The genus Erica ( Ericaceae ) in southern Africa : taxonomic notes 1

This is the first in a series of notes on the southern African species of Erica L. which are currently recognized. Brief synonymy citing only the important revisions and synonyms is given, as w ell as comments on the status and diagnostic char ​acters of each species. One new species, E. petrusiana E.G.H.Oliv. & I.M.Oliv. and 14 new subspecies are described, and11 species are reduced to subspecific status. Where necessary drawings of important features and variations are given. Part 1 covers the first 35 species (Sections 1-3).


INTRODUCTION
The last revision of the genus Erica L. covering the species in southern Africa was undertaken by Dulfer (1965) and published as a brief conspectus-excluding descriptions and with very little or no discussion about problems of delimitation and the decisions he took.All of this was based almost exclusively on the very small collection o f Erica housed in the Natural History Museum in Vienna (W), in many cases with species rep resented by only one or two specimens.He had only a few small loans from the Bolus Herbarium (BOL) and the Royal Botanic Gardens, Kew (K).having been refused large loans.His work was based on that of Guthrie & Bolus in Flora capensis (1905) and included the 220 species described since that publication.We esti mate that he had far less material on w hich to base his decisions than did Guthrie & Bolus, w ho also consulted the large collections at Kew and the Natural History Museum.London (BM) and also at Trinity College.Dublin (TCD).Bolus also consulted the important Ericaceae collections in Berlin, which, except for the Willdenow herbarium, were totally destroyed during World War II Fortunately he was allowed to remove a few flowers from several of the types and these are in BOL.
Dulfer's work is valuable in that he spent much time checking the synonymy and old references that Guthrie & Bolus had put together.This remains a much used source of reference in our work on Erica.Since Dulfer's revision, another 63 species have been added, and all the species formerly included under the 23 minor genera have also been added, w ith the relegation of these genera to synonymy under Erica (Oliver 1988(Oliver . 1993a(Oliver . 1993b(Oliver . 2000)).The number of species in the genus currently stands at 765 in the area covered by the Flora o f south ern Africa.The number of specimens located in the two Cape herbaria.Bolus Herbarium (BOL) and the Compton Herbarium (NBG).which now includes the old South African M useum Herbarium (SAM ) and Government Herbarium, Stellenbosch (STE).greatly increases the amount of material that exhibits variation, compared to that which was available lor Guthrie & Bolus and considerably more so in the case of Dulfer.
Having studied the genus in the herbarium and in the field for 42 years (EGHO) and 15 years (IMO) respec tively.w e are in a position to update the species concepts currently recognized in the literature (Dulfer 1965: Oliver & Van Wyk 1993).and to publish our concepts which are incorporated as curatorial practice in the Compton Herbarium (NBG).
It has been decided to produce a series of taxonomic notes on the species following the short format used by Dulfer.but with short to detailed explanations of varia tions. of problems w ith delimitation and of new delim i tation of some taxa.Where necessary illustrations of the variable taxa have been included.We have followed the numbered order started by Guthrie & Bolus (1905) and adapted by Dulfer (1965) and then by ourselves in the Compton Herbarium.This system follows the sectional arrangement used by Guthrie & Bolus in w hich the genus is divided into 43 sections based on the earlier work of Bentham (1839).These sections were an attempt to group the species into supposedly related assemblages.In some cases this is clearly the case, but in others there are no close relationships and species are. in our opinion, sometimes w idely separated from their nearest relatives.An example is a species which was described twice by Guthrie & Bolus (1905).firstly as E. auriculata, w hich is no.135.and secondly as E. greyii.which is no.338.This sectional/species arrangement is used here for w ant of a better, natural system w hich is as yet not forth coming.The available morphological characters are not sufficiently adequate for this purpose due to the high degree of homoplasy present within the genus.To attempt to resolve this impasse, a study is to be under taken into the molecular relationships of some 300 species in collaboration w ith overseas researchers to try to ascertain the basic clades w ithin the genus and to cou ple this with an analysis of morphological characters.These results should give a good indication as to which the important morphological characters are in a reassess ment ol the genus at the level of subgenus and section.
In a tew cases the species have not been fully resolved and we are forced to list these in complexes under the oldest name with pointers towards the problems that need to be resolved through more detailed fieldwork and molecular studies.
The genus Erica is by far the largest represented in the Cape Flora and it is. in our opinion, clearly still evolving actively.There are many species which are isolated in their relationships, and these one may regard as palaeoendemics, whereas there are many that form extremely complicated complexes that currently, and perhaps always will, defy satisfactory resolution.
To date, the only subspecific category that has been used in the main revisions is that of variety (varietas).With the current trend to drop that category, we have resorted to recognizing worthy subspecific groupings as subspecies, especially when there is, in addition to some morphological disjunction, also a disjunction in distribu tion, ecology or postulated pollination biology.In other cases we have refrained from a proliferation of sub species and have resorted to listing unnamed variants which reflect what we believe are low level groupings that have been noted by us.
The concentration of so many species in the Cape Floral Region is shown in counts we have done.Often we have found four species growing in one square metre with the highest count thus far obtained being seven, or. in one case, four species in a quarter square metre-all not closely related species.In a number of cases the disjunc tions between taxa can occur over very short distances with variants being confined to relatively small areas.
With such a large genus-currently 760 species in southern Africa, we are forced to publish this review in a series of parts covering a varying number of species per part depending on the complexity of the species and the species complexes in order to provide data on species delimitation soon, rather than wait many years for a com plete coverage of all 680 capsular species in one publi cation.The 84 indehiscent fruited species have recently been dealt with in detail (Oliver 2000).
