ILLUSTRATIONS OF CARPEL POLYMORPHISM. II BY EDITH R. SAUNDERS,

Transcription

1 175 ILLUSTRATIONS OF CARPEL POLYMORPHISM. II BY EDITH R. SAUNDERS, Sometime Fellow of Newnham College, Cambridge (With 62 figures in the text,) BERBEEIDACEAE HE farnily of the Berberidaceae, according to the former view T that the carpel is monomorphic, is characterised by having an ovary composed of a single carpel. As was, however, shown in an early statement of the polymorphic theory^ one genus belonging to this family Fpimedium affords the clearest example of the polymorphic condition, the ovary being obviously not monomerous, but dimerous. It was pointed out also that Jeffersonia was evidently constructed upon a similar plan. Examination of other genera {Podophyllum, Berberis and Mahonia, Leontice, Nandina) has now made it abundantly clear that this dimerous condition is found throughout the whole family. There is, consequently, no longer need to formulate some plausible explanation of the solitary terminal carpellary leaf for here it clearly does not exist. It is hoped in a later account to show that in other families in which this anomaly has been believed to occur the i-carpel interpretation is similarly at variance with the evidence afforded by various structural features. Fpimedium. The structure of the gynoecium in this genus has already been fully described I The main features are recalled here merely for comparison with the other genera dealt with below. The ovary is composed of two carpels, one of the valve type, small, sterile, and destitute of style and stigma, the other, large, semi-solid and fertile, with style and capitate stigma. The membranous ovary wall shows an obliquely vertical furrow on each side indicating the lines of junction of the two members. The wall here is thinner than elsewhere and destitute of vascular tissue, the venation system of both carpels falling short of their boundaries. The dry fruit dehisces along the two sutures when ripe; the sterile valve is then shed exposing the seeds stdl attached to the inner face of the persistent, larger carpel. ' Carpel Polymorphism, I, A>i7ials of Botany, 39, x)p Figs , " Loc. cit.

2 176 EDITH R. SAUNDERS

3 Illustrations of Carpel Polymorphism. II 177 Podophyllum peltatum L. (Figs. 1-4). The much larger gynoecium of Podophyllum has the vascular system particularly well developed. Transverse sections through the ovary base show on the one side the midrib and wing bundles of the expanded, sterile, valve member, and on the opposite side the bundle system of the consolidated, fertile carpel (Fig. i). This latter system shortly becomes concentrated into the twin strands of the midrib and the double bundle of the placenta (Figs. 2, 3). At this and higher levels the numerous lateral veins of the valve are seen cut obliquely. As is commonly the case with the consolidated carpel type the ovules, which extend almost to the top of the placental cushion, are borne in several rows. The only other feature calhng for comment is the stigma which appears as a series of convolutions crowning the ovary, and so intercoiled as to obscure any definite relation to the two carpel members (Fig. 4). Figs Berberidaceae Podophyllum peltatum L, i. Transverse section of the ovary below the level of the loculus. In the upper half are seen the midrib, lateral (wing) bundles and two smaller strands belonging to the sterile valve carpel, on the lower side the bundles destined for the fertile carpel. 2. The same at the level of development of the loculus; the fertile carpel shows twin bundles constituting the midrib and a double bundle in the placenta. 3. The same at a higher level with the placental cushion now well developed. 4. The ovary apex viewed from above showing the stigma Berberis vulgaris L. 5. Transverse section of the ovary taken through the extreme base. Towards the upper side, the midrib and secondary veins of the sterile valve; opposite, the large cord (midrib) of the fertile carpel. 6. The same just below the level of origin of the loeulus. Above, the midrib, and at the periphery, the secondary veins of the valve; below, the midrib of the fertile earpel; in the eentre, the numerous strands given off from the midrib which traverse the placenta and serve the ovules. 7. The same at the ovule-beating level. 8. The gynoecium showing the circular form of the stigma. 9, ro. Berberis (Mahonia) Aquifolium Pursli. The gynoecium and accessory carpel structures in the flowering (9) and fruiting (10) stage Leontice Chrysogonum L. {Bongardia Rauwolfii C. A. Mey.). 11, 12. Two views of the gynoeciuiti. 11. Side view of the two carpels. 12. Face view of the solid, fertile carpel Transverse sections of the gynoecium at successively higher levels. 13. The ovary below the level of the loculus. The midrib and lateral veins of the sterile, valve carpel are seen eut longitudinally as they run out to the periphery; the strands destined for the fertile carpel are not yet fully differentiated. 14. The same, as the two carpels begin to separate along their inner faces to form the loculus. 15. The same, in the ovule-bearing region. 16. The same, above the ovule-bearing region. 17. The stigma after the stylar canal has opened on to the surface; the larger lip is formed of the median portion of the solid carpel, the smaller, of the lateral portions of the same connected by a narrow bridge of tissue derived from the valve earpel. The regions corresponding to the solid earpel are abundantly supplied with vascular tissue. ac, accessory carpels; fcb, fertile carpel bundles; fee, fertile carpel cord; fcm, fertile carpel midrib; I, loculns; Iv, lateral veins of the valve carpel; ov, ovule; plb, plaeental bundle; scm, sterile earpel midrib.

