Systematics of the southern African genus Ixia (Iridaceae). 1. The I. rapunculoides complex

Field and laboratory investigation of the six varieties of Ixia rapunculoides Delile (recognized in the current revisions and flora accounts of this southern African genus o f some 65 species) show that a revised taxonomy better reflects the biology of the complex. The complex is defined by the funnel-shaped perianth tube, filaments fully and anthers partly included in the perianth tube. Our conclusions indicate that I. rapunculoides (currently var. rapunculoides) is restricted to the western Karoo and has broad, leathery leaves, a short perianth tube, mostly 5.5-8.0 mm long, and half nodding blue-mauve flowers. Plants currently referred to I. rapunculoides var. flaccida G.J.Lewis, include four sets of populations that we regard as two separate species. These are: I. flaccida with small, short-tubed whiteor pale blue-flushed flowers, sofl-textured leaves and corms with basal cormlets from the Olifants River Valley and nearby; and I. sobolifera from the Western and Little Karoo, which has linear leaves, nodding spikes and flowers and corms with stolons. We divide I. sobolifera into three subspecies: subsp. carnea with pink flowers, is restricted to the Bokkeveld Plateau and nearby; subsp sobolifera with slate-blue flowers occurs in the Klein Roggeveld and nearby; and subsp albiflora with strongly scented white flowers occurs in the central Little Karoo. The taxon called I. rapunculoides var. namaquana (L.Bolus) G.J.Lewis, defined by a longer perianth tube, mostly 13-16 mm long, horizontally oriented, white, pale lilac or pink flowers and few-flowered lateral branchlets. is treated as I. namaquana, first described by H.M.L. Bolus in 1931. Plants referred to this taxon from south of its range at Hex River Pass and nearby, however, have fully included anthers, a longer perianth tube. 16-20 mm. and ascending purple-pink flowers with a white cup and are referred to the new /. oxalidiflora. Two more varieties, I. rapunculoides var. subpendula G.J.Lewis and var. rigida sensu G.J.Lewis, which have upright flowers and distinctively branched stems are treated here as I. divaricata and I. contorta. Plants in the past included in var. rapunculoides from the Klein Roggeveld, south of the range of typical I. rapunculoides, have a longer perianth tube. 10-14 mm long, and attenuate, slightly lacerate. 5-veined. dry. rust-tipped bracts and comprise the new 1 lacerata In addition, plants from streambeds in the Roggeveld that have large, white flowers, are not included in current accounts of the genus, and also represent a novel taxon, I. rivulicola. Lastly, I. rapunculoides var. robusta G.J.Lewis, the range of which falls entirely within that o f var. rapunculoides and is usually sympatric with it. has pink flowers of similar structure, but four or five leaves and deep-seated corms with a collar of coarse fibres around the stem base. We raise this plant to species rank as I. robusta


INTRODUCTION
As currently circumscribed. Ixia rapunculoides Delile (Iridaceae: Crocoideae) is a widespread species of the southern African winter rainfall zone. It is a member of Ixia subgenus Morphixia (G.J.Lewis) J.C.Manning & Goldblatt, one of two subgenera of this southern African genus of some 65 species (Goldblatt & Manning 1999). The species is variable and comprises six vari eties, which together constitute a complex that may be defined by its specialized bell-like flowers with unusu ally short stamens and short style branches, often less than 1 mm long. The filaments are included in the fun nel-shaped perianth tube, and the anthers are usually partly included. Whether partly included in the tube or exserted. the anthers remain enclosed within a cup formed by the lower half of the tepals. The upper half of the tepals spreads outward ± horizontally, except in I. rapunculoides var. subpendula, in which the upper half of the tepals is often not fully patent. As treated by Lewis (1962), whose account was followed exactly by M.P de Vos (1999), the complex included just I. rapunculoides with its six varieties. The range of I. rapunculoides as understood by Lewis extended from the Richtersveld of northern Namaqualand in the northwest, through interior Northern and Western Cape Provinces of South Africa to the Kammanassie Mountains of the Little Karoo in the southeast, a distance of over 800 km. and covering almost the entire range of the genus. One more species that has been described in the complex. I. namaquana L.Bolus (Bolus 1931), was reduced to varietal rank in I. rapun-culoides by Lewis (1962) on the basis of its included fila ments and partly included anthers.
Vegetatively, all members of the complex have slen der, erect flowering stems bearing few to many, short, ± wiry, often thread-like lateral branchlets and typi cally, but not exclusively, just three (rarely two or four to six) foliage leaves. The lower two (rarely one or up to four) leaves have well-developed blades, and the upper one (or two) completely sheath the stem and reach to between the middle of the stem and the base of the spike. The main spike bears relatively few flowers and the lat eral spikes have fewer, sometimes only a single flower. Plants referred to I. rapunculoides var. robusta normally have four leaves but sometimes up to six, with three or four basal and one or two sheathing the stem, and two other varieties rarely have three expanded foliage leaves. None of these vegetative features are, however, unique to the complex, but are shared by several other species of subgenus Morphixia (Goldblatt & Manning 1999), notably Ixia latifolia D.Delaroche and I. marginifolia G.J.Lewis.
Lewis's taxonomic treatment seems to reflect a per ception that the morphological variation within the com plex was less significant than that encountered between species elsewhere in the genus, presumably because the distinctive flowers of the complex were given undue weight in a genus that is relatively conservative in floral morphology. Some of the variation, especially floral scent, corm features, and the nature of the bracts subtending the branchlets was simply overlooked. The morphological variation in the complex is, nevertheless, extensive and includes: 1, corm tunics that range from coarsely to finely fibrous, or softly papery; 2, cormlets that are either sessile at the base of the main corm or borne on flattened, ribbon-like stolons; 3, flower colour that is often blue to blue-mauve, but also purple, pink or white; 4, leaf shape, which may be narrow, linear, and erect, or sword-shaped and erect, or arching outward and falcate; 5, degree of thickening of the margins of the basal leaves, which is least developed in I. rapunculoi des var. flaccida and most pronounced in var. robusta and var. rapunculoides; 6, shape and degree of develop ment of the bracts and prophylls subtending the thread like, lateral branches (here called branchlets when short, or branches when well-developed); 7, length of the peri anth tube, mostly 5.5-8.0 mm in var. rapunculoides, and 12-16(-18) mm in var. namaquana, 8, length of the fila ments and anthers, features linked to tube length, and together range from 4.3-6.5 mm in var. rapunculoides to a maximum of 9.5 mm in var. namaquana and 10.5 mm in var. subpendula.
The morphological variation often correlates with ecological and geographical differences but the pat terns have until now been inadequately understood. Particularly noteworthy, and from a systematic point of view unsatisfactory, is the local sympatry of some vari eties or the parapatric and edaphic or geographic sepa ration of others, in all cases without the occurrence of plants of intermediate morphology.
For example, on dolerite-derived clay soils on the Bokkeveld Plateau and on the Hantamsberg at Calvinia in Northern Cape, both Ixia rapunculoides var. rapun culoides and plants that conform to var. flaccida (sensu Lewis 1962), flower contemporaneously within a few metres of one another with no intermediates, indicating complete genetic isolation. Sympatry and synchrony of flowering of two distinct races without the presence of intermediate individuals is usually accepted as an indi cation of genetic isolation and therefore evidence for their recognition as separate species. Similarly, Lewis (1962) noted that at one of its two localities then known, var. robusta grew on rocky dolerite slopes a few metres from typical var. rapunculoides, growing on flat, stony loam near a seasonal watercourse. We have confirmed Lewis's observation and amplify it, having seen var. robusta growing side by side with var. rapunculoides, both taxa maintaining their typical morphology. A com parable situation occurs on the Bokkeveld Plateau near Nieuwoudtville, where typical var. rapunculoides grows on tillite-derived clay soils and var. namaquana on adja cent sandy soils.