NOTE 1: in the numbering system the whole numbers for the species are those that were originally allocated by Guthrie & Bolus (1905).The species described since then until 1965 were placed by Dulfer nearest their sup posedly nearest ally as a & b numbers.These we have altered to decimal notation since we use the a/b notation for subspecies.Species described since D ulfer's 1965 revision have been placed by us in their postulated alliances.Species we recognize are printed in bold Roman type.Those that are reduced to synonymy, or that we regard as cultivated forms, or dubious, are printed in bold italics.
NOTE 2: the inclusion under the synonymy of cita tions of publications additional to the original one, is restricted to the most important revisions of the genus.For all other citations and more detailed synonymy, ref erence should be made to Guthrie & Bolus (1905) and to Dulfer (1965).
NOTE 3: in the case of citations involving the two works of Andrews, Coloured engravings o f heaths  and The Heathery • (1804-1812), the rele vant work has to be cited in full in each instance, due to the considerable overlapping of publication dates (see Cleevely & Oliver 2002), and cannot be cross-referenced in the references.

ERICA COCCINEA COMPLEX
This complex, consisting of E. coeeinea, E. melastom a. E. intermedia and E. monadelphia, is character ized by an enlarged petaloid bract and bracteoles adpressed to the calyx, long, attenuated, well-exserted anthers with rounded apices and basal attachment, the often flap-like placenta with a naked abaxial zone, shiny smooth seeds (not alveolate or reticulate) and leaves with a large sclerenchyma bundle on either side of the sul cus-these visible as white stripes in fresh or pressed material.
The inflorescence consists of 3-nate flowers, except in E. melastoma where they are borne singly.
The 'coccinea' com plex's nearest relatives are the species belonging to the small-flowered E. imbricata complex (no.369) and not to any of the other large-flow ered species, e.g.E. plukenetii, where similarities are the result of convergent evolution in pollination syndromes.It is at times difficult to differentiate between the various species, especially in the 'imbrieata-placentiflora' com plex which has evolved numerous variants and in some cases these are sympatric, thus adding to the confusion.There could very well be a case for considering hybridization between some of these variants.
These two complexes cover a wide range of flower sizes from a large tubular corolla ± 20 mm long, to a small globose one ± 2 mm long.The pollination syn dromes vary from bird pollination in the large-flowered species to insect and wind pollination in the small-flow ered species.
Erica banksii is very similar to members of this com plex, but lacks the distinctive sclerenchyma bundles in the leaves, and the anthers have the filaments attached more or less dorsally.
A yellow-flowered variant with long narrow sepals which has a dark corolla mouth, occurs in the Bredasdorp to Gansbaai area.
The typical subspecies is the common and w idespread taxon in this species occurring from the Cederberg to the Cape Peninsula and eastwards as far as ihe Kammanassie Mountains.Diagnostic features: flowers borne singly at ends of short, leafy lateral branches, usually yellow, sometimes orange-red; leaves semi-erect to reflexed.This is a lowland taxon occurring on hills and flats not far from the coast from the Cape Peninsula eastwards to near Mossel Bay.

1.1.
E. m elastom a Andrews, Coloured engravings of heaths: t. 37 (1799).E. petiveri var.melastoma (Andrews) Benth.: 622 (1839); Guthrie & Bolus: 47 (1905).E. follicularis Salisb.var.melastoma (Andrews) Dulfer: 29 (1965).Iconotype: Andrews: t. 37 (1799).D iagnostic features: leaves erect to sem i-erect, straight; flowers single on short, leafy side branches; sul cus on sepals, bract and bracteoles very short, apical, broadly open.V-shaped (not narrow, slit-like and form ing a keel) or absent.This is a very variable species in the size of the (low ers.size and shape of the sepals, shape and stickiness of the corolla and colour of the corolla apex.Within all this variation there is only one distinct discontinuity, namely in the size of the corolla, and we propose to recognize this at subspecific level.Diagnostic features: corolla ± 18 mm long (Figure 1B4).
Within the subspecies there is a considerable amount of variation in several characters, mainly the corolla shape, colour and stickiness.Despite there being some collections which appear to be very distinct, we are unable to recognize any formal categories and thus list them here as unnamed variants to show possible groups that need to be worked on in detail in the field and through DNA analyses.
VARIANT A: the corolla is yellow w ith a dark, almost black, distal end.It may, however, be greenish yellow with the distal end fading to brown soon after anthesis.The typical variant, shown in Andrews' painting, has large sepals, i.e. broad and more than half the length of the corolla tube, the corolla slightly to much inflated towards the base and mostly not sticky although small sessile glands may occur on the margins of the sepals (Figure 1B?).The leaves can be short to very long.The distribution is from Stellenbosch to Bredasdorp.
VARIANT B: this variant has flowers similar to Variant A. but they are very sticky and yellowish w ith a black mouth (Figure 1B4).It includes the var.inflata and occurs in the Bredasdorp District.
VARIANT C: with similar flowers to A. but not sticky, usually light orange-yellow, mostly with no tuft ed small side branches typical of the other variants.It occurs on limestone flats of the Bredasdorp District.
VARIANT D: flowers yellow-orange with no black mouth, with long narrow tube and a short calyx less than half the length of the corolla tube, which is not sticky (Figure 1 B i).This occurs in the region from Tulbagh to the Cederberg.
VARIANT E: flowers the same as those in D above, but very sticky.The stickiness is produced by a remark able array of sessile glands in several zones around the margins of the sepals (Figure 1    Diagnostic features: anther apices acuminate; corolla distinctly longer than calyx.± 6 -1 1 mm long.
