418 AMERICAN JOURNAL OF BOTANY [Vol. 43

(indoleacetic acid) is the aUXIn of importance In GOWING, D. P., AND R. W. LEEPER. 1955. Induction of pineapple flowering. flowering in pineapple by heta-hydroxyethylhydrazine. Science 122:i267. PINEAPPLE RESEARCH INSTITUTE OF HAWAll, HONOLULU, HAWAII LEOPOLD, A. C. 1955. In Auxins and growth, PI'. 258­ 260. Univ. California Press. Los Angeles.. LITERATURE CITED --, AND K. V. THIMANN. 1949. The effect of auxin on flower initiation. Amer. Jour. Bot. 36:342-347. BONNER, J., AND J. LIVERMAN. 1953. Hormonal control of LINFORD, M. B. 1933. Fruit quality studies. II. Eye number flower initiation. In Growth and differentiation in and eye weight. Pineapple Quart. 3:185-188. , p, 299-300. Iowa State College Press. Ames. THIMANN, K. V. 1937. On the nature of inhibitions caused CLARK, H. E. AND K. R. KERNS.. 1942. Control of flowering by auxin. Amer, Jour. Bot. 24:407-412. with phytohormones. Science 95 :536-537. COOPER, W. C. AND P. C. REECE. 1942. Induced flowering of SKOOG, F., AND C. TSUI. 1951. Growth substances and the formation of buds in plant tissues. In Plant growth pineapples under Florida conditions. Proc. Florida State Hort. Soc. 54 (1941):132-138. substances, PI'. 263-285. Univ. Wisconsin Press. Madison. FOSTER, R. J., D. H. McRAE, AND 1. BONNER. 1952. Auxin­ induced growth inhibition a natural consequence of VAN OVERBEEK, J. 1951. Use of growth substances in tropi­ two-point attachment. Proc. National Acad. Sci. (U.S.) cal agriculture. In Plant growth substances, p. 225­ 38:1014-1022. 244. Univ. Wisconsin Press. Madison. GORDON, S. A., AND F. S. NIEVA. 1949. Biosynthesis of --, E. S. DE VASQUEZ, AND S. A. GORDON. 1947. Free auxin in the vegetative pineapple. I., II. Arch. and bound auxin in the vegetative pineapple plant. Biochem. and Biophys, 20:356-385. Amer, Jour. Bot. 34:266-270.

STUDIES IN THE HIPPOCASTANACEAE II. INFLORESCENCE STRUCTURE AND DISTRIBUTION OF PERFECT FLOWERS 1 James W. Hardin

IN CONJUNCTION with a detailed study of the the ends and along the sides; they are in part phylogeny and systematics of the American species monopodial, Each lateral branch from the central of the Hippocastanaceae, two questions concerning axis is subtended by a bract and at the node of the inflorescences have arisen. First, just what each successive branching there is a pair of small basic type of inflorescence is represented and what subtending bracteoles. Deviations from this type changes in this basic type cause the conspicuous of inflorescence result in a wide variation in size, variations in size and shape? Second, what is the shape, and general aspect. distribution of perfect flowers within an inflores­ The illustrations of Billia inflorescences (fig .. cence? 1-3) show branches of only one plane for sake INFLORESCENCE STRUCTURE.-A brief look at the of clarity. The branches of the central axis, how­ literature concerning the inflorescences of the Hip­ ever, are actually decussate, i.e., succeeding pairs 0 pocastanaceae, or, for that matter, inflorescences are turned more or less 90 • in general, illustrates how thoroughly confused Very commonly in Billia columbiana PI. & Lind. terminology can become when early misinterpreta­ the inflorescence is very large, diffuse, and leafy tions are wrongly "corrected" or reproduced over at the base, as illustrated in fig. 1. The lowermost and over in textbooks and manuals. The inflores­ branches of this inflorescence are subtended by cence of has been described as a raceme, unmodified leaves rather than the caducous bracts panicle, or a thyrse with lateral helicoid or scorpi­ that are found at the upper nodes. The first inter­ oid cymes depending upon which author one nodes of these lowest branches are often very long, follows. Thanks to the recent studies by Rickett producing a very broad and nearly flat-topped (1944, 1955) the terminology of inflorescences is inflorescence. now more understandable and usable. I am there­ Within the lateral branches, abortion of the fore following the proposals of his latest paper central or lateral member of a dichasial cluster, or (1955) for the terms used here. both, has resulted in an apparent dichotomy or The inflorescence type of the Latin American monopodium in many positions throughout the Billia is a diffuse terminal panicle (fig. 1-3). inflorescence. The lost member in either situation The central axis exhibits opposite branching; the is usually abortive and caducous. At each node lateral branches are paniculate, bearing dichasia at there is always found the subtending bracteole of the abortive branch and usually a scar left by the 1 Received for publication December 1, 1955. I am deeply abortive and caduceus member. The occurrence grateful to Dr. Warren H. Wagner, Jr., Department of Botany, University of Michigan, for his real interest and of these abortive or little developed branches sug­ constructive criticism during the preparation of the manu­ gests various degrees of suppression of branches script. of monopodial or dichasial axes. June 1956] HARDIN-STUDIES IN THE HIPPOCASTANACEAE. II 419

