Taxonomic significance of inflorescences , floral morphology and anatomy in Passerina ( Thymelaeaceae )

Comparative studies were undertaken on the inflorescence, bracts and floral morphology of all taxa of the genus Passerina L. in southern Africa. Information is given in tabular form and a key based on bract morphology is presented. Floral morphology supported the status of the intrageneric taxa and also proved to be of taxonomic significance in the genus. Controversy surrounding the interpretation of a number o f floral morphological structures in Passerina has been resolved. Morphological and anatomical evidence allowed a re-interpretation of the structure of the receptacle, hypanthium and sepals, ovary type and position, structure of the seed coat, ovule type and position, obturator, fruit and seed. On this basis an authentic generic description o f the floral morphology was compiled. Passerina is distinguished by the following set o f characters, a very short floral receptacle, tubular hypanthium, petaloid calyx, absence of petals and petaloid scales, diplostemonous dimorphic androecium, extrorse anthers, superior ovary, anatropous, ventrally epitropous ovule, an obturator of elongated cells, a I-seeded berry or an achene and tegmic seed with nuclear endosperm becoming cellular throughout. On this basis the flower in Passerina is considered a phylogenetically advanced structure, supporting the view that the genus is advanced within the Thymelaeoideae. The proposed taxonomic relationship between Thymelaeaceae and Malvales is confirmed by floral morphological evidence.


INTRODUCTION
The infrageneric taxonomy of Passerina L. is a prob lem, due to the apparent lack of marked morphological differences between species.The latest revision by Thoday (1924). is now mostly outdated.
The controversy surrounding the interpretation of certain floral structures in the genus became obvious during the present study.Heinig (1951) did not include Passerina in her study of the floral morphology of the Thymelaeaceae.From the sexual system of Linnaeus (1781,1784) to Domke (1934), floral morphology played an integral part in the intrafamilial classification of the Thymelaeaceae and in the circumscription of the family (Table I).Between ± I960 and 1996, vast leaps were taken in the classification of the flowering plants by making use of anatomical, floral, palynological, embryological and chemical evidence.During this period, dis agreement on the circumscription of the Thymelaeaceae was common (Table 2).Since 1990, many higher-level phylogenetic relationships in angiosperms were based on evidence from rbcL and I8S nuclear ribosomal DNA sequence data.In many cases these studies confirmed previously proposed phylogenetic patterns and trends within the family, based on morphological evidence; in other cases, however, profound changes in the circum scription and relationships of the Thymelaeaceae were suggested (Table 3).
The present paper emanated from a monographic study of the genus currently in progress.Available evi dence suggests the presence of at least four new species and four new subspecies, to be added to 16 previously recognized species, mostly endemic to southern Africa (Thoday 1924;Bond & Goldblatt 1984) (Table 4).Here we report on a comparative morphological study of the inflorescence, bract, flower, fruit and seed.Specific  (Bredenkamp & Van Wyk 2001), are associated with floral morphological structures as well as fruit and seed types in the present study.

MATERIALS AND METHODS
As far as possible, material was collected from at least five different localities for every taxon.Live and pre served (dried and in liquid preservatives) material of all the species, subspecies and varieties in Passerina was studied (Table 4).Illustrations were made from herbari um material by means of a drawing tube.Measurements were taken by using a dissection microscope and a cali brated eyepiece.
Flowers were fixed and stored in a 0.1 M phosphatebuffered solution at pH 7.4, containing 2.5% formalde hyde, 0.1% glutaraldehyde and 0.5% caffeine [modified Kamovsky fixative; Kamovsky (1965)].Light micro scopy (LM) was used to study the floral anatomy of P. ericoides (Bredenkamp 956. 962).which has tleshy fruit and P. vulgaris (Bredenkamp 944.951) with dry fruit.As the flowers are quite small, whole flower buds, flowers directly after anthesis and young fruits enveloped in the floral tube were washed in water, dehydrated and embed ded in glycol methacrylate (GMA) following the meth ods of Feder & O'Brien (1968).Embedded floral mater ial was serially sectioned from the base of the receptacle to the anthers.Sections were stained in toluidine blue "O', subjected to the periodic acid-Schiff's (PAS) reaction and mounted in Entellan (Art.7961.E. Merck.Darmstadt).

Floral envelope
The authors regard the floral envelope ('outer floral whorl') as a hypanthium (fused calyx and androecium).differentiating into four petaloid sepals and a diplostemonous androecium.arising from the hypanthium rim at the separation of the sepals.For the description of colour, texture and measurement of total floral length, only the hypanthium and sepals are considered-the stamens are excluded.

Phylogeny
Speculations on phylogeny are based on prevailing family characters representing the ancestral state and derived characters, indicating a reduction in tissue at genus and species levels, considered as possibly advanced.A cladistic analysis based on anatomical and morphological characters in Passerina is scheduled for the final stages of the study.