These examples are particularly surprising because the flowers of all these taxa are apparently adapted for pollination by the same insects, a range of mediumand large-bodied bees, mostly Apidae: Anthophorinae, that forage for nectar and acquire loads of pollen while probing the perianth tube (Goldblatt et al. 2000 and con firmed here).
Field observations suggest to us that varietal sepa ration within Ixia rapunculoides of var. namaquana, var. flaccida and var. robusta is unsatisfactory because the biological situation is not concordant with this tax onomy. The recognition of varieties of a single species is notoriously unevenly applied (Stuessy 1990), but is most often used to denote trivial regional variation, and sometimes only small genetic differences such as flower colour or leaf shape. In our estimation it should never be associated with genetic isolation and effective crossing barriers at the microgeographic level. Raising the rank at which members of the complex are separated to that of subspecies has similar problems. The rank of subspe cies is most often used for sets of populations that are isolated geographically and have relatively modest, and often overlapping, morphological differences (Davis & Heywood 1973). As with varietal rank, subspecific rank is inconsistently applied. The examples in I. rapuncu loides listed above thus accord poorly with treatment at either varietal or subspecific rank.

MATERIALS AND METHODS
We have attempted to resolve this unsatisfactory taxo nomic situation outlined above where so-called varie ties of a species co-occur and flower synchronously yet maintain their morphological differences by examination of the morphology and ecology of plant populations in the field. These studies were complemented by a study of herbarium collections of the genus at BOL, K, MO, NBG and PRE, the herbaria with the most complete col lections of species from the southern African winter rain fall zone. Then, using a morphological species concept, we have compared all significant taxonomic characters of the different sets of populations with their ecology and geography to establish a revised taxonomy of the complex.

RESULTS
Our results, obtained from measuring both living plants in the field and carefully preserved specimens with laid out flowers are presented in tabular form (Table 1). We do not use Lewis's (1962) or De Vos's (1999) measures for any taxa because we apply names in different ways. We have also been careful to include measurements of only well-pressed herbarium speci mens because floral parts are particularly susceptible to distortion and can shrink by as much as 20% as they dry, less so when rapidly pressed on glue-covered paper or between the pages of a heavy book.
Members of the Ixia rapunculoides complex are not reported in the literature to have scented flowers (Lewis 1962;De Vos 1999;Manning et al. 2002). Nevertheless, we have found that plants of most populations have distinctive floral odours. Plants from the eastern part of the range of the complex with nodding white flow ers have a particularly intensely sweet, carnation-like odour; typical var. rapunculoides has a faint rose scent; var. namaquana usually has a violet scent or are occa sionally unscented; and var. robusta has a faint, unpleas ant chemical odour. A population of white-flowered plants allied to var. rapunculoides from the Roggeveld Escarpment has a strong fruity odour recalling a combi nation of banana and grenadilla, very similar to Virgilia oroboides (Fabaceae).
Another feature ignored in the past, is the nature of the bracts and prophylls subtending the branchlets. These range from minute and truncate (var. rapunculoi des, var. robusta) to forked with attenuate tips (Western Karoo plants assigned to var. flaccida), to filiform, elon gate and recurved (plants from the Klein Roggeveld referred by Lewis (1962) to var. rapunculoides or var. namaquana).
Corms of members of the complex have until now been described as 'having fairly coarse, often subligneous, reticulate fibres' (Lewis 1962) and neither Lewis nor De Vos (1999) mention the presence of cormlets (cormels) or underground stolons. Texture and compo sition of the tunics, however, varies considerably and more so than is ever found within a single species of any allied genus of Iridaceae. Hard, coarsely fibrous tunics, matching Lewis's description, are characteristic of var. rapunculoides, var. robusta and var. namaquana and these taxa bear ± sessile cormlets at the base of the main corm. Soft, short-lived tunics o f ± membranous layers characterize some plants included in var. flaccida and the southern populations of pink-flowered plants referred to var. namaquana, and in addition these plants produce long, flattened, ribbon-like stolons each terminating in a large cormlet. Typical var. flaccida has tunics of fine to medium textured, netted fibres and produces cormlets at the base of the parent corm.
1. Var. rapunculoides: the typical form of the species is based on a painting in P.J. Redoute's Les Liliacees (Delile 1816), and we associate Ixia rapunculoides (the campanula-like Ixia) with plants from the Bokkeveld Plateau and Hantamsberg with a blue perianth and pale yellow cup. The painting, not known to be associated with any preserved material, shows the characteristic blue-mauve perianth with a yellow cup, ± horizontally oriented flower, numerous multi-flowered branchlets, and most important, the short, funnel-shaped perianth tube with a wide upper half. Measurements of fresh, field-collected flowers from the Bokkeveld Plateau (e.g. Goldblatt & Porter 12162) show a perianth tube 7-8 mm long (but only 5.5 mm at the southern end of its range on the Roggeveld Escarpment), filaments ± 2 mm long, and anthers 2.3-3.5 mm long. Well-grown plants typi cally have three leaves, occasionally four. The lower two (or three) are usually relatively broad and lanceolate to falcate, with hyaline margins that appear thickened when dry, and the upper leaf is entirely sheathing. The corm tunics are coarse-textured and accumulate in a thick mass with age, occasionally extending upward as a col lar around the stem base. A few small cormlets are usu ally developed at the base of the main corm.
The taxon favours clay soils, either heavy red clay derived from dolerite or lighter Ecca shale-or Dwyka tillite-derived soils and grows in renosterveld or karroid scrub. Plants conforming to these features extend from the Langberg near Loeriesfontein in the Western Karoo through the Bokkeveld Plateau to the Roggeveld Escarpment near Sutherland and are morphologically fairly uniform throughout this range. Plants from the Kamiesberg in Namaqualand that were included by Lewis in var. rapunculoides have well-exserted anthers and are actually a local, blue-or mauve-flowered vari ant of the plant currently called I. latifolia var. ramulosa G.J.Lewis.
2. Var. robusta: Ixia rapunculoides var. robusta G.J.Lewis was known to Lewis (1962) from two locali ties around the Hantamsberg Massif at Calvinia. The taxon exhibits two features associated with dry habitats: thick, almost succulent, leathery leaves with thickened, hyaline margins and deeply seated corms with a collar of fibres around the underground part of the stem. The flowers are similar to those of var. rapunculoides, but are consistently pale pink with a white cup, and are slightly to much larger than those of var. rapunculoides grow ing nearby. They have a somewhat longer perianth tube, 7.5-10.2 mm long, and tepals 17-20 mm long. Plants also differ sharply from var. rapunculoides in having three or occasionally four leaves with expanded blades in addition to one, or sometimes two, sheathing upper leaves. Most other variants of I. rapunculoides, includ ing var. rapunculoides, normally have two expanded foliage leaves (rarely one or three are present) and the latter has a perianth tube usually 5.5-8.0 mm long. Plants with large pink flowers, at least four leaves and a fibrous collar occur at numerous sites east and north of the Hantamsberg at the northern end of the Bokkeveld Plateau and plants with four leaves and a fibrous collar are also recorded at the southern end of the Roggeveld Escarpment (see exsiccatae listed below under I. rapun culoides). Flowers of these plants are fairly large, and have anthers ± 4 mm long and tepals (15-) 17-20 x 5.0-9.5 mm, thus in the upper range for var. rapuncu loides. Plants of the type collection of var. robusta from Moordenaarspoort, 38 km northwest of Calvinia, have bracts with blunt tips that contrast with the toothed tips of other varieties of I. rapunculoides but we note that some populations of typical var. rapunculoides also have blunt-tipped inner and outer bracts (see illustration by Lewis 1962).

3.
