A monograph of the fern genus Pyrrosia ( Polypodiaceae )

This monograph deals with the systematics and phylogeny of the genus Pyrrosia (Polypodiaceae). It is one of the results of the Polypodiaceae-project that is being carried out under the supervision of Prof. Dr. Hennipman (State University, Utrecht). With this treatment of Pyrrosia, the group of the Platycerioid ferns (sometimes regarded as a subfamily) will be completed. The total number of genera in this group is reduced to two: Platycerium and Pyrrosia. Other publications dealing with this group are Hennipman & Roos (1982), dealing with Platycerium; and Ravensberg & Hennipman (1986), dealing with the former genera Drymoglossum and Saxiglossum, now in Pyrrosia. Together with Hennipman & Ravensberg’s account this forms the first complete account of Pyrrosia. The genus has been subject to revision on a world-wide scale only once before (Giesenhagen, 1901). In that work, however, several species were omitted and the genera Drymoglossum and Saxiglossum were not included. After Giesenhagen’s monograph, Pyrrosia has been the subject of several regional studies, most notably Ching’s (1935), which deals with the species from the mainland of Asia including Japan and Taiwan. The African species were revised by Schelpe (1952), the Australian ones by Tindale (1961). Nayar & Chandra’s (1965) account of the species from India emphasizes the morphology rather than the taxonomy of the species treated.

found that in the material at hand the criteria for the delimitation and discrimination of species that were used by Giesenhagen and Ching could not always be applied. In Giesenhagen's work, undue stress is placed on minute variations in lamina anatomy; in Ching's work a similar, and in my opinion unjustified, emphasis is put on details of the indument. Therefore, a morphological analysis was carried out, resulting in the recognition as important taxonomic characters of e.g. the distribution of sclerenchyma in the rhizome and the sculpture of the perispore. The range of the variation in Pyrrosia for both these characters appears to be unique in the Polypodiaceae. Details of the sculpture of the perispore are treated separately by Van Uffelen & Hennipman (1985). Other characters that were found to be important are the morphology of the rhizome, of the rhizome scales, as well as the presence or absence of specialized paraphyses.  (Giesenhagen, 1901). In that work, however, several species were omitted and the genera Drymoglossum and Saxiglossum were not included.
After Giesenhagen's monograph, Pyrrosia has been the subject of several regional studies, most notably , which deals with the species from the mainland of Asia including Japan and Taiwan. The African species were revised by Schelpe (1952), the Australian ones by Tindale (1961). Nayar  account of the species from India emphasizes the morphology rather than the taxonomy of the species treated.
2 Using these and other characters 51 species were recognized, two of which were new (Hovenkamp, 1984) and 6 other species were reduced to the level of variety.
Evidence was sought for the occurrence of hybrids among in the genus, but no indications could be found that hybridization plays an important role in the development of new species in this genus. The available evidence from chromosomes indicated, however, that in several species polyploid complexes occur.
As a result of the morphological analysis hypotheses could be formulated about the probable course of evolution of several characters. With the aid of these hypotheses, and by using cladistic techniques, an explicit idea could be formulated regarding the phylogeny within the genus. This, in its turn, led to an elaboration of ideas concerning the historical geography of the genus.
As a result, a hypothesis is formulated according to which the origin of the genus must be placed in what is now Africa, some time in the Jurassic. The present distributon of the genus was reached in two ways: one via 'rafting' on the Indian subcontinent, one via 'rafting' on Australia. 1 : 3), then lifted from the rhizome. In order to obtain whole scales it is sometimes necessary to dissect carefully as the scales break off easily above the point of insertion. The scales were then rinsed in water and semi-permanent preparations were made by embedding them in glycerin-jelly.
Fronds. The hairs on the stipe and lamina were treated in a similar way, but careful dissection is not necessary here.

Anatomy
Rhizome, stipe. For anatomical study pieces of rhizome or stipe were boiled in water, usually after adding a few drops of detergent (see above), for a few minutes, or until they sank when transferred to cold water. They were sectioned at ± 30 pm with a Reichert slide microtome using a Gilette double-edged razorblade. The sections were embedded in glycerin jelly. No staining or other treatment of the sections was found to be necessary to study the distribution of the vascular strands and/or the sclerenchyma.
Rhizome scales. Cross-sections through the rhizome scales were often found to be included in cross-sections of the rhizome. Thinner sections were prepared 4 by boiling the scales in water for several minutes and then sectioning the softened scales at 5-10 (xm using a slide microtome with a double-edged razor blade and elder-pith. Sections of scales obtained in this way were found to be superior to sections obtained after embedding the scales following routine procedures. Lamina. Cross-sections of the lamina were prepared in a similar way as those of the rhizome (see above) but were sectioned at 20-30 pm. The sections were bleached in a commercial solution of hypochlorite, stained for a few minutes in Astra-blue, and embedded in glycerin jelly. To obtain epidermis peels, as a rule small pieces of lamina were bleached following a procedure slightly modified from that used for studying the venation (see below). They were simply boiled in a 5 % solution of KOH, then bleached in hypochlorite. Either the lamina in its entire thickness was then embedded in glycerin-jelly, or the mesophyll was scraped away as much as possible. Staining with Astra-blue was sometimes applied. Due to the diversity in anatomical structure of the lamina the results thus obtained were highly variable in quality. More conventional methods of obtaining epidermis preparations by maceration were found to yield equally unsatisfactory results and are more laborious.

Venation
For study of the venation sufficiently large pieces of lamina were cleared following the procedure developed by O'Brien & Von Teichmann (1974). This involves autoclaving the pieces in a 5 % solution of KOH and then bleaching them in commercial hypochlorite. For Pyrrosia this was found to yield better results than bleaching with chloro-lactophenol according to Hennipman (1977).
Unfortunately, after the treatment the tissue had become so brittle that remounting the pieces on the herbarium sheets was impracticable. The venation patterns were photographed following the method developed by Hennipman (1977).

Sporangia, spores
Sporangia and spores were treated in a similar way as the indument of the lamina.

Juvenile stages, gametophytes
To study the stages of the heteroblastic development, spores were sown on an Agar medium. In various stages of development gametophytes or juvenile sporophytes were fixed in FAPA. This was found to clear the tissue sufficiently in the course of a few days, so that details of venation, epidermis and indument could be observed without further treatment. For observation, specimens were transferred to water or observed in FAPA.

Ontogeny
The development of hairs and scales was studied in a few species. Actively growing shootor frond-apices were macerated and stained in acetocarmine and squash-preparations were made in 45 % acetic acid. No attempt was made to convert these preparations into permanent ones.

Chromosomes
For cytological study actively growing root-tips were fixed in Carnoy's fluid (alcohol:chloroform:acetic acid -6:3:1) for one hour to several days at room temperature and stained and macerated by boiling for several minutes in acetocarmine. Squashes were made and photographed. No attempt towards more permanent preparations was made. Pretreatment with alpha-bromo-naphtalene was not found to produce visible results.

Observations
Light-microscopical observations were made with a Leitz Laborlux microscope with drawing-arm; for photography a Leitz Ortholux microscope was used.
For scanning-electron microscopy, rhizome scales were dissected as indicated above, air-dried, and then mounted on aluminium stubs using double-stick tape or silver-containing conducting glue. For study of the indument, small pieces of lamina were taken from herbarium-specimens and mounted (lower side up) as indicated above. In some cases fresh pieces of lamina were dehydrated in a graded acetone series and critical-point dried using C02. All preparations were coated with gold (using a Polaron E 5100 series II sputter-coater) and observed with a JEOL JSM-35 scanning electron microscope. 6 3.

Taxonomic history
The genus Pyrrosia was established by Mirbel in 1803, but its history goes back at least one more century.
A plant that is almost certainly a Pyrrosia (most likely P.  stigmosum Swartz, 1801).
In 1803 Mirbel (Hist. Nat. Veg. 5) established the genus Pyrrosia with P. chinensis ( = Pyrrosia stigmosa) as the only species. In the same work he established the genus Candollea with 4 species, three of which are now included in Pyrrosia. C. heterophylla is probably a mixture: Mirbel cited Acrostichum heterophyllum L.
( = Pyrrosia heterophylla (L.) Price) but also included plants from Java, where P. heterophylla does not occur, and the description that he gave of C. heterophylla cannot be matched with any species now known from either India, Africa, or Java.
The next two species, Candollea longifolia and the 'candolline lanceolee' (Mirbel did not make the combination C. lanceolata) probably represent  Cyclophorus. Throughout the 19th century, most of the species of Pyrrosia were referred to Niphobolus by authors who considered the genus distinct from Polypodium J. Smith, 1842;Fee, 1853-4;J. Smith, 1875;Beddome, 1863Beddome, , 1892Giesenhagen, 1901); whereas the name Cyclophorus was largely neglected until it was brought up again by Underwood (1903). Other authors, among whom Mettenius (1856) and Hooker (1863) were the most influential, followed Swartz (1806)  In 1906 Christensen (Index Filicum) followed Underwood (1903) and reinstated the older name Cyclophorus, making all necessary and some unnecessary new combinations. Neither Giesenhagen nor Christensen altered the circumscription of the genus, in which they have been followed by most modern authors.
In 1931 Farwell revived the by then completely forgotten name Pyrrosia, making several new combinations. A number of others were made by Ching (1935) in his revision of the species from China.
Several authors have attempted to divide the genus.
Presl recognized two groups in 1836; in 1851 he distinguished 8 genera. In this he has not been followed by other authors. Apart from that, in 1836 he transferred one species to the new genus Drymoglossum, where it was placed together with a species that is not a Pyrrosia. Subsequent authors expanded the concept of Drymoglossum into a heterogeneous assembly of species, many of which were removed to other genera by Christensen (1929). Drymoglossum in Christensen's (/. c. ) narrower circumscription has been accepted by many modern authors (Backer & Posthumus, 1939;Copeland, 1947;Holttum, 1954;Nayar, 1957;. All species, however, were transferred to Pyrrosia by Price (1974), and subsequent investigations tend to confirm this view (Ravensberg & Hennipman, 1986).
In this he was not followed, but the name Niphopsis was taken up by  and Shing (1983) for a subgenus and a section within Pyrrosia, respectively, though with circumscriptions that differed widely from each other.
This was also rejected by other authors.
In the present work, Pyrrosia is considered as including both Drymoglossum (in the circumscription of Christensen, 1929) and Saxiglossum, but the species formerly assigned to these two genera are treated in detail by Ravensberg & Hennipman (1986).

Exploration
The total number of taxa accepted in the present work is 57. A historical review of the number of described taxa is given in Table 1, compared with the number of taxa accepted here.
As shown by this table the highest number of taxa now accepted (second column) has been described during the 19th century, and the number dwindles rapidly in the 20th century.
The effect is more pronounced if it is taken into account that relatively many species are only described a considerable time after the first specimens have been collected. These species have gone unnoticed for some time and are recognized only when a critical revision is made of a large number of species. Thus, Giesenhagen (1901) described three new species from material that  ( Niphobolus christii, N. ceylanicus and N. mannii); the present revision has resulted in the description of two new species, neither of which is based only on recently collected material.
A comparison with the total number of taxa described in each period ( Table   1, first column) shows that the diminishing number of accepted taxa is not the effect of diminishing efforts on the part of descriptive taxonomists. Almost the reverse seems to be the case: increasing numbers of taxa described do only occasionally result in increasing numbers of accepted taxa. Particularly in the 20th century the "law of diminishing returns" seems to be at work. If this general concept applies, it can be considered an indication that the total number of Pyrrosia species awaiting discovery and description is rapidly decreasing in relation to the number of taxa already described. Expectations that each large collection will yield at least a few new species seem to be unfounded, at least for Pyrrosia.
A closer examination of the history of the exploration of the genus tends to confirm this.
The earliest species known were all collected in the vicinity of trading posts in Ceylon (Herman), Japan (Thunberg), or Java (Thunberg, Burmann); or else they were found during the early exploration of the Pacific (Forster). Exploration of these relatively easily accessible places was sufficiently intense, so that when Blume enumerated the ferns of Java (1828), he included almost all species now known from that island (Hennipman, 1979). Many of the species of Pyrrosia occurring on the Malesian islands were collected relatively early during the exploration by, e.g., Cuming, Zollinger, and Korthals; and the intensive exploration during the 20th century (e.g., by Clemens, collectors from Buitenzorg, the Philippine Bureau of Science) did not result in the discovery of a proportionally large number of new species.
In New Guinea, all species now known had been collected already at least once before the exploration by the Archbold expeditions, and recent large collections by Croft and others in the LAE series have not yielded any new species. This is in striking contrast to the situation in some other genera, of which Grammitis (recently revised for New Guinea by Parris, 1983) may be taken as an example.
In Parris' revision, out of a total number of 64 accepted species, some 30 are based on material collected by Brass or other collectors active since around 1925. This is a large proportion compared to Pyrrosia, and the difference cannot only be due to a different species concept that may have been used in the present work.
In other areas a similar situation prevails.
In India and Ceylon, for instance, by far the largest number of species was already collected by Wallich, Griffith, Thomson, Hooker, Mann, and other collectors active in the 19th century.
In China the species were discovered mainly by the early explorers (Henry, David, Shearer), intensive exploration in the 20th century by Forrest, Wilson, 10 Rock and Chinese collectors (C.W. Wang, W.T. Tsang and others) has not resulted in the discovery of new species accepted here. Therefore, the conclusion must be that further exploration is not likely to lead to the discovery of a considerable number of new species. Further efforts in the field therefore might more profitably be directed towards more detailed investigation of some of the problems in the genus, e.g., the description and elucidation of possible mechanisms that may be concerned in maintaining or not maintaining the distinct entities found in P. lanceolata and P. porosa.
The reason for the relatively complete knowledge of the genus may be in the ecology of the species. Most species are either common species, widely distributed throughout the tropical lowlands, extending frequently into gardens, plantations, etc.; or else they are species with a distinct preference for the middle elevations in hilly areas, areas that have been favoured as summer resorts in many cases on account of the pleasurable climate in contrast to the hot, humid lowlands. None of the species is confined to the less accessible higher montane areas, and there are hardly any species that are both rare and restricted to the lowlands.

SUBDIVISION OF THE GENUS
In this work I do not present a formal subdivision of the genus. Although several groups of species can be recognized that have distinctive characters, recognition of these groups would leave us with a considerable number of "species incertae seats : species tnat lack all characteristics necessary to assign them to any ol the groups; or species that have combinations of characters that would make assignment difficult, arbitrary, or even impossible. Such a subdivision of the genus in my opinion is better not formalized.
Another reason is that, although several of the easily recognizable groups are so distinctive that they can safely be assumed to be monophyletic, at least some groups would be paraphyletic or polyphyletic. Recognition of paraphyletic groups is not desirable. It is defended sometimes with the argument that 'natural' groups do not necessarily have a synapomorphy (Geesink, 1984). The existence of a monophyletic group, however, is evidenced only by an autapomorphy, and in the absence of other knowledge about the course of the evolution, autapomorphies are the only arguments for the monophyletic character of a group. In some cases it may be necessary to recognize para-or polyphyletic groups to comply with the basic requirements of taxonomic practice. Complete avoidance of para-or polyphyletic groups would make classifications unwieldy with a very high number of groups of different ranks. The formal recognition of series, subgenera, etc., in the case of Pyrrosia, where a relatively small genus is involved, is not one of those cases where more would be gained than lost.
For the same reasons, the arrangement of the species in the Taxonomic Part is strictly alphabetical. To facilitate discussion, however, those groups that can easily be recognized are recognized informally. Throughout the discussion in the general chapters they are called by the name of one of their most characteristic, or most common, members; or by the number assigned to them here. A discussion of the monophyly of these groups is given on pp. 93-99.
These species are similar in rhizome structure, rhizome scales, frond shape, venation, stomata, and spores.
The rhizome is distinctly elongated between the phyllopodia; the lateral buds are close to the phyllopodia; the tissue is completely parenchymatous. The scales are pseudopeltate. The fronds are estipitate and 12 often slightly falcate; with an irregular venation without distinct secondary veins.
The stomata are polocytic. The spores have a very thin, tightly adhering perispore; the exospore is often strongly ornamented.
These species are similar in rhizome structure, rhizome scales, venation, sori and spores.
The rhizome is generally rather thick and completely sclerified; the scales are basifix and not thickened.
The venation of the lamina is characteristically complex, with very strongly developed secondary veins and an intricately anastomosing pattern of included veins. The sori have few sporangia each, and sporangium structure is peculiar (p. 62). The spores are smooth, with a thin, tightly adhering perispore. 3. The P. porosa-group: P. porosa, P. assimilis, P. linearifolia, P. stolzii, P. rhodesiana.
These species have few characters in common; they are nevertheless all rather similar in appearance: the rhizome is shortly elongated, with peltate scales. The fronds are estipitate, with a simple venation pattern with mainly simple, excurrent included veins.
4. The P. sheareri-group: P. sheareri, P. drakeana, P. hastata, P. polydactyla, P. sub furfuracea, P. boothii, P. flocculosa. This group is characterized by the combination of a thick, short rhizome and distinctly stipitate fronds. P. flocculosa is in some ways aberrant: it has fewer sclerenchyma strands in the rhizome than the other species; it has dentate, instead of ciliate-dentate, rhizome scales; and the perispore is not as densely granulate as those of the other species. 5. The P. lingua-group: P. lingua, P. petiolosa, P. christii, P. sphaerosticha, P. ab breviata.
The typical combination of characters for this group is the long-creeping rizome with ciliate scales; the slightly to distinctly dimorphic, distinctly stipitate fronds; and the indument with appressed, boat-shaped rays.
The group, however, is markedly heterogeneous with regard to perispore morphology. 6. The P.
These species are similar with regard to certain aspects of rhizome structure, rhizome scales, frond, and spores. 8. The P. confluens-group: P. serpens, P. confluens, P. rupestris, P. eleagnifolia. This is a group of similar-looking species but with relatively few unique characters. The rhizome scales of many specimens from species in this group show annular figures as described on p. 27. The fronds are small, and dimorphic in various ways.
There is a tendency for the hydathodes to be restricted to a single marginal row, and they are absent in P. eleagnifolia. The indument is uniformly monomorphic and appressed. The sori are relatively large, often more or less confluent, and usually contain many paraphyses with short, straight rays.
9. The P. lanceolata-group: P. lanceolata, P. ceylanica, P. longifolia, P. fallax. This is a distinct group characterized by the deeply sunken sori with centrally situated paraphyses that are arranged in a bundle in most species, or in a central row in P. fallax. In many other characters this groups also is homogeneous.
It is characterized by the presence of coenosori, the characteristic venation with recurrent included veins, and the homogeneous lamina structure. Although some of the characters of this group (e.g. the peculiar venation) are unique in Pyrrosia, in other characters (spore morphology, rhizome structure, indument) it is similar to the P. confluensor the P. lanceolata-group.
The following species are not included in any of the groups mentioned above: P. mannii, P. penangiana, P. pannosa, P. angustissima, P. gardneri, P. laevis, and P. oveolata. P. mannii and P. penangiana are aberrant in having the combination of a short rhizome with pseudopeltate scales, estipitate fronds, and polocytic stomata. They differ from each other in indument and do not form an obvious group themselves.
P. angustissima shares many characters with the P.
aneustata-group (e.g. the absence of hydathodes, the presence of a coenosorus, the entire rhizome scales), but differs markedly in perispore sculpture. In this preliminary analysis P. angustissima is not incorporated in any group.
P. gardneri is more or less intermediate between the P. porosa-group and the p.
linguagroup: the short-creeping rhizome and the hardly stipitate fronds indicate a connection with the P. porosa-group.
The spores and the structure of the rhizome scales are similar to what is found in the P. linguagroup. 14 P. pannosa has stipitate fronds like the P. linguagroup, but the scales are often pseudopeltate and more like those of P. schimperiana in structure.
P. laevis can be included in the P. lingua-e roup on the basis of indument and perispore sculpture, but is aberrant there on account of the absence of frond dimorphism. P. foveolata has the deeply sunken sori that are characteristic of the P. lanceolatagroup, without any tendency to become confluent; it lacks the central position of the paraphyses in the sorus that is the other character used to delimit the p.

Rhizome
The morphology and anatomy of the rhizome have been studied by Giesenhagen (1901, pp. 23-29) and by Nayar and collaborators (Nayar, 1957, 1961. They chiefly paid attention to the vasculature of the rhizome, which was illustrated by these authors for both short-and long-creeping rhizomes (Giesenhagen, I. c. p. 23 fig 3, p. 27 fig. 4; Nayar, 1957, p. 170 fig. 3-5;1961 fig. 19-21;Nayar & Chandra, 1967, p. 616 fig. 1-3), and this character plays an important part in the phylogenetic considerations of Nayar & Chandra (1967). These authors paid relatively little attention to the various ways in which the sclerenchyma may be distributed in the rhizome.
The comparative morphology of the rhizome has not been as intensively studied.
John Smith (1842) used the gross morphology of the rhizome to distinguish two sections in the genus: the Repentes, and the Caespitosae.

