A biosystematic study of Pentameris ( Arundineae , Poaceae )

A biosystematic study of the endemic southwestern Cape grass genus Pentameris Beauv. is presented. Results of studies on the macroand micromorphology, leaf blade anatomy and cytology are discussed and illustrated. The results of a cladistic study indicate that the genus is monophyletic, united by the synapomorphies of ovary and fruit characters. The conservation status of the taxa in the genus is assessed, and conservation status codes allocated. A key to the taxa in the genus is presented, and each species is described. Five new species, Pentameris glacialis N.P. Barker, P. hirtiglumis N.P. Barker, P. oreophila N.P. Barker, P. swart be rgensis N.P. Barker and P. uniflora N.P. Barker, and one new subspecies, P. longiglumis (Nees) Stapf subsp. gymnocolea N.P. Barker, are described and illustrated.


INTRODUCTION
The genus Pentameris Beauv.occurs in the winter rainfall region of the Cape Province, South Africa, where it is restricted to soils derived from Table Mountain Sandstone or the shale bands associated with this geology.It may therefore be considered an endemic of the Cape Flora (Goldblatt 1978).All known species of the genus are perennial C3 plants.However, the unusual leaf blade anatomy (Ellis 1985d;Barker 1990) and fruit morphology (Barker 1986(Barker , 1989(Barker , 1990) ) of P. obtusifolia places this species within the genus Pseudopentameris Conert (Barker in prep.a).Pentameris obtusifolia is therefore not further considered in this study.

MORPHOLOGY
The basal parts of all the species are woody, but the woolly and/or swollen underground parts that occur in certain species of Pentaschistis (Nees) Spach and Merxmuellera Conert are absent.The species are generally tufted, with stems sometimes branching, but generally only Bothalia 23,1 (1993) in the basal quarter.In older plants branching results in a cushion or a bush-like growth form.These branched structures are referred to here as 'culms', although they do not possess typical culm anatomy.Instead, the anatomy resembles that of a rhizome (H.P. Linder pers.comm.).The term 'aerial culm' will be used here to denote the annually produced, few-noded structure which terminates in an inflorescence.
In P. thuarii, branching can occur at any node on the culm, and some specimens comprise long, decumbent, scandent or vertical culms with many shorter vegetative and reproductive branches arising from the upper regions.Such branching from cauline innovation shoots results in an 'evergreen', perennial plant, which takes the form of a divaricate herb.This branching pattern might allow the plant to grow taller than species possessing only basal innovation shoots (Linder & Ellis 1990a).Branching and growth from cauline innovation shoots is also found in a few species of Pentaschistis, Chaetobromus Nees and Pseudopentameris.In the latter genus the branched condition may approximate the extreme development found in Pentameris thuarii.
The leaf structure of Pentameris is typical of the Arundineae.The leaf sheaths are persistent, although the leaf blades may not be.The older culms are thus frequently covered in the remains of sheaths from previous years of growth.These sheaths may be appressed to or free from the culm.The sheath indumentum varies between the spe cies, ranging from glabrous to pubescent along the mar gins to uniformly pubescent (and sometimes woolly).These sheath characters are also found in other arundinoid genera.The ligule is, as in many other arundinoid genera, a fringe of hairs.The sheath mouth may be beard ed in some species, but this, too, is not unusual.Two spe cies, P. thuarii and P. longiglum is have fairly well-developed auricles.In the former species these are a characteristic purple or brown colour.The leaf blade varies in its internal and epidermal anatomy.Many charac ters are shared, although not consistently so, with taxa of Pentaschistis, and (to a lesser extent) Merxmuellera.Ellis (1985cEllis ( , 1986;;Ellis & Linder 1992), on the basis of leaf anatomical studies, is of the opinion that Pentameris is closely allied to Pentaschistis.
The inflorescence is a panicle which varies from com pactly lanceolate to laxly globose.It must, however, be noted that the panicles of all the species are open and somewhat lax during the period of pollination.Before and after this period, the panicles contract to a lanceolate shape.Panicle shape is therefore difficult to describe as it depends on the reproductive phase of the plant at the time of observation or collection.This variation is also found in numerous other arundinoid taxa, for example Pentaschistis (Linder & Ellis 1990b).No spicate panicles are known from this genus.
The spikelets are two-flowered (one exception; P. uniflora), and the partially developed remains of a third floret may sometimes be present, particularly in P. macrocalycina.The two-flowered state is shared with the vast majority o f the species in Pentaschistis (Linder & Ellis 1990b).When two florets are present, the basal floret is sessile, the apical floret pedicellate.The florets are otherwise morphologically identical, and both are hermaphroditic and fertile, as in Pentaschistis.
The lemma is generally nine-to eleven-nerved.The majority of the veins anastomose in the basal region of the central awn and the two lateral bristles.The lemma lobes on the outside of the bristles are variously adnate to the bristle, ranging from free (in P. thuarii) to almost completely adnate (in P. macrocalycina).The shape of this lobe is also variable, being acuminate, acute or dentate.Similar variation is found in Pentaschistis (Linder & Ellis 1990b).The central awn is geniculate in all species, the basal portion is flattened but twisted and shorter than the attenuating apical portion.The margin of both the basal and apical regions o f the awn are finely serrated.This awn structure is shared with genera such as Pentaschistis, Pseudopentameris, Merxmuellera and Chaetobromus as well as the non-African genus Danthonia.
The palea is bicarinate, apically bilobed and pubescent between the veins.It exceeds the lemma body in length, but seldom exceeds the lemma lobes, a situation also found in species of Pentaschistis, Pseudopentameris and Merx muellera.
The lodicules in certain arundinoid genera have been examined and used for taxonomic purposes (Tomlinson 1985).The lodicules of Pentameris are generally glabrous and cuneate, but differences have been noted in some specimens of P. swartbergensis and P. distichophylla.The lodicules in these species are generally apically ciliolate, and sometimes an arm-like extension arising from one of the lateral margins is present.Ciliolate lodicules are also known from species of Pseudopentameris, Merxmuellera and Pentaschistis (Linder & Ellis 1990b).
All species of the genus have three anthers, which are usually purple in colour.These vary in size in relation to the size differences of the florets, but this has not been incorporated as a taxonomically meaningful character.
The above morphological features can therefore not be used to clearly differentiate Pentameris from several other southern African arundinoid genera, notably Pentaschistis.There are, however, two micromorphological features which have, in the past, been used to distinguish Penta meris from all other southern African arundinoids.Both features are characters of fruit and ovary.Firstly, the fruit is an achene, and secondly, the fruit of all taxa possesses a dense tuft of hairs at the apex of the ovary.These hairs are retained in developing and mature fruit where they become reluctantly deciduous.
Stapf (1897: 512) separated Pentameris from Pentaschistis on the basis of fruit morphology: the crustaceous pericarp and free seed.He described the structure of the ovary as being '.. .s o alike in the five species of this genus that it is very probable that they agree in the peculiarities of the ripe fruit which is known only in P. thuarii' and further stated that there is in Pentaschistis no approach to the characteristic structure of the ovary and the fruit of Pentameris This unusual fruit type, the achene, is described by Clayton & Renvoize (1986)  Scanning electron microscope (SEM) studies on the fruit of Pentameris have shown the pericarp to be sculptured (Barker 1986(Barker , 1989(Barker , 1990) ) as illustrated here in Figure 1A, and partly or almost completely free from the seed coat when viewed in section (Barker 1990)  key to the grass genera.Chippindall (1955) describes Pentameris as having an ovary which is hairy on top, the hairs being deciduous.This character was also used in the key to the genera of the tribe Danthonieae.The nature of these unusual structures has been clarified to a certain extent by Barker (1986Barker ( , 1989Barker ( , 1990), but their ontogeny and function is not known.These structures, visible on the apex of the fruit (Figure 1A) appear to be unicellular, and arise from the apex of the ovule around the base of the styles.