This occurs along the Langeberg and Outeniqua Ranges from Swellendam to George.Diagnostic features: flowers white; corolla broadly ovoid.5.5-6.0 mm long; bract, bracteoles and sepals with only small sessile glands on margins and no hairs admixed (Figure ID:).
The subspecies occurs in the Robinson Pass/Ruitersberg to Jonkersberg area of the Outeniqua Mountains.Note: Andrews published the name as 'monodelphia' in the protologue (text and plate), but later changed it to 'monadelphia' in the index which was published some five years later in 1802 when the complet ed volume was bound.He also used the spelling 'Monadelphia'.This latter spelling was taken up by all subsequent authors who did not query the '-ia' ending.There is no indication why Andrews used this substantival epithet in apposition.This could be assumed to be Linnaeus' major Class name, Monadelphia (stamens united by their fil aments into one binly) (Steam 1957: 31  Diagnostic features: anthers w ith a distinct boundary between thecae and filament, and a slight basal adaxial 'nose'; corolla tube straight.± II mm long, bright red (Figure 1C).This is a distinct species within the 'coccinea' com plex and not w ith E. banksii as its number would suggest.There are two branching forms w ithin the species-one with numerous very short side branches as in E. coc cinea.the other without these.The species is a resprouter after fires.

ERICA PLUKENETII COMPLEX
Three separate species used to be recognized in this complex-E.plukenetii, E. lineata and E. breviflora.
The main variation in this species is in the length and shape of the corolla and sepals, size and texture of the leaves, and in its habit.The flowers are 7-16(-28) mm long with a tubular or inflated tubular corolla with farexserted stamens (Figure 2).Flowers change shape as they mature, otten being rather slender in bud and quite inflated at maturity.The sepals range from 2-12 mm long, lanceolate to ovate in shape w ith the sulcus as long as the sepal.The plants are mostly reseeders, but can be resprouters in some areas, and can v an from woody, rounded shrubs 0.75 m tall near the coast, to tall delicate erect shrubs in the Bredasdorp District.In the Kamies berg the multi-stemmed resprouters can be up to 3 m tall.
This species is often confused with species in the E. coccinea complex.The position and size of the bract and bracteoles is the most visible character to distinguish it from that group.5a.subsp.plukenetii Diagnostic features: sepals 2-8 mm long, short and ovate to longer and lanceolate (Figure 2A
The taxon occurs on limestone hills and lateritic flats from Heuningrug south of Bredasdorp, eastwards as far as Cape Infanta and also on the sandstone of Potberg.There are numerous collections from this area.Diagnostic features: sepals up to 12 mm long, thick and fleshy, with distinctive non-revolute thick margins forming two longitudinal keels like a catamaran (Figure 2C), bases of these keels produced into small lobes beyond point of attachment to pedicel.These features are well displayed in Andrews' drawing.Bartling's descrip tion fits this taxon-'sepalis base productis solutis, margine revolutis'.Diagnostic features: leaves very long, narrow; sepals (Figure 2D) broad and flat sepals like those in subsp.hredensis.
The subspecies occurs on sandy soils associated with coastal limestone deposits from Gansbaai to Zoetanysberg.Diagnostic features: corolla short, ovoid to almost globose, mostly ± 7 mm but less than 12 mm, white, occasionally pink; synflorescence typically very long spike-like up to 250 mm long (Figure 2F, G).
There is an increase in flower size in some collections from the Cederberg together with a reduction in the length of the synflorescences.These can merge with some short variants o f subsp.plukenetii in the Franschhoek Mountains.We have noted that populations in the Porterville Mountains have flowers that are sweet ly scented.This coupled with the smaller flowers would indicate that the subspecies is pollinated by insects as opposed to the postulated bird-pollinated, longer flowers of all the other subspecies.The plants are single stemmed reseeders.
Guthrie & Bolus (1905) noted under E. scariosa 'we admit this species with doubt* and kept it as distinct 'with some reluctance'.
The subspecies occurs in the Cederberg to Porterville and the southern Cold Bokkeveld area.

6.
E. monadelpltia Andrews-this has been placed in the E. coccinea group after E. intermedia (1.2).
The typical subspecies forms a small, compact, woody shrublet growing on rock ledges and occurs in the mountains around the Elgin Basin.Diagnostic features: corolla w hite w ith purple lobes; leaves short.± 5 -9 mm long: ovary glabrous.
This subspecies is confined to the higher mountain peaks from Kogelberg to Hangklip.In the Kogelberg complex the subsp.banksii occurs at lower altitudes at the northeastern end.
Subsp.comptonii differs from the other two sub species in the larger, more open habit (up to 500 mm tall) in open ground between rocks and in the longer leaves (up to 26 mm) with longer mucros on the leaves and sepals (Figure 3Bi).The sepals can vary from about half to the full length of the corolla tube.
The variation in this species occurs in the flower colours, the habit, habitat, length of leaves and length of the mucro on the leaves and sepals.Disjunctions occur regionally, but warrant recognition only at subspecific level.
This distinct species has no alliances with any other long-tubed species, but has, rather, an alliance with sev eral species with much shorter flowers, such as E. monsoniana L.f. (no.402) and E. goatcheriana L. Bolus (no. 405.1).The possession of plumose hairs on the pedicel are shared by all of them.
The typical subspecies forms an erect shrub, 0.5-1.0m tall, growing in open ground or between rocks.It occurs on the coastal mountains from George through to Humansdorp.
