Evolution, 38(5), 1984, pp. 1130-1143

OUTCROSSING AND POLLINATOR LIMITATION OF FRUIT SET: BREEDING SYSTEMS OF NEOTROPICAL INGA TREES (: )

SUZANNE KOPTURI Department ofBotany, University ofCalifornia, Berkeley, California 94720

Received October 17, 1983. Revised January 25, 1984

Many hermaphroditic flowering flowers on a are visited, and pol­ produce many more flowers than fruit. linated with pollen appropriate for fer­ Low fruit set from a large number ofpo­ tilization, the plant may be unable to ma­ tential fruit may result from a variety of ture every fruit because of resource, factors. In -pollinated species, a spatial, or physical restrictions. Resource large floral display may be advantageous limitation may cause abortion of some in attracting pollinators (Gentry, 1974; developing ovules or ovaries (Stephen­ Willson and Rathcke, 1974; Schaffer and son, 1981), and can provide a logistical Schaffer, 1979; Stephenson, 1979; Aug­ basis for mate choice in plants (Janzen, spurger, 1980; Udovic, 1981), especially 1977; Willson, 1979) and sibling com­ if the density of plants is low, the dura­ petition between developing embryos tion ofbloom ofthe species is short, and/ (Kress, 1981). or there are many other species blooming Inga is a large genus ofneotropicalle­ simultaneously. If pollinators visit only gume trees that have alternate, parapin­ some of the flowers, not all flowers will nately compound leaves (often with fo­ receive pollen, and fruit set without fer­ liar nectaries: Leon, 1966; Bentley, 1977; tilization will not take place in plants that Koptur, 1984) and large floral displays are non-agamospermous and without au­ (Croat, 1978; Koptur, 1983). The showy tomatic selfing; in this way the plants can white inflorescences are composed of be pollinator-limited, by virtue of low many flowers that have reduced perianth visitation (Bierzychudek, 1981). Incom­ parts and many white stamens; the sta­ patibility systems can cause pollinator­ mens are basally connate, forming a tube limitation ofa more complex nature: the within which nectar accumulates. The fruit set ofa self-incompatible species will open flowers are visited by a wide array be limited by the kinds of visitors, and ofvisitors ofwhich bats, hummingbirds, the kinds ofvisits (intraplant movements hawkmoths, butterflies, and settling resulting in intrafloral selfing or geito­ are pollinators (Snow and Snow, nogamy versus interplant movements 1972; Salas, 1974; Toledo, 1975; Fein­ resulting in xenogamy) (Arroyo, 1976). singer, 1976, 1978; Koptur, 1983). Inga Interspecific pollinator movements can pollen grains are clustered into globular result in stigma-clogging, with foreign polyads of 16 to 32 grains; the po1yad is pollen impairing the adherence and ger­ dispersed as a unit. Most Inga produce mination ofpollen ofthe correct species few fruit in relation to the number of (Levin and Anderson, 1970; Wissel, 1977; flowers they bear (Leon, 1966; pers. ob­ Waser, 1978). Allelopathic effects offor­ serv.). The focus of this study is the fac­ eign pollen may also depress fecundity of tors influencing fruit set in some Costa some species (Sukada and Jayachandra, Rican species ofInga. The floral biology, 1980; Thompson et al., 1981). Even ifall flowering and fruiting phenology ofthese Inga species is described by Koptur (1983). The breeding systems of these I Present address: Department of Biological Sci­ ences, Florida International University, Miami, trees are elucidated here as a basis for Florida 33199. understanding patterns offlowering, and 1130 BREEDING SYSTEMS OF NEOTROPICAL TREES 1131 interpreting pollinator movements in po1yads were mounted in polyvinyl lac­ terms of potential fertilization of Inga tophenol cotton blue, and the pollen flowers. The opposing evolutionary con­ grains were separated by applying pres­ straints of massive flowering to attract sure on the cover slip. The number of pollinators and the resulting excessive pollen grains per polyad was counted un­ self-pollination are conflicts common to der a compound microscope. Twenty many tropical trees (Bawa, 1983). In this five-30 polyads from 2-5 individuals per study I consider the importance of pol­ species were counted. Ovules were dis­ lination phenomena in limiting fruit set sected from ovaries and counted under a in Inga and possibly many other species. dissecting microscope. Twenty five-40 ovaries from 2-5 individuals per species MATERIALS AND METHODS were dissected. The pollen ovule ratio (P/ The study area is in Puntarenas Prov­ 0) for each species was calculated using ince, Costa Rica, situated between 1,320 the formula shown in Table 1. m and 1,600 m elevation, and includes Activity of visitors was observed at two forest types and the transition zone flowers of selected Inga brenesii and I. between. The lower areas consist largely punctata individuals for 10 min every 30 offorest remnants bordering cleared pas­ min over 8-h periods on several different tures in the communities of Santa Elena days. Visitor frequency and the number and Monteverde. The transition area is of flowers visited in each foraging bout included in a large tract of privately were noted where possible, and in some owned forest contiguous with the Mon­ cases it was possible to watch where the teverde Cloud Forest Biological Reserve, visitors went after they left the individual a large area of pristine forest, which in­ tree under observation. cludes the uppermost locations. The sites The amount ofpollination in the field have been described in detail (Buskirk was measured by looking for polyads on and Buskirk, 1976; Feinsinger, 1976; stigmas of randomly collected flowers Powell, 1979; Lawton and Dryer, 1980; (both fresh and recently wilted) from 3­ Hartshorn, 1983). 5 individuals ofall the species. The flow­ Seven species of Inga were studied: ers were preserved in ETOH or FAA and Inga brenesii StandI., I. densiflora Benth., not examined until later, so that some I. longispica StandI., I. mortoniana J. polyads may have fallen off if not ger­ Leon, I. oerstediana Benth. ex. Seem., 1. minated. The measure is therefore a con­ punctata Willd., and 1. quaternata Poep­ servative estimate on two counts: fresh pig. Voucher specimens are in the her­ flowers may have had additional polli­ baria of the University of California, nation opportunities prior to wilting, and Berkeley (UC), the Missouri Botanical ungerminated polyads may have been Garden (MO), and the Museo Nacional disturbed prior to counting. de Costa Rica (CR). All these species are Crossing experiments to investigate the canopy or subcanopy trees. Trees used in breeding systems ofsix species were per­ the study were growing at the forest's edge, formed on inflorescences bagged with along trails, or in open pastureland, as Pollen-Tector heavy paper bags (Carpen­ the flowers ofthese individuals were ac­ ter Paper Co., Des Moines, IA). Inflo­ cessible. Flowers were reached by climb­ rescences were bagged the day before the ing trees, using a ladder, or occasionally opening offlowers to be hand-pollinated. from the ground. There are biases that Flowers that were open previously were may have arisen by working on semi­ removed at the time ofbagging. Approx­ isolated trees (altered pollinator activity, imately equal numbers were used for each resource availability) resulting in false treatment on each tree. Flowers were first (perhaps) estimates of actual fruit num­ emasculated, andpolyads from eitherdif­ bers in nature. ferent flowers on the same tree (self-pol­ To determine pollen/ovule ratios, lination) or from flowers on two or three 1132 SUZANNE KOPTUR other trees (cross-pollination) were ap­ Interspecific cross-pollinations were plied to the receptive stigmas. All flowers performed between simultaneously pollinated on an inflorescence received blooming Inga species: I. brenesii x I. the same kind ofpollen; only 1-4 flowers oerstediana; I. brenesii x I. punctata; I. were pollinated per inflorescence, as it punctata x I. densiflora; I. brenesii x I. was found (see below, maximum fruit set mortoniana. 20-30 crosses were per­ capacity) that maximum fruit set did not formed on at least two individuals ofeach exceed 4-6 fruit per inflorescence. Pollen species in all cases. was collected from flowers that had In all experiments, inflorescences were opened inside bags; stamens were re­ re-bagged for 3-5 days after pollination, moved with clean forceps and put in glas­ until the non-fertilized flowers began fall­ sine envelopes. Polyads were visible on ing. Inflorescences with flowers remain­ the hairs of fine paint brushes that were ing (presumably fertilized) were tagged used to place polyads on stigmas. Care and labeled, and subsequent fruit devel­ was always taken to apply at least three opment was monitored on a monthly ba­ polyads to a stigma. On other bagged in­ sis. When fruits appeared fully mature, florescences on the same trees, newly­ they were removed and opened to see if opened flowers were left untouched to they had mature seeds; only then was a test for autogamy (automatic selfing). An flower counted as successful in producing additional group of unbagged inflores­ a fruit. All seeds germinated after sowing cences on the same trees were unobtru­ unless they had been damaged in the pod sively tagged and left untouched for mon­ by larvae. itoring fruit set on "open pollinated" In order to examine more closely the flowers, in order to estimate the actual nature ofthe incompatibility system, the amount ofsuccessful fertilization that oc­ styles of I. brenesii flowers pollinated ac­ curs under natural conditions. cording to the preceding regime (401 The fertilization effectiveness ofpollen treatment) were fixed in lactophenol at from different distances away from stig­ 2-h intervals from the time ofpollination matic parent was investigated for Inga (2 h, 4, 6, 8, 12, and 24 h). These styles brenesii and I. punctata. Pollen was col­ were cleared and stained with aniline blue lected from three sources (Self, Near­ using the method of C. Johnson and D. Cross, and Far-Cross-2 or 3 individuals Mulcahy (adapted from Ramming et aI., at each distance) from flowers newly 1973, Stain Tech. 48:133-134). Pollen opened in bagged inflorescences. Polli­ tube growth was observed using a fluo­ nations were performed within 3-4 h of rescence microscope. collecting the pollen, and all three types of crosses were done more-or-less si­ RESULTS multaneously (same tree, sametime, same day) so that all conditions were as similar Pollen and Ovule Counts, PIO Ratios as possible. For "near" cross-pollina­ All ofthe Inga species have 8 polyads tions, the pollen source tree was at a dis­ per anther (therefore, 8 polyads per sta­ tance less than .5 km (.2-.4 km) from the men), as reported by Leon (1966) and stigmatic parent; for "far" cross-polli­ Elias (1981). Previously (Leon, 1966) it nations, the pollen-parent was at a dis­ had been reported that the number of tance more than 1 km away (1-3 km). pollen grains per polyad was either 16, The maximum fruit-set capacity ofan 24, or 32, presumably constant within a inflorescence was determined in Inga species. I found considerable variation in brenesii by cross-pollinating every flower this number for Inga species with greater on ten inflorescences as the flowers than 16 pollen grains per polyad (Table opened over 3-10 days. Pollen from a 1-I. brenesii, I. densiflora, I. oerstedi­ distant source was applied (these were all ana, I. punctata, I. quaternata). Varia­ "far" crosses). tion in the number of pollen grains per BREEDING SYSTEMS OF NEOTROPICAL TREES 1133