LEAF BLADE ANATOMY
Leaf blade anatomy has contributed much to the tribal and subfamilial classification of the grass family.However, within the tribe Arundineae, the variation o f observed anatomical characters led Renvoize (1981) to comment that the arundinoid genera could not be readily divided into the tribal groups on the basis o f their leaf anatomy, and he concluded that other characters such as spikelet morphology would have to be used to divide the subfamily further.
The leaf blade anatomy of many of the southern African arundinoid genera has been documented (Schweickerdt 1942;De Wet 1956;Conert & TUerpe 1969;Ellis 1980;1981;1985a-d).Ellis (1985aEllis ( -d, 1986) examined many specimens of the then five known and one undescribed species of Pentameris.Additional anatomical information for the new and undersampled species was obtained by sectioning a small portion of leaf material obtained from herbarium specimens.This material was heated in a soapy solution prior to sectioning by means o f a sledge microtome.Sections thus obtained were 30 to 50 microns thick.The sections were stained in safranin and fast green and mounted in Euparol.
To complement the data obtained from the work o f Ellis (1985aEllis ( -d, 1986) and additional sections, leaf material from 95 of these previously studied specimens was prepared from herbarium specimens, mounted on twosided tape, coated in gold-palladium and examined using a ISI-SX-25 scanning electron microscope.The salient anatomical features obtained from the SEM and light microscope studies are presented in Table 1.
The abaxial surface is relatively uniform throughout the genus.It is generally smooth, with ribs observed only occasionally.No abaxial microhairs or stomata are present, and the silica bodies are usually rounded, but are tall and narrow in P. thuarii (Ellis 1985b(Ellis -d, 1986).Abaxial macrohairs were occasionally present in specimens of some taxa (Table 1).The adaxial epidermis is far better endowed with micromorphological structures, including prickles, microhairs and macrohairs.These epidermal features are discussed in some detail below.