4 178 EDITH R. SAUNDERS Berberis vulgaris L. (Euherberis), Berberis Aquifolium Pursh. {Mahonia) (Figs. 5^10). The ovary in Berberis is constructed upon the same plan as that of Podophyllum, but here the ovules, similarly borne in several rows, are restricted to the basal region of the placenta. At the ovary base the two carpel midribs are easily distinguished (Fig. 5), but the repeated branching of the fertile vascular cord (Fig. 6) to form the several strands which later become condensed in the placental cushion to correspond with the number of ovules (Fig. 7) renders the construction at this intermediate level less clear (Fig. 6) than in the preceding genus; but as the loculus level is reached the bicarpellary ground plan is again clearly apparent (Fig. 7). The whole of this placental vascular tissue is used up in supplying the ovules but the midrib is continued to the top, as well as the venation system of the valve carpel. As the stigma region is approached the two systems give rise to a ring of small equal bundles, a reflection of this symmetrical distribution being seen in the circular form of the stigma, the soldering of the two carpels being here complete to the top (Fig. 8). Occasionally individual branches on a Mahonia bush are found bearing flowers with an extra whorl of floral structures between the androecium and the gynoecium proper. I am indebted to Miss D. Milner Brown for some specimens showing this peculiarity in the flowering (Fig. 9) and fruiting (Fig. 10) stages. These supernumerary structures, which vary in number, are obviously carpellary in nature. They may be separate or partially united; in the latter case they form a lobed "collar" round the normal ovary. The separate structures, in their most reduced condition, are thread-like with a capitate stigma and correspond each to a single carpel. More often they are plate-like and stigmatic along the extent of the upper edge. From the present point of view these latter structural forms are of considerable interest. They are unable, doubtless owing to spatial relations, to form so many individual, closed ovaries, but they are, nevertheless, evidently bicarpellary for they show two separate vascular cords. In one of these exceptional flowers one of these accessory organs was found to be as weu developed as the normal ovary to which it was adherent, producing in this way the appearance of "twinning." Although distorted in form the accessor}' ovary contained ovule-like structures. Leontice Leontopetalnnt L., L. Chrysogonum L. (Bongardia Rauwolfii C. A. Mey.) (Figs ). Leontice resembles Fpimediuni in that the bicarpellary construction is indicated in the outward appear-

