Roots. Although roots have not been used traditionally as taxonomic characters, they display a wide variety of qualitative characters. Typically, roots develop one or more at each node. In section Polyphyllium Engl., the roots are numerous, short, and slender and arc borne along the internodes. In most species of section Pachyneurium Scroll, and in other short-stemmed species as well, the roots per node are much more numerous. Since the internodes are very short, the root mass can be large and dense, with the roots contiguous or nearly so. In many cases, there has been a divergence in the function of roots so that some are utilized exclusively for support while others are negatively geotropic and are directed upward where they collect debris, or they extend into the rosette of leaves and collect debris. In other cases, the support roots of the plant are negatively geotropic, such as in the case of pendent species (Madison, 1978). Commonly, roots appear smooth when fresh; when they are dry it can be seen that they actually consist of a dense mass of woolly trichomes that are so closely matted they look contiguous before shrinkage of the root. Anthurium gracile (Rudge) Lindl. has roots with a thick white layer of velamen useful in absorption of moisture from the air. Other species that have been investigated lack this velamen layer.

Leaves. Among the more useful taxonomic characters that have been overlooked previously is the cross-sectional shape of petioles (Fig. 1). Most Anthurium species have^ petioles that are terete or subterete in cross-section and frequently also narrowly and obtusely channelled. However, considerable variation exists, and at least sometimes this is believed to be taxonomically significant. Many Mexican species have petioles that are D-shaped or broadly sulcate in cross-section. Petiole cross-sectional shape, within limits, has proven useful in the separation of species of Anthurium in the section Pachyneu-rium Schott, where leaf blades of many of the bird's-nest species are otherwise almost identical. However, petiole shape cannot be relied on entirely. Some of the more variable, wide-ranging species, such as A. schlechtendalii Kunth, have considerable variation in petiole cross-sectional shape even within a single population (see that species for a discussion of this). Some species of Anthurium have sharply triangular petioles, others trapezoidal, square, or terete with several to many sharp ridges. The degree of channelling of the petiole also provides useful taxonomic characters. The length of petioles, like overall size of any morphological part of Anthurium, is highly unreliable because plants often flower when quite young and thus the overall size of leaves, inflorescences, and other parts vary remarkably in size over the course of the life of the plant. More useful is the relative length of the petiole in comparison to the length of the inflorescence. The most widely used taxonomic characters in Anthurium arc the shape and venation of the leaves, since these are perhaps more diverse than in any other genus of plants (Croat & Bunting, 1979). In addition, glandular punctations on the leaf blades provide useful characters and are valid, at least sometimes, at the sectional level. For example, sections Porphyrochitonium Schott, Tetraspermium Scholt, and Digitinervium So-diro consistently have glandular punctations on one or both surfaces. Still, some species in otherwise cpunctale sections, such as section Cal-omystrium Schott, have evolved glandular punctations independently. While many members of section Calomyslrium Schott have brownish spots resembling the glandular punctations of section Porphyrochitonium Schott and others, at least one unknown isolated Brazilian species (Rosa & Lira 2325) has definite glandular punctations on the lower blade surface. A group of coriaceous, cordate species from Mexico (A. verapazcnse al-liance) has also developed glandular punctations.Other useful features in the leaf include the presence of papillate epidermal cells, such as in section Cardiolonchium Schott, and the presence of easily visible raphide cells so prevalent in section Calomystrium Schott.

Inflorescences. The inflorescence, although generally critical for identification of species of Anthurium, has been under utilized, perhaps befigure 1. Crosssectional petiole shapes in Anthurium.-- A. Basically terete: ranging from esulcatc (1), to shallowly and acutely sulcate (2 & 3), to obtusely and broadly sulcate (4), to narrowly and obtusely sulcate (5), to narrowly and acutely sulcate (6), broadly and acutely sulcate (7), shallowly and broadly sulcate (8), flat adaxially (9), flat adaxially with marginal ribs (10), to flat adaxially with marginal and medial ribs (11).--B. D-Shapcd or broader than thick: ranging from flat adaxially with obtuse margins (1), to broadly sulcate with obtuse margins (2), broadly sulcate with acute margins (3), flat adaxially with acute margins (4), flat adaxially with erect margins (5), flat adaxially with erect margins and with 1 or more medial ribs (6 & 7).--C. U-shapcd or thicker than broad: ranging from broadly and obtusely sulcate (1) to narrowly and obtusely sulcate (2), narrowly and acutely sulcate with acute margins* (3), shallowly and acutely sulcate* (4), sulcate with acute margins (5), sulcate with sharp margins and a medial rib* (6).--D. Markedly angular: ranging from quadrangular with acute angles (1) to quadrangular with obtuse angles (2), trape/.oidal (3), obtusely triangular (4), acutely triangular (5), acutely triangular with two marginal ribs (6), basically triangular with two abaxial ribs (7), obtusely triangular, narrowly and sharply sulcate with convex sides (8).--E. Markedly ribbed abaxially: ranging from trape/.oidal or quadrangular, sharply and broadly sulcate adaxially, 3-ribbcd abaxially (1), to obtusely and broadly sulcate adaxially, 3-ribbed abaxially (2), broadly and obtusely sulcate adaxially 5-ribbcd abaxially (3), broadly and sharply sulcate adaxially, narrowly and sharply ribbed around the remaining circumference (4), or with one or more ribs on the sulcus (5).