Microhairs
The genus Pentameris has a 'festucoid' type o f leaf epidermis and leaf anatomy (implying the absence of abaxial microhairs), a poorly differentiated parenchyma sheath and evenly distributed chlorenchyma sheath (De Wet 1956).However, Renvoize (1986) places Pentameris in the 'core' of the arundinoids, a group possessing micro hairs, and describes the lower epidermis of Pentameris and Pseudopentameris as possessing long slender papillae which occasionally bear the remains o f a small thin-walled apical cell.Clayton & Renvoize (1986) note that these structures distinguish these genera (and certain others such as Cortaderia Stapf and Centropodia Reichb.f.) from the rest of the Arundineae.These 'papillae', as described by Renvoize (1986) and Clayton & Renvoize (1986), appear to differ from those described by Ellis (1979), who considers papillae to be protuberances of the cell wall rather than structures consisting of separate cells.From observations discussed below, it appears that Renvoize is using the term 'papillae' to describe what others, such as De Wet (1956) and Ellis (1979), call microhairs.
In addition to their structure and appellation, the recorded distribution of these microhairs is also controver sial.Renvoize (1986, discussed above) reported them from the lower (abaxial) surface of specimens of Pentameris.However, Ellis (1985aEllis ( -d, 1986) has found no evidence of abaxial microhairs on any of the specimens he has examined.Microhairs were however found on the sides and bottoms of the furrows of the adaxial surfaces of many specimens in all the taxa observed (Ellis I.e.).My obser vations on material examined, using the SEM, corroborate Ellis's observations, and it is clear that there are no abaxial microhairs present in any of the species of Pentameris.
As illustrated in Figure 1C, D and E, these microhairs are indeed bicellular, with a minute, deflated apical cell (shown at high magnification in Figure IE).The length of the basal cell varies, but it is always very much longer than the apical cell.Microhairs with this structure were found on the adaxial surface of all specimens of all taxa in Pentameris (Barker 1990), and were also reported from certain species of Pentaschistis (Ellis & Linder 1992).
In addition to their distribution and morphology, varia tion in the size of the microhairs is taxonomically useful.The microhairs were measured from a number of SEM micrographs in order to determine whether or not there was any measurable size difference.These varying sizes are recorded in Table 1.Unfortunately, the limited sample size does not allow a statistically meaningful compari son to be carried out on this potentially very valuable taxonomic character.
The definite bicellular nature of these microhairs is comparable to those reported and illustrated by Amarasinghe & Watson (1988, 1989).However, microhairs with such unequal cell sizes have not been previously documented, and perhaps merit recognition as a separate type of microhair, the 'pentameroid' type.

Abaxial macrohairs
Abaxial macrohairs were only observed on the abaxial surface of some specimens of Pentameris distichophylla, P. hirtiglumis and P. glacialis.In all these taxa, the hairs were long, usually produced from the intercostal regions, had a distinctly swollen base and were surrounded by four or more cells, termed 'modified cells' by Ellis (1986).The taxonomic value of these structures is limited, as it appears that they are not universally present throughout all specimens examined of all taxa.Inconsistent sampling (variation in the region of the leaf from which the samples were taken, or the age of the leaf blade) may explain the lack of macrohairs in some samples.

Adaxial macrohairs
Two different hair-like structures were found on the adaxial leaf surface in some of the taxa of Pentameris, sometimes both occurring on the same leaf.Conventional macrohairs, those which have a distinct basal structure comprised of modified cells, were frequently observed on the adaxial epidermis in Pentameris thuarii, P. glacialis and P distichophylla, but were only occasional in P swartbergensis.Figure IF illustrates these hairs on the adaxial leaf blade surface in P. thuarii.
The second type of macrohair, termed here a 'filament', is a long, unicellular hair without any obvious basal differentiation or associated cells.Although no measure ments were taken, the filament type of macrohair appeared to be shorter than the conventional type.Filaments were frequently observed in the bottom of the adaxial furrows in specimens of P macrocalycina and occasionally in P distichophylla and P. swartbergensis.In P. oreophila filaments were only found in the furrow nearest the edge of the lamina.One of these filaments observed in P. distichophylla is shown in Figure 2A.
Adaxial macrohairs are of slightly greater taxonomic significance than abaxial macrohairs.The presence and type of macrohair are useful characters in Pentameris, although some variability within the species exists.It appears that the species with open or folded leaves (P.thuarii, P. distichophylla and P. swartbergensis) possess conventional macrohairs, whereas those taxa with perma nently rolled or folded leaf blades have the second type of macrohair (P.macrocalycina, P. oreophila, P. disticho phylla and P. swartbergensis).P. swartbergensis and P distichophylla thus have both types of macrohair.It is possible that the filament macrohairs are conventional macrohairs that have lost the modified cells around the hair base.The loss of these modified cells may be related to the evolution of the permanently rolled leaf blade, where the adaxial surface is not as exposed to the environment.