Erica clavata Andrews (Coloured engravings of heaths: t. 159, 1809) was included by Guthrie & Bolus (1905) and Dulfer (1965) in synonymy.Careful exami nation of his painting leads to the conclusion that E. clavata could be a hybrid with E. viridiflora as a possi ble parent.Characters which do not fit the current taxon are the erect imbricate leaves, the broadly ovate sepals with long attenuate apices, and the muticous anthers.Diagnostic features: corolla cream to greenish, nonsticky; shrubs small, woody shrublets, erect to 300 mm high but sometimes prostrate (Figure 4B).Subsp.primulina occurs only in the Swartberg Range from the Klein Swartberg to near Willowmore.A collec tion from Gamkaberg.which is further to the south and west than the rest of the collections, is an intermediate variant having the flowers of this subspecies, but the grow th of subsp.viridiflora-it is recorded as a woody erect shrub 500 mm tall.
In the protologue.Bolus mentioned the affinity with E. viridiflora.but surprisingly stressed a closer likeness to E. banksii and gave the differences from that species.Diagnostic features: differs from subsp.primulina only in size of flowers and their parts, corolla 4.5-5.0mm long (Figure 4C).This small-flowered form is restricted to a single popu lation o f some 100 plants occurring in crevices on a large steep slab o f quartzite rock.The plants were about 200 mm tall and bore creamy white to sometimes greentinged flowers.Subsp.primulina grows in similar habi tats in the vicinity.This is clearly a case of a short-tubed variant adapted to a different pollination syndrome (insect versus bird).No material matching the description and type has been found in the Cape.We postulate that this collection is o f hybrid origin in the w ild.Parentage could include E. abietina because of the similar anthers and ovary.Jacob Mulder is known only through the citation by Salisbury of 37 species of Erica from the Cape (Gunn & Codd 1981).Only one specimen is labelled as a Mulder col lection by Salisbury.Diagnostic features: anthers exserted.with short stub by appendages below thecae; ovary puberulous; corolla glabrous; stems, leaves, pedicel, bract, bracteoles and sepals all puberulous.
This species has been collected only once.There are no indications of its alliances.There is a possibility that it is a naturally occurring hybrid, as found in several other supposedly rare species that have been investigat ed in the wild.e.g.E. x flavisepala (Oliver 1977) and E.
x vinacea (Oliver 1986).The remaining 24 species with whole numbers dealt with in this paper were placed by Guthrie & Bolus (1905) in §Evanthe which was based on 'Inflorescence mostly axillary, more rarely also terminal on the same plant.Corolla tubular, mostly over 6 lin.[12 mm] long, rarely shorter.*The inflorescence is 1-flowered on a very short lateral branchlet often less than 1 mm long and sometimes bearing very reduced scarious leaflets.These occur in the axils of leaves on the main branch (hence axillary) and are arranged in dense spike-like synflores cences towards the ends of the main branches w hich con tinue their vegetative growth.This contrasts with the often 3-nate inflorescences on leafy lateral branchlets found in §Pleurocallis which will be dealt with in Part 2.
The inclusion of species in this section.§Evanthe.has been rather arbitrary, being based mainly on the size of the corolla.Species with smaller flowers, but with the same inflorescence structure, have been included in §Hermes.and one of these is certainly very closely relat ed to some species in this section, e.g.E. axilliflora L. Bolus.12.
E. m am m osa L.. Mantissa plantarum altera: 234 ( 1771 Diagnostic features: corolla narrow ed at base with 4 indentations or folds, glabrous: bract and bracteoles small and remote from calyx: anthers with long thin appendages and a basal, projecting "nose"; leaves erect and green: ovary short, rounded and glabrous. We agree w ith Salter's reduction of E. gilva to svnonymy under E. mammosa, both having four broad to narrow indentations at the base of the corolla.The former occurs only on Table Mountain on the Cape Peninsula and was separated by Guthrie & Bolus (1905) on the longer bract, bracteoles and sepals-it has w hite flowers sometimes w ith a green tip.There is a lot of variation in the colour of the flowers on the Peninsula, particularly in the southern parts and some plants in the Cape of Good Hope Nature Reserve have very long spike-like synflor escences.
Erica mammosa is widespread from the Cape Peninsula to the Cold Bokkeveld and eastwards to Bredasdorp with flowers orange or deep purple-red.greenish yellow to white.Variations occur in the length of the spike-like synflorescence from short and dense to very long (up to 250 mm) and rather open.The pedicel can be glabrous or hairy and the sepals 3 -6 mm long.On the mainland the plants are resprouters, whereas on the Peninsula they may be resprouters or single-stemmed reseeders.The most variation occurs on the mountains above Simonstown.

13.
E. broadleyana Andrews, Coloured engravings of heaths: t. 154 (1809).Type: Andrews: t. 154 (1809).This is regarded as a hybrid of garden origin in England.The plant looks like E. plukenetii, but has included anthers.Andrews' drawing shows the upper part of the filaments zig-zagged, which condition would indi cate that the anthers could become exserted from the very narrow mouth.
14a. subsp.b aueri D iagnostic features: leaves short, spreading to recurved; flowers usually 10-20 in few short dense sub terminal spike-like synflorescences, white to pale pink.
This com m only cultivated and well-known sub species is now becoming rare in its habitat due to the spread of agriculture and alien vegetation (mainly Acacia cyclops) in the sandy flats which are derived from quartzitic sandstone and lie west of Albertinia.Diagnostic features: leaves erect; flowers 5-10 in loose, spike-like synflorescences.more numerous and scattered over shrub, pale to deep pink (Figure 5).