Var. nam aquana: first described by H.M.L. Bolus in 1931 as Ixia namaquana, the species was based on pink-flowered plants from Klipfontein near Steinkopf in northern Namaqualand in Northern Cape. It was reduced to varietal rank by Lewis (1962), who included not only plants from Namaqualand in the taxon, but also collec tions from the Bokkeveld and the Cedarberg as well as from the Worcester and Laingsburg Districts. Except for these southern populations, var. namaquana has flow ers held horizontally or slightly above the horizontal and a relatively long perianth tube, (12-)14-16(-18) mm, and anthers 3.5-4.5 mm long. The flowers are whitish, mostly oyster-coloured flushed pale mauve, or deep pink, and the pale yellow throat has a ring of dark lines at the top of the perianth tube. Plants we have examined in the wild from the Kamiesberg and the Bokkeveld Mtns have the style dividing at ± mid-anther level and this feature is confirmed in the few preserved specimens that show the character. Some plants we have examined alive have flowers with a strong scent of violets (e.g. Grasberg road, northwest of Nieuwoudtville, Goldblatt & Porter 12407A;Goldblatt 12684), whereas others appear scent less (e.g. 10 km west of Nieuwoudtville, Goldblatt 12680). The taxon is consistently different from other varieties of I. rapunculoides and does not overlap them in the critical feature of perianth tube length and differs from most others in anther length. We regarded it as a separate species.
Southern populations of Ixia namaquana from the Bokkeveld Mtns to the Cedarberg favour sandstone-derived soils, whereas those in the north, in the Kamiesberg, occur in gritty granitic gravel, typical of central Namaqualand habitats. Both soil types are nutrient-poor and well drained. The type population and several more from the Steinkopf area of northern Namaqualand and the southern Richtersveld apparently grow on clay, an odd ecological shift. The latter popula tions also differ in their deep pink flowers.
Among specimens until now included in the taxon, those from the south of the range, in the Hex River Mtns and nearby, stand out in having unusually large, pink flowers that are either ascending or upright and have the stamens fully included in the funnel-shaped tube, itself unusually long, 16-22 mm. These plants are also unlike typical Ixia namaquana in their narrow, soft-textured leaves and small corms, about 8 mm in diameter. The corms have tunics of fine fibres, and produce conspicu ous stolons. These plants cannot be included in I. nama-quana but represent a novel species, which we describe below as I. oxalidiflora.

4.
Var. flaccida: the type of this variety, from the Olifants River Valley near Clanwilliam in Western Cape, is distinguished by the fairly thin-textured, nar rowly lanceolate leaves less than 10 mm wide and ± half as long as the stem. The white to palest blue flow ers are carried ± horizontally, have a perianth tube 7-9 mm long, while the stamens are inserted ± 2 mm from the base of the tube, the filaments are 2.5-3.0 mm long and the anthers ± 2 .5 mm long. Plants have three or four leaves, the uppermost one (or sometimes the upper two when four leaves are present) is typically entirely sheathing. The corms have tunics composed of fine or medium-textured, netted fibres and bear cormlets at the base. Plants grow in light sandy soil, among sandstone rocks, on south-trending slopes, in light shade, in thicket communities. They most closely match var. rapunculoi des in tube and stamen length but apart from differences in leaf shape and texture, they have narrower tepals and corm tunics of finer fibres. Dimensions given by Lewis (1962) and De Vos (1999) for the variety have a wider range because they included additional sets of popula tions within the taxon.
The first of these populations composes slender, pinkflowered plants from the Bokkeveld Plateau, the second slate-blue-flowered plants from the Klein Roggeveld, and the third represents white-flowered plants from the central Little Karoo. These variants are recognized by the taller than usual flowering stem, up to 750 mm long when growing under conditions of adequate rainfall, a half or fully nodding flower, and perianth with a pale yellow cup. The leaves of these plants are narrow, lin ear to narrowly sword-shaped, slightly twisted and often trailing above. Most significantly, the corms differ from all other populations of I. rapunculoides sensu Lewis including typical var. flaccida, in having soft, soft-papery corm tunics and broad, flat stolons 30-100 mm long ter minating in a large cormlet.
At two sites where we have seen the pink-flowered variant, at Glenlyon near Nieuwoudtville and on the Hantamsberg at Calvinia, plants were sympatric with typical var. rapunculoides. At a third site, near Grasberg Farm, northwest of Nieuwoudtville, pink-flowered plants were sympatric with var. namaquana. The taller pinkflowered plants grew in grass tufts or in bush clumps, whereas var. rapunculoides and var. namaquana grew on open ground. We conclude that these pink-flowered plants represent an unnamed taxon. Apart from the nod ding flower, pink perianth and narrow tepals, flowers of this plant have a tube ± 9 mm long, filaments ± 3 mm long, and anthers ± 4 mm long. The style divides at the anther apices in fully open flowers and the style branches are short, slightly exceeding 1 mm long, thus visible above the anthers. Blue-flowered var. rapunculoides growing nearby has a perianth tube 7-8 mm long, fila ments ± 2 mm long, and anthers ± 3 mm long, whereas var. namaquana has a tube (12-) 14-16 mm long, anthers ± 4 mm long and a shorter style.
We have also re-collected and examined the greyblue-and the white-flowered plants referred by Lewis to var. flaccida at sites in the Klein Roggeveld and Little Karoo. Plants from the latter area grow on south-trend ing shale slopes in light loamy ground near Oudtshoom and in the foothills of the Kammanassie Mountains (e.g. Goldblatt & Porter 12291, MO, NBG;Vlok & Schutte 494, MO, NBG), and at a few other sites. These plants resemble most closely the pink-flowered plants from the Bokkeveld Plateau in their tall stature, and narrow, ± lin ear leaves and, most significantly in their soft, somewhat papery corm tunics and long, slender stolons bearing a terminal cormlet. The Klein Roggeveld plants, which also have corms with soft tunics and stolons, favour clay or sandy soils in more mesic sites in this semi-arid area, often south-facing slopes. Because these three sets of populations of stoloniferous plants differ in flower colour, branching pattern, and in the strength and quality of their scents, as well as in small differences in size of floral parts, they are probably best regarded as subspe cies of a single species, which we call I. sobolifera. We propose the names subsp. albiflora for the Little Karoo populations, subsp. carnea for those from the Bokkveld Plateau, while the Klein Roggeveld populations con stitute subsp. sobolifera. The disjunction between the northern subsp. carnea and Klein Roggeveld subsp. sobolifera may prove to be apparent rather than real as more collecting on the Roggeveld Escarpment is done. Likewise the disjunction between the ranges of subsp. sobolifera and the central Little Karoo subsp. albiflora will likely be bridged when more exploration at suitable sites between their ranges is undertaken. 5. Var. subpendula: plants assigned to this taxon have a discrete ecogeographic range, encompassing the Cold Bokkeveld and surrounding Grootwinterhoek and Hex River Mtns where they grow in seasonally waterlogged sandstone-derived soils. They also comprise a coherent morphological unit, having moderate-sized, white to pale pink flowers with a perianth tube 8-11 mm long, sub equal tepals 11-14 x 6-7 mm, anthers that are 4.0-5.5 mm long and usually fully exserted from the floral tube (sometimes the bases are included), and style branches 1-2 mm long. The narrow leaves are typically 3-5 mm wide (but sometimes up to 10 mm). The remarkable long, straight lateral branches diverge at more than 45° from the main axis and bear flowers in the distal half. These populations seem out of place in the Ixia rapunculoides complex, particularly in the large, excluded anthers and relatively longer style branches, and the flowers match most closely those of the related I. capillaris which is distinguished not only by the shortly exserted anthers but by the narrow, linear leaves, and particularly by the slender lateral branchlets, which are fairly short, rarely exceeding 10 mm. and usually bearing only one or two, rarely three, flowers. We suspect that var. subpendula is not immediately allied to I. rapunculoides. While its flowers resemble most closely those of I. capillaris, the unusual branching pattern recalls the pink-or red-flow ered I. latifolia.