Prickles
The adaxial surfaces of all the specimens examined had prickles of one form or another.These different prickle types (Table 1) are invariant within the taxa of the genus.The prickles of P. oreophila are long and erect, with the tips sometimes bent or recurved.In addition to their unusual length, the prickles possess a basal collar like structure (Figure 2B), a character shared with P. hirtiglumis.The prickles of the latter species are, however, much shorter, and are almost appressed to the surface of the leaf, a character shared with P. longiglumis (Figure 2C).
Of the remaining species, P. macrocalycina, P. thuarii, P. glacialis and P. swartbergensis have short, erect prickles, the tips of which may be bent or curved; are swollen basally, and the sides are convexly shaped; are densely distributed, particularly so in P. macrocalycina.The prickles of P. uniflora are also quite densely distributed and almost erect, as indicated in Figure 2D.
When the two distantly related taxa are used as the out group, one shortest length tree (1.= 60, c.i. = 50, r.i.= 67) is generated.In this tree (not presented), P. cur vifolia is the taxon most closely related to the study group, and was used as the outgroup (the other Pentaschistis spe cies were removed from the data set); the data were reanalysed after the autapomorphies and invariant charac ters had been removed.Once again, HENNIG86 found one shortest tree, presented in Figure 3 (1.= 34, c.i. = 64, r.i, = 70).The phylogeny o f the study group in this tree is identical to that obtained in the initial tree (not presented) where the two glandular taxa are used as an outgroup.

Character distribution
In discussing the distribution of the characters on the tree, the characters from the complete database are used, although in the smaller data set (where Pentaschistis curvifolia is the outgroup) some are excluded because they are constant or autapomorphic.The distribution of the characters is shown on this tree (Figure 3).

Taxonomic implications
The monophyletic status of Pentameris is supported by the fruit and ovary characters.However, as Pentameris is undoubtedly closely related to Pentaschistis (at least the eglandular taxa), the possibility that the latter genus is paraphyletic ought not to be ruled out.The observation that Pentameris was placed within, and terminal to, the Pentaschistis clade in the analysis of the complete data set provides some evidence as to this possibility.If this is the case, the name Pentameris would be retained for the monophyletic assemblage containing the taxa with the unique fruit characters, while other genera may have to be erected to accommodate monophyletic groups within the Pentaschistis assemblage.Only a detailed phylogenetic study of both Pentaschistis and Pentameris can provide sufficient data to properly test any hypothesis of paraphyly.

CONSERVATION STATUS
Of the previously known taxa o f Pentameris, only P. obtusifolia is listed in the Red Data Book as a rare, threatened or extinct species, where it is listed under the synonym lP squarrosa' (Hall & Veldhuis 1985).The conservation status for this taxon is given as 'uncertain'.However, as discussed above, the nomenclature of this taxon has been found to be confused, and P. obtusifolia is not included in the genus Pentameris.
Certain species of Pentameris nonetheless require consideration and recognition as taxa requiring conserva tion.Most of the new taxa described below are known from only a few localities, and should perhaps be regarded as 'rare'.However, as many of these taxa are found in inaccessible and thus rarely collected areas, the true distribution is not fully known.In addition to the problem of varying collecting intensity, the effects o f fire on the germination, growth and re-establishment of fynbos grass species are poorly known.Linder & Ellis (1990a) discuss the various strategies that fynbos grasses have evolved to escape or adapt to fire.These authors note that grasses appear in abundance in the first few years after fire, but are then almost inevitably outcompeted by members of the Restionaceae and woody fynbos elements.The collection of fynbos grasses is therefore best carried out in areas that have been burnt within the previous few years.Such areas are not always attractive to collectors intent on obtaining other woody or non-graminoid taxa.Mature, woody fyn bos may therefore hide many new or apparently rare grass species which are present as dormant seeds or under ground vegetative structures.Actual species distributions may therefore be wider than presently known.
In assessing conservation status of the taxa of Penta meris, the categories proposed by Rabinowitz (1981) and modified by Karron (1987)   This species is the only taxon in the genus with oneflowered spikelets.Mature fruit have not been seen, but stylar hairs on the apex of the ovary confirm the position of this species in the genus Pentameris (Barker 1990).
P uniflora is known from only three localities which are quite widely separated (Figure 5).Populations at these localities appear to be quite abundant, but the habitat of this species (damp, rocky, southern aspects of the Cape fold mountains), habit (decumbent, hidden under taller plants), and small, somewhat inconspicuous inflorescences may account for the paucity o f herbarium specimens.February is the time of year when many fynbos grasses, including P. thuarii, are in fruit (pers.obs.).As the annotated specimen from Paris is in full fruit, this specimen is regarded as type material.However, this evidence is considered to be insufficient to warrant holotype status for this specimen, which is therefore selected as a lectotype.Stapf added that this latter variety has the appearance of a robust annual.Chippindall (1955) considers the characters on which these distinctions are made to be too variable, and rejects the two varieties, a conclusion supported by Barker (1990).This species grows abundantly in moist environments, and is usually found alongside streams, seeps and drainage lines.Geographically widespread, it occurs from Stellen bosch in the west to Montagu Pass in the east (Figure 6).

