a an pe Oa ne eS ee ee en tt iD Sie etna are it cnn a eagwe ect Dee oe Cretan Site mm tee hp A CADE ALE NG DOAE Dee FA; ian in Rented ip De we ma RS dh simi Rg hms Ta Poland al . "orton . foots Pa Met ~ Rew e SoD signa ‘ . ‘ . Nae a ath seg, fe Rr en ay a tomee oe * aotee ~~ eR ie i ot tof | Pee ee ne ee ee ee ee er 2 a ~ i a a ed CoP tees z ”_ -—, > aim K —e AO er eer an se see A ieathe er ee en rr I th Nihe Mh at Ahr Bh a aide tetra ane bp a eg ee ene oe ‘ XXXVIT. 1947- Archaeology : “(Part I): eo. ee XXXVIII. 1950 Zoology : - - 9 15 0= XXXIX. In Press. : except the Geplenee and Palaeontological parts, which are obtainable from the GOVERNMENT PRINTER, PRETORIA. Copies may be obtained from— The LIBRARIAN, Sout Arrican Museum, Cape Town, : 3 OF THE _ SOUTH AFRICAN MUSEUM VOLUME XL PART II, containing:— ; . 3. Some aspects of the Morphology, Phylogeny and Taxonomy of the South _ African Iridaceae. By G. J. Lewis, B.A., PH.D. (With Plates II-XXIV.) ne ISSUED FEBRUARY 1954 PRICE 17s. 6d. raat PRINTED FOR THE | - TRUSTEES OF THE SOUTH AFRICAN MUSEUM BY THE RUSTICA PRESS (PTY.) LIMITED, COURT ROAD, WYNBERG, GAPE Bt ee ye eee) PS ix Wa ee 3 Ehuinass hase 3. Some Aspects of the Morphology, Phylogeny and Taxonomy of the South African Inidaceae.* By G. J. Lewis, B.A., PH.D. (With Plates II-X XIV) INTRODUCTION For the taxonomist the South African members of the Iridaceae present many problems and a study of this family has shown that in many cases it is impossible to describe any large group, such as a tribe or genus, without some proviso to cover the exceptions or apparent anomalies which occur. Even the description of the family, according to Baker in the Flora Capensis and other botanists, requires some slight qualification to include a few of the exceptions which occur in two or three of the genera. There has been a considerable difference of opinion not only in the conception of the genera but also of the main groups into which the family has been divided. In 1883 Bentham and Hooker, in their “Genera Plantarum’, divided the family into three tribes, Moraceae, Sisyrinchieae and Ixieae; the first two of these tribes they subdivided into several subtribes. Most of the botanists who have worked on the classification of the Iridaceae since that date have agreed in dividing the family into three main groups, but have not agreed about the combination of the genera in two of the groups. Pax, in the first edition of Engler and Prantl’s Pflanzenfamilien (1889), raised one of Bentham and Hooker’s subtribes of the Szsyrinchieae, Croceae, to subfamily status, Cyrocoideae, and combined the remaining subtribes of Szsyrinchieae with those of Moraeeae, treating this also as a subfamily to which he gave the name Iridoideae. This subfamily he subdivided into several tribes and subtribes. The third group, corresponding with Bentham and Hooker’s Jxieae, he also raised to subfamily status and divided into three tribes. Baker’s arrangement in the Flora Capensis is the same as that of Bentham and Hooker, except that he treated the tribes as suborders and the subtribes as tribes. Diels, in the second edition of Engler and Prantl’s Pflanzenfamilien (1930), also followed Bentham and Hooker’s arrangement, treating the three main groups as tribes. The only alteration he made was to change the order in which the tribes were placed by putting Szsyrinchieae first, Ixieae second and Moraeeae third. * ‘Thesis approved by the University of Cape Town for the degree of Ph.D., 1952. MOT cli PART, 2: 16 ANNALS OF THE SOUTH AFRICAN MUSEUM Rendle, in his Classification of Flowering Plants, and Marloth, in The Flora of South Africa, both followed Pax’s arrangement. A very different arrangement is that of Hutchinson, who, in his Families of Flowering Plants, published in 1934, divided the family into eleven tribes of equal status. In 1940 Weimarck estab- lished the tribe Niventeae (Niveniineae, apud Weimarck), thus bringing the number of tribes up to twelve. In view of the various discrepancies that exist with regard to the classification of this family it is clear that the characters which have been relied on for differentiation cannot be entirely satisfactory. A distinct tendency in the past to attach too much importance to the morphology of the flower alone, while other important characters were overlooked or neglected, has led to considerable confusion. A few examples which illustrate the fallacy of relying on floral characters alone for grouping the species together are the following: Antholyza. In 1932 N. E. Brown pointed out the confusion that had resulted from a wrong conception of Linné’s genus Antholyza, and that the genus, according to Baker in the Flora Capensis, comprised members of nine different genera, having in common their conspicuous red or red-and-orange flowers with the perianth tube somewhat abruptly constricted at or below the middle. In 1941 E. P. Phillips grouped four of Brown’s genera together and so reduced the number to six. Acidanthera. In 1941, after a critical examination of the South African species of Acidanthera, I found that it was not possible to uphold this genus in South Africa as the fourteen species placed under Acidanthera in the Flora Capensis belonged to five or six different genera, the only feature common to them all being the comparatively long and slender perianth tube. Ixia. As arranged in the Flora Capensis this included members of the genera Gladiolus, Dierama, Tritonia (all these placed in Ixia on account of their actino- morphic flowers), and the monotypic genus Tanaosolen, the last placed here on account of a superficial resemblance of the flowers to those of the section Hyalis, but actually a very distinct plant. The distinct genus Exohebea was treated as a subgenus of Gladiolus from which it differs in its corm, leaves, stem, bracts, fruits and seeds, as well as floral characters, such as the marked protandry of the flowers. Tritonia. This genus, like many others, is badly in need of revision, and as it stands at present it includes two distinct genera. Species belonging to five other genera have already been removed from Tritonza. It is obvious that the structure of the flower alone does not always provide sufficient characters for differentiation and that it is necessary to take into consideration the other morphological members of the whole plant. After working on the taxonomy of this family over a period of several years it became increasingly obvious that the only way to arrive at a satisfactory solution to some of the problems was by means of a morphological study of the entire plant and by taking into consideration the various correlated morphological characters of the vegetative and floral organs. This undoubtedly provides more definite MORPHOLOGY, PHYLOGENY, TAXONOMY OF S.A. IRIDACEAE 17 indications of phylogeny than relying merely on floral characters, and it is in fact the only way to arrive at the true affinities of the genera and tribes and to place with any degree of certainty the doubtful or borderline species which appear in some cases to connect the genera. It also became apparent that although a certain amount of work has been done on the morphology of this family, a great deal still remained to be done. Arber has described and discussed many of the different types of leaf which occur in the Iridaceae. Scott and Brebner, and to a greater extent Adamson, described the anatomy and secondary thickening of some of the shrubby genera. Ingeborg Haeckel worked on the inflorescences of some of the genera and Pax mainly on the morphology of the flowers. Marloth, who spent so much time studying and making observations in the field, made many interesting and helpful remarks about the morphology of some of the genera, many of which are illustrated, and another South African botanist who has studied this family in the field and by her comments has added much to our knowledge is L. Bolus. Two botanists outside South Africa who, during recent years, have done some valuable work on three genera of the South African Iridaceae, namely Weimarck, in his monograph of Aristea and revision of Nivenia, and Foster, in his revision of Geissorrhiza, have both discussed some of the morphological features of these three genera, especially the former, who discussed in considerable detail some of the different types of inflorescence which occur in Aristea, Nivenia and some allied genera. Weimarck included with Aristea the plant which was formerly treated as a separate monotypic genus, Cleanthe, which was separated from Aristea only on account of the difference in the size of the inner and outer perianth segments. It was found that the works mentioned above did not offer any explanation for some of the structures observed in certain genera of the South African Iridaceae, and a few of these morphological features which I have examined are discussed in the following paper. Various processes of reduction and adnation suggest an explanation for some of the anomalous features. In making a morphological study of some members of this family the taxonomist is confronted by one or two problems concerning which the morphologists, although they have propounded various theories, have not succeeded as yet in finding any conclusive evidence. These are mentioned in connection with the various organs concerned. The paper is divided into four parts. In the first the structure of some of the leaves and prophylls is discussed. Part II concerns the bracts and inflorescence. In Part III the subterranean stems are described and discussed. Part IV is divided into two sections; the first deals rather briefly with the flowers and one or two controversial taxonomic groups, and the second with the phylogeny and classification. A bibliography is given at the end of Part IV. For the sake of convenience the names Iridoideae, Crocoideae and Ixioideae, given by Pax to the three main groups variously treated as subfamilies or tribes by the botanists mentioned above, are used. 18 ANNALS OF THE SOUTH AFRICAN MUSEUM ACKNOWLEDGEMENTS The writer wishes to express her sincere gratitude to Prof. R. S. Adamson, who was in charge of the Department of Botany at the University of Cape Town when this research was commenced. Although he retired from the University a few months later, he kindly continued to assist with advice up to the final preparation of this paper for publication. Acknowledgement is made to the Fourcade Fund, University of Cape Town, for a grant towards the cost of publication of this paper. ParTI. ‘THE LEAVEs 1. Introductory remarks The most disputed organ in the Monocotyledons is almost certainly the leaf, and several theories concerning the morphological value of this organ have been advanced. It is not proposed to discuss these various theories in this paper as this subject has been dealt with in great detail by Arber, who also devoted much attention to the several different types of leaf which occur in the Iridaceae. My remarks will be confined to a few features observed in some of the leaves, bracts and prophylls examined which have not been discussed by Arber. As a result of my investigations I am inclined to accept Arber’s ‘phyllode theory’ as applied to the leaves in this family, that is, that the leaf of the Iridaceae has no true lamina but represents in every case either a petiole and leaf base or a leaf base alone. ; With the exception of the three shrubby genera, Klattia, Nivenia and Witsenza, and of the genera Aristea, Bobartia, Dietes, Pillansia and Dierama, all of which are evergreen, the South African Iridaceae are geophytic and in the majority of the genera the leaves and inflorescences are produced together and wither and disappear together at the end of the season. In the western Cape Province ‘hysteranthous’* species occur in a few of the geophytic genera, such as Gladiolus, Anapalina and Exohebea. These so-called hysteranthous species flower during the dry late summer or early autumn months, and in nearly all of them only one basal leafis produced and the cauline leaves present at the time of flowering are very much reduced. Except for a few of the ‘hysteranthous’ species the leaves in all cases have a sheathing base and in most of the genera they are of the so-called isobilateral equitant type. In some species they are terete and in several genera in the Jridoideae they are dorsiventral. Exceptions to the typical sheathing base are found in two species of Anapalina and one Exohebea which have a very distinct, slender, solid ‘petiole’ and an expanded pseudo-lamina. These, as will be shown later, are derived from the typical isobilatera] equitant type of leaf. * The term ‘hysteranthous’, often used to describe these leaves, is almost certainly incorrect as the leaves really precede the flowering shoot. They nearly always start to appear as soon as the plant has flowered, but they arise from an axillary bud near the top of the old corm which develops into the new corm and bears a terminal flowering shoot after a dormant period. MORPHOLOGY, PHYLOGENY, TAXONOMY OF S.A. IRIDACEAE 19 As the structure of the mature leaves does not reveal any evidence of their ontogeny it is necessary to look for the earliest stages which can be found in the development of these organs. For this purpose various bud cones and shoots were examined and in one or two of those selected some very early stages were found which seem to offer some explanation for the development and structure of the leaves. One of these is Witsenta maura, one of the shrubby genera, in which some of the prophylls could just be differentiated on the axillary vegetative bud cones on some of the branches. The other plants to which special attention was paid were two or three species of Anapalina and Exohebea in which rudimentary cauline leaves are present on the flowering axis. Some of the smallest of these, just below the spike, gave some indications of development similar to that of the Witsenia prophylls. 2. The development of the prophylls and leaves of Witsenia maura and the presence of a laminar rudiment in the early stages. In looking for indications of the ontogeny of the leaves the prophylls of Witsenia maura were examined in various stages, from when they could first be differentiated, at which stage they were little more than 1 mm. long, until they reached their full development, when they varied from 6 to 10 mm. in length. As these prophylls represent the first leaves borne on an axis in which it was possible to observe one or two very early stages in development they are of particular interest. The fusion of the prophylls is discussed later in connection with the bracts and inflorescence in Part II of this paper. There is no doubt that the prophylls in Witsenia maura, as in most of the genera of the Iridaceae, are two more or less united foliar members. The youngest buds of Witsenia maura to be examined consisted only of a bud cone and two immature prophylls, the latter about 1:25 mm. long, and there was some evidence of the prophylls being united as well as what appeared to be a rudimentary lamina separated from the leaf base in each of the prophylls. A much enlarged abaxial view of one of these very young buds, removed from the axis, is shown on pl. II, fig. 1. The bud cone is situated between the two prophylls which are shown in side view in the figure. On the adaxial side the overlapping margins of the sheaths are united and under the microscope it could be seen quite clearly near the top that such a fusion had taken place. The united overlapping margins near the top are indicated by small dotted lines; the larger dotted lines represent the free posterior margins of the sheaths which in some cases overlap near the base but are not united. At the base, outside each of the prophyll sheaths, a very distinct protuberance was present, resembling the rudimentary lamina which, according to Eichler, is first differentiated from the primordial leaf. In the next stage examined it was found that this protuberance was carried upwards for a short distance, adnate to the base of the colourless sheath of the prophyll, but after this no further development takes place in this region. In mature prophylls there is sometimes a slight indication of the minute dorsal projection which is evident in the very early stages, but in mature leaves there is no sign of it. Amin, ©.) Air Nigis) aol, Dal: ie Plate II 1-4, Witsenia maura Thunb. 1, young prophylls starting to develop, abaxial view, X 40; i, bud cone; 11, sheath; iii, laminar rudiment. 2 and 2a, immature prophylls, x2. 3, prophylls, side view, 24; 1, bud cone; u, sheath; iii, short monofacial tip; iv, laminar rudiment; v, main lateral vein; vi, median vein; vii, bud. 3a, plan showing position of prophylls and leaf of shoot. 4, diagram of mature prophylls; i, sheath; ii, monofacial tip; iii, laminar rudiment; iv, main lateral vein; v, median vein. 5, Klattia partita Baker, part of old stem and leaves; 5a, part of young stem and leaves, «14; i, leaf base; ii, absciss layer; 5b, prophylls, adaxial view, x 34; 5¢, prophylls, side view, «34. 6-7, ivenia Stokoe: (Guthrie) N. E. Brown. 6, prophylls with rudimentary lamina, side view, x 23. 7, part of stem and leaves; i, internode of bud adnate to axis. MORPHOLOGY, PHYLOGENY, TAXONOMY OF S.A. IRIDACEAE 21 On pl. II, fig. 2, is shown a pair of half-developed prophylls, removed from the axis and more or less flattened. In these the limbs have started to develop and in the smaller prophy]ll on the right there is still a very distinct protuberance, no longer at the base but extending up nearly to the middle of the sheath, while in the larger prophyll on the left there is a mere trace of it near the base. The shaded part is green and the remainder of the two sheaths is colourless. In fig. 2a is shown another pair of prophylls at a slightly later stage, with a minute protu- berance near the base of each. A still later stage in the development of the prophylls is illustrated in fig. 3. At this stage the prophylls are almost fully erown and the first leaf of the bud above the prophylls had started to develop; its position is shown in a diagram in fig. 3a, where the axis is marked with a cross and the bud, with its united prophylls and first single leaf, is enclosed within the sheath of the outer subtending leaf. This leaf above the prophylls was removed from the bud cone which was cut in half, so that the prophylls in fig. 3 are shown separated and in side view. The first vascular strands to enter the prophylls from the axis cylinder are the main lateral ones and in the prophyll on the right in fig. 3 only these main lateral vascular strands are present. In the larger prophyll on the left the median strand is also present and the limb extends above the sheath into a very short monofacial tip, flattened in the vertical plane, in fact the beginning of the so-called ensiform isobilateral type of leaf which occurs in Witsenia and so many other members of this family. It can be seen at this stage that the main lateral veins enter the apex before the median vein which, at the base, within the bud cone, is attached to an axillary bud. ‘This is the only sign of a bud in the axil of these prophylls but it does not develop and the prophylls in Witsenia, as in many of the other genera, are always sterile. In the smaller prophyll on the right there is a fairly distinct protuberance near the middle and in the one on the left a slight trace of it. A diagrammatic representation of one of the largest prophylls found (1 cm. long) is given in fig. 4. A few prophylls from some herbarium specimens of Alattia partita and Nivenia . Stokoe, two other members of the small group of shrubby South African Iridaceae, were examined for comparison. In Klattia partita the two sheaths overlap much further and are almost, or sometimes completely, united, having a single common apex, and it is not so evident that the prophylls are two separate foliar members. At the base of each is a projecting portion corresponding with the protuberances in the Witsenia prophylls. ‘Two views of these are shown on pl. I, figs. 56 and 5¢. In Nivenza Stokoe: the prophylls are also very closely united and on the poscerior side the overlapping margins at the base are united, forming a closed sheath round the bud cone. In one of those examined the structure which appears to be the rudimentary lamina had developed as far as the top of the prophyll, adnate to it for three-quarters of its length but with the upper quarter free. A side view of this is shown in fig. 6, on pl. IT. These protuberances may actually be traces of a rudimentary lamina. As far as can be observed in the prophylls, the petiole starts to develop and the laminar rudiment is carried up for a very short distance but no further development 22 ANNALS OF THE SOUTH AFRICAN MUSEUM takes place in the region of the laminar rudiment which gradually disappears. At about the point where the growth in this region has stopped the median vascular strand diverges out towards the dorsal margin and the growth continues in the dorsal area which is now laterally flattened and without a laminar rudiment. If this is correct then it bears out Arber’s phyllode theory, for the structure which continues to develop and forms the limb is petiolar, with a flat dorsal wing-like region and no laminar rudiment. As has been shown in the case of the prophylls, the protuberance begins to disappear at an early stage, and in mature leaves there is no trace of it. There are two possible interpretations for the sheath which is so conspicuous in the early stages in the development of the W2tsenia prophylls. One is that it represents the whole of the remainder of the primordial leaf, that it is in fact the leaf base, and the other that it is a stipular or ligular ‘axillary’ outgrowth from the leaf base, and that the petiole develops outside but completely adnate to this sheath. The position of the protuberance which possibly represents the laminar rudiment and the development of the prophylls in Witsenza maura, also the fact that an absciss layer is formed near or at the base of the limb in Witsenza, Klattia and Nivenia, seem to suggest that the latter is the correct explanation. This view perhaps appears similar to that maintained by Gluck, who, according to Arber, regarded the sheaths of the Monocotyledon leaves as ‘stipulae adnatae’ which have arisen by the fusion of a pair of free stipules with the base of the petiole; but my hypothesis is not the same, for judging by the manner in which the petiole starts to develop outside the sheath in the Witsenza prophylls, the sheath is definitely not a pair of lateral stipules united to the petiole but a single stipule or ligule of the axillary type which is not uncommon in the Monocotyledons and is found in a few Dicotyledons as well. In the fully developed leaves this ligular sheath is entirely or almost entirely enclosed within and adnate to the amplexicaul base of the petiole. There seems to be a possibility that the structure of these leaves, in which the main lateral veins enter the leaves from the axis cylinder before the median vascular strand, is associated with the partial adnation of the buds to the axis. In some of the shrubby species it is apparent that the lowest internode of the bud is to a greater or lesser extent adnate to the axis. In Nivenza Stokoe: the buds are adnate to the axis for a considerable distance above the base of the subtending leaf, in some specimens as far as the next leaf above. ‘This was most conspicuous in a particularly robust specimen of WV. Stokoe: with longer internodes than are usual in this species. A portion of this axis is shown on pl. I], fig. 7, with two of the leaves removed so that the buds can be seen. The branching of Witsenta maura was examined and in this genus also it is evident that the first internode of each branch is adnate to the axis. This is shown diagrammatically on pl. III, fig. 1. Here, as in the rhizomes and corms of the Iridaceae, the branching is sympodial. After the main axis has flowered the buds at two consecutive nodes near the base of the stem develop and form lateral branches. Neither Scott and Brebner nor Adamson, in their descriptions of the anatomy of some of the shrubby Iridaceae, mentioned this adnation, and ‘Ann’ S. Afr. Mus. Vol. XL. Plate III 1-5, Witsenia maura Thunb. 1, diagram of stem with two branches partly adnate at the base. 2, apex of stem, with lower leaves removed; 2a, lower part of outer leaf shown in fig. 2, <3; i, internode; 11, node; ii, leaf. 3, part of old stem with bases of two leaves; i, bud; ii, trans- lucent margin; iti, leaf base; iv, absciss layer. 4, transverse section of internode at top of stem, X10; 4a, t. s. of leaf base, x 10; i, median vein; 40, t. s. of base of leaf, x 54; i, meristematic tissue; il, sclerenchyma; ili, median vein; iv, secondary vein; 4c, t. s. about 5 mm. above base of leaf shown in fig. 2, X 10; 1, median vein; ii, main lateral veins; iii, secondary veins. 5, part of meristematic tissue, X c. 1353; 1, sclerenchyma; ii, rudimentary vascular strand. 6, Aristea oligocephala Baker, bractecle. 24. ANNALS OF THE SOUTH AFRICAN MUSEUM the whole matter requires investigation. Adamson referred to the fact that the central parts of the stem are formed before the outer parts, and that the laying down of the outer parts and the thickening of the outer ground tissue takes place after the complete differentiation of the central portions. This delay in the formation of the buds and the subsequent elongation of the internodes with the still undifferentiated bud tissue adnate to the axis, probably due to intercalary growth commencing at an early stage, is possibly connected with the delay in the appearance of the median vascular strand which, instead of passing directly into the outer region of the leaf base, remains adnate at the base to the axillary bud and only diverges out towards the dorsal margin after the leaf has started to develop and the main lateral veins have extended up to the apex. The median vascular strand in the leaves of the Iridaceae is nearly always very small in comparison with the lateral veins and in some of the dorsiventral leaves there is no sign of it. The internodes just below the apex of the stem in Witsenca maura are extremely short, but in the older part of the stem they are fairly long, considerably longer than in Klattia or Nivena. The apex of a stem of W. maura showing part of the upper leaves and two internodes and nodes (with the lower leaves removed) is figured on pl. III, fig. 2. Fig. 2a is the upper of the two internodes and base of the outer leaf, enlarged to show the veins in the base of the leaf, the dotted lines indicating secondary veins. ‘Two longer internodes near the base of the same stem are shown in fig. 3. In all these internodes there is a narrow, colourless, and somewhat translucent margin to the axis just below the dorsal region of the leaves, which extends up for a short distance into the limb before it gradually disappears (marked with dotted lines in fig. 3), and on the other side of the elliptical axis is a similar but shorter and narrower wing-like area on the axis below the node. In the short internode examined at the top of the stem the outgrowth of the axis below the dorsal margin of the leaf appears to correspond with the position occupied by the base of the protuberance in the prophylls. In the longer internodes it is the same, but owing to intercalary growth the basal portion is decurrent on the axis throughout the whole internode, while the upper part can be seen to extend up for a short distance on the dorsal margin of the limb. Some transverse sections of the stem and bases of young leaves of Watsenia maura were examined and, as in the case of the prophylls, it was found that the first vascular strands to leave the axis cylinder and pass out into the base of the limb are lateral ones. Some of these sections are shown on pl. III. The first, fig. 4, is a section of an internode in which two lateral bundles are starting to leave the central cylinder. In fig. 4a two lateral bundles have entered the arms of the limb which are already separated from the axis, but the remaining bundles have not yet left the axis cylinder in the region where the leaf base is still attached, and the median strand is still some distance inside the cylinder. In the lower half of this section the vascular strands which will enter into the next leaf above are oblique as they are just leaving the centre of the cylinder MORPHOLOGY, PHYLOGENY, TAXONOMY OF S.A. IRIDACEAE Pus and curving out to the margins. (This corresponds with the course of the bundles in Nivenia corymbosa (Aristea corymbosa) described by Scott and Brebnevr.) The third section (fig. 40) is through the base of one of the slightly older leaves, immediately above the node. The median and twelve lateral vascular bundles which have entered the leaf from the axis are present, also a few traces of secondary strands and, near the dorsal margin, a slightly larger vascular bundle. This dorsal bundle is not the median one and has not come from the central cylinder but arose in the tissue in the wing-like projection of the axis (in this section it is the triangular region outside the semicircle of main vascular bundles). The tissue in this area was found to be meristematic; asmall portion, very much magnified, is shown in fig. 5. A few strands of sclerenchyma are formed, especially towards the margins, and also a few rudimentary vascular strands, one of which, with a distinct xylem vessel, is present in the portion illustrated. A few of these bundles were slightly larger than the one shown, the largest being the one near the dorsal margin. A few similar rather rudimentary bundles are formed in the external wing- like portion of the leaf base below the two arms of the limb which, as can be seen in fig. 4a, are not supplied with vascular bundles from the axis cylinder. The small rudimentary bundles in the margins are not connected with the main vascular strands and seem to disappear towards the top of the sheathing part of the limb where they are apparently replaced by the primary lateral vascular strands which diverge slightly outwards towards the margins (see figs. 2a and 3 on pl. III). Shortly above the base the median vein starts to diverge out towards the dorsal margin. A fourth section of a young Witsenza leaf is illustrated in fig. 4¢ on pl. I1I, which shows the median and two of the main lateral veins starting to diverge outwards and might be compared with the figure of the prophyll on pl. I, fig. 3, in which the course of the median vascular strand is shown. In fig. 4c (pl.. III) the shaded parts represent the assimilatory tissue and the fairly numerous secondary veins are also shown. The outer dorsal meristematic tissue is starting to disappear and no trace of it was found in sections cut near the middle of the leaf. Near the base of the leaf a fairly distinct absciss layer is formed as the leaves develop. The position of this layer is marked on the leaf in fig. 3. Scott and Brebner mentioned that there were distinct indications of such a layer in the leaves of Nivenia (Aristea) corymbosa. In Klattia partita it is still more evident (see pl. II, figs. 5 and 52), and in this plant the whole amplexicaul base extends up for a short distance above the node, the lateral arms more than the dorsal region, and it is persistent on the axis after the leaf has fallen, as can be seen in the lower part of the stem in fig. 5 on pl. II. The formation of an absciss layer near the base of the limb seems to suggest that it is at this point that the leaf base ends and the leaf, or rather the petiole, begins, and lends some support to my hypothesis that the petiole has developed outside and completely adnate to an axillary ligule. With the exception of some 26 ANNALS OF THE SOUTH AFRICAN MUSEUM of the dorsiventral leaves, probably nearly all the leaves in the Iridaceae have developed in a similar manner. In this respect my views are not entirely in accordance with those of Arber who regarded the sheathing portion in all cases as the leaf base, whereas if my interpretation of the sheath is correct, then the whole limb above the absciss layer, which includes the sheathing portion of the limb, is petiolar, the sheathing base being due to the presence of the ligule. In the Witsenia prophylls no development takes place in the marginal regions of the leaf base and only the dorsal part of the limb develops. The development of the bracts and bracteoles of some species of Aristea is very similar, as a comparison of the bracteole of Aristea oligocephala, illustrated on plate III, fig. 6, with the prophylls of W2tsenia maura (fig. 2 on pl. II) will show, allowing for the fact that the Witsenia prophylls are two partly united members whereas the Aristea bracteole is a single member. The membranous apex of the sheath in the Aristea bracts and bracteoles is ruptured as the flower-buds develop and the single axillary sheath thus appears to be two lateral wing-like stipules. The leaves of Nivenia Stokoe: appear to be exactly the same as those of NVivenza corymbosa. I have not examined either of these in detail but the latter were described by Scott and Brebner and according to their description they are very similar to those of Witsenia maura. Scott and Brebner stated that there were two median bundles lying one behind the other in the same radial plane, and also referred to the presence of isolated strands of sclerenchyma in the leaf, especially toward its edges. There can be little doubt that the outer of the two bundles which they called median corresponds with the outer dorsal bundle in Witsenia maura. According to them the inner of the two median bundles (i.e. the true median bundle), on entering the stem from the leaf, turns sharply upwards, then sharply down again. This is probably due to its attachment to the base of the bud which in Nivenia corymbosa, as in N. Stokoez, is carried up above the base of the leaf. In Nivenia Stokoe: and other species of Nivenza the dorsal region of the leaf base is separated from the two lateral arms and the limb develops obliquely with the base of the dorsal part decurrent on the axis (see pl. II, fig. 7). This may have something to do with the stage at which intercalary growth commences and is probably connected with the way in which the bud is carried up on the axis. The winging of the aerial axis in the plane of flattening of the limb which, as Arber has remarked, is associated with the equitant leafin some of the Ividaceae, is very conspicuous in several species of one or two genera, notably Aristea and Lapeyrousia. 3. The structure of the rudimentary cauline leaves and the bracts of Exohebea and Anapalina In the majority of species in the closely related genera Anapalina and Exohebea there are rarely any lateral branches and the cauline leaves are very much reduced, so much so in some species that there is only a trace of them, sometimes less than 1 mm. long. Several of these rudimentary leaves were examined in Ann. S. Afr. Mus. Vol. XL. Plate IV Ie 5a 1, diagram of typical herbaceous bract with reduced tricuspidate bract in dotted lines in lower half. 2, Tritonia scillaris ( L.) Baker, membranous tricuspidate bract, x 2. 3, Ixia micrandra Baker, membranous bract, X 14. 4, 4a and 4b, Anapalina revoluta (Burm.) N.E. Br., rudimentary cauline leaves, <2 1, laminar rudiment; ii, prophylls. 5, Exohebea Dodi (Lewis) Foster, rudimentary cauline leaf, adaxial view, X30; 1; ‘bud cone; i, laminar rudiment; iii , prophylls adnate to sheath of leaf (v); iv, bud in axil of prophyll; 5a, same, abaxial view, x 30; 1, laminar rudiment; ii, prophyll; ti, sheath of leaf; 5b, rudimentary leaf, more advanced stage than puaaeC Oy, Lom 7e ia "short monofacial apex; ii, ? vestigial trace of laminar rudiment; ii, “prophyll: 5¢, ? vestigial trace of same laminar rudiment, x c. 65. 6, Anapalina revoluta, bract and bracteole, x 14; ian bract.<1e5 GD, bracteole, x 14; 1, prophylls adnate to bract and bracteole. 28 ANNALS OF THE SOUTH AFRICAN MUSEUM specimens of Anapalina revoluta, Exohebea Dodii and one or two other species and two or three very early stages in development were found which seem to confirm my observations regarding the development of the prophylls in W2tsenia maura, that is, that a laminar rudiment is differentiated from the leaf primordium at the beginning but no further development takes place in this region and it very soon disappears. In some of the smallest leaves near the top of the stem in Anapalina revoluta (just below the spike) a minute cusp-like projection was observed between two somewhat flattened wing-like structures, the whole less than 1 mm. long. This first stage is shown in figs. 4 and 4a on pl. IV. In the second stage (shown in fig. 4b) there is a green and slightly fleshy swelling at the base and the central cusp is brown. It was thought at first that the wing-like structures on either side of the cusp were probably free lateral stipules which became adnate to the limb as it developed but a further examination proved that this was not correct. A rudimentary leaf from a specimen of Exohebea Dodit was removed from the stem and when examined under a low-powered microscope it was found that the lateral wing-like structures are the prophylls of the axillary bud and that they are almost completely adnate to the sheath of the rudimentary subtending leaf. ‘Two views of the whole structure are shown on pl. IV, the adaxial side in fig. 5 and the abaxial in fig. 5a. A bud cone is situated between the two prophylls and in the axil of each prophyll a minute bud is present (see fig. 5). Although the prophylls are united to the sheath of the leaf they are entirely free from each other and on the adaxial side they are widely separated. Outside the sheath to which the prophylls are adnate is the small cusp-like protuberance which is possibly the laminar rudiment, and which corresponds more or less with the protuberances found at the base of the prophylls in W2ztsenta maura. The protuberance at this stage has already turned brown and begun to shrivel. This can be seen in fig. 5a. At a slightly more advanced stage, in which the petiole had started to develop and extended above the sheath in a very short monofacial flattened tip, there was no visible sign of the cusp-like protuberance found in the very early stages, but in one of the leaves of Exohebea Dodi examined under a low-powered microscope a minute hair-like structure was observed a short distance below the apex. The position of this structure is shown in fig. 56 on pl. IV, and the minute projection, very much enlarged, in fig. 5c. Until there has been an opportunity of examining a great many more of the leaves in these genera it is not possible to state with any certainty that this is actually a vestigial trace of a rudimentary lamina which has been differentiated from the leaf primordium but it seems to be a possible explanation for this unusual hair-like outgrowth in a plant which is entirely glabrous. The stage of development reached by this particular leaf is about the same as that of the left-hand prophyll of Witsenia maura shown in fig. 3 on pl. II, and the position of the minute hair-like projection more or less corresponds with the position of the projection on the prophylls, which is most apparent in the prophyll of Nivenia Stokoe: shown in fig. 6 on pl. II, in which MORPHOLOGY, PHYLOGENY, TAXONOMY OF S.A. IRIDACEAE 29 there is a small brown dorsal outgrowth separated from the remainder of the prophyll. The bracts and bracteoles of most species of Hxohebea and Anapalina, as well as those of the closely related monotypic genus Tanaosolen, are cartilaginous and usually brown at the time of flowering. The bracteoles of these genera differ from those in all the other genera in the /xiozdeae in being longer than the bracts, sometimes, as in Anapalina triticea and A. Burchellii, very considerably so, but apart from this the bracts and bracteoles in these genera are almost identical. A feature which can be observed in most of them, for which at first it was difficult to find an explanation, is the way in which the fairly distinct outer margins which appear to be stipular are folded in at the base of the bract and arise between the axis and the bracteole. The bract and bracteole of Anapalina revoluta are shown on pl. IV; in fig. 6 they are attached to the axis and in figs. 6a and 6) removed and flattened out. The development of the minute cauline leaves below the spike provides an explanation for the structure of the bracts and there can be little doubt that, like the cauline leaves, the bracts have developed adnate to the prophylls of the axillary bud. This will explain the thick texture of the bracts, and why the margins arise between the axis and the bracteole, and also offers an explanation for the anomalous bracteoles in these genera which, as well as being longer than the bracts, differ from those of the other genera with a spicate inflorescence in not being bifid or bicarinate. In the genera Anapalina and Exohebea, as well as in the closely allied genera Tanaosolen and Tritoniopsis, the prophylls are free from each other but united to the bract, and the bracteole is therefore almost certainly a third foliar member produced on the reduced lateral shoot, above the two prophylls and below the flower. Like the bract, this bracteole has also developed adnate to two prophylls. The origin of the bifid or bicarinate bracteoles in the other genera with a spicate inflorescence is discussed later in connection with the inflorescence. 