Generic description of floral anatomy (as seen in transverse section)
Receptacle base: vascular tissue arranged in a central stele from which 8 traces are derived in a single whorl (Figure 5A).cular bundles, fused sepal and stamen traces occupying peripheral position (Figure 5B); cells at perifery of cor tex arranged in rows (abscission tissue), differentiation of inner epidermis of hypanthium and outer epidermis of ovary wall, separating hypanthium from ovary wall (Figure 5C, D).Hypanthium irregularly lobed or scal loped; outer and inner epidermis variously hairy; cuticle sometimes strongly developed; 8 vascular bundles stretching over entire length (Figure 5E, F).Calyx with 4 imbricate lobes developing at hypanthium rim, each containing 3 vascular bundles; epidermal and hypodermal layers containing large amounts of pigment; spongy parenchyma aerenchymatic (Figure 8C, D; Table 6).Androecium : each of the 4 fused commisural sepal and antipetalous stamen bundles (cs-pst) split into 2 sepal lateral bundles (si) and 1 antipetalous stamen bundle (pst), resulting in the first whorl of 4 antipetalous sta mens (situated slightly lower in the hypanthium) and each sepal containing 3 vascular bundles (Figure 8A, B); each of the 4 fused sepal midrib and antisepalous stamen bundles (s-sst) split into a sepal midrib bundle (s) and an antisepalous stamen bundle (sst), forming the second whorl of 4 antisepalous stamens (Figure 8B, C); anthers extrorse, with wall of locule comprising epider mis and endothecium only, periclinal walls of epidermis thin and folded inwards, cell wall thickenings of endothecium ± stellate, with rib-like extensions directed towards epidermis (Figure 9B), partitions between loculi withered and ruptured (Figure 9C), accompanied by final rupturing of outer walls of thecae (Figure 9D).Ovary b ase: wall independent of hypanthium or loosely adhering to hypanthium at distal side away from placen ta (Figure 5E, F; Table 6); outer and inner epidermal walls strongly developed, containing ample amounts of tanniniferous substances (Figure 5E) or less sturdy (Figure 5G); mesophyll of densely arranged parenchy-   7C) or inconspicuously arranged in a row along the long axis of elliptic style (Figure 7D;  (Davis 1966), but later becoming cellular throughout (Figure 10D), absorbed by cotyledons con taining no starch but copious amounts of oil.

Floral morphology at species level
Floral morphological characters and taxonomically important fruit characters are summarized in Table 7, and specialized hypanthium and sepal characters in Table 8.All these are associated with specific leaf anatomical characters (Bredenkamp & Van Wyk 2001).

Inflorescences
Weberling (1989) regards polytelic synflorescences as dominant w ithin the Thymelaeaceae.He found monotelic synflorescences in the Gonystyloideae.a relatively primi tive group, as well as certain genera of the Thymelaeoideae and Aquilarioideae.In the Gnidieae.it was found in Lachnaea L. (= Cryptadenia Meisn.)(Beyers & Van der Walt 1995;Beyers 1997), a genus endemic to the Cape Floristic Region (Beyers 1992).Weberling (1989) never theless concluded that it appears impossible to draw any taxonomic conclusions from the existence of monotelic synflorescences within these taxa.Passerina is character ized by polytelic synflorescences.Most species have multiflowered main and co-florescences, and a reduction of florescences to single and subcapitulate spikes is clearly show n (Table 5).

Bracts
In their descriptions of the Thymelaeaceae.Domke (1934) reports the presence or absence of bracts and bracteoles, sometimes involucral.and Peterson (1978) mentions that deciduous or persistent bracts are often present.In Passerina, single flowers are always enveloped by persis tent bracts.In the present study, this constant taxonomic character has been employed in a key for application in herbarium and field work (Figures 3. 4; Table 5).

Receptacle
Historically the interpretation of the receptacle in the Thymelaeaceae has been controversial.Tables 1 and 2 •p 3 Q .show that Meisner (1857) regards the floral envelope as perigynous and hypocrateriform, implying a cup-shaped receptacle or hypanthium.Gilg (1891, 1894) describes the floral envelope as a cylindric receptacle which is articulate in the upper half, and Endlicher (1837, 1847), Leandri (1930) and Dahlgren & Thome (1984) regard the floral arrangement as perigynous.The present study indicates that the receptacle (in t/s) is very short (Figure 5A, B) and definitely not cup-shaped.This is evident from the arrangement of peripheral cortex cells in rows, followed by differentiation into the inner epidermis of the hypanthium and the outer epidermis of the ovary wall (Figure 5D), finally by the separation of the hypanthium (including the vascular bundles differentiated from the stele) from the ovary wall and the presence of trichomes in the space subsequently formed (Figure 5D-F).

