A generic classification of the Restioneae ( Restionaceae ) , southern Africa

We propose a new generic classification of the African Restionaceae. tribe Restioneae (subfamily Restionoideae). based on the phylogeny and on extensive morphological data. The phylogeny is based on both plastid sequence data and morphological data. We delimit the genera to be monophyletic, to minimize the nomenclatural changes, and to maximize the ability to diagnose the genera. We recognize eight genera, one of w hich w ith nine subgenera, in the tribe. Of the currently accepted genera, only three need changes. We provide descriptions for all genera and subgenera, and include a kev to them. In this paper we erect one new genus, Soro\ eta. redelimit Platycaulos and Restio, and reduce Calopsis and Ischyrolepis to synonomy under Restio. We list the species which we recognize under each genus, make 37 new combinations, propose eight new names, and also describe eight new species that belong in these genera.


INTRODUCTION
The African Restionaceae (subfam.Restionoideae, ± 350 species) comprise one o f the dominant elements o f the fynbos vegetation o f the Cape Floristic Region (CFR.Goldblatt 1978) (Taylor 1978;Rebelo et al. 2006), and is regarded as one o f the most important clades o f the larger Cape flora (Linder 2003).W ithin the African Res tionaceae.the tribe Restioneae (Briggs & Linder 2009) includes most species (288).all o f which are endemic to sub-Saharan Africa or Madagascar.A user-friendly, maximally informativ e generic classification o f the tribe is thus essential for the effective study o f the Cape flora.
The taxonomy o f the Restioneae has been inv estigated several times over the past centuries, using different data sets and different taxonomic concepts.Most o f the publi cations from the 19th and 20th centuries w ere concerned w ith describing the many species which were brought to Europe by early collectors: for a summary, see Linder (1985).Howev er, sev eral publications also focused on the generic delimitations.In the 19th century, the morphol ogy o f the species was the primary ev idence on w hich the generic delimitation was based, resulting in the clas sification proposed by Masters (1878;1897) and used by Pillans (1928).This was based heavily on the gynoecial characters and remained in use until 1984 (Adamson & Salter 1950: Dyer 1976).Citing problems with the Mas ters classification.Gilg-Benedict (1930) based an alter native classification on the culm anatomical data o f her husband.Gilg (1891).but this classification was unfor tunately ignored.By 1980 it was ev ident that there w ere two major problems with the Masters-Pillans generic classification.Firstly, it did not effectiv ely summarize the new anatomical and palynological data, and so was not natural in a phenetic sense.Secondly, many o f the gen era were clearly based on plesiomorphic features, and so could not be monophyletic.In 1984.Linder proposed a new generic classification, incorporating a much wider data base than had previously been used-including ana tomical (Gilg 1891: Cutler 1969).palynological (Chanda 1966: Chanda & Rowley 1967: Linder & Ferguson 1985).and phytochemical (Harbome & Clifford 1969;Harbome 1979: Harbome et al. 1985) data in addition to morphology.He also attempted to define monophyletic genera.However, this classification still contained prob lems.Firstly, the data sampling was not complete, and the eventual classification proposed was based on the congru ence o f partial data-set-specific cladograms.Secondly, for several genera, no explicit morphological or anatomical synapomorphies could be identified, and so these genera could not be justified in a classic Hennigian sense.The first set o f problems was in the Elegia clade, where Chondropetalum lacked synapomorphies.Moline & Linder (2005) solved this problem by combining Chondropetalum, Elegia and Dovea.The next set o f problems was in the Restio clade.Under Restio, Linder noted: 'How ever.it is still not possible to demonstrate that Restio, as delimited here, is monophyletic'.Under Calopsis, Linder noted: 'the exact boundary between Restio and Calopsis is not yet resolved'.These problems remain unresolved.
Recently, Hardy et al. (2008) published an almost com pletely sampled phylogeny o f the African Restionaceae (subfam.Restionoideae), expanding the previously pub lished phylogeny for the nested Elegia group (Moline & Linder 2005) to include a complete species sampling of the entire tribe Restioneae (Figures 1-5).This study cor roborated earlier indications from a much more sparsely sampled phylogeny (Eldenas & Linder 2000) that several genera were para-or polyphyletic.It also provided the sampling density that enabled a test o f the monophyly of all previously recognized genera, as well as the discov ery o f many new' clades, some o f w hich w arrant descrip tion as new taxa.As such, the objective o f this article is to integrate this new information into a revised generic classification o f the Restioneae.In order to facilitate the use o f the new generic classification, we list the species accepted under each genus and subgenus, and we simul taneously make all necessary new combinations, and for mally describe several as yet undescribed species.

Theoretical criteria
The criteria for the circumscription and ranking of genera have received little attention and most authors simply indicate that they follow the criteria proposed by Backlund & Bremer (1998).These were expanded by Humphreys & Linder (2009), who argued for larger gen era.more consistent with the generic concept used by Bentham & Hooker in Genera plantarum.Thus, genera are not necessarily the smallest monophyletic or diagnosable groups.These, they suggest, may be better recog nized by formal infrageneric taxa.Our classification that follows represents our attempt to optimize our adherence to the following criteria: 1. monophyly; 2, diagnosability; 3. nomenclatural stability; and 4. informativeness.

1.
M onophyly.All genera and subgenera recognized here, were resolved as (potentially) monophyletic (i.e. as clades) by Hardy et al. (2008;summarized in Figures 1-5).Monophyletic taxa contain a closed segment of evo lutionary history; therefore, such taxa can be expected to have a much greater predictive power than taxa that are not monophyletic.Furthermore, monophyletic taxa are more comparable for evolutionary studies or conservation efforts which seek to account for phylogenetic diversity in addition to species diversity in management decisions.
However, in this context monophyly refers to the 'true' phylogeny.The data collected by phylogeneticists only allow us to estimate this phylogeny.These estimates con tain several sources of error, two o f which are pertinent to our study.Firstly, different genome partitions may reflect different phylogenetic histories (Doyle 1992;Maddison 1997).Such incongruent phylogenetic histories may result from the transfer o f a chloroplast from one species to another during hybridization.In this case, a phylogeny based on the chloroplast plastid genome may not reflect the phylogeny of the nuclear genome.The second error may occur when stochastic variation in DNA base pairs results in the retrieval of clades.In this case, the shared DNA base pairs that lead to the retrieval of a clade are not the result of a common history.An extreme, and pos sibly quite rare form o f this phenomenon results in long branch attraction (Felsenstein 1978).Much more com mon are nodes with relatively little statistical support.
The phylogenetic inferences of Hardy et al. (2008) were based on both parsimony and Bayesian analysis of a mod erately large dataset consisting of plastid DNA sequences (± 7.3 Kb) and 150 morphological and anatomical charac ters for all 292 species and subspecies of the Restioneae.The robustness of these inferences, how ever, is limited by the inability of Hardy et al. (2008) to successfully generate nuclear DNA sequences.As such, the phylogenetic infer ences of Hardy et al. (2008) were strongly influenced by evolutionary history of the plastid genome w hich, because of problems associated w ith phenomena such as introgressive hybridization (e.g.chloroplast capture) or lineage sort ing.may not faithfully reflect the organismal phylogeny in all aspects (Doyle 1992;Maddison 1997).Additionally, a plastid-specific bias or imbalance in molecular evolution ary rates across the phylogeny could result in phyloge netic error unchecked by another, unlinked nuclear DNA dataset (e.g.Felsenstein 1978;Doyle 1992).Although the morphological and anatomical data are thought to provide phylogenetic signal under the influence of the nuclear genome and therefore independent of the plastid dataset, the relatively large size of the plastid dataset may have had a disproportionate influence on the resulting phylogenetic inferences.However, because the addition of our mor phological data to the plastid data in a combined analysis resulted in increased support (via bootstrap and posterior probabilities) for most clades than did either data set alone (Hardy et al. 2008), we are confident that the phylogeny is a good approximation of the species phylogeny.
A second problem with our phylogeny is that several of the deeper nodes from the strict consensus trees have very low bootstrap support.This could mean that these nodes will not be retrieved if a larger dataset were to be generated.