TABLE 1. Pollen ovule ratios. Numbers given are x ± standard deviation. Sample size is 25-40 for each count. P/O = pollen ovule ratio; p.g. = pollen grains.

# p.g.zpolyad x number x number .enumber stamens/flower pollen grains/polyad ovules P/O' # ovules 1. brenesii 92.07 ± 14.50 33.8 ± 4.48 21.13 ± 2.72 1178.22 1.60 I. densiflora 29.65 ± 7.32 30.5 ± 3.06 17.50 ± 2.09 413.41 1.74 I. longispica 80.29 ± 12.16 16.0 ± .00 12.71 ± 1.50 808.59 1.26 I. mortoniana 50.00 ± 11.83 16.0 ± .00 16.80 ± 1.93 380.95 .95 I. oerstediana 60.02 ± 7.39 31.3 ± 1.92 18.07 ± 2.52 830.38 1.73 I. punctata 37.00 ± 6.90 23.2 ± 1.76 19.40 ± 1.35 353.98 1.20 1. quaternata 35.30 ± 4.99 21.8 ± 4.53 19.80 ± 1.62 310.93 1.10

8 polyads X # pollen grains x # stamens x x ---=---~- stamen polyad * Pollen ovule ratio figured by formula P/O = ------x # ovules polyad in species has been de­ than is hawkmoth activity during eve­ scribed by Leach and Whiffin (1978) and ning and night hours (Table 2), but both Ali and Quaiser (1980), associated with of these kinds of visitors visit a large hybridization. Guinet (1981) also de­ number offiowers. The number ofvisits scribes asymmetric 28-grain polyads for per fiower per day for each visitor type some Inga species. was estimated by multiplying (the mean Mean ovule numbers range from 12.7 # visits during the 10 min observation in 1. longispica to 21.1 in 1. brenesii (Ta­ periods) x (6 periods/h) x 12 h ofact iv­ ble 1). Pollen/ovule ratios are all within ity per day) x (the mean number offiow­ the same order ofmagnitude, but the ra­ ers per visit) to obtain the estimated tios of 1. brenesii, 1. longispica, and I. number offiowers visited each day, which oerstediana are roughly twice that ofthe was then divided by the number ofopen other species. These three species have fiowers counted on the study trees. Hum­ the greatest number of stamens as well. mingbirds provide an estimated 4.1 vis­ The final column in Table 1 shows the its/fir-day, hawkmoths .7, and skippers mean number ofpollen grains per polyad .1; a total ofclose to five visits per fiower divided by the mean number of ovules per day. Chances are that very few flow­ for each species. The data in this column ers go unvisited! Feinsinger (1976) re­ address the question: Can one polyad be ports hummingbird visits to this species enough to fertilize all the ovules and form average 1.68 visits/fir' day, (from 1.24­ a full-seeded fruit? In all cases (except 1. 5.09). mortoniana, which is .95) the ratio is Inga punctata has most of its flowers greater than one, indicating that if the opening around dusk, but a small pro­ pollen grains are all ofa compatible type, portion have early morning anthesis all the ovules could be fertilized with one (Koptur, 1983). This accounts for the in­ polyad, creased importance of hawkmoth polli­ nators (an estimated 2.02 visits/fir' day) Pollinator Activity over hummingbirds (.4). Pollinators fur­ Inga brenesii has fiowers that open nish an estimated total of 2.4 visits/fir' continuously throughout the day and day, ample for most fiowers to receive night (Koptur, 1983), and both hum­ pollen. Feinsinger (1976) reported a fig­ mingbirds and hawkmoths are abundant ure of .12 visits/fir-day for humming­ and effective pollinators. Hummingbird birds on Inga sp. 2 (later identified as I. activity is higher during daylight hours punctata). 1134 SUZANNE KOPTUR

TABLE 2. Visitor activity on two Inga species. Numbers given are: x ± standard deviation (range). Fifteen observation periods averaged for each visitor type; flower counts averaged for 20 visits/visitor.

Estimated # visits # flowers # visits during 10 min. Hours of # flowers visited # flowers per flower Visitor obs. period activity per visit each day on tree per day Inga brenesii Hummingbirds 4.5 ± 2.6 12 36.1 ± 26.5 11,696 2,860 4.09 (diurnal) (1-9) (4-91) Hawkmoths 2.7 ± 2.3 12 10.2 ± 8.8 1,983 2,860 .69 (crepuscular and nocturnal) (0-8) (1-32) Skippers .3 ± .4 12 9.2 ± 6.8 199 2,860 .07 (diurnal) (0-2) (1-21) Total for brenesii visitors...... 4.85 Inga punctata Hummingbirds 1.2 ± 1.0 12 6.4 ± 6.9 553 1,440 .38 (diurnal) (0-4) (1-18) Hawkmoths 3.7 ± 3.2 12 10.9 ± 5.6 2,904 1,440 2.02 (crepuscular and nocturnal) (0-10) (2-22) Total for punctata visitors 2.40

Polyads on Stigmas amount of effective outcrossing that oc­ Approximately 66% of the stigmas on curs in nature; few flowers that receive Inga brenesii had at least one polyad; ap­ pollen set fruit. In the inflorescences of proximately 49% of I. punctata stigmas Inga brenesii in which every flower was were pollinated. The other Inga species cross pollinated (25-40 flowers), fruit set were found to have approximately half was never more than six (3-6), from 10­ of the stigmas with polyads (I. oerstedi­ 15% within each inflorescence. ana 66%; I. longispica 57%; I. morton­ Interspecific crosses were all unsuc­ iana 48%; I. quaternata 46%) except for cessful, indicating that the species tested I. densiflora, which had only 28% (Table are unlikely to be interfertile. No foreign 3). These are conservative estimates (see polyads were seen to germinate on Inga Methods), so it is likely that even more stigmas. Pollen grains in all polyads ex­ flowers than these received polyads. amined were entirely viable (100% stain­ ing dark with lactophenol cotton blue). Breeding Systems The variable number ofpollen grains per Hand-pollinations were performed on polyad in some Inga species may be an six of the seven Inga species, and the indication ofhybridization, but it seems results (Table 4) indicate that these species unlikely in view of these crosses. are self-incompatible, using the criteria Both self and cross I. brenesii polyads of Bawa (1974) and Zapata and Arroyo germinated. In all cases of germination, (1978). For I. quaternata no fruit was set the pollen tubes penetrated the stigmatic with either self- or cross-pollination; it is surface and grew into the style. In no likely that the two individuals used were preparation, however, did the pollen close relatives, and therefore incompat­ tubes reach the ovules. Apparently, fer­ ible (see "Pollen source distance," be­ tilization occurs after 24 h from the time low). Fruit set on open-pollinated (con­ of pollination. In approximately 50% of trol) flowers can be compared with that the self-pollinations, self pollen tubes on hand-pollinated flowers (Table 4). grew the length of the style, indicating These figures provide an estimate ofthe that the incompatibility system is ga- BREEDING SYSTEMS OF NEOTROPICAL TREES 1135

TABLE 3. Polyads on stigmas of Inga flowers at Monteverde. Flowers were sampled randomly from 3- 5 individuals per species, and preserved prior to counting.

# flowers with Total # Proportion flowers with Species No polyad 1 polyad >3 examined polyads I. brenesii 50 65 16 12 6 149 .66 I. punctata 83 46 23 6 6 164 .49 I. densiflora 118 24 10 6 5 163 .28 I. longispica 86 80 28 3 3 200 .57 I. mortoniana 65 38 14 4 4 125 .48 I. oerstediana 49 71 14 4 6 144 .66 I. quaternata 29 17 6 I 1 54 .46