Whereas Ixia capillaris and I. latifolia favour clay soils and are fairly widespread in the southwestern Cape, I. rapunculoides var. subpendula is restricted to season ally waterlogged, sandy, stony flats and rocky sites. We recognize var. subpendula as a separate species, which we call I. divaricata for the unusual, stiff, straight lat eral branches that diverge at a sharp angle from the main axis. Use of the epithet subpendula seems inappropri ate, for the lateral branches are rarely subpendulous, and usually held above the horizontal.

6.
Var. rigida: the type of var. rigida from the Hex River Valley, does not differ in any significant respect from var. subpendula and we therefore include it in that taxon. Other collections from the Cedarberg and Cold Bokkeveld that were assigned to var. rigida by Lewis (1962), however, appear to represent a different spe cies. Despite Lewis's comment that the androecium and gynoecium are like those of var. rapunculoides, we find the stamens of these plants rather different: the filaments are ± 3 mm long (versus ± 2-3 mm in var. rapunculoi des), and the anthers are 3-4 mm long, again longer than is usual for var. rapunculoides. The flowers also differ from typical var. rapunculoides (Table 1) in their upright orientation, and the slightly longer perianth tube, 9-12 mm long (versus 5.5-8(-10) mm in var. rapunculoides). We see no particularly close relationship of these plants to typical I. rapunculoides or any other member of the complex and treat them as a separate species, I. contorta.
Additional variants: there are two more sets of popu lations allied to Ixia rapunculoides. The first of these comprises white-flowered plants from the Roggeveld Escarpment, unknown to Lewis when she revised Ixia in 1962. Flowering in October (when all other members of the I. rapunculoides complex in the Western Karoo and Roggeveld are in fruit), these plants are confined to the edges of streams that usually have running water until October and remain moist at least until December. Plants are tall, up to 1.2 m high, have membranous corm tunics that do not accumulate with age, and bear small corm lets at the base of the main corm. The relatively large, nodding flowers have a tube 8-9 mm long, tepals 17-18 x 5.0-7.5 mm, and anthers 4.5-5.5 mm long, the lower halves included in the perianth tube. The flowers have a strong fruity odour, reminiscent of banana and grenadilla (passion fruit), unique in Ixia, but closely matched by the scent produced by Virgilia oroboides.
We suggest that the most appropriate treatment for this plant is recognition at species rank and propose the name Ixia rivulicola. While its features recall other varieties of the I. rapunculoides complex, notably var. rapunculoides and var. flaccida, it is distinct in flower colour and fragrance, leaf number and shape, in the poorly developed corm tunics, and in the unique riparian habitat.
The second set of populations comprises short plants from the Klein Roggeveld with pale bluish grey flow ers. Assigned previously either to var. rapunculoides or to var. namaquana (Lewis 1962;De Vos 1999) the two lower leaves of the plant have expanded blades and a third sheathing leaf, flowers with a narrowly fun nel-shaped perianth tube 10-14 mm long, branchlets subtended by long, recurved, thread-like bracts and prophylls that are bifurcate, and unusual floral bracts with dry, often tom tips, the outer usually with five veins and the inner with three or four. The corms bear basal corm lets and have tunics composed of coarse fibres typical of Ixia rapunculoides. We treat these plants as the new spe cies, I. lacerata.

Key to Ixia rapunculoides and its close allies in subgenus Morphixia
Note: care must be taken in measuring floral parts of preserved specimens: depending on the method of drying, the perianth can shrink as much as 20%. Presence of stolons is difficult to establish as they are often left in the ground unless corms are removed with particular care. Filaments are measured from point of insertion on the perianth tube (although they are decurrent) to the base of the anther. The perianth tube is measured from the top of the ovary to the point at which the tepals separate from the upper portion of the tube.
la Foliage leaves linear-filiform, less than 2 mm wide when alive; midvein lying closer to abaxial margin but not evident when alive unless 7a Leaf blades of lower two or three leaves sword-shaped to falcate (rarely sublinear), mostly 8-15 mm wide (rarely 3-7 mm wide), usually less than one third as long as stem (longer if growing through bush); flowers ascending to horizontal; filaments 2-4 mm long; anthers 2.3-4.0 mm long: 8a Bracts and prophylls subtending branchlets bifurcate and attenuate, often directed downward distally; leaves 2( 1) with well-deve loped blades; perianth tube 10-12(-14) mm long; outer bracts with (4)5 major veins, 5-toothed, teeth attenuate ............. I. lacerata 8b Bracts and prophylls subtending branchlets short, obtuse to truncate, less than 2 mm long; leaves 2 or 3(4) with well-developed blades; perianth tube 5.5-8.0(-10.2) mm long; outer bracts with 3 major veins, shortly 3-toothed to bluntly 3-lobed: 9a Flowers blue-mauve, occasionally pale pink, with yellow to white throat; expanded foliage leaves 2(3); blades usually falcate; base of stem rarely with fibrous collar; perianth tube (5.5-)8.0-10 mm long; anthers 2.3-3.5 mm long ............. I. rapunculoides 9b Flowers pale pink with white throat; expanded foliage leaves 3(4); blades usually sword-shaped; base o f stem covered by col lar o f fibres; perianth tube 7.5-10.2 mm long; anthers 3. Plants mostly 150-400 mm high. Corm subglobose, mostly 14-18 mm diam., with tunics of firm, wiry fibres accumulating in dense mass, bearing small cormlets at base. Leaves (2)3(4), lower 2 lanceolate to falcate, mostly (3-)7-12 mm wide, ± one quarter to half as long as stem, margins thickened, plane or lightly crisped, uppermost leaf sheathing stem below spike, rarely unifa cial in upper 10-30 mm. Stem erect, nodding above, with several to many (up to 10) short, thread-like, ascending to horizontal branchlets subtended by obtuse bracts and prophylls ± l(-2 ) mm long. Main spike half-nodding, mostly 3-5-flowered, lateral spikes (1)2-5-flowered; bracts translucent sometimes tinged brown especially in fruit, outer with three dark veins, mostly 5-6 mm long, 3-toothed or bluntly 3-lobed, inner about as long, with two dark veins and forked apically. Flowers ± hori zontally oriented to half nodding, pale blue (mauve or blue-grey) with pale yellow cup, or pink with white cup. faintly rose-scented or evidently unscented; perianth tube 5.5-8.0(-10.0) mm long, widely funnel-shaped, flaring in upper 3-5 mm; tepals subequal, ovate, 11-16 x 4.5-6.0 mm, proximal ± 3 mm forming part of floral cup. spreading at right angles to tube distally. Stamens paral lel; filaments 2-3 mm long, inserted 2^4 mm above base of tube; anthers 2.3-3.5 mm long, upper half exserted from tube but included in floral cup, pale yellow. Style dividing shortly below anther tips, branches 0.7-1.2 mm long, not reaching anther tips. Flowering time: August to mid-September, sometimes July. Figure  Diagnosis and variation: Ixia rapunculoides, as cir cumscribed here, is fairly uniform across its range: it is recognized by the horizontally oriented, mostly blue flowers with a yellow cup (locally pale pink with a white cup), short perianth tube mostly 5.5-8.0 mm long, short filaments 2-3 mm long and anthers 2.3-3.5 mm long and half included in the perianth tube. In the vicinity of the Hantamsberg, the species co-occurs with I. robusta, another member of the complex, which was distinguished at the varietal level by Lewis (1962) by the presence of a short collar of fibres sheathing the base of the stem, and three or sometimes four, rather than the usual two leaves with expanded blades. These plants are restricted to dolerite outcrops along the eastern and northern end of the Hantamsberg and wherever we have seen them, typical /. rapunculoides grew nearby or among the I. robusta plants but was always shorter in stature and with grey-blue flowers, in marked contrast to the tall, pink-flowered I. robusta. The occasional presence of three, rather than the usual two expanded leaves, must be regarded as no more than part of the pattern of variation and not as evidence for intergradation with I. robusta. Leaf number occasion ally varies in typical I. rapunculoides, thus some speci mens of Lewis 5808 have three broad basal leaves (in two specimens the uppermost of these partly sheaths the stem) very like those of I. robusta', and all four specimens of De Vos 2561 have four leaves, the penultimate one partly to almost entirely sheathing the stem. Some plants from the northeastern end of the Hantamsberg (Moordenaarspoort and the Farm Vanrhynshoek) stand out in the species in their unusually narrow leaves (± 3 mm wide), narrow' tepals and single flowered branchlets (e.g. Snijman 2068) but at present seem merely to be depauperate rather than representing a separate race or genotype.