2.
Diagnosahility.In addition to monophyletic groups, we sought to recognize genera that were diagnosable mor phologically.Essential and differential morphological attributes can also be used for the assignation o f species to the genera, and to dev elop a morphological concept of a genus.These characters were sought by using the parsi mony options in Mesquite (Maddison & Maddison 2003) to optimize the morphological and anatomical character matrix used in Hardy et al. (2008; available at http://herbarium.millersville.edu/pubs-support.php)over the com plete cladogram from Hardy et al. (2008).We defined as diagnostic characters those character states that were opti mized to the crown-node (base) o f each clade that could be treated as a genus or subgenus, i.e. we ignored variation within the clade.We then coded these diagnostic charac ters (Table 1) for the clades into a new matrix (Table 2) and then mapped them onto a summary tree o f the clades (Figure 1) using CLADOS (Nixon 1993).Morphological concepts and states were largely as used by Linder (1984), and are explained, illustrated, and fully documented in the interactive key to the species (Linder 2001a) (http://www.systbot.uzh.ch/Bestim m ungsschluessel/R estionaceae.html).Species for which molecular data were not available (20 out of 289) were placed on the basis o f morphological and anatomical data, by their inclusion in a cladistic analy sis with the full data set.Essential and differential morpho logical and anatomical attributes were extracted from the morphological data set for each postulated clade.Essential attributes are typical (albeit sometimes with exceptions) o f each clade, but not unique to it.Differential attributes, often as combinations o f characters, are unique to each clade.and can be used to diagnose the clades.Differential attributes can be used to assign previously unstudied spe cies to their appropriate clades.
A second important component o f diagnosability is the ability o f users to be able to assign species to the right genus without too complex a procedure, ideally even without having to use a hand-lens.Thus diagnos ability that is only possible on anatomical attributes is not optimal.3. N om enclatural stability.Genera should be recog nized at nodes that will minimize nomenclatural changes.Nomenclatural changes fall into three elements.Firstly, minimizing the number o f new combinations: these are most common when genera are divided into segregate genera; in these instances the specific epithet stays the same.Secondly, reducing the number o f new names that need to be proposed: these are most common when gen era are combined, and result when a specific epithet is already occupied; these are much more confusing than new combinations, as the species now has a completely Culms: round or flattened = 0; square or angular = 1.
different name (both genus and species have changed).
Thirdly, frequent changes in the combinations: this results when there is a frequent changing of the generic concepts.
In some circumstances no new formal names are required, but from the user point of view there are still changes.
A second consideration should be not only to mini mize the number o f nomenclatural changes required, but also to consider the nomenclatural consequences o f future changes in our understanding o f the phylogeny.Ideally, genera should be erected at nodes which are likely to remain robust to future additional data, be it new species or new data sources.Theoretically, at least, larger genera should be less sensitive to smaller changes in the phylogenetic hypothesis.

4.
Inform ativeness.Genera should have a morpho logical and ecological "coherence', and not consist of mere collections o f species that happen to share a plastid genome.This maximizes the number o f attributes that we can assign to a genus, and makes the genus concept more useful.

Genera in the Restioneae
With the cladistic and morphological data available for the Restioneae.several generic classifications for the Restioneae are possible.In order to facilitate the discus sion here, w e refer to the names o f clades described later in the paper.

1.
The maximal option is to include the whole sub tribe in one genus.This genus would be strongly sup ported as being monophyletic by the plastid data (Eldenas & Linder 2000: Hardy et al. 2008).but w ould be difficult to diagnose, as the morphological characters separating Willdenow ieae and Restioneae are cryptic anatomical and palynological attributes (Linder 1984).and the clade has no diagnostic attributes.The characters listed in Figure 1 as synapomorphies for the clade (the numbers o f styles and carpels), should actually be listed under the Willde-  nowieae-the Restioneae have the ancestral condition of three styles and three carpels.The genus would not be exceptionally large (some 300 species), but would be mor phologically very heterogenous.It would require numer ous name changes, and would lose on informativeness, as several very distinct genera, such as Elegia, Thamnochortus and Staberoha, would be lost.This broad approach has been followed in several Cape taxa, such as Erica (Oliver 2000) and Disa (Bytebier et al. 2007(Bytebier et al. , 2008)).
2. The m inim al option is to retain all existing genera that were retrieved as monophyletic (i.Note that Platycaulos is here somewhat unusual, as this genus receives several species misplaced to Restio in the previous classifications.The recognition o f more than one genus in the tribe implies that the basal branch leading to Restio ambiguus has to be recognized at generic level (as the genus Soroveta), as the node excluding this species from the rest o f the tribe is very strongly supported.This results in 16 genera, all o f which are strongly supported as being monophyletic by the combined plastid DNA and morphological datasets, but not all by the morphological dataset alone.There are great difficulties in diagnosing several o f these clades.In some instances no essential or differential characters could be found (e.g.Restio subgen.Varirestio), in others (e.g.Restio subgen.Craspedolepis, Restio subgen.Eremorestio, Restio subgen.Locapsis, Res tio s.str., Restio subgen.Simplicaulos) there are essential but no differential characters, i.e. they also occur in other genera albeit relatively rarely.As a result we can have a vague concept o f the genus, but cannot key it out.Predict ably, these smaller genera are mostly morphologically quite homogeneous.This approach requires more than 70 new combinations, but no new names.
The advantage o f this approach is that the monophyly o f all genera is strongly supported by the available phy logenetic hypothesis.However, there are two problems.The first is that the generic classification will not be user friendly, as it will be difficult to assign species to the gen era.either in the field or in a herbarium.Secondly, there is a very heavy reliance on plastid data in the absence of nuclear DNA data (Doyle 1992).Any mistakes in the phylogeny will result in generic redelimitation.

3.
The interm ediate option is to retain all existing genera that are not proven to be para-or polyphyletic, and not to start w ith the assumption that all existing genera that are retrieved as monophyletic have to be retained.In this option, the distinctive and well-supported genera Askid iosperma, Elegia, Platy caulos, Rhodocoma, Soroveta, Sta beroha and Thamnochortus are retained, and a larger Res tio s.lat. is assembled.The major advantage o f this option is that the two difficult segregates.Eremorestio and Vari restio.will not be recognized as distinct genera.Due to the structure o f the phylogeny, in all these solutions, Ischy rolepis will need to be included in Restio s.lat., together w ith most species o f Calopsis, as the nodes embedding these two genera within Restio s.lat.are strongly supported in the combined analysis.In the morphological phylogeny of Linder (1984) Ischyrolepis w as not included in the Restio-Calopsis clade, due to its very different pollen mor phology.A major insight from the molecular phylogeny is that the pollen data are not phylogenetically conservative, as a result Ischyrolepis is found to be nested in the Restio-Calopsis clade.Three variants of Restio s.lat.can be proposed: 3a.Include Calopsis s. str., Craspedolepis, Eremores tio, Ischyrolepis, Locapsis, Pendulostemon, Restio s.str., Simplicaulos and Varirestio in one genus.There is weak evidence that this grouping is paraphyletic: in the com bined phylogeny, but with no bootstrap support, Pendu lostemon groups with the Thamnochortus-Rhodocoma clade, and Simplicaulos with Staberoha).This large genus is easy to diagnose by exclusion (not Rhodocoma, Thamnochortus, Soroveta, Platycaulos, Elegia or Askid iosperma).It is the most conservative approach, requir ing statistically significant evidence to dismantle a genus.The disadvantage o f this classification is that there is no evidence for the monophyly o f the genus Restio s.lat.(but also no evidence that it is not monophyletic).The advantage is that the genus is easy to recognize in the field and herbarium and that it is nomenclaturally con servative (if eventually there is sufficient evidence that to separate Simplicaulos and Pendulostemon will not require undoing a set o f nomenclatural changes).The strongly supported monophyletic units can then be rec ognized as subgenera o f Restio s.lat., these should be used for evolutionary investigations.3b.As above, but retain Simplicaulos and Penduloste mon as distinct genera on the basis that there is weak evidence linking these to Staberoha and Thamnochortus respectively.It removes all evidence that Restio s.lat.could be paraphyletic.but there is also no evidence that the remaining Restio s.lat. is monophyletic.Penduloste mon can be readily diagnosed, but Simplicaulos lacks differential characters.The most important attribute is the unbranched culms: this is a highly variable character.The advantage is that no genera are recognized for which there is (even poorly supported) evidence for paraphyly.The disadvantages are two-fold: due to the low' support values, it remains possible that additional datasets will group the two segregates again with Restio s.lat., leading to name changes.Furthermore, one o f the two segregate genera cannot be keyed out.3c.Recognize, in addition to Simplicaulos and Pen dulostemon, also Craspedolepis.This leaves the rest of Restio s.lat.with positive (albeit weak) evidence of its monophyly, but it adds another genus that is difficult to diagnose.Frustratingly, Craspedolepis does have a strik ing essential character, a band of hollow (concave) cells at the apical margins of the floral bracts, but this is in some species poorly developed, and also occurs in a few species outside the genus.It is thus of little use as a differential character.
We follow here option 3a, which recognizes eight genera in the tribe.The eight strictly monophyletic seg regates o f Restio s.lat.are recognized as subgenera.We hope that this will combine the need for strictly mono phyletic taxa for evolutionary analysis (the subgenera).with easy-to-recognize, nomenclaturally stable genera.It is important, though, that no taxa that are demonstrably para-or polyphyletic be recognized.