metophytic in nature, and that inhibition The same trend was shown in polli­ may occur in the ovary. nations done on two individuals of I. densiflora (sample sizes smaller): cross Pollen Source Distance pollinations between neighboring indi­ The results of experimental pollina­ viduals, .3 km apart, failed to yield fruit; tions with pollen from sources at varying whereas crosses with pollen from a source distances from the stigmatic parent were 1.5 km away yielded 25% fruit. examined with contingency table analy­ sis (BMD stat. pack.; Sokal and Rohlf, DISCUSSION 1982), to take possible variation between At one time it was assumed that be­ individuals into account. For Inga bre­ cause of the enormous diversity and nesii (Table 5), there is a significant in­ complexity of tropical wet forests, and teraction between pollination treatment the large interindividual distances in and fruit set, selfs yielding the lowest, many species oftropical forest trees, most near crosses intermediate, and far crosses species would be self-compatible and the highest fruit set. For each tree taken inbred (Comer, 1954; Baker, 1959; Fed­ separately, this difference is upheld erov, 1966). Ashton (1969) was perhaps (Freeman-Tukey deviates all significant). the first to suggest that pollen and seed Inga punctata shows the same overall dispersal occur over large distances in result (Table 6); however, only three of tropical forests, and that rain forest trees four trees show greater fruit set with "far" are largely outcrossed. Janzen (1971) pollen. The fourth tree set very little fruit demonstrated that pollinating Euglossine at all; this may have resulted from weath­ bees moved great distances in tropical er conditions unfavorable to fertilization forests; subsequent studies (e.g., Stiles, (cool, misty) on the days of pollination. 1975; Frankie et al., 1976; Linhart and

TABLE 4. Results of hand pollinations. Self- and cross-pollinated flowers were bagged prior to opening, emasculated, and hand-pollinated. Control flowers were not bagged and not hand-pollinated, but were left open for visitors. Fruit set is # fruit/# flowers.

Self-pollinated Cross-pollinated Control # trees N Fruit set N Fruit set N Fruit set I. brenesii 11 319 .01 205 .35 960 .03 I. densiflora 2 88 .01 79 .25 505 .05 I. mortoniana 2 33 .03 53 .49 480 .02 I. oerstediana 3 65 .02 67 .12 325 .05 I. punctata 4 130 .02 146 .20 360 .02 I. quaternata I 50 .00 52 .00 430 .03 1136 SUZANNE KOPTUR

TABLE 5. Pollen source distance experiment with I. brenesii. Near Cross (Near X) is with pollen from a source at a distance of less than .5 km from the stigmatic parent; Far Cross is from a source greater than I km away.

Flower fate Total % fruit Tree Cross w/fruit w/o fruit # flowers set Pearson x2 dj Probability Self 0 39 39 0 Near X 8 42 50 16 21.26 2 .0000 Far X II 14 25 44 ------.--...... ------...... Total 19 95 114 2 Self 0 39 39 0 Near X 3 30 33 9 20.49 2 .0000 Far X 8 12 20 40 ...... __ .__ .._------...... _-_.._------Total II 81 92 3 Self I 33 34 3 Near X 0 15 15 0 19.41 2 .0001 Far X 18 30 48 38 ------...... Total 19 78 97 4 Self 0 40 40 0 Near X 5 44 49 10 9.10 2 .0106 Far X 5 17 22 23 -----.---...-...... _------..--...... Total 10 101 III 5 Self I 45 46 2 Near X 2 33 35 6 31.37 2 .0000 Far X 25 35 60 42 -...... ------.-...... Total 28 113 141 6 Self 0 30 30 0 Near X 3 23 26 II 15.98 2 .0003 Far X 10 16 26 38 -----.-...... _---_...._--- Total 13 69 82 7 Self I 175 176

Mendenhall, 1977) have revealed that (Bawa, 1974), lowland wet forest ofCosta many tropical forest pollinators are ca­ Rica (Bawa, 1979), and lowland semi­ pable ofmoving large distances between deciduous forest in Venezuela (Zapata individuals of the same species. Inves­ and Arroyo, 1978) have discovered that tigations of many species of lowland the majority of hermaphroditic species (semi-deciduous) dry forests ofCosta Rica are self-incompatible. A recent investi- BREEDING SYSTEMS OF NEOTROPICAL TREES 1137

TABLE 6. Pollen source distance experiment with 1. punctata.

Hower fate Total % fruit Tree Cross wi fruit wlo frt # flowers set Pearson x2 df Probability Self 0 29 29 0 Near X 0 16 16 0 16.16 2 .0003 Far X 10 22 32 31 -----_...... ------Total 10 67 77 2 Self I 37 38 3 Near X 1 16 17 6 5.76 2 .0561 Far X 7 31 38 18 -----_..... _--_...... Total 9 84 93 3 Self 0 29 29 0 Near X I 19 20 5 3.24 2 .1980 Far X 0 35 35 0 (inflated value due NS to small expecteds) .-....._-.-...... -..-.....-.---.-.-.------. ------.---- Total 1 83 84 4 Self 2 32 34 6 Near X 2 9 11 18 6.71 2 .0348 Far X 12 29 41 29

Total 16 70 86 All Self 3 127 130 2 trees Near X 4 60 64 6 23.95 2 .0000 Far X 29 117 146 20

Total 36 304 340 gation of montane tropical cloud forest tubes into ovules in any of my prepara­ in Venezuela (Sobrevila and Arroyo, tions, the suggestion is that the incom­ 1978) found that a lower proportion of patibility reaction takes place in the ovary. hermaphroditic species possess self-in­ Bawa (1979) suggests that such pollen­ compatibility systems than species of ovule incompatibility may occurin many lowland forest, a consequence, perhaps, tropical trees; Kress (1983) reports a sim­ ofweather-influenced unpredictability of ilar situation in Heliconia species, trop­ pollination pattern. ical herbs. The Inga species studied here in mon­ Self-incompatibility is more wide­ tane cloud forest of Costa Rica are all spread in woody than herbaceous le­ self-incompatible. Arroyo (1976) sug­ gumes (Arroyo, 1981). The tribe Ingeae gested that self-incompatibility in mas­ is a largely woody group, and all species sively flowering tropical trees has evolved previously tested have proven self-in­ as a response to excessive geitonogamy compatible. Of the three (out of 5) mi­ (self-pollination effected by pollinators mosoid tribes in which some experimen­ moving between flowers on the same tation has been done, all contain some tree). Inga fits a typical pattern for a mass­ self-incompatible species. The phenom­ flowering tree, in having a large number enon of few fruit set from many flowers of small, generalized flowers that attract has been noted in other mimosoid le­ many species of pollinators. Although I gumes besides Inga: Parkia clapperton­ did not observe the penetration ofpollen iana (Baker and Harris, 1957), Prosopis 1138 SUZANNE KOPTUR