Hypanthium and androecial position
Owing to reduction of the receptacle, the hypanthium in Passerina is here interpreted as being formed by the fused calyx and androecium only.The vascular tissue of the hypanthium constitutes the fused sepal and stamen traces (Heinig 1951), separating from the central stele in a single whorl and forming a peripheral ring of eight vas cular bundles (Figure 5A, B), which persist throughout the length of the hypanthium.A similar pattern of fusion and distribution of vascular tissue has been reported by Heinig (1951) and for the genus Lachnaea (= Cryptadenia) by Beyers (1992) & Beyers & Van der Walt (1995).In Passerina the central stele differentiates into carpellary bundles after the separation of the fused sepal and stamen traces (Figure 5C-E).It can therefore be concluded that the segments of the floral envelope and the androecium arise below the gynoecium, the floral arrangement being hypogynous and the ovary superior (Weberling 1989).

Identity o f floral envelope
From Wikstrom (1818) to Takhtajan (1997) (Tables 1; 2) the floral envelope in Thymelaeaceae and, in many cases Passerina, was variously interpreted as an in fundibular corolla, hypocrateriform hypanthium.infun dibular perigone, perianth, cylindric receptacle, floral tube formed by coalescence of four external whorls, perianth tube, or a hypanthium.Heinig (1951) supports the interpretation of Leandri (1930) and Domke (1934) of the floral tube as appendicular in origin, composed of the fused bases of the sepals and adherent stamen filaments, also pointing out that the sepal is with few exceptions a three-trace organ.Bunniger (1972) is of the opinion that the hypanthiums in families of the Myrtales and Thymelaeales have a similar origin.Our results show eight vascular bundles running along the length of the hypanthium and separating into sepal and stamen bundles, each sepal lobe eventually with three vascular bundles (Figure 8A, B).We regard the floral envelope as a hypanthium (fused calyx and androecium), differen tiating into four petaloid sepals and a diplostemonous androecium, arising from the hypanthium rim at the separation of the sepals.A very short receptacle (Heinig 1951), which does not contribute to the hypanthium, indicates a reduction in tissue and a possible advanced state.This is indeed the case in Passerina.
A study of petaloid scales in Thymelaeaceae has added further evidence to the interpretation of the floral envelope.These scales have been regarded by various authors as perigynous scales or glands, perigynous nec taries, petals or staminodes, petaloid scales and petaloid appendages (Tables 1; 2).Heinig (1951) is convinced that the morphology and vascularization of the petaloid scales resemble that of stipules, an opinion shared by Rao & Dahlgren (1969) on the floral anatomy and rela tionships of the Oliniaceae.In their floral description of O linia, Dahlgren & Van Wyk (1988) consider the petaloid scales as true petals.Heinig (1951) concludes that the Thymelaeaceae is apetalous.In Passerina there are no petaloid scales or corolla (Tables 1; 2).Our results have shown the separation of antipetalous as well as antisepalous stamens, but petaloid scales or even ves tiges of them were not observed.We therefore regard Passerina as truly apetalous.The complete reduction of the corolla or the absence of petaloid scales can be regarded as an advanced state in the Thymelaeaceae; it could also be interpreted as part of the anemophilous syndrome displayed by the genus.Based on both the pattern of the vascular tissue and the absence of petaloid scales, we consider the floral envelope in Passerina as a hypanthium consisting of the fused calyx and androecium.differentiating into four sepals and the diplostemonous stamens.

Fragmentation o f hypanthium
The fruit in Passerina is enveloped by a persistent, loosely attached hypanthium.Bentham & Hooker (1880), Gilg (1891, 1894), Domke (1934) and Melchior (1964) mention that the hypanthium is articulated above the ovary.We found no definite articulation point in the hypanthium neck (narrowed part of hypanthium between apex of ovary and sepals) in Passerina, possibly because of the absence of receptacle tissue and the appendicular nature of the hypanthium.The base of the neck fragmented as a result of desiccation and torsification of cells (Figure 2E).In some species, fragmentation of the hypanthium takes place over the widest circumference of the fruit, shedding the frag mented distal part of the hypanthium, sepals and androeci um (Figure 1D).A strong association was found between flowers with a short hypanthium neck and fragmentation of the hypanthium over the widest part of the fruit, and also between flowers with a long hypanthium neck and frag mentation at the base of the neck (Table 8).
A plane of circumscission, dividing the floral tube into a basal and upper portion, is clearly illustrated in Gnidia and Struthiola (Peterson 1978) and Lachnaea (Beyers 1992: Beyers & Van der Walt 1995).We hypothesize that the plane of circumscission possibly indicates a differ ence in tissue composition between the basal and upper portions of the hypanthium and that this articulation can be of morphological importance in the Thymelaeaceae.The basal portion of the floral tube below the plane of cir cumscission possibly indicates the inclusion of receptacle tissue in the hypanthium, whereas the upper portion con sists of calyx and androecium tissue only (accepting the apetalous state).An alternative interpretation, offered by one of the referees of this paper, regards the vasculature as a prerequisite to decide w hether one is dealing with a hypanthium (appendicular in origin) or a receptacle (axial in origin).The significance of an articulation indicating a distinction between parts of the hypanthium of different derivation, should be further investigated.