TAXONOMY
The full nomenclature and synonymy o f the species is not repeated: it was published in 1985.consequently only the accepted species and their place o f publica tion is given.However, for the new combinations, the basionym and the type is indicated, as required.Again no full synonymy is presented.The complete synonymy, descriptions, illustrations to all species, and an interac tive key is available online at http: Vwww.systbot.uzh. ch Bestimmungsschluessel Restionaceae.html.or can be bought as a CD from the Bolus Herbarium.University o f Cape Town.The species are presented in a 'natural' sequence, partially based on the phylogeny, partially based on their similarity.Notes: the isolated position o f this species was already recognized by Masters (1897) who gave it the very prescient specific epithet.Morphologically, the species shows several unusual features in the male and female inflorescences, notably resembling Elegia in the naked female flowers without any lateral (dorsiventral) flattening, but without the caducous sheaths and other attributes typical o f that clade.Furthermore, the dark brown floral bracts w ith transverse pitting are unique in the subfamily.However, there are no obvious synapo morphies for the rest o f the Restioneae, consequently the basal position o f this species cannot be detected morpho logically.In the absence o f molecular data, the species would probably have been retained as a highly special ized species o f Restio.P.Linder in Bo thalia 15: 436 (1985).2.12.Platy caulos callistachvus (Kunth) H. P.Linder in Bo thalia 15:436(1985).

Staberoha-Elegia clade
Although molecular support for this clade is poor (see also Hardy et al. 2008), this clade makes sense morpho logically, based largely on its combination of glabrous, conduplicate tepals and, mostly, simple, unbranched culms.Interestingly, the species in the family with pen dulous male spikelets (e.g.  2) is weak, and an alternative posi tion is for Simplicaulos to be sister to Elegia, and for Sta beroha to occupy an isolated position (Hardy et al. 2008).The two taxa may be linked morphologically by their sim ple culms, but that could be plesiomorphic in the whole clade, with a reversal in Restio subgen.Pendulostemon.Many species in these two taxa also have large female bracts that completely obscure the female flowers.
Restio subgen.Pendulostemon, consisting of just two species, is both phylogenetically and morphologically isolated in the clade.In particular, the branching culms o f this genus are unusual for the clade, although species with branching culms also occur occasionally in Elegia and Thamnochortus.
The Thamnochortus-Rhodocoma subclade was previ ously recognized by Linder (1984).The two genera share pendulous male spikelets (also present in Staberoha and Restio subgen.Pendulostemon), scattered cavities in the central ground tissue, and sheaths o f which the upper half is more or less membranous and soon decays.The genera have very different gynoecia: in Rhodocoma the ovaries are dehiscent, and the diaspore is a seed, w hereas in Thamnochortus the ovaries are unilocular and indehis cent.and the fruit a soft-walled nut, included in an often w idely w inged, persistent perianth.
The Elegia-Askidiosperma subclade also dates back to the morphological phylogeny o f Linder (1984).and is recognized by the caducous, or w hen persistent, poorly structured sheaths, and the inflorescences with numerous partial panicles originating from several nodes along a central axis.Furthermore, the tepals are quite bony, w ith generally no differentiation between the lateral and odd tepals.The central ground tissue lacks a cavity.Elegia is distinguished from Askidiosperma by a double-layered culm epidermis, whereas Askidiosperma is diagnosed by the very long, membranous bracts, and its more welldefined spikelets.
Notes: a very distinctive genus, almost restricted to the Western Cape, where often several species are found co-occurring.The ball-like, pendulous male spikelets and the small female flowers obscured behind the large female bracts allow an easy recognition of the genus.All species are found on well-drained soils, both on deep sandy soils (such as Staberoha distachyos) and on rocky slopes (S. remota).S. stokoei is a remarkable exception to many o f the morphological attributes (the male spike lets are erect, and almost identical to the female spike lets), and is a geographically restricted species in the Great Swartberg, where it occupies almost vertical rock ledges on the upper south-facing slopes.It is satisfying to see that it occupies a topological position sister to the rest o f the genus, indicating that many o f the distinctive attributes of the genus evolved after the origin o f the genus.If subdiv ision of the genus is appropriate, then S. stokoei could be regarded as one division, and the rest of the genus the second.
Notes: mostly associated with more arid areas, espe cially in the northwestern Cape, along the coastal plains ( Thamnochortus baclwiannii).Nieuwoudtv ille ( T. platypteris).but also typical o f the wettest fynbos vegeta tion (T.cinereus).Also noted for the important thatch ing plants ( T. insignis).The plants often form extensive stands, and can locally dominate the vegetation.Striking in some species are the w idely w inged, persistent peri anths.making for efficient wind dispersal o f the seed.
This genus is very clearly demarcated, and is very readily diagnosed by the slender, pendulous male spike lets, the closely convoluted sheaths with a decaying upper half, the many-flowered, erect female spikelets.and the single plumose style.The closest relative is Rhodocoma.with which it shares features o f the culm anatomy and pollen structure, as well as the pendulous male spikelets.However.Thamnochortus has an inde hiscent.unilocular ovary, whereas Rhodocoma has a dehiscent, three-locular ovary.In many ways this rela tionship is similar to that between Chondropetalum and Elegia s.str.. but in the case o f Thamnochortus there are no morphological intermediates and there is stronu molecular evidence for the reciprocal monophyly o f the two genera.
The biogeography and evolution in this genus was investigated by Linder & Mann (1998).It is possible to assign the species to five main groups, not all o f w hich are monophyletic (Figure 2
Plants caespitose or with straight rhizomes.Fertile culms round, unbranched or in R. capensis much branched with branches whorled at each node; sheaths persistent, upper half chartaceous and soon decaying, leaving a lac erated remnant; hyaline shoulders absent.Male inflores cence racemose from several nodes, each partial inflor escence with several to numerous pendulous spikelets on flexible pedicels; spikelets ± elliptical; bracts without distinct membranous upper margin; anthers included in flowers at anthesis.Female inflorescence similar to male, but partial inflorescences stiffly erect with few to several spikelets; spathes sometimes prominent and membranous or chartaceous; spikelets elliptical to ovate, w ith 1-5 flow ers; bracts shorter than flowers.Female flowers not later ally compressed; tepals bony, glabrous and smooth; stami nodes present.Gynoecium w ith 3, white to red.feathery styles free to base; ovary 3-locular, dehiscent or rarely falling with seed.Seed brown, grey or tan, colliculate or rugose.Culm anatomy: epidermal cells I-layered, lateral walls straight and unthickened or sinuose and thickened; parenchyma of 1-4 rows of cells; central ground tissue w ith scattered cavities; tannin present in epidermis, paren chyma, sclerenchyma and central ground tissue; silica usually absent, or when present, in parenchyma.
Notes: most o f the species are found in the south ern Cape, w here a clear pattern o f ecological and geo graphical replacement is evident.Rhodocoma fruticosa is w idespread from the Cape to the Drakensberg, and is very common over large areas along the eastern margins o f the fynbos.
Superficially the species are similar to Thamnochor tus, especially because o f the pendulous male spikelets and the similar culm and sheath construction.However, they can be separated by the dehiscent female ovaries, by the short bracts in the females, and less confidently by the shorter and more elliptical spikelets.
Etymology: Vleiberg.a little known peak in the Swartruggens, from which this new species was first recorded.
Distribution and ecology: endemic to the western margins o f the Tanqua Karoo, from the Bonteberg at Touws River, northwards via Baviaansberg.Vleiberg to the Bokkeveld Sneeukop.The altitude is from 1 000-1 600 m. the habitat restricted to sandstone-or quartzite-derived soils, where populations are found in welldrained habitats on rocky slopes.The plants are killed by fire, and populations are re-established from seed.
Discussion: this species is morphologically very simi lar to Rhodocoma arida.but w ith a different 'gestalt'.the plants being smaller, tattier looking, and spreading with a more developed rhizome.The spikelets are generally larger.This applies particularly to the female spikelets.Furthermore, there is a geographical separation, with R. vleibergensis found north of the Laingsburg Witteberg.and R. arida east of this line.This also determines the pro portion of summer rain in the area, w ith R. arida receiving a substantial proportion of its rain in summer, whereas R. vleibergensis receiv es almost no summer rain.Despite the morphological similarity to R. arida.the phylogeny of Hardy & Linder (2007) indicates that the species is more closely related to the R. alpina-R.fruticosa subclade.Plants tufted, clumped or mat-forming, often with well-developed, spreading rhizomes.Fertile culms mostly unbranched, in some species branched, and in two spe cies branches w horled at each node, mostly round, some times ± compressed: sheaths sometimes persistent, but more commonly caducous with distinct abscission line.when present, most commonly loosely convoluted and same texture over whole structure, rarely upper margins membranous or otherwise different from rest of organ.Male inflorescence o f several clusters o f partial, paniculate inflorescences, each w ith several to very numerous flowers only poorly organized into spikelets; young partial inflo rescences generally hidden behind large spathes, which drop before anthesis; large spathe-like spathellae occasion ally present; male bracts usually shorter than perianth, but in appearance similar to tepals.Male flowers w ith undif ferentiated.glabrous tepals; anthers included in flowers at anthesis.Female inflorescence structurally similar to male inflorescences; each partial inflorescence with 1-many flowers, these poorly organized into spikelets; spathes large, usually persistent, often completely obscuring flow ers at anthesis; spikelets often clustered in paniculate par tial inflorescences, making them difficult to delimit; bracts mostly shorter than flowers w ith no differentiation between body and margin, rarely acuminate and/or lacerated on margins.Female flowers w ith undifferentiated, glabrous tepals; staminodes usually present.Gynoecium with 2 or 3 mostly feathery styles free to base or more rarely seated on a stylopodium; ovary with 3 usually dehiscent or 1 inde hiscent locule.Nuts generally triangular, often black, with out elaiosomes, perianth sometimes persistent on nuts but never w inged; seeds from dehiscent ovaries often brown and smooth.Culm anatomy: epidermal cells almost always in 2 layers, lateral wall straight, unthickened; parenchyma tous layer o f 1-6 cells; central ground tissue solid or with scattered cavities, rarely w ith a single, central cavity; tan nin usually in epidermis and central ground tissue, rarely in parenchyma or sclerenchyma; silica absent.