(Solbrig and Cantino, 1975; Simpson et on individuals of Calliandra callistemon a1., 1977), Calliandra (Crudenet a1., 1976; were pollinated (as evidenced by pres­ Cruden, 1977). ence ofpolyads on stigmas), but fruit set Pollen grains dispersed as a unit (in in this species was only 5.2% (Cruden et polyads, or pollinia) provides great effi­ a1., 1976). ciency offertilization ifthe pollen grains Resource limitation alone does not ex­ are ofa compatible nature-one good po­ plain the low levels of fruit set in Inga. lyad can potentially fertilize all the ovules This factor was controlled by doing dif­ needed to fill a fruit. Cruden (1977) has ferent pollinations on the same individ­ called this strategy "sweepstakes repro­ ual trees; hand cross-pollinations with duction." The Inga pollen/ovule ratios pollen from far sources were much more are low for a typical obligately out-cross­ successful than "open pollinations" on ing species (Cruden, 1977), but are very the same trees, indicating that it was not similar to those Cruden found for species for lack ofenergy reserves that fruit were ofCalliandra, another mimosoid legume not forming. Although an Inga inflores­ with polyads. Asclepias have pollinia, and cence may bear up to 40 flowers, the exhibit low fruit sets in many situations maximum number offruits matured with (Willson and Price, 1980). Kress (1981) multiple optimal hand cross-pollinations argues that this strategy reduces intra­ was six (more usually, 4 or 5); there may ovary competition between sibling seeds be physical/spatial restrictions on how as all are sure to be genetically similar. many large fruit will fit on one peduncle. Pollen packaging in polyads assures high Resource limitation has been found to be seed set if one compatible unit is trans­ important in Asclepias species (Willson ferred. These sorts ofpollination systems and Price, 1980): when fertilizer was may allow a plant to recognize and choose added, fruit set increased; with defolia­ between mates more readily than in sit­ tion and shading, fruit set decreased. Ca­ uations where the number ofpollen grains talpa speciosa responded to defoliation reaching a stigma is highly variable. Here with lower fruit set due to increased fruit the most important factor may be finding abortion (Stephenson, 1980). Baker maximum compatibility. Clearly, more (1970) suggested that resource limitation experimentation needs to be done before of fruit production is not infrequent these speculations can be tested. among tropical woody plants. There are What factors limit fruit set in Inga? many examples ofresource limitation in Lack of pollinator visitation is not the cultivated plants (reviewed by Stephen­ explanation: not only were approximate­ son, 1981) and reports ofincreasing fruit ly halfofthe stigmas ofthe various species set with fertilizing at the time of flow­ found to have polyads on them, but ob­ ering. Overall fruit set of Inga would servations of all visitors to two species probably not be enhanced by application gave estimates ofvisits per flower per day ofsoil fertilizers, etc... , unless the flow­ in excess of 1, indicating that most flow­ ers receive increased numbers of com­ ers are probably visited. A number of patible polyads (see below). other plants have been shown to have Inga are pollinator-limited (sensu natural pollinations much in excess of Bierzychudek, 1981), as open-pollinated fruit set. In milkweeds (requiring only one flowers and inflorescences have much pollinium for fruit set), only 9.6% ofAs­ lower fruit set than those flowers and in­ clepias syriaca flowers set fruit, although florescences cross-pollinated by hand. 95% of the flowers had received ample The limitation is not strictly a result of pollination (Moore, 1947); A. solanoana geitonogamy; it is determined also by the fruit set was only 3%, although there was distance ofthe source ofcross-pollen from an average of 1.73 pollinia per flower the stigmatic parent. The mechanism of (Lynch, 1977). From 73-93% of flowers this phenomenon in Inga has not yet been BREEDING SYSTEMS OF NEOTROPICAL TREES 1139 elucidated; it may be that flowers out­ revealed many legumes, and plants of crossed with remote pollen donors are other families also, to have low open­ selectively matured. Selective abortion of pollinated fruit set (less than 10%), which fruits with less than average seed num­ could result from a variety of factors. bers has been found in other plants, in­ However, some of his cross-pollinations cluding the annual legume, Cassia fas­ (performed with pollen from several ciculata (Lee and Bazzaz, 1982b). Further sources, presumably near) yielded low experimentation on Inga is needed be­ amounts offruit-e.g., Enterolobium cy­ fore anything definite can be said regard­ c!ocarpum (Mimosoideae) with crosses ing selective maturation-the strategies yielding only 28% fruit; Caesalpinia er­ may well be different in long-lived pe­ iostachys 10%; Lonchocarpus eriocar­ rennials and annuals. What can be as­ inalis 28%; Hemiangium excelsum 17%; serted is: since crosses are more success­ Hirtella racemosa 18%; and Godmania ful when pollen comes from a substantial aesculifolia 14% (Bawa, 1974). A likely distance, the most effective pollen trans­ explanation for this phenomenon is that fer results from pollinators that move not cross-pollinations between genetically only from tree to tree, but to trees some similar individuals are less fecund than distance away. those between genetically distinct indi­ Many Inga pollinators are known to viduals. be strong fliers, and are likely to travel Inga population structure is the result substantial distances in their daily for­ of gene dispersal effected not only via aging bouts. Hummingbirds have been pollen movement, but also by seed dis­ implicated in traplining by Stiles (1975), persal. Seed dispersal of Inga is per­ and are capable ofmovinglong distances. formed by mammals (monkeys, squir­ Frankie et al. (1976) found that territorial rels) and perhaps by birds (pers. observ.). behavior by Centris bees in crowns of Dispersers eat the sweet pulp surround­ Andira inermis cause more inter-tree