Morphology
The rhizome of Pyrrosia is always horizontally creeping and varies from shortly so to long-creeping ( fig. 1).
Dorsally, the rhizome bears two alternating rows of phyllopodia. The width of these generally differs not much from that of the rhizome, the height is variable but rarely exceeds 5 mm. Only in P. gardneri the phyllopodia are distinctly higher On each side the rhizome bears a row of lateral buds. These are regularly present even if they do not develop into branches, which they do more frequently in shortly elongated rhizome, buds ± halfway down the internodia. 17 some species than in others. As Nayar and Chandra remarked (1967, p.617), each bud is associated with a phyllopodium. In short-creeping rhizomes with closely set phyllopodia ( fig. 1 b) the buds are situated basally on the phyllopodia, in an abaxial-lateral position, but in species with more or less elongated rhizomes and longer internodia the position of the bud varies. In some species in which the phyllopodia are up to several cm apart (e.g., P. africana, P. platyphylla, fig. 1 e) the buds are placed very close to or on the phyllopodia, just as they are in species with short-creeping rhizomes. In other species the buds are shifted downwards to a position up to halfway down the internodia (e.g., P. rhodesiana, P. pannosa). In most species with long-creeping rhizomes (e.g., P. abbreviata, fig. 1 d) the buds are constantly situated from halfway to farther down the internodia, in some species (e.g., P. nummulariifolia, fig. 1 c, P. foveolata) they usually occupy a position opposite the phyllopodia and apparently are shifted downwards over the whole length of the internodium.
Other ferns with a vining habit have buds in the same position, e.g., Microgramma (Hirsch & Kaplan, 1974). A detailed analysis of these cases might reveal a similar remote association of buds and phyllopodia, although Hirsch & Kaplan (I. c.) in their morphological analysis concluded that there is no relation between buds and fronds.
In cross-section the rhizome is often ± rounded, but in very short rhizomes the dorsal side is often somewhat flattened due to the very closely set phyllopodia.
In many long-creeping rhizomes, on the other hand, the ventral side is often more or less flattened or even distinctly furrowed. Often this is more distinct in dried specimens, but the frequent occurrence of rhizomes that are round in dried state indicates that the ventral groove is not an artifact of drying.
Branching of the rhizome is exclusively by means of development of the lateral buds. In many species the rhizome is always profusely branched, but in others the parts collected are often unbranched. This is particularly so in the species with short creeping rhizomes. In many cases where branches were present they have probably been distributed as duplicates, in other cases truly unbranched rhizomes may occur. Specimens of P. polydactyla and P. sheareri did not show any branching even after a long period of cultivation in the greenhouse .
Apical dichotomous branching occurs occasionally in galled rhizomes of P. eleagnifolia and P. confluens, both species with a long-creeping rhizome.
Anatomy Vascular structure (plate 1; fig. 2). The stele of Pyrrosia has been called a "highly dissected siphonostele" by ) and a "perforated dictyostele" by Schmid (1982, p. 864 overlap, solenostelic if they do not, and the two types are considered as closely related, and difficult to distinguish in highly perforated steles (Schmid, I. c. p. 900). From the drawings of stelar structure given by  it appears that in Pyrrosia the frond-gaps may overlap (their fig. 1) or not (their fig.   3), or only just a little ( fig. 2), which confirms the fundamental unity of the two types postulated by Schmid.
In cross-sections 3 to c. 13 vascular strands may be seen, the number of strands being roughly correlated with the thickness of the rhizome. In thin, long-creeping rhizomes (especially those of the P. lanceolata-and P. confluens-group) often five strands (pi. If) are present in a fixed arrangement: two ventral strands, two lateral ones, and a relatively thick dorsal strand. In thicker rhizomes of these species a larger number of strands may be present, but this may be due to a larger number of anastomoses between the five main strands.
Branch traces are composed of one to several strands. Branch traces with several strands occur in the P. costata-group (e.g., P. stigmosa, fig. 2 a-d). Traces with a double strand occur in P. subfurfuracea and P. schimperiana, traces with either a double or a single strand in P. porosa (double strand: fig. 2 i-1) and P. penangiana.
A double strand is also found in P. nummulariifolia and in P. piloselloides, but here the two strands quickly fuse to a single one entering the branch. In all other species that were investigated a single strand was found.
This single strand is often U-shaped but sometimes it is fully cylindrical for a short distance before splitting up into separate strands in the base of the branch ( fig. 2 t). Due to the relative inaccessibility of this character (either laborious dissection is necessary or large series of consecutive sections must be made) no fully representative survey of the genus has been made, and results should be interpreted with caution.
Frond traces have 2-9 strands, the number depending mainly on the size of the frond. When more than 2 strands are present in the frond trace they are in a Ushaped configuration, with two prominent strands on the open side. This arrangement is similar to the vascular arrangement in the stipe, and is already distinct at a short distance from the rhizome bundles. In species where the phyllopodia are very closely associated with the lateral branches, some of the strands to the branch may originate from the frond trace instead of directly from the rhizome bundles (e.g., in P. stigmosa, P. princeps); in other cases branch trace and frond trace are completely separate.
Non-vascular tissue. In most species the non-vascular tissue is distinctly differentiated into parenchymatous ground tissue and sclerenchyma. Parenchyma cells have thin, uncoloured, walls; the sclerenchyma cells have more or less thickened walls and are often almost completely filled with brown or dark-brown wall-material (pi. 1 h). Generally a distinct sheath of sclerenchyma is present situated peripherally to the vascular cylinder and several cell layers below the 20 epidermis. In some species (e.g., P. rhodesiana, pi. 1 c, P. pannosa) this is less distinct but it is rarely completely absent (P. schimperiana, P. africana, pi. 1 b). In the latter case sometimes a very faint collenchymatous sheath may be found instead. When present, the sclerenchyma sheath is more distinctly demarcated from the peripheral parenchyma than from the inner, central parenchyma. The structure of the cell walls in these strands is similar to that of the walls in the sclerenchyma sheath, but the wall material is often distinctly darker. This is particularly evident in the specimen of P. penangiana mentioned above in which the sclerenchyma sheath extends unusually far inwards: not only the vascular strands are embedded in it, but also some of the many sclerenchyma strands, and these stand out distinctly from the surrounding material. Due to the strong coloration of the wall material I have not been able to assess by way of standard diagnostic staining to what extent lignification is involved in either or both types of wall-thickening in sclerenchyma cells.
In many species the sclerenchyma strands occur in large numbers, scattered in the inner parenchyma (e.g., P. penangiana, the P. subfurfuracea-group) .
In others, fewer strands are similarly scattered in the inner parenchyma (e.g., in P. mannii, P. porosa). Between these two states there is a gradual transition: within several species the number of strands varies considerably (e.g., from 0 to over 20 in P. albicans and P. distichocarpa, from 0 to c. 15 in P. stolzii).
In some cases the sclerenchyma strands are not scattered irregularly in the parenchyma but are restricted to a peripheral zone, and thus occur more or less in between the vascular strands (e.g., in P. angustata, pi. 1 g, P. samarensis, P. abbreviated. Several species (e.g., P. rasamalae, P. asterosora) have a variable but still lesser number of strands scattered in the parenchyma. Sometimes, the absence of strands seems to be fixed in geographical races (P. confluens, P. rasamalae). This can be considered as a step towards complete loss of the sclerenchyma strands, as has apparently occurred in, e. g., P. eleagnifolia, P. christii, P. rhodesiana.
A condition that seems to be strongly fixed in many species with relatively thin, long-creeping rhizomes is the presence of a single sclerenchyma strand, which is usually distinctly dorsiventrally flattened and occupies a central position. This is characteristic for most of the species with a strongly grooved rhizome (e.g., P. 21 foveolata, P. lanceolata, pi. If, P. ceylanica) with 5 vascular strands in a fixed arrangement (see above).
Exceptions to this correlation between rhizome morphology, vascular and nonvascular anatomy occur, however. Some specimens of P. lingua have a single central sclerenchyma strand but not the fixed arrangement of 5 vascular strands, and P. nummulariifolia sometimes has a central strand but less than 5 vascular strands owing to its very thin rhizomes. In other forms of P. lingua sclerenchyma strands are completely absent, or else there may be up to 10 scattered strands. Similarly, in some forms of P. nummulariifolia the sclerenchyma strand is absent. In P. confluens and P. rupestris the vascular strands usually show the fixed arrangement of 5, but the central sclerenchyma strand is often absent or replaced by a small number of strands occupying the same position.
Species of the P. costata-group (pi. 1 a) have a distinctly different rhizome structure. The rhizome is not differentiated into sclerenchyma and parenchyma, but is almost entirely composed of sclerenchymatous tissue. Only a thin peripheral zone is more or less parenchymatous, possibly corresponding to the thin parenchymatous layer outside the sclerenchyma sheaths of other species. The cell walls of the sclerenchyma in this type of rhizome, however, are not as strongly thickened as they are in the distinctly differentiated sheaths of other species. In this they are comparable to the walls of the rather ill-defined sclerenchyma sheath of, e.g., P. rhodesiana, P. porosa, or P. flocculosa on the one hand; on the other to the walls of parenchyma cells.
The lack of differentiation in the rhizomes of the P. costata-sxowa can be compared to that in the P. africana-eroup. The species of that group have fully parenchymatous rhizomes in which no or only very slight differentiation is visible. The abscission pad of the phyllopodia, however, is more or less sclerified in a way similar to that in the rhizomes of the P. costatagroup.
In all other species, the structure of the phyllopodia is similar to that of the rhizome. They have a sclerenchyma sheath in direct continuation of the sheath of the rhizome, but lack the scattered strands of sclerenchyma in the central parenchyme.

Development
The development of the rhizome structure could be studied in some species by way of serial sections through the rhizomes of young sporelings grown in the Leiden Botanical Garden. In P. lingua there is from very early on a more or less distinct sclerenchyma sheath; the stele at that stage is composed of only two vascular strands. As the rhizome increases in thickness the sclerenchyma sheath becomes gradually more distinct. The adult situation in P. lingua would be a very distinct sclerenchyma 22 sheath with or without a number of scattered sclerenchyma strands, but this stage was not reached in the plant studied.
In P. angustata the mature situation is more or less similar to that in P. lingua, but sclerenchyma strands are more constantly present and usually very strongly developed. In the young sporeling a sclerenchyma sheath was present from about P. lingua, though in a less distinct form, but the inner parenchyma was slightly sclerifled as well. As the thickness of the rhizomes increases, the distinction between sheath and inner parenchyma becomes more clear. Sclerenchyma strands did not develop in the stages studied and may be supposed to develop in a later stage.
In P. princeps (P. costata-group) the adult condition is a fully sclerihed rhizome, and this is also found in the young sporeling. Only directly behind the apical meristem the structure is parenchymatous as in all other species studied. Directly behind this apex the sclerenchyma is restricted to a peripheral zone, thus indicating that sclerification develops first in the form of a sheath. The same has been observed in Platycerium-species with a fully sclerified rhizome; here also the sclerenchyma takes on the form of a sheath directly behind the apex of the rhizome.

Roots
Roots are scattered over the ventral side of the rhizome, either very densely set in short-creeping rhizomes, or more sparsely, often in distinct tufts, in longcreeping ones. The roots appear to be initiated continuously at the growing apex of the rhizome. A root, however, soon ceases growth if no contact is made with a suitable substrate. Apparently the ability to develop is then lost permanently: even if contact with a substrate is made afterwards, the new roots arise from new initials directly behind the apex of the rhizome or on new branches. In species with short-creeping rhizomes the rhizome is permanently in contact with the substrate so that all roots can grow out. All these roots thus form a dense, spongy mass around the rhizome. By retaining moisture this may again facilitate the growth of new roots. In species with a long-creeping rhizome the tufted occurrence of the roots is a reflection of the degree to which the rhizome apex has been intermittently in contact with a suitable substrate.
In those species that have a long-creeping rhizome with a distinct ventral furrow the roots arise in two distinct rows on the ridges on both sides of that groove (e.g., in the P. lanceolata-group) . In some species the rhizome is less distinctly and more shallowly grooved (e.g., in P. abbreviata) and here the roots are again scattered over the surface of the rhizome. In thin, elongated and terete rhizomes it is not possible to assess whether the roots arise in two rows or not.

23
Root-traces branch off from the vascular strands running ventrally through the rhizome at irregular distances. Whether the roots are irregularly scattered or arranged in two more or less distinct rows thus depends directly on the number of ventral vascular strands.
The structure of the roots is uniform throughout the genus and has been described by Nayar (1961, p. 166 fig. 23, p. 168). They have a distinct sclerenchyma sheath, which is structurally similar to and continuous with that of the rhizome. This sheath is also present in those cases in which the rhizome does not have a distinct sheath, and is then acquired by the root as it passes through the zone where in other species the sclerenchyma sheath is situated.

Rhizome scales
Rhizome as well as phyllopodia are densely covered with overlapping scales.
Shape, colour, and indument of these scales are highly variable and provide many taxonomically useful characters. More or less extensive descriptions of the scales are given by most authors. Scales were described in some detail by Nayar (1961, p. 166 fig. 1-18).

Morphology
Shape. In the P. costata-group the scales are basally attached (basifix) to the rhizome by a more or less semicircular attachment strip ( fig. 11  In most aspects they are similar to the large scales but they have only a short acumen or sometimes completely lack one. These small scales are usually completely hidden under the larger ones, but in some forms of P. lanceolata small scales are relatively more numerous, and visible without removal of the larger ones. Marginal indument. The margin of the scales may be entire or very slightly sinuose-denticulate, in other cases the margin is more distinctly dentate with fine teeth composed of the protruding cell ends of single marginal cells (as in P. eleagnifolia, P. assimilis, P. porosa p.p.) or with coarser teeth composed of the protruding cell ends of two adjacent marginal cells (as is often the case in P. africana, P. distichocarpa, pi. 2 g). These two-celled teeth are often forked at their tips.
In many species the margin is set with long or short cilia. The short cilia (pi. 2 f) may not be very distinct from long unicellular teeth (as in some specimens of P. porosa, P. drakeana), but very long and curled cilia occur in other species (e.g., P. lingua, P. abbreviata, pi. 2 h). In that case they are easily lost with age and often absent from older parts of the rhizome, but they are often distinct as a more or less dense web covering the apex of the rhizome. Most species have cilia that are intermediate between short teeth and these long, curly cilia. In all cases they are composed of a single cell with thickened walls. They are always unicellular and never obscurely uniseriate, as is often the case with the apparently unicellular cilia in other ferns (e.g., most Grammitidaceae).

25
In some species entire and ciliate scales can be found in one collection. Partly this may be due to abrasion of the marginal indument as in the case mentioned above, but in other cases differences in relative development may be involved.
This last explanation is indicated in P. foveolata (fig. 35 d): in this species the thicker parts of the rhizome tend to have scales with more cilia than thinner parts.
In peltate scales the marginal indument is usually confined to the acumen, and the base of the scales is more or less entire. Several species, however, have forms in which the base of the scale is also more or less deeply ciliate or lacerate (P. nummulariifolia, P. linearifolia, P. lanceolata, P. kinabaluensis), and the species of the P.
piloselloides-group have deeply lacerate scale bases (Ravensberg & Hennipman, 1986). In this last group, the long cilia all around the scales give a characteristic, "woolly" appearance to the rhizome. In P. sphaerosticha on the other hand, long cilia are absent from the apex and confined to the basal part of the acumen.
One or two sessile glands are often present near the apex of the scales ( fig. 11 f, j). Either one or both of these may be replaced by a long, acicular hair which is usually quite distinct from the more or less sinuose marginal cilia below the apex ( fig. 11 k). In many cases the absence or presence of these glands is difficult to ascertain, as the apex is abraded easily, and the gland itself may have collapsed, or be completely hidden by and inconspicuous among the long marginal cilia surrounding it. Probably the glands are present in all species. The acicular hairs sometimes found instead probably are structures in some way equivalent to the glands.
The apical glands or hairs are best observed on the relatively short scales that have been protected by the larger ones.
In several species glands are also present at the base of the scales. In the P.
aetata-group they can often be found on long, multicellular, apex-like extensions of the base of the scale. In P. schimperiana, P. eleagnifolia, and P. rupestris they are more or less regularly present, in number varying from 0-5, at (1972, p. 390-391) has introduced the term "isotoechous", as opposed to 'clathrate'. However, especially in the P. costatagroup a few cells may be present, mostly at the margin of the scale, in which the periclinal walls are almost hyaline and the transverse walls more prominent, so that the clathrate state is approached.
A more complicated structure can be reached in two ways: multiplication of the number of cell layers constituting the scale; or thickening of the walls of the cells.
Usually, but not always, both processes have been involved in the same scale. In a few species only the number of cell-layers has increased without an increase in thickness of the walls, resulting in scales with a more or less parenchymatous structure (P. pannosa, P. schimperiana).
The thickening of the cell walls is often confined to the abaxial side of the scale.
In all cases at least the outer periclinal walls are thickened. The adaxial side of the cells may be thickened as well but usually less strongly. With regard to the thickness of the anticlinal walls, there seem to be two different patterns of cell wall thickening: one in which the anticlinal walls are thickened as well (pi. 2 b) and one in which they are not involved (pi. 2 c).
Thickened anticlinal walls occur, among the species investigated, in the scales of P. sheareri, P. polydactyla, and P. porosa ( fig. 2 b). All of these three species have relatively dull, rather tough and flexible scales. Thickening of the outer periclinal walls only has been found in P. abbreviata, P. albicans, P. gardneri, P. angustata, and P. serpens, ( fig. 2 c). The scales of these species are relatively shiny, and, if strongly thickened (as in P. abbreviata, P. gardneri), stiff and brittle. There is thus a clear indication that the type of wall-thickening determines the texture and stiffness of the scales. The brittleness of the last type of scales may be caused by the more continuous layer of wall material that is present in these scales, contrasting with the interrupted layer of material found in scales of the first type.
The colour of the scales varies with the thickness. They are light to dark brown in most species, but mostly blackish when they are strongly thickened. Usually the area around the attachment is most strongly thickened and accordingly darkest in colour, in transmitted light it is often completely opaque. The margins, being relatively thin, are generally translucent and light brown or hyaline. In some species a large part of the acumen is also hyaline, in others the dark part around the base extends upwards far into the acumen as a wide, ill-defined, pseudocosta. 27 The material with which the cell walls are thickened is yellowish-brown or blackish in most cases, and is distinctly different from the brown material that often fills up the lumina of the cells. In a number of species (P. confluens-g roup, also

Development
The development of the rhizome scales of P. lanceolata has been illustrated by Nayar (1961, p. 166, fig. 7-15 The fronds are here considered as differentiated into stipe and lamina if a stipe is more distinct, or more constantly present. It is difficult to make a sharp distinction: in P. gardneri there is an indistinct stipe, but it is nevertheless almost constantly present, and accordingly P. gardneri is considered as stipitate. A stipe may also be indistinct or almost absent when the fronds are very small and the lamina gradually attenuated (as is often the case in, e.g., the P. lanceolata-group, the P.
confluens-group, and the P. piloselloides-group). In some cases a stipe may be distinctly demarcated from the lamina but still be very short and inconspicuous below the cordate base (e.g., the sterile fonds of strongly dimorphic forms of P.
nummulariifolia, some forms of P.
distichocarpa). In all these cases the stipitate character of the fronds may not be evident at once, but only after comparison with different forms of the same species.
In most of the stipitate species, however, a stipe is both distinct and constantly present. In length it varies from a few mm (P. distichocarpa) to 40 cm (P. abbreviata hastata, and P. polydactyla, the stipe may be distinctly longer than the lamina. 30 In cross-section the stipe is abaxially usually rounded and adaxially slightly rounded, flat, or distinctly grooved. In continuation of the decurrent lamina there may be two narrow lateral ridges at the distal end.
Colour. In most species the stipe is straw-coloured to brown when dry, but it may be blackish in P. hastata and P. polydactyla.
Anatomy. Directly below the epidermis of the stipe there is a colourless collenchymatous sheath. Distally, this is often interrupted by two narrow, lateral parenchymatous bands, in continuation of the narrow wings decurrent from the lamina. These aerating bands do not run along the whole length of the stipe.
Within the collenchymatous sheath the ground tissue is parenchymatous and only very rarely contains sclerenchymatous strands like those that occur frequently in the rhizome, but often there is a sclerenchyma-sheath enclosing, or partly enclosing, the vascular strands.
Centrally in the stipe in many species of the P. 32 fronds and a similarly sized lamina (e.g., the P. lingua-group) usually at least one lateral strand is present from the base of the stipe upwards.