Additional collections
Etymology '. Elegia (f.), from the Greek elegos.mean ing weeping or lament, and this was then used by Plinius in the Latinized form o f Elegia to refer to a kind o f reed.
Notes', this genus contains a remarkably w ide range of variation in morphology, previously recognized as four genera.However, this variation does not readily group into reciprocally monophyletic groups, and consequently the whole group is best treated as one large genus, as sug gested by Moline & Linder (2005).This larger clade is readily diagnosed by anatomical attributes (epidermis in two layers: Figure 1).The morphological attributes (cadu cous sheaths; paniculate inflorescences) have numerous exceptions, which makes them more difficult to use.
The genus contains several well-known species, such as Elegia tectorum, widely cultivated, previously w idely used as a thatching reed: E. capensis, w idely cultivated, dominant in many areas in regions w ith some groundwa ter.where the stoloniferous growth allows the species to take over the vegetation; E. fla c e a .w hich has an enor mous ecological range, and is found in almost all veg etation types.
The species are organized into four clades (Figure 3).

a)
Elegia mucronata clade, without obvious charac ters:   Discussion: this species is very close to Elegia tectorum, from which it primarily differs in being bigger in all dimensions.Furthermore, the inner and outer whorls o f the female tepals are the same length, whereas in E. tectonim the inner tepals are longer than the outer.Both species co-occur on the west coast at Blouberg. and
Plants caespitose.Fertile culms unbranched, round; sheaths dropping off, with an abscission line present, loosely convoluted or flat and standing free from culm, margins coriaceous, like rest o f body, hyaline shoul ders absent.Male inflorescence with I to several partial inflorescences subtended by large spathes; each partial inflorescence with I to several, many-flowered spikelets; bracts much taller than flowers, papery or membranous, margins entire or lacerated.Male flowers not laterally compressed; tepals glabrous, inner and outer often o f different lengths; anthers included in flowers at anthesis.Female inflorescence, spikelets and flowers similar to male.Female flowers with staminodes.Gynoecium: styles 2 or 3, white, feathery, free to base; ovary I 3-locular.dehiscent or sometimes dropping with seed.Seed silvery or more commonly brown and smooth.Culm anatomy: epidermis with a single layer o f cells, lateral walls straight, unthickened; parenchyma o f 1-3 layers o f cells; central ground tissue solid, rarely w ith a single, central cavity; tannin usually absent, when present in epidermis, sclerenchyma or (rarely) in central ground tis sue; silica usually absent, when present, in parenchyma or sclerenchyma.
Etymology> : Askidiosperma (n.): from Greek askion or askidion, which is the diminutive o f askos.w hich is a bag, sac, wine skin or hide; and sperma (n" Greek), which is a seed.The connection to the genus is unclear.
Notes: this genus shares with Elegia the caducous sheaths, and the tepals which are all the same (outer lat eral tepals are not flattened and keeled).The species had been included in Chondropetalum by Pillans (1928).but were again separated by Linder (1984), as they differ from Elegia by the culms with simple (not doubled-layered) epidermis, and by the floral bracts which are hyaline-membranous and much taller than the flowers, as well as by the phytochemical composition (Harbome et al. 1985).
The species are mostly found in the western moun tains, rather rarely in the southern mountains o f the Western Cape, and are absent from the coastal flats and plains.Species range from seepages over bedrock (Askidiosperma insigne) to alluvial soils along streams and on well-drained rocky slopes.