5 Illustrations of Carpel Polymorphism. II 179 ance of the gynoecium, but the size relation of the carpels is reversed in the two cases. For whereas in Epimedium the fertile member is semi-solid and has considerable lateral extension, the valve, in consequence, being small, in Leontice the fertile carpel is contracted to the completely sohd form and is seen on one side of the ovary as a stout rib or column of tissue which is continued upwards as the style (Figs, 11, 12), The rest of the wall, which is thin and membranous, and served by a series of fine secondary veins, is formed by the large valve carpel (Fig, 15), In the style column, however, the valve carpel component becomes greatly reduced until at the stigma level it is represented by a short non-vascular bridge connecting the incurved edges of the solid member. In both species the dry fruit wau sooner or later ruptures, becoming irregularly torn, but in neither case is there true dehiscence. This is no doubt attributable to the extreme paucity of such mechanical elements'^ in the vascular bundles as would be capable of setting up the degree of tension which one assumes to be required to bring about dehiscence at some particular locus. In a cross-section taken through the ovary base of L. Chrysogonum strands are seen running out from the residual vascular tissue to the periphery round the greater part of the circumference. These become the midrib and secondary veins of the sterile valve member (Fig, 13), At a slightly higher level the loculus makes its appearance as a periclinal split (Fig, 14) between an arc of thin wall formed of the valve carpel and a bulky mass of tissue which separates more and more from the valve wall as it becomes differentiated into the solid carpel with midrib and placcntal bundles (Fig, 15), The ovules, which are confined to the basal portion of the placenta, arise on several radii. The whole of the vascular tissue in the placenta is utilised in their formation so that above the ovule level the midrib alone persists (Fig, 16), In the valve carpel, on the other hand, the midrib is weakly developed and soon comes to an end, a pair of lateral veins alone persisting towards the apex. As the stigma level is reached the stylar canal opens out on to the surface right and left of the fertile cord and at some distance from it. Of the two lips thus formed the smaller, in its middle region, is destitute of vascular tissue. This region represents the last remnant ' In the flowering stage these consist of no more than some two or three very small vessels in each vein in an ovary wall which, in L. Chrysogonum, mostly shows but one single thin cell layer between the inner and outer epidermis, and in L. Leontopetalum, generally not more than three or four similar layers.

6 i8o EDITH R. SAUNDERS wb 34

7 Illustrations of Carpel Polymorphism. II of the valve carpel. The margins of this lip and the whole of the larger one, which together represent the solid carpel, are well supplied with vascular elements (Fig. 17). Nandina domestica Thunb. (Figs ). The occurrence of certain variations from the normal in the gynoecium of Nandina renders this genus one of the most interesting in the family from the present viewpoint. The ovary at first is radially symmetrical with erect stigma lobes (Fig. 18) but as development proceeds it becomes markedly bilateral, the apical portion being bent to one side at a right angle so that it roughly resembles the head and beak of a bird (Figs. 19, 20). This zygomorphism is due to unequal growth in length of the two carpels, the longer forming the upper, the shorter, the lower lip of the "beak." Here we have developed but slightly that inequality in size between the two component carpels which is exhibited in such a much more pronounced form in Jeffersonia, Leontice and Epimedium. Both stigmatic lips show a number of blunt marginal lobes, reflecting probably the pattern of the lateral vein system (Figs. 21, 22). The most striking feature of the pistil, however, is that both carpels are of the valve type though generally one alone is fertile, this member being then more fully expanded as well as longer than the other (Figs ). Exceptionally, however, both may expand equally and both bear ovules (Figs. 29, 30). Ovaries of this type were observed by Citerne^ and correctly interpreted, but he noted them merely as abnormalities and missed their significance in relation to the ordinary type. In these exceptional cases the ovary may show a furrow down each side along the line 1 Berberidees et Eryihrosperni6es, p. 28. Paris, i8i Figs. i8-. O. Berberidaceae (cohi'ihherf). Nandina domcsticathunh. 18. Gynoecium apex from a young bud, showing erect stigma lobes (highly magnified). 19. Whole gynoecium at an older stage with the apex bent over to one side. 20. The same more highly magnified. 21, 22. Apex of the same still further magnified. 21. The "beak" in side view. 22. The same in face view All from transverse sections taken at successively higher levels From a well-developed ovary in which the two carpels are slightly unequal From a well-developed ovary of two equal carpels From a poorly-developed ovary. The infertile carpel is much reduced and destitute of vascular tissue; the second loculus is represented by a narrow channel leading out of the main chamber From an ovary composed of three equal valve carpels which ultimately become disunited. 40. Portion of the apex of one of the three carpels cut off and laid fiat showing the lobing of the stigmatic margin. /, loculus; m, m', m", midrib bundle; pc, papillose cells of the stigma; sic, stylar canal; wb, wb', wb", wing bundles.