Lamina
Shape. In species in which the fronds are not distinctly differentiated into a stipe and a lamina, the fronds are usually oblanceolate in shape, with an index of c. 5-20 or more. The largest width is situated in the upper half of the frond. When the fronds are differentiated into a stipe and a lamina, the latter is more variable in shape, varying from orbicular (in e.g. P. abbreviata, fig. 29 a; P. nummulariifolia) to very long and strap-shaped (P. longifolia). The largest width in these cases usually is below or about the middle of the lamina. In both stipitate and estipitate species, however, very narrow, linear fronds may be found (e.g., in P. linearifolia, P. rasamalae).
At the base of the lamina there is a transition to the ridges running downwards along the stipe. This transition is gradual in species with an attenuate lamina base as well as in species with a truncate laminabase. In the latter case the laminabase is also often distinctly unequal, and sometimes there is a difference of 2 cm or more between both sides of the lamina-base.
The shape of the apex varies widely, from rounded to acute to distinctly acuminate. In some species (P. lingua var. heteracta, P. abbreviata) this whole range can be found, but in most only part of the range is encountered.
The texture of the lamina in most species is coriaceous or thick-leathery. It may be thinner in specimens probably originating from sheltered situations. In living plants the lamina is usually more or less succulent, in extreme cases (P. 5 a, b; fig. 30, 31, 32, 34, 37). In these monomorphic species sterile fronds are usually present only in young plants or in plants growing under unfavourable circumstances. Once the plant has matured all fronds produced subsequently will be fertile. These fertile fronds are similar in shape to the earlier produced sterile ones, though often distinctly larger. The soriferous area is generally situated apically, extending downwards to a varying degree; sometimes the lamina is entirely covered with sori. Sterile fronds of these species are often absent from collections, and presumably were wanting at the time of collecting.
In principle, the formation of sterile fronds can be regulated by two possible mechanisms: either some internal mechanism may be active which is largely independent from the external circumstances, or the formation of either sterile or fertile fronds may be a more direct response to changes in the environment. Some indication for the presence of a mechanism of the second type is found in the  34 monomorphic species, which occasionally produce only sterile fronds. They often do so for long periods in the greenhouse, and it seems likely that this is due to unfavourable circumstances.
On the other hand, most of the dimorphic species seem to produce sterile and fertile fronds in an irregular alternation. Many do so also in the greenhouse, where they may grow alongside monomorphic species that continue to form fronds of only one type, be it sterile or fertile. At first sight, therefore, it seems that two different mechanisms are active. However, both apparently different growth-forms may be explained by the same mechanism if it is considered that dimorphic species tend to be fast-growing plants with longcreeping rhizomes, on which many, quickly developing, fronds are produced.
The monomorphic species are generally short-creeping, with fewer, more slowly developing fronds ( fig. 6).
If sterile or fertile fronds are produced in response to environmental changes, the dimorphic species may thus be able to react more quickly to a changing environment by the formation of sterile or fertile fronds. More drastic or longer-lasting changes are necessary to induce the "monomorphic" species to form another type of fronds. Dimorphic species therefore can adapt their mode of reproduction (vegetative or sexual) to slight or brief changes that pass by unnoticed by monomorphic species. Dimorphic species have then the double advantage of quick growth and a flexible reproductive strategy, and it is not surprising that 35 some of the most widespread species of Pyrrosia (P. lanceolata, P. piloselloides, P. nummulariifolia) have long, vining rhizomes with mostly dimorphic fronds.
Whatever mechanism is involved in the ability of plants to produce sterile or fertile fronds, only when sterile fronds occur regularly in mature plants alongside with fertile fronds a morphological dimorphism of the type encountered in Pyrrosia can develop.
As is often the case with fertile/sterile frond dimorphy in Polypodiacaea (Wagner & Wagner, 1977), the morphological differences between the two types of fronds seem to be related to specialization as either assimilating or spore-producing organs. In Pyrrosia, there is always at least a slight difference in size or shape between sterile and fertile fronds, and in some groups there is a pronounced difference. Only in P. pannosa no difference whatsoever between sterile and fertile fronds was found. This may be due to the small number of available collections, so that possibly a slight difference (as is present in, e.g., P. petiolosa) could not be discerned.
In the P. lingua-groupi (fig. 5 c-e) the sterile fronds are only slightly wider and shorter than the fertile fronds. The length/width index shows a considerable overlap in, e.g., P. lingua, P. petiolosa, but less so in, e.g., P. abbreviata. Fertile fronds of the P. linguagroup are usually fertile all over the lamina, or else the sori are arranged in an irregularly shaped patch that is not always situated apically.
Sterile fronds may have been formed in this group by suppression of the development of the sori without other changes in the morphogenesis of the fronds than those that produce the slight dimorphism.
In the P. lanceolatagroup the fertile fronds are usually distinctly longer and narrower than the sterile ones ( fig. 5 f-h). The sori are often situated apically, in a more or less contracted part of the lamina. Often there is a distinctly wider, sterile part at the base of the lamina, similar in shape and size to the sterile fronds.
Sterile fronds may thus be supposed to have developed from fertile ones through suppression of apical growth before the sori are initiated in the apical meristem.
If so, that is a marked difference between this group and the P. lingua-group, in which only soral development seems to be suppressed and the development of the lamina is only weakly affected.
A similar mechanism seems to occur in the P. angustata-group ). In P. angustata the situation prevails as described for the P. lanceolata-group and a wide sterile area is often present at the base of fertile fronds. Fronds that are fully fertile and narrow throughout also occur. In P. samarensis a sterile widened basal area is present in all fertile fronds, and the fertile area is contracted to an apical spike, similar to the apical spikes occurring in the genus Belvisia. This condition can be described as "hemidimorphic" in the sense of . In P. samarensis completely sterile fonds occur sometimes which are indistinguishable from fertile fronds before these develop an apical spike; they may be derived directly from fer-36 tile fronds through suppression of growth before the formation of sori. In P. novoguineae, the situation is at first sight different. In this species the sterile fronds are wide and relatively short, the fertile fronds are uniformly narrow and mostly longer than the sterile ones. This condition, holomorphic sensu Dickason (/. c.), is connected to the hemidimorphic condition of P. samarensis through the intermediate situation in P. angustata.
Frond dimorphism in the P. confluens-group i is more variable. Partly (in P. confluens and P. serpens) it follows the pattern of the P. lanceolata-group, partly (in P. eleagnifolia and P. rupestris) a somewhat different situation prevails. In these two species (and in P. nummulariifolia, from the P. albicans-group, as well) the fertile fronds distinctly differ in shape from the sterile ones. The fertile fronds are longer, narrower, and often erect, whereas the sterile fronds are usually more or less appressed to the substrate. However, intermediate fronds occur that are sterile but similar in shape to the fertile fronds; or similar to the sterile ones but (partly) fertile (fig. 5 j, k). This appears to occur more often in cultivated plants than in material collected in the field. Some sort of dimorphism thus seems to have developed but to a certain degree it has become independent of the actual sterile/fertile dimorphism.
Dimorphism in the P. piloselloides-group is more pronounced than in most species from other groups (Ravensberg & Hennipman, 1986). Similarly strongly dimorphic species, however, occur in the P. angustata-group (P. novo-guineae) and in the P.
lanceolata-g roup (P. fallax). This is an indication that this pronounced dimorphism may have been reached by different evolutionary pathways. The comparatively weak dimorphism in P. niphoboloides, if compared with the other species in this group, appears to be of a similar type as the dimorphism in P. angustata or the P. lanceolatagroup.
Dissection. Normally developed fronds are simple and entire in all species except P.
hastata and P. polydactyla. The fronds of these two species are pedately dissected to a varying degree, and the shape of the lamina has been compared to that of Dipteris and of Platycerium (Holttum, 1954;Jarrett, 1980, . 7). This chain can be detected by comparing the closely related species P. drakeana, P. sheareri, and P. hastata. In P. sheareri the lamina often has a number of distinct 37 lateral lobes at the base ( fig. 7 b). Into each lobe runs a secondary vein that is slightly stronger than the veins in the rest of the lamina. Once this tendency is recognized in P. sheareri, it becomes apparent that it is also, but less distinctly, present in some specimens of P. drakeana ( fig. 7 a).
The presence of these lateral lobes in P. sheareri apparently is not a fixed condition (see discussion under P. sheareri), and usually the lobes are developed slightly asymmetrically. In P. hastata (fig. 7 c) the presence of two approximately equally large lobes is fixed, and into both lobes a vein runs that is almost as strongly developed as is the costa. A small tooth may be present at the base of both lobes.  (1984) suggested to restrict the term "ontogeny" to the second process, and to use the term "primordial development" for the first. I find this use of the term "ontogeny" confusing considering the common usage of the term. I shall refer to the individual development as "ontogeny", or "ontogenetic" development, and to the second process as "heteroblastic" development. polydactyla a distinct stipe was found to be present in all the first fronds. In the other species the lamina of the first fronds was variously narrowed towards the base. With regard to the degree to which a stipe is present in the first fronds there is little difference between species with distinctly stipitate adult fronds (e.g., P. christii) and species in which the adult fronds are estipitate (e.g., P. princeps). This uniformity persists roughly through the stage in which the rhizome is not yet elongated. As the rhizome starts to elongate, in the species where the mature plant has an elongate rhizome, the fronds formed successively on the elongating rhizome gradually take on the mature morphology. Modifications of the indument have always played an important part in systematic accounts of the genus, and keys given by, e.g., Giesenhagen (1901), , Holttum (1954),  rely heavily on characters of the indument. Shing (1983) has presented a subdivision of the genus based almost exclusively on characters of the indument.
In phylogenetic discussions of the affinities of Pyrrosia the indument is also an important character: the stellate hairs have been compared to the stellate hairs occurring in Gleicheniaceae ; to the peltate scales of Pleopeltis (Lepisorus) by Copeland (1947) and Jarrett (1980) [26][27]. The rays of each crown are attached to the apical cell of the stalk without any apparent regularity of insertion. This apical stalk-cell, serving as a point of attachment for up to 20 rays, is accordingly often distinctly enlarged compared with the other stalk cells.
The stellate rays of the hairs may have become modified in the following ways: Boat-shaped rays (pi. 3 b, c, h): The rays are broadly grooved (pi. 3 b) or flat (pi. 3 c), generally less than 0.5 mm long, and are sometimes more or less distinctly striate. Usually, the rays of a single hair are all in the same plane, parallel to the surface of the lamina. Hairs with this type of stellate ray generally have short stalks (pi. 3 h), so that the indument as a whole is rather appressed.
Acicular rays (pi. 3 a, d, g): The rays are much narrower and longer than those of the boat-shaped type.
Examined by scanning electron microscopy they appear to be narrowly grooved and often somewhat spiralized. The diameter of hairs with these rays is usually larger than that of hairs with boat-shaped rays, and in many cases their stalk is much longer. Induments with this type of hairs are thus usually rather loose and shaggy, and at first sight quite different from the appressed, monomorphic ones with boat-shaped rays.
With SEM, however, it becomes apparent that there is no fundamental difference between both types of rays.
The acicular type of rays appears to be a narrow, twisted form of a basic structure, of which the boat-shaped type may be considered a wider and flatter form. All intermediates are possible and actually can be found in Pyrrosia as well as in Platycerium.
From the variability in some species it also is obvious that there is no fundamental difference between acicular and boat-shaped rays, which can be illustrated by the situation in, e.g., P. flocculosa: here in some plants appressed, boat-shaped rays prevail, whereas in others the hairs are very long and distinctly acicular. A similar variablity is found in P. kinabaluensis, and in, e.g., P. angustata all different types of rays can be found in only a small sample of hairs.
Woolly rays (pi. 3 d, e, f, g): These are very long, ribbon-shaped, and strongly and irregularly curled.
Usually these woolly rays form, closely appressed to the epidermis, a more or less dense felt in which the individual rays are strongly intertwined. That there is no fundamental difference between straight rays and woolly ones is apparent from 44 the occurrence of intermediate forms. These intermediates are more or less distinctly acicular at their base, but gradually become elongated and ribbonshaped towards the apex. Another intermediate condition is present in P. africana where the rays are mostly acicular, but may be more or less sinuose without ever becoming completely ribbon-shaped.
The modifications described above may occur in varying combinations, thus forming induments that look widely different. Two basic types of indument have generally been recognized: monomorphic and dimorphic induments.
Monomorphic induments (pi. 3 b, c). In this type of indument all the hairs have only straight stellate rays. The rays may be boat-shaped or acicular. Boatshaped rays are characteristic for many species of the P. lingua-group, for the P.
confluens-group and the P. lanceolata-group. However, this type occurs also in several other apparently unrelated species (P. schimperiana, P. sheareri, P. hastata).
Monomorphic induments composed entirely of hairs with only acicular rays occur only in P. africana, P. penangiana and in the P. piloselloides-group. This type of indument occurs only incidentally in other species, and then obviously originated through the incidental loss of hairs with woolly rays.
Dimorphic induments (pi. 3 d, f, g). In this type straight rays are always present (either acicular or boat-shaped), together with at least some hairs with woolly rays.
The woolly rays either form a dense felt, close to the epidermis of the lamina (pi. 3 g), or they occur more scattered on hairs with predominantly straight rays (pi. 3d). In the first case, the straight rays are usually appressed to the felt or are more or less raised above it, but if neither of the layers is very dense, the two layers may be mixed. Although often the straight rays appear to occur in a separate layer distinct from that of the woolly rays, closer examination shows that in most cases the two layers are not strictly separated. Most hairs have straight rays as well as woolly ones, and these two types do not occupy different "whorls", but they may be completely mixed, and not infrequently some of the uppermost rays are of the woolly type (pi. 3 e). Usually, though, the hairs closer to the epidemis have a larger proportion of woolly rays.
In the straight rays of the dimorphic indument the same differentiation is seen as in the rays of the monomorphic induments: they may be either boat-shaped (pi. 3 e) or acicular (pi. 3 d), whereas intermediate forms also occur. The combination of woolly rays and boat-shaped straight ones is less common than the combination of woolly and acicular rays.
The latter combination occurs in almost all members of the P.
albicans-group and the P. porosa-group, and in several members of other groups. The combination of boat-shaped and woolly rays is restricted to members of the P. costata-grroup, to P. lingua var. heteracta and P. laevis.

45
In many species the proportion of hairs with woolly rays varies strongly. In P. porosa and P. rhodesiana forms without woolly rays are rather frequent. In P. distichocarpa and P. drakeana similar monomorphic forms are rarer, and in P.
A very distinct modification of the hairs is found in P. laevis (pi. 3 f) and P.
splendens (and occasionally, in a less distinct form, in P. costata and P. princeps).
Here the straight rays overlying the woolly layer have developed distinct "dorsal spines", that is, one of the rays occupying a centred position in a hair-crown has become distinctly modified. It stands out erect and is conspicuously longer than the other rays in the same hair-crown. This central ray has moreover become acicular in P. laevis (in P. splendens all rays are acicular), and in both species it contrasts strongly with the other rays that lie closely appressed to the lower, woolly layer.

Development
Ontogeny. The ontogeny of the stellate hairs of Pyrrosia has been studied by Nayar (1961, p. 169 fig. 35-37, p. 171) in unspecified material. The development of the stellate hairs present in the sori in P. lanceolata (as "P. nuda") was studied by Wilson (1958, p. 486-489, fig. 32-37). I have studied the development of the hairs of P. christii (grown from spores in the Leiden Botancial Garden) and have found a similar course of development as described by Nayar ( fig. 3 v-z).
In P. christii development of a hair starts with a cell of the epidermis that swells and begins to protrude. From this cell a number of cells are split off by repeated oblique divisions, each one in a different direction. These cells then elongate and form the crown of the hair. Only after this crown is almost fully differentiatedthe uniseriate stalk develops. It is not clear whether the initial epidermis cell or the apical cell of the stalk is involved in this process.
Heteroblastic development. In the heteroblastic development the firstformed indument on the juvenile fronds is composed almost entirely of glandular trichomes. These trichomes are similar to those occurring on the midrib of the prothallia of Pyrrosia ; and in the first stages there is no distinction between the indument of the lamina and that of the prothallium. The trichomes are composed of a basal cell with two glandular, swollen cells attached to it, one apically, one subapically-laterally. On the earliest fronds these trichomes form a sparse cover on both sides of the lamina, and only on the margin of the lamina a few hairs are present that are short-stalked, branched, and have one or two elongated rays apart from a glandular cell ( fig. 11 b). These hairs are similar to the other glandular trichomes, but apically they have a few acicular rays instead of only two glands.
In later fronds, the trichomes are replaced gradually by hairs with acicular and woolly rays similar to those of the mature plant, provided, of course, that both 46 these two types are present in the mature plant. This cover is at first sparse, but as more mature fronds are formed the indument increases in density and the mature condition is approached gradually. The transformation of elongate, acicular rays to boat-shaped ones is the last to take place, and hairs with acicular rays usually persist as an upper cover over mature fronds, to be shed as the lamina expands. These acicular hairs are often slightly larger and darker in colour than the hairs of the lower strata and they can often be seen as a dense cover of the circinnate apices of developing fronds. In several species they are rather persistent, and on fully grown fronds they remain visible as a sparse cover of blackish hairs, giving the lower surface of the lamina a distinctly "punctate" appearance (often distinct in P. lingua var. heteracta). The uppermost layer of acicular hairs found in some species as a constant feature (e.g., in P. angustata, P. samarensis) may have a similar origin, as may also be true for the upper layer of hairs in P.
princeps and P. splendens. This upper layer with patent rays (P. princeps)  These scale-like filaments occur in all species of the P. sheareri-g roup, except P.
hastata and P. polydactyla; also in the P. costatagroup and in most species of the P.  (Copeland, 1947;Jarrett, 1980) to exist between the stellate crown of a hair and the blade of a peltate scale is then false.

Venation
In most species the venation of the fronds, in fresh as well as in dried state, is obscured by the thickly coriaceous or succulent texture of the lamina. At most the secondary veins are more or less evident on the lower side of the lamina. Only when the lamina is comparatively thin the venation pattern is directly evident, in all other cases some treatment for clearing the fronds is necessary before the venation pattern exhibited by the tertiary and higher order veins can be studied (see page 4). The following descriptions of the venation therefore refer to the pattern as it is visible in cleared parts of the lamina. As there is often a slight difference between the patterns of sterile and fertile parts, the description refers only to sterile parts of the lamina. The difference between the patterns in the sterile and fertile parts will be discussed in connection with the insertion of the sori. Tertiary veins. The tertiary veins connect the secondary veins at regular intervals, thus delimitating series of areoles between the costa and the margin. They are straight (pi. 5 c) or more or less arched outwards (pi. 4 a), and run perpendicular or more or less oblique to the secondary veins. The areoles thus formed vary from more or less regular rectangles to parallellograms or are more irregularly shaped. The latter is mostly the case in small or narrow fronds, in which the secondary veins are indistinct. The number of areoles between one pair of secondary veins may be as much as 10 to 20 in wide fronds (pi. 5 c). The tertiary veins are usually distinctly thicker than the higher-order veins while they are thinner than the secondary ones, but in some species the distinction is not clear. This is the case in P. africana (pi. 4 d), P. schimperiana, and in the narrow or small fronds mentioned before.
Included veins. Most of the variation in venation pattern is found in the patterns formed by the higher-order veins included in the areoles delimitated by the secondary and the tertiary ones.
In the simplest patterns the included veins are simple, straight, and excurrent (pi. 5 a, c, d). The number in each areole varies from 2 to 5 (occasionally 10).
This simple type of venation is characteristic of the P. porosa-group and occurs in many species scattered through the other groups (P. longifolia, P. gardneri, P. albicans In such patterns the included veins form a more or less anastomosing pattern of varying complexity (pi. 4 a). The remaining free veinends are still mostly excurrent, but incidentally some recurrent vein-ends are also present. This more complicated pattern is often present in large fronds of species the smaller fronds of which show a more simple pattern like that described above (this is often the case in P. lanceolata). Apart from that, it is the prevalent pattern 50 in the P.
lingua-group, in P. pannosa, P. eleagnifolia, in the larger-fronded species of the P. albicans-group, and in many species of the P. shearen-group.
The most complicated pattern occurs if the included veins branch and anastomose frequently and irregularly (pi. 4 c, 5 e). In that case a very fine mesh is formed in which the free vein-ends are no longer mainly excurrent, but point to all directions. This pattern is characteristic of the P. costata-group, but a similar pattern occurs also in some species of the P. linguagroup (P. abbreviata, P. christii, P. sphaerosticha).
Costal areole. In most patterns described above the situation is different in the areole directly adjoining the costa (the costal areole): here the included veins are predominantly recurrent, as they spring from the tertiary vein instead of from the costa (pi. 4

a).
This costal areole often cannot be distinguished in basal parts of the lamina, where it may be completely obscured by and incorporated into the thick costa. It is usually evident near the apex, where the costa is less thick than it is at the base. A costal areole with these characteristics seems to be absent from the P. africana-group, or else it is indistinguishable there in the irregular mesh formed by the veins.
"Costular areoles". In some species of the P. linguagroup (especially the three with the most complicated pattern, viz., P. abbreviata, (pi. 4 c), P. christii, P. sphaerosticha) there is a tendency for the included veins to form a row of "costular'' areoles alongside the secondary veins, analogous to the costal areoles alongside the costa. These costular areoles, in contrast to the areoles usually formed by secondary and tertiary veins, are themselves devoid of included veinlets.
Marginal veins. At the very margin of the lamina there is a distinct row of excurrent, free veins. This row is present in the P.
piloselloides-group as well, although in this group the free veins are otherwise recurrent.