Restio clade
This large clade is readily defined by the laterally compressed female flowers, where the keels o f the outer lateral tepals are more or less villous.In most species the culms are branching.This clade largely includes most o f the segregates o f the genus Restio sensu Linder (1984) and Calopsis sensu Linder (1984).as well as Ischyrolepis.Several strongly supported subclades are recognized: Craspedolepis, Calopsis s.str.Restio s.str.
Locapsis and Ischyrolepis which are characterized by different apomorphies and apomorphic tendencies: Craspedolepis: diagnosed by the bracts with hollow (deeply concave) cells along the upper margin (honey comb cells), and the plants and spikelets tend to be robust, the ovaries mostly dehiscent, with two locules.
In many species the spikelets are relatively massive.
Calopsis: highly paniculate inflorescences with very many small spikelets; culms mostly angular or even square.
Restio: sheaths mostly with large membranous lobes flanking the mucro; the plants and spikelets tend to be small, the ovary mostly unilocular and often indehis cent.
Ischyrolepis: ovaries with two styles which are basally fused, this fused portion persisting as a peg on the dehisced ovary.
Varirestio and Eremorestio, however, lack characters, and are recognized in order to retain monophyly.
Morphologically.Eremorestio would fit with Locap sis, while Varirestio could be polymorphic.It remains possible that the apparent monophyly o f these two segre gates is an artifact o f ancient hybridization, and that the plastid phylogeny is not tracking the species phylogeny.
However, the phylogenetic structure within the clade is unclear.There is little molecular support for the group ings.and we cannot postulate good morphological char acters for the various molecular clades.
Also included in this clade are Pendulostemon and Simplicaulos.The balance o f the ev idence places these two segregates into the Staheroha-Elegia clade.but taxonomically they are better kept in the large Restio s.lat.Plants caespitose.clumped, tangled or mat-forming.Fertile culms simple or branching, round, or rarely square or compressed; sheaths persistent, closely or loosely convoluted, usually w ith a narrow membranous margin, often with tall, acute membranous lobes flank ing mucro.Male and fem ale inflorescences mostly very similar: mostly paniculate, more rarely racemose, with 1 to numerous spikelets; spathes usually persistent: bracts usually coriaceous, varying from shorter to taller than flowers; anthers exserted from flowers at anthesis.Female inflorescence occasionally with fewer spikelets than males; spikelets with up to 30 laterally compressed flowers; tepals chartaceous or cartilaginous, glabrous or sparsely (rarely densely) villous on keels o f lateral sepals; staminodes present.Gynoecium with 1, 2 or 3 feathery styles free to base, fused into a basal peg, or rarely seated on a stylopodium; ovary 1-, 2-or 3-locular, either dehiscent or indehiscent.Diaspore either a seed or a soft-walled nut; seed variously coloured, usually smooth, occasionally colliculate; nut dropped with a per sistent papery perianth that is sometimes winged.Culm anatomy: epidermal cells 1-layered, lateral walls straight and unthickened or sinuose and thickened; parenchyma tous layer o f up to 6 cells; central ground tissue either with no cavity, a single central cavity, or with scattered cavities; tannin rarely absent, when present then in the epidermis, sclerenchyma, or central ground tissue; silica either absent, or when present, in chlorenchyma or paren chyma.
Etymology: Restio (m., Latin), a ropemaker; it is unclear why Linnaeus referred to restios as such, since they are not to our knowledge useful for the making of ropes, due to the brittle nodes.
The extensive variation in the genus is best expressed in the eight subgenera recognized.This subgenus lacks any striking synapomorphies.However, optimization on the phylogeny shows that the subgenus ancestrally had simple culms, and that branched culms evolved only in the Restio bifarius group.Even in this group, branching is often absent, and when present, is sparse.The unbranched culms are shared with its possibly nearest relatives, Elegia and Sta beroha, but differentiate the subgenus from the rest of the genus Restio.Although it is easy to differentiate sub gen.Simplicaulos from its putative phylogenetic relatives Staberoha or Elegia by the persistent sheaths, by stami nodes still present in the female spikelets, the (mostly) erect male spikelets and the male flowers aggregated into clear spikelets, these are all plesiomorphic features.It is remarkable how variable the subgenus is regarding pre viously used generic characters, such as the number of locules and styles and the type o f diaspore (nut or seed).Not surprisingly, subgen.Simplicaulos contains spe cies previously placed in Restio sensu Linder 1985 and Calopsis sensu Linder 1985.
Although the clade is strongly supported, its position within the Restioneae is not.The plastid data suggests (weakly) a phylogenetic relationship to Staberoha or Elegia, but the possibility that it is related to Restio s.lat.cannot be rejected, even by plastid data.Ety mology: Pendulostemon (m .):pendulo( Latin).pen dulous; stemon (Greek), stamen: referring to the pendu lous male spikelets.
Notes: the two species o f this subgenus mentioned below were previously included in Restio s.str because o f their dehiscent ovaries, laterally flattened flowers and branched culms.Their current position is uncertain, but plastid data weakly indicate a relationship to Tham nochortus and Rhodocoma (Figure 2) with which they share the pendulous male spikelets.However, a closer relationship to Restio s.str cannot be rejected.
Although the two species had not previously been asso ciated.they share a number of morphological attributes.The plants are rather untidy, erect tussocks with robust, erect, branching culms.The sheaths are coriaceous with out a substantial membranous margin.The spikelets.both male and female, are spindle-shaped and relatively large.The male spikelets are pendulous.The only unique feature is the longitudinal striation on the bracts, but this feature is not always easy to observe.Restio micans is a rare spe cies on coastal sands, R. egregius is more w idespread in the w etter mountains of the Western Cape.Plants caespitose to tangled, without spreading rhi zomes.Fertile culms mostly branching, round or rarely compressed; sheaths persistent, usually closely convo luted.apical margin ± membranous, sometimes w ith large membranous shoulders flanking mucro.Male inflores cence racemose or paniculate, with up to 10 (rarely more) usually erect, mostly elliptical, spikelets (2 spp.w ith pen dulous spikelets); male bract with upper margin usually with honeycombed cells, which eventually decay, taller than flowers; anthers exserted from flowers at anthesis.Female inflorescence similar to males, but spikelets are somewhat bigger, mostly elliptical, with up to 40 flow ers; female bracts imbricate, shorter to taller than flowers, apical margin with honeycombed cells.Female flowers: outer tepals differentiated into conduplicate lateral tepals and a flat odd tepal.rarely glabrous and smooth, usually densely v illous on keels o f conduplicate tepals.occasion ally also on back o f outer odd tepal: staminodes present.Gynoecium with 3. white, feathery styles free to base: ovary usually w ith 2. rarely 3 or 1. fertile locules.dehis cent.Seed variously coloured, surface smooth or colliculate.Culm anatomy: epidermal cells 1-layered, lateral wall straight and unthickened or sinuose and thickened: parenchyma up to 4-lavered: central ground tissue usually solid: tannin usually present in epidermis, sclerenchyma and central ground tissue: silica usually absent.
Etymology: Craspedolepis (f.): kraspedon (Greek), edge, border or fringe; lepis (Greek), scale.This name may be interpreted to refer to the upper margin o f the floral bracts (scales) which differ from the body o f the bract by the honeycombed cells, and are thus bordered.
Notes: this is a very distinctive segregate from Res tio s.str.The most notable synapomorphy are the hollow cells (honeycomb cells) that form the upper margins of the floral bracts.This is absent from one species in the subgenus, and similar structures are also found in a few species in two other subgenera.The other characters are more in the nature o f'ten d en cies': large, elliptical spike lets (where Restio insignis and R. strobilifer are extreme examples); very hairy tepals, with often all three outer tepals hairy along the keels; almost always with two fertile locules.This contrasts with Restio s.str with its tendency to unilocular, indehiscent or tardily dehiscent ovaries.
The subgenus is widespread especially in the west ern half o f the Cape Floristic Region (CFR), and is almost absent from the eastern parts o f the CFR.There is a remarkable range o f seed surface ornamentation, from colliculate to smooth.The inflorescence structure is also highly variable, from solitary spikelets, to manyspikeletted.paniculate structures.
Notes: this is a dramatically redefined concept of Calop sis, as it excludes the whole genus as currently recognized except the type species, and includes two species formerly placed in Restio.However, the three species share a very similar growth form: one can think o f them as being huge paniculate inflorescences o f which the lower portion is sterile.Segregating them into separate genera was based on an over-reliance on the ovary dehiscence (Restio with capsules, Calopsis with nuts).Furthermore, in all three species the culms are more or less angular, and in two of them they are sharply square.All three species are asso ciated with ground water, and C. paniculata is a typical streambank species, widespread in the CFR.8.5.1. Restio  Male inflorescence racemose or sparsely paniculate with 2-20 ± elliptical spikelets; spathes persistent, sometimes taller than spikelets; bracts acute to acuminate, slightly taller than flowers, apices somewhat darker than bod ies; anthers exserted from flowers at anthesis.Female inflorescence similar to males, but always racemose and with fewer, somewhat larger spikelets; bracts taller than flowers, bony or cartilaginous, acute to acuminate, apices recurved somewhat from spikelet, giving it a bristly appearance; flowers laterally compressed; tepals chartaceous, rounded, glabrous or keels o f lateral sepals sparsely villous; staminodes present.Gynoecium with 3, white, feathery styles tree to base; ovary unilocu lar, indehiscent.Diaspore a soft-walled nut enclosed in perianth, triangular, 2-3 mm long, perianth not winged.Culm anatomy: epidermal cells 1-layered, lateral wall straight, unthickened: parenchymatous layer o f 2 or 3 cells; central ground tissue solid; tannin, when present, in epidermis or central ground tissue; silica absent.
Etymology: Eremorestio (m.): Eremia (Greek), desert; Restio (Latin), ropemaker; restios living in arid areas, referring to the habitat o f the two species in this genus, occurring along the arid margins o f the Cape Floristic Region.
Notes: similar to the situation in Varirestio, this sub genus o f two species is recognized in order to be able to separate the larger subgenera Restio s.str and Ischy rolepis.Contrary to the situation in Varirestio.however, there are numerous similarities between the two species.They have similar diaspores (soft-walled nuts dispersed in a papery perianth).Such diaspores defined the genus Calopsis sensu Linder 1984.The papery perianth and rather brittle bracts are also typical o f this group.Mor phologically the genus is very close to Locapsis.and it is not clear whether monophyly would be significantly rejected it the two were combined.The morphological distinction between Eremorestio and Locapsis is not easy, as most of the characteristics of Eremorestio are also found in the larger subgen.Locapsis.Possibly the best attribute could be that the tips o f the floral bracts in the two species o f Eremorestio are darker than the bodies o f the bracts, and these tips are recurved away from the axis o f the spikelet.In Locapsis the bract tips are generally the same colour as the body o f the bracts, and the tips are closely rolled around the spikelets, resulting in slender spindle-shaped spikelets.However, it is not clear whether this attribute will be a useful, consistent key character.
These two species also occupy a typical Locapsis habitat, along the arid fringes o f the Cape.They have a disjunct distribution, with Restio levynsiae in the Swartruggens Mountains, and R. rigidus from the Voetpadsberg near Laingsburg to the Great Swartberg.15: 470(1985)]; BM!, BOL!, BR!, K!, P!, S!, WAG!, Z!).8.7. Restio subgen. Restio Plants caespitose, clumped, tangled or mat-form ing.Fertile culms branching (with the sole exception o f R. strictus), round; sheaths persistent, closely or loosely convoluted, usually with a narrow membranous mar gin, often with tall, acute, membranous lobes flank ing mucro.Male and female inflorescences mostly very similar; mostly paniculate, more rarely racemose, with 1 to numerous spikelets; spathes usually persistent, often overtopping spikelets; bracts varying from shorter to taller than flowers, without honeycombed cells in api cal parts; anthers exserted from flowers at anthesis.Female inflorescence occasionally with fewer spikelets than males; spikelets with up to 16 laterally compressed flowers; tepals chartaceous or cartilaginous, glabrous or sparsely (rarely densely) villous on' keels o f lateral sepals; staminodes present.Gynoecium with 3, white, feathery styles free to base or rarely seated on a stylopodium, ovary mostly unilocular, rarely with 2 locules, only in R. sejunctus with 3 locules.either dehiscent or inde hiscent.Diaspore either a seed or a soft-walled nut; seed variously coloured, usually smooth, occasionallv colliculate, nut dropped with a persistent papery perianth that is sometimes winged.Culm anatomy, epidermal cells 1-lay ered, lateral walls usually straight and unthickened, occa sionally sinuose and thickened; parenchymatous layer o f up to 6 cells; central ground tissue either with no cavity, a single central cavity, or with scattered cavities; tannin rarely absent, when present then in the epidermis, scler enchyma.or central ground tissue; silica either absent, or when present, in chlorenchyma or parenchyma.
Notes: Restio, as defined here, is still a very broad subgenus, including species from Calopsis and Restio sensu Linder 1984.Broadly speaking, it includes the generally more slender members o f Restio s.lat.. w ith a tendency towards small spikelets, to unilocular and often indehiscent ovaries, and to more membranous sheaths and bracts.This contrasts to subgen.Craspedolepis w hich includes the more robust Restio species.
Although there are no clearcut synapomorphies for the subgenus, there are a number o f attributes w hich in combination will often lead to a correct identification.Most species in the subgenus have tall membranous lobes flanking the mucro on the sheath.These are also found in many other genera (e.g.subgen.Ischyrolepis), but are rare in the other segregates o f Restio s.lat.In Restio s.str they are optimized as being ancestrally present.As such, this character can be seen as a synapomorphy for the subgenus, although it cannot be used to key out the subgenus, due to the absence o f this feature from many species o f Restio, and its presence in many species in other genera.In most species in the subgenus there is a single functional locule.Furthermore, in most cases this locule is also dehiscent.Typical o f the subge nus are rather small spikelets.often with few flowers, rather than the more massive spikelets characteristic of Craspedolepis.
Ecologically most species are found along the wetter coastal mountains, w here they often form an understorey under other vegetation.
Despite the substantial variation in the subgenus, it is not possible to formally recognize sections, but to facili tate an understanding o f the variation in this large taxon, the species are placed into informal groups.
Distribution and ecology: restricted to the Kleinrivierberge behind Hermanus.in the Western Cape, where it is found at an altitude o f 300-700 m, on Table Moun tain Sandstone.The species is restricted to wet habi tats. in seepages with Roridula, and along stream mar gins.These are found on the south-facing upper slopes of these low mountains, where they receive ample rain throughout the year.The plants grow into a dense tan gled understorey under the other vegetation.
Notes: the new species is very close to Restio communis from the Cape Peninsula, but it differs in the less robust spikelets.the more slender flowers, and the smaller female bracts.The male spikelets are 8-10 mm long (instead o f 15-20 mm), and have only I or 2 flowers (instead o f 6-12 flowers), and the male flowers are 2.0-3.5 mm long (instead o f 3.6-4.0mm long).There is a single collection from the Caledon Swartberg.w hich might be intermediate, or w hich might constitute yet another segregate.
Etymology:paluster(Latin), marshy; cola (Latin), dwel ler in; referring to the habitat of the species in boggy places.Plants mostly tufted, occasionally tending to mat-forming from spreading underground stolons.Fertile culms mostly branching, round; sheaths persistent, closely or loosely convoluted or often flat and standing free from culm, margins generally same texture as body of sheath, w ithout membranous upper margin or membranous shoul ders.Male and female inflorescences similar, racemose or paniculate, w ith up to 20 erect spikelets; spathes persist ent, very rarely overtopping spikelets; bracts concolorous.margins same texture and colour as body of bract, usually cartilaginous, as tall as or somewhat taller than flowers.Male flow ers w ith exserted anthers.Female spikelet gen erally narrow ly elliptical, w ith up to 12 flowers, as flowers mature bracts usually diverge, transforming spikelet from slender to bristly.Female flowers laterally compressed; female tepals chartaceous and apically rounded, keels of lateral sepals generally ± villous; staminodes present or absent.Gynoecium with 3. feathery, white to red styles free to base or seated on a stylopodium; ovary unilocular, indehiscent.Diaspores small, soft nuts, perianth persistent and ± enclosing nuts.Culm anatomy: epidermal cells in 1 layer, lateral walls straight and unthickened or sinuose and thickened; parenchyma (1)2( 3) cells wide; central ground tissue w ith a single central cavity (this lacking in R. ramosissimus) tannin usually absent, when present, in epidermis and central ground tissue; silica usually absent, but present in parenchyma o f R. rigoratus.
The name refers to the slender culms, also expressed in the specific epithet filiformis.Plants tufted or tangled, often with spreading sto lons, but very rarely with rhizomes.Fertile culms usu ally branching, rarely unbranched or branches whorled at nodes; round or very rarely compressed; sheaths per sistent, closely or more rarely loosely convoluted, very rarely standing free from culm; margins coriaceous or membranous, sometimes with tall membranous lobes flanking mucro.Male and fem ale inflorescences basi cally quite similar, racemose to paniculate, 1 to numerous spikelets, except that females generally have fewer and larger spikelets than males; in some species male spikelets tightly clustered, while females tend to have single spike lets.Male spikelets sometimes linear and curved, bracts obtuse to often acuminate and recurved, or extended into a slender awn; anthers exserted from flowers at anthesis.Female spikelets with up to 40 flowers; bracts shorter to taller than flowers, apically obtuse to often acuminate and recurved, or extended into a fine awn; flowers laterally compressed; tepals chartaceous or cartilaginous; glabrous or keels of lateral sepals more or less v illous; staminodes absent.Gynoecium with 2, white to red.feathery styles; style bases fused to form a pillar; ovary with 2( 1) locules, dehiscent.Seed translucent or variously coloured, smooth, pitted, colliculate.striate or rarely rugose.Culm anatomy: epidermal cells 1-layered, lateral walls straight and unthickened or sinuose and thickened; parenchyma tous layer 1-3 cells deep; central ground tissue solid or with a single, central cavity; tannin absent or present in epidermis, sclerenchyma or central ground tissue; silica absent or present in parenchyma and/or rarely in central ground tissue.
Etymology: Ischyrolepis (f.): Ischvros (Greek), strong; lepis (Greek), scale; presumably this refers to the usually quite tough floral scales which enclose the flowers.N otes: Restio subgen.Ischyrolepis is readily diag nosed by a basal stylar peg, formed by the fusion o f the two style arms.This persists on the ripe ovary, even after the seed has been released, as a distinctive peg.This is unique in the Restioneae.Furthermore, it usually has two styles (com pare to the three or one more typically found).The species can look superficially very similar to Restio s.str.or Restio subgen.Craspedolepis.The pollen type is held in common with Staberoha, but the two gen era are not phylogenetically closely related, according to the molecular data.
It is particularly common in the more arid parts o f the Greater Cape Floristic Region (Born et al. 2007), and is replaced along the w etter coastal mountains by the sub genera Craspedolepis and Restio.
There are a number o f groups in this large subgenus, but it is not possible to clearly diagnose them.Fur thermore, the subgenus, as currently defined, is so eas ily diagnosed that there is very little sense recognizing smaller groups at subgeneric level.However, further research may indicate that these subgroups deserve sec tional recognition.

