms, which range from apocarpous to syncarpous, have correspondingly alismatoid to septal nectaries. Presumably, a similar change occurred in the line(s) leading to the Zingiberidae and Liliidae.

The Subclasses of Monocots

None of the subclasses of monocots can be considered ancestral to any of the others (fig. 14.8). The Alismatidae, Commelinidae, Zingiberidae, and Liliidae are fairly well characterized, but the Arecidae are a more loosely knit affinity group.

Alistmatidae

The Alismatidae have often been considered to be the most archaic group of Liliopsida. They can scarcely be on the main line of evolution of the class, however, because a primitive monocot should have binucleate pollen and endospermous seeds. The anomalous small order Triuridales does have endospermous seeds, but the mycotrophic, nongreen habit of this group sets it well apart from any possible mainstream of monocot evolution.

The Alismatidae are here considered to be a near-basal side branch of the monocots, a relictual group that has retained a number of primitive characteristics. The apocarpous gynoecium of most members of the Alismatidae, combined with the mainly 1-aperturate pollen of the Liliopsida as a whole, indicates that any connection of the Liliopsida to the Magnoliopsida must be to the archaic subclass Magnoliidae. It should be noted, however, that the ontogeny of the pleiomerous androecium in the Alismatidae is quite different from that in the Magnoliidae, so that the evolutionary homology can be questioned. The putative relationships among the orders of Alismatidae are given in figure 14.9.

Arecidae

Our two orders of the subclass Arecidae, the Arecales (palms) and Arales (aroids), are only loosely linked and are best discussed separately (fig. 14.10). The palms are the only monocotyledons to combine an arborescent habit, a broad leaf blade, and a well-developed vascular system that has vessels in all vegetative organs. This obviously functional syndrome approaches that of woody dicotyledons, but palms lack an adequate means of secondary growth and a means of expanding the coverage of the crown. Furthermore, palms have never developed the deciduous habit, and with minor exceptions they have not adapted to temperate or cold climates. Thus, their ecological amplitude is limited, as compared with that of woody dicots. They do well in tropical regions that are moist enough to support evergreen tree growth but not moist enough to support a dense forest, and they are also common components of the understory in tropical rainforests.

The Araceae form the principal family of Arales, with the Lemnaceae and Acoraceae as much smaller appendages. The leaves are usually more or less expanded and tend to be net-veined, but they lack the ontogenetic peculiarity previously noted for the palms. Most aroids are herbs of the forest floor or are vines climbing on forest trees. Pistia, a free-floating aquatic aroid, has a relatively small spadix. It is seen as pointing the way to the free-floating, thalloid Lemnaceae, which reproduce mainly vegetatively but occasionally produce very small, few-flowered inflorescences. Acorus, with ensiform, unifacial leaves, has usually been included in the Araceae, but it also differs from typical Araceae in a number of technical features, including the presence of ethereal oil cells and absence of raphides. Its treatment as a separate family Acoraceae now seems well justified (M. H. Grayum 1987).

Commelinidae

The Commelinidae are a well-marked group characterized by progressive floral reduction, absence of septal nectaries, and eventually the abandonment of insect pollination in favor of anemophily. The Commelinales, standing at the base of the subclass (fig. 14.11), have entomophilous flowers of normal appearance but without nectar or nectaries. The other orders have small flowers of more or less reduced structure, but Eriocaulon (Eriocaulales) has reverted to entomophily and has nectariferous glands just within the tip of the tiny petals. The Typhales (Typhaceae and the closely related Sparganiaceae) have been variously disposed in different schemes. They are here included in the Commelinidae because of their floral reduction associated with anemophily, their paracytic stomates, the presence of vessels in all vegetative organs, their mealy, starch-bearing endosperm, and the bound ferulic acid in their cell walls. All of these features are perfectly fine for the Commelinidae, but collectively they are difficult to reconcile with any other subclass.

Although the Cyperaceae and Poaceae have traditionally been associated in older systems of classification, in recent years several authors have taken each family to represent a separate, unifamilial order, or have associated the Cyperaceae with the Juncaceae and the Poaceae with the Restionales. Neither of these alternatives is really wrong, but I am not convinced that the change is yet necessary. When the taxonomic distribution of diffuse centromeres in this group of families is better known, it may be time to reconsider the ordinal alignments.

Zingiberidae

The two sharply defined orders (Bromeliales and Zingiberales) that collectively make up the subclass Zingiberidae have not usually been closely associated in past systems. Nevertheless, they would be discordant elements in any other subclass, and they have certain features in common that collectively distinguish them from other subclasses. They resemble the Liliidae (and differ from the Commelinidae) in commonly having septal nectaries and in usually having the vessels confined to the roots.

They resemble Commelinidae (and differ from most Liliidae) in their starchy endosperm with compound starch grains, and they further differ from typical Liliidae (and resemble the order Commelinales) in having the sepals well differentiated from the petals, often green and herbaceous in texture. They differ from both the Liliidae and the Commelinidae in that the number of subsidiary cells around the stomates is usually four or more.

Liliidae

The Liliidae characteristically (although not always) have showy flowers, with the tepals all petaloid, and they have intensively exploited insect pollination. With few exceptions, they have not taken the path of floral reduction, and none of them has a spadix. Although some few Liliidae are arborescent, some have broad, net-veined leaves, and some have vessels throughout the shoot as well as in the root, none of them has coordinated these features into a working system that could present a competitive challenge to woody dicotyledons.