4. The tricuspidate bracts The tricuspidate bracts which are so conspicuous a feature in Tritona subgenus Dichone, as well as many species of Jxza and one or two other genera, all in the Jxiozdeae, cannot, in my opinion, be regarded as providing evidence of lateral stipules. As far as it has been possible to ascertain, the tricuspidation is due to reduction, the upper half of the bract not being developed. As a result of this the median and two main lateral veins which are derived from the axis cylinder remain fairly widely separated instead of converging towards the apex. This is shown in a diagram in fig. 1 on pl. IV. The large member represents the type of comparatively large, herbaceous bract found in several genera in the Ixioideae, such as Gladiolus, Watsonia and Babiana, with only the median and two main lateral veins marked, the other fairly numerous and uniform veins present in these bracts being omitted. A line is drawn across the middle and in the lower half, in dotted lines, is indicated the type of reduced bract found in Tnitoma scillaris, T. trinervata and a few allied species. The actual small reduced 30 ANNALS OF THE SOUTH AFRICAN MUSEUM bract of Ttoma scillaris, which is about 4. mm. long, is shown in fig. 2 on the same plate, and a bract of [xa micrandra (5-6 mm. long) in fig. 3. In the slightly larger bract of the latter, in addition to the median and two lateral veins there are two more lateral veins present and traces of six secondary veins starting to develop at the base. The bracts of Trztonza scillaris and allied species are membranous and colour- less, those of [xia micrandra being herbaceous only at the base where the short intervening secondary veins are present. From this it might perhaps be deduced that the bracts in Gladiolus, Watsonia, Babiana, etc., represent the base of a petiole, while the small reduced bracts of Tritonia scillaris, etc., are only the ligular outgrowth of the leaf base, with the three (or in Ixia micrandra five) vascular bundles derived from the axis cylinder, the small intervening veins present at the base of the bract in [xza micrandra only appearing as the base of the petiole begins to develop. This seems to correspond more or less with the early development of the prophylls in Witsenia maura (except for the presence of the laminar rudiment in Witsenia), and conforms with the hypothesis that the petiole in these plants, which all have the isobilateral equitant type of leaf, has developed outside and adnate to a ligular outgrowth from the leaf base. The bracts of Sparaxis and Synnotia which are scariose but longer than those of Ixia and Tritonia subgenus Dichone, are probably derived from the same reduced type of bract but with the veins more strongly developed and projecting up in three or five acuminate points. 5. Lhe pseudo-lamina of Exohebea flexuosa and Anapalina triticea In addition to the unusual character of the cauline leaves of most species of Anapalina and Exohebea, the basal leaves of three species in these genera at first sight appear to differ so markedly in structure from all the other leaves in the family that they deserve mention. The ‘foliated’ leaves of the genus Babiana have been described by Arber, who pointed out that although they appear to be divided into a fairly conspicuous ‘lamina’ and ‘petiole’, they do not really differ fundamentally from the ensiform type, and the same is true of the leaves of Anapalina triticea, A. Burchell and Exohebea flexuosa (L.f.) Lewis (Gladiolus flexuosus L.f.), all of which have an expanded laterally flattened pseudo-lamina sharply marked off from a solid, slender petiole-like base. A leaf of A. triticea is illustrated in fig. 1 on pl. V and two leaves of E. flexuosa in figs. 3 and 3a on the same plate, the one with an obtuse and the other an acute apex. The leaves of the latter, with their slender ‘petiole’, comparatively broad and short pseudo- lamina with undulate margins and only two prominent veins, and in many cases an obtuse apex, are unique in the family. In A. trtecea and A. Burchelli there are either two or three or occasionally four prominent veins; only two were present in the leaf illustrated. In most species of Anapalina and Exohebea three or four basal leaves are produced and are present at the time of flowering, but in mature plants of A, triticea, A. Burchellit and E. flexuosa only one basal leaf is produced and this is 1 mmn. >. Afr. Mus. Vol. XL. Plate V 1, Anapalina triticea (Burm.) N. E. Br., basal leaf, x4. 2, A. revoluta (Burm.) N. E. Br., basal leaf, x4. 3 and 3a, Exohebea flexuosa (L. f.) Lewis, basal leaves, x4. 4, E. flexuosa, transverse section of base of petiole, x 35; 1, median vein; 1i, main lateral veins; iii, external lateral veins; iv, fibres; v, phloem; vi, xylem; 4a, t. s. of base of pseudo-lamina, x 30: i, median vein; ll, primary lateral veins; i11, secondary veins; iv, transverse lateral vein; 40, t. s. of middle part of pseudo-lamina about 2 mm. above previous section, x 30. 5, Exohebea lata (L. Bol.) Foster, t. s. of sheathing base of leaf, x 30, 32 ANNALS OF THE SOUTH AFRICAN MUSEUM ‘hysteranthous’. In all except these three species the linear or linear-ensiform ‘blade’ tapers to a long, narrow sheathing base. ‘The corm in these genera and in the closely allied genera Tanaosolen and Tritoniopsis is deep-seated, and the basal leaves therefore arise a few inches below the surface of the ground, as is also the case in the genus Babzana. On comparing the anatomy of the leaves of A. triticea, A. Burchell and E. flexuosa with that of some other members of the same genera it was found that they are very similar, the only significant difference being the disappearance of the sheath. As a result of this the basal part is solid and the upper laterally expanded pseudo-lamina develops equally on either side instead of somewhat obliquely, as is the case when the sheath is present. (Compare the figures of the leaves of A. triticea and A. revoluta on pl. V, figs. 1 and 2.) Some plants of E. flexuosa were grown at the South African Museum Herbarium for observation so this species has been studied in more detail than any of the others. On some of the young cormlets which had not yet produced an inflorescence there were two or three very small leaves and signs of a very much reduced sheath in these. In mature leaves of E. flexuosa the slender basal ‘petiole’ is solid, with no opening or sheathing region on the ventral side. ‘The disappearance of the sheath in this species is almost certainly a recent development as is indicated by the fact that there are still some signs of a sheath in the immature leaves of young cormlets, and also by the fact that in the ‘petiole’ of the mature leaves the very small external lateral veins are still separate. The median and external veins are very small in comparison with the two main lateral veins on either side. The most conspicuous feature in the large lateral veins is the mass of scleren- chymatous tissue which entirely surrounds and encloses the xylem and phloem. Between the veins is a small amount of ground tissue consisting of rounded cells; towards the top of the ‘petiole’, in the short portion which is above the ground, these cells contain chlorophyll. A section near the base of the ‘petiole’ is shown in fig. 4 on pl. V and for comparison a section of the slender basal sheathing part of a leaf of Exohebea lata is given in fig. 5. Apart from the narrow opening of the sheath the basal part of the leaf of FE. lata differs from that of E. flexuosa in having three large lateral veins on either side of the median vein, as well as five or six minor veins. At the top of the ‘petiole’ in E. flexuosa the rounded cells of the ground tissue on either side of the main lateral veins become laterally elongated as the small median and external lateral veins (the latter now united) curve outwards and give rise to two or three small secondary veins which are connected with each other and with the median and external lateral veins by small transverse veins. The two pairs of large main lateral veins become united to form two pairs of collateral veins, with the xylem of each separated by a band of sclerenchyma. At the base of the pseudo-lamina the parenchyma cells between these two united pairs of main lateral veins are still rounded, as is shown in fig. 4a on pl. V, but about 2 mm. above the base they also become laterally elongated as the main lateral veins diverge outwards. The small secondary veins which arise in this MORPHOLOGY, PHYLOGENY, TAXONOMY OF S.A. IRIDACEAE 33 middle portion of the pseudo-lamina are connected with the large primary veins, as well as with each other, by transverse lateral veins. After curving outwards near the base of the pseudo-lamina the prominent primary veins remain widely separated, then in some leaves they converge gradually towards the apex, in which case the leaf is acute, or somewhat abruptly, in which case the leaf is obtuse. The course of these prominent veins can be seen in the figures of the two leaves on pl. V, figs. 3 and 3a. The small median and secondary veins are not marked in these figures as they are only faintly visible in dried material and are not perceptible in the living state. Fig. 46 on the same plate is a somewhat diagrammatic transverse section about 2 mm. above the base of the pseudo-lamina, showing the main lateral veins starting to diverge outwards, with two secondary veins present between them. Chloro- phyll is present in all the laterally elongated parenchyma cells, though it is -more concentrated in the outer layers, the cells of which are slightly smaller than those in the middle. A striking feature in these leaves is the large number of small transverse lateral veins connecting the xylem of the secondary veins with that of the median and main lateral veins and with each other. Traces of them are evident near the base of the ‘lamina’, as can be seen in figs. 4a and 46 on pl. V, but above the base they are very much more numerous, as can be seen in a section of the leaf on pl. VI, fig. 1. One of the secondary veins is shown in more detail in fig. 1a (pl. V1) with a transverse lateral vein connected with the secondary xylem. In the linear-ensiform leaves of E. /ata no transverse veins were found, but they are present in Anapalina revoluta, although not as numerous as in EL. flexuosa. In A. revoluta some of the very small secondary veins are united and a few of them are free, as can be seen in fig. 3 on pl. VI. Fig. 2 on the same plate is a transverse section of a leaf of Exohebea lata. The large primary veins of all the species of Exohebea and Anapalina examined are composed of two united collateral vascular bundles with a small band of sclerenchyma separating the xylem of the two bundles, and a large mass of sclerenchymatous tissue between the phloem and epidermis and extending in a narrow band around the bundles. The median and external veins have no sclerenchymatous tissue, nor have the small secondary veins which consist of two united concentric bundles, except some of those in A. revoluta in which species they are sometimes separate. The median vein of E. lata, which is identical with that of F. flexuosa, is shown in detail in fig. 2a on pl. VI. In their small size and particularly in the complete absence of sclerenchymatous tissue the secondary veins in these genera differ from those in most of the other genera in the Jxioideae where in many cases they are as Jarge as the primary veins and cannot be distinguished from them. For comparison the upper part of a leaf of Chasmanthe aethiopica is illustrated in fig. 4 on pl. VI. This type of leaf is very common in the Jxzoideae and has what Arber described as a ‘pseudo-mid-rib’ in the region of the main lateral veins. A section of part of this leaf is shown in fig. 4a, in which it can be seen that there is no difference in size between the Ann. S. AfrmeiMius UVol. Su: Plate VI on, - =F 8 recat (ti oO ii KC ({ ES CU oS oO AO EZ O° 20 off 1, Exohebea flexuosa (L. f.) Lewis, transverse section of part of pseudo-lamina about 2 cm. above base, * 30; 1, median vein; ii, primary lateral vein; iii, secondary veins; iv, transverse lateral veins; 1a, secondary vein with transverse lateral vein, < c. 135; i, primary xylem; ii, secondary xylem; ui, phloem. 2, £. lata (L. Bol.) Foster, t. s. of leaf, x 30; i, median vein; ii, primary lateral veins; iii, secondary lateral veins; iv, fibres; v, phloem; vi, xylem; 2a, median vein, x ¢. 135. 3, Anapalina revoluta (Burm.) N. E. Br., t. s. of part of leaf, x 30; numbers i—vi as in fig. 2; vii, transverse lateral veins; 3a, part of epidermis and cuticle, xX c¢c. 135. 4, Chasmanthe aethiopica (L.) N. E. Br., upper part of leaf; i, pseudo-mid-rib; 4a, t. s. of part of leaf, x 30; i, fibres; ii, phloem; ili, xylem; iv, assimilatory tissue; v, colourless mesophyll in region of pseudo-mid-rib. 5, 5a and 5b, E. flexuosa, abnormal leaves, <4; 5c, transverse section of the upper part of the leaf shown in fig. 5a, X 30. MORPHOLOGY, PHYLOGENY, TAXONOMY OF S.A. IRIDACEAE 35 numerous veins, all of which are composed of two united collateral vascular bundles, except those in the region of the pseudo-mid-rib. All the veins have a large mass of sclerenchymatous tissue outside the phloem, extending out to the epidermis. The leaves of the species of Exohebea and Anapalina examined all have a thick cuticle and in E£. lata and A. revoluta the whole surface is covered with minute papillae, as shown in a transverse section of a small portion of the epidermis of A. revoluta in fig. 3a (pl. VI). In E. flexuosa similar protuberances occur only on the dorsal and ventral margins which are slightly horny. An interesting irregularity was observed in the leaves of some specimens of E. flexuosa grown at the South African Museum Herbarium which seems to suggest that the short and broad pseudo-lamina in this species is of comparatively recent origin. The corms were collected in the Caledon Division when the leaves had just appeared but were not fully developed. After being transplanted the leaves continued to grow and attained the average, or slightly more than average, length, but the upper half which developed after the corms had been transplanted was very much narrower than the lower half, as the growth continued only in the region between the large primary veins. In one of the leaves there was a very narrow border outside each of the primary veins (see pl. VI, fig. 5@), and in another none at all (see fig. 5). A third leaf which developed after the corms were moved was linear-ensiform with almost flat margins (see fig. 55), and, except that it was considerably shorter, the ‘lamina’ of this leaf resembled that of F. lata. It appears therefore that as a result of disturbance during the growing-period these leaves reverted to the linear- ensiform type from which they originated. On comparing a transverse section of the upper abnormal part of the leaf of E. flexuosa illustrated in fig. 5a with a section of a leaf of E. lata it was found that the former differed from the latter only in having two instead of three pairs of united primary veins. The transverse lateral veins characteristic of the broad lower part of the leaf were not present in the narrow upper part. These two sections are shown on pl. VI, figs. 5¢ and 2. The structure and development of the leaves in Exohebea flexuosa lend further support to the theory that the pseudo-lamina of these and many other leaves of the isobilateral equitant type which occur in many genera of the Iridaceae arises as a winged outgrowth of the petiole. 6. The dorsiwentral leaves of Hexaglottis, Homeria and Moraea The dorsiventral type of leaf which occurs in most species of Moraea and in the closely related genera Gynandriris, Hexaglottis and Homerza, has been discussed by Arber and will only be mentioned briefly here. As Arber remarked, the leaves of many species of Moraea, etc., are prevailingly dorsiventral, but as a rule they terminate in a solid monofacial apex, which may often be reduced to a mere trace. This type of leaf Arber regards as a leaf base crowned by a petiole and there is little doubt that this is correct. In many of the species the growth continues for a considerable time in the region of the leaf base which sometimes Ann S. Afr. Mus. Vol. XL. Plate VIT \\ | \\ clipe 3 wey ; AN 1, Hexaglottis flexuosa (L. f.) Sweet, upper half of young leaf, x2; i, leaf base; ii, petiole; iii, ? laminar rudiment; 1a, transverse section of leaf base, x 17; 10, t. s. of petiole 1 cm. below apex, X50; 1, fibres; 11, phloem; itl, xylem; iv, assimilatory tissue; v, colourless mesophyll; Ic, t. s. of swollen tip 5 mm. below apex, x 40; 1, assimilatory tissue; ii, group of colourless cells; iii, colourless mesophyll. 2, Moraea neglecta Lewis, stem with inflorescence and lower half of leaf; i, leaf base; 1, stipules; iii, petiole; iv, viscid patch on stem below inflorescence; 2a, lower part of same leaf, x 13. 3, Watsonia meriana Mill., cauline leaf with one stipule; 3a, same, x4; 3), bract near top of spike with one stipule, x #. 4, W. humilis Mill., bract near top of spike with indications of two stipules, x #. DP MORPHOLOGY, PHYLOGENY, TAXONOMY OF S.A. IRIDACEAE 37 attains a great length, often up to 1 metre or more in Hexaglottis longifolia, H. virgata and H. flexuosa, as well as in a few species of Moraea and Homeria. Although the growth continues in the region of the leaf base, the growth in the upper petiolar region often stops at a fairly early stage and as a rule this portion of the leaf turns brown and disappears when the plants are still quite young. In the young leaves of Hexaglottis flexuosa and one or two species of Moraea a feature was observed which was not recorded by Arber, that is, a slightly bulbous swelling in the petiolar region just below the apex. In young leaves of H., flexuosa varying from about 6 to 12 cm. in length the upper more or less elliptical petiolar region 1s only about 1-5 cm. long and the swelling about 5 mm. below the apex (see pl. VII, fig. 1). As the leaf base elongates the whole petiolar tip dies away and in leaves between 15 and 20 cm. long only a brown, shrivelled tip remains, but even this dead tip disappears as the leaves become longer. Some sections of the three regions in the young leaves were examined and some- what diagrammatic sketches of them are shown on pl. VII. The first, near the middle of the dorsiventral leaf base (fig. 1a) is similar in structure to some of the dorsiventral Moraea leaves described by Arber (e.g. M. bituminosa), with two very small vascular strands in the mid-rib region. As remarked by Arber, a striking feature of the dorsiventral Moraea leaves is the unusual character of the mid-rib region which is thinner than the rest of the leaf and has smaller bundles which are somewhat irregularly placed, or in some cases they are absent. The second section (fig. 10) is of the petiolar region, about 7 or 8 mm. below the apex, and the third is of the slightly bulbous tip, about 4 or 5 mm. below the apex (fig. 1c). A few differences are apparent in these last two sections. In the petiolar region the ribbing, which is fairly prominent in the leaf base, is still evident, though less pronounced, while in the tip, which is nearly cylindrical with smooth margins, there is no ribbing and the veins in this swollen region are arranged slightly towards the centre and are less clearly defined than in the petiolar region where they are well defined and situated near the periphery. In the swollen tip six small patches of colourless cells are present between the epidermis and the green assimilatory tissue which are not present in the petiolar region, where these patches of colourless cells have evidently broken down and in the position occupied by four of them there are four grooves. In the region of the leaf base there are six grooves. It is possible that the tip, which is differentiated from the petiolar region, might be a laminar rudiment. The position of this laminar rudiment, if it is correct to interpret it as such, does not at first appear to be compatible with that of the protuberances which it has been suggested are probably the laminar rudiments in the prophylls of Witsenia maura, but actually it lends some support to the theories already discussed regarding the ontogeny of the leaves. If the dorsiventral leaf of Hexaglottis, Moraea, etc., is an elongated leaf base, then one would expect to find a laminar rudiment at the apex. In H. flexuosa a short petiole is interposed between the leaf base and laminar rudiment. On the other hand, in the Witsenia prophylls the laminar rudiment differentiated from the 38 ANNALS OF THE SOUTH AFRICAN MUSEUM leaf base is carried up for a short distance by the petiole as it starts to develop, though it very soon disappears, but the actual leaf base does not develop any further, as has already been discussed. In the dorsiventral leaves the petiolar region is solid and cylindrical (fide Arber) or somewhat elliptical as in Hexaglottis flexuosa, and it is very probable that in these leaves there is no ligule; it is possible that the disappearance of the ligule is associated with the development of the dorsiventral type of leaf. The leaf base entirely encircles the axis and at the base forms a short closed sheath which extends upwards for a short distance. In Moraea neglecta Lewis the dorsiventral leaf base is comparatively short (see pl. VII, fig. 2), and the greater part of the leaf consists of a solid cylindrical petiole. A suggestion, which has not so far been investigated, but which might perhaps account for the viscid patches on the axis below the nodes in Moraea viscaria, M. bituminosa, Bobartia lilacina, B. paniculata, as well as a species of Homerita and one or two species of Ferraria from tropical Africa, is that an axillary stipule, which completely encircles the stem as in Polygonum as well as one or two other Dicotyledons, is formed at an early stage in these plants, but that it becomes detached from the leaf base and disintegrates or dissolves into a viscid substance which is carried up on the axis as it develops. In two of the type specimens of Bobartia lilacina Lewis in the South African Museum Herbarium (No. 56521), instead of a viscid patch on the stem below the nodes there are in two cases the remains of very delicate membranous transparent structures which encircle the stem immediately below the nodes. One is about 1-5 cm. long and the other 3 cm. long and there are distinct traces of veins on them. These structures are too delicate to remove from the dried specimens and living material is required for investigation. 7. Stipules There are no completely free lateral stipules in the South African Iridaceae but in a few rare cases the external margins of the sheathing part of the limb, which is sometimes considerably broader than the upper part, become detached towards the top of the sheath and are prolonged slightly upwards to form what might be regarded as short stipular outgrowths. This was found in a specimen of Watsonia Meriana, but occurred only on one side of one of the lower sheaths (see pl. VII, fig. 3). At first it was thought that this might simply be due to a rupture, but in a bract near the top of the spike the same one-sided lobing was repeated (fig. 3b). In a bract near the top of a spike of Watsonia humilis two slight indications of stipular wings were found (pl. VII, fig. 4) and also in the leaf of a single specimen of Moraea neglecta (pl. VII, figs. 2 and 2a), but these represent the only indications of lateral stipules which have been observed. As has been mentioned in connection with the leaf of W2tsenia maura, the extremities of the wide amplexicaul arms of the limb are not supplied with vascular bundles from the axis cylinder and the rather rudimentary bundles which develop in MORPHOLOGY, PHYLOGENY, TAXONOMY OF S.A. IRIDACEAE 39 this region of the leaf base seem to disappear at the top of the sheath, so that the lateral extremities of the sheathing base of the leaf might perhaps in some cases be regarded as being stipular. Part II. THe Bracts AND INFLORESCENCE 1. Introductory remarks and a brief review of previous works on the inflorescence In his classification of the flowering plants Hutchinson has attached great importance to the type of inflorescence in the Monocotyledons, even more in some cases than to the long-accepted differentiation based on the position of the ovary. He considered that a classification based on the type of inflorescence resulted in a nearer approximation of allied genera. This is undoubtedly true of the Iridaceae, for in this large family the various types of inflorescences which occur are most important characters for distin- guishing the tribes, and also, in certain genera, especially in the Jrzdoideae, for distinguishing species or groups of species. In some of the larger genera in the Tridoideae, such as Aristea, Moraea, and one or two others, several different types of inflorescence occur. It is essential therefore to understand how these various forms have developed, for they provide most important clues to the phylogeny of the genera and species. An extensive examination of the inflorescences of many of the South African genera has shown that the great majority of the various types have evolved along parallel lines, and that very often links which are missing in one genus may be found in another. The interpretation of some of the inflorescences of the Iridaceae by various botanists in the second half of the nineteenth century has, in some cases, been proved by more recent workers to be not quite correct from the morphological viewpoint. This is no doubt due to the fact that their conclusions were derived mainly from the insufficient material available in the Northern hemisphere, where a comparatively small number of the genera occur, and these all fairly highly evolved. Most of the more primitive genera, in which the earlier evolu- tionary stages can be traced, occur in the Southern hemisphere. Of these a fairly large number exist in South Africa, particularly in the mountain districts of the Cape Province. ‘The Cape Province is especially favoured, as in addition to these primitive forms some of the most highly evolved are found in the coastal region, e.g. Antholyza, Anaclanthe and Anomalesia. It is only in the past twenty years that a more thorough examination of the inflorescences of some of the genera of the Southern hemisphere has been made, and some of the earlier misconceptions have been corrected and a truer picture presented of what has taken place. The most valuable work in this connection is that of Ingeborg Haeckel, published in 1931. The main basis of Haeckel’s work was an anatomical examination of the vegetative cones, following their development from the earliest stages of differentiation until after the morpho- logical values of the various parts could be determined. By this method she arrived at the following conclusions: 40 ANNALS OF THE SOUTH AFRICAN MUSEUM 1. That in many cases the buds of flowers or whole inflorescences which are apparent in the early evolutionary stages do not develop and reach maturity. 2. That the spathes or bracts in the groups of the family investigated are lateral and never terminal, as was previously considered to be the case in the Iridoideae and Crocoideae in contradistinction to the Lxioideae. 3. The most primitive inflorescence is a much-branched one, the more composite types being derived from this by various processes of reduction. In 1939 and 1940 Weimarck published the results of his investigations of some types of inflorescence in Avistea and some allied genera. On one fundamental point he does not agree with Haeckel. According to Weimarck the simplest inflorescences are those in the [xzozdeae where they are almost always composed of a single spike or in some cases lateral branches terminating in few-flowered spikelets are present as well. This he regarded as a primitive arrangement and he tried to show that the more complicated and composite inflorescences were derived from this type of spike; that from the single flower is derived the single rhipidium in such a way that the prophylls become fertile, and that the binate two- to several-flowered rhipidia typical of the genus Avistea are similarly derived from twin or binate flowers. 2. The prophylls and bracteoles Perhaps the most important point to be considered is the presence of the bifid or bicarinate prophylls in the spicate inflorescence of the Ixzoideae. One of Weimarck’s criticisms of Haeckel’s work is that she underestimated or was unaware of the importance of the bracts and prophylls which remain and, in the majority of cases, indicate that even the more or less simple inflorescence is often to be regarded as being derived from a more complicated one. The type of spike in the Iridaceae differs from a typical or simple spike in having these bracteoles and their significance was not explained either by Haeckel or Weimarck. | There has been much discussion about the bicarinate, often bifid, prophylls and bracteoles which are so common in the Ividaceae and are present in many other Monocotyledons as well, and two opposing theories have been advanced in explanation. The first is that the prophylls and bracteoles represent a pair of foliar members which have become fused together, while the second theory, favoured by many botanists, among them Arber who made such extensive studies of the morphology of the Monocotyledons, is that the prophylls and bracteoles are derived from a single leaf, the two-keeled form being due to a compression between the main axis and the branch or flower-bud. It is argued by those supporting this theory that if the prophyll or bracteole were the equi- valent of two leaves whose mid-ribs formed the keels, it would be expected to reveal its bifoliar nature by subtending two axillary buds. According to Arber, however, there is no trace of a second bud. Aan S. Afr. Mus. Vol. XL: Plate VITI 1, Babiana stricta (Ait.) Ker, bract and bracteoles, adaxial view, «14; 1a, bracteoles, x 14; 1b, part of branch with prophylls at base (adaxial view), x14; i, part of axis; ii, bract; 111, bracteoles; iv, top of ovary; v, prophylls. 2, B. plicata (L.) Ker, bract; 2a, partly united bracteoles; 20, free bracteoles. 3, B. Sprengellii Baker, bract; 3a, united bracteoles. 4, Babiana species (aff. B. stricta), transverse section of bract, x 16; 4a, t. s. of one of the bracteoles, x 16; i, hair; wu, epidermis; iii, assimilatory tissue; iv, vascular bundle; v, xylem; vi, membranous margin. 42 ANNALS OF THE SOUTH AFRICAN MUSEUM The genus Babiana is the only one of the Iridaceae known to me in which, in some of the species, each flower has two distinct and entirely free bracteoles. An examination of the prophylls and bracteoles of all the species of Babiana convinced me that these are two distinct foliar members which, in the majority of species in the genus, have become variously fused. Some of these are illustrated on pl. VIII. In B. stricta and its varieties and allied species there are two completely free bracteoles, each strongly carinate, with narrow membranous margins. At the base, on the side next to the axis, the membranous margins of the bracteoles overlap slightly, as can be seen in fig. 1 on pl. VIII, which shows an adaxial view of the bract and bracteoles of the first flower removed from a spike of B. stricta. In the case of the prophylls of a lateral branch from the same plant, shown in fig. 15, the margins do not quite meet. In some other species two stages have been observed; in some members the bracteoles are separate while in others the membranous margins are fused at the base, as in B. plicata (see figs. 2a and 2b). Various further stages of fusion are found in other species. In B. Sprengelli, for instance, the fusion is almost complete, only the two keels and bifid brown membranous tips remaining to indicate their bifoliar origin (shown in fig. 3a on pl. VIII). These various stages seem to indicate clearly that this was a case of fusion having taken place due to juxtaposition and pressure, and not a case of a single member having become divided into two free members by pressure. It remained to find some proof in support of this conviction, but none could be found in the genus Babiana. A great number of plants from several species were examined but there was no visible sign of any buds in the axils of the prophylls or bracteoles. Many sections of the single and fused bracteoles were examined but these produced no evidence of any actual fusion. The free bracteoles are smaller in size, more strongly keeled and have conspicuous membranous margins, but otherwise do not differ in any respect from the bract. Transverse sections of a bract and bracteole of an undescribed species of Babzana, allied to B. stricta, are shown on pl. VIII, figs. 4 and 4a. It was necessary therefore to turn to some other genus in the hope of finding some evidence of significance in connection with the bracteoles. With the idea in mind that the spike was derived by reduction or suppression from a more densely branched type of inflorescence, the first plant examined was Watsonia bulbifera, which provided some definite proof of a process of fusion taking place. 3. Fusion of the prophylls in the cormlets of Watsonia bulbifera Watsonia bulbifera Mathews and L. Bolus is one of the tallest and most robust members of the genus and a characteristic feature is the large number of aerial cormlets or bulbils produced in the axils of the upper sheathing leaves and bracts. As the authors remarked, the habit of producing cormlets is rare in the genus, occurring only in two or three species, and in none are the cormlets so numerous as in this. The few capsules that form (usually with one or more cormlets in the same axil) contain only one or two seeds which are perfect. Ann. S. Afr. Mus. Vol XL. Plate IX T 5c. Watsonia bulbifera Mathews & L. Bol. 1, node near base of inflorescence; 1a, same node with bract removed showing immature ‘bulbils’; i, prophylls. 2, pair of partly united bracts sub- tending two buds, x6; 2a, same bracts, back view. 3, partly united bracts with two buds, x 6; 3a, same bracts with buds removed. 4, pair of partly united bracts with one bud, x6; 4a, same bracts with bud removed; i, free margin of bract on left. 5, pair of united bracts with common apex, front view, <6; 5a, same bracts, back view; i, free margin. 6, immature ‘bulbil’, x 7; 6a, bract, x6; 6, united prophylls, x6; 6c, first scale leaf, x16. 7, prophylls shown at i in fig. 1a, X 2; 1, free margin of prophyll on left. 44 ANNALS OF THE SOUTH AFRICAN MUSEUM The number of cormlets in the axils of the bracts at the base of the inflorescence, where more or less erect lateral branches are present, 1s greatest, up to about thirty to a bract, decreasing in number towards the top of the spike. The final three or four bracts of the spike usually subtend flower-buds only, those just below having one or two cormlets as well as the flower. One of the nodes near the base of an inflorescence was examined at the beginning of the flowering season, at which stage the cormlets are present as small immature buds. An erect lateral branch was developed, next to the axis, and between this and the subtending bract were about thirty immature cormlets which even at this early stage had exerted sufficient pressure on the subtending bract to rupture it. This is illustrated on pl. TX, fig. 1. The same node is shown in fig. 1a but with the bract removed. Immature cormlets were also found in the axils of the lower bracts of the branch in some specimens. Many of the cormlet buds were found to be subtended by more or less enveloping pairs of bracts or prophylls in various stages of fusion. Some of the bracts were only partially united and some of these subtended two separate buds, while in others, where the fusion was more complete, only one bud had developed. Some of these stages are shown on pl. [X. In fig. 2 two bracts can be seen subtending two buds; in this case the bracts are almost free, with only the margins united. The outer side of the same pair of bracts is shown in fig. 2a. In fig. 3 two buds are present but the bracts are more closely united, as can be seen in fig. 3a, which represents the same pair of bracts with the buds removed. Fig. 4 shows a pair of fused bracts with only one bud developed. This was removed and the two partially united bracts ‘are shown in fig. 4a. Here the inner margin of the smaller bract on the left is free and can be seen as a small overlapping flap in the centre. Figs. 5 and 5a are the outer and inner sides of two bracts which are still more closely united, though the overlapping edges of both are still free and can be seen as flaps on the outer and inner sides. In this case the two bracts are more or less united in a common apex and only one bud was present. Fig. 6 represents a single young cormlet enveloped in its subtending bract (fig. 6a). Within this but on the opposite side of the bud are two completely united sterile prophylls (fig. 6b), which look like a single slightly bicarinate bract. The next foliar member, which encloses the bud and is united around it at the base, is the first scale leaf which remains attached to the young cormlet. There is an opening at the side through which the next scale leaf emerges as it develops (see fig. 6c). When mature cormlets were examined about two months Jater (shown on pl. X, fig. 1) the outer bracts and prophylls were membranous, very thin and brittle, and tended to fall apart and break up when touched, so that they evidently do not take any further part in the protection of the young cormlets after they have developed. This function is performed by the first and four or five succeeding scale leaves, which at this time were brown and more or less fibrous. ‘These are shown in a longitudinal section of a bulbil in fig. ra on jl >.< ann. >. Air. Mus.. Vol. XL: Plate X Watsonia bulbifera Mathews & L. Bol. 1, mature cormlets, x 2; 1, remains of bract; 1, branch; 1a, longitudinal section of cormlet, x 13. 2, node of first flower of spike with bract removed, x 3; i, immature cormlets; ii, bracteoles of flower; iii, perianth tube; 2a, bract from same node, x3; 2b, bracteoles of flower, x14. 3, node of first flower (two months later) with mature cormlets, x #; i, capsule. 4, node of second flower of spike, x 2; i, internode on lateral bud; 4a, bud with internode, removed from axis, x 14. 5, sixth bud of spike (adaxial view), x 3; 5a, bracteoles of same bud, with cormlet buds, x ?. 6 and 6a, front and back views of bracteoles of seventh flower, x2. 7, terminal bud, x%; 7a, bract, x3; 7b, bracteoles, back view, x §. 8 and 8a, top flower-buds of spikes with vestigial tips, x 3. 