a)
Female spikelets small and m ostly with single flowers; styles often red: 8.9.1.
Restio anom alus H .P .L in d e r sp.nov., a R. fe m in e u s marginibus vaginarum late m em branaceis, tepalis interioribus florum feminearum 3.0-4.5 mm longis, ovario biloculari statim dignoscenda.3419 (Caledon): Pheasantshoek near Viljoenshof, in a marshy area near E le g ia fis tu lo s a and R e stio fe s tu c a e fo r m is , and along a small stream (or furrow), in peaty soil and on clayish bank, no male found, (-D A ), 15-09-1976, E ste r h u v se n 3 4 3 7 4 (BOL, holo.).
D istrib u tio n a n d e c o lo g y , endemic on the Agulhas plains between Elim, V iljoenshof and the Soetanysberg, at an altitude o f 50-150 m, on either sandstone or silcretes.The species is locally common on gravelly soils or damp sand, often with impeded drainage.
N otes: this species is known currently only from the female material, and since it can be very common locally, it is presumably apomictic.It is closely related to R estio fe m in e u s , which is also known only from female material, and which forms sim ilar neat little branching tufts.However, R. a n o m a lu s differs in many details from R. fe m in e u s.The sheath margins are broadly m em bra nous (instead o f coriaceus or narrowly m em branous), the spathes overtop the spikelets (instead o f being as tall as the spikelets), the inner and outer tepals are the same length (instead o f the inner tepals being shorter than the outer tepals), the inner tepals are 3.0-4.5 mm long, com pared to 2.2-2.6 mm in R. fe m in e u s , and the ovary has two locules instead o f one.In addition, there are num er ous small anatom ical differences in the culm.Plants clumped or tangled, stoloniferous, forming dense cushions among boulders, 0.2-0.5 m tall, or spreading mats on open slopes.Fertile culms branching, smooth, green or olivaceous, 0.5-1.5 mm diam.at apex; sheaths closely convoluted, 10-25 mm long, dark brown, apical margins narrow ly membranous, hyaline shoulders absent, apex acute to acuminate; mucro peniciMate, straight and erect, 2-10 mm long.Male inflorescence with I spikelet (very rarely, in the Swartberg) or 2-10 spikelets, forming a tight button, 8-25 x 7-25 mm; spathes persistent, as tall as spikelets, coriaceous, rather similar to lower bracts; male spikelets sessile, oblong or elliptical, 7-12 x 3 -6 mm, 3 -10-flowered; bracts taller than flowers, 5.5-10 mm long, oblong, acuminate, cartilaginous.Male flower 4.5-5.0mm long, both tepal whorls equally long, hyaline or membra nous, linear or oblong, outer lateral tepals conduplicate, sparsely villous on keels; anthers 2-3 mm long, exserted from flowers; pistillode absent.Female inflorescence with I spikelet (rarely, then mostly at western and eastern extremes o f range) or 2 10 spikelets, forming a tight but ton, 10-25 x 3-25 mm; spathes equalling spikelets, per sistent.coriaceous; spikelets sessile, elliptical, obtuse or acute, 10-20 mm long, 5 15-flowered, with 4 or 5 sterile bracts; bracts taller than flowers, erect or reflexed, 5.5-10 mm long, oblong, acuminate, cartilaginous, apical margin like rest o f bract, same as body o f bract, awn less than half as long as bract body.Female flower 5.0-6.5 mm long, tepals cartilaginous, all equally long, outer lateral tepals conduplicate with sparsely villous keels, odd outer and inner tepals oblong; staminodes absent.Gynoecium: styles 2, feathery, white; style bases free but adjacent; ovary dehiscent, bilocular.Seed 1. 3-1.5 x 1 mm, in side view elliptical, in diam.round, silvery, pitted.Flowering time: September to December.
Distribution and ecology': widespread on the sum mits o f the inner ranges o f the Cape Fold Mountains, from Sneeukop in the Cedarberg to Mannetjiesberg in the Kamannassie and Meiringspoortberg in the Groot Swartberg.The species is absent from the lower coastal mountains: the Kogelberg, Kleinriviersberg and the Cape Peninsula.The altitude range is 1 500-2 100 m, all col lections are from sandstone habitats.This is a species of rocky summits and exposed ridges.On the Matroosberg, the first author found it growing on shady rock ledges, well protected from fire, and quite cool, but dry.This is consistent with most habitat notes, which indicate cool, dry places, associated with rocks or rocky sites, or ledges.Possibly these are all fire-protected habitats.
Notes: the new species is similar to Restio witteber gensis', both species have both male and female spikelets aggregated into button-like heads.However, it differs from R. wittebergensis by its branching culms (R. witte bergensis has simple culms), and the stoloniferous habit (R. wittebergensis is caespitose).The stoloniferous habit and aggregated spikelets suggest a similarity to R. sch oenoides, but this species is more slender, carpet-form ing rather than tangled.There is a wide range o f vari ation within R. constipatus, and in its current definition it might include two entities: the typical boulder-habitat (rocky summit) plants from the western part o f the range, and more mat-forming plants from the Swartberg, which are very common along the summit ridges o f these inland mountains.The differences in growth form are corroborated by a subtle difference in the flower size and a more slender growth form.This needs more, criti cal, investigation.These collections have in the past been assigned to R. schoenoides; however, the latter species is better understood as a sum m er rainfall species wide spread in the Drakensberg and reaching to the Blouberg in the Limpopo Province.The interface between R. sch oenoides and R. wittebergensis remains taxonomically difficult.