I take the narrow, parallel-veined leaf to be primitive within the Liliales. Broader, more or less net-veined types have evolved repeatedly but still show traces of the ancestral condition. Often they have several main veins that arch out from the base and converge toward the tip. It is not difficult to envisage the evolution of this type of venation by increase in the width of the blade, without any increase in the number of main veins. Concomitantly, the connecting cross-veins are elaborated to vascularize the expanded space between the main veins.

I take the primitive type of endosperm in the Liliales to be fleshy or cartilaginous, with food reserves of protein, oil, and possibly some starch, but not hemicellulose. This type is readily compatible with the endosperm of archaic dicotyledons. Both strongly starchy endosperm and very hard endosperm with reserves of hemicellulose in the thickened cell walls appear to be advanced features.

Although the Dioscoreaceae and the geographically extralimital family Taccaceae do not appear to be primitive in other respects, they have a more primitive embryo than other families of the Liliales. The embryo is slipper-shaped or obliquely ovate, the cotyledon is evidently lateral, and the plumule is more or less distinctly terminal. In the other families, the embryo is mostly barrel-shaped or ellipsoid to ovoid or cylindric, with a terminal cotyledon and a tiny, lateral, often scarcely distinguishable plumule that may be sunken into a small pocket.

Despite the embryo, the Dioscoreaceae are a poor candidate for a position near the base of the monocotyledons, as recently proposed by a few authors. They have a compound, inferior ovary with septal nectaries, and a horny endosperm with hemicellulose deposited in the thickened cell walls. An extraordinary reversal of current views on polarity would be required to consider these as plesiomorphic features in monocotyledons in general. Furthermore, if the views here presented about evolution of leaf form and venation in the monocots are correct, the broad, more or less net-veined leaves of the Dioscoreaceae must remove them even farther from the ancestry of the class.

The delimitation of the family Liliaceae is much disputed. In the traditional Englerian system most of the lilioid monocots with a superior ovary were put into the Liliaceae, and those with an inferior ovary into the Amaryllidaceae. All hands now agree, however, that a cleavage on this basis is unnatural, because the genera with inferior ovaries properly belong in several groups that relate separately to the ancestral group with a superior ovary.

There is enough diversity within the traditional Liliaceae plus Amaryllidaceae to provide for several families, but the problem is how to go about the dismemberment. The separation of Smilax and its allies as a distinct family (or cluster of smaller families) now seems to be generally accepted. Most authors now also accept the Agavaceae, but they are not agreed as to its delimitation. If the Agavaceae are defined cytologically (x = 5 large and 25 small chromosomes), then they include about 8 genera of coarse plants with firm, perennial leaves, plus the habitally very different genus Hosta. If, on the other hand, the Agavaceae are defined largely on the basis of habit and chemistry, then the number of genera goes up to about 18, and the chromosome number is more variable but still the karotype comprises a few large chromosomes and many small ones. A consistent treatment then requires the recognition of the mainly African family Aloeaceae as another line that has independently undergone the same sort of habital changes as the Agavaceae.

I have chosen to recognize the Agavaceae, Aloeaceae, and Smilacaceae as distinct families, and to relegate the remainder of the traditional Liliaceae plus Amaryllidaceae to a single rather amorphous family Liliaceae. I would be happy enough to divide this group into several families, if I could find a reasonable way to do it, but I have not found the way. R. M. T. Dahlgren et al. (1985) distributed the Liliaceae sensu meo into 27 families in 4 orders (Melanthiales, part of the Liliales, part of the Dioscoreales, and part of the Asparagales) with such minimal differences that I have not been able to comprehend the essential nature of the groups. We still await a comprehensive reorganization of the lilies into several families more comparable to other recognized families of angiosperms.

The Orchidales differ from the Liliales essentially in their strongly mycotrophic habit, and in their very numerous, tiny seeds with a minute, mostly undifferentiated embryo and no endosperm. The reduction of the embryo is at least in part a consequence of mycotrophy; these two features are also associated in some other groups of angiosperms. The ovary of the Orchidales is always inferior and only seldom has typical septal nectaries, although the ovarian nectaries of some Burmanniaceae and Orchidaceae may well be derived from septal nectaries.

The combination of mycotrophy and numerous tiny seeds offers certain evolutionary opportunities as well as imposing some limitations. The plants are physiologically dependent on their fungal symbionts, sometimes for food, sometimes only for other factors as yet not fully understood, but in any case they can grow only where their fungal symbiont finds suitable conditions. The dustlike seeds of the Orchidaceae are admirably adapted to being carried by the wind and lodging in the bark of trees, and many orchids are epiphytes. The production of many ovules is, of course, of no value if they are not fertilized. One way to increase the likelihood of fertilization is to offer special attractants to a limited set of pollinators and to have the pollen grains stick together in masses so that many are transported at once. This strategy puts the plants in thrall to their pollinators, but it opens the door to explosive speciation.

Only the Orchidaceae have efficiently exploited the evolutionary possibilities of the order. The floral characteristics that distinguish the Orchidaceae from their immediate allies clearly reflect progressive specialization for the massive transfer of pollen by specific pollinators.

The Orchidales are evidently derived from the Liliales as here defined. All the characteristics in which the Orchidales differ from the Liliales are apomorphies. Within the Liliales, only the epigynous segment of the Liliaceae has the characteristics from which the Orchidaceae, Burmanniaceae, and Corsiaceae (extraterritorial to our flora) might well have arisen. Although the Orchidaceae never have more than three stamens, two of these are considered to come from the ancestral inner cycle, and one from the outer. Thus a hexandrous ancestry seems likely.

Epilogue

Nothing can be more certain than that future studies will lead to changes in the system here discussed. I can hope that considerable segments of it are now stable, but only time will tell. Just as we stand on the shoulders of our predecessors, future taxonomists will stand on ours.