ing the seeds, and drop the seeds (which movement by other bees; territorial and germinate almost immediately). aggressive behavior ofhummingbirds on feeding in the parent tree may drop many Inga may cause others to move greater seeds directly below, resulting in off­ distances than they might otherwise, cur­ spring clustered around the parent tree. tailing not only geitonogamy by the oth­ Seed dispersal may also occur to a great ers, but also cross-pollination of near distance when whole fruits are carried off neighbors. Sphingidae moved Lindenia by dispersers that may range far and wide, rivalis pollen hundreds of meters (Lin­ but clumps ofrelatives may be deposited hart and Mendenhall, 1977); it is con­ where the frugivore sits and feeds. It is ceivable that pollen may be carried on common to see a dense local group of hawkmoth bodies for days, during which Inga seedlings of the same age and size time they could move quite far (W. A. on the forest floor, far from any potential Haber, pers. comm.). Hesperiidae are "parent" tree, providing circumstantial strong, fast, and direct fliers, and are im­ evidence that this type of dispersal may portant in cross-pollination of many be occurring. If this is indeed the case, plants (Klots, 1976). All ofthese visitors crosses between adult individuals close can serve as effective pollinators, but in together are likely to be crosses between many cases they transfer incompatible close relatives. This could be why far­ pollen from near neighbors. crosses are more fruitful than near-cross­ This sort ofpollinator limitation (many es in the Inga species examined. Price selfs and near-neighbor pollinations not and Waser (1979; Waser and Price, 1983) setting fruit) may be common in tropical found the "optimal outcrossing dis­ trees. Open-pollinated fruit set in all Inga tance" ofthe herbaceous perennials Del­ species here was less than 5%; Bawa's phinium nelsonii and Ipomopsis aggre­ (1974) survey of many lowland species gata to be between 1 and 100 m, 1140 SUZANNE KOPTUR determined by the interaction ofrestrict­ cific plants low as the result of selection ed pollen and seed dispersal. If there is over evolutionary time by seed predators an optimal distance for outcrossing in and other enemies (Janzen, 1970; Con­ Inga, it is likely to be greater than 10 m nell, 1978), long-distance pollinators are because of the greater distances between clearly necessary for any fruit to be pro­ individuals in nature, as well as greater duced in self-incompatible species. Re­ distances involved in pollen and some cent quantitative studies (Hubbell, 1979, sorts of seed dispersal. 1980; Hubbell and Foster, 1983) indicate The large floral displays ofInga species that hyperdispersion is not usual, but are successful in attracting many visitors rather than individuals of most species away from other species of flowering are clumped. Ifclumps are comprised of plants in the forest, but may be advan­ individuals with a higher degree of re­ tageous in other ways as well. Interspe­ latedness than that found between cific hand-crosses yielded no fruit, and I clumps, there should be selection for ad­ saw no germination offoreign po1yads on aptations in self-incompatible tree species Inga brenesii stigmas; however, foreign to support the energetic needs of excep­ po1yads may interfere with the attach­ tionally long-flying visitors, and to offer ment and germination of po1yads of the these rewards over an extended period of appropriate species. The same visitor time. The perfection of this strategy is species (from hummingbirds to hawk­ seen in species with very large flowers moths to skippers) will visit whatever opening a few at a time over a long pe­ Inga species have flowers open at their riod, taking advantage of specialized foraging time, and although a certain "trapliners." Phylogenetic constraints amount of separation is afforded simul­ may prevent many species from adopting taneously blooming, sympatric Inga this strategy-an alternative available to species by virtue of differences in floral many (e.g., the mimosoid legumes like behavior (flower opening time, anther Inga) is to produce large numbers ofre1­ dehiscence, nectar secretion; Koptur, ative1y unspecia1ized flowers that are vis­ 1983), some interspecific cross-pollina­ ited by a variety of pollinators over an tions are inevitable. Se1f-po1yadsand in­ extended blooming season. Most ofthese compatible cross-polyads can also serve flowers will be visited, but the vast ma­ to "clog" stigmas (Waser, 1978), and vis­ jority ofthese will not receive pollen suit­ itors were frequently observed to visit able for fertilization. Since most ofthese many flowers on the same tree, and/or plants exist within a clump of close rel­ move to the nearest neighbor tree after atives, these costly features may be leaving the first tree. The many thou­ worthwhile for maintenance of genetic sands of flowers produced by an Inga variability in the species. individual may be needed to provide enough stigmas so that at least a few flow­ SUMMARY ers will be successfully fertilized. Species of Inga characteristically have The findings of this study underscore large floral displays, but few ofthe flow­ the importance of long-distance polli­ ers set fruit. Seven species were studied nators in the reproductive biology of in lower montane wet forest at Monte­ tropical trees. Self-incompatibility is the verde, Costa Rica (Inga brenesii, I. den­ rule rather than the exception among siflora, I. longispica, I. mortoniana, I. tropical trees that have been studied; oerstediana, I. punctata, and I. quater­ where complex self-incompatibility sys­ nata) to elucidate breeding systems and tems are involved, the movement ofpo1­ determine what factors limit fruit set. len between individuals is not always the Observations ofvisitor activity (hawk­ only requirement for fertilization. Rea­ moths, hummingbirds, and skippers) on soning that inter-individual distances are Inga brenesii and 1. punctata indicated maximized, and the density of conspe- that there was no shortage ofpollination. BREEDING SYSTEMS OF NEOTROPICAL TREES 1141

