Diffusely branched ascending or sprawling perennial herbs to 60 cm high; stems hirsute, eglandular; leaves petiolate, 2–6 cm long, bipinnately or tripinnately parted; flowering spikes 2–4 cm long, usually much elongating in fruit; bractlets mostly linear-lanceolate, eglandular, as long as or longer than the subtended calycies; calycies mostly 7–10 mm long, hispid-hirsute; corollas pale lavender to purple, the tubes up to 1.5 times as long as the calyx, the limb 8–10 mm across; nutlets unappendaged.

Umber (1979) noted that his concept of Glandularia bipinnatifida included “... several growth forms or ecotypes which have been given specific or varietal status by other authors.” Taxonomic recognition of these forms in his view was “. . . unwarranted based on morphological and chemical characters. The characters in this species show clinal and more or less random variation.”

I cannot agree with Umber's assessments, as quoted in the paragraph above. My examination of nearly 1,000 sheets of this complex housed at LL, TEX shows very clearly that the mass of the specimens can be divided into two morphogeographical units, as noted in the key and Fig. 1. Indeed, variation of the characters concerned are not at all clinal or random in distribution over the range of the species. Variation in Glandularia bipinnatifida var. bipinnatifida is not nearly so great as occurs in var. ciliata, discussed below, most of the synonymic names having been proposed for that element of the species. Glandularia bipinnatifida var. bipinnatifida does, however, appear to intergrade with G. b. var. ciliata along their regions of contact, as suggested in Fig. 1.

Glandularia bipinnatifida var. bipinnatifida is relatively uniform over most of its distribution, but just west of the Pecos River in trans-Pecos Texas it appears to intergrade over a relatively short distance into G. b. var. ciliata. Indeed, at a single site 56.6 mi E of Fort Stockton in easternmost Pecos Co. along IH 10,1 collected what appeared to be clear intermediates between the two taxa (Turner 97–43), these occurring next to fairly typical individuals of G. b. var. pinnatifida (Turner 97–44). Once the Pecos River is reached going eastward (only ca 15 mi or so), the populations of G. bipinnatifida var. ciliata are largely replaced by G. b.var. bipinnatifida.

During the spring of 1997, I again examined the zone of contact between Glandularia bipinnatifida var. bipinnatifida and G. b. var. ciliata in trans-Pecos Texas. The latter variety was abundant along U.S. Highway 90 between Marfa (Presidio Co.) and Sanderson (Terrell Co.), occurring in relatively small patchy populations having mostly pale lavender corollas. Immediately to the north of Sanderson and continuing northeastwards along State Highway 2400 towards Sheffield, G. b. var. ciliata was replaced by nearly continuous populations of G. b. var. bipinnatifida having darker lavender flowers (Fig. 8). The spring of 1997 was a banner year for Glandularia in Texas. Indeed, I drove from the Pecos River eastwards along IH 10 to Kerrville (some 300+ km), and there was only rarely a gap of more than 500 meters between populations of G. b. var. bipinnatifida.

Westward from the Pecos River along IH 10, Glandularia bipinnatifida var. ciliata replaces G. b. var. bipinnatifida, as already noted. Populations of the former are only rarely continuous, the taxon occurring mostly in small patches 2–5 km apart, much as along highway 90, as mentioned in the previous paragraph. North of Interstate 10, along State Highway 67, the G. b. var. ciliata was first detected about 5 mi W of Given (Turner 97–58). Variable populations with variable individuals occur in the contact zone of the two taxa, these mostly confined to a relatively narrow band 8–50 km along either side of the Pecos River in Texas. I surmise that this intergradation is a selection phenomenon along this transition zone between the more mesic habitats of central Texas and those of a more xeric nature in western Texas (i.e., such variation being due to primary intergradation, as opposed to secondary intergradation with allopatric hybridization a factor). At least I never detected the co-occurrence of distinct populations of these two taxa with clear-cut putative hybrids in the area concerned.

Verbena ciliata Benth., PL Hartw. 21. 1839. Type: MEXICO. GUANAJUATO: Leon, 1839, Hartweg 176 (holotype K; isotypes GH, NY, according to Umber, 1979, none of these examined by me).

Essentially as described for Glandularia bipinnatifida var. bipinnatifida but the bractlets subtending the calyces narrowly lanceolate and mostly shorter than the subtended calyces; additionally the vestiture is much more variable, this usually hirsute with spreading glandular hairs or various lengths, or these intermixed with eglandular hairs, rarely eglandular throughout.