8 182 EDITH R. SAUNDERS of the conjoined edges thus produeing an effect as of "twinning" of the gynoecium. Both carpels e.xpand to enclose a chamber and the ovary may then for a short distance at the base be bilocular (Fig. 29). But the locuh enlarge at once, the intervening tissue disappears, and the two become continuous. Where the partition previously appeared there are now to be seen four projecting points, two in close juxtaposition on each side of the now single extended loculus, each of these pairs developing hairs and representing the conjoined margins, on the one side and the ottier, of the two carpels. The carpels show a midrib, two prominent wing bundles and a number of smaller veins running more superficially (Fig. 30). In all cases the secondary veins gather together towards the apex into a smaller number of larger bundles (Figs ), and their distribution possibly determines the lobing of the stigmatie margins referred to above (Figs. 21, 22). In poorly developed ovaries the infertile carpel may be reduced to such an extent as to show no vascular elements at all, only undifferentiated tissue occupying the area between the wing bundles of the fertile member^ (Figs ). In these cases the ovary is unilocular throughout, the main chamber representing the loculus of the fertile earpel. From this there opens out on a radius between the wing bundles a narrow bay or channel which is au that the reduced carpel is able to aehieve in the way of a loculus (Fig. 32). I was for long puzzled by this indentation in this tj^pe of ovary until I sueceeded in obtaining more favourable material in whieh the bicarpellary ground plan was unmistakable. Occasionally tricarpellary ovaries occur, each of the three carpels being of the valve type and enclosing a loculus (Figs ). The fitting together of three members seems, however, to present some difficulty. One of the three may sooner or later become disjoined from the other two and suffer some distortion. Or all three may become free and assume the form of folded leaves, the disunited portions dovetailing one with another (Fig. 39). The evidence furnished by the series of types described above affords the clearest proof that the former monocarpellary hiterpretation of the gynoecium is untenable even apart from the fundamental difficulty involved in the coneeption of a terminal fohar organ. A point deserving of notice is that when two placentae belonging to different carpels bear eaeh an ovule, as undoubtedly is occasionally ' A small group of a few xylem elements seen in one case close to one of the wing bundles, which came to an end below the level of the loculus floor, possibly represented a last remnant of the residual vascular tissue from which the second carpel cord is diiferentiated in better developed ovaries.

9 Illustrations of Carpel Polymorphism. II 183 the case, we have a genuine instance of parietal placentation in a syncarpotts gynoecium, a condition which in the light of our present knowledge must be regarded, contrary to received opinion, as of rare occurrence. This feature of his exceptional cases was appreciated by Citerne but he failed to perceive that these very cases provide us with a most valuable clue to the plan of construction of the ordinary type of gjmoecium. For from these instances of equal carpel development we can trace a series showing increasing disparity in size between the two carpel members and consequent asymmetry of external contour as we pass from the various Nandina forms to Epimedium, Leontice and Jeffersonia. In marked contrast to this series is the Berberis-Podophylhtm group, for in these genera the two carpels although dissimilar in tj'pe are of equal height and so evenly united that the mature gynoeeium is radially symmetrical. On the basis of the carpel characters the genera dealt with above can be arranged in series thus: Nandina. Rare variation. Carpels 3, all of the valve type and equal. Nandina. Occasional variation. Carpels 2, both of the valve type, and both equal and fertile. Nandina type. Carpels 2, both of the valve type, slightly unequal, one only being fertile. Nandina, depauperated. Carpels 2, both of the valve type, the infertile one much reduced and destitute of vaseular tissue. Berberis (including Mahonia) and Podophyllum. Carpels 2, unequal, one of the valve type and sterile, one solid and fertile. Ovary radially symmetrical. Leontice (including Bongardia). Carpels 2, unequal, one of the valve type and sterile, one solid and fertile. Ovary bilaterally symmetrical. Epimedium and (probably) Jeffersonia. Carpels 2, unequal, one of the valve type and sterile, one semi-solid and fertile. Ovary bilaterally symmetrical. FUMARIACEAE Ceratocapnos [Corydalis) palaestina Boiss. (Figs. 41, 42). The two or three species included in the section Ceratocapnos are of extreme interest on account of their heteromorphic fruits, urn-shaped in the first few flowers (Fig. 41), sihqua-like in all the rest (Fig. 42). Examination of material of C. palaestina recently sent to me by Mr J. E. Dinsmore from Jerusalem has shown that the lower fruits owe their shape to two short and broad valves, and two solid carpels, not, as I had earlier suggested^ to four similar carpels. In the many-ovuled species of Corydalis and in Dicentra the midribs of the two narrow, sterile, valve carpels give off a succession of lateral veins. The amount of xylem in the bundles is small and 1 From such evidence as is afforded by descriptions and figures. See Annals of Botany, 37, p. 474 and Figs. 52, 53 (p. 477). The sihqua has narrow, long valves.

