The genus Erica (Ericaceae) in southern Africa: taxonomic notes 2

This is a continuation in the series of notes from Bothalia 32 (2002) reflecting the status of species of Erica L. recognized in the Compton Herbarium. These notes cover the 43 species currently included in the section  Evanthe of Flora capensis . Five new species, E. ceraria . E. croceovirens . E. gerhardii . E. prolata and E. viridimontana and six new sub­species are described.


INTRODUCTION
The first paper in this series (Oliver & Oliver 2002a) dealt with the large-flowered species with tubular corol las that are borne near the ends of extremely reduced (almost invisible), non-leafy lateral branchlets which are then aggregated into dense, spike-like synflorescences. The term 'axillary' was applied by previous workers to these inflorescences. The next set of species with long tubular corollas has the flowers also borne near the ends Franschhoek. It is most closely related to the new species, E. gerhardii under which the differences are discussed. A small flowered collection, Esterhuysen 9614, from Winterberg in the Wemmershoek/Du Toitskloof area is placed here with some uncertainty. It needs further inves tigation before any definite taxonomic position can be assigned to it. woody, valves laterally splitting to base, septa remaining attached to columella. Seeds ellipsoid, ± 1 x 0.7 mm, shallowly alveolate, brown; testa cells rectangular to somewhat square, ± 100 x 100 ^m, anticlinal walls jigsawed, inner periclinal wall with numerous small pits. Figure 2.
Diagnostic features: no noticeably spike-like synflor escence, inflorescences more scattered; flowers 1-or 2nate at ends of short to long, leafy lateral branches, most ly erect to suberect; bract and bracteoles small ovate; sepals relatively long lanceolate; corolla greenish yellow, slightly 4-angled; anther appendages long with numer ous upward-pointing short hairs.
This new species falls within the E. patersonii-sacciflora-nana alliance with its 4-nate leaves, hard yellowgreen corolla, long anther appendages and similar ovary and fruit (Figure 1). It appears to be most similar to E. sacciflora due to the few-flowered inflorescences, and the small and more remote bract and bracteoles. The lat ter differs in having spike-like synflorescences with flowers spreading to pendulous, shorter, broader sepals, ± 3.5 x 3.5 mm (not 5 x 2 mm), and shorter leaves, 4-5 mm long (not 8-10 mm) and hairy stems. E. sacciflora occurs some 60 km to the northwest in the Franschhoek Mountains.
Erica gerhardii is very localized on the mountains above Stanford: we saw only 24 plants in one population ( Figure 3). These grew on a steep southeast-facing slope in very old, moist fynbos vegetation. The mountain was subsequently burnt in the summer of 2001. The most noticeable feature of the plants in the wild was the growth form-tall, erect, very sparsely branched, with the leaves confined to the uppermost parts and the erect to suberect, greenish yellow flowers.
The species was first collected by Niven two hundred years ago and then turned up again by Gerhard Kirsten , the co-author of the finely illustrated book, Ericas o f South Africa , who was unable to be with us when the type material was col lected due to illness, but gave us directions to find the population. We name this species in honour of his dedi cation to the collection and study of Erica over a period of 25 years. He is also commemorated in E. kirstenii E.G.H.Oliv. (Oliver & Oliver 2000a).
Note: the larger plate in Andrews' Coloured engravings o f heaths, t. 235, must have been published later than 1822 since in the text accompanying the plate he cites plants he had seen in 1822, but the vol ume is dated 1805-the publication cannot be dated accurately (Cleevely & Oliver 2002). See discussion under E. colorans (63) regarding dating these works.
Diagnostic features: synflorescence loose to dense, flowers 2^4-nate on short, leafy side branches; bract and bracteoles long; sepals relatively short, lanceolate; corol la yellow and orange-red, appendages long. The intemodes are longer than in the related species, ± 1.5-4.0 mm long and are finely hairy.
The plate of Andrews clearly does not depict the plant that was until now known as E. foliacea-it is now being described as E. ceraria. His protologue notes-'flowers nearly an inch long (± 25 mm), yellow, and transparent'. The many collections currently recognized under this name have flowers which are ± 18 mm long, yellowish green and very hard, smooth and wax-like in texture, not the pure yellow with longitudinal stripes/ridges and do not have the ovary with bulbous upper lobes, as illustrat ed by Andrews.  had only one of these collections, Bolus 6870, to which Andrews' name was applied thus setting the identification of all subse quent collections. All this material is here described as £. ceraria (38.3).
Andrews' iconotype matches reasonably well the few collections that, until now, have been identified as E. foliacea var fulgens. This latter name in fact refers to a distinct variant based solely on the density of the synflor escences, the flowers being identical in all the collections of the species. These variants are recognized here at sub specific level. 38a. subsp. foliacea Diagnostic features: loose synflorescence with 2-4 flowers on occasional leafy side branches; corolla entire ly yellow or yellow below and orange-red above where facing the sun (in wild-collected material).