Umber (1979) recognized Glandularia wrightii as a distinct species, distinguishing this from his concept of G. bipinnatifida (s.l.) by a number of fuzzy characters that, in my opinion, do not serve to distinguish anything but fuzzy concepts. Most of the collections from the trans-Pecos area of Texas were referred by Umber (through annotation, LL-TEX) to G. bipinnatifida. Perry (1933) also recognized G. wrightii as a distinct species but perceptively noted that “The main characters [which separate the two] appear to be the glandular somewhat viscid pubescence of the calyx, the very short acute-subulate calyx-teeth, and the somewhat compact spikes ” These are the very characters that distinguish G. b. var. ciliata from G. b. var. bipinnatifida. She noted, in addition, that “On account of its strong variability, this species has been difficult to define” Careful examination of plants belonging to this complex from throughout the southwestern U.S.A. and Mexico has convinced me that the more western glandular elements of G. bipinnatifida (including G. wrightii, the type from near El Paso, Texas) belong to but a single widespread taxon, G. b. var. ciliata, the latter varietal name legitimized with the description of Verbena ciliata var. longidentata Perry.

Umber placed Verbena ciliata var. longidentata in synonymy under his broad concept of G. bipinnatifida, but since the former name is typified by material from near Matamoros, Mexico (Tamaulipas), it is more likely that this name is synonymous with G. polyantha Umber, the latter essentially endemic to the more lowland, Gulf Coastal regions of southern Texas and Mexico. At least I know of no collections of G. bipinnatifida from this region, although Umber shows several such collections from this area. I take the latter to be but misiden-tifications of G. polyantha, to judge from his annotation labels (LL-TEX).

Umber (1979) also placed Verbena andrieuxii Schauer in synonymy under Glandularia bipinnatifida, but examinations of a photoholotype, LL (also examined by Umber!), and comparison of recent collections from the state of Oaxaca, Mexico, whence the type, leads me to place the plant concerned in synonymy with the pubescent phase of G. teucriifolia (Mart. & Gal.) Umber (Turner, in prep.).

Finally, it should be noted that Umber's Glandularia bipinnatifida var. brevispicata (the type from Otero Co., New Mexico) is, in my opinion, but a regional variant of G. b. var. ciliata , if that. His description of the short spikes, calyx teeth, viscid inflorescence, ascending habit, dark foliage, and acrid-smelling flowers, applies to my concept of G. b. var. ciliata (except for the acrid-smelling flowers, which I cannot evaluate). Umber (in his Fig. 7) also maps specimens of his var. brevispicata as occurring in the Central Plateau region of northern Mexico, all of these referrable to G. b. var. ciliata, in my treatment (Turner, in prep.).

Sparsely or diffusely branched ascending or decumbent perennial herbs to 50 cm high; stems glabrous to variously hirsute, but not densely so; leaves extremely variable mostly 3–10 cm long, 1.5–4.5 cm wide, deeply pinnately divided to merely 3-part-ed, appressed pubescent to glabrate; flowering spikes 2–5 cm long, elongating in fruit; bractlets linear-lanceolate, as long as or shorter than the subtended calyces; calyces 10–14 mm long, sparsely glandular-pubescent, rarely not; corollas varying from pink to purple, the limbs mostly 10–16 mm across; nutlets exappendiculate.

My concept of this taxon is about the same as that of Perry (1933) and Umber (1979). Because of its very large corollas and relatively elongate weakly dissected leaves, it is readily distinguished in the field, occurring primarily in sandy soils of eastern Texas (cf. Fig. 2).

My concept of this taxon is about the same as that of Perry (1933) and Umber (1979). It is superficially similar to G. polyantha but readily distinguished from the latter by its much smaller corollas and seemingly prostrate habit, as noted by Umber. While occurring over a broad range of eastern Mexico, it is known to Texas only by collections from along the Rio Grande River, where it is mostly associated with palm groves.

As noted by Umber (1979), plants of his newly described species, the type from Cameron County, Texas, were long placed under G. delticola. Except for its distinctive leaves it might with some justification be treated as part of the G. pinnatifida complex. Indeed, Perry (1933) treated at least some of the specimens positioned by Umber in his concept of G. polyantha in her concept of Verbena ciliata var. longidentata; the type of the latter is from Matamoros, Mexico, and has all of the attributes of Umber's G. polyantha, except for its more deeply divided leaves. Umber provided an excellent full description which I need not improve upon here.