In order to fix the application of the name Ixia rapun culoides, the type of which is a painting in Redoute's Les Liliacees that is not, as far as we know, associated with preserved specimens and possibly of doubtful iden tity. we have designated an epitype. Plants from the Bokkeveld Escarpment near Nieuwoudtville match the illustration best and of the collections available from there, Lewis 5853 is representative of the species and distributed in several herbaria.
Flowers of Ixia rapunculoides have been shown to be pollinated by large-bodied Anthophorine bees, includ ing Anthophora diversipes and Pachymelus peringueyi (Goldblatt et al. 2000). We have confirmed this initial report at a second population of the species, south of Nieuwoudtville, where only female Anthophora diver sipes individuals were captured visiting the flowers. The reward for visiting bees is nectar, moderate amounts of which are secreted from septal nectaries and retained in the base of the perianth tube. Plants mostly 0.5-1.0 m high. Corm subglobose, mostly 14-18 mm diam., tunics of firm, wiry fibres accu mulating in dense mass, bearing small cormlets at base. Leaves usually 4(5 or 6), lower 3 or 4 lanceolate to subfalcate. mostly 8-14 mm wide, ± one quarter to one third as long as stem, margins thickened, usually plane, upper most 1 or 2 leaves sheathing stem below spike. Stem erect, sheathed below ground by collar of fibres, with up to 10 short, slender, ascending branchlets subtended by truncate-apiculate bracts and prophylls 0.5-1,0(-2.0) mm long. Main spike suberect. mostly 4-6-flowered. lat eral spikes ascending, mostly 2-6-flowered; bracts trans lucent becoming tinged brown at tips in fruit, outer with three dark veins, (6-)8-10 mm long, 3-toothed or bluntly 3-lobed. subequal or inner slightly longer, with two dark veins and forked apically. Flowers ± horizontally ori ented, pale pink with white cup, faintly acrid-metallicscented or apparently unscented; perianth tube 7.5-10.2 mm long, widely funnel-shaped, flaring in upper 4-5 mm; tepals subequal, narrowly ovate to subspathulate, (15-) 17-20 x 7-9 mm. proximal ± 3 mm forming part of floral cup. spreading at right angles to tube distally. Stamens parallel; filaments ± 4 mm long, inserted ± 3.5 mm above base o f tube; anthers 3.5^4.0 mm long, half exserted from tube but included in floral cup, yellow. Sty le dividing opposite middle third of anthers, branches 1.5-2.0 mm long, extending between anthers. Flowering time: late August to late September. Figure 1D-F. Distribution: locally in the western Karoo east and north of the Hantamsberg, on flats and lower slopes in heavy, dolerite clay, among dolerite boulders (Figure 3).
Diagnosis and variation: Ixia robusta is recognized by its tall stature, pink flowers with a white cup. and at least three (and often four) lanceolate basal leaves that are thick and leathery to almost succulent in texture. The corms, which are difficult to remove from the rocky ground, are large, and covered with wiry tunics, and the underground part of the stem is enclosed by a collar of fibres, a feature often not preserv ed when the deepseated corms are removed from the rocky ground.
Wherever we have seen Ixia robusta, typical I. rapun culoides has been growing nearby and sometimes among plants of I. robusta, never with any sign of intergrada tion. Ixia robusta is always taller, has 3 or 4 expanded leaf blades, and branches with more and larger, pink flowers. Associated with the larger flower, the anthers are ± 4 mm long, the perianth tube is 7.5-10.2 mm long and the style branches are 1.5-2.0 mm long, contrasting with anthers 2.3-3.5 mm long and style branches 0.7-1.0 mm long in I. rapunculoides. Lewis's comment that what she called var. robusta and var. rapunculoides favoured dif ferent habitats, the latter in flatter, more sandy situations, is not fully borne out by our observations. We could usually find I. rapunculoides growing among I. robusta plants, although I. rapunculoides also grew in flat, less rocky places where I. robusta did not occur. The mor phological differences between Lewis's var. rapunculoi des and var. robusta. combined with their svmpatry. syn chronous flowering and the absence of morphological intermediates, provides convincing evidence that they are separate species.
Although Ixia rapunculoides usually has bluemauve to grey-blue flowers, plants from west of the Hamtamsberg and near Loeriesfontein have a pale pink perianth, identically coloured to I. robusta and care must be taken not to confuse I. robusta with pink-flowered I. rapunculoides: the latter always has short, falcate leaves, usually only two with expanded blades, and lacks a collar of fibres around the stem base. Flower colour alone is not a consistent difference between the two species but where they co-occur, flower colour and scent always differ.  Plants 450-800 mm high. Corm subglobose, tunics of submembranous fibres soon disintegrating, bearing 1few sessile cormlets at base. Leaves 3(4 or 5), lower 2-4 narrowly sword-shaped to linear, 3-7 mm wide, mar gins moderately thickened, hyaline when dry, straight, usually ± one third to half as long as stem, uppermost leaf sheathing stem, penultimate leaf often sheathing in lower half. Stem with 3-5 straight lateral branchlets mostly held at ± 30° to main axis, filiform, bearing flow ers in upper half. Main spike mostly 2-or 3-flowered; lateral spikes 1-or 2-flowered; bracts translucent, outer with three dark veins, mostly ± 8 mm long, inner with two dark veins and forked apically. Flowers ascending to nearly horizontal, white with pale yellow cup, with strong fruity scent; perianth tube funnel-shaped, 8-9 mm long; tepals subequal, 17-18 x 5.0-7.5 mm, proximal 3-4 mm forming part of floral cup, spreading at right angles to tube distally. Stamens parallel; filaments 7>-A mm long, included, inserted ± 3 mm above base of tube; anthers 4.5-5.5 mm long, lower half included in tube. Style dividing opposite upper third of anthers, branches ± 0.8 mm long, extending between anthers. Capsules and seeds unknown. Flowering time: mainly October. Figure 4A-C.
Distribution: Ixia rivulicola is apparently restricted to the upper reaches of the Bo-Visrivier (Upper Fish River) on the Roggeveld Escarpment (Figure 3). First collected in 1981 by Harry Hall, this plant was not known to Lewis (1962) at the time that she completed her revision of Ixia, nor was it dealt with by De Vos (1999). Revisiting Hall's locality in the Roggeveld in October 2004 we found plants growing in shallow water close to the banks of the Visrivier near the entrance to Voelfontein Farm. Further downstream at the drift across the river at Noudrif Farm an extensive population was flowering particularly well where the reeds and cattails in the riverbed had been burned the past summer. The plants are tallest when growing in water among reeds but shorter when the surrounding vegetation is removed. The essentially aquatic habitat of this species is unique in the genus.