This subspecies forms a taller, denser shrub up to 2.5 m.It grows very localized in sandy areas associated with limestone between the hills to the south of Albertinia near the coast, an area called Gouriqua.Diagnostic features: pedicel very short; bract, bracte oles and sepals equally long and becoming enlarged, fleshy and persistent in fruiting stage; corolla pale green to yellowish green; sepals spathulate with entire to ser rated margins; anther appendages narrow; ovary short, rounded and glabrous.This is a very distinctive species being the only one in the genus exhibiting serotiny-the fruiting synflores cences remain on the plants for several seasons as cone like structures with the fruits protected by the thickened calyces.The species is very variable in the size of the corolla and in the shape of the sepals from narrowly oblanceolate to very broadly spathulate-characters used to separate several varieties by previous authors.There is, however, no clear-cut disjunction between these shapes that warrant recognition.A very showy variant occurs on the Riviersonderend Mountains with green corolla and bright red sepals compared to the normal green sepals.Diagnostic features: corolla cyathiform, 8-10 mm long, yellow or dark pink; anthers situated in middle of corolla.Figure 6B.
The typical subspecies occurs on sandy hills between Viljoenshof and Pearly Beach in the Bredasdorp District.Both colour forms have been noted growing sympatrically.Diagnostic features: corolla white to pale pink, urceolate-cyathiform, 5 -7 mm long; anthers positioned higher up in corolla than in typical subspecies.Figures 6C; 7.
The type population had flowers varying from white to pink on the same plant and grew with low plants of the green-flowered form of E. penduliflora (17.2).The subspecies occurs on sandy flats and hills between Elim and Zoetanysberg.There are numerous collections from this region.This species forms a complex with the next two species, which were all treated as one species by Guthrie & Bolus (1905).It varies in a number of characters-size and colour of flowers, form of anther appendages, anther shape and position in the flower (Figure 5).Bolus' var.major was removed as a distinct species, E. penduliflora (17.2) and Dulfer (1965)   Diagnostic features: sepals ovate, V5-V4 as long as corolla; anther appendages broadly based and much toothed; corolla purplish pink, 3-5 mm long.
The larger-flowered collections southwest of Bredas dorp have very reduced anther noses, whereas the collec tions from the De Hoop area have smaller flowers and remarkable tumed-up noses, a character not found in any other species (Figure 6D).D ulfer only exam ined the single collection, Schlechter 10472 (W), when he raised E. filipendula var.minor to specific status and selected that as the lectotype, unaware that the other three syntypes constituted anoth er distinct taxon (see 17b).Diagnostic features: corolla 12-18 mm long; sepals broadly ovate; anthers with projecting nose and long thin appendages; ovary glabrous with long stipe.Figure 6A.
This was regarded as a large-flowered variety of £ fili pendula by Bolus (1905), but was deemed sufficiently distinct to be recognized as a distinct species based on the above characters.The species has two distinct colour variants, white or yellowish green, which are allopatric.In the Pearly Beach area a white variant has recently been recorded with green tips.18. E. grandiflora L.f.-see E. abietina subsp.aurantiaca (23e).Guthrie & Bolus: 57 (1905);Dulfer: 35 (1965).This is just a large-flowered form of E. parilis Salisb.(Guthrie & Bolus 1905: sp. no. 301).
The inflorescence is umbel-like on the main stems and can have, in addition, below the umbel, short leafy branchlets ending with a 3-or 4-nate inflorescence.This latter situation is shown in Andrews' painting.
The main axis can sometimes continue growth.Diagnostic features: corolla sticky, with longitudinal ridges, densely to sparsely covered with hair-like spicules or pustules; flowers in a dense spike-like syn-florescence.with 1-flowered florescences on vestigial lateral branchlets bearing small scarious leaves; bract and bracteoles small, tough and situated in middle of pedicel: anthers included with ear-like appendages below thecae; filaments with spicule-like hairs; ovary rounded and glabrous.
We retain the epithet thomae in recognition of Thomas Stokoe's considerable collections of fynbos plants made from the 1920s to the early 1950s-both L. Bolus' species were described on the same page.
The species shows much variation in the size, thick ness and degree of spreading of the leaves, the length of pedicels (from 4 -1 6 mm long), and the colour and size of the flowers.There are no clear-cut discontinuities to war rant formal taxonomic rank and the three species have been grouped here as variants of a single species.All occur in the mountains from Kogelberg to Hangklip to Kleinmond.Some detailed molecular studies may help to elucidate the relationships between the variants.
VARIANT A (formerly E. thomae sensu stricto): medium-long pedicels, 6-10 mm long: corolla 22-30 mm long, rose pink to dark reddish pink with or w ithout paler tips.It occurs in the southwestern parts of the distribution range.
VARIANT B (formerly E. tenax): long pedicels, 13-16 mm long; corolla 22-30 mm long, green to white.This is the variant from the northern and eastern parts of the distribution range.
VARIANT C (formerly E. porteri): short pedicels.4-7 mm long; corolla 20-25 mm long, more delicate and dark reddish pink w ith white mouth; leaves more spreading up to 90°.This is restricted to a single small population in the Buffelsrivier Valley near Pringle Bay where it grows with Variant A and appears to produce hybrids.However, some collections of Variant A from other areas possess similar small flowers, but not the bicoloured corolla.The Buffelsrivier population has plants more delicate than the other two variants.Compton- Diagnostic features: corolla tube with basal restriction zone: inflorescence spike-like and/or umbel-like and w ith no continuing apical grow th: ovary emarginate, glabrous.