Heteroblastic development
The fronds formed successively by the developing sporeling show an increasing complexity of venation pattern. In the fronds of a single sporeling not all the stages of this process are necessarily represented (Wagner, 1952), so that large numbers of early sporophytic fronds have to be studied in order to obtain an image of the entire process.
Large numbers of sporelings from evidently axenic cultures were available to me of P. princeps, P. polydactyla, and P. christii. The results obtained through study of the heteroblastic development of these species could be supplemented with the results obtained by studying smaller numbers of sporelings of the rest of the species already mentioned on p. 39 (figs. 9, 10, 12).
In all species studied the development of the very simple venation of the firstformed fronds into the complex patterns of the mature stages is a gradual one.  (Wagner, I. c.: "an individual leaf is not a morphological entity"). In the following, I shall accordingly consider the increase in complexity rather than the individual succession of the fronds.
The simplest fronds have a medial vein (costa) that is simple or pinnately branched, with few, simple, free branches. The number of branches may be 5 to 10 before the first anastomoses start to appear.
In fronds with a larger number of branches usually one or more of these are forked or anastomose to form a closed areole alongside the costa.
The number of areoles then increases until most of the branches are involved in the formation of areoles, so that the costa is lined with a row of areoles on each side (the "costal" areoles). In this stage there is usually also a row of marginal, simple excurrent veinlets inserted on these areoles (the marginal excurrent veins). These two rows persist more or less unchanged throughout the heteroblastic series, while the more complex, mature patterns appear to be "interpolated" between them as more mature fronds are formed.
As the width of the lamina gradually increases, the number of areoles increases accordingly, and successive rows of areoles are added to the costal row. The areoles are then filled with an increasing number of included veinlets, so that gradually the pattern of the mature fronds develops.
When the first included veins appear, there is usually only one per areole. The pattern is therefore necessarily a simple one and reflects the mature pattern only to a limited extent.
In those species in which the mature fronds have a venation pattern with predominantly excurrent free veins (P. angustata, P. lanceolata, P. longifolia, P. lingua, P. gardneri, fig. 12 b, P. nummulariifolia, P. polydactyla) the included veins occur first in the outer areoles (as opposed to the costal areoles). They are mostly excurrent, occasionally recurrent or anastomosing. In the costal areoles the included veins appear later and are mainly recurrent. This is similar to the situation in the mature fronds.

53
In species with a more complex venation pattern in the mature fronds, with many anastomoses and free veins pointing to all directions (P. princeps, fig. 12 (Ogura, 1972, p. 124), but in Pyrrosia they are usually slightly sunken instead of being superficial, and occasionally deeply pitted or somewhat raised above the surface, the latter especially in dried fronds. Often they are covered with a small white scale, which dissolves in dilute hydrochloric acid and is probably composed of CaC0 3 . The hydathodes are usually evenly scattered over the lamina. They are borne on the end of the free veins, but in the species with a strongly anastomosing venation pattern with relatively few free vein-endings they are also borne dorsally on the included veins. In the P. confluem-group, where free vein-endings are also relatively infrequent, the hydathodes occur on the ends of the marginal excurrent veins, so that they are restricted to a marginal row. In P. rupestris some hydathodes are occasionally present scattered over the lamina, and the marginal row is situated on the upper surface of the lamina. In P. serpens and P. confluens scattered hydathodes are completely absent and the marginal row is shifted towards the lower surface of the lamina. In P. serpens the hydathodes are located on the extreme margin of the frond, in P. confluens they occupy a position just inside the margin on the lower surface of the lamina.
Hydathodes are completely absent in the P. albicans-group i and in most species of the P. lanceolata-group. In this last group some hydathodes are occasionally present in P. lanceolata, and more regularly so in forms of P. foveolata. (see also fig. 12). Of these species, P. angustata, P. lanceolata, P. nummulariifolia, P. longifolia, and P. piloselloides lack hydathodes in the mature fronds. Nevertheless, in the early stages of the heteroblastic series hydathodes were found to be present 54 in P.
angustata, P. lanceolata and in P.
nummulariifolia. In P. lanceolata and P. angustata it could be observed that the hydathodes in a marginal position tended to persist to later stages than those situated on the lamina. Only in P. longifolia and P. piloselloides hydathodes were not found throughout the heteroblastic series.
In all other species hydathodes were present from the very first stages onwards.

Lamina anatomy
The anatomy of the lamina varies within species as well as between species.
Giesenhagen (1901, p. 65-85, fig. 12-17) gave an extensive account, and described most of the variation encountered in the genus. He stressed the importance of anatomical characters for the delimitation of the species. However, the large variability of this character, which may well be due to environmental circumstances, requires caution in applying anatomical characters for this purpose.

Upper epidermis
In surface view the cells of the upper epidermis are oriented longitudinally; they usually have more or less strongly sinuose walls. In a few species only (P. stigmosa, P. pannosa), the walls of the epidermis cells are practically straight; straight walls occur only occasionally in other species. In transverse sections the outer walls at the surface of the epidermis are either flat (pi. 6 d, f, g, h), or convex and then projecting outwards (pi. 6 a, b, c) to a variable degree. If they are strongly convex, the surface of the epidermis has a distinct papillose appearance resulting in a characteristic dull surface of the lamina. A flat epidermis gives a certain sheen to the surface of the lamina. Together with the overall thickness of the lamina, the latter influencing the "toughness" of the fronds, the shape of the epidermis cells appears to be responsible for most of the differences in texture between dried specimens. The extremes in this character are distinct, but many species have a more or less intermediate state, with only slightly projecting cells (e.g., P. serpens, P. sphaerosticha), or with both extremes occurring (e.g., P. lanceolata, P. lingua, P. foveolata). Despite the occurrence of these intermediate stages, some species similar in other characters, e.g., P. costata and P. platyphylla, or P. drakeana and P. boothii may be distinguished by the shape of the epidermis cells.
The walls of the epidermis cells are variously thickened. In many species the outer walls are particularly thickened; in others all walls are more or less equally thickened, or all walls are equally thin (the degree of thickening in the mesophyll may be used for comparison). In some cases the epidermis is separated from the mesophyll by distinctly thickened lower periclinal walls of the epidermis cells (pi. 6 d). 55 However, the variability in the degree to which the walls are thickened in many species is an indication that this character is probably easily modified by external circumstances.

Mesophyll
In most species the mesophyll is differentiated into three distinct layers. The uppermost of these layers, directly below the epidermis, is called the hypodermis (Payne & Peterson, 1973) and may be composed of one or more layers of cells (pi. 6 c-f). The size of the hypodermis cells as well as the degree to which the walls are thickened varies between species. If the hypodermis is composed of only one layer, the cells usually are rather small in transverse section and the walls are thickened in a similar way as are the walls of the epidermis. The aspect of the hypodermis in that case is similar to that of the epidermis (e.g., P. drakeana, pi. 6 c).
Often more layers of hypodermal tissue are present, and then the cells of the lower layers are usually distinctly enlarged and often have thinner walls than those of the upper layers. These large, thin-walled cells (pi. 6 f, g) probably serve as a water-storage tissue. Towards the epidermis the less specialized character of the layer may still be evident, but in many species all cells of the hypodermis are specialized as water-storage tissue, and a distinct hypodermis is at first sight absent.
In extreme cases (e.g., P. nummulariifolia, pi. 6 g) the water-tissue can take up more than half of the total thickness of the lamina. In other cases the watertissue may be thin, and hardly different from a normal hypodermis. A structurally similar hypodermis probably is the most important water-retentive structure in Platycerium (Froebe & Strank, 1981).
In several species a hypodermis is completely absent, and the palissade tissue directly adjoins the upper epidermis (e.g., in P. assimilis, P. mannii, P. pannosa).
In other species the degree to which a hypodermis is present varies. In some the presence of a hypodermis is restricted to geographically or morphologically more or less distinct forms. In P. flocculosa (pi. 6 b) a hypodermis is normally absent, but it is present in most specimens from Indo-China, which also differ in a few other characters.
porosa a hypodermis is usually absent from the type variety; it is present in var. tonkinensis and indistinctly present in var. stenophylla.
rhodesiana a layer which is indistinctly and weakly specialized as a waterstorage tissue is present only in specimens from Madagascar, and absent from specimens from continental Africa.
In some species the hypodermis is not continuous, but occurs patchily in the lamina, so that there are patches where a hypodermis is distinctly present and patches where it is absent (e.g., in P. stigmosa: Hennipman 3015; in other specimens of the same species a hypodermis is competely absent: Hennipman 3203). In other species the hypodermis shows a similar irregular occurrence (P. costata, P.

56
sheareri). Such a discontinuous hypodermis may occur in the form of incidental groups of hypodermis cells or even isolated cells scattered through the lamina (P. hastata, P. polydactyla, P. schimperiana).
Below the hypodermis, or, when a hypodermis is absent, directly below the epidermis, a layer of palissade parenchyma is present in most species, and/or a third layer, the spongy parenchyma.
The degree to which the mesophyll is differentiated, and thus the sharpness of the distinction between hypodermis (including water-tissue) or palissade-and spongy parenchyma cells, varies strongly. Especially in the more strongly succulent forms the mesophyll is often only weakly differentiated, and all cells are more or less similar to the cells of the water-tissue in other species.
In several species (e.g., P. flocculosa, P. penangiana, pi. 6 a, P. princeps) the degree to which a distinct palissade tissue is developed varies strongly within the species. External circumstances may play a role in this, but it is also possible that differentiation of the mesophyll is the last character for which the mature situation is reached in the heteroblastic development. In that case, differences may represent only juvenile aspects of otherwise apparently fully mature fronds (compare also p. 39).
If infraspecific variation is disregarded as much as possible, the least differentiated fronds are those with a relatively thin lamina (P. laevis, P. christii, P. penangiana), and those with a thick strongly succulent lamina in which most of the cells probably have a function as water-storage cells rather than as specific palissade or water-storage cells (e.g., P. longifolia, P. serpens). This is only a rough correlation, to which there are several exceptions (e.g., P. costata, with a thin lamina but a well-differentiated mesophyll; P. nummulariifolia, with a succulent lamina but also a well-differentiated mesophyll).
The structure of the mesophyll in the P. piloselloidesgroup is aberrant (pi. 6 h) in being completely undifferentiated. All the cells are large, thin-walled, and parenchymatous in character. This is either a separate, distinct development, or this state represents the final stage of the trend towards undifferentiated, succulent fronds as observed in other species.

Lower epidermis
The structure of the lower epidermis is rather similar to that of the upper one.
In surface view the cell walls are more often straight, and in transverse section they are often more strongly thickened. In extreme cases more than half of the entire height of the cell may be taken up by the thickened wall (e.g., P. albicans, P. princeps, P. angustata). The lower epidermis usually directly adjoins the spongy parenchyma layer. Only in P. confluens a layer of hypodermis-like cells is present between spongy parenchyma and lower epidermis; this layer is interrupted at the stomata.

Stomata
In mature fronds the stomata are restricted to the lower surface. Configuration of guard-cells and the ontogeny of the stomata has been studied by Giesenhagen (1901, 81-85) andSen &Hennipman (1980). The terminology used by Sen & Hennipman will be adopted here.
The occurrence of perior desmocytic stomata does not correlate with other characters. Polocytic stomata occur only in the P. africana-group and in P. mannii and P.
penangiana. The configuration of the guard cells in the groups of P.
lanceolata and P. confluens could not be investigated in all species as the stomata in these species are deeply sunken. Incidental observations, however, confirm the pericytic character of the stomata in these species.
In transverse view all intermediate stages can be seen between stomata that are slightly raised above the surface of the lamina and deeply sunken stomata.
Stomata of the first type occur only in P. sheareri and P. drakeana (pi. 6 c). Stomata that are flush with the surface of the epidermis or slightly sunken below it (the difference is often not clear owing to the protrusion of the epidermis-cells, compare pi. 6 b, h) occur in most other species. Deeply sunken stomata (pi. 6 e, f, g) are common in the P. albicans-, the P. lingua-, the P. confluens-and the P.
lanceolata-group. They occur also in some forms of P. subfurfuracea. Often the stoma-groove is narrowed above the stoma. In most species the narrowest part is situated approximately halfway the depth of the stoma-groove, but in some species the narrowest part is often at the surface of the epidermis (e.g., in forms of P. lanceolata and in P. confluens). In some species the insertion of the sori is shifted downwards along the included veins until they are situated on the junction of tertiary veins and included veins. This is accompanied by a "break-down" of the regular venation pattern in P.
distichocarpa: the distinction between tertiary and included veinlets becomes less clear, and the areoles become irregular in shape.
Usually the sori are superficial on the lamina, or slightly impressed in mostly ill-defined and shallow depressions (e.g., P. schimperiana). More distincdy sunken sori occur in species of the P. confluens-group (particularly P. serpens) and in P. angustata. Sori that are deeply sunken in distinct, steep-sided pits are characteristic of the P. lanceolata-g roup.
The number of sori in each areole is highly variable. This seems to depend mainly on the size of the sori relative to the size of the lamina, so that it depends strongly on the size of the sori as well as on the size of the lamina. Mettenius (1856) used the number of sori in each areole as the main criterion for his subdivision of the genus; the distinction is indeed a useful one, even though in many species this number is not constant.
The very small sori of the P. africanagroup, the P. costata-group, of P. pannosa, and of many species of the P. linguaand P. sheareri-g roup occur in large numbers in one areole (pi. 4 d, 5 e) and are irregularly scattered over most of it. In some of these species there is a distinct tendency towards the contraction of the sori into two rows in each areole (e.g., P. lingua var. lingua, P. christii) or sometimes even one row ( P. lingua var. heteracta). In other species the sori are constantly situated in one row in each areole (pi. 5 c), although sometimes a second row is weakly and irregularly present (e.g., P. flocculosa).
In the species with one row of sori in each areole the number of sori in each row varies from c. 10 (P. longifolia) to one (P. angustata, P. serpens). The number of 2 seems to be fixed in many species, although incidentally a third sorus may be present. In some cases this low number of sori in each areole is correlated with enlargement of the sori (P. foveolata, P. distichocarpa, P. asterosora), in others the 59 reduction in lamina size seems to have been the most important factor (P. assimilis). The single sorus per areole that is characteristic for P. angustata, pi. 4 b, P. serpens, and P. confluens is probably derived from a situation with two sori per areole through lateral contraction of the lamina and concomitant merging of the sori. This can be concluded if an aberrant specimen of P. angustata, in which 2 or 3 sori per areole are present ( Wright & Ismawi 32884), is interpreted as atavistic. Lateral contraction and fusion of the sori is also evident in P. serpens, in which species the intermediate condition of two sori per areole is rather frequent. The reduction of the number of sori to a single one in each areole is always accompanied by a reduction in the number of soriferous areoles between costa and margin, so that the sori are ultimately reduced to a single row on each side of the costa.
In a number of species the separate sori are replaced by a longitudinal sorus on each side of and parallel to the costa. This coenosorus is situated close to the margin or midway between costa and margin. The presence of a coenosorus is often accompanied by a strong reduction in width of the soriferous part of the lamina. In some species (e.g., P. heterophylla, P. fallax, P. novo-guineae) this has resulted in a strong dimorphism, in other cases the coenosorus is borne on an apical spike of the lamina (P. samarensis).
In some cases it is evident that the presence of a coenosorus is directly derived from a situation with a single row of slightly elongate sori. In P. samarensis forms occur with an interrupted coenosorus (e.g., Ramos 946), which can be considered as transitional to the single row of sori as present in P. angustata. In P. confluens forms with a coenosorus and with separate sori both occur, and the conclusion seems obvious that here, too, the coenosorus originated through a similar process of fusion.
In the P.
lanceolatas-g roup the situation seems to be different. Here, intermediates between the small, circular sori of P. lanceolata and the coenosorus of P. fallax are absent. In P. lanceolata there is no tendency towards fusion of the sori.
Thus the coenosorus of P. fallax, which is structurally similar to the sori of P.
lanceolata, must be supposed to have originated by a more sudden transition from a circular to an elongate structure.

Size and shape
The smallest sori occur in the P. costala-group. Measured when ripe and with all sporangia present the sori in this group are c. 0.5 mm in diameter, and round in outline. In most other species the sori are more variable in size and shape. The diameter varies there from 1-2.5 mm, but distinctly larger sori occur in the larger species of the P. albicans-group ((especially in P. distichocarpa), in the P. angustatagroup, and in the P. confluens-group. In these last two groups the large sori are often elongated longitudinally and may attain sizes of 3 x 4 mm. The size of the 60 receptacle varies accordingly from small and insignificant to large and often strongly vaulted (in P. distichocarpa and P. confluens). In species with small sori the receptacle often spreads along the veins, so that the sori are irregularly elongated and often somewhat confluent (e.g., in P. lingua, strongly so in P. laevis, P. christii, P. petiolosa, pi. 5 b). Small, closely packed sori may also appear to become confluent when they ripen and the sporangia increase in size while the sporangial stalks elongate. The spreading receptacles and the enlarged sori together may make ripe, fertile fronds appear completely acrostichoid (particularly in P. abbreviata, P. christii, and P. sphaerosticha).
The occurrence of larger, elongated sori in the P. confluens-group and in P.
angustata is also associated with confluence of the sori (leading, as shown above, to a coenosorus), but in this case the confluence is not along the soriferous veins, but generally across the secondary veins separating the areoles.

Innervation
Small sori, situated terminally or dorsally on. the veins, are each innervated by a simple vein. In larger sori such a vein may be forked in the receptacle (P. distichocarpa, P. confluens). The very large sori of P. angustata have apparently spread over a large part of the areole, and thus are innervated by several veins, which give off small branches that form a complicated network in the receptacle.
A similar network of veins is present in the receptacle of the coenosorus of P. samarensis.

Paraphyses
When young, the sori of all species are covered with hairs similar to those covering the sterile lamina. The stalks of these hairs are slender and fragile, and it is difficult to ascertain whether the hairs are actually inserted on the receptacle or not, but in some cases they obviously are. According to common usage, these hairs should be called "paraphyses" (Wagner, 1964). The generality of that concept, however, makes it practically useless (Tryon, R., 1965;Tryon, A., 1965), and at least in Pyrrosia I would prefer to restrict the term "paraphyses" to trichomes on the receptacle that are homologous to but also at least slightly modified from the indument of the sterile lamina. Baayen & Hennipman, however, in a survey of paraphyses in the Polypodiaceae (in prep.), will use the term conform to Wagner (/. c.), and this usage will, for the sake of conformity, be adopted here. Paraphyses often differ from the hairs on the sterile lamina in the following aspects: -the stalk is longer, -the stellate rays are fewer in number and they are often slenderer, -the density of hairs (paraphyses) on the receptacle is higher than that on the lamina.
Paraphyses modified in one or more of these ways occur in the P. albicansgroup. , the P. angustata-gr oup, the P. confluens-group, and the P. lanceolata-group.
Their presence is not always evident, and they may be constantly absent in P.
In the P. albicans-group the paraphyses, where present, have stellate rays that are less strongly curled than the woolly rays of the sterile indument, but more so than the straight rays. In P. rasamalae small tufts of woolly hairs are often present loosely appressed to the sori, which may be paraphyses or alternatively, they may be "sterile" woolly hairs pushed away by the developing sporangia. Due to the fragile character of the stalks it is not possible to ascertain which is the case. If the hairs forming these tufts are true paraphyses, they form an intermediate stage between the distinctly modified paraphyses in, e.g., P. albicans and the unmodified hairs in the sori of P. nummulariifolia.
In the P. confluens-and P. lanceolatagroup the paraphyses have stellate rays that are shorter and wider than the rays of the sterile hairs. Especially in the P.
confluens-group paraphyses may be very numerous: only a few sporangia are sometimes present in each sorus, interspersed with large masses of paraphyses.
In this group it is also evident that the paraphyses are distinct from the "sterile" hairs. Young sori have a thin cover of normal hairs, as the sorus develops these hairs are pushed away by the distinctly different paraphyses.
In the P. lanceolatagroup the paraphyses are situated in a central bundle in the sorus.
The sporangia are situated around this bundle in a distinct rosette-like arrangement.
In P. fallax the coenosorus has a central row of paraphyses. This probably equivalent to the central bundle of paraphyses in other species of this group (Ravensberg & Hennipman, 1986).