The typical subspecies is known only from a few col lections on the Paardeberg west of Houwhoek. Bot River. This material has flowers shorter than Andrews' 25 mm in the range 15-20 mm, all of them do not have the sin gle colour of the iconotype. Andrews (± 1824) shows the ovary with exaggerated thickened apical processes and these are present in the Middelmann collections from Honingklip. These are very similar to the bulges occur ring in E. curx iflora (45).
Diagnostic features: long, dense, spike-like synflores cences on long main stems; flowers 4-nate on short lat eral branchlets borne at every node; corolla red-orange and yellow. The main branches appear to be fewer and much longer with longer intemodes (2-4 mm).
This subspecies is apparently extinct, having been last collected by Guthrie/Grisbrook/Bolus and Schlechter in the 1890s and never again despite searches in the area by ourselves. It must have been a very striking plant with its 'very pronounced red-orange and yellow flowers' (noted by Guthrie). Klotzsch gave the diagnostic character as 1 coroll is lateritis' (with corollas brick red like roof tiles) that he must have seen on the recently collected Ecklon material since Ecklon never recorded colour on his col lections.
The locality must have been on the moist southern side of Sir Lowry's Pass (Gantouws Pass) where Grietjiesgat was a known stop-off point for hungry, thirsty travellers after the arduous ascent up the pass. This whole area has been under alien pine plantations for many years. Guthrie must have travelled that route many times to and from his family home at Caledon to which he had retired from Cape Town and Bolus went many times to the Houw Hock estate which was owned by Guthrie's brother-in-law, C.S. Grisbrook, where they met and discussed plants, especially Erica (L. Bolus pers. comm. 1963).
The collection by Schlechter as 'Koue Bokkeveld' and the excellent one by MacOwan as 'Zwartebergen, Caledon' are certainly incorrect localities. MacOwan collected several species from that locality which we have found to be incorrect (see E. macowanii No. 57).
This species is the taxon currently well known as 'E.foli-acea\ It is closely allied to E. foliacea, E. nana and E. galpinii but is the only one with 5-nate flowers (3-5-nate), slightly prognathous anthers and glabrous stems. The calyx is also hidden by the imbricate leaves and the greenish yel low corolla is very hard and wax-like in texture and not transparent nor ridged. E. foliacea has hairy stems. In E. foli acea subsp.fulgens, nearly every node on the main branch es bears a short, secondary branchlet, terminating in a flor escence and is therefore very' spike-like, as in E. galpinii.  had a single unlocalized spec imen of this species w hich they placed under E. foliacea (Bolus 6X70). This was the first collection of E. ceraria.
The name is derived from cerarius, of wax, waxy (Latin), in allusion to the hard, wax-like texture of the flowers.
This species is confined to the rocky peaks and ridges around Kogelberg Peak down the west side of the Biosphere Reserve to above Betty's Bay (Figure 3).
The plants are low, compact, much-branched, spreading shrublets, covered with numerous inflorescences. They grow hanging over rocks. This habit is the most distinctive in this group of species. The flowers when young are greenish yellow but soon turn bright yellow at anthesis.
A chance hybrid E. nana x E. patersonii occurred in the Harold Porter National Botanical Garden at Betty's Bay and has been subsequently propagated and given the cultivar name 'Gengold'. It has slightly larger flowers than E. nana and is more floriferous. Subsequent backcrosses onto E. nana have been made in Germany (H. Kramer pers. comm.) and these have even longer flowers.

NOTE ON ORDER OF REMAINING SPECIES
The numbering order given by  and added to by  does not reflect the correct relationships between the remainder of the species in this paper. We have deviated here from this order and placed the species in what we postulate are their more natural alliances. The original numbers have been retained for cross-referencing, only the order is changed here.

ERICA CURV1FLORA-PERSP1CUA GROUP
This group of species is characterized by 4-nate leaves, loose to dense, leafy synflorescences, a long, tubular, hairy corolla (except E. ignita) with spreading recurved lobes, short pedicel (except in E. conspicua), anthers lacking appendages or sometimes possessing the remnants of appendages, ovary 4 8-locular, mostly obovoid and a capsule opening widely ( Figure 5).  (Figure 6). These species are, in our opinion, not the only ones in this alliance, since there are numerous small-flowered species in the section Ephebus which are clearly closely related to several of the above species.
Erica curviflora is by far the most widespread species occurring from the Gifberg in the north to the Cape Peninsula and eastwards as far as Grahamstown. All the other species have much more limited ranges and in some cases are very localized in the region between the Hottentots Holland Mtns and Klein River Mtns.
Several species were described by Andrews in his Coloured engravings o f heaths (1794-± 1830) using cultivated material that bore flowers on short, lateral, leafy branches arranged in loose to dense, spike-like synflorescences and had tubular, hairy corollas that were pink and white. These look similar to the form and arrangement in currently recognized species such as E. perspicua, E. macowanii, E. leucotrachela and E. latitu ba known from the wild. With the specific differences based among others on microcharacters of ovary com plement and seed structure that are not shown in Andrews' drawings, it is not possible to identify them with any certainty. Also, with numerous cultivated vari ants and hybrids being raised in England at the time (Nelson & Oliver 2004), we are very hesitant to ascribe these names to currently known wild populations. We have therefore decided to regard them as cultivars or arti ficially raised hybrids. Among these are Andrews' E. linnaea, I.e.: t. 106 (1801) Diagnostic features: dense, short to long synflores cence; flowers 1 -or 2-nate on ends of short, leafy side branchlets; bract and bracteoles minute; sepals small, ovate; corolla very finely hairy, cream to salmon/peach pink; ovary 4-locular ( Figure 6C).