10 184 EDITH R. SAUNDERS mcc 44

11 Illustrations of Carpel Polymorphism. II 185 there is almost a complete absence of mechanical tissue. In both the fruit types of C. palaestina, which are r-ovuled, the chief lateral veins (usually six in the urceolate form) all take their rise at the ovary base and almost at the same level. Both types differ from the fruits of other species not classed under Ceratocapnos in being what may be described as "two-storeyed," herein showing some resemblance to certain genera of the Cruciferae, Both also develop a large amount of mechanical tissue in the form of longitudinal bands which come to an end just above the junction of the two storeys, anastomosing at this level so as to form a horizontal girdle. These bands are associated with the vascular bundles. In the lower seetion of the urn-shaped fruits the chief lateral veins leave the main cord of the valve as a single trunk which runs out on either side horizontally, Figs, 41-48, Fumariaceae, 41, 42, Ceratocapnos [Corydalis) palaestina Boiss. 41, A npe urn-shaped fruit with the upper triangular portion entire but with one of the two lower valves removed, the remaining valve seen from the outside, 42, A whole immature fruit of the later-formed siliquoid type. Near the edge to the right and left the main cords of the two median solid carpels, the one on the left shows the branch given off to the solitary ovule. Between, the midribs and the strongest of the lateral veins of the two sterile valve carpels, 43-48, Corydalis lutea L. All from transverse sections taken at successively higher levels, 43, Flower stalk. The vascular mass ou the right in process of giving off the composite cord for the spurred petal and superposed stamen, 44-48, Flower base. Part of the spurred (right) petal has been cut away, the cut edges being indicated by the interrupted portions of the contour line, 44, On the right the midrib and lateral veins of the spurred petal; nearer the centre and on the same radius the bundle of the lateral (whole) stamen. The compound cord for the left (unspurred) petal and whole stamen, and the two compound cords consisting of the conjoined bundles of a sepal and an inner petal are leaving the central ring on the left, and iu the median plane back and front, respectively, and are seen cut longitudinall}', 45. At the level of exsertion of the two small median sepals. The two small strands destined for the two half stamens on the right are about to leave the ceutral ring, 46, The conjoined bundles of the sepals and median petals have separated. The small bundles for the two half stamens on the left and the larger cords for the two median solid carpels and the right valve carpel can now be identified. The composite cord on the left has already given off the wing bundles of the petal but the midrib is not yet disjoined from the stamen bundle, 47, Differentiation of the petal, stamen and carpel bundles on the left follows the same course as that already completed on the right where the ovary surface is now exposed, 48 A, The ovary, now almost free, is enclosed on the right by the coherent bases of the stamen triplet, which in turn is encompassed by the spurred petal. On the left tbe tissues of tbe petal (uuspurred), stamen triplet and ovary are still continuous, 48 E, Portion of the same more highly magnified, cwh, sterile carpel wing bundle ;/cc, fertile carpel corci; hsb, half stamen bundle; Ice, lateral conjoined cord of petal and stamen; Isb, lateral stamen bundle; mcc, median conjoined cord of sepal and petal; pnih, pmb', petal midrib bundle; pwb, pwb', petal wing bundle; s, sepal; scr. sterile carpel cord; sctn, sterile carpel midrib; sp, spurred petal.