Androecium
In his description of Thymelaeaceae.Peterson (1978) describes the stamens as twice as many or as many as the sepals (rarely reduced to two or one), in one or two whorls, the outer whorl antisepalous.In the subfamily Aquilarioideae (Heinig 1951), stamens are of approximately equal length and the traces supplying them separate from the sepal traces, practically in a single whorl.The Thymelaeoideae is characterized by stamens arising as two distinct whorls at two distinct levels in the floral tube.
The dimorphic diplostemonous nature of stamens in Passerina (Thymelaeoideae), in which the antipetalous stamens are shorter than the antisepalous ones, has been confirmed by our observations (Figure 8A-C).According to Heinig (1951) evolution within the androecium has been from polystemony to diplostemony to dimorphic diplostemony, indicating the advanced state of the androe cium in Passerina.Peterson (1978) describes the anthers as usually introrse, rarely extrorse.Species of Passerina have large, exserted, extrorse anthers, clearly an adapta tion to the anemophilous syndrome of the genus.

Separation o f fu sed sepal and stamen traces
Heinig ( 1951) is of the opinion that a foreshortening of the floral axis has resulted in a fusion of the calyx and androecium and that progressive stages of adnation can be observed in the family.In all species the antipetalous stamen traces are fused to the commisural sepal traces and the antisepalous stamen traces to the sepal midrib traces at their point of origin from the stele.In the Aquilarioideae, in Octolepis dinklagei, they become separated in the receptacle or, in other cases, low down in the floral tube.Except for Gnidia splendens (= Lasiosiphon splendens), in which the separation of the antipetalous and antisepalous stamen traces takes place in the receptacle, Thymelaeoideae is characterized by separation higher up in the floral tube, e.g. at the top of the ovary in Dirca occidentalis and below the origin of the sepals in Gnidia subulata (Heinig 1951).In Passe rina the separation of the antipetalous stamen traces takes place at the origin of the sepals (Figure 8A, B) and separation of the antisepalous traces slightly higher up (Figure 8C), indicating what appears to be a phylogenetically advanced tendency.

Pollen
Pollen grains of most members of Thymelaeaceae are monads, spheroid and pantoporate, characterized by a typical croton pattern, comprising rings of more or less trihedral sexine units mounted on an underlying reticulum of circular muri (Bredenkamp & Van Wyk 1996).In Passerina, the basal reticulum, as in the typical croton pattern, is no longer discernible as it is replaced by a sec ondary reticulum derived from fused sexine.The supratectal subunits are fused completely to form a continuous reticulum which replaces the underlying reticulum.The reticulum in Passerina is therefore secondary in origin and considered phylogenetically advanced.This modifi cation of the crotonoid pattern is probably also of func tional significance as pollen in Passerina is adapted to anemophily.

Disc
In Aquilarioideae a hypogynous disc is generally absent, but is almost always present in Thymelaeoideae ( Heinig 1951;Peterson 1978).Possibly because of a reduction of tissue, no disc was observed in Passerina, a state confirmed by Bunniger (1972).

Ovary
The ovary in Passerina was described as unilocular up to the time of Domke (1934) (Table 1).The pseudomonomerous state is mentioned by Eckardt (1937), Melchior (1964), Dahlgren & Thorne (1984) and Cronquist (1988) (Table 2), while most authors agree that ovules are solitary and pendulous.According to Takhtajan (1997) the ovary in Thymelaeoideae consists of two carpels, it is monolocular and the ovule is solitary.Heinig ( 1951) is convinced that within the carpellary whorl a reduction series may be observed, ranging from four-or more-carpellate members in the Aquilarioideae to the two-carpellate members of the Thymelaeoideae in which one carpel is markedly contracted, thus a pseudomonomerous ovary.According to Heinig (1951) the ovules have been reduced to one per locule in the entire family.In Passerina, at the base of the ovary, the dorsal and median carpellary bundles initially separate from the central stele (Figure 5C.D).After differentia tion has taken place, the dorsal, median and commissur al carpellary bundles (Figure 5E) can be distinguished.Because of a redistribution of carpellary bundles, the sin gle dorsal carpellary bundle is arranged at the opposite side of the horseshoe-shaped median and commissural carpellary bundles (Figure 5G).In Dirca palustris, Heinig (1951) has illustrated the authenticity of the pseudomonomerous ovary by the presence of two dorsal carpellary bundles, one in the fertile carpel and one in the second, reduced, sterile carpel positioned between two groups of commissural carpellary bundles.Bunniger (1972) showed the presence of two carpels in the flower primordia of P. filiformis.In the present study, which included the young bud stage of P. vulgaris (Figure 5F), no indication of a second carpel or a second dorsal carpellary bundle was found, possibly because of reduc tion and fusion processes, which had already taken place in the formation of the young ovary, consequently we consider the ovary of Passerina as pseudomonomerous.Domke (1934) describes the ovule in Thymelaeaceae as pendulous, anatropous, with a ventral funiculus, ex ceptionally hemi-anatropous or orthotropous, indicating a phylogenetic tendency.Our study clearly indicates a pendulous, anatropous ovule in Passerina (Figure 6C,D).The funiculus is ventral and has been sectioned from the base of the ovule (Figures 5E-H; 6A) to the point of attachment with the placenta (Figure 6C, D).Close to the embryo sac the micropyle is a trilete opening formed by the inner integument (Figures 6C, D; 7A), facing upward.Towards the micropyle, the outer and inner integuments become horseshoe-shaped (in t/s), resulting in the micropyle being directed towards the elongated obturator cells, located at the base of the style (Figure 7B).Based on these results, we agree with Dahlgren (1975b), who regards the ovule as pendulous and epitropous.