Fig. 1-3. Diagrammatic representation of the inflorescences of Billia.-Fig. 1. B. columbiana PI. & Lind.-Fig. 2, 3. B. hippocastanum Peyr,

In Billia hippocastanum Peyr, (fig. 2, 3) the The sympodial axis is composed of axillary shoots. subtending leaves have been replaced by small Each of the axillary shoots is subtended by a small bracts and there is more tendency toward a reduc­ and caducous bracteole, and the terminal extension tion of the inflorescence through a loss of the of each shoot is always limited by a flower or a branches of the second and third order. Also in flower rudiment. It is important in distinguishing this species, most inflorescences are narrow and this type of lateral axis from a racemose type to small owing to a lesser development of the lateral understand that in the sympodial axis each bracte­ branches, and to shorter internodes. ole subtends the next sector of the whole "false" In Aesculus, the terms panicle and thyrse are axis whereas in the racemose type each bracteole applicable in designating the inflorescences. Ac­ merely subtends the pedicel of one of the lateral cording to Rickett (1955), a panicle is a loosely flowers. The opposing sterile bracteole, if present, branched inflorescence of which the ultimate units varies in position from a point opposite the fertile may be of various types. A thyrse is a compact and subtending bracteole up to the position of the panicle of more or less cylindrical form. These pedicel joint. When this point of articulation is definitions are very broad and general. If they quite low, the sterile bracteole is usually absent. are to be used in manuals and texts to describe The subtending bracteoles, usually larger than specific structures, they should be followed by a the sterile ones, are nearly always present and qualification as to just what kind of branching and arise alternately to right and left along the sympo­ ultimate units are found in the particular taxon dial axis-indicating a cincinnus rather than a being described. In Aesculus the panicle is a prin­ drepanium or bostryx. (In a drepanium all sub­ cipal elongated axis with irregular monopodial tending bracteoles are on one side of the sympodial branching. The lateral branches are, in this genus, axis; in a bostryx they are spiraled around the cincinni. The distinction made between the panicle axis.) In Aesculus, however, the pedicels of the of Billia and that of Aesculus is based upon the cincinnus are usually bent through an angle of 90 0 difference in the branches-these being lateral or less which causes them to be aligned nearly in panicles in Billia and lateral cincinni in Aesculus. one plane, or directed upward, and which gives the The basic structure of the lateral cincinnus is appearance of a drepanium. The conspicuous illustrated diagrammatically in side view by fig. 4. downward coiling of the cincinnus in Aesculus is A top view of such a cincinnus is shown by Law­ partially due to a slight accentuated growth on one rence (1951, p. 63, fig. 3d) although called a side of the axis, but primarily caused by a decrease "scorpioid cyme" by him. Each cincinnus is sub­ in pedicel lengths away from the central axis of tended by a bract which is either caducous or the panicle. Since the flowers are nearly in one which may persist through the flowering period. plane, due to the bending of the pedicels, the true 420 AMERICAN JOURNAL OF BOTANY [Vol. 43

4

6 8 eM. .