46 ; ANNALS OF THE SOUTH AFRICAN MUSEUM Fig. 7 on pl. IX shows the two prophylls of the shoot. They are situated between the branch and the axis, as can be seen in fig. 1a on the same plate. The smaller one on the left subtends a bud and is partly united with the larger one, with the free margin present as a small flap. The larger prophyll is actually the one which subtends the branch. (The more or less tricuspidate form of this prophyll is very unusual in the genus Watsonia and is probably due to the crowded conditions at the node as a result of which only the median and two main lateral veins are developed and project up in three short cusps. The tricuspidate bracts of some other genera are discussed in Part I of this paper.) Evidently some sort of proliferation has occurred in Watsonia bulbifera which has resulted in the development of those buds which, in the [xiozdeae, are almost invariably suppressed. ‘Thus buds have developed in the axils of many of the prophylls which are normally sterile. The prophylls shown in fig. 7 are those of the branch which is subtended by the large outer bract at the node. All the internodes of this branch have been suppressed and the lateral branch which is present is not a primary branch of the main axis but a secondary branchlet, the first branchlet of the primary branch. The second is present as a bud which develops into a cormlet through the accumulation of food reserves in its abbre- viated stem. The next node examined on the same inflorescence was that of the first flower. When the outer bract was removed two buds were disclosed, one on either side of the flower. These are shown in fig. 2 on pl. X. The bracteoles of the flower, which envelop the ovary and base of the perianth tube, are united for about three-quarters of their length, with free acute tips. When removed from the flower and opened up, a bud was found between them and the ovary (see pl. X, fig. 2b). In fig. 3 on the same plate a later stage in development is shown. At this stage, about two months after flowering, there are three mature cormlets. There were no fertile seeds in the capsule. The second flower in the same spike was the same as the first except that in the bud on the left an internode had developed, producing a rudimentary branchlet (see pl. X, 4 and 4a). The next four bracts subtended flower-buds, but here no lateral buds were present next to the flower-buds, though in each case a bud was found between the ovary and the bracteoles. In the sixth a second very small bud was present (fig. 5a). The last four bracts of the spike subtended flower-buds only, with no visible sign of any other buds. The fusion of the bracteoles of the spike becomes progressively more complete towards the top of the spike. At the base they have very distinct free tips but compare figs. 2b and 54 on pl. X, the latter the bracteoles of the sixth flower in which the apex is very shortly bifid. The inner and outer sides of the bracteoles of the seventh flower (figs. 6 and 6a) show the minutely bifid apex and towards the base two shallow pouches. The bracteoles of the terminal flower are com- pletely united and have only two rather obscure pouches at the base (shown in fig. 7b). In some of the spikes a vestigial tip was present above the terminal flower. Two of these tips are illustrated in figs. 8 and 8a. MORPHOLOGY, PHYLOGENY, TAXONOMY OF S.A. IRIDACEAE 47 No living material of Watsonia vivipara Mathews and L. Bolus was available for examination but in this species a similar proliferation has resulted in the production of numerous cormlets. The development of the cormlets is not quite the same as in W. bulbifera, and the authors commented on this as follows: “The name alludes to the “‘bulbils’? which while still on the parent plant put forth a green leaf projecting for about 1 cm. beyond the basal sheath of the “bulbil’”’, suggesting the idea that this leaf may be functioning in order to permit the “bulbils” to start growth without a preliminary resting-stage. In the large number of bulbils produced this species resembles W. bulbifera, but this differ- ence must be noted: the development of the bulbils is concurrent with the development of the flowers in W. vivipara, whereas in the former their develop- ment starts when the flowering is nearly, if not quite, over.’ In the allied genus Micranthus a similar proliferation occurs in one of the species, M. junceus N. E. Br. In this very much smaller plant with its much denser spike the number of cormlets produced is not as great as in Watsonia bulbifera, not more than seven to nine being found sometimes in the axils of the lowest bracts; more often there are solitary cormlets, or from two to five to a bract near the base of the spike, or, in some specimens, no cormlets are produced. The prophylls of Watsonia bulbifera provide fairly conclusive evidence of a fusion taking place between two closely adjacent overlapping members, the closer union of the two members resulting in the suppression of the lower of the two axillary buds. 4. Vestigial traces of bracts in Watsonia pyramidata and a hybrid Some vestigial traces of bracts were found in one or two species of Watsonia, as well as in Pillansia Templemanni, which furnish further evidence of the spike having been evolved from a more complex inflorescence through various stages of reduction. When collecting fresh flowers of Watsonia pyramidata (Andr.) Stapf (W. rosea Ker) in the field it was frequently noticed that within the bracts of some of the specimens was a colourless mucilaginous substance. What appeared to be a natural hybrid between W. pyramidata and W. Schlechteri, from the top of the French Hoek Pass, was sent to me and here again the viscid substance was observed inside some of the bracts subtending the flowers. One inflorescence was therefore examined in detail. It was found that the one or two short, erect branches at the base have their lowest internode partially adnate to the axis (see pl. XI, fig. 12). Between these branches and the axis were the short, united but distinctly bifid prophylls which are shown in fig. 2 on pl. XI. In the flowering parts of the inflorescence small and very delicate imperfect structures were found in many cases between the closely adpressed bracteoles and the axis. Some of these are shown in figs. 3 and 3a on pl. XI. These structures were translucent, pale pink in colour, and when moistened they immediately dissolved and became colourless and muci- laginous. Under the microscope they were seen to be only one or two cells in Ann. S. Afr. Mus. Vol. XL. Plate XI 6c A Li\ @ 6b \ \ 6a 1-4, Watsonia hybrid. 1, first node of inflorescence, bract removed, xX %; Ia, second node; i, internode of branch partly adnate to axis. 2, prophylls of branch in fig. 1a, X §. 3, 3a and 36, adaxial views of bract and bracteoles of flower-buds showing vestigial remains of prophylls, x 2; i, bract; i, bracteoles; iii, remains of prophylls. 4, part of vestigial remains of prophylls, xX ¢. 135. 5, Watsonia Middlemostii L. Bol., part of spike, x %; i, outer bract; u, prophylls; ii, inner bract; iv, bracteoles. 6-11, Pillansia Templemanni (Bak.) L. Bol. 6, branch of inflorescence, <#%; 6a, bract subtending branch, x 14; 6, prophylls of branch, x 1}; 6c, prophylls of secon- dary branchlet, « 13. 7, apex of primary branch with tip developed, x %. 8, bud with vestigial traces of inner bracts, x 1}; 8a, bract of bud, x 14; 8b, bracteole, x 14; 8c, first vestigial bract, x 13; 8d, second ditto, x 2; 8e, ovary and second vestigial bract, x 2. 9, diagram of bracts and flower shown in fig. 8. 10 and 11, diagrams of two theoretical earlier stages before reduced stage shown in fig. 9. 12, Nivenia corymbosa (Ker) Bak., diagram of bracts and branches at first node of inflorescence; X, main axis; Bi—Bv, bracts. MORPHOLOGY, PHYLOGENY, TAXONOMY OF S.A. IRIDACEAE 49 thickness, consisting only of one or two epidermal layers, with elongated cells and distinct traces of stomata (see pl. XI, fig. 4), resembling the epidermis of the bracteoles which was examined for comparison. In a bud near the top of the spike these translucent structures were present between the bracteoles and the flower as well as between the bracteoles and the axis (see fig. 35). These extremely delicate structures are almost certainly the vestigial traces of prophylls of which only the epidermal layers are partially developed. ‘That this is possible seems to be borne out by the arrangement of the bracts in a few specimens of Watsonia Middlemostu L. Bol. This species is closely allied to W. pyramidata from which it differs in being smaller in all its parts, with narrower leaves and nearly always an unbranched axis. In specimens collected by Schlechter at Onrust River, Klein River Mountains (No. 9492), it was observed that the flowers in the spike were subtended by two bracts and had two pairs of united bracteoles between the flowers and the axis, the lower pair of the latter very short and not very noticeable (see pl. XI, fig. 5), whereas the upper ones are as long as the inner of the two bracts. One of the outer bracts was removed and an extremely short internode could just be discerned between the outer and inner bracts. The reduction from a branched inflorescence to a simple spike has for some reason been arrested in these plants, and the lower of the two pairs of united members between the flowers and the axis are therefore the prophylls of the reduced lateral branch, the upper two, which are almost completely united, being the bracteoles of the flowers. This is comparable with the arrange- ment in W. bulbifera where the lower prophylls are present on either side of the lower flowers in the spike, but instead of being sterile and united they are free and each subtends a bud. 5. Vestigial traces of bracts in the inflorescence of Pillansia Templemanni From the phylogenetic point of view Pillansia Templemann (Bak.) L. Bolus (Tritonia Templemanm Bak.) is of particular interest. This monotypic genus, which is confined to a small area near the coast in the Caledon Division, is indeed not easy to place as it is intermediate between the Iridozdeae and. Ixtotdeae. In many respects it is most closely allied to Watsonia and may be regarded as the ancestral type of that genus, but, as noted by L. Bolus, it differs among other things in having flowers which are regular and arranged in a ‘cymose panicle’. Baker and L. Bolus evidently were not aware of the name Wredowza pulchra given to this plant in 1827 by Ecklon, as neither of them cited this name, which must, however, be regarded as a nomen nudum. (Apart from stating that it was intermediate between Stsyrinchium and Aristea Ecklon gave no description of the genus or species.) In 1840 Steudel, in Nomencl. ed. II, 1, 1300, changed the name to Aristea Wredowia, but in 1940 Weimarck quite rightly excluded this plant from the genus Arstea. Hutchinson placed Pillansia in the tribe Ixieae. The systematic position of this plant is referred to again in Part IV. Some vestigial traces of bracts were found in the inflorescence which provide some evidence of a link between the spicate inflorescence and a more richly 50 ANNALS OF THE SOUTH AFRICAN MUSEUM ramified one such as occurs in some species of Nivenia and Aristea. L. Bolus described the inflorescence as follows: ‘Flowers solitary, terminal on the primary, secondary, tertiary and quaternary branches, forming a somewhat lax panicle 25-40 cm. in length, with ascending or suberect branches 4—9 cm. long.’ In living material from the Palmiet River Valley which was examined there were fewer branchlets than in the inflorescence described above. A primary branch is shown in fig. 6 on pl. XI. This is subtended by a short bract (fig. 6a) and between the branch and axis are a pair of united prophylls, one slightly longer than the other (fig. 6b). Similar prophylls are also present at the bases of the secondary branchlets (fig. 6c). Each of the secondary branchlets apparently terminates in a solitary flower, with a bract and a similar bracteole and no sign of any vestigial tip, but in some instances there was evidence of the axis extending above the apparently terminal flower of the primary branches. In fig. 6 a vestigial tip is shown, and in fig. 7 the apex of a branch is shown in which the tip has developed and terminates in a flower, the penultimate flower being sessile. This development of the tip is not common and was only observed in one or two of the many branches examined. The outer bract subtending each flower is, as described by L. Bolus, herbaceous in the lower half, chartaceous round the margin, 0-8-1 cm. long (pl. XI, 8a). The inner sterile bract or bracteole is similar but broader and rounded towards the apex (fig. 8b). Between these bracts and the flowers what appeared to be the membranous tips of a second pair of bracts could be seen in almost every case. Fig. 8 is the terminal bud of the primary branch sketched in fig. 6, slightly enlarged, showing the tips of the two inner bracts between the bud and the two outer bracts. When the outer bracts were removed with care the traces of two extremely delicate inner bracts were revealed. The position of these bracts is shown diagrammatically in fig. 9. Within the outer bract, 1, is a vestigial bract, iil, which, like the outer one, partially envelops the second outer bract, iii. Between the bract ii and the ovary is the trace of a fourth bract, iv. The two vestigial bracts are shown in figs. 8¢ and 8d. The inner of the two vestigial bracts was so delicate that it tended to disintegrate when touched, but a number of them were examined and it was found that at the base they were more or less adnate to the base of the ovary. The ovary, with the remains of a bract attached, is shown in fig. 8e. This second bract is larger than the lower vestigial bract and might be compared with the bracteoles of some species of Vivenia which are conspicuously longer and wider than the subtending bracts, and are brown and membranous whereas the bracts are firm and green. These very delicate structures which are presumed to be traces of bracts are, like those found in the Watsonia hybrid, pale pink and translucent, and when moistened they immediately dissolved into a colourless mucilage. There is no sign of any actual vestigial bracts in herbarium specimens which have been examined, but in many of them the buds and exterior surface of the perianth tube and lobes of open flowers have a shining, lacquered appearance. ‘This MORPHOLOGY, PHYLOGENY, TAXONOMY OF S.A. IRIDACEAE pit undoubtedly comes from the mucilage from the moistened vestigial bracts, either due to rain having fallen on the inflorescence before the buds opened, or to the process of drying and poisoning. The flowers of some herbarium specimens of Watsona pyramidata have the same lacquered appearance. Assuming that these structures observed in Pillansia Templemanni are indeed the vestigial traces of bracts, then the inflorescence in this plant fairly obviously represents an intermediate stage between the spicate inflorescence of the Ixioideae and a compound inflorescence such as that of Nivenia binata or N. corymbosa. On pl. XI, in figs. 10 and 11, are shown diagrammatically two theoretical earlier stages in the development of the seemingly apical bud of the primary branch of Pillansia Templemanni. If, in addition, the prophylls at the bases of the primary and secondary branchlets, instead of being united and sterile, were free and one or both of them subtended axillary branchlets, then the inflorescence would scarcely differ from that of Nivenia corymbosa (Ker) Baker. Weimarck, in his revision of /Vzvenza, illustrated the theoretical develop- ment of the upper, umbel-like part of the main inflorescences of WV. corymbosa and WN. binata Klatt. A plan showing the arrangement of the bracts and branches at the first node of the main axis of an inflorescence of V. corymbosa which was examined is given in fig. 12 on pl. XI. Bi is the first bract of the main axis (marked X) and it subtends the lowest primary branch, 1. The lowest internodes of this branch are contracted and the two small bracts or prophylls at the base, Bii and Bi, are thus extremely close together. The upper of these two bracts is sterile but the lower one, Bu, subtends a secondary branchlet which, like the primary branch, has its lowest internode contracted. It bears a tertiary branch- let, 3, in the axil of the bract Biv. This in turn has a bract, Bv, at its base, with a bud in its axil. Above the very short internodes at the bases of the primary, secondary and tertiary branchlets are long internodes, but there is no reason to describe these branchlets as this has already been done by Weimarck. The point that I wanted to illustrate in the diagram is the proximity of the two bracts or prophylls at the base of the primary branch. Only a very slight further abbreviation of the basal internodes would, in all probability, result in the fusion of the overlapping margins next to the axis of the bracts Bu and Bin, and the suppression of the bud in the axil of the lower bract (Bi). This would leave a primary branch with a pair of united sterile bracts or prophylls at the base, between the branch and the axis, such as occurs in Pillansia Templemanni. On the other hand, only one further stage in reduction in the branch of Pillansia Templemanni shown on pl. XI, fig. 6, would result in the suppression of the internodes of the secondary branchlets. If, instead of branchlets, flower- buds were produced at these nodes, this would give a short lateral] spikelet with two sessile flowers each with its subtending bract and a pair of united prophylls, or bracteoles, as they would now be called. The vestigial trace of a bract within the bracteoles of one of the buds in the Watsonia hybrid (pl. XI, 34) seems to suggest that a similar reduction has taken place in that plant. 52 ANNALS OF THE SOUTH AFRICAN MUSEUM As observed by Haeckel, the apex of the branch generally disappears in an early stage, but it occasionally happens that the uppermost flower is developed, in which case, as recorded by Boehni, the flower is regular even if all those below it are zygomorphic. I have observed this myself in a few specimens of Watsonia Mervana. 6. The inflorescences of some species of Lapeyrousia Another genus of interest from the phylogenetic and morphological view- points, which requires a great deal more study, is Lapeyrousza. It is a fairly large genus, well represented in South and tropical Africa, and in some species the inflorescences present features which are associated with those found in the Iridoideae and Crocoideae, while in others, probably the majority, the inflorescence is reduced to the spicate type which is typical of the Jxzoideae. In the subgenus Sophronia, for instance, all the internodes of the main axis and the branches are abbreviated and the leaves and flowers are congested in a dense ‘sessile rosette’, asin Galaxia. In some species in the subgenus Ovieda intermediate stages between a prolific, laxly branched inflorescence and a simple or branched spicate one occur. The subgenus Anomatheca is the only one of the three subgenera in which the basal leaves arise from the corm, as is almost invariably the case in the Ixiozdeae. The corms in this subgenus also differ in other respects from those of the other two subgenera, in which the lowest green leaves do not arise from the corm itself but from the aerial axis, usually shortly above the corm, as they do in Galaxia and also all those members of the Jridozdeae which have corms. There is another feature of interest which is probably more conspicuous in some species of Lapeyrousia than in any of the other genera. This is the matter of adnation, which requires further investigation and has been mentioned before in connection with the leaves. In several species of Lapeyrousia, notably L. corymbosa, L. Fabrici and L. Facquiniz, it can be seen quite clearly that the branches (also the flowers) with their subtending bracts are more or less adnate to the axis. In L. Fabricii, a species which grows in sandy ground, the corm is deep- seated and the portion of the aerial axis below the ground is frequently elongated. Very often cormlets are produced in the axils of the lower leaves which arise below the ground-level, one in each leaf axil. Such cormlets almost invariably appear some distance below the node (see pl. XII, fig. 1), breaking through the leaf skin or lower portion of the sheath which is adnate to the axis. In one or two specimens of L. Fabricii examined it was found that some of the upper branches remain adnate to the axis for 6 or 7 mm. above the node, while in one or two specimens of L. Facquinit the lowest branch was found to emerge from the axil of its subtending leaf below the node, rupturing the lower dorsal part of the sheath as in the case of the cormlets mentioned above. In this species also the rather large, conspicuously bicarinate and bifid prophylls are frequently carried out of their normal position, at the base of the branch, and appear half-way up the branch. The axis in these species mentioned above, as well as Ann. S. Afr. Mus. Vol. XL. Plate XII N 6 2 5a h 1, Lapeyrousia Fabrict (Thunb.) Ker, base of stem, with cormlets; i, node. 2, L. Facquini N. E. Br., diagram of spike; i, bifid bracteoles. 3, L. Vaupeliana Dinter, diagram of top of inflorescence. 4, L. Bainesii Bak., diagram of top of inflorescence; i, bifid bracteoles. 5, L. corymbosa (L.) Ker, inflorescence, x %; i, prophylls; 5a, branchlet, x14; 56, bract, x14; 5c, bracteole, x 14; 5d, bract subtending branch near top of inflorescence, x 64. 6, L. corymbosa, branch of inflo- rescence, the lowest flower with united bracteoles. 7, L. micrantha Bak., bract of branch near top of inflorescence, x 6$. 54 ANNALS OF THE SOUTH AFRICAN MUSEUM in several others, is often extremely flexuose, a condition which usually seems to be associated with adnation. In some weak or drought forms of L. Facquinit, in which the lower branches are not developed, the axis is almost straight, but in more robust plants with several branches it is extremely flexuose, the branching often appearing divaricate as the primary branches are usually as long as the main axis. A diagram of the spike of L. Jacquin is given on pl. XII, fig. 2, to illustrate the adherence of the lower part of each bract to the axis, suggesting a sympodial rather than a spicate arrangement. A vestigial tip is present and the bracteoles are all distinctly bicarinate and bifid (forked in diagram). The fact that the bracteoles of the apical flower here are also strongly bicarinate and bifid refutes the argument of those botanists who maintain that this is due to compression between the bud and axis. L. Vaupeliana Dinter, from South West Africa, is among the tallest and is undoubtedly the most richly ramified species in the genus. Some specimens in the South African Museum Herbarium collected by Dinter (Dinter 2787 and 3374) are from 2 to 4 ft. high, and from shortly above the base upwards the axis, which is distinctly flexuose, bears numerous branches, each in turn bearing several branchlets, so that the whole plant is a dense mass of innumerable branchlets, each as a rule terminating in a single flower with a membranous bract and an identical bracteole. In one or two forms which are less vigorous the upper parts of some of the lower branches are present only as abortive buds and towards the top of the inflorescence on some of the branchlets there is a penultimate sessile flower (compare the lateral branch in the diagram of L. Bainesu on pl. XII, fig. 4). The upper part of an inflorescence of L. Vaupeliana is shown diagrammatically on pl. XII, fig. 3. All the flowers in the upper part of the inflorescence open together, those of the lower branches being still in bud. The lower internodes of the primary branches are all elongated, so that in effect the branching is more or less corymbose, but towards the ends of the branchlets the internodes are almost equal and here the branching is dichotomous (see diagram). In the lower part of the inflorescence there is a pair of united prophylls at the base of the primary and most of the lower secondary branches, but in the upper part these have not developed, although abortive buds can be seen quite distinctly between the main axis and primary branches as well as between the primary and lower secondary branchlets. With a lens it can be seen in some cases that these buds have two collateral growing points. The primary branches arise alternately from the axis, in accordance with the distichous arrangement of the leaves, but the first secondary branchlet which is developed is always in the axil of a bract which is on the same side as the preceding bract or leaf. The same is true of the tertiary branchlets and finally of the flowers. This arrangement is shown in the diagram, although in the plants examined it is not immediately obvious owing to a slight spiral twist in the stems. It can, however, be fairly readily traced as the stems are sharply angled. ‘This inflorescence might be compared with that of Nivenia corymbosa MORPHOLOGY, PHYLOGENY, TAXONOMY OF S.A. IRIDACEAE 55 with the following differences noted: (1) The prophylls at the bases of the branches, when developed, are united and always sterile; (2) each flower has a bracteole but there is no upper sterile bract. In his very brief description of the species, which he stated was near L. Bainesii, Dinter merely referred to the inflorescence as prolific and laxly branched. He made no reference to the style branches which in all the specimens I have seen are simple and not bifid as they are almost invariably throughout the genus. The corm, described by Dinter as ‘broadly bell-shaped’, is typical of the sub- genera Sophronia and Ovieda, flat at the base and broadly conical above, and the lower leaves arise from the axis above the corm. The inflorescence described above and the simple-style branches seem to indicate that these characters are the most primitive in the genus. Comparing the inflorescence of L. Bainesit with that of L. Vaupeliana, it is evident that the former represents a transition stage between the much-branched inflorescence of the latter and the spicate inflorescence which occurs in many species of Lapeyrousia. A diagram of the top of an inflorescence of L. Bainesii is sketched on pl. XII, fig. 4. In this the main axis terminates in a four-flowered spike. ‘The apical flower and the flower below it have single bracteoles, identical with the bract, but the bracteoles of the two lower flowers are bifid (forked in the diagram). The two branches have two flowers each, all of them with a single bracteole. On the lower branch there is a slight indication of a vestigial tip above the upper flower. The lower internode of both branches is long but the upper one abbreviated, so that the two flowers are comparatively close together, a circumstance which was commented on by Weimarck in connection with L. plagiostoma. This arrangement of the flowers in pairs is even more striking in L. corymbosa Ker, a species fairly common on the Cape Peninsula and in some of the neighbouring districts. In this species the upper internode of the branches is almost completely suppressed and the two flowers are very close together (see figs. 5 and 5a on pl. XII), resembling the arrangement which Weimarck referred to as “binate flowers’. In living material examined there was no visible sign of the apex extending above the upper flowers and in practically every case the bracteole was found to be a single member, slightly narrower than the bract but otherwise identical, as can be seen in figs. 56 and 5c. In only one specimen was a bifid bracteole found on one of the lower branches, where a reduction had taken place and instead of a two-flowered branchlet there was a single sessile flower with a pair of united bracteoles (see fig. 6). On the lowest branch of the plant figured in the diagram (XII, 5) the united prophylls have been displaced and are carried out of the position they normally occupy, a circum- stance referred to above in connection with L. JFacquinit. The small, very much reduced bracts (about 1 mm. long) subtending some of the upper branches of the inflorescences in L. corymbosa and L. micrantha, two closely related, well-branched species, are shown in figs. 5d and 7 on pl. XII. The first shows a tendency towards tricuspidation and the second is distinctly 56 ANNALS OF THE SOUTH AFRICAN MUSEUM tricuspidate. In the lower fully developed bracts in the same inflorescences there is no sign of this lobing, which is probably due to reduction. | 7. Reduction of internodes and the rhipidia in the more primitive genera The most simple and primitive combination of characters in the South African genera is found in the genus Avistea. The systematic position of this genus and the confusion which ensued from its having been combined with the genus Nivenia was discussed by Weimarck in his monograph. He placed it in its correct position, in the tribe Szsyrinchieae, with Sisyrinchium and several other genera, none of which, with the exception of Bobartia, are South African. Weimarck regarded the binate rhipidia as the typical inflorescence type in Aristea and the various deviations in the types of inflorescence he attributed as a rule to reduction of flowers, so that bracts and prophylls remain sterile. My own investigations seem to show that bracts and prophylls which have become sterile through reduction either disappear or remain as sterile members which do not again become fertile except in the rare case of proliferation such as has occurred in Watsonia bulbifera. In Lapeyrousia Vaupeliana, for instance, in spite of all the ramification, the prophylls and bracteoles are all sterile, as they are throughout the genus. | Probably the most primitive inflorescence to be found in the South African Iridaceae is that which occurs in some of the larger species of Aristea, such as A. Bakert and A. capitata. In these two tall, robust species, with their compound racemose branching, sterile prophylls are present only in some of the weaker forms in which the branchlets have not developed. ‘The branching of a fairly well developed specimen of A. capitata is shown diagrammatically on pl. XIII, fig. 1. The branches and branchlets throughout end in ‘binate rhipidia’, as Weimarck has named this type of inflorescence. The lowest primary branch bears four secondary branchlets and these sometimes give rise to tertiary branch- lets, as is shown in the diagram. ‘The lowest secondary branchlet arises at the extreme base of the primary branch, next to the axis. The lateral branches become progressively shorter towards the top of the axis, but as can be seen in the third, fourth and fifth, each still bears two or three branchlets, although in these the two lower internodes are extremely abbreviated and the two lower branchlets arise next to the axis. In the diagram it is only possible to show one plane, but the actual position of the branchlets on either side of the branch is shown in fig. 1a. ‘This reduction of the internodes and position of the bracts or prophylls is significant, for only a very slight further contraction of the basal internodes, or the adnation of these abbreviated internodes to the axis, would almost undoubtedly result in the fusion of the bracts and the suppression of one or both of the buds. | The end of a branch of a recently described species of Aristea (A. latifolia Lewis), slightly enlarged and without the flowers which terminate each branchlet or pedicel, is shown on pl. XIII, fig. 2. This tall and well-branched species, somewhat similar in appearance to A. Bakert but quite distinct in many respects, Ann. S. Afr. Mus. Vol. XL. Plate XIII i U tt eb aay wie “ a & ee, tf. Pi Era 9 a iy f a sey 4 t AS ll AB f ya | if i Tp t aii . fi ‘ JOR K\\ p, A yp i M, \ Lie é 4 ws , Zi i) \\ f y ’ ip Z Se WAZ NY ? kK. {Zig i Lo } (hat i ‘ 1, Aristea capitata Ker, diagram of inflorescence; 1a, third node of inflorescence, bract removed. 2, Aristea latifolia Lewis, binate rhipidia, 2 (the bract and two bracteoles removed from the rhipidium on the left); 2a, bracteole, x2; 2b, plan of binate four-flowered rhipidia. 3, Aristea oligocephala Baker, binate rhipidia with vestigial tip, x 3; 3a, same, with one rhipidium removed; 3b, bracteole. 58 ANNALS OF THE SOUTH AFRICAN MUSEUM grows only at fairly high altitudes (about 4,000—5,500 ft.) on mountains in the Paarl, Worcester and Caledon Divisions. It was selected for illustrating the inflorescence as the bracts were small and narrow and the internodes slightly longer than in any of the other species. In the figure the outer bract and two succeeding bracteoles on the left are removed and only the upper one, which is sterile in this case, is shown. Except that it is very slightly smaller, there is no difference at all between the first bracteole, shown in fig. 2a, and the bract. There is a slight indication of an apex between the two rhipidia. The inflorescences of several species of Aristea were examined and as a rule there is little if any visible sign of an apex between the two rhipidia, but in some specimens of A. oligocephala an actual vestigial tip was found. This is illustrated on pl. XIII, fig. 3. Fig. 3a is the same, slightly enlarged, with one rhipidium removed. This vestigial tip was only found in three inflorescences out of about fifty that were examined; in one of them it was more robust and terminated in a normal flower. One of the strongly carinate bracteoles with its wide mem- branous, colourless margins is shown in fig. 3d. A point that should be mentioned in connection with the rhipidia of A. laizfolia is that the flowers are more or less spirally arranged. A plan showing the arrange- ment of a pair of four-flowered rhipidia in this species is given on pl. XIII, 20. In each rhipidium the first and second flowers are more or less opposite and the third and fourth opposite each other and almost at right angles to the first pair. The fifth bracteole in each is sterile. ‘he resemblance of these rhipidia (if this term is indeed the correct one in this case) to the umbellate inflorescences of the Amaryllidaceae, with their condensed bostrychoid cymes, led to a compara- tive examination of some of these umbels. The first one available was Tulbaghia violacea Harv., a species from the eastern Cape Province, and as this happened to provide some further evidence of the fusion of the bracteoles it will be described briefly here. The genus Tulbaghia is one which Hutchinson, in his Families of Flowering Plants, transferred from the family Liliaceae to Amaryllidaceae on account of its umbellate inflorescence. In nearly all the inflorescences of Tulbaghia violacea examined the umbel consisted of two cymes, usually with four or five perfect flowers in each and one or two abortive buds. In one specimen there were three such cymes and a trace of a fourth. In all of them the inner bracts, or the bracteoles, were variously united, with only the short acute tips free, either one free and three or four united, or in pairs, or all united. In no case were they all free. The first specimen examined was slightly abnormal, with the pedicels of two flowers in each of the two cymes united up to the base of the perianth. This specimen is illustrated on pl. XIV, figs. 1 and 1a. Fig. 16 shows four normal buds with their united bracteoles, and in fig. 1¢ these same bracteoles are shown opened out, with the buds removed. A diagram of the arrangement of the flowers and bracteoles in one of the umbels is shown in fig. 1d. It has been observed by Arber and others that the conditions of pressure under which lateral buds arise in some of the Monocotyledons sometimes results in MORPHOLOGY, PHYLOGENY, TAXONOMY OF S.A. IRIDACEAE 59 anomalous fusions. Such a fusion has occurred in this species of Tulbaghia. In the umbels of other members of the Amaryllidaceae examined, such as Crinum, Cyrtanthus and Nerine, the bracteoles are all free but they are very much reduced in size and are all more or less filiform and functionless, the function of protecting the buds being performed by the much enlarged outer bracts or spathes. In Bobartia, the other South African member of the tribe Sisyrinchieae, the discovery during the past twelve years of two very distinct species is of particular interest, as they provide some important ‘missing links’. These species, B. paniculata Lewis (Journ. S.A. Bot. VII, 50 (1941)) and B. lilacina Lewis (Journ. S.A. Bot. X1, 108 (1945)), both from mountain districts in the Cape Province, R la PLATE XIV Tulbaghia violacea Harv.; 1, umbel, x ¥; i, united pedicels; 1a, two flowers with united pedicels, x 2%; 1b, four united bracteoles subtending four flower-buds, x 3; 1c, same bracteoles with buds removed, x2; 1d, plan of umbel with bracteoles variously united. 60 ANNALS OF THE SOUTH AFRICAN MUSEUM will necessitate some alteration to the definition of the genus of which the inflorescence has hitherto been described as compact or unbranched. It might be mentioned here that during a visit to Kew and some herbaria in Europe in 1949 it was found that the laxly branched species named B. /ilacina was actually first collected by Drége over a hundred years ago. The extremely delicate and fugacious pale mauve flowers are missing from the specimens (Drége 1538) in the herbarium at Kew and in Paris. Drége and Meyer placed this plant in the genus Moraea, but it was not described or named. Bobartia lilacina is a laxly branched species with slender stems and small bracts and prophylls, the latter united, distinctly bifid and bicarinate. In most cases the prophylls are sterile but occasionally a branchlet develops in the axil of the upper of the two united prophylls. Two such branchlets are shown in a diagram of the inflorescence of B. lzlacina in fig. 3 on pl. XV. In B. paniculata the one to three slender branches are borne near the top of the stem. It is evident that in this species a considerable reduction has taken place in most of the internodes. There is little doubt that B. /ilacina is the most primitive species in the genus and that the inflorescence of B. paniculata represents an intermediate stage between the laxly branched inflorescence of B. lilacina and the congested type which occurs in the other twelve to fifteen species, in which a further reduction has taken place in the upper internodes of the main axis and the lateral branches. On pl. XV, fig. 1, is illustrated an inflorescence (without flowers) of B. gladiata Ker, and in fig. 1a a diagram of the same inflorescence. This is more or less the typical congested type of inflorescence most common to the genus; in some species it is very much denser and in others, such as B. filiformis, more reduced, but the arrangement is more or less the same throughout, except in B. paniculata and B. lilacina. ‘The internodes are all extremely abbreviated and the lowest bract, which is sometimes much enlarged, often appears as a terminal extension of the axis, the inflorescence above this bract being forced out of position and more or less at right angles to the lower part of the axis. As can be seen in the diagram, however, the apparently terminal bract is actually lateral, with a bud in its axil which, in more robust forms of B. gladiata, frequently develops. In B. indica L. the lateral position of this bract is more evident and the considerably denser inflorescence occupies the terminal position. The thickened lines in fig. 1 represent the strongly bicarinate prophylls, which are forked in the diagram. Fig. 2 is the same as fig. 1a but with the internodes extended and with the two Jower buds developed into branches which are indicated by dotted lines, as well as a branchlet in the axil of one of the pairs of prophylls. With the addition of these branches it represents a diagram of the inflorescence of B. paniculata. Through B. lilacina the genus Bobartia is more closely allied to Aristea. It differs from Aristea, among other things, in having the prophylls united throughout, even in the rhipidia. In this respect the rhipidia of Bobartza are the same as those of Homeria, Ferraria, Moraea, Dietes, etc., and as some of them were examined in mas. Air. Mus. Vol. XL. Plate XV @ On ! Ti S ESS. i | ENS © a la ; Bobartia gladiata (L. f.) Ker, inflorescence; 1a, diagram of same inflorescence. 2, same as 1a put with internodes extended; with addition of branches marked with dotted lines this represents the inflorescence of B. paniculata Lewis. 3, B. lilacina Lewis, diagram of inflorescence. 62 ANNALS OF THE SOUTH AFRICAN MUSEUM some detail this occurrence of the bicarinate bifid prophylls or bracteoles should be discussed. In all these genera (1.e. Bobartia, Moraea, etc.) the lowest bract or spathe of the rhipidium completely encircles the inflorescence or rhipidium, and has a short, closed sheath at the base. ‘The second bract is almost always longer and broader and also encircles the inflorescence, but the margins, although they overlap at the base, remain free. In none of the specimens examined was this bract found to be bifid or bicarinate and there is little doubt that it is a single bract. Each of these lower bracts subtends a flower, but the upper flowers in the rhipidia are subtended by more or less bicarinate and often distinctly bifid bracteoles. There seems every reason to believe that, just as the prophylls of the branches in Bobartia, Moraea, Homerza, etc., consist of two united bracts, the lower of which is sterile (often both are sterile), the same is true of the bracteoles in the rhipidia of these genera. ‘That such fusions of two overlapping members can and do take place, due to the reduction of internodes and conditions of congestion and pressure, has already been shown. In the plan of the Aristea rhipidia on pl. XIII, fig. 2b, the somewhat spiral arrangement of the flowers and overlapping of their bracteoles is illustrated, and reference was made to the fact that the internodes in this species are slightly longer than is usual in the genus. A contraction of the internodes in a similar spirally arranged inflorescence might well have resulted in the fusion of the bracteoles and consequent suppression of the lower of the two buds, and it 1s possible that such a fusion having once taken place it has become an hereditary character. I am also of the opinion that the rhipidia in these genera (Bobartia, etc.) are derived from binate rhipidia through a closer adnation of the two lateral branches, resulting in the complete disappearance of the apex and of all except the first flower of the lower of the two rhipidia, the flowering being continued only in the upper one. This seems to be the only explanation for the structure of the rhipidia, in which each of the two lower flowers is subtended by an outer bract, or spathe, whereas the upper flowers are all subtended by united bracteoles. A diagram illustrating this theory is given.on pl. XVI. Fig. 1 on pl. XVI isa diagram of the inflorescence of Dietes bicolor (Lindl.) Sweet. The first flower is subtended by the lowest bract, the second by the second much larger and longer bract, and those above by bifid bicarinate bracteoles (forked in the diagram). Fig. ta is a plan of the same inflorescence showing the relative positions of the bracts and flowers. (The arrangement here is exactly the same as in Iris pseudacorus as shown. by Haeckel.) Fig. 15 is a theoretical diagram of the same inflorescence but with the internodes slightly extended and with a vestigial tip added in the position I presume that it would occupy. For comparison I have added a diagram (fig. 2) of the actual inflorescence of the Arstea shown on pl. XIII, fig. 2. It has already been pointed out that in Aristea there are only traces of a vestigial tip between the two rhipidia in a very few cases. In most of Ann. S. Afr. Mus. Vol. XL. Plate XVI 1, Dietes bicolor (Lindl.) Sweet, diagram of rhipidium; theoretical diagram of same inflorescence with internodes extended and a vestigial tip (i) added. 