Ovule type and position
Obturator Davis (1966) defines an obturator as any structure asso ciated with directing the growing pollen tube towards the micropyle.but elongated cells extending from the base of the style to the micropyle are considered exclusive to the Thymelaeaceae.In Passerina such elongated obturator cells can be clearly seen at the level of the placenta, at the departure of the funiculus, touching on the inner integu ment (Figures 6C; 7A) and finally these cells extend from the base of the style entering the micropyle (Figure 7B).

Fruit
Most authors (Tables 1; 2) agree that the fruit in Thymelaeaceae is indehiscent.In Passerina, Wikstrom (1818), Meisner (1857) and Endlicher (1837, 1847) con sider the fruit as a nut or a nutlet.Domke ( 1934) concludes that the fruit of Dais.Gnidia, Lachnaea (= Cryptadenia) and many taxa of Passerina can be defined as an achene, and that of P. ericoides as a berry.Dahlgren ( 1975b) con siders the fruit of Thymelaeaceae as a nut or drupe.Peter son (1978) regards it as a berry, a nut. a drupe or a loculicidal capsule and Takhtajan (1997) describes them as nut like, baccate or drupaceous.A relevant family character, that the outer integument of the ovule disintegrates and the inner integument becomes palisade-like and hardens to torm a seed coat or tegmen (Figure 10A-D).is illustrated by the present study.Structurally the dry.membranous fruit in Passerina conforms to an achene (Spuyt 1994) and the reduction in pericarp tissue can probably be considered as a specialized adaptation.A reduction in tissue from a drupe to a membranous 1-seeded berry or an achene can be illus trated in Thymelaeaceae and therefore the fruit in Passerina could be considered advanced within the family.We agree with Domke (1934) that P. ericoides (Figures 1F;10A) and P. rigida are characterized by a fleshy 1 -seeded berry, while all the other species have an achene (Figure 2F; Table 7).The achene remains enveloped in the rem nants of the papyraceous hypanthium, nestled adaxially in the tomentum of the concave, persistent cymbiform bracts.

Seed
The existing confusion concerning the state of the tegmen in Passerina is a reflection of the total lack of information of this aspect in descriptions of the group by various authors (Tables 1; 2).Meisner (1857) describes a crustaceous pericarp and Domke (1934) a black, crustaceous testa.The ovule in Passerina is bitegmic (Figures 6;7;10). Comer (1976) refers to outer integument (oi) and inner integument (ii), the product of the outer integu ment becoming the testa and that of the inner integu ment.the tegmen.In Passerina the outer integument dis integrates, whereas the inner integument remains (Figure 10A), its outer epidermis becoming palisade-like (Figure 10A, B).Hence, the seeds of Passerina are exotegmic with a palisade, a state common to the family (Comer 1976).In Passerina the tegmen is black and lignified.and in t/s, still portraying its origin from the palisade-like epidermis (Figure 10C, D). 7; 8) In the present study, specific results in leaf anatomy indicating the arrangement of taxa in Passerina (Bredenkamp & Van Wyk 2001), are associated with flo ral morphological structures as well as fruit and seed types.In Table 7, leaf structural types A and especially B3 are associated with four species that have smaller, yellow, membraneous flowers (up to 5.3 mm long).Leaf structural types B4.B5, B6.C and D are associated with most species having larger, yellow-pink, papyraceous flowers (up to 8.4 mm long).The same tendency in the length of the neck, and the size of the inner sepals is shown in Table 8.

