Aliso: A Journal of Systematic and Floristic Botany Aliso: A Journal of Systematic and Floristic Botany

The genus Chilopsis contains one species, C. linearis, with two subspecies: subsp. linearis of the Chihuahuan Desert characterized by erect, straight leaves, and subsp. arcuata of the Sonoran and Mojave deserts with longer, arcuate leaves and other distinguishing characteristics. Populations of subsp. linearis east of the Sierra Madre Oriental in eastern Mexico have woolly stems and are described as a new variety, tomenticaulis. Relationships between Chilopsis and Catalpa are discussed and data are presented supporting continued recognition of Chilopsis as a distinct genus.


INTRODUCTION
The genus Chi/apsis was proposed by D. Don (1823) based on a Mexican collection from Pavon deposited in the Lambert Herbarium. Dan's genus contained one species, C. saligna; however, the species was previously published as "Bignonia? linearis" by Cavanilles (1795). The combination Chi/apsis linearis (D. Don) Sweet was first made in Sweet's Hortus Britannicus in 1827. In 1848 Engelmann noted that Wislizenus's specimens from Chihuahua, Mexico, had narrow, glutinous leaves and branches, and, although not sure of the characteristics of C. linearis, he suggested that these may represent a new species Chi/apsis glutinosa Engelm.
The first taxonomic treatment of Chi/apsis by Fosberg ( 19 36) accepted Chi/apsis as monotypic, consisting of three varieties: an eastern C. linearis var. linearis with "woolly" branchlets and straight leaves with prominent venation; a western C. linearis var. arcuata Fosberg with glabrous branches and distinctly arcuate leafblades usually without prominent venation; and a more centrally located C. linearis var. glutinosa (Engelm.) Fosberg with branches and leaves glutinous and leaf-blades either erect or arcuate.
Fosberg's 1936 work was based largely on field studies and specimens at UC. His study was stimulated by the discovery of plants with strongly glutinous stems from south-central New Mexico. He correctly separated specimens with straight leaves from central Mexico and Texas from plants more to the west having arcuate leaves. But in recognizing his glutinous-stemmed taxon he was forced to include in that taxon material with both straight and curved leaves, which resulted in a series of rather ambiguously defined taxa.
Since its publication Fosberg's treatment has been variously followed in major floras. His taxa are referred to in Kearney and Peebles (1951), Correll and Johnston (1970), and Martin and Hutchins (1981), are questioned in Benson and Darrow (1954), but are accepted in Benson and Darrow (1981). Munz and Keck (1959) and Munz (1974) refer to his var. arcuata as a form while Abrams and Ferris (1960), Shreve and Wiggins (1964), Wiggins (1980), and Cronquist, Holmgren, Holmgren, Reveal and Holmgren (1984) have all placed his varieties into synonymy.
This reevaluation of Chilopsis is an outgrowth of the preparation of the treatment ofBignoniaceae for the Chihuahuan Desert flora. In addition to field studies through the southwestern United States and Mexico, specimens were examined or borrowed from A, ARIZ, ASU, CSLA, ENCB, GH, JEPS, LL, MEXU, MO, NMC, NY, POM, RSA, SRSC, TEX, UC, UNM, UNLV, and US, photographs of authentic specimens were provided by MA.
In addition to a systematic revision of Chilopsis, this paper also addresses questions regarding the relationships to the genus Catalpa.

Relationships ofChilopsis and Catalpa
Chilopsis is most closely related to the genus Catalpa Scop. Within the Bignoniaceae Chilopsis and Catalpa are placed in the largely pantropical tribe Tecomeae Endl., characterized by a tree-shrub growth habit, 2-loculed ovaries, and fruits that dehisce perpendicular to the septum. Within the Tecomeae they have been placed by Bentham and Hooker (1876), along with Sparattosperma Mart. in DC., in subtribe Trichosperminae, characterized by seeds having separate, long hairs extending from each end. The Brazilian Sparattosperma, however, has palmately compound leaves and differs in many other features, and Gentry (pers. comm.) rejects the notion of relationship with Sparattosperma and considers that Chilopsis and Catalpa represent a distinct and separate lineage within the Tecomeae of unknown affinities (Gentry 1980).
Catalpa is sometimes allied with Paulownia Sieb. & Zucc. (e.g., Airy Shaw 1966), but Armstrong (1985) has shown that Paulownia has ovary characteristics typical of Scrophulariaceae, while Catalpa clearly shows character states of Bignoniaceae.