The species is confined to small seepage areas at the heads of kloofs along the western end o f the Riviersonderend Mountains. Baker overlooked E. xan thina when describing his species.  Diagnostic features: a very dense, long, spike-like synflorescence; flowers mostly 4-nate on ends of short, leafy side branchlets; sepals short, ovate with elongate, acute apex; corolla glabrous, dull brick-red flushed with orange; ovary (5)6(7)-locular ( Figures 5B; 6A).
This striking species is highly restricted in its distri bution, being confined to a few steep slopes near the summit of the Riviersonderend Mountains above the town. 42. E. xpa/lens Andrews, Coloured engravings of heaths: t. 194 (1806); Benth.: 635 (1839); Guthrie & Bolus: 70 (1905); Dulfer: 43 (1965). Type: I.e.: t. 194. This taxon is regarded as a hybrid of garden origin in England. No material matching Andrews' plate has been located in the wild in South Africa.
Erica bibax occurs in the area between Kogelberg and Sir Lowry's Pass on streambanks and in seeps.  Diagnostic features: flowers 1^-nate at ends of short, lateral branchlets arranged in a dense, spike-like inflores cence, 40-100 mm long; corolla very finely hairy, plain yellow or bicoloured (red with yellow apex); sepals elon gate, lanceolate; anthers dorsally attached near base, with small decurrent appendages on apex of filament; ovary 4(-6)-locular; leaves hard in texture ( Figure 6G).
The localities given for the syntypes are clearly erro neous information derived from the flower sellers. It is pos tulated that this was done purposely to protect their sources.

45.
E. curviflora Species plantarum, edn 1,1: 354 (1753); Benth.: 633 (1839); Guthrie & Bolus: 71 (1905); Dulfer: 43 (1965 Diagnostic features: flowers 1-nate on leafy lateral branches in a loose arrangement up main branches; corolla mostly hairy with long, distinct hairs or glabrous, pink to orange to red, spreading to slightly recurved; sepals varying from 0.4-1.2 mm long; anthers manifest to just exserted, dorsally attached, with appendages dis tinct and short or just visible as remnants on apex of fila ment; ovary 4-[8]-locular, emarginate, with apical part having distinctive enlarged callous-like bosses ( Figures  5C; 6N). This is the most widespread and variable of the species in this group, occurring from the Gifberg near Vanrhynsdorp, throughout the Cape Floral Region as far east as Grahamstown. There is much variation in the size, colour and indumentum of the flowers and in the size of the leaves. In corolla indumentum there are some distinct variants confined to certain localities but also variable within a locality. We have found no reliable characters that can be used to subdivide the species. The most distinctive variant is one with very long leaves (up to 18 mm) which are covered with very long simple hairs. It occurs in the Kogelberg area and surprisingly also in the mountains above George. Diagnostic features: similar to E. curviflora but corolla bright yellow and more sparsely hairy with shaggy hairs; ovary 8-locular, broadly obovoid with no apical bosses.
The typical subspecies occurs in the mountains from the Houw Hoek Mtns, the Hottentots Holland to the Franschhoek Pass area.  This new subspecies is restricted to the Groenlandberge which are just east of Viljoen's Pass and the populations of the typical subspecies. A single collection, Stehle 272, from just east of Sir Lowry's Pass is west of the ranges of both subspecies. The subspecies is very similar to E. chrysocodon (141), a highly restricted endemic near Franschhoek Pass, but that has 4-celled ovaries. Further investigations, including molecular analyses need to be done to assess their relationship. Diagnostic features: flowers l(-4)-nate on occasional short, lateral branches; sepals always glabrous, hard and cartilaginous, broadly ovate; corolla with loose, long, sparse hairs; anthers prognathous at base (chin and nose).
Two subspecies are recognized based on the corolla length and colour. Both share the distinctive sepals of the species.
Subsp. conspicua has the longest flowers among the tubularflowered species-up to 36 mm long, and is second only to E. junonia var. junonia in the genus with its flowers 40-50 mm long. In the Cold Bokkeveld where this subspecies grows in seeps with the similar looking E. curviflora, a few plants of obvious hybrid origin have been recorded Other 11518 (NBG).
The subspecies occurs in the Franschhoek Mtns, near Wolesley and in the Cold Bokkeveld.  Diagnostic features: corolla 8-15 mm long, pale pink ( Figure 6P).
At first investigation, the material cited below seemed to be an obvious new species, but in the process of writing up the taxa for these notes it was decided to regard it as a subspecies of E. conspicua, the sepals and prognathous anthers having the distinctive features of the species.