Fruit
Fleshy fruit in P. ericoides (Figure 1D,F) and P. rigida is possibly correlated with the moist maritime climate of the coastal habitat of these species; it is possibly also an adaptation to bird dispersal.Both species occur in the Western Cape, and the range of P. rigida extends along the coast to St Lucia.P. ericoides has red berries and P. rigida has yellow berries.All other species are char acterized by achenes and are adapted to drier habitats, from mountainous areas along the Great Escarpment to the arid Karoo.In Passerina.each achene is enveloped by papyraceous remnants of the fragmented hypanthium and enclosed within the tomentum of an enlarged bract (Figure 2D.E), which often takes on a rounded shape and turns yellow, red or brownish.
The fruits of Passerina clearly illustrate the phenome non of transfer of function from protection to dispersal (Stebbins 1974).In P. ericoides and P. rigida, with fleshy fruit, the pericarp has the double function of protecting the ovules during early stages of development and disper sal.The mature fruit enlarges beyond the bracts and is protected by the coriaceous pericarp, while the patent bract does not have a protective function (Figure ID ).The pericarp of the fruit is yellow or red when it is ready for dispersal, probably by birds or small mammals (Richards 1986), and the dispersal unit is the berry.In all other species which are characterized by achenes, the protec tion of the ovule is apparently transferred from the peri carp to the persistent bract.The bract enlarges around the achene.protecting it in the woolly tomentum of the con cave adaxial surface (Figure 2D).The mature fruit is often still enveloped by the reddish, papyraceous rem nants of the hypanthium.Both P. montana and P. glom erata are characterized by subcapitulate inflorescences, with proliferating growth more common in the latter species.In P. glomerata (growing in the arid Karoo) the accompanying bract turns yellow and becomes more patent when the fruit is mature; the achene is shed after abscission.The yellow colour is associated with senes cence of the bracts and these structures are eventually shed, leaving conspicuous bract scars on the remaining, often terminal, woolly inflorescence axis.The unit of dis persal. in this case, is the achene, assisted by the patent senescing bract.The achene falls the ground where it could either be dispersed by ants or small mammals or germinate under favourable conditions.In P. montana (growing along the Great Escarpment), the margin of the fruiting bract turns red and it becomes more patent, exposing the achene enveloped in the beak-like, reddish, papyraceous hypanthium, which fragments at the neck base.Perhaps birds, attracted by the red colour (Richards 1986).feed on the exposed achenes.Leafless, terminal, woolly branchlets, with terminal scars are a conspicuous feature of the plant after fruiting and it therefore also seems possible that the subterminal capitulum w ith sever al achenes is broken off as a unit.In this case the disper sal unit could be the achene or the achene accompanied by the bract or perhaps even the entire subcapitulum.There is a need for further observations on seed dispersal in the field to test some of these suggestions.

Floral envelope
The hypanthium and sepals in P. ericoides are char acterized by their coriaceous (almost fleshy) appearance and dull green to pinkish colour.The floral envelope in P. rigida.P. paleacea.P. sp. nov. 1 and P. sp. nov. 2 is pale yellowish and quite membranous.P. pendula is dis tinguished by a pinkish floral envelope, with a membra nous texture.In all the other species the floral envelope is yellow-pink at anthesis, with a papyraceous texture.After pollination these flowers turn red and the hypan thium and sepals become thinly papyraceous and dry.
For practical purposes the total length of the floral envelope indicates flower size, and its taxonomic impor tance is clearly illustrated by the general increase in size from species 1-20 (Table 7) A strong correlation was found between the indumen tum of the lower hypanthium and of the bract.There is a tendency for species characterized by a glabrous hypan thium base (Table 8) to have a protective bract with a very hairy adaxial surface (Table 5), whereas species with a hairy hypanthium base have a less hairy to almost glabrous adaxial bract surface.When the hypanthium neck is not covered by a bract, it tends to be hairy in varying degrees.Only P. paleacea has a completely glabrous hypanthium.In this species the entire hypanthi um is completely covered by the hairy adaxial surface of the bract because of the very short hypanthium neck.This tendency shows that the function of protection of the ovule is partly transferred from the hypanthium to the hairy bract (Stebbins 1974).

c. Indumentum o f sepals
Thoday (1924) uses the character 'outer sepals beard ed behind the apex' in his key to distinguish between species.The abaxial surfaces of the outer and inner lobes of seven species are setose at the apex only and the adax ial surfaces range from glabrous to variously hairy (Table 8).In P. com osa abaxial surfaces of both outer and inner lobes are tomentose and adaxial surfaces are glabrous, P. drakensbergensis is similar except for the tomentose adaxial surface o f the inner lobes.A ll sepals are com pletely glabrous in P. paleacea.The indumentum of the sepals varies infraspecifically and should, however, be used with discretion to distinguish between groups of species.

d. Size and shape of sepals
The size of the inner sepals is of taxonomic impor tance (

Less important taxonomic characters
Ovary size (Table 7) has been considered less impor tant.as it is difficult to measure all ovaries at the same developmental phase.In Passerina the size of the ovary increases markedly after anthesis and the enlarged ovary, enveloped by the persistent hypanthium.can already be observed in older flowers, rapidly followed by matura tion of the fruit.

Taxonomic relationships
Up to 1930.priority was given to the definition of subordinal or subfamilial taxa in Thymelaeaceae, based mainly on floral morphology (Table 1)

Family level
The exotegmic palisade and the distinctive obturator are regarded as family characters.They form the basis of C om er's (1976) Euphorbiales-Malvales-Thymelaeales-Tiliales complex.

Subfamily level
The

Genus level
The present study indicates the exserted.extrorse anthers and the anemophilous habit as unique to P asse rina.