8

7

10 II 12 13 14 June 19.56] HARDIN-STUDIES IN THE HIPPOCASTANACEAE. II 421 nature of the lateral branch as a cincinnus is indi­ The pedicel elongates as the flower matures; thus cated only by the alternating lateral bracteoles. the longest pedicel is normally the one closest to The conspicuous differences among Aesculus in­ the central axis of the panicle. Resulting from this florescences (fig. 5-14) rcsult from five modifica­ (as previously mentioned) is a decrease in pedicel tions of the basic form. These modes of specializa­ length away from the central axis which causes an tion occur singly or in combination. Although some apparent curvature or downward coiling of the of the species have an inflorescence of a character­ cincinnus, although the sympodial axis itself may istic and fairly uniform shape, many of these vari­ be essentially straight. ations may be found within a single species. For In addition to the variation in pedicel length on the most part then, the names of the species in fig. a single cincinnus, one finds that some species are .'>-14 serve merely as references to the specimens more or less characterized by their pedicel lengths. used for the illustrations. For example, A. parviflora has pedicels usually 1-2 The first of these modifications involves the ern. long while in most of the other species the length of the main axis which is a function of the flowers are nearly sessile or on pedicels 1-5 mm. number of nodes and the lengths of the internodes. long. There seems to be no close correlation between A fourth modification is in the loss of bracteoles these two variables so that all four possible com­ and bracts. The sterile bracteoles are frequently binations may be found-many nodes and short absent. The subtending bracteoles, on the other internodes, or few nodes and long internodes, etc. hand, are nearly always present, and whether they Two obvious extremes in size and shape are the are entirely lost is questionable. A few inflores­ very short and broad type of panicle -found fre­ cences examined apparently lacked all bracteoles, quently in A. pavia, A. glahra, and A. sylvatica, at least toward the top, but the cincinni in these and the very long columnar type found most were reduced to such a point that evidence of the typically in A. parviflora and A. californica. I have presence of bracteoles or their scars may have been noticed that this type of variation may be found obscured. Neither bract nor bracteole was found also on a single tree or shrub, in which case it is on a few inflorescences of A. hippocastanum; how­ correlated with the position and age of the shoot. ever, their minute vestiges were probably hidden Vigorous shoots or suckers, arising at the bases of within the dense, reddish indument covering the older trees or shrubs, will usually have very long axes of the panicle. columnar inflorescences. This type of inflorescence The fifth type of modification is in the pattern would fit the definition of a thyrse. However, since of branching of the central axis. In Aesculus this this term, as recently defined by Rickett (1955), does not seem to fit any regular arrangement or implies only a general shape, I see no point in phvllotactic fraction. Occasionally an individual using it for a few inflorescences of Aesculus. inflorescence will have an apparent 2/5 phyllotaxy. A second modification within the basic form is In the majority of specimens the branches are in the length of the lateral cincinni and the number whorled, opposite, sub-opposite, or spiral at differ­ of nodes in each. Eight flowers appears to be the ent levels on the same axis. The diagrammatic hrgest number horne on a single cincinnus of illustrations (fig. 5-14) show the branches of only Aesculus and this is at the bottom of the panicle. one plane, therefore all nodes are not indicated. The presence of only one to three flowers on any DISTRIBUTION OF PERFECT FLOWERS.-Much has cincinnus in a panicle is more or less characteristic been written concerning the different types of of A. parryi, A. caliiornica, and A. parviflora. Reg­ flowers found on an inflorescence of Aesculus (Pax, ularly in Aesculus the distal end of the cincinnus 189.5; Sargent, 1895; Knuth, 1908). Different bears one to three abortive flower buds. authors have used a great variety of terms, pre­ The first segment of the lateral cincinnus is sumably describing the same situation. Such terms elongated at least slightly in most inflorescences. as bisexual, polygamo-monoecious, polygamo-dioe­ Extreme cases are occasionally found (fig. 9-14) cious, andromonoecious, and coenomonoecious which produce a more or less "hollow" inflores­ have been applied to the Hippocastanaceae, al­ cence, i.e., a panicle in which the flowers occupy though all species seem to have the same sexual the periphery only. differentiation of the flowers. Most authors do The third modification has to do with the length agree that there are two kinds of flowers: (1) sta­ of the pedicels,. No pedicels of Aesculus have been minate, bearing a vestigial pistil, and (2) perfect illustrated, except in fig. 4, since the presence of or bisexual, bearing a fully developed pistil and the flowers and pedicels obscures the branching functional anthers. Arulromonoecious would seem pattern, at least for a two-dimensional illustration. best to fit the condition since the plants are func-