2, Aristea latifolia Lewis, diagram of binate rhipidia shown in fig. 2 on pl. XII. 3, Moraea ramo- sissima (L. f.) Druce, part of inflorescence; 3a, diagram of branching at the lower node in fig. 3, showing bracts and short and long internodes; 34, diagram of branching from the same node but with the lower internodes of the branches extended; 3¢, plan of bracts and branches at the same node. 1a, plan of same inflorescence; 1), 64 ANNALS OF THE SOUTH AFRICAN MUSEUM the species in that genus there is no visible sign of a tip and the two rhipidia are extremely close together. | It is stated above that in Bobartia, Moraea, etc., the lower of the two fused prophylls of the branches is always sterile, but an exception was found in an unusually tall and robust specimen of Moraea ramosissima; in this branchlets were present in the axils of both of the lowest prophylls. Moraea ramosissima (L.f.) Druce is usually 2 to 4 ft. high, and the very apt specific epithet refers to the numerous branches and branchlets. A part of the southern area of the Cape Peninsula which had been completely burnt out a few months previously was visited in 1950 and in a small marshy area there was a dense growth of this species, many of the plants 6 to 7 ft. high. On examining one of these plants two branchlets were found in the axils of two partly united prophylls. A sketch of the upper part of the axis with the branching at two of the nodes is given on pl. XVI, fig. 3. Fig. 3a is a diagram of the branching at the lower of these two nodes, which illustrates the congestion of the branchlets and their subtending bracts and the succession of short and long internodes. The lowest bract, B, subtends the primary branch, 1. Bracts Bi and Bii, although shown as separate members, are actually partly united and each subtends a secondary branchlet (numbered 2 and 3). Bit represents a pair of united bracts subtending a single tertiary branchlet, 4, and Biv also united bracts with a single branchlet, 5. Between the fourth and fifth branchlets is a bud. Each branchlet terminates in a two- or three-flowered rhipidium. Fig. 3b represents the branching from the same node but with the lower inter- nodes of the branches extended to show the arrangement more clearly. Since only one plane can be shown in these diagrams I have added a plan (fig. 3) to show the position of the branches and bracts in relation to the axis (marked x). The base of the bract subtending the primary branch encircles and forms a short closed sheath round the axis. 8. Conclusion and a reference to the value of the bracts as a diagnostic character in taxonomic work The various inflorescences which have been discussed in the preceding pages all provide some evidence in support of Haeckel’s assertion that the more composite types of inflorescences in the Iridaceae are derived from a primitive much-branched type. Reductions in the number of flowers and suppression of some of the internodes have finally resulted in the simple or branched spicate inflorescence of the Ixzoideae. ‘The fusion of the prophylls has been shown to be due to the reductions of internodes, as well as in part to the width of the sheathing bases of these bracts so that, when reductions have taken place, these broad sheathing bases have first overlapped, then finally become united. These conclusions are not in accordance with Weimarck’s theory that the spicate inflorescence of the /xiozdeae is the most primitive arrangement in the family, and that the more complex types of inflorescence were derived from this. ‘Two or three factors seem to preclude the correctness of this theory. One is the MORPHOLOGY, PHYLOGENY, TAXONOMY OF S.A. IRIDACEAE 65 presence of the bifid or bicarinate bracteoles in the spicate inflorescence of the Ixioideae. ‘Vhis type of spicate inflorescence occurs only in the Jxzozdeae, which is generally considered to be the most highly evolved of the three groups into which the family has been divided (the development of the corm, leaves, inflorescence and flowers all bear this out). Weimarck’s assumption that the type of spike in the Jxzordeae was the most primitive and simple arrangement, would seem to imply that those genera with more complex inflorescences, such as Aristea, Nivenia, Bobartia, etc., which according to Weimarck are evolved from this type of spike, were all at a more advanced evolutionary stage than those in the Ixioideae, such as Gladiolus, Antholyza, etc. There is little doubt that the reduction in the number of flowers in the rhipidia or of some of the internodes, such as has occurred in some species of Avistea as well as in other genera, is associated with other variations, such as either the enlargement of the bracts and prophylls or, in some species in which the pedicel is reduced and the flower almost sessile, this is accompanied by an elongation of the ovary or, in those genera in which the pedicel has entirely disappeared, the development of a perianth tube. Throughout the [xzozdeae the flowers are sessile and all have a perianth tube, even though in one or two of the genera, such as Melasphaerula and Geissorrhiza, it is very short. ‘The reduction in the number of flowers has also resulted in the formation of flowers of very much firmer texture with an anthesis of two or three days instead of only a few hours. These apparently correlated variations will be referred to again when discussing the flowers. The bracts and bracteoles, although occasional exceptions do occur, on the whole provide fairly reliable diagnostic characters for use in taxonomic work and deserve more attention than they have so far received. In the majority of the genera in the Ixiozdeae with a spicate inflorescence there is some evidence (though sometimes barely perceptible) of the bracteoles being bifid or bicarinate and in nearly every case they are either slightly or distinctly shorter than the bracts. One group of closely allied genera has been mentioned in Part I in which this order is reversed In Anapalina, Exohebea, Tanaosolen and Tritoniopsis the bracteoles are nearly always longer than the bracts (in Anapalina triticea and A. Burchellit considerably so) but otherwise they do not differ from them. The membranous tricuspidate bracts of Ixia and Tritonia subgenus Dichone, and of one or two other genera, have also been mentioned in Part I. The vestigial traces of bracts in Pillansia Templemanni and the species of Watsonia discussed and the occurrence of cormlets in Watsonia bulbifera seem to indicate that these plants are of comparatively recent origin. Another reduction probably of fairly recent origin is that of the terminal spike in Antholyza ringens L. In this species the long terminal axis overtopping the short lateral spike usually has near its apex only a single barren bract, but occasionally it reverts to its ancestral form and there are two or three bracts, sometimes even a small flower. This monotypic genus is so closely allied to the genus Anaclanthe N. E. Br. that I doubt whether the latter should be maintained 66 ANNALS OF THE SOUTH AFRICAN MUSEUM as a separate genus. In Anaclanthe spicata (Thunb.) N. E. Br. there are three or four short lateral branches with secund spikes borne more or less at right angles to the axis, which terminates in a spike, but apart from this there is no real difference between the two genera of which the corms, leaves and bracts are almost identica]. The flowers of Anaclanthe spicata are slightly smaller than the typical flowers of Antholyza ringens but the size of the plants and flowers of the latter is extremely variable, and in the smaller forms there is little if any dif- ference to be seen in the flowers of these two plants. The disappearance of the terminal spike in Antholyza ringens is probably due to the comparatively recent adaptation of these flowers to bird-pollination, as a result of which the flowers at the top of the spike, which were not so easily accessible to visiting birds, tended to remain unfertilized and finally became sterile and have now almost entirely disappeared. Part III. THE SUBTERRANEAN STEMS 1. Introductory remarks In his Classification of the Flowering Planis Rendle observed, regarding the corms of the Iridaceae, that ‘the corm is especially characteristic of the Iridaceae, varying in size in different genera and species. The size of the corm, its shape and more especially the character of the sheathing scales, afford means of distinguishing the very numerous species of Crocus and Gladiolus.’ Except for the shrubby genera and three or four which have rhizomes, such as Aristea, Dietes, etc., the great majority of the South African genera have corms, and Rendle’s observation is very apt, for those who are familiar with this family can frequently recognize the genus, or in some cases the species, or eroup of species, to which a plant belongs by the characteristic shape or distinct fibrous tunics of the corm. This is not always the case, however, for since the character of the fibrous tunics or sheathing scales of the corm is naturally associated with those of the leaves and bracts, so it is only to be expected that genera which have very similar or identical leaves and bracts should also have very similar or almost identical corms. Examples of some genera which it is difficult or impossible to distinguish by the corm alone are: Synnotia and Sparaxis; Lapeyrousta subgenus Anomatheca and Freesia; Exohebea and Anapalina; Ixia and Tritonia subgenus Dichone; Babiana, Anaclanthe and Antholyza; Homo- glossum and some species of Gladiolus. In all these cases the respective genera linked together are very closely allied and are separated mainly on account of differences in their floral characters. Although there is little doubt that they furnish some distinct evidence of phylogenetic trends in the family, up to the present comparatively little use has been made of the character of the corm in taxonomic work, nor has the mor- phology of these organs been studied in much detail. Foster, in his revision of Geissorrhiza, has used the nature of the corm tunics, which he observed to be associated with other morphological features, as a diagnostic characteristic. MORPHOLOGY, PHYLOGENY, TAXONOMY OF S.A. IRIDACEAE 67 Hutchinson made some use of the nature of the underground stems in the keys to the tribes and genera of the Iridaceae in his Families of Flowering Plants, and to some extent I have also made use of the character of the corm, in combination with other characters, in a revision of the Iridaceae of the Cape Peninsula. One obvious drawback to relying too much on the character of the corm or rhizome alone in taxonomic work is the fact that in herbarium material, more especially in the older collections, these organs are not always present. In 1949, when examining some of the old collections in England and Europe such as those of Linneé, ‘Thunberg, Burmann, Lamarck and others, it was found that in hardly any of the specimens are the corms or rhizomes present. ‘Though this is regrettable, as so many of them are type specimens, it is hardly surprising, since Linné and the early botanists following him used only the floral or sexual characters as a means of identifying the families and genera. An examination of some of the subterranean stems of the South African Iridaceae has shown that more than one type of corm occurs and that there are one or two intermediate stages between a rhizome and corm. The swollen underground stems of several species of the shrubby genera have been described by Adamson and will not be described here, but some of the structures in various other genera which have not hitherto been described in any detail are mentioned in the following pages. 2. Transition from rhizome to corm Of the forty to fifty genera of Iridaceae which occur in South Africa, the three shrubby genera form a small, distinct and fairly homogeneous group, with woody aerial stems with secondary thickening and swollen more or less sub- terranean basal stems in which secondary thickening also takes place. Many of the species in these genera grow in wet or fairly moist situations. Of the other genera only four have been recorded as having rhizomes, the remainder, which constitute the great majority of genera and species, having a corm or an under- ground storage stem which approaches a corm in structure. Three of the genera with rhizomes are in the Iridoideae, namely Aristea, Bobartia and Dietes, and all these are evergreen herbs. Schizostylis, the fourth genus with a rhizome, belongs to the /xzozdeae and differs from the other three genera in losing its leaves, together with the aerial axis, at the end of the flowering season. The subterranean parts of the plant are probably the last to be affected by the various changes which take place in the process of evolution and it is very possible that in Schzzostylis the habitat may account for the persistence of the rhizome. The two closely allied species (perhaps only one species and a variety) are found only in wet situations, along the margins of streams, where the subterranean stem is not subject to any great seasonal fluctuations in moisture. This genus differs from the fairly large and widely distributed genus Hesperantha only in having a rhizome instead ofa corm. The corms of Hesperantha, which seem to provide some evidence of having been derived from a rhizome, are discussed later. 68 ANNALS OF THE SOUTH AFRICAN MUSEUM Arber suggested that there was some degree of probability that in the Iridaceae the rhizome may have preceded the bulb. So far no bulbs have been found in any of the South African genera, but there seems little doubt that the more primitive organ is the rhizome and that the corm is a later development, in all probability evolved as a result of changes in climatic conditions. With the exception of Schzzostylis which is found only in wet places, the few South African genera which have rhizomes are evergreen, and their distribution is significant for they are found only in the coastal or mountain districts with a fairly high and regular rainfall, and not in the dry regions such as the Karoo or Namaqua- land where only the geophytic genera with corms, in which the aerial portions die away at the end of each season, are able to survive. Of the South African Iridaceae the most simple and primitive combination of characters occurs in the genus Aristea, in which the subterranean stem in all the species in which it has been seen is either a short, somewhat woody rhizome with numerous, very short internodes, or else more slender and elongated, with longer internodes and more in the nature of a sucker. The leaves in most of the species are the simple isobilateral equitant type which Arber regards as being the most primitive in the family. In Bobartia, the other South African member of the Szsyrinchinae (in the Iridoideae), most of the species have a somewhat woody rhizome, but in some of them, instead of growing horizontally, the underground stem is vertical and more swollen and in structure it shows distinct indications of a transition to a corm. Bobartia lilacina Lewis, which is considered to be the most primitive member of the genus (see Part IT), has a rhizome which scarcely differs from that of a few species of Aristea. In Dietes, the other member of the Iridoideae with a rhizome, the rhizome is also very similar to that of some species of Aristea, i.e. rather woody with short internodes. The type of corm which occurs in some genera in the Irzdoideae will be dis- cussed later, as it differs in some respects from the type which is most common in the [xzoideae. In most of the genera in the latter group the abbreviated portion of the aerial axis below the ground becomes swollen and forms a new corm each season, while in almost every case the previous season’s corm dies away as its food reserves are used up, though it frequently remains at the base of the new corm as a flat, hard, dry structure. There are three exceptions in this group, namely Schizostylis, which has been mentioned above, Dzerama, in which the corms are perennial and the leaves evergreen, and Pullansia, which is also evergreen. The systematic position of the genus Pillansia has already been mentioned in connection with the inflorescence. On account of certain features observed in the inflorescence it is considered to be intermediate between the Jridozdeae and Ixioideae, and in the subterranean stem of this plant there is also evidence of a link between these two groups, for it presents a rather striking transition from a rhizome to the type of corm which is most common in the Ixiozdeae. Each year a swollen corm-like structure is formed which perennates, and the single long, green, isobilateral type of leaf to which it gives rise is persistent for two or Ann. S. Afr. Mus. Vol. XL. Plate XVII 1-4, Pillansia Templemanni (Bak.) L. Bol. 1, part of underground stem, x 3; i, old flowering axis; il, base of dead leaf; ili, bases of green leaves; iv, base of flowering axis; v, sheathing leaves. 2, longitudinal section of part of underground stem, with leaves removed, x 3; i, short scale leaves. 3, part of underground stem and base of aerial stem, with outer leaves removed, x2; i, base of leaf; ii, short scale leaves; iii, bud; iv, leaf scar; v, scar of sheathing leaf. 4, transverse section of root, X16; 4a, t. s. of part of root, X50; i, cork; ii, epidermis; iil, cortex; iv, endodermis; v, pericycle; vi, phloem. 5, Dierama pendulum (L. f.) Baker, longitudinal section of corm, x 14; 1, contractile roots. 70 ANNALS OF THE SOUTH AFRICAN MUSEUM three years. The growth of the underground stem is in a horizontal direction and in many ways it approximates more closely to a rhizome than a corm. A portion of one of these subterranean stems is shown in fig. 1 on pl. XVII, with the fibrous tunics removed. This represents only a fifth of the actual specimen in which there were twenty-five swollen corm-like bodies, half of them lying immediately below the portion illustrated. The short sheathing leaf produced near the base of each ‘corm’ does not become detached but, after the soft living tissues have died and broken down, the coarse fibrous strands continue to form a protective sheath around the corm. The second leaf produced is the long green one and above this there are two or three very short brown scale leaves, similar to those usually found on rhizomes. The position of these small scale leaves is shown in figs. 2 and 3, the former a longitudinal section of two of the corms and the latter one of the corms and the base of the flowering axis, with all except the scale leaves removed. (For convenience these two are shown in a vertical position but they should actually be horizontal.) ‘The portion of the axis which gives rise to the scale leaves is so very much reduced that no internodes can be discerned in this region. Each corm-body has a single, rather large root with a polyarch cylinder (see fig. 4) and very large vessels. A transverse section of a portion of the root, showing the anatomy in more detail, is given in fig. 4a. The cells of the cortex are pitted and of the endodermis much thickened, and the outer walls of the epidermis are suberised. ‘The vascular bundles of the well-defined stele of the underground axis are concentric. A section of a corm of Dierama pendulum,* the other evergreen member of the Ixtoideae, is shown for comparison on the same plate (fig. 5). As in the case of Pillansia this also was probably derived from a rhizome but it represents a later transition stage and differs in several respects from the simpler corm-like structure of Pzllansia. ‘The corm illustrated was attached to nine living corms growing more or less in a vertical direction one above the other and forming a dense mass. In order to maintain its position or level in the ground the few adventitious roots developed at the base are strong and contractile. The corm gives rise to several leaves, all of which, even the outer ones, are the simple linear isobilateral equitant type with no secondary winging or pseudo-mid-rib, and there are no distinct scale or sheathing leaves. They differ in behaviour, however, for the inner leaves, which are more swollen at the base than the outer ones, become detached at the base, possibly through the formation of an absciss layer, although they remain held in position by the outer leaves which remain attached to the axis after they have died. These outer leaves, although similar in appearance to the inner ones, are probably more in the nature of elongated sheathing leaves. The corms in the majority of genera in the I[xzoideae are of the same type as that of D. pendulum. Vhey differ only in being annual instead of perennial and in having the outer leaves reduced to short leaf-sheaths. In all of these the corm * According to N. E. Brown the name Dierama is neuter, not feminine. MORPHOLOGY, PHYLOGENY, TAXONOMY OF 5.A. IRIDACEAE Te! is formed from the abbreviated base of the aerial axis, with two or three or more swollen internodes and a well-defined stele, the reproduction of the plant being maintained by the development of the bud in the axil of the upper leaf at the commencement of the next growing-period. As far as is known, all except one of the species of Dierama are evergreen. The exception is D. trichorhizum (Baker) N. E. Brown, in which the previous season’s corm dies away as the new one develops. Through this species the section Morphixia of the genus Ixia is closely connected with Dierama and in fact D. trichorhizum was originally described as an Ixia by Baker. It is of interest to note that although so closely related these two genera have quite different distribu- tions which scarcely overlap. Duzerama covers a wide area extending from Tropical Africa to the Transvaal, Natal and eastern Cape Province and as far south-west as the Humansdorp Division, while /xza is confined to the south- western districts of the Gape Province, with one or two species in Namaqualand and the Karoo and only one extending eastwards as far as Port Elizabeth. The transition from a perennial corm, such as that of Dzerama, to an annual one is probably due mostly to changes in climatic conditions, but it may also to some extent be brought about by the conditions of extreme congestion which result from the persistence of the older corms. These are pushed out of position by the growth of the new corms and although they continue to live for several years their lower buds appear to remain dormant and they do not produce new leaves or flowering shoots. ‘These congested conditions may have resulted eventually in the formation of smaller corms, often with fewer leaves, followed by the complete absorption of the food reserves of the older corms as the new ones developed and the dying-off of the leaves and older sterile corms. In the Pillansia specimen described, although twenty-five corm-like structures were present, more than half of the older ones were doubled up beneath the newer ones above and although these structures still contained living tissue their buds remained dormant and no longer produced leaves or made any further growth. In this respect they differ from the perennial swollen subterranean basal stems of the three shrubby genera, in which buds continue to develop and give rise to new aerial shoots for an undetermined number of years. Two views of one of the annual types of corm are shown in figs. 6 and 6a on pl. XX. This is a corm of Chasmanthe floribunda and is among the largest of the corms of this type found in the Ixiozdeae. It is depressed-globose and bears several basal leaves with fibrous bases which are persistent on the corm during the dormant period (the ‘tunics’), though the upper green parts of these leaves die off and disappear at the end of the flowering season. Fig. 6 is a view of the top of the corm, with the fibrous tunics removed; the dark rings are the scars left by the bases of the tunics. Fig. 6a is a longitudinal section of the corm, showing the large, well-defined stele and the bud at the top, on the left of the scar of the old flowering axis, which will give rise to the new flowering shoot at the beginning of the next season. In this species one or two of the outer or lower buds scmetimes develop at the same time and give rise to new plants. As in the 72 ANNALS OF THE SOUTH AFRICAN MUSEUM case of Dierama pendulum, the adventitious roots produced at the base of the corm at the beginning of the growing period are strong and contractile. 3. The tuberous ‘corm of Ferrara The very distinct subterranean stem of Ferraria, which probably represents a transition stage from a rhizome to the type of corm which occurs in many genera in the Iridoideae, such as Hexaglottis, Moraea, etc., differs from all others in the family in having no persistent scale leaves or tunics, and is actually more in the nature of a tuber. ‘The underground parts of two plants of Ferraria ferrariola (Jacq.) Willd. are shown in figs. 1 and 1a on pl. XVIII. In this species the growth is in a horizontal direction, a new, more or less rounded tuber being formed each season while the older tubers of the two or three preceding years to which it remains attached become somewhat flattened as they begin to shrivel. The bud which develops at the beginning of each season is at first enclosed in a short sheathing leaf, but this becomes ruptured as the internode immediately above it begins to swell and usually in the older tubers no trace of it remains. More material of another species, /’. undulata, was available for a more detailed examination, and it was found that the tuber in this species is perennial, and for three or four succeeding years a new bud develops on the upper side of the slightly compressed body. A young tuber with its first shoot is shown in fig. 2 on pl. XVIII. Each shoot has three colourless and slightly fleshy sheathing leaves, produced at the base of the bud, and above them is a fourth, longer, and half-green sheathing leaf which is followed by one or two long isobilateral equitant green leaves. A new tuber is formed by the swelling of the internode between the three basal sheaths and the next two or three leaves. A longitudinal section of a young tuber and shoot showing the position of the leaves and the internode which develops into the new tuber is illustrated in fig. 2a. The basal sheaths consist of parenchymatous tissue, without chlorophyll, and small vascular strands without any sclerenchyma, and as the tuber increases in size it bursts through these sheaths, which shrivel and disappear. ‘Traces of them can be seen at the base of the tuber in fig. 2a. A few ordinary adventitious roots develop immediately above the basal sheaths of the young shoot (shown in a transverse section of a bud in fig. 3) and a single, larger adventitious root, which is contractile, slightly higher up. A contractile root starting to develop is shown in the transverse section in fig. 3a; part of it can also be seen in fig. 3. The leaves and roots borne by the shoot die off and disappear at the end of the first season and no further leaves or roots are produced on the tuber itself, which is probably able to absorb water. The succeeding buds which develop on the tuber are not adventitious but arise in the axils of the original leaves, of which only the scars remain, as can be seen in fig. 4. The bud shown towards the base in this tuber is apparently not in a leaf axil and might be adventitious but it is more likely that it was originally in the axil of the upper of the three basal sheaths and was carried up mam. s. Afr, Mus. Vol. XL: Plate XVII7 1, Ferraria ferrariola (Jacq.) Willd., new corm attached to old corm, x32; 1a, new corm and three old corms. 2-6, Ferraria undulata LL. 2, young corm, x 14; i, colourless sheaths; ii, half- green sheath; iii, leaf; iv, contractile root; v, ordinary adventitious roots; 2a, longitudinal section of young corm, xX 34; 1, contractile root; ii, outer sheathing leaves; 111, leaf; iv, new corm; v, ordinary adventitious root; vi, remains of old sheaths. 3, transverse section near base of new corm, X7; 3a, t. s. near top of same, x7; 1, contractile root; ii, adventitious roots; ili, colourless sheathing leaves; iv, base of leaf. 4, older corm, from above, <2; i, scars of shoots; ii, bud. 5, corm with shoot attached to older corm, x #; i, contractile root; 11, ordinary adventitious roots. 6, mature corm with flowering shoot, x 3. 74. ANNALS OF THE SOUTH AFRICAN MUSEUM out of position by the swelling of the tuber. A new tuber is formed by each bud, which either remains attached to its parent, as shown in fig. 5, or else the thin connection between the two tubers is severed by the action of the contractile root of the new shoot. Sometimes two or three tubers develop in this way, somewhat obliquely one above the other, and in this case no further develop- ment seems to take place in the lower tubers, only the upper producing a new shoot or flowering axis. The tubers in this species grow in a more or less vertical position but are frequently pulled into a horizontal position by the contractile root. It is only after the third or fourth year that the tuber produces an inflorescence, after which it begins to shrivel (see fig. 6). A few ordinary adventitious roots are produced at the base of the flowering shoot, but no contractile roots, and the shoot remains attached to the old tuber until it withers and finally disappears at the end of the season. The base of the flowering axis does not become swollen or form a new tuber, but a bud on the old parent tuber, in the axil of the lowest sheathing leaf of the flowering shoot, provides for the continuity of the plant. There is probably just a sufficient food reserve left in the old and somewhat shrivelled parent tuber to enable the bud to commence its growth at the beginning of the next season. The next stage in development might well be a transition from the perennial structure just described to the annual type of corm which occurs in several genera in the Iridoideae, such as Hexaglottis, Moraea, etc., through the storage of food reserves in an internode of the axillary bud and the complete shrinking and almost entire disappearance of the parent corm or tuber. 4. The single internode corms of Hexaglottis, Homeria and Moraea In the corms of the Iridoideae, as in the rhizomes and corms throughout the family, the branching is sympodial and every year the terminal bud gives rise to an aerial shoot while an axillary bud provides for the continuity of the plant. The axillary buds of the members of the Iridordeae which have been examined are morphologically not quite the same as those of the majority of the [xzozdeae, nor do they develop in the same way. With only one or two exceptions the type of corm which occurs in the Iridoideae is associated with a dorsiventral leaf, whereas in the Jxiozdeae the leaves are the isobilateral equitant type. In most of the genera in the Jxzozdeae the food reserve 1s accumulated in two or three or more abbreviated internodes at the base of the aerial axis and the axillary bud which gives rise to the next season’s shoot only develops the following season, but in genera in the Jrzdoideae such as Moraea, Homeria, Hexa- glottis and one or two others, the whole of the aerial axis dies away right down to the base at the end of each flowering season and the food reserve is stored in an internode of the axillary bud which develops at the same time as the aerial axis, though the terminal growing point of the bud remains dormant until the _ following growing-season. The growing point of the bud is often protected by two or three very small scale leaves in which strong fibrous strands are usually MORPHOLOGY, PHYLOGENY, TAXONOMY OF S.A. IRIDACEAE 75 developed at an early stage, and the whole bud is generally enclosed within a single persistent fibrous scale leaf which arises immediately below the swollen internode. Between this leaf and the aerial axis there are usually a pair of strongly keeled united prophylls which have no fibres and do not envelop the bud except perhaps before it starts to swell. ‘The whole structure, i.e. the base of the aerial axis and the new axillary corm, is enclosed within another fibrous scale leaf, the subtending leaf of the axillary bud, and outside this is a thin, membranous sheath which extends up for a short distance and encircles the base of the aerial axis. In the Ivxioideae, as in the Iridoideae, the first leaf at the base of the corm is nearly always a membranous sheath, either pale golden-brown or colourless. The fibrous scale leaves outside this belong to the previous season’s corm. Some of the corms of the Jridoideae are illustrated on pl. XIX. Fig. 1 is a slightly enlarged longitudinal section of the old and new corms of Gynandriris setifolia (L. f.) Foster, with the old outer corm tunics removed. At this stage, about two months before flowering, the old corm is rather shrivelled and reduced to half its original size as much of its food reserve has been used up. In this, as in all these single internode corms, there is little if any trace of a stele. The terminal bud of the old corm has developed into a shoot with a terminal aerial axis (i) and the terminal bud of the next season’s aerial shoot (viii) 1s already formed at the apex of the swollen internode of the axillary corm-bud (11). As this new corm continues to swell the base of the aerial stem is forced outwards and becomes curved. At the base of the shoot are several fine adventitious roots and above these the first two leaves of the shoot. The outer one (iii) is sterile. It is very thin and colourless and soon becomes membranous. The inner one (iv), which subtends the axillary corm-bud, is already firm and its fibrous strands are well developed. The first leaves of the axillary bud are the united prophylls (vi) which are not shown in this figure but can be seen in a transverse section of the corm in fig. 1a. Above the prophylls, shown in both figures, is the first scale leaf of the axillary shoot (vii), which envelops the new corm and is sterile. The first Jeaf borne at the base of the aerial axis above the corm-bud is a short, thin, membranous sheath which is not shown in either of the figures as it was not observed in this early stage, though it was present in all the plants examined about two months later. The leaf (v) shown in figs. 1 and 1a near the base of the aerial axis is the first green leaf of the shoot and is of the dorsiventral type. The internode below this leaf lengthens as the flowering axis develops and in mature plants the node of the first green leaf is at or near the surface of the ground. ‘ Fig. 2 is a corm of Hexaglottis flexuosa with the outer scale leaf removed. At the time when this was examined the plant was flowering and at this stage all that remained of the previous season’s corm was a very small, flat disc at the base. The membranous remains of the prophylls can be seen between the aerial axis and the new corm. These were removed and are shown in fig. 2a. Fig. 26 Ann. S. Afr. Mus. Vol. XL. Plate XIX 1, Gynandriris setifolia (L. f. ) Foster, longitudinal section of old and new corms, X 2; Ia, transverse section of new corm, <2; i, aerial axis; ii, new corm; iii, membranous sheath; iv, lignified scale leaf; v, base of first green leaf; vi, prophylis; vii, first scale leaf of new corm; viii, growing point of new corm. 2, Hexaglottis flexuosa (L. f.) Sweet, corm with outer tunics removed, x #; i, prophylls; 2a, prophylls, x 14; 26, 1.s. of corm; 2c, part of scale leaf, x 3. 3, Moraea edulis (L. f.) Ker, part of old scale leaf, x 14. 4, M. bituminosa (L. f.) Ker, l. s. of corm, X 14; i, aerial axis; 11, corm partly adnate to axis; iii, lignified base of first leaf; 4a, base of leaf with cormlet attached, x 1$; 4b, same, x %; i, brown, viscid lining; 4c, part of old outer tunic with cormlet attached. 5, Lapeyrousia corymbosa (L.) Ker, corm, X %; 5a, flat base of corm, x 3. 6-7, Romulea bulbocodioides (De la R.) Baker. 6, corm, X14; 6a, same, X%; 65, corm with outer tunics removed, Xx 14; i, lignified base of first green leaf; 6c, same corm with base of leaf removed, x #; i, bud of next season’s shoot. 7, new corm starting to develop, x14; i, base of leaf; 7a, |. s. of same corms, X 24; i, membranous sheath; ii, lignified scale leaf. MORPHOLOGY, PHYLOGENY, TAXONOMY OF S.A. IRIDACEAE 77 is a longitudinal section of the corm and in this it can be seen that the base of the aerial axis has been pushed outwards by the new corm so that, although it is terminal and the corm axillary, the positions appear to be reversed. The first leaf shown on the aerial axis is the membranous sheath which was not shown on the younger axis of Gynandrirts setifolia. The only difference between the corm of Hexaglottis flexuosa and that of Gynandriris setifolia is in the character of the fibrous tunics. In the latter they are pale in colour and more or less of the ‘herring-bone’ type which is found in several species of Moraea (see fig. 3, a portion of a scale leaf of M. edulis). In Hexaglottis the fibres are much coarser and almost black. A portion of a corm- tunic of 1. flexuosa is shown in fig. 2c. The corm of Moraea bituminosa differs in two or three respects from those of Gynandriris and Hexaglottis and most other species of Moraea, although funda- mentally it is the same type of corm. One difference is that all the scale leaves in this species are not sterile but each subtends an axillary cormlet borne on a short pedicel. A longitudinal section of a corm is shown in fig. 4 (pl. XTX), with the old outer corm tunics removed. In M. bdituminosa the leaf which appears to subtend the new corm-bud is not a short scale leaf, as in Gynandriris and Hexa- glottis, but though the base is fibrous and resembles a scale leaf, it is the first long, green, dorsiventral leaf of the plant. When removed from the axis it was found that it does not subtend the new corm, but at its base is attached to the pedicel of a small axillary cormlet. ‘Two views of this are shown in figs. 4a and 4b. (The closely allied species, M/. viscaria, has several of these cormlets with more or less united pedicels.) One of these axillary cormlets still attached to — one of the old outer corm tunics is shown in fig. 4c. The axillary bud which has developed into the new corm appears to be without a subtending leaf, but within the base of the dorsiventral leaf below is a brown, viscid lining and, towards the top of the wide basal portion which enclosed the corm, there are distinct traces of an inner skin. This is shown in figs. 4a and 4b. The explanation for this somewhat anomalous structure appears to be as follows: When the apical bud starts to develop at the commencement of the growing-period cormlets borne on short pedicels arise in the axils of the short scale leaves which have protected the apical bud during the dormant period. ‘The first leaf which develops at the base of the axis also subtends a cormlet borne on a short pedicel which is partly adnate to the base of the leaf. These cormlets might or might not develop and give rise to new plants at a later date. From the number of old cormlets found among the old scale leaves which envelop the corm it seems that many of them do not develop. The bud which forms the new corm and which will give rise to the following season’s aerial shoot has probably arisen in the axil of one of a pair of united prophylls, such as were found in the corms of Hexaglottis and Gynandriris, though not in Moraea bituminosa. It is possible that such prophylls were present originally but that they became completely detached as the corm started to swell and disintegrated into the brown viscid substance lining the base of the leaf below. 