Pentameris longiglum is (Nees) Stapf
The following description is based on all the specimens seen, and the variation reported includes that for both known subspecies, described below.The first three axes resulting from the PCA accounted for 94.2% of the variation in the data, with the first axis contributing 75.7 %.The first two axes (Figure 7) account for 88.3% o f the variation in the data.The two groups comprise specimens from two geographically distinct localities, Kogelberg and Table Mountain.On the basis of this evidence, the allocation of the rank of subspecies to these two groups is considered to be justified.---------------------------♦

Plants densely tufted
Axis 1 This subspecies is known from a few specimens collected from Table Mountain, the distribution shown in Figure 5. Until recently, this taxon was thought to be extinct, as the most recent specimen seen among the holdings of eight South African herbaria is dated 1918.However, a small population has been located in a recent ly burnt area on Table Mountain.This population is growing in a slightly sloping seepage area.The moist habitat requirement matches that of the other subspecies, which is, however, found on steeper slopes.As shown in Table 5, this subspecies differs from P. longiglumis subsp.longiglumis in a number of features.The subspecific epithet chosen for this species describes one of these distinguishing characters: the glabrous leaf sheaths.
This subspecies is known only from the Kogelberg Forest Reserve (Figure 5).It inhabits south-facing moun tain slopes in seepage areas where it forifts large tussocks.This species is similar in appearance to P. oreophila, but can be readily separated from that species by the presence of the hirsute glumes (Figure 10B).Additional differences between these two taxa are the acute lemma lobes (Figure 10D) and softer, non-pungent leaf apices in P. hirtiglumis.However, as is apparent from the cladogram discussed above (Figure 3), this species is sister to P. longiglumis on the basis of several anatomical features including swollen leaf margins, inflated abaxial epidermal cells and dense, strongly inclined adaxial prickles.
Like P. oreophila, this taxon has a geographically restricted distribution, and is known only from high montane regions of the Hottentots Holland Mts (Figure 5).Where found, this species is abundant and locally dominant.However, it appears to have narrow habitat requirements, and is restricted to shale bands.Ellis (1985d), Barker (1986Barker ( , 1989)).Note that P. obtusifolia (Hochst.)Schweick.is to be transferred to Pseudopentameris.This species may be confused with P distichophylla, but differs in such features as the nature of the colliculate surface of the caryopsis (which is rugose in P. disticho phylla), the leaf sheaths (glabrous in P. swartbergensis, pubescent to woolly in P. distichophylla) and the lateral lemma bristles, which are substantially shorter in P. swartbergensis.