Fig. 4-14. Diagrammatic representation of the inflorescences of Aesculus.-Fig. 4. Basic structure of the lateral cincin­ nus in side view. The pedicels and flowers are indicated by dotted lines. (The scale does not apply here.)-Fig. 5. A. pauia L.-Fig. 6. A. sylcatica Bartr.-Fig. 7. A. pauia L.-Fig. 8. A. octarulra Marsh.-Fig. 9. A. pauia L.-Fig. 10. A. parviflora Walt.-Fig. 11. A. cali/ornica (Spach) Nutt.-Fig. 12. A. parryi Gray.-Fig. 13. A. glabra Willd.-Fig. 14. A. hippocastanum L. 422 AMERICAN JOURNAL OF BOTANY [Vol. 43 tionally monoecious but have both staminate and single infructescence is occasionally found in A. perfect flowers. sylvatica, A. pavia, and A. glabra and is not always The significance of the different flowers and the the condition in A. octandra. Distribution of cap­ protandrous condition in terms of effective cross­ sules or distribution of perfect flowers is not cor­ pollination has been throughly discussed elsewhere related with species differences. (Pax, 1895; Knuth, 1908) and does not need The long columnar panicles (as in A. parviflora repeating here. and A. calijornicas very often have capsules scat­ There is much variation in the relative numbers tered throughout the the entire length or occa­ of bisexual and staminate flowers in a single sionally only at the top. Figure 20 of A. X du­ panicle. Coker and Totten (194.5) reported that pontii (pavia X sylvatica) shows this situation in in A. sylvatica the perfect flowers were few or which a capsule terminates the infructescence. rarely entirely absent. In a total of 67 inflores­ DISCUSSION.-There has been much speculation cences and 2,757 flowers they found an average of as to the origin of inflorescences. Three theories only 2.7 perfect flowers per inflorescence. Knuth have been proposed postulating the panicle, single (1908) stated that most of the flowers of A. oc­ terminal flower, or the dichasium as the most tandra were perfect. Recent dissections of flowers primitive type (Lawrence, 1951). Since, in the from panicles of A. pavia, A. glabra, and A. parvi­ Hippocastanaceae, we are dealing with the mor­ flora have revealed a paucity of perfect flowers per phology and evolutionary changes within only one inflorescence, but lack of sufficient fresh or pre­ basic type of inflorescence (the panicle), it is of no served material prevents my making statistically immediate concern what relative position the pan­ significant counts like those of Coker and Totten. icle holds in the phylogeny of all inflorescence Granting that the perfect flowers are usually rela­ types. One may, however, extrapolate from this tively few per panicle, the distribution of these in series of panicles to a primitive leafy branch bear­ a single inflorescence becomes the main point of ing many dichasial clusters, but this would be only interest. an hypothesis. In Billia, I have noticed from herbarium sheets In the Hippocastanaceae, my comparative studies that the perfect flowers may be at any position in indicate a simple reduction series from a diffuse the inflorescence. In Aesculus, however, according panicle with many lateral paniculate and primarily to some authors the perfect flowers