Chilopsis and Catalpa share many characteristics. They are both small to large trees. They are nearly identical in fruit, seed, embryo, style, and anther characteristics. They have similar thickened, obovoid calyces that open irregularly into upper and lower lips. Their corollas are similar in texture, coloration, and lobing though those of Catalpa are often larger with more ampliate throats and vary from white to yellowish in background coloration. They also share a unique pollen type in the family (Fig. 4d) consisting of tetrads with sculpturing limited to coarsely reticulate areoles (Gentry and Tomb 1979), and both have 20 pairs of chromosomes (Goldblatt and Gentry 1979) though that number occurs throughout the family. In addition sterile hybrids have been formed between Chilopsis linearis and Catalpa cf. bignonioides (Rusanov 1964).
Most North American botanists know Catalpa from our two native southeastern species: C. bignonioides Walt. and C. speciosa Warder. Catalpa, however, consists of about 11 species placed in two sections (Paclt 1952): section Catalpa with six species showing an Asian-eastern North America disjunction (Li 1952) with two species native to southeastern North America and four to temperate China; and section Macrocatalpa Grise bach with five species native to the West Indies. Species of Catalpa share many basic flower, fruit, seed, and growth habit characteristics. Flowers are borne in terminal racemes or panicles, and are distinguished from most other Bignoniaceae in having only two fertile stamens and three staminodia. The five-lobed corollas have ampliate throats with a pair of basal ridges and are variously spotted or striped with contrasting color patterns. Leaves are usually opposite to verticellate, petiolate, and dorsi ventral in structure with bundle sheath extensions continuing to the minor veins. Leaf blades bear clusters of distinctive capitate glands sunken in pockets in the axils of midrib and secondary (and sometimes tertiary) veins on the lower surface that serve as extrafloral nectaries attracting ants, which protect the leaves from various sucking insects (Elias and Newcombe 1979). Species of section Catalpa are winter deciduous, distinguished by having large, rather soft, ovate to broadly ovate, entire to variously coarsely lobed, long-petiolate leaves, and, as in Chilopsis, the seed body bears hairs only at the ends of the seeds. Species of the West Indian section Macrocatalpa, in contrast, are semideciduous, have smaller, coriaceous to chartaceous, ellipticlanceolate to ovate, entire to denticulate, short-petiolate leaves, and the bicomose seeds are also hairy on their dorsal surfaces. They also have fewer nectariferous glands on their leaves, and tend to have more delicate pedicels, etc.
Chilopsis linearis differs fundamentally from Catalpa in having flowers with four stamens and one staminode (Fig. 3f)-a plesiomorphy occurring in all but two genera in the Bignoniaceae. Other differences appear associated with the adaptation of Chilopsis to semiarid habitats in southwestern United States and Mexico where it occurs as a phreatophyte mostly along dry streambeds. Chilopsis linearis is a small tree. Its leaves are linear to linear-lanceolate, attenuate, isolateral (with stomata and palisade cells on both surfaces- Fig. 4e-f) and they lack the concentrations of nectariferous glands in depressions in the vein axils on the abaxial surface (as leaves are isolateral, they lack a homologous abaxial leaf surface). They have reduced racemose to racemose-paniculate inflorescences and their corollas have narrower throats, etc. .