A dditional collections
Distribution and ecology, endemic in the Hex River Mountains between Worcester and Ceres, where it is locally very common between 1 600-2 100 m.It is found on stony, well-drained mountain slopes.It appears to occupy the same habitat as Restio virgeus, and the two species alternate on the stony slopes between Zebasberg and Mt Superior.The plants appear to be killed by fire, regenerating from seed.
N otes: Restio elsieae species is a high-altitude seg regate from R. ocreata.It differs by the simple, stout culms, and sm aller growth form.The single female spikelet with recurved bracts and the unbranched culms are sim ilar to R. curvibracteatus, but R. elsieae can immediately be recognized by the flat, spreading sheaths.

Additional collections
Distribution and ecology: this new species is restricted to the southern Cedarberg and the northern Bokkeveld, and has been collected from Bloukop, north o f the Bokkeveld Sneeukop, to G abriel's Pass in the cen tral Cedarberg.In this region, Restio cedarbergensis is found from 900-1 500 m, on well-drained soils, often described as rocky.Several collections are from shale bands, and the first author has seen the species grow next to renostcrveld.where sandstone meets shale, as well as on typical dry TMS soils.The species appears to pre fer hot, dry, well-drained localities.Plants appear to be killed by fire, regenerating from seed.
N otes: although the spikelet structure is very similar to the hyper-variable (and probably too broadly defined) species Restio sieberi, the plants can always be distin guished by growth form, by the woolly prophylls, and by the stouter culms and somewhat larger spikelets.Ecologically it is also distinct, and is always associated with large rocks, or even more commonly with passages through the sandstone bedrock, where it often forms large monospecific stands.These habitats are presum ably more shaded and colder than the typically sunny open habitat o f the numerous forms of' R. sieberi.

AC KNOW L E DC. E M EN TS
This study was funded by the Swiss Science Foun dation grant SNF 31-66594-01, and the field work was supported by the C'laraz Foundation and National G eo graphic Society Grant 7289-02.CapeNature provided collecting permission.We particularly want to thank Philip Moline (field work, collaboration on the Elegia complex), the Bolus Herbarium and Terry Trinder-Smith for handling our collections in Cape Town.Peter Gold blatt, Barbara Briggs and Nick Helme are thanked for critical input into the paper, and Peter Wilson and Otto Leistner for checking the Latin.

FIGURE I . -
FIGURE I.-Cladograms summarizing generic and subgeneric relationships in the Restioneae (Hardy et al. 2008).A. relationships in the strict consensus tree from the analysis o f all 292 species and subspecies.B. relationships in the strict consensus tree from the analysis o f 212 spe cies and subspecies omitting 20 species/subspecies for which DNA data were lacking.Bootstrap percentage values (in square brackets) are based on the analysis o f Hardy et al. (2008).Characters from Tables 1 and 2 are optimized using the slow (deltran) options in WinClada: the character number appears above the solid and open squares.The state o f the character appears below the solid and open squares.An open square indicates a character that evolved at least tw ice, w hereas a solid quare indicates a character that evolved only once.
). a) Basal grade: includes the only two species with branching culms: 4.1.T h am n o ch o rtu s levynsiae Pillans in Transactions o f the Royal Society o f South Africa 16: 364 (1928).4.2.T h am n o ch o rtu s p u lch er Pillans in Transactions o f the Royal Society o f South Africa 29: 350 (1942).4.3.T ham nochortus nutans (Thunb.)Pillans in Transac tions o f the Royal Society of South Africa 16: 365 (1928).4.4.T h am n o ch o rtu s gracilis Mast, in A.DC., Monographiae phanerogamarum 1: 327 (1878).b) Limestone clade: all species associated with lime stones or coastal dunes, except T. karooica: 4.5.T h am n o ch o rtu s karooica H.P.