12 186 EDITH R. SAUNDERS from which the individual veins then separate in succession at the same time, turning again at right angles to follow a parallel, upward course. Hence the characteristic "gridiron" pattern seen in Fig. 41. In the upper section, which in this type forms a triangular cap or lid, only the four carpel midribs persist. This is the case also in the siliqua type, in which the two sections are of about equal length and the veins of the lower section fewer and less prominent. In both types the solitary ovule is borne in the lower section. It is possible that the "two-storeyed" condition is the outcome of this extreme reduction in ovule formation. With respect to the dimorphic character of the fruits it may be said that we are already familiar with instances in other genera where the gynoecium of early flowers frequently shows a deviation from the type form significant of the phylogenetic history. These variations, it is true, are generally in the direction of an increase in carpel number (many Cruciferae, some of the Gramineae), nevertheless it seems scarcely doubtful that in the present case the sudden alteration in the dimensional relations of the two pairs of carpels, which results in the abrupt change from the urceolate to the siliquoid form, is similarly of phylogenetic significance and points to an ancestral condition in which the urn-shaped fruit was typical; or in other words to a phylogenetic history in which a fumarioid type of ovary preceded the present-day siliquoid type. Although a consideration of the interpretation of the androecium lies wholly outside the scope of our present enquiry, the unique distribution of the stamens of the Fumariaceae in two groups, each composed of one central whole stamen flanked on either side by a half stamen, has so great an interest in the present connection as to justify a brief digression into this field. For this J i J grouping affords a striking parallel with the 11 carpel combinations forming the unit components of the gynoecium in certain genera belonging to other families, as e.g. Rosaceae and Liliaceae among others, a parallelism which is fully borne out by the similar mode of development in the two cases. Each 11 J stamen group, according to Payer', arises as a single protuberance. This fact is quite in accord with what is to be observed in carpel development in such cases as those instanced above. That these staminal triplets are not the outcome of the branching of two lateral stamens is shown by the mode of origin of their vascular bundles. As the cords for the two outer lateral petals leave the central vascular ring they carry out ' Organoginie, 1, p. 228.

13 Illustrations of Carpel Polymorphism. II 187 conjoined with them the cords for the two lateral whole stamens (Fig. 44). The two cords originally passing out from the centre in the median plane are also composite and shortly divide to furnish the bundle for the small scale-like sepal and the superposed inner petal (Fig. 46). These are followed immediately by four very small bundles situated one on each side of both cords and destined for the four half stamens (Fig. 47). The petal tissue blocks the way to the union and consolidation of each pair of these small bundles into a single median cord. This congenital separation into two portions of the vascular component of both median stamens persists in the later development, hence the triplet arrangement. This further confirmation of the ordinary explanation of the appearance of the adult androecium of the Fumariaceae might have seemed superfluous were it not that some writers still treat this question as though the above interpretation were open to doubt. CRUCIFERAE I have already dealt in an earlier communication with the construction of the gynoecium of the Cruciferae in general^, and in particular with that of Capsella Bursa-pastoris Moench. and its derivatives C. Heegeri Solms and C. Viguieri Blar.^. At that time however the true nature of the character which distinguishes C. Viguieri was not wholly appreciated. From the further account here given it will be seen that an added interest attaches to this form in the light of the new facts. Capsella Vigttieri Blar. (Figs ). According to Blaringhem, who thus named a sohtary individual found by him growing wild in 1908, C. Viguieri is primarily to be distinguished from the type form, C. Bursa-pastoris, by its 4-valved fruit. Individuals also show varying and often extreme degrees of fasciation of the axes. Examination of even a small population of this form renders it clear, however, that neither of these characters is altogether constant; for 3-valved fruits frequently occur and even 2-valved ones are not uncommon. [The 6-, 7- and even 8-valved specimens also occasionally met with in early flowers, as Blaringhem notes, arise from the union of two ovaries.] The fact that some Viguieri fruits are only 2-valved has more interest than that of a mere numerical variation, for MarchaP gives the number of chromosomes in the 2-valved type 1 Annals of Botany, 37, p ; 39, p Op. cit. 39, pp. 13S-41 and 7 figure.s (p. 136). ^ Mini. Acad. Roy. Belg. Gl. Sci. 4, pp. 8 and 22.