Sporangia
Structurally the sporangia are of the common polypodiaceous type as recognized by Wilson (1959). There is some, probably minor, variation in the configuration of the cells on the faces of the capsule of the sporangia. The shape of the capsule varies from more or less spherical in most species to distinctly elongated or ellipsoid in, e.g., P. rupestris or more or less obovate, especially in the P. costata-group.
Most of the variation concerns the relation between stalk length and capsule height, and the number of indurated cells of the annulus (the "bow", Wilson, I. c.).
Capsule height. The height of the capsule varies from 0.3-0.4 mm, but distinctly smaller as well as distinctly larger sporangia (to 0.6 mm) occur scattered through the genus. There is little taxonomic significance in this character: none of the species-groups recognized can be characterized by sporangia of a certain size except the P.
africanagroup, of which both species have relatively small sporangia.
62 Stalk length. The most striking variation in the sporangia is present in the length of the stalk relative to the capsule. Sessile or very shortly stalked sporangia (stalk at most half as long as the capsule) occur in the P. costata-group and in some species of the P. sheareri-group (P. drakeana, P. sheareri, P. hastata), and in a few other species (P. mannii, P. porosa var. stenophylla). In most species the stalk is 1 to 1 V2 x as long as the capsule, with lengths of c. 1 x the capsule restricted mainly to the P. sheareri-, the P. africana-and the P. porosa-group.
The longest sporangium-stalks (2 V2 to 3 x the height of the capsule) occur in P. albicans, in P. angustata and P. samarensis, and in P. rupestris.
Modifications. In Pyrrosia the only modifications of the polypodiaceous type of sporangium are those found in the P. costatagroup and in P. schimperiana. The sporangia of the P. costata-group are different in that the number of indurated annulus-cells is distinctly lower than in most other species; typically hardly more than 10 cells are indurated. This is not a consequence of the small sporangia in this group (as it is in, e.g., P. gardneri), but a result of the reduction of the induration of the annulus near the attachment of the sporangia. In the most extreme cases (present in P. princeps, P. platyphylla, and P. splendens ) there is only a small apical part of the annulus that is indurated and the basal, non-indurated part is hardly recognizable as part of the annulus. are not torn apart completely as in the other species, but the basal part remains intact and forms a hollow cup, whereas the apical part is lifted off and turned outwards. In these species there are, in each sorus, only c. 10 sporangia, arranged with the mouths facing towards the centre of the sorus. The ripe sori appear like distinct, small craters, the bottom being formed by the entire sporangium-bases, the sides by the apical parts that are turned outwards. This arrangement may be functional in that it effectively pushes away the dense mat of hairs overlying the young sori, thus completely exposing the spores. In herbarium specimens it can indeed be observed that old fronds often have hardly any spores left trapped in the tomentum, in contrast to what is found in many other species.
Another modification of the sporangial structure is found in P. schimperiana.
Here the one or two annulus cells closest to the stomium are distinctly darker in colour and often swollen, and sometimes they are divided by a longitudinal wall.
Arrangement. Most species have an indefinite number of sporangia in each sorus, which do not ripen or shed their spores with any apparent, regular order.
In some groups, however, there is some regularity in the number or arrangement of the sporangia.
The peculiar arrangement of the sporangia in the sori in the P. costata-g roup is already discussed above.
A similar arrangement, though probably of independent origin, is found in several species of the P. lanceolata-g roup. Here the sporangia are arranged around a central bundle of paraphyses. In contrast to the situation in the P. costata-group, the number of sporangia is indefinite and rather large. The sporangia ripen in a centripetal order; first the outermost ones grow out and shed their spores, later the more centrally situated ones develop. In this way the sorus gradually forms a distinct rosette of ripe, empty sporangia around a central tuft of paraphyses.
This central tuft may sometimes be hidden completely among the sporangia, as gradually more sporangia grow out and ripen.
Development of the sori. The sori are initiated at the growing apices of the fronds, and usually ripen at a constant rate. As a result, on a developing frond a succession can be found from very young sori at the apex to ripe sori closer to the base. The same succession is often found within a single elongated sorus: the young sporangia are present near the apex, the older ones near the base. In some species however (most notably in some species of the P. lingua-group) the sori seem to ripen simultaneously at apex and basis of the lamina. Development of the sori here is apparendy arrested for some time after the initiation, until the lamina has developed fully. Thus most sporangia on a frond are at the same stage of development and all spores are shed simultaneously. The occurrence of this process, however, needs experimental confirmation, as the exact order in which the sporangia ripen often can not be ascertained in the herbarium.
Spore output. Usually 64 spores are formed in each sporangium. Deviating numbers are extremely rare and were found only in two cases during the present study. Once an aberrant sporangium of P. hastata {Ito 67) with an unusual number of indurated annulus cells (ca. 30) was found, which contained double the normal number of spores, but other sporangia from the same collection appeared to be normal. In the other case, in a sample of spores of P. christii collected by Geesink in 1982, several sporangia were found that contained 32 spores each, the spores all slightly larger than normal. In a larger sample of already shed spores the proportion of these larger spores was found to be less than 1 %. The sporangia producing these spores were not aberrant in any way except that the stomial region was relatively undifferentiated.

Gametophytes
Gametophytes of Pyrrosia have been studied by several authors, i.a. Nayar (1957;1961)  According to Nayar & Kaur (/. c. ) the mature gametophyte of the Platycerioidae is uniformly cordate-thalloid, with a thin median midrib. The margin is beset with unicellular, occasionally multicellular glands ; on the surface of the gametophyte multicellular glandular trichomes are present, usually on the lower, but sometimes on the upper surface as well.
Development of the gametophyte is also reported to be uniform in the Platycerioidae. After germination of the spore a short "germ filament" is formed; subsequently an apical meristem is established and the cordate thallus begins to develop. Both antheridia and archegonia are formed on the same gametophyte (Nayar & Kaur, 1971).
In the course of the present study only a few incidental observations were made on the gametophytes. From these the following results were obtained: In at least three species (P. princeps, P. gardneri, P. nummulariifolia) the gametophytes appear to pass through a phase during which they are elongated and relatively narrow, before they attain a cordate shape. It should be stressed that a complete continuity exists between the indument of the gametophyte and of the juvenile sporophyte. The variety of different kinds of hair found on the gametophyte matches that of the first sporophytic fronds, up to the (identical) occurrence of acicular, elongated cells. This continuity probably has implications for the assessment of the homology between various types of dermal appendages in the polypodiaceous ferns; these considerations, however, lie outside the scope of this study.

Karyology
Knowledge of the cytology and the chromosome numbers in Pyrrosia is incomplete. Several reasons can be adduced to explain this, most of which are applicable to tropical ferns in general. One reason, however, is peculiar to Pyrrosia, viz., that the indument of the lamina effectively protects the young sori. Meiotic chromosome numbers, obtained from developing sporangia, are therefore difficult to establish. Mitotic counts, taken from developing roottips, are easier to obtain, but on account of the double number of chromosomes that has to be counted they are less reliable. In the literature mitotic counts appear more frequently.

Results
In  The exact distribution of either number, however, is far from certain. This is due both to the incompleteness of the survey and to the uncertainty of some of the counts. In many of my own counts it was not possible to decide with certainty between counts of 72 and 74; in other published counts some uncertainty is similarly evident. For instance, in P. linearifolia a meiotic number of 37 is given by Mitui (1966), a mitotic number of 72 by Takei (1969). As both counts were taken from plants originating from Japan, this inconsistency is not likely to be due to the heterogeneity that is shown by this species between plants from Japan and from Taiwan. Most likely one of the counts is erroneous. In another case, a number of 2n = 72 is given by Takei (1972)  Polyploidy. More reliable than the determination of base numbers is the identification of polyploid series. There is no necessity to count the number of chromosomes with an accuracy of 1 (nor is it possible with the large numbers involved), but a reliable estimate of the level of polyploidy can be made relatively easily.
Polyploid series occur in Pyrrosia in at least three not closely related species groups.
P. porosa-group. Within  heterophylla from Madagascar is tetraploid. The number found, 2n = + 149, indicates (though weakly) a base number of 37 rather than 36. This suggests an autopolyploid origin with P. piloselloides as the possible parent species.
In contrast with the two polyploid complexes considered above, the species are morphologically distinct, and geographically separated.
The occurrence in Madagascar of a polyploid species with its closest possible parent species in Southeast Asia is curious. Polyploidy is thought to be associated, in a general way, with the occupation of new habitats, and this is contrary to the commonly held view that the flora of Madagascar is mainly composed of ancient relict species.
Assuming that P. niphoboloides originated by way of autopolyploidy from P.
piloselloides, it is not obvious how it could have reached Madagascar. There is an unresolved conflict here between the probable phylogeny as reconstructed by cytological evidence and probabilities based on geographical distribution.

Recognition of hybridization
One of the problems encountered in the study of herbarium material is that it is difficult to assess whether hybridization has occurred in the plants studied.
Some criteria that can be used to recognize hybridization in herbarium specimens are given by Hennipman (1977) and Wagner (1983). Hennipman (/. c.) gives four criteria, but points out that not all four criteria are equally useful and reliable, so that in his opinion each suspected case of hybridity should be confirmed by experimental work, or at least by direct observation of the meiotic 69 behaviour of the chromosomes. Wagner's (/. c.) main criterion is based on intermediacy in morphological characters; other criteria are used as checks.
The role of hybridization has been thoroughly analyzed in a few groups, mainly from temperate regions (e.g., Asplenium and Dryopteris). From these investigations it has become clear that a thorough knowledge of the cytological aspects is a prerequisite for any analysis of the sometimes subtle and intricate patterns of morphological variation displayed in these "problem"-groups. In the absence of such detailed knowledge, any assessment of the role of hybridization is bound to remain speculative.
Morphological evidence. The morphological criteria, that according to Hennipman (1977) and Wagner (1983) can be used to detect hybridity, can be regarded as morphological expressions of a hybrid genome, through disturbance of the normal course of morphogenesis.
The first of these criteria, in Hennipman's (/. c.) enumeration, is the one of morphological instability. By this criterion, extreme morphological variation in a species is interpreted as an indication of the presence of two different genomes.
This extreme variability, however, is in .practice difficult to distinguish from variation due to plasticity under different circumstances. A wide range of variation is expected in most wide-ranging species growing in a variety of different environments, irrespective of any possible hybrid origin of these species. Study of large samples of well-annotated plants, combined with experimental work, is necessary to distinguish between the two possible causes of variability.
In Pyrrosia there are several instances of such wide-ranging species with a large range of morphological variation, e.g., P. lanceolata, P. confluens, P. lingua, and P. porosa. All these species may be suspected to contain "hidden" hybrid complexes but only in the case of P. lanceolata and P. porosa are there other reasons (see p. 74) for assuming that this is actually the case.
A second morphological criterion is the one of morphological intermediacy. According to this criterion, a hybrid is generally expected to be morphologically intermediate between its parents. This intermediacy, however, as Hennipman (1977) points out, cannot be expected to take on a predictable form, and, if occurring in an unpredictable form, is difficult to recognize as an intermediate state.
In morphologically rather simple plants like Pyrrosia, moreover, the assessment of intermediacy is particularly difficult, as many characters can be seen as intermediate between others.
The vagueness of the criterion may be obviated by considering as possible parent species only those that are morphologically similar; or those that grow in the immediate neighbourhood of the plant under consideration. On the other hand, the number of hybrids that can be detected is directly related to the number of parent species that are compared with it. Any restriction of the latter will diminish the general usefulness of the comparison.

70
The last criterion is the one of "structural irregularity", first formulated by Wagner (1962). This criterion is relatively easy to observe, but it is not met with in all hybrids, nor is it with certainty absent in normal plants. In Pyrrosia "structural irregularity" is possibly present in the form of asymmetrical fronds in the P. sheareri-group; perhaps the occasional laciniate fronds of P. mannii are another instance.
Spore sterility. The presence of large numbers of aborted or otherwise abnormal spores is one of the criteria that can be used to detect hybrids in herbarium material. Spore sterility is a direct reflection of the incompatibility of two different genomes if these are combined in a hybrid plant. Nevertheless, the criterion has no absolute value. Aborted spores may be present in small amounts even if no hybridity is involved (Wagner & Lim Chen, 1965); spore abortion may also be caused by disturbance of the normal process of spore development by external factors. On the other hand, several mechanisms (autopolyploidization, various forms of apospory, cf. Walker, 1979) may result in the formation of normal spores in hybrid plants. In Pyrrosia I have investigated the occurrence of spore sterility in two ways.
by examining samples of shed spores, whenever possible several samples for each species, -by examining the contents of one or more, apparently well-developed, sporangia, whenever possible also of several specimens of each species.
A summary of the results is given in the following list, in which each species is accounted for that had a large number of abnormal spores in at least one sample. From this list it is apparent that the occurrence of sometimes large numbers of abnormal spores is by no means restricted to those taxa that may, on account of other criteria, be suspected of hybrid origin. Several species that are apparently normal if evaluated by morphological criteria (e.g. P. asterosora, P. distichocarpa, P. longifolia, P. penangiana) turn out to have a fairly high incidence of abnormal spores. On the other hand, species with an irregular morphology, which may accordingly be suspected of hybridity (e.g. P. sheareri, P. drakeana) also contain specimens with mostly normal spores. This means that either hybridization is a much more common phenomenon in Pyrrosia than is apparent from the morphology of the specimens examined, or that spore sterility is only a poor indication of the occurrence of hybridization. In view of what is known about spore formation and the various ways in which that process may be interrupted by changes in the external circumstances, I suspect that the latter of these two possibilities is more probable, and that any correlation between the phenomena of spore sterility and hybridity is weak at best.

Possible cases of hybridization
If all these criteria are considered, the possibility of hybridization in Pyrrosia appears to be restricted to a few groups.
One of these groups is the P. sheareri-group. Almost all species in this group are variable in several aspects, though not strikingly more so than many species in other groups. This variability may be interpreted as morphological instability.
The species may be arranged in a series, in which each is morphologically intermediate between the preceding and the following one, as follows: P.
Moreover, in many of these species the lamina is markedly asymmetrical at the base (see p. 37). The degree of asymmetry in itself is not constant, and there are indications that it may be influenced by external factors.
Examination of the spores of these species, however, does not give any indication that hybridization in this group is more frequent than in other species- Independent evidence for the occurrence of hybridization is found in the occurrence of aberrant chromosome numbers in P. lanceolata and P.
porosa. Spore abortion as a more common phenomenon than in other groups of Pyrrosia, however, is found only in the P. porosagroup. Within this last group p.
porosa var. stenophylla may more specifically be suspected of hybrid origin: the irregular venation-pattern and position of the sori most characteristic of this variety can be interpreted as structural irregularity.
Hybrid origin is also a possible explanation for the large variability shown in the few available specimens of P. kinabaluensis. Morphologically this species is intermediate between P. rasamalae and P. nummulariifolia, but of these two possible parent species only P. nummulariifolia occurs throughout the area occupied by P. kinabaluensis, whereas P. rasamalae is restricted to a relatively small part of it. The spores of P. kinabaluensis do not seem to be significantly abnormal.
Relatively few hybrids in Pyrrosia have been reported in the literature. Pyrrosia x pseudopolydactylis Serizawa (P. matsudae x P. polydactyla) has been reported from Taiwan (Serizawa, 1970). It appears to be highly similar to P. polydactyla, differing from that supposed parent species only in the somewhat lesser number of lamina lobes; also the spores are reported to be irregular. In my opinion neither the slightly aberrant frond shape (similar fronds occur not infrequently in P. polydactyla together with completely normally developed fronds); nor the irregular spores (which may have been aborted due to external circumstances) are sufficient evidence for a hybridogenous origin of this plant. On the other hand, the possibility that these specimens are aberrant forms of P. hastata or P. polydactyla, grown under unfavourable circumstances, cannot be ruled out. As far as could be investigated, the spores of these specimens were normal and well-developed (see also Van Uffelen & Hennipman, 1985). 76 7. PHYLOGENY A hypothesis concerning the genealogical relations within the genus Pyrrosia is presented in the form of a cladogram (Fig. 13). In this figure, the cladogram is presented as a 'synapomorphy scheme': a branching scheme superimposed on a species/character matrix that has been ordered according to certain principles.
In theory, a taxon/character matrix can be ordered according to many different principles, of which the hierarchic principle generally used in taxonomy is only one (Ball, 1983). Also, any ordering of a group of taxa may be interpreted as some kind of a hypothesis concerning these taxa. The advantage of the particular type of ordering used in taxonomy, a hierarchical ordering, is that it is easily converted into a genealogical hypothesis. To do so, it is only necessary to accept the basic premise of evolutionary theory, viz., that taxa have come about by descent with modification. A hierarchical system represents genealogy in that each group in the hierarchy on any level represents a group of taxa descended from a common most recent ancestor. It has been argued not only that the initial assumption of evolution is a sufficient reason to interprete cladograms as genealogical hypotheses, but also that it is the only possible reason to prefer a hierarchical system over other systems expressing relationships (Beatty, 1982;Ball, 1983;Hull, 1983).

Methodology
Once the desirability of a hierarchical system has been established, a method must be chosen by which this hierarchy is to be constructed. In how far a genealogical interpretation of the hierarchic system is useful as a biological hypothesis depends on the following two factors: -In how far the method used for the construction of the hierarchy can be considered to reflect the relevant evolutionary processes; -The explanatory power of the ordering achieved.
The two factors are, of course, not fully independent of each other. If the explanatory power of a certain ordering (genealogical hypothesis) is great, that in itself may be taken as an indication that the ordering principles in some way reflect the processes in nature. On the other hand, if there are strong reasons to suppose that certain processes have been active, this may be taken as an indication that the hypothesis produced by using principles reflecting those processes 77 will be a useful one. For the construction of hierarchical systems in biology it is desirable to choose a method that reflects as much as possible the course of historical events during speciation processes.
Such a method has been developed by Wagner (1961Wagner ( , 1969 and Hennig (1966); it has been used since without important modifications. This method is derived by applying the following considerations: 1) that during the course of evolution new species are formed out of older ones; 2) that any new species must have at least one new character to distinguish it from the parental species (this is not to say that species can always be "based on" a single character; the process of species recognition is a different one from that of phylogeny reconstruction); 3) that this newly acquired character persists unchanged or is modified uniformly throughout the species (anagenesis) until this uniformity is disrupted because the character is modified again, either in the process of the formation of a new daughter-species (cladogenesis) or afterwards, in one of the daughter-species.
Thus is the principle derived that only the shared possession of newly acquired (apomorphic) characters can be used as an argument for the sharing of a common most recent ancestor. The method consists of the search for a hierarchical pattern in the distribution of these apomorphic characters. The sharing between two taxa of an apomorphic character then is a direct reflection of the sharing of the ancestor in which that character evolved for the first time.
The recognition of apomorphic from plesiomorphic characters is thus an important element of the method.
There are two ways in which it is possible to approach this problem. The first, traditional, approach (advocated by Wagner, 1961, 1969, andapplied, e.g., to the genus Platycerium by Hoshizaki, 1972, andHennipman andRoos, 1982) is that the a priori recognition of transformation series is a prerequisite for the construction of a cladogram. The second possible approach (advocated by, e.g., Gaffney, 1979) is that transformation series, and thus the relative apomorphy of each character in such a series, are to be reconstructed in some way from the cladogram. This latter approach is here rejected, because of inconsistencies that are inherent to it. A more detailed comparison of both approaches to phylogeny reconstruction will be published elsewhere (Hovenkamp, in prep.) For the purpose of analysing the consequences of either approach, phylogeny reconstruction can be conceived of as a process composed of three independent steps.
The first step is the construction of a data matrix in which the observations on the organisms are incorporated in a suitably coded form.
The second step is the construction of a network that connects the taxa in such a way that the total number of character changes along the segments of the netwerk is minimized (Wagner-network, Farris, 1970;Jensen, 1980). The third step, relevant only if the second approach is adopted, is the evaluation of this network as a 'synapomorphy scheme': for each group that is specified by the network it must be evaluated whether there are apomorphic characters that can be used as arguments for the monophyly of the group.
This step has two different aspects: the assignment of characters to the nodes of the network, from which assignments the patterns of character change (character transformation series) can then be read off; and the establishing of a 'root' to the network, by which the direction of character change can be found. These two aspects are independent of each other.
The root of a network can be found by applying some kind of 'out-group analysis' (Lundberg, 1972;Farris, 1982). The network is then compared to a supposed sister-group of the group that is being analyzed, and the segment of the network is found where the sister-group can best be connected to it. This method relies heavily on a hypothesis about the phylogeny of a higher order; and, in principle, leads to an infinite regress (Colless, 1967(Colless, , 1969 (Farris, 1970;Mickevich, 1982) to eliminate these ambiguities, but not all possible ambiguities can be eliminated by these methods, nor is it clear whether these methods do not make some arbitrary choices among the several possible solutions of the ambiguities.
The inclusion of this last, third, step in the reconstruction is, as already indicated above, necessary only if the transformation series are to be 'read off from the data. If transformation series are specified in advance they can be incorporated in the first step. This "reading off' afterwards appears to be not always possible; from there it follows that we also have to reconsider the problems inherent in the first of the three steps.
In order to construct the network (step two), it is necessary to code the data (step one). This can be done either in such a way that no transformation series are implied ('neutral coding'), and above it has been argued that then the correct transformation series can not always be reconstructed. The only alternative is to code the data in a way that already incorporates transformation series ('additive 79 coding'). Thus, the choice seems to be one between adoption of 'a priori' assumptions and unavoidable loss of information.
Moreover, even a method that tries to code the data in a way as 'neutral' as possible has to make assumptions about which two characters (or 'characterstates' as they are usually called) should be connected in a one-step transformation. It seems inconsequent to accept the ability to perceive a connection between two characters, and at the same time, to deny oneself the ability to perceive connections between more characters.
Another problem that follows directly from the avoidance of a priori assumptions is inherent in the second step, the construction of the network from the datamatrix. In principle, the problem of finding the minimum length network connecting the taxa is NP-complete (Felsenstein, 1982). This means that an efficient algorithm to find that network probably can never be constructed. The search for the minimum-length cladogram then must be made by comparing all possible cladograms; which is practically impossible if more than a few taxa are involved. The restriction that results from the adoption of a priori transformation series can reduce the number of cladograms that have to be compared.
Summarizing, I would like to stress that the search for an objective method of cladogram construction, avoiding a priori transformation series, has a number of disadvantages: 1) An arbitrariness in the choice of the cladogram from among all possible cladograms, a choice that is not explicit in most algorithms (exceptions are all fully implemented 'exhaustive search' algorithms).
2) The reliance, for the rooting of a cladogram, on a higher level phylogeny.
Apart from leading to infinite regress, this will usually involve comparisons with groups less intensively studied than the group for which the cladogram is being constructed, and consequently, may weaken the conclusions rather than strengthen them.
3) The assumption that transformation series can be reconstructed from the data in an unambiguous way is unwarranted. Any results arrived at may in fact be due to arbitrary decisions that are not explicitly accounted for.