The subspecies is known only from two small areas, the Keeromsberg, with shorter flowers (8-10 mm) and the Langeberg above Robertson, with longer flowers (13-15 mm). Diagnostic features: flowers erect-spreading; corolla with broad tube, ± 22-25 mm long, with no subapical bulge and large, more spreading lobes, yellow or red. w ith yellow or white apical portion; bract, bracteoles and sep als relatively large and broad; ovary 8-locular, seeds al veolate, finely pitted, without small scales; leaves large, ± 1 mm broad, no scale-like leaves below the flowers.
This forms a complex with the following six species and their subspecies and only careful examination and dissection will provide correct identifications. This species complex should be re-evaluated when molecular studies have been undertaken of all the relevant popula tions in order to ascertain the true relationships of all the taxa. The delimitation provided here is the best we can find in the currently available material based solely on morphological grounds. Diagnostic features: hairs on corolla fine; sepals ellip tic-ovate, ± 9 mm long ( Figure 6K).
The typical subspecies occurs at higher altitudes on most of the mountains in the Kogelberg Biosphere Reserve.  Diagnostic features: hairs on corolla longer and shag gy; sepals narrow-lanceolate, ± 5 mm long.
This taxon is confined to the upper reaches of the Klein River Mtns above Hermanus where two closely related taxa also occur, E. latituba at higher altitudes and E. perspicua subsp. latifolia on the flats. The types of both subspecies are reported to have come from the Swartberg at Caledon where the species has never been recorded. These localities are therefore suspect.  Dulfer: 48 (1965). Type: Caledon Div., in large colonies usually above 2000 ft. on steep, marshy slopes on the seaward side of all major peaks from Rooiels to Palmiet River. Not found below the mist of cloud belt characteristic of this district. Baker 1415 (BOL, holo.!; BM!, NBG!. PRE!, W).
Note: no precise locality was cited for the holotype. only general dis tributional details for the holotype and the five paratypes given above.
Diagnostic features: flowers 16-18 mm long with bulbous subapical zone; bract-like leaves usually on the flowering branchlet just below the flower; seeds with scale-like flaps and with the normal small pits coalescing into long, snake-like pits; corolla dark purplish/cerise pink with white apical portion, hairy ( Figure 6F).    Diagnostic features: flowers on very short, lateral branchlets arranged in loose, long, spike-like synflores cences; corolla short. ± 12 mm long, pale cream, with very fine short hairs; bract, bracteoles and sepals small, broad-ovate; ovary 4-locular ( Figure 6D).
The species is divided into two subspecies based on differences in the corolla and seed.
The typical subspecies was first recorded at the Caledon Wildflower Show in 1991 with no reliable local ity details. Then it appeared at the Flora 93 Wildflower Show in Cape Town with details provided to us by the supplier, Mr B. Morkel of Vyeboom and again at the Caledon Show due to the collecting of Mr Pitte who informed us of the exact locality.
It is confined to the summit regions of the Groenlandberg east of Grabouw (Figure 8), hence the specific epithet-viridis, green, montanus, of mountains (Latin); Groenlandherg (Afrikaans), green-land-mountain. Unfortunately all the range was burnt off in December 1999 and we will have to wait a few more years for the plants to grow to full maturity. Diagnostic features: corolla bright crimson, 22-30 x 5 mm; seeds with large, round pits and no scales, shiny, pink-brown.
This taxon was last collected in 1948 by Thomas Stokoe. Several attempts to re-collect it have not been successful despite the reasonably small but rugged area of Somerset Sneeukop. Unfortunately several fires in rather close succession have ravaged the Hottentots Holland Mtns during the last 10 years.
In the Cape herbaria the material cited below was identified as E. bibax, E. macowanii and E. leucotra chela. It has the flower colour of E. leucotrachela and general facies of that species but does not have the very characteristic seeds with their scales and slit-like pits which are also found in subsp. viridimontana. We have thus placed this material as a subspecies of E. viridimon tana rather than in E. leucotrachela.
It is confined to the highest reaches of the Hottentots Holland Mtns, mainly on Somerset Sneeukop (Figure 8), hence the epithet, nix/nivis, snow, -cola, dweller (Latin). Diagnostic features: flowers (l)2^4-nate at ends of short, leafy, lateral branches, spreading to subpendulous, ± 16 mm long, without small, bract-like leaves below the inflorescence; sepals long and narrow, leaf-like, from a dilated, pinkish, semitransparent, small basal portion hardly visible; corolla with stiffish scattered hairs, lobes erect, mouth not widened; ovary 5-8-locular; seeds with slight scales, pits small not coalescing; no bract-like leaves below the flowers (Figure 9). This species belongs in the E. curviflora/perspicua complex due to the 4-nate leaves, hairy corolla, anther shape and ovary details of shape and variable number of locules. Unlike most of the group with 1-nate flowers, this has (1 )2-4-nate flowers. The shape of the sepals is similar to E. curviflora which, however, always has cal losities on the ovary, and to E. perspicua subsp. latifolia which has a corolla with an open mouth. Both of these species have the 1-nate flowers. The seeds with small scales are similar to those of E. leucotrachela but the pits do not coalesce as in the species. E. leucotrachela also differs in having small, bract-like leaves on the stems just below the inflorescence.