14 188 EDITH R. SAUNDERS form, Bnrsa-pastoris, as i6 and in Viguieri as 8, It is not proposed, however, to pursue this point further here (material for a full investigation of the chromosome problem involved has been placed in more competent hands), but to confine our attention to the special character of the carpel scheme. From microscopical investigation it becomes apparent that the really constant feature distinguishing C. Viguieri is neither the precise number of the fruit valves nor the occurrence of fasciation, both of which characters, as stated above, are variable, hut the fact that the formation of two carpel whorls, one of sterile valves alternating with one composed of fertile solid members, does not use iip the whole of the vascular tissue. A sufficient amoimt remains over to permit of the formation of a third carpel whorl. The members of this third whorl lie on the same radii as the valve members of the first whorl, that is to say in the orthogonal planes, and alternate with the solid members of the second whorl which lie in the diagonal planes. The unusual character of the members of this third whorl is evidently the result of the peculiar environmental relations consequent upon their position. The outer epidermis and mesophyll of these innermost carpels form irregular protrusions into the loculi enclosed by the outer valve carpels, this being the only direction in which expansion is possible. Their vascular cords abut on the pith which after a while comes to an end, giving place to a cavity in which ovules may be developed by one or more of the members of this third whorl. The three whorls are isomerous and usually tetramerous but most individuals produce some flowers in which the Figs, 49-59, Capselta Viguieri Blar, All from transverse sections, 49-51, From a typical gynoecium (whorls tetramerous), 49, Ovary base at the level of origin of the loculi of which three are already in being. The vascular cords of the three carpel whorls are clearly defined, 50, The same at a higher level showing the protrusions of the tissues of the third carpel whorl into the loculi formed by the outermost whorl, 51, The same at a still higher level. In the loculi the developing ovules arise on either side of the above-mentioned protrusions, 52-55, From a trimerous gynoecium. The stages represented in 52 and 53 correspond with those shown in 49 and 50 respectively, 54, The same at a higher level. The carpels of the third whorl have formed a second ovary within the first, and have produced ovules in the loculus which appears as the pith comes to an end. The ovules in the original ovary are borne as usual by the solid members of the second whorl, 55, The same inner ovary at a higher level, now disjoined from the wall of the outer o\ary and standing free in the loculus, 56-59, From a 6-winged gynoecium, 56. The ovary base with single vascular c^'linder, 57, The same from a higher level showing bifurcation of the vascular cylinder, 58, The resulting double ovarj' composed of two sets of three tetramerous whorls so disposed as to render the whole gynoecium bilaterally symmetrical, 59, The same at a higher level.