Construction of the cladogram
In order to avoid the problems indicated above I have chosen for a method in which several choices are made a priori. I consider a limited number of explicit though subjective decisions preferable over an unknown number of hidden, arbitrary choices. In this way, further discussion is made possible of these decisions and their consequences; in the other option neither the assumptions nor their consequences for the structure of the cladogram are clear.

80
The a priori establishment of transformation series can in my opinion best be considered as the formation of hypotheses; a process that, according to many philosophers, defies rational analysis anyway. The procedure outlined here can relatively easily be fitted into a hypothetico-deductive model of scientific reasoning (Popper, 1959;compare also Gaffney, 1981 andCartmill, 1981). The 'a priori' postulation of transformation series is regarded as the formulation of the hypothesis; the construction of the cladogram, following the rules as set out above, is then the deductive phase by which testable predictions are (hopefully) derived.
The a priori choices made here are the following: 1) For each character used a transformation series is specified and given direction in advance of further analysis.
2) Also, prior to analysis, several groups of species are regarded as monophyletic units and the monophyly of these units has been considered a fixed matter. Only when in the course of the construction of the cladogram it was not possible to proceed further otherwise, the monophyly of these groups was reconsidered.
With regard to these points, the following remarks can be made. In several cases, the characters occurring in small, restricted groups have been regarded as apomorphic without further arguments. This admittedly dubious '"in-group" common = primitive' argument was not used, however, in all these cases.
An assessment of the primitive state for rhizome structure is not made as the possibility of a more complex transformation series than a one-step series could not be excluded. ad 2. The initial acceptance of supposedly monophyletic groups is not justifiable by reference to some high principle. It is, however, firmly established taxonomic practice, and is equivalent to the initial recognition of species as units of evolution; it can also be regarded as equivalent to giving a heavy 'weight' to the characters that are used in recognizing the groups.
Application of these considerations have made it feasible to handle a dataset of this size without relegating the actual constructing to a computer. The number of possibilities is drastically reduced by accepting species-groups as cladogenetic units. Relatively little extra work was found to be necessary to analyze the groups themselves in order to arrive at a hypothesis about the ancestral states for those characters for which the groups were found to be heterogeneous.
Initial specification of the transformation series similarly reduced the number of possible cladograms.

Character selection
The characters used for this analysis were taken from as wide a range of morphological structures as possible. Some selection, however, is necessary.   Characters of which the use was rejected for these reasons were: The characters left after this selection was applied are given in Table 3.

Unknown transformations
Characters were finally rejected if no transformation series could be specified with at least a little plausibility. Characters that were initially considered but finally rejected for this reason were, e.g., the structure of the ground tissue of the rhizome; the surface structure of the spores (at the time under investigation, Van Uffelen & Hennipman, 1985). These characters were used, however, in the initial delimitation of the species-groups, but confirmation of the monophyly of these groups was sought as much as possible in other characters with a specified transformation.
The postulated transformation series were used in recoding the characters in table   3; the results are presented as background to the cladogram (Fig. 13). A few additional characters, introduced later in the construction of the cladogram are given as 'a', 'b' and 'c'. 85 Proposed transformations Rhizome morphology (see p. 15).
For the assignment of values for these characters, the situation in the majority of specimens in each species is taken as a criterion for the assignment of a particular character code. Incidentally, aberrant specimens occur. A maximum internode length of 2 cm is taken as upper limit for the delimitation of 'shortly elongated' rhizomes. At first sight this value appeared to give the taxonomically most useful separation; this view is confirmed by the absence of parallelisms for this character in the final cladogram. "Buds situated opposite the phyllopodia" is not considered as a separate character, as this condition is rarely constantly present in a species. 5C -Rhizome differentiated in parenchyma and sclerenchyma.
Characters 3A, 3B and 3C can easily be combined into a transformation series.
The direction is indicated by the observation that in many species the loss of sclerenchyma in the rhizome is a not uncommon phenomenon so that reduction here seems a more plausible possibility than repeated 'de novo' origin of a similar sclerenchyma structure. As sclerenchyma strands appear rather late during the development of the rhizome a mechanism that is possibly involved is retention of the juvenile state.
The character transformation series thus constructed shows a distinct correlation with that hypothesized for characters 1A, IB, ID and 4A, 4B, 4C.
For character 3B, the number of 10 strands is taken as a more or less arbitrary distinction between few and many; with regard to the frequent reduction pointed out above the highest number of strands found in a species is used as a criterion to assign a character to that species.
A transformation for characters 4A, 4B and 4C is proposed following the principle that structurally more complex features are to be considered derived if other arguments are lacking.
No transformation is proposed for the completely sclerified rhizomes of the P.
costata-group (5B) and the almost completely parenchymatous rhizomes of the P.
africana-group (5A). It is difficult to fit these two characters into a transformation scheme together with the differentiated rhizomes of the other species (5C). On the one hand, the lack of differentiation in both groups can be taken as a unifying character for the two groups; on the other hand, the presence of sclerenchyma can be considered a character shared between the P. 8B-Margin entire.
9A -Scales with glands only at the apex.
9B -Glands present at the base of the scale.
The direction of the transformation series connecting the characters 6A, 6B and 6C, 7A and 7B is supposed to be towards increasing complexity of the structures involved (see also p. 80). The postulated series 6A -6B-*6C implies that the presence of an acumen is to be considered plesiomorphic, the (occasional) absence as derived; the distinct shape of 6E is considered another derived character.
Fragmentary data from cross-sections of rhizome scales indicate that character 7B may actually be composed of two different, possibly independent characters (see p. 26). The distinction is not evident without sectioning the scales. The distinction between 7 A and 7B that is recognized here is the following: in 7B the scales are distinctly thickened at least near the point of attachment, which is evident in translucent light; in 7A the scales are equally thickened and translucent throughout. 7C is a further modification.
Characters 8B, 8C, 8D, and 9B are restricted to small groups of species.
Entire scales (8B) are found mainly in the P. albicansgroup p (also in the P. lanceolatagroup: coarsely dentate scales (8C) are characteristic for P. asterosora and P.
distichocarpa (but occur also in P. africana); long, curly cilia (8D) only occur in the P.
lingua-group. Basal marginal glands (9B) are characteristic for P. eleagnifolia and P. rupestris, they occur also in P. schimperiana and scattered throughout the P.
costatagroup. In this last group they may be more frequent than is observed as the scales are often severely damaged while being dissected from the rhizome.
No transformation is proposed involving character 8E: on the one hand, the presence of superficial papillae can be considered as indicating an increasing complexity, on the other hand it can be regarded as indicating a lack of differentiation between surface and margin of the scales.
10B-Fronds differentiated into stipe and lamina, stipe with lateral vascular strands.
10D-Fronds differentiated, stipe with lateral vascular strands, dorsal strands in stipe fusing below the lamina.
11A -Stipe without a central bundle of collenchyma.
11B -Stipe with a central bundle of collenchyma.
Here again, as in character IB, the situation in the majority of specimens is taken as a criterion for the assignment of a character code to a species. Incidentally, an indistinct stipe may be found in otherwise estipitate species; conversely, stipes may be absent in highly reduced forms of stipitate species. In some species with constantly small fronds (P. angustissima, P.  developments. An indication that such is the case is found in the apparently different ways in which the sterile fronds seem to be derived from fertile ones (p. 35): either by complete suppression of the formation of sporangia (P. linguagroup), or by suppression of apical growth (e.g., P. angustata-group). Considering the very widespread occurrence of forms of frond dimorphism and the various expressions of it throughout the ferns it is not surprising that some parallellism apparently has occurred also within a single genus.
The general direction of the trend is set by the consideration that in spore producing species the presence of specialized sterile fronds is of necessity a secondary development.
13C -Lamina often lobed near the base with several basal teeth.
The transformation series involving characters 13A, B, C and D is more fully discussed on p.
37. The recognition of the first step in this series in P. drakeana was possible only after the general trend had been recognized. Due to considerable plasticity the characters in P. drakeana and P. sheareri are not evident in all specimens.
14F-Some rays of the upper layer forming dorsal spines.
In contrast to the method used in characters IB and 10B, for the assignment of a character to a species the presence of woolly hairs in some specimens was considered sufficient here. Thus, the potential to form woolly rays is considered as more important than the degree to which this potential is actually realized. This is because of the notable plasticity with which this character may be expressed, and the possibility of asynchronous development of lamina shape and indument during the heteroblastic series. The procedure is crucial, however, only in assigning characters to P. assimilis and P. rhodesiana: in all other species a 'majority count' of the available specimens would have yielded the same results as the procedure followed.
Character 14D may be composed of two different characters: one with boatshaped, one with acicular straight rays (in the P. lingua-and the P. costata-group resp).
The direction of the transformations of these characters is indicated by the heteroblastic development (see p. 45). The thin, often sparse induments showing character 15B, and the distinct structure of character 14F obviously represent a reduction resp. a specialization. The reduction leading to character 15B apparently is a common phenomenon, as highly similar induments of this type are 90 found in widely different groups (e.g., P. schimperiana in the P. africana-group, P. sheareri in the P. shearerigroup, P. petiolosa in the P. lingua-group, P. novo-guineae in the P. angustata-group). A similar indument can also be found in some species of Platycerium (Hennipman & Roos, 1982, p. 68, PI. 11: c). In a strongly reduced form (as in forms of P. lanceolata and in the P. piloselloides-group) I the difference between boat-shaped and acicular rays is obscured.
No transformation series is postulated for these characters. The pattern that emerges in the heteroblastic development is ambiguous: character 16C seems to be often preceded by a stage with a pattern similar to 16D, whereas this seems not to be the case for characters 16A and 16B. Apart from that, it would be most likely that a series 16A-• 16B-T6C is present. The occurrence of venation patterns with recurrent veinlets in a genus with predominantly excurrent veinlets is curious, as the difference seems to be a consistent one between large groups in the Polypodiaceae. In the absence of a more general theory about the development of venation patterns I desist from postulating theories about transformations in Pyrrosia.
17E-Hydathodes in a row on the lower surface of the lamina. Character 18B is restricted to the P. piloselloides-group. It seems to be an adaptation directly related to the succulent character of the lamina. Similar succulent fronds of, e.g., P. lanceolata or P. longifolia usually show some, if not much, differentiation; succulent fronds in the P. albicans-group (e.g., in P. nummulariifolia) show a completely different development, with a well-differentiated mesophyll with a very thick water-tissue (see also p. 56).
The direction of the transformation of character 19A and 19B is indicated by ontogenetic studies (Sen & Hennipman, 1981).
Impressed or immersed stomata are an obvious adaptation to a xeric habitat; in the species were this character is present probably to the periodically dry conditions experienced by high epiphytes. The various degrees to which the stomata are sunken are accordingly interpreted as representing successive stages in a transformation series.
Sori and sporangia (see also p. 57) 21A-Sori small, several to many in each areole.
2ID-Sori in a single row between costa and margin.
Small sori are 1-1.5 mm diam. when ripe, larger sori may be up to 4 mm diam.
In very small-fronded species the number of sori may be reduced to 2 per areole without a concomitant increase in sorus size; this is considered a parallel development and is not scored as apomorphic (P. assimilis).

92
A single row of sori may be due to extreme reduction of the width of the lamina (as in P. schimperiana var. liebuschii, or in P. linearifolia), but this is regarded as a different character from the apomorphic character 2ID.
Presence of character 21D is considered a further modification from character 21B. The connection is beautifully shown in an atavistic form of P. angustata (see p. 59). The last step in this transformation series composed of characters 2IB and 21D is represented by character 21E. Nevertheless, it is also likely that character 2IE has arisen more than one time, and also through other mechanisms than those implied in the series 21A -21B-■ 2 ID -21E. In the P. angustata, P. rupestris, P. eleagnifolia, P. serpens. In many other species very slightly immersed sori occur, but the exact degree of immersion can perhaps not be reconstructed from herbarium material.
23A-Indument of the receptacle similar to that of the sterile lamina.

23C -Paraphyses in
A partly indurated annulus is restricted to the P. costata-group (see p. 62). It is an obvious specialization.
Are the groups monophyletic?
Prior to the construction of the cladogram several species-groups were recognized (chapter 4) as possibly monophyletic ones. An assessment of the arguments for monophyly of these groups can be given at this stage. The acceptance of certain groups as monophyletic can break down the complexity of the cladogram into manageable parts.

P. africana-group
In the interpretation of the characters presented above, this group has no autapomorphies. Nevertheless, it is provisionally accepted as a monophyletic one on acount of the homogeneity with regard to rhizome morphology, rhizome anatomy and sporoderm structure.

P. costata-group
This group has the obvious autapomorphy of character 24B; moreover, it is homogeneous with regard to rhizome anatomy, venation, sorus structure and spores.
It can confidently be regarded as a monophyletic one.

P. porosa-group
This group lacks autapomorphies. Moreover, none of the characters used for the delimitation of this group is an apomorphy in the interpretation presented above.
The group cannot be regarded as monophyletic unless some characters are interpreted in a different way.

P. sheareri-group
This group has no autapomorphies. It is provisionally accepted as monophyletic because of the rather weak delimitations between the species in this group, suggesting that speciation is not fully completed.

P. lingua-group
This group has no evident autapomorphies among the characters enumerated above, but possibly a single one in the peculiar form of frond dimorphism (p. 94 35). In other aspects, particularly spore morphology, the group is markedly heterogeneous. It is certainly not a well-established monophyletic group.

P. albicans-group
Although autapomorphies are absent from this group as well, it is provisionally accepted as a monophyletic one on account of the uniformity with regard to the rhizome scales, the lamina anatomy and the spore sculpture. This last character, however, is shared between the species in this group and P. sphaerosticha, here provisionally placed in the P. lingua-group on account of rhizome scales and frond shape.
7. P. angustata-group I confidently regard this group as monophyletic on account of great similarities between the species, e.g., in rhizome scales, sorus structure and spore sculpture.
A similar sorus structure, regarded as apomorphic, is also found outside this group, but the other two characters are unique to it.

P. confluens-group
In this group a similar tendency towards the formation of coenosori is present as in the last one. The monophyly of the present group is accepted also on account of similarities in structure of the rhizome scales (cj. p. 27) and the reduction of the hydathodes to a marginal row (17C + E). Although this last character is absent from P. eleagnifolia, the similarities between this species and P. rupestris are so strong that P. eleagnifolia is confidently included in the monophyletic group.

P. lanceolata-group
This group has the obvious apomorphy of character 23C. P. foveolata is here added to this group on basis of character 23B.

P. piloselloides-group
This group has character 18B as an autapomorphy. Its monophyly is confirmed by the homogeneity for characters 3C, 2IE, the unique venation pattern, and the marginal sori (Ravensberg & Hennipman, 1986).

Relationships within the genus
The actual construction of the cladogram ( fig. 13) starts with those of the groups enumerated above that are most confidently regarded as monophyletic. Of these groups, the P. confluens-(8), P. lanceolata-(9) and P. piloselloides-group (10)  albicans-group, some members of the P. lingua-group 'here possibly in a slightly different form) and in P. angustissima. These occurrences have to be explained as parallelisms.
Character 12B has parallel occurrences in P. pannosa (weakly), in the P. linguagroup (but, again, possibly in a different, not homologous, form), in the P.
angustatagroup ; and in several species of the P. albicans-group. Resolution of the relations within the P. albicans-group (see below), however, indicates that the occurrences of character 12B are best explained as parallelisms, and that the ancestral state of the group probably is the corresponding plesiomorphic character 12A. The occurrence of this character in the P. angustata-group is discussed below.
It has already been indicated that character 14E probably arose more than once, and the occurrences outside the group considered here are accordingly considered to be parallelisms.
Within each group a higher resolution is possible with the aid of the characters 8B and 23C, 9B, 21D, and 17E. Character 17D is not used for this purpose, due to the uncertain situation in P. lanceolata (see Taxonomic part,p. 195).
The next group considered is the P. angustatagroup (7). This group shares with the joined group (8,9,10) the character 12B, with group 8 only it shares the tendency towards the formation of a coenosorus (characters 21B -21D -21E).
With the P. albicans-group (6) it shares character 17D, and the placement of group 7 is thus a matter of weighing character 12B against character 17D, unless other arguments can be found. One possible reason to unite group 7 with group 6 instead of with (8,9,10) is that the perispore sculpture possibly is more easily brought into a relation to that of group 6 than to that of group (8,9,10).
If this argument is accepted and group 7 is accordingly placed next to group 6, the tendency towards formation of a coenosorus occurs at least three times (in group 7, 8 and 10).
The group formed by joining group 7 to group 6 has an apomorphy in character 17D. The ocurrences of this character outside the group can be explained by three parallel events.
Within group (6,7) further resolution is possible if the initial assumption of monophyly of group 6 is abandoned. It is not supported by any apomorphies anyway. On basis of character 2IB the group formed by P. asterosora and P.
distichocarpa is then considered to be the sister-group of group 7. P. albicans can be joined to this group on basis of character 6E, but that would require two other 97 events: one in 6E and another concerning 21B. P. nummulariifolia, P. kinabaluensis and P. rasamalae share the character 2B and are considered as a subgroup. The occurrence of character 2B here is parallel to its occurrence outside group (6,7).
These three species also share a secondary reduction of character 23B, parallel to a similar reduction in P. novo-guineae.
From the resolution thus obtained for groups (6,7) and (8,9,10) it follows that the ancestral state of both groups is likely to be 23B. Both groups can accordingly be united with character 23B and 10C as autapomorphy. A different interpretation of character 23B would require that it developed independently at least three times, with several reversals elsewhere. In the interpretation adopted here it arose only once, and reverted three times, in some cases probably as a result of reduction in sorus size.
The group thus formed shares with the P.
lingua-group (5) the apomorphic characters ID, 3B, and 20C and is at the next lower level united with that group.
Many individual species in group (6-10) also share with group 5 the apomorphic characters 12B and 15B, but from the distribution of these characters it follows that they probably arose as parallelisms in group ((6,7)(8,9,10)). On the next lower level the P. porosa-group (3) comes into view. With the groups (5-10) it shares apomorphies IB and 6C, and accordingly it can be united with (5-10) on basis of these two characters. The initial supposition that the P.
borosa-group as defined in Chapter 4 lacks autapomorphies is here confirmed, and the group (3, 5-10) therefore has a basal polytomy with 6 branches. In this polytomy also P. pannosa can be included on basis of characters IB and 6C, as it lacks characters 20C and ID, necessary to include it in a group higher up the cladogram. Reduction of the number of branches in this polytomy can be attained by recognizing small monophyletic groups within it: P. assimilis and P. linearifolia can be united on basis of the joint possession of narrow fronds (additional character "a"), in the remaining species a subgroup can be recognized formed by P. pannosa and P. rhodesiana, sharing characters 3C and 17B.
At this point in the construction of the cladogram the position of P. angustissima and P. gardneri can also be considered. P. gardneri shares all its synapomorphies with group (3,(5)(6)(7)(8)(9)(10) and clearly should be included in it. With group (5-10) it shares one synapomorphy (character 10B), but it lacks all synapomorphies of the constituting groups. It can therefore be placed as a sister-group to group (5-10), but if characters of the perispore are taken into consideration (Van Uffelen & Hennipman, 1985) another position is also possible. P. gardneri shares a peculiar spore type with P. laevis, P. lingua and other species of the P. linguagroup, and can accordingly be placed as a sister-group to the group formed by P. laevis and the P. linguagroup.
The position of P. angustissima is more equivocal.
It shares a number of apomorphies with species from the P. angustata-group i (characters 2B, IOC, 2IE, 17D). On the other hand, on basis of the spore type (Van Uffelen & Hennipman, I. c.) P. angustissima should be placed with the P. lineua-group, and that is the position here adopted. The characters in common with the P. angustata-group are then supposed to be parallelisms. Inclusion of P. angustissima in the P. angustatagroup would require a number of reversals in other characters.
The next group added is the P.
sheareri-group (4). This group shares with groups (5-10) the apomorphic character 10B (stipitate fronds); however, it lacks apomorphies IB and 6C, common to group (5-10) and the P. porosa-group. . Placing group 4 with group 5-10 would result in one homoplasy more than the adopted place. The P. sheareri-group is joined to group (3,(5)(6)(7)(8)(9)(10) with characters 7B, 19B and 20B as synapomorphies. Further resolution within the group is obtained with the aid of characters 13B, 13C, 13D (dissection of the lamina) and by a character not contained in the full dataset, viz. the presence of a distinct hypoderm (marked "b"). This character has not been evaluated for other groups than the P. shearerigroup, as there appeared to be no way to assess the presence or absence of a hypoderm in all species in a comparable way (cf. p. 84). This is caused by the presence in many species of a hypoderm modified into water-tissue, and the obscure differentiation of the mesophyll in many of the more succulent species (p. 56).
In the P.
sheareri-group , however, the mesophyll is usually welldifferentiated and a water-tissue is never present; accordingly within this group presence or absence of a hypoderm could be assessed relatively easily.
The presence of a hypoderm is here considered the apomorphic character, on account of the increased complexity of the mesophyll in these cases. The absence in P. hastata and P.
polydactyla is then a secondary reduction, confirming the monophyly of these two species that is already established by the curious lamina shape.
At the next lower level in the cladogram there is a trichotomy formed by groups (4-10), P. penangiana and P. mannii. These last two species share character 6B with group (4-10). Another possibility is that on this level the P. costata-group (2) is joined to (4-10) on basis of character 19B. It is difficult to assess the relative merit of either possibility, and the distribution of other characters gives no clear support for the choice made here. That I have chosen to unite P. mannii/P. penangiana to (4-10) on this level rests mainly on similarities in rhizome structure; these similarities, however, have not been interpreted in terms of plesioor apomorphic. The decision therefore remains arbitrary to a relatively high degree.
All the characters that are apomorphies in Pyrrosia thus fail to delimit Pyrrosia as a genus in the circumscription that has so far been used.
This uncertainty regarding the monophyletic status of Pyrrosia is expressed in the cladogram by the basal trichotomy joining Platycerium, the Pyrrosia africanagroup (1) and the rest of Pyrrosia (2-10). As is indicated above, with only slightly less support another trichotomy can also be constructed at the base of the cladogram, with resp. Platycerium, the P. costata-group, and the rest of Pyrrosia on the three branches.