E. porteri
With the bipartite anthers and emarginate turbinate ovary, this species clearly belongs in the E. abietina group.It is a restricted endemic on the Cape Peninsula.22.2.E. q u a d risu lc a ta LB olus in Annals of the Bolus Herbarium 3: 172, t. 7D (1923): Dulfer: 37 (1965) Diagnostic features: ovary acute, glabrous; inflores cence umbel-like, not continuing w ith vegetative elonga tion.
The species is a very restricted endemic on the Cape Peninsula.
The considerable range of variation and lack of clear disjunctions have resulted in the reduction of several well-know n species to subspecific rank and w ith two new taxa warranting only subspecific rank.Diagnostic features: corolla dark red.tubular.18-26 mm long, spiculed to sparsely puberulous and slightly viscid: sepals subovate, sparsely pilose with adaxial ses sile glands; anthers included to exserted (Figure 8B).
The subspecies is confined to the upper rocky slopes and plateau of Table Mountain  Diagnostic features: corolla rose to deep rose, tubular.18-22 mm long.± glabrous and somewhat sticky; sepals broadly lanceolate, sparsely puberulous w ith adaxial ses sile glands; anthers included, occasionally manifest (Figure 8C).
The taxon is confined to the Cape Peninsula and occurs from the lower slopes of Table Mountain   Diagnostic features: corolla shortly obconical, rosepink, 11-14 mm long, subglabrous, subviscid; sepals ovate, pilose with adaxial sessile glands (Figure 8E).
The subspecies is a very restricted endemic occurring only on the saddle between D evil's Peak and Table Mountain.
Andrews' E. echiiflora is placed under E. viscaria subsp.gallorum (34d).His paintings show a ridged corolla with nipped-in mouth and spreading corolla lobes and an ovary with longish erect hairs-all typical of that species.Diagnostic features: corolla pale to deeper rose-pink.obconic, 8-11 mm long, glabrous, subviscid; sepals lanceolate-ovate, with adaxial sessile glands; anthers always included, situated about :A way up tube (Figure 8D).
This subspecies is confined to the Cape Peninsula occurring on the mountains from Constantia Neck to Chapm an's Peak.
Guthrie & Bolus (1905) noted that there is little to separate E. conica from subsp diabolis (quoted by them as E. abietina var.echiiflora).We retain it as a subspecies in this complex.Diagnostic features: corolla tubular, (1 0 -)25-30[-34] mm long, glabrous, sometimes with few hairs on lobes, orange to orange-red, sticky to non-sticky; sepals long acuminate from ovate base, with large area of adaxial ses sile glands; anthers included to far exserted (Figure 8A).This is the most widespread, common and variable of the subspecies, occurring from the hills just northeast and east of Cape Town to as far inland as the Witteberg at Matjiesfontein and southeast to near Ashton, but absent from the Cape Peninsula.The flowers can be pale to dark orange with zones of yellow below, to complete ly deep orange-red.Diagnostic features: corolla pure yellow, 20-25 mm long, densely, finely hairy, non-sticky; sepals broadly elliptic and long acuminate, with adaxial non-sticky ses sile glands; anthers included to m anifest; leaves 20-30(-42) mm long (Figure 9).
The material of this taxon was at one stage classified as E. exsurgens Andrews.The taxon does not match any of Andrews' plates and is a distinct, very localized entity occurring only in the Jonkershoek Valley near Stellen bosch where it is occasional to locally common on the moister granitic slopes facing south and southwest.Diagnostic features: corolla pure yellow, ± 20 mm long, densely, finely hairy, non-sticky; sepals narrowlanceolate with no adaxial sessile glands; anthers mani fest to exserted (Figure 10).
The subspecies is restricted to rocky outcrops on the mountains above Porterville where the common and widespread subsp.aurantiaca is not known to occur.Erica abietina is highly variable in flower size and colour, indumentum of calyx and corolla, stickiness of the corolla, degree of inclusion/exsertion of the stamens, anther shape, and leaf length and habitat preferences.It used to consist of four separate species that were long established in the literature-E.abietina, E. phylicifolia, E. grandiflora and E. conica.There are no clear disjunc tions in the ranges of characters that were formerly used to separate them, but there are some slight discontinuities which warrant only subspecific recognition.
The whole complex is held together by flowers borne 1-nate or occasionally 2-nate on vestigial lateral branch lets arranged in compact, spike-like synflorescences towards the ends of main branches.The leaves are all apiculate and vary from 10-30(-^2) mm long, the se pals vary from very narrow, long-lanceolate to lanceo late, to broadly lanceolate, to almost ovate and longacuminate and have a flattened, raised area below the sulcus.They may be villous to pilose or glabrous and often bear numerous sessile glands over their inner sur face, thus rendering the corolla viscid.The sepals all have sessile glands on the margins and.except for subsp.petraea, numerous sessile glands adaxially in the middle zone next to the margins (Figure 9H)-these may be sticky or non-sticky.
The corolla is mostly long-tubular varying from 18 to 30(-34) mm in length, but in two subspecies, obconical and only 8-14 mm long.It varies from glabrous to subglabrous to densely and finely hairy and may be red.orange-red, orange, deep pink, pink or yellow.In the fresh state these colours are very distinctive, and would clearly lead one to use them as specific characters, but in dried material without colour notes, identification is nigh impossible and one has to resort to a few morphological characters.The anthers can be distinctly bipartite, even splitting in the apex of the filament, to having the thecae closely adpressed-sometimes varying in a single col lection.They are mostly attached basally to slightly subbasally.The filaments are often sparsely strigulose with the style sparsely hairy.