The species was known only from the unlocalized type collection until our studies and was misunderstood by both L. Bolus and Dulfer-the former likened it to E. colorans and E. glandulosa. We therefore include here detailed drawings of the species (Figure 9). The material collected in the latter half of the last century clearly con stituted a distinct species that we were planning to describe, until examination of the type and only collec tion of E. latituba revealed its true identity. Fortunately, a seed found in the flower dissected, showed the charac teristic testa cells. A figure is published here to highlight the existence of this taxon that was otherwise very poor ly understood. The material figured by Andrews and named by him as E. linnaeoides {Coloured engravings o f heaths: t. 107, 1805) looks very much like this species in general facies. However, the sepals are shown as lanceolate, not the dis tinctive linear from a small dilated base, and the distinc tive seeds are not illustrated. Bentham (1839) suggested this was a hybrid of garden origin. Diagnostic features', corolla finely but perceptibly hairy with narrow tube, ± 22 mm long, having no distinct subapical bulge, white to pink with white tips; flowers erectspreading; leaves small, ± 4.5 x 0.5 mm ( Figure 5D).
The species forms a variable complex with E. macow anii (57) and E. leucotrachela (57.1). All three occur in the Betty's Bay/Kleinmond area with the latter two on the mountains. Only careful examination of the flowers can provide the distinguishing characters with corolla colour being a problem-easily noticed in fresh materi al, but not always recorded in dried herbarium material.
The species is commonly known as the Prince of Wales heath. Diagnostic features: flowers mostly 1-nate, occasion ally 2-nate, on short, lateral branches; anthers with hairs ( Figure 6J).
The typical subspecies occurs abundantly in marshy places on the coastal flats from Rooiels to Kleinmond. However, one collection in BOL, Burman 978, is record ed from Onrust Mtn. This subspecies is confined to the lowland flats between Hermanus and Stanford.
Most herbarium collections of this species have been derived from flower shows or flower sellers with no locality details being provided which complicates the assessment of diversity of regional populations in this species complex. Diagnostic features', corolla broadly funnel-shaped to obconical, ± 6 mm long, dark to pale pink; filaments :/i length of corolla; ovary 8-locular; stems pilose; leaves rather flat.
This species is clearly very closely allied to E. propendens (139) and E. trichophora (142). See discus sion above under 62.1. It is known only from the Klein River Mtns.

Note: the larger format plate was published by Andrews in vol. IV of Coloured engravings o f heaths which is impossible to date accurate ly: it is unfortunately one o f those undated plates in the last two vol umes. The date of The Heathery plate is also uncertain. Following the researches o f C'leevely et al (2003)-'As a consequence, it can be con cluded that most, if not all. o f H.C. Andrews' new names for
The species occurs in the area from Stanford to Elim and is not sympatric with E. perspicua. E. colorans var. breviflora H.A.Baker is transferred to E. plena L.Bolus (265.3). We regard this as a species dis tinct from, but close to. E. colorans and not just as a short-tubed variant, since the flowers have a different shape and the anthers a different position. Diagnostic features: flowers 2-4-nate on leafy side branches arranged in bunches along main stems often with long, almost naked areas in between and long con tinuous growth: corolla finely hairy, rosy purple: anthers dorsally attached in midregion, filaments with distinct cobra-like stance; ovary (4-)7 or 8-celled. obconical, with erect apical calluses (not bulging laterally as in E. curviflora) (Figure 61).
This species is extinct in the wild having formerly been collected only on the sandy Cape Flats (I.M. . It is still cultivated in several botanical gardens, in South Africa and Europe, and from nurseries in Australia and western USA. It can be grown very eas ily (Hitchcock 2003). Diagnostic features', flowers l-3(4)-nate on short, lat eral, leafy branches, densely aggregated along main branches; corolla finely hairy, bright scarlet, narrow tube 6-16 mm long; anthers with small basal appendages; ovary 4-locular. with erect hairs on upper half.
There is considerable variation in the length of the corolla with most of the long-tubed variants flowering in late summer autumn and the short-tubed variants flower ing in spring, early summer. Diagnostic features: flowers in a spike-like synflorescence, 120-260 mm long; corolla 12-16 mm long, tube not narrowed below the mouth.

66.1a. subsp pillansii
Pillans said (pers. comm.) that he had never seen this species in the wild, only in flower sellers' buckets in Cape Town and was told that the material came from the Caledon District.
The typical subspecies occurs in the Kogelberg Bio sphere Reserve from low to middle altitudes from the Platberg area to near Kleinmond and flowers mainly in autumn. Diagnostic features: flowers in a spike-like synflorescence, 60-120 mm long; corolla 6-8(-10) mm long, con stricted below the mouth.
The protologue of E. pyrantha gives the corolla length in the Latin diagnosis as 5 mm [0.5 cm] long but in the description, as 10 mm [1 cm] long. Bolus referred to this species as 'unlike any known to me' and surpris ingly did not refer to E. pillansii which he described in the same paper. It is known only from the type collection. There is a possibility that E. pyrantha may be a chance hybrid between subsp. pillansii or subsp. fe r\'ida and the yellow-flowered E. campanularis, since the corolla is described as 'orange with a red base' and the anthers are much lower down in the corolla tube as in that species, with several other floral characters being similar in both species. It is tentatively placed in synonymy under this subspecies. The types of all three taxa were 'collected' in flower-sellers' buckets in Cape Town.