P. obtusifolia sensu
This taxon is known from only two localities in the Klein Swartberg, and is probably endemic to this mountain range This species, first collected in 1987 by Ellis, bears superficial resemblance to P macrocalycina, but is distinguishable from that species by the slender, geniculate culms, the rolled but not acicular leaf blades, the panicles with few spikelets, as well as by the size of the floral structures, which are substantially smaller in P glacialis.The species is generally found to be locally abundant, growing in narrow gullies, rock ledges and overhangs in black, humic but sandy soils.Such habitats are thought to become particularly heavily snowed up in winter.This habitat appears to preclude P macrocalycina, which is found in more stony sites, as well as P. distichophylla, which appears to prefer northern aspects and rock crevices.The woolly leaf bases and sheaths o f P distichophylla separate this species from P. glacialis (Figure 15).When visited in mid-October 1991, the plants and rocks at the type locality were covered in a thick layer of ice.Such conditions may persist for three or more months of the year in the winter and spring, depending on the severity of the frontal systems associated with the winter rainfall regime of the southwestern Cape.This harsh, icy environ ment gave rise to the specific epithet.
A subsequent visit in mid-December 1991 provided an indication of the range of climatic extremes in which this species survives, with very hot, dry conditions prevailing.The steep-sided, rocky gullies in which this species is found were, however, cooler, being shaded for much of the day, and the humic soil was still damp.Figure 5 shows the distribution of this species.
Palisot de Beauvois in 1812 on the basis of a specimen sent to him by Du Petit-Thouars.No collection number or locality is provided in this description.However, it is known that Du Petit-Thouars, a French traveller and botanist, obtained material from the Cape during a visit in February 1793 (Gunn & Codd 1981).The genus name is Greek, meaning 'five parts', probably a reference to the apical region of the lemma which is divided into what Palisot De Beauvois (1812) described as four bristles (two of which may be more accurately described as lemma lobes) and a central awn.Only one species is mentioned under the generic description: P. thuarii Beauv.(Palisot de Beauvois I.e.).This single species was subsequently placed in Danthonia DC. by a number of early taxonomists (Desvaux 1831; Nees 1841; Steudel 1855; Durand & Schinz 1895), whereas others retained it in the genus Pentameris (Roemer & Schultes 1817; Kunth 1833, 1835).Stapf (1897) expanded the genus Pentameris to include four other taxa, all charac terised by fruit with a free pericarp.Chippindall (1955) retained the genus in this format, but incorporated two nomenclatural corrections published by Schweickerdt (1938).Gibbs Russell et al. (1985) list these five taxa as Pentameris dregeana Stapf, P. longiglumis (Nees) Stapf, P. macrocalycina (Steud.)Schweick., P. obtusifolia (Hochst.)Schweick.and P. thuarii Beauv.
as shown here in Figure IB.The colliculate surface sculpturing is also unique among the fruit o f the southern African genera of the Arundineae (Barker in prep.b).The second feature that characterises the genus is the presence of hairs on the apical region of the ovary and fruit.This was first observed by Palisot de Beauvois (1812), who described the fruit as being crowned with stellate hairs.Nees (1841) also noted this feature in his descrip tions of P. thuarii and the taxon now known as P. macrocalycina.Phillips (1931) considered these hairs to be characteristic o f the genus, using this feature in his Bothalia 23,1

CYTOLOGY
Preliminary cytological studies have shown thatPentameris distichophylla has a chromosome count of 2n=36, whereas P. thuarii has a count of 2n=12.The latter figure has been corroborated by H. du Plessis (pers.comm.).Both these counts were obtained from meiotic material.P. distichophylla is therefore hexaploid.Such polyploidy is not unusual in the southern African Arundineae (see for example counts for Merxmuellera, Pentaschistis and other genera given by Du Plessis & Spies 1988; Spies & Du Plessis 1988; Spies et al. 1990).These counts further support the contention that the base chromosome number for the Arundineae appears to be x=6 (Davidse 1988), and not x = l2 as proposed by Clayton & Renvoize (1986).Despite the apparent similarities between Pentameris and Pentaschistis, the different base chromosome number of the latter genus (x=7 in many of the known instances) does not suggest a close relation ship between these two taxa.It is, however, possible that Pentameris evolved by means of aneuploidy from a pentaschistoid ancestor.PHYLOGENY The data presented in Table 1 were converted into a data set suitable for cladistic analysis.The anatomical characters were augmented by a few morphological characters, resulting in a data set o f 26 characters (presented in macrohair type: 0 = absent, 1 = present 13.Filament macrohair type: 0 = absent, 1 = present 14.Prickle type: 0 = inflated basally, 1 = markedly elongated, 2 = apically knobbed 15.Prickle orientation: 0 = erect, 1 = inclined/appressed 16.Prickle base: 0 = unornamented, 1 = collar present 17.Prickle density: 0 = sparse, 1 = dense 18. Microhair type: 0 = basal cell length » apical cell length 1 = basal cell length = apical cell length 19.Microhair length: 0 = short ( < 6 0 /tm), 1 margin: 0 = entire, 1 = ciliolate 24.Achene surface: 0 = rugose, 1 = colliculate, 2 = reticulate 25.Fruit type: 0 = caryopsis, 1 = achene 26.Ovary appendages: 0 = absent, 1 = present multistate, and are treated as undirectional in the analysis.The 'ie' option of the cladistic package HENNIG86 version 1.5 was used to analyse the data.Using this method, the complete set of most parsimonious trees is found (Platnick 1989).Choice o f outgroupPentaschistis is the obvious outgroup, although the recognition of a particular species or group of species within this genus as the closest to the study group is not possible on an a priori basis.Six species of Pentaschistis were therefore chosen as possible outgroup taxa.These taxa were chosen on the basis of inferences made by Ellis in his anatomical papers(1985c, 1986), and after discus sions with Dr H. P. Linder.The taxa chosen were P. aspera (Thunb.)Stapf, P. colorata (Steud.)Stapf, P. curvifolia (Schrad.)Stapf, P. eriostoma (Nees) Stapf, P. glandulosa (Schrad.)Linder and P. tortuosa (Trin.)Stapf.
used in cladistic analyses.Characters marked by asterisks above the data set were removed in the data set where Pentaschistis curvifolia was used as the outgroup (data set delimited below the dashed line).Characters appear in the order presented above.The question mark (? FIGURE 3.-Shortest cladogram produced by HENNIG86 using the 'ie' option.The outgroup used was Pentaschistis curvifolia (Schrad.)Stapf.Length = 34, c.i. * 64, r.i.= 70.Numbers represent characters (given in Tables 2 & 3; 'a' = state 0, 'b' = state I and 'c' = state 2).Solid bars represent apomorphies.parallel lines represent parallelisms and crosses represent character state reversals.Star at base of Pentameris clade represents apomorphic fruit characters: achene fruit type and presence of apical hairs on ovary.
FIGURE 4 .-Pentamerisuniflora N.P. Barker.A, habit with decumbent, branching culm basally covered in dead, appressed leaf sheaths; B, ligule, which is a row of hairs, and surrounding leaf parts; C. panicle subcontracted to lax. with few spikelets D -G .parts of spikelet and florets: D, glumes; E, lemma (in its normal conformation) showing lemma bristles, lobes (which are adnate to the bristle for most of their length) and geniculate awn; F, palea; G, ovary with apical hairs and stigmas.A. x 0.5; B, D -F , x 4.7; C, x 1.6; G. x 11.8.