are present monopodial branches to a columnar panicle in only near the base of the inflorescence (Gray, 1849; which the lateral branches are greatly reduced Sargent, 1895; Coker and Totten, 1945). On the cincinni. I regard, then, the panicle of Billia as other hand, Knuth (1908) and Hutchinson (1926) primitive and the panicle of Aesculus as having stated that they were variously arranged, and that been derived from this, or a somewhat similar the arrangement differed among species. Variation ancestral form. in this character is apparent from observations of This reduction series is illustrated by tracing the the infructescences (i.e., fruiting inflorescences) in modifications in four morphological characteristics. late summer (fig. 15-20). Although the peduncles It is interesting to note in this regard, that many' and pedicels supporting the mature capsules enlarge of these modifications appear to have taken place greatly, the relative position of the capsules which at the top and periphery of the inflorescence first. mature in the panicle remains quite obvious. Also, The bottom nearly always exhibits more primitive many of the abortive ovaries remain on the in­ characteristics than the top. The central axis and fructescence for some time. Some of the function­ the bottom, therefore, represent the part of the in­ ally pistillate flowers, however, drop prematurely florescence which is stable and resistant to morpho­ or fail to set fruit. The presence, then, of the logical modification. The periphery and top appear remaining capsules (mature or abortive) does not to be the most plastic and subject to change. show the complete distribution of the perfect The first quite obvious change has been in the flowers, but it does show (as in fig. 15-20) that subtending leaf structures-from the typical, pal­ perfect flowers are not limited to the base of the mately compound, vegetative leaf at the base of inflorescence or to any particular position of the some Billia inflorescences to the small bracts at panicle as once thought. the top of these inflorescences and throughout all One similarity among all six infructescences panicles of Aesculus. The second noticeable change illustrated is that the majority of mature capsules has been from the monopodial and dichasial are produced from the perfect flowers which were clusters found in the lateral branches of Billia to located near the distal end of the cincinni, or at the cincinni of Aesculus. the periphery of the inflorescence. It would be The third modification in this reduction series interesting to know if this is a result of selective has affected the scars and subtending bracteoles pollination. of the lost members of the primitive dichasial The illustration of the infructescence of A. oc­ clusters. In Billia the sterile bracteoles are nearly tandra (fig. 16) substantiates the statement of always present and there is usually a conspicuous Knuth (1908) that nearly all flowers in this species scar remaining after the abortion and abscission were perfect. This large number of capsules in a of the member. In Aesculus, however, the sterile June 1956] HARDIN-STUDIES IN THE HIPPOCASTANACEAE. II 423