This subspecies also occurs in the Kogelberg Biosphere Reserve but mainly at middle altitudes from Kogelberg to above Betty's Bay. It flowers mainly in spring. In two well-separated populations, hybrids have been recorded between this subspecies and E. fastigiata and were described under the name, E. xvinacea L.Bolus (L.Bolus 1928;Oliver 1986 49. E. densifolia Willd., Caroli a Linné species plantarum 2: 359 (1799); Guthrie & Bolus: 73 (1905); Dulfer: 45 (1965). Type: Herb. Willdenow (B!). Diagnostic features: flowers mostly single on ends of short, lateral, leafy branchlets, aggregated into short, spike-like synflorescences below ends of main branches; sepals with sessile glands on adaxial surface; corolla shortly hairy, pink to red with green tips; anthers with long appendages; ovary 4-locular, long, narrow and not emarginate; the style slightly swollen at base.
There is considerable variation in a number of char acters-length of the pedicel (0.8-10 mm), position of the bract and bracteoles, sepal shape, texture and indu mentum (narrow to broadly lanceolate, glabrous to lanate, simple to gland-tipped hairs, leaf-like to very hard and rigid). Some nine variants ( Figure 10A-I) have been noted, but not named, since additional research, includ ing DNA analyses, needs to be done on the species.
The species ranges from the Grootvadersbosch area on the Langeberg, eastwards to Humansdorp and inland on the Kammanassie and eastern Groot Swartberg ranges.  Diagnostic features: flow ers 1 or 2(3)-nate on ends of leafy, lateral branchlets, scattered over plant or sometimes aggregated into loose, spike-like synflorescences; corolla sparsely and finely hairy, pale orange; ovary 4-locular, long and narrow, not emarginate, with very distinctive large bulge at base of style, sometimes w ider than ovary.

E. xerophila
The identity of E. wendlandiana was formerly uncer tain and it was regarded as a possible variant of E. cur viflora until we examined the lectotype at Kew. The dis-  Diagnostic features: leaves 4-nate; corolla hairy with distinct swelling at base; bract and bracteoles small and remote; anthers well exserted, muticous, long and nar row; sparse hairs on filaments and style; pedicel relative ly long (10-15 mm) with small, sticky red glands; ovary densely hairy, cylindrical.

A -J . 4mm . K-N . 2mm .
Diagnostic features: open groupings of flowers to long, rather open, spike-like synflorescences; flowers 1nate on short, leafy, side branchlets or 3-nate on ends of main to long side branches; leaves and sepals glabrous to woolly; corolla glabrous, pale yellow to greenish yellow; anthers included to exserted, long and narrow with long, thin appendages; ovary 4-locular, hairy.
This species is widespread on the inland mountains from the Cederberg to the Klein Swartberg.

ERICA DISCOLOR-VERSICOLOR-UN1COLOR GROUP
There is a group of long, narrowly tubed species that has proven very problematic to identify, due to the con siderable variation in most organs. They all have similar ovules and seeds flattened laterally in the vertical plane, and the testa has a similar warty structure on the outer periclinal walls. They occur mainly in the southern part of the Cape Floral Region particularly the southern coastal mountain ranges of the Langeberg-Outeniqua-Tsitsikamma and the adjacent foothills and coastal plain.
They were introduced into horticulture in Europe in the late 1700s from which several plants were described by Andrews as distinct species. The oldest name in the complex is his E. discolor of 15 October 1794.
There has been much confusion among the taxa in this group. We have found considerable variation to the extent that we retain four common and widespread species, £. discolor, E. versicolor, E. unicolor and E. diaphana, note several variants under the first one, and describe two new species, E. croceovirens and E. prolata.
Key to species in the Erica discolor-versicolor group 52. E. hebecalyx  53. E. discolor complex Diagnostic features: shrubs single-stemmed or multi stemmed; leaves 3-nate; main branches with usually every node producing lateral branches in the flowering zone; flowers (l-)3-nate on ends of leafy side branches in short to long, spike-like synflorescences; sepals from ovate to lanceolate-acuminate, glabrous to hairy all over; corolla glabrous, often viscid, pink to red with greenish or yel lowish tips, occasionally totally green; anthers with long, thin, pendulous appendages ± 213 length of theca, rarely only '/2 as long or as long as theca, often kinked, thecae sometimes prognathous at base either bluntly or sharply so; ovary 4-locular, slightly emarginate, glabrous. This is a very common, widespread and highly variable complex. It used to consist of two very well-known and widespread species, E. discolor and E. speciosa, both Andrewsian names and the more localized E. hebecalyx. Despite their apparently well-known status as species, many persons had difficulty in assigning a name to plants from this complex, even finding it problematic to separate them from E. versicolor or E. unicolor (viridescens) and its variants. With the large amount of collections at our disposal we have found no satisfactory morphological characters to separate the variants at species level and felt that they all belonged to one large, widespread and com mon species with E. discolor being the oldest name and having three, rather indistinct variants, A, B & C, that need further investigation in the field and DNA studies to assess correctly for subspecific status. We are not prepared at this stage to give them any formal taxonomic ranking with new combinations or new names. This should be done when the above studies have been completed.