Species level
Characters useful at species level are summarized in Tables 7 and 8.

CONCLUSIONS
The evidence on floral morphology not only con firmed the identity of 20 species and four subspecies, but also proved significant in the taxonomy of the genus.The status of the following taxa is confirmed by the present floral morphological study; P. burchellii anatomy, indicating the arrangement of taxa in Passerina
FIGURE 5.-LM photographs of P. ericoides, Bredenkamp 956, with fleshy fruit, and P vulgaris, Bredenkamp 951, with dry fruit showing structure of receptacle and base of ovary.A-D, I/s at different levels of receptacle.A-C, P. vulgaris: A. departure of eight traces from central stele; B. carpel lary bundles arranged in ring, fused stamen and sepal traces in peripheral position; C. differentiation of cortex tissue into inner epidermis of hypan thium and outer epidermis of ovary wall, separation of dorsal carpellary bundle.D, comparison to similar stage in P. ericoides.E-H.t/s at base of ovary showing vasculation of ovary wall, funiculus and chalaza of pendulous ovule: E, P. ericoides, ovary wall independent of hypanthium.carpellary bundles in a single whorl; F, P. vulgaris, young bud, ovary wall loosely adhering to hypanthium distally: G. P. ericoides, one cell of functional macrospore; H, comparison to similar stage in P. vulgaris, reduction of dorsal carpellary bundle, c, carpellary bundle; cc, commissur al carpellary bundles; cs-pst, fused commissural sepal and antipetalous stamen bundles; d, area of tissue differentiation: dc.dorsal carpellary bun dle; f, funiculus; fm, functional macrospore: h, hypanthium; ieh, inner epidermis of hypanthium; ieow.inner epidermis of ovary wall: ii.inner integument; me, median carpellary bundle: n, nucellus: o, ovule; oeow, outer epidermis of ovary wall: oi, outer integument; ow.ovary wall; p, placenta; s-sst, fused sepal midrib and antisepalous stamen bundle; t, trichome: vh.vascular bundle of hypanthium.Scale bars: A-H, 100 pm.