The relationship of Chilopsis with Catalpa is not evident. Is Chilopsis just a xeromorphic derivative of Catalpa section Catalpa, or is it a sister group of the genus Catalpa? To test the relationships of Chilopsis with Catalpa, character states were recorded for Chilopsis and Catalpa using data from Paclt's (1952) synopsis of Catalpa and other descriptions. As no reliable outgroup is known (Gentry, pers. comm.) it was not possible to polarize the data using outgroup comparison so three separate data sets were developed considering the character states of section Catalpa, section Macrocatalpa, or Chilopsis as plesiomorphic. The data were analyzed with the WAGNER option of PHYSYS on the California State University central Cyber computer. The data sets included 11 to 10 OTU's and 13 to 11 characters-the reduced sets eliminated closely related taxa and those character states whose polarity was more difficult to establish. Without evidence as to the correct ancestral group we end up with an uprooted tree of three convex groups sensu Duncan (1980;Meacham 1980) as shown in Figure 1a. However, it may be possible to establish a root within this tree. Only one character used in this data set can be polarized with some degree of confidence on the basis of external evidence-that of four stamens vs. two stamens. As noted above, only two genera in the Bignoniaceae exhibit the two stamen-three staminodia character state-Catalpa in the Tecomeae and the unrelated genus Pseudocatalpa in the Bignoneae (A. Gentry, pers. comm.). I believe this strongly implies that the twostamened characteristic is apomorphic and the four-stamened character state found in Chi/apsis is plesiomorphic. This single character can be used to root the tree between Chi/apsis (four stamens) and Catalpa (two stamens) which makes Chilopsis a sister group to the lineage of Catalpa and Macrocatalpa (Fig. 1 b). While it could be argued that it is rather risky to root a tree on the basis of a single character it is, in my opinion, the most concordant option available. To root the tree with either section Catalpa or section Macrocatalpa would make Chi/apsis part of the sister group with the alternate section, which would either demand that the two-stamen characteristic evolved independently in the two sections of Catalpa or that the four-stamened character of Chi/apsis would represent a reversal from a two-stamened character state (as in Catalpa) to a four-stamened state and would thus be homoplastic.
It should be noted that while Catalpa is characterized by having two stamens and three staminodia, occasional plants of our native southeastern Catalpas have four fertile stamens and a single staminode (Fernald 1950;Gleason 1952)-apparently representing an atavism. However, in these flowers, one anther sac of the proximal anther pair is usually not completely developed. This raises a question whether or not the four stamens of Chi/apsis could represent a fixed atavistic character state? This question is basically unresolvable with the data at hand.
The resulting generic treatment allows for retention of the current classification recognizing Chi/apsis as a distinct genus forming a sister group to Catalpa with its two sections Catalpa and Macrocatalpa.

Classification of Taxa within Chilopsis
Field studies and analysis of the herbarium material show that three well-defined taxa exist over the range of the species. Plants from the Sonoran and Mojavean deserts in Baja California, Sonora, California, Arizona, and extreme northwestern portion of the Chihuahuan Desert in the western half of New Mexico (Fig. 6) have arcuate, long leaves, light-colored corollas, glabrous to sparsely pilose, glandular or somewhat glutinous stems, and a tendency to have widely spreading to decurved branches ( Fig. 2c-d, 3b). This taxon, named C. linearis var. arcuata by Fosberg, is easily recognized by its curved leaf blades alone. In contrast the type taxon from the Chihuahuan Desert region in central Mexico, trans-Pecos Texas, and the eastern half ofNew Mexico (Fig. 6) is characterized by straight, moderately short, erect-ascending leaves, more strongly colored corollas, and a tendency to have erect-ascending branches ( Fig. 2a-b, 3a). Through most of this range their stems are strongly to weakly hirtellous to nearly glabrous and variously glandular to glutinous. This constitutes C. linearis var. linearis of Fosberg in part. Populations east of the Sierra Madre Oriental in Tamaulipas and Nuevo Leon (Fig.  6), however, represent a third, heretofore undescribed taxon that has slightly larger flowers and distinctly woolly stems-taxon "tomenticaulis." The rationale for establishing this classification is explained in detail in the systematic section of this paper.
To reflect this pattern, taxa arcuata and linearis are here recognized at the rank of subspecies while "tomenticaulis," which differs from taxon linearis mainly in stem vestiture, is considered a variety of linearis. The subspecific rank in this classification is used to allow taxon tomenticaulis to be relegated to a subordinate (varietal) rank thus providing for a hierarchial classification. From the Greek Cheilos, labium, and opsis, resemblance, and thus pronounced Kilopsis (fide Shinners 1961). With one species. A much amplified species description is here presented in lieu of a separate discussion of the characteristics of Chilopsis.
The species is here divided into three subordinate taxa, two subspecies, one with two varieties as indicated in the following key.
Chi/apsis linearis subsp. linearis is easily distinguished from subsp. arcuata by its straight, erect-ascending leaves (Fig. 2b, 3a). Several other quantitative trends are also apparent in subsp. linearis: (1) young stems tend to be more erect as are major branches giving at least young plants more erect growth habits (Fig. 2a); (2) leaves tend to be shorter, mostly 4-11 em in length, and may be very narrow (2.8-4 mm wide) or broader (5-11 mm wide) within a population; (3) inflorescences tend to have longer, less noticeably glandular peduncles and pedicels; (4) corollas tend to be larger and are accompanied with longer styles, filaments, staminodia, etc. , and though variable in color even within a population they tend to be more strongly colored often having rather vivid maroon-red outer corolla tubes and more extensive coloration on the inner corolla lobe surfaces (Fig. 4b); and (5) fruit are usually noticeably thicker than those of subsp. arcuata.