Examination of stigmas of flowers col­ J. Leon identified the Inga species. Help­ lected from all species showed that far ful comments on earlier drafts of the more flowers had received pollen than manuscript were made by H. G. Baker, normally set fruit. Hand pollination of G. W. Frankie, W. A. Haber, C. Lumer, six spp. revealed them to be self-incom­ T. P. Spira, C. E. Turner, A. Cockburn, patible. Observations ofpollen grain ger­ K. Grove, R. J. Marquis, E. J. Trapp, P. mination and pollen tube growth in I. Davidar, D. Costich, D. Baldwin, H. F. brenesii indicated its incompatibility sys­ Howe, S. D. Hendrix, J. L. Hayes, S. J. tem is gametophytic. Pollen/ovule ratios Wright, R. W. Cruden, L. K. Wagner, N. were much lower than those usually as­ M. Waser, J. O. Washburn, M. A. Wha­ sociated with self-incompatible species, len, and N. Wheelright. The research was but this may be explained by the greater supported by a grant from the Center for efficiency of fertilization provided by Latin American Studies ofthe University having pollen in polyads. None of the of California, NSF grant DEB 78-10612 species that bloom simultaneously were to H. G. Baker and S. Koptur, NSF grant cross-compatible. Intraspecific cross­ DEB 78-11728 to G. W. Frankie and H. pollinations in I. brenesii, I. punctata, and G. Baker, and a fellowship from the I. densiflora were more successful when American Association of University the pollen sources were more than 1 km Women. All this is gratefully acknowl­ away from the stigmatic parent than when edged. This research was part of a dis­ pollen sources were less than.5 km away. sertation submitted in partial fulfillment Long-distance pollinator movements are of requirements for the Ph.D. in Botany therefore likely to be of greatest conse­ at the University ofCalifornia, Berkeley. quence in fruit setting. The low fruit set of Inga is a result of a special case of LITERATURE CITED pollinator limitation: much more geito­ ALI, S. I., AND M. QAISER. 1980. Hybridization nogamy and unsuitable xenogamy is ef­ in Acacia nilotica (Mimosoideae) complex. Bot. fected by pollinators than optimal out­ J. Linnean Soc. 80:69-77. crossing to greater distances. This ARROYO, M. T. K. 1976. Geitonogamy in animal phenomenon may be widespread among pollinated tropical angiosperms: a stimulus for the evolution ofself-incompatibility. Taxon 25: tropical trees; long-distance pollinators 543-548. may be even more important than for­ ARROYO, M. T. KALIN. 1981. Breeding systems merly thought. The flowering strategy of and pollination biology in Leguminosae, p. 723­ Inga has evolved to attract many polli­ 770. In Polhill, R. M., and P. H. Raven (eds.), nators over extended periods of time, Advances in Legume Systematics, Part 2. In­ ternational Legume Conference, 1978, Kew, while the breeding systems ensure main­ London, England. tenance of genetic variability despite ASHTON, P. S. 1969. Speciation among tropical much selfing and crosses between near rain forest trees: some deductions in the light of neighbors. recent evidence. BioI. J. Linnean Soc. 1:155­ 196. ACKNOWLEDGMENTS AUGSPURGER, C. K. 1980. Mass-flowering of a tropical shrub (Hybanthus prunifolius): influ­ H. G. Baker, G. W. Frankie, I. Baker, ence on pollinator attraction and movement. W. A. Haber, J. Leon, S. Salas, P. Fein­ Evolution 34:475-488. singer, C. Lumer, and D. Feener provid­ BAKER, H. G. 1959. Reproductive methods as factors in speciation in flowering plants. Cold ed helpful discussion, advice, and criti­ Spring Harbour Symp. Quant. BioI. 24:177-199. cism during the course ofthis study. Field --. 1970. Two cases ofbat pollination in Cen­ assistance was provided by N. Murray, tral America. Revista de Biologia Tropical 17: L. Westley, B. Yandell, C. Guindon, and 187-197. BAKER, H. G., AND B. J. HARRIS. 1957. The pol­ T. Szaro. Many residents ofMonteverde lination ofParkia by bats and its attendant evo­ and Santa Elena gave their kind permis­ lutionary problems. Evolution 11:449-460. sion for me to study trees on their land. BAWA, K. S. 1974. Breeding systems oftree species 1142 SUZANNE KOPTUR