78 ANNALS OF THE SOUTH AFRICAN MUSEUM On the aerial axis of M. bituminosa there are nearly always two branchlets at each node (occasionally three), apparently arising in the axil of a single sheathing leaf, though when this is removed it can be seen that the second branchlet (and the third when present) is subtended by one of a pair of brown membranous prophylls. Although the first cormlets formed are on short stalks or pedicels, the new corm itself is sessile and, as can be seen in fig. 4, it is partly adnate to the base of the aerial axis, yet another example of the various fusions and adnations which occur in the family. The new corm has no sterile enveloping scale leaf at the base, as in Gynandriris and Hexaglottis; in other words it is the first internode of the axillary bud which forms the corm, the only leaves which it bears being the very small scale leaves at the apex which protect the apical growing point. The three cormlets seen at the base in fig. 4 probably arose in the axils of the small apical scale leaves of the previous season’s corm. 5. Lhe flat-based corms of Lapeyrousta, Romulea and Hesperantha In the majority of genera in the Iridaceae the corm is more or less globose (sometimes depressed-globose), but in several species of Hesperantha and Romulea, one or two species of Gezssorrhiza, and in all the species of Lapeyrousia in the subgenera Sophronia and Ovieda, they have a very distinct and characteristic shape which Dinter, in describing the corm of Lapeyrousia Vaupeliana, called ‘broadly bell-shaped’. The scale leaves have a flat, circular base and articulated to it is a broadly conical upper part. A corm of Lapeyrousia corymbosa is shown in fig. 5 on pl. XTX, and its flat, solid, fibrous base in fig. 5a. The circular bases of the old scale leaves become detached and eventually disappear with the remains of the old corms, while the upper parts, of which the fibres are usually solid and regularly notched at the base and coarsely and closely reticulate above, are persistent and remain as an outer covering on the old corm. A succession of the upper portions of several old scale leaves can be seen on the corm in fig. 5. In Lapeyrousia the corms themselves, in the species of the subgenera Sophronia and. Ovieda which have been examined, are of the same type as those of Gynan- driris and other members of the Jrzdoideae, that is, they are formed from a swollen internode of an axillary bud which develops simultaneously with the aerial axis. They differ only in shape and in having a small, well-defined stele. In these two subgenera the node of the first green leaf, which is the isobilateral and not the dorsiventral type, is on the aerial axis at or near to the surface of the ground, as in Gynandriris, Hexaglottis and most species of Moraea and Homerza. In the third subgenus of Lapeyrousta (Anomatheca) the corm is of the type most common in the Jxioideae, with coarse or fine reticulate, fibrous tunics and no distinct flat disc at the base, though in one or two specimens examined the corm is slightly flat at the base. There are two or three basal] leaves which arise from nodes on the corm itself, their lignified bases forming some of the fibrous tunics which envelop the corm, and the new growth commences the following season MORPHOLOGY, PHYLOGENY, TAXONOMY OF S.A. IRIDACEAE 79 from a bud in the axil of the upper of these leaves. This subgenus has diverged so far from the other two subgenera that it should perhaps be treated as a separate genus. Its affinity is with Freesva rather than with the other two sub- genera of Lapeyrousia. In Romulea triflora (Burm.) N. E. Brown and two or three other species the corm is similar in shape to that of Lapeyrousia corymbosa but the scale leaves are different, being solid and smooth, and the corm itself is of the [xiozdeae type. The first green leaf of the plant arises from a node on the corm which is the swollen base of the acrial axis, and its axillary bud does not develop at all until the following season. In several species of Romulea the corm is slightly or distinctly obliquely flattened at the base. There is a small or sometimes fairly large, flat, and somewhat rounded projecting portion on one side of the outer scale leaf, and the upper part of the solid concentric corm tunic is very distinctly articu- lated to the flat basal part in this region. The corm of R. bulbocodioides, which is of this type, is shown in figs. 6 and 6a on pl. XIX. Fig. 6d is the same corm with the outer tunics removed, showing the fibrous base of the first green leaf, and fig. 6¢ is the same but with the base of the leaf removed to show the base of the aerial deciduous part of the axis and the bud of the next season’s shoot. At the commencement of the next growing period roots arise from the lateral slightly bulging portion of the corm on one side of the axis, as can be seen in figs. 7 and 7a (pl. XIX). Fig. 7 shows the old corm with the new corm starting to develop. The old outer tunics have been removed and the small fibrous scale leaf above the new corm is the lignified base of the first green leaf of the old corm, removed from the lower corm and carried up by the new corm as it starts to develop. Fig. 7a is a longitudinal section, slightly enlarged, of the two corms. ‘he first leaf which envelops the new corm is a membranous, colourless sheath. It has been mentioned before that the outer sheath of the corm seems almost invariably to be membranous. In many of the species of Lapeyrousia and Hesperantha which have been examined the roots arise only on one side of the axis, growing out between the flat base of the corm tunic and the region where the upper part is articulated to it (see L. corymbosa, fig. 5). The one-sided arrangement of the roots in these species seems to indicate that these corms have originated fairly recently from an underground axis which grew in a horizontal direction, and they probably represent another case of a transition from a rhizome to a corm. As has been mentioned earlier, there is a very close affinity between Schizostylis and Hesperan- tha, the only difference between these two genera being that the former has a rhizome and the latter a corm. Schizostylis probably represents the final stages of a rhizome, in which the aerial portions of the plant are no longer perennial but annual, and the species of Hesperantha with a flat-based corm represent the first stage in a transition from this type of rhizome to a corm. The next stage seems to be the disappearance of the flat base and a transition to the more or less globose type of corm which is typical of the Ixioideae, in which the lignified So ANNALS OF THE SOUTH AFRICAN MUSEUM bases of the green leaves arising from the corm form some of the protective fibrous tunics of the corm. In Romulea bulbocodioides the base of the first green leaf, which arises from the corm, is lignified and similar to the upper portion of the outer scale leaf of the corm. The green upper part of this leaf dies away with all the aerial parts of the plant at the end of the season, but the lignified basal portion is persistent on the corm throughout the dormant period, and finally becomes detached from the corm by a clean-cut absciss layer at the base when the new bud starts to develop (see figs. 6b, 6¢ and 7 on pl. XIX). In many species of Romulea, such as R. rosea and its numerous allies, the corm is almost globose, with solid concentric tunics, and only a slight trace remains of the flat basal disc which is so conspicuous in R. bulbocodioides. In these species there are two or three basal leaves arising from the corm and the aerial portion of the axis is extremely abbreviated, usually only a few millimetres long and not reaching to the surface of the ground. ‘The same reduction of the basal disc has taken place in some species of Hesperantha and the closely allied genus Geissorrhiza. In describing the corms of the latter Foster stated that they are more or less elobose, or ovoid, or even conic in shape, always with a flattened base although in some this is barely perceptible. The flat basai part of the corm tunics in the corms described above might be interpreted as the leaf base, the upper part which is articulated to it being the lignified base of a rudimentary petiole. A reduction of the leaf base takes place until it almost, or finally completely, disappears, and the lignified base of the petiole is sessile on the axis. An analogy might be found in the leaf bases of the three shrubby genera which have been discussed in Part I of the paper. In Klattia partita there is a conspicuous leaf base which projects above the node and is persistent on the axis after the leaf has fallen (see pl. II, fig. 5); in Witsenza maura the leaf base is reduced, especially in the dorsal region (see pl. III, fig. 3); finally in Nivenia Stokoe: (and other species of Nivenia) only a slight trace of a leaf base remains on the axis after the leaf has fallen. (Fig. 7, pl. 11, does not show this as the leaves were cut off just above the base to show the buds.) In several species in the subgenus Ovieda of Lapeyrousia the flat-based corm is associated with a more primitive type of inflorescence (see Part II of this paper), and there is other evidence to suggest that the flat-based type of corm found in the four genera mentioned above is a fairly primitive one. ‘There are distinct indications in some of the species in these genera that it develops into the globose type of corm which is typical of the great majority of genera and species in the Ixioideae. In many genera in the Ixiozdeae, however, the corm has probably evolved along different lines, through an earlier transition from the rhizome with evergreen leaves, to a perennial corm with evergreen leaves such as those of Pillansia and Dierama, which have already been discussed, and from this to the typical annual type of corm which is formed during one season, then passes through a resting-period and dies away at the beginning of the next season as its food reserves are used up. MORPHOLOGY, PHYLOGENY, TAXONOMY OF S.A. IRIDACEAE SI Lapeyrousia, Hesperantha and Geissorrhiza have been placed in the Ixzoideae (or Ixieae) in the various systems of classification, but Romulea, on account of its abbreviated aerial axis and different type of inflorescence, was placed in a different subfamily or tribe by the various botanists who have worked on the family (i.e. in Sisyrinchieae by Bentham and Hooker and others; in Crocoideae by Pax and others). When describing the leaves of some species of Romulea Arber suggested that the affinity was with some genera in the Ixzoideae, such as Hesperantha and Geissorrhiza, and in the structure of the corm as well as the leaves, also in many aspects of the inflorescence and flower, there is evidence of a distinct affinity with some of the genera in the Ixiozdeae. The leaf itself is of the phyllode type which, according to Arber, is elaborated by secondary differentia- tion from the fundamental flat equitant type. 6. The spinous roots of Moraea ramosissima The corms of Moraea ramosissima (L. f.) Druce (M. ramosa Ker) appear to differ so markedly from the typical Moraea corms already described that they deserve some mention here. The peculiar spinous roots borne on the subterra- nean axis of this plant have been described by Scott and were also discussed by Arber (the observations of the latter based on herbarium material), but as more material was at hand for examination than was available to either of them, one or two features were observed which were not recorded by Scott or Arber. This species, the largest and most ramified in the genus, is geophytic and grows in fairly moist, often semi-shaded places; it is sometimes abundant after fires in really damp situations, such as stream-sides or slightly marshy ground. The underground stem of one of the youngest plants examined is illustrated in fig. 1 on pl. XX. This probably represents the first season’s growth but it is not known whether this plant originated from a seed or cormlet, though it was most probably from the latter. A puzzling aspect about it is the length of the vertical underground axis. There are no contractile roots to account for it, and in older plants the axis is considerably shorter and the dense mass of cormlets and roots is situated just below the surface of the ground, as can be seen in figs. 1a and 10. The subterranean axis of the young plant consists of a series of nodes and fairly long internodes, with a cluster of axillary cormlets at each of the nodes, though only the uppermost, which is at the surface of the ground, is in the axil of a subtending leaf. At the base of the first two nodes are ordinary adventitious roots but at the third node there are a few normal adventitious roots and between them and the cormlets three spinous roots have started to develop, growing in an upwards direction. At the fourth node there are no ordinary roots but a single spinous root has started to develop. Fig. 1a shows the lower part of the axis of a slightly older plant which did not produce any flowers and is probably the second season’s growth. At the base are several ordinary adventitious roots, but near the middle of the dense mass of cormlets is a distinct whorl of spinous roots, the majority of them growing Ann. S. Afr. Mus. Vol. XL. Plate XX 1-3, Moraea ramosissima (L. f.) Druce. 1, underground stem with cormlets, first season, x $; Id, same, second season, <2; 1b, longitudinal section of same, third season (flowering), xX 3; i, spinous roots; 11, normal adventitious roots. 2, longitudinal section of large cormlet, x 13; i, membranous sheath; u, lignified scale leaf; 111, inner lignified layer partly adnate to scale leaf. 3, 3a and 3b, remains of membranous sheaths with spinous roots, x 3%. 4-5, AZ. plumaria (Thunb.) Ker. 4, Corm, x2; 4a, part of inner fibrous layer of scale leaf, x 14; 4b, cormlets, x 14. 5, part of inner and outer fibrous layers of scale leaf, x 2; 5a, 1. s. of cormlet, X 2. 6, Chas- manthe floribunda (Salisb.) N. E. Br., corm with leaves removed, from above, x #; 6a, 1. s. of same corm, X¥%; i, scars of leaves; ii, bud. MORPHOLOGY, PHYLOGENY, TAXONOMY OF S.A. IRIDACEAE 83 upwards though one of them is growing downwards. Near the top of the mass of cormlets is one which is considerably larger than the rest and it is almost certainly the corm which will give rise to the next season’s shoot. At the two nodes above are leaves with their sheathing bases ruptured by the pressure of the numerous axillary cormlets. A longitudinal section of the base of a mature flowering plant is shown in fig. 16. The first green leaf, which is the dorsiventral type, arises from the axis just below the ground and above it is a succession of leaves, each subtending a short axillary branchlet bearing numerous cormlets. The branchlets arising from the abbreviated portion of the axis below the ground do not appear to have subtending leaves but in the position they should occupy are the stout spinous roots. The majority of these roots are at first negatively geotropic but when they have reached a certain level, just below the surface of the ground, they curve over and become positively geotropic. There are two cormlets which are considerably larger than all the others and probably one or both of these will give rise to a new flowering shoot the following season. Fig. 2 is a longitudinal section of one of them. The outer sheath which envelops the cormlet is thin, membranous and colourless. These thin outer sheaths have either no visible vascular strands or else very fine strands, apparently without any sclerenchyma, and soon break up and disappear. The second scale leaf of the cormlet is firm and lignified but this also breaks up when the corm begins to develop and either disappears or in some cases parts of it remain and can be seen mixed up with the spinous roots. The spinous roots were described by Scott, who stated that the only difference of importance between them and the normal adventitious roots is that the more internal vessels of the spinous roots are smaller. Arber suggested that the association of one or more of the groups of cormlets with the basket-work of interwoven spinous roots was probably fortuitous but it is doubtful whether this view is correct. In all the specimens examined these spinous roots arise immediately below the clusters of cormlets, and the position occupied by the spinous roots and the apparent absence of subtending leaves to the underground branchlets seems to have some significance. It is possible that the subterranean clusters of cormlets originally arose in the axils of thin scale leaves, but that the soft tissues between the vascular strands became ruptured and disappeared at an early stage, although the vascular strands derived from the axis cylinder remained and continued to.grow. In one or two of the specimens examined the membranous remains of leaves or sheaths were found which seem to lend some support to this view. These are shown in figs. 3, 34 and 3b. In all of them the spinous roots appeared to be growing between two membranous skins. As has been mentioned above, the lower green aerial leaves of the younger plants become ruptured at the base as the cormlets develop. These can be seen in figs. 1 and ra. The fact that the spinous roots, as they start to develop, are almost always negatively geotropic seems to lend further support to this view. As they no 84 ANNALS OF THE SOUTH AFRICAN MUSEUM longer function as leaf veins it is possible that they have undergone various modifications, both internally and externally, and the upper extremities of these roots when they develop are usually positively geotropic and appear to function as normal adventitious roots, which they frequently resemble in appearance (see fig. 3). The lower parts nearer to the axis have developed sharp spines and the function of this portion is probably to form a protective basket-work surrounding the cormlets, just as the lignified strands of the outer scale leaves which persist after the softer tissues have died away, do in the case of the other species of Moraea. Unfortunately an attempt made to cultivate some of the cormlets for observation was not successful, so that up to the present it has not been possible to observe any of the very early stages in the development of the shoots. There is a great deal of variation in the size of the plants in M. ramosissima, as well as in the size of the basal clusters of cormlets and development of the spinous roots. In many of the herbarium specimens examined the spinous roots are fairly short and all are negatively geotropic. ‘The numerous small cormlets and larger cormlets or corms in M. ramosissima are all of the single internode type which is found in other species of Moraea, Hexaglottis, etc. 7. The possible ligular origin of the inner part of the scale leaf of Moraea plumaria The scale leaves of Moraea plumaria (Thunb.) Ker are unusual in the genus, in fact in the family, and seem to lend some support to the hypothesis put forward in the first part of this paper that in the Iridaceae there is an axillary stipular outgrowth from the leaf base to which the base of the leaf is almost invariably more or less adnate. In Moraea plumara there appear to be two entirely different types of scale leaf on the same corm, the outer a basket-like arrangement of a few coarse reticulate fibrous strands (see pl. XX, fig. 4) and inside this a second very different type with wider and more solid fibrous strands, which is shown in fig. 4a. In the older corms, between the inner layer and the corm itself, there are sometimes clusters of axillary cormlets borne on short stalks which usually force their way between the two layers, as can be seen in fig. 4. Some of these cormlets, slightly enlarged, are shown in fig. 40. At first the inner and outer fibrous layers are connected by a layer of slightly fleshy tissue but these thin-walled cells soon break up and completely disappear, leaving the inner fibrous layer attached to the outer only at the base, and appearing like an appendage of the outer. This is illustrated in fig. 5, a longi- tudinal section of a young corm with the corm itself removed so that only the two fibrous layers are shown. A section of one of the cormlets is shown in fig. 5a. In this last figure it can be seen that the axis extends above the first apparently double scale leaf and there is an internode between this and the corm which is enclosed in a brown membranous sheath. ‘That the lower leaf is a single scale leaf there can be no doubt, and the inner of the two fibrous layers might be interpreted as an ‘axillary’ stipule or ligule which, in this species, is lignified. The section of the young corm of M. ramosissima (pl. XX, fig. 2) shows a Plate XXI Vol. XL. Ann. S. Afr. Mus. YZATOVANY : “ARLINNG TDYNY “WNWISVa AWLNYWSYHD “SNNOLYND “YAWSODOAUD | SISAO\LAYSINOW YVNOLRS LL NOWAISOMLNAA “WISSTIWWONY SSNAWWLAd “WINSSOTIOWOH NOWd ISONITUY ‘SNIO\AWAYD “UNI ANY “"NSTOSOYNYVL *YAGAHOXR “SISJONOLIAL VINOSLY NEY ~~~. SNHLNYVYDW ‘SNHLNYISYSHL “WV LONNAS ‘S\RWA VAS “YUARINYLdSYLS yyy Soo wale "NOWAISADNSA = ‘YZINYNOSs1ayD~ ‘yinyavigswaWw = WALNW¥S dSal\~ ‘WISABY4 “(WDAHLYWONY) WIsnoy Aady <1 ‘YZIHWYOSSI3 YHLNYY3d SAH™ WISNnOdhadW) (Saprs we)S) aa Bere me Z\ ¢ 1 SYVALSOZIAIDSG YWHUSBIC >., \ \ ienwatia LOSS ) SIN37S39O14SN “YWISNOSMh3dAWI suoiyo) “YINASL IN “wordy “YATOONIYAS “MES, YVINAAIN nog ON a . [ye] SS" "=" Y3ANWOS em) “WILLY Ryysovd SVU CNUINA Dy |e mea oy ee ee SU ee ee eee ee ‘SALI : OW Rese W “WILY BOG WV N30} “W3ALS\VYY “"S@A02) ‘ywayasrisaop Servd]i *Suis09 »padous Je1se0 Snonprep "png Faro up "SUNOD JO\UL 2294 WEAn P2S°U FLA | YZIM PWweZYY| wary ‘six poIrre sD} ; oS) JEruvo 9s0q0}b $521 20 soy |: abujs Vor}su0d abeis UdIZISUBAL Peyy MS Zo pedojan 2 ; Wh : 9 “2 Sussod BSOQ%y S “SSANWS1 Wav Snondi2ad HALIM saLAHdO’»D “siko por1ss0 Ou 4o “2S0q 2yY Wrost ¢ Ppowsot Sswuw209 “UWI0D «O ocr woa seboqs Vol}, ASVoAL "28eis5 Vol SUDA, + “SQNUHS YO S@y3aH N3AYDYIAANSA 86 ANNALS OF THE SOUTH AFRICAN MUSEUM somewhat similar structure, but in this case the inner layer is completely united with the outer layer of the scale leaf for two-thirds of its length and is only free in the upper third. 8. Relationships of the tribes considered mainly in connection with the underground organs The various types of underground stems found in the South African Iridaceae which have been mentioned in the preceding pages are arranged in tabular form on pl. XXI, with the genera possessing these different types of organ placed in the columns below. In drawing up this table one other factor has been taken into consideration, that is, whether the inflorescence is spicate or not. The great importance attached to the type of inflorescence in this family has been discussed in Part II of this paper. It is necessary to consider the inflorescence in conjunction with the subterranean parts of the stem as the changes and reductions which have taken place in the aerial parts of the plant, which finally led to the development of the spicate inflorescence in the [xiozdeae, are associated with changes in the underground parts of the stem, although this portion of the plant has probably been the last to be affected by these changes. Just as a few intermediate or transition stages exist which provide some clues to the lines of development in the aerial parts of the plant, so also are a few still found in the subterranean organs. In the genera Pillansia and Lapeyrousia the intermediate conditions observed in the inflorescences are correlated with what are considered to be transition stages in the subterranean axis. All the evidence provided by the aerial and subterranean parts of the plants indicates that the most primitive members of this family were perennial much-branched evergreen herbs, or possibly small shrubs, with erect aerial stems arising from swollen basal stems which developed in a horizontal direction on or just below the surface of the ground. In the diagram (pl. X XI) the types of underground stems have been divided into eight groups, four of them having aerial leaves which are persistent for more than one season, and four in which the aerial leaves are deciduous. Of these eight groups, four are considered to represent distinct transition stages (see columns 3, 4, 6 and 7). The rhizome with evergreen leaves is regarded as being the most primitive type of subterranean stem and the others are believed to have evolved from this. Aristea is considered to be the most primitive genus found in South Africa. The next is Bobartia, especially the two branched species, B. lilacina Lewis and B. paniculata Lewis, which are mentioned in Part II. In one or two of the species of Bobartia with a very much reduced inflorescence there is distinct evidence of the rhizome undergoing a transition to a corm. This genus has therefore been placed in two of the columns. Dietes, although it still has a rhizome and evergreen leaves, and an inflores- cence similar in some respects to that of Bobartia lilacina, shows a considerable advance in the structure of the flower, which is like that of Moraea. There is undoubtedly a close affinity between Dietes and Moraea and in the diagram this MORPHOLOGY, PHYLOGENY, TAXONOMY OF. S.A. IRIDACEAE 87 is indicated by a dotted arrow. ‘The type of corm found in Moraea and its allied genera in column 5 has been described. It is the same in all the genera in this column and in all of them it is associated with a dorsiventral leaf. No actual intermediate conditions between a rhizome and a corm have been observed in the large genus Moraea but it has been suggested that the type of development in the ‘tuberous corm’ of Ferraria might indicate the lines along which the corm of Moraea, etc., has evolved. All these genera, i.e. Aristea, Bobartia and Dietes in column 1, Bobartia and Ferraria in column 3, and Hexaglottis, Homeria, Moraea, Gynandriris and Galaxia in column 5, belong to the Jridoideae. (According to Pax, Galaxia belongs to the Crocoideae; the reasons for including it with the Jridoideae will be given in the final part of this paper.) The limits of this tribe, number 1, are shown by the double lines in the diagram. ‘This is regarded as being the most primitive tribe of the Iridaceae. In this tribe, though some of the genera have a rhizome and others a corm, there is less evidence of the transition from the former to the latter than there is in the more advanced Jxiozdeae. The origin of the three shrubby genera in the second column has been the subject of a good deal of discussion, Marloth and a few other botanists regarding them as being of exceedingly ancient origin and possibly representing relics of an older flora. They exhibit certain features, however, which are more advanced than those of Avistea, as in the inflorescences which in the shrubby genera give some evidence of having undergone further reductions than those of some of the more primitive species of Arzstea, as well as in the sessile flowers, all with a perianth tube (in Alatiza very short). They form a small but very distinct group which Weimarck placed in a separate tribe, Nivenieae (together with three genera found in South America and one in Australia). I agree with Weimarck in regarding these shrubby genera as constituting a distinct tribe and consider that this tribe, which is confined to the Southern hemisphere, is inter- mediate between the Jrzdoideae and Ixioideae, though nearer to the former. It is very possible that at one time this tribe was considerably larger than it is to-day. The three South African genera, although they have very similar subterranean stems and leaves, are quite distinct from each other, with very different flowers which are arranged in various types of reduced complex inflorescences. Their development has probably been greatly influenced by their habitat; the majority of the comparatively few species which occur in South Africa (one in Witsenia, two in Alattia and eight in Nivenia) grow in moist situations, mostly along stream-sides or in marshy places, so that on the whole they have not been as greatly affected by seasonal changes in moisture and temperature as nearly all the other genera in the Jridoideae and Ixioideae. It is possible that some of the more primitive features have survived in the vegetative parts of these plants and their characteristic swollen, more or less subterranean stems might even represent a more primitive type of subterranean axis than the rhizome. Adamson, who studied the stems of the shrubby genera, noted that although secondary thickening takes place in the swollen basal stems of these 88 ANNALS OF THE SOUTH AFRICAN MUSEUM plants, as in their aerial stems, there is no sign of it in the rhizomes of some of the species of Avistea which he examined. The Ixioideae (No. III in the diagram) is the most advanced tribe and is by far the largest. As will be seen in the diagram, the great majority of the genera fall into the last column, in which the more or less globose corm is formed through the accumulation of food reserves in the abbreviated subterranean portion of the aerial axis. In these genera in column 8 the inflorescence is spicate (except in Romulea, Syringodea and Crocus—the last not a South African genus), and in all of them the aerial part of the axis as well as the aerial parts of the basal leaves arising from the corm disappear at the end of the flowering- season. There are a few exceptions in the [xiozdeae which provide some distinct clues as to the lines of development in this large group. One of the most important is the monotypic genus Pillansia. ‘The inflorescence of this genus was discussed in Part IJ. Although the inflorescence is not spicate, but more or less intermediate between the type of inflorescence found in one or two species of Nivenia and a spicate inflorescence, and the subterranean stem is intermediate between a rhizome and corm, Pillansia cannot rightly be included with the Jridozdeae, nor with the Nzvenieae, and is best placed in the Ixiozdeae. It represents the most primitive genus in this tribe but is undoubtedly closely allied to Watsoma (indicated by a dotted arrow) and almost certainly represents a survivor of the ancestral type of that genus. ‘There is some reason to believe that the corm of Watsonia passed through an intermediate stage like that of Dierama, that is a perennial corm with evergreen leaves. The corm of Dierama has been discussed, and its transition to the annual type found in D. trichorhizum. ‘This genus has therefore been placed in two columns, 4 and 8, connected by an arrow, though all except one of the species belong in column 4. The affinity between Dierama and Ixia (section Morphixia) has also been mentioned; this is indicated by a dotted arrow. The only genus in the Jxiozdeae with a rhizome is Schizostylis, though this rhizome differs from those in the Iridoideae in having deciduous aerial leaves. It has been suggested that the survival of this primitive type of subterranean axis is probably due to the habitat, in wet situations, and that it represents an intermediate condition, being the final stages of a direct transition from a rhizome to a corm without passing through intermediate stages such as those in Pillansia and Dierama. There is a very close affinity between Schizostylis and Hesperantha (indicated by a dotted arrow), and the flat-based type of corm found in Hesperantha, as well as the other genera in column 7, is also regarded as a transition stage, which in turn passes into the more or less globose corm which is typical of the great majority of the [xioideae. As both types of corms are found in Hesperantha this genus has been placed in columns 7 and 8, as are also Geissorrhiza, Romulea and Lapeyrousia. Geissorrhiza is closely allied to Hesperantha, and in this slightly more advanced genus there are only one or two species with a distinct flat base, the majority of the species falling into the last column. MORPHOLOGY, PHYLOGENY, TAXONOMY OF S.A. IRIDACEAE 89 The close affinity between Schizostylis and Hesperantha and the transition from the rhizome of the former to the corm of the latter leads one to believe that the flat-based corms of Romulea and Lapeyrousia probably evolved along similar lines, but that in these two genera the transition to a corm took place at a slightly earlier stage, before the inflorescence had reached its final reduction to a spike, whereas in Schizostylis the spicate inflorescence was evolved before the final transition from a rhizome to a corm. Pax’s tribe Crocoideae is indicated by dotted lines, group Ila. This tribe, according to Pax, consists of the genera Galaxia, Romulea, Syringodea and Crocus. It is not proposed to discuss this group at present as it will be mentioned in more detail in the next part of this paper; in my opinion it is an artificial one consisting of one member of the Iridoideae (Galaxia) and three of the Ixiozdeae. It will be noticed that Lapeyrousia has been placed in three positions. Only the subgenus Anomatheca belongs in column 8, but, as pointed out before, in the subgenus Ovieda there is more than one type of inflorescence, many of the species having a simple or branched spicate inflorescence, while in others there is a -more primitive type, with intermediate stages between the two. All the genera placed in column 7 should almost certainly be included with the Ixzoideae. A system of classification based on correlated morphological characters is suggested in Part IV; for the present it is intended merely to point out that the type of corm in Romulea and Syringodea is the same as that of a few genera in the /xiozdeae, and not like that of the [rdozdeae. Part IV. ‘THE FLOWERS AND SYSTEMS OF CLASSIFICATION 1. Introductory remarks This final part is divided into two sections; the first deals rather briefly with the flowers and one or two controversial taxonomic groups and the second with the phylogeny and classification. In the second section the correlated morpho- logical characters of the vegetative and floral organs are considered in connection with a proposal for an amended system of classification which is not entirely in accordance with any of the previous systems. In the introduction to this paper it was pointed out that a great deal of confusion has resulted from relying too much on the morphology of the flowers alone, without taking into consideration the morphological character of the other parts of the plant. It is not intended to underestimate the value of the floral characters for there is no doubt that they are of great importance to the systematist, but in this family it is not always possible to differentiate the tribes or genera merely by the shape or form of the flowers or their styles and stamens. Where this was done in the past it resulted sometimes in an artificial grouping of a number of species which actually belonged to two or three or more different genera, as in the case of Antholyza, Acidanthera, Ixia and Tritonza. It is felt that the floral characters of the Iridaceae are on the whole fairly well known, since great importance has been attached to them in systematic gO ANNALS OF THE SOUTH AFRICAN MUSEUM work, and it is not proposed therefore to analyse in detail the morphological features of the flowers, but only to discuss one or two of them, and to mention a few of the anomalies which have been observed in the flowers of some of the genera. In the more primitive genera in the Jridozdeae, though there is a great deal of variation in the type of inflorescence in some of the genera, the flowers in the various genera are as a rule fairly homogeneous, although variations do occur, such as the occasional development of a perianth tube in a single species of a genus which normally has completely free segments, also a tendency to a reduc- tion of the inner whorl of the perianth segments in one or more species in one or two of the genera. In the more advanced Jxiozdeae the inflorescence in almost all of the genera is reduced to its simplest form, either a simple or branched spike (except in Pillansia and some species of Lapeyrousia), but a greater degree of variation occurs in the flowers in some of the genera and it is this group, with its intermediate forms and borderline species apparently connecting some of the genera, which presents the most problems for the systematist. In many of the larger genera there are almost invariably one or two species in which the flowers do not conform with the description given of the genus, although the vegetative characters are usually more or less typical. The flowers of these genera sometimes cover a fairly wide range, varying from the simple, more primitive forms to more highly evolved forms, and often provide important phyletic indications of actual affinities as well as of evolutionary trends. SECTION I. [HE FLOWERS 2. Symmetry A character to which some importance has been attached in systematic work is the symmetry of the flowers. The Iridozdeae and Crocoideae present no problems in this connection as the four genera placed in the Crocoideae by Pax all have actinomorphic flowers and throughout the Jridoideae, with probably only one exception, the flowers are also actinomorphic. The only genus which is known to have zygomorphic flowers in this tribe is endemic in Australia (Dzplarrhena). In the Ixtozdeae, however, both actinomorphic and medianly zygomorphic flowers occur, and it has been found that in some of the genera previously described as having either one or the other, both types occur though usually one or the other predominates. For instance, in Lapeyrousia and Babiana both types occur but in the majority of species in these genera the flowers are zygo- morphic. The genus Babiana has been divided into two subgenera, the one with subequal perianth lobes and more or less actinomorphic flowers, the other with more or less unequal perianth lobes and zygomorphic flowers, but an examina- tion of all the species in this genus has proved this arrangement too artificial to be upheld as the first group merges into the second through intermediate stages. In Hesperantha and Getssorrhiza nearly all the species have actinomorphic flowers with equilateral stamens, but a few exceptions have been recorded in rE ————— ee Ann. S. Afr. Mus. Vol. XL. Plate XXII (i ‘la 1, Watsonia marginata (L. f.) Ker, flower, front view, x2; 1a, perianth tube with upper half opened to show base of filaments (i) and staminodes (ii), x 1$; 16, part of perianth tube, side view, X is; 1, base of filament; 11, staminode; iii, callus. 2, Gladiolus species (aff. G. trichonemi- folius), flower laid open, x #; 2a, perianth tube with stamens and stigmas, x 1}; 26, bract, x 3; 2c, bracteoles, x $; 2d, corm, x 3. 3, Watsonia Pillansii L. Bolus, perianth tube opened to show filaments and staminodes, < 2; 3a, part of perianth tube with base of filament (i) and staminode (ii), x13. 4, Moraea Cooperi Baker, flower, x $; i, pedicel; ii, ovary; ii, perianth tube. 5, Tritonia scillaris (L.) Bak., stamen, front view, x 34; 5a, same, side view; i, opening. 6, Babiana stricta (Ait.) Ker, stamen, front view, X 1%; 1, connective. 7, Exohebea lata (L. Bol.) Foster, anther, front view, x 1}; 7a, same, back view. 8, E. parviflora (Jacq.) Foster, stamen, back view, x 14; 8a, same, front view. 9, E. lata, style branches and stigmas, x 14; 9a, same, from above. 10, Moraea plumaria (Vhunb.) Ker, style branches and stigmas (i), x 13. 92 ANNALS OF THE SOUTH AFRICAN MUSEUM which the stamens and style are more or less arched or all directed to one side, as in Hesperantha grandiflora, Geissorrhiza rochensis and G. Lewisae. Although the position of the style and stamens can be seen clearly in living flowers, it is often obscured or lost in dried material, from which many of the original descriptions were drawn. The asymmetry of the stamens in flowers which are otherwise regular is undoubtedly the first step towards the more highly evolved types of medianly zygomorphic flowers in which the perianth tube is curved and the flowers tend towards a horizontal position with the stamens and style arched under the uppermost lobe which often becomes expanded and hooded, as in the great majority of species of Gladiolus, or elongated, as in Chasmanthe, Anomalesia and one or two other genera. In the more highly evolved flowers of these genera the elongation of the uppermost perianth lobe is accomplished at the expense of the other lobes which are nearly always very much reduced. This is apparent in some species of Gladiolus, especially in the orchidiflorus group, as well as in Antholyza, Anomalesia and Kentrosiphon. In Anomalesia the three lower lobes are extremely reduced and remain curved inside the perianth tube (see pl. X XIII, fig. 4), while in Kentrosiphon a pouch has developed below the three lower lobes and only a trace remains of the lowest lobe, the four lateral lobes being very much reduced and only the conspicuous scarlet uppermost lobe well developed (see pl. XXIII, fig. 5). In most species of Watsonia, although the uppermost perianth lobe is usually arched above the stamens, it is scarcely at all expanded or wider than the other lobes. There are one or two species of Waisonia in which the flowers are very nearly regular, as in W. marginata, where, except that the perianth tube is slightly curved, the flowers are actinomorphic (see figs. 1 and 14 on pl. XXII). In Gladiolus, the genus with the greatest number of species and certainly the greatest range of distribution (South and tropical Africa, Europe, and Asia), there are in South Africa a few species in which the flowers have subequal perianth lobes and are only very slightly zygomorphic, as well as one or two in which they are definitely actinomorphic. Botanists have hesitated about placing these anomalous species with actinomorphic flowers in the genus, yet the corm, leaves and bracts, as well as the flattened alate seeds, are typical of the genus, or rather of the particular group of species to which they are clearly related within this large and heterogeneous genus. One example which might be given is the plant known as Geissorrhiza Patersoniae L. Bolus. Foster, in his revision of Geissorrhiza, followed L. Bolus in treating this as a Gezssorrhiza, but placed it in a separate subgenus. In my opinion, however, although it obviously represents a link between Gerssorrhiza and Gladiolus, it should rather be placed in the latter genus near to G. edulis and G. permeabilis, from which species it differs only in having actinomorphic flowers. (Incidentally this species was first collected by Thunberg and named by him Gladiolus elongatus. N. E. Brown upheld this species although he did not associate it with Gezssorrhiza Patersoniae.) Another borderline species which MORPHOLOGY, PHYLOGENY, TAXONOMY OF S.A. IRIDACEAE 93 seems to connect these two genera is one which has not so far been described. The flowers are completely regular (see figs. 2 and 2a on pl. XXII), yet all its other characters, such as the flat alate seeds, the bract, bracteoles and corm (see figs. 2b, 2c and 2d) are typical of the group of Gladioli comprising the species G. linearis, G. erectiflorus and G. trichonemifolius. If the flower alone were to be used as a diagnostic characteristic this plant would be placed in the genus Geissorrhiza, although the other characters do not conform with those of that genus. 