Diagnosis and variation: Ixia rivulicola is not sim ply a tall, white-flowered variant of I. rapunculoides, for the flowers are larger, with a tube 8-9 mm long, tepals 17-18 mm long, and anthers 4.5-5.5 mm long, com pared with the shorter tube, 5.5-8.0(-10.0) mm long in I. rapunculoides, tepals 11-16 mm long, and anthers 2.3-3.5 mm long. The ascending to horizontally oriented flowers are white with a pale yellow cup and are particu larly sweetly scented, the odour like that of Virgilia oroboides (Fabaceae), a combination of grenadilla fruit and violets. Typical blue-flowered I. rapunculoides grows on the slopes above the Visrivier and blooms in August and early September, and is in fruit by the time I. rivulicola comes into flower at the end of September or in early October. Ixia rivulicola responds well to cultivation, and survives outdoors in several degrees of freezing. Plants reproduce liberally by vegetative reproduction, unlike typical /. rapunculoides, although both produce cormlets at the base of the main corm.
Distribution: Ixia lacerata is restricted to slopes and flats in the Klein Roggeveld, which lies south of the main Roggeveld Escarpment, and the Koedoes Mountains to the west (Figure 3). Plants mostly grow on shales or mudstones of the Ecca System, and only occasionally on rocky, dolerite slopes in heavy clay, the habitat in which typical I. rapunculoides most frequently occurs.
Diagnosis and variation: until now the handful of specimens resembling typical Ixia rapunculoides from the Klein Roggeveld and Koedoes Mountains have been referred either to this taxon or to I. namaquana. Careful examination of flowers and bracts, however, shows that these plants differ consistently in several respects. The perianth tube is 10-14 mm long (vs 5.5-8(-10) mm in var. rapunculoides and (12-) 13-16(-18) mm in I. nama quana); the outer bracts usually have five major veins and the inner, two main and often two smaller veins, with the veins terminating in attenuate teeth; and the margins are dry and often somewhat tom. The bracts in particular are highly diagnostic, contrasting with I. rapunculoides and most other species of Ixia subgenus Morphixia, which have 1-or 3-veined outer bracts and 2-veined inner bracts, the veins terminating in short, acute or blunt teeth. In addition to the floral differences, the flowering stem has only 1-3 branchlets, each bear ing 1-3(4) flowers, and the branchlets are subtended by forked, thread-like bracts and prophylls that are spread ing or directed downward distally, unlike the short, trun cate and inconspicuous bracts in most other members of the I. rapunculoides complex, including I. namaquana and I. rapunculoides itself.
Ixia lacerata is poorly represented in herbaria and appears to have first been collected by the travellerexplorer William Burchell. who found the species on the Windheuwel in the Koedoes Mountains in July, 1811. Surprisingly, Burchell's collection attracted no botanical attention, although at the time the only Ixia known with short stamens and included filaments was I. rapuncu loides, described in 1816 and not then known from any wild locality.

Ixia sobolifera
Flowering time: August to mid-September.
Distribution: Ixia sobolifera extends from the Langberg and Kubiskou ranges northwest of Loeriesfontein in Northern Cape though the western Karoo and inte rior Western Cape to the Little Karoo ( Figure 5). More detailed distribution and habitat information is provided under the subspecies. Lewis (1962) in Ixia rapunculoides var. flaccida, I. sobolifera is rec- ognized by the combination of corms with soft papery (rather than fibrous) tunics, producing prominent, rib bon-like stolons from the base; a main spike bearing 2-5(-7), half to fully nodding flowers; filiform bracts and prophylls subtending the branchlets; pink, slate-blue or white flowers with a white or yellow cup; and a peri anth tube 6-10 mm long. We recognize three subspecies, each with a separate geographic range and minor distin guishing characters. Subsp. carnea from the Bokkeveld Plateau and its extensions, the Kubiskou and Langberg.. has a pink perianth with white cup. bracts somewhat dry and dark brown at the tips with age, and anthers marked with dark longitudinal lines; subsp. sobolifera, from the Klein Roggeveld and nearby, has slate blue flowers and particularly slender stems; and subsp. albiflora from the Little Karoo has white, intensely fragrant flowers.
Distribution: so far recorded only from the Klein Roggeveld and adjacent hills between the Witteberg in the south and the Komsberg and Verlate Kloof in the north ( Figure 5). Until now, subsp. sobolifera was hardly known, and we have found just three earlier collec tions, Compton 9272 and Snijman 539, both from Ngaap (Gwaap) Kop near Matjiesfontein. and Levyns 1368 from Verlate Kloof, in the several herbaria that we have consulted. Our field research in 2006, a year of ample rainfall in the Western Karoo, yielded small populations at five separate sites, from Komsberg Pass in the north to Memorial west of Matjiesfontein in the south. Additional collecting in the Koedoes Mtns and on the Roggeveld Escarpment will likely extend the known range of subsp. sobolifera.
Distribution: subsp. carnea is restricted to the north ern end of the range of Ixia sobolifera, where it has been recorded from a few sites on the Bokkeveld Plateau, mostly near Nieuwoudtv ille, on the Hantamsberg some 80 km to the east, and also on Kubiskou Mountain near Loeriesfontein and the Langberg to the west ( Figure 5). Plants are most often found on dolerite-derived clay or on shale slopes but northwest of Nieuwoudtville, near the Farm Biekoes, plants grow on sandy tillite soils among sandstone rocks. The species is fairly common in the rocky dolerite hills of the Nieuwoudtville Wildflower Reserve and in similar habitats to the south on the farms Glenlyon and Oorlogskloof.
Distribution: mainly restricted to the central Little Karoo, subsp. albiflora is known from just a handful of sites, mostly south and east of Oudtshoom, with one record from the northern foothills of the Swartberg, east of Prince Albert ( Figure 5). We also provisionally include here a collection from Montagu {Page s.n.), which is isolated some 150 km west of the remaining stations. The plants have white flowers, typical of albi flora, with a perianth tube ± 7.5 mm long and anthers ± 4 mm long, but the corm has finely fibrous tunics resem bling those of I. flaccida and shows no evidence of sto lons, critical in recognizing I. sobolifera. The bracts are ± 7 mm long, also consistent with /. sobolifera rather than I. flaccida which has slightly shorter bracts, typi cally ± 6 mm long. The locality of the Page collection, the Little Karoo, accords better with subsp. albiflora than 1. flaccida, which occurs to the west in the Olifants River Valley. Subsp. albiflora favours moist, south-fac ing slopes and grows in pockets of loamy soil on shale in renosterveld. The habitat for the collection from Montagu is from rocky slopes and cliffs, but the rock type is not recorded.