ofa lowland tropical community. Evolution 28: p. 118-157. In D. H. Janzen (ed.), Costa Rican 85-92. Natural History. Univ. Chicago Press, Chicago. ---. 1979. Breeding systems oftrees in a trop­ HUBBELL, S. P. 1979. Tree dispersion, abundance, ical wet forest. New ZealandJ. Bot. 17:521-524. and diversity in a tropical dry forest. Science ---. 1983. Patterns of flowering in tropical 203:1299-1309. plants, p. 394-410. In C. E. Jones and R. J. ---. 1980. Seed predation and the coexistence Little (eds.), Handbook of Experimental Polli­ oftree species in tropical forests. Oikos 35:214­ nation Biology. Van Nostrand Reinhold Co., 229. N.Y. HUBBELL, S. P., AND R. B. FOSTER. 1983. Diversity BENTLEY, B. L. 1977. Extrafloral nectaries and of canopy trees in a neotropical forest and im­ protection by pugnacious bodyguards. Ann. Rev. plications for conservation, p. 25-41. In S. L. Eco!. Syst. 8:407-427. Sutton, T. C. Whitmore, and A. C. Chadwick BIERZYCHUDEK, P. 1981. Pollinator limitation of (eds.), The Tropical Rain Forest: Ecology and plant reproductive effort. Amer. Natur. 117:838­ Management. Blackwell Scientific Publ., Ox­ 840. ford. BUSKIRK, R. E., AND W. H. BUSKIRK. 1976. JANZEN, D. H. 1970. Herbivores and the number Changes in abundance in a highland oftree species in tropical forests. Amer. Natur. Costa Rican forest. Amer. Midl. Natur. 95:288­ 104:501-528. 298. ---. 1971. Euglossine bees as long-distance CONNELL, J. H. 1978. Diversity in tropical rain pollinators oftropical plants. Science 171:203­ forests and coral reefs. Science 199: 1302-131O. 205. CORNER, E. J. H. 1954. The evolution oftropical ---. 1977. A note on optimal mate selection forests, p. 34-46. In J. Huxley, A. C. Hardy, by plants. Amer. Natur. 111:365-371. and E. B. Ford (eds.), Evolution as a Process. KLOTS, A. B. 1976. Butterflies ofthe World. Ban­ Allen and Unwin, London. tam Books, The Ridge Press, N.Y. CROAT, T. B. 1978. Flora of Barro Colorado Is­ KOPTUR, S. 1983. Flowering phenology and floral land. Stanford Univ. Press, Stanford. biology of Inga (Fabaceae: Mimosoideae). Syst. CRUDEN, R. W. 1977. Pollen ovule ratios as a Bot. 8:354-368. conservative indicator of breeding systems in ---. 1984. Experimental evidence for defense flowering plants. Evolution 31:32-46. of Inga saplings by ants. Ecology. In press. CRUDEN, R. W., S. KINSMAN, R. E. STOCKHOUSE II, KRESS, W. J. 1981. Sibling competition and evo­ AND Y. B. LINHART. 1976. Pollination, fecun­ lution ofpollen unit, ovule number, and pollen dity, and the distribution of -flowered vector in Angiosperms. Syst, Bot. 6:101-112. plants. Biotropica 8:204-210. 1983. Self-incompatibility in Central ELIAS, T. S. 1981. Mimosoideae, p. 143-151. In American Heliconia. Evolution 37:735-744. R. M. Polhill and P. H. Raven (eds.), Advances LAWTON, R.,ANDV. DRYER. 1980. Thevegetation in Legume Systematics, Part 1. International Le­ ofthe Monteverde cloud forest reserve. Brenesia gume Conference, 1978, Kew, London, En­ 18:101-116. gland. LEACH, G. H., AND T. WHIFFIN. 1978. Analysis FEDEROV, A. A. 1966. The structure ofthe tropical ofa hybrid swarm between Acacia brachybotrya rain forest and speciation in humid tropics. J. and A. calamifolia (Leguminosae). Bot. J. Lin­ Eco!. 54:1-11. nean Soc. 76:53-69. FEINSINGER, P. 1976. Organization of a tropical LEE, T. D., AND F. A. BAZZAZ. 1982a. Regulation guild ofnectarivorous birds. Eco!. Monogr. 46: of fruit and seed production in an annual le­ 257-291. gume, Cassiafasciculata. Ecology 63:1363-1373. 1978. Ecological interactions between 1982b. Regulation of fruit maturation plants and hummingbirds in a successional trop­ pattern in an annual legume, Cassiafasciculata. ical community. Ecol, Monogr. 48:269-287. Ecology 63:1374-1388. FRANKIE, G. W., P. A. OPLER, AND K. S. BAWA. LEON,J. 1966. Central American and West Indian 1976. Foraging behavior ofsolitary bees-im­ species of Inga (Leguminosae). Ann. Miss. Bot. plications for outcrossing ofa neotropical forest Gard. 53:365-359. tree species. J. Ecology 64: 1049-1957. LEVIN, D. A., AND W. W. ANDERSON. 1970. Com­ GENTRY, A. H. 1974. Flowering phenology and petition for pollinators between simultaneously diversity in tropical Bignoniaceae. Biotropica flowering species. Amer. Natur. 104:455-467. 6:54-58. LINHART, Y. B., AND J. A. MENDENHALL. 1977. GUINET, PH. 1981. Comparative account ofpollen Pollen dispersal by hawkmoths in a Lindenia characters in the Leguminosae, p, 789-800. in rivalis Benth. population in Belize. Biotropica R. M. Pohill and P. H. Raven (eds.), Advances 9:143. in Legume Systematics, Part 2. International Le­ LYNCH, S. P. 1977. The floral ecology ofAsclepias gume Conference, 1978, Kew, London, En­ solanoana Woods. Madrofio 24:159-177. gland. MOORE, R. J. 1947. Investigations on rubber­ HARTSHORN, G. S. 1983. Introduction to plants, bearing plants. V. Notes on the flower biology BREEDING SYSTEMS OF NEOTROPICAL TREES 1143