Chilopsis linearis subsp. linearis var. linearis can be distinguished from subsp. arcuata by its straight, erect-ascending, noncurved leaves and from var. tomenticaulis by its glabrous to moderately hirtellous-pilose young stems with scattered hairs only 0.05-0.4 mm long. Fosberg (1936) recognized two varieties within this taxon: C. linearis var. "originaria" was distinguished in his key as having "sterile branches somewhat woolly, veins in leaves prominent"; and C. linearis var. glutinosa was distinguished by having "sterile branchlets and leaves glabrous, glutinous, the veins in leaves not usually prominent." This he contrasted with nonglutinous stems and young leaves of his variety arcuata. I have been confused by Fosberg's use of the term woolly to describe stem vestiture in this taxon as young stems tend to be only hirtellous to pilose or they may lack nonglandular hairs altogether ( Fig. Sab), though inflorescences do tend to be densely villous to woolly as do the young stems ofvar. tomenticaulis. I have also been confused by his term glutinous (sticky) as I have found specimens throughout the range of Chilopsis (except in var. tomenticaulis where the woolly vestiture obscures this characteristic on stems) to have some degree of stickiness expressed at least in young stems, e.g., all stick to the newsprint when removed from the press. With regards to vestiture, stems and leaves of all taxa of Chilopsis have two types of trichomes. There are series of scattered, subsessile glands with conspicuous, flattened heads (0.3-)0.4-0.6( -1.0) mm in diameter that usually consist of a single series of 16 vertically-oriented cells (Fig. 4g). Occasional glands, however, are larger and have 30-60 vertical cells (Fig. 4h). These glands contain a viscid substance and they may eventually rupture along their outer margins exuding these contents onto stem internodes and leaf surfaces causing the structures to become variously glutinous (Fig. 5e-f). The exudate, however, gradually dries and becomes less viscid over time. If the glands are still intact, they can easily be seen on stems and leaf surfaces. After the glands rupture on leaf surfaces the glutinous contents remain in the slightly recessed pockets that contained the glands and appear as series of glandular punctae. The glands along nodes, on basal portions of leaves, and on axillary buds tend not to break down and thus persist on specimens. Whether these serve as extra-floral nectaries as in Catalpa (Elias and Newcombe 1979) is unknown. These glands occur on plants throughout the range of the species.
Stems and leaves also contain series of erect or upcurved, whitish, uniseriate, multicellular hairs. These are usually only scattered on leaves confined to lower margins and along the mid vein and petioles. Their development on stems is highly variable. In var. linearis plants from much of Texas and along the eastern border of the Chihuahuan Desert (generally from near Muzquiz to Cuatro Cienegas and the Sierra Santa Fe del Pino in Coahuila) tend to have a hirtellous vestiture on young stems with moderately dense, erect or upcurved, uniseriate trichomes 0.05-0.3(-0.4) mm long mixed with glandular trichomes (Fig. 5a). Only rare plants ever develop a stem vestiture that could be termed woolly. Plants around El Paso and the Chisos Mountains and in the central Chihuahuan Desert in Chihuahua, Durango, and adjacent Coahuila, San Luis Potosi, in contrast, tend to have more scattered, shorter or no uniseriate trichomes on young stems (Fig. 5b) and in this characteristic they blend towards the more western subsp. arcuata that has only widely scattered uniseriate trichomes on young stems. These plants also tend to have slightly longer leaves. Engelmann, when describing Wislizenus's collections from northern Mexico (Engelmann 1848) noted that the plants "from New Mexico and Chihuahua, with longer, narrower glutinous leaves, perfectly glabrous, glutinous branchlets, and darker and smaller flowers may be Ch. linearis, DC., or a new species, Ch. glutinosa." While admitting that Engelmann's g/utinosa was rather dubiously published, Fosberg (1936) accepted the taxon at the varietal level mainly as he had collected specimens north ofEl Paso, Texas, with very glutinous stems and leaves. In so circumscribing this taxon on the basis of glutinous stems, Fosberg was forced to accept into his var. glutinosa some glutinous-stemmed specimens with arcuate leaves that would otherwise fit well into his var. arcuata. This caused him to deemphasize the leaf character in his keys.