Diagnosis and variation: subsp. albiflora is distin guished from its sister subspecies mainly by the white, strongly carnation-scented flowers with a yellow cup, and also by the 4-or 5(-7)-flowered main spike, peri anth tube 6-7 mm long (perianth tube 9-10 mm long in subsp. carnea and 5-6 mm in subsp. sobolifera), and often broader leaves, 4-8 mm wide. Lewis (1962) included the two collections of Ixia sobolifera subsp. albiflora known to her {Page s.n. and Thorne s.n.) in I. rapunculoides var. flaccida (now I. flaccida), which occurs in the Olifants River Valley, well to the west. As we have discussed above, this is distinct from the stoloniferous I. sobolifera, which also has larger flowers with a longer perianth tube and longer anthers. Plants mostly 300-500 mm high. Corm 12-15 mm diam., with tunics of medium-textured to coarse fibres. Leaves 3, lower 2 sword-shaped to falcate, mostly 8-18 mm wide (particularly broad in the type), margins thick ened and hyaline, sometimes minutely crisped, usually ± one third as long as stem but sometimes much shorter, uppermost leaf sheathing stem. Stem with several short, filiform, twisted lateral branchlets, subtended by short bracts ± 2 mm long. Main spike erect, 2-or 3-flowered, lateral spikes 1-or 2-flowered; bracts translucent, outer with 3 dark veins, ± 10 mm long, with 3 short, sub equal teeth, inner with 2 dark veins and forked at apex. Flowers held ± horizontally, whitish to pale mauve, lilac, pale blue or pink with yellow cup, rim of cup marked with a band of short vertical lines, often flushed lilac to mauve outside, violet-scented or unscented; perianth tube 12-16(-18) mm long, narrowly funnel-shaped; tepals subequal, ovate, (12-) 14-17 x 5-7 mm (inner), 6-8 mm (outer), proximal 4-6 mm forming part of floral cup, spreading at right angles to tube in distal 10-12 mm. Stamens parallel; filaments erect, 4-5 mm long, inserted ± 6-10 mm above base of tube; anthers 3.5-4.5 mm long, upper half exserted from tube but included in floral cup. Style mostly dividing opposite upper third of anthers, branches ± 1 mm long, extending between anthers (in the type, dividing at anther tips with branches arching above anthers and 1.5 mm long). Capsules ovoid, ± 9 mm long. Seeds unknown. Flowering time: mainly August to mid-September, rarely October. Figure  6A-D.
Distribution: Ixia namaquana has a scattered dis tribution in Northern and Western Cape, and has been recorded from the Richtersveld and the high ground around Steinkopf in northern Namaqualand, from the southern Kamiesberg in central Namaqualand, and from the Bokkeveld Mtns and northern Cedarberg to the south (Figure 7). The disjunct distribution between the Steinkopf area of Namaqualand and the southern Kamiesberg is a pattern shared by Brunsvigia pulchra (Amaryllidaceae) and is probably related to the some what lower altitude and associated reduction in precipita tion of the country between these two high-lying points. Another species pair with this distribution is Hessea pilosula (Amaryllidaceae) from the Steinkopf area and its sister species H. incana from the Kamiesberg. There are no significant differences between the northern Namaqualand populations of I. namaquana and those from the south except for the bright pink flower colour (rather than pale mauve to blue) in the type form, which occurs on clay soils. In the Kamiesberg, I. namaquana occurs on granitic gravel but in the Bokkeveld Mtns, Gifberg and Cedarberg it grows in sands derived from Cape Sandstone formations. Plants that we assign to I. namaquana from the Wiedouw River (Lewis 1980), south of Vanrhynsdorp, probably grow on loam among limestone outcrops, the main formation there but we have not seen the species at this site ourselves.
Diagnosis and variation: Ixia namaquana is readily distinguished from other members of the I. rapunculoi des complex by the longer perianth tube, mostly 13-16 mm, short branchlets bearing 1 or 2 flowers each, and conn with tunics of coarse, netted fibres. The flowers are relatively large with tepals mostly 14-16 mm long and anthers 3.5r4.5 mm long. The proximal part of the tepals is directed forward, forming part of the floral cup. which fully encloses the stamens. The flowers are usually sweetly scented but are apparently unscented in plants from Vanrhvn's Pass and Nieuwoudtville in the Bokkeveld Mtns. Pink-flowered plants with a peri anth tube 16-22 mm long from the Hex River Pass and nearby that were identified as I. namaquana by De Vos (1999) and others, are here referred to the new I. oxalidiflora. This species can be distinguished from I. nama quana by the longer perianth tube, suberect rather than horizontally spreading flowers, fully included anthers, and perhaps most significant, by the corms with papery to finely fibrous tunics and bearing ribbon-like stolons.
Plants from a seasonally wet site near Leliefontein in the Kamiesberg, referred to Ixia namaquana by De Vos in various herbaria, differ in their later flowering time, October and November, and in having a white peri anth with tepals pale blue at the tips, filaments shortly exserted, a perianth tube ± 7 mm long, and narrow leaves. They are obviously misplaced in Ixia namaquana and apparently represent a late-flowering ecotype of I. latifolia var. ramulosa from moist habitats.  Plants 200-350 mm high. Corm subglobose, tunics of medium-textured, netted fibres, bearing 1-few sessile, basal cormlets and rarely a short stolon. Leaves 3(4), lower 2(3) sword-shaped to linear, 4-9 mm wide, mar gins hardly thickened, straight, usually about half as long as stem, uppermost leaf sheathing stem. Stem erect, slen der-filiform, with 4-6 loosely twisted, filiform branch lets up to 35 mm long, spreading at right angles to main axis, subtended by minute, acute bracts and prophylls ± 1 mm long. Main spike 2-^4-flowered, lateral spikes 1-3flowered; bracts translucent light brown, outer with three dark veins, mostly 5-6 mm long, inner as long or slightly longer than outer, with two dark veins and forked at apex. Flowers horizontally oriented, white or pale blue with white to greenish cup, apparently unscented; peri anth tube funnel-shaped, 7-9 mm long; tepals subequal, ± 14.0 x 3.5-4.5 mm, proximal 3-4 mm forming part of floral cup, spreading at right angles to cup distally. Stamens parallel; filaments 2.5-3.0 mm long, included in tube, inserted 3^4 mm above base of tube; anthers ± 2.5 mm long, lower half included in tube. Style divid ing at mid-anther level, branches ± 1 mm long, extend ing between anthers. Capsules and seeds unknown. Flowering time: late July to early September.
Distribution: restricted to the Olifants River Valley and nearby, Ixia flaccida grows on seasonally moist, south-facing, sandy, or sometimes light, loamy clay slopes, among rocks usually in the shade of shrubs and small trees (Figure 8).
Diagnosis and variation: the flowers of Ixia flaccida are relatively small, with tepals ± 14 x 3.4-4.5 mm, peri anth tube 7-9 mm long, and anthers ± 2.5 mm long. The tunics of medium-textured to fine fibres and the ± linear or narrowly sword-shaped, soft-textured leaves, as well as the white or pale blue flowers, readily distinguish the species from typical I. rapunculoides, which has blue to blue-mauve or sometimes pink flowers with a yellow cup.
In addition to the Olifants River Valley collections that closely match the type of I. rapunculoides var.flac- albiflora. Both differ from var. flaccida in their corms, which have soft-textured, short-lived tunics and produce long, ribbon-like stolons. They also differ in some flo ral features, notably the longer bracts, 6-10 mm long (vs 5-6 mm long in I. flaccida) and, in subsp. carnea, a lon ger perianth tube 9-10 mm long and longer anthers 3^4 mm long.
Distribution: Ixia oxalidiflora is currently known from two sites, the top of Hex River Pass, and near Tweedside Station, ± 60 km to the east (Figure 8). It favours southtrending slopes and clay or loamy soils, where it typi cally grows wedged in crevices o f shale outcrops. The purple-pink flowers with a white cup (the latter some times described as pale yellow) closely resemble in shape and colouring those of co-occurring Oxalis heterophylla (Oxalidaceae) which blooms at the same time.
Diagnosis and variation: the earliest record of the spe cies that we have traced is one made by F.A. Rogers in 1915. Early collections of Ixia oxalidiflora were referred to I. rapunculoides var. namaquana without comment by Lewis (1962) and De Vos (1999), although these speci mens equally resemble similarly large-flowered I. pauciflora G.J.Lewis in general aspect. Careful examination of the flowers, however, shows that they differ markedly from both taxa in having the stamens fully included in the perianth tube, with the anther tips reaching 0.2-0.5 mm below the tube apex, and a perianth tube 16-22 mm long. In I. namaquana, only the filaments and bases of the anthers are included in the tube, which is shorter, mostly 13-16 mm long, and in /. pauciflora, the tips of the filaments are typically exserted from the tube. The corms, which have soft-textured tunics and long stolons, resemble neither species, and the lanceolate, soft-tex tured leaves are quite different to the linear, very nar row, firm leaves of I. pauciflora or the broader, leathery leaves of I. namaquana. Ixia oxalidiflora is thus rec ognized by the combination of fully included stamens, a perianth tube 16-22 mm long, soft-textured leaves, 6-12(-15) mm wide, and the small corms that bear long, flat stolons. The flowers are also held upright, a feature only apparent when plants are seen alive, thus unlike the horizontal flowers of I. namaquana and I. pauciflora.