The Peninsula taxa tend to form a group having the apex of the corolla lobes a little more rounded, whereas the taxa from the mainland have more acute apices to the corolla lobes.
This species is characterized by the 6-8-locular ovary which is stipitate and glabrous, the glabrous dry smooth corolla and no glands on the adaxial surface of sepals.The flowers may be yellow with a white tip or pure w hite.It can easily be mistaken for white-flowered forms of E. viscaria subsp.longifolia.With only a slight disjunction in one morphological character coupled w ith a clear difference in habitat pref erences, we are reducing this complex o f three species and three varieties to one species with two subspecies.Diagnostic features: corolla nipped in at apex just below spreading lobes.
The variant with variegated flowers (basal white zone, middle mauve/purple zone and bright red upper zone) is very striking and well known as the Elim Heath, but has no morphological differences from the unicoloured forms, some of w hich can have a slightly paler w hitish basal portion.
The var. williana described by Bolus is a problem in that it was considered by us as a short-tubed form (flow ers 6-12 mm long versus 14-20 mm in the typical sub species) with a postulated different pollination syndrome and worthy of subspecific status.However, we have recorded it as a few plants growing together with the orange-red long-tubed typical subspecies and the dark pink E. a.xilliflora Baill.just northwest of Zoetanysberg (see Note above under synonymy).The flowers varied from pale to darker purple.At the time w e concluded that the plants were possible hybrids between the two species.Further "populations' of this form need to be located and investigated thoroughly before any definite conclusions can be made as to its identity.Diagnostic features: corolla not nipped in at apex.This subspecies occurs only on limestone ridges from southwest of Bredasdorp to Stilbaai.In the Mierkraal area this taxon grow s on the limestone ridge, whereas the variegated variant of subsp.regia grows on the nearby lateritic/sandy Hats.
Guthrie & Bolus separated the species in this variable complex on the colour and shape of the corolla and the degree of protrusion of the nose at the base o f the anthers.
The protologue for the species described the type with 'flores purpurei'.We have not seen a purple-flowered long-tubed plant in the wild, but have seen fresh material of the purple-flowered short-tubed form (var. uUliana).The Elim hills and Zoetanysberg possess a form with bright orange-red flowers and further south towards the sea the white-flowered (sometimes tinged pinkish) form, formerly E. casta.occurs, both on sandy substrates.On the lateritic flats southeast and east o f Elim the form with variegated flowers occurs.On the limestone ridges from Mierkraal right through to Stilbaai.there occur plants with dark red flowers.The colours are very distinctive and recognizable, but have no major taxonomic signifi cance to warrant subspecific ranking.Herbarium materi al in which colour has not been recorded cannot be iden tified with any certainty.
The corolla shape in E. regia and E. casta is almost identical, both having a restriction below the spreading corolla lobes and sometimes a tapering tow ards the base.The flowers o f E. mariae do not have the distinct restric tion.There is no clear-cut boundary between the more extreme anther noses in some material of E. regia to the almost lack of a nose in E. casta and E. mariae.Diagnostic features: long leaves with long petiole (i.e.versatile leaves); corolla trumpet-shaped.16-24 mm long, finely hairy, not ridged: ovary hairy on top.hairs erect.

E. casta
The variation in this species occurs in the shape of the sepals from narrow-lanceolate to those with a broader base, the corolla indumentum from very finely downy to pilose and flower colour-w hite, pink, purple to red.The northern groups of populations have distinctly geographi cal colours-white at Ezeljacht hills, red on the Riviersonderend M ountains and purplish pink on the Langeberg near Swellendam.They all have finely downy corollas, whereas the southern populations from the Klein River M ountains to Cape Agulhas have more pilose corollas.The population north of Stanford has white flowers, whereas there is a mixture of purplepink to reddish in the colour of the flow ers from the hills east of Stanford to Agulhas.The type is described as white-flowered and would probably match the Ezeljacht material.
Under E. longifolia Bolus commented on the close similarity between his var.maritima and E. vestita.We find it inseparable from the various forms of E. vestita found in the Bredasdorp District and transfer it as a syno nym under this species-the smooth corolla (without ridges) being sparsely pilose and the long petioles mak ing the leaves versatile, are the defining characters.Diagnostic features: corolla tubular.10-12 mm long, narrowed slightly towards apex, finely hairy to subglabrous.w hite or pink.
The long-flowered variety of E. filamentosa has the closed corolla mouth of this species, and even though the corolla is pink, it is transferred to this species.There is a clear disjunction in the distribution range with the pink variant on the eastern Riviersonderend Range and the white variant on the Langeberg at Grootvadersbos and Garcia's Pass.Diagnostic features: corolla longitudinally ridged with very short bristle-like hairs or pustules; tube slightly con stricted below spreading lobes, not or rarely obconical to broadly so; ovary obconical.not stipitate, covered with erect dense fairly long, white hairs (Figure 11).This is a very variable species which occurs from the Cape Peninsula to Franschhoek and to the Bredasdorp coastal flats.It is probably the most variable species of Erica with respect to flower colour-white, green, yellow ish.pink, purple, red.or combinations of pink with a w hite mouth or red with a yellow mouth.Some of these colour variants are very striking and showy.The corolla varies in shape and size over a w ide range.The indumentum may be almost absent, finely puberulous, strigose, spiculate, strigose from pustules, or markedly pustulate.The flowers may be very viscid, partially so or non-viscid and the plants may be reseeders or resprouters with long to short leaves.
Colour is an unsuitable character since this is not retained by older herbarium material and can therefore not be used as a sole distinguishing feature.