For a comparison o f
Filament length, it is easier to measure the antisepalous filaments as they are ± twice the length of the antipetalous ones.Both P. sp. nov. 1 and P. sp. nov. 2 have short antisepalous filaments of ± 1.2 mm, corresponding to their small flowers.P. paludosa, P. filiform is and P. falcifolia have exserted stamens because o f their long (2 .1-2.4 mm) filaments.
. P. rigida.P. paleacea and P. sp. nov. 2 are characterized by small flowers, the length of the floral envelope 4.0 -4 .6 mm.In most other species it ranges from 5.3-7.3mm long.P. rubra and P. falcifolia are characterized by large flowers, der Walt 1995), different trichome types are found below and above the articulation plane in the hypan thium.This state could possibly also be present in other genera of the Thymelaeaceae.In Passerina the trichome type remains constant over the length of the hypanthium.possibly because there is no articulation plane in the hypanthium.As Passerina is distributed over a w ide range of habitats, the density of the indumentum has been impor tant in the distinction of certain species (Table 8).P. eri coides is distinguished by the strigose indumentum over the length of the hypanthium.whereas the indumentum of the neck is strigose in P. paludosa.A completely glabrous hypanthium is characteristic of P. paleacea.In 12 of the species, the hypanthium surrounding the ovary is glabrous, scantily tomentose or tomentose at the apex, with the neck scantily tomentose or tomentose.In P. sp. nov. 1. P. sp. nov. 2 and P. filiformis subsp.filiformis the hypanthium is tomentose over its entire length.In P. galpinii the whole of the hypanthium is pubescent, w hereas in P. rubra only the neck is pubescent and the rest o f the hypanthium is glabrous.
. Applying both morphological and anatomical evidence.Domke (1934) proposed a satisfactory subfamilial classification system and envisages a phylogenetic relationship between the Thymelaeaceae.Malvaceae and Euphorbiaceae.Modem techniques have enabled taxonomists to find relation ships between families and to arrange them into higher hierarchies, with ranks such as superorders or account o f the strongly perigynous polypetalous to apetalous flowers, internal phloem, ves tured pits and obturator.However, he admited a possible relationship with other families, based on the pseudomonomerous ovary and crotonoid pollen.Dahl gren (1975a, b) placed the superorder Thymelaeanae between the Dillenianae (Dilleniales, Cistales, Malvales, Urticales, Euphorbiales) and the Myrtanae.Within the superorder Malviflorae, Dahlgren (1980) recognized a close affinity between the Malvales and Euphorbiales, and indicated a strong relationship with the Urticales and the Thymelaeales, but a phylogenetic affinity between the Malviflorae and Myrtiflorae was not supported.The inclusion of the Thymelaeaceae in the Myrtales was reviewed by Dahlgren & Thome (1984).They argued that anatomically, most members of the family possess Myrtalean characters.On the other hand, embryological.chemical and palynological evidence strongly indicates an affinity with the Malviflorae.Thome (1992a) accept ed the superorder Malvanae, but included the Thyme laeaceae in the order Euphorbiales.Takhtajan (1969) considered the Thymelaeales to have a common origin with the Euphorbiales and Malvales, all arising from a Flacourtiaceae-type ancestor, and Takhtajan (1997) placed the Thymelaeales in the superorder Euphorbianae, with the Gonystylaceae as the only other family in the order.Palynological evidence indicates that the very distinc tive pollen in Thymelaeaceae is totally different from that o f any Myrtales and similar to that o outer integument and the differentiation of the inner integument into a palisade-like outer epidermis, a mesophyl layer and an inner epidermis (Figure 10A.B).It is also clear that the mechanical part o f the seed coat is derived from the palisade-like outer epidermis, hence it is an exotegmen (Figure IOC, D).Corner (1976) distin guished the Euphorbiales-M alvales-Thym elaeales-in context with the evidence from other botanical fields.The primary focus of The Angiosperm Phylogeny Group (APG 1998) is on orders, with a secondary focus on families of flowering plants.Above the ordinal level, ranks are defined as subgroups, clades or supraordinal subgroups.Magallon et al. (1999) attempted to compare these groups to the existing systems mentioned in the preceding paragraphs.Both APG (1998) and Magallon et al. (1999) recognized the Eudicots, a group characterized by tricolpate pollen, as well as the supraordinal group Core Eudicots, supported by pentamerous and isomerous flowers.Thymelaeaceae is desig nated to the Rosid clade by most authors.APG (1998) placed the Thymelaeaceae in the subgroup Eurosids II and order Malvales, whereas all the other authors placed it in the Expanded Malvales.The Malvalean relationship of the Thymelaeaceae seems to be strongly supported by molec ular phylogeny, as well as floral morphology, anatomy, embryology and palynology.The Euphorbiales-Malvales-Thymelaeales relationship indicated by embryology and palynology is, however, not supported by molecular data.APG (1998) placed the Euphorbiaceae in the order Malpighiales in the subgroup Eurosids I and Magallon et al. (1999) placed it in the Core Rosid Clade.Cronquist (1968.1981, 1988) was convinced that if the Thymelaeaceae is not placed in the Myrtales, it would stand next to it.Conti et al. (1996), APG (1998), and Alverson et al. (1999) all regarded the Myrtales as a sister group of the Malvales or the Expand ed Malvales to which the Thymelaeaceae is designated.Speculations on phylogeny Within Thymelaeaceae, both Domke (1934) and Heinig (1951) agreed that the subfamily Thvmelaeoideae is phylogenetically more advanced than the Aquilarioideae.On the basis of the advanced pollen.Bredenkamp & Van Wyk (1996) raised the subtribe Passerininae to the tribe Passerineae.a decision supported by the present study.Although many of the follow ing advanced charac ter states are present in other genera of the Thymelaeoideae.these advanced character states are all found together in Passerina: receptacle reduced to a ± lenticu lar structure; departure of the fused sepal and stamen bundles before carpellary bundles; hypogynous floral arrangement; petal-like floral envelope comprising a hypanthium (fused calyx and androecium), differentiat ing into four sepals and a diplostemonous androecium; separation of stamen bundles high up in hypanthium.at formation of sepals; exserted.extrorse anthers; anemophilous habit; secondary reticulum o f pollen; complete absence of petals or petaloid scales; asymmetric devel opment of the style; superior, pseudomonomerous.uni locular ovary; asymmetric attachment of ovule at top of ovary; ventrally epitropous ovule; distinctive obturator; bitegmic ovule with exotegmic palisade; fruit a 1-seeded berry or an achene; seed with lignified.black exotegmen.Considering all the characters mentioned.Passerina is considered highly advanced in relation to other genera in Thymelaeoideae.
On the basis o f the secondary reticulum, unique to the pollen o f Passerina, Bredenkamp & Van Wyk (1996) raised the subtribe Passerininae to the tribe Passerineae.

For
almost three centuries evidence from floral mor phology has been basic to plant taxonomy and applied at all hierachical levels.Our research on the flowers in Passerina has produced new morphological and anatomi cal evidence, especially as Heinig's classical study of floral morphology in Thymelaeaceae (1951) did not include Passerina.The present study has succeeded in resolving the floral morphology in Passerina, as many mistakes have been perpetuated by previous authors.We conclude that the flower in P asserina is a phylogenetically advanced structure and consider the genus advanced within the Thymelaeoideae.Possible taxonomic relation ships of the Thymelaeaceae with the Malvales are strong ly supported by the presentation of floral morphological, anatomical, embryological and palynological information to which this study has also contributed.Finally, all the above-mentioned evidence serves to form a firm taxo nomic basis for future comparative studies, especially in the expanding molecular field.
Proliferating spikes with inflores cence apex growing out and returning to vegetative growth, common; main florescences and co-florescences subtermi nal.