Distribution: Ixia divaricata is scattered through the mountains of the southwestern Cape, from Elandskloof, east of Citrusdal in the southwestern Cedarberg, through the Cold Bokkeveld to Ceres and Tulbagh ( Figure 10). Plants favour seasonally wet, stony sandstone flats and rocky sites.
Diagnosis and variation: Ixia divaricata is unmistak able in its unusual, divaricate branching pattern, with dis tinctive, stiff, straight lateral branches held at angles of 30-50° to the main axis and bearing mostly 3-5 flowers crowded in the distal half. The white or pink, or really purple flowers are held erect, with the tepals loosely spreading but not fully patent. Why this plant was treated as var. subpendula of I. rapunculoides is puzzling for not only is it distinctive in its branching pattern but the flow ers are upright, not horizontal, and the anthers, 4.0-5.5 mm long, are usually fully exserted from the tube or rarely have their bases included. Specimens with anthers exserted 1-2 mm were often included in I. latifolia by De Vos (1999) but true /. latifolia has larger, deep pink flowers with a perianth tube 14-17 mm long, filaments ± 10 mm long, and broad basal leaves, mostly 10-18 mm wide. We regard var. subpendula as a separate species, which we name I. divaricata for the divaricate branch ing pattern. Although valid synonyms at varietal rank exist for the species (see above), we have chosen the new name divaricata at species rank. The name I. angustifolia is preoccupied in the genus by I. angustifolia (Andrews) Klatt (1895), a later synonym of I. monadelpha (Lewis 1962). We prefer not to use the epithets sub pendula (based on I. latifolia var. subpendula) or rigida (based on I. rapunculoides var. rigida) for the species because the former is misleading and the latter is based on an atypical type specimen.
The immediate relationships of Ixia divaricata are most likely with I. latifolia. Confusion with that spe cies is due in part to some longer-tubed collections with the tube 12-14 mm long (e.g. De Vos 2693 and Leipoldt 4070) that also have relatively short, broad leaves, up to 18 mm wide (a feature of I. latifolia) but fairly slen der branches (characteristic o f I. divaricata). These are almost certainly hybrids between the two species, which grow within sight of one another, although always on different soils, I. latifolia favouring heavy clay or loam.
Ixia divaricata was associated historically with I. latifolia, and one of the two specimens cited by Baker (1892) under var. longifolia of I. scariosa, the name by which I. latifolia was then known (the other specimen is I. orientalis and is the lectotype of the name) is I. divaricata. Although the type of I. latifolia var. angus tifolia, from the Witzenberg is I. divaricata, several other specimens assigned here by Lewis (1962) and De Vos (1999) are a different species, I. monticola, which we describe elsewhere (Goldblatt & Manning in prep.). Although these two taxa have been confused in the lit erature they may not be particularly closely allied. The most important differences are that I. monticola has a corm with membranous, non-accumulating tunics, few or no branches, and four or five leaves. I. divaricata, in contrast, has fibrous tunics of medium texture and a dis tinctive branching pattern, and plants from high eleva tions typically have two or three leaves in most speci mens, although sometimes more in those from middle elevations.
Another synonym may be Ixia stellata (Andrews) Klatt (1882), based on I. capitata var. stellata Andrews (1802), but the quality and detail in the painting on which the taxon must be based makes a firm identifi cation impossible and no preserved material has been found at K and BM, the herbaria where they would most likely have been preserved.
We also include the type of Ixia rapunculoides var. rigida from the Hex River Valley in /. divaricata, which it closely resembles in the relatively long straight branches with distally crowded flowers, unlike the remaining specimens assigned to var. rigida, which have shorter, flexuose, horizontal branchlets and flow ers evenly spaced along their length. This collection is the only record from the Hex River Valley, where we assume it no longer occurs as a result of intensive viti culture there. Ixia divaricata is still common in the Cold Bokkeveld, where it is found in sites presently protected from agriculture, although below Gydo Pass in the Warm Bokkeveld, it is now threatened by the expansion of orchards from Ceres and Prince Alfred's Hamlet to the foot of the Pass.
Provisionally we also include in Ixia divaricata plants from upper Moraine Kloof in the Hex River Mountains (Helme 2864), flowering in January. They have a simi lar branching pattern but somewhat smaller, purple flow ers with a tube ± 5 mm long, tepals ± 10 mm long, and anthers ± 3 .2 mm long. The style divides at the top of the anthers and has branches ± 1.3 mm long that extend above the anther tips, whereas in typical I. divaricata the style divides opposite the middle of the anthers. Unfortunately, the specimens lack corms and leaves without which a fair decision about the status of this plant, flowering some two months after other collections of I. divaricata, cannot be made.
Distribution: Ixia contorta is centred in the Cold Bokkeveld but extends north to the Cedarberg and east to Touws River, growing in well-drained, sandy ground ( Figure 12). (1962) as Ixia rapunculoides var. rigida, this plant from the inte rior Western Cape mountains is recognized among those species with included filaments by the markedly flexuose spikes and by the main axis strongly flexed below the first flower, thus inclined to nearly horizontal. The pur ple flowers (also sometimes described as blue or lilac) have a yellow cup sometimes edged in dark purple, and are held upright on the spikes, unique among the species with included filaments and anthers held partly within the perianth tube. The perianth tube is 9-13(-14) mm long and the tepals are 9-12 mm long. The very different appearance of the flowers suggests that these plants are not immediately allied to I. rapunculoides and we not only regard them as representing a separate species, but question their relationship to the complex. We describe this plant here as I. contorta, so named for the twisted lateral branchlets and flexuose spike, a feature par ticularly well-developed in this species. The new name is required because the type of var. rigida is I. divari cata (see synonymy of the latter species). Ixia contorta can be distinguished from I. divaricata by the short, twisted branchlets, quite different from the long, straight branches of I. divaricata, the flowers of which are white to pale pink, rarely purple, and the filaments are usually shortly excluded from the tube.

Diagnosis and variation: treated by Lewis
Plants are usually fairly short, seldom exceeding 150 mm, but collections from Gydo Pass (e.g. Leipoldt 3023) are taller, some reaching 300 mm. Typical Ixia contorta has also been collected on Gydo Pass, which indicates to us that the taller plants with narrow leaves are not sim ply a local variant of the species. We suspect they are hybrids, the result of crossing with I. divaricata, which also grows there. The longer, narrower leaves, paler flower colour and slightly larger perianth in these plants may also reflect the influence of genes of I. divaricata.
Plants from the north of the range near Alberta Farm, and in the Matjiesrivier Valley in the Cedarberg stand out in having a longer perianth tube, 12-13 mm in Nanni 81, and 12-14 mm in Gillett 4108. The latter plants, also exceptional in their slightly wider perianth tube, were referred to I. rapunculoides var. rigida by Lewis (1962) but evidently puzzled De Vos (1999), who iden tified a duplicate of the Gillett collection at PRE as I. rapunculoides var. namaquana (i.e. /. namaquana). The latter taxon differs particularly in having flowers facing to the side, rather than upright, longer floral bracts, and a straight and erect rather than flexuose, inclined spike axis. The broad leaves, up to 10 mm in one plant of the Gillett collection, are exceptional for I. contorta and it remains possible that the gathering represents a novelty. Additional collections of I. contorta from the north of its range are needed to better assess the variation in the spe cies.