Interestingly, none of the four Chi/apsis specimens collected by Wislizenus during his journey and seen by Engelmann had entirely glabrous stems, i.e., completely lacking uniseriate hairs-all were sparsely hirtellous with very short hairs. Also the glutinous condition dissipates in herbarium specimens as the exudate dries and at the present time none of Wislizenus's now 136-year-old specimens show any indication of a glutinous nature though they undoubtedly did the year after their collection when observed by Engelmann.
Generally, stem vestiture exhibits too much variation to be useful in classification here. Development of nonglandular hairs shows intrapopulational variation, and the glutinous character is variably expressed during the season. Fosberg (1936) also noted that his varieties could be divided on the basis of  (Fig. 4e). In dried leaves this may be evidenced by mere darkened lines following minor venation or by variations in blade thickness over and between minor venation. The character is often not expressed in young and very narrow leaves.
There are several very good characteristics that can be used to separate subsp. linearis from subsp. arcuata (see above) with the leaf shape (whether straight or arcuate) being the most easily recognized. However, even the leaf shape characteristic is not consistent. Occasional specimens will have mostly erect leaves with a few, scattered, somewhat arcuate leaves. I interpret the western populations of var.linearis-those with reduced young stem vestiture and slightly longer leavesas exhibiting some degree of intergradation with subsp. arcual a.
Variety linearis occurs mostly in the Chihuahuan Desert ranging from southcentral New Mexico where it co-occurs and intergrades with subsp. arcuata, trans-Pecos Texas south through east-central Chihuahua, Coahuila with scattered localities through Durango, in northern Zacatecas, and northern San Luis Potosi (Fig. 6). It is cultivated well outside its range in Texas, Oklahoma, Kansas, Louisiana, Missouri, and California as well as in Europe and the Soviet Union and in many places in west Texas (Randall, Hall, and Lynn counties), Oklahoma (Payne and Jefferson counties) and in Kansas (Comanche and Barber counties) it appears to have naturalized. It grows in sandy and gravelly washes along streams and canyons from areas of Larrea scrub to oak and juniper woodlands, and grasslands occurring with such species as Prosopis glandulosa Torr., Brickellia laciniata Gray, Fallugia paradoxa (D. Don) Endl., Celtis pallida Torr., Acacia berlandieri Benth., A. greggii Gray, Forestiera angustifolia Torr., Juglans microcarpa Berl., Pistacia texana Swingle, Baccharis salicifolia (Ruiz & Pavon) Pers., etc. from 550 to 1650 m elevation. It flowers primarily in the spring from May to July but some flowering may continue until September.
The tomentose stem characteristic is conspicuous and usually present in all plants of a population though occasional plants within a population, e.g., Henrickson 19071 (TEX), or all plants of a gathering (Palmer 390) have glabrous stems and calyces indicating perhaps a simple genetic basis of the character. As in var. linearis occasional plants show a scattering of falcate leaves. The stem vestiture character is not completely restricted to these plants as occasional isolated collections of subsp. linearis such as Pope s.n. (NY) from Texas (exact locality not given) have hairy stems as in subsp. tomenticaulis. It appears that this woolly stem characteristic merely became fixed in the populations east of the Sierra Madre Occidental perhaps when the populations were greatly restricted in extent during the holocene.
Chilo psis linearis subsp. arcuata can be distinguished entirely by leaf orientation and shape. Leaves are usually twisted at the base with leaf-blades oriented with one edge towards the stem with the blades arching outward or downward (Fig.  2d, 3b). As young stems also tend to be more spreading in this taxon, all blades of a stem commonly droop in the same direction (Fig. 3b). Leaves also tend to be much longer than in subsp. linearis, tapering to often elongated, winged, petiolelike bases. As in subsp. linearis leaf-blades are isolateral in structure (Fig. 4f), they initially are covered with sessile glands on both surfaces but with age the glands break down leaving distinct glandular punctae. Young leaves tend to have scattered, upcurved marginal hairs but these seldom persist. Mature leaves also tend to be thinner than in subsp. linearis with bundle sheath extensions forming only along mid and secondary veins (Fig. 4f).