3. The perianth tube Another floral character which should be used with discretion is the length of the perianth tube. As was pointed out in a paper published by myselfin 1941, elongation of the perianth tube occurs in many genera, such as Lapeyrousia, Ixia, Babiana, Gladiolus and even Romulea. A characteristic of the last-named genus is the short, funnel-shaped perianth tube, yet two or three species with a long perianth tube have been found in South Africa. Very often intermediate forms connect the short and long-tubed forms so that this character alone is not a satisfactory or sufficiently limiting factor to use in subdividing a tribe or genus. Although the length of the perianth tube is of relative unimportance, the presence or absence of a perianth tube is an important character. In the Ixioideae the flowers are sessile and all have a perianth tube, even though in some cases it is very short, as in Melasphaerula and a few species of Geissorrhiza. In the Crocoideae both Galaxia and Syringodea have very short pedicels and long perianth tubes. In the South African members of the /ridoideae, excluding the shrubby genera Alattia, Witsenia and Nivenia, the flowers are always more or less pedicel- late and the perianth segments nearly always free or almost so. There are, however, three anomalous species in this group in which the flowers, as well as being distinctly pedicellate, have a short but quite distinct perianth tube. These three species, belonging to three different genera in which the perianth segments are described as free, have been the cause of some confusion owing to the fact that, in the case of two of them at any rate, the original descriptions were made from poorly dried herbarium material in which the flowers were evidently not dissected, for no mention was made of the perianth tube. One of these is Bobartia macrospatha Baker (syn. B. tubata Gillett), which has a pedicel up to 5 mm. long and a perianth tube up to 1 cm. long. Gillett has pointed out to me that the plant he described as B. tuwbata was later found to be the same as Baker’s B. macrospatha. He had dissected a flower of the type and found that it had a perianth tube although Baker made no mention of it in the specific description and the generic description in the Flora Capensis seemed to deny its existence. Another species, which necessitates some amendment to the generic descrip- tion, is Hexaglottis virgata (Jacq.) Sweet. This has a pedicel about 5 mm. long and also a perianth tube 5 mm. long, although in the generic description in the Flora Capensis the perianth is described as ‘cut down to the ovary’. In the Flora 94. ANNALS OF THE SOUTH AFRICAN MUSEUM of the Cape Peninsula it is regretted that I also omitted to observe the presence of a perianth tube in this species. The third species is undoubtedly a Moraea although it appears to be quite anomalous in that genus. At present it is known by three different names, Moraea Coopert Baker, Gynandriris stenocarpa (Schlitr.) Foster (Moraea stenocarpa Schltr.) and G. apetala (L. Bol.) Foster (Moraea apetala L. Bol.), but the three type specimens have been examined and there is no doubt that they all represent the same species; so that Baker’s name, the first given to this species, must be retained and the other two become synonyms. Baker’s description is incorrect and quite misleading. He described the flowers as ‘lilac’, whereas they are pale yellow, though they usually turn mauvish when dried. He made no mention of the perianth tube and described the inner perianth segments as ‘oblanceolate’, but in this species there are no inner segments! Working with poorly dried specimens it is very possible that he mistook one of the style crests for an inner segment. He also described the spathes as two- to three-flowered, whereas they are one-flowered, a character which is probably unique in the genus. In this respect, as well as in having a perianth tube, Moraea Cooper: differs from all other species in the genus. The most complete description, from living specimens, was given by L. Bolus, but after having examined a large number of living specimens it was found that there is one point which requires clarification. In describing Moraea apetala in 1929 Dr. Bolus stated that the ovary is produced into a beak like a perianth tube, while in all the specimens examined by myself in 1950 it was found that there was an articulation at the top of the ovary, below the short cylindrical perianth tube which is continuous with the segments (see fig. 4 on pl. XXII). The genus Gynandriris (Helixyra Salisb. ex N. E. Br.) is separated from Moraea on account of its membranous spathes and the beak-like elongation above the ovary, with the perianth segments articulated to its apex, so that if L. Bolus’s description were correct then this species should belong rather to the genus Gynandriris; this combination was actually made by Foster. In Moraea Coopert the spathes are not membranous, nor does the ovary extend up in a short, sterile projection, so that this species cannot be included with Gynandriris but should remain in the genus Moraea, where it should be placed in the subgenus Vieusseuxta which is characterized by having the three inner perianth segments much smaller than the outer, sometimes tricuspidate or, as in M. tripetala, extremely reduced and filiform, sometimes barely perceptible. In M. Barnardi L. Bol. they are com- pletely suppressed. M. Coopert and M. Barnardii are the only two South African species so far known which have only three perianth segments, with no trace of the inner whorl. One of the characters distinguishing Moraea from Irs is the presence of a perianth tube in Js and the absence of one in Moraea, so that M. Coopert, with its short perianth tube, provides a link between these two closely allied genera. It was mentioned in Part II of this paper that the reduction from a compound inflorescence to a simple spicate one appeared to be correlated with other MORPHOLOGY, PHYLOGENY, TAXONOMY OF S.A. IRIDACEAE 95 variations. One of these was that the reduction in the number of flowers resulted in the production of a smaller number of flowers of very much firmer texture, each individual having a flowering period of a few days instead of only a few hours. Another variation to which reference was made was that the reduction of the internodes of the branches and of the pedicels was sometimes accompanied by an elongation of the ovary, as in the section Pseudaristea of Aristea, or by the development of a perianth tube. In the three anomalous members of the /rzdoideae mentioned above considerable reductions have taken place in each case. In Hexaglotiis virgata the internodes of the rather numerous lateral branches are completely or almost completely suppressed, so that the flower-clusters or rhipidia (each two-flowered) are sessile on the axis and the inflorescence has the appearance of a spike. In Moraea Coopert and Bobartia macrospatha the cymes are one-flowered, and in both, and Hexaglottis virgata, the pedicel is considerably shorter than is usual in these genera, so that the ovary, instead of being exserted, remains enclosed in the spathes or bracts. In a discussion of the morphology of the genus Nivenia Weimarck made the following statement: “The tube of Nivenia flowers corresponds biologically in some Aristea species to the pedicels (e.g. A. alata), in other species to the cylin- drical or prismatic ovary (sect. Pseudaristea).’ This interpretation of the perianth tube is probably correct, and the slender, stalk-like sterile extension of the ovary in Gynandriris, and observed also in one or two species of ferraria, seems to provide some evidence of a preliminary step towards the formation of the perianth tube. It is very probable that this stage is followed by the transition of the point of articulation from the top of the ‘beak’ to the top of the functioning part of the ovary, so that the sterile upper part, with no articulation between it and the segments, becomes a part of the perianth instead of an extension above the ovary. Part of the perianth tube in the [xzozdeae, possibly the whole of the lower portion of the tube up to the point where the stamens are inserted, was probably derived in a similar manner from the pedicel, the upper part, which is present in many of the genera, being formed by the fusion of the lower part of the perianth segments. In some genera the stamens are inserted near the top of a slender cylindrical tube (e.g. Exohebea, Thereianthus and several species of /xza), in others at or near the middle of the tube, as in Watsonia, Gladiolus, etc., but in either case the portion of the tube below the stamens is slender and more or less erect. When the tube is curved the bend is almost invariably at the junction of the upper and lower parts, as in Watsonia and Gladiolus. In some cases the lower part passes gradually into the upper part which is often funnel-shaped, as in some species of Gladiolus; in others the perianth tube is clearly differentiated into two distinct portions, the upper either funnel-shaped, as in many species of Gladiolus, or broadly cylindrical as in many species of Watsonia, also in Homoglossum, Anapalina and Chasmanthe. . In two species of Chasmanthe, C. aethiopica and C. Peglerae, the differentiation of the perianth tube into two distinct parts is further emphasized by the develop- Ann. S. Afr. Mus. Vol. XL. Plate XXIII 1-2, Chasmanthe acthiopica (L.) N. E. Br. 1, spike, x 3, arrows indicate direction of twist; 1a, part of perianth tube of bud, x 2; 15, part of perianth tube of open flower, with spiral twist in basal portion, X2; 1c, bract, x #%; 1d, bracteoles, x. 2, capsules, x43) 2a, OpenGansalew a. 2b, seed, x 14; 1, point of attachment; 2c, transverse section of seed, x 14; 1, thick yellow testa. 3, Anapalina revoluta (Burm.) N. E. Br., flower, first stage, x3; 3a, same, second stage. 4, Anomalesia Cunonia (L.) N. E. Br., flower, x 2; i, three lower perianth lobes folded inside tube. 5, Kentrosiphon saccatus (Klatt) N. E. Br., flower, x 2; i, lowest perianth lobe. MORPHOLOGY, PHYLOGENY, TAXONOMY OF S.A. IRIDACEAE Q7 ment of three pouches at the base of the upper part. Nectar, secreted in the ovary, passes up the short slender basal part of the perianth tube and collects in these pouches. The stamens are inserted at the base of the upper part of the perianth tube and the basal part becomes spirally twisted as the flowers open. In this respect it behaves somewhat in the same manner as the pedicel and ovary, or ovary alone, of some genera in the Orchidaceae. The secund position of the fully open flowers in C’. aethiopica is brought about by the twisting of the basal part of the tube. The buds, as they first emerge from the axis, are more or less distichously arranged but slightly twisted, one of the upper side lobes being opposite the subtending bract. As the bud begins to open the basal part of the tube begins to twist in towards the axis. This is shown in fig. 1 on pl. XXIII. At one stage the half-open flowers face towards the axis and for a short time they are inverted but the tube continues to twist and when the flowers are fully open the perianth tube has more or less completed a spiral twist. ‘The ovary and lower half of the perianth of a bud, before the basal part of the tube has started to twist, is shown in fig. 1a, and the spiral twist in the tube of a mature flower in fig. 1b. In C. floribunda the flowers are more or less distichously arranged in the spike and the perianth tube is less twisted than in C. aethiopica, and in this species there is only one rather shallow pouch at the base of the broadly cylindrical upper part of the tube. 4. The genus Chasmanthe As there has been some difference of opinion about the status of Chasmanthe and two or three of the other genera established by N. E. Brown twenty years ago, it seems advisable to discuss them briefly here. Chasmanthe is one of the six new genera described by Dr. Brown in 1932 when he divided up the species which had previously been included in Antholyza. In a note on N. E. Brown’s subdivision of the genus Antholyza, published in 1941, E. P. Phillips commented as follows: “The writer considers that Brown has succeeded in clearing up much of the prevailing confusion though he has gone too far in proposing so many genera. In the writer’s opinion the genera Petamenes Salisb., Kentrosiphon N. E. Br., Chasmanthe N. E. Br., and Anomalesia N. E. Br. should be grouped together under the oldest name Petamenes Salisb. All these genera are characterized by the elongated upper perianth-lobe which is differently shaped to the other five lobes and are concave or hooded. This appears to be a very natural grouping, but the subsidiary characters used by Brown to separate the genera, e.g. the saccate perianth-tube of Kentrosiphon, the reflexed perianth-lobes of Anomalesia, and the slight differences he gives between Petamenes and Chasmanthe do not warrant generic status.’ The fallacy of relying entirely on floral characters for grouping species together in this family, more especially in the Ixzozdeae, has already been pointed out in this paper. In placing Chasmanthe with Petamenes, Kentrosiphon and Anomalesia, Phillips has taken into consideration only the floral characters of these genera, 98 ANNALS OF THE SOUTH AFRICAN MUSEUM which are somewhat similar. Hutchinson, in his Families of Flowering Plants, published in 1934, also grouped Chasmanthe with Petamenes though he maintained both Anomalesia and Kentrosiphon as separate genera. I cannot agree with Phillips or Hutchinson, and consider that Brown’s genus Chasmanthe, with the exception of three of the species which he placed in this genus, 1s well defined and quite distinct from Petamenes. ‘Two or three important characters do not agree with the generic description of Petamenes given by Phillips. For instance, he described the corm as ‘small, globose or sub-globose’ ; this applies to the corm of Petamenes but the corm of Chasmanthe is not small, that of C. floribunda being among the largest in the family, nor is it globose. Two views of the somewhat flattened corm of C’.. floribunda, with the corm tunics removed, are illustrated on pl. XX, figs. 6 and 6a. The capsule Phillips described as ‘ellipsoid, shorter than the bracts’, and the seeds as ‘many, flat, broadly winged’, but again this applies to Petamenes only, for in Chasmanthe the capsule is somewhat rounded and considerably longer than the rather short bract and the few seeds produced are fairly large and oval, with a smooth, shining, thick and brightly coloured testa. They are distributed by birds and not by wind, as are the winged seeds of Petamenes. ‘The capsule and seeds of C. aethiopica are illustrated in figs. 2 and 2c on pl. XXIII. The testa in this species is bright yellow or orange. The three species which should be excluded from Chasmanthe are C. fucata (Bak.) N. E. Br., C. caffra (Bak.) N. E. Br. and C. intermedia (Bak.) N. E. Br. Both Hutchinson and Phillips upheld Brown’s genus Curtonus, but L. Bolus considered that it should be combined with Chasmanthe, according to the follow- ing note made by her in 1933 (S.A. Gard. and Country Life XXIII, p. 46): ‘Another genus which rests upon too weak a basis to be stable is Curtonus, containing only one species, C. paniculatus, and occurring in the Orange Free State, Transvaal and Natal. It was placed by Baker next to what is now Chasmanthe fucata, from the Khamiesberg (Namaqualand), and there seems to be no ground for upsetting this arrangement.’ There is no doubt that Dr. Bolus is correct in recognizing that these two species should not be separated, but in my opinion they should be placed together in the genus Curtonus, which, although closely allied to Chasmanthe, seems to be sufficiently distinct to merit generic status. The other two species, C. caffra and C. intermedia, should be transferred to the genus Anapalina. The corms, leaves, bracts, fruits and seeds of these two species are all typical of the genus Anapalina and none of these organs are the same as those of Chasmanthe. ‘The corms are deep-seated, as in Anapalina, and enclosed in wiry reddish-brown fibrous tunics which are not reticulate and extend up in a long neck. These characters do not agree with those of Chasmanthe, though they are typical of Anapalina and the closely allied genus Exohebea. The difference between the leaves of Anapalina and Chasmanthe was pointed out in the first part of this paper. The leaves of C. caffra and C. intermedia are typical of the former genus, with only a few prominent primary veins, instead of numerous veins and MORPHOLOGY, PHYLOGENY, TAXONOMY OF S.A. IRIDACEAE 99 a pseudo-mid-rib as in Chasmanthe. ‘The bracts and bracteoles of C. aethiopica are shown in pl. XXIII, figs. 1c and 1d; the bifid bracteoles are of the type most commonly found in the Jxzozdeae. In C. caffra and C. intermedia the bracteoles are slightly longer than the bracts and they are of the type peculiar to Anapalina and Exohebea, which were discussed in the first part of this paper. The fruits and seeds also are quite unlike those of Chasmanthe, described above, and are exactly like those of Anapalina. N. E. Brown’s description of the seeds of Anapalina was incorrect, as was pointed out by L. Bolus: ‘Dr. Brown describes the seeds as “flattened, winged all round”: those I have seen are four-sided, each face being sculptured, or bearing processes, and being surrounded by a wing.’ It was only on account of the analogous flowers, with the uppermost lobe much longer than the other five, that the two species just mentioned were placed in Chasmanthe. In Anapalina the perianth lobes vary from subequal, as in A. triticea, to distinctly unequal. In A. revoluta there is a clear indication that the uppermost lobe tends to become elongated and in some of the forms of this widespread and variable species it is considerably longer than the other lobes, as can be seen in the flower illustrated in fig. 3 on pl. XXIII. In some specimens collected in the Bredasdorp Division the uppermost lobe is even longer than in the flower illustrated. ‘The flowers are protandrous and a later stage with the perianth lobes recurved is shown in fig. 3a. With regard to two of N. E. Brown’s genera, Anomalesia and Kentrosiphon, I am inclined to agree with Phillips in considering that Brown created too many genera, and consider that Kentrostphon and Anomalesia should be combined under the latter name. The only difference between them is the presence of the pouch or spur in Kentrosiphon and the reduction of the two upper lateral perianth lobes. The corms, leaves, bracts, fruits and seeds of these two genera are the same. On pl. XXIII, figs. 4 and 5, are shown a flower of Anomalesia cunonia and one of Kentrosiphon saccatus. Whether Phillips is right in sinking both genera in Petamenes I am not prepared to say until there has been an opportunity of examining more material of Petamenes, but it is very likely that Anomalesia (including Kentrosiphon) is sufficiently distinct to be maintained as a separate genus. L. Bolus remarked of Petamenes abbreviatus, on which the genus was founded: “This barely escapes being placed in Homoglossum and is one of the critical species that are difficult to place: if it were not for the short side-sepals and oblique tube-mouth it could be put into Homoglossum.’ Homoglossum, in its turn, is extremely closely allied to Gladiolus and in fact there are two or three records of a natural hybrid between Homoglossum Priortt and Gladiolus maculatus. The whole group of the most advanced members of the Iridaceae, which Hutchinson placed together in the tribe Antholyzeae, presents more problems than any other and requires further critical study. The tribe is actually rather an artificial one based entirely on the fact that the flowers are, on the whole, more zygomorphic than in the Gladioleae tribe, but it is very doubtful whether there is sufficient justification for this distinction, This matter will be discussed again in the final section. 100 ANNALS OF THE SOUTH AFRICAN MUSEUM 5. Ihe stamens Throughout the family the number of stamens is three, but occasionally in abnormal flowers one or other of the stamens of the inner whorl is developed. The genus Watsonia is the only one known to me in which, in three or four of the species, there are what appear to be rudimentary staminodes at the base of the inner perianth lobes, alternating with the filaments at their point of attach- ment to the perianth tube. The staminodes of W. Pillansi are shown in fig. 3 on pl. XXII and in more detail in fig. 3a. In figs. 1 and 1a on the same plate are shown the staminodes of W. marginata. In this species, in addition to the structures which are presumed to be staminodes, there are small callus-like projections below the filaments of the three stamens as well as below the staminodes, those below the staminodes being smaller and more or less adnate to the staminodes (see fig. 15). In one or two species of Lapeyrousia (e.g. L. divaricata and L. anceps) there is a small callus on each of the three lower perianth lobes and similar calli occur in one or two species of Trztonia (e.g. T. securigera). ‘These calli should probably be interpreted as ligular or stipular in origin. In many species of Moraea (also in Ferraria and Homeria) there is a small shallow nectary at the base of the three outer perianth segments, and in one or two species, such as M. tricuspidata (L. f.) Lewis and M. confusa Lewis, there is a very small flap at the base of the outer segments which is clearly ligular. In almost all members of the Iridaceae the anther cells are parallel and slit open from top to bottom to shed their pollen. Exceptions occur in Ferraria, in which the anther-cells in some species are distinctly divaricated, and also in the subgenus Dichone of Tritonia in which the cells open only by a small slit at the base (see figs. 5 and 5a on pl. XXII). In this as well as in other characters the subgenus Dichone differs from the other subgenera of Tritonia, and in my opinion it should be established as a separate genus. It is far more closely allied to Jxza than to Tritoma. In Babiana stricta and its varieties and allied species the anther-cells are widely separated by the broad connective (see fig. 6 on pl. XXII). A characteristic feature of the stamens in Exohebea is the apiculus, formed either by an extension of the connective or, as in EF. lata, by an extension of the anther-cells (see figs. 7 and 7a on pl. XXII). An exception in this genus is E. parviflora, in which, as can be seen in figs. 8 and 8a, there is no apiculus. Heterostyly has been recorded in two genera of the South African Iridaceae, in Nivenia, and also in one species of Gezssorrhiza, namely G. heterostyla L, Bol. It is possible that it also occurs in Syringodea but it is necessary to examine more material in this genus before any conclusion can be reached. In a few species of Geissorrhiza, though the flowers are not heterostylous, the filaments in each individual flower are of different lengths, two long and one considerably shorter. This perhaps is a step towards heterostyly. The flowers are protandrous in three or four genera, as for example Exohebea and Anapalina. a ———_ MORPHOLOGY, PHYLOGENY, TAXONOMY OF S.A. IRIDACEAE IOI One other important feature in connection with the stamens should be mentioned. In some genera in the Irdotdeae the filaments are completely connate in a tube round the style, as in Homeria and Ferraria. In Moraea they are more or less connate at the base. In Galaxia, in the Crocoideae, they are also completely connate and in /xia monadelpha in the Ixtoideae they are connate in the lower half. With the exception of the various cases mentioned above the anthers or stamens do not present many characteristics which are of much value for diagnosis. 6. The style and style branches The structure of the style branches shows considerable variation and these afford useful characters which have been largely used in systematic work for the distinction of genera or small groups of genera. Many of the different forms have been illustrated by Pax and Marloth and others who have worked on this family, so it would be superfluous to discuss them in any detail here. On the whole the structure of the style branches within the different genera is fairly homogeneous so that this character is a fairly reliable one, although a few exceptions have been recorded in some of the genera. In species in which these exceptions occur a comparative examination of the vegetative organs will, as a rule, remove any doubt as to the genus to which the plant belongs. In two or three species of Lapeyrousia, for instance, such as L. Vaupeliana Dinter and L. effurcata Lewis, the style branches are simple instead of bifid, yet the corm, leaves, etc., are typical of the genus, or of the particular group of species within the genus. Two exceptions also occur in Geissorrhiza in which genus the style branches are characteristically short, recurved, narrow and conduplicate; in G. Mathewsi L. Bol. and G. Mathewsii var. eurystigma the style branches and stigmas are broad, flat, thick, and crenate-edged. In Exohebea the stigmas are retuse (see figs. 9 and ga on pl. XXII) but in most of the specimens of £. ramosa which have been examined they were found to be shortly but distinctly bifid. The style branches and stigmas of Moraca plumaria, shown in fig. 10 on pl. XXII, are quite different from the crested style branches which characterize the genus. A few other anomalies might be mentioned but those which have been commented on in connection with the various organs of the flowers will serve to confirm the statement made in the introduction to this paper, that it is often impossible to describe any large group in this family, such as a tribe or genus, without some proviso to cover the exceptions which occur. ‘They also serve to stress the necessity of taking into consideration the morphological characters of the entire plant. SECTION II. SysTEMs OF CLASSIFICATION The different opinions held with regard to the classification of the Iridaceae have already been mentioned in the introduction. To recapitulate, the three main systems generally recognized are briefly as follows: vaALsiay *snoi\20bn WLW oer Cees Scent Uzi “Bisse sAvmo} 4 “ay oads jou woussos0t ft “Ssqnayuc Yiluysog “sqnays 10 SowozryUsr : ; AV al NIAIN Z WEBB: Ssqrouy vaesbrong ——— h st, np Mien a= e-ink Bae = 4 Se ee eee =e {|_| sata ‘SQaay usesbsvrg Yilyydo>d *Us109 CF BWIOTZ YA WW yalE "SOAT9] joI100 SnoN snA \WOS Webewad “pirap yim seyhydoey WALNUY 343 = Sa ae Soe *SaA0d\ YAGOONIYAS SLLISTOYYSH Yo1se0 = =snonpisep pub Vix) S209 ype sahudoary WIYAWOK vavsow YZ\WYBOSS\39 / pales NA YuanInvidaurc “>1Ndsowoulzr2°qns 210 oh ENS VIEvAUID Aanonaec DYdaowou 13 90 Saomo\ | (YINOLIBL) , smolduID AW INY BEAL ee ce go a ee -—[- —Qysehgus SE Seay (oat ee ee, eee NS a ee phe sede : SORNOON Se ee a eee | ee eee SIXKVAVAS WNoled ;Wnwwractnaw SHWLNG DNA / VINOSIUNS SISdO1L3WOL NOW) sayd 10 of fz Qns 30 smo WIS3a ys WRIA avNW re ce WIAWLUWO RY (N;OSOYNVL) rYdsowobhz Sx9Mo) 4 YILONN AS WINoOLIss (WISMOYARANT) W3SavOXS YNViIeVa NOWASONIAYYE ISnNOYA32441 ; nD Pp mea eS gies Lage te | : Taye ¥ 2 YINoslyM Base fa ae i SE ee, a tee Bale te ee ee an ee ee. Bomar p10 ou| -s3s0.> pry! BHLNYTOUNY WNSsorpowoy pris tronaieua> | SNMOL END (wowarsoOLways) | “aaenids Ou MOEN faduys Sua |-120h 29 Pasig [arm plojsyed VZAITOWLNY BYLNYUWSYHD VISA WhoNny SaNaWy 1 ad Bee IO AS ee ee SWav 21N{SA Stusv shag eee “Syuoyd sso "PYFIQ Swso a1Ayo ISa\osyxr0ag Word Burqbuse9| © “Swi! 2 Ris { ae “ways hyr0e | -yisnoyA3dN jo Saweds ewes uy EERO Coes) SUBMWe}c | ay\soddo SUBWO}S { ‘a\duais SWALO aFiS “BOD ids B2VDSALO\ FUT frome yde2%o ayouds DOMIISOAOL UT ards Suse yUT ; aed \ SES "PE Sie 2€ =I qt Ea ‘ “YK D Ma qdooxs ae wa 6 : ‘ 5 eqn UdIAe "Saw uy u O | snoizvb nf jou oa yy ryrad YIN — -BNISSeS Ssavmoy 4 Wsnoysady] f° Seeds 1 aaa asic “Snois e6ry easeoja | | wos puo YACOINIYAS VaInNwoy Y \SNUTVN ul ad aDKXo ayooids OU SIS310} eb : Sq eH “zynords Jou BWwuasssstopuy —-“Squoy ‘| = 2 ~~ San axl] € Gh s/c at ys 3 Ge AIXX °3?I1d AX TON SHINY Se Uby MORPHOLOGY, PHYLOGENY, TAXONOMY OF S.A. IRIDACEAE 103 (1) Bentham and Hooker’s classification of the Iridaceae in Genera Plantarum (1883). The family is divided into three tribes, Moraeeae, Stsyrinchieae and Ixieae. The first two tribes are subdivided into several subtribes. This system is followed by Baker in the Flora Capensis (1896) and by Diels in the second edition of Engler and Prantl’s Pflanzenfamilien (1930), with only one or two minor alterations. (2) Pax, in the first edition of Engler and Prantl’s Pflanzenfamilien (1889), raised one of Bentham and Hooker’s subtribes of the Stsyrinchieae to subfamily status, Crocoideae, and combined the remaining subtribes of Sisyrinchieae with those of Moraeeae, treating this also as a subfamily to which he gave the name Iridoideae. This subfamily he subdivided into several tribes and subtribes. ‘The third group, corresponding with Bentham and Hooker’s Jxzeae, he also raised to subfamily status, [xzoideae, and divided into three tribes. This system is followed by Rendle in his Classification of Flowering Plants (1904) and by Marloth in: The Flora of South Africa (1915). (3) Hutchinson, in his Families of Flowering Plants (1934), divided the family into eleven tribes of equal status. Weimarck, in 1940, placed the shrubby genera together in a separate tribe, Nivenzeae, this bringing the number of tribes up to twelve. In attempting to work out a phylogenetic diagram to illustrate the natural relationships of the South African members of the Iridaceae the various morphological features discussed in this paper were all taken into consideration, such as the type of inflorescence, the presence or absence of a pedicel, the nature of the subterranean stem, the form of the leaf and bracts, and the structure of the flower and its organs. Only the genera which occur in South Africa are recorded in the diagram, which is given on plate XXIV. Although the position of the majority of the genera which occur in other countries is fairly clear from the descriptions, it is not easy to place all of them satisfactorily without having examined them in person; since it has not been possible to do this I do not feel justified in attempting to place them all in their correct positions. The fairly numerous genera in South Africa include the most advanced as well as some of the most primitive found in the family; in addition there are many genera which display characters intermediate between the latter and the former and thus provide some evidence of the lines along which most of the more highly evolved genera have developed. These South African genera are representative of the whole family, and it is possible therefore to work out from them the relationships of the genera and the groups to which they belong. It was found that they fall fairly naturally into a number of groups, most of them distinct and well defined. By taking into consideration the correlated morphological characters in the various groups they in turn can be divided into three major groups, two of them large and consisting of three or more of the minor groups, and one small group by itself which is morphologically quite distinct from the others. It is proposed 104 ANNALS OF THE SOUTH AFRICAN MUSEUM to treat these three major divisions as tribes, Jrideae, Nivenieae and Ixieae, and the smaller groups in two of the tribes (i.e. Jrideae and Ixieae) as subtribes. In the diagram the boundaries of the tribes are indicated by double lines and those of the subtribes by single lines. On the whole, with some alterations, the arrangement arrived at approxi- mates most closely to that of Pax, though Pax’s subfamily Crocozdeae has fallen away and Weimarck’s tribe Nivenieae is retained as distinct from the Jrideae. Pax combined the shrubby genera with Avistea, treating this group as one of the tribes of the Iridoideae. ‘The tribes [rideae and Ixieae correspond fairly closely with Pax’s subfamilies Iridozdeae and Ixioideae, though with two important changes, namely the removal of the shrubby genera from the former and the division of the four genera of the Crocoideae between the two tribes. Pax’s subfamily Crocoideae presents a few problems and seems to be an artificial group linking the Jrideae and Ixieae together. The four genera placed in this group, Galaxia, Syringodea, Romulea and Crocus, were separated from the Irdoideae and Ixiordeae respectively only on account of their abbreviated axis and reduced types of inflorescences, but if all the morphological features of these plants are taken into consideration it becomes apparent that the affinity of Galaxia is with the Jrideae and of the other three genera with the Ixieae, and that although similar reductions have taken place in these plants it is very unlikely that they have developed along the same direct line. : Galaxia has the same type of corm as Moraea and its allied genera, with corm tunics similar to those of some species of Moraea. ‘The dorsiventral leaves of Galaxia, although reduced and very short, are also of the same type as those of Hexaglottis, Moraea, etc. It differs from most other members of the Indeae only in having undergone further reductions, resulting in the almost complete suppression of the internodes of the axis and lateral branches, as well as a reduction in the number of flowers in the cyme. Although the flowers have a comparatively long perianth tube they are not sessile, for in all the specimens examined a short pedicel was found, sometimes up to 5 mm. long, though more often only 2 to 3 mm. long. Three species of the Jrideae have been mentioned in which there is distinct evidence that a similar reduction is taking place. In Bobartia macrospatha and Moraea Coopert the cymes are one-flowered, the pedicel considerably reduced, and the flowers have a short perianth tube. In Hexa- glottis virgata similar reductions have also occurred and the two-flowered cymes are sessile in that species and the flowers have a short perianth tube. In Galaxia the cymes are sessile and one-flowered, with the inner bract or spathe consider- ably shorter than the outer one. |The outer spathe resembles the leaf which subtends the cyme and between the cyme and the axis is a pair of united prophylls. Another character of Galaxia which links this genus more closely to Homeria and its allied genera is that the filaments are united around the style and the anthers opposite the style arms, instead of alternating with them as they do in Romulea and Syringodea. In Moraea, Homeria, Iris and allied genera (i.e. Bentham MORPHOLOGY, PHYLOGENY, TAXONOMY OF S.A. IRIDACEAE 105 and Hooker’s tribe Moraeeae) the style arms are opposite the stamens, but in all the other genera they alternate with them and are above the septa of the ovary. Pax suggested that the second position is due to a twisting, the original position as in Moraea, etc., being the normal, namely above the dorsal suture of the carpels. This phenomenon requires further investigation, though it is very probable that Pax’s theory is correct. Bentham and Hooker attached a great deal of importance to this character and it was their main basis for establishing Moraceae and Sisyrinchieae as separate tribes. The morphological characters of Romulea and Syringodea differ in several respects from those of Galaxia but are the same as, or very similar to, those found in a few species of Hesperantha and Geissorrhiza in the Ixieae. ‘The corms have been described and their similarity to those of Geissorrhiza and Hesperantha discussed. The structure of the leaves and the evidence they provide of an affinity with the Ixieae were discussed by Arber. Although the inflorescence in Romulea and Syringodea is not spicate as in nearly all the genera in the Ixeae, it has undergone very extensive reductions, especially of some of the internodes of the main axis, and the flowers are not arranged in cymes as in the Irideae. In Romulea the comparatively long lateral branches, like the main axis, all terminate in a solitary flower, which has a perianth tube but no pedicel, and is subtended by a bract with what appears to be a pair of united bracteoles opposite, instead of a single inner bract or spathe as in Galaxia. In some species of Romulea the bifid or bicarinate nature of the bracteoles is fairly evident, in others not or barely perceptible. In one or two species of Hesperantha and Getssorrhiza reduced inflorescences very similar to those of Romulea occur, with the main axis and lateral branches all terminating in a solitary flower. Still further reductions have taken place in the inflorescence and entire plant in Syringodea, although this genus possesses some characters which are regarded as being more primitive than those of Romulea, such as the simple style branches (bifid in Romulea) and the presence of a short stipe below the ovary. The latter probably represents the final remnant of a pedicel, such as is found in Galaxia, but in Syringodea the flowers are not in sessile one-flowered cymes as in Galaxia. Instead each flower has a subtending bract with a pair of united bracteoles between the flower and axis, but no outer and inner spathes as in Galaxia. The rather brief analysis of the characters of Galaxia, Romulea and Syringodea serves to illustrate the fact that Galaxia is morphologically distinct from the other two genera and it is for this reason that it is proposed to abolish Pax’s Crocozdeae and to place Galaxia in the Irideae, in the same subtribe as Homeria to which it is probably most closely allied, and to place Romulea and Syringodea together in a separate subtribe with the Ixzeae. Crocus, the fourth genus of the Crocoideae, is confined to the Northern hemisphere, where some species of Romulea also occur. It is undoubtedly evolved from the same ancestral stock as Romulea, through further reductions in the internodes of the aerial axis, accompanied by an elongation of the perianth tube, as in Syringodea. As Arber remarked: “The extreme reduction of the whole Crocus plant and the length of the flower tube 106 ANNALS OF THE SOUTH AFRICAN MUSEUM may be regarded as characters in which the genus has progressed further than Romulea on the path of specialization.’ Crocus should clearly be placed in the subtribe of Jxieae with Romulea and Syringodea. The main basis for separating the three tribes is the inflorescence, but there are other correlated characters as well. In the Jrzdeae the flowers are pedicellate and arranged in cymes (rhipidia) with an outer and inner bract or spathe. The perianth segments are usually free, though in a few of the genera there is a perianth tube. In a few species in two or three of the genera and all species of Galaxia the cymes are reduced to a single flower, this reduction being correlated with a reduction in the length of the pedicel and the development of a perianth tube. In Galaxia the perianth tube is long and the pedicel very short. In a few species of Aristea the pedicels are extremely short (about 1 mm. long), though throughout the genus the perianth segments are free. The plants are herbs, with a rhizome and evergreen leaves, or a corm and deciduous aerial leaves. The corm, in nearly all of the genera which have a corm, is formed through the accumulation of food reserves in a single internode of an axillary bud which develops at the same time as the aerial axis, the aerial axis dying away right down to the base. This type of corm is associated with a dorsiventral leaf. In the Nivenieae the flowers are sessile, all have a perianth tube (very short in Klattia), and are variously arranged in reduced compound inflorescences, not in rhipidia as in the /rzdeae, nor in spikes as in nearly all of the Ixzeae. The plants are small perennial evergreen shrubs with a large basal, more or less subterra- nean stem and woody aerial stems bearing numerous distichous imbricate leaves of the simple isobilateral equitant type; secondary thickening takes place in the underground and aerial stems. The three shrubby South African genera in this tribe form a fairly homogeneous systematic group with many morphological features in common which are distinct from those of the other two tribes. The flowers in the Ixieae are sessile, all have a perianth tube, and in nearly all of the genera they are arranged in spikes, in almost every case with a pair of united bracteoles between the flower and the axis. Exceptions occur in Pullansia and Lapeyrousia in which there is evidence of a transition from a more complex inflorescence to a spike. In Romulea and Syringodea, one or two species of Hesperantha and Geissorrhiza, and also in the subgenus Sophronia of Lapeyrousia, the whole inflorescence is considerably reduced and the aerial axis extremely abbreviated. In Syringodea there is a short stipe or pedicel below the ovary. In Exohebea, Anapalina and two allied genera the bracteoles are not derived from a pair of united bracts or prophylls as in the other genera. All except one of the genera in the Jxzeae have corms, nearly all of which are formed through the accumulation of food reserves in two or more abbreviated internodes of the basal underground portion of the aerial axis. Pillansia and Dierama are evergreen herbs but in all the other genera the aerial leaves are deciduous. ‘The leaves vary considerably in structure but all are derived from the isobilateral equitant type. Dorsiventral leaves do not occur in the Ixzeae. MORPHOLOGY, PHYLOGENY, TAXONOMY OF S.A. IRIDACEAE 107 The genera in the Jrideae are segregated into three subtribes, marked 1a, 16 and tc in the diagram (pl. XXIV). The confusion resulting from the misinter- pretation of the tribe Szsyrinchieae was mentioned by Weimarck when discussing the systematic position of Aristea in his monograph of that genus. He placed both Aristea and Bobartia in the Sisyrinchieae and this arrangement is followed here, except that this group is treated here as a subtribe, Szsyrinchinae (1a). The other two subtribes correspond more or less with two of Pax’s subtribes of the Tridoideae, namely Cipurinae (16) and Inidinae (1c), the only change of any impor- tance being the inclusion of Galaxia in the former. It is rather doubtful whether there is sufficient justification for Pax’s division of the tribe T2gridieae into two subtribes (Tigridinae and Cipurinae), but this can only be decided when it is possible to examine material of the various genera concerned. The classification in the Jrdeae is based almost entirely on the form and differentiation of the flowers. In the Szsyrinchinae the style branches are simple and alternate with the stamens, while in the other two subtribes the style arms are opposite the stamens and variously modified, i.e. bifid, crested, flattened, etc. The Iridinae are separated from the Cipurnae on account of the more elaborate style arms of the former, which are flattened, winged and petaloid, with con- spicuous bifid crests and the stigmatic surface on a small projecting flap below the crests. The largest genus in the Jrzdinae is Iris, which is widely distributed in the Northern hemisphere; Moraea, closely allied to Jris, is well represented throughout South Africa and slightly less so in Tropical and North Africa. Pax divided the Jxioideae into three tribes, but Bentham and Hooker, Baker, Diels, and one or two others, grouped all the genera together and treated them as a single tribe. Hutchinson also segregated these genera into three tribes, though his tribes are not the same as Pax’s. One of them, Antholyzeae, was established by Hutchinson for the most advanced genera of the Iridaceae. ‘This last arrangement will be referred to again shortly; it is not followed here as it is considered to be an artificial one, based entirely on the symmetry of the flowers. As will be seen in the diagram (pl. X XIV), it is proposed to divide the Ixzeae into six subtribes. One of these, marked 3d, has been discussed already. Excluding Galaxia this corresponds with Bentham and Hooker’s tribe Croceae, or Pax’s subfamily Crocoideae, but since the conception of this subtribe is not in accordance with the earlier ideas it would probably be better to rename it and the name Romulineae is suggested as being the most suitable. Romulea is a large genus and is well represented both in the Northern and Southern hemi- spheres. ‘The plants in this subtribe are all small and very much reduced, with an extremely abbreviated axis (except in one or two South African species of Romulea), and the inflorescence is not spicate, the flowers being borne singly and each arising between its subtending bract and a pair of united bracteoles. : The name Watsoniineae could be adopted for the subtribe 36 since this cor- responds almost exactly with Pax’s tribe Watsonieae. he inflorescence is spicate, except in several species of Lapeyrousia, and the style branches bifid in all the genera, the only exceptions being two or three species of Lapeyrousia. 108 ANNALS OF THE SOUTH AFRICAN MUSEUM In Lapeyrousia and Watsonia, especially the former, there is some evidence of the inflorescence being derived from a more primitive ramified one. This has been discussed in the section dealing with the inflorescence. The subgenus Anomatheca has diverged so far from -the other two subgenera of Lapeyrousia that it should probably be treated as a separate genus, closely allied to Freesza. In the diagram the range covered by Lapeyrousia is indicated by placing this genus in two positions, linked together by a bracket. This same method is employed with some of the other genera, such as Watsonia, Gladiolus, Babiana, etc., to indicate that various progressive stages occur in these genera. Fairly detailed descriptions of the subterranean stem and inflorescence of the monotypic genus Pillansia have been given in this paper and it has been pointed out that the subterranean stem provides some evidence that this plant is derived from a perennial evergreen ancestral stock, either a herb with a rhizome or possibly a small shrub with a thick undezground stem such as occurs in members of the tribe Niventeae. In the inflorescence also there is evidence that a reduction has taken place, the presence of vestigial traces of bracts suggesting an affinity with Nivenia. It is considered that Pillansia represents an intermediate stage in development between the tribes Nivenieae and Ixteae, but that it is nearer to the latter and should be included in that tribe. It seems to be sufficiently distinct from the other genera in the Jxzeae to be placed in a separate subtribe by itself (marked 3a in the diagram), the most primitive in the Jxieae, but directly linked with the tribe marked 36 through Watsoma. ‘The only applicable name for this subtribe is Pillanstineae. A group of morphologically closely related genera, which is also regarded as constituting a distinct subtribe, are the genera in the column marked 3c (Tritoniopsis, Exohebea and Anapalina). ‘This subtribe is characterized by having a deep-seated corm, with tunics of coarse, wiry, fibrous strands, not reticulated and extending up in a long neck, by the bracteoles being longer than the bracts, and by the leaves which, in their external appearance as well as their internal structure, are quite distinct from those of any other members of the Jxieae. ‘The structure and development of the basal leaves and the rudimentary cauline leaves of some species of Anapalina and Exohebea have been discussed. In these genera the bracts have developed adnate to a pair of prophylls and the bracteole is not derived from a pair of united prophylls or bracteoles as in the other genera in the Ixieae. It has been suggested that two species which N. E. Brown included in the genus Chasmanthe (i.e. C. caffra and C. intermedia) should be transferred to Anapalina as all the morphological features of these two species are the same as those of Anapalina and, except for the similarity of the flowers, quite distinct from those of Chasmanthe. By transferring these two species to Anapalina this genus extends up into the ranks of the most advanced members of the Jxzeae in which the flowers are very irregular, with three or five of the perianth lobes much smaller than the uppermost. The name Exohebineae is suggested as being the most suitable for this subtribe. There is little doubt that Brown’s monotypic genus Tanaosolen should be combined with Exohebea, the only difference between them being the fact that MORPHOLOGY, PHYLOGENY, TAXONOMY OF S.A. IRIDACEAE 109 the perianth tube is long in Tanaosolen and short in most species of Exohebea. It has been pointed out that long- and short-tubed flowers occur in several genera in the /xieae and this character alone is not considered to be of much importance. In most species of Exohebea the perianth tube is short but there are two species with a long tube as well as one with a tube intermediate between the long and short forms. Another small group of genera with a combination of morphological features in common which do not occur together in the other members of the Jxieae are Babiana, Anaclanthe and Antholyza. They differ from the other genera in having deep-seated corms with reticulate fibrous tunics (quite distinct from those of Anapalina and Exohebea) and plicate leaves which are divided into two distinct portions, a long, narrow sheathing base and an expanded pseudo-lamina. In most species of Babiana the stem, leaves and bracts are pubescent; in Anaclanthe and Antholyza only the stem and bases of the bracts are pubescent. Another character peculiar to this group is the very short, compact and secund spike, with the internodes between the flowers extremely reduced. This type of spike occurs in Antholyza, Anaclanthe and the majority of species of Babiana. These genera, which have clearly developed along the same line, the most advanced being Antholyza, seem sufficiently distinct from the other genera in the Jxieae to warrant being placed in a separate subtribe (marked 3/ in the diagram), to which the name Babianuneae could be given, since this is by far the largest and most widely distributed genus in the group. Plicate leaves occur in one other genus in South Africa, in Curtonus in the Ixieae, but the corm and corm tunics, bracts, inflorescence and flowers are not the same as those of Babiana, Anaclanthe and Antholyza. According to Arber, plicate leaves also occur in two or three genera in the Jrideae which are not found in South Africa. Arber commented on this occurrence of foliated leaves in the two main groups of the family, which she considered to represent a case of parallel development. The subtribe of the /xzeae marked 3e in the diagram is a large one, but it was found that any attempt to divide it into subtribes, as was done by Pax, and toa further extent by Hutchinson, results in an artificial grouping of the genera based entirely on the development or symmetry of the flowers. Pax divided this group (which included Babiana and Antholyza, the latter a mixture of several genera as was later pointed out by N. E. Brown) into two tribes, Jxieae and Gladioleae, the division following the horizontal line in the diagram between the groups of genera with actinomorphic or subactinomorphic flowers and those with zygomorphic or subzygomorphic flowers. Hutchinson distributed the genera of Pax’s tribe Watsonieae between the tribes [xieae and Gladioleae, though he went a step further than Pax in establishing the tribe Antholyzeae in which he placed all the most advanced genera which are shown in the uppermost section in the diagram (i.e. from Anapalina to Antholyza), all of which have the three or five lower perianth lobes considerably smaller than the uppermost. Although the more primitive genera near the base of this group, i.e. Dierama, Schizostylis, Hesperantha, Geissorrhiza and Ixia, are quite distinct from the advanced IIo ANNALS OF THE SOUTH AFRICAN MUSEUM genera at the top, various progressive stages in development can be traced in this group where it is evident that many of the genera are closely linked together. Dierama and Ixia are closely allied, as has been pointed out, and in Dzerama there is a transition from an evergreen plant with a perennial corm to a geophyte with deciduous leaves and a corm which does not persist for more than a year. Schizostylis and Hesperantha are also very closely allied, the only difference being that the former has a rhizome and the latter a corm. The flat-based corms which occur in some species of Hesperantha (as well as Romulea and Lapeyrousia) have been discussed and it was pointed out that these give some indication of a transi- tion from a rhizome to a corm. The borderline species between Geissorrhiza and Gladiolus have also been mentioned and the fact that in some species of Gezssor- rhiza and Hesperantha the stamens and style are all directed to one side, which is obviously a preliminary step towards the formation of a medianly zygomorphic flower. Pax included Babiana in the Gladioleae but in this genus there are species with actinomorphic and others with zygomorphic flowers. In the genus Lapeyrousia there are also species which fall into both categories. Pax’s division also resulted in the separation of Streptanthera from Sparaxis and Synnotia. On account of its actinomorphic flowers Streptanthera was placed with the Jxzeae and the other two genera with the Gladioleae. ‘These three genera are morphologically closely allied, especially Streptanthera and Sparaxis which have almost identical corms, corm tunics, leaves and bracts, and natural hybrids between these two genera have been recorded in the National Botanic Gardens, Kirstenbosch. It is evident that an attempt to divide the genera in the Jxzeae into subtribes based on the symmetry of the flowers results in a false division not only of closely allied genera but also of species within cne or two of the genera. The same is true of the more advanced group of genera. It has been pointed out that Gladiolus is extremely closely allied to Homoglossum and that Homoglossum passes into Petamenes. In the section marked 3c, Anapalina passes upwards into the most advanced region through the species at present known as Chasmanthe caffra and C’. intermedia. { consider therefore that the whole group of genera in the section marked 3¢ can only be regarded as a subtribe of the /xzeae, in which various progressive stages occur as they do in the other subtribes. In three of them, 3c, ge and 3f, the flowers have all developed along similar lines, resulting, in the most advanced members, in medianly zygomorphic, very irregular flowers in which three or five of the lower perianth lobes are reduced, sometimes consider- ably so, while the uppermost is elongated and expanded. In each of these respective groups or subtribes the other morphological features of the most advanced genera are the same as those of the lower genera in the same group and distinct from those of the other more advanced genera. For instance, Antholyza and Anaclanthe (which are scarcely separable) are very closely con- nected with Babiana but morphologically quite distinct from Anapalina, Homo- glossum and Chasmanthe. It seems preferable, therefore, to place these most advanced genera, as is shown in the diagram, at the top of the subtribes 3c, 3¢ MORPHOLOGY, PHYLOGENY, TAXONOMY OF S.A. IRIDACEAE III and 3 f. In my opinion this results in a more natural phylogenetic grouping of the genera. The name Ixineae is suggested for the subtribe 3¢, though the conception of this subtribe is much broader than that of the tribe /xzeae of Pax and Hutchinson and includes members of their tribe Gladioleae, as well as some of the genera of Hutchinson’s tribe Antholy zeae. In the diagram one or two other slight alterations are proposed which will necessitate some changes to the previous conception of one or two of the genera. For instance, Anomalesia and Kentrostphon are combined into one genus and the subgenus Dichone is separated from Tritona and raised to generic status, also the subgenus Anomatheca of Lapeyrousta. N. E. Brown’s genus Radinosiphon has been included with a query as this requires further investigation. It is doubtful whether it is sufficiently distinct from Gladiolus to warrant generic status. With a few additions the scheme for classification which has been outlined in the preceding pages could be made applicable to the whole family, though this cannot properly be done without making an examination of some of the genera which occur in other countries. SUMMARY 1. In the prophylls of Witsenia maura what appears to be a rudimentary lamina is apparent in the very early stages but it tends to disappear as the prophylls develop. 2. From the development of the prophylls in Witsenia maura it appears that the amplexicaul base of the petiole develops outside and adnate to an ‘axillary’ ligule. In the shrubby genera, Wiisenia, Nivenia and Klattia, an absciss layer is formed at the base of the limb, below the sheathing portion. 3. It is suggested that the delay in the entry of the median vascular strand into the leaf base is associated with the partial adnation of the axillary buds to the axis. This adnation is evident in some of the shrubby genera, especially in Nivenia Stokoe. 4. In the minute rudimentary cauline leaves of some species of Anapalina and Exohebea there is further evidence of a laminar rudiment. In these plants the bracts and rudimentary cauline leaves are adnate to the prophylls. 5. The basal leaves of Exohebea flexuosa, Anapalina triticea and A. Burchellu are divided into a very slender solid petiole and an expanded pseudo-lamina. 6. ‘The young dorsiventral leaves of Hexaglottis flexuosa (also other species of Hexaglottis and several species of Moraea and Homeria) have a slightly swollen area at the apex which is possibly a laminar rudiment. ‘The apex of the leaf dies off at an early stage. 7. Evidence is brought forward to substantiate Ingeborg Haeckel’s theory that the more primitive inflorescence in the Iridaceae is a much-branched one, the spicate inflorescence of the Ixioideae being derived from this by various stages of reduction. I12 ANNALS OF THE SOUTH AFRICAN MUSEUM 8. In some species of Babiana the prophylls and bracteoles are entirely free and separate, providing some evidence that these are two distinct foliar members which in most of the genera are more or less united owing to reduction of the internodes. g. Prophylls in various stages of fusion were found subtending cormlets in Watsonia bulbifera. Where the fusion is only partial, two buds are present, but a further overlapping and fusion results in the suppression of one of the buds. 10. Some vestigial traces of bracts in one or two species of Watsonia and in Pillansia Templemanni furnish evidence that a reduction has taken place. The inflorescence of Pillansia represents an intermediate stage between a compound inflorescence like that of some species of Nzvenza and the spicate inflorescence of Watsonia. 11. In some species of Lapeyrousia the branches and flowers, with their sub- tending leaves or bracts, are more or less adnate to the axis. The inflorescence is richly ramified in some species of Lapeyrousta and spicate in others, and there are indications of a transition from the former type to the latter. 12. The fused bracteoles of Tulbaghia violacea are described briefly as they provide further proof of fusion of adjacent overlapping bracts. 13. The reduction of internodes and the occurrence of short and long inter- nodes in the Iridaceae are discussed. 14. The rhipidia and inflorescences of one or two species of Aristea, Bobartia and Moraea are described. The apparently single rhipidium of the two latter genera is probably derived by further reductions from ‘binate rhipidia’ such as are common in Arisiea. 15. The rhizome is the more primitive type of subterranean stem in the Iridaceae and the corm a later development. The subterranean stem of Pillansia represents a transition stage between a rhizome and corm. 16. The unusual ‘corms’ of Ferraria, which have no persistent scale leaves, are described. These probably represent a transition stage from a rhizome to the type of corm which occurs in some members of the Iridozdeae. 17. In the Jxzordeae the corm is formed through the accumulation of food reserves in two or more abbreviated internodes in the underground basal part of the aerial axis; in the /ridozdeae the food reserve is stored in a single internode of an axillary bud. 18. The flat-based scale leaves of the corms of Lapeyrousia, Romulea and Hesperantha, the spinous roots of Moraea ramosissima and the possible ligular origin of the distinct inner layer of the scale leaf of Moraca plumaria are discussed. 19. Some anomalies observed in the flowers in some of the genera are mentioned. In three genera in the Jridozdeae previously described as having the perianth segments free there are species (one in each genus) with a perianth tube. MORPHOLOGY, PHYLOGENY, TAXONOMY OF S.A. IRIDACEAE 113 20. The origin of the perianth tube and the probable derivation of either the whole or part of it from the pedicel is discussed. A few controversial taxonomic groups are mentioned. 21. A phylogenetic diagram is drawn to illustrate the natural relationships of the South African genera of the Iridaceae and an amended system of classifica- tion is proposed. BIBLIOGRAPHY Adamson, R. S., ‘Anatomy of Some Shrubby Iridaceae’. Trans. Roy. Soc. S. Afr., XIII, 175 1926). sae ee Leaf Structure of the Iridaceae considered in relation to the Phyllode Theory’. Ann. Bot., XXXV, 301 (1921). , Monocotyledons. A Morphological Study. (1925.) Baker, J. G., Flora Capensis, VI (Iridaceae) (1896). Bentham, G., and Hooker, D. H., Gen. Plant., III, 681. (Classification of Iridaceae.) (1883.) Bolus, L., ‘Pillansia Templemanni’, Ann. Bolus Herb., 1, 20 (1915). , Moraea apetala, S.A. Gard. and Country Life, XIX, 385 (1929). ——., ‘Notes on “‘Antholyza’’ (N. E. Brown’s subdivision of the genus)’, S.A. Gard. and Country Life, XXIII, 46 (1933). Bolus, L., and Mathews, W. F., ‘Watsonia bulbifera’, Ann. Bolus Herb., III, 140 (1923). , Watsonia vivipara’, Ann. Bolus Herb., 1, 117 (1925-8). Brown, N. E., “Contributions to a Knowledge of the Transvaal Iridaceae’. Trans. Roy. Soc. NEVA ero, 261. (1932). , ‘The genus Dierama’. Journ. Royal Hort. Soc., LIV, 1 (1929). Diels, L. ‘Classification of Iridaceae in Engler & Prantl Nat. Pflanzf.’, ed. 2, 15a, 469 (1930). Dinter, K., ‘Lapeyrousia Vaupeliana’, Fedde Rep., XVIII, 436 (1922). Foster, R. C., Revision of Geissorrhiza. Contr. from the Gray Herb. Harvard Univ. CXXXV, 1-78 (1941). ——, Notes on Nomenclature of Iridaceae. Contr. from Gray Herb. Harvard Univ. CXIV, 38-41 (1936). ae BE teres New Species and New Combinations. Contr. from Gray Herb. Harvard Univ. CXXVITI, 47-8 (1939). Haeckel, I., Flora, 125, 1-82 (1931). (Inflorescences of some of the Iridaceae.) Hutchinson, J., Families of Flowering Plants, 11 (Monocotyledons) (1934). Lewis, G. J., “The genus Engysiphon and a Revision of the species placed under Acidanthera in the Fl. Cap.’ Journ. S.A. Bot., VII, 19 (1941). ——., ‘Notes on Bobartia lilacina and B. paniculata’. Journ. S.A. Bot., X1, 108 (1945). ——., Iridaceae in Adamson and Salter’s Flora of the Cape Peninsula (1950). ——, ‘A note on the rediscovery of Thunberg’s “Gladiolus flexuosus” (= Exohebea flexuosa) and a description of the plant’. Ann. S. Afr. Mus., XL, 1 (1952). Marloth, R., Flora of S. Africa, IV (Iridaceae) (1915). Pax, F., Classification of Iridaceae in Engl. & Prantl ‘Nat. Pflanzf.’, II, 5, 137 (1888). Phillips, E. P., “A note on N. E. Brown’s sub-division of the genus Antholyza’. Bothalia, IV, 43 (1941). Rendle, A. B., Classification of Flowering Plants, I (Iridaceae) (1904). Scott, D. H., “Notes on Spinous Roots’. Ann. Bot., XI, 327 (1897). Scott, D. H., and Brebner, G., ‘Secondary Tissue in certain Monocotyledons’. Ann. Bot., VII, 21 (1893). Weimarck, H., “Types of Inflorescences in Aristea and some Allied Genera’. Bot. Notiser, 616 (1939). (Lund.) ——., ‘Monograph of the genus Aristea’. Lund. Univ. Arsskrft. N. F. Avd., 2, Bd. 36, Nr. 1 (1940). ——.,, ‘A Revision of the genus Nivenia Vent.’. Svensk. Botanisk. Tidskrift, Bd. 34, H. 4, 355 (1940). i gently ma ‘anton be ae Rise 4 AM ee Pin! ws < At Vie ir a Be he i mee. oe Lrg Path hee Y" of ey © Avis e A i Ses ua fh ME as ean ; ¥ fan i 1 } f 4 i t Ath 4 ay he : OE ae ‘ 12) pate ake ROP I Ae a Ae lle ii a -bceogpale «oa stave es Tew. vty AL raga fy eee auuntiidl, “yee mods Tg yh a ont at me Pe Cob esle on tiled Shae tee e mite ay Lean y hawk eae fie Oe Me eee il a f 7 Ne vy 4 ce OR a) SOP 4 + ss age af ie eae : ve ; wi denna ab = a =< — Fast 7. uit Dai Losses ) aOR seiqueniein) Ade tea Nye big ih pe WAS ss pte | ie A Nea ete Gh en ; sbhiae Pied cy is aig j t aa u ‘ is ht Pes jee Pala i Wh again damd: sPhamter. aia iadtie | arg ee weal pide ‘evcliien tsi SOA AR is aE AE Ween he ie fo ee Ms hw Hei Le yi Ho Pe iho Orie hi pant: ; | ' t 5 | a * : * ; \ \ | : ‘ ) : t nee = ; | , Z 3 | ass | ¢ . : ‘ } , ‘ j ! . : | x ‘ i ; | ' i ) % i | i : 2 : é : | | . ——d Fl : : ¥ > & Gru t | : | | ) os wf j f | | : : 3 P | : | : aaa i uae i + 7 x - 7 * ¢ - me 4 } . P ‘ : ; | ; : : Sik ie 7 Shines ered ED PS PE The ANNALS OF THE SOUTH AFRICAN MUSEUM are issued in parts at irregular intervals as material becomes available. As far as possible each volume is devoted exclusively to a particular subject (Zoology, Botany, ey Two or more volumes may be in course of publication concurrently. Most of the Geological and Palaeontological papers are issued in conjunction — the — Geological Survey of the Union of South Africa. Some volumes and parts are out of print, and others are only sold as parts of aset, or volume, et respectively. The prices of parts published prior to 1940 have been increased. Out of print: Vols. I, II, V (Parts 1, 2, 9), VII, VIII, IX (Part a XII Shei: eS XXII, XXIV (Part 2), XX XI (Parts 1, 2, 3). Vol. ! Pa ier oe III. 1903-1905 Zoology I 1g 0 IV. 1903-1908 Palaeontology _ sega apo V. 1906-1910 Geology, Palaeontology, Zoology, Anthropology RE Ke Bee VI. 1908-1910 Zoology ae, 3 3 6 IX. 1911-1918 Botany .. ey oh ee. ‘Part 1) 2.160 X. 1911-1914 Zoology ? a ue we) ey ee Sea oe ope XI. 1911-1918 Zoology .. oe unt ai XII. 1913-1924 Palaeontology and Coo: = .. (excl. Part 7) 913 0 XIII. 1913-1923 Archaeology and Zoology aes Bey: Betsey ees XIV. 1915-1924 Zoology : 3-6. 6 XV. 1914-1916 Zoology 4 5 0. XVI. 1917-1933 Botany BS OF XVII. 1917-1920 Zoology ‘ 3 10° .o. XVIIT- 1921 Zoology 2 Nee Os XIX. 1924-1925 Zoology 3 7.0 XX. 1924-1926 Zoology .. Seah eects be: 2 on 21220 XXII. 1925-1927 Zoology .. ee Sy ie as iy Sea es OED XXIII. 1925-1926 Zoology Bees es ti ees XXIV. 1929-1938 Aniiecpologs oad Eeincogy 2-0. (excl: Part 2)o2 eae @ XXV. 1927-1928 Zoology os Ee XXVI. 1928 Zoology .. a ie 110 0 XXVIII. ——_—srg2g__—Ss Anthropology I 10 0 XXVIII. 1929-1932 Palaeontology 212 0 XXIX. 1929-1931 Zoology . Paice» ae 9 XXX. 1931-1935 Zoology 319 6 INDEX of papers, authors, and subjects, ubisheds in Vals L-XXX. Ree ta: Me XXXI. 1934-1950 Palaeontology Le ie ATE (Rake 4 only) o1m4 O- XXXII. 1935-1940 Zoology een So XXXITI. 1939 Zoology 2 XXXIV. 1938 Zoology 2 8 o XXXV. Reserved for conclusion of one pbanin in Vol. XXXIV. XXXVI. 1942-1948 Zoology 2 Ei XXXVIT. 1947-1952 Archaeology — 1 16 XXXVI. 1950 Zoology 315 KXXXIX. 1952 Zoology a ie x3 rs oo Wea ee XL. 1952- Botany... Bis A Part 71s. Part! 2x0 89 “XLI. = 1.952- Zoology ... i 5 Part 1, -75. 6d. Part "a> warre XLII. 1953- Palaeontology a ey ee oe Part’? G38 a eee Copies may be obtained from— : The LIBRARIAN, SoutH Arrican Museum, Carz Town, except the Geological and Palaeontological parts, which are obtainable from the GOVERNMENT PRINTER, PRETORIA. HMADRMCOO. FE a ee CHEE ag wea EAS as he eat a toe ee Ce Oe” Eh RIN I She ee ie at OE Ed re ee ee ee ae | Re ANNALS OF THE SOUTH AFRICAN MUSEUM VOLUME XL - PART III, containing :— 4, TIridaceae— New Species and Miscellaneous Notes. By G. J. Lewis, (With five text-figures.) = AT HSO Ni4 4y\ ( NOV 15 1954 64) the WFIE BRARY, oe a ISSUED SEPTEMBER 1954 PRICE 53. | PRINTED FOR THE TRUSTEES OF THE SOUTH AFRICAN MUSEUM BY THE RUSTICA PRESS (PTY.) LIMITED, COURT ROAD, WYNBERG, CAPE 4. Iridaceae—New Species and Miscellaneous Notes. By G. J. Lewts. (With 5 text-figures) MorAEA MACRONYX Lewis sp. nov. (Iridaceae—Moraeeae) M. macrochlamys affinis sed floribus majoribus et unguibus perianthii segmen- torum multo longioribus differt. Cormus globosus, 1~1-5 cm. diam., fere bulbillos emittens, tunicis extimis e fibris paucibus pallidis crassis reticulatis compositis. Caulis brevissimus, omnino foliis opertus; planta (cormo et flore incluso) 15-25 cm. alta. Vaginae basales 2, pallidae membranaceae, obtusae, 1-2 et 3:5—6°5 cm. longae. Folia 2 caulina, glabra, falcata, 12-25 cm. longa, tertia parte inferiore late vaginante, parte superiore conduplicata, marginibus planis vel partim undulatis. Cyma unicum, 3-4 - florum, floribus sat grandibus, flavis immaculatis. Spatha exterior foliis similis sed breviora dimidio parte inferiore vaginante; spatha interior recta, saepe ventricosa, obtusa vel subacuta, 8-11 cm. longa, plerumque 1-1-5 cm. diam., sed in plantis minoribus interdum 6 mm. diam. Pertanthii segmenta longiunguiculata, exteriora 4-5 cm. longa ungue 2-5-3 cm. longo, lamina ovata, obtusa vel parum emarginata, 1-4-1°8 cm. lata; interiora 3-6—-3-9 cm. longa, lamina oblongo-ovata obtusa, 7-8 mm. lata. Filamenta 1-3-1°5 cm. longa, per 6-8 mm. coalita; antherae 0-8-1 cm. longae. Rami styli (cristis inclusis) 375-4 cm. longi, cristis acutis circa 2 cm. longis, 4-6 mm. latis. Ovarium 1-5 cm. longum; pedicelli pallidi tenuissimi, 8-11 cm. longi. Hab. Cape Province. Ceres Division: Cold Bokkeveld, near Elandsfontein, about six miles north from top of Gydo Pass, in sandy ground, Lewis, South African Museum No. 61920 (Type, in S.A.M. Herb.), 22 Sept. 1952; Cold Bokkeveld, Leeurivier, Lewis, S.A.M. No. 57966, Sept. 1945; Cold Bokkeveld, Schlechter 8890 (Kew); Cold Bokkeveld, top of Gydo Pass, Lezpoldt 4082 (in Bolus Herb.), Sept.; Cold Bokkeveld, near Elands Kloof, Le:ghton 1287 (Bolus Herb.), Sept. Corm globose, 1-1-5 cm. diam., with the outer tunics composed of a few pale, coarse, reticulate fibres; cormlets usually present. Stem simple, very short, enclosed within the leaf sheaths; height (including corm and flower) 15-25 cm. Basal sheaths 2, membranous, colourless, obtuse, 1-2 and 3°5—-6-5 cm. long. Leaves 2, cauline, glabrous, falcate, 12-25 cm. long, the lower third wide and sheathing, conduplicate above, the margins smooth or partly undulate; lowest leaf sometimes with an axillary cormlet. Cyme solitary, 3-4 - flowered, the flowers fairly large, yellow, without markings. Outer spathe like the leaves but slightly shorter, with the lower half ventricose; inner spathe erect, often ventri- cose, obtuse or subacute, 8-11 cm. long, usually 1-1-5 cm. diam., in smaller plants sometimes only 6 mm. diam. Perianth segments with long claws, the outer Os VOL. XL. PART 3 116 ANNALS OF THE SOUTH AFRICAN MUSEUM Ficure 1. Moraea macronyx Lewis. 1. Plant in flower. 2. Outer perianth segment. 3. Inner perianth segment. 4. Flower with perianth segments removed, showing pedicel, ovary, stamens and style branches. 5. Ovary and part of pedicel of faded flower, showing slight contractions in the pedicel. All drawings natural size. Del. G. J. Lewis. IRIDACEAE—-NEW SPECIES AND MISCELLANEOUS NOTES U7) ones 4:5 cm. long with the claw 2-5-3 cm. long and an ovate obtuse, or slightly emarginate blade 1-4-1-8 cm. wide; inner segments 3:6—3-9 cm. long, the claw widening gradually into an oblong-ovate blade 7-8 mm. wide. Filaments I°-3-1°5 cm. long, united for half their length; anthers 0-8-1 cm. long. Style branches (with crests) 3:5—4 cm. long, the crests acute, about 2 cm. long, 4-6 mm. wide. Ovary 1-5 cm. long; pedicels colourless, very slender, 8-11 cm. long. This species differs from M. macrochlamys Baker, to which it is most closely allied, in having larger flowers and very much longer claws to the perianth segments. ‘The claw of the inner segments widens gradually into the blade, but in the outer ones the long, narrow claw passes abruptly into the broad, ovate blade. The mature flower, with its ovary, is well exserted from the spathes but as the flower fades the long, slender pedicel begins to contract and the ovary is withdrawn into the spathes, the capsule developing and ripening within the ventricose sheath of the inner spathe. When the flower is mature the pedicel has smooth margins but slight undulations and wrinkles have been observed in the pedicels of flowers which have faded. This phenomenon of a contractile pedicel is not common to the whole genus but it is almost certain that it occurs also in the allied species M. macrochlamys and M. ciliata, as in herbarium speci- mens examined the only capsules found were enclosed within the spathes, near the base. FERRARIA FOLIOSA Lewis sp. nov. (Iridaceae—Moraeeae) F’, undulatae L. affinis sed sine foliis basalibus et foliis caulinis valde patentibus differt. Cormus globosus, 2:5-3 cm. diam. Caulis circa 40 cm. altus, foliis caulinis imbricatis opertus, spiraliter tortus, prope apicem aliquot ramis brevibus instructus. Vaginae basales 2, submembranaceae, 6 et 11 cm. longae. Folia 14-20, omnia caulina, glauca, 17-6 cm. longa, obtusa et mucronulata vel subacuta, tertia parte inferna vaginante ad 5 cm. lata marginibus pallidis membranaceis, parte superiore valde patente, ad 6 mm. lata. Cymata 2-flora. Spatha exterior 2-8-3 cm. longa, 2 cm. lata apice acuta leviter incurva; spatha intertor 4-8-5 cm. longa, ad 3 cm. lata. Perianthii segmenta purpureo-rubida marginibus fulvis, foetida, exteriora 3:3-3:8 cm. longa, supra unguem nectari- ferum 1-1-4 cm. lata, interiora 2-8—3-3 cm. longa, 0-8-1 cm. lata. Filamenta 8-9 mm. longa, per 7 mm. coalita; antherae suborbiculares, circa 2 mm. longae, lobis parallelis. Rami stylt 4-5 mm. longi cristis inclusis. Ovarium 1-6-1-:8 cm. longum, 3-3-5 mm. diam.; pedicelli 2 cm. longi. Hab. Cape Province. Piketberg Division: Elands Bay, near Verloren Vlei, in sandy ground, Lewzs, South African Museum No. 60869 ( Type, in S.A.M. Herb.), Sept. 1951; north-east side of Verloren Vlei at Matjesgoed Drift, Pillans 8127 (in Bolus Herb.), Oct. Corm globose, 2-5~3 cm. diam., attached to 2 or 3 older corms and growing more or less horizontally. Stem about 40 cm. high, spirally twisted, entirely enclosed within the wide imbricating sheaths of the cauline leaves, with several 118 ANNALS OF THE SOUTH AFRICAN MUSEUM Ficure 2. Ferraria foliosa Lewis. 1. Plant—half natural size. 2. Flower—natural size. 3. Outer perianth segment—natural size. 4. Flower with perianth segments removed, showing pedicel, ovary, stamens and style crests—natural size. 5, Stamens and style crests— X 13. 6. Style arms and crests, from above, X 2. Del. G. J. Lewis. IRIDACEAE—NEW SPECIES AND MISCELLANEOUS NOTES i ee) short flowering branches at the top. Basal sheaths 2, more or less membranous, 6 and 11 cm. long. Leaves 14-20, all cauline, glaucous, 17-6 cm. long, obtuse and mucronulate or subacute, the lower third forming an inflated sheath, up to 5 cm. wide, with white membranous margins, the upper flat part spreading more or less horizontally, 6 mm. wide. Cymes several, 2-flowered. Outer spathe 2-8-3 cm. long, 2 cm. wide, the apex acute, slightly incurved; inner spathe 4°8-5 cm. long, up to 3 cm. wide, more inflated than the outer. Flowers dark maroon with yellow-brown margins, unpleasantly scented, with the nectari- ferous claws of the segments curved up to form a cup about 1°5 cm. diam.; outer segments 3°3-3°8 cm. long, I-1:4 cm. wide just above the claw; znner segments 2°8—3-3 cm. long, 0-8-1 cm. wide. Filaments 8-g mm. long, united for 7mm.; anthers suborbicular, about 2 mm. long, with parallel cells. Style arms 4-5 mm. long, including the fimbriate crests. Ovary 1-6-1:8 cm. long, 3-3-5 mm. diam.; fedicels 2 cm. long. F. foliosa is allied to F. undulata, a species fairly common along the south- western coast from the Malmesbury Division to Danger Point in the Caledon Division, but differs in having no basal leaves and the cauline leaves spreading out more or less horizontally above their ventricose bases. Owing to a slight twist in the stem, the distichous leaves are spirally arranged, an arrangement which is not easy to represent in a drawing and which is not apparent in dried herbarium specimens. Corms were collected during July 1950, when the plants were not fully developed, and grown at the South African Museum Herbarium, where they flowered from the end of August to September 1951, and again the following year. These plants are probably smaller than they would be if grown in their natural habitat; a plant with fruits in the Bolus Herbarium, collected near the type locality during October, is considerably taller and more robust (more than 1 metre high) with correspondingly longer leaves and branches. Its natural habitat is in coastal sand, and the same species was observed at two or three localities along the coast from Elands Bay to Lamberts Bay in the Clanwilliam Division. FERRARIA BREVIFOLIA Lewis sp. nov. (Iridaceae—Moraeeae) Cormus depresso-globosus, 1°5-2 cm. diam. Caulis brevissimus omnino foltis opertus; planta (cormo et inflorescentia incluso) 12-19 cm. alta. Vaginae basales 2 vel 3, pallidae membranaceae, acutae, 2°5-8:5 cm. longae. Folia plura omnia caulina brevissima, patentia, 2-6 cm. longa laminis 1-1°5 cm. longis ad 1 cm. latis oblique ovatis marginibus crassis et apicem versus acutis leviter incurvis. Cymata pauca, 2-flora. Spatha exterior foliis similis; spatha interior 5-5°5 cm. longa, obtusa vel acuta marginibus pallidis membranaceis sed apicem versus crassis. Perianthi segmenta acuminata, flava, marginibus aurantiacis; exteriora basi viride maculata, 3-3-3:7 cm. longa ungue circa 1-3 cm. longo, lamina supra unguem ad 1 cm. lata; interiora immaculata, circa 3-3 cm. longa, 4—5 mm. lata. Filamenta 1 cm. longa, omnino coalita; antherae aurantiacae, I20 ANNALS OF THE SOUTH AFRICAN MUSEUM LASTS car a ag LP cal 33 22 FicurE 3. Ferraria brevifolia Lewis. 1. Plant. 2. Flower, from above. 3. Outer perianth segment. 4. Inner perianth segment. 5. Stamens and style crests, X 2. 6. Anther cells, x 2. 7. Gynaeceum and pedicel. 8. Style arms and crests, x2. All drawings natural size except where stated. Del. G. J. Lewis. IRIDACEAE—NEW SPECIES AND MISCELLANEOUS NOTES I21 divaricatae. Ovarium 2 cm. longum, apicem versus in rostro sterili 8-g mm. longo productum; fedicelli 2-5-3 cm. longi. Hab. Namaqualand: 2 miles south of Nieuwerust,* Lewis, South Afr. Mus. No. 57943 (ype, in S.A.M. Herb.), 7 Sept. 1945; between Bitterfontein and Garies, Lewis, N.B.G. No. 2065/32 (in Bolus Herb.), Sept. 1932; Bitterfontein, Schlechter 11031 (Kew and Bolus Herb.). Corm depressed-globose, 1-5-2 cm. diam. Stem very short, hidden by the leaves; height (including corm and inflorescence) 12-19 cm. Basal sheaths 2 or 3, colourless, membranous, acute, 2°5—8-5 cm. long. Leaves several, all short and cauline, the lower ones spreading horizontally, the upper suberect, 2-6 cm. long, the blades 1-1-5 cm. long and up to 1 cm. wide, obliquely ovate with slightly incurved acute tips and thickened margins. Cymes few, 2-flowered. Outer spathe like the leaves; inner spathe 5—5:5 cm. long, obtuse or acute, with the margins membranous and colourless but thickened towards the apex. Perianth segments acuminate, yellow with orange margins; outer segments with a few small green spots on the claw and base of the blade, 3-3—3-7 cm. long, the claw about 1°3 cm. long, the blade above the claw up to 1 cm. wide; inner segments not spotted, about 3-3 cm. long, 4-5 mm. wide. Filamenis 1 cm. long, com- pletely united; anthers bright orange, divaricate. Ovary enclosed within the spathes, 2 cm. long, with a sterile beak-like prolongation 8-9 mm. long above; pedicels 2-5-3 cm. long. A very distinct species readily recognized by its short spreading leaves with very short obliquely ovate blades which have conspicuous cartilaginous margins. The attractive yellow and orange flowers have a fairly pleasant scent. IxIA LEIPOLDTH Lewis sp. nov. (Iridaceae—Ixieae) Cormus globosus, 1-1-4 cm. diam., tunicis e fibris numerosis sat tenuibus reticulatis compositis. Caulis 11-25 cm. altus, simplex vel ramo brevissimo suberecto, interdum 2 ramis. Vaginae basales 1 vel 2, 1°5-4°5 cm. longae. Folia 3, mediocriter robusta, 7-20 cm. longa, 3-5 mm. lata, infimum sub- falcatum, secundum suberectum, supremum erectum fere vel omnino vaginans. Spica 2-4 flora. Bractea membranacea, fulva vel fulvo-lilacina, 0-7-1 cm. longa, conspicue 3-nervata, acuta vel obtusa vel truncata, plerumque 3-cuspidata sed interdum cuspide unica induta; bracteolae bracteae similes sed 2-nervatae et breviter bicuspidatae. Perianthi tubus flavus, 1-1-2 cm. longus basi filiformis superne infundibuliformis; segmenta aequalia, alba, tertia parte inferna rubra, orbicularia vel suborbicularia, 1-5-1-7 cm. longa, ad 1-4 cm. lata, basi ungue brevissimo et lato. Stamina brevia, antherarum basibus in tubo perianthii inclusis; filamenta 3-4 mm. longa; antherae 3-4 mm. longae. Stylus apicem fila- mentorum attingens ramis circa 2 mm. longis. Ovarium 3 mm. longum, 2 mm. diam. * Now spelt Nuwerus. 