This widely introduced cultivar, native to South America, has become relatively common in the southeastern U.S.A., as noted by Umber (1979). The species is readily distinguished from yet other Texas species by its deeply dissected leaves, the ultimate segments linear.

Diffusely branched mostly prostrate or procumbent perennial herbs 2–5 cm high; stems hirsute with mostly short glandular hairs, these after intermixed with eglandular hairs; leaves mostly 1.5–3.5 cm long, variously dissected but oftentimes merely tripartite with lobed divisions; spikes mostly 1–3 cm long, not much elongating in fruit; bractlets linear-lanceolate to narrowly lanceolate, as long as the subtended calyces, hispid-hirsute with mostly glandular hairs; calyces 5–6 mm long, usually to some extent glandular-pubescent; corollas deep lavender to purple, the tubes slightly extending beyond the calyx, the limbs mostly 2.5–4.5 mm across; nutlets exappendiculate.

My concept of this taxon is about the same as that of Umber (1979), except that he recognized the species as occurring in southern Texas. I take most of the latter to be misidentifications of exappendiculate forms of Glandularia quadrangulata (cf. discussion under the latter).

Glandularia pumila is relatively rare in trans-Pecos Texas; I know it from this region by relatively few collections: El Paso Co., Worthington 4295 (TEX); and Brewster Co., Ertter 4723, (TEX); Turner 97-15, (TEX).

Diffusely branched prostrate to semierect sprawling perennial herbs mostly 2–15 cm high; stems hirsute, eglandular; leaves short-petiolate to sessile, variously pinnately parted with narrow to relatively broad divisions; flowering spikes mostly 1–3 cm long, not much elongating in fruit; bractlets narrowly lanceolate, hirsute with eglandular hairs, rarely not, as long as or somewhat shorter than the calyces; calyces 5–6 mm long, hirsute with eglandular hairs; corollas white or creamy white to very pale lavender, the tubes emerging from the calyces for 0.1–4.0 mm; nutlets with conspicuous ovate appendages, or these absent.

My concept of this taxon is about the same as that of Umber, except that I recognize the species as occurring in two forms: a typical form having nutlets with well-defined ovate appendages, and an atypical form lacking such appendages. The two forms occur together over a broad region (Fig. 6) and may occur in the same population, as noted by Umber on Umber 3, (TEX) (which he identified as a Glandularia verecunda): “indistinguishable from G. quadrangulata except for the seeds.” I treat G. verecunda as an exappendiculate form of G. quadrangulata; the plants from south Texas and elsewhere which Umber identified as G. verecunda are clearly exappendiculate forms of G. quadrangulata. This variation in nutlet shape has also plagued nearly all previous workers, the exappendiculate forms being misidentified as either G. pumila (which to my knowledge does not occur in southernmost Texas), G. racemosa (including G. verecunda sensu Umber), G. delticola, or G. racemosa.

In the early stages of this study I was inclined to accept Glandularia racemosa (the type from near Sierra Blanca, Hudspeth Co., Texas, not examined) as distinct, but subsequent field work at or near the type locality, and examination of a large suite of herbarium specimens has convinced me that it readily nestles into G. quadrangulata, differing from the latter only by its exappendiculate nutlets. Umber (1979) distinguished G. racemosa from G. quadrangulata (in his key to species) by the nutlet commissure, which was said to be 1/2 the width of the nutlet and “definitely not reaching the top of the nutlet” and further, that the nutlet at its base was cylindrical and without a broadened base. Examination of a large suite of collections (LL, TEX) annotated by Umber as G. racemosa shows that these “key characters” do not hold. The commissure width of these specimens range from 0.4–0.6 mm, which is about the same as the commissure variation found in his G. verecunda and G. quadrangulata (0.4–0.8). Indeed, other than the appendiculate nutlets in the latter collection, I can find nothing to distinguish between the various names I have listed in synonymy with the present. Perry (1933) also maintained G. racemosa but distinguished this from G. quadrangulata (in addition to the nutlet appendages) by its more exserted corolla tubes, a character that will not serve to discriminate between these; in the specimens from trans-Pecos Texas, the tube exsertion varies from 0.5 to 5.0 mm, with an average of about 3.0 mm (13 specimens measured).