15 Illustrations of Carpel Polymorphism, II 189 PHYT. XXVII. 3.

16 190 EDITH R. SAUNDERS gynoecium is composed of trimerous or occasionally even of dimerous whorls. Examination of 6-, 7- and 8-winged ovaries shows that they arise from "twinning" consequent upon bifurcation of the central vascular cyhnder after the giving off of the stamen cords, and before differentiation of the carpels. In the fully formed double ovary two opposite valve carpels extend between and serve to connect the two 3-valved portions which make up the rest of the structure. According as these two connecting valve carpels are both flattened (with linear cross-section), or one is flattened and one folded (with V-shaped cross-section), or both are folded, the gynoecium appears as a 6-, 7- or 8-winged structure. As noted in an earlier account^, 5- and 6-valved fruits also occur occasionally in Tetrapoma. These instances were observed in herbarium material, and the question whether bifurcation of the vascular cylinder accounted for the extra carpels here too could not be determined. But the fact that these exceptional Tetrapoma fruits appeared to exhibit strict radial symmetry does not support this interpretation, in view of the marked bilateral symmetry characterising those fruits of C. Viguieri in which bifurcation was observed (see Figs. 58, 59). CARYOPHYLLACEAE Saponaria officinalis L. The normal gynoeeium of Saponaria, as shown already for Dianthus^, is composed of two median, sterile, valve carpels, each surmounted by a style, and two lateral, fertile, solid, styleless carpels. A semi-double form of this species is of considerable interest in the present connection for many of the flowers show an increase in the number of carpels accompanied by various malformations of the pistil such as partial disjunction of carpel groups with resultant production of ovules on the margins of the valves as well as on the solid carpels. The largest number of carpels observed in these flowers was eight. A case of this kind, where the gynoecium consisted apparently of two tetramerous whorls, the outer of valve, the inner of solid carpels, is shown in Figs In the basal region of this gynoecium the whole vascular system destined for the inner carpel whorl was consolidated into a single irregular central mass, but at higher levels the partial disjunction of some of the carpels allowed the general ground plan to become apparent. A couple of carpels, one valve and one solid, became separated off on the one side and together formed a small, distinct 1 Annals of Botany, 37, p See Carpel Polymorphism, I, yl nnals of Botany, 39, p. 142, and Figs. 36,

17 Illustrations of Carpel Polymorphism. II 191 ovary. A similar pair behaved in like manner on the otlier side, leaving the eentral ovary now fairly normal in appearanee and showing plainly its two valve and two solid members. Five styles were present, one being borne by eaeh of the four proper valves and one by the second member of one of the two small dimerous ovaries, this member having undergone a change in character from the consolidated form below to valve form above. This faet alone is suffieient proof of the true earpel nature of this seeond member, but as is so often the ease with aberrant forms other fresh and eonvineing evidence of earpel polymorphism was furnished by this specimen^. Figs. 60-G2. Saponaria officinalis, semi-double variety. All from transverse sections taken at successively higher levels of a gynoecium of eight carpels. 60, 61. The two valve carpels to right and left already enclose each a loculus. Below, a valve and a solid carpel are about to separate from the rest of the gynoecium and to form a small, separate, dimerous ovary. Above, the remaining valve carpel is already partly free from the main body of the gynoecium and has developed an ovule on the disjoined margin. The solid carpels of the inner whorl are still in process of differentiation. 62. One of the two small, dimerous ovaries has beeome completely disjoined and is no longer seen, the other will also shortly be completely free. The two solid carpels of the central ovary, now fully differentiated, have separated from each other, the one on the left giving a false appearance of parietal placentation (as formerly understood). It was long ago noted by Masters- that both marginal and free central (so-called) placentation were often to be met with in the double form of Saponaria. On the old monomorphic view such a combination should be a morphological impossibility since a fertile carpel margin cannot be supposed both to remain the aetual edge of the carpel lamina, and at the same time to be separated off and ' See e.g. the case of Cheiranthus Cheiri var. gynanthcrus. New PhyintogisI, 27, p. 50, and Figs ' Vegetable Teratology, p Also Jourii. Linn. Soc. p. i

18 192 EDITH R. SAUNDERS form a part of the central vascular column, as this view would require in these cases. On the theory of carpel polymorphism, on the other hand, there is no fundamental difficulty in a valve member, ordinarily sterile when conjoined with fertile, solid members, itself becoming fertile along its free margins on becoming disjoined from these members. A somewhat comparable case has already been described in the case of the single and double forms of Kerria japonica'^. In concluding I wish to express my very grateful thanks to Miss D. F. M. Pertz for making the drawings here reproduced. ' Annals of Botany, 41, p. 604 and Fig

19