Not yet observed but to be expected. cause of too much reliance on herbarium material, which greatly diminished the value of the inflorescence. Peduncle length has frequently been used as a taxonomic character, yet length alone is extremely variable on plants of different ages and even in populations of plants of the same age. However, the ratio of the peduncle length to the petiole length is usually relatively constant and often provides a useful character. Like most vegetative parts, the overall dimensions of the spathe and spadix are not very useful, but again the ratio of the length of the spathe to the length of the spadix is relatively reliable. The shape of both the spathe and spadix and the texture of the spathe are usually much more reliable than size.

The spadix and the many characters it embodies is perhaps one of the least used taxonomically. This is no doubt due to the fact that too few species have been critically compared in live condition. The dried spadix is difficult to study and loses many of the characters that are apparent before drying. One of the most useful characters exhibited by the spadix is color. Spadix color when seen at anthesis is usually stable, and species vary considerably in the colors exhibited. In some species, however, the spadix varies from green to violet-purple. While many spadices are green before the spathe opens, most quickly assume the color they will have at anthesis. Some species may be green when stamens begin to emerge and later may turn violet-purple. Some species change colors after anthesis, so it is important to make note of the color at anthesis. Anthurium huixtlense Matuda and A.formosum Schott, for example, have spadices that become noticeably darker after anthesis. Other characters exhibited by the spadix include scent, size, shape, and texture of the flowers; the number of flower spirals and the number of flowers per spiral; the disposition of the stigmas and timing of the stigmatic droplets; the rate of emergence and disposition of the stamens, as well as the shape of anthers and pollen color. Many of these characters are discussed in detail in Croat (1980), but a summary of the important points will be included here as well.

Although there has been no success as yet in determining scents of flowering Anthurium, there is little doubt that a large percentage of them have distinctive aromas. Those with attractive bright spadices have sweet aromas, such as of mint or lilac, and are no doubt bee-pollinated. Euglossine bees have been seen in several cases visiting Anthurium (Croat, 1980). Species with purple or drab colored spadices generally have yeasty or rotting fruit aromas (the scent of one species is like that of fresh cucumbers) and are probably fly-pollinated. Scents are usually not evident at night, and some species are phenologically active, presenting their aromas only during midday. Aromas of some species are sometimes so intense as to be smelled easily from a distance of several meters, but most species give only faint but distinct aromas.

Flowers are closely aggregated in spirals and are quite variable in size and shape. Their overall shape at the apex may be square, rhombic, or 4-lobed. The tepals themselves may be 3, 4, or 5-sided. The texture of the surface, presence of droplets, raphide cells, and other features, all may be taxonomically important.

Stigmas may be slitlike or brushlike and exserted. The production of stigmatic droplets varies considerably from species to species, with both timing and duration being important. The sequence and duration of stamens as well as their method of presentation is similarly important. While most anthers are merely exserted just above the tepals, the stamens in some species are exserted and later the filament shrinks and retracts the anther to the surface of the tepals. Stamens in other species do not retract and the filament and anthers remain exposed. Stamens in still others retract completely below the surface of the tepals and thus out of sight. The sequence ofstaminal emergence is also quite variable (Croat, 1980), with the rate of emergence being an additional variable. Most species with typically long-lapered spadices have emergence beginning at the base of the spadix and progressing toward the apex. Species with cylindroid or clavate spadices have staminal emergence generally scattered or even beginning at the apex or middle. In all types of spadices the rate of emergence may be quite different. For example some species develop all of their lateral stamens before the third (alternate) stamen and finally the fourth (alternate) stamen emerge. Still other species have all stamens of the flowers in any spiral fully exserted before many flowers in succeeding spirals have their lateral stamens exserted. In other words, the leading spirals in the staminal development progression are few. In contrast, other species may have, for example, ten or more of the leading spirals with only lateral stamens exposed, with another ten or more of the preceding spirals bearing three stamens (i.e., the laterals plus the third, usually anterior, stamen) and finally all ofthe lower (basal) spirals with a full complement of four stamens exposed.

All of the characters mentioned here have proven to be useful taxonomic characters, yet most have never been mentioned previously by other workers.

Fruits. Anthurium fruits consist of 2-loculed berries. The berries provide a number of useful characters including size, shape, and color, the nature of the pericarp and mesocarp, the presence or absence of raphide cells, as well as the number, size, and shape of seeds. Engler (1905) appeared to realize the significance of the berries in classification, but unfortunately he had not seen berries on the majority of the species he studied. Because Anthurium has markedly protogynous flowers (Croat, 1980), they do not readily set fruit on greenhouse collections without manipulation. Often this is not possible without a second individual of the species. A great deal of attention has been given to those species in the living collections for which fruits were not known, in order to encourage them to set fruit.

The number of seeds per locule has long been considered an important character, with such groups as sections Tetraspermium Schott, Gymnopodium Engl., and Porphyrochitonium Schott having more than one seed per locule, while the remaining sections usually have one per locule. Seeds themselves are as individually different from species to species as are the berries. Most have sticky, mucilaginous appendages firmly associated with them (Croat, 1980) often attached to one or both ends. In some cases the seed is completely enclosed in a mucilaginous envelope. These extremely sticky structures seem ideally suited to being attached to bird beaks, and Anthurium berries seem ideally suited for bird dispersal. Many of the Mexican species of Anthurium lack a sticky appendage on the seeds.