Fig. 15-20. Photographs of infructescences showing variation in position of the mature capsules in Aesculus.-Fig. 15. A. glabra Willd.-Fig. 16. A. octandra Marsh.-Fig. 17. A. sylvatica Bartr.-Fig. 18. A. hippocastanum L.-Fig. 19. A. sylvatica Bartr.-Fig. 20. A. X dupontii Sarg. bracteole is frequently absent and there is never there has been a complete loss of regularity in a remaining scar. The loss of these structures branching pattern of the central axis in Aesculus. might be described as being "ontogenetic" in Billia If the main axis of these inflorescences were de­ and "phylogenetic" in Aesculus. rived from a leafy branch bearing many dichasial A fourth character which has undergone a con­ clusters, then Billia is not far removed from this. siderable change in this reduction series is the In Aesculus, on the other hand, the extreme irregu­ mode of branching of the central axis. From the larity in phyllotaxy is completely different from exactly opposite and decussate branching in Billia the vegetative shoot. 424 AMERICAN JOURNAL OF BOTANY [Vol. 43

Since one frequently finds in the literature-texts SUMMARY and manuals-reference to "determinate" and "in­ In connection with the detailed study of the determinate" inflorescences, it seems important to phylogeny and systematics of the Americanspecies mention this briefly in relation to these inflores­ of the Hippocastanaceae, analyses of the inflores­ cences of the Hippocastanaceae. We now realize cence and the distribution of the unisexual and that this classification and description of inflores­ bisexual flowers have been made. The inflorescence cences on the basis of order of flowering is invalid of the Latin American genus Billia is a diffuse (Rickett, 1944, 19.5.5). The use, however, of this terminal panicle with a central axis and lateral criterion, dating back to the middle eighteenth paniculate branches bearing dichasia at the ends century, seems to persist even in the most modern and along the sides; they are in part monopodial. texts. Specialization of this has resulted in reduction of In Billia and Aesculus it is very difficult to say the subtending vegetative leaves to small bracts whether the inflorescence shows a basipetal or and the loss of many branches of the second and acropetal anthesis. In Billia the terminal flower of third order. The inflorescence of Aesculus is a a dichasial cluster should theoretically mature first, panicle with the principal elongated axis monopo­ and yet this is seldom so. Within a single lateral dially and irregularly branched and bearing lateral cincinni. The conspicuous variation within the cincinnus of Aesculus, which is composed of inflorescence of Aesculus results from five types of "determinate" units, the development of flowers specialization involving changes in (1) the length should theoretically follow a pattern which, in of the central axis, (2) the length of the lateral appearance only, resembles a simple racemose axis cincinni, (3) the length of the pedicels, (4) the of "indeterminate" growth. This expected acropetal development of bracts and bracteoles, and (.5) the anthesis in each lateral cincinnus as well as the branching pattern of the central axis. Although entire inflorescence, however, is usually disrupted some species of Aesculus have fairly constant in­ by the early development of the staminate flowers, florescence structure, others are extremely variable. i.e., a protandrous condition. It has been noticed The flowers on a single inflorescence are both stam­ in specimens of A. parryi and others which I have inate and perfect; the plants are therefore andro­ seen that the capsules frequently mature basipetally. monoecious. Staminate flowers usually outnumber Irrespective of which opens or develops first, the the perfect ones, although nearly absent in some important morphological question is which are inflorescences. The perfect flowers, as shown by initiated first. Lack of ontogenetic study prevents studies of the mature capsules in the infructes­ a positive answer to this question. cences, may either be grouped at the bottom, or more often, not limited to any particular region of Within the Hippocastanaceae there appears to the panicle. Although an evolutionary sequence be a living representative for many theoretical steps may be shown for the inflorescences within the in the evolutionary sequence from the primitive family, there can be no attempt to correlate this diffuse panicle of Billia to the advanced panicle directly with species relationships and origins until of Aesculus. Any attempt, however, to correlate considerably more evidence is at hand. this directly with species relationships and origins DEPARTMENT OF BOTANY, should wait until considerably more evidence is UNIVEHSITY OF MICHIGAN, at hand. ANN ARBOR, MICHIGAN

LITERATURE CITED

COKER, W. c., AND H. R. TOTTEN. 1945. Trees of the south­ PAX, F. 1895. Hippocastanaceae. ln ENGLEH and PRANTL, eastern states. Univ. North Carolina Press. Chapel Hill. Die natiirlichen Pflanzenfamilien. 3 (5): 273-276. GRA Y, A. 1849. The genera of the plants of the United RICKETT, H. W. 1944. The classification of inflorescences. States. 2: 203-204. George P. Putnam. New York. Bot. Rev. 10: 187-231. HUTCHINSON, J. 1926. The families of flowering plants. J. Dicotyledons. Macmillan & Company. London. ---. 1955. Materials for a dictionary of hotanical tern" KNUTH, P. 1908. Handbook of flower pollination. (Transla­ -III. Inflorescences. Bull. Torrey Bot. Club 82: 419- tion.) 2: 253-255. Clarendon Press. Oxford. 445. LAWRENCE, G. H. M. 1951. of vascular plants. SAHGENT, C. S. 1895. The silva of Norlh America. 2: 51-62. The Macmillan Company. New York. Houghton, Miffiin & Company. New York.