Vouchers:
Bond 1581 (SAAS); Ellis 2546 (PRE); Esterhuysen 25025, 32718a (BOL); Esterhuysen 31771 (BOL, PRE).Plants caespitose or decumbent; culms 0.46-1.63m long; leaf sheaths pubescent, appressed to culm, with purple auricle at mouth, persistent; leaf blades up to 500 mm long, folded or flat (rarely rolled); panicle globose, lax, 7 0 -2 2 0 x 40-1'TO mm; spikelets 16-90, twoflowered; glumes 15.5-21.5 x 1.8-3.0mm ; lemma body 2 the epithet thuarii, as used in the protologue, differs so obviously from that of the name of the collector, Du Petit-Thouars, that it is not considered a typographic error.The epithet thouarsii, used in a number of subsequent accounts, is therefore not adopted.It is plausible that Palisot de Beauvois wished to maintain the phonetic pronunciation of the French name 'Thouars', and thus spelt it 'thuarii' in latinised form.No locality or specimen number of the type specimen was cited by Palisot de Beauvois.However, two specimens of P thuarii were obtained from the Paris herbarium (P).One of these specimens is annotated with 'Herb.Du Petit Thouars', which is written on the bottom of the label.The identity of the handwriting is not known.That one (or both) of these specimens are type material is further supported by the observation that the illustration accom panying the original description includes a representation of a mature fruit, a description of which also appears in the text (Palisot de Beauvois 1812).The type specimen must therefore have borne mature fruit.TTiis deduction is compatible with the historical record which indicates that Du Petit Thouars visited the Cape in Febuary 1793.
from a woody base; culms erect, up to 1.4 m tall, unbranched or branched close to woody base; leaf sheaths loose and free from culm, wide, persistent, clustered at base of plant, shortly pubescent or glabrous, straw-coloured or purple, sheath mouth some times with green auricles; leaf blades rigid, rolled, up to 550 mm long; panicle lax, globose, up to 300 x 140 mm; spikelets 30-100 or more, two-flowered; glumes 15recognised on the basis of a number of floral and vegetative characters, as listed in Table5.The significance of these differences is demonstrated by the results of a principal components analysis (PCA) subsp.g y m n o c o le a

Vouchers:
FIGURE Yd.-Pentameris hirtiglumis N.P. Barker, type specimen Kerfoot 6092.A, habit: decumbent, branching culm basally covered in dead leaf sheaths; note that flowering culm is not normally curved back upon itself, but appeared so on specimen.B -G , parts of spikelet and florets: B, hirsute glumes; C, whole basal floret; D, lemma (opened, flattened and viewed from inside of floret) showing venation, lemma bristles, lemma lobes (adnate to bristle for half or less o f their length) and geniculate awn; E, palea; F. anthers, with no filaments; G, ovary with apical hairs and stigmas.A, x 0.5; B -F , x 4.7; G, x 14.