and pod yield ofAsclepias syriaca L. Can. Field STILES, G. F. 1975. Ecology, flowering phenology, Natur.61:40-46. and hummingbird pollination ofsome Costa Ri­ POWELL, G. V. N. 1979. Structure and dynamics can Heliconia species. Ecology 56:285-301. of interspecific flocks in a neotropical mid-ele­ SUKADA, D. K., AND S. JAYACHANDRA. 1980. Pol­ vation forest. Auk 96:375-390. len allelopathy-a new phenomenon. New Phy­ PRICE, M. V., AND N. M. WASER. 1979. Pollen tol. 84:739-746. dispersal and optimal outcrossing in Delphin­ THOMSON, J. D., B. J. ANDREWS, AND R. C. ium nelsonii. Nature 277:294-297. PLOWRIGHT. 198 I. The effect ofa foreign pol­ SALAS, D. S. 1974. Analisis del Sistema de Polin­ len on ovule development in Diervilla lonicera izacion de Inga vera subespecie spuria. Thesis. (Caprifoliaceae). New Phytol. 90:777-783. Universidad de Costa Rica. TOLEDO, V. M. 1975. La estacionalidad de las SCHAFFER, W. M., AND M. V. SCHAFFER. 1979. flores utilizadas por los colibries de una selva The adaptive significance of variations in re­ tropical humedo en Mexico. Biotropica 9:262­ productive habit in the Agavaceae II: pollinator 267. foraging behavior and selection for increased UDOVIC, D. 1981. Determinants of fruit set in reproductive expenditure. Ecology 60: 1051­ Yucca whipplei: reproductive expenditure vs. 1069. pollinator availability. Oecologia 48:389-399. SIMPSON, B. B., J. L. NEFF, AND A. R. MOLDENKE. WASER, N. 1978. Interspecific pollen transfer and 1977. Prosopis flowers as a resource, 84-107. competitionbetween co-occurringplant species. In B. B. Simpson (ed.), Mesquite: Its Biology in Oecologia 36:223-236. Two Desert Scrub Ecosystems. US/IBP Synthe­ WASER, N. M., AND M. V. PRICE. 1983. Optimal sis Series 4. Dowden, Hutchinson, & Ross, Inc. and actual outcrossing in plants, and the nature Stroudsburg, Pennsylvania. of plant-pollinator interaction, p. 341-359. In SNOW, B. K., AND D. W. SNOW. 1972. Feeding C. E. Jones and R. J. Little (eds.), Handbook of niches ofhummingbirds in a Trinidad valley. J. Experimental Pollination Biology. Van Nos­ Anim. Ecol. 41:471-485. trand Reinhold Co., N.Y. SOBREVILA, c., AND M. T. K. ARROYO. 1978. WILLSON, M. F. 1979. Sexual selection in plants. Breeding systems in a montane tropical cloud Amer. Natur. 113:777-790. forest in Venezuela. Plant Syst. Evol. 140: 19­ WILLSON, M. F., AND P. W. PRICE. 1980. Resource 37. limitation offruit and seed production in some SOKAL, R. R., AND F. J.ROHLF. 1982. Biometry, Asclepias species. Can. J. Bot. 58:2229-2233. 2nd ed. W. H. Freeman and Co., San Francisco. WILLSON, M. F., AND B. J. RATHCKE. 1974. Adap­ SOLBRIG, O. T., AND P. D. CANTINO. 1975. Re­ tive design of the floral display in Asclepias productive adaptations in Prosopis (Legumino­ syriaca L. Amer. MidI. Natur. 92:47-57. sae, Mimosoideae). J. Arnold Arboretum 56: WISSEL, C. 1977. On the advantage of the spe­ 185-210. cialization of flowers on particular pollinator STEPHENSON, A. G. 1979. An evolutionary ex­ species. J. Theoret. BioI. 69: 11-22. amination of the floral display of Catalpa spe­ ZAPATA, T.R.,ANDM. T. K. ARROYO. 1978. Plant ciosa (Bignoniaceae). Evolution 33:1200-1209. reproductive ecology of a secondary deciduous ---. 1980. Fruitset, herbivory, fruit reduction, tropical forest in Venezuela. Biotropica 10:221­ and the fruiting strategy of Catalpa speciosa 230. (Bignoniaceae). Ecology 61:57-64. ---. 198 I. Flower and fruit abortion: proxi­ Corresponding Editor: Stephen G. Weller mate causes and ultimate functions. Ann. Rev. Ecol. System. 12:253-279.