Populational forms of exappendiculate G. quadrangulata occur throughout most of trans-Pecos Texas and northern Mexico, but in southern Texas and along the Pecos River in trans-Pecos Texas, mixed “populations” occur (e.g., Turner 97-1 and Turner 97-45 from Crockett Co., and near Sheffield, Texas, Pecos Co., Turner 97-167 [all TEX]).

Finally, it should be noted that both appendiculate and exappendiculate forms of G. quadrangulata have been collected at the same locality on the same date by the same collector in sandy soils along the Rio Grande in trans-Pecos Texas (Fletcher 1224, 1226 [SRSC]). Except for the appendage variability, these two collections are identical.

Because of its pecular ovoid nutlets, this species is unlikely to be confused with another. In early flower, however, it is superficially similar to Glandularia bipinnatifida, and has been identified as such by various authors, especially among Mexican collections. Indeed, Umber himself misidentified a Coahuilan collection of G. tumidula as G. gooddingii (Briq.) Solbrig, mapping this as such in his Fig. 8, considerably out of range of the latter, an element of the Sonoran Desert flora.

This is a common, very attractive and commonly cultivated ornamental. It is grown in gardens throughout the temperate world and its origin is moot, as discussed in more detail by Nesom 8c Pruski (1992). Moldenke (1963) has provided a prolonged account (32 pages!) of the history and nomenclature, including common names, of this taxon (this reference not cited by Nesom and Pruski).

I have not accepted this taxon as part of the Texas flora because it does not appear to persist when escaping cultivation. I have, however, examined two atypical collections from Texas (Bexar Co.: Crockett 231 [LL]; Cameron CO. Knobloch s.n. 20 Feb 1981, [TEX]) that were apparently collected along highways; both were identified as Verbena canadensis initially, but these appear to be atypical cultivars of Glandularia × hybrida, although it is possible that these are exceptional disjunct populations and/or individuals of G. canadensis. Both Moldenke and Umber annotated the Crockett collection as G. canadense, but I sense the two plants to be but aberrant leaf forms of the cultivated G. × hybrida.

Glandularia bipinnatifida var. brevispicata Umber = G. bipinnatifida var. ciliata

Glandularia gooddingii Briq.: Reported for Texas by Moldenke (1970, as Verbena) but the material concerned is mostly eglandular forms of G. bipinnatifida var. ciliata.

Glandularia wrightii (A. Gray) Umber = G. bipinnatifida var. ciliata

Verbena ambrosifolia Rydberg ex Small = G. bipinnatifida var. ciliata

Verbena cameroensis L. I. Davis = G. delticola

Verbena ciliata Benth. = G. bipinnatifida var. ciliata.

Verbena ciliata var. longidentata Perry = G. polyantha?

Verbena elegans HBK.: So far as known, this handsome species, which is native to the higher elevations of Mexico (1000–2300 m) does not occur in Texas, although reported by Moldenke (1970) to occur in Duval and Hidalgo Counties. The latter collections (Lewis & Oliver 5410 and E. J. Walker 32, not examined; cited by Moldenke in Phytologia 9:88.1963) are either misidentifications of cultivated and escaped forms of V. × hybrida and/or introduced cultivar forms of V. canadensis. I have referred similar cultivars from south Texas (e.g., Knobloch s.n.-, Cameron Co.) to Glandularia × hybrida, although these may, indeed, be cultivars of G. canadensis, if not introduced (but not persistent) elements of G. elegans.

Verbena lundelliorum Moldenke = G. delticola.

Verbena oklahomensis Moldenke: According to Moldenke (1970) this name applies to putative hybrids between Glandularia bipinnatifida var. bipinnatifida and G. canadensis. While inadvertently treated without the hybrid notation in Correll & Johnston (1970), Moldenke (Phytologia 224: 27.1972) corrected this editorial lapse. Moldenke reports a single collection from Lampasas Co., Texas (Mahler 1241; cf. Phytologia 10: 284.1964). Such hybrids might well occur in Texas, but I suspect the latter collection to be but an atypical element of G. bipinnatifida var. bipinnatifida, mainly because populations of G. canadensis do not occur in Lampasas County to my knowledge, thus the natural hybrid between the putative taxa concerned is unlikely to occur there.

Verbena hybrida Grönland & Rumpler = G.x hybrida.

Verbena wrightii A. Gray = Glandularia bipinnatifida var. ciliata.