The structure of the main branches in the flowering region seemed to be a useful character to subdivide this complex-lateral flowering branches at every node with no intervening leafy nodes versus lateral flowering branches not at every node with few to many leafy nodes in-between. Many specimens fitted these characters in conjunction with the other one of elongated vegetative growth or not, but there were exceptions-the iconotype of E. discolor being one of these.
We hope that collectors will take note of all these problems when accurately noting population and plant details in the field.
VARIANT A: E. discolor Andrews, Coloured engravings of heaths 1: 20 (1794); Benth.: 629 (1839); Guthrie & Bolus: 75 (1905); Dulfer: 46 (1965 Plants that were associated with this name have main ly a bushy growth coming from a multi-stemmed base and the main branches not usually continuing with long vegetative growth beyond the flowering zone. This is seen in wild populations in the region from Betty's Bay to Agulhas. We surmise that the occasional elongated vegetative growth with a more floriferous flowering zone would attract a collector more readily than the more sparsely flowering, short branches, hence the presence of this feature in a number of herbarium collections. The iconotype of this name has the main branches continuing with vegetative growth beyond the inflores cence zone, several whorls of leaves between the lateral flowering branches (i.e. flowering branches not at every node), both of which can be found in some collections. We feel that this condition may have arisen due to very favourable growing conditions of unbumt young plants in England. The other variants, B & C, listed below, all have this clear growth. The ovate, glabrous sepals with a short sulcus, ± '/3 the length of the sepal and paler coloured flowers of the iconotype, match those of mate rial from the western part of the distribution from Betty's Bay to Agulhas.
The leaves also tend to be shorter in this variant, 2-4 mm long.
This variant is widespread, occurring from Betty's Bay along the southern coastal lowlands to Humansdorp. It is more common in the west. This variant is usually a single-stemmed reseeder and has the main branches continuing with elongated vegeta tive growth beyond the flowering zone on every plant. The leaves are 6-10 mm long and the sepals narrower. 4-8 x 1-2 mm. with an attenuate apex. The corolla is always bicoloured, from dark pink to orange-red with green or yellowish tips. We have seen only single stemmed plants, but have had reports of multistemmed resprouters in some inland populations.
A minor variation of this variant has slightly longer leaves and sepals which are either hairy or glabrous. Superficially it looks rather like E. unicolor subsp. georgensis (55c), which, however, has 4-nate leaves. It occurs along the coastal plains east of George from Knysna to Humansdorp.
Variant B is also common and widespread, mainly along the coastal mountain ranges from George to Humansdorp, but extends onto the much drier inland mountains of the Little Karoo from Laingsburg to Willowmore. VARIANT C: E. hebecalyx Benth. in DC'., Prodromus 7: 630 (1839); Guthrie & Bolus: 75 (1905); Dulfer: 46 (1965 The tomentose sepals were regarded by  as giving this variant 'a distinct appear ance' and referred to those plants occurring in the moun tains above George. Subsequent collections have shown populations with tomentose sepals in the Kammanassie Mtns and as far afield as the northern side of Seweweekspoort. It forms single-stemmed plants with the main branches producing elongated vegetative growth in the George area where the flowers are a pale yellow-green, occasionally reddish pink with green tips. We have not studied the other populations in the field. Diagnostic features: leaves 6-10 mm long, villous; sepals ± 11 mm long. ± V3 length of corolla: corolla uni coloured. yellow-green: anther appendages 1.3-1.7 mm long, usually slightly kinked.
This subspecies occurs on the southern slopes of the Outeniqua Mtns between Robinson Pass and George.  Diagnostic features: leaves 5-6 mm long, glabrous to sparsely hirsute; sepals 5-7 mm long. ± '/4 length of corolla; corolla bicoloured, red with green tips; anthers muticous or occasionally with very' short appendages.
Subsp. mutica occurs on the lowlands south of the Outeniqua Mtns in the Herbertsdale Mossel Bay area and lower southern and northern slopes of the mountains from Herbertsdale to George. Diagnostic features: leaves 7-11 mm long, glabrous and sparsely ciliate; sepals ± 11 mm long; corolla bicoloured, shades of pink-red with green/yellow tips; anthers muticous or with thin straight appendages up to 2 mm long.
This new species is related to the group of tubularflowered species (E. unicolor, E. discolor complex, E. versicolor. E. diaphana. E. berzelioides) which all have the same seed structure and arrangement of bract, bracte oles and sepals and viscid corollas caused by sessile glands on the inner, adaxial surface of the sepals. It is most closely related to E. unicolor in the 4-nate leaves, narrow sepals and muticous to appendiculate anthers. The flow ers are yellow -green to green in subsp. unicolor to bicoloured, orange-red and green yellow in subsp. mutica and subsp. georgensis. The latter differs in having only simple hairs when these are present. The other species listed above all have 3-nate leaves.