122 ANNALS OF THE SOUTH AFRICAN MUSEUM Hab. Cape Province. Prince Albert Division: Eikerkraal, between Prince Albert and Klaarstroom, Lezpoldt, Bolus Herb. No. 21808 (Type, in Bolus Herb.), 27 Sept. 1935. Corm globose, 1-1-4 cm. diam., with numerous tunics of fairly fine reticulate fibres. Stem simple or with 1 or sometimes 2 very short suberect branches, 11-25 cm. high. Basal sheaths 1 or 2, 1°5-4°5 cm. long. Leaves 3, fairly firm, 7-20 cm. long, 3-5 mm. wide, the lowest subfalcate, the second suberect, the third erect and entirely or almost entirely sheathing. Spike 2-4 - flowered. Bract membranous, brownish or brownish purple, 0-7-1 cm. long, conspicuously 3-veined, acute, obtuse or truncate, usually shortly 3-cuspidate but sometimes with a single cusp; bracteoles like the bracts but 2-veined and shortly bicuspidate. Perianth tube yellow, 1-1-2 cm. long, filiform at the base, the upper half funnel- shaped; lobes equal, white with the lower third claret-coloured, orbicular or suborbicular, 1-5—-1-7 cm. long, up to 1-4 cm. wide, with a very short broad claw. Stamens short, with the basal quarter or third of the anthers not exserted from the perianth tube; filaments 3-4 mm. long; anthers 3-4 mm. long. Style reaching to the top of the filaments with branches about 2 mm. long. Ovary 3 mm. long, 2 mm. diam. Although a bright, contrasting patch of colour at the base of the perianth lobes is a fairly common feature in the section Euzxia, it has not hitherto been recorded in the section Morphixia, to which this species, with its distinctly funnel-shaped perianth tube, belongs. The lobes are white, with a conspicuous claret-coloured patch at the base, and short coloured veins extending upwards into the base of the white part. Another character which distinguishes this species from all others in the section is the width of the perianth lobes, the broad, almost orbicular, lobes presenting a striking contrast to the more or less oblong lobes of the other species in the group. This attractive and apparently rare species has been collected only once and is named in honour of the collector, the late Dr. C. L. Leipoldt, who will always be remembered in this country as a writer and poet and a keen observer and lover of nature. Thanks are due to the Curator of the Bolus Herbarium, Dr. L. Bolus, for the loan of the type material and for the notes made by her on the colour and shape of the living flowers. GLADIOLUS SYMMETRANTHUS Lewis sp. nov. (Iridaceae—Ixiae) G. erectifloro et G. trichonemifolio affinis sed floribus omnino aequilateralibus flavis, in faucibus purpureis, differt. Cormus globosus vel subglobosus, 1-1°5 cm. diam. Caulis gracilis, 16-26 cm. altus. Vaginae basales 2, obtusae, 2-3 cm. et 3-6 cm. longae. Folia 2 vel 3, infimum dimidiam partem caulis vaginans, lamina subtereti valde striata, quam caule duplo longiore; folia cetera multo breviora laminis brevibus subulatis spicam non superantibus. Spica 1-3 - flora, floribus aequilateralibus flavis, in faucibus purpureis. Bractea mediocriter rigida, crebre nervosa, acuta, 2°5-3 cm. longa, 7-8 mm. lata; bracteolae parum angustiores obscure 2-dentatae. Perzanthiu IRIDACEAE—NEW SPECIES AND MISCELLANEOUS NOTES 123 tubus rectus, infundibuliformis, circa 1-5 cm. longus; segmenta subaequalia oblonga obtusa vel interdum leviter emarginata, 2-5-3 cm. longa, 1-1-1-3 cm. lata. Stamina recta, aequilateralia, filamentis 8-9 mm. longis antherisque 8-9 mm. longis. Stylus rectus, prope apicem antherarum attingens ramis 3-4 mm. longis; ovarium 4 mm. longum. Hab. Cape Province. Stellenbosch Division: near Koelenhof, between Stellenbosch and Mulders Vlei, in marshy ground on flats, Lewis, South Afr. Mus. No. 60700 (Type, in S.A.M. Herb.), 14 Sept. 1950; Mulders Vlei, in somewhat marshy ground, Duthie 940 (Bolus Herb.), Sept.; near Koelenhof, 4 2 1 Ficure 4. Gladiolus symmetranthus Lewis. 1. Corm and base of stem. 2. Bract. 3. Bracteoles. 4. Flower, laid open. 5, Flower, side view, with upper part of the perianth lobes removed to show the stamens and style, x 2. All drawings natural size except where stated. Del. G. J. Lewis. Leighton, s.n. (Bolus Herb.), Sept.; Elsenburg, Grant 2512 (Bolus Herb.), Sept.; between Paarlberg and Paardeberg, Drége 8457 (Geneva), Aug.—Sept. Corm globose or subglobose, 1-1°5 cm. diam.; tunics coarse, usually con- centric above with the lower half toothed. Stem slender, simple, 16-26 cm. high. Basal sheaths 2, obtuse, 2-3 and 3-6 cm. long. Leaves 2 or 3, the lowest sheathing the lower half of the stem, with a strongly ribbed subterete blade, more than twice as long as the stem, the upper 1 or 2 leaves very much shorter with subulate acuminate blades not reaching above the inflorescence. Spzke 1-3 - flowered; flowers actinomorphic, bright yellow with the throat dark purple-maroon. Bract firm, closely veined, acute, 2-5-3 cm. long, 7-8 mm. wide; bracteoles similar but slightly narrower, obscurely bifid. Perzanth tube straight, funnel-shaped, about 1-5 cm. long; lobes subequal, oblong, obtuse or sometimes slightly emarginate, 2-5-3 cm. long, 1-1-1:3 cm. wide. Stamens erect, symmetrically arranged, reaching to the middle of the lobes; filaments 8-9 mm. 124 ANNALS OF THE SOUTH AFRICAN MUSEUM long; anthers 8-g mm. long. Style erect, reaching nearly to the top of the anthers; style branches alternating with the anthers, 3-4 mm. long; ovary 4 mm. long. This species is closely allied to G. erectiflorus Baker and G. trichonemifolius Ker, and G. linearis (L. f.) N. E. Br. It differs from the two former in having actino- morphic flowers, as well as in the colour of the flowers, those of G. erectiflorus and G. trichonemifolius being yellow without a purple throat. In these two species the perianth tube is usually somewhat longer than in G. symmetranthus and the flowers are slightly zygomorphic with the three upper perianth lobes a little broader than the lower ones. The third species previously described in this small and very distinct group is G. linearis (L. f.) N. E. Br. (G. baflorus Klatt), and in this species, as in G. symmetranthus, the flowers are actinomorphic, though they are considerably smaller and of a different colour, either pale or bright lilac. In the Flora Capensis (vi, p. 146), Baker remarked that this species was on the edge of the genus in the direction of Geissorrhiza. There is undoubtedly a close affinity between these four species and all of them have the same habitat, in marshy ground on flats, either in or on the margins of shallow depressions which are filled with water during the late winter and spring months. GLADIOLUS JONQUILODORUS Eckl. ex Lewis sp. nov. (Iridaceae—Ixieae) G. Pillansi proxime accedit sed foliis radicalibus 2, bracteis leviter brevioribus, perianthii segmento supremo obtuso, et perianthii colore, distinguitur. Cormus globosus, 1-5-2 cm. diam., tunicis e fibris pallide brunneis papyraceo- membranaceis compositis. Caulis gracilis, 28-65 cm. altus. Vaginae basales 2, obtusae, 6°5-13 cm. longae. Folia radicalia plerumque 2, glabra, teretia, 4-canaliculata, ad 60 cm. longa, 1-5-2 mm. diam., post anthesin evoluta; folia caulina 3, omnino vaginantia, acuta, supremum 2-5-8 cm. longum, 2 inferiora longiora. Spica secunda, 5-9 - flora, floribus odoratis pallide flavis vel gilvis, segmentis inferioribus citrinis maculatis. Bractea acuta, 1-2 cm. longa, 6-7 mm. lata; bracteolae similes sed leviter breviores. Perianthi tubus curvatus superne infundibuliformis, 0-8-1 cm. longus; segmenta 3 superiora obovata, supremum obtusum, 2-2°5 cm. longum, 1:2—1-4 cm. latum, cetera acuta vel subacuta; 3 inferiora 1-4-1:°8 cm. longa, 6-8 mm. lata, acuta, recurva, unguiculata, unguibus per 5 mm. connatis. Stamina arcuata filamentis circa 1°3 cm. longis antherisque 6-8 mm. longis. Stylus apicem antherarum attingens, ramis 2 mm. longis, stigmata obcordata. Hab. Cape Province. Cape Division: Cape Flats, in damp places, 45 oie aes eee XII. 1913-1924 Palaeontology and Gedtony .. (excl. Part 7) XIII. 1913-1923. Archaeology and Zoology : ‘aera ine XIV. 1915-1924 Zoology : XV. 1914-1916 Zoology XVI. 1917-1933 Botany XVII. 1917-1920 Zoology XVIII. 1921 Zoology XIX. 1924-1925 Zoology XX. 1924-1926 Zoology XXI. 1925-1927 Zoology XXIII. 1925-1926 Zoology an sie ar XXIV. 1929-1938 Antheapoliey aye Ethnology . (excl. Part.2) — XXV. 1927-1928 Zoology 5 : XXVI. 1928 Zoology ae a ae bs XXVIT. 1929. Anthropology .. ot ae aie Ge XXVIII. 1929-1932 Palaeontology — XXIX. 1929-1931 Zoology XXX. 1931-1935 Zoology INDEX of papers, authors, and subjects, publeheds in Vols. I-XXX. XXXI. 1934-1950 Palaeontology on So “(Pat 4 on) XXXII. 1935-1940 Zoology : XXXITI. 1939 ©Zoology es ae oe = XXXIV. 1938 Zoology .. ee vs XXXV. Reserved for conclusion of monograph i in Vol. XXXIV. XXXVI. 1942-1948 Zoology .. cs we .“ os vie awe XXXVIT. 1947-1952 Archaeology . XXXVIII. 1950 Zoology XXXIX. 1952 Zoology : ; : XL. 1952- Botany i “Part Ij 55 Patits 2 XLI. 1952- Zoology Parts I & 2, 75. 6d. each. Part 3, 155. Part 4 XLIT. 1953- Palaeontology Part 1, 12s. 6d. Part 2, Copies may be obtained from— The LIBRARIAN, Sourn Arrican Museum, Care Town, except the Geological and Palacontological parts, which are obtainable from the GOVERNMENT PRINTER, PRETORIA. ~NNWO COW DNR H HD mMOKDHORWHOROWWO DR DO Mo We oOoOON WW = WN COMGPAMIDGDCOFDCORVIGCDOMOCCADTO|AS Ae ae 4 90 07M0 00 as a en) sea O77, GS ‘eles Ae Aes ~ ANNALS SOUTH AFRICAN MUSEUM PART IV, containing :— ‘5. A revision of the Genus Synnotia. By G. J. Lewrs, B.A., Ph.D. ( betel? UN13 1956 S/BRARY 4 : ISSUED APRIL 1956 PRICE 3s. PRINTED FOR THE ita a TRUSTEES OF THE SOUTH AFRICAN MUSEUM _ BY THE RUSTICA PRESS (PTY.) LIMITED, COURT ROAD, WYNBERG, CAPE "ia ‘ Cae ded » ' Vs te +t ener gt ea ote e COMA 5. A revision of the Genus Synnotia. By G. J. Lewis, B.A., Ph.D. INTRODUCTION AND HistroricAL NOTE The history of the first species found at the Cape goes back more than two centuries, to 1739, when Breynius described and figured the species which was named Gladiolus villosus by Burmann f. in 1768. In 1784 this same species was described and figured by Thunberg as Gladiolus bicolor. Sweet established the genus Synnotia in 1826, with an illustration and description of Synnotia variegata, and the following year he transferred Gladiolus bicolor to Synnotia, retaining Thunberg’s specific name; from that date until 1929, when the older specific name was restored by Dr. N. E. Brown, the species was known as Synnotia bicolor. It was evidently fairly well known in Europe by the end of the eighteenth century, and in the first half of the nineteenth century S. villosa, and, after its discovery, S. variegata as well, became popular garden plants in England and Europe, judging from the illustrations published in botanical and horticultural works of that period. In England figures of one or both species appeared in the Botanical Magazine, Sweet’s British Flower Garden and Ornamental Flower Garden and Loudon’s Ladies’ Flower Garden, while in Europe Thunberg and Jacquin published illustrations of S. villosa before the end of the eighteenth century. Another species, illustrated by Jacquin and named by him Gladiolus galeatus, was named Synnotia galeata by Sweet, but after that little seems to have been known about it and it does not appear to have been rediscovered in the field until about a century after the publication of Jacquin’s figure. Although Durand and Schinz retained it as a distinct species in their Conspectus Florae Africae, vol. v (1895), Baker evidently considered it to be a form of S. villosa, for he placed it as a synonym of S. bicolor in the Flora Capensis, and in this country it has remained confused with that species until now. In this revision it is restored to its proper status as a distinct species. In spite of its early history very little herbarium material of the genus seems to have been collected before 1896, when Baker’s work on the Iridaceae was published in vol. vi of the Flora Capensis. Baker himself saw and cited only seven specimens, which he placed under two species and one variety. In 1go1 he described a third species but this has to be excluded from the genus as it is a mixture of Synnotia and Sparaxis (see note on S. stenophylla Bak. at the end of this paper). Except for the publication of a new species by Dr. L. Bolus in 1923, no work has been done on the genus during the past sixty years, although a great deal more material has been collected, and it is not surprising to find that Baker’s work, based on his meagre knowledge of the genus and the few speci- mens available to him, is inadequate and requires revision. ‘This indeed is true of most of the genera of Iridaceae described in the Flora Capensis. 137 WOME. PART 4 JUN 4 195 138 ANNALS OF THE SOUTH AFRICAN MUSEUM The present revision is based on a study of the plants in the field and an examination of eighty-two collectings, the majority of which are housed in herbaria in South Africa. Unfortunately it has not been possible to examine all the material in herbaria in England and Europe, though some of it has been seen. For the loan of specimens and for facilities granted me to examine material in their herbaria sincere thanks are expressed to the Directors and Curators of the following institutions: Bolus Herbarium, University of Cape Town; National Herbarium, Pretoria; National Botanic Gardens, Kirstenbosch; Botanical Museum, Uppsala; Conservatoire de Botanique, Geneva; Botanical Museum, Zurich; Musée d’Histoire Naturelle, Paris. GEOGRAPHICAL NOTE Synnotia is a small genus, endemic in South Africa where it is confined to a few districts in the south-western part of the country, from the northern area of the Cape Peninsula northwards to the Van Rhynsdorp Division, with one record from the Calvinia Division. With the exception of the record from near Calvinia, all the others are from low or fairly low altitudes, on flats, hills or lower mountain slopes in the coastal districts, the Olifants River Valley and the base of the escarpment below Van Rhyn’s Pass. (See map.) S. Metelerkampiae L. Bolus (treated as a variety of S. variegata in this revision) has been recorded from the western slopes of the Cedarberg, up to an altitude of 2,000 ft., but the plants found at these higher altitudes are obviously dwarfed and stunted compared with those from lower altitudes. The most easterly record and the highest is one of S. galeata, from about 12 miles south-west of Calvinia, at an altitude of about 3,000 ft. The genus Sparaxis is very closely related to Synnotia, from which it differs mainly in having more regular flowers, with equal or subequal perianth lobes. This also is a small and endemic genus, with much the same range of distri- bution as Synnotia, though it does not go further north than the Clanwilliam district and extends further to the south-east, to the Caledon and Swellendam Divisions. Like Synnotza it occurs at low altitudes, where in some parts it is extremely common, but is not found on the higher mountain slopes. DESCRIPTION OF THE GENUS Synnotia Sweet, Brit. Flow. Gard., t. 150 (1826); Baker, Fl. Cap., vi, 134 (1896); Anactorion Raf., Fl. Tellur., iv, 34 (1836). Named in honour of W. Synnot who collected a number of plants at the Cape. Corms small to medium sized, globose or ovoid, with pale brownish tunics of short, hard reticulate fibres or long fine hair-like fibres extending up in a A REVISION OF THE GENUS SYNNOTIA 1390 neck. Basal sheaths 2, acute or acuminate, membranous and colourless or the upper sometimes green-tipped. Leaves 5 to 9 in a fan-like distichous rosette, ensiform or narrow-ensiform, acute, subacute or obtuse, glabrous, closely veined, the veins seldom conspicuous; cauline leaves 1 to 3, acute or acuminate. _ Stem erect, glabrous, simple or branched. Inflorescence spicate, the spikes laxly 2 to 7-flowered, or occasionally the flowers solitary. Bracts membranous, ovate, cuspidate and irregularly denticulate, with numerous reddish-brown vein-like streaks above the base; bracteoles similar and almost as long as the bracts, united nearly to the top, bicuspidate with acute to setaceous-acuminate cusps. Flowers distinctly zygomorphic; perianth tube short or long, the lower half or more cylindrical, curved above and becoming gradually or abruptly infundi- buliform; perzanth lobes obtuse or rarely subacute, the uppermost larger than the others, erect in fully open flowers, the 3 lower connivent, forming a more or less projecting reflexed lip. Stamens contiguous, arcuate, inserted at the base of the expanded part of the perianth tube, usually reaching to the middle of the uppermost perianth lobe; filaments filiform; anthers linear or linear- oblong, attached shortly above the base. Ovary small, oblong or ovate-oblong; style reaching to the middle of the anthers, with short entire stigmas either expanding from near the middle upwards and conduplicate or slightly expanded at the apex only. Capsules subglobose or turbinate. Seeds fairly numerous, comparatively large, smooth and shining, rounded or slightly angled by pressure. The type species is S. variegata Sweet, Brit. Fl. Gard., t. 150. Key to the species 1. Flowers small, not more than 2:5 cm. from base of ovary to top of uppermost perianth lobe. ... 1. parviflora 1. Flowers not less than 3 cm. from base of ovary to top of uppermost perianth lobe, usually considerably more. 2. Cylindrical part of perianth tube up to 1 cm. long, rarely slightly longer. 3. Corm tunics of hard reticulate fibres not extending up in a neck; lower side perianth lobes auriculate above the base; uppermost lobe more or less obovate. pia QeUslOsa 3. Corm tunics of numerous fine matted fibres extending up in a neck; lower side perianth lobes not auriculate; uppermost lobe not obovate, up to twice as long as wide. 4. Cylindrical part of perianth tube not more than 7 mm. long; uppermost perianth lobe of mature flowers usually slightly more than twice as long as wide. ... 3. galeata 4. Cylindrical part of perianth tube not less than 1 cm. long, usually more; uppermost perianth lobe up to twice as long as wide. ... 4. Roxburghit 2. Cylindrical part of perianth tube 2 cm. or more long. 5. Corm tunics of fine hair-like fibres extending up in a neck; cylindrical part of perianth tube seldom more than 2 cm. long, curved above but not geniculate. ... 4. Roxburgh 140 ANNALS OF THE SOUTH AFRICAN MUSEUM 5. Corm tunics of hard reticulate fibres not extending up in a neck; cylindrical part of perianth tube 2-5 cm. long or more, geniculate at the top. 6. Upper part of perianth tube broadly funnel-shaped, 1 to 1°5 cm. diam. at the top; length from bend in perianth tube to top of uppermost lobe 3 to 3°5 cm. ... 5. variegata 6. Perianth tube usually 0-5 to 0-8 cm. diam. at the top, rarely up to 1 cm.; length from bend in perianth tube to top of uppermost lobe rarely more than 2°5 cm. ... 5. variegata var. Metelerkampiae I. S. parviflora Lewis sp. nov. Cormus ovoideus, 1:2-1°5 cm. diam., tunicis e fibris numerosis tenuibus compositis. Caulis plerumque 15-30 cm. altus, simplex vel basin versus uno ramo. Folia 7-9, acuta, 3-18 cm. longa, 0-3—1-2 cm. lata, nervis inconspicuis. Spica 1-4-flora, plerumque 2-3. Bractea 1-1-5 cm. longa, 0-8-1 cm. lata, plus minusve tricuspidata, lineis rubrido-brunneis distincta; bracteolae bractea aequilongae, cuspidibus acutis vel acuminatis. Flores gilvi extrinsecus lilacino- suffusi, segmentis inferioribus lateralibus croceis basi minute brunneo- maculatis. Pertanthi tubus 9 mm. longus, parum curvatus, parte cylindrica 6 mm. longa, parte superiore leviter infundibuliform1; segmenta obtusa, 1-1-2 cm. longa, 5—7 mm. lata, supremum quam cetera leviter longius et latius, 3 inferiora basin versus breviter attenuata. [lamenita 6 mm. longa; antherae 4 mm. longae. Ovarium 4. mm. longum, 2-5-3 mm. diam.; stylz ram: 2 mm. longi. Corm ovoid, 1:2-1°5 cm. diam.; tunics of very fine matted fibres. Stem usually 15-30 cm. high, simple or with 1 branch arising fairly near the base. Leaves 7-9, acute, 3-18 cm. Jong, 0-3-1-2 cm. wide, the veins inconspicuous. Spikes 1-4-flowered, usually 2-3. Bract 1-1-5 cm. long, 0-8-1 cm. wide, more or less tricuspidate, streaked with dark red-brown; bracteoles as long as the bract, with acute or acuminate cusps. Flowers cream-coloured, flushed with mauve outside, the lower side Jobes bright yellow with a pair of small brownish comma-shaped marks at the base of each; pertanth tube 9 mm. long, the cylin- drical part 6 mm. long, slightly curved and expanded above; perianth lobes obtuse, 1-1:2 cm. long, 5-7 mm. wide, the uppermost only slightly longer and wider than the others, the 3 lower very shortly attenuate at the base. Filaments 6 mm. long; anthers 4 mm. long. Ovary 4 mm. long, 2:5-3 mm. diam.; style branches 2 mmm. long, the upper half flattened, conduplicate and recurved. Type. Lewis (65637 in the 8. Afr. Mus. Herb.). Malmesbury Division: between Mamre and Darling, on sandy ground, Lewis (S.A.M. 65637); near Darling, Lewis (S.A.M. 56745); near Hopefield, Schlechter 5304; Bolus (B.H. 12853); Langebaan, Salter 3025; Lewis (N.B.G. 2032/32 in Bol. Herb.). Flowering season. September. A very distinct species. The flowers are by far the smallest in the genus and the perianth lobes less irregular than in the other species. Its nearest ally is S. villosa from which it differs in having finer corm tunics and smaller flowers, A REVISION OF THE GENUS SYNNOTIA I41I x Figure 1. Synnotia parviflora Lewis. 1. Plant in flower. 2. Bract. 3. Bracteoles. 4. Flower, laid open. 5. Stamen, x2. 6. Gynaeceum, x 2. Del. G. J. Lewis. 142 ANNALS OF THE SOUTH AFRICAN MUSEUM with the uppermost lobe only slightly larger than the others and the lower side lobes not auriculate. Among the specimens in the Bolus Herbarium collected by Bolus (B.H. no. 12853) there is one very robust plant, about 50 cm. high, with leaves about 1°5 cm. wide, but with the normal small flowers. All the others on the same sheet are the typical rather small plants described above. 2. §. villosa (Burm. f.) N.E. Br., Kew Bull., 1929, 133; Gladiolus villosus Burm. f., Fl. Cap. Prodr., 2 (1768); Gladiolus bicolor Thunb., Diss. Glad., 16, t. 2, f. 1 (1784); Prodr. 8 (1800); Jacq., Ic., 11, t. 240 (1786-93); Coll. Suppl., 25 (1796); Ixza bicolor Sims, Bot. Mag., t. 548 (1802); Sparaxis bicolor Ker, Konig & Sims Ann., 1, 225 (1805); Synnotia bicolor (Thunb.) Sweet, Hort. Brit., ed. 1, 398 (1827); ed. 2, 501 (1830); Klatt, Linnaea, xxxii, 750 (1863); Baker, FI. Cap., vi, 134 (1806). Corm globose, 1-1°5 cm. diam., with hard tunics, reticulate above and the lower half of rather coarse vertical parallel strands. Stem 12-35 cm. high, usually slightly flexuose, simple or with 1 or 2 branches. Leaves usually 7 in the rosette, acute or subacute, rarely obtuse, 3-21 cm. long, 0-4-1°5 cm. wide, usually 10-18 cm. long and 1 cm. wide, the veins inconspicuous. Spike 1-5- flowered, usually 2-3-flowered. Bract 1-5-2:5 cm. long, 1-1-4 cm. wide, setaceous-cuspidate, the margins irregularly denticulate, sometimes more or less cuspidate or lacerate, colourless or pale straw-coloured at the base with reddish-brown streaks above; bracteoles similar, setaceous-bicuspidate. Flowers pale yellow, the uppermost lobe pale mauve, sometimes the tips of the lower lobes as well; perianth tube 1-5-2 cm. long, the cylindrical part 0-7-1 cm. long, the upper part broadly funnel-shaped: perianth lobes usually with the lower half more or less connivent, the uppermost erect and the other 5 with the upper half somewhat patent or recurved; uppermost lobe obovate, 1-4—1-9 cm. long, 0-g-1°3 cm. wide, sometimes slightly emarginate; upper side lobes 1-1°4 cm. long, 5-6 mm. wide; lower side lobes shortly unguiculate, auriculate above the claw, 1-1°3 cm. long, 5-8 mm. wide above the claw; lowest lobe smaller, 7-g mm. long, 4-5 mm. wide, not unguiculate nor auriculate. Stamens reaching to or slightly above the middle of the uppermost lobe, slightly arcuate; filaments about 1-5 cm. long; anthers 4 mm. long. Ovary 4—5 mm. long, 2°5-3 mm. diam.; style branches 3 mm. long, the upper half expanded, conduplicate, minutely ciliate. Lectotype. As there is no specimen in Burmann’s collection and the figure on which his name is based (Breyne, Jc. Rar. Pl., t. viii, f. 2) dates back to 1739, it seems advisable to retain Thunberg’s type of Gladiolus bicolor as the type of this species. There are two sheets in his herbarium in Uppsala, from Groene- kloof, in the Malmesbury Division. Cape Division: Lion’s Head, MacOwan 798; Signal Hill, Marloth 188; fields at Observatory, W. Dod 1443; between Salt River and Kalabaskraal, 13 miles from Cape Town, Hutchinson 175; Vissershok, Barker 1811. Malmesbury A REVISION OF THE GENUS SYNNOTIA 143 Division: Groenekloof, Thunberg; north-east slopes of Dassenberg, Pzllans 6859; Malmesbury, Barker 4667, 6394; Guthrie 2386; Lewis (S.A.M. 60674, 65642) ; between Mamre and Darling, Lewis (S.A.M. 56747); Darling, Compton 19880. Piketberg Division: De Hoek, Steyn 532. Paarl Division: Paarlberg, Drége 8347; Dal Josafat, near Wellington, Grant 2333; near Wellington, Martin, s.n. Tulbagh Division: near Saron, Schlechter 4869; Tulbaghskloof, Ecklon and Keyher 121 (77:9); near Artois, Bolus 7586. Stellenbosch Division: Klapmuts, Prior. Without locality: Fusszeu 3608; Lamarck (named Ixia aristata). Flowering season. August to September. It is rather unfortunate that the old and seemingly inappropriate specific epithet used by Burmann has to be retained for this species which, like the others, is entirely glabrous. There is no specimen in Burmann’s collection and he gave no description of the plant, merely citing Breynius’ figure. According to N. E. Brown this is the plant known as Synnotza bicolor (Thunb.) Sweet, but as Burmann’s name is older it should be called S. villosa. Brown added that the figure of Breynius is a fairly good one, and quite unmistakable. More than a century earlier Sims was of the same opinion, for in Bot. Mag., t. 548 (1802), under Ixia bicolor (=Gladiolus bicolor Thunb.) he gave as a synonym Gladiolus villosus Burm. and cited the figure of Breynius. Although there is no doubt that the figure is of this species, and the des- cription given by Breynius applies to it, his use of the adjective villosus in the description, which was adopted by Burmann for the specific epithet, is at first rather puzzling. It is evident from the description and illustration, how- ever, that it applies to the hair-like cusps at the apex and on the margins of the bracts and bracteoles, and bearing in mind that at the time when he described the species it was probably the only plant of its kind in Europe, this ‘character of the bracts was a striking one and distinguished this species from all other species of Gladiolus, to which genus it was then considered to belong. 3. §. galeata Sweet, Hort. Brit., ed. 1, 398 (1827); ed. 2, 501 (1830); Dur. and Schinz, Consp. Fl. Afr., v, 211 (1895); Gladiolus galeatus Jacq., Ic., ii, t. 258 (1786-93) and Coll. iv, 167 (1790), non Burm. f. (1768); Sparaxis galeata Ker, Konig. & Sims Ann., 1, 225 (1805); Synnotza bicolor Pole Evans in Fl. Pl. of S. Afr., v, t. 162 (1925), non S. bicolor (Thunb.) Sweet. Corm oblong-ovate or ovoid, 1-2-2:2 cm. diam.; tunics of numerous fine hair-like fibres, nearly always extending up in a neck 2-5-3:5 cm. long. Stem 12-35 cm. high, somewhat flexuose, simple or with a branch arising near the base, rarely with 2 branches. Leaves 5-9, acute or obtuse, usually obtuse and apiculate, 2-16 cm. long, 0-5-1°5 cm. wide, sometimes spotted or banded with purple at the base, the marginal veins usually slightly thickened, the other veins not very conspicuous. Spike 2—-5-flowered. Bract 1-2-2 cm. long, 0-7-1 cm. wide, setaceous-cuspidate, the margins irregularly denticulate, colourless or the upper half pale reddish-brown, speckled all over with fine 144 ANNALS OF THE SOUTH AFRICAN MUSEUM red-brown streaks; bracteoles similar, setaceous-bicuspidate. Flowers cream- coloured more or less suffused with mauve, with purplish patches at the base of the 2 upper side lobes, the 3 lower lobes bright yellow, sometimes tipped with pale mauve; perianth tube up to 1-3 cm. long, the cylindrical part 6-7 mm. long, narrowly infundibuliform above; perzanth lobes very unequal, the uppermost longer than the others, erect and well separated from the other 5 lobes when the flowers are fully open, 1-8—-2-6 cm. long, o-6—1-2 cm. wide, shortly unguicu- late and more or less oblong above the claw; upper side lobes patent, 1-5-2 cm. long, 6-8 mm. wide; 3 lower lobes alike, forming a distinct deflexed lip, 1°3-1°6 cm. long, 4-5 mm. wide, usually very shortly unguiculate. Stamens very arcuate; filaments 2 cm. long; anthers 5-6 mm. long. Ovary 2-5-3 mm. long, 2 mm. diam.; style branches 4—5 mm. long, filiform, slightly expanded at the apex. Mypew a )acquin. Hc nih at 250: Clanwilliam Division: west side of Olifants River, about 2 miles east of Claypan and 7 miles south-west of Klaver, stony koppie on farm Driefontein, Nieuwoudt (S.A.M. 67740); Claypan, Ponder (N.B.G. 1466/28 in Bol. Herb.). Van Rhynsdorp Division: near Van Rhynsdorp, Rood, N.H. 2860 (Pretoria) ; N.B.G. 2306/23 (Bol. Herb.); between Van Rhynsdorp and Clanwilliam, Marloth 7469; foot of Van Rhyn’s Pass, in hard reddish soil, Lewis (S.A.M. 60136); Barker 6445. Calvinia Division: Driefontein, south-west of Calvinia, Marloth 10426. Flowering season. End of July to middle of September. No specimens were seen by Baker, who placed Synnotia galeata Sweet and Gladiolus galeatus Jacq. as synonyms under Synnotia bicolor in the Flora Capensis (vi, 135), in spite of Jacquin’s excellent figure which depicts a plant quite unlike the one figured by him as Gladiolus bicolor in the same volume (Jc., t. 240). The two species are quite distinct, with very different corm tunics and flowers. The corm tunics of S. villosa are of hard, reticulate fibres, not extending up in a neck, while those of S. galeata are fine, soft and thickly matted, and nearly always extend up in a neck to ground level. In S. villosa all the perianth lobes are more or less connivent in the lower half, giving the flower a somewhat closed and broadly funnel-shaped appearance, whereas in S. galeata, though the immature flower somewhat resembles that of S. villosa, the mature flower has a wide open appearance, with the uppermost lobe well apart from the others and the upper side lobes also widely separated from the 3 lower. In addition the lower side lobes of S. villosa are auriculate near the base and the uppermost lobe is obovate, whereas in S. galeata the lower lobes are not auriculate and the uppermost, which elongates as the flower matures, is finally much longer and comparatively narrower, at least twice as long as wide. For comparison a reproduction of the inflorescence of S. villosa (from Bot. Mag., t. 548) and S. galeata (from Jacq., Jc., t. 258) is given in figure 2, nos. 1 and 2. A REVISION OF THE GENUS SYNNOTIA 145 I Figure 2. 1. Synnotia villosa (Burm. f.) N.E. Br., from Bot. Mag., t. 548 (Ixia bicolor Sims). 2. Synnotia galeata Sweet, from Jacq., Ic., i, t. 258 (Gladiolus galeatus Jacy.). The purple mottling near the base of the leaves does not appear to be a constant feature in S. galeata. The plant figured by Jacquin shows no trace of any marks and on only two of eight specimens seen from near Claypan on the west side of the Olifants River are there rather faint purple spots. In the majority of specimens from other localities, however, the leaves are con- spicuously spotted or banded with purple, as is shown on the plant figured in Flowering Planis of S. Africa (v, t. 162). The leaves are often obtuse, usually obtuse and apiculate, more or less as shown by Jacquin, and are not all very 146 ANNALS OF THE SOUTH AFRICAN MUSEUM acute as described in Flowering Plants. As well as having most of the leaves marked with purple, plants from the foot of Van Rhyn’s Pass and near Calvinia have slightly larger and firmer bracts than those from near Claypan, but they are not considered to be more than a local form. It is very probable that Jacquin’s type came from the west side of the Olifants River as the plants recently collected on the farm Driefontein, near Claypan (Sept. 1955), most closely match his figure, and this locality was on the old route to Namaqualand followed by the early travellers, near the Company’s Drift where the Olifants River was crossed. 4. S. Roxburghii (Baker) Lewis sp. nov. S. bicolor (Thunb.) Sweet var. Roxburgh Baker, Handbk. Irid., 198 (1892) ; Fils Cap.; vi... 135 (1896). Cormus globosus, 2-5-3 cm. diam. tunicis e fibris copiosis capillaribus com- positis in collum 7-8 cm. longum productis. Caulis 28-42 cm. altus, simplex vel uno ramo interdum 2 ramis. Folia 7-9 basalia, 5-23 cm. longa, 0:5-1-4 cm. lata, acuta vel subacuta, nervis conspicuis sed non crassis; folium caulinum 1, 4-6 cm. longum, acutum vel acuminatum. Spica 2—-7-flora. Bractea 1-5-2°5 cm. longa, o-8-1-1 cm. lata, setaceo-cuspidata marginibus denticulatis vel leviter laceratis; bracteolae similes, setaceo-bicuspidatae. Flores pallide lilacini plerum- que segmentis tribus inferioribus in parte inferiore luteis; perzanthi tubus 2-3 cm. longus, parte cylindrica 1-3-2-3 cm. longa, superne curvatus gradatim ampliatus et enguste infundibuliformis; segmenia obtusa, supremum 2-2-4 cm. longum, 1-1-6 cm. latum breviter et late unguiculatum, lateralia superiora 1-6—2-2 cm. longa, 0:5-I'1 cm. lata; segmenta inferiora 1-3-2 cm. longa, 4-6 mm. lata, basin versus leviter attenuata. Stamina clare arcuata; filamenta 1-5-2 cm. ae antherae 4 mm. longae. Ovarium 4 mm. longum, 2-3 mm. diam.; stylus 9-3-5 cm. longus ramis filiformibus apice leviter ampliatis. Corm globose, 2:5-3 cm. diam.; tunics of numerous fine hair-like fibres extending up in a neck 7-8 cm. ion Stem 28-42 cm. high, simple or with 1 or 2 branches. Leaves 7-9 in the basal rosette, 5-23 cm. long, 0:5—1-4 cm. wide, acute or subacute, the veins visible but not prominent; 1 cauline leaf near the middle of the stem, 4-6 cm. long, acuminate. Spzkes 2—7-flowered. Bract 1°5-2°5 cm. long, o-8-1-1 cm. wide, setaceous-cuspidate with denticulate or slightly lacerate margins, colourless near the base, pale reddish-brown above with darker red-brown stripes; bracieoles similar, setaceous-bicuspidate. Flowers mauve or lilac-coloured, usually with the lower half or two-thirds of the lower lobes yellow; perianth tube 2-3 cm. long, the cylindrical part 1-3-2-3 cm. long, curved and expanding gradually into the narrowly infundibuliform part; perianth lobes obtuse, the uppermost 2—2°4 cm. long, 1-1-6 cm. wide near the middle, shortly and broadly unguiculate; upper side lobes 1-6—2-2 cm. long, o-5-I't cm. wide; 3 lower lobes 1-3-2 cm. long, 4-6 mm. wide, slightly narrowed to the base but not unguiculate. Stamens distinctly arcuate: filaments A REVISION OF THE GENUS SYNNOTIA 147 1-5-2 cm. long; anthers 4 mm. long. Ovary 4 mm. long, 2-3 mm. diam.; style 3-3-5 cm. long, with filiform branches slightly expanded at the apex. Type. South Africa, without locality, Roxburgh (not seen). Clanwilliam Division: Olifants River Valley, 20 miles south of Clanwilliam, Salter 7495; 15 miles north of Citrusdal, Lewis (S.A.M. 57911); between Clan- william and Citrusdal, Wilman 850; Kriedouwkrans, between Clanwilliam and Citrusdal, Barker 5756; Olifants River Valley, Steyn 383. Flowering season. August to September. This species is closely allied to S. galeata from which it differs in having larger flowers with a longer perianth tube, the uppermost lobe comparatively wider, and the lower lobes not clawed at the base. Baker placed it as a variety of S. bicolor, with a very brief description—‘Bulb tunics of fine threads; flower all lilac-purple’. Unfortunately I have not seen the type but it is almost certain that the plants described and cited above must be the same. Baker’s description can scarcely apply to S. galeata, in which the flowers, even in dried material, are not all lilac-purple, and his description of the corm tunics pre- cludes any likelihood of it being S. Meielerkampiae L. Bol. All the records of this species are from the same small area in the Olifants River Valley. 5. S. variegata Sweet, Brit. Flow. Gard., t. 150 (1826); Klatt, Linnaea, XXXII, 750 (1863); Baker, Handbk. Ind., 198 (1892); Fl. Cap., vi, 135 (1806); S. versicolor Steud., Nom., ed. 2, 11, 614 (1840); 8. Waltham Hort. Kew, cf. Gard. Chron. (1881), 1, 370, nomen. Corm globose or ovoid, 1-2-2 cm. diam.; tunics of hard fibres closely reticulate above with numerous rather coarse vertical parallel strands in the lower half, not extending up in a neck but sometimes with one or two old tunics present above the corm giving the appearance of a neck. Stem 8-40 cm. high, usually 15-20 cm., simple or more often 1 to 3-branched. Leaves 7-5 in the rosette, 2-15 cm. long, 0-5-2 cm. wide, usually 5-10 cm. long and 1-1-5 cm. wide, obtuse or acute, the veins visible but only the mid-rib slightly prominent; cauline leaves 1-3, 4-10 cm. long, acute or acuminate. Spike 2—7-flowered, usually 2—-4-flowered. Bract and bracteoles 2-2:5 cm. long, 1-1-4 cm. wide, setaceous-cuspidate with irregularly denticulate margins, usually colourless in the lower half and pale red-brown above, with a few conspicuous dark red- brown streaks. Flowers varying in colour from pale mauve and yellow or cream to deep purple and yellow, the lower lobes sometimes all yellow or tipped with mauve or sometimes with the upper half or more purple, with narrow purple and yellow stripes running down into the perianth tube; pertanth tube with the cylindrical part 2-5-4 cm. long, usually 2-5-3 cm., geniculate at the top and abruptly dilating into a broad oblique funnel 1-1-5 cm. long and 1-1°5 cm. diam. at the top, the length from the bend in the perianth tube to apex of uppermost lobe 3-3:5 cm.; uppermost perianth lobe obovate or oblong-ovate, 2-3 cm. long, 1-2 cm. wide; upper side lobes 148 ANNALS OF THE SOUTH AFRICAN MUSEUM 1:5-2 cm. long, 0-7-1 cm. wide; lower side lobes unguiculate, auriculate above the claw, 1:3-2 cm. long, 6-8 mm. wide; lowest lobe 1-2--1-5 cm. long, 5-7 mm. wide. . Filaments 1-5-2°2 cm. long; anthers 5-7 mm. long. Ovary ‘—5 mm. long, 2°5-3 mm. diam.; style branches 4-5-6 mm. long, the upper half expanded, conduplicate and minutely ciliate. Type. Sweet, Brit. Flow. Gard., t. 150. Clanwilliam Division: Clanwilliam, Leipoldi, 163, 226; Mader (in Herb. MacOwan 2138); Olifants River Valley, near Rondegat, Schlechter 5037; Clanwilliam Barrage, Esterhuysen 5809; Barker 14.77; in kloof near Clanwilliam Barrage, Lewis (S.A.M. 60116); near Vanrhynsdorp, Lezpoldt 809; Nardouw Pass, Barker 7435, 3624; Lewis (S.A.M. 59851); Olifants River Valley, near Nardouw road, Lewis (S.A.M. 57912); near Langkloof, south of Nardouw road, Leighton 1110; 9 miles north of Clanwilliam, Lewis (S.A.M. 60673); Olifants River Valley Cave, south of Clanwilliam, Compton 22772; 5 miles south of Clanwilliam, Lewis (S.A.M. 67867). Flowering season. August to September. S. variegata Sweet var. Metelerkampiae (L. Bol.) Lewis, comb. nov. S. Metzlerkampiae L. Bol., Ann. Bol. Herb., ii, 77 (1923); Fl. Pl. of S. Afr., i, t. 98 (1923); Sparaxis luteo-violacea Eckl., Top. Verz., 27 (1827), nomen nudum. Corm, leaves, inflorescence and bracts as in S. variegata, the corm slightly smaller. Flowers with shorter and narrower perianth lobes, the length from the bend in the perianth tube to the top of the uppermost lobe rarely more than 2-5 cm.; cylindrical part of the perianth tube as in S. variegata, the upper part narrower, usually 0-5-0-8 cm. diam. at the top, occasionally up to 1 cm.; uppermost perianth lobe oblong or ovate-oblong, 1-3-2-3 cm. long, 0-7—1-3 cm. wide; upper side lobes 1-1-5 cm. long, 4-6 mm. wide; lower side lobes 0-9-1°3 cm. long, 4-6 mm. wide, varying from shortly unguiculate and auriculate above the claw to subunguiculate and not auriculate; lowest lobe o-8-1-1 cm. long, 4-5 mm. wide. Type. Clanwilliam Division: near Eendekuil, Metelerkamp (B.H. 16039, in Bolus Herb.). Clanwilliam Division: near Eendekuil, Metelerkamp (B.H. 16039); Olifants River, near Brakfontein, Ecklon and