44
Elegia filaceaCiL RL 3 .-Portion of strict consensus tree of total combined analysis o f Hardy el al. (2008) containing Askidiosperma and Elegia sensu Moline & Linder (2005).Bootstrap \alues above the branches indicate support based on a 500-replicate bootstrap analysis that included 20 species lor which DNA data were lacking (asterisked taxa).Bootstrap \alues below branches indicate support based upon a 500-replicate analysis that included only taxa for which both DNA data and morphology data were available.Where no value is given below a branch, it indicates that the clade did not appear in the analysis excluding morphology-only species.

FIGURE 4 .
FIGURE 4. Portion of strict consensus tree of total combined analysts o f Hardy et al. (2<K)8) containing Restio subgen.Craspedolepis.R subgen.Varirestio, R. subgen.Eremorestio, R subgen.Calopsis, and R. subgen.Restio.Names in parentheses are the now synonymous names o f Linder (2001a) when different from the ones here proposed.Bootstrap \alues above the branches indicate support based on a 500-replicate bootstrap analysis that included 20 species lor which DNA data were lacking (asterisked taxa).Bootstrap values below branches indicate support based upon a 500-replicate analysis that included only taxa for which both DNA data and morphology data were available.Where no value is given below a branch, it indicates that the clade did not appear in the analysis excluding morphology-only species.

FIGURE 5 .
FIGURE 5. Portion o f strict consensus tree o f the total combined analysis o f Hardy et al. (2008) containing Restio subgen.Locapsis and R. subgen.Isc hyrolepis.A single species o f Restio subgen.Ischyrolepis (Restio fuscidulus) that was missing many data (including all molecular data), was resolved in this tree as sister to Staberoha (see Figure I), albeit with .1 % bootstrap support which we do not consider substantial enough to warrant disputing the monophyly o f Restio subgen.Ischyrolepis.Names in parentheses are the now synonymous names o f Linder (2001a) when different from the ones here proposed.Bootstrap values above branches indicate support based on a 500-replicate bootstrap analysis that included 20 species for which DNA data were lacking (asterisked taxa).Bootstrap values below branches indicate support based upon a 500-replicate analysis that included only taxa for which both DNA data and morphology data were available.Where no value is given below a branch it indicates that the clade did not appear in the analysis excluding morphology-only species.
Flowering time: September and October.Distribution and ecology", this new, apparently apomictic species is known from only two localities in the Cold Bokkeveld, from the Suurvlakte at the north ern end o f the Cold Bokkeveld.and from the summit o f the Skurweberg Pass at the southern end o f the Cold Bokkeveld.The two locations are at 750-1 100 m, on sandstone, in deep sand over groundwater.Both loca tions showed some disturbance; the Skurweberg Pass population showed substantial disturbance.

. Portion o f strict consensus tree o f total combined analysis o f all species by Hardy et ul. (2008) containing Sorvveta. Platy caulos. Staberoha. Restio subgen. Simplicaulos. Restio subgen. Pendulostemon. Rhodocoma. and Thamnochortus. Names in parentheses are the now synonymous names o f Linder (2001a) when different from the ones here proposed. A single species o f Restio subgen. Ischyrolepis (Restio fuscidulus) that was missing many data (including all molecular data), was resolved in this tree as sister to Staberoha. albeit with 3 % boot strap support which we do not consider substantial enough to warrant disputing the monophyly o f Restio subgen. Ischyrolepis or the genus Restio as circumscribed here. Bootstrap values above branches indicate support based on a 500-replicate bootstrap analysis that included 20 species for which DNA data were lacking (asterisked taxa). Bootstrap values below branches indicate support based on an analysis that excluded these morphology-only taxa. Where no value is given below a branch, it indicates that the clade did not appear in the analysis excluding morphology-only species.
banksii Pillans in Transactions o f the Royal Society o f South Africa 16: 385 ( 1928).3.7.S taberoha m ultispicula Pillans in Journal o f South African Botany 18: 117(1952).3.8.Staberoha o rn ata Linder in South African Journal o f Botany 56: 450(1990).4.6.T h am n o ch o rtu s m uirii Pillans in Transactions of the Royal Society o f South Africa 16: 366 (1928).4.7.T h am n o ch o rtu s paniculatus Mast, in Botanische Jahrbiicher 29, Beiblatt 66: 12 ( 1900b).4.8.T h am n o ch o rtu s spicigerus (Thunb.)Spreng., Systema vegetabilium 1: 187(1824).4.9.T h am n o ch o rtu s fra te rn u s Pillans in Transactions o f the Royal Society o f South Africa 16: 370 (1928).4.10.T h am n o ch o rtu s p luristachyus Mast, in Botanis che Jahrbiicher 29.Beiblatt 66: 1 2(1900b).Linder in Bothalia 15:473 (1985).4.12.T h am n o ch o rtu s rigidus Esterh. in Linder in Both alia 15:477 (1985).4.13.T ham nochortus glaber (Mast.)Pillans in Transac tions o f the Royal Society o f South Africa 16: 363 (1928).4.14.T h am n o ch o rtu s p ap yraceus Pillans in Trans actions o f the Royal Society o f South Africa 29: 350 (1942).4.15.T h am n o ch o rtu s am oena H.P.Linder in South African Journal o f Botany 56: 451 (1990).4.16.T h am n o ch o rtu s acum inatus Pillans in Trans actions o f the Royal Society o f South Africa 29: 349 (1942).4.17.T h am n o ch o rtu s fruticosus PJ.Bergius, Descrip tiones plantarum ex Capite Bonae Spei 353 (1767).
c) Big bract group: species with large, often silvercoloured female bracts: 4.11.T h am n o ch o rtu s cinereus H.P.
Chondropetalum clade: includes most species pre viously included in Chondropetalum.Morphologically this clade is defined by the dehiscent ovaries.However, it also includes Elegia verreauxii, which has an indehis cent ovary: Rhizome spreading, unbranched or sparsely branched, culms evenly spaced.Fertile culms unbranched, round, smooth, olivaceous, 4-6 mm diam.atbase, 1.4-2.5 mm diam.atapex; sheaths caducous, with an abscission line present, loosely convoluted.30-60mm long, margins entire, red dish to dark brown with fine tan speckles and tan margins, slightly glossy; sheath mucro penicillate, straight and erect, 6-10 mm long.Distribution and ecology: endemic to the west coast of South Africa, from Blouberg to Elands Bay.Local in damp hollows and seepages on well-leached sand below 200 m.
b)Etymology: this species is known as olifantsriet on the west coast.Unfortunately, this name appears to be applied to any very large member o f the Restionaceae.thus is not very useful.The name is derived from the Latin for elephant: elephas.
Fruit: in species with indehiscent ovaries, perianth winged, persistent around ovary, in spe cies with dehiscent ovaries, seed silvery, brown or grey; surface pitted, colliculate or striate.Culm anatomy: epi dermal cells in 1 layer, lateral wall straight, unthickened or rarely sinuose and thickened; parenchymatous layer o f 1-3 cells; central ground tissue solid or rarely with scat tered cavities or with single, central cavity; tannin absent or present in epidermis, sclerenchyma or central ground tissue; silica absent.Notes: many species in this subgenus are associated with impeded drainage, and Restio miser and R. confusus are often dominants in convex-domed seepages on shale bands and on sandstone.Restio bifidus is typical of shal low seepages, and can be dominant over much o f the upper plateau of Table Mountain.However, many of the other species are found in well-drained habitats.There are very distinct morphological groups in this subgenus.The first group has large, almost flat, bracts (Restio bifarius, R. bifidus, R. nuwebergensis and R. papyraceus), that are quite similar to the female spikelets ot the possible sister clade Staberoha.The second group has small button-like spikelets, borne on fine-stemmed tussocky plants, often with only one spikelet per culm, and round female flowers in which the lateral tepals are not conduplicate (R. subtilis, R. confusus, R. miser).The third group has thick spindle-shaped spikelets, and nut lets with a persistent perianth forming a wing (R. mon ostylis.R. esterhuyseniae).Finally.R. echinatus and R. pedicellatus have acuminate bracts.