E. croceovirens has distinctively orange flowers w ith the upper third of the tube being yellow-green, hence the epithet: croceus, saffron orange; virens, green (Latin). There is little variation in this colour combination other than in the intensity of the orange.
The species is confined to the Doringrivier Wilderness Area on the northern slopes of the Outeniqua Mountains ( Figure 12) where it grows on sandy, dry flats or rocky, north-facing slopes with short fynbos vegeta tion. In the area, scattered plants of the glandular E. glan dulosa occur, as well as plants of E. discolor w ith its flowers being more pink. The species could w ell be over looked as being another form of E. glandulosa but it is easily distinguished on closer examination and assess ment of the characters mentioned above. This probably explains its somewhat recent discovery by Jan Vlok.
Andrews described the plants as having 4-nate leaves w ith flowers 2-or 3-nate but show s in the draw ing. five as 2-nate, 13 as 3-nate and two as 4-nate. He says the flow ers are 'orange yellow' but shows in his painting the open flowers with a distinct, red base ('/«-'A) and clear yellow upper portion, therefore bicoloured, and the buds all dis-  Diagnostic features: lateral branchlets not at every node; leaves 3-occasionally 4-nate; sepals with nar rowed leafy apex, often slightly spreading; filaments spoon-shaped at base; anthers often with sharp project ing point at base, muticous. This is a variable species especially in the habit and habitat, the indumentum, the leaf arrangement, colour of the corolla and in the base of the anthers. Bolus' var. monticola, although growing in inland, rocky, dry areas from Matroosberg to near Citrusdal, has no morphologi cal features to distinguish it from typical E. versicolor from the moist Langeberg range. The Shand collection, syntype of var. monticola, is clearly the Langeberg vari ant and not the gnarled inland variant. Detailed work is required to ascertain whether these variants should receive taxonomic status.
The species occurs from Citrusdal to the Anysberg (Laingsburg) and along the Langeberg and western Outeniqua Mtns and the coastal plains from Swellendam to Gouritz River.  Diagnostic features: lateral branchlets not at every node; leaves 3-nate, elliptic/ellipsoid, short, ± 3 mm long; flowers 1-3-nate, scattered; corolla red-pink in lower half with white upper half; sepals with sessile glands on adaxial surface densely packed; anther appendages small; ovary 4 locular, emarginate, glabrous.
Diagnostic features', habit of a few, very long, erect, subnaked stems, bearing the flowers and arising from a bushy, leafy base; main stems with occasional side branches (not every node); leaves 3-nate; anthers muti cous; ovary densely and finely hairy.
The species is most similar to E. diaphana and E. berzelioides but differs mainly in the marked dimor phism of the branching system-the other species having rounded, open to closed, bushy habits. It is the only species in this complex with a hairy ovary.
E. prolata has an unusual and distinctive growth form which is similar to that found in E. barrydalensis (285.1) from the northern slopes o f the Langeberg near Barrydale. The single-stemmed shrub forms a rounded,  Diagnostic features: like E. glandulosa but anthers differently shaped and subbasally attached, with longer decurrent appendages; flowers borne more terminally on main branches. This is known only from the type collection, the local ity of which is rather vague. No additional collections have been made. With the areas of unspoiled indigenous vegetation now virtually absent on the sandy flats west of Port Elizabeth due to the spread of housing, farming and alien plants (Australian Acacia spp.) the likelihood of finding material to match the type is very small.
We postulate that this material could have been of hybrid origin with the parent species being E. glandulosa and E. chloroloma both of which have been recorded in the Port Elizabeth area.
This species is variable in corolla length. 10-24 mm. with collections at the lower end having been described as var. buccinula. There is no clear-cut distinction in this range to warrant taxonomic recognition.
E. cruenta occurs on heavy shale soils at low altitudes where it is often found in renosterveid or such-like arid vegetation and as such is the only species of Erica with this habitat preference. It grows from Grabouw to Riversdale.
E. cruenta var. campanulata Bolus is transferred to E. elimensis L.Bolus (261.2 Diagnostic features: leaves 4-nate; flowers 1 ^4-nate on ends of very short to longer lateral branchlets, some times aggregated at ends of main branches but not spike like; pedicel hairy; bract and bracteoles small, remote; corolla glabrous, dark red, shiny; anthers with very small appendages; ovary emarginate, with slight stipe, glabrous.
This species occurs in rocky areas or on rocks in the Cold Bokkeveld and Cederberg.
Diagnostic features: shrubs very woody; corolla hairy internally and externally, yellowish green; bract and bracteoles approximate; sepals with adaxial central patch of sessile glands; anthers relatively large with or without very small, upward-pointing appendages.
This is a very distinctive species with possible alliances in the E. discolor complex. It is restricted to rocky areas close to the sea near Cape Point and across False Bay from Rooi Els to Hangklip. The shrubs are low but very woody with trunks as much as 100 mm in diam. and in some cases growing just above the spray zone of the waves.