Relationships with other Polypodiaceae
At this basal level in the cladogram it may be relevant to consider also the possible relationships of the subfamily Platycerioidae to other polypodiaceous ferns. All remarks concerning these relationships necessarily are preliminary, as no full survey of the Polypodiaceae with regard to the relevant characters is as yet available.
When the Platycerioidae are compared to the rest of the Polypodiaceae, the following characters emerge as possible synapomorphies for the subfamily:  (Nayar & Kaur, 1971). These characters must be interpreted as either apo-or plesiomorphic before their implications can be compared to those of the characters considered above.

Postscript
The subdivision of the genus and the phylogenetic analysis presented in this chapter are based on a set of data from which the sporoderm characters were left out, as a detailed analysis of the sporoderm was in preparation but not yet available. These data now being published (Van Uffelen & Hennipman, 1985), it emerges that a classification can be based on spore type alone that is to a large extent concordant with the subdivision presented in Chapter 4. There are, however, a few discrepancies.
A subdivision based solely on spores would classify P. angustissima, P. gardneri, and P. laevis together with P. lingua, and this is the position adopted during the construction of the cladogram. Another difference between a classification exclusively on basis of the spores and the one presented here is that in the former, P. abbreviata and P. sphaerosticha would be classified not with P. lingua, but in the P. albicansgroup.
In the following paragraphs I shall try to evaluate the result of the adoption of such a subdivision for the phylogenetic hypothesis presented here.
In all relevant characters P. abbreviata and P. sphaerosticha are similar, and transfer to the P. albicans-g roup is therefore most plausible in the form of the transfer of a single, supposedly monophyletic group. The interpretation of the following 104 characters is affected: 2B (the development of a shallowly grooved rhizome cannot be explained any more by a single origin for P. christii, P. abbreviata and P. sphaerosticha); 12B (as indicated on p. 35, the form of frond dimorphism that has developed in the P. albicans-group seems to differ from that developed in P. abbreviata and P. sphaerosticha); 15B (one more separate origin of the monomorphic indument with boat-shaped hairs has to be assumed); 21C (the densely packed sori cannot be explained as synapomorphy for P. christii, P. abbreviata and P. sphaerosticha); 17D (if the absence of hydathodes should be considered a synapomorphy for the P. albicans-group, the presence in P. abbreviata and P. sphaerosticha must be assumed to be a reversal).
Frond dimorphism (12B) in P. abbreviata and P. sphaerosticha can then be linked to the dimorphism in P. kinabaluensis, as can be the densely packed sori (21C); so that P. abbreviata and P. sphaerosticha can be placed as a sister-group to a group formed by P. nummulariifolia and P. kinabaluensis.
This leads to an alternative arrangement of the species in the group as in fig. 14. In this arrangement even more homoplasies are present than in the original arrangement. The somewhat ambiguous position of P. sphaerosticha in the accepted cladogram, however, is emphasized by the occurrence of character IOC.

Distribution and ecology
The present distribution of Pyrrosia is shown in fig. 15 The distribution of the individual species is shown in Figs. 16-26. Most taxa inhabit continuous areas of varying sizes. Circa 10 taxa are truly wide-spread, The boundaries between the restricted areas are drawn purely on Fig. 15     confluensgroup ( fig. 26), which inhabit relatively large areas extending over islands in the Pacific Ocean. In this group, a particularly wide disjunction, from New Caledonia to Tahiti, is shown by P.

113
Many species of Pyrrosia have adapted to more or less extended periods of drought. Two different growth-forms seem to have evolved with regard to drought resistance: a poikilohydrous and a succulent growth-form.
The poikilohydrous form is best exemplified by P. schimperiana. This species has been observed in the field to behave like a "resurrection" plant: in a period of drought it dries out without shedding its fronds; after rain the fronds take up water quickly, resuming metabolic activity probably within a few hours (Kornas, 1978 Judging by the distribution of these species, combined with the exhibition of one of the two morphological syndromes discussed above, it seems likely that the poikilohydrous habit has evolved in response to a seasonal climate with long dry periods. Most species that clearly exhibit this syndrome (e.g., P. porosa, P. flocculosa, P. costata) are restricted to the mainly seasonally dry areas of continental Asia. The succulent habit may be a response to the circumstances under which high epiphytes occur, even in everwet climates: short, severe dry spells between torrential showers. Distinctly succulent plants are often found as high epiphytes, and the xerophytic anatomy of many species is at first sight in contradiction to their preference for everwet areas (e.g., P. angustata, P. rasamalae, P. num-

mulariifolia)
In general, the species of Pyrrosia are epiphytes, and they have a preference for hilly or mountainous areas. A considerable number, however, seems to be more often (but never exclusively) epilithic or terrestrial: from the available data this preference could be established for P. angustissima, P. assimilis, P. drakeana, P. hastata, P. lingua var. lingua, P. pannosa, P. petiolosa, P. princeps, P. rupestris, P. sheareri, and P. subfurfuracea. Most species that are preferentially epilithic occur in Geographical analysis of Pyrrosia.
The first stage in the biogeographical analysis of Pyrrosia is the search for similar patterns of speciation in different monophyletic groups.
In order to have any relevance for this kind of analysis, the distribution of the members of these groups should be mainly allopatric. Sympatric species either have become so through dispersal after speciation, or directly by way of some sympatric speciation process. In the former case it will generally not be possible to identify the area where the species originated, in the latter case the speciation obviously is uninformative with regard to vicariance processes. The explanation of patterns of this type should be left for the second of the two stages of analysis recognized above.
A further restriction is that comparisons can be made only for groups roughly inhabiting the same area.
In Fig. 27 lingua-group a similar dichotomy is seen between the continental Asian members of the group around P. lingua and the Malesian members P. abbreviata, P. christii, and P. sphaerosticha, both groups supposed to be monophyletic ones.
The other group under consideration here, formed by uniting the P.
aneustata-g roup with P. distichocarpa and P. asterosora (see p. 96) has no representatives on continental Asia.

A dichotomy on some level between West/Central Malesia (Sunda Shelf)
and East Malesia is visible in all three cladograms: In the P.
costata-group this dichotomy is present between P. platyphylla, on Borneo, and the monophyletic group formed by P. splendens and P. princeps, occurring in the Philippines and on New Guinea, Celebes and the Moluccas, respectively.
In the P. lingua-group the same dichotomy is present between P. abbreviata, occurring on Sumatra, Java and the Lesser Sunda Islands, and P. sphaerosticha, from Celebes and the Philippines.
In the P. angustata-g roup the Sundaland-East Malesia dichotomy is present in the sister-group relationship between P. angustata on the one hand and the monophyletic group formed by P. samarensis and P. novo-guineae on the other.  (1) above. An identification of the two vicariance events discussed under (2) and (3) is more difficult to find: the exact tectonic history of the area following the collision of Papaustralia with southeast Asia is far from clear (Powell et al., I. c., fig.   4). There is thus plenty of scope for speculation, and the possibility that the last two events indeed correspond to some episode in the history of this unstable area cannot be ruled out. The timing of this event would have to take into account that it has to be dated later than the first event. It seems obvious that the collision of Papaustralia with the Malesian area may have been concerned. A possible course 119 of events is that an extensive island arc on the east side of the Southeast Asian subcontinent was involved, parts of which may have been incorporated into New Guinea, parts into the Philippines.
Starting from these assumptions, the evolutionary history of Pyrrosia may tentatively be filled in in the next stage of the analysis. In this stage, explanations can also be sought for unique patterns. The result of this stage is a scenario, constructed with the aid of the following considerations.
-From the assumption that the basal fork in the P. costatagroup and the p.
lingua-group refers to an event in which the Indian subcontinent was involved, it follows that the ancestors of both groups must have been present on this subcontinent before it came into close contact with Southeast Asia. We may then suppose that also the ancestors of the groups that split off on the cladogram between these two were also present on this subcontinent, and not in any The most striking of these are the wide distribution of the P. lanceolatagroup, and of other species that are of recent origin according to this scenario.
In the scenario presented here, these distributions have to be due entirely to long-range dispersal. Therefore, the possibility has to be considered seriously that the place assigned to the P. lanceolata-group in the cladogram is wrong and that the age of this group as it emerges from the scenario has been significantly underestimated. The first implication is that the present distribution of Pyrrosia on continental Asia has not been reached by overland dispersal around the mesozoic Tethys sea, but by rafting on the Indian subcontinent. The prediction derived from this is that fossils will be absent from western Asia, but that they may be found yet on the Indian subcontinent.
Another implication is that Pyrrosia must have been absent from the Malesian area before the Cretaceous. This implication is testable by way of fossils in a similar way as the one above.
Similarly, the presence of fossil Pyrrosia on the Australian mainland may be predicted from the assumption of widespread but relatively recent extinction in this area. This extinction is implicit from events (3) and (6)  Another feature of the present distribution of Pyrrosia that can be tied in with this scenario is the species-richness of Sumatra. As discussed above, it is difficult to account for this relative richness with ecological considerations only. There may be a historical aspect, too: in the scenario presented here Sumatra has served as a centre of speciation in severed groups, and in the past may have been in contact with both the Indian subcontinent and with Australia. This stands in contrast with New Guinea, which only relatively recently has become a contact zone between Malesia and Papaustralia, and is possibly a centre of speciation in only one group (the P. lanceolatagroup).  COTTHEM, W. R. J. VAN, 1970. Comparative morphological study of the stomata in the Filicales.
Bull. Jard. Bot Hist.) Bot. 2 (5) STEENIS, C. G. G. J. VAN, 1981 Thus, the various dimensions of the fronds are expressed in cm, with further precision in decimals only when necessary. The height of the capsule of the sporangia is not expressed in micrometres but in fractions of millimetres, as it had often to be reconstructed from more or less widely opened sporangia.

Rhizome
Thickness of the rhizome as given in the descriptions refers to the largest crosssection in a rhizome that has been boiled in water for several minutes. As the rhizome is often dorsiventrally somewhat flattened, in most cases this represents the horizontal width. The position of the lateral buds is indicated relative to the next phyllopodium apical to the bud on the same side of the rhizome. Rhizome anatomy is described from cross-sections from parts at some distance behind the growing tip.

Scales
Scales were preferentially taken from the rhizome at 1-2 cm behind the growing apex.
Here they are fully developed and not yet eroded in any way.
In short, slow-growing rhizomes scales may be taken closer to the apex. If an apex was not present in a collection, scales were taken, if necessary, from a part of the rhizome where they had apparently been well-protected during the process of drying and mounting. Scales from the phyllopodia are often different and are not included in the descriptions, but in some cases they are described separately in the notes.

137
Stipe As length of the stipe the distance is given between the phyllopodium and the first dilated part of the lamina, viz., where the lamina is more than ± 1 mm wide on each side of the stipe. If the lamina base is unequal, only the distance to the lower side is given. The anatomy of the stipe refers to the situation at ca. 1 cm above the phyllopodium, or, in cases of short or absent stipes, directly below or at the base of the lamina.

Lamina
Length and width of the lamina were measured in dry material (fresh material may be wider by a factor of 1 V2!) and were reconstructed as much as possible if the lamina was strongly rolled inwards. Indument is described from the sterile fronds or from sterile parts of the lamina; the venation is described from cleared parts of the lamina (see section Methods in the General Part) and is also taken from sterile parts of the lamina or sterile fronds. For the purpose of description, the hypodermis and water-tissue are treated as separate layers (see, however, p. p. 55). The thickness of the lamina has been measured in cross-sections prepared as indicated in the section Methods.

Sori
Size of the sori refers to the diameter of well-developed, mature sori and includes the outermost sporangia. Spore ornamentation is described as far as made visible by light microscopy. A more detailed description and a structural analysis is given by Van Uffelen & Hennipman (1985). Spore size has usually been measured in samples of 10 spores from three collections. In some cases a different number of spores has been measured, depending on the availability of material. The dimensions given include the perispore and its protuberances. The following generalization has been derived: the dimensions given represent the range of the means of the samples, between brackets the extremes are given. The first value given is the length, the second the width in equatorial view.

Glossary
In the descriptions the following terms are used in a way that may not be directly evident from common usage: paraphyses: vegetative structures in the sori. In many species distinctly different from the hairs on sterile parts of the lamina, either in size, in shape, or density; or in more than one of these characters. shortly elongated (of the rhizome): with distinct but short internodes, intermediate between short-creeping and long-creeping ( fig. 1 Herbs, epiphytic or epilithic, occasionally terrestrial, in small tufts or forming extensive colonies. Rhizome 0.5-±7.0 mm thick, shortly creeping to widely trailing, appressed to or just immersed in the substrate, densely set with scales, rarely (Pprinceps) with some woolly hairs on the phyllopodia; ventrally rooting diffusely or in two rows, dorsally with two alternating rows of more or less prominent phyllopodia, the phyllopodia contiguous or up to 9 cm distant; branching sparsely to regularly from lateral buds placed basally on the phyllopodia or up to a full internode behind each phyllopodium on the rhizome. Rhizome anatomy. In cross-section with a peripheral parenchyma several cells thick; central region either completely sclerified, or with a more or less distinct sclerenchyma sheath and a varying number of sclerenchyma strands within the sheath, or completely parenchymatous. Stele enclosed within the sclerenchyma sheath, composed of 3-± 12 vascular strands; frond traces with 2-4(-9: P. princeps) strands; branch traces with a single, U-shaped or cylindrical strand to 4-6 strands. Rhizome scales appressed to patent or squarrose, ± orbicular to lanceolate, 2-14x0.5-1.5(-3.3) mm, widest at or below (occasionally above) the attachment; either basifix with a more or less deeply cordate base, or pseudopeltate, or peltate with a short, more or less stout stalk; blade thin and translucent to variously thickened and opaque, cell walls either thin or all more or less strongly but equally thickened, rarely (P. costata-g roup) a few cells ± clathrate; margin entire, dentate or ciliate, at the base of the scale sometimes with one or more, sessile or stalked, glandular cell(s); apex narrowly (to broadly) acute (occasionally rounded), terminated by 1 or 2 glandular cells and/or 1 or 2 acicular cells; abaxial surface smooth or finely striate, rarely with cilia or papillae similar to the cilia on the margin; colour varying from hyaline to light or dark brown or blackish, darker near the point of attachment. Fronds articulated to the phyllopodia, mono-or dimorphic; sessile to stipitate, entire or rarely hastately ( P. hastata) or pedately (P polydactyla) dissected; covered with stellate hairs; texture pergamentaceous to (thick-) coriaceous, often succulent to 2 mm thick in vivo. Indument sparse to dense on either side of the lamina, usually denser on the lower side, often particu-larly thick and dense at the base of the stipe and containing scale-like filaments there; hairs stellate; stellate rays broad and flat or shallowly grooved, to narrow and narrowly grooved, or long, ribbon-shaped, curly and intertwined. Venation. Veins (in sicca) prominent on the abaxial side of the lamina or immersed, flat or occasionally grooved on the adaxial side; secondary veins usually distinct, running from the costa under an angle of (35-)40-50(-60)°, not reaching the margin; tertiary veins connecting the secondary veins, thus forming series of few to many more or less rectangular areoles; included veins either free, simple or forked, or anastomosing into a more or less complicated network, often      Notes.
They are somewhat similar in outline to the juvenile foliage fronds of Platycerium.
The soriferous region is often somewhat contracted.
Habitat: Epiphytic (sometimes as a high epiphyte), mostly in primary forest, but also on roadside trees, in abandoned coffee-plantations, etc.; occasionally epilithic or terrestrial. In everwet to seasonally dry areas.
Altitudinal range: 900-1800 m.  Much confusion of the present species with Pyrrosia penangiana (Hooker) Holttum and P. porosa (Presl) Hovenk. seems to have been due to another mixed collection by Zollinger. Zollinger 3183, syntype of N. mollis Kunze, in most herbaria is represented by specimens of P. penangiana, but in K by a specimen of P. albicans, and in P by specimens of P. rasamalae and P. lanceolata.
In P. porosa the rays of a single hair rarely differ as much as a factor 2 in length, in P. assimilis almost each hair has a few very short as well as some long rays.
Older fronds however, particularly those of P. assimilis, may have lost most hairs.   Altitudinal range: 300-1150 m.
BORNEO. -Sarawak: Brooks s.n.   Notes. 1. The thin indument is easily overlooked. On young fronds a more conspicuous indument is also present, composed of hairs larger than those left on older fronds, with patent, acicular rays. This probably corresponds to the juvenile indument of other species but in P. christii it appears to be frequently present in 167 mature fronds that are not yet fully developed. A more distinct woolly layer is also present on the stipe of young fronds.
2. The almost glabrous, often thin fronds at first sight are characteristic for a hygromorphic plant. According to Geesink, however, (pers. comm.) it may grow in full sun on dry rocks; in the greenhouse it also appears to be quite resistant to desiccation.
2. More or less stipitate forms of P. costata can be confused with P. stigmosa, especially as the two species sometimes occur together (as in Thailand). P. costata can best be distinguished from P. stigmosa by the flat or only shallowly grooved costa and the absence of collenchyma centrally in the stipe. In (3)  Habitat: Mostly epiphytic, in various situations, (tall trees in primary forest, hedgerows, orchards, and wayside trees); occasionally epilithic. Probably confined to seasonally dry 'pockets' in Sumatra.
SUMATRA. 43 collections (mainly from the mountains around Lake Tawar, G. Leuser, Lake Toba, and Padang Pandjang). Notes. 1. An easily recognized species. The rhizome scales are large and conspicuous, and have their greatest width distinctly above the point of attachment.
The only other snecies with rhizome scales thus shaned is P. albicans, which has entire scales. P. distichocarpa has coarse marginal teeth on the scales, each one formed by two projecting cells (plate 2 f; fig. 34 b).
2. The woolly hairs of the indument do not form a distinct layer. In some specimens (e.g., Rahmat si Boeea 11013, Surbeck 666) woolly hairs are absent; in most specimens, however, a few woolly rays can always be found interspersed with the acicular rays.
3. Variability: The base of the lamina is either distinctly truncate-cordate or gradually attenuate. Both shapes, however, may occur in a single collection (e.g., Lorzing 6078), the distinction thus cannot be used to separate a fo. attenuata (Rosenst., ined.). strands 6-10. Scales pseudopeltate, 2.5-6.2 x 0.6-1.2 mm; base irregularly dentate to ciliate; acumen dull brown to blackish with a straw-coloured margin, den-   characters). These have relatively small hairs (diam. less than 1 mm) with more appressed and more distinctly boat-shaped rays than is usual in P. drakeana, and the sori are more exserted from the indument. With P. drakeana these specimens share the dimorphism of the indument, the relatively long stipes, and the wide lamina.
2. Confusion is also possible between P. drakeana and P. boothii. Usually the much larger fronds of P. boothii are characteristic; however, small fronds may be distinguished from P. drakeana by the more cuneate base, the less dense indument, the more exserted sori and the slightly sunken stomata. P. boothii is not a common species, and more extensive collecting may show that the transition between P.
drakeana and P. boothii is more gradual than can be anticipated from the material at hand.
3. Due to the very coriaceous texture the detailed configuration of the included veins could not be studied.
The confusion with P. serpens is due to Forster's mislabelling of his specimen as collected in New Zealand. That his specimen differs from the common New Zealand species was already noted by Giesenhagen (1901), who cited it under Niphobolus tricholepis, but did not draw the necessary nomenclatural consequences.
As a similar mislabelling of a Forster specimen is known in P. longifolia (Copeland, B. P. Bish. Mus. Bull. 93 (1932) 66), it is almost certain that Forster's specimen is not from New Zealand but from one of the other Pacific Islands he also visited.