Developing leaves on young shoots will commonly be erect and straight; they become increasingly falcate with age. Occasional specimens with only developing leaves or with very short leaves will have only erect leaves as in subsp. linearis.
As noted in the discussion under subsp. linearis, inflorescences in subsp. arcuata tend to have shorter peduncles-pedicels, the pair seldom exceeding 8 mm in total length, usually with only one flower per peduncle, and corollas tend to be shorter as are filaments, staminodia, styles, etc., though they are structurally identical. Corollas also tend to be less vividly colored with the throat and lobes often being white, light pink, or lavender marked with the characteristic yellow and purplemaroon lines along and between the ridges on the corolla floor. Considerable intra-and interpopulation variation is also observed in corolla color with some plants having nearly white corollas and others having somewhat lavender or strongly-colored corollas with vivid purple-maroon markings on the lower lobes. Orientation of the mid lower corolla lobe is also variable within a population, declining in some individuals, spreading in others.
Young stems in most specimens appear glabrous but are usually initially covered with scattered, sessile glands that usually break down along internodes but persist near nodes, lower petioles, and on axillary buds. Young stems may also be weakly pilose with a scattering of spreading to slightly curved hairs 0.1-0. 5(-0. 7) mm long (Fig. 5d) and in occasional plants young stems are moderately hirtellouspilose as in subsp. linearis.
Inflorescence axes, pedicels-peduncles, bracts-bracteoles, and calyces are always somewhat densely pilose to pilose-villous or occasionally strongly villous with scattered to dense, curved or crinkled, simple, occasionally branched hairs 0.2-0.7(-1.2) mm long. This feature is often accompanied by an understory of sessile or stipitate glands that are particularly common on the pedicels and calyx bases. Occasionally plants have nearly glabrous inflorescences and calyces.
Chilopsis linearis subsp. arcuata ranges from southern California, through Arizona to southern Nevada, southwestern Utah, southeastern New Mexico (where it co-occurs and intergrades with subsp. linearis), south to northern Sonora, and into Baja California (Fig. 6). It is considered a facultative phreatophyte typically occurring in sandy washes in the Sonoran and Mojave deserts in association with such species as Onleya tesota Gray, Acacia greggii, Cercidium microphyllum (Torr.) Rose & Jtn., C.jloridum Benth., Psorothamnus spinosa (Gray) Barneby, Hymenoclea sa/sola T. & G., Prunus fasciculata (Torr.) Gray, Rhus trilobata T. & G., Fallugia paradoxa, Chrysothamnus Nutt. spp., and Brickellia Ell. spp. from near sea level to 5000 ft. It also ranges into Joshua tree, juniper, and oak woodlands, and in Graham County in eastern Arizona it has been recorded from oak-ponderosa pine woodlands up to about 6200 ft elevation [fide Moore et a!. 6471 (ASU)]. Plants become deciduous in November, leaves are produced in late March (to May at higher elevations), and the new shoots flower from late April through May-June (into July at higher elevations) with some flowering continuing into September if conditions permit.
The three taxa appear to represent distinct vicariants-subsp. arcuata in the west, var. linearis in central Mexico, with var. tomenticaulis isolated east of the Sierra Madre Occidental. It would be reasonable to consider that the taxa have been separated and of a more southern distribution at some point during the quaternary, and with reduction in xeric habitats their distributions would be commensurately reduced. This isolation and reduced population size would allow for differentiation and character fixation. In the present post-pluvial period their ranges are apparently expanding northward (the species is winter-deciduous and thus somewhat frost tolerant) and the distribution pattern shown in Figure 6 can be considered to show subsp. arcuata moving northward into Nevada and western New Mexico meeting with var. linearis in eastern New Mexico while var. tomenticaulis, a more recent derivitive from var. linearis differing only in stem vestiture, is restricted to the east side of the Sierra Madre Oriental-fitting well into a vicariance pattern. ACKNOWLEDGMENTS I thank the curators of the listed herbaria for loans of specimens or courtesies extended; J. Seto for use of the Jeol JSM T200 Scanning Electron Microscope (CSLA); C. Clark (CSPU) for advice and guidance on PHYSYS and cladistics (though I accept all responsibility for the statements presented); T. Elias (RSA) and A. Gentry (MO) for discussions on Bignoniaceae; M. C. Johnston (TEX) for the latin diagnosis; the University of Texas Plant Resources Center for use of facilities; and Bobbi Angell (NY) for line drawings.