Corolla length we regard as important in distinguish ing subspecies, since the open corolla shape of these taxa coupled with a lower position of the stamens in their flowers would suggest a different pollination syndrome from the bird-pollinated long-flowered variants.Diagnostic features: inflorescence mostly lax; corolla short (5-9 mm long), urceolate to campanulate-obconic, sott and more transparent with slightly softer and shorter hairs; anthers often adhering laterally: thecae with more prominent basal bulges towards filament (Figure 1IG).Bolus (1905) noted the slight difference between subsp.viscaria and subsp.gallorum by describing E. vis caria var.hispida from the Sir Lowry's Pass area (see below under 34d).The inflorescence is mostly laxer and longer and the corolla soft and more transparent in the typical subspecies.
This subspecies occurs on the Cape Peninsula moun tains and also the surrounding flats from where it is now mostly extinct except in the northern parts.Two variants occur the commoner one with narrower, hairy sepals on the mountains and the rarer with broader, subglabrous sepals, as represented by the lectotype selected above, mostly on the flats.Diagnostic features: corolla ± 12-20 mm long, tubu lar, non-viscid.hairy or pustulate, red, pink, purple, white, yellowish or green or. in some cases, bicolouredpink with a white mouth or red with a yellow mouth; sepals, bract and bracteoles long, linear-lanceolate (Figure 11 A. B).This is the most variable subspecies, which it may, on more detailed population studies coupled with molecular analyses, be possible to divide into more subspecific taxa.All plants appear to be single-stemmed reseeders.
There are problems in distinguishing material of the pink/purple-flowered form in the Bredasdorp District from similarly coloured forms of E. vestita.It is very pos sible that the two species hybridize where they are sympatric.The flowers of E. viscaria subsp.longifolia have a longitudinally ridged corolla and do not possess the longer, soft hairs of E. vestita.E. longifolia var.maritima Bolus has been transferred to E. vestita (31).Diagnostic features: sepals lanceolate to ovate-lanceolate, often with a broader base and attentuated apex, corolla ± 15-20 mm long, tubular (Figure I ID).
The corolla is yellow ish to green, occasionally w hite and the plants are mostly resprouters.Some variants may have very viscid flowers.The subspecies occurs in the region from Betty's Bay to Elim.Diagnostic features: corolla markedly pustulate, yel low, ± 7 mm long, ovoid-urceolate (Figure 1 IE).
Subsp.pustulata is closely related to the long-tubed form of subsp.longifolia with yellowish to green flowers from the same locality.Diagnostic features: flowers pendulous (not spreading to semi-erect); corolla ± 12-18 mm long, tubular, white, sometimes tinged pink, non-sticky, with short hairs only along longitudinal ridges and around base; plants sparse ly branched and up to 1.5 m tall, with relatively short grey-green leaves (Figure 11C).
This subspecies is confined to the eastern end of the Paardeberg Range west of Bot River Village.
Erica petrusiana is closely related to the short-tubed subspecies of E. viscaria.but differs in the shape of the corolla w ith its open mouth.It shares this character with E. latiflora L. Bolus (303.1) which has an even more open corolla with longer lobes.This latter relationship points to the problem in the genus of the long-tubed and short-tubed species being closely related, but placed very far apart in the current sectional system.
The name is derived from the generic name of the fish.Petrus rupestris.the 'red steenbras' (red rock-bream), after the locality, the Steenbras area which is the type and only known locality.Esterhuysen records the plants as very local, but common, in stony shaly soil and also on sandstone ridge.She also records 'corolla was yellow, not very bright, but certainly yellow, slightly sticky'.

Note:
Dulfer overlooked the problem of the dates of publication of Andrews' plates and selected Salisbury's name o f 1802 for the species, thereby relegating Andrews' E. melastoma to varietal status.Dulfer (1965) cited 1802 as the date o f publication for E. melastoma which is the date o f the bound volume and not the date which is printed on the plate.A paper detailing the publication dates of all Andrews' species is in preparation (Cleevely & Oliver 2002).
Note: as with several other epithets coined by Andrews, the '-ia' ending was thought to reflect the use o f a generic name used in appo sition, in this case Banksia (Australian Proteaceae).and should not be corrected.However, there are some cases where a generic name did not exist or was coined later than the publication o f the species name.These -ia endings are being replaced by the relevant genitive ending (i, -ii, -ae or -iae) for all Andrews' names commemorating persons.The date o f the Bauer plate is actually 1 Jan. 1801 and not the date o f the first published fascicle in the bound publication, i.e. 1796, as used by Dulfer to antedate Bauer's name over that o f Andrews.However, fas cicle 3 containing this plate only appeared in 1803 (Britten 1899).

FIGURE 7 .-
FIGURE 7.-Erica filipendula subsp.par\ a. type.Oliver 11254.A. flowering branch, x 1; B. stem with leaves removed: C, leaf: D. vestigial lat eral branch in axil o f vegetative leaf, showing vestigial leaves at base and flower with corolla removed: E. flower: F. anther, side, back and front views.Scale bars: B -G . 2 mm.
are two specimens in MEL-one the original from Wendland's herbarium determined by Klotzsch.the other a branch bro ken off by Klotzsch and retained in B and seen by Bentham.Both were borrowed by Sonder who was working on the genus for his Flora capensis w hen he died, and were never returned to the original herbaria.His herbarium was sold in two parts with the Ericaceae going to Baron Ferdinand von Mueller in MEL.Illu'tration: Schumann & Kirsten: 45. t. 23 (1992).

a paper covering the typification of all Linnaean names of Cape Erica spp. is in preparation by Oliver. Jarvis & Cafferty.
Note:E.