2.
Variability: Specimens from the Kermadec Islands deviate in having rhizome scales with a narrowly ± tubular acumen that is more strongly dentate than in typical P. eleagnifolia. The base of the scale of these specimens also consistently lacks marginal glands. 3. In a few specimens {Helms, s.n., 1882-1883 (M), Lambrechtsen 10 A (L)) the rhizome is in some places densely and fastigiately branched. According to a note to Lambrechtsen 10 A this is a gall due to a mite (Aceria spec.). A similar aberration is found in several specimens of P. confluens (see note 2 under that species). In P. eleagnifolia the gall is apparently not restricted to a single locality. 4. The rhizome scales may show a few annular figures in the cells as described under P. confluens.
180 the inner parenchyma, 10-± 25; vascular strands 10-13. Scales pseudopeltate, [4][5].7 x 0.5-1.0 mm; base entire to irregularly dentate; acumen dull brown to blackish, dentate, entire towards the apex. diam., with erecto-patent to appressed, boat-shaped to acicular rays, ± mixed with a lower layer composed of hairs with mainly woolly rays. Sori apical to all over the lamina, slightly spaced in an ill-defined patch, superficial; several in a single row in each soriferous areole, often elongated along the veins, 1-1. Notes. 1. Variability: the indument varies from a very thin, appressed mat to a dense one with shaggy hairs, but no other characters can be found that correlate with this difference. The transition between the two types of indument is gradual.
In northern Indo-China a form occurs (which has been called P. annamensis and P. rhomboidalis) with a more coriaceous texture, often narrowly inrolled fronds, and relatively longer stipes. If this form has a thin indument, it is at first sight difficult to distinguish from long-stipitate forms of P. subfurfuracea; however, the ciliate scales of the latter are characteristic. Similar plants occur more sporadically in Bhutan.
1. P. foveolata var. foveolata is often confused with P. lanceolata. It can be distinguished from P. lanceolata by the larger sori, in which paraphyses and sporangia are completely mixed. Apart from that, the rhizome scales of P. foveolata are, if ciliate, more sparsely so than those of P. lanceolata.
2. Variability: entire or dentate rhizome scales are found mainly on the thinner rhizomes (or rhizome parts), whereas on thicker rhizomes some cilia almost always are present. This seems to be a difference in development rather than a taxonomic distinction; both types of scales may occur in the same collection.
Hydathodes may be absent or present, but there is no correlation with the heterogeneity in rhizome scales nor with any other character.
b. var. lauterbachii. Fig. 35 fig. 2, 3; C. Chr., Ind. Fil. Suppl. (1913)   Habitat: Epiphytic, on mossy branches, silted trunks, leaning trees; generally at lower altitudes than the type variety. Notes. 1. The most pronouncedly dimorphic fronds occur in collections from low altitudes (up to 500 m). Less distinctly dimorphic forms (e.g., Clemens 155, Croft LAE 68294 ) occur at higher altitudes (to 1000 m). These are ± transitional to the type variety, which occurs generally above 1000 m. It may be that the transition from the type variety to var. lauterbachii is a more or less gradual one. In that case, the apparent discontinuity in the variation represented in the collections at hand may be due to the scarcity of specimens from intermediate altitudes; this, in its turn, may be due to the extensive deforestation at these elevations due to population pressures (R.J. Johns, pers. comm.). These galls occur in Versteeg 1254, Versteeg 1532, Pulle 128, Schlechter 19020. Obviously they are not restricted to a single locality. It is curious that the gall is completely absent from the type variety, which is much more common at higher altitudes than var. lauterbachii is at lower altitudes. This indicates that the galling fly is restricted to lower altitudes.
19. Pyrrosia gardneri (Mett.) Sledge. Fig. 21  Notes. 1. P. gardneri is superficially similar to P. porosa. The rhizome, however, is characteristic, with unusually high (up to 2.5 cm) phyllopodia, and covered 186 with conspicuously shining, appressed, black scales. The lamina anatomy is also characteristic, with the cell-walls in the hypodermis peculiarly thickened (illustrated by Giesenhagen, 1901: fig. 14 g). 2. The lamina is often narrowly attenuate at base, so that the transition to the stipe is very gradual, but the stipe is always free of the decurrent lamina for a short distance at its base.
Altitudinal range: 200-400 m, but probably occurring at higher altitudes as well.
Notes. 1. P. hastata can be very similar to forms of P. polydactyla; the distinctions between these two species are discussed under the latter.

The lamina may
have three or five lobes, in the latter case the two smallest ones near the base are often no more than small teeth.
3. The two main vascular strands of the stipe fuse some distance before entering the lamina (in other species except P. polydactyla they unite a short distance before or after entering the lamina). This may be an indication that the, for the genus unusually strongly, truncate base is formed by a basal constriction of the Sterile fronds: stipes 0.5-3 cm, V10-V2 x as long as the lamina; lamina, index 3-8; widest below the middle, 3-13 x 1-1.8 cm; otherwise similar to the fertile ones. Venation: secondary veins distinct, with the tertiary veins forming regular areoles; included veins forked and anastomosing; free veins ? (note 3).
Hydathodes absent. Anatomy: stipe with 3-5 central and no lateral vascular strands; lamina 0.3-0.9 mm thick, upper epidermis with flat cells with thin to moderately thickened walls, hypodermis absent or composed of 1 cell-layer, water tissue indistinct or thick, palissade and spongy parenchyma distinct or indistinct, lower epidermis with thin or thickened walls; stomata slightly to strongly sunken, pericytic.
Indument dimorphic, a dense mat, persistent, whitish brown; upper layer composed of hairs 0.4-1.2 mm in diam., with erecto-patent to appressed, boat-shaped to acicular rays, separate from or appressed to a lower layer composed of hairs with mainly woolly rays. Sori all over the lamina, closely packed, superficial; many in each soriferous areole, occasionally confluent along the veins; ± description, however, clearly refers to entire rhizome scales; his type is conspecific with P. rasamalae.
P. kinabaluensis differs from P. rasamalae in the ciliate rhizome scales as well as in the markedly dimorphic fronds and the more closely packed sori. It is in many characters intermediate between P. nummulariifolia and P. rasamalae, but distinct from either in the closely packed sori. Both the intermediate character and the relatively strong heterogeneity indicate a hybridogenous origin. This is not supported by examination of the spores: the percentage of abnormal spores is not higher than in other species (see also p. 71).
3. Venation: due to the very thick-coriaceous character of many fronds the venation could not be studied in detail.
Many species have been distinguished in this large and variable aggregate. In the following survey of the variation a number of extremes is recognized and identified by names derived from the species under which these entities have been separated. It is not my intention to give a formal status to any of these entities.
An extremely large number of cells seems to be characteristic for 'dimorpha'.
Stomata with such a large number of adjoining cells tend to stand out distinctly in dried fronds, probably due to the large numbers of cell-walls strengthening the area immediately around the pore.
Thus, with the aid of this character the otherwise often rather similar 'varia' and 'dimorpha' can be separated, but otherwise no useful distinction is found. The variation is almost continuous, and is even more so if the the total range of the numbers observed is considered.
-Thickness of hypoderm and water-tissue layer. In 'vittarioides' this combined layer is one cell thick; in 'varia' it is 1-3 cells thick, in 'pustulosa' and 'stellata' 2-3 cells; in 'adnascens' and 'dimorpha '2-4 cells  . The specimen in the Linnean herbarium (LINN) is a fertile frond of Elaphoglossum sp. and probably a later acquisition.
Habitat: Epiphytic, mostly on forest trees; or epilithic; in sheltered to exposed situations.
Altitudinal range: Sea level to 1400 m. Notes.

Variability. Plants from Taiwan differ from those occurring on
Cheju-do (Quelpaert) and Japan in having slightly shorter and wider rhizome scales (index to + 5, 2-4 mm long; compared with index to ± 9, 3-5 mm long) with a base that is less deeply lacerate-ciliate.
2. Lamina indument: the indument is sometimes somewhat similar to that of P. assimilis, which has hairs with unequally long branches. In P. linearifolia also very unequal branch-lengths occur on a single hair (differing up to a factor 7).   4. Numerous furcate and crispate forms have been cultivated, mainly in Japan.
They are often described as formae, varieties or monstrosities. For a review of many of these forms see Nakaike, Enum. Pt. Jap. (1975) Soc. 1 (1935) 52; Nayar, J. Ind. Bot. Soc. 40 (1961)  J. Bot. 45 (1967)    lingua woolly hairs are usually absent from the lamina, but occasionally present in small numbers at the base of the stipe and sometimes higher upwards on the costa (e.g., in Steward & Cheo 409). Var. heteractis is often considered specifically distinct from the type variety, the main differential character being the presence of a woolly layer in the indument. Although this is a conspicous character, it is not sufficiently consistent to warrant specific recognition.
The same holds for other characters.
The variation in frond shape is slightly different in the two varieties, but there is an almost complete overlap: In both varieties the shape of the lamina apex varies from rounded-obtuse to acuminate, but the extremes of this range (both rounded and acuminate) are more distinct and more frequent in var. heteractis.
Scattered black hairs occur in both varieties in varying density. They are most conspicuous in young fronds, and tend to be more persistent on stipe and costa.
They are most likely homologous to the upper layer of hairs that bear the dorsal spines in P. laevis.
In both varieties the rhizome scales can be patent or appressed. In var. lingua appressed scales are more common, in var. heteractis patent scales dominate.
In var. heteractis, the sori are more often distinctly spaced, in the var. lingua the sori are mostly contiguous, though individually distinct when old.
Thus, it appears that the weak distinction that can be made on basis of the indu-   3. P. pannosa has been erroneously reported from Burma by Beddome (1892) and . These reports are probably all based on the same few collections by Parish and Lobb. These collections are here referred to and discussed in  Naumann s.n., (24-5-1875)  Habitat: Almost exclusively epilithic, on rocks, walls, cliffs etc., often in exposed, sunny places, occasionally in sheltered situations in forest; only rarely epiphytic. Extending northwards to regions with up to 5 months of frost yearly. Notes.
1. P. petiolosa is usually distinct from P. lingua in its small size, more distinctly pitted hydathodes, more frequently confluent sori, and more thickly coriaceous fronds.
1. P. princeps can be confused with P. splendens and P. platyphylla. P. princeps can usually be distinguished from both by the indument, which has an upper layer of hairs with patent, acicular rays. Forms occur with more or less distinct dorsal spines, which are superficially similar to P. splendens; other forms (e.g., Kanis 1153, Brass 8859) lack the characteristic upper layer and then strongly resemble P. platyphylla. Detached fronds, from which both rhizome and rhizome scales are lacking, cannot be identified with certainty in all cases. The most constant distinguishing character is found in the rhizome scales, which in P. princeps have an acumen that is at least slightly dentate, but usually distinctly dentate or ciliate. In P. splendens and P. platyphylla the scales have an entire acumen. From P.
platyphylla the present species can moreover be distinguished by the short rhizome (shortly elongated in P. platyphylla).
A slightly aberrant form occurs on Celebes. These specimens have practically entire rhizome scales. They have the indument characteristic for P. princeps.
2. The distinctionbetween the indument on the rhizome and that on the lamina is slightly less sharp in P. princeps than in other species: sometimes a distinct layer of woolly hairs is present on the phyllopodium among the scales.  6 (1906) 991.
Altitudinal range: Low altitudes to 750 m. part of fertile fronds is expanded and similar in shape to a sterile frond and probably corresponds morphogenetically to a sterile frond; sterile fronds probably being derived from fertile ones by the arrestment of the apical development.

PHILIPPINES. Samar
The lamina may be abruptly constricted into the apical spike (e.g., Ramos & Edano 31467) or the transition may be more gradual (Madulid et al. 858).
Fully sterile fronds appear to be rare, and were present in only a few of the specimens studied. Most of the fully sterile fronds are conspicuously larger than fertile ones of other collections; collections with both fertile and sterile fronds are 241 few. A reliable evaluation of the morphological difference and co-occurrence between sterile and fertile fronds is especially difficult because of the following factors: -The apical fertile part develops only after the basal sterile part of the lamina is fully expanded; the absence of a fertile part on a developing frond is therefore no indication that the frond is a sterile one. It might not yet have become fertile at the time of collecting.
-The apical fertile spike is apparently fragile as in many collections it is broken off somewhere. All fronds with a damaged apex therefore could have been fertile ones.
Even after considering this, there appears to be a tendency for sterile fronds to be distinctly larger than the corresponding basal part of fertile fronds. This may be an indication that in this species luxurious vegetative growth in some way inhibits the formation of sori.
2. Apart from the presence of a coenosorus (which may be interrupted for some or all of its length) the following characters distinguish P. samarensis from P. angustata: -The indument is more distincdy differentiated into an upper layer with acicular rays only and a lower layer with woolly rays; in P. angustata these two layers merge more gradually.
-The fertile part of the lamina is more strongly contracted than in P. angustata.
3. According to , Presl's name Gyrosorium samarense cannot be applied to the present species, as it refers to a species more similar to P. lanceolata. , however, pointed out that the type specimen in PRC is P.
Habitat: Mainly epiphytic, on trunks and branches, on a large variety of host trees (e.g., Ficus spp., Mimusops sylvestris, Entandrophragma utile, Coffea), most frequently in forest or in gallery forest, also in wooded savannas, gardens, or in 243 degraded forest; also epilithic, on mossy rocks, in crevices, etc.; occasionally terrestrial; once reported growing on a termite hill. ANGOLA. 6 collections.
Notes. 1. The ripe sori stand out conspicuously against the light background of the indument. The contrast is particularly striking as the sori are more blackish than is usual in Pyrrosia; this is due to a modification of the one or two cells of the indurated annulus closest to the stoma. These cells are very dark and slightly swollen, Rhizome long-creeping, occasionally grooved ventrally,  Notes. 1. The name P. serpens has been used widely for P. eleagnifolia. The confusion is caused by a probably erroneous location given by Forster. Although the locality given for the type of Polypodium serpens is New Zealand, the specimen concerned is definitely not conspeciflc with the New Zealand species. It most likely was collected on one of the Pacific Islands that Forster visited. A similar mistake was probably made in labelling the type specimen of Polypodium acrostichoides 246 Forster (see under P. longifolia). P. serpens has been known under the names P. tricholepis and P. blepharolepis, these are antedated by P.  Island, and it seems likely that the specimen concerned has been mislabelled.
Pyrrosia sheareri Ching, Bull. Chin. Bot. Soc. 1 (1935) fig. 179; C. Chr., Ind. Fil. Suppl. 3 (1934) 10. -Pyrrosia grandissima Ching, Bull. Chin. Bot. Soc. 1 (1935) 63; Tagawa, J. Jap. Bot. 24 (1949)   In a few specimens, however, these characters are combined in a different way, which makes identification sometimes difficult. Some of these specimens have fronds shaped like P. sheareri but a more or less dimorphic indument (e.g., Henry 9114), or a monomorphic indument but with relatively large, ± acicular hairs {Kramer et al. 8130). In other cases the lamina may be similar in shape to that of P. drakeana, but the hairs are small and the sori more or less exserted {Dai 104588, Fang 3966, 12848, Wilson 5323). Although these specimens can be considered intermediate between P. sheareri and P. drakeana, there are no other reasons for supposing that they are of hybrid origin (see also note 3).
2. The lamina of P. sheareri is often characteristically lobed at the base with a varying number of lateral lobes or teeth on one or both sides. Often the base itself is highly asymmetric. Similarly asymmetric, unstable frond shapes are sometimes supposed to be an indication of hybridity ('structural irregularity ' Wagner, Phytomorphology 12, 1962: 87-100). In this case there is, however, no reason to assume that P. sheareri, as recognized here, contains a large number of hybrid specimens, as neither the spores are more often abnormal than in other species of Pyrrosia, nor is the chromosome number of one plant that was examined aberrant in the genus. Moreover, there are indications that the shape of the lamina is liable to variation that depends more on environmental circumstances than on the genetic constitution of the plant (see note 3). symmetrical (though still unequal) base. This type of plasticity is comparable to that discussed under P. polydactyla, in which a diminishing degree of lamina dissection could also artificially be induced.
Between the two plants of P. sheareri a slight difference also developed with regard to the diameter of the hairs: The one grown under cool conditions had hairs most of which were 0.25-0.40 mm in diam.; the other, from the warmer phytotron, had hairs most of which were 0.2-0.25 mm in diam. This variation falls completely within the range of P. sheareri, and therefore is not a sufficient explanation for the difference between P. sheareri and P. drakeana found for this character.
It should be noted, however, that the correlation found between hair diameter and climate when P. drakeana and P. sheareri are compared is similar to the correlation found when the two plants of P. sheareri grown under different climatic regimes are compared. Wherever P. drakeana and P. sheareri occur together, P. drakeana, which has the larger hairs of the two, tends to occur at higher altitudes than P. sheareri, where the climate may be supposed to be cooler. In the plants from cultivation similarly larger hairs were found on the plant from a cooler regime. It is curious that if the same correlation is considered between P. hastata and P. polydactyla, the converse effect is found: larger hairs are found on P. polydactyla, which grows at a considerably lower latitudes than P. hastata.
These considerations show that, although environmental conditions certainly play a part in the expression of characters, they nevertheless cannot fully explain the differences found between P. sheareri and P. drakeana.
Indument monomorphic, a thin mat, persistent, light brown to whitish; hairs 0.3-0.5 mm in diam., with appressed, boat-shaped rays. Sori all over the lamina or in a sharply defined, irregularly shaped patch, very closely packed, superficial; many scattered through each soriferous areole, not confluent; developing all more or less simultaneously, when old pseudoacrostichoid, exserted from the indument. Sporangia on stalks V2-I x as long as the capsule, capsule ± 0.4 mm high, with 21-24 indurated annulus cells.
Hab itat: Epiphytic, in forest or sometimes on exposed trees; occasionally epilithic. Notes.

P.
sphaerosticha is superficially similar to P. lingua, but is easily recognized by the very closely packed, pseudoacrostichoid sori and by the characteristic rhizome scales.

251
The rhizome scales are constantly ciliate in the lower half of the acumen; the apex, which is entire again, is moreover strongly and irregularly crisped and ± squarrosely curved outwards.
The scales on the phyllopodia are often enlarged and form a more or less distinct tuft around the base of the stipe similar to that in P. abbreviata.
2. Alston 16995 from Batjan is a slender plant with a rhizome only 1 mm thick, almost entire scales and small fronds.

3.
In SING is sphaerosticha. The same locality is indicated on two specimens of P. christii (which is restricted to Borneo) collected by Ch. Hose, his nephew. According to Van Steenis-Kruseman (Fl. Males. 1, 1950: 246) Ch. Hose also made some collections on Celebes, whereas his uncle did not. Therefore, it seems most likely that somewhere the labels of two collections, one by Ch. Hose and one by his uncle, have been changed. 4. The specimen in L of Cuming 127 represents P. rasamalae.
Altitudinal range: Sea level to 1500 m. Notes.
1. P. stigmosa can be deceptively similar to P. costata. Generally the elongated rhizome and the distinctly stipitate fronds are distinctive, but specimens with only partially collected rhizome and strongly inrolled fronds may cause confusion, especially with the more or less stipitate forms that may occur in P. costata. These specimens can be identified with the aid of one or more of the following characters: -In P. stigmosa the costa and secondary veins are narrowly grooved above (more distinct in dried than in fresh fronds), in P. costata costa and veins are practically flat or shallowly and broadly grooved; -In P. stigmosa a central strand of collenchyma is present in the stipe, in P. costata the stipe or lower part of the costa lacks a central strand of collenchyma; -In P. stigmosa the sori are shortly spaced and usually spread over the whole length of the lamina, in P. costata the sori are more closely packed and usually restricted to the apical part of the lamina; -In P. stigmosa the rhizome scales are more distinctly dentate at the base than in P. costata.
2.  identified Pyrrosia chinensis Mirbel with P. lingua (Thunb.) Ching. He may have been influenced by the fact that several cultivated forms that rowed, apex more or less acuminate. Venation: secondary veins distinct, with the tertiary veins forming regular areoles; included veins frequently forked and anastomosing; free veins many, mainly excurrent. Hydathodes distinct, scattered over the lamina, ± superficial. Anatomy: stipe with ± 9 central and ± 3 lateral vascular strands; lamina 0.3-0.7 mm thick, upper epidermis with flat cells with moderately to strongly thickened walls, hypodermis composed of a single celllayer, water-tissue absent, palissade and spongy parenchyma distinct or occasionally indistinct, lower epidermis with moderately thickened cell-walls; stomata superficial to slightly sunken, pericytic. Indument dimorphic, a thin or occasionally ± dense mat, persistent or fugacious, dirty greyish; upper layer composed of hairs 0.8-1.3 mm in diam., with erecto-patent, acicular rays, ± mixed with a lower layer composed of hairs with mainly woolly rays. Sori apical in a more or less sharply defined patch, closely packed, superficial; several to many in each soriferous areole, in 1 or 2 rays or irregularly scattered, occasionally confluent;