TFigv
Y P E .-Cape,Worcester Div., Jona's Kop, common on shale band, or on peaty slopes, after fire, in different aspects but not on steep southern slopes, forming dense and, in some places, quite extensive patches, leaf tips very sharp, 5000 ft (1 500 m), 19 Dec. 1971, Esterhuysen 32681 (PRE, holo.!; BOL, iso.!).Plants cushion-like or densely bushy, branched basally; culms erect or somewhat decumbent in older, larger tufts and then with numerous nodes, up to 530 mm long; leaf sheaths persistent, pubescent along margins, appressed or free from culm, especially when dead, often purple when young; leaf blades up to 105 mm long, permanently rolled, falcate (less so in young plants), strongly pungent; panicle lanceolatecolliculately sculptured; flowering time September to December.Figures IE & 2B.The species can be distinguished by its permanently rolled, short (< 105 mm long), falcate leaf blades and the leaf sheaths which are pubescent at least along the margins.The specific epithet oreophila, or 'mountain-loving', is chosen because this species is only found at high altitudes in the Hottentots Holland, Riviersondereinde and Hex River Mountain ranges (Figure 5).Where found, the species is abundant, especially in years immediately after fire.Its high altitude habitat receives snow in the ' /inter months, and the cushion-like growth form (to tip; palea 4 -6 mm long; lodicules cuneate, glabrous or shortly ciliolate, rarely with arm-like extensions; anthers 3 .0-4 .5 mm long; fruit cuneate, 2.0 x 0.9 mm, surface rugosely sculptured; flowering time September to December.Figures 1 & 2. Stapf (1897) created the epithet dregeana for the partial concept of the taxon which Nees (1841) called Danthonia distichophylla.Stapf did not agree with Nees on the inclusion of Lehmann's description of Danthonia disticho phylla, and in a footnote states that Lehmann's descrip tion probably refers to a Pentaschistis.It thus appears that Stapf did not see the type of Lehmann's Danthonia distichophylla.Lehmann's description (1831) states that he has seen the specimen in a dried state (v.s.= vide siccam).According to Nordenstam (1980), Lehmann's herbarium comprised many specimens collected by Ecklon & Zeyher.The collections of Lehmann now reside in the Swedish Museum of Natural History (S), from which the type of Danthonia distichophylla was obtained.The label of this specimen had been annotated (possibly by Nordenstam) to the effect that it is written in Lehmann's hand.Further confirmation of this was obtained by matching the script with that of published examples of Lehmann's handwriting (Burdet 1976; Nordenstam 1980).As there are no other labels on the specimen, the original collector of this specimen is unknown, but as Lehmann never visited southern Africa, it is possible that the specimen is an Ecklon & Zeyher collection.Despite Stapfs comment, the Lehmann specimen is a good match to the specimens of P. dregeana cited by Stapf (1897).The name P dregeana is therefore illegitimate as it is antedated by Lehmann's P. distichophylla.This taxon is therefore correctly named Pentameris distichophylla (Lehm.)Nees.As there is no conclusive evidence that the Lehmann specimen from S is the holotype, it is designated as the lectotype.This species is one of the most widely distributed in the genus, ranging from the northern Cedarberg to Raarl and eastwards to the Tsitsikamma Mountains (Figure 13).Vouchers: Esterhuysen 22352, 27321 (BOL.PRE); Hafstrom A Aithks 45 (PRE); Gillett 3835 (STE); Taylor 11590 (PRE.Plants caespitose or somewhat decumbent, basally branched; culms up to 560 mm long; leaf sheaths glabrous, appressed to culm; leaf blades up to 230 mm long, folded or rolled, sparsely pubescent near base; panicle lanceo late

FIGURE
FIGURE 12. -Pentameris oreophila N.R Barker.A, habit: erect culms and prickly, cushion-like basal leaf growth; note also pubescent leaf sheaths loosely appressed to stems, falcate, permanently rolled leaf blades with pungent apices.B-G, parts of spikelet and florets: B, glumes; C, whole basal floret; D, lemma (opened, flattened and viewed from inside of floret) showing venation, lem ma bristles, lemma lobes (adnate to bristles for most of their length) and geniculate awn; E, palea; F, anthers with no filaments; G, ovary with apical hairs and stigmas.A, X 0.3; B -F , x 3.7; G, X II.
specimen, undoubtedly belonging to the genus Pentameris (possessing an achene with apical hairs) was collected from Cockscomb Peak by Ellis in 1987.This specimen, Ellis 5605 (PRE), is morphologi cally similar in certain respects to P glacialis, but differs in its leaf anatomy.

FIGURE 14 .
FIGURE 14.-Pentameris swartbergensis N.P. Barker.A, habit: erect culms (folded back, based on type material) and branched basal parts.B-F, parts of spikelet and florets: B, glumes; C, whole basal floret; D, lemma (opened, flattened and viewed from inside of floret) showing venation, lemma bristles, lemma lobes (adnate to bristles for most of their length) and geniculate awn; E, palea; F, developing fruit with apical hairs, note that fruit appears to develop in a basipetal direction, uppermost region becoming swollen and associated surface sculptured, appearing at apex first.A, X 